UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Modelling of kraft pulp mill total reduced sulphur emissions Jensen, Allan Stewart 2007

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2007-317914.pdf [ 16.97MB ]
Metadata
JSON: 831-1.0059010.json
JSON-LD: 831-1.0059010-ld.json
RDF/XML (Pretty): 831-1.0059010-rdf.xml
RDF/JSON: 831-1.0059010-rdf.json
Turtle: 831-1.0059010-turtle.txt
N-Triples: 831-1.0059010-rdf-ntriples.txt
Original Record: 831-1.0059010-source.json
Full Text
831-1.0059010-fulltext.txt
Citation
831-1.0059010.ris

Full Text

M O D E L L I N G OF K R A F T P U L P M I L L T O T A L R E D U C E D S U L P H U R E M I S S I O N S by A L L A N S T E W A R T J E N S E N B . A . S c , University of British Columbia, 1989 A THESIS S U B M I T T E D IN P A R T I A L F U L F I L L M E N T OF T H E R E Q U I R E M E N T S F O R T H E D E G R E E OF D O C T O R OF P H I L O S O P H Y in T H E F A C U L T Y OF G R A D U A T E STUDIES (Chemical and Biological Engineering) T H E U N I V E R S I T Y OF B R I T I S H C O L U M B I A October 2007 © Al lan Stewart Jensen, 2007 A b s t r a c t Atmospheric release of odorous total reduced sulphur (TRS) emissions from kraft pulp mills has been an ongoing concern worldwide. Organic TRS compounds are formed in the digester by ^undesired side reactions during the kraft pulping of wood. These, along with H 2 S , being highly volatile, are released from mil l processes, such as brown stock washing. T R S gases are extremely noxious, described as rotting eggs or rotting cabbage, and have a threshold of odour detectability in the range of a few parts per billion. The general objective of this work was to develop and test a method to predict emissions of these T R S compounds from kraft pulp mills using a vapour-liquid equilibrium ( V L E ) model. This technique could then be utilized to optimize the process to minimize emissions, or to optimize the design and operation of T R S control systems. A mil l T R S sampling and testing program was conducted around the brown stock washing area of the Howe Sound pulp mi l l located in Port Mel lon, British Columbia. Over 90% of the sulphur emissions from the T R S compounds were in the form of dimethyl sulphide. This compound was subsequently chosen as a surrogate compound for modelling of T R S . Analytical testing was done on black liquor and various solutions containing the substances found in the highest concentrations in black liquor, including mixed sodium salts solutions to 6 wt% and lignin mixtures to 6 wt%. The V L E of dimethyl sulphide, in the form of an activity coefficient, for these solutions was determined using a headspace gas chromatographic method. The adjustable parameters that provided the best fit for the electrolyte non-random two-liquid (eNRTL) equation for the experimental activity coefficients were determined using statistical regression techniques. Simulation software, incorporating a V L E model using the e N R T L correlation, was used to construct a base-case heat and mass balance representing current operation of the Howe Sound brown stock washing process. This balance was used to evaluate potential process modifications and their effect on T R S emissions. T a b l e o f C o n t e n t s A b s t r a c t i L i s t o f T a b l e s vi L i s t o f F i g u r e s v i i i N o m e n c l a t u r e x i i G l o s s a r y x iv A c k n o w l e d g m e n t s xx i i i D e d i c a t i o n xxiv C h a p t e r 1 I n t r o d u c t i o n 1 C h a p t e r 2 B a c k g r o u n d 4 2.1 Kraft Pulp M i l l Non-Combustion Source T R S Emissions 4 2.2 TRS Formation 5 2.2.1 Hydrogen Sulphide Formation 6 2.2.2 Organic TRS Formation 7 2.2.3 Kinetics of Organic T R S Formation 9 2.2.4 T R S Formation in Black Liquor Evaporators 13 2.3 T R S Properties 13 2.4 TRS Health Effects 15 2.4.1 T R S Odour and Health Threshold Levels 18 2.4.2 Odour Detection Threshold 20 2.4.3 Odour Identification Threshold 20 2.4.4 Nuisance Odour Threshold 21 2.4.5 Health Effects Threshold 22 2.4.6 What to Make of These Threshold Levels? 22 2.5 Kraft Pulp M i l l Environmental Regulations 23 2.5.1 Environmental Guidelines in Canada 23 2.5.1.1 British Columbia 25 2.5.1.2 Alberta 25 . 2.5.1.3 Ontario 26 2.5.1.4 Quebec 27 ii 2.5.1.5 Other Provinces 27 2.5.1.6 Canadian National Pollutant Release Inventory (NPRI) 28 2.5.2 Environmental Guidelines in the U . S . A 28 2.5.3 Environmental Guidelines in Rest of the World 30 2.6 N C G Collection and Treatment Systems 31 2.7 Kraft M i l l Process Streams 37 2.7.1 Black Liquor Composition 37 C h a p t e r 3 L i t e r a t u r e R e v i e w 41 3.1 T R S Phase Equilibria Behaviour 41 3.1.1 Vapour-Liquid Equilibrium 41 3.1.2 Activity Coefficient Models 45 • 3.1.3 Factors Affecting T R S Systems 48 3.1.3.1 pH Effects on V L E 49 3.1.3.2 Inorganic Dissolved Solids Effects on V L E 51 3.1.3.3 Organic Dissolved Solids Effects on V L E 51 3.1.3.4 Other Effects on V L E 52 3.1.4 Henry's Constants and Activity Coefficients for T R S 52 3.1.5 T R S Vapour Pressure 58 3.2 Modell ing of the Kraft Pulping Process 58 3.2.1 Computer Software Modell ing Tools 60 3.2.2 Previous Work Modelling Non-Combustion T R S Emissions 61 3.3 T R S Emissions Factors 66 3.4 T R S Sampling and Testing Considerations 68 C h a p t e r 4 R e s e a r c h O b j e c t i v e s 71 C h a p t e r 5 M a t e r i a l s a n d M e t h o d s 72 5.1 Gas Chromatographic Equipment 72 5.2 Phase Equilibria Testing 74 5.2.1 Phase Equilibria Testing Method 74 5.2.2 Recovery of A lka l i Lignin From Black Liquor 76 5.2.3 Phase Equilibria Sample Solution Preparation 76 5.2.4 Phase Equilibria Testing Procedure 80 in 5.2.5 Testing of Solutions for Other Properties 82 5.3 M i l l Testing 83 5.3.1 The Howe Sound Pulp and Paper M i l l 84 5.3.2 M i l l Testing Method 86 5.3.3 Sample Collection Procedure 87 5.3.4 Sample Collection Considerations 88 5.3.5 M i l l Testing Procedure 89 5.3.6 M i l l Testing Considerations 91 5.4 Activity Coefficient Modelling 95 5.4.1 Regression Analysis of Phase Equilibria Testing Data 97 5.5 V L E Modelling 98 5.5.1 Phase Equilibrium Equation 99 5.5.2 Mole Balance Equation 100 5.5.3 Energy Balance Equation 100 5.5.4 Temperature (Boiling Point Rise) Equation 101 5.5.5 Solution of V L E Model 101 5.5.6 Commercial Software V L E Modules 104 Chapter 6 Results and Discussion 105 6.1 Phase Equilibria Testing using Published Data 105 6.2 Phase Equilibria Testing Results 110 6.2.1 Other Properties of Tested Solutions 113 6.2.2 Activity Coefficient as a Function of Conductivity 115 6.2.3 Activity Coefficient as a Function of Sodium Concentration 117 6.3 Phase Equilibria Testing Data Regression 121 6.3.1 Fitting of Henry's Constant Equation 121 6.3.2 Fitting of N R T L Equation for DMS-Water System 124 6.3.3 Comparison of N R T L Fit to Other Sources 126 6.3.4 Range of Validity for Infinite Dilution 128 6.3.5 Fitting of e N R T L Equation for DMS-Water-Sodium Salts System . . 128 6.3.6 Fitting of e N R T L Equation for H2S-, M M - and DMDS-Water -Sodium Salts Systems 131 iv 6.4 M i l l Sampling and Testing Results 133 6.4.1 Howe Sound Equipment Dimensions and Operating Conditions . . . . 140 6.4.2 Time to Equilibrium for D M S in Black Liquor 144 6.5 Modell ing of Howe Sound Equipment 146 6.5.1 Model Sensitivity Analysis 147 6.5.2 Equipment Modell ing Results for D M S 148 6.5.3 Equipment Modell ing Results for other T R S Compounds 152 6.5.4 Equipment Modell ing Analysis 154 6.6 M i l l Testing Data Compared to Emission Factors 160 6.7 Howe Sound M i l l Modelling 165 6.7.1 Base-Case Balance of Howe Sound Process 167 6.7.2 Predicted Effect on D M S Emissions from Process Changes 175 6.7.3 Reducing Brown Stock Washing D M S Emissions 180 6.8 The Future of N C G Collection Systems 182 C h a p t e r 7 C o n c l u s i o n s a n d F u t u r e W o r k 183 7.1 Conclusions 183 7.2 Future Work 187 R e f e r e n c e s 189 A p p e n d i x A U . S . E P A C l u s t e r R u l e S u m m a r y 206 A p p e n d i x B N C G C o l l e c t i o n a n d T r e a t m e n t S y s t e m s 209 A p p e n d i x C M a t l a b V L E M o d u l e P r o g r a m 226 A p p e n d i x D N C A S I M i l l T e s t i n g D a t a 235 A p p e n d i x E P h a s e E q u i l i b r i a T e s t i n g R a w D a t a 250 A p p e n d i x F H o w e S o u n d M i l l T e s t i n g R a w D a t a 261 A p p e n d i x G H o w e S o u n d M O P S H i s t o r i a n D a t a 286 v Lis t of Tables 2.1 Total organic T R S formed during kraft pulping of softwood (spruce) at a cooking temperature of 170°C for 1, 2, 3, and 4 hour cooks (kg S/tonne dry wood) . . . 9 2.2 Physical properties of selected T R S compounds 14 2.3 TRS health effects data 16 2.4 H 2 S threshold concentrations 19 2.5 Ontario POI and A A Q C limits 26 2.6 Typical combined C N C G composition 32 2.7 U .S . E P A typical T R S emissions for C N C G and D N C G sources (ppmv) 33 2.8 N C A S I measured and averaged T R S emissions for C N C G and D N C G sources(ppmv) 34 2.9 Components of weak black liquor solids (wt% on water-free basis) 38 2.10 Typical inorganic salt composition of black liquor 39 2.11 Elemental composition of softwood black liquor solids 40 3.1 Henry's constants for hydrogen sulphide in water 54 3.2 Henry's constants for hydrogen sulphide in various salt solutions 55 3.3 Henry's constants for methyl mercaptan in water 55 3.4 Henry's constants for methyl mercaptan in various salt solutions 55 3.5 Henry's constants for dimethyl sulphide in water 56 3.6 Henry's constants for dimethyl sulphide in various salt solutions 56 3.7 Henry's constants for dimethyl disulphide in water 57 3.8 Henry's constants for dimethyl disulphide in various salt solutions 57 3.9 Constants for vapour pressure equation 3.21 58 3.10 U .S . E P A Emissions Factors for Chemical Wood Pulping 67 5.1 Sodium salt composition used for preparing salt solutions 78 5.2 Dimethyl sulphide solutions used for phase equilibria testing 79 6.1 Methanol(I) in water(j) parameters for N R T L equation 106 6.2 Vapour-liquid equilibrium activity coefficients for D M S solutions tested I l l 6.3 pH, total dissolved solids, ash, and sodium content of solutions tested 114 6.4 Conductivities of alkali lignin mixtures, sodium salt solutions, and black liquor . . . . 115 6.5 Regressed Henry's constants for dimethyl sulphide in solutions tested 121 vi 6.6 Regressed N R T L equation parameters for DMS(i)-water(j) system 124 6.7 DMS( i ) in waterfj) parameters for N R T L equation from other sources 126 6.8 Regressed e N R T L equation parameters for DMS(i)-water(j)-black liquor sodium salts(ca) system 129 6.9 DMS-black liquor activity coefficients 130 6.10 Regressed e N R T L equation parameters for H2S-, M M - and DMDS(i)-water(j)-sodium salts (ca) system . 132 6.11 M i l l test results for September 16, 2005 134 6.12 M i l l test results for September 21, 2005 134 6.13 M i l l test results for September 22, 2005 135 6.14 M i l l test results for September 28, 2005 135 6.15 M i l l test results for September 29, 2005 135 6.16 M i l l test results for September 30, 2005 136 6.17 M i l l test results for November 22, 2005 136 6.18 M i l l test results for November 23, 2005 137 6.19 M i l l test results for November 24, 2005 137 6.20 M i l l test results for November 30, 2005 138 6.21 M i l l test results for December 1,2005 138 6.22 M i l l test results for December 2, 2005 139 6.23 Equipment dimensions and operating conditions 141 6.24 Root mean square error ( R M S E ) of V L E model vent vapour concentration prediction 154 6.25 V L E model correction factor for systematic error 155 6.26 Brown stock washing D M S emission factors 163 6.27 Vapour sample testing data for November 23, 24, 30, and December 1, 2005 168 6.28 Liquid sample testing data for November 23, 24, 30, and December 1, 2005 169 6.29 M O P S historian data for November 23, 24, 30, and December 1,2005 171 6.30 Predicted effect on Howe Sound brown stock washing D M S emissions as a result of operational or equipment changes 176 vii Lis t of Figures 2.1 Organic T R S formation as a function of time for kraft pulping of softwood (spruce) at 170°C and a sulphidity level of 30.5% 10 2.2 Organic T R S formation as a function of sulphidity for kraft pulping of softwood (spruce) at 170°C for 4 hours 11 2.3 Organic T R S formation as a function of H-factor for kraft pulping of softwood (loblolly pine) at 170°C 12 3.1 Relative volatility of common kraft mil l contaminants 45 3.2 Dissociation of hydrogen sulphide and methyl mercaptan 50 3.3 Effect on activity coefficient of methanol due to the presence of dissolved inorganic and organic matter in black liquor 63 3.4 Comparison of methanol vent stack model predictions with mi l l measurements for various process equipment 64 3.5 Comparison of G E M S prediction versus mi l l data for M i l l E for methanol concentration in the vapour phase 65 3.6 Comparison of G E M S prediction versus mi l l data for M i l l E for T R S concentration in the vapour phase 66 5.1 Typical gas chromatograph output 73 5.2 Time for a 1 ppm (mol) solution of D M S in 10 ml of water in a 24.1 ml sample vial at 90°C to reach equilibrium 81 5.3 Howe Sound mi l l brown stock washing area overview showing liquid sample (LS) and vapour sample (VS) point locations 85 5.4 Gas sampling equipment, including phosphoric acid impinger, Teflon tubing, 3-way valves, vacuum pump and Tedlar bag 87 5.5 Typical sample point (blow tank vent) being tested using the velometer 88 5.6 Typical liquid standard D M S calibration curve at the testing temperature of 80°C . . . 91 5.7 Degradation of TRS liquid standard held at 20°C and at an initial concentration o f30ppm(mol ) 92 5.8 Degradation of T R S gas standard held at 80°C and at an initial concentration of 5 ppm (mol) 93 v i i i 5.9 V L E module used to predict emissions of volatile compounds 98 5.10 Summary of V L E module calculation block non-linear equations that require simultaneous solution using a Newton-Raphson or equivalent technique 102 5.11 V L E Module vapour and liquid by-passes 103 6.1 Vacuum drum washer filtrate tank vent stack measured methanol concentration compared to predicted concentration based on outlet filtrate concentration 107 6.2 Decker washer hood vent stack measured methanol concentration compared to predicted concentration based on inlet wash liquor concentration 108 6.3 Diffusion washer vent stack measured methanol concentration compared to predicted concentration based on inlet wash liquor concentration 109 6.4 Weak black liquor tanks vent stack measured methanol concentration compared to predicted concentration based on liquor outlet concentration 110 6.5 Temperature effect on DMS-water activity coefficient at various alkali lignin concentrations (solutions 4 and 5) 112 6.6 Temperature effect on DMS-water activity coefficient at various sodium salts concentrations (solutions 6, 7, and 8) 112 6.7 Temperature effect on DMS-water activity coefficient at various total dissolved solids concentrations in black liquor (solutions 11 and 12) 113 6.8 Conductivity of alkali lignin mixtures, sodium salts solutions, and black liquor as a function of sodium concentration 116 6.9 Activity coefficient at 90°C of alkali lignin mixtures, sodium salt solutions and black liquor, as a function of sodium concentration 117 6.10 Activity coefficient at 90°C (up to 0.5 wt% sodium) of alkali lignin mixtures, sodium salt solutions and black liquor, as a function of sodium concentration 118 6.11 Sodium concentration effect on DMS-water activity coefficient for sodium salts and black liquor at 20°C 119 6.12 Sodium concentration effect on DMS-water activity coefficient for sodium salts and black liquor at 50°C 119 6.13 Sodium concentration effect on DMS-water activity coefficient for sodium salts and black liquor at 70°C 120 ix 6.14 Sodium concentration effect on DMS-water activity coefficient for sodium salts and black liquor at 90°C 120 6.15 Henry's constant for D M S in water as a function of temperature 122 6.16 Henry's constant for D M S in water as a function of sodium salts concentration at 40°C and 70°C 123 6.17 DMS-water activity coefficient experimental data and N R T L equation best fit (R 2 = 0.95) 125 6.18 Activity coefficients determined from Olsson and Zacchi (2001) N R T L parameters, U N I F A C parameters and best fit to experimental data and holding a^  and a^  constant at zero (R 2 = 0.26) 127 6.19 DMS-water-sodium salts activity coefficient experimental data and e N R T L equation best fit (R 2 = 0.95) 129 6.20 Comparison of activity coefficient determined from e N R T L equation to black liquor experimental data (solutions 9 to 13) 131 6.21 Activity coefficients determined from the e N R T L equation for the T R S compounds at 80°C 133 6.22 Decker washer at the Howe Sound pulp mi l l 143 6.23 Time required for D M S phase equilibrium to be established in a black liquor sample h e l d a t 8 0 ° C 145 6.24 Sensitivity analysis for prediction of D M S vapour phase concentration 148 6.25 Measured versus predicted D M S concentrations for decker washer filtrate tank N C G vent 149 6.26 Measured versus predicted D M S concentrations for 2 n d stage diffusion washer N C G vent 150 6.27 Measured versus predicted D M S concentrations for blow tank N C G vent 151 6.28 Measured versus predicted D M S concentrations for decker washer hood N C G vent . 152 6.29 Measured versus predicted D M D S concentrations for decker washer filtrate tank N C G vent 153 6.30 Comparison between decker washer filtrate tank measured D M S vapour concentration and predicted equilibrium value using the V L E model, and predicted value corrected for systematic error 156 x 6.31 Comparison between blow tank measured D M S vapour concentration and predicted equilibrium value using the V L E model, and predicted value corrected for systematic error 157 6.32 Comparison between diffusion washer filtrate tanks measured D M S vapour concentration and predicted equilibrium value using the V L E model, and predicted value corrected for systematic error 158 6.33 Comparison between decker washer hood measured D M S vapour concentration and predicted equilibrium value using the V L E model, and predicted value corrected for systematic error 159 6.34 Total measured H 2 S emissions (left axis) compared to predicted emissions using emission factor (right axis) for brown stock washing process (kg of sulphur per day) 161 6.35 Total measured organic T R S ( M M , D M S and D M D S ) emissions (left axis) compared to predicted emissions using emission factor (right axis) for brown stock washing process (kg of sulphur per day) 162 6.36 Decker filtrate tank emissions based on measured data and predicted using V L E correlations and emission factors 164 6.37 Comparison between decker washer filtrate tank measured D M S vapour concentration and predicted equilibrium value using the V L E model, and predicted value using emission factor 165 6.38 Base-case heat and mass balance for the brown stock washing area of the Howe Sound mi l l 173 6.39 Detailed view of a blow tank C A D S i m V L E module heat and mass balance 174 6.40 Base-case D M S mass balance for the brown stock washing system of the Howe Sound mi l l ( D M S in kg/day as sulphur) 175 xi Nomenclature - Activity of component i a, c - Components (anion, cation only, respectively) a, b - N R T L equation parameters (temperature dependency for r) B P R K Boil ing point rise C - Ionic charge (absolute value) C1.. .C5 - Extended Antoine vapour pressure equation parameters ca - Salt (composed of anion, a, and cation, c) C P - Specific heat Co w t % Stock consistency f kPa Fugacity F mol/s Feed, inlet flow G - N R T L equation parameter H J/mol Enthalpy A s o l H J/mol Enthalpy of solution - Components (any species, molecular and ionic) k H (mol/mol)/MPa Henry's law constant K a mol/m 3 Acid dissociation constant K - Distribution coefficient . L mol/s Liquid, outlet flow m, m ' - Components (molecular species only) n - Number of components 0 - Standard state P kPa Absolute pressure, vapour pressure psat kPa Saturation or vapour pressure pH - -log[H +] p K a - -log(Ka) R = 8.314 kPa m 3 /mol /K Universal gas constant S - Weight percent solids in a liquid sat _ Saturation state x i i T K Temperature 'psat K Saturation temperature T D S w t % Total Dissolved Solids V mol/m 3 Molar volume of liquid V mol/s Vapour, outlet flow X - Liquid phase mole fraction X - Effective liquid phase mole fraction - Mole fraction undissociated y - Vapour phase mole fraction z - Feed flow mole fraction a - Relative volatility / N R T L equation parameter Y - Liquid phase activity coefficient <f> - Vapour phase fugacity coefficient r - N R T L equation parameter Xll l Glossary A A Active A lka l i A A Q C Ambient A i r Quality Criteria. Ontario government policy. A A Q S Ambient A i r Quality Standards. Canada government policy. A C G I H American Conference of Government Industrial Hygienists. A A Active A lka l i . Concentration of N a O H plus Na 2 S , expressed as N a 2 0 . A D T P A i r Dry Tonne Pulp A I H A American Industrial Hygiene Association. B A D T Best Available Demonstrated Technology. Alberta government policy. B D T D Bone Dry Tonne Per Day B O D Biochemical Oxygen Demand. C A D S i m Modeling and simulation steady-state and dynamic software package supplied by Aurel Systems of Burnaby, B . C . Designed specifically for the pulp and paper industry. C A S Chemical Abstracts Service. Compilers of a database of chemical substance information. A C A S registry number is assigned to a substance when it enters the database. They are assigned in sequential order to unique, new substances identified by C A S scientists. xiv C C M E Canadian Council of the Ministers of the Environment. C C O H S Canadian Centre for Occupational Health and Safety. C E P A Canadian Environmental Protection Act. Chip B i n N C G liberated from the chip bin. Typically dealt with separately from C N C G and N C G D N C G as it is a transitional source that can be below the L E L or above the U E L . Cluster Rule U .S . E P A regulations promulgated in 1998 and with full compliance due by 2006. To regulate the release of H A P to the environment. For the pulp and paper industry, the controlling documents include M A C T I for non-combustion sources and M A C T II for combustion sources. C N C G Concentrated Noncondensible Gases. Also referred to a L V H C N C G . Malodorous gases collected in a kraft pulp mil l that theoretically wi l l be above the U E L . Typically consists of gases collected from the black liquor evaporator and from the digester relief gas condenser. Co Consistency. For example, this may refer to the consistency of the stock in the product from the digester. Often referred to as percent consistency or % C o . C W S Canada-Wide Standards. D F Dilution Factor. A term used to quantify washing of pulp stock. Equals the mass of wash waster over bone-dry mass of pulp. D R Displacement Ratio. A term used to quantify washing of pulp stock. Equals the mass of dissolved solids removed from stock over the maximum possible amount available to be removed. xv D M S Dimethyl sulphide. (CH 3 ) 2 S. A component of T R S . Also referred to as methyl sulphide, D M S and R S R . D M D S Dimethyl disulphide. (CH 3 ) 2 S 2 . A component of T R S . Also referred to as methyl disulphide, D M D S , and RSSR. D N C G Dilute Noncondensible Gases. Also referred to a H V L C N C G . Malodorous gases collected that wi l l theoretically be below the L E L . Typically consists of gases collected from tankage and washers. DS Dissolved Solids. For example, this may refer to the dissolved inorganic or organic matter in black liquor. Often referred to as percent DS or % D S . E A Effective Alka l i e N R T L Electrolyte N R T L E O S Equation of State. E P A Environmental Protection Agency. U.S . government organization. E P E A Environmental Protection and Enhancement Act . Alberta government policy. E R P G Emergency Response Planning Guidelines. ESP Electrostatic Precipitator. Particulate capture device typically used on combustion vent sources such as recovery boilers. G C Gas Chromatograph. H B L Heavy Black Liquor xvi H-factor A single physical variable that represents the net delignification effect of both cooking time and temperature during the kraft pulping process. H 2 S Hydrogen sulphide. A component of TRS . H A P Hazardous A i r Pollutants. Consists of 188 contaminants, to define air emissions for regulation of industry by the E P A through the Cluster Rule. H A P is a general grouping of chemicals that have been identified as causing serious illness, including cancer. H S G C Headspace gas chromatographic. H V L C High Volume Low Concentration N C G . Also called D N C G . ICP Inductively Coupled Plasma IPST Institute of Paper Science and Technology. U .S . organization located in Atlanta, Georgia. L C 5 0 Concentration of toxin that wi l l result in a 50% mortality rate in a set exposure time. L E L Lower Explosive Limit . The concentration limit below which there are insufficient combustibles to sustain combustion. L L E Liquid-Liquid Equilibrium. L V H C L o w Volume High Concentration N C G . Also called C N C G . M A C T Maximum Available Control Technology. xvi i M e O H Methanol. Methanol Also known as methyl alcohol, wood alcohol or M e O H . Formed, mainly in the digester, primarily from alkaline hydrolysis of 4 -0 methyl glucuronic acid residues in hemicelluloses and to a smaller extent from demethylation of lignin. A V O C and included in the definition of H A P . It is used as a surrogate for H A P to determine Cluster Rule compliance. p,m / p,g One one-millionth of a metre / gram. M M Methyl mercaptan. C H 3 S H . A component of T R S . Also referred to as M M , M e S H , R S H , and methanethiol. M O E Ministry of the Environment. Canadian government ministry. M O P S Mi l lwide Optimization System. Operating data historian system. mS 1000 m S = 1 Siemen M S A Mean Spherical Approximation. N A A Q O National Ambient A i r Quality Objectives. N A C National Advisory Committee. N C A S I National Council for A i r and Stream Improvement. U .S . pulp and paper industry research organization. N C G Noncondensible Gases. A general term for Kraft mi l l odorous gases liberated from non-combustion process equipment in the mi l l . Composed of T R S , V O C (methanol, terpenes, etc.), air, and water vapour. Subcategories include C N C G , xviii D N C G and Chip B i n N C G . N C G sources are often collected into one or several systems for disposal by chemical modification or incineration. NIST National Institute of Standards and Technology. U .S . federal agency. N O x Nitrous Oxides. Typically formed during combustion. Mainly nitrous oxide and nitrous dioxide with negligible nitrous oxide. Smog and acid rain causing air emission. NPRI National Pollution Release Inventory. Canadian federal government program which requires all industrial facilities to report their emissions to air, land and water. N R T L Non-Random Two-Liquid. Activity coefficient model. O S H A Occupational Safety and Health Administration. U.S. Government organization. Paprican Pulp and Paper Research Institute of Canada. Paptac Pulp and Paper Technical Association of Canada. P M Particulate matter. P M 2 5 Particulate matter below 2.5 pm aerodynamic diameter. P M 1 0 Particulate matter below 10 pm aerodynamic diameter. POI Point of Impingement. Ontario government policy. ppb / ppm Parts per bill ion / mill ion, typically on volume basis when discussing gases. xix P v M S E Root Mean Squared Error R S C Reduced sulphur compound. R S H Methyl mercaptan R S R Dimethyl sulphide R S S R Dimethyl disulphide R T O Regenerative Thermal Oxidizer. Used to treat thermally oxidize N C G . S B L O x Strong Black Liquor Oxidation. S O G Stripper Off Gases. Foul gases liberated from a foul condensate steam stripping system. Consists mainly of methanol, water vapour and N C G . S O x Sulphur Oxides. Mainly sulphur dioxide and sulphur trioxide. Smog and acid rain causing air emission. STFI Swedish Pulp and Paper Research Institute. Located in Stockholm, Sweden. Sulphidity The percentage ratio of Na 2 S to active alkali, expressed as N a 2 0 . Sweep air A i r deliberately drawn into process equipment to "sweep" volatile compounds into the N C G collection system. Through stripping action, sweep air can increase emissions of volatile compounds. Undesired air entering is referred to as tramp air.. Tappi Technical Association of the Pulp and Paper Industry. U .S . industry organization. xx T D S Total Dissolved Solids. For example, this may refer to the inorganic and organic matter in black liquor. Often referred to as percent TDS or % T D S . Tramp air Unwanted air drawn into process equipment through openings such as leaking flanges, open manholes covers and poorly sealed overflows or inspection hatches. Tramp air w i l l increase the volumetric flow of N C G collected from the equipment. Through stripping action, tramp air can increase emissions of volatile compounds. Also referred to as sweep air i f deliberately introduced to the process equipment. T R S Total Reduced Sulphur. A general term used to describe Kraft mi l l odorous bivalent sulphur compounds. Typically defined as including hydrogen sulphide, methyl mercaptan, dimethyl sulphide and dimethyl disulphide. Other compounds found in lower concentrations include dimethyl trisulphide, dimethyl tetrasulphide, ethyl sulphide, ethyl disulphide, ethyl methyl disulphide, 2-furfuryl methyl sulphide, and allyl methyl sulphide. T T N Technology Transfer Network. Operated by the U .S . E P A . Turpentine Sulphate turpentine is a volatile oi l present in wood and liberated during the Kraft pulping process. It is highly combustible. U E L Upper Explosive Limit . The concentration limit above which there is insufficient oxygen to sustain combustion. U N I F A C U N I Q U A C Functional-group Activity Coefficient. A group contribution method that combines the solution of functional groups concept and the U N I Q U A C model for thermodynamic design of highly non-ideal multi-component systems. U N I Q U A C Universal Quasi-Chemical Activity Coefficient. xxi V L E Vapour-Liquid Equilibrium. V O C Volatile Organic Carbon compounds. W B L Weak Black Liquor W B L O x Weak Black Liquor Oxidation. W C B Workers Compensation Board. W G A Q O G Working Group on A i r Quality Objectives and Guidelines. W H O World Health Organization. xxi i Acknowledgments This work could not have been completed without a lot of help and support from many people. First and foremost I would like to thank the members of my supervisory committee, Drs. Richard Branion, Sheldon Duff, Dusko Posarac and Brian Blackwell for their guidance, support and encouragement. Thank you to the fellow students in my group, Nicole Bennett, Charles Liao and John Ruffel, not only for assistance and encouragement, but for providing good company during those long tedious days in the lab. And special thanks to Steve Helle for his advice and assistance with lab procedures; it would have taken a lot longer without someone who knew their way around a G C . Thank you to the Natural Sciences and Engineering Research Council of Canada who partly funded this research via a Industrial Postgraduate Scholarship. Thank you to Bruce Der and A . H . Lundberg Systems Limited for providing time and support for completing this work, and for financial support as the industrial sponsor of the scholarship. Thank you to Howe Sound Pulp and Paper for providing lab space, equipment, assistance in collecting samples, and access to all data required for this work. In particular, Gerry Pageau for coordinating the mi l l aspects of this work and co-authoring a paper, and George Zhang, who provided assistance in collection and testing of samples. Thank you to Paprican for the loan of testing equipment and advice on procedures. Thank you to Larry Wasik and Aurel Systems Limited for providing copies of C A D S i m Plus to the University of British Columbia for research purposes. Thank you to my M o m and Dad for everything. M y Dad passed away in 2002 just before I began the PhD program, but he and my M o m were instrumental in putting me on the path that led to this challenge and their influence during my early years instilled in me the focus and patience required to complete it. They always said I would become an engineer; they were right. A n d finally a very special thank you to Susie, who stood by me in solidarity these last five years. I w i l l endeavour to support her as much as she did me as she works to complete her own doctorate. xxiii Dedication To my Dad, Birger Wolder Jensen xxiv Chapter 1 Introduction Chapter I: Introduction Kraft pulp mills have historically been associated with foul odour. Many of us may remember the obnoxious odour that was apparent whenever you passed near apulp mi l l , with it often being noticeable from 10 or 20 kilometres away. For a decade from the early 1960's, a significant amount of research was conducted on the origin of this foul odour. Much of the impetus for the research was to lower the overall odour effects of a kraft mi l l , which were found to be caused mainly by reduced sulphur compounds generated as a by-product of the kraft pulping process. The odorous sulphur compounds originate from combustion sources such as the recovery boiler and lime kiln, and from non-combustion sources such as the digester, brown stock washing, and black liquor processing equipment, the latter being the focus of this research. Emissions for non-combustion sources include the smog and odour-causing volatile organic compounds (VOCs) and the odorous total reduced sulphur (TRS) compounds. These sulphur gases are the main component of what give a kraft mi l l its distinctive odour, which has been described as being like rotten eggs or rotting cabbage, but is typically just called "that pulp m i l l " smell. In the late 1960's, pulp mills began to install noncondensible gas (NCG) systems, to reduce the impact of non-combustion emissions (Sarkanen et al., 1970). These were simple end-of-pipe treatment systems, consisting of collection piping and a fan, designed to collect the vents from the two or three lowest volumetric sources containing the highest concentration of odour-causing emissions and deliver them to treatment: usually alkaline scrubbing or incineration. During the last decade a number of mills, typically newly-constructed mills or those undergoing major expansions, installed a second, much larger, odour control system. High volume N C G systems collect vents from up to thirty sources containing lower, but still significant concentrations, of the odour causing sulphur compounds. The gases are delivered to treatment, typically incineration in an existing mil l boiler. Before beginning this doctorate, the author worked for over ten years for A . H . Lundberg Systems Limited ( A . H . Lundberg), designing kraft mil l N C G systems. Section 2.6 and Appendix B contain descriptions of the design of N C G systems, and the extent that they exist in kraft mills; many of these details are not available in the open literature. These descriptions are primarily based 1 Chapter 1: Introduction on the author's personal experience, so they are un-referenced. Installation of N C G systems was often driven by. complaints from the general population, usually those living in the vicinity of the mi l l . A s ambient odour concentrations were reduced, the residents adapted to the new conditions, with a resulting strong intolerance for a return to the higher concentrations. Efficient operation of these systems became important because even a short period of venting these N C G s would often draw a litany of complaints from the nearby residents. The human nose can detect the odorous sulphur compounds at concentrations of a few parts per billion. Thus, a reduction of odour to undetectable concentrations in the vicinity of a mi l l is probably not achievable. Nonetheless, incremental reduction has been an ongoing focus over the years. Today, there is a wide range of installed odour control systems, with some mills, often in isolated areas, having none whatsoever, while others, usually in more populated areas, have elaborate systems that collect and treat numerous odorous vent sources. Over time, the focus of kraft pulp mi l l air emissions has gradually shifted from basic "nuisance" odour reduction to include a wide variety of hazardous air pollutants with potentially serious effects to human heath. Historically there have not been any consistent requirements between Canadian provinces or between U.S . states to control odour levels; thus, the installation of N C G systems has been highly location specific. This is changing in the U.S . with the introduction of new blanket environmental regulations, referred to as the Cluster Rule (U.S. E P A , 1998), that standardize requirements for reduction in air emissions for all U .S . kraft mills. In Canada, community complaints still appear to be the main driving force behind odour reduction, with each mil l under more or less pressure depending on its vicinity to a population centre. In a recent Paprican survey of 24 Canadian mills, 21 mills reported present or recent problems with odorous T R S emissions, with 15 of these expressing a need for additional information or research in the field (Allen, 1998a). Research continues to better understand kraft mil l odour, including this work attempting to model emissions from process systems within the mi l l . Other areas of research include modelling emissions from effluent treatment systems (Crawford et al., 2006) and dispersion modelling to better predict ambient air quality near pulp mills (O'Conner and Ledoux, 2002; Freeburn and Redmond, 1998; Jarvensivu et al., 1997). Collection and treatment of non-combustion sources can be capital intensive, with the need 2 Chapter 1: Introduction to install N C G collection systems, scrubbers, incineration systems, and tall stacks. These systems can also have significant associated operating costs such as loss of cooking chemicals, consumption of scrubbing chemicals, additional requirements for power, steam, water and fossil fuel, maintenance issues, and lost pulp production due to downtime caused by outages of these systems. More cost effective methods to reduce odorous emissions may be found either through operational or equipment changes to reduce the formation of odorous compounds, or through modifications to the design and/or operation of the vent source equipment operation to reduce the quantity of odorous gas released to atmosphere. By providing a method to model the release of T R S compounds from process equipment, the design of N C G systems could be optimized and the capital and operating costs could be minimized. One method to predict emissions is the use of simulation software to construct heat and mass balances of the process. Balances can be constructed using commercially available simulation software packages such as W i n G E M S (Pacific Simulation) and C A D S i m Plus (Aurel Systems), which are widely used in the North American pulp and paper industry. Often, the mil l heat and mass balance capabilities of commercially available software are limited to the liquid and solid phases, mainly because these cover the majority of the pulping process. The objective of this work is to extend these balances to include N C G emissions from process equipment and tankage. Emissions can be modelled by incorporating an emissions module, based on vapour-liquid equilibrium ( V L E ) , into the balances. These enhanced heat and mass balance calculations could then be used to optimize the design and operation of N C G systems and to improve emissions reporting. To this end, two testing programs were completed. Analytical testing was conducted in the lab to determine phase equilibria behaviour of T R S compounds in kraft pulp mi l l process liquids. A mi l l sampling and testing program was conducted at the Howe Sound Pulp and Paper Limited Partnership (Howe Sound) mi l l located in Port Mellon, British Columbia, to determine the concentration of T R S compounds in vapour and liquid streams in the brown stock washing area. This data was used to test and validate a V L E emissions model. To demonstrate an industrial application of this model, a base-case heat and mass balance of the brown stock washing area of the Howe Sound mil l was constructed, and the V L E emissions module was incorporated to model N C G emissions. 3 C h a p t e r 2 B a c k g r o u n d Chapter 2: Background 2.1 K r a f t P u l p M i l l N o n - C o m b u s t i o n S o u r c e T R S E m i s s i o n s A i r emissions from kraft pulp mills are often classified into one of two categories, combustion and non-combustion sources. Combustion sources include the recovery boiler, lime kiln and oi l - , natural gas- and wood-waste-fired power boilers. Non-combustion source emissions include vents from the digester, brown stock washing, evaporators, and tankage for unwashed stock, black liquor and foul condensate. Non-combustion source emissions of concern include the odorous total reduced sulphur (TRS) compounds, also referred to as reduced sulphur compounds (RSCs), along with the smog and odour causing volatile organic compounds (VOCs) . The TRS compounds encountered in the highest concentrations include hydrogen sulphide (H 2 S) along with the organic compounds, methyl mercaptan ( C H 3 S H or M M ) , dimethyl sulphide ((CH 3 ) 2 S or D M S ) , and dimethyl disulphide ( (CH 3 ) 2 S 2 or D M D S ) (Niemala, 2001; Cook and Hoy, 2003). These four compounds have historically been used by researchers when discussing T R S in kraft mills; this has generally been accepted as a fairly accurate simplification. Niemala (2001) and Cook and Hoy (2003) identify other T R S compounds besides the four typically studied, although these are usually found in much lower concentrations. These include dimethyl trisulphide, ethyl sulphide, ethyl disulphide, methyl propenyl sulphide, ethyl methyl disulphide, allyl methyl sulphide, allyl methyl disulphide and a previously unidentified disulphide of the form C 3 H g S 2 (or isomer). O f the numerous V O C s present, methanol typically makes up the largest fraction, followed by acetone and methyl ethyl ketone (Zhu et al., 1999a). Other emissions include wood extractives such as turpentine, chemicals formed during the pulping process, such as ammonia, and chemicals formed during the bleaching process, such as chloroform. Most kraft pulp mills have installed some form of odour control system to collect and treat non-combustion source T R S emissions. These systems are commonly referred to as noncondensible gas (NCG) collection and treatment systems. N C G is a general term used by the industry for those gases released from non-combustion process equipment and tankage around a kraft pulp mil l during 4 Chapter 2: Background normal production. The design of these systems, discussed in detail in Section 2.6, has historically focussed on reduction of the main odour contributors, the TRS compounds, which are generated in undesired side reactions in the kraft pulping process (Yoon et al., 2003). 2 .2 T R S Formation The formation of the organic sulphur compounds occurs mainly in the digester. The kraft pulping process consists of "cooking" wood chips in white liquor, a solution of sodium hydroxide and sodium sulphide in water, in a digester at about 170°C for three to four hours, using a batch or continuous process. The hydroxide ion and hydrosulphide ion (formed from the hydrolysis of sodium sulphide) in the white liquor attack and dissolve the lignin, the "glue" which holds the cellulose fibres together in the wood. The digester product, called brown stock, consists of a mixture of pulp stock and black liquor. The black liquor is composed of the residual pulping chemicals, dissolved lignin, extractives, and unwanted reaction products. The main undesirable reaction products include methanol and the organic T R S compounds. The pulp stock is separated from black liquor in the brown stock washing process, with the pulp stock directed to bleaching, and the black liquor directed to the recovery process, where it is converted back to white liquor. A series of papers on the topic of TRS formation in kraft digesters was published from the mid 1960's to the early 1970's in Tappi Journal by groups based at the University of Washington and the University of Maine (McKean et al., 1965; 1967; 1968: Douglass and Price, 1966; 1968; Douglass et al., 1969; Sarkanen et al., 1970; Wilson and Hrutfiord, 1971; Tsuchiya and Johanson, 1972; Wilson et al., 1972). These groups carried out a series of batch micro-kraft cooks using a 7.5 m L stainless steel digester charged with 1.0 gram of wood and 4.0 m L of cooking liquor at temperatures ranging from 150 to 180°C. At the completion of the cook, the contents were acidified and vapour samples were taken and analysed using a gas chromatograph (GC). Their results, along with those from other researchers, wi l l be discussed below. At about the same time, Andersson and Bergstrom (1969; 1970) heading up a group at the Royal Institute of Technology in Stockholm Sweden, published data on the same topic. They also used one gram of wood and a liquor charge at a 4:1 ratio in 10 m L glass ampoules at a cooking temperature of 70 to 170°C. The organic TRS compounds were extracted into carbon tetrachloride 5 Chapter 2: Background and analysed using a G C . Their results agreed quite closely with those from the American groups. In the experiments conducted by all of these groups, the amount of wood used was very small, and there is some question about how representative the samples would have been, especially when compared to the much larger wood chips used as feed in the industrial process. There were also questions about scale-up and applicability to the continuous process. A couple of decades later, a group based mainly at the Institute of Paper Science and Technology (IPST) in Atlanta published another series of papers on the same topic (Zhu et al., 1999a; 1999b; 2000b; 2001; 2002; Yoon et al., 2003). Their initial work focussed on methanol formation, while their latter work focussed on T R S formation. They used 50 grams of oven dried wood (mainly loblolly pine) at a 4:1 liquor charge with 18% active alkali on wood and cooking liquor sulphidities ranging from 0 to 30%. Cooking times were varied but cooking temperatures were held at 170°C. The concentrations of methanol and T R S were measured using an indirect headspace G C method (Chai et al., 1998; 2000). They used a simple linear regression method to determine the relationship between variables and to determine the parameters affecting T R S formation. These parameters were expressed in terms of the H-factor, a pulping expression that combines cooking time and temperature into one variable (Yoon et al., 2003). A l l of the research conducted on TRS formation was done using a batch process. Continuous digesters are designed as pseudo plug-flow type reactors, and operate under cooking time and temperature conditions, similar to those used for batch digesters; thus, the data for batch digesters were also expected to be relevant to this process. The conclusions reached by these researchers on the mechanism and rate of TRS formation are summarized in Sections 2.2.1 to 2.2.3. 2.2.1 Hydrogen Sulphide Formation O f the TRS compounds studied, hydrogen sulphide is unique in that it is not an organic compound and it is not formed as a reaction product. The potential for the presence of hydrogen sulphide exists due to its equilibrium with the hydrosulphide ion. The first step in this process is the hydrolysis of sodium sulphide to sodium hydrosulphide: N a 2 S + H 2 0 >NaHS+ N a O H (2.1) 6 Chapter 2: Background The sodium hydrosulphide dissociates in the liquor according to the equilibrium: NaHS< >Na + + H S " (2.2) The hydrosulphide ions in the liquor exist in equilibrium with hydrogen sulphide: H S " + H 2 0 < > H 2 S + O H ~ (2.3) Due to the alkaline conditions of the kraft pulping process, this equilibrium lies almost completely to the left. Equilibrium conditions wi l l be discussed in detail in Section 3.1.3.1. In strongly alkaline conditions, the hydrosulphide ion can further dissociate to the sulphide ion: H S " + O H " < >S= + H 2 0 (2.4) 2 . 2 . 2 Organic T R S Formation Methyl mercaptan and dimethyl sulphide are produced through a series of reactions in the digester, and also to a lesser extent in other process equipment. Dimethyl disulphide is produced by the oxidation of methyl mercaptan, mainly downstream of the digester where the oxygen content is higher (Douglass and Price, 1966). The rates of formation of methyl mercaptan and dimethyl sulphide are a function of sulphide and methoxyl concentration and degree of delignification, along with the related factors, cooking time, temperature, and pH (Andersson and Bergstrom, 1969). The reactions that form these two compounds wi l l only proceed to a measurable degree above about 140°C; thus, the digester, which typically operates at about 170°C, provides ideal conditions for the formation of methyl mercaptan and dimethyl sulphide (McKean et al., 1968). The mechanism for TRS formation has been studied quite extensively and is well understood, with a good overview provided by the University of Washington group (Sarkanen et al. 1970). Methyl mercaptan is formed by the reaction with a lignin methoxyl group and the sulphide and hydrosulphide ions: lignin • O C H 3 + S = >CH 3S~ + lignin • O" (2.5) lignin OCH3 + S H " >CH 3 SH+ l ig in in -O" (2.6) 7 Chapter 2: Background The mercaptan is weakly acidic and is soluble in the strongly alkaline kraft liquor, so the two reaction products exist in equilibrium (Shih et al. 1967b): C H 3 S H + O H " < >CH,S" + H 2 0 (2.7) With the mercaptide ion present, dimethyl sulphide is formed by a sequential reaction with a lignin methoxyl group: lignin - O C H 3 + C H 3 S " > C H 3 S C H 3 + lignin - 0~ (2.8) Initially, the mercaptan is formed at a constant rate as the sulphide and hydrosulphide ions are responsible for the initial attack on the lignin methoxyl group. As the concentration of the resulting mercaptide ion builds up, it attacks the methoxyl group producing dimethyl sulphide and as the reaction proceeds, the mercaptide ion is eventually consumed as fast as it is formed. At this point in the cook, reported to be at about the three hour mark (Douglass and Price, 1966), the mercaptan concentration wi l l level off while the dimethyl sulphide concentration steadily increases. Dimethyl disulphide, the last T R S compound of interest, is formed by the oxidation of the mercaptide ion (Cooper, 1974): 1 2 C H 3 S " + - 0 2 + H 2 0 > C H 3 S S C H 3 + 2 0 H " (2.9) This last reaction does not occur to any significant extent in the digester, compared to the formation of methyl mercaptan and dimethyl sulphide. Douglass and Price (1966) speculated that this likely results from the lack of oxygen in the cooking zone of the digester due to it being consumed in other reactions before the digester contents reach 140°C. Even though there is some oxygen present in the digester feed from residual air carried through with the wood chips, the hydrosulphide ion wi l l readily consume this oxygen through oxidation to thiosulphate, then onto sulphite, and eventually to sulphate (Cooper, 1974): 2HS" + 2 0 2 > S 2 0 ; + H 2 0 (2.10) S 2 0 ; + 0 2 + 2 0 H " >2SO: + H 2 0 (2.11) 2so;; + o 2 — + 2 s o ; (2.12) 8 Chapter 2: Background These reactions occur at lower temperatures and consume essentially all of the oxygen before any methyl mercaptan, required for dimethyl disulphide formation, is formed. Any dimethyl disulphide that is found in the system is likely formed in processing equipment downstream of the digester. 2.2.3 Kinetics of Organic T R S Formation Douglass and Price (1966) studied organic TRS formation for two types of softwood, spruce and pine, and two types of hardwood, birch and maple, at cooking times of 1, 2, 3, and 4 hours, cooking temperatures of 150, 160, 170 and 180°C, and initial cooking liquor sulphidities of 14.7, 22.2 and 30.5% on an active alkali ( A A ) basis. Active alkali is defined as the concentration of N a O H plus Na 2 S and sulphidity is defined as the percentage ratio of Na 2 S to active alkali (all expressed as N a 2 0 ) . They reported that an increase in the cooking temperature, cooking time and sulphidity w i l l increase the formation of methyl mercaptan and dimethyl sulphide. They also found that more organic T R S compounds are produced when pulping hardwoods rather than softwoods, presumably due to the presence of additional labile methoxyl groups that demethylated rapidly at the beginning of the cook. Their data for spruce at 170°C, converted to a sulphur basis, are summarized in Table 2.1. Table 2.1: Total organic T R S formed during kraft pulping of softwood (spruce) at a cooking temperature of 170"C for 1,2,3, and 4 hour cooks (kg S/tonne dry wood) (Douglass and Price, 1966) Sulph-idity Methyl Mercaptan Dimethyl Sulphide Dimethyl Disulphide % o n A A 1 hr 2h r 3 hr 4 h r 1 hr 2h r 3 hr 4 h r 1 hr 2h r 3 hr 4 h r 14.7% 0.07 0.14 0.21 0.28 0.03 0.09 0.15 0.24 0.00 0.01 0.01 0.00 22.2% 0.15 0.33 0.47 0.47 0.04 0.12 0.22 0.32 0.01 0.01 0.01 0.03 30.5% 0.28 0.49 0.49 0.68 0.05 0.17 0.27 0.38 0.00 0.01 0.02 0.03 9 Chapter 2: Background The methyl mercaptan concentration increases as a consequence of Reactions (2.5) and (2.6). After some methyl mercaptan has formed, the dimethyl sulphide concentration wi l l begin to increase by Reaction (2.7). Ultimately, a steady state condition may be reached, at which point the mercaptan is formed and depleted at equal rates and its concentration remains constant (McKean et al., 1967). The rate of dimethyl sulphide formation wi l l gradually increase, approaching a constant rate at steady state. Due to a lack of oxygen in the digester, formation of dimethyl disulphide from the Reaction (2.9) is negligible. In Figure 2.1, formation of the organic TRS compounds is plotted as a function of cooking time. The mercaptan curve should be convex approaching a horizontal line, the dimethyl sulphide curve should to be concave approaching a straight line, and the sum of both of these species should increase in a linear manner (McKean et al., 1967). 1.50 -i Time (hours) Figure 2.1: Organic TRS formation as a function of time for kraft pulping of softwood (spruce) at 170°C and a sulphidity level of 30.5% (drawn from data from Douglass and Price, 1966) The predicted trends appear to be born out in the plotted data. In Figure 2.2, formation of the organic T R S compounds is plotted as a function of the sulphidity of the initial cooking liquor. Linear trends are observed for increasing organic TRS formation with increasing sulphidity. 10 Chapter 2: Background 1.50 r Sulphidity (%) Figure 2 . 2 : Organic TRS formation as a function of sulphidity for kraft pulping of softwood (spruce) at 170° C for 4 hours (drawn from data from Douglass and Price, 1966) The rate of organic TRS formation is also proportional to the methoxyl content in the system. The lignin methoxyl group is typically very stable and Sarkanen et al. (1970) report that only up to 5% is converted to organic T R S in a typical cook. The increase in TRS formation based on the methoxyl concentration is mainly noticed in the practice of recycling black liquor to the digester. When a batch digester is charged, some black liquor is usually added, along with the white liquor and wood chips. For continuous digesters using the "Lo-Solids" cooking process, black liquor is also recycled to the feed stream. This modified cooking process is used at the Howe Sound mil l where testing for this work was conducted. Depending on the amount of black liquor introduced at the start of the cook, the methoxyl concentration associated with dissolved lignin in the cooking liquor can be as much as doubled (McKean et al., 1967). This additional methoxyl group is then available to react with the hydrosulphide ion, with a corresponding increase in the formation of organic TRS . 11 Chapter 2: Background Yoon et al. (2003) reported similar results. They presented their data versus H-factor rather than time, thus encompassing the time and temperature factors in one variable, as illustrated in Figure 2.3. The H-factor is a single physical variable that represents the net effect of both cooking time and temperature during kraft pulping. 340 1360 1700 680 1020 H-factor Figure 2.3: Organic T R S formation as a function of H-factor for kraft pulping of softwood (loblolly pine) at 170°C (Yoon et al., 2003) The quantity of organic T R S formed increases with increasing cooking temperature, cooking time, sulphide concentration, and methoxyl concentration (increasing black liquor recycle to the digester). For typical digester operating conditions (170°C, 30 % sulphidity, 4 hours, softwood), the formation of organic TRS ranges from about 1 to 2 kg per tonne of dry wood (on a sulphur basis), based on the results of all research papers reviewed by the American and Swedish groups. Stated another way, about 2.5 to 5% of the total sulphur charge is converted to the organic sulphur compounds in the digester. McKean et al. (1967) reported that pulping of hardwood wi l l result in the formation of about 30% more TRS than a softwood cook. 12 Chapter 2: Background 2.2.4 T R S Formation in Black L i q u o r Evaporators Other than the digester, the only other equipment in a mi l l where the temperature is high enough for the organic TRS formation reactions to proceed, is the black liquor multiple effect evaporator set. The T R S compounds are volatile and thus are almost completely stripped from the black liquor in the first few effects in the evaporator set. Weak black liquor is usually introduced to the back end of the evaporator set where high vacuums result in boiling at temperatures as low as 70°C. The liquor is then pumped up through each effect to the first effect where the high pressure steam is introduced. As expected, the effects at the back end of the evaporator set, processing fresh liquor, have the highest concentrations of T R S in the vapour, while the vapour from the middle effects contains almost no TRS . Compared to previous effects, recent mil l testing has revealed that there is typically a spike in TRS concentration in the vapour from the first effect or concentrator of the evaporator set (Jaakkola et al., 1998; Klarin-Henricson, 2004). The liquor solids concentration can surpass 60% in these effects and, due to the boiling point rise, the liquor is processed at a temperature up around 140°C. These conditions are favourable to T R S formation; this spike is attributed to TRS generated in the black liquor in these effects. Although the amount of T R S generated in the evaporator has not been quantified, preliminary results by Klarin-Henricson (2004) indicate that it is minor when compared to T R S generation in the digester. A s discussed in Section 2.6, the N C G from concentrated sources, including the black liquor evaporators and concentrators, is typically collected and treated, so quantifying the amount formed in this equipment has not been an industry priority. 2.3 T R S Properties Table 2.2 lists some common properties of the TRS compounds; these data were extracted from the C H E M I N F O database available on the Canadian Centre for Occupational Health and Safety website (CCOHS) , and from Yaws ' Handbook of Thermodynamic and Physical Properties of Chemical Compounds (Yaws, 2006). 13 Chapter 2: Background Table 2 . 2 : Physical properties of selected T R S compounds ( C C O H S ; Yaws, 2006) Compound Hydrogen Sulphide Methyl Mercaptan Dimethyl Sulphide Dimethyl Disulphide C A S Registry Number 2148875 74-93-1 75-18-3 624-92-0 Structural formula H - S - H C H 3 - S - H C H 3 - S - C H 3 C H 3 - S - S - C H 3 Molecular weight 34.082 48.108 62.134 94.199 Melting point (°C) -85.5 -123 -83.2 -84.7 Boil ing point (°C) -60.3 5.95 37.3 109.7 Vapour pressure (kPa) 1840 at 21 °C 205 a t21°C 56 at 20°C 3.81 a t 2 5 ° C Vapour specific gravity (compared to air) 1.19 1.66 2.14 3.25 Liquid density (g/mL) 0.99 at -60°C 0.867 at 20°C 0.848 at 20°C 1.063 at 20°C Solubility in water (g/L) 4.0 at 20°C 23.3 a t20°C 6.3 at 20°C 2.5 a t 2 0 ° C p K a Value a t25°C pKa, = 6.97 p K a 2 = 12.9 p K a = 10.3 does not dissociate does not dissociate Critical temp. (°C) 100.4 196.8 230 333 Critical pressure (kPa) 9010 7235 5530 5360 L E L (% by volume in air) 4 3.9 2.2 1.1 U E L (% by volume in air) 46 21.8 19.7 16 Flame speed (m/s) 0.5 0.55 not available not available Auto ignition temp. (°C) 260 300 206 >300 A l l of the T R S compounds have low solubilities in water, and relatively high vapour pressures compared to water. Hydrogen sulphide and methyl mercaptan exist as gases at room temperature, although both wi l l dissociate in alkaline solutions, as shown in Reactions (2.3) and (2.7); in dissociated form at ambient temperature and pressure, they exert negligible vapour pressure. The pH effect is discussed in Section 3.1.3.1. Dimethyl sulphide and dimethyl disulphide exist as liquids at room temperature, although the former exist as a gas at the normal operating temperatures of the kraft pulping process. A l l of the T R S compounds are flammable and can sustain combustion at a concentration of only a few percent in air and all have relatively low auto-ignition temperatures, 14 Chapter 2: Background so they are dangerous to handle. As well as the dangers due to possible fire and explosion, all are toxic with significant adverse health effects. 2.4 T R S Health Effects Concern over emissions from kraft pulp mills has historically focused on the nuisance odour they cause in the surrounding community. To counter this, pulp companies began to install odour control systems to collect T R S laden gases to reduce the impact of emissions to a tolerable level, one that does not cause annoyance to the surrounding population. Odour levels are very difficult to quantify; thus, they are commonly described using these subjective terms. In this discussion, a "nuisance" level is considered synonymous with a level causing "annoyance" or no longer considered "tolerable"; these terms are used interchangeably. When combined with the term "threshold", measured on a quantitative scale of "parts per billion volume" or "ppbv," this is defined as the level above which there wi l l be complaints from a significant number of citizens. This level can vary; it is often defined by a local authority for the purposes of regulation. A s an example, California specifies it at the level where 40% of the population is expected to be annoyed by odour (Shusterman, 1992b). Defining threshold levels is difficult. For example some people may have a more acute sense of smell, a raised sensitivity to the sulphide compounds, a deadened sense of smell through smoking, or a higher tolerance to odour through repeated exposure as an employee at a mi l l . Another complicating factor is that a person's threshold level can change depending on the odour level. For example, at a higher concentration, a person's sense of smell may be deadened, and when the odour level is reduced, their sense of smell returns, and their perceived nuisance threshold level is reduced. T R S compounds are extremely noxious with a low threshold of odour detectability (Table 2.3). The extreme sensitivity of the human nose to these compounds may be due to the evolutionary process, as trace amounts of these compounds are emitted by decaying protein, serving as a warning signal to toxins present in food that has gone bad (Shusterman, 1992a). 15 Chapter 2: Background Table 2.3: TRS health ef fects data ( C C O H S ; A C G I H ; Ruth, 986; Tanseyetal. , 1981) Compound Hydrogen Sulphide Methyl Mercaptan Dimethyl Sulphide Dimethyl Disulphide State at room temperature colourless gas colourless gas colourless liquid pale yellow liquid Odour rotten eggs rotten cabbage wi ld radish, cabbage like disagreeable Odour Threshold (ppbv) 1 - 130 1 -41 1 -20 0.8-3.6 Occupational 8-hour exposure limit (ppmv) 10 0.5 Not set by the A C G I H Not set by the A C G I H Rat 4-hour inhalation LC50 (ppmv) 444 675 40250 805 T R S is toxic at high concentrations and has been responsible for a number of injuries and deaths in pulp and paper mills, mainly in confined space incidents. In a number of cases, multiple deaths have occurred when rescuers have also succumbed when entering a confined space without respiratory protection while attempting to save an unconscious coworker (CCOHS) . The most abundant data available are for H 2 S and thus it is often used as a surrogate when discussing the health effects of TRS . This seems to be a reasonable simplification as other T R S compounds have similar health effects but appear to be toxic to a lesser degree, based on L C 5 0 studies conducted on rats (Table 2.3). These studies determined the maximum concentration to which rats could be exposed for 4 hours that would result in 50% mortality (Tansey et al., 1981). The human nose can detect H 2 S in the range of a few parts per billion on a volume basis (ppbv). Exposure up to 10 parts per mil l ion (ppmv) wi l l affect sensory perception and wi l l cause irritation of the eyes and throat. The American Conference of Governmental Industrial Hygienists (ACGIH) has established exposure limits for H 2 S of 15 ppmv for 15 minutes and 10 ppmv for 8 hours (ACGIH) . These have been adopted by most provincial authorities in Canada, including the Workers Compensation Board ( W C B ) of British Columbia. 16 Chapter 2: Background Loss of the ability to smell H 2 S begins at 50 ppmv and at this concentration there wi l l be severe nose, throat and lung irritation. H 2 S wi l l cause damage to olfactory senses at 250 ppmv, so the presence of the gas can no longer be sensed. Exposure to 250 to 800 ppmv H 2 S wi l l cause severe sickness, permanent damage to the respiratory system and mucous membranes, unconsciousness and then death in 4 to 8 hours. Exposure for 15 minutes at 1000 ppmv can be fatal ( C C O H S ) . Some sources in a kraft mi l l , such as the digester and evaporator vents, typically contain upwards of 100,000 ppmv T R S (Burgess, 1992; U.S . E P A , 1976). Some local vent sources within the mi l l are thus extremely toxic. These sources are collected and treated at almost all mills. A t those few mills where these concentrated sources are not treated, the actual vents are piped to a high point in the mi l l and atmospheric dispersion is used to reduce their concentration to "safe" levels. A number of studies have been conducted on the effects of pulp mi l l emissions on the health of the population in nearby communities. The National Council for A i r and Stream Improvement ( N C A S I ) has compiled a number of these into one report (Tatum, 2001). Even though one of the studies reports "a higher incidence of symptoms of eye and upper respiratory tract infection" in children living near a pulp mil l in the South Karelia area of Finland (Marttila et al., 1994), N C A S I states that the results are statistically inconclusive. They conclude that "none of the kraft pulp mi l l community health studies described here provides any conclusive evidence that the emissions of a modern pulp mi l l pose any serious health risk to the residents of surrounding communities." N C A S I also conducted their own fairly extensive study on the health effects of T R S and concluded that "although high concentrations of H 2 S are acutely toxic, exposure to low (less than 20 ppmv) concentrations of H 2 S is not generally associated with significant health effects" (Tatum, 1995). The key word here is "significant," as some people had "various physical complaints of a more subjective nature, such as headaches, nasal congestion, fatigue and eye and throat irritation." The "mechanism(s) ...is uncertain", with some research suggesting "that such symptoms may be psychologically-based responses to the perception of unpleasant odors," while others have suggested that "odorant chemicals are capable of triggering neurological events leading to the development of physiologically-mediated symptoms." If N C A S I , a research organization mainly funded by the pulp and paper industry, appears to downplay the health impact of pulp mills on the community, then these views are balanced by those expressed by Nei l Carman, a clean air director for the Sierra Club. He stated, in a news release on 17 Chapter 2: Background the Sierra Club website (Sierra), that "public health scientists now recognize that hydrogen sulphide is a potent neurotoxin, and that chronic exposure to even low ambient levels causes irreversible damage to the brain and central nervous system." Because the evidence is inconclusive regarding the effects of T R S at the very dilute concentrations found in ambient air near pulp mills, the focus has historically been on the nuisance factor of these odours in the surrounding community. This appears to be a reasonable approach, since reducing the TRS concentrations to a tolerable level, i.e., to the 10s of ppbv, would likely also be a concentration well below that which most scientists and health officers would be concerned about the adverse health effects. 2.4.1 T R S O d o u r and Health Threshold Levels Thresholds are a difficult thing to quantify, given their subjective nature. The first threshold is the concentration at which most people can detect an odour, the second is when most people can identify the substance causing the odour, the third is when this odour becomes a nuisance and the fourth threshold is when the substance adversely affects health. A l l of these thresholds can have wide ranges, even overlapping, depending on the substance and local circumstances. Since most is known about hydrogen sulphide and since it can make up a significant fraction of T R S , it is often used as a surrogate for TRS when discussing threshold levels. The range of interest appears to be about 1 ppbv, at which point these odours can be detected, to 10,000 ppbv, the maximum 8-hour exposure level specified by the A C G I H (Table 2.3). This is presumably the concentration above which adverse heath effects would become apparent, although it is in many ways meaningless since many workers would refuse to work in those conditions. M i l l s in Canada typically set their H 2 S alarms in work areas at around 2500 ppbv, at which point people must evacuate the area. Even approaching this concentration, the odour is significant and would generate numerous complaints among workers. I have observed that Canadian mil l workers would not enter a work area i f the odour concentration, as measured on a handheld H 2 S detector, was above about 500 ppbv, and even near this level they would be reluctant. Various H 2 S threshold concentrations are summarized in Table 2.4. 18 Chapter 2: Background Table 2.4: H>S threshold concentrations ppbv H 2 S Threshold 1 Odour detection lower limit 4.5 Odour recognition lower limit 5 World Health Organization (WHO): "odour nuisance" 5 - 10 Typical ambient near mi l l with L V H C N C G and H V L C N C G collection systems 1 0 - 4 0 Canadian provincial Ambient A i r Quality Objectives ( A A Q O s ) 30 California objective: "assumes only 40% of people annoyed" 3 0 - 8 0 Typical ambient near mi l l with only L V H C N C G collection system 100+ Typical ambient near mi l l with no N C G collection system 100 American Industrial Hygienist Ass.: "transient adverse health effects" 110 W H O : "health hazard" 500 Typical maximum workplace concentration tolerated in a pulp mi l l 2500 Typical alarm setting for ambient detectors in pulp mills 10000 W C B and O S H A maximum 8-hr exposure limit The highest ambient TRS concentrations can usually be found in the mi l l production areas, but mi l l workers acclimatize to these odour levels. Leach and Chung (1982) collected a total of 1645 gas samples from four locations within eight B . C . pulp mills. O f the areas tested, they found the highest T R S concentrations around the digester, and of the T R S compounds, they found dimethyl sulphide in the highest concentrations. The normal range for dimethyl sulphide was measured at 0.5 to 6.0 ppmv, with a peak of 64 ppmv. The normal range for total TRS was 1.0 to 8.3 ppmv, with a peak of 73 ppmv. The Leach and Chung study is over two decades old and much progress has been made in odour reduction within mills, but it does give some idea to what extent people w i l l acclimatize to an unpleasant working environment. There is certainly an acceptance among mi l l workers and others dependant on the mi l l for their livelihood that they must put up with a certain level of odour as part of a means to earn a living. The presence of a mil l brings prosperity to the area and provides jobs; thus, the odour in and around the mi l l is tolerated. 19 Chapter 2: Background The actual ambient air concentrations in the geographical area surrounding the mil l can vary widely depending on the actual amount of contaminants released and the dispersion characteristics. The dispersion wi l l depend upon many factors such as the number, location and height of stacks, the wind direction and velocity, topography such as valleys and hills, and climate conditions causing such things as inversions. Older mills with only rudimentary odour collection systems may create TRS concentrations up to 100 ppbv at distance of 1 to 2 kilometers from the mi l l (O'Conner and Ledoux, 2002). A similar sized mil l with a more complete odour collection system may create ambient T R S concentrations of only 5 ppbv under similar circumstances (Jarvensivu et al., 1997; Freeburn and Redmond, 1998). It may be impossible to eliminate odour completely from kraft pulp mills, but well operated mills using the latest technology can get close. 2.4.2 O d o u r D e t e c t i o n T h r e s h o l d Odour detection thresholds for TRS compounds are given in Table 2.3. The odour detection for hydrogen sulphide ranges from 1 to 130 ppbv, with this variance depending on the olfactory sensitivity of the observer.' For example, a smoker living in a pulp mil l town wi l l likely not detect the odour until it reaches a much higher concentration than a trapper living in the northern wilderness. Prolonged exposure to low concentrations may cause olfactory fatigue. A s well , the sense of smell may be damaged from short term exposure to higher concentrations. It is not known to what extent the different TRS substances may be cumulative when it comes to odour detection. It seems likely that the mixed compound odour threshold would be near 1.0 ppbv i f each of the four T R S substances was at 0.25 ppbv level, but no research on this topic could be found in the literature. 2.4.3 O d o u r I d e n t i f i c a t i o n T h r e s h o l d Odour identification levels are typically somewhat higher than detection levels. For example a recognition threshold of 4.5 ppbv is given by the same source referenced for detection levels of 1 20 Chapter 2: Background to 130 ppbv in Table 2.3. That said, this threshold does not hold much importance to this issue as most people don't care what it is they smell, only that they know it smells bad. 2.4.4 Nuisance Odour Threshold The World Health Organization (WHO) has specified guidelines for half-hour exposure of 5 ppbv for "odour nuisance" for H 2 S (Shusterman, 1992b). They have obviously set this level on the basis that i f you can smell it, then it is a nuisance. California has established an Ambient A i r Quality Standard ( A A Q S ) for H 2 S that is "based on the endpoint of odour annoyance" (Shusterman, 1992b). They state that this standard, 30 ppbv on a one hour average, is set at approximately four times the population-mean odour threshold for hydrogen sulphide. Their assumption in establishing this standard was that approximately 40% of the population is expected to be annoyed at this odour level. What complicates defining a nuisance threshold is that not only is it different between people (eg. smokers and non-smokers), but it can be different for each person based on the initial conditions. For a person in a pristine environment exposed to an increasing odour, their threshold is likely to be much lower than someone acclimatized to a high concentration that is lowered. This can result from a desensitizing effect of the odour substance, either by simply getting used to it, or a physiological effect where the sense of smell is temporarily or permanently damaged. There are also varying responses to reduction in odour, with the nuisance threshold level often a moving target. For a community member in a pulp mi l l town, lowering the odour concentration may result in them becoming more sensitive to the odour, and thus their threshold level is lowered as the concentration is lowered. A t mills which have installed odour control systems, a short term venting incident wi l l often result in numerous complaints, whereas previously the local population tolerated long term higher concentrations. The nuisance threshold can be related to the next level since there is evidence that odour alone can have adverse health effects. Shusterman (1992a) of the E P A stated that "noxious environmental odours may trigger symptoms by a variety of physiologic mechanisms, including exacerbation of underlying medical conditions, innate odour aversions, aversive conditioning phenomena, stress-induced illness, and possible pheromonal reactions." 21 2 . 4 . 5 H e a l t h E f f e c t s T h r e s h o l d Chapter 2: Background W H O has specified guidelines for 24-hour exposure of 110 ppbv for "health hazard" (Shusterman, 1992b). The Emergency Response Planning Guidelines (ERPGs) issued by the American Industrial Hygiene Association (AIHA) in 2001 specify three levels for exposure to H 2 S . The lowest level, E R P G - 1 , is specified at 100 ppbv, and is defined as "the maximum airborne concentration below which it is believed that nearly all individuals could be exposed for up to 1 hour without experiencing other than mild transient adverse health effects or perceiving a clearly defined, objectionable odour" (CCOHS) . The U .S . Occupational Safety and Health Administration (OSHA) and the Workers Compensation Board ( W C B ) of B . C . have set a maximum average 8-hour occupational exposure level of 10000 ppbv ( C C O H S ) . The two orders of magnitude difference between the levels specified by W H O and A I H A for maximum ambient conditions and the exposure level set by O S H A and W C B for the workplace is eye-opening. This highlights the lack of scientific data available regarding the health effects of long term exposure to hydrogen sulphide and exposes the somewhat arbitrary nature of these limits. 2 . 4 . 6 W h a t t o M a k e o f T h e s e T h r e s h o l d L e v e l s ? Odour problems generate a significant fraction of the publicly-initiated complaints received by the polluter and air pollution control authorities. Given the predominantly subjective nature of associated health complaints, Shusterman (1992b) stated that "given our current state of knowledge, any differential regulatory response to environmental odour pollution, which is based upon the distinction between community annoyance reactions and health effects, is a matter of legal - not scientific - interpretation." He appears to be saying that given the inconclusive evidence of adverse health effects at the dilute concentrations in question, regulations for emissions should be set based on annoyance levels within the community rather than on perceived health effects. Based upon the above data, we can comfortably reduce our range of discussion for the ambient air concentration of T R S near pulp mills from 4 orders of magnitude (1 to 10000 ppbv) to two orders (1 to 100 ppbv). It is obvious that ambient concentrations above 100 ppbv, i.e. a mi l l 22 Chapter 2: Background with no odour control system or one that frequently vents their system to atmosphere, wi l l generate numerous complaints, while ambient concentrations around 5 ppbv, i.e., a mi l l with a complete well operated system, wi l l generate very few complaints (Jarvensivu et al., 1997; Freeburn and Redmond, 1998; O'Conner and Ledoux, 2002). The goal is to define the range of the nuisance threshold within this two orders of magnitude range; it appears that this is the range of discussion for most regulators. 2 . 5 K r a f t P u l p M i l l E n v i r o n m e n t a l R e g u l a t i o n s The most stringent environmental regulations with respect to pulp and paper making facilities are in North America and Europe, although what is defined as pollutants, how much is allowed to be released, and how they are regulated, vary greatly, not only from country to country, but between states, provinces, and even districts and regions. Some regulators refer to parts per billion on a volume basis (ppbv), while others refer to pg/m 3 . For H 2 S , the most commonly regulated compound, 1 pg/m 3 of H 2 S in air equals 0.717 ppbv at standard conditions of 25°C and 101.325 kPa. 2.5 .1 E n v i r o n m e n t a l G u i d e l i n e s i n C a n a d a Both provincial and federal statutes govern environmental legislation in Canada. Provinces exercise proprietary rights over the resources and therefore have the right to legislate on all aspects of water and air use and contamination. In general, each mil l is regulated on an operating permit issued by the local region and these requirements differ greatly not only from province to province, but from mil l to mi l l within provinces. Federally, the release of contaminants to the environment is regulated by the Health Act, the Fisheries Act and the Canadian Environmental Protection Act ( C E P A ) . National Ambient A i r Quality Objectives ( N A A Q O ) are maximum contaminant concentrations set with the goal of protecting public health, the environment, or aesthetic properties of the environment. These concentrations are set by the National Advisory Committee ( N A C ) Working Group on A i r Quality Objectives and Guidelines ( W G A Q O G ) , a group comprising representatives of federal, provincial and territorial departments of environment and health. 23 Chapter 2: Background Historically, there has been a three-tiered structure for defining air quality objectives, maximum desirable (tier 1), maximum acceptable (tier 2) and maximum tolerable (tier 3). The first tier was defined as ambient air quality for residential areas, the second tier for industrial locations and the third tier as an unacceptable level requiring action. The listing for hydrogen sulphide under N A A Q O specifies tier 2 limits of 15 p.g/m3 for a 1 hour average and 5 p.g/m3 for a daily average. These figures were extracted from the document "The Table of National Ambient A i r Quality Objectives and Guidelines", undated, available on the federal government website (Canada Government). The three-tiered structure is being phased out in favour of a single tiered system used by most countries. In 1998, the Canadian Council of Ministers of the Environment ( C C M E ) signed the Canada-Wide Accord on Environmental Harmonization and its sub-agreement on Canada-Wide Standards (CWS) . These standards are intended to be achievable targets that wi l l reduce environmental risks within a specific time-frame. It is not clear why two, apparently similar, standards have been developed, but the C C M E website states that "air pollutants that have been identified by governments as needing to be managed wi l l be targeted for either N A A Q O or C W S development, not both." Thus, as would be expected since it is already listed under N A A Q O , hydrogen sulphide is not listed under the C W S document "Canadian Environmental Quality Guidelines," updated 2002, available on the C C M E website ( C C M E ) . The numerous Canadian objectives and standards and the provincial guidelines discussed below are not legally binding. Emissions to the environment are enforced through permits issued by the provincial regional authority authorizing companies to discharge contaminants into the air. The objectives and guidelines only become legally binding i f they are written into the individual mi l l operating permit. Federal air quality documents refer only to H 2 S , with no mention of any of the organic T R S compounds. This is also true for provincial documentation, with the exception of Ontario. 24 Chapter 2: Background 2.5.1.1 Br i t i sh Co lumbia In British Columbia, emissions to land, water and air are regulated under the provisions of the Environmental Management Act. A s part of this act, each mil l is regulated on an operating permit issued by the local region. These permits may specify maximum T R S , N O x , S O x and particulate emissions from designated equipment or identified stacks. Since there are no uniform provincial guidelines for collection and treatment of kraft pulp mi l l air emissions, there is a huge diversity in type and scope of collection and treatment systems installed at B . C . mills. The installed systems range from one mil l having no collection and treatment whatsoever, to another mil l collecting N C G from about twenty sources. Geographical location, weather conditions, proximity to populated areas, vintage of the mil l and political factors have an influence on the extent of the emissions control systems installed in a particular mi l l . Some provinces, including B . C . , specify general ambient air quality objectives in the vicinity of pulp mills, although, as previously mentioned, these are not legally binding. B . C . air quality objectives and standards also use a three-tiered system (level A , B , and C) roughly corresponding to the Canadian tiered objectives. A t the industrial " B " level, pulp and paper mills in B . C . capable of demonstrating that the H 2 S concentration in ambient air does not exceed 28 pg/m 3 on a one hour average and 6 pg/m 3 on a daily average at monitored stations (whose location are undefined), are considered to be in compliance. These figures are extracted from the document " A i r Quality Objectives and Standards," undated, available on the provincial government website (British Columbia Government). 2.5.1.2 Alber ta Alberta's ambient air quality guidelines are established under Section 14 of the Environmental Protection and Enhancement Act (EPEA) ; all documentation is available on the provincial website (Alberta Government). The purpose of this act is to ensure that emissions are minimized through the use of Best Available Demonstrated Technology ( B A D T ) and that residual emissions are dispersed so that the guidelines are met. Alberta has graduated to the one tier system and mills are judged to be in compliance i f the H 2 S concentration in ambient air did not exceed 14 25 Chapter 2: Background pg/m 3 on a one hour average and 4 | ig /m 3 on a daily average at monitored stations as specified in the document "Alberta A i r Quality Guidelines," dated February 2000. Alberta also has an additional guideline for "static" H 2 S , defined as a maximum one month accumulated loading of 0.10 mg S 0 3 equivalent per day per 100 cm 2 land area.. 2.5.1.3 Ontario Ontario has similar guidelines to B . C . and Alberta, but with more detail as they also refer to the organic TRS compounds. The enabling legislation is Regulation 346 of the Environmental Protection Act, which specifies maximum ambient Point of Impingement (POI) limits and Ambient A i r Quality Criteria ( A A Q C ) for over 350 substances. These are detailed in the document "Summary of Point of Impingement Standards, Point of Impingement Guidelines and Ambient A i r Quality Criteria ( A A Q C s ) " dated September 2001, available on the provincial website (Ontario Government). A A Q C standards can also be set for 10 minute, daily and annual averages, but for TRS , it has only been set for the 1 hour average listed in Table 2.5. Table 2.5: Ontario POI and A A Q C limits (Ontario Government) Compound Hydrogen Sulphide Methyl Mercaptan Dimethyl Sulphide Dimethyl Disulphide POI Vi hour limit (pg/m 3) 30 20 30 40 A A Q C 1 hour limit (p.g/m3) 30 20 30 40 The POI standards are those used for compliance monitoring of industrial facilities. The POI limits are typically derived by mathematical scaling, from A A Q C s which represent human health or environmental effects-based values. The POI monitoring locations are not specified in the guidelines; instead, they are determined on a mi l l to mi l l basis for each operating permit issued by the local authorities. Typically, the location for these criteria w i l l be defined as the property limit, but in cases where the mi l l is located in more populated areas, the monitors may be placed in a "sensitive" location such as at an 26 Chapter 2: Background elementary school or hospital. The location of the monitoring stations and the prevailing wind direction have a great influence on whether the POI limits are met. A recent study by the Sierra club reported that for air pollution violations in Ontario during 2001 (the latest statistics available), kraft pulp mills held at least the top six slots (the list included only six) with the highest having 90 violations for that year alone (MacDonald and DeMarco, 2004). 2 . 5 . 1 . 4 Q u e b e c Quebec pulp mills are regulated under schedule III of the "Reglement sur les fabriques de pates et papiers Q-2, r.12.1," available on the provincial website (Quebec Government). This legislation is legally binding and is much more detailed than the regulations for the other provinces, with emission limits specified for individual process equipment. The legislation states that all mechanically vented emissions from cooking systems, evaporation systems, condensate stripping systems and brown stock washing systems that have concentrations higher than 10 ppmv T R S must be collected and treated. The result of this requirement is a large variability in installed systems and the effectiveness of these systems between mills. For example, tank vents do not require collection unless they are "mechanically" vented, presumably using a fan. A s well , a mi l l with an open hood and a large ventilation fan on a brown stock washer drawing in large volumes of tramp air can stay below the limit while another mi l l with a sealed hood and a lower vent volume wi l l have to install a collection and treatment system. This may be the case in spite of the fact that the first mi l l , all else being equal, emits more T R S on a weight per time basis due to the stripping effect on the volatile substances. Quebec also has an air quality objective stating that the ambient air quality should not exceed 10 ppbv T R S at the mill property limit. 2 . 5 . 1 . 5 O t h e r P r o v i n c e s The other provinces have similar policies to those discussed for B . C . , Alberta, Ontario and Quebec, although with some variance. There appears to be a move to standardizing the provincial guidelines based on the Canadian objectives; this was one of the driving forces behind the formation of the C C M E . Most provincial ministries appear to be adopting the approach of stipulating T R S 27 Chapter 2: Background emission levels at the mil l property limit or other monitored stations and allowing each company to install facilities that best meet the guidelines. 2 . 5 . 1 . 6 C a n a d i a n N a t i o n a l P o l l u t a n t R e l e a s e I n v e n t o r y ( N P R I ) Federally, under C E P A , the National Pollutant Release Inventory (NPRI) was established in 1992. This program is a legislated, nation-wide, publicly-accessible inventory of pollutants released to the environment. It was created to provide Canadians with information on pollutants released from facilities in their communities. It is meant to assist governments to identify priorities for action, encourage industries to take proactive measures to reduce releases, allow for tracking of progress in reducing releases and support regulatory initiatives. This program requires all industrial facilities to report their emissions to air, land and water, i f the reportable substances used or produced onsite exceed given threshold limits in quantity or concentration. The latest "Guide for Reporting to the N P R I " is dated 2005 and is available on the Environment Canada website (Canada Government). This document lists over 300 substances that require reporting, including hydrogen sulphide, but not the organic T R S compounds. N P R I states that facilities are allowed to calculate their reportable releases in either of four ways, listed in order of preference (Allen, 2002): (i) direct measurement, (ii) mass balance calculations, (iii) emissions factors, or (iv) engineering estimates. The last option is not defined, but is presumably based on an expert consultant's experience. 2 . 5 . 2 E n v i r o n m e n t a l G u i d e l i n e s i n t h e U.S.A. The U.S . Environmental Protection Agency (EPA) uses the term Hazardous A i r Pollutants (HAP) , which includes 188 contaminants, to define emissions for regulation of pulp and paper mills. H A P is a general grouping of chemicals that have been identified as causing serious illness, including cancer. The regulations associated with H A P are typically referred to as the "Cluster Rule," regulated under the Maximum Achievable Control Technology ( M A C T ) emissions standards document, available on the E P A website (U.S. E P A , 1998). Non-combustion sources, which are regulated under M A C T Part I, were promulgated in 28 Chapter 2: Background Apr i l 1998 at which point a three and eight year countdown began for achieving these standards. The portion covering combustion sources such as the recovery boiler, smelt dissolving tank, lime kiln, and power boiler, are regulated under a separate document referred to as M A C T Part 2, which was promulgated in October 2005, at which time a three year countdown for compliance began. The E P A lists the following as the H A P compounds emitted in the highest quantities on a mass basis in a typical kraft mi l l : • Methanol • Methylene chloride • Methyl ethyl ketone • Phenol • Propionaldehyde • 1,2,4-Trichlorobenzene • o-Xylene • Acrolein • Acetaldehyde • o-Cresol • Carbon tetrachloride • Chloroform • Cumene • Formaldehyde Some of these substances may be in higher or lower concentrations at some mills depending on the specific process. For example, chloroform emissions from a mi l l using a chlorine-free bleaching process would probably be insignificant. Volatile Organic Compound ( V O C ) emissions, which are linked to smog formation through chemical reaction with nitrogen oxides (NO x ) , are another group which are of particular concern. These compounds, by definition, are highly volatile, so escape into the atmosphere easily. Many V O C s , but not all , are listed as H A P . Recent research aimed towards Cluster Rule compliance has focussed on methanol since it typically makes up a large fraction of H A P and has thus been designated as a surrogate compound. Although methanol may make a good surrogate for the H A P compounds, it is not a good surrogate for the T R S compounds due to their much higher relative volatilities, as discussed in Section 3.1. A number of volatile substances emitted in relatively high quantities, such as the terpenes, are not included in H A P . Presumably, this is because they are wood extractives, and thus considered a "natural" emission rather than a manufactured chemical release. Most of the turpentine compounds can easily be condensed and are a potentially valuable by-product, either as a fuel or for sale to a refining plant; therefore, many mills practice turpentine recovery. Not only are terpenes highly volatile, but in the liquid phase they can be very toxic and can cause upsets in the effluent treatment system i f released in large quantities. None of the TRS compounds are among those defined as H A P , even though they are obviously hazardous, as discussed in Section 2.4. It is not clear why these compounds were 29 Chapter 2: Background excluded from the national regulations, but it may have been a political decision as T R S release was already being regulated through local state operating permits. Although the Cluster Rule requirements do not pertain directly to T R S , a correlation can be made with reduction in H A P to reduction in T R S , since the targeted emission sources that are high in H A P are also typically high in T R S ( N C A S I , 1994a; 1994b; 1996; 2002b). That said, there may soon be a direct link because hydrogen sulphide may eventually be classified a H A P . The E P A is presently being lobbied by "130 public health and environmental organizations and community groups," including the Sierra Club, to "formally list hydrogen sulfide as a hazardous air pollutant," as reported on their website (Sierra). This has renewed interest in research in the area of hydrogen sulphide release, particularly from effluent treatment lagoons, ln the U.S. , the National Council for A i r and Stream Improvement (NCASI) , in 2003, began an ongoing program of study on this topic (Cook and Hoy, 2003; Crawford et al., 2006). There has also been some pressure on the E P A to include the organic TRS compounds in H A P , but the consensus seems to be that this wi l l occur in the longer term, at least 10 years down the road. More details are provided in Appendix A regarding the Cluster Rule M A C T Part 1. These details have been extracted and summarized from the hundreds of pages that make up the actual Cluster Rule document. There has been speculation about the potential for U.S. companies to seek a requirement for Canadian competitors to meet the Cluster Rules (Vice, 1998). Two concerns were expressed: (a) a trade barrier in the form of a border tax on products crossing into the U.S . or; (b) Canadian companies unable to secure cross border contracts until they meet the same environmental criteria as U.S . mills. To date, there is little evidence to suggest that this may be the case, but once U.S . mills complete their upgrades, there may be political pressure for Canadian mills to follow. 2 . 5 . 3 E n v i r o n m e n t a l G u i d e l i n e s i n R e s t o f W o r l d There are many kraft mills besides those in North America, with the largest number located in Brazil , China, Indonesia and the Scandinavia countries. The latter are generally perceived to have some of the strictest standards and regulations. In Finland, a new amendment to the A i r Pollution Act came into force in Apr i l 1996 30 Chapter 2: Background (Hynninen, 1999). Kraft pulp mills must ensure that ambient air quality in the vicinity of the mi l l does not exceed 10 p g / W T R S (as sulphur) as the second highest daily value of one month, meaning that a mi l l is allowed to exceed this limit for one day per month. It also requires that total gaseous sulphur emissions from pulping processes be less than 2 kg/t (as S 0 2 ) for new and rebuilt mills. The amendment also has limits for carbon monoxide, nitrogen dioxide, sulphur dioxide, total suspended particulate and respiratory particulate ( P M ] 0 ) , all in terms of pg/m 3 ambient air quality. In Sweden, no general standards are applied at present (Hynninen, 1999). Instead, the authorities grant a specific environmental discharge permit to each mi l l . Typical restrictions include maximum annual average gaseous sulphur emissions of 0.5 kg/t (as sulphur), a minimum availability of odour control systems of 99% (venting of odorous gases limited to 1% of time that the mi l l process is operating), maximum H 2 S concentration in the recovery boiler stack of 10 mg/m 3 (normal dry gas, maximum of 5% of time exceedance allowed per month), as well as limitations on nitrogen oxides, sulphur oxides and particulate from the recovery boiler, lime kiln, power boilers, etc. 2.6 N C G Collection and Treatment Systems In most countries, in order to meet air quality requirements, it is necessary to collect noncondensible gases (NCG) from various emission sources into one or several systems for disposal by chemical modification or incineration. As discussed in the previous section, environmental regulations governing kraft pulp mills are highly dependent on the local regulatory authority; thus, worldwide there is a huge variability in the extent of installed N C G systems. This can range from no odour control system whatsoever, to very elaborate systems collecting up to thirty vent sources (Pinkerton, 1999; Bruce and Van der Vooren, 2003; A . H . Lundberg). N C G is a general term used by the industry for those gases released from non-combustion process equipment and tankage in a kraft pulp mi l l during normal production. The use of "noncondensible" is applied somewhat inaccurately, since, of all the major components of N C G , nitrogen and oxygen are the only two that fit the definition. Noncondensibles are defined as being above their critical temperature. The critical temperature of a material is the temperature above which distinct liquid and gas phases do not exist. A s the critical temperature is approached, the properties of the gas and liquid phases become the same. The critical pressure is the vapor pressure 31 Chapter 2: Background at the critical temperature. Above the critical temperature, there is only one phase. The noncondensible components of N C G often enter the system as "tramp" air through openings in tanks and equipment hoods. The other components of N C G , including water vapour, methanol, turpentine, and the TRS compounds are vapours that have been volatilized from brown stock, black liquor, or foul condensate in the process. For safety reasons, because most of the volatile components are also flammable, to avoid handling of explosive gas mixtures, it is common practice to segregate the vent sources into so called Concentrated NonCondensible Gases ( C N C G ) , also referred to as L o w Volume High Concentration ( L V H C ) N C G , and Dilute NonCondensible Gases ( D N C G ) , also referred to as High Volume L o w Concentration ( H V L C ) N C G collection systems (Burgess, 1992). In such an arrangement, the combustible components of the N C G are generally to be found in concentrations above the Upper Explosive Limit (UEL) in the C N C G system and below the Lower Explosive Limit ( L E L ) in the D N C G system (Table 2.2). The lack of oxygen in C N C G , typically as low as 8% by volume, also contributes to maintaining the mixture above the U E L . In Table 2.6, Burgess (1992) provides a typical analysis for C N C G combined from various sources. He states that composition wi l l vary widely from system to system, and occasionally within the same system. T a b l e 2 . 6 : Typical combined C N C G composition (Burgess, 1992) Compound % by volume Hydrogen sulphide 1.7 Methyl mercaptan 2.1 Dimethyl sulphide 2.1 Dimethyl disulphide 1.7 Turpentine 0.1 Methanol 0.2 Water vapour 6 Nitrogen 77.2 Oxygen 8.9 32 Chapter 2: Background This variability in concentration is illustrated by the U.S . E P A (1976), who provide typical concentration ranges for C N C G and D N C G sources shown in Table 2.7. Table 2.7: U .S . E P A typical T R S emissions for C N C G and D N C G sources (ppmv) (U.S. E P A , 1976) Source Hydrogen Sulphide Methyl Mercaptan Dimethyl Sulphide Dimethyl Disulphide C N C G Batch digester blow gases 0 - 1000 0 - 10,000 100-45,000 10-10,000 Batch digester relief gases 0 - 2000 10-5,000 100-60,000 100-60,000 Continuous digester 10-300 500- 10,000 1500 - 7,500 500 - 6,000 Evaporator 600 - 9000 300 - 3,000 500 - 5,000 10- 10,000 D N C G Drum washer hood 0 - 5 0 - 5 0 - 15 0 - 3 Washer seal tank 0 - 2 1 0 - 5 0 10-700 1 - 150 Black liquor oxidation 0 - 1 0 0 - 2 5 10-500 2 - 8 5 When companies are considering the installation of N C G systems, they wi l l sometimes do sampling and testing of vent sources to assist in the design process. This data, often considered sensitive, is typically not published in the open literature. Often, the testing is conducted by an industry research organization such as N C A S I or Paprican. These organizations publish technical reports which are only available to member companies and to some educational organizations. In one such report, N C A S I compiled T R S emissions data from all testing they had conducted during the preceding decade ( N C A S I , 2002). The emission figures in Table 2.8 are averaged from up to 24 mills. 33 Chapter 2: Background T a b l e 2 .8 : N C A S I measured and averaged T R S emissions for C N C G and D N C G sources (ppmv) ( N C A S I , 2002b) , Source H 2 S M M D M S D M D S T R S as S avg. avg. avg. avg. avg. range C N C G Batch digester blow gases 3702 54100 35971 632 95000 64 to 683000 Batch digester relief gases 452 4678 4481 456 10500 10500 to 10600 Continuous digester 122 4128 2072 935 8200 26 to 34000 Evaporator 65281 69766 32874 2321 173000 17 to 719000 Combined C N C G 17639 42282 20292 1493 83200 1730 to 701000 D N C G Drum washer hood (open) 1 2 12 3 21 2 to 84 Drum washer hood (sealed) 2 8 209 20 260 Oto 1950 Unbleached pulp storage 0.4 1 371 58 490 4 to 3340 Weak black liquor storage 9 7 134 18 190 9 to 650 Strong black liquor storage 340 2259 1085 370 4430 9 to 22600 Black liquor oxidation 7 7 14 19 66 7 to 180 Combined D N C G 32 9 178 16 250 81 to 440 34 Chapter 2: Background In general, the averaged concentration data presented by N C A S I falls into the concentration ranges specified by the U .S . E P A . There are a few inconsistencies. The U.S . E P A ranges for dimethyl disulphide appear high and their concentration ranges for all species for evaporators appears low. The N C A S I data for combined C N C G sources also agree quite well with the typical C N C G concentrations given by Burgess (Table 2.6) although, again, with dimethyl disulphide too high. C N C G sources w i l l always include the two highest emitters, the digester and the evaporator, and sometimes the blow tank, turpentine recovery system, foul condensate tank and foul condensate stripping system. A C N C G system collects the highest emitters into a relatively small system, with this system collecting roughly 80% of the odorous substances emitted from non-combustion sources. Without a C N C G system, the ambient air T R S concentration near a mi l l w i l l typically be in the 100s of ppbv; all but two mills in Canada have these systems installed. A mil l with a C N C G system wi l l typically have ambient air conditions near the mi l l in the range of 30 to 80 ppbv T R S (O'Conner and Ledoux, 2002), although this depends greatly on vent stack heights and local climatic conditions. Typical D N C G sources include the continuous digester chip bin, diffusion washer, brown stock washer, decker washer, knotter hood, screen hood, black liquor oxidizer, soap skimmer, sewer vent and tankage such as the filtrate, stock, knot, screen feed, screen rejects, refiner feed, black liquor, spill, soap, tall o i l , precipitator mix and chemical ash mix tanks. These systems collect the numerous lower concentration emitters into a relatively large system and, combined with a C N C G system, wi l l typically reduce ambient air conditions around a mi l l to less than 10 ppbv T R S (Freeburn and Redmond, 1998; Jarvensivu et al., 1997). Appendix B includes a more in-depth discussion of C N C G and D N C G collection and treatment system design and operation. In the U.S. , partial compliance with the new Cluster Rule, required as of Apr i l 2001, was typically achieved with the installation of C N C G and foul condensate collection and treatment systems. Full compliance, which was required as of Apr i l 2006, typically required installation of a D N C G system. The extent of installed D N C G systems at Canadian mills is highly dependent on mil l location, with the main dependency being political, i.e., in which province the mi l l is located. For example, Alberta has strict H 2 S ambient air quality standards (presumably originating from flaring at oil refineries and sour gas processing but covering all industrial facilities); thus, all pulp mills in 35 Chapter 2: Background Alberta have extensive C N C G and D N C G collection systems. This contrasts with the only kraft pulp mill in Manitoba which has no N C G systems whatsoever. In Ontario, all mills installed C N C G systems in the 1970s, but only two mills operate D N C G collection systems, these having been only recently constructed. Within each province, each mi l l is issued an operating permit by the local regional authority. Due to pressure from local citizens and town councils, mills located near urban centres often have more extensive odour control systems. This may not be only a result of a larger population generating more complaints,, but also differing social demographics. In large urban centres, there may be a higher population of people with little invested in the mi l l (i.e., those that do not depend on the mi l l for their livelihood), environmental activists, or those with political influence. Local climate and topography have a large influence on permitting and thus on N C G system design. Climate has an impact, mainly in the form of prevailing wind direction. Many mills are located downwind of the local community, a pattern repeated so often that one suspects the owners deliberately located them in this way. Even though this may seem preferable, when the wind direction does change and blow into the community, more complaints may be triggered because the severe change in ambient air quality is often more noticeable than a constant odour. Climate issues also include inversions where dispersion into the upper atmosphere is inhibited and the odorous compounds are pushed down towards the ground. Topography has a significant effect; hills and mountains or ocean or river currents can affect local air flow and atmospheric dispersion patterns. Another factor influencing the operating permit of a mil l is its vintage. Older mills wi l l often have a "grandfathered" operating permit; they are allowed to exceed provincial objectives because the mil l was constructed under previous laxer standards. More recently-constructed mills have been required to engineer for D N C G systems during planning stages, often as a result of community consultation processes, before they receive their development permit. The last Canadian mills to be constructed or go through major rebuilds, in the late 80s and early 90s, are located in Castlegar, Kamloops, and Port Mel lon in British Columbia, Boyle and Peace River in Alberta and Windsor in Quebec. A l l these mills have extensive D N C G systems. There does not appear to be an accepted methodology in the industry for estimating T R S emissions. Designers of N C G systems typically rely on limited field testing, and in the absence of this, in-house engineering estimates based on previous systems installed at similar mills (Burgess, 36 Chapter 2: Background 1992; A . H . Lundberg). A more accurate, yet relatively inexpensive method of predicting emissions is to use simulation software to construct mass balances across the equipment, and validate and tune this model using mil l testing data. One method to predict vent emissions using simulation software is to incorporate an emissions module based on the vapour-liquid equilibria of the target contaminant in kraft mil l liquid process streams. 2 . 7 K r a f t M i l l P r o c e s s S t r e a m s TRS can be found in many mil l process streams including black liquor and foul condensates. Foul condensates originate from condensing vapour from the digester extraction black liquor flash tanks and from the black liquor evaporators., Foul condensates consist mostly of water, but are referred to as foul because they also contain odorous compounds such as turpentine, methanol and TRS . Foul condensates only contain dissolved solids when there is carryover of liquor droplets with the vapour to the condensing surface. Black liquor contains many different dissolved solids, including many salts. 2.7 .1 B l a c k L i q u o r C o m p o s i t i o n Wash water is used counter-currently in the brown stock washing process. The total dissolved solids in the wash water increases as it passes through each washing stage. The filtrate from the final wash stage is combined with the black liquor from the digester and sent to the evaporators. The total dissolved solids in weak black liquor is typically in the region of 15 wt%, but can reach as high as 20 wt%. Table 2.9 shows a typical analysis for softwood black liquor from three mills (Grace et al., 1989), on a water-free basis. 37 Chapter 2: Background Table 2.9: Components of weak black liquor solids (wt% on water-free basis) (Grace et al., 1989) Component M i l l A M i l l B M i l l C Lignin 28.9 30.7 31.1 Hemicellulose and sugars 1.14 0.11 1.3 Extractives 6.69 2.53 5.7 Saccharinic acids not measured not measured 18.8 Acetic acid 3.52 2.08 5.2 Formic acid 4.48 2.7 3.1 Other organic acids 5.5 2.22 0 Unknown organic compounds 19 29.5 5.8 Inorganic salts 18.6 18.5 20.3 Organically combined N a 10.1 10.3 8.7 Unknown inorganic compounds 2.08 1.35 0 The saccharinic acids were not measured for M i l l A and B ; these would be included in either "other organic acids," or "unknown organic compounds." Lignin is a natural branched polymeric macromolecule, present in most species of plants. In wood, lignin consists of a three dimensional, cross-linked network comprised mainly of phenylpropane units connected by ether linkages or carbon-carbon bonds (Ohman, 2006). The lignin structure also contains methoxyl and phenolic hydroxyl groups. During kraft pulping, the alkali and sulphide attack the lignin and polysaccharides in wood, reducing their molecular size and dissolving them. The degradation of lignin is due mainly to the cleavage of ether linkages between the phenylpropane units. Most of the alkali used in pulping is consumed in neutralizing the acidic phenolic groups on the lignin and its degradation products, the organic acids formed from the polysaccharides, and the resin acids. Post degradation, the lignin fragments, referred to as alkali lignin, diffuse out into the black liquor, dissolved as sodium phenolates. The other dissolved organic matter in black liquor consists the sodium salts of the polysaccharinic acids, resin acids and fatty acids. 38 Chapter 2: Background The phenolic hydroxyl groups of alkali lignin are characterized by p K ' s of 9.4 to 10.8. When the black liquor p H is in this range, alkali lignin wi l l precipitate. Because of the heterogeneous nature of alkali lignin, this effect comes on gradually, proceeding from a slow thickening at higher pH values, to actual precipitation at lower values (Grace et al., 1989). During kraft cooking, some sulphur becomes irreversibly bound to the dissolved lignin, decreasing the overall hydrosulphide concentration in the cooking liquor. McKean et al. (1968) found that the sulphur lost to lignin to be a maximum of 0.4%, while at the same time stating that others estimate the loss to vary from almost negligible up to 3%. Ideally, the total hydrosulphide concentration through the cook wi l l only decrease by the reactions to organic T R S plus the amount lost to the lignin, i.e., the sum of 2.5 to 5% and 0 to 3%, for a total loss of 2.5 to 8%. In practice, the amount of hydrosulphide lost during the cook and in downstream processing equipment is typically in the range of 30 to 40% due to oxidation reactions to thiosulphate, sulphite and sulphate (Grace et al., 1989). A typical breakdown of the inorganic salts in black liquor is shown in Table 2.10 (Grace et al., 1989). The concentrations of sodium salts in kraft liquors are often given in terms of N a 2 0 . To convert these to actual concentrations, multiply by their molecular weight and then divide by the molecular weight of N a 2 0 . For example, to convert Na 2 S from g/L as N a 2 0 to g/L, multiply by 78 and divide by 62. Table 2.10: Typical inorganic salt composition of black liquor (Grace et al., 1989) Chemical g/L as N a 2 0 g/L as N a 2 0 g/L as itself wt% (median) (range) (median) (median) N a O H 1.0-4.5 1.4 1.8 4.6 Na 2 S 1.6-5.6 4.2 5.3 13.4 N a 2 C 0 3 5.0-12 7.8 13.3 33.7 N a 2 S 0 3 0.4-3.8 2 4.1 10.3 N a 2 S 0 4 0.5-6.0 2.8 6.4 16.2 N a 2 S 2 0 3 1.8-5.1 3.4 8.7 21.9 39 Chapter 2: Background The elemental composition of the black liquor is also of interest. Table 2.11 shows the typical values of the elemental composition of virgin black liquor derived from softwood (Vakilainen, 1999). Table 2 . 1 1 : Elemental composition of softwood black liquor solids (Vakilainen, 1999) Element Typical amount (wt%) Typical range (wt%) Carbon 35 32 to 37 Hydrogen 3.6 3.2 to 3.7 Nitrogen 0.1 0.06 to 0.12 Oxygen 33.9 33 to 36 Sodium 19 18 to 22 Potassium 2.2 1.5 to 2.5 Sulphur 5.5 4 to 7 Chlorine 0.5 0.1 to 0.8 Inert matter 0.2 0.1 to 0.3 A s wi l l be discussed in Chapter 3, the presence of the dissolved inorganic solids has a significant impact on the vapour-liquid equilibrium of the TRS compounds. 40 C h a p t e r 3 L i t e r a t u r e R e v i e w Chapter 3: Literature Review 3.1 T R S P h a s e E q u i l i b r i a B e h a v i o u r Emissions of TRS are contingent on the inter-phase mass transfer between the liquid (e.g. black liquor) and the vapour (e.g. N C G vent). In this case, a flux of the T R S compounds wi l l begin with a transfer from the liquid to the vapour, but as more T R S enters the vapour phase, with a consequent increase in concentration within the vapour, the rate at which T R S returns to the liquid increases, until eventually the rate at which it enters the vapour is exactly equal to that which it leaves. At the same time, through the mechanism of diffusion, the concentrations throughout each phase become uniform and a state of dynamic equilibrium is said to exist. If the flux between phases is not limiting, the release of volatile substances from process equipment in a kraft pulp mil l can be modelled based on their vapour-liquid equilibria. Thermodynamic models to describe vapour-liquid equilibria are well established and detailed descriptions can be found in many reference sources (Balzhiser et al., 1972; Perry's, 1997). The description below is a condensed version for the specific case of a low pressure water-based system. 3.1 .1 V a p o u r - L i q u i d E q u i l i b r i u m For a system to be in thermodynamic equilibrium, it wi l l be at a state of maximum entropy, or lowest possible Gibbs free energy. Classical thermodynamics can provide relationships based on Gibbs energy and chemical potentials, but these are defined in relationship to internal energy and entropy, both quantities for which absolute values are unknown. Models, based on what is known as molecular thermodynamics, have been created to relate these variables to one another. Homogenous systems at internal equilibrium can be modelled using these macroscopic properties expressed as functions of temperature, pressure and composition. One of these macroscopic properties is called fugacity, f, which represents the tendency of a component to leave a phase (Jugare in Latin means escape). Phase equilibrium exists at a given temperature, T, and pressure, P, when each component shows the same chemical potential in the 41 Chapter 3: Literature Review vapour and liquid phases. From this criterion it can be deduced that the fugacities of each component are equal in each phase when equilibrium prevails: f , v (T ,P ,y , )= f , L (T ,P ,x , ) (3.1) f f v is the fugacity and y is the mole fraction of component i in the vapour phase, and f^ and x represent the same for the liquid phase. Since many real mixtures are not ideal solutions, a thermodynamic variable called "activity" was introduced to facilitate calculations. The activity, a, of component i is defined as: a, = | r (3-2) The fugacity, fj0, is the fugacity of component i at the standard state, 0. The most commonly used standard state, and the one used here, is the pure component at the same temperature, pressure and phase as the mixture. Two further thermodynamic variables based on the activity were also introduced, the vapour phase fugacity coefficient, 0„ and the liquid phase activity coefficient, y„ defined by the following equations: These coefficients are used to relate fugacity to measurable quantities such as temperature, pressure, molar volume and composition. They are a measure of the deviation of the fugacity from its standard state. The most computationally straightforward and thermodynamically consistent method for calculating phase equilibria is to choose an equation of state (EOS), such as the ideal gas law, to calculate the fugacity of the vapour phase. At higher pressures, a cubic E O S , such as the Peng-Robinson equation or Soave-Redlich-Kwong equation, should be used, but near atmospheric pressure, as is the case for this work, it is reasonable to assume that gases form ideal mixtures. The 42 Chapter 3: Literature Review vapour standard state is pure gas at the temperature and pressure of the mixture. If the ideal gas law is used, the vapour phase fugacity coefficient is set to unity: f,v = ^ y , f , v ' ° = y , P (3-5) Although good for hydrocarbon systems, equations of state have proved inadequate for modelling solutions with strong interactions in the liquid phase, such as aqueous mixtures of polar compounds. The phase equilibria of dilute aqueous solutions are therefore treated differently from those of dilute organic systems, due to water's unique structure and hydrogen-bonding characteristics. For aqueous systems, liquid phase behaviour can be described using the activity coefficients and a liquid standard state of the pure liquid at the temperature and pressure of the mixture: ffL = Yhfi* = KXiPrexp Pj S a t is the saturation or vapour pressure and v is the molar volume, of species i at temperature, T. The exponent term, referred to as the Poynting correction, accounts for the pressure induced deviation of fugacity from its saturated state. A t low pressures, including this work at or near atmospheric pressure, this factor approaches unity and can be taken as 1.0. At equilibrium, when the vapour and liquid phase fugacities are equal, it follows that the controlling equation for vapour-liquid equilibrium is: y i P = y i x i P i - (3.7) This equation relates the partial pressure of the compound in the vapour phase to the vapour pressure exerted by the compound in the liquid phase at the system temperature. Since ideal behaviour is assumed for the vapour phase, the compound vapour pressure is simply the product of the vapour phase mole fraction, y, of compound i , and the total pressure, P. The pressure exerted by the compound in the liquid phase is the product of the liquid phase mole fraction, x, of compound i , and the saturated state vapour pressure, Pj S a t , "corrected" by the activity coefficient, Y J . This approach presumes knowledge of the vapour pressure of each species at the temperature of interest. Correlations for the vapour pressure of all of the TRS compounds are given in Section 3.1.5. , ( P - P , s a ' ) R T (3.6) 43 Chapter 3: Literature Review In an ideal system, the activity coefficient is unity, and the above equation reduces to Raoult's law. In most real situations the activity coefficient deviates significantly from unity due to interactions between molecules in multi-component systems. Equation (3.7) also reduces to Raoult's law at the extreme limit where compound i approaches 100% concentration and the total pressure equals the vapour pressure of this compound. At the other extreme, where the concentration of compound i approaches zero, this equation can be simplified to Henry's law: y . p = - j ^ (3.8) K H i Henry's law states that the concentration of a solute gas in solution is directly proportional to the partial pressure of that gas above the solution, related by the Henry's constant, k H i . The accuracy of Henry's law increases as the liquid phase concentration of the solute decreases; thus, it is often used to describe the solubility of gases such as nitrogen and oxygen, which almost always exist in very low concentrations in the liquid phase of water-based systems. Henry's law can also be used to describe any substance at very low concentration, typically in the range of parts per mill ion, often referred to as the condition of "infinite dilution." The distribution coefficient, K , is defined as the ratio between the vapour and liquid mole fractions for each component: K{ = — (3.9) This term defines the volatility of an individual substance. The relative volatility, a, of a substance, is defined as the ratio between the vapour and liquid molar fractions for the solute compared to the solvent. For an aqueous system, the former is referred to as the contaminant and the latter as water: « i - w a , e r = ~ = ~ ( 3 - 1 0 ) ^ water * water X . water 44 Chapter 3: Literature Review The relative volatility of an aqueous system defines the relative ease with which the contaminant can be stripped from water, with the temperature dependency for common kraft mi l l contaminants shown graphically by Blackwell et al. (1980) in Figure 3.1. 50 85 120 155 190 T E M P E R A T U R E , °F Figure 3.1: Relative volatility of common kraft mil l contaminants (Blackwell et al., 1980) 3.1.2 Act ivi ty Coefficient Models The activity coefficient, V j , for substance i in a solution can be estimated using a number of correlations such as the Wilson equation developed by Wilson (1964), the non-random two-liquid ( N R T L ) equation developed by Renon and Prausnitz (1968) and the universal quasi-chemical 45 , Chapter 3: Literature Review ( U N I Q U A C ) equation developed by Abrams and Prausnitz (1975). These models are able to represent, with a reasonable number of adjustable binary parameters, the phase equilibria of highly non-ideal non-electrolyte systems. Experimental measurements must be carried out to determine these binary parameters; unfortunately, available data are limited. Lohmann and Gmehling (2001) state that there are about half a mil l ion binary systems of interest to the chemical industry and V L E data for less than 6000 have been published. If experimental data are unavailable, group contribution methods can be used to predict the activity coefficient. The U N I Q U A C functional-group activity coefficient ( U N I F A C ) model, developed by Fredenslund et al. (1975) and the newer Modified U N I F A C (Dortmund) model developed by Weidlich and Gmehling (1987) combine the solution of a functional group concept and the U N I Q U A C model. The group contribution method is based on the concept that a molecule consists of different functional groups, such as C H 3 , O H , or C H , and that the thermodynamic properties of a solution can be correlated in terms of these functional groups. The advantage of this method is that a very large number of mixtures can be described by a relatively small number of functional groups. The disadvantage, when compared to the other models, is that the parameters are not fitted from actual experimental data for the molecular solutions, so accuracy may be sacrificed. The addition of salts can alter the phase equilibria behaviour of systems containing one or more solvents. These effects include changing the vapour pressure for single solvent electrolyte systems and alteration of relative volatilities for mixed solvent electrolyte systems. Many correlations have been proposed to describe the effect of salts on the vapour pressure of mixed solvent electrolyte systems. Good overviews of empirical and semi-empirical equations that can be used for industrial applications are provided by Renon (1986), Leohe and Donohue (1997) and Anderko et al. (2002). In general, these models contain several contributions that determine the excess Gibbs free energy including: a short-range term represented by models developed for non-electrolyte systems, such as the N R T L equation discussed above; a long-range term represented by an equation based on Debye-Huckel or mean spherical approximation ( M S A ) theories; and an equation based on the Born model for the electrostatic contribution to ion solvation (Chen and Song, 2004). The most well established electrolyte model is the electrolyte N R T L (eNRTL) equation developed by Mock et al. (1986). This model extends the N R T L equation to include electrolytes 46 Chapter 3: Literature Review through the addition of solvent-salt binary parameters; therefore it builds on the extensive databank of solvent-solvent parameters available for the N R T L model. The e N R T L equation is the recommended option when using Aspen Plus (Aspentech), one of the simulation software packages used to model the results of this work.. The e N R T L equation is based on the idea that for relatively dilute electrolyte solutions, the interactions between the ions and the solvent can be represented locally and then generalized to the entire solution; thus, a long-range Debye-Huckel term and a Born term are not required. Mock et al. (1986) found that this approach reproduces experimental activity coefficients very well for dilute and moderately concentrated (up to 3 M ) solutions of strong electrolytes. A n updated and comprehensive overview of the e N R T L model was given by Chen and Song (2004). In this latest work, a Debye-Huckel term and a Born term were included to extend the accuracy of the e N R T L model to the entire concentration range. The e N R T L equation, incorporating adjustable binary parameters, AT and r, is expressed as follows: Z X J G Hrm)=~ J jm jm X • G m mm I x k G km m Z X k • G k m . • r k m . Z X k ' G k n V , k V c a £ X k - G k i kc.ac m m' S x k - G k I X k ' G km' kc.ac kc.ac ^ \p a ' ma.ca Z x k - G k Z x k - G kc.ac kaxa ka,ca r mc.ca Z X k - G k a c i (3.11) 47 Chapter 3: Literature Review For the e N R T L equation, the subscript m includes any solvent species, while the subscript c and a refer to the cation and the anion, respectively, of any ionic species. The subscripts i , j and k are defined as any species including both solvent and ionic. For molecular species, Xj = x j 5 while for ionic species, Xj = C^x,, where Cj is the absolute value of the charge number of ion i . The parameter G is determined from the energy parameter, r, and the non-randomness factor, a. G ^ e x p ^ - r , , ) (3.12) If the temperature is varied, rcan be modelled using the adjustable parameters a and b: ^ = a , i + Y ( 3 - 1 3 ) It should be noted that in the absence of electrolytes, the e N R T L equation reduces to the N R T L equation, i.e., the first two terms of Equation (3.11). Binary parameters for the N R T L equation are available from reference books and published articles for some solvent pairs, although, as discussed in Section 3.1.4, there are limited data for the T R S compounds in water. The data that do exist for T R S compounds suggest that the activity coefficients are large, ranging from about 20 for hydrogen sulphide up to about 1000 for dimethyl disulphide in water at 80°C (Blackwell et al., 1980); thus, they are said to have positive deviations from Raoult's law. Because the T R S compounds have relatively high vapour pressures and large activity coefficients, they are considered highly volatile; i.e., a relatively low concentration in the liquid phase wi l l be in equilibrium with a relatively high concentration in the vapour phase. For example, based on Figure 3.1, for a H 2 S -water solution at 80°C under a sealed air space at atmospheric pressure, 1 ppm of H 2 S in water wi l l be in equilibrium with about 1500 ppm H 2 S in the vapour phase. 3 . 1 . 3 Factors Affecting T R S Systems For a system including T R S in water there are a number of factors affecting V L E including: (i) pH effects, e.g., dissociation of hydrogen sulphide and methyl mercaptan in alkaline black liquor solutions, (ii) "binding" of these substances with metal and lignin ions in the liquor, (iii) the presence of electrolytes which alter phase equilibria, e.g., salting-in and salting-out effects of the pulping chemicals, (iv) reactions, e.g., oxidation of methyl mercaptan, (v) local physical constraints, e.g., 48 Chapter 3: Literature Review incomplete mixing or stratification in tankage, (vi) "tramp air" ingress, and (vii) insufficient time to reach equilibrium. 3 .1 .3 .1 p H E f f e c t s o n V L E In electrolyte systems, molecular species can remain undissociated, or dissociate partly or completely in solution. This phenomena is important to the determination of V L E since, near ambient temperature and pressure, the vapour pressure of ionic liquids is negligible, and only the undissociated fraction contributes to the vapour pressure (Earle et al., 2006). O f the TRS compounds, hydrogen sulphide and methyl mercaptan are weakly acidic and wi l l dissociate to the hydrosulphide and mercaptide ion, respectively (Shih et al., 1967a, 1967b). The equilibria formed by the ionization of hydrogen sulphide and methyl mercaptan are strongly dependent on the pH of the solution and only slightly affected by temperature. The vapour pressure of hydrogen sulphide is a function of p H because of the successive ionizations of hydrogen sulphide to the hydrosulphide and sulphide anions, which result in a decrease in the concentration of the undissociated hydrogen sulphide, and thus a decrease in its vapour pressure: H 2 S < K a ' >H + + HS"< K a ; > 2 H + + S ° (3.14) The extent of the dissociation is quantified by the acid dissociation constant, K a . For example, for the dissociation of hydrogen sulphide, Ka , = [H +][HS"]/[H 2S]. The hydrosulphide anion is an active cooking chemical that is formed when sodium sulphide hydrolyzes to hydrosulphide, which in turn dissociates, along with sodium hydroxide, to the sodium cation and the hydrosulphide and hydroxyl anions: N a 2 S + H 2 0 < >NaHS + NaOH<——>2Na + + H S " + O H " (3.15) Shih et al. (1967a) showed that the vapour pressure of hydrogen sulphide over its aqueous solution at various p H values is dependent on the concentration of undissociated hydrogen sulphide present in the solution. The fraction of hydrogen sulphide that exerts a vapour pressure, i.e., the fraction undissociated, x±, can be expressed as: H + ] 2 + Ka, [H + ]+ K a , K a 2 49 Chapter 3: Literature Review pH is defined as -log[H +] and pKa as -log(Ka). From these relationships, the concentration of the hydrogen ion, [FT], can be determined from the measured p H , and the value for K a can be determined from the p K a values given in Table 2.2. Shih et al. (1967b) also found that methyl mercaptan wi l l ionize according to the following equilibrium: C H 3 S H ( K a >CH 3 S~ + F T (3.17) The undissociated fraction of methyl mercaptan can be expressed as: + [ H + ] X M M = [ H + ] + K a ( 3 1 8 ) Based on the p K a values given in Table 2.2, above a pH of 9, very little undissociated hydrogen sulphide wi l l exist, and above a p H of 12, very little undissociated methyl mercaptan wi l l exist. This effect is represented graphically in Figure 3.2. Maintaining a high level of residual alkali in kraft pulping wi l l maintain high p H which increases the dissociated fraction of hydrogen sulphide and methyl mercaptan and decreases their vapour 50 Chapter 3: Literature Review pressure. The other organic T R S compounds, dimethyl sulphide and dimethyl disulphide, do not dissociate, therefore their vapour pressures are not a function of pH. 3 . 1 . 3 . 2 I n o r g a n i c D i s s o l v e d S o l i d s E f f e c t s o n V L E Electrolyte chemistry includes the dissociation effects discussed above and also the so-called "salt effects," where partial or complete dissociation of salts in solution alters the vapour pressure of the volatile components. This is commonly referred to as the "salting-in" or "salting-out" of a volatile species. When salt is added to a solvent mixture, a resulting increase in the relative volatility of one of the species is referred to as a salting-out of this component, while salting-in refers to the opposite effect. A s discussed in Section 3.1.2, this effect can be modelled by extending the activity coefficient N R T L equation to the e N R T L equation. A s described in Section 2.7, kraft black liquor contains many salts, including sodium hydroxide, sulphide, carbonate, sulphite, sulphate, and thiosulphate, along with others in smaller concentrations (Table 2.10). The total dissolved solids in black liquor include about 20 wt% inorganic salts (Table 2.9). Zhu et al. (2000) found that for typical softwood black liquor more than 95 wt% of the total dissolved inorganic materials (excluding organically bound sodium and sulphur) is sodium salts. Data found in the literature describing the effect of dissolved salts on the phase equilibrium of the TRS compounds are included in Tables 3.1 to 3.8. 3 . 1 . 3 . 3 O r g a n i c D i s s o l v e d S o l i d s E f f e c t s o n V L E The total dissolved solids in black liquor include about 80 wt% organic material, with about 30 wt% lignin (sodium free basis) included in this fraction (Table 2.9). Dissolved alkali lignin in black liquor consists mainly of sodium phenolates, so the total alkali lignin fraction, including the "organically combined sodium" fraction is about 40 wt % of the total dissolved solids in black liquor. The balance of the organic material consists of hemicellulose, sugars, extractives, acids, and numerous other unidentified organics. N o information could be found in the literature describing the effect of dissolved organic substances on the equilibria of any of the T R S compounds. 51 3 . 1 . 3 . 4 O t h e r E f f e c t s o n V L E Chapter 3: Literature Review A n under-reported phenomenon is "binding" of reduced sulphur compounds with metal or lignin ions in kraft liquor or condensate. Since only the unbound fraction of these compounds would contribute to the vapour pressure, this has a direct impact on phase equilibria. This phenomenon is mentioned in the literature (Wardencki, 1998), and although this effect has not been quantified, it is expected to be significant only at very low sulphide concentrations. The final effects to be considered are the deviations from equilibrium due to incomplete mixing, stratification in the liquid phase, tramp air ingress, or insufficient residence time to reach equilibrium. These potential effects wi l l be different for each type of process system or tankage in a kraft mi l l and wi l l depend on the configuration and residence time for the particular equipment. These effects wi l l be accounted for as part of the modelling process, which is discussed in greater detail in Chapter 6. 3 . 1 . 4 H e n r y ' s C o n s t a n t a n d A c t i v i t y C o e f f i c i e n t s f o r T R S A n extensive search of the literature was conducted to find information relating to the vapour-liquid equilibria of the TRS compounds in water, black liquor, and salt solutions. In almost all cases, these data were reported as a Henry's constant. For the most part, only sources containing original data are cited; i.e., Henry's constants or activity coefficients regressed from data generated by others were ignored when the original work could be found in the literature. Hydrogen sulphide is a common compound found in oi l and gas processing, effluent treatment, composting, and in food preparation (or spoilage such as from rotten eggs); thus, it is not surprising that a number of researchers have studied the properties of aqueous H 2 S systems. Less data are available for the organic T R S compounds, but some do exist as a result of research for specific issues. In atmospheric research, recent models of the global sulphur cycle point to a significant flux of sulphur from the oceans to the atmosphere and it has been proposed that dimethyl sulphide may provide the source for such a flux (Dacey et al. 1984). Knowledge of vapour-liquid equilibrium for dimethyl sulphide in salt water is helpful in study of this phenomenon; this has provided the impetus to study the effects of salt on the phase equilibria of a DMS-water system. 52 Chapter 3: Literature Review Przyjazny et al. (1983) reported Henry's constant for the three organic TRS compounds over a temperature range of 25 to 70°C at various concentrations of N a C l and N a 2 S 0 4 . Dacey et al. (1984) reported Henry's constant for D M S over a temperature range of -0.8 to 32.4°C for distilled water, "55% seawater" and "Sargasso seawater." Barrett et al. (1988) reported Henry's constants for H 2 S over a temperature range of 25 to 95°C at various N a C l concentrations. Carroll and Mather (1989) reported Henry's constants for H 2 S over a temperature range of 0 to 90°C, while Suleimenov and Krupp (1994) reported them for 20 to 90°C. De Bruyn et al. (1995) reported Henry's constants for H 2 S , M M and D M S over a temperature range of 5 to 25°C. Wong and Wang (1997) reported Henry's constants for D M S over a temperature range of 18 to 44°C but for an unspecified "seawater" concentration. Tormund (1997) reported Henry's constants for all of the T R S compounds over a temperature range of 40 to 80°C for N a + concentrations of 0.3 M and 3.0 M , but did not report for a salt-free system. Two sources not based on original work, instead composed of large databanks of compiled sources, were considered. The Aspen Plus version 11.2 (Aspentech) databank contains Henry's constant temperature dependent parameters for only one system, H 2 S in water for -0.15 to 59.5°C. Yaws et al. (2003) compiled Henry's law constants for many sulphur compounds in water from multiple sources including the C R C Handbook of Chemistry and Physics and his own Yaws ' Handbook of Thermodynamic and Physical Properties; data are provided for M M and D M S , but only at 25°C. Only Olsson and Zacchi (2001) have reported N R T L parameters for calculating activity coefficients, but they regressed their parameters based on data generated by others, with the exception of D M D S , for which they conducted some lab work. Unfortunately, they did not supply the experimental data on which they based their results. Nonetheless, a system approaching infinite dilution was assumed and these parameters were converted to a Henry's constant so that their data could be compared to others. Very little V L E data could be found for volatiles in black liquor. Zhu and Chai (1999) and Zhu et al.(2000a) published results for the phase partitioning in black liquor for methanol, but not for the TRS compounds. This data was used for simulation work for Cluster Rule compliance, with this discussed in more detail in Section 3.2.2. Olsson et al. (2000) published results for 15 common V O C s found in black liquor, including the TRS compounds. Unfortunately their results were limited 53 Chapter 3: Literature Review to only two samples and were based on testing TRS concentrations after the "aging" effect had stopped; aging can significantly alter results. For example, the concentration of dimethyl disulphide in one of their black liquor samples increased from 26 to 127 mg/kg over a few hours, with a resulting decrease from 225 to 120 mg/kg for methyl mercaptans. This aging effect is described in more detail in Section 3.4. The reported Henry's constant values in Tables 3.1 to 3.8 have all been converted to the same units of (mol/mol)/MPa, to facilitate comparisons. Henry's constant is typically presented at 25°C, but some researchers have provided temperature dependency data. The enthalpy of solution, - A s o l H , is the enthalpy change associated with the process when a component enters solution. The temperature dependance of Henry's constant can be expressed in the following form (Sanders, 1999): - d n ( k H ) A s o l H d(l/T) R Integrating from 298 K , the temperature dependency of Henry's constant is given by: (3.19) k „ = k- 8 K exp^ A S 0 l H 7 l 1 R V T 298 (3.20) Henry's constants at 298 K , and the temperature dependency constants, - A s o l H / R , for all literature sources are summarized in Tables 3.1 to 3.8. T a b l e 3 . 1 : Henry's constants for hydrogen sulp lide in water Source 1, 298K K H (mol/mol)/MPa -A S „ ,H /R ( K ) pH Temperature range (°C) Barrett et al. (1988) 0.018 2550 1.5 to 2.5 25 to 95 Carroll and Mather (1989) 0.019 2110 not specified Oto 90 Suleimenov and Krupp (1994) 0.01.9 1820 not specified 20 to 90 De Bruyn et al. (1995) 0.014 2080 <6.0 5 to 25 Aspen Databank 11. .2 0.019 2160 not specified -0.15 to 59.5 Olsson and Zacchi (2001) 0.016 1330 not specified not specified 54 Chapter 3: Literature Review T a b l e 3 . 2 : Henry's constants for hydrogen sulp lide in various salt solutions Source 1, 298K (mol/mol)/MPa - A S 0 , H / R ( K ) Salt content Temperature range (°C) Barrett et al. (1988) 0.015 2520 1 mol/kg N a C l 25 to 95 Barrett et al. (1988) 0.013 2460 2 mol/kg N a C l 25 to 95 Barrett et al. (1988) 0.012 2440 3 mol/kg N a C l 25 to 95 Barrett et al. (1988) 0.010 2390 4 mol/kg N a C l 25 to 95 Barrett et al. (1988) 0.009 2370 5 mol/kg N a C l 25 to 95 Tormund (1997) 0.025 2200 0.3 M N a + 40 to 80 Tormund(1997) 0.026 2530 3 . 0 M N a + 40 to 80 T a b l e 3 . 3 : Henry's constants for methyl mercaptan in water Source K 298K (mol/mol)/MPa - A s o l H / R (K) pH Temperature range (°C) Przyjazny et al. (1983) 0.070 3420 not specified 25 to 70 De Bruyn et al. (1995) 0.036 2800 <9.0 5 to 25 Yaws et al. (2003) 0.044 - not specified 25 Olsson and Zacchi (2001) 0.051 1630 not specified not specified T a b l e 3 .4 : Henry's constants for methyl mercaptan in various salt solutions Source u 298K K H (mol/mol)/MPa - A s o l H / R ( K ) Salt content Temperature range (°C) Przyjazny et al. (1983) 0.058 2400 0.70 M N a C l 25 to 70 Tormund (1997) 0.063 3490 0.30 M N a + 40 to 80 Tormund (1997) 0.030 2580 3 . 0 M N a + 40 to 80 55 Chapter 3: Literature Review Table 3 . 5 : Henry's constants for dimethyl sulphide in water Source 1, 298K (mol/mol)/MPa - A s o l H / R (K) Temperature range (°C) Przyjazny et al. (1983) 0.100 4080 25 to 70 Dacey et al. (1984) 0.099 3490 -0.8 to 32.4 De Bruyn et al. (1995) 0.085 3070 5 to 25 Yaws et al. (2003) 0.099 - 25 Straver et al. (2005) 0.088 - 25 Olsson and Zacchi (2001) 0.026 1620 not specified Table 3 . 6 : Henry's constants for dimethyl sulphide in various salt solutions Source K 298K (mol/mol)/MPa - A S 0 , H / R (K) Salt content Temperature range (°C) Przyjazny et al. (1983) 0.078 3880 0.7 M N a C l 25 to 70 Przyjazny et al. (1983) 0.087 4160 0.33 M N a 2 S 0 4 25 to 70 Przyjazny et al. (1983) 0.062 4230 0.66 M N a 2 S 0 4 25 to 70 Przyjazny et al. (1983) 0.044 3740 1 . 0 0 M N a 2 S O 4 25 to 70 Przyjazny et al. (1983) 0.032 3510 1.33 M N a 2 S 0 4 25 to 70 Dacey et al. (1984) 0.094 3250 10.5 ppt CT 0.2 to 28 Dacey et al. (1984) 0.085 3490 19.1 ppt Cl" Oto 29.1 Wong and Wang (1997) 0.077 4300 "seawater" 18 to 44 Tormund(1997) 0.071 3480 0.3 M N a + 40 to 80 Tormund(1997) 0.045 3130 3 . 0 M N a + 40 to 80 56 Chapter 3: Literature Review Table 3 . 7 : Henry's constants for dimethyl disulphide in water Source K 298K (mol/mol)/MPa - A S 0 , H / R (K) Temperature range (°C) Przyjazny et al. (1983) 0.162 4130 25 to 70 Olsson and Zacchi (2001) 0.046 2480 not specified Table 3 . 8 : Henry's constants for dimethyl disul ?hide in various salt solutions Source K 298K (mol/mol)/MPa - A S 0 | H / R (K) Salt content Temperature range (°C) Przyjazny et al. (1983) 0.117 4020 0.70 M N a C l 25 to 70 Przyjazny et al. (1983) 0.147 4110 0.33 M N a 2 S 0 4 25 to 70 Przyjazny et al. (1983) 0.098 4470 0.66 M N a 2 S 0 4 25 to 70 Przyjazny et al. (1983) 0.074 4620 1.00 M N a 2 S 0 4 25 to 70 Przyjazny et al. (1983) 0.064 4090 1.33 M N a 2 S 0 4 25 to 70 Tormund (1997) 0.102 3480 0.30 M N a + 40 to 80 Tormund(1997) 0.065 3690 3 . 0 M N a + 40 to 80 From inspection of the data presented in the tables above, it is clear that the addition of salt to a water system containing any of the TRS compounds wi l l increase the relative volatility of the T R S compound with respect to water (salting-out of TRS) . The data from Olsson and Zacchi (2001) do not agree very well with other sources. Their focus was on modelling of these compounds for foul condensate steam stripping systems; thus, they were concerned with distillation of these compounds at much higher concentrations, well outside the valid range for Henry's law. They may have regressed their activity coefficient parameters from experimental data only at higher concentrations without considering the accuracy at the infinite dilution extreme. O f the five parameters used to fit the N R T L equation, they held three of them constant (ay and a^  were set to zero, and a was set to 0.3) for their data regression. It is not clear why they did this, but, as discussed in Chapter 6, when fitting experimental data from this work, a good fit could not be achieved when holding these same parameters constant. 57 3 . 1 . 5 T R S V a p o u r P r e s s u r e Chapter 3: Literature Review The vapour pressure of a pure compound can be estimated using the extended Antoine's equation of the form (Perry's, 1997): C 2 l n f P i 8 " ^ C l j + ~y- + C3i -ln(T)+ C4 , - T C 5 i (3.21) Parameters for this equation for the T R S compounds and methanol are available in reference books, with the values given in Table 3.9 taken from Perry's (1997). One exception to this is dimethyl disulphide which was taken from the National Institute of Standards and Technology website (NIST); these parameters were given for a slightly different equation, so data generated from this correlation was regressed to determine the best fit parameters for Equation (3.21) and these are given in Table 3.9. T a b l e 3 . 9 : Constants for vapour pressure Equation (3.21) (Perry's, 1997; NIST) Compound C I C2 C3 C4 C5 Temperature range (K) H 2 S 85.584 -3839.9 -11.199 0.018848 1 187 to 373 M M 54.15 -4337.7 -4.8127 4.5000e-17 6 150 to 469 D M S 83.485 -5711.7 -9.4999 9.8449e-06 2 174 to 503 D M D S 4.573 -3841.4 3.2245 -0.00574 1 273 to 401 Methanol 81.768 -6876.0 -8.7078 7.1926e-06 2 • 175 to 512 3 .2 M o d e l l i n g o f t h e K r a f t P u l p i n g P r o c e s s Since the advent of computer technology, modelling of the kraft process using computer simulation has become popular with many researchers and engineers in the industry. Modell ing of industrial facilities can be used for many purposes including process optimization, sensitivity analysis, development of process control strategies, training and environmental reporting. Optimization in kraft pulp mills is often focussed on maximizing pulp yield and quality and energy efficiency. Efforts toward closure (e.g., recovery and re-use of hot contaminated process water) for 58 Chapter 3: Literature Review energy savings and minimizing environmental impact have increased the complexity of the process. With closure, solving heat and mass balances becomes more of a challenge because recycling of recovered waste streams back into the process increases the complexity of the iterative nature of the solution. Commercial simulation packages provide powerful tools that allow for analysis of very complex systems with minimal effort. Use of simulation software in the pulp industry to study environmental impact has recently become more popular, mainly due to the continuing improvements in computer hardware and software. Also driving this are new regulations and emissions reporting requirements: in the U.S . , the Cluster Rule, and in Canada, the N P R I , both discussed in detail in Section 2.5. Increased computing power, combined with user friendly commercially available software packages, provides the means, and the more stringent environmental reporting and regulations provides the impetus. After the initial investment to design and test the model, computer simulation could potentially be as inexpensive as the application of emissions factors, while providing much more accurate estimates of actual emissions. The first requirement of modelling the release of contaminants to the atmosphere is understanding how the contaminants enter the process or how they are generated within the system. The organic TRS compounds are mainly generated within the digester and there are a number of research papers which were discussed in Section 2.2 that provide details for estimating the amount formed. The second requirement is to understand how these compounds behave in equilibrium within the system. With this information, it is possible to estimate how the target contaminants move between each individual process unit and where they wi l l leave the system. T R S generation and release modules can then be integrated into a general heat and mass balance to produce an overall T R S emissions model. The T R S model can be used to generate a "base-case" heat and mass balance based on the existing process. The base-case model can be used to predict changes in emissions for changes to the operation or changes to any process equipment. The model can be used to determine the most effective and cost-efficient way to reduce emissions, either through modification to the equipment, changes to operation, or through installation of collection and treatment systems. 59 Chapter 3: Literature Review 3.2.1 Computer Software Mode l l ing Tools There are many software tools that can be used for modelling, with the most common being spreadsheet programs such as Lotus 123 or M S Excel. These are extensively used for their simplicity and flexibility. They often provide all the computing power and functions required to prepare heat and mass balances of simple processes. The drawbacks to spreadsheets include their lack of an integral thermodynamic properties database and their limitations for presentation. M S Excel can be linked to M S Vis io , a flowsheet drawing program, for presentation purposes, but more powerful simulators combine these features. The next step up are the commercially available general simulation programs such as Aspen Plus (Aspentech). These are very powerful programs which are relatively simple to use; they typically operate on an input screen user interface. These packages incorporate a back to front approach when compared to spreadsheets; i.e., with these packages the flowsheet drawing is typically completed first and the specifications for each stream are added later. This process is more intuitive and allows for greater ease of programming for more complex systems. They are designed to deal with processes that require iterative solutions, something that can be done on a spreadsheet but with greater difficulty. Typical features available with most commercial simulation packages include built-in properties databases, integrated thermodynamic mathematical models designed to represent the actual behaviour of many different fluid mixtures, integrated mathematical models for standard equipment such as heat exchangers and distillation columns, and dynamic capabilities. A number of modelling software packages are available that were originally developed for specific industries such as Aspen Plus, H Y S Y S , Chemcad and P R O S I M ; these were all originally aimed at the oi l and gas industry. In general, these include a database and modules that are tailored to their particular specific industry requirements. The pulp and paper industry has a number of unique requirements for process simulation. Typically, the simulation packages include industry-specific modules such as digesters, washers, screens, etc., and a properties database that includes the chemicals used in kraft pulping. They should also include other substances where the properties must be estimated, such as for fibre, lignin, dissolved solids, black liquor, etc. Many of these substances are not molar definable compounds, 60 Chapter 3: Literature Review and their exact composition is unknown; thus, they wi l l not be found in any standard substances database. This lack of specific compound information leads to the significant feature of all pulp industry simulation packages, that they are typically done on a weight rather than molar basis. The most common pulp mil l modelling packages include I D E A S (IC&S), Pulpmac (Pulpmac), W i n G E M S (Pacsim) and C A D S I M Plus (Aurel); details can be found on their respective websites. I D E A S was developed by the H . A . Simons Consulting group in Vancouver, B . C . , but was recently sold to the Andritz group based in Austria. I D E A S was used by H . A . Simons (later A M E C ) as a marketing device and a design tool for their own engineers in a construction or consulting role, while Andritz appears to be marketing the software package as a tool for researchers and mi l l personnel. Pulpmac was developed by the consulting group Papermac A B in Sweden. Both I D E A S and Pulpmac are based on the Extend Player platform, a commercially available software programming package supplied by Imagine That! (Imagine). Extend Player is a generic simulation platform that can be extended by programming modules to create a package for a specific requirement. W i n G E M S is a Fortran-based program which was developed at the University of Idaho in the early 1970's. W i n G E M S is now being marketed by Pacific Simulation of Moscow, Idaho, a subsidiary of Metso Automation. The software was developed for the pulp and paper industry and is used, as its name implies, for general energy and material balance systems ( G E M S ) . C A D S i m was originally conceived in the 1980's by the H . A . Simons process simulation group, beginning as a graphical interface, essentially a computer-aided design ( C A D ) flowsheet drawing, integrated with M A S S B A L , one of the original simulation software packages. Because of limitations integrating the two, H . A . Simons moved onto I D E A S development and an offshoot company, Aurel Systems of Burnaby, B . C . , moved on to develop C A D S i m Plus, based on the C programming language. C A D S i m Plus and W i n G E M S presently appear to be the dominant players in the North America pulping industry. 3.2.2 Previous W o r k Mode l l ing Non-Combust ion T R S Emissions The most extensive work to date on modelling kraft pulp mi l l non-combustion sources was done by a group led by Yongxiang Gu and Lou Edwards of the University of Idaho. They used the 61 Chapter 3: Literature Review computer simulation software package, W i n G E M S , which was originally developed by their group. For this work, they recently expanded its capability with the addition of methanol generation and release modules. They presented this work in series of four papers titled "Cluster Rule Compliance Tools" (Gu et al., 2001). The work focussed on methanol, since this is specified as a surrogate to determine compliance with the Cluster Rule for H A P emissions (U.S. E P A , 1998). This modelling work was supported by experimental work conducted at the Institute of Paper Science and Technology (IPST) in Atlanta led by Jun Yong Zhu. This group conducted research on methanol formation ((Zhu et al, 1999a; 1999b; 2000b; 2001) and methanol vapour-liquid equilibrium (Zhu and Chai,1999; Zhu et al., 2000a). They also did some work on T R S formation (Zhu et al., 2002; Yoon et al., 2003), although not on TRS vapour-liquid equilibria. In addition to the experimental work done at IPST, Gu et al. (2001) conducted experiments to determine the effect of alkali lignin on methanol equilibrium. They added an unspecified "commercial kraft l ignin" to a solution containing 800 mg/L methanol. They found that the addition of this alkali lignin had a relatively minor effect compared to the dissolved salts, with a mixture of 10 wt% alkali lignin increasing the relative volatility of methanol by about 15%. In fact the effect was so minor that the seven data points below 7 wt% actually showed the opposite effect, with the six data points between 7 and 12 wt% just pulling their linear fitted equation up to a slightly positive effect on the relative volatility. They also conducted equilibrium experiments with dissolved salts, preparing solutions containing either N a 2 S 0 4 , N a 2 S 2 0 3 or N a C l . They used the data from all of these salt solutions to develop an empirical correlation. They found that these dissolved salts had a relatively large effect, with a solution of 10 wt% salts increasing the relative volatility of methanol by about 60%. Figure 3.3 illustrates the order of magnitude effect on methanol vapour-liquid equilibrium that could be expected from the solids content of weak black liquor. The "total solids" line is the graphical representation of a purely empirical correlation developed by Gu et al., (2001), from laboratory tests on weak black liquor. For example, they determined that the 15 wt% solids content of a typical weak black liquor would increase the activity coefficient of methanol by about 50%). 62 Chapter 3: Literature Review weight % Figure 3 . 3 : Effect on activity coefficient of methanol due to the presence of dissolved inorganic and organic matter in black liquor (drawn from data from Gu et al., 2001) The "inorganics" line in Figure 3.3 represents their empirical correlation based on the data generated from the sodium salt solutions, while the "organics" line represents their empirical correlation for an alkali lignin mixture. One of the questions raised was whether this latter correlation could be extended for the entire organics content since some of these substances likely wi l l not have the same effect as alkali lignin. In fact, the extractives, being sparingly soluble in water, may have the opposite effect, suppressing the methanol activity coefficient through surfactant or other effects, but due to the great complexity of the black liquor mixture, simplifications had to be made to allow a model to be applied. For the next stage of their Cluster Rules work, Gu et al. (2001) incorporated an emissions module into W i n G E M S . The module uses the N R T L equation to calculate the binary methanol-water activity coefficient. This value is then adjusted using the empirical correlations described above to account for the shift due to the dissolved solids. The application of this module is limited to a system with no temperature change, i.e., the liquid and vapour outlet temperatures equal the 63 ; Chapter 3: Literature Review liquid inlet temperature. With this limitation in place, calculations can be simplified by assuming the outlet liquid flow equals the inlet liquid flow; i.e., the outlet vapour flow is very small compared to the liquid flow. The emissions module could then be programmed in such a way that the mass, energy and phase equilibria equations could be solved directly; otherwise the solution includes multiple non-linear equations, for which an iterative process would be required. This module can only be used for equipment where there is negligible temperature change across the system and where the vent flow is very low compared to the liquid flow, such as a storage tank. It can not be used to model process equipment such as a liquor flash tank, where there is a temperature change, and thus flashing of the liquid. For the final stage, they used mil l measurements taken from a single fibre-line of a southern softwood kraft mi l l to test their model. A l l air samples were collected and analysed by N C A S I and the liquor samples collected by N C A S I were delivered to IPST for testing. For this mi l l , they found they could predict methanol emissions reasonably accurately; their results for various process equipment are summarized in Figure 3.4. •a 1 a. a o c 1600 1400 1200 1000 800 600 400 - H 200 1 2 , 2 0 0 1 2 , 2 0 0 • Mill Measurement H Model, Eq.(l) Brn Atm 0 2 Blow Diffuser Tank 0 2 Atm Decker D 0 Tower D 0 Seal Eo Seal Diffuser Seal Tank Tank Tank Figure 3 . 4 : Comparison of methanol vent stack model predictions with mi l l measurements for various process equipment (Gu et al., 2001) 64 Chapter 3: Literature Review In the period from 1993 to 1997, process simulation studies for V O C emissions using W i n G E M S were completed for all the Mead mills in the U.S . (Venkatesh et al., 1997; Venkatesh, 1999; Venkatesh et al., 2000). Models were developed for a total of eight mills, referred to as M i l l A to H . Modelling was done with the aim of determining the most cost effective option for Cluster Rule compliance. The predictions made by these models were verified by comparing with sampling and testing results for the specific mi l l . They began by preparing a base-case model of existing operation and then developed alternative operating scenarios, including potential modifications to existing equipment, to determine compliance. For one Mead mil l ( M i l l E), Venkatesh (1999) reported data for T R S , although no details were provided on which T R S compounds were tested. No details were given regarding what correlations and what parameter values were used to model the phase equilibria. Figure 3.5 and 3.6 are scatter diagrams for M i l l E testing data versus predicted values for methanol and T R S . 950 -900 850 -800 -CO Q — 750 -700 -650 -650 700 750 800 850 900 950 G E M S Prediction Figure 3 . 5 : Comparison of G E M S prediction versus mi l l data for M i l l E for methanol concentration in the vapour phase (Venkatesh, 1999) 65 Chapter 3: Literature Review G E M S Prediction Figure 3 . 6 : Comparison of G E M S prediction versus mi l l data for M i l l E for TRS concentration in the vapour phase (Venkatesh, 1 9 9 9 ) No details were provided for the sampling location for the points on these scatter diagrams, only that they are for selected process streams at current conditions. The T R S predictions vary by as much as three times too high or too low; Venkatesh ( 1 9 9 9 ) states that some scatter in the T R S data was attributed to sampling issues. 3 .3 T R S Emissions Factors A n emissions factor is a representative value that attempts to relate the quantity of a pollutant released to atmosphere with an activity associated with the release of the pollutant. These factors are often expressed as the weight of a pollutant divided by a unit weight or volume of product produced. For example, the estimate of reduced sulphur emissions for kraft pulping equipment is expressed as kg sulphur emissions per tonne pulp produced. The NPRI reporting program directs Canadian facilities to use emissions factors developed by the U.S . Environmental Protection Agency (EPA) through their Technology Transfer Network (TTN). Emissions factors for a kraft pulping brown stock washer shown in Table 3.10 were extracted from Table 10.2-1 of "Compilation of A i r 66 Chapter 3: Literature Review Pollutant Emission Factors, Volume 1: Stationary Point and Area Sources, A P 42, Fifth Edition," available on the E P A Technology Transfer Network website (TTN). Table 3 . 1 0 : U .S . E P A Emissions Factors for Chemical Wood Pulping. S m refers to kg of pollutants released as sulphur ( R S H = methyl mercaptan, R S R = dimethyl sulphide, R S S R = dimethyl disulphide), and M g refers to millions of grams (i.e. tonne) of air-dried unbleached pulp (TTN) Location H 2 S (S m ) kg /Mg R S H , R S R , R S S R (S m ) kg /Mg Digester relief and blow tank 0.02 0.6 Brown stock washer 0.01 0.2 These emissions factors are based on the use of contaminated condensate for brown stock washing; i f fresh water is used, then the emissions factors are reduced by a factor of four. Emissions factors are also provided for other equipment such as the multiple-effect evaporators, recovery boilers, smelt dissolving tanks and lime kilns. These emissions factors, last updated in September 1990, are based on a mi l l where the pulp from a batch digester is directed to a blow tank and then to a washing line. There are no details given on the type of washing equipment, but the typical equipment used in that era was a series of vacuum drum type washers. A more modern mil l configuration, such as that used at the Howe Sound mil l where testing for this research was completed, consists of a continuous digester, with the pulp from the digester passing through a two-stage atmospheric diffusion washer before entering the blow tank. The stock is then processed through a vacuum drum type washer, called a decker washer, before going to the oxygen delignification system. It is questionable how valid the above emissions factors are for a mil l with a continuous digester and more modern washing equipment, but lacking other emission factor options, these are still in common use. Between 1996 and 2003, the Swedish Pulp and Paper Institute (STFI) led a U S $ 1 3 M program called the "Eco-Cycl ic Pulp M i l l . " This project brought together a number of research organizations and universities to study the possibilities and limitations of sustainable eco-balanced production of kraft pulp. A n overview of the project was presented by Peter Axegard, who headed the program (Axegard, et al., 2001). This work was completed and published in July 2003 as a 260 page report, one section of which provides information for a "reference m i l l " (Ledung, et al., 2002). 67 Chapter 3: Literature Review The reference mil l is a hypothetical pulp mi l l for the production of fully bleached market kraft pulp, consisting of sub-processes, each of which represents existing best available technology. It includes the latest in emissions control technology including C N C G and D N C G collection systems. The reference data are provided in the form of design or emissions factors. The basis for all calculations for the reference mil l are 11 9 0 , with this term defined as 1 air-dry metric ton of pulp. The factor specified for "sulphur to air" is 0.2 kg S/t9 0. 3 .4 T R S Sampling and Testing Considerations It is well documented that T R S samples are unstable and that analytical testing is a challenging proposition since these sulphides have absorptive, adsorptive, photo-oxidative, and metal catalytic oxidative features (Wardencki, 1998). O f the degradation reactions, oxidation is the main concern. Hydrogen sulphide and methyl mercaptan wi l l dissociate at high pH, such as in black liquor, to the non-volatile hydrosulphide and mercaptide ions, respectively (Reactions (2.3) and (2.7)). The hydrosulphide ion can oxidize to thiosulphate (Reaction (2.10)), while the mercaptide ion can oxidize to dimethyl disulphide (Reaction (2.9)). Once formed, dimethyl disulphide does not oxidize to any significant extent, but it wi l l undergo hydrolytic disproportionation under the influence of hydrosulphide and hydroxide ions (Bentvelsen et al., 1975). The complex mechanism is believed to involve an initial nucleophilic displacement of the sulphide bond. Following this rate limiting step, a series of reactions lead ultimately to formation of non-volatile methane sulphinic acid and regeneration of some methyl mercaptan: 3 C H 3 S S C H HS" >2CH 3 SO~ + 4 C H 3 S (3.22) 2 C H 3 S S C H OH" >CH 3SO;" + 3 C H 3 S (3.23) 68 Chapter 3: Literature Review The mercaptan formed in these reactions can then oxidize back to dimethyl disulphide: 1 2 C H 3 S " + - 0 2 + H 2 0 > C H 3 S S C H 3 + 2 0 H " (2.9) Given enough time, essentially all of the methyl mercaptan and dimethyl disulphide wi l l convert to methane sulphinic acid ( N C A S I , 2000). These oxidation reactions are the basis for common commercial processes referred to as oxidation systems (Bentvelzen et al., 1976). For example, air or oxygen is reacted with black liquor in weak or strong black liquor oxidation systems ( W B L O x or S B L O x ) to convert the T R S compounds to non-volatile forms, to prevent venting of these odorous gases to atmosphere in downstream process equipment. The extent of these oxidation reactions is limited by the quantity of oxygen in the liquor samples. Dimethyl sulphide appears to be slightly more stable, but it also degrades over time. McKean et al. (1965) state that dimethyl sulphide may disproportionate under the influence of hydroxide ions to form methyl mercaptan and methanol, although they found the extent of this reaction to be negligible: C H 3 S C H 3 + O H " » C H 3 S " + C H 3 O H (3.24) Another possible reaction, in the presence of hydroxide ions, is that methyl mercaptan may disproportionate into dimethyl sulphide and hydrogen sulphide: 2 C H 3 S H 0 H" ) C H 3 S C H 3 + H 2 S (3.25) McKean et al. (1965) found this reaction to have a slow but significant rate, while Douglass and Price (1966) found this reaction to be negligible. For black liquor, due to the alkaline conditions, the pH is typically above 11.5 (Grace et al., 1989) and the concentrations of hydrogen sulphide and methyl mercaptan wi l l theoretically be near zero (Figure 3.2). Since methyl mercaptan is expected to be below a measurable level, the loss of the mercaptide ion through oxidation is not a direct concern, but the generation of dimethyl disulphide from this oxidation reaction (Equation (2.9)) can compromise results i f it is formed after the sample is collected. To avoid this, during collection and storage, care should be taken to avoid introducing oxygen to black liquor samples. 69 Chapter 3: Literature Review Sulphides wi l l strongly adsorb to glass, although glass surfaces can be deactivated to minimize this effect ( N C A S I , 2002). N C A S I states that deactivation should only be required for glassware that comes into contact with fluids containing sulphides at a concentration less than 50 pg/L ( ~28 ppb) as sulphur. A l l of these effects discussed above are time dependent; thus, to minimize degradation effects, for the testing programs, the samples should be tested as soon as possible after preparation or collection. 70 Chapter 4: Research Objectives Chapter 4 Research Objectives The general objective of this work was to develop and test a method to predict emissions of TRS compounds from kraft pulp mi l l non-combustion sources based on modelling using vapour-liquid equilibrium correlations. To determine i f this was possible, phase equilibria behaviour for the T R S compounds must be known, and operating data from a kraft pulp mi l l must be available to test the model. The specific objectives of this research were: 1. To develop a V L E emissions model that accounts for the effects of the dissolved solids in kraft black liquor. 2. To conduct phase equilibria testing to determine activity coefficients that represent the V L E of the T R S compounds in kraft black liquor. 3. To identify a method to describe the effects of kraft black liquor composition and temperature on the V L E of the T R S compounds and to find the best fit for this correlation using the phase equilibria testing data. 4. To conduct a T R S sampling and testing program at a kraft pulp mi l l . 5. To test the V L E emissions model using the results of the mi l l testing program. 6. To demonstrate a practical use for the V L E emissions model by constructing a base-case heat and mass balance based on the mi l l testing data and then modifying this simulation to evaluate the effect on T R S emissions from potential process modifications. Descriptions of the material and methods, required to complete the objectives concerned with sampling and analytical testing, are given in Chapter 5, with results and discussion in Chapter 6. 71 Chapter 5: Materials and Methods C h a p t e r 5 M a t e r i a l s a n d M e t h o d s To support the development of a V L E model used to predict T R S emissions from kraft pulp mills, two testing programs were conducted. The mil l testing, conducted at the Howe Sound mill located in Port Mellon, British Columbia, was conducted first; the results are tabulated in Section 6.4. During mi l l testing, it was found that over 90% of the sulphur in the vent gases from the brown stock washing area was in the form of D M S . Based on this, the decision was made to use D M S as a surrogate for modelling of TRS emissions and all subsequent testing focussed on this compound. Phase equilibria testing was done in a lab environment to investigate vapour-liquid equilibrium behaviour of D M S in various solutions, including black liquor. For both testing programs, a gas chromatograph was used to measure concentrations of the T R S compounds. 5.1 G a s C h r o m a t o g r a p h i c E q u i p m e n t A Varian 3800 gas chromatograph (GC) fitted with an 1177 split/splitless injector and a pulsed flame photometric detector (PFPD) was used for testing. A P F P D has the unique characteristic of high selectivity of sulphur compounds compared to hydrocarbons (105:1). The detector was operated in sulphur mode, which is highly specific to the sulphides tested and provided a detection limit of about 0.1 ppm (mol). A s described in Section 2.7, black liquor is a complex mixture of hundreds of compounds, including many hydrocarbons and sulphides. A P F P D , when compared to the more common flame ionization detector (FID), wi l l provide distinct sulphide signal peaks with little interference from the hydrocarbon peaks, thus providing a more accurate result. The G C was fitted with an AT-5 column, 30 m long and 0.53 mm diameter, with a 5.00 pm film thickness (Alltech), which provided good separation of sulphur compounds. Helium at 2.5 mL/min was used as the carrier gas. The P F P D hydrogen flow and two air flows were set to 13, 18 and 10 mL/min respectively, the gate delay to 4.0 msec, the gate width to 3.0 msec, the trigger level to 200 m V , and the photomultiplier voltage to 550 V . Figure 5.1 shows the typical G C output for a gas sample, in this case the gas standard used to prepare the calibration curves. 72 Chapter 5: Materials and Methods mvol 300 200 100 DMDS H2S MM DMS 1 2 3 4 5 6 7 Figure 5.1: Typical gas chromatograph output mm The square root of the P F P D signal (the area under the peak) provides a linear relationship to the quantity of sulphur injected. For example, the concentration of each sulphide in the gas standard is equal at nominal 50 ppm (mol) each, but the area under the peak for D M D S is four times larger due to the presence of 2 moles of sulphur per mole of compound (a disulphide), whereas the others have 1 mole of sulphur per mole of compound. Borosilicate sample vials (Cole-Parmer), fitted with blue Teflon and silicone septa and polypropylene caps, certified for volatile organic analysis, nominally 20 m L in volume, but measured at 24.1 mL, were used for the testing techniques described below, including preparation of liquid standards. Gas samples were collected into 1 litre Tedlar bags (Mandel, P10 5, 7" x 7") fitted with dual on-off / septum valves. These bags were also used for preparing dilutions of the gas standard. Tedlar bags are widely used at kraft pulp mills for collecting T R S samples because they have very low permeability and are inert and solvent resistant. They are recommended for many U.S . E P A sampling methods and are specifically required by Environment Canada (1992) for use in their T R S reference method EPS l / R M / 6 , which was used here for the mi l l testing program. 73 Chapter 5: Materials and Methods 5 .2 P h a s e E q u i l i b r i a T e s t i n g As discussed in Section 3.1.4, available data from the literature describing the V L E of D M S in water are limited, especially when effects due to temperature and dissolved solids, such as those in kraft black liquor, are taken into account. The addition of dissolved solids, and in particular electrolytes such as the sodium salts found in black liquor, is known to alter the phase equilibrium behaviour of the system. Phase equilibria testing was done to attempt to quantify the effect on V L E of the presence of the dissolved inorganic and organic substances found in black liquor. The first purpose of phase equilibria testing was to determine Henry's constants and activity coefficients for a binary solution of D M S in water over a temperature range that covered typical operating temperatures for the brown stock washing area of a kraft mi l l , i.e., 70 to 90°C. The range of study was expanded to cover 25°C to allow comparison of results to the previously published results. To cover this range, and to provide enough data points to fit correlations used for modelling, all solutions were studied at 20, 50, 70 and 90°C. The second purpose was to quantify the effects on the V L E when organic or inorganic matter such as that contained in black liquor is added to a DMS-water solution. This was done by preparing artificial liquor solutions with some of the dissolved substances found in the highest concentrations in black liquor, including sodium salts and alkali lignin. These were tested, along with black liquor samples collected from the mi l l . The data from the artificial liquor solutions could then compared to the black liquor results to see i f any conclusions could be drawn as to what substances in black liquor contributed to the altering of D M S phase equilibrium. 5.2.1 P h a s e E q u i l i b r i a T e s t i n g M e t h o d Headspace gas chromatographic (HSGC) methods give a direct quantitative analysis of the vapour of a liquid sample matrix and are therefore commonly used for vapour-liquid equilibria studies (Zhu and Chai, 2005). The traditional H S G C method requires quantitative determination of the solute concentration both in the vapour and liquid phases through direct measurement. This has the drawback of requiring preparation of both liquid and gas standards with calibration procedures, all o f which introduce potential errors into the experimental measurements. This method also has 74 Chapter 5: Materials and Methods the drawback of requiring direct injection of the liquid sample, in this case black liquor, into the G C . Black liquor is an aqueous mixture of many compounds and direct injection introduces complex matrices. Black liquor is also viscous and corrosive which can plug or damage the G C column. Chai and Zhu (1998) developed an indirect headspace method that avoids these potential issues. At no time is the absolute concentration of T R S quantified in either the vapour or liquid phase, and only vapour samples are injected into the G C . The method consists of injecting a known volume of liquid sample into a sealed sample vial, and injecting a known, but lesser volume, of liquid sample into a second vial. These are then gently shaken until they come to equilibrium. A n equal volume of headspace sample is then withdrawn from each vial and injected into the G C . From the vial volume data, along with the G C output, the "dimensionless" Henry's constant, H c , can be determined: (5.1) C g and C, are the concentration of the solute (DMS) in the gas and liquid phases, V, and V , are the liquid sample injection volume and the vial empty volume, A is the G C signal output (the area under the curve representing the mass of D M S injected), and the superscripts 1 and 2 refer to the first and second sample vials. Even though the Henry's constant used here is referred to as dimensionless, this is not strictly true; the actual units are in the form of molar volume in the gas phase over molar volume in the liquid phase. From the dimensionless Henry's constant, the Henry's constant referred to in Equation (3.8), as well as the activity coefficient referred to in Equation (3.6), can be calculated. Although other solutes, such as methanol, can influence the V L E of D M S , they are typically in such low concentrations that the effect would be expected to be negligible. Nevertheless, as described in Section 5.2.3, to negate any possible effect, volatile solutes were stripped from the black liquor samples. A known amount of D M S was then reintroduced to the sample to be tested. For best accuracy with a system involving a highly volatile substance such as D M S with a 75 Chapter 5: Materials and Methods relatively large H c (ranging from about 0.05 to 0.5 in the temperature range 20°C to 90°C), a volume ratio of about 30 was recommended; thus, 10 m L of sample was injected into vial 1, and 0.3 m L was injected into vial 2. 5.2.2 R e c o v e r y o f A l k a l i L i g n i n F r o m B l a c k L i q u o r Two sources of alkali lignin powder were used for phase equilibria testing, one purchased from Sigma-Aldrich and the second recovered from black liquor collected at the Howe Sound mi l l . The alkali lignin was recovered from two litres of l s l stage diffusion washer filtrate liquor, collected on September 22,2006, when the mi l l was pulping 100% Hemlock. A method described by Ohman (2006) was used to precipitate lignin from the black liquor. The liquor was placed in a 3 litre flask, held at 80°C and continuously stirred on a combination hotplate / mixer. The solution was sparged with carbon dioxide from a gas cylinder via VA" tubing and a porous stone until it reached a p H of 9. A t this pH, a large fraction of the lignin is expected to have precipitated. The recovered alkali lignin was filtered (Whatman N o . l filter paper). It was then washed with a 0.005 M (pH = 2) solution of sulphuric acid, followed by a wash with distilled water, with this washing sequence repeated. The alkali lignin, resembling a light brown paste, was dried overnight in an oven at 105°C. The dried lignin, in the form of crumbly chunks, was pulverized into a fine powder using a mortar and pestle and stored in a sealed jar. Approximately 35 grams of alkali lignin were recovered. 5.2.3 P h a s e E q u i l i b r i a S a m p l e S o l u t i o n P r e p a r a t i o n A sample solution of D M S in water was prepared in a 1000 m L flask (referred to as solution 1). The flask was filled to 1000 m L with distilled water and sealed with a cap fitted with a septum. The headspace was purged for 10 minutes using 250 mL/minute of nitrogen gas. A n approximately 1000 ppm (mol) solution of D M S in distilled water was prepared using liquid D M S (Sigma-Aldrich, >99%). One m L of this solution was injected into the 1000 m L flask to create a test solution of approximately one ppm (mol) D M S . A s mentioned earlier, the exact concentration does not need to be known for the analytical technique used, only that the solution be in the infinite dilution range. 76 Chapter 5: Materials and Methods Several more D M S solutions were prepared, each of these spiked with organic and inorganic substances found in the highest concentrations in black liquor. These substances include alkali lignin, which typically makes up about 40 wt% of total dissolved solids, and sodium salts which typically make up about 20 % (Table 2.9). Weighed amounts of these substances were added to the flask before it was brought to volume with distilled water. Two alkali lignin mixtures of 3 and 6 wt% were prepared by adding 30 g and 60 g of the Sigma-Aldrich alkali lignin to distilled water (solutions 4 and 5). This range was chosen to represent typical black liquor lignin concentrations found in the brown stock washing area of a mi l l . The source and purity of this lignin could only be partly confirmed. The supplier's data sheet stated that it was isolated from a commercial pulp mi l l using Norway spruce as the wood source; it may contain ash (3-7%); it can form a 10% aqueous slurry at a p H of 6 to 7; and it is soluble in aqueous alkali solutions above a pH of 11. The pHs of the 3 and 6 wt% alkali lignin mixtures were measured at 8.7 and 9.1 respectively. Although these mixtures had the appearance of a dissolved solution (no turbidity and no settling of suspended solids, even days later), they are possibly slurries, but for simplicity of terminology, they wi l l be referred to as solutions. Another 3 wt% lignin mixture was prepared using the lignin isolated from black liquor collected at the Howe Sound pulp mil l (solution 3). When preparing a mixture from the recovered alkali lignin, to ensure that this lignin was completely dissolved, the p H was adjusted to 12 with the addition of 0.08 wt% N a O H . A "baseline" solution was also prepared by adjusting distilled water to a pH of 12 using 0.04 wt% N a O H (solution 2). Three sodium salt solutions of 2,4 and 6 wt% were prepared by adding 20,40, and 60 grams of a mixture of sodium salts to distilled water (solutions 6, 7 and 8). The total concentration was chosen to represent typical total sodium salt levels in brown stock washing liquor, while the composition was chosen to be representative of typical softwood black liquor (Table 2.10). The composition of the salt solution is shown in Table 5.1, along with the amount of each salt added for the 2 wt% solution (double and triple these amounts were added to prepare the 4 and 6 wt% solutions). Sodium sulphide was added as N a 2 S 9 H 2 0 ; the total weight of this salt added was adjusted accordingly. 77 Chapter 5: Materials and Methods T a b l e 5 . 1 : Sodium salt composition used for preparing salt solutions S a l t C o m p o s i t i o n , w t % G r a m s p e r 1 0 0 0 m L N a O H 10.0 2 N a 2 S - 9 H 2 G 17.5 3.5 (10.8 with H 2 0 ) N a 2 C 0 3 32.5 6.5 N a 2 S 0 3 10.0 2 N a 2 S 0 4 17.5 3.5 N a 2 S 2 0 3 12.5 2.5 Black liquor samples were collected from the Howe Sound mi l l on three occasions, June 8, June 20 and September 22, 2006 (solutions 9 to 13). Samples from the 1 s t stage diffusion washer filtrate tank and the decker washer filtrate tank were collected into 3 litre glass bottles, brought back to the lab and stored in a refrigerator at 4°C. The June 8 samples were collected when the mi l l was running at reduced capacity; it appears that the diffusion filtrate sample was more dilute than usual. The mi l l was running at full capacity and under stable operation when the June 20 samples were collected and at 90% of capacity and stable operation during the September 22 sample collection. The wood species being pulped were 50% Hemlock and 50% Interior Douglas Fir for the June samples and 100% Hemlock for the September samples. To ensure there was no interference with the G C signal for D M S from the other sulphur compounds, the black liquor samples were stripped of volatiles by purging with nitrogen gas at 250 mL/minute (foaming was noticed at higher rates) for 30 minutes. This was done by filling a 1000 m L graduated cylinder with black liquor and placing this in a fume hood. Nitrogen was introduced through a porous stone supplied by W tubing from a nitrogen cylinder. Periodic testing of liquor samples during purging of the first sample indicated that 30 minutes purging was sufficient to reduce the volatiles to a level where they did not interfere with the dimethly sulphide signal. The black liquor sample was transferred to a sealed flask, and the flask headspace was purged with nitrogen for about 10 minutes to minimize potential oxidation reactions in the black liquor. A s with the other solutions, one m L of the 1000 ppm (mol) D M S solution was injected into the black liquor flask to 78 Chapter 5: Materials and Methods produce an approximately I ppm (mol) D M S black liquor solution. A l l solutions tested are summarized in Table 5.2. Table 5.2: Dimethyl sulphide solutions used for phase equilibria testing Solution Details 1 Distilled water 2 Distilled water, adjusted to a pH of 12 using 50% N a O H 3 3 wt% alkali lignin in distilled water, adjusted to a pH of 12 using 50% N a O H , lignin powder recovered from 1 s t stage diffusion washer filtrate (solution 13) 4 3 wt% alkali lignin in distilled water, lignin powder purchased from Sigma-Aldrich 5 6 wt% alkali lignin in distilled water, lignin powder purchased from Sigma-Aldrich 6 2 wt% sodium salts in distilled water, salt composition given in Table 5.1 7 4 wt% sodium salts in distilled water, salt composition given in Table 5.1 8 6 wt% sodium salts in distilled water, salt composition given in Table 5.1 9 Decker washer filtrate, collected June 8, 2006 from the Howe Sound pulp mi l l , when pulping 50% hemlock / 50% Interior Douglas Fir 10 1s t stage diffusion washer filtrate, collected June 8, 2006 from the Howe Sound pulp mi l l , when pulping 50% hemlock / 50% Interior Douglas Fir 11 Decker washer filtrate, collected June 20, 2006 from the Howe Sound pulp mi l l , when pulping 100% Hemlock 12 1 s l stage diffusion washer filtrate, collected June 20, 2006 from the Howe Sound pulp mi l l , when pulping 100% Hemlock 13 1s t stage diffusion washer filtrate, collected September 22, 2006 from the Howe Sound pulp mi l l , when pulping 100% Hemlock 79 Chapter 5: Materials and Methods 5.2.4 Phase E q u i l i b r i a Testing Procedure On testing days, the solutions in the 1000 m L flasks were stored at room temperature and in a dark cupboard to minimize photo-oxidation effects. The sample vials were purged with 250 mL/min of nitrogen for a minimum of three minutes and then preheated to the testing temperature for a minimum of 30 minutes. The volume of sample injected into each sample vial was determined by weight difference. A preheated sample vial was weighed, and either 0.3 or 10 m L of sample was injected into the vial using disposable 1 m L and 10 m L syringes (Cole-Parmer), respectively. The vial was equalized to atmospheric pressure by inserting a syringe needle into the vial septum while the sample was being injected. The sample vial was again weighed, with the difference being the amount of sample injected. Densities of all solutions were determined, by weight difference using 50 mL flasks, to allow for conversion between volume and weight of solution. Once the sample was injected into the vial, the vial was fastened to a tumbler (Barnstead / Thermolyne Labquake) rotating at 15 rpm and placed in an oven ( V W R 1350FM, ±1 °C). The slow tumbling action was sufficient to provide good contacting between the liquid and vapour phase without heating the sample, which could be a factor at more vigorous rates of mixing. The time to reach equilibrium varied depending on the volume of sample injected into the vial and the testing temperature. It was found that six minutes was sufficient for the smaller sample size (0.3 mL) to reach equilibrium at 20°C, while eighteen minutes was the longest time required, in this case for the larger sample size (10 mL) at 90°C (Figure 5.2). The longer time at higher temperatures may be due to the fact that more of the volatile substance must be transferred to the vapour phase at higher temperature to reach equilibrium. It may also have to do with the fact that the sample must first be heated from room temperature to the final temperature before equilibrium can be established. 80 Chapter 5: Materials and Methods Time (minutes) Figure 5.2: Time for a 1 ppm (mol) solution of D M S in 10 m L of water in a 24.1 m L sample vial at 90°C to reach equilibrium Once the vial in the oven had come to equilibrium, a known volume of headspace sample was manually injected into the G C using a gas-tight syringe (Hamilton, 500 pL) , also kept in the oven. This syringe was fitted with a Chaney adaptor (Cole-Parmer), which promised reproducibility of ±1 % when injections of identical volumes are required, which was the case for this procedure. To minimize heat loss during injection, the area of the syringe containing the sample was insulated with foam. The sample vial, when removed from the oven, was slipped into an insulating foam sleeve also kept in the oven. To minimize cooling of the sample between removal of the vial from the oven and injection of the headspace sample into the G C , a quick procedure was adopted. First the G C was prepared to receive an injection, then the syringe was removed from the oven along with the sample vial, which was placed in the foam sleeve, and then a known volume of headspace sample was drawn into the syringe and injected into the G C . The time between removing the sample from the oven, to injection of the sample into the G C was approximately 10 seconds. The sample vial was then placed back in the oven in preparation for the next injection. 81 Chapter 5: Materials and Methods The G C injector was set at 200°C at a 75:1 injection split ratio. To minimize the time between injections, the G C oven was programmed to hold at 220°C, which resulted in a D M S peak at about two minutes and a total turnaround time between injections of three minutes. To keep the amount of D M S injected within the detection range of the G C , the injection volume ranged from 30 p L to 100 p,L, with the smaller volume used at 90°C where the D M S is more volatile and the headspace concentration higher. The exact volume injected is not important, only that the exact same amount is injected for each sample pair, ie., the vials containing 0.3 m L and 10 m L of the same sample. The injection volume was maximized to push the signal to the upper edge of the detection limit for the 10 m L sample without overloading the G C . This was done to maximize the strength of the signal for the 0.3 mL sample, which was near the lower detection limit. A t least three injections were done from each vial, with the G C output from these averaged to give the results ( A l and A2) required in Equation (5.1). Each run, including time to prepare the two vials, time in oven and time for G C operation averaged out at about 45 minutes. To minimize random error, at least three runs per sample per temperature were conducted. 5 . 2 . 5 T e s t i n g o f S o l u t i o n s f o r O t h e r P r o p e r t i e s The ash content of alkali lignin powder was determined by weighing out a 2 gram sample into a crucible and placing it in the oven at 105 "C for 4 hours to remove any moisture. The crucible was then re-weighed and placed in a muffle furnace at 650°C for 20 hours, after which it was weighed again, with the ash portion determined by weight difference. The ash resembled a light blue crystalline powder. Ash determination procedures typically specify 550°C (Grace et al., 1989), but 650°C was used on the recommendation of Paula Parkinson (Civ i l Environmental Lab, University of British Columbia) on the basis that sodium wi l l not volatilize at this higher temperature, and it w i l l assist in driving off the organics, providing a better sample for further tests. The purchased and recovered alkali lignin ash samples were then tested by the C i v i l Environmental Lab by atomic emission flame photometry using an air-acetylene flame to determine the sodium content. 82 Chapter 5: Materials and Methods The total dissolved solids of the black liquor samples was determined by weighing out a 10 gram sample into a crucible and placing in the oven at 105°C overnight. The water content was determined by weight difference after drying. The ash fraction was determined by placing the crucible in a muffle furnace at 550°C overnight. For mil l testing, the black liquor total dissolved solids concentration was determined using a moisture analyser (Omnimark Inst. Corp., Mark 3). A sample of black liquor collected at the Howe Sound mi l l was tested by Econotech (Delta, B . C . , Canada) by inductively coupled plasma (ICP) atomic emission spectrometry to determine the elementary composition of the dissolved solids. The pH of the solutions was measured using a bench-top meter (ThermoOrion, model 710). The pH meter was calibrated using buffered solutions of pH 4.0, 7.0, and 12.0. The conductivity of the solutions was measured using a hand-held meter (ThermoOrion, model 105). Distilled water dilutions of 60, 20, and 10:1 of the 6 wt% sodium salts solution (solution 8), and 20, 10, and 5:1 of the black liquor (solution 13) were prepared. These dilution ratios were used to bring the conductivity of these solutions into the same range as the alkali lignin mixtures (solutions 3, 4, and 5). Dilutions were prepared by weighting out 100 grams of distilled water in a flask and adding the appropriate amount of solution. 5 .3 M i l l T e s t i n g A mil l sampling and testing program was conducted around the brown stock washing area of the Howe Sound pulp mi l l . Process liquid and vapour samples were collected and analytical testing was completed to determine the concentration of T R S compounds. Because T R S samples are time sensitive, with significant degradation in less than an hour, as discussed in Section 3.4, the testing equipment, including the gas chromatograph, was relocated to the mi l l site. A number of challenges were overcome to conduct this testing including transportation of fragile equipment to the mi l l site, lack of sample points, poorly located sample points, sample instability, process instability, and unscheduled mil l shutdowns. 83 5.3.1 The Howe Sound Pulp and Paper M i l l Chapter 5: Materials and Methods , The kraft mi l l was completely modernized in 1989. Fibre-line equipment includes: * Kamyr two-vessel hydraulic continuous digester rated for 1300 ADt /d * Kamyr brown stock two-stage atmospheric diffusion washer (located on top of the blow tank) * Ingersol-Rand pressure knotting and screening * Corrudek brown stock decker washer * Kamyr two-stage medium consistency oxygen delignification * Kamyr post oxygen delignification single-stage atmospheric diffusion washer * Corrudek post oxygen delignification washer * DEopDnD bleach plant * Fourdrinier pulp machine and Flakt dryer. The kraft power and recovery equipment includes: * B & W recovery boiler rated for 386 t/h * B & W bark power boiler rated for 239 t/h * Kamyr falling film evaporators rated for 456 t/h * Al l i s Chalmers lime kiln rated for 388 t/d * Dorr Oliver recausticizing. Odorous non-combustion source T R S emissions are collected into two N C G systems. The concentrated N C G ( C N C G ) system sources include the digester flash steam condenser and the black liquor evaporator. The dilute N C G ( D N C G ) system sources include the digester steaming bin (chip bin), brown stock washer hoods and storage tanks for unwashed brown stock fibre and black liquor. The C N C G is burned in the lime kiln and the D N C G in the power boiler. A l l samples tested for this research were collected from D N C G sources in the brown stock washing area. Figure 5.3 is a simplified overview of the brown stock washing process and shows the location of the seven vapour and eight liquor sample points. 84 Chapter 5: Materials and Methods VS1 DIFFUSER 2nd STAGE DIFFUSION WASHER FILTRATE TANK < L S 2 ) | DIFFUSION WASHER FILTRATE 1st STAGE DIFFUSION WASHER FILTRATE TANK STOCK FROM DIGESTER VS = Vapour Sample Point LS = Liquid Sample Point DECKER WASHER FILTRATE TANK Figure 5 . 3 : Howe Sound mil l brown stock washing area overview showing liquid sample (LS) and vapour sample (VS) point locations The brown stock washing process is designed to separate the residual cooking chemicals, dissolved lignin and other undesired products from the pulp stock. The operating pressure of the digester pushes the stock through the blow-line to the bottom of the 1 s t stage diffusion washer (located on top of the blow tank). The stock passes up through the washer through screens that introduce wash liquor and extract washer filtrate. This is repeated in the 2 n d stage diffusion washer, with the product stock falling through two stock chutes into the blow tank. The stock in the blow tank is diluted and pumped through screens and knotters to the decker washer. Filtrate from the post-oxygen delignification system washer (referred to as 0 2 filtrate) is used as wash water on the vacuum drum type decker washer. Some decker washer filtrate is used for dilution in the blow tank, knotters, screens and decker washer, with the balance introduced as wash water to the 2 n d stage diffusion washer. The filtrate from the 2 n d stage diffusion washer filtrate tank is used for wash water in the 1 s t stage diffusion washer, with the filtrate from the 1 s t stage diffusion washer filtrate tank sent to the chemical recovery process. A s can be expected when conducting testing in an operating industrial environment, operation of the mi l l may be variable at times, including an unscheduled mil l shutdown in late 85 Chapter 5: Materials and Methods September that delayed testing, but in general, the process was stable on the most days that samples were collected. The only exception to this would September 28, 2005, when the mi l l was coming out of an unscheduled six-day shutdown. Most of the samples were collected around the diffusion washer and filtrate tanks and around the decker washer and filtrate tank. Knotters and screens are located between these washers, but due to lack of time, the complexity of the systems (four stages of screening with cascading flows) and lack of sampling points, only a few samples were collected around this equipment. This decision appears to be justified; the limited testing and material balances conducted confirmed that this area was only a minor source of emissions, as w i l l be discussed in Section 6.4. 5 .3 .2 M i l l T e s t i n g M e t h o d For determination of volatile organic sulphur compound concentrations in kraft black liquor, a full evaporation headspace technique, as described by Chai et al. (2000), was used. Methods for directed injection of liquid samples into the G C are presented by Berube et al. (1999) and N C A S I (2002), but these were only used for waste-water samples, not black liquor. A s mentioned earlier, black liquor is corrosive and has a complex sample matrix containing many inorganic salts that discourage G C analysis by direct injection of the liquid sample. A method presented by O'Conner and Genest (1997) of Paprican avoids these issues by sparging the black liquor sample with a measured amount of nitrogen to strip out the volatiles, with the collected nitrogen tested for volatiles concentration, and the liquor concentration calculated from this information. The main drawback to this method is the testing time, with up to an hour of sparging required per sample. A headspace technique is relatively quick, avoids the interference from the sample matrix, and avoids G C column damage by the corrosive materials. The full evaporation headspace technique uses a very small sample size injected into a sample vial to achieve near complete transfer of the volatile compound from a condensed matrix into the vapour phase in a very short period of time. A sample of the vial headspace is then injected into the G C . Gas samples were collected and tested using Environment Canada reference method EPS l / R M / 6 (Environment Canada, 1992). This method describes sample collection procedures for measurement of releases of TRS compounds from pulp and paper operations. 86 5.3.3 Sample Collection Procedure Chapter 5: Materials and Methods A sample collection vacuum pump (Supelco 1060) was used to draw the samples via lA" Teflon tubing through a phosphoric acid impinger (to remove water vapour), into a Tedlar bag. Figure 5.4 shows a photograph of this sample collection equipment. Three-way valves (Whitey, lA'\ SS) were used to draw a sample through the gas pump and vent it to atmosphere until a representative sample was being was being drawn through. A n extra-long-stem thermometer (Control Company, Model 94460-40), inserted into the gas stream in the vent stack, was used to measure wet and dry bulb temperatures for determination of relative humidity. A pitot-tube velometer (Alnor. Series 6000) was used to determine the vent line velocity, from which the volumetric flow was calculated. Figure 5.4: Gas sampling equipment, including phosphoric acid impinger. Teflon tubing, 3-way valves, vacuum pump and Tedlar bag The Tedlar bags were fitted with special dual valve fittings, which allowed collection of the sample through the valve, and syringe withdrawal of a sample through the septum to be injected in the gas chromatograph. These bags were used to collect the gas samples as well as to prepare the 87 Chapter 5: Materials and Methods gas standards required for testing. Each Tedlar bag was designated for one sample location, or for one gas standard, with the bags flushed with nitrogen between uses. Liquor samples were collected into narrow necked sample jars. These jars were filled completely and were immediately capped with minimal vapour headspace to minimize loss of the volatile sulphur compounds. 5.3.4 Sample Collection Considerations There was no opportunity for installation of new sample points, thus existing locations were used. For vapour sampling, the typical sample point consisted of a 3A" pipe fitted with a ball valve, originally installed for a pressure gauge, but with the gauge temporarily removed for our sampling purposes. A typical sample point, in this case the blow tank, being tested using the velometer, is shown in figure 5.5. Figure 5.5: Typical sample point (blow tank vent) being tested using the velometer 8H Chapter 5: Materials and Methods The locations of a most sample points was less than ideal, some being located right on the tank outlet (blow tank), upstream of an elbow (diffusion washer filtrate tank), upstream of a valve and strainer (blow tank), or right on a "T" section (decker washer hood). With any of these non-idealities, it is difficult to get an accurate averaged flow reading using the velometer, which wi l l compromise the accuracy of the calculated flow. As well, the operation of some equipment, such as the diffusion washer, is cyclical; thus, the liquor flow to the filtrate tanks is cyclical, causing cycling in vent flow, temperature, and concentration. The diffusion washer sample point was better placed, but the vent flow reading was almost zero. This was, at the time, thought to be due to a plugged fibre strainer immediately downstream of the sample point, but inspection during a subsequent shutdown revealed an in-line valve with a reversed handle (thus closed). It appears that the vent gases from the diffusion washer were being drawn down the stock chute to the blow tank. Unfortunately a shutdown was required to fit new sample points or clean in-line equipment, and with the next shutdown scheduled for many months later, the best was made of what was available. 5 . 3 . 5 M i l l T e s t i n g P r o c e d u r e For testing of the gas samples, a 100 p L sample was withdrawn from the Tedlar bag using a gas-tight syringe (Hamilton, 500 pL) fitted with a Chaney adaptor. This sample was immediately injected into the G C , which was set at a 50:1 injection split ratio. To achieve good separation of the TRS compounds, as illustrated in Figure 5.1, the G C oven was held at 40°C for 2 minutes, then ramped at 30°C per minute to 220°C, and then held for another 2 minutes, for a total run time of 10 minutes. Gas standards were prepared from a size 32 (7" x 21") certified gas standard cylinder containing nominal 50 ppm (mol) of each of the four TRS compounds with balance nitrogen ( B O C Gases, Langley, B.C. ) . Three additional standards of 15, 5 and 1 ppm (mol) were prepared with dilutions of the gas standard with nitrogen into Tedlar bags. These dilutions were made by metering a set amount of nitrogen into a Tedlar bag using a mass-flow meter calibrated for nitrogen ( M K S , model 2179), then using a gas-tight syringe (Hamilton, 50 mL) to inject the required amount of gas 89 Chapter 5: Materials and Methods standard. Due to degradation problems discussed below, a fresh set of these standards was prepared each morning and again in the afternoon. On the day preceding testing, sample vials were purged with 250 mL/min of nitrogen for a minimum of three minutes and then placed in an oven at 80°C. On the day of testing, a vial was taken from the oven, equalized to atmospheric pressure with a syringe needle and 10 p L of a liquor sample was injected using a micro-syringe (Hamilton, 10 pL) fitted with a Chaney adaptor. The sample vial was placed in a shaker in an oven at 80°C for five minutes. The vial was then taken from the oven and a 1.0 m L headspace sample was withdrawn from the vial and injected into the G C at a 20:1 injection split ratio. The same oven ramp schedule and helium carrier gas flow as used for gas testing were also used for the liquor headspace technique. A set of liquid standards was prepared using 99+% pure liquid D M S and 99+% pure liquid D M D S (Sigma-Aldrich). A sample vial was filled with nanopure water, capped and weighed. A small amount of D M S was injected into the sealed vial, and the vial weighed. A small amount of D M D S was then injected and the vial again weighed, with this being the base solution standard. The procedure was repeated, instead using the base solution in place of the pure D M S and D M D S to produce the next standard. This was repeated twice more with the previous solution used to prepare the next. A total of four standards with D M S and D M S concentrations ranging from about 30 to 0.2 ppm (mol) were prepared. Linear calibration curves from gas and liquid standards were prepared by plotting the square root of the P F P D signal against concentration. A typical calibration curve is shown in Figure 5.6. 90 Chapter 5: Materials and Methods 0 10 20 30 40 50 Sq.Root P P F D Count R-square = 0.991 # pts = 17 y = 0.051 + 0.717x F i g u r e 5.6: Typical liquid standard D M S calibration curve at the testing temperature of 80°C 5.3.6 M i l l T e s t i n g C o n s i d e r a t i o n s During mil l testing, significant degradation of liquid standards was noticed after a few hours (Figure 5.7). It is not clear why the sulphide concentration in the liquid standard dropped to such an extent but one possibility is that sulphides may have absorbed onto the glass vial. 91 Chapter 5: Materials and Methods 100 80 Time (hours) Figure 5.7: Degradation of T R S liquid standard held at 20°C and at an initial concentration of 30 ppm (mol) Rather than deactivating glassware or attempting to preserve the liquor samples against the degradation effects, samples were tested as quickly as possible after collection, typically within an hour. For liquid standards, new standards were prepared each day and used within an hour. Most of the complications discussed, including adsorption to glass, are time dependent and can be minimized by minimizing the time between collecting and testing the samples. The Environment Canada method requires that all testing of all gas samples must be done within an hour of the sample being collected. To test this requirement, a sample of the gas standard containing nominal 50 ppm (mol) of each of the TRS compounds was injected into a Tedlar bag. This bag was placed in an oven at 80°C, to simulate mi l l operating conditions, and tested at 5,35 and 65 minutes, with the results shown in Figure 5.8. 92 Chapter 5: Materials and Methods 50 Time (minutes) Figure 5.8: Degradation of TRS gas standard held at 80°C and at an initial concentration of 5 ppm (mol) Figure 5.8 illustrates that there is significant potential degradation, even within the prescribed one hour time limit. There should not have been any oxygen in this sample, so the degradation effects should not be due to oxidation, but rather one or a combination of the other effects discussed by Wardencki (1998) and summarized in Section 3.4. For mill process gas samples, the extent of the oxidation reactions may be significant as the vent samples can contain a high proportion of air. This can be exacerbated by ultraviolet light catalysed oxidation of the TRS compounds in the gas phase. These sample instabilities, all of them time dependent, drove the decision to re-locate the testing equipment to the mi l l site. To meet the Environment Canada one-hour time requirement, and to minimize the degradation effects, gas samples were tested as quickly as practically possible. Once samples were collected, they were taken directly to the lab and injected into the pre-prepared gas chromatograph, typically within the hour. The gas standards were tested immediately after preparation, again within the hour. 93 Chapter 5: Materials and Methods The following should be noted when viewing the analytical test results tabulated in Section 6.4: 1. The gas chromatograph detection limit was approximately 0.1 ppm (mol); values below this are reported as non-detectible (nd). 2. Vapour sample concentrations were corrected to include water vapour, based on the measured relative humidity level, to account for water vapour that was removed (in the phosphoric acid impinger) for testing, i.e., reported values are in ppm (mol) on a wet basis. 3. Liquor sample concentrations are given on a stock-free basis. 4. Liquor sample concentrations for liquid squeezed from stock may be low due to potential dilution in the sample collection pot and sampling handling technique. Liquor was squeezed from stock samples in the field and some of the T R S compounds, being highly volatile, may have been lost from the liquid. These potential sampling problems were only discovered when attempting preliminary mass balances, so those samples collected prior to November 30, 2005 may be compromised. In a particular, L S I was very difficult to sample due to the high operating pressure of the blow-line. There was a dedicated sample line and sample pot with flush connections to draw a sample from this line. To prevent plugging, the sample line had to be flushed before and after sampling and this would cause dilution of the sample unless extreme care was taken. It was only on the last day of testing (December 1) that what appeared to be an undiluted sample was collected. 5. The 1s t stage diffusion washer filtrate tank vents through 2 n d stage diffusion washer filtrate tank via the overflow line. Only one vapour sample point was available, located on the combined vent from the 2 n d stage tank. 6. The diffusion filtrate tanks vent flow cycles (reversing the vapour flow) due to the cycling of the diffusion washer operation and liquor flow to these tanks. A vapour flow out of the tank was recorded for about 60 seconds and peak at 6.4 m 3 /min, after which the flow reversed for about 10 seconds and peaked at about 8.6 m 3 /min. The recorded flows are averaged values. 7. The diffusion washer vent registered little or no flow. At the time this was thought to be due to a plugged fibre strainer, but later confirmed to be a closed valve (due to a reversed handle). The vent gases from this washer were likely venting through the stock chute to the blow tank. 94 Chapter 5: Materials and Methods 8. The screen dilution tank flow was estimated because the velometer could not be used on the sample point due to interference with a handrail. The vent flow was assumed to be equal to the volume displacement caused by the greatest rate of level change in the tank. 9. Some early variability in results was attributed to standard type 2 syringe needles damaging the Tedlar bag septum and the G C injector septum, with the resulting rubber material plugging the needle. This was rectified by replacing the needles with type 5 rounded tip type with the orifice in the side of the tip of the needle. 5.4 A c t i v i t y C o e f f i c i e n t M o d e l l i n g The activity coefficient, y{, for a volatile substance i can be modelled using the N R T L equation, which consists of the first two terms of Equation (3.11), presented in Section 3.1.2. For a binary system at infinite dilution, i.e., when the solute, i , concentration, x i 5 approaches zero, and the solvent, j , concentration, x j 5 approaches unity, the N R T L equation reduces to: ln(r,)= ^, + G ^ ( 5 . 2 ) This equation can be used for a binary non-electrolyte system such as for i = D M S and j = water. A s described in Section 3.1.2, the N R T L equation can be extended to the e N R T L equation to describe the effect due to the presence of electrolytes (such as dissolved salts), by the addition of the last two terms in Equation (3.11). Black liquor is a complex aqueous mixture including many dissolved salts (Table 2.10). N o e N R T L binary parameters could be found in the literature for the TRS compounds and these salts. To simplify the system, the black liquor dissolved salts were treated as a single substance, and the system was modelled as ternary system consisting of D M S -water-black liquor sodium salts. The dissolved black liquor salts were assumed to consist of a single cation (sodium) associated with a single anion, i.e., x c a = x c = x a , where x is the mole fraction, c refers to the cation, a to the anion, and ca to the salt. This terminology is used for modelling of electrolyte systems using the e N R T L equation (Chen and Song, 2004). A s described in Section 2.7.1, the major anions in black liquor are hydroxide, hydrosulphide, carbonate, sulphite, sulphate, thiosulphate and organic salts. These latter substances include the 95 Chapter 5: Materials and Methods sodium salts of phenolates (dissolved alkali lignin), polysaccharinic acids, resin acids and fatty acids. At low salt concentrations, negligible error is introduced by assuming all salts consist of a single cation associated with a single anion, even when considering the disodium salts. This is because the e N R T L equation is based on the effective mole fraction, X , rather than the actual mole fraction, x. For example, for N a 2 C 0 3 , which dissociates into two cations and a single anion, i.e., x c = 2x a , the absolute value of the ionic charge of the anion wi l l be double, i.e., C c = 1 and C a = 2. In this case, the effective mole fraction of the anion wi l l be double the actual mole fraction, i.e., X a = C a x a = 2x a = 2(x c/2) = x c , and the effective mole fraction of the cation w i l l equal the actual mole fraction, i.e., X c =C c x c = x c . Also note that for molecular species, such as water, that do not dissociate, C, = 1, thus Xj = x,. With these assumptions, the system can be modelled as a three component system with i = D M S , j = water, and ca = black liquor sodium salts. For a ternary system, r c a i = rci = r a i, r i c a = r i c a c = r i a c a , r c a j = rC J= r a j, and r i c a = r J c a c = z"a c a. There are nine adjustable parameters: the solute-solvent pairs, rtJ, aV]; the solute-salt pairs, r i c a , r c a j , <rica; and the solvent-salt pairs, T- , r c a j , <rjca. For this system, with a solute, i , at infinite dilution in a solvent, j , with a single salt, ca, added, the e N R T L Equation (3.11) reduces to: , / \ x J - G J , T J I + 2 - x c a - G c a , - r c a , x , - G M l r V i ) = — T T X J • G j i + 2 • xca • x J + 2 - x c a G c a j v - ^ - - ^ - 0 ^ 2 ' X ca ' Gcaj ' Tcaj r , J ~ X ; + 2• X „ • G 2 ' X c a - G , c a — 1 *ica X j ^jca "^jca T- ~ ica f~\ -j — jca • "ca V X j ' ^ j c a + X c a J x . G - + x„ (5.3) The G parameters are determined from the r a n d a parameters using Equation (3.12). The temperature dependencies of the r parameters are modelled using the adjustable parameters a and b and Equation (3.13). 96 Chapter 5: Materials and Methods 5.4.1 R e g r e s s i o n A n a l y s i s o f P h a s e E q u i l i b r i a T e s t i n g D a t a Excel was used to compile and analyse the phase equilibria testing data. The dimensionless Henry's constants were experimentally determined for all solutions at 20, 50, 70 and 90°C; from the Henry's constants the activity coefficients were calculated. The rand a parameters that provided the best fit for the e N R T L equation for these experimental activity coefficients were determined using regression; a statistical technique used to explain or predict the behavior of a dependent variable. Regression was used to find the relationship between variables using the regression equation, in this case the e N R T L equation. This mathematical model was fitted to a set of experimental data (activity coefficients) using the least squares method. The experimental data values for the activity coefficients at the various temperatures were entered into the first column of the Excel spreadsheet. The calculated value for the activity coefficient based on the e N R T L Equation (3.11) were entered in the second column. The activity coefficient calculations were based on rand a parameters values entered into separate cells in the spreadsheet, with the "first guess"values arbitrarily set at one. The square of the difference between the experimental value and the calculated value was entered into the third column; the sum of squares was totalled below this column. The best fit for the e N R T L equation was achieved when the sum of squares was minimized (least squares criteria). Excel includes a "solver" tool which can be used to manipulate specified variables to maximize or minimize a calculated value. In this case, the sum of squares cell was specified as the "Set Target C e l l " to be minimized. The rand a parameter cells were specified as the " B y Changing Cells," i.e., the manipulated variables. The solver tool determines the values for r a n d a that minimize the sum of squares, i.e., it would determine the values for the adjustable parameters that provided a best fit for the e N R T L equation to the experimental data. Aspen Plus was also used to regress the experimental data to find the parameters that achieved the best fit for the e N R T L equation. When using Aspen in this way, the data regression option is chosen when setting up the new file. Experimental data is entered as " T P X Y " , i.e., temperature, pressure and mole fraction compositions of the liquid and vapour stream are entered in the data file. The E L E C N R T L (another way to refer to e N R T L ) properties method is chosen, and the e N R T L parameters to be fitted are chosen, while others not used are set to zero. The Aspen Plus 97 Chapter 5: Materials and Methods data regression file is then run to determine the best fit e N R T L rand a parameters. The experimental data were regressed using both the Excel solver tool and Aspen Plus, partly to check each procedure to ensure no mistakes were made in the process. This proved to be a good decision as different results were initially produced, as wi l l be discussed in Section 6.3.5. 5 .5 V L E M o d e l l i n g The mil l testing data were modelled using a V L E module based on the e N R T L activity coefficient correlation. For an adiabatic V L E module, i.e., where there is zero heat duty, emissions from process equipment can be predicted based on feed stream properties and the outlet pressure or temperature. The general form of the solution of the V L E module, also called a "flash" module, that can be incorporated into a mi l l heat and mass balance is shown in Figure 5.9. Vapour (V,y,T,P) Feed (F,z,T F ) J V L E Module calculation block Liquid (L,x,T) F i g u r e 5 . 9 : V L E module used to predict emissions of volatile compounds For an adiabatic V L E module, the properties of the inlet feed stream must be known, including flow, F, composition, z , and temperature, T F , along with the outlet pressure, P (or temperature, T). From this information, i f vapour liquid equilibrium is established, the outlet temperature (or pressure), and the outlet vapour flow, V , and composition, y, and the outlet liquid 98 Chapter 5: Materials and Methods flow, L , and composition, x , can be calculated. The main component of the V L E module is the phase equilibrium Equation (3.7) presented in Section 3.1.1. 5.5.1 Phase E q u i l i b r i u m Equat ion The V L E of each condensible component is determined using Equation (3.7): y i P = r i x i P i s " , , i = l , n (3.7) The V L E of each noncondensible component can also be determined using the above equation, but since these components by definition wi l l be in very low concentrations, the simpler Henry's law Equation (3.8), described in Section 3.1.1, is often used instead: y , P = - ^ - , i = l , n (3.8) K H i Henry's law constants, k H , for many noncondensibles, such as oxygen and nitrogen, have been collected together from many literature sources by Sander (1999). To allow conversion between Henry's constant and an activity coefficient, Equations (3.7) and (3.8) are combined to get: k H i = T - ^ a T (5-4) The vapour pressure, P; s a t at the corresponding temperature, T, can be determined using the extended Antoine Equation (3.21) given in Section 3.1.5, with the parameters for the T R S compounds given in Table 3.9. 99 5 .5 .2 M o l e B a l a n c e E q u a t i o n Chapter 5: Materials and Methods Assuming there are no reactions and that steady state has been established, the total moles entering the system must equal the total moles leaving the system: Fz , = V y , + L x i , i = l , n (5.5) I x, = 1 (5.6) i=l,n 1 y, = i (5.7) i = l,n X z, = 1 (5.8) i = l,n F, V and L are the molar flows, and z,y, and x are the mole fraction composition of component i for the feed, and outlet vapour and liquid streams, respectively, and n is the total number of components. 5 . 5 . 3 E n e r g y B a l a n c e E q u a t i o n In steady state, the total energy entering the system must equal that exiting the system: F ! Z , H : = v X y i H r + LXx.H, 1- (5.9) i = l,n i = l,n i=l,n To solve this equation, the enthalpy, H , of each component must be known. For a case involving black liquor, which consists of many substances which are often not defined, the enthalpy of the feed liquor stream and the outlet liquor stream can determined using an empirical specific heat correlation given by Grace et al. (1989): Cp,B,ac kL,c.uor = 1.0 - (l - C P f S ) • S (5.10) S is the percent total dissolved solids in the black liquor, and C P s is the specific heat of the dissolved solids, which can range from 0.3 to 0.5 (Grace et al. 1989). 100 5 .5 .4 T e m p e r a t u r e ( B o i l i n g P o i n t R i s e ) E q u a t i o n Chapter 5: Materials and Methods The boiling point rise (BPR) is defined as the difference between the boiling temperature of a liquid containing dissolved solids and that of the pure liquid with no dissolved solids. For a flashing liquid, the temperature of the vapour wi l l match that of the liquid, with the vapour superheated by the amount of the B P R . The B P R for black liquor can be estimated using the following equation (Grace et al., 1989): The fraction S/(l-S) is the solids to water ratio in black liquor. The B P R is given in units of degrees Celsius. The average constant for this equation is 7, but it can range from 5 to 9, depending on the composition of the dissolved solids in the black liquor (Grace et al. 1989). The superheated vapour leaving the liquid vapour interface is at temperature T, with the saturated vapour temperature, T s a t , determined from the temperature and the B P R : 5 . 5 . 5 S o l u t i o n o f V L E M o d e l The solution of the adiabatic V L E flash model involves the simultaneous solution of multiple non-linear equations. This includes activity coefficient, vapour pressure, phase equilibrium and mole balances for each component, along with overall energy balances. The mathematical solution may also require correlations to describe other properties such dissociation effects and liquid boiling point rise. These equations are summarized in Figure 5.10. (5.11) sat T - B P R (5.12) 101 Chapter 5: Materials and Methods Feed stream (F,z,TF,P) Electrolyte effects using eNRTL equation = fi(x,T) I | Activity coefficients using NRTL equation = «[x,T) I Pure component vapour pressure using extended Antoine equation = f(T) I Dissociation effects = f(x,T) I W-M Boiling point rise effects using Grace equation = f(x,T) olution of multiple^ non-linear equations (Newton-Raphson or equivalent) . I I Energy balance = fi(F,V,L,x,y,z,T, T F,H) I Mole balances Phase equilibria balances =f(7,x,y,P,Psat) = f(F,V,L,x,y,z) / Outlet / stream / (V,L,x,y,T, / T s a t ,P) Figure 5.10: Summary of V L E module calculation block non-linear equations that require simultaneous solution using a Newton-Raphson or equivalent technique In the case where the liquid feed stream is below its bubble point at the vessel operating pressure, the vapour flow wi l l be zero, unless a noncondensible component such as nitrogen or oxygen is present. These can enter either entrained with the feed liquid or as a separate feed stream, such as air entering a tank through an opening. This air stream may be introduced deliberately as sweep air or it may be undesired, such as tramp air entering through a poorly sealed overflow opening. Tanks at Howe Sound, including the filtrate tanks and blow tank, include vacuum breakers 102 Chapter 5: Materials and Methods and are designed to be sealed so that any air that does enter the system is considered tramp air. The tramp air can be incorporated into a V L E balance as shown in Figure 5.11. F i g u r e 5 .11 V L E Module vapour and liquid <by-passes A tramp air by-pass around the V L E module can be used to account for insufficient time for the tramp air to reach equilibrium with the liquid phase. A liquid by-pass can be used to account for incomplete mixing in the liquid phase. For example, liquor entering a tank may short-circuit to the suction of the liquor outlet causing stratification of the concentration of the contaminants in the liquid phase in the tank. If these by-passes are required, the tramp air and liquor by-passes wi l l have to be determined for each individual piece of equipment. Part of the objective of this work was to investigate whether the process equipment tested during the mi l l testing program was operating at equilibrium, or i f some type of by-pass was required to fit the data. 103 Chapter 5: Materials and Methods 5.5.6 Commerc ia l Software V L E Modules A V L E module incorporating the Newton-Raphson method was programmed in Matlab (Mathworks) to solve a ternary system involving the simultaneous solution of ten non-linear equations. The components for this system include water, methanol and nitrogen, with nitrogen classified as a noncondensible. This Matlab program is included in Appendix C. This program is limited to three components, so is of limited practical use, but it was useful as a training tool with the programming of the equations leading to a better understanding all of the factors governing the mathematical solution. Simulation software such as Aspen Plus and C A D S i m Plus include pre-programmed V L E modules designed for multiple component systems. Aspen Plus includes a V L E module referred to as "Flash2" with their simulation software. For this module, Aspen provides a number of correlations, including the e N R T L equation, their recommended method to model electrolyte systems. If experimental data are unavailable, Aspen Plus also includes the U N I F A C group contribution method to estimate activity coefficients. C A D S i m Plus was also used because it has a strong focus on the pulp industry. C A D S i m incorporates many empirically fitted correlations unique to the kraft pulping industry, including a properties database for fibre, alkali lignin, organic and inorganic dissolved solids and black liquor. For example, it includes Equations (5.10) and (5.11), the specific heat and B P R correlations for kraft black liquor. C A D S i m also includes many modules specific to kraft pulp mills such as diffusion washers, blow tanks, drum-type decker washers, etc., which make programming heat and material balances much simpler. C A D S i m Plus is used at many mills, particularly in Canada, including Howe Sound Pulp and Paper, where mil l testing for this research was conducted. The offices of Aurel Systems, the developers of C A D S i m , are located near the University of British Columbia; therefore, it was very convenient for support services. They were also wil l ing to extend and modify C A D S i m to meet the requirements of this work, including the incorporation of a V L E module. Commercial simulation software can be costly; Matlab, C A D S i m Plus and Aspen Plus were made available to University of British Columbia students for research purposes. 104 Chapter 6: Results and Discussion C h a p t e r 6 R e s u l t s a n d D i s c u s s i o n The premise for this work is that emissions of T R S compounds from kraft pulp mi l l processes can be predicted based on modelling of their vapour-liquid equilibria behaviour. Before any phase equilibria or mi l l testing was conducted, preliminary phase equilibria calculations were performed using published data to determine i f volatile components in kraft pulp mi l l processes were at or near vapour-liquid equilibrium. 6.1 P h a s e E q u i l i b r i a T e s t i n g u s i n g P u b l i s h e d D a t a Since no relevant published mil l testing data could be found for the T R S compounds, preliminary analysis was done using methanol. Methanol, like methyl mercaptan and dimethyl sulphide, is formed in the digester in an undesired side reaction. It is relatively volatile compared to water, but less so than the T R S compounds (Figure 3.1), making it a good candidate for testing the relative merits of this method. Due to the recent implementation of the Cluster Rule in the U.S. , which specifies methanol as a surrogate for emissions of hazardous air pollutants, fairly extensive methanol emissions testing has been completed by the National Council for A i r and Stream Improvement (NCASI) . Data was extracted from N C A S I reports ( N C A S I , 1994a; 1994b; 1996) and from testing conducted by N C A S I for Gu etal. (1998a). If the liquid and vapour are in equilibrium, with methanol defined as i , then its vapour mole fraction, y i 5 can be predicted from the liquid mole fraction, x i 5 using Equation (3.7). y i P = / i x i P i , B (3.7) The total pressure of the system, P, is atmospheric in all cases for the published data. The vapour pressure, P s a t , of methanol at the process temperature, can be determined using the extended Antoine Equation (3.21) in Section 3.1.5, using parameters in Table 3.9. The activity coefficient, y{, for methanol in water can be determined from the N R T L equation, the first two terms of Equation (3.11) in Section 3.1.2 using parameters in Table 6.1. These parameters were taken from Chen and Song 105 Chapter 6: Results and Discussion (2004), who extracted these figures from the databank of Aspen Plus version 12.1 (Aspentech). Aspen Plus is based on D E T H E R M , an extensive thermophysical properties databank compiled from numerous published sources ( D E T H E R M ) . Table 6.1: Methanol(i) in waterQ) parameters for N R T L equation (Chen and Song, 2004) N R T L Parameter Value a u -2.626 4.824 h 828.387 h -1329.544 0.3 Because the published methanol testing data did not include details on the composition of the liquid streams, except for the methanol concentration, a binary methanol-water system was assumed and any potential effects from other solvents or dissolved solids were ignored. The published data also did not include details on liquid flows to and from equipment, so mass and energy balances could not be completed, and methanol emissions could not be predicted on a mass basis. The calculations that follow are not done to predict mass emissions, but solely to test i f methanol, as a surrogate volatile compound, is at or near equilibrium in any kraft mi l l equipment. Data from over 225 vapour and 250 liquid samples were collected from the N C A S I reports into an Excel spreadsheet for analysis (Appendix D). O f these data, only a subset was used, due to a number of limitations. Data were considered i f they could be matched together in pairs; e.g., for a tank, corresponding samples of the vent stack vapour stream and the liquid outlet stream were required, as these would be the streams potentially at equilibrium. O f these pairs, only those vapour and liquid samples collected at exactly the same time were used, to exclude variability in the process. N C A S I noted that their liquid measurements might be questionable, as a direct injection G C method was used to determine the methanol concentrations. This method is susceptible to matrix interference, especially in the black liquor samples. For vacuum drum washer filtrate tanks, Figure 6.1 shows a comparison of measured 106 Chapter 6: Results and Discussion methanol vent stack vapour outlet concentration compared to the predicted equilibrium concentration based on the outlet filtrate (black liquor) methanol concentration using Equation (3.7). 2000 > o. 1500 Q. C > O c CO sz 0> 1000 500 Measured LU Predicted C O cn T t T t oo oo 0 0 > > > > > > > _l —J < < o o Equipment Identification Code Figure 6.1: Vacuum drum washer filtrate tank vent stack measured methanol concentration compared to predicted concentration based on outlet filtrate concentration The letter in the identification code refer to the mil l and the numbers refer to the specific equipment where the measurements were made. N C A S I does not state where these mills were located or provide any specific details for the equipment. A l l that is known is that for the vacuum drum washers measurements, shown in Figure 6.1, the 10 pairs of measurements were taken from seven different washers at four different mills. In the majority of cases, the vapour-liquid equilibrium method predicts fairly well , with 6 of 10 pairs predicting within 25%. The measured values range from 225 to 1493 ppm (mol) (average 710) with an average root mean squared error ( R M S E ) of 361 ppm (mol) for the predicted values. Better agreement might be expected with a more accurate methanol testing technique and incorporation of the dissolved solids effects. 107 Chapter 6: Results and Discussion The vacuum drum washer hood and diffusion washer comparisons are shown in Figure 6.2 and 6.3, respectively. For these, the predicted equilibrium was based on the inlet wash liquor concentration. This might not seem intuitive as the first thought might be that outlet streams should be in equilibrium, but in displacement washing, the inlet wash liquor is the liquid in most intimate contact with the outlet vent gas. Gu et at. (2001) used this basis to design their emissions module for W i n G E M S , except for the vacuum drum washer they also included a term for the inlet vat area of the washer based on the inlet stock concentration. Since the published data did not include this latter information, the predicted value shown in Figure 6.2 are based only on the shower water concentration. 1000 mv) 800 o_ Q. C 600 > 75 400 c CO 0) 200 0 Measured D Predicted! ^ ^ b * * > > > > <o <o <o (O (/)(/)</) CQ DO DQ < < < < CD CO. CO. X X X CO CO r m io w m in fflllloooooJJJ * * * * * X X Equipment Identification Code Figure 6.2: Decker washer hood vent stack measured methanol concentration compared to predicted concentration based on inlet wash liquor concentration 108 Chapter 6: Results and Discussion 1500 > E CL — 1000 .*-» c CD > c I 500 0) > > > > > > _ I _ I J < < O O (D O Equipment Identification Code Figure 6.3: Diffusion washer vent stack measured methanol concentration compared to predicted concentration based on inlet wash liquor concentration For the decker washer hoods, 10 of 17 pairs predict values within 50% of the measured values. The measured values range from 64 to 916 ppm (mol) (average 256) with average R M S E of 260 ppm (mol) for the predicted values. For the diffusion washers, 7 of 11 pairs predict within 50%. The measured values range from 57 to 1350 ppm (mol) (average 576) with an average R M S E of 308 ppm (mol) for the predicted values. Results were more scattered for weak black liquor tanks, as can be seen in Figure 6.4. The measured values ranged from 1 to 2600 ppm (mol) (average 556) with an average R M S E of 913 ppm (mol) for the predicted values. The poor prediction may be a result of the more variable operation of these liquor tanks compared to brown stock washing equipment. For example, i f more liquor is being withdrawn from a liquor tank than is being pumped in, there may be a reverse flow in the vent pipe, drawing in atmospheric air, which would result in a low or non-detectable methanol concentration reading in the vent gas. 109 Chapter 6: Results and Discussion 3000 Measured • Predicted JLrJ • i l l § § § ^ s | | | < < < - - o o « i i i i n t i 2 0 0 0 0 0 0 & > & « « « « < \ - i - \ - \ - \ -2 2 2 O O O O O Equipment Identification Code Figure 6 .4: Weak black liquor tanks vent stack measured methanol concentration compared to predicted concentration based on liquor outlet concentration The poor prediction for weak black liquor tanks (Figure 6.4) can be contrasted with the results for vacuum drum washer filtrate tanks (Figure 6.1). The level is typically held constant in filtrate tanks, giving a more consistent positive vapour flow, which should result in a more accurate prediction. As described in Section 6.7, testing at the Howe Sound mil l indicated that liquor storage tanks were only minor contributors to overall TRS emissions (less than 3%); therefore, the decision was made to focus testing and modelling work on the brown stock washing area. Taking into account variability in the pulping process, sampling and testing limitations, and lack of reported details in the published data (such as dissolved solids concentration in the liquid), the results using published data for methanol (Figures 6.1 to 6.3) showed enough promise to proceed with a study plan for the T R S compounds. 6.2 Phase E q u i l i b r i a Testing Results The complete results of phase equilibria testing of D M S in all solutions are shown in the form of an activity coefficient in Table 6 .2 . The raw data for phase equilibria testing are provided 110 Chapter 6: Results and Discussion in Appendix E. Each activity coefficient given in this table was averaged from a minimum of three data points, with each data point calculated from the G C output averaged from three separate injections. The standard deviations given in this table, and the 95% confidence bars used in the following figures, were calculated from these three or more data points. T a b l e 6 . 2 : Vapour-liquid equilibrium activity coefficients for D M S solutions tested Solution Activity coefficient (± standard deviation) 20°C 50°C 70°C 90°C 1 173 ± 6 145 ± 2 128 ± 5 124 ± 1 2 176 ± 4 154 ± 2 133 ± 6 122 ± 3 3 181 ± 7 159 ± 7 142 ± 4 126 ± 5 4 200 ± 14 154 ± 4 133 ± 3 129 ± 6 5 220 ± 19 160 ± 9 139 ± 10 132 ± 4 6 237 ± 2 1 186 ± 14 158 ± 6 151 ± 10 7 276 ± 24 225 ± 0 180 ± 7 171 ± 5 8 351 ± 6 292 ± 16 243 ± 3 202 ± 4 9 255 ± 7 208 ± 6 183 ± 4 170 ± 14 10 261 ± 4 211 ± 2 189 ± 0 174 ± 3 11 269 ± 18 203 ± 2 168 ± 4 164 ± 13 12 314 ± 2 4 224 ± 1 1 187 ± 6 179 ± 14 13 251 ± 6 223 ± 5 214 ± 18 173 ± 10 To illustrate the effect of the addition of increasing amounts of alkali lignin (solutions 4 and 5), the shift in the activity coefficient of a binary DMS-water system is shown as a function of temperature in Figure 6.5. To provide a baseline for comparison purposes, binary DMS-water data (solution 1) are shown in Figures 6.5, 6.6 and 6.7. A l l data points on the three figures are shown with 95% confidence bars. I l l Chapter 6: Results and Discussion 400 - 3 5 0 Q) 1300 **-<j250 & >200 *-> u <150 100 290 -6 wt% Alkali Lignin in DW -3 wt% Alkali Lignin in DW - Distilled Water (DW) 310 330 350 Temperature (K) 370 Figure 6.5: Temperature effect on DMS-water activity coefficient at various alkali lignin concentrations (solutions 4 and 5) The increase in the activity coefficient of a binary DMS-water system caused by the addition of a mixture of sodium salts (solutions 6, 7 and 8), is illustrated in Figure 6.6. 400 100 290 «-6 wt% Sodium Salts in DW »- 4 wt% Sodium Salts in DW Sodium Salts in DW Water (DW) 310 T 330 ,350 Temperature (K) 370 Figure 6.6: Temperature effect on DMS-water activity coefficient at various sodium salts concentrations (solutions 6, 7, and 8) 112 Chapter 6: Results and Discussion The increase in the activity coefficient of a binary DMS-water system caused by the dissolved solids in black liquor is illustrated in Figure 6.7. 400 100 -I , 1 1 1 290 310 330 350 370 Temperature (K) Figure 6.7: Temperature effect on DMS-water activity coefficient at various total dissolved solids concentrations in black liquor (solutions 11 and 12) From Figures 6.5 and 6.6 it can be seen that the addition of sodium salts to a DMS-water solution shifts the equilibrium to a much greater degree than the addition of an equal amount of alkali lignin. This suggests that the shift in activity coefficient in black liquor seen in Figure 6.7 would be mainly due to the dissolved inorganic salts. 6.2.1 Other Properties of Tested Solutions The ash content of the Sigma-Aldrich alkali lignin powder (used to prepare solutions 4 and 5) was measured at 20.9 wt%, while the Howe Sound alkali lignin powder (used to prepare solution 3) was measured at 6.4 wt%. These lignin ash samples were found to contain 21.4 and 15.7 wt% sodium, respectively. From this data, the total sodium in the Sigma-Aldrich alkali lignin sample was calculated to be 4.5 wt%, while the Howe Sound alkali lignin was calculated to be 1.0 wt%. 113 Chapter 6: Results and Discussion The measured ash content of the Sigma-Aldrich alkali lignin was much higher than the 3 to 7 wt% stated in the supplier's data sheet. Alka l i lignin is typically recovered by acid precipitation, with the resulting "paste" washed with weak acid before drying into powdered form (Oilman, 2006). This was the method used to recover alkali lignin from the Howe Sound black liquor sample, as described in Section 5.2.2. The total sodium left in the alkali lignin w i l l be a factor of how well this paste is washed, because the sodium is stripped out by the acid. One possible explanation for the high ash content of the purchased alkali lignin sample would be inefficient acid washing. . A n elementary composition analysis was done on the dissolved solids of a sample of Howe Sound weak black liquor. It was found to contain 20.7 wt% sodium, 0.8 wt% potassium and less than 0.1 wt % of other metals. This sodium content falls into the typical range of 18 to 22 wt% given for typical softwood black liquor (Table 2.11). The measured pH, total dissolved solids, ash and sodium content of all solutions is given in Table 6.3. Table 6.3: pH, tota dissolved solids, ash, and sodium content of solutions tested Solution pH Total dissolved Ash Sodium solids (wt%) (wt%) (wt%) 1 7.0 0 0 0 2 12.0 0.04 0.04 0.02 3 12.0 3.1 0.27 0.08 4 8.7 3.0 0.63 0.13 5 9.1 6.0 1.3 0.27 6 12.8 2.0 2.0 0.86 7 13.0 4.0 4.0 1.7 8 13.2 6.0 6.0 2.6 9 12.8 6.6 3.3 1.4 10 12.9 7.8 3.9 1.6 11 12.5 6.6 3.2 1.4 12 12.6 9.7 4.8 2.0 13 12.9 8.2 4.1 1.7 114 Chapter 6: Results and Discussion The addition of electrolytes, such as sodium salts, alters phase equilibrium behaviour, while no similar effect is theorized for non-ionic dissolved organic substances (Chen and Song, 2004). It is not clear what causes the equilibrium shift in the dissolved alkali lignin mixtures, but it was thought to be an ionic effect, either from residual inorganic salts in the alkali lignin, or due to the dissociation of sodium phenolates, the main component of alkali lignin (Grace et al., 1989). 6 .2 .2 C o n d u c t i v i t y as a F u n c t i o n o f S o d i u m C o n c e n t r a t i o n One measure of the ionic strength of a solution is its conductivity. The term conductance refers to the ability of materials to carry an electric current, with solutions referred to as electrolytic conductors. Under the influence of an electric field, the flow of current through an electrolytic conductor is accomplished by the movement of cations and anions. The alkali lignin mixtures were tested for conductivity, with the results shown in Table 6.4. For comparison purposes, dilutions of the sodium salts solutions and black liquor samples were prepared that gave conductivities in a similar range to the alkali lignin mixtures. T a b l e 6 . 4 : Conductivities of alkali lignin mixtures, sodium salt solutions, and black liquor Solution Dilution Sodium Conductivity (wt %) (mS) 1 0 0 0 3 0 0.08 3.8 4 0 0.13 6.7 5 0 0.27 11.9 8 60:1 0.04 2.2 8 20:1 0.13 6.0 8 10:1 0.26 11.3 13 - 20:1 0.08 3.7 13 10:1 0.17 6.9 13 5:1 0.34 14.1 115 Chapter 6: Results and Discussion In black liquor, sodium makes up a large fraction of the potential cation content (Table 2.11), with the balance mainly consisting of potassium. This was confirmed by the results of the ICP test conducted on the Howe Sound black liquor which indicated that sodium makes up 96 wt% of the total metals in the sample. The ionic strength associated with sodium includes the salts and the hydroxyl and hydrosulphide ions (Equation 3.15), the latter of which is in equilibrium with hydrogen sulphide (Equation 3.14). Not all ions in black liquor are associated with sodium, such as methyl mercaptan (Equation 3.17), although these are at negligible concentrations compared to the ions associated with sodium. This raised the question, can the conductivity of these solutions be related to their concentration of sodium? To illustrate this, the conductivity is plotted as a function of sodium concentration in Figure 6.8. Conductivity should increase linearly with increasing ionic strength, and this seems to be born out in the plot. Sodium (wt%) Figure 6.8: Conductivity of alkali lignin mixtures, sodium salts solutions, and black liquor as a function of sodium concentration The conductivities correlate reasonably well , within 10%, with the sodium content of these solutions. This leads to the question, can the activity coefficient of these solutions be related to their concentration of sodium? 116 Chapter 6: Results and Discussion 6 . 2 . 3 A c t i v i t y C o e f f i c i e n t as a F u n c t i o n o f S o d i u m C o n c e n t r a t i o n Activity coefficients are plotted as a function of sodium concentration for 90°C in Figure 6.9 (results for other temperatures were similar). For clarity, the data points for each solution are connected. Also for clarity, only the alkali lignin data is shown with 95% confidence bars. 220 | 2 0 0 o ? 180 o o >>160 > I 140 - Black liquor —•— Sodium salts solution —•—Lignin mixture 120 0.0% 0 .5% 1.0% 1.5% 2 .0% 2 . 5 % 3.0% Sodium (wt%) F i g u r e 6 . 9 : Activity coefficient at 90°C of alkali lignin mixtures, sodium salt solutions and black liquor, as a function of sodium concentration Figure 6.10 provides a close-up of the same data as Figure 6.9, but focussing on the bottom left corner of the graph with a sodium concentration up to 0.5 wt%. 117 Chapter 6: Results and Discussion 140 .1 135 u 120 —*r - Black liquor —•—Sodium salts solution —•—Lignin mixture 0.0% 0 .1% 0 . 1 % 0 .2% 0 .2% Sodium (wt%) 0 .3% 0 .3% Figure 6 .10: Activity coefficient at 90°C (up to 0.5 wt% sodium) of alkali lignin mixtures, sodium salt solutions and black liquor, as a function of sodium concentration The increases in the activity coefficient shown in Figures 6.9 and 6.10 correlate well with an increasing sodium content for all solutions. It appears that there may be a correlation between the activity coefficient and the sodium cation content of the solution, irrespective of which anion with which the sodium is associated. This suggested that the shift in vapour-liquid equilibrium in black liquor could be modelled based solely on the sodium concentration. To test this hypothesis, the data for the sodium salts were overlain with the data points for all of the black liquor samples (solutions 9 to 13) and plotted against their inorganic solids concentration in Figures 6.11 to 6.14, at 20, 50, 70 and 90°C, respectively. Even though there appears to be a linear response, there is no theoretical basis for this, so for clarity, the sodium salts solution data points are simply connected. 118 Chapter 6: Results and Discussion 400 350 £ 300 250 C O o o o > o < 200 150 100 •Sodium salts solution Black liquor 0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% Sodium (wt%) Figure 6.11: Sodium concentration effect on DMS-water activity coefficient for sodium salts and black liquor at 20°C 400 *- 350 100 •Sodium salts solution Black liquor 0.0% 0.5% 1.0% 1.5% 2.0% Sodium (wt%) 2.5% 3.0% Figure 6.12: Sodium concentration effect on DMS-water activity coefficient for sodium salts and black liquor at 50°C 119 Chapter 6: Results and Discussion 400 350 .2 o 300 -!fc <u o o 250 200 -o < 150 -100 -•Sodium salts solution Black liquor 0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% Sodium (wt%) Figure 6.13: Sodium concentration effect on DMS-water activity coefficient for sodium salts and black liquor at 70°C c .2 "o £ o o > o < 400 350 300 250 200 150 100 •Sodium salts solution Black liquor 0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% Sodium (wt%) Figure 6.14: Sodium concentration effect on DMS-water activity coefficient for sodium salts and black liquor at 90°C 120 Chapter 6: Results and Discussion There appears to be a very good correlation between the increase in the activity coefficient and the sodium concentration, with the 95% confidence interval of the black liquor activity coefficient data consistently overlapping the sodium salt activity coefficient data. 6 .3 P h a s e E q u i l i b r i a T e s t i n g D a t a R e g r e s s i o n The phase equilibria experimental data were regressed to fit functions using the software tools described in Section 5.2.6. 6.3 .1 F i t t i n g o f H e n r y ' s C o n s t a n t E q u a t i o n Henry's constants for all solutions tested at all temperatures were calculated from the activity coefficients in Table 6.2 using Equation (5.4). A regression analysis, similar to the technique described in Section 5.4.1, was used to determine the best fit parameters for the temperature dependent Henry's constant Equation (3.20). The adjustable parameters for this equation include Henry's constant at 298 K , and a temperature dependency constant, - A S 0 , H /R. The best fit for these parameters are shown in Table 6.5. T a b l e 6 . 5 : Regressed Henry's constants for dimethyl sulphide in solutions tested Solution 1, 2 9 8 K K H (mol fract/mol fract)/MPa - A S 0 , H / R (K) Temperature range (°C) 1 0.093 2780 20 to 90 2 0.091 2820 20 to 90 3 0.088 2870 20 to 90 4 0.082 2560 20 to 90 5 0.075 2440 20 to 90 6 0.069 2570 20 to 90 7 0.059 2590 20 to 90 121 Chapter 6: Results and Discussion Table 6.5 (cont.): Regressed Henry's constants for dimethyl sulphide in solutions tested Solution u 2 9 8 K K H (mol fract/mol fract)/MPa - A s o l H /R (K) Temperature range (°C) 8 0.046 2600 20 to 90 9 0.063 2700 20 to 90 10 0.062 2660 20 to 90 11 0.061 2470 20 to 90 12 0.053 2350 20 to 90 13 0.063 2930 20 to 90 Henry's constant for a binary DMS-water system (solution 1) was found to be 0.093 (mol/mol)/MPa at 298 K , which compares favourably to the literature values presented in Table 3.5 in Section 3.1.4, which range from 0.085 to 0.100 (mol/mol)/MPa. The literature sources that also provided Henry's constant data for temperatures other than 298 K are shown on Figure 6.15 along with the experimental data and best fit line from this work. 0.35 ~ 0 JS a 5 f w £ 0 fro S ^ o i 30 25 20 15 10 0.05 0.00 • This work This work (best fit) Przyjazny et al. (1983) Dacey et al. (1984) De Bruyn et al. (1995) 270 290 310 330 Temperature (K) 350 370 Figure 6.15: Henry's constant for D M S in water as a function of temperature 122 \ Chapter 6: Results and Discussion In Figure 6.15, there is good agreement between the Henry's constant Equation (3.20) fitted from the experimental results from this work and that from the literature sources. There are two literature sources that give Henry's constant data for a DMS-water system with sodium salts added, with details given in Table 3.6. Przyjazny et al. (1983) provide data for solutions containing 0.7 M N a C l and 0.33, 0.66, 1.00, and 1.33 M N a 2 S 0 4 , and Tormund (1997) provides data for solutions containing 0.3 and 3.0 M N a C l . Henry's constant data for the sodium salts solutions from this work (solutions 6, 7, and 8) and for the two literature sources are plotted in Figure 6.16 against sodium concentration. Data are plotted at 40°C as this is the lower end of the valid range for the Tormund (1997) data (40 to 80°C), and at 70°C, as this is the upper end of the valid range for the Przyjazny et al. (1983) data (25 to 70°C). • This work 40°C —•—Tormund (1997) 40°C —-Przy j azny et al. (1983) 40°C • This work 70°C —O—Tormund (1997) 70°C —O -Przyjazny etal. (1983) 70°C - — - - — - © - - . _ . 0.00 0 0%. 1.0% 2.0% 3.0% 4.0% 5.0% 6.0% 7.0% Sodium (wt%) F i g u r e 6 . 1 6 : Henry's constant for D M S in water as a function of sodium salts concentration at 40°C and 70°C Taking into account the limited data, there is reasonably good agreement between the experimental results from this work and that from the literature sources. 123 6.3.2 Fi t t ing of N R T L Equat ion for D M S - W a t e r System Chapter 6: Results and Discussion A regression analysis, described in Section 5.4.1, was used to find the best fit for the N R T L equation, using the adjustable rand a parameters, to the experimental binary-DMS water activity coefficient data (solution 1). The N R T L equation consists of the first two terms of the e N R T L Equation (3.11), which can be simplified to Equation (5.2) for a system at infinite dilution. Table 6.6: Regressed N R T L equation parameters for DMS(i)-water(j) system N R T L Parameter Aspen Plus Results Excel Results a u 1.00 2.93 by 4.04 10.1 2.48 2.06 \ 567 549 0.3 (fixed) 0.3 (fixed) The regression analysis was done using the data regression tool in Aspen Plus and a technique using the solver tool in Excel , with each returning a different set of parameters (Table 6.6). Even though the reported parameters are different, the fit is identical; i.e., the resulting best fit line calculated using the N R T L equation and shown in Figure 6.17 is the same using either of the sets of parameters. 124 Chapter 6: Results and Discussion 200 190 c 180 o • • Experimental Data — NRTL Equation Best Fit t 110 100 290 310 330 350 370 Temperature (K) Figure 6.17: DMS-water activity coefficient experimental data and N R T L equation best fit (R 2 = 0.95) It is common for different regression tools to report different best fit parameters. This is due to the fact that with four adjustable parameters, there are multiple sets of parameters that can provide a best fit. It is also possible for the same regression tool to report different sets of parameters, with the result depending on a number of factors including the specified starting values, fit tolerances and regression criteria (objective function). Both tools used here incorporate a least squares criteria; the adjustable parameters that provide the best fit are determined by minimizing the sum of squares between the experimental value and the calculated value. The Excel technique was set up using a simple sum of squares objective function incorporating the activity coefficient; Aspen Plus uses a more complex weighted sum of squares objective function which also incorporates the second measured variable, temperature. Normally this objective function may be expected to produce a better fit, but because the experimental data was reported at set temperatures with varying activity coefficients, the objective function was heavily weighted to the activity coefficient, and both tools reported a set of parameters (albeit different) that provide an equally good fit. 125 Chapter 6: Results and Discussion 6 . 3 . 3 C o m p a r i s o n o f N R T L F i t t o O t h e r S o u r c e s The results of the regression fit for this work were compared to the results of a fit using the N R T L parameters of Olsson and Zacchi (2001). Because there are multiple solutions, the parameters can not be compared directly; instead, the results calculated from the regressed parameters must be compared. As described in Section 3.1.4, Olsson and Zacchi held a^  and a^  constant at zero. Their parameters are given in Table 6.7 and the activity coefficients determined from these parameters are shown in Figure 6.18. The experimental results were also compared to the activity coefficient estimated using the U N I F A C model. The U N I F A C method is not based on experimental data for a DMS-water system; instead, it is based on a group contribution method discussed in Section 3.1.2. The U N I F A C method is incorporated into Aspen Plus, so this was used to generate the N R T L parameters for a binary DMS-water system. Aspen includes D M S in its databank, with the U N I F A C groups identified as C H 3 S - and - C H 3 based on a structure of C H 3 - S - C H 3 . The N R T L parameters from the U N I F A C method are given in Table 6.7 and the resulting activity coefficients are shown in Figure 6.18. Both Olsson and Zacchi (2001) and the U N I F A C method hold a^  and a^  constant at zero. To check the validity of this, a regression was attempted using the experimental data from this work, with the results also show in Table 6.7 and Figure 6.18. T a b l e 6 . 7 : D M S (i) in water (j) parameters for N R T L equation from other sources N R T L Parameter Olsson and Zacchi (2001) U N I F A C (using Aspen Plus) This work (holding a^  and a,; at zero) \ 0 (fixed) 0 (fixed) 0 (fixed) by 696.07 37.31 2143.15 h 0 (fixed) 0 (fixed) 0 (fixed) 1576.49 746.81 1301.87 0.3 (fixed) 0.3 (fixed) 0.3 (fixed) 126 Chapter 6: Results and Discussion For comparison purposes, the experimental activity coefficient data points determined from this work are shown in Figure 6.18. 700 ^ 600 | 500 % 400 o o >,300 + J £ 200 u < 100 0 290 s s s • Experimental Data NRTL Best Fit, aij = aji = 0 - - Olsson and Zacchi — 'UNIFAC s s T. s s s 1 — 310 330 350 Temperature (K) 370 Figure 6.18: Activity coefficients determined from Olsson and Zacchi (2001) N R T L parameters, U N I F A C parameters and best fit to experimental data and holding a^  and ajf constant at zero (R 2 = 0.26) The N R T L parameters supplied by Olsson and Zacchi (2001) overestimate the activity coefficient by a large degree, while the U N I F A C parameters underestimate by almost equally large degree. It is not clear why either provide such a poor fit to the experimental data. The work of Olsson and Zacchi (2001) involved mixtures at much higher concentrations, so they may not have been concerned with the accuracy at infinite dilution when regressing their data. The DMS-water mixture is highly non-ideal, which likely explains the error in the U N I F A C method, which in general loses accuracy as the mixture moves away from the ideal. In either case, it does not appear that holding afj and constant at zero is a good simplification. The best fit that could be achieved with the experimental data from this work with a^  and a^  fixed at zero resulted in an R 2 of 0.26 (Figure 6.18) versus an R 2 of 0.95 when these parameters are allowed to vary (Figure 6.17). 127 Chapter 6: Results and Discussion 6 .3 .4 R a n g e o f V a l i d i t y f o r I n f i n i t e D i l u t i o n The activity coefficient for a DMS-water system, using the parameters from Table 6.6, was calculated using the N R T L Equation (3.11) and a simplified version of the N R T L Equation (5.2). The latter equation was simplified by assuming that the solute D M S was at infinite dilution in the solvent water. Over the temperature range of 20 to 90°C, the maximum error introduced by assuming infinite dilution is 0.01% at 100 ppm (mol) and 0.1% at 1000 ppm (mol). The same assumption is made when using Henry's constant to model a system, with the same error introduced. The highest concentration measured as part of this work is less than 100 ppm (mol), thus the assumption of infinite dilution is applicable because the introduced error is negligible. 6 . 3 . 5 F i t t i n g o f e N R T L E q u a t i o n f o r D M S - W a t e r - S o d i u m S a l t s S y s t e m The binary DMS-water N R T L parameters determined from this work (Table 6.6) were then used in the e N R T L equation to allow solute-salt parameters to be fitted. The sodium salts (solutions 6, 7, and 8) activity coefficient data were fitted using the e N R T L Equation (3.11). A s discussed in Section 5.4, the system was modelled as a ternary DMS(i)-water(j)-black liquor sodium salts(ca) system. The e N R T L equation is based on mole fractions, with the mole fraction sodium content, x c a , calculated from the weight fraction based on the composition given in Table 5.1. The mole fractions of sodium used were 0.0067, 0.0133, and 0.0200 for 2, 4 and 6 wt%, respectively. For a dilute weak electrolyte system, a good fit can usually be achieved by fixing the salt-solute and salt-solvent parameters to zero. A s well , it is normal to fix the solvent-salt arparameters at 0.2 (Chen and Song, 2004). For regression using Aspen Plus, it was discovered, after much trial and error, that fixing the solvent-salt and solute-salt r parameters at zero resulted in the r parameters defaulting to 8 and -4 for water-ion and ion-water, respectively, and 10 and -2 for non-water-ion and ion-non-water, respectively. This was rectified by fixing the r parameters at 1 e-6 instead of zero. The best fit e N R T L parameters are shown in Table 6.8. 128 Chapter 6: Results and Discussion Table 6.8: Regressed e N R T L equation parameters for DMS(i)-water(j)-black liquor sodium salts (ca) system e N R T L Parameter D M S r i ca a , c a 39.1 b i c a 163 T ca i a c a i 0(fixed) b c a , 0(fixed) ^ica ^ c a i 0.05 * j c a a i c a 2.77 b , c a 2292 r caj a c a j 0 (fixed) b c a , 0 (fixed) ^ i c a ^ c a j 0.2 (fixed) For consistency, the experimental data and the best fit lines for the e N R T L equation are shown in weight percent sodium content, in Figure 6.19. 400 • Exp. 20°C 100 '— - i - r - - T -0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% Sodium (wt%) Figure 6.19: DMS-water-sodium salts activity coefficient experimental data and e N R T L equation best fit (R 2 = 0.95) 129 Chapter 6: Results and Discussion The e N R T L equation, using the parameters in Table 6.6 and 6.8, was used to estimate the activity coefficients for the literature sources for the black liquor samples studied (solutions 9 to 13). The calculated activity coefficient values and experimental values are shown in Table 6.9, along with the percent difference between these figures. Table 6.9: DMS-black liquor activity coefficients Solution Temperature (K) Sodium (mole frac.) Activity Coefficient Calculated Experimental Difference (%) 9 293 0.0063 252 255 0.9 10 293 0.0074 269 261 -3.1 11 293 0.0061 251 269 6.8 12 293 0.0090 299 314 5.1 13 293 0.0078 275 251 -8.5 9 323 0.0063 202 208 2.9 10 323 0.0074 214 211 -1.4 11 323 0.0061 202 203 0.6 12 323 0.0090 235 224 -4.5 13 323 0.0078 218 223 2.5 9 343 0.0063 179 183 2.4 10 343 0.0074 189 189 0.3 11 343 0.0061 178 168 -5.8 12 343 0.0090 206 187 -9.3 13 343 0.0078 192 214 11.4 9 363 0.0063 161 170 6.0 10 363 0.0074 169 174 3.3 11 363 0.0061 160 164 2.7 12 363 0.0090 183 179 -2.1 13 363 0.0078 171 173 0.9 130 Chapter 6: Results and Discussion There is good agreement, typically with only a few percent difference and no more than 12%, between the activity coefficient calculated using the e N R T L equation and the experimental values. The data are also shown on a scatter diagram in Figure 6.20. 350 300 ro co 250 Q 2 200 c <D .1150 i_ o * 1 0 0 LU 50 0 0 50 100 150 200 250 300 350 eNRTL Prediction F i g u r e 6 . 2 0 : Comparison of activity coefficient determined from e N R T L equation to black liquor experimental data (solutions 9 to 13) The e N R T L Equation (3.11), using the parameters in Table 6.6 and 6.8, was used for modelling the D M S vapour-liquid equilibrium for black liquor systems for the mi l l testing data presented in Section 6.4. 6 .3 .6 F i t t i n g o f e N R T L E q u a t i o n f o r H 2 S - , M M - a n d D M D S - W a t e r - S o d i u m S a l t s S y s t e m s Although D M S was used as a surrogate compound for modelling T R S emission in the brown stock washing area, there are other areas of the mi l l , such as the black liquor evaporators, where modelling for each individual TRS compound may be desirable. Activity coefficients for H 2 S , M M 131 Chapter 6: Results and Discussion and D M D S were calculated using Equation 5.4 from the Henry's constants found in the literature (Tables 3.1 to 3.8). The data of Barrett et al. (1988), Przyjazny et al. (1983) and Tormund (1997) were used because solutions with sodium salts were also tested. Using the same regression technique that was used for D M S , the best fit parameters for the e N R T L equation were determined; these are given in Table 6.10. Table 6.10: Regressed e N R T L equation parameters for H 2 S - , M M - and DMDS(i)-water(j)-sodium salts (ca) system e N R T L Parameter H 2 S M M D M D S a u 42.8 1.87 2.76 \ -190 -104 24.6 h % 4.5 4.71 4.99 \ -350 -414 353 alj = asl 0.3 (fixed) 0.3 (fixed) 0.3 (fixed) T-ica 60.1 -0.22 75.8 bica -78.6 3929 -9102 T • cai •^cai 0 (fixed) 0 (fixed) 0 (fixed) t>cai 0 (fixed) 0 (fixed) 0 (fixed) ^ica ^cai 0.05 (fixed) 0.05 (fixed) 0.05 (fixed) j^ca jca 8.78 17.5 19.5 bjca -145 -2370 -1699 r caj O acaj 0 (fixed) 0 (fixed) 0 (fixed) b c a J 0 (fixed) 0 (fixed) 0 (fixed) j^ca ^cai 0.2 (fixed) 0.2 (fixed) 0.2 (fixed) The parameters given in Table 6.10, and those determined for D M S and given in Tables 6.6 and 6.8, were used to calculate activity coefficients at a typical brown stock washing temperature of 80°C; the results are shown in Figure 6.21. 132 Chapter 6: Results and Discussion o o © 00 c o o O o > o < 2500 2000 1500 1000 500 0 — DMDS - - D M S --- MM - H2S 0.0% 1.0% 2.0% Sodium (wt%) 3.0% Figure 6 .21: Activity coefficients determined from the e N R T L equation for the T R S compounds at 80°C The D M D S activity coefficients are an order of a magnitude higher than the next highest, D M S , but this is still considerably less than for wood extractives such as the terpenes, which have activity coefficients that exceed 100,000 (Fichan et al., 1999). Hydrogen sulphide and methyl mercaptan wi l l dissociate at high p H (Figure 3.2). The undissociated fraction, i.e., the fraction that wi l l potentially be equilibrium with the vapour phase, can be estimated using Equations 3.16 and 3.18. 6.4 Mill Sampling and Testing Results Sample collection and testing was conducted at the Howe Sound Pulp and Paper mi l l located at Port Mellon, British Columbia. Composition data for liquid and vapour streams at a pulp mi l l were required to test the concept of using vapour-liquid equilibrium correlations to estimate T R S emissions. Since no such data could be found in the literature, a mi l l testing program was organized. In the kraft pulping process, for non-combustion sources, the T R S compounds are mainly emitted from the digester area and from the black liquor evaporation plant. Not including sources collected to the C N C G system, the brown stock washing TRS emissions are much higher than those 133 Chapter 6: Results and Discussion from the black liquor evaporation area, which was confirmed as part of this work, so mi l l testing was focussed in the brown stock washing area. Testing around the brown stock washing area of the mi l l was done from mid September to early December 2005, with samples collected on twelve days spread over this period. The raw data from these testing sessions are provided in Appendix F. Operating data, required to develop heat and mass balances across the system, were collected from the mill-wide optimization system (MOPS) historian. The M O P S historian data are included in Appendix G . The results of the twelve days of testing are summarized in Tables 6.11 to 6.22. T a b l e 6 . 1 1 : M i l l test results for September 16, 2005 VAPOUR SAMPLES Time Flow Temp. Relative H2S M M DMS DMDS AnrVmin °C Humidity ppm (mol) ppm (mol) ppm (mol) ppm (mol) kg S / day kg S / day kg S/day kg S/day VS6 Decker Washer Filtrate Tank 13:30 15.1 83.0 100% nd nd 716 52.4 0.0 0.0 17.1 2.5 VS6 Decker Washer Filtrate Tank 15:45 15.1 83.0 100% nd nd 545 27.3 0.0 0.0 13.0 1.3 LIQUOR SAMPLES Time pH Temp. Dissolved H2S M M DMS DMDS °C Solids ppm (mol) ppm (mol) ppm (mol) ppm (mol) LS6 Decker Washer Filtrate - 13:30 11.7 80.0 nd nd 1.84 0.17 LS6 Decker Washer Filtrate 15:45 11.7 80.0 nd nd 1.26 0.01 T a b l e 6 . 1 2 : M i l l test results for September 21, 2005 VAPOUR SAMPLES Time Flow Temp. Relative H2S M M DMS DMDS AiwVmin °c Humidity ppm (mol) ppm (mol) ppm (mol) ppm (mol) kg S/day kg S / day kg S / day kgS/day VS 1 Diffusion Washer Filtrate Tanks 14:10 3.2 60.0 77% nd nd 2187 20.2 0.0 0.0 11.9 0.2 LIQUOR SAMPLES Time PH Temp. Dissolved H2S M M DMS DMDS °C Solids ppm (mol) ppm (mol) ppm (mol) ppm (mol) LS2 1 st Stage Diffusion Washer Filtrate 14:10 11.7 80.0 nd nd 8.74 0.33 134 Chapter 6: Results and Discussion T a b l e 6 . 1 3 : M i l l test results for September 22, 2005 VAPOUR SAMPLES Time Flow Temp. Relative H2S M M DMS DMDS AnvYmin ° C Humidity ppm (mol) ppm (mol) ppm (mol) ppm (mol) kg S / day kg S / day kg S/day kg S/day VS6 Decker Washer Filtrate Tank 10:00 15.9 75.0 100% 1.9 2.3 1396 226 0.0 0.1 35.9 11.6 LIQUOR SAMPLES Time pH Temp. Dissolved H2S M M DMS DMDS °C Solids ppm (mol) ppm (mol) ppm (mol) ppm (mol) LS6 Decker Washer Filtrate 10:00 12.0 80.0 nd nd 2.91 0.65 T a b l e 6 . 1 4 : M i l l test results for September 28, 2005 VAPOUR SAMPLES Time Flow Temp. Relative H2S M M DMS DMDS AmVmin °c Humidity ppm (mol) ppm (mol) ppm (mol) ppm (mol) kg S / day kg SI day kg S / day kg S / day VS2 Diffusion Washer 16:05 3.2 52.0 71% nd nd 132 5.4 0.0 0.0 0.7 0.1 VS6 Decker Washer Filtrate Tank 11:30 15.1 52.0 71% nd nd 251 3.1 0.0 0.0 6.5 0.2 VS6 Decker Washer Filtrate Tank 16:05 15.1 62.0 70% nd nd 194 4.1 0.0 0.0 4.9 0.2 LIQUOR SAMPLES Time pH Temp. Dissolved H2S M M DMS DMDS °C Solids ppm (mol) ppm (mol) ppm (mol) ppm (mol) LS6 Decker Washer Filtrate 11:30 12.0 52.0 nd nd 0.82 nd LS6 Decker Washer Filtrate 16:05 12.0 62.0 nd nd 0.93 nd T a b l e 6 . 1 5 : M i l l test results for September 29, 2005 VAPOUR SAMPLES Time Flow Temp. Relative H2S M M DMS DMDS AmVmin °C Humidity ppm (mol) ppm (mol) ppm (mol) ppm (mol) kg S / day kg S / day kg S / day kg S/day VS2 Diffusion Washer 16:00 0.6 63.0 100% 3.3 6.8 486 27.6 0.0 0.0 0.5 0.1 VS3 Blow Tank 16:00 54.5 75.0 100% 2.1 3.3 810 84.4 0.2 0.3 71.4 14.9 VS4 Screen Feed Tank 09:40 10.9 62.0 82% nd nd 42.6 2.8 0.0 0.0 0.8 0.1 VS5 Screen Dilution Tank 09:40 3.2 36.0 75% nd nd 12.0 nd 0.0 0.0 0.1 0.0 VS6 Decker Washer Filtrate Tank 13:00 15.1 76.0 100% nd 11.6 2056 110 0.0 0.3 50.0 5.3 VS7 Decker Washer Hood 13:00 159.9 63.0 86% nd nd 51.5 4.2 0.0 0.0 13.8 2.3 LIQUOR SAMPLES Time pH Temp. Dissolved H2S M M DMS DMDS °C Solids ppm (mot) ppm (mol) ppm (mol) ppm (mol) LS4 Stock from Diffusion Washer 16:00 11.7 81.0 6.7% nd nd 6.96 1.30 LS6 Decker Washer Filtrate 13:00 11.7 76.0 4.9% nd nd 4.94 0.40 LS8 02 Filtrate 13:00 11.4 77.0 2.0% nd nd 0.33 nd 135 Chapter 6: Results and Discussion Table 6.16: M i l l test results for September 30, 2005 VAPOUR SAMPLES Time Flow Temp. Relative 112 S M M DMS DMDS AnrVmin °C Humidity ppm (mol) ppm (mol) ppm (mol) ppm (mol) kgS/day kg S / day kg S / day kg S / day VS1 Diffusion Washer Filtrate Tanks 08:30 4.3 80.0 81% 3.9 35.2 14938 332 0.0 0.2 102.0 4.5 VS1 Diffusion Washer Filtrate Tanks 10:30 4.8 80.0 81% 5.4 54.1 17336 216 0.0 0.4 133.1 3.3 VS2 Diffusion Washer 14:00 2.4 70.0 83% 3.4 5.8 952 52.0 0.0 0.0 3.7 0.4 VS3 Blow Tank 14:00 56.9 78.0 88% nd 4.6 1169 87.7 0.0 0.4 106.5 16.0 LIQUOR SAMPLES Time p H Temp. Dissolved H2S M M DMS DMDS °C Solids ppm (mol) ppm (mol) ppm (mol) ppm (mol) LS2 1 st Stage Diffusion Washer Filtrate 08:30 11.9 83.0 nd nd 47.5 6.25 LS3 2nd Stage Diffusion Washer Filtrate 08:30 11.9 82.0 nd nd 19.0 2.53 LS2 1 st Stage Diffusion Washer Filtrate 10:30 12.0 83.0 11.7% nd nd 39.7 6.15 LS3 2nd Stage Diffusion Washer Filtrate 10:30 12.0 82.0 9.3% nd nd 20.5 2.45 LS4 Stock from Diffusion Washer 14:00 12.0 81.0 9.8% nd nd 7.36 1.56 LS6 Decker Washer Filtrate 15:00 11.9 79.0 8.2% nd nd 4.32 1.11 Table 6.17: M i l l test results for November 22, 2005 VAPOUR SAMPLES Time Flow Temp. Relative 112 S M M DMS DMDS Am'/min °C Humidity ppm (mol) ppm (mol) ppm (mol) ppm (mol) kg S / day kg S / day kg S / day kg S / day VS1 Diffusion Washer Filtrate Tanks 15:00 3.2 70.0 83% 2.7 21.7 16605 518 0.0 0.1 87.5 5.5 VS2 Diffusion Washer 15:00 0.6 61.0 95% nd 4.0 522 31.0 0.0 0.0 0.5 0.1 VS6 Decker Washer Filtrate Tank 10:00 19.3 75.0 100% nd 10.5 2471 162 0.0 0.3 77.0 10.1 VS7 Decker Washer Hood 10:00 186.6 66.0 100% nd nd 64.8 3.9 0.0 0.0 20.1 2.4 LIQUOR SAMPLES Time p H Temp. Dissolved ICS M M DMS DMDS °c Solids ppm (mol) ppm (mol) ppm (mol) ppm (mol) LSI Stock from Digester 15:00 80.0 nd nd 27.3 5.57 LS2 1 st Stage Diffusion Washer Filtrate 15:00 79.0 nd nd 36.0 4.64 LS3 2nd Stage Diffusion Washer Filtrate 15:00 78.0 nd nd 20.0 2.40 LS4 Stock from Diffusion Washer 15:00 77.0 nd nd 7.69 0.86 LS5 Stock to Decker Washer 10:00 76.0 nd nd 5.10 0.52 LS6 Decker Washer Filtrate 10:00 76.0 nd nd 5.02 0.63 LS7 Stock from Decker Washer 10:00 76.0 nd nd 0.21 0.03 LS8 02 Filtrate 10:00 81.0 nd nd 0.50 nd 136 Chapter 6: Results and Discussion Table 6.18: M i l l test results for November 23, 2005 VAPOUR SAMPLES Time Flow Temp. Relative H2S M M DMS DMDS Ani'/min °C Humidity ppm (mol) ppm (mol) ppm (mol) ppm (mol) kg S / day kg S / day kg SI day kg S/day VS1 Diffusion Washer Filtrate Tanks 09:00 3.2 65.0 100% 2.2 12.1 8976 154 0.01 0.06 48.0 1.7 VS3 Blow Tank 09:00 54.5 71.0 100% nd 2.6 1124 71.5 0.00 0.23 100.2 12.7 VS6 Decker Washer Filtrate Tank 13:30 17.6 75.0 100% nd nd 2725 122 0.00 0.00 77.5 6.9 VS7 Decker Washer Hood 13:30 186.6 65.0 91% nd nd 204 7.0 0.00 0.00 63.4 4.4 LIQUOR SAMPLES Time pH Temp. Dissolved H2S M M DMS DMDS °C Solids ppm (mol) ppm (mol) ppm (mol) ppm (mol) LS1 Stock from Digester 09:00 11.7 80.0 13.4% nd nd 35.5 5.39 LS2 1 st Stage Diffusion Washer Filtrate 09:00 11.6 79.0 10.8% nd nd 43.5 5.18 LS3 2nd Stage Diffusion Washer Filtrate 09:00 11.6 78.0 8.9% nd nd 23.5 2.39 LS4 Stock from Diffusion Washer 09:00 11.7 77.0 8.6% nd nd 8.90 1.08 LS5 Stock to Decker Washer 13:30 11.6 76.0 7.6% nd nd 4.64 0.68 LS6 Decker Washer Filtrate 13:30 11.6 76.0 6.9% nd nd 5.57 0.92 LS7 Stock from Decker Washer 13:30 11.6 76.0 4.5% nd nd 0.44 0.08 LS8 02 Filtrate 13:30 11.5 80.0 4.1% nd nd 0.39 0.02 Table 6.19: M i l l test results for November 24, 2005 VAPOUR SAMPLES Time Flow Temp. Relative H2S M M DMS DMDS Am3/min °C Humidity ppm (mol) ppm (mol) ppm (mol) ppm (mol) kg S/day kg S / day kg S/day kg S/day VS1 Diffusion Washer Filtrate Tanks 09:30 3.2 65.0 100% 3.0 17.0 10138 125 0.02 0.09 54.2 1.3 VS3 Blow Tank 09:30 56.9 72.0 100% nd 15.1 1295 130 0.00 1.40 120.1 24.1 VS6 Decker Washer Filtrate Tank 13:30 18.4 72.0 100% 2.6 26.8 3789 69.5 0.08 0.81 113.9 4.2 VS7 Decker Washer Hood 13:30 186.6 63.0 100% nd nd 185 8.1 0.00 0.00 57.8 5.1 LIQUOR SAMPLES Time pH Temp. Dissolved H2S M M DMS DMDS °C Solids ppm (mol) ppm (mol) ppm (mol) ppm (mol) LSI Stock from Digester 09:30 11.7 82.0 nd nd 28.1 3.94 LS2 1 st Stage Diffusion Washer Filtrate 09:30 11.6 80.0 nd nd 31.3 2.59 LS3 2nd Stage Diffusion Washer Filtrate 09:30 11.6 79.0 nd nd 19.7 1.79 LS4 Stock from Diffusion Washer 09:30 11.7 78.0 nd nd 7.32 0.89 LS5 Stock to Decker Washer 13:30 11.6 74.0 nd nd 6.97 1.12 LS6 Decker Washer Filtrate 13:30 11.6 74.0 nd nd 7.88 1.31 LS7 Stock from Decker Washer 13:30 11.6 74.0 nd nd 0.82 0.21 LS8 02 Filtrate 13:30 11.6 79.0 nd nd 0.31 nd 137 Chapter 6: Results and Discussion Table 6.20: M i l l test results for November 30, 2005 V A P O U R S A M P L E S Time Flow Temp. Relative H2S M M D M S D M D S Am 7m in °C Humidity ppm (mol) ppm (mol) ppm (mol) ppm (mol) kg S/day kg S / day kg S / day kg S / day VS1 Diffusion Washer Filtrate Tanks 15:00 3.2 64.0 100% 2.2 14.6 12886 157 0.01 0.08 69.1 1.7 VS3 Blow Tank 15:00 56.9 70.0 100% nd nd 692 58.2 0.00 0.00 64.5 10.9 VS6 Decker Washer Filtrate Tank 11:00 13.4 74.0 100% nd 8.4 1987 111 0.00 0.18 43.2 4.8 VS7 Decker Washer Hood 11:00 186.6 64.0 95% nd nd 56.8 4.3 0.00 0.00 17.7 2.7 L I Q U O R S A M P L E S Time PH Temp. Dissolved H2S M M D M S D M D S °C Solids ppm (mol) ppm (mol) ppm (mol) ppm (mol) LSI Stock from Digester 15:00 11.6 80.0 nd nd 12.9 2.23 LS2 1st Stage Diffusion Washer Filtrate 15:00 11.6 79.0 nd nd 33.8 6.75 LS3 2nd Stage Diffusion Washer Filtrate 15:00 11.6 78.0 nd nd 22.6 3.15 LS4 Stock from Diffusion Washer 15:00 11.6 77.0 nd nd 7.20 0.85 LS5 Stock to Decker Washer 11:00 11.4 75.0 nd nd 5.18 1.07 LS6 Decker Washer Filtrate 1.1:00 11.4 75.0 nd nd 4.38 0.46 LS7 Stock from Decker Washer 11:00 11.4 75.0 nd nd 0.43 0.10 LS8 02 Filtrate 11:00 11.4 80.0 nd nd 0.13 nd Table 6.21: M i l l test results for December 1, 2005 V A P O U R S A M P L E S Time Flow Temp. Relative H2S M M D M S D M D S Am'/min °c Humidity ppm (mol) ppm (mol) ppm (mol) ppm (mol) kg S / day kg S/day kg S / day kg S/day VS1 Diffusion Washer Filtrate Tanks 14:00 2.7 63.0 100% 3.2 14.2 16299 150 0.01 0.06 73.1 1.3 VS3 Blow Tank 14:00 54.5 71.0 96% nd 8.1 1095 65.7 0.00 0.72 97.6 11.7 VS6 Decker Washer Filtrate Tank 11:00 11.7 77.0 100% nd 4.6 1590 61.0 0.00 0.09 30.0 2.3 VS7 Decker Washer Hood 11:00 186.6 65.0 100% nd nd 95.9 4.5 0.00 0.00 29.8 2.8 L I Q U O R S A M P L E S Time pH Temp. Dissolved H2S M M D M S D M D S ° C Solids ppm (mol) ppm (mol) ppm (mol) ppm (mol) LSI Stock from Digester 14:00 11.6 81.0 13.8% nd nd 50.3 6.52 LS2 1 st Stage Diffusion Washer Filtrate 14:00 11.6 80.0 12.7% nd nd 40.9 5.19 LS3 2nd Stage Diffusion Washer Filtrate 14:00 11.4 79.0 9.8% nd nd 21.8 3.40 LS4 Diffusion Washer Stock Out 14:00 11.5 78.0 8.5% nd nd 7.86 1.00 LS5 Stock to Decker Washer 11:00 11.7 76.0 7.0% nd nd 4.77 0.66 LS6 Decker Washer Filtrate 11:00 11.7 76.0 6.7% nd nd 4.09 0.68 LS7 Stock from Decker Washer 11:00 11.7 76.0 5.1% nd nd 0.57 0.11 LS8 02 Filtrate 11:00 11.7 81.0 4.5% nd nd 0.19 nd 138 Chapter 6: Results and Discussion Table 6.22: M i l l test results for December 2, 2005 VAPOUR SAMPLES Time Flow Am'/min Temp. °C Relative Humidity H2S ppm (mol) kg S / day M M DMS DMDS ppm (mol) ppm (mol) ppm (mol) kg S/day kg S/day kg S/day Chip Bin (23/11/05) 13:30 48.0 78.0 88% 385 911 1486 11.5 29.6 70,0 114.2 1.8 Chip Bin (24/11/05) 09:30 33.6 77.0 100% 231 599 1033 4.2 12.5 32.3 55.7 0.5 Chip Bin (01/12/05) 14:00 43.2 80.0 100% 267 1145 2854 5.9 18.3 78.7 196.3 0.8 D N C G - total to boiler 09:45 255.9 19.0 100% 79.6 245 767 28.5 39.2 120.7 378.1 28.1 HBL Tanks 08:30 6.7 94.5 100% 239 387 264 94.2 2.5 4.0 2.7 1.9 W B L & Cond. Tanks (& sweep air) 09:00 10.1 35.0 100% 18.5 32.3 218 14.6 0.3 0.6 4.0 0.5 D N C G - total recovery area 09:20 16.8 48.0 100% 85.2 163 215 38.3 2.5 4.8 6.3 2.2 LIQUOR SAMPLES Time pH Temp. °C Dissolved Solids H2S M M DMS DMDS ppm (mol) ppm (mol) ppm (mol) ppm (mol) Screen Feed Tank Liquor (29/09/05) 09:40 Screen Dilution Tank Stock (29/09/05 09:40 11.7 11.7 76.0 76.0 nd nd nd nd 2.24 2.32 nd nd The mil l operation was stable for all testing days, with one exception. The mil l was shut-down from September 22 n d to the early morning of the 28 l h . The samples with the lowest concentrations were collected on the 28 t h in the late morning and middle afternoon, apparently before the dimethyl sulphide concentrations could build up in the system. The concentrations were back up to "normal" levels by the next day. To keep dissolved alkali lignin from precipitating, the kraft pulping process is operated with sufficient excess alkali to keep the brown stock washing process above a p H of about 11 and more typically between 11.5 and 12.5 (Grace etal., 1989). This was confirmed during this testing session, with the liquor samples ranging in pH from 11.4 to 12.0 (Tables 6.11 to 6.22). A s discussed in Section 3.1.4, hydrogen sulphide and methyl mercaptan are weak acids and are essentially fully dissociated above a pH of 11. A s expected, the hydrogen sulphide and methyl mercaptan concentrations in the vapour samples were either below the detection limit or at very low concentrations, the highest measured being 5.4 and 54.1 ppm (mol) (Table 6.16), respectively. 139 Chapter 6: Results and Discussion With hydrogen sulphide and methyl mercaptan concentrations typically below detection levels due to the high p H , and with very little oxygen in the system resulting in relatively low concentrations of dimethyl disulphide, dimethyl sulphide was the greatest contributor to the T R S in the vent gases. On all testing days, dimethyl sulphide contributed over 90% of the sulphur in the vent gases. Thus, for modelling purposes, it was decided to use dimethyl sulphide as a surrogate compound for T R S and all phase equilibria testing focussed on this compound, as discussed in Section 6.2. 6.4 .1 H o w e S o u n d E q u i p m e n t D i m e n s i o n s a n d O p e r a t i n g C o n d i t i o n s Some D N C G systems are designed to include "sweep air." A n opening is included in the tank and the vacuum created by the D N C G collection system draws outside air through the tank, sweeping along the volatiles into the D N C G system. The sources at Howe Sound do not include sweep air, instead they are designed to be sealed against air ingress. Any air that is drawn into the system is undesired and is referred to as "tramp air." A sealed system such as Howe Sound requires vacuum breakers to protect tankage, but the volumetric flow of N C G collected wi l l be lower and the mass loading of volatiles wi l l be lower because there wi l l be less stripping of volatiles. Equipment dimensions, vapour-to-liquid volume ratios, and liquid phase residence times are provided for all equipment in Table 6.23. 140 Chapter 6: Results and Discussion Table 6.23: Equipment dimensions and operating conditions Equipment Type Tank inside c 'G dimensions and volume Typical liquid level (m) Liquid volume (r Vapour-liquid volume ratio Liquid flow (mVm Residence time (m: l s l Stage Vertical 4.9 m 4.4 83 0.33 10.2 8.1 Diffusion cylindrical steel diameter by Washer Filtrate tank 5.8 m high Tank (110 m 3) 2 n d Stage Vertical 4.9 m 2.9 58 0.24 8.1 6.8 Diffusion cylindrical steel diameter by Washer Filtrate tank 3.8 m high Tank (72 m 3) B low Tank Vertical cylindrical steel tank, partial cone top/bottom 14.4 m diameter by 34 m high (4600 m 3) 10 1000 3.6 12 83 Decker Washer Drum washer n/a n/a n/a n/a n/a n/a Hood Decker Washer Rectangular 14.3 m by 3.35 390 0.41 87 4.5 Filtrate Tank concrete pit 8.2 m by 4.7 m high (550 m 3) 141 Chapter 6: Results and Discussion A simplified process flow diagram (Figure 5.3) and process description of the brown stock washing area of the mi l l was provided in Section 5.3.1. Besides the mechanical design of the equipment, the process operation can also influence the potential equilibrium that may establish between the liquid and vapour phases in this equipment. The 2 n d stage diffusion washer filtrate tank is located on top of the 1 s t stage tank with both sharing a common overflow line, which is sealed where it enters the sewer. The 1 s t stage tank is vented via this overflow line to the 2 n d stage filtrate tank, with the vent from the 2 n d stage tank collected to the D N C G system. Therefore, the vent vapour sampled on the "common" vent for these tanks wi l l be in contact, and potential equilibrium with, the contents of the 2 n d stage tank. There was also the potential added complication caused by the diffusion washer operating on a 60 to 90 second cycle. During the last 10 seconds of this cycle, the extraction liquor flow from the diffusion washer to the filtrate tanks is stopped, causing a rapid drop in the liquor level in both of these tanks (the flow of liquor pumped from the tanks remain constant). The operating level for the l s l stage filtrate tank is typically controlled at 75% (4.4 m), while the 2 n d stage filtrate tank normally operates at 100% level (2.9 m) with a portion of the filtrate overflowing to the 1 s t stage tank. During the part of the wash cycle when the liquor flow to these tanks is stopped, the 1 s t and 2 n d stage filtrate tank levels drop to about 2.3 m and 1.3 m, respectively. The vapour flow in the vent lines on these two tanks reverses during this period (this effect was confirmed during sampling of the vent vapour). By watching the liquor valves open and close, it was possible to collect the vent vapour samples during the period of the cycle that liquor was flowing to the tank. A complicating factor for the blow tank is the potential for tramp air ingress through poorly sealed overflow openings. These overflows are large rectangular openings near the top of the tank that are "sealed" with plastic sheets attached at the top and weighted at the bottom; openings were quite visible around the edges of these sheets. The D N C G system draws a slight vacuum on the blow tank, which in turn can draw tramp air in through these overflow openings. The D N C G vent nozzle is located on top of the blow tank, very near the overflow openings, so it is likely that most of the tramp air is drawn into the tank and right back out via the D N C G system. The Howe Sound decker washer hood is shown in Figure 6.22. A fibreglass hood encloses the drum washer, with the D N C G vent stack just visible exiting on top of the hood at the far end. 142 Chapter 6: Results and Discussion Figure 6.22: Decker washer at the Howe Sound pulp mi l l For the decker washer, use of a single wash liquor source simplifies the analysis, so there are only two liquid streams in contact with the vapour: the liquor in the stock feed and the wash liquor used for washing and for repulping. The wash liquor is filtrate from the post-oxygen delignification system washer ( 0 2 filtrate). Brown stock is fed to the inlet vat of the washer as a slurry at a consistency of about 1% (similar to runny porridge). The stock is pulled onto the turning drum (which is half submerged in the feed brown stock), and the liquor is drawn from the stock through the drum screens by the vacuum inside the drum. This vacuum is formed by the filtrate liquor dropping down the seal leg to the filtrate tank. A series of shower nozzles, with the first set located immediately after the stock has been pulled onto the drum, spray wash liquor onto the brown stock mat on the turning drum. The mat of stock, at a consistency of about 12% (similar to wet cardboard), is scraped off the drum screens into the outlet vat (repulper) of the washer. The brown stock slurry surface area at the inlet vat is 15.3 m 2 based on a width of 1.3 m and a length of 11.8 m. The rotating washer drum is 4.7 m diameter by 11.5 m long with 85 m 2 o f exposed brown stock area. The brown stock surface area at the outlet vat is 14.2 m 2 based on a width of 1.2 m and a length of 11.8 m. The inlet vat makes up only about 13% of the total exposed 143 Chapter 6: Results and Discussion surface area, whereas the washer drum and outlet vat make up about 87%. The 0 2 filtrate is sprayed through the vapour space in the washer hood, and onto the surface of the brown stock mat on the washer drum. The liquor in the feed stock that is pulled out of the inlet vat and onto the surface of the drum is almost immediately displaced with 0 2 filtrate. Because 0 2 filtrate is sprayed through the vapour space and is in contact with the vapour for a large fraction of the exposed liquid surface, it is expected to have the greatest impact on the vapour concentration. Based on this, the wash liquid was used as the liquid phase for modelling phase equilibrium. 6.4.2 Time to E q u i l i b r i u m for D M S in Black L i q u o r Process equipment in a mi l l may have limited liquid phase residence times, as low as a few minutes in some cases (Table 6.23). To determine i f typical residence times are sufficient time for equilibrium to be established, lab testing was conducted to determine the time required for D M S in black liquor to come to equilibrium with its vapour headspace. The equilibrium tests were done using black liquor samples collected September 22, 2006. Vapour to liquid volume ratios and intensity of mixing were varied to determine their effects. Some of these tests were replicated using both 1 s t stage diffusion washer filtrate and decker washer filtrate to determine i f the total dissolved solids concentration in the liquor was a factor. The operating temperature of the brown stock washing process varies little during normal operation, typically only within a few degrees of 80°C, so all testing was done at this temperature. Two vapour-liquid volume ratios and two agitation levels were tested. The first vapour-liquid volume ratio of 5 m L liquid in a 24 m L vial (ratio of 3.8) roughly represents typical blow tank conditions. The second vapour-liquid volume ratio of 20 mL liquid in a 24 m L vial (ratio of 0.2) roughly represents typical filtrate tank conditions. The agitation levels were simulated by using a tumbler, slowly rotating at about 15 rpm to provide "light" agitation, while a variable speed shaker was used to provide a "vigorous" level of mixing. It is difficult to quantify the level of agitation in equipment, but the blow tank would likely be on the light end, while the filtrate tanks would be somewhere between light and vigorous. For all tests, a liquor sample was preheated to 80°C and injected into a preheated 24 m L vial, with the vial placed in the agitator located in the oven. A headspace sample was withdrawn and 144 Chapter 6: Results and Discussion injected into the G C about every 3 minutes (the time required to complete a G C run). Theoretically equilibrium is never reached; instead, the practical equilibrium concentration is reached when the change in concentration over time can no longer be detected with the available measuring device. For this test, practical equilibrium was judged to have been reached when there was less than 5% change (at which point the change became indistinguishable from G C signal noise) in three consecutive injections. The D M S headspace concentration data for each run have been normalized to the largest value for each run, with the results shown in Figure 6.23. 1.00 O CD CD o ^ I 0.80 0.90 8 g ro t £ o 0.70 TD - S CD CD X C/) 0.60 Q 0.50 - • -Decker , Vigorous, 5 ml - * - Decker, Light, 5 ml - A T - Diffusion, Light, 5 ml - • - Decker, Vigorous, 20 ml - * - Diffusion, Vigorous, 20 ml -a-Diffusion, Light, 20 ml 0 2 4 6 8 10 12 14 16 18 20 22 Time (minutes) Figure 6.23: Time required for D M S phase equilibrium to be established in a black liquor sample held at 80°C Equilibrium was established more rapidly as the vapour-liquid volume ratio and the level of agitation increased. There was no discernable difference in time-to-equilibrium between the decker filtrate and the diffusion filtrate. For all liquor samples, equilibrium was established in less than 10 minutes for the 5 m L sample at vigorous agitation, while 10 to 15 minutes were required for the 5 m L at light agitation and the 20 m L at vigorous agitation, and just over 15 minutes for the 20 mL sample at light agitation. Based on the laboratory results, for the blow tank with a high vapour-liquid volume ratio and a low level of agitation, equilibrium should be established in 15 minutes or less, and with 83 minutes 145 Chapter 6: Results and Discussion liquid phase residence time (Table 6.23), there should be more than adequate time. For the filtrate tanks, with a low vapour-liquid volume ratio and a higher level of agitation, equilibrium should also be established in under 15 minutes. With liquid phase residence times ranging between 4.5 and 8.1 minutes (Table 6.23), there is insufficient time for equilibrium to be established. Referring to Figure 6.22, the measured vent vapour concentration for the filtrate tanks would be expected to be in the range of 80 to 95% of the theoretical equilibrium value. 6.5 Mode l l ing of Howe Sound Equipment The V L E model, based on the e N R T L parameters determined in Section 6.3, was applied to the mil l sampling and testing data presented in Tables 6.11 to 6.22. Similar to the analysis of methanol data from literature sources described in Section 6.1, the mi l l testing data were analysed to determine i f process conditions in the equipment were at or near vapour-liquid equilibrium. If the liquid and vapour were in equilibrium, with D M S defined as i , then the vapour mole fraction, y j 5 could be predicted from the liquid mole fraction, x j 5 using Equation (3.7): v . P ^ x . P , 5 3 ' (3.7) The total pressure of the system, P, was atmospheric for the equipment tested. The D M S vapour pressure, P s a t , was determined for the equipment operating temperature using the extended Antoine Equation (3.21) using the parameters in Table 3.9. The activity coefficient for D M S in black liquor was determined using the e N R T L Equation (3.11) using the parameters determined from phase equilibria testing given in Tables 6.6 and 6.8. 146 Chapter 6: Results and Discussion 6.5.1 Mode l Sensitivity Analysis A sensitivity analysis was performed to see how the model output changed with the input parameters. The input parameters were varied over their range of uncertainty and the change in the model response was calculated. Uncertainty may be introduced by possible errors in measurement (instrument sensitivity or measuring technique), absence of information, or poor or partial understanding of the driving forces and mechanisms. The model output, i.e., the predicted D M S vapour phase concentration, y j 5 was calculated using Equation (3.7) from the measured variables: D M S concentration, x i 5 and pressure, P. The model output is also dependent on two other variables, temperature, T, and sodium concentration, x c a , with temperature used to calculate D M S vapour pressure, Pj S a l , and both used to calculate the activity coefficient, YJ- For temperature, the process was operating near 80°C, with field transmitters providing readings typically within a degree of the measured temperature. To be conservative, the sensitivity range for temperature was set at 353 ± 2 K , i.e., ± 0.6%. For pressure, the process was operating near atmospheric pressure with the D N C G system drawing a slight vacuum. The vacuum at the sources could not exceed the setting of the vacuum breakers installed to protect the tanks, with these set at 0.5 kPag. The sensitivity range for pressure was set at 101.3 ± 1 kPa, i.e., ± 1%. For determination of D M S concentration, x i 5 for the black liquor samples collected, the range of confidence must take into account sample degradation issues (see Section 3.4), syringe reproducibility (± 1% when using a Chaney adaptor), and calibration curve accuracy (Figure 5.6, R 2 - 0.99). The sensitivity range for this parameter was set at ± 5%. For the total dissolved solids level, the moisture analyser provided by Howe Sound mi l l had an accuracy level of 0.1 wt% and the lowest measured value was 2.0 wt%, so the sensitivity range was set at ± 5%. The results of the sensitivity analysis are shown in Figure 6.24. 147 Chapter 6: Results and Discussion g, 96% c CD O 94% 0.96 0.98 1 1.02 1.04 Parameter Multiplier Figure 6.24: Sensitivity analysis for prediction of D M S vapour phase concentration Temperature has the greatest effect, but the range of uncertainty is small, introducing a potential error of up to 5%. The liquid phase concentration has a linear effect; e.g., a 5% error w i l l result in a 5% error in the model prediction. It can be seen that confidence in the model results are in the range of ± 5% based on all input variables, although these may be cumulative. A s discussed in Section 6.5.2, the error in the predicted values were greater than can be explained with the factors studied in the sensitivity analysis. To explain this, two other parameters were considered when modelling the Howe Sound equipment: insufficient liquid residence time for equilibrium to be established and dilution of the vent vapour stream with tramp air drawn in through openings in the equipment. 6.5.2 Equipment Mode l l ing Results for D M S Comparisons between measured and predicted D M S concentrations in the vent vapour stream were made for the decker washer filtrate tank, decker washer hood, blow tank and diffusion washer 148 Chapter 6: Results and Discussion filtrate tanks. The diffusion washer N C G vent line was closed off, so no comparisons could be made for this source. For the testing days where the total dissolved solids were not measured, the average for that particular liquor sample point from all other testing days was used. In Figure 6.25, the measured D M S vent vapour concentrations from the decker washer filtrate tank are compared to the predicted equilibrium vent concentrations based on the filtrate concentration. 5000 > E Q. Q. C > C CO 4000 3000 2000 Q 1000 Measured • Predicted C J ) C T ) C J > 0 } C J ) C J > T — T — T — 1 Date Measured Figure 6.25: Measured versus predicted D M S concentrations for decker washer filtrate tank N C G vent There appears to be a good response by the model to the changing concentration of D M S in the liquid phase. The model fit for the decker washer filtrate tank and other equipment is quantified in Section 6.5.3. In Figure 6.26, the measured D M S vent vapour concentration from the diffusion washer filtrate tanks are compared to the predicted equilibrium vent concentrations based on the filtrate concentration in the 2 n d stage tank, again with good results. 1 4 9 Chapter 6: Results and Discussion 20000 <J> T - T - T - p«l ° ^ ^ ^ ^ *~ 3> c i to ^- o <M «M o i CO o Date Measured Figure 6.26: Measured versus predicted D M S concentrations for 2 n d stage diffusion washer N C G vent Comparison of measured data to predicted values for the blow tank vent was not straightforward. The diffusion washer is mounted on top of the blow tank and the brown stock from the washer drops down through a stock chute, into the blow tank. The brown stock, at a consistency of about 8% (clumps of wet fibre saturated with black liquor), drops from the stock chute, lands and settles on top of the stock pile in the blow tank. The stock consistency is too high for it to act like a liquid, so the stock moves down the blow tank in a plug flow type motion until it is removed from the bottom of the tank. The stock is diluted right at the bottom of the blow tank with decker washer filtrate so that it can be pumped to the knotters and screens and onto the decker washer. The stock going to the blow tank was sampled as it dropped down the stock chute, but the stock in the blow tank itself could not be sampled because the stock is diluted ahead of any potential sample point. The vapour from the blow tank would theoretically be in equilibrium with the stock from the blow tank, but since this was not sampled, the D M S concentration of the liquor in the stock had to be determined indirectly. The D M S concentration of the stock leaving the blow tank (prior to dilution) was determined from a mass balance calculation conducted around this tank. The D M S concentrations in the stock 150 Chapter 6: Results and Discussion to the blow tank, and in the vapour from the blow tank, were measured and the mass flow of each stream was calculated. From this, the mass flow of D M S in the stock from the blow tank could be calculated along with its concentration. In Figure 6.27, the measured D M S vent vapour concentrations from the blow tank are compared to the predicted equilibrium vent concentration based on the calculated D M S concentration in the stock leaving the blow tank. The comparison is not favourable; the predicted values are about three times greater than the measured values (that were determined by the mass balance around the blow tank). A s discussed in Section 6.5.3, the measured values were thought to be low due to dilution from tramp air ingress. > E Q. Q. 4000 c CD c 2000 CO Measured • Predicted o C N C M C M C M C O C M C M CO Date Measured Figure 6.27: Measured versus predicted D M S concentration for blow tank N C G vent In Figure 6.28, the measured D M S vapour concentrations in the vent from the decker washer hood are compared to the predicted equilibrium vent concentrations based on the inlet wash liquor ( 0 2 filtrate) D M S concentration. For the latter four sets of data there appears to be a favourable comparison; although it is unclear why the first two predict so poorly. 151 Chapter 6: Results and Discussion 300 s 250 E a. 200 -4—* | 150 w 1 0 0 Q 50 0 Date Measured j F i g u r e 6 . 2 8 : Measured versus predicted D M S concentration for decker washer hood N C G vent 6 . 5 . 3 E q u i p m e n t M o d e l l i n g R e s u l t s f o r o t h e r T R S C o m p o u n d s Hydrogen sulphide and methyl mercaptan were below detection limits for all liquor samples. Other than D M S , only D M D S was measured in high enough concentrations for modelling purposes. In Figure 6.29, the measured D M D S vent vapour concentrations from the decker washer filtrate tank are compared to the predicted equilibrium vent concentrations based on the filtrate concentration. 152 Chapter 6: Results and Discussion 1000 800 > E Q. Q. ^ 600 CD > .£ 400 CO Q i 200 Measured • Predicted o <B o C M o CM C M C M C O C M C M O CO C M o Date Measured F i g u r e 6 . 2 9 : Measured versus predicted D M D S concentrations for decker washer filtrate tank N C G vent It was very difficult to avoid mixing of air with the liquor sample when it was collected, because the liquor line was under high pressure and it sprayed from the sample valve. A s discussed in Section 3.4, significant oxidation of the methyl mercaptide to D M D S occurs after only a few minutes (Reaction 2.9). D M D S concentrations are very low, ranging from 0.17 to 1.3 ppm in the liquor samples (Tables 6.11 to 6.22); whereas, the concentration of the mercaptide is expected to be in the range of double that of D M S (Figure 2.1 and 2.2), i.e., up to about 100 ppm (mol). Therefore, oxidation of only a very minor fraction of the mercaptide would have a large impact on the D M D S concentration. This oxidation effect is suspected of causing the over-prediction seen in Figure 6.29. A false high measured concentration of the D M D S in the liquor w i l l lead to a false high D M D S prediction in the vapour phase. Without this complication, the quality of the prediction for the decker washer filtrate tank for D M D S would be expected to be similar to that for D M S (Figure 6.25). If this effect was confirmed, it would provide further reason to choose D M S as a surrogate for modelling of TRS emissions. 153 Chapter 6: Results and Discussion 6.5.4 Equipment Mode l l ing Analysis Quantification of the accuracy of the predicted vent vapour concentration for the Howe Sound equipment was done using the root mean square error ( R M S E ) . This is defined as the square root of the sum of the square of the error, where the error is the amount by which the predicted value differs from the measured quantity. The difference occurs because of randomness or because the predicted value does not account for information that could produce a more accurate estimate. Table 6.24: Root mean square error ( R M S E ) of V L E model vent vapour concentration prediction Equipment Measured vapour Predicted vapour R M S E concentration (ppm (mol)) concentration (ppm (mol)) (ppm (mol)) range average range average Decker Washer 194 to 1611 247 to 1977 412 Filtrate Tank 3789 4202 Diffusion Washer 8976 to 13707 13473 to 14651 3552 Filtrate Tanks 17336 15485 Blow Tank 692 to 1295 1031 2532 to 3946 3215 2261 Decker Washer 52 to 110 67 to 162 95 Hood 204 272 For all equipment, the error appears to be a systematic over-prediction, rather than random error. The error is greater than what could be explained by the sensitivity analysis, i.e., there are unaccounted factors involved. These factors are suspected to be insufficient liquid phase residence time in the tank for equilibrium to be established, and/or dilution of the vent vapour stream by tramp air entering the equipment through openings such as poorly sealed overflows. The filtrate tanks are sealed against tramp air ingress, so the error was likely introduced due to insufficient liquid phase residence time in the tank. For the decker washer filtrate tank, this is supported by the data in Figure 6.23, which predict that the vapour should be in the range of 80 to 90% of the equilibrium value after 4.5 minutes; the average of the measured values was 82% of the 154 Chapter 6: Results and Discussion average of the predicted values. The 2 n d stage diffusion washer filtrate tank data are also supported by the data in Figure 6.23, with this predicting that the vapour should be in the range of 85 to 95% of the equilibrium value after 6.8 minutes; the average of the measured values for these tanks was 94% of the average of the predicted values. There was more scatter in this data compared to the decker washer filtrate tank, which is likely attributable to the cyclical operation of the diffusion washer, and venting of the 1 s t stage filtrate tank via the 2 n d stage tank. The blow tank has more than sufficient time to reach equilibrium. The error was likely introduced by tramp air ingress through poorly sealed overflow openings. The averaged measured values were 32% of the average predicted value, indicating that almost 70% of the vent vapour flow from this tank was drawn in as tramp air. The tramp air is likely drawn into the tank and right back out via the D N C G system; therefore, it has insufficient contact with the stock in the tank for equilibrium to be established. For the decker washer hood, tests done on the latter four days have measured results which agree well with the predicted results (Figure 6.28), but it is unclear why there was an over-prediction for the first two testing days. It appears that random error is an issue, but is not clear with the limited data for this source i f there is a systematic error, such as tramp air ingress. To further analyse the data for systematic error, the D M S vapour phase predictions were adjusted by a set percentage of the predicted equilibrium value. This percentage was determined by minimizing the R M S E of the corrected value. Table 6.25: V L E model correction factor for systematic error Equipment Correction Factor for Systematic Error (%) R M S E (ppm (mol)) Decker Washer Filtrate Tank 84 168 Diffusion Washer Filtrate Tanks 93 3408 Blow Tank 31 300 Decker Washer Hood 61 66 155 Chapter 6: Results and Discussion The predicted values based on the V L E model and predicted values adjusted to account for systematic error, are compared to measured data on scatter plots in Figures 6.30 to 6.33. Measured (ppmv) Figure 6.30: Comparison between decker washer filtrate tank measured D M S vapour concentration and predicted equilibrium value using the V L E model, and predicted value corrected for systematic error 156 Chapter 6: Results and Discussion Measured (ppmv) Figure 6.31: Comparison between diffusion washer filtrate tanks measured D M S vapour concentration and predicted equilibrium value using the V L E model, and predicted value corrected for systematic error 157 Chapter 6: Results and Discussion Measu red (ppmv) Figure 6.32: Comparison between blow tank measured D M S vapour concentration and predicted equilibrium value using the V L E model, and predicted value corrected for systematic error 158 Chapter 6: Results and Discussion 300 250 > E o 200 i J Q . T3 Q) 150 O TJ CD 100 CL 50 0 Q Model Model (corrected) 50 100 150 200 250 300 Measured (ppmv) Figure 6.33: Comparison between decker washer hood measured D M S vapour concentration and predicted equilibrium value using the V L E model, and predicted value corrected for systematic error For the decker washer filtrate tank and the blow tank, the scatter plots appear to confirm a systematic error, insufficient residence time for the former and tramp air ingress for the latter. For equipment with insufficient residence time, the V L E model could be extended to include gas-liquid mass transfer limitations, although more extensive testing would be required to fit such a model. For modelling of the Howe Sound process using a V L E model, which is discussed in Section 6.7, a liquor by-pass technique was used to account for the lack of residence time and a vapour by-pass technique was used to account for tramp air effects. 159 Chapter 6: Results and Discussion 6.6 M i l l T e s t i n g D a t a C o m p a r e d t o E m i s s i o n F a c t o r s Emission factors are provided by the U.S . E P A for estimating T R S emissions from chemical wood pulping facilities. The relevant factors are discussed in Section 3.3, with details provided in Table 3.10. Two sets of emission factors are provided, one for "digester relief and blow tank," and a second for a "brown stock washer." These factors were last updated in September 1990 and are based on an older process configuration consisting of batch digesters blowing directly to a blow tank, followed by a series of vacuum drum washers used for brown stock washing. In this process, the first emission factor refers to the C N C G sources. The equivalent sources in the continuous digester process used at Howe Sound are the extraction liquor flash tanks, which were not tested as part of this work. The second factor refers to the D N C G sources, which are equivalent to the sources tested in this work. The Howe Sound process consists of a continuous digester followed by a two stage diffusion washer, blow tank, knotters, screens, and decker washer. The blow tank in this position can not be considered equivalent to the blow tank referred to in the emissions factors provided for a batch mi l l . For this comparison, the blow tank has been included as part of the D N C G brown stock washing emissions. To allow comparison to emission factors, the test data from Tables 6.11 to 6.22 were used to calculate the emissions of each T R S compound on a mass basis; this information is included in these tables. Emission factors are given for H 2 S , and for the organic T R S compounds, i.e., for the sum of M M , D M S and D M D S . The emission factors for the brown stock washing process are 0.01 and 0.2 kg sulphur per air dry tonne pulp ( A D T P ) for H 2 S and the organic T R S compounds, respectively. Figures 6.34 and 6.35 compare the measured emissions and those determined using this emission factor. Only the testing days when all major sources were tested are shown, i.e., November 23, 24, 30 and December 1, 2005 (Tables 6.18 to 6.21). 160 Chapter 6: Results and Discussion 0.2 5 5 0.15 "D CD i _ CO cc (D 0.1 co 0.05 C N X CO CM C M O C O Date Measured C M o 20 15 33 CO 10 % UL E o * % C N X Figure 6 .34: Total measured H 2 S emissions (left axis) compared to predicted emissions using emission factor (right axis) for brown stock washing process (kg of sulphur per day) The measured H 2 S emissions are two to three orders of magnitude lower than those expected based on emission factors (Figure 6.34). The undissociated H 2 S concentration in the liquor, the fraction that contributes to vapour pressure and thus to emissions, is strongly dependent on p H . The undissociated fraction is typically below detection limits above a pH of 9 (Figure 3.2). The lowest measured pH during the Howe Sound testing was 11.4; therefore, the expectation agrees with the result, i.e., the H 2 S in the vapour phase was negligible. To keep dissolved alkali lignin from precipitating, the brown stock process must be operated at pH higher than 11; thus, it is questionable i f an H 2 S emission factor for this process is necessary. 161 Chapter 6: Results and Discussion C O C M Measured G Emiss ion Factor C M o C O 500 400 300 200 100 0 5 to CO L _ o o CO u_ E o v H— CO or CO CO Date Measured Figure 6 .35: Total measured organic T R S ( M M , D M S and D M D S ) emissions (left axis) compared to predicted emissions using emissions factor (right axis) for brown stock washing process (kg of sulphur per day) The organic T R S emission factor provides a better estimate of emissions, with measured values ranging from 215 to 383 kg/day as sulphur (average 291) with an average R M S E of 84 for the predicted values. A n emission factor is not given for D M S alone, so a direct comparison to the D M S mi l l data is not possible. Summing the total organic T R S emissions from all brown stock washing sources, 90.7% of the sulphur emissions originated from D M S . For comparison purposes, the emission factor was adjusted by this amount; i.e, the brown stock washing D M S emission factor used was 0.2 * 0.907 = 0.181 kg sulphur per A D T P . It is difficult to directly compare how well these emission factors predict D M S emissions when compared to a V L E model. A n emission factor is given for the entire brown stock washing line, not for each piece of washing equipment. To allow comparison, the emission factor was divided among each of the washing equipment process elements. For Howe Sound, the potential major emitters include the diffusion washer filtrate tank, diffusion washer, blow tank, screen feed 162 Chapter 6: Results and Discussion tank, screen dilution tank, decker washer filtrate tank and decker washer hood. Of these, the diffusion washer was valved off and the screen feed and dilution tanks were found to have negligible emissions. The DMS emission factor was divided up between the rest of the sources based on the fraction of measure emissions from each source. For example, based on measured data from all testing days, 23.9% of the DMS emissions originated from the decker washer filtrate tank. The prorated DMS emissions factors for each source are listed in Table 6.26. Table 6.26: Brown stock washing DMS emission factors Sample Point Process Element % of total DMS (avg.) DMS Emission Factor kg S / ADTP VS1 Diffusion Washer Filtrate Tanks 29.6 0.054 VS2 Diffusion Washer (valve closed) negligible -VS3 Blow Tank 34.2 0.062 VS4 Screen Feed Tank negligible -VS5 Screen Dilution Tank negligible -VS6 Decker Washer Filtrate Tank 23.9 0.043 VS7 Decker Washer Hood 12.3 0.022 Brown Stock Washing Area (total) 100 0.181 For the decker washer filtrate tank, Figure 6.36 show a comparison between DMS emissions based on measured data, predicted emissions based on V L E correlations and a prorated emission factor. 163 Chapter 6: Results and Discussion 1 5 0 5 1 0 0 oo CD GO 5 0 Measured D Predicted Emiss ion Factor n T - T - C N J C N C N I C M f N C ^ t ^ f O o Date Measured Figure 6.36: Decker filtrate tank emissions based on measured data and predicted using V L E correlations and emission factors Figure 6.36 illustrates the advantage of the V L E model over emission factors for predicting emissions of volatile contaminants. The measured values range from 4.9 to 114 kg/day as sulphur (average 42.6) with a R M S E of 10 kg/day for the predicted values using V L E and a R M S E of 31 kg/day for the predicted values using the emission factor. A n emission factor is independent of D M S concentration in the process. It is based solely on production levels, thus it wi l l tend to over-predict when the D M S concentration in the process is low, and under-predict when the D M S concentration is high. This is illustrated in Figure 6.37, a scatter plot of predicted values based on V L E and on a prorated emission factor. 164 Chapter 6: Results and Discussion o r • i ' i . i 0 20 40 60 80 100 120 M e a s u r e d (kg/day sulphur) Figure 6.37: Comparison between decker washer filtrate tank measured D M S vapour concentration and predicted equilibrium value using the V L E model, and predicted value using emission factor Emission factors may provide a good averaged emissions estimate over time, but being independent of D M S concentrations, they can not be used for modelling changes to the process. 6.7 Howe Sound M i l l Mode l l ing The final section of this work provides an example on how this modelling technique can be applied to an industrial situation. According to Gerry Pageau of Howe Sound, high S 0 2 emissions from the power boiler have been an ongoing concern at the mi l l . The mil l has a D N C G system and the collected gases are incinerated in the power boiler. These gases were suspected to be the main contributor of sulphur to the combustion process that results in the S 0 2 emissions. Some means of reducing the sulphur loading in the D N C G would contribute towards meeting power boiler S 0 2 165 Chapter 6: Results and Discussion emissions requirements. One of reasons that Howe Sound agreed to assist in this research was that they were hoping to gain some insight into where the sulphur in the D N C G originated and possibly find solutions on how to reduce the sulphur loading to the power boiler. Based on mil l testing data from the days when all major D N C G sources were tested (Tables 6.18 to 6.22), the average sum of the sulphur loading from all D N C G sources was 512 kg/day. The brown stock washing area and the digester chip bin were suspected to be the main contributors to sulphur in the D N C G . The brown stock area average sulphur loading was 291 kg/day. These sources include about 57% of the total sulphur loading in about 82% of the total D N C G flow. The chip bin average sulphur loading was 204 kg/day. This source includes about 40% of the total sulphur loading in about 13% of the total D N C G flow. The recovery area, which includes the heavy black liquor ( H B L ) , weak black liquor ( W B L ) and contaminated condensate tanks was confirmed as a low sulphur contributor at 17 kg/day. This tankage includes about 3% of the total sulphur loading in about 5% of the total D N C G flow. The combined D N C G flow to the boiler was also tested yielding a total sulphur loading of 566 kg/day. Considering the variability in the process and the potential errors in measurement, there is reasonably good agreement, within 10%, between the total sulphur loading based on summing of the sources and the total loading measured in the combined D N C G . Contaminated steam from the extraction liquor flash tanks is used for pre-steaming in the chip bin, and when there is a low chip bin level, this TRS laden steam breaks through the chip pile and is collected by the D N C G system. The sulphur loading in the D N C G from the chip bin is highly variable due to the sporadic nature of this contaminated steam break-through. The frequency of the steam break-through can be minimized by controlling the chip pile at a higher level in the chip bin. Other means can be used to reduce the sulphur loading from this source, including switching to clean steam (although at an increased operating cost), or switching to clean steam when the bin level is low. Because brown stock washing is the largest sulphur contributor to the D N C G , sampling and testing, and subsequent modelling, was focussed on this area. Where changes to the process are being considered or improvements are desired, it may be desirable to "bench-mark" the existing operation. A base-case heat and mass balance model can be constructed which reflects current 166 Chapter 6: Results and Discussion operation. Once calibrated using the mi l l testing and operating data, the base-case model can be modified to evaluate process changes and their effect on D M S emissions. 6.7.1 Base-Case Balance of the Howe Sound Process A base-case heat and mass balance was constructed for the Howe Sound brown stock washing system, using C A D S i m Plus. To account for vapour phase non-idealities, such as tramp air entering the system through the poorly sealed overflows on the blow tank, a modelling technique was incorporated where a percentage of the vapour flow was by-passed around the V L E module. This by-pass was used to simulate tramp air that is drawn into the tank and promptly out through the vent stack, with essentially no contact with the brown stock. Where there was suspected to be insufficient liquid phase residence time to establish equilibrium, a percentage of the liquid flow was by-passed around the V L E module. The base-case balance was based on mi l l testing data averaged from the days when all major brown stock washing area sources were tested (Tables 6.18 to 6.21). This was the same subset of testing days that were used to do the emission factor comparisons in the previous section. These days were chosen because the mill was operating at full capacity, the process was stable, a full set of samples was collected and there was generally the highest confidence in the experimental data collected (the mi l l sampling and testing technique had been refined over the previous testing days). The vapour and liquid sample test results are shown in Tables 6.27 and 6.28, respectively. 167 Chapter 6: Results and Discussion Table 6.27: Vapour sample testing data for November 23, 24, 30, and December 1, 2005 Location Parameter Units Average Range VS1 Diffusion Flow m 3 /min 3.1 2.7 to 3.2 Washer Temperature °C 64 63 to 65 Filtrate Relative Humidity % 100 100 to 100 Tanks D M S Concentration ppm (mol) 12075 8976 to 16299 D M S flow kg /day as S 61.1 48.0 to 73.1 VS3 Blow Tank Flow m 3 /min 55.7 54.5 to 56.9 Temperature °C 71 70 to 72 Relative Humidity % 99 96 to 100 D M S Concentration ppm (mol) 1051 692 to 1295 D M S flow kg /day as S 95.6 64.5 to 120.1 V S 6 Decker Flow nrVmin 15.3 11.7 to 18.4 Washer Temperature °C 74.5 72 to 77 Filtrate Tank Relative Humidity % 100 100 to 100 D M S Concentration ppm (mol) 2523 1590 to 3789 D M S flow kg /day as S 66.2 30.0 to 113.9 V S 7 Decker Flow nrVmin 186.6 186.6 to 186.6 Washer Temperature °C 64.3 63 to 65 Hood Relative Humidity % 97 91 to 100 D M S Concentration ppm (mol) 135.5 56.8 to 204 D M S flow kg /day as S 42.1 17.7 to 63.4 168 Chapter 6: Results and Discussion Table 6.28: Liquid sample testing data for November 23, 24, 30, and December 1, 2005 Location Parameter Units Average Range L S I Stock from p H - 11.7 11.6 to 11.7 Digester Temperature °C 80.8 80 to 82 Total Dissolved Solids % 13.6 13.4 to 13.8 D M S Concentration ppm (mol) 50.3 1/12/07 only LS2 1 s t Stage p H - 11.6 11.6 to 11.6 Diffusion Temperature °C 79.5 79 to 80 Washer Total Dissolved Solids % 11.8 10.8 to 12.7 Filtrate D M S Concentration ppm (mol) 37.4 31.3 to 43.5 LS3 2 n d Stage p H - 11.6 11.4 to 11.6 Diffusion Temperature °C 78.5 78 to 79 Washer Total Dissolved Solids % 9.3 8.9 to 9.8 Filtrate D M S Concentration ppm (mol) 21.9 19.7 to 23.5 LS4 Diffusion p H - 11.6 11.5 to 11.7 Washer Temperature °C 77.5 77 to 78 Stock out Total Dissolved Solids % 8.5 8.5 to 8.6 D M S Concentration ppm (mol) 7.82 7.2 to 8.9 LS5 Stock to pH - 11.6 11.4 to 11.7 Decker Temperature °C 75.3 74 to 76 Washer Total Dissolved Solids % 7.3 7.0 to 7.6 D M S Concentration ppm (mol) 5.39 4.64 to 6.97 169 Chapter 6: Results and Discussion Table 6.28 (cont.): Liquid sample testing data for November 23, 24, 30, and December 1, 2005 Location Parameter Units Average Range LS6 Decker p H - 11.6 11.4 to 11.7 Washer Temperature °C 75.3 74 to 76 Filtrate Total Dissolved Solids % 6.8 6.7 to 6.9 D M S Concentration ppm (mol) 5.48 4.09 to 7.88 LS7 Stock from pH - 11.6 11.4 to 11.7 Decker Temperature °C 75.3 74 to 76 Washer Total Dissolved Solids % 4.8 4.5 to 5.1 D M S Concentration ppm (mol) 0.57 0.43 to 0.82 LS8 0 2 Filtrate pH - 11.6 11.4 to 11.7 Temperature °C 80.0 79 to 81 Total Dissolved Solids % 4.3 4.1 to 4.5 D M S Concentration ppm (mol) 0.25 0.13 to 0.39 Relevant operating data for the base-case balance was extracted from the mi l l M O P S historian data (Appendix G). Operating data for the four days were averaged, with the results shown in Table 6.29. 170 Chapter 6: Results and Discussion Table 6.29: M O P S historian data for November 2 3 , 2 4 , 3 0 , and December 1, 2 0 0 5 Location Parameter Units Average Range Digester Chip feed Flow bone dry tonne/day 2 6 7 3 2 5 1 8 to 2 9 0 2 Stock from digester Yie ld (on bone dry wood) % 4 3 -Flow (stock only) bone dry tonne/day 1 1 5 0 1 0 8 3 to 1 2 4 8 Flow (total) L/s 1 2 7 1 2 0 to 1 3 7 Temperature °C 8 0 . 6 7 9 . 6 to 8 2 . 1 Consistency % 9 . 6 3 9 . 2 to 9 . 8 1 s t stage diffusion washer Filtrate flow to L/s 1 6 3 1 4 8 to 1 7 0 Filtrate flow from L/s 1 5 3 1 3 2 to 1 6 1 Tank level % 7 5 7 4 to 7 7 2 n d stage diffusion washer Filtrate flow to L/s 1 3 1 1 2 1 to 1 4 2 Filtrate flow from L/s 1 2 8 1 1 5 to 1 3 8 Tank level % 1 0 2 1 0 1 to 1 0 2 Blow tank Tank level % 2 4 1 8 to 3 5 Knotters Feed consistency % 4 . 1 3 . 8 to 4 . 4 Screens Feed consistency % 1 . 8 5 1 . 6 7 to 1 . 9 8 Decker washer Product flow L/s 1 3 1 1 2 3 to 1 4 3 Product consistency (after dilution in repulper) % 9 . 1 8 . 8 to 9 . 3 Feed wash liquor flow L/s 1 2 1 1 0 1 to 1 4 2 Feed wash liquor temperature °C 7 9 . 7 7 9 . 4 to 8 0 . 0 Tank level % 7 0 6 9 to 7 0 Filtrate temperature ° c 7 5 . 6 74.4 to 7 6 . 4 171 Chapter 6: Results and Discussion The base-case heat and mass balance for the brown stock washing system was set up based on the fibre, water and total dissolved solids (TDS) flow measurements. The data used included flow, temperature and consistency readings from the M O P S data, vent vapour flows from mil l testing and total dissolved solids measurements from collected samples. The variables used to model washer performance were the displacement ratio (DR) and washed pulp consistency, both of which are related to the dilution factor (DF). The washed pulp consistency from the diffusion washer is not measured, but recent manual lab tests conducted by the mil l showed the consistency was averaging about 8%. This is lower than expected and indicates the diffusion washer was operating at lower than expected efficiencies. This was confirmed when an attempt was made to fit the total dissolved solids data to the base-case model; the diffusion washer displacement ratio that best fit the data was 62%. A n atmospheric diffusion washer would normally be expected to operate at a displacement ratio of about 80%. A study was recently completed on decker washer performance by a consultant hired by the mi l l . They found that the decker washer was operating at an average displacement ratio of 75%, which is typical for a vacuum drum washer. The feed and washed stock consistencies were not measured, but this information, based on manual lab tests, was included in the consultant's report. The average feed stock consistency was 1.26%, and the average washed stock consistency was 12.3%; these values are fairly typical for a vacuum drum washer. "Tramp water", also called "cheater flow", was added to balance the inlet and outlet water of the system. This tramp water enters the system at various locations such as water hoses, foam showers or leaking glands on seal water for pumps. For simplicity, the tramp water was added as one stream to the decker washer filtrate tank. The base-case balance is shown in Figure 6.38; the data was extracted from the C A D S i m model. 172 4.0|3.1176.5 0|0|0 77|55.7|76.5 0|0|0 0|0|0 19|15.3|75.3 241|186.6|68.6 DIFFUSION WASHER FILTRATE A — J 2nd STAGE DIFFUSION WASHER FILTRATE TANK 1st STAGE DIFFUSION WASHER FILTRATE TANK 14184|157|10.7|0|78.4 STOCK FROM DIGESTER 120021126113.619.6180.6 « CO L o ro o cn to cn SCREENS KNOTTERS REJECTS 130|7.2|8.4|76 DECKER WASHER 02 DIFFUSER FILTRATE 10448|121|4.3|0|79.7 STOCK TO 02 DELIGNIFICATION ro C O ro oo on cn 9212|109|5.0|12.3|78.5 DECKER WASHER FILTRATE TANK TRAMP WATER 1200|14|0|0|20 Vapour flows: t/d|m 3/min|°C Liquid flows: t/d|l/s|%TDS|%Co|°C Os ?0 a s a. Figure 6.38: Base-case heat and mass balance for the brown stock washing area of the Howe Sound mil l « Co Co S' 3 Chapter 6: Results and Discussion After the base-case balance had been adjusted to best fit the fibre, water and TDS data, a V L E module was added at all of the potential vent points. D M S was added to the mass balance and the base-case balance was adjusted using the vapour and liquid by-passes to best match the measurements from the samples collected. A n example of the by-pass technique, for the blow tank, is shown in Figure 6.39. TRAMP AIR 34.8 m3/min 10.0 n C 12.9 m3/min NCG VENT 54.5 m3/min 72.2 °C 1095 ppmv DMS DIFFUSION STOCK PRODUCT 156 l/s 78.0 "C 7.90 ppmv DMS VAP 101.3 kPag LIQ 156 l / s 77.5 °C 3.52 ppmv DMS STOCK TO KNOTTERS Figure 6.39: Detailed view of a blow tank C A D S i m V L E module heat and mass balance A s discussed in Section 5.3.6, there was less confidence in the measurements for D M S concentrations in the black liquor samples squeezed from stock (LS1, LS4, and LS7). The measured concentrations of these samples were thought to be low due to the sample collection method used. This was particularly true for the stock from the digester (LSI) , where there were added difficulties collecting a sample from a line under high pressure. This appears to be confirmed in the base-case balance where a good fit was achieved with simulation values for L S I , LS4 and L S 7 consistently higher than the measured values. A n overview of the D M S results for the base-case balance taken from Figure 6.38 is shown in Figure 6.40, with the D M S balance shown in kg/day (as sulphur). 174 Chapter 6: Results and Discussion o2 DIFFUSER FILTRATE DECKER WASHER FILTRATE TANK F i g u r e 6 . 4 0 : Base-case D M S mass balance for the brown stock washing system of the Howe Sound mil l ( D M S in kg/day as sulphur) For the base-case operating conditions, a large fraction of the D M S was washed from the • digester brown stock in the diffusion washer, with 75% of the D M S exiting the system with the diffusion washer filtrate. Only about 2% of the D M S passed through to the 0 2 delignification system with the washed stock. 23% of the total D M S entering the brown stock washing system was flashed off with the vent gases and collected in the D N C G system. 6 .7 .2 P r e d i c t e d E f f e c t o n D M S E m i s s i o n s f r o m P r o c e s s C h a n g e s Various operational and equipment changes to the brown stock washing process were investigated to determine their effect on overall D M S emissions. The base-case model was modified to simulate these changes and to predict their effect on D M S emissions. The changes investigated are given in Table 6.30 with descriptions and analysis following. 175 Chapter 6: Results and Discussion Table 6.30: Predicted effect on Howe Sound brown stock washing D M S emissions as a result of operational or equipment changes Changes made to base-case heat and mass balance D M S emissions (kg/day as sulphur) Diffusion Filtrate Tanks Blow Tank Decker Washer Hood Decker Washer Filtrate Tank Total Change from . base-case (%) Base-case 61 96 42 66 265 0% Worst-case (by-passes set to zero) 52 227 13 24 316 19% Emission factor 69 79 29 55 232 -12% Decrease temperature of stock from digester by 10°C 56 85 45 68 254 -4.2% Decrease temperature of decker wash water ( 0 2 filtrate) by 10°C 58 87 45 66 256 -3.4% Reduce D N C G flow from blow tank by 50% (seal over-flow openings) 63 57 51 79 250 -5.7% Reduce D M S in stock from digester by 10% 55 86 38 60 239 -10% Increase diffusion washer displacement ratio from 62% to 75% 62 52 27 40 181 -33% Increase decker washer displacement ratio from 75% to 80% 61 96 43 67 267 0.7% Increase diffusion washer washed stock consistency from 8% to 10% 68 88 34 52 242 -8.7% Increase decker washer washed stock consistency from 12.3% to 14% 55 87 39 60 241 -9.1% Sparge diffusion washer filtrate tank D N C G in decker washer filtrate tank 0 101 48 106 255 -3.7% Increase temperature of stock from digester to 105°C (hot blow) 72 129 36 61 298 12% Stock by-pass of diffusion washer to blow tank on hot blow of 105"C 11 989 24 52 1076 306% 176 Chapter 6: Results and Discussion For process modelling, when designing a new mil l or for emissions reporting, it may be desirable to ignore the residence time and tramp air effects so that the model reflects a "worst-case" scenario for total emissions. If the V L E model is used without accounting for these effects, it w i l l over-predict D M S concentrations in the vent vapour. The worst-case scenario was determined by setting all V L E module by-passes to zero. This balance may be preferable for design or environmental reporting purposes since D M S emissions are over-predicted by about 20%, therefore providing a contingency factor. The base-case model was also compared to estimated emissions based on an emission factor (Table 6.26), with the factor predicting low by 12%. Two trials were conducted to determine temperature effects on overall D M S emissions from the brown stock washing area. The temperatures of the two inlet streams that could be manipulated in a brown stock washing process, the stock from the digester and the wash water ( 0 2 filtrate) were each reduced by 10°C. For both trials, the change in D M S emissions was relatively insignificant with about a 4% reduction. There are other consequences of reducing these temperatures, such as reduced washing efficiencies, but since this process change does not appear to be promising for reducing D M S emissions, discussion of possible side-effects are not necessary. A relatively high D N C G flow was noted from the blow tank, with a large fraction likely entering through poorly sealed over-flow openings. A trial was conducted to determine the effect of better sealing these overflow openings. It is assumed that tramp air ingress could be reduced by an amount that would be sufficient to reduce the D N C G flow from the blow tank by half. With a 50% decrease in the volume of D N C G from the blow tank, the amount of D M S in the vent gas from the blow tank was reduced by about 40%, from 96 to 57 kg/day. Even though the blow tank emissions were reduced by 39 kg/day, total emissions from the brown stock washing system were reduced by only 15 kg/day. A cursory look at the situation may lead one to conclude that halving the vent gas flow out the blow tank would lead to a reduction in D M S sulphur emissions of 48 kg/day (half of the base-case total from the blow tank), or 18% overall. According to the model, a reduction in D M S emissions from upstream process equipment wi l l increase the D M S concentration in stock and liquor flowing to downstream equipment, increasing emissions from this equipment, and the overall emissions reduction would be only 15 kg/day, or 5.7% overall. A trial was then conducted to determine the effects of reducing D M S loading to the brown 177 Chapter 6: Results and Discussion stock washing system. On a practical level, this could be accomplished in a number of ways including a reduction in sulphidity of the cooking process, which would in turn reduce the formation of D M S in the digester (Figure 2.2). A s discussed in Section 2.2.2, D M S is formed from a reaction between a mercaptide ion with a lignin methoxyl group. For this reaction to proceed, the mercaptide ion must first be formed from a reaction between the hydrosulphide ion and a lignin methoxyl group. Sulphidity is a measure related to the initial concentration of the hydrosulphide ion in the white liquor. Determining the extent that these reactions proceed in the digester is difficult, with the complexity increased by the practice of recycling black liquor back to the digester, such as at Howe Sound which uses the "Lo-Solids" cooking process. For this trial, the D M S concentration in the stock from the digester was reduced by 10% to determine the magnitude of the effect on the brown stock washing process. The effect was found to be linear; i.e., a 10% reduction in D M S in the stock from the digester w i l l result in a 10% reduction in emissions (although T R S emissions from other parts of the mi l l would likely also be reduced with a reduction in sulphidity). There would be other potentially significant consequences of sulphidity reduction; therefore, reducing sulphidity for possibly only modest environmental gains would likely not be considered. Another possible way to reduce D M S loading to the brown stock washing process is to optimize the washing efficiency in the digester washing zone prior to this stock being released to the brown stock washing process. The D M S is washed from the liquor along with the dissolved lignin and excess cooking chemicals. A n increase in washing efficiency wi l l decrease the D M S in the stock from the digester. This same principal holds true for stock processing within the brown stock washing system; i f the efficiency of the diffusion washer could be increased, more D M S would be captured into the filtrate and less would be carried through to downstream process equipment. According to the base-case balance, the Howe Sound diffusion washer was not operating as efficiently as would typically be expected. The reasons for this are not fully understood; the equipment supplier should be consulted for advice on how to optimize operation. Because there appears to be room for improved operation, trials were conducted to determine the effect on D M S emissions i f some operating parameters of the washers could be improved. A trail was conducted by increasing the displacement ratio of the two stages of diffusion washing from 62% to 75%. The overall D M S emissions from brown stock washing were reduced by 33% with this change. A s expected, the emissions from the diffusion washer filtrate tanks were 178 Chapter 6: Results and Discussion slightly higher because the filtrate D M S concentrations were higher, but this increase was more than offset by a decrease in emissions in downstream equipment. A similar trial, increasing the displacement ratio, was then done for the decker washer. Even though the decker washer appeared to be operating at a reasonable efficiency, a trial was conducted to determine the magnitude of the effect that could be expected i f there was a change. When the decker washer displacement ratio was increased from 75% to 80%, the overall D M S emissions actually increased. Increasing the washer efficiency wi l l reduce the D M S in the washed stock going to 0 2 delignification system, but at the expense of an increased D M S concentration in the decker washer filtrate, with the subsequent higher emissions in equipment processing this liquor. Washing efficiency is also related to washed stock consistency; i.e., a washer operating at a higher washed stock consistency wi l l contain less liquor (higher dilution factor); therefore, the washed stock wi l l contain less dissolved solids (and D M S ) . A trial was conducted by increasing the consistency of the washed stock that exits the diffusion washer, from 8% consistency to a more typical value of 10%. The overall D M S emissions from brown stock washing were reduced by 8.7% with this change. Similar to the trial where the displacement ratio was increased, the emissions from the diffusion washer filtrate tanks were higher, but again this increase was more than offset by a decrease in emissions in downstream equipment. When the washed stock consistency from the decker washer was increased from 12.3% to 14% (still a reasonable design target), the overall D M S emissions were decreased by 9.1%. In this case, the higher dilution factor from increasing the washed stock consistency results in a lower D M S concentration in the decker washer filtrate and a subsequent reduction in D M S emissions from other equipment using this liquor for processing. For emissions reduction, the final trial was to return D M S emissions back to the process. High D M S concentration D N C G from the diffusion washer filtrate tanks was "scrubbed" by sparging it into the decker washer filtrate tank. Based on phase equilibrium, the relatively low concentration decker filtrate w i l l absorb D M S from this gas. This does occur, but the resulting higher D M S concentration in the decker washer filtrate results in higher emissions not only from this tank, but from other process equipment as well , and the overall reduction in D M S emissions was only 3.8%. Two trial were conducted to determine the effect of "hot blowing" on D M S emissions. This term refers to when the stock from the digester climbs above about 90°C. With an increase in stock 179 Chapter 6: Results and Discussion temperature to 105°C, D M S emissions increase by only 12%. The stock from the digester first flows to the 1s t stage of the diffusion washer where it is washed (and cooled) by the washing liquor (although at this temperature there may be flashing inside the diffusion washer). This results in hotter liquor temperatures (increasing from 78°C for the base-case to 95°C) in the 1 s t stage diffusion washer filtrate tank, which results in the higher D M S emissions. During hot blowing, to avoid damaging the diffusion washer due to flashing, some mills w i l l temporarily by-pass the diffusion washer and blow the stock directly to the blow tank. If the stock from the digester goes directly to the blow tank at 105°C, emissions wi l l go up dramatically. The stock wi l l flash in the blow tank from 105°C down to 101 °C (there is a boiling point rise of about 1.1°C due to the dissolved solids), with approximately 0.8% of the water fraction vaporizing, resulting in a vent vapour volumetric flow of 80 Am 3 /m in . This would likely overwhelm the D N C G collection system, resulting in odorous vapour spilling out the poorly sealed overflow openings to atmosphere. Because D M S relative volatility compared to water is high, about 90% wi l l flash off into the vent vapour. This wi l l quadruple overall D M S emissions from 265 to 1076 kg/day (as sulphur). It is upset situations such as this, often occurring during startup or shutdown situations, that result in many of the odour complaints from the surrounding community. 6.73 Reducing B r o w n Stock Washing D M S Emissions The brown stock washing process is a counter-current washing operation beginning with the wash water ( 0 2 filtrate) introduced at the decker washer and finishing with the diffusion washer filtrate exiting the 1 s t stage diffusion washer. There is recycling of decker washer filtrate within the washing process, but only downstream of the diffusion washer; decker washer filtrate is used for dilution of stock in the bottom of the blow tank and ahead of the decker washer, and for de-knotting and screening. The recycling of liquor results in the recycling of D M S ; from the base-case model in Figure 6.40, it can be seen that the D M S loading in the decker filtrate is 828 kg/day, more than three times higher than that contained in the stock passing through the blow tank. Due to liquor recycling, any D M S in the washed stock leaving the diffusion washer wi l l have multiple opportunities to be 180 Chapter 6: Results and Discussion emitted; therefore, operational or equipment changes aimed at reducing D M S emissions should be focussed on the process upstream of the blow tank. The most significant reductions in D M S emissions are based on achieving better washing, either in the digester washing zone or in the diffusion washer. For Howe Sound, there appears to be potential for improving the operation of the diffusion washer. Increasing the efficiency of this washer would have production benefits (such as reduced operating costs from lower consumption of bleaching chemicals) and it would have the beneficial side-effect of reducing overall emissions from the washing process. More efficient operation of the diffusion washer w i l l result in the capture of more D M S (and other volatiles) into the washer filtrate, with this liquor processed in the evaporators, where the vent is collected to the C N C G system. For Howe Sound, this w i l l have the benefit o f shifting sulphur loading from the D N C G to the C N C G system. Unlike the D N C G which is fired in the power boiler, potentially causing S 0 2 emissions, C N C G is incinerated in the lime kiln. The lime kiln can generally tolerate sulphur in the combustion process, with most of the S 0 2 absorbed into the lime mud forming sodium and calcium sulphate. For other mills, capturing higher levels of D M S and other volatiles into the washer filtrate may have an even more profound impact. Not only are C N C G systems much smaller, therefore cheaper to install and operate, but most mills already have these systems. For a mi l l without a D N C G system, emissions from the brown stock washing area to atmosphere could be reduced, and the necessity of collecting D N C G sources from downstream of a highly efficient first washing stage could be questioned. A s a first stage of an emissions reduction program, a mi l l could optimize their initial stages of brown stock washing. Then, since the N C G flow from the first washing stage filtrate tanks is low and the concentration is already relatively high, averaging 13,700 ppm (mol) at Howe Sound (Table 6.24), which is near the L E L of 22,000 ppm (mol) (Table 2.2), this source could be added to their existing C N C G system. / If the m i l l is planning to install a D N C G system, some mil l testing would be required to construct a base-case balance similar to the one presented here for the Howe Sound mi l l . D N C G systems collect up to 30 sources and can cost up to 10 mill ion dollars to install, so elimination of just a few sources and reduction in the D N C G flow from other sources could result in a significant savings, not only in installation costs, but also in operating costs. 181 6.8 The Future of N C G Collection Systems Chapter 6: Results and Discussion Testing at Howe Sound showed that only about 2% of the D M S contained in the stock from the digester going to brown stock washing remained in the washed stock (Figure 6.40). Therefore, with efficient washing, collection of N C G vent sources after the brown stock washing process is not necessary. Testing at Howe Sound also showed that the N C G flow from sealed equipment, such as the filtrate tanks, was relatively low; whereas, flow from unsealed sources such as the blow tank and decker washer hood was relatively high (Table 6.27). The high flows from these sources was mainly due to tramp air ingress. Therefore, to minimize the volume of N C G that requires collection and treatment, washing equipment should be sealed; and to further prevent tramp air ingress, all washing equipment could be operated pressurized. This equipment is commercially available, including pressure diffusion and drum washers and sealed screens and knotters. Historically low-volume high-concentration N C G sources have been collected into a separate dedicated concentrated N C G ( C N C G ) system with the theory being that the concentration of combustibles in this gas would be above the upper explosive limit ( U E L in Table 2.2); i.e., there is insufficient oxygen available to sustain combustion. In reality, in the Author's experience, many i f not most C N C G systems operate in the explosive range; this is taken into account during design of these systems. Design features for C N C G systems include non-sparking motive equipment (steam ejectors rather than fans), flame arresters, rupture disks, and numerous safety interlocks. For the design of a new kraft pulp mi l l , all potential N C G sources, including washing equipment and black liquor tankage, could be designed as sealed and pressurized with all sources treated as C N C G . A l l sources could then be collected into a single C N C G system collecting a relatively low volume of gases. This would greatly reduce installation and operating costs as only a single relatively small N C G system would be required. 182 Chapter 7: Conclusions and Future Work C h a p t e r 7 C o n c l u s i o n s a n d F u t u r e W o r k 7.1 C o n c l u s i o n s A method using a vapour-liquid equilibrium ( V L E ) model was developed to predict emissions of volatile compounds from kraft pulp mills. From the Howe Sound mi l l sampling and testing results, the following conclusion was made: 1. Over 90% of the sulphur emissions from the brown stock washing area were in the form of D M S , with the balance as D M D S (Tables 6.11 to 6.22). The lowest p H measured was 11.4; above a p H of about 11 H 2 S and M M are essentially fully dissociated (Figure 3.2). A s expected, H 2 S and M M were either below the detection limit or at very low concentrations in the vent vapour stream. Since kraft mills must operate above a p H of 11 to ensure that lignin does not precipitate, a similar trend for TRS composition of brown stock washing emissions would likely be observed at other mills. Based on this observation, the decision was made to use D M S as a surrogate compound for T R S and all subsequent lab work focussed on this compound. From phase equilibria analytical testing, the following conclusion was reached: 2. The activity coefficient for D M S in black liquor could be modelled using the e N R T L equation based on a DMS-water-sodium salts system. Since sodium makes up over 96% of the cation loading of the black liquor dissolved solids, it was found that the electrolyte effect could be modelled based solely on the sodium concentration. The electrolyte effect was independent of the anion associated with the sodium cation; this was true whether the anion originated from the inorganic sodium salts or from the organic salts such as the sodium phenolates (alkali lignin). This conclusion is based on the following: 2a. Activity coefficients describing the V L E of D M S in water were determined from 183 Chapter 7: Conclusions and Future Work phase equilibria testing. D M S was found to form a highly non-ideal, aqueous solution, with activity coefficients ranging from 173 to 124 in the temperature range of 20 to 90°C (Table 6.2). 2b Regression techniques were used to determine N R T L parameters for a binary D M S -water system from experimental data. The best fit N R T L parameters are given in Table 6.6. 2c. Activity coefficients were also determined for a DMS-water system containing dissolved solids, including solutions containing sodium salts, alkali lignin and black liquor solids. The addition of electrolytes, such as sodium salts, to a DMS-water system resulted in a significant "salting-out" effect on D M S ; i.e., the relative volatility of D M S with respect to water was increased. For example, addition of 6 wt% sodium salts increased the activity coefficient of a DMS-water solution at 20°C by over 100%, from 173 to 351 (Table 6.2). 2d Regression techniques were used to determine e N R T L parameters for a ternary DMS-water-sodium salts system from experimental data. The best fit e N R T L parameters are given in Table 6.8. 2e The e N R T L equation, using the parameters determined for the DMS-water-sodium system, was used to predict the activity coefficient of D M S in black liquor. There was good agreement, typically with only a few percent difference and no more than 12%, between the black liquor activity coefficient experimental data and the predicted value using the e N R T L equation (Table 6.9). The V L E model, using the e N R T L activity coefficient correlation and the parameters determined for D M S in black liquor, was applied to the Howe Sound mil l sampling and testing results, from which it was concluded: 184 Chapter 7: Conclusions and Future Work The V L E model could be used to model D M S emissions from individual process equipment and these could be combined to model the entire brown stock washing line. For all equipment there was a systematic over-prediction of D M S concentration in the vent vapour; this was attributed to insufficient liquid phase residence time for equilibrium to be established and/or dilution of the vent vapour from tramp air ingress. Modell ing resulted in an over-prediction of 19% for D M S emissions (Table 6.30); this over-prediction could be beneficial as it reflects a "worst-case-scenario," which would provide a level of contingency in the design of N C G collection and treatment systems. This conclusion is based on the following: 3a. In lab tests for a DMS-water system, equilibrium was established more rapidly as the vapour-to-liquid ratio and the level of agitation increased (Figure 6.23). Based on this work, for a high vapour-to-liquid ratio and a low level of agitation (eg. blow tank), and for a low vapour-to-liquid ratio and a higher level of agitation (eg. filtrate tank), equilibrium is expected to be established in 15 minutes or less. 3b. For the Howe Sound decker washer filtrate tank, with only 4.5 minutes liquid residence time, the measured D M S vent vapour concentrations were, on average, 84% of the predicted equilibrium concentration (Figure 6.30). This agrees with the lab test results which indicate that concentration of D M S in the vapour should reach 80 to 90% of the equilibrium value after this time. 3c. For the Howe Sound 2 n d stage diffusion washer filtrate tank, with only 6.8 minutes liquid residence time, the measured D M S vent vapour concentrations were, on average, 94% of the predicted equilibrium concentration (Figure 6.31). This agrees with the lab tests which indicate that concentration of D M S in the vapour should reach 85 to 95% of the equilibrium value after this time. 3d. For the Howe Sound blow tank, the measured D M S vent vapour concentrations were, on average, 31% of the predicted equilibrium concentration (Figure 6.32). Based on 185 Chapter 7: Conclusions and Future Work lab tests, the liquid residence time of 83 minutes should be more than sufficient for equilibrium to be established. The over-prediction of D M S concentration in the vapour was attributed to tramp air being drawn into the blow tank through poorly sealed overflow openings, significantly diluting the vent vapour. 3e. For the Howe Sound decker washer hood, the predicted vapour equilibrium concentration was based on the inlet wash liquor D M S concentration. The measured D M S vent vapour concentrations were, on average, 61 % of the predicted equilibrium concentration (Figure 6.33). The over-prediction of D M S concentration in the vapour can likely be attributed to tramp air ingress diluting the vent vapour, although more testing is required to confirm this. The Howe Sound mil l sampling and testing results were compared to emission factors to determine their accuracy for this particular mi l l : 4. Industry standard emission factors predict reasonably well for the organic T R S emissions observed at Howe Sound, although they greatly over-predicted for H 2 S (Figures 6.34 and 6.35). Emission factors are independent of installed equipment and T R S concentrations in the process, being based solely on production rates; therefore, they can not be used to predict changes in emissions as a result of equipment or operational changes. To demonstrate how the results of this work could be applied to an industrial situation, and to show how emission factors could be improved upon, commercially-available simulation software incorporating the V L E model was used to model the mi l l testing results. C A D S i m Plus was used to construct a base-case heat and mass balance reflecting the current operation of the Howe Sound brown stock washing process. To determine the changes to overall D M S emissions as a result of potential operational or equipment modifications, this base-case balance was modified accordingly. From the mi l l modelling results, the following conclusions were drawn: 186 Chapter 7: Conclusions and Future Work 5. The most significant reductions in D M S emissions from the brown stock washing area were found to be based on achieving better washing, either in the digester washing zone or in the first stages of brown stock washing. For Howe Sound, the first washing stage, the two stage atmospheric diffusion washer, appears to be operating at a lower than expected efficiency. If the diffusion washer operation could be improved, increasing the displacement ratio from 62% to 75%, overall D M S emissions from the brown stock washing area could be reduced by 31%. 6. If all potential sources of N C G were operated pressurized, ensuring no tramp air entered the system, a single low-volume high-concentration N C G system could be installed to collect all N C G sources. This would have multiple benefits over the traditional two systems including: (a) lower installed cost, (b) lower operating cost, (c) less impact on incineration point, (d) higher reliability; i.e., less venting of N C G , and (e) less operator attention required. 7.1 Future W o r k The following activities are recommended for further research work on this subject: 1. Phase equilibria testing, including determination of dissolved solids effects, for hydrogen sulphide, methyl mercaptan and dimethyl disulphide. 2. Additional phase equilibria testing could be conducted at higher black liquor total dissolved solids concentrations. The correlations developed from this work could then be used for modelling T R S in black liquor evaporators and downstream tankage. 3. More detailed mil l testing is required around the decker washer. The current model may not represent very well what is happening inside the decker washer hood. Due to the limited data collected, a simple model was used that was based on equilibrium being established between the inlet wash liquor and the vent vapour. 187 Chapter 7: Conclusions and Future Work 4. More detailed mil l testing is required to quantify tramp air effects, such as the suspected air ingress through the overflow openings on the blow tank. 5. Testing at a number of mills is required to build up a database for specific equipment such as the blow tank. This could be used to determine typical blow tank conditions which could then be used for general modelling purposes when no mi l l testing can be done, such as for the design of a new mil l . 6. For equipment that has limited residence time, a mass transfer model could be developed based on volumetric mass transfer coefficients for each type of equipment. 188 References References Abrams, D.S. and Prausnitz, J . M . (1975), "Statistical thermodynamics of liquid mixtures: A new expression for the excess Gibbs energy of partly of completely miscible systems," A I C h E Journal, Vol .21 , pp. 116-121. A C G I H , American Conference of Governmental Industrial Hygienists, "Guide to Occupational Exposure Values," adopted by the Workers Compensation Board of British Columbia and most other provincial authorities, <http://www.acgih.org/home.htm> (December 1, 2006). A . H . Lundberg Systems Limited, Richmond, B . C . . http://www.ahlundberg.com (December 1,2006). Alberta Government, air quality management system, <http://www3.gov.ab.ca/env/air/index.html> (December 1,2006). Al len , L . (1998a), " A i r emission challenges and priorities in the Canadian pulp and paper industry," Paprican Miscellaneous Report 375. Al len , L . (1998b), "Control of total reduced sulphur emissions from pulp and paper mil ls ," Paprican Miscellaneous Report 389. Al len , L . (2002), " A comparison of direct measurement vs. emission factors for N P R I reporting: methanol air emissions case studies," Tappi Environmental Conference Proceedings. Anderko, A . ; Wang, P.; and Rafal, M . (2002), "Electrolyte solutions: from thermodynamic and transport property models to the simulation of industrial processes," Fluid Phase Equilibria, V o l . 194-197, pp.123-142. Andersson, K and Bergstrom, J.G.T. (1969), "The formation of organic sulphur compounds during kraft pulping, Part 1," Svensk Papperstidning, Vol.11.15, pp.375-379. 189 References Andersson, K . (1970), "The formation of organic sulphur compounds during kraft pulping, Part 2," Svensk Papperstidning, V o l . 1.15, pp. 1-10. Aspentech, <http://www.aspentech.com/> (December 1, 2006). Axegard, P.; Backlund, T . L . B . ; and Warnqvist, B . (2001), "The eco-cyclic pulp mi l l - with a focus on closure, energy - efficiency and chemical recovery development," Paptac / Tappi International Chemical Recovery Conference Proceedings, Whistler, British Columbia, Canada, pp. 1-6. Aurel Systems Limited, <http://www.aurelsystems.com/> (December 1, 2006). Balzhiser, R .E . ; Samuels, M . R . ; and Eliassen, J.D. (1972), "Chemical Engineering Thermodynamics," Prentice Hal l , Inc. Englewood Cliffs, New Jersey. Barrett, T.J.; Anderson, G . M . ; and Lugowski, J. (1988), "The solubility of hydrogen sulphide in 0-5 m N a C l solutions at 25°-95°C and one atmosphere", Geochimica et Cosmochimica Acta, Vol .52, pp.807-811. British Columbia Government, environmental protection division, <http://www.env.gov.bc.ca/air/> (December 1,2006). Bentvelzen, J . M . ; McKean, W.T.; Gratzl, J.S.; L i n , S.Y.; and Tucker, W.P. (1975), "Kinetics of methyl mercaptan oxidation and dimethyl disulfide hydrolysis in alkaline solutions," Tappi Journal, Vol.58, No . l ,pp . l02-105 . Berube, P.R.; Parkinson, P.D. ; and Hal l , E. (1999), "Measurement of reduced sulphur compounds contained in aqueous matrices by direct injection into a gas chromatograph with a flame photometric detector", Journal of Chromatography A , Vol.830, pp.485-489. 190 References Blackwell , B .R . ; MacKay, W . B . ; Murray, F .E . ; and Oldham, W . K . (1979), "Review of kraft foul condensates," Tappi Journal, Vol .62, No. 10 pp.33-37. Blackwell , B .R. ; Murray, F .E. ; and Oldham, W . K . (1980), "Vapour/liquid equilibrium relationships for kraft foul condensates," Tappi Journal, Vol .63, No.2, pp.151-154. B L R B A C , Black Liquor Recovery Boiler Advisory Committee, Waste Streams Subcommittee, "Recommended good practice for the thermal oxidation of waste streams in a black liquor recovery boiler," R e v . l , May 10, 1999, <http://www.blrbac.org/> (December 1, 2006). Bruce, D . and van der Vooren, T. (2003)," Trends in air emission limits for world class mil ls ," Pulp and Paper Canada, Vol.104, No.7, pp.51-54. Burgess, T. (1992), "Noncondensible Gases," Chemical Recovery in the Alkaline Pulping Processes, 3 r d edition," Chapter 3, Tappi Press. Canada Government, health and air quality, <http://www.hc-sc.gc.ca/ewh-semt/air/index_e.html> (December 1,2006). Carroll, J.J. and Mather, A . E . (1989), "The solubility of hydrogen sulphide in water from 0 to 90°C and pressures to 1 Mpa , " Geochimica et Cosmochimica Acta, Vol .53, pp.1163-1170. C C M E , Canadian Council of Ministers of the Environment, <http://www.ccme.ca/> (December 1, 2006). C C O H S , Canadian Centre for Occupational Health and Safety, C H E M I N F O database, <http://ccinfoweb.ccohs.ca/cheminfo/search.html> (December 1, 2006). 191 References Chai, X . S . ; Dhasmana, B . ; and Zhu, J .Y. (1998), "Determination of volatile organic compound contents in kraft mi l l streams using headspace gas chromatography", Journal of Pulp and Paper Science, V o l . 24, No.2, pp.50-54. Chai, X . S . and Zhu, J .Y . (1998), "Indirect headspace gas chromatographic method for vapor-liquid phase equilibrium study", Journal of Chromatography A , V o l . 799, pp.207-214. Chai, X . S . ; L u i , P . - H ; and Zhu, J .Y. (2000), "Analysis of volatile organic sulphur compounds in kraft liquors by full evaporation headspace gas chromatography", Journal of Pulp and Paper Science, V o l . 2 6 , No. 5,pp.l67-172. Chamberlin, D.S . and Clarke, D.R. (1928), "Flame speed of hydrogen sulphide," Journal of Industrial and Engineering Chemistry, Vol.20, pp.1016-1018. Chen, C . -C . and Song, Y . (2004), "Generalized Electrolyte-NRTL Model for Mixed-Solvent Electrolyte Systems", A I C h E Journal, Vol.50, No.8, pp.1928-1941. Cook, D. and Hoy, D . (2003), "Analytical techniques for investigating odors at pulp and paper facilities: preliminary results," T A P P I International Environmental Conference Proceedings. Cooper, H . B . H , (1974), "Kinetics of Inorganic Sulphur Oxidation during Black Liquor Oxidation with Oxygen", Tappi Journal, Vol.57, No.10, pp.130-133. Crawford, R.J. ; Carlson, L . W . ; Crapo, A.M.;Drese, J .H. ; Sain, J.D. (2006), "Reduced sulphur compound and methane emissions from kraft mil l wastewater treatment plants," Pulp and Paper Technical Association of Canada (Paptac), 92 n d Annual Meeting Proceedings, Montreal, Quebec, Canada, pp.B263-B268. 192 References Crawford, R.J. ; Rovel l -Rixx, D .C . ; Jett, S.W.; Jain, A . K . ; Dillard, D