Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Catalytic recovery of the oxidative power of the chlorate residual in the chlorine dioxide delignification Heynen, Ian F. 1995

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
ubc_1995-0155.pdf [ 5.67MB ]
Metadata
JSON: 1.0058603.json
JSON-LD: 1.0058603+ld.json
RDF/XML (Pretty): 1.0058603.xml
RDF/JSON: 1.0058603+rdf.json
Turtle: 1.0058603+rdf-turtle.txt
N-Triples: 1.0058603+rdf-ntriples.txt
Original Record: 1.0058603 +original-record.json
Full Text
1.0058603.txt
Citation
1.0058603.ris

Full Text

CATALYTIC RECOVERY OF THE OXIDATIVE POWER OF THE CHLORATE RESIDUAL IN CHLORINE DIOXIDE DELIGNJEICATION By Ian F. Heynen B.A.Sc. University of Waterloo, 1992  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE STUDIES Department of Chemical Engineering  We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA April 1995  ©Ian F. Heynen  In presenting this thesis degree  in partial fulfilment  of the requirements  for an advanced  at the University of British Columbia, I agree that the Library shall make it  freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes department  may be granted  or by his or her representatives.  by the head of my  It is understood  that  copying or  publication of this thesis for financial gain shall not be allowed without my written permission.  Department of  C V^Qsw\{c.oJL ^"v^-Inee/''  The University of British Columbia Vancouver, Canada  DE-6 (2788)  ABSTRACT  Chlorine dioxide has traditionally been used extensively in the brightening stages of kraft pulp bleaching. More recently, however, environmental pressures have resulted in increased substitution of chlorine dioxide for chlorine in thefirstbleaching stage, otherwise known as the chlorination or delignification stage. It is now common, in Canadian bleached kraft mills, for the bleach plant to be operated with 100 percent chlorine dioxide substitution in thefirststage. One of the few disadvantages of using chlorine dioxide in the bleaching of chemical pulp is the formation of chlorate. Chlorate formation accounts for up to 30 percent of the chlorine dioxide charge. Since chlorate is inactive as a bleaching chemical, chlorate formation represents a loss in chlorine dioxide bleaching power, and increased bleaching costs. The formation of chlorate has an environmental cost as well. Chlorate is a known fungicide and herbicide, and the chlorate contained in bleached kraft mill effluent has resulted in the elimination of the Bladder Wrack population in the areas surrounding bleached kraft pulp mill outfalls in the Baltic Sea.  The purpose of this study was to examine the potential for vanadium pentoxide to recover the bleaching power of the chlorate residual in the chlorination stage, when the chlorination stage was operated at high substitution. Vanadium pentoxide has been shown to catalyze the transfer of electronsfromchlorate to pulp by cycling between the +5 and +4 oxidation states. This catalytic transfer of electrons has loosely been termed the "activation" of the chlorate residual. By "activating" the chlorate residual, we  ii  hoped to decrease thefinalchlorate concentration and recapture the bleaching power lost as chlorate. The investigation was conducted using factorial experimental design at two levels. The experimental conditions were chosen to reflect practical bleaching conditions. The experiments were performed using the bag bleaching method with an oxygen delignified pulp. The factors included in the factorial design were temperature, retention time, pH, chlorine dioxide substitution, and vanadium pentoxide charge. In order to measure the "activation" of the chlorate residual, the response of the following variables were measured: final chlorate concentration, brightness, kappa number, pulp viscosity, and pulp physical properties.  The addition of vanadium pentoxide to the chlorination stage had a large impact on the final chlorate concentration. For bleaching runs performed at 80°C, for four hours and at a catalyst charge of 0.005 to 0.01 weight percent, reductions in chlorate concentration of up to 95 % were obtained.  Despite the large drops in chlorate concentration, the addition of vanadium did not substantially increase the oxidizing power of thefirstbleaching stage. When compared to uncatalyzed bleaching runs at the same temperature, the addition of the catalyst resulted in brightness gains of up to 1 percent. The catalyst had no significant impact on the (D+C)E kappa number of the bleached pulp.  A serious disadvantage of using vanadium pentoxide to "activate" the chlorate residual in chlorine dioxide delignification is the loss in pulp viscosity. The viscosity dropped  iii  from the uncatalyzed values of 22 to 23 cps to as low as 16 cps. These viscosity losses, however, did not result in significant losses in pulp strength properties. From this study it is recommended that further investigations of recovery of the bleaching power of chlorate in the delignification stage be performed. Of particular interest are organic catalysts for application in closed cycle mills where the solids content of the bleach plant effluent will befiredin the recovery boilers.  Another recommendation focused on the use of vanadium pentoxide for chlorate elimination and effluent improvement. The chlorination stage recycle would be passed through a vanadium pentoxide catalyst stage prior to re-entering the chlorination stage. The vanadium pentoxide could be in the form of plates or immobilized catalyst. This would result in reductions in chlorate concentration in the chlorination stage without entraining vanadium pentoxide in the effluent stream.  iv  T A B L E OF CONTENTS ABSTRACT  ii  T A B L E OF CONTENTS  v  LIST OF TABLES  vii  LIST OF FIGURES  viii  ACKNOWLEDGMENTS  ix  1. INTRODUCTION  1  1.1. CHLORATE FORMATION 1.2. RECOVERING THE BLEACHING POWER OF THE CHLORATE RESIDUAL 1.3. OBJECTIVES OF THIS WORK  2 3 5  2. LITERATURE REVIEW  7  2.1. CHLORATE FORMATION 2.1.1. Residual Oxidants in Chlorine Dioxide Bleaching 2.1.2. The mechanism of Chlorate Formation 2.1.2.1. Chlorate Formation and Pulp Properties 2.1.2.2. The Inorganic Reactions of Chlorate Formation 2.1.3. The Impact of Operating Conditions on Chlorate Formation 2.1.3.1. Chlorine Dioxide Substitution 2.1.3.2. Order of Oxidant Addition 2.1.3.3. Reaction Time 2.1.3.4. Bleaching Temperature.... 2.1.3.5. Delignification Stage pH 2.1.3.6. Chlorine Contamination of the Chlorine Dioxide Charge 2.1.3.7. Chlorination Stage Recycle 2.1.3.8. Pulp Consistency 2.1.3.9. Oxidant Charge 2.1.3.10. Chloride Ion Concentration 2.1.3.11. The Impact of Chlorate Formation on the Receiving Environment 2.1.4. Reduction of Chlorate during Effluent Treatment 2.2. VANADIUM PENTOXIDE 2.2.1. Vanadium Pentoxide Catalysis 2.2.2. Vanadium Pentoxide Oxidations of Organic Compounds 2.2.3. Kinetics of Vanadium (V) Oxidations 2.2.4. Aqueous Solutions of Vanadium in the +IV and +V Oxidation State 2.3. VANADIUM PENTOXIDE "ACTIVATION" OF CHLORATE 2.3.1. The Mechanism of the Catalytic Reaction  7 7 10 10 11 15 15 16 18 18 20 21 22 22 23 23 24 25 26 27 28 29 29 31 33  3. EXPERIMENTAL  34  3.1. EXPERIMENTAL DESIGN 3.1.1. Factor Variables 3.1.2. Summary of Runs 3.1.3. Response Variables 3.2. MATERIALS AND METHODS  34 34 36 37 40  v  3.2.1. The Bleaching Runs 3.2.2. The Pulp Source 3.2.3. Vanadium Catalyst Preparation 3.2.4. Chlorine Solution Preparation 3.2.5. Chlorine Dioxide Generation 3.2.6. Bleached Pulp Property Testing (CPPA 1966-) 3.2.7. Bleaching Liquor Analysis 3.2.7.1. Chlorate Analysis , 3.2.7.2. The Air Oxidation of Iodide to Iodine during Chlorate Determination  4. RESULTS AND DISCUSSION  40 40 40 41 41 42 42 43 44  46  4.1. THE RESPONSE OF THE RESIDUAL CHLORATE CONCENTRATION 48 4.1.1. The Effect of Catalyst Charge 49 4.1.2. The Effect of Temperature 51 4.1.3. The Effect of Reaction Time 52 4.1.4. The Effect of Chlorine Dioxide Substitution 52 4.1.4.1. Effect on Chlorate Formation in Uncatalyzed Bleaching 54 4.1.4.2. Effect on Chlorate Concentration in the Catalyzed Runs 55 4.1.4.3. Impact of Substitution on the Percent Reduction of Chlorate in Catalytic Bleaching.... 57 4.1.5. The Effect of pH 57 4.2. THE RECOVERY OF THE BLEACHING POWER LOST TO CHLORATE 58 4.2.1. The Kappa Number Response 59 4.2.2. The Brightness Response 61 4.2.2.1. Brightness from Catalyzed and Uncatalyzed Bleaching Runs 61 4.3. THE IMPACT OF THE CATALYTIC BLEACHING ON PULP PROPERTIES 63 4.3.1. Pulp Viscosity 64 4.3.1.1. The Effect of Chlorine Dioxide Substitution 64 4.3.1.2. The Effect of Catalyst Charge 65 4.3.1.3. The Effects of Temperature and Time 67 4.3.2. The Response of the Physical Properties 67 4.4. POTENTIAL APPLICATIONS FOR VANADIUM IN THE DELIGNIFICATION STAGE 69  5. CONCLUSIONS  73  6. RECOMMENDATIONS  75  PREFERENCES  76  APPENDIX A THE DETAILED CATALYTIC BLEACHING PROCEDURE  83  APPENDIX B THE CHLORINE DIOXIDE GENERATION PROCEDURE  86  APPENDIX C DESIGN AND ANALYSIS OF INDUSTRIAL EXPERIMENTS (MURPHY, 1977)  89  APPENDIX D SUMMARY AND FACTORIAL ANALYSIS OF EXPERIMENTAL RESULTS  vi  105  LIST OF TABLES Table 2.1: A summary of the literature concerning chlorate formation in the chlorination stage, (expandedfromFair 1991)  19  Table 2.2: A summary of the literature concerning vanadium pentoxide oxidations of organic substrates 30 f Table 3.1: Factor level table for Experimental Program 1  38  Table 3.2: Factor level table for Experimental Program 2  39  Table 3.3: A summary of the uncatalyzed bleaching runs  39  Table 3.4: A summary of the pulp testing methods used in this study  42  Table 4.1: A summary of the experimental programs  47  Table 4.2: Results of the factorial analysis of chlorate concentration for Experimental Program 1 50 Table 4.3: Results of the factorial analysis of chlorate concentration for Experimental Program 2 51 Table 4.4: Chlorate production in the uncatalyzed bleaching runs  54  Table 4.5: The factorial analysis of the (D+C)E kappa number and brightness responses  60  Table 4.6: The factorial analysis of the (D+C)E viscosity response  65  Table 4.7: Factorial analysis of the physical testing results for the runs at 80°C  68  Table 4.8: A comparison of the physical testing results for the 40°C uncatalyzed runs and the 80°C runs with a 0.015 weight percent catalyst charge 69  vii  LIST OF FIGURES Figure 1.1: The prevalent bleaching sequence used in Canadian kraft mills  1  Figure 2.1: The concentration profiles for the inorganic chlorine species in pure chlorine dioxide bleaching of kraft pulp at 45°C  9  Figure 2.2: The mechanism of chlorate formation during chlorine dioxide bleacWng. 13 Figure 2.3: The impact of substitution and mode of addition on chlorate production. 17 Figure 2.4: The formation of chlorate at 25°C and 45°C  20  Figure 2.5: The aqueous chlorine equilibrium distribution  21  Figure 2.6: The distribution of vanadium (+5) in aqueous solution  31  Figure 2.7: The mechanism of the vanadium catalyzed bleaching reaction  33  Figure 4.1: Chlorate concentration as a function of catalyst charge, temperature, and substitution 53 Figure 4.2: An illustration of chlorate concentration profiles in catalyzed and uncatalyzed chlorine dioxide bleaching  56  Figure 4.3: Brightness as a function of temperature and substitution  63  Figure 4.4: The impact of vanadium addition on viscosity.  66  Figure 4.5: Conceptual flow sheet for the chlorination stage operated with vanadium pentoxide addition 71 Figure 4.6: Conceptual flow sheet for the vanadium treatment of the chlorination stage recycle 72  viii  ACKNOWLEDGMENTS  Many people have helped to make this work successful. First and foremost, I must thank my supervisors Dr. K.L. Pinder and Dr. C-L. Lee for their advice and encouragement. I must also thank the PAPRICAN staff at the U.B.C. Pulp and Paper Centre and the Vancouver Laboratory. Special thanks to Sandy Reath for her instruction in standard laboratory techniques. I would like to offer the following thanks to a few others for their less tangible contributions to this work: my parents for their belief in me; Janice for her support, encouragement, and help in meeting the deadlines; my friends in Vancouver, at the house, in the office, and elsewhere, for making the time unforgettable.  Thefinancialsupport provided by the following organizations is gratefully acknowledged:  *  The Science Council of British Columbia - G.R.E. A.T. Grant Program  *  The Vancouver Laboratory of PAPRICAN  *  The Province of British Columbia  ix  1. I N T R O D U C T I O N  The kraft pulping process is used globally for the production of high quality chemical pulp. Generally, kraft pulps are chemically treated in pulp mill bleach plants to produce a high brightnessfinalproduct.  Figure 1.1 illustrates the prevalent bleaching sequence used in Canadian pulp mills (Prykeefor/. 1993). As seen in Figure 1.1, chlorine dioxide is used extensively in the bleaching of kraft pulp  (D, D+C)EoDED  where each bleaching stage is described as follows:  (D,D+C):  A delignification stage operated with most of the chlorine dioxide (D) added more than 10s before chlorine (C) but with a small amount added with chlorine. This stage focuses on lignin removal.  E:  An oxygen reinforced alkaline extraction stage  DED:  The two chlorine dioxide brightening stages separated by an alkaline  0  extraction stage. These stages focus on pulp brightening. Figure 1.1: The prevalent bleaching sequence used in Canadian kraft mills (Pryke et al. 1993)  1  Introduction in Canada. The (D,D+C) stage is known as the chlorination or delignification stage. As the name implies, the purpose of thefirstbleaching stage is to remove most of the lignin that remains in the pulp when it enters the bleach plant. Although the following two D stages do result in further lignin removal, they focus on pulp brightening through oxidation of the chromophores in lignin (Reeve 1992).  To improve effluent quality and to respond to market demands for pulp bleached without molecular chlorine (ECF bleaching), bleach plants are being operated with higher chlorine dioxide substitution in thefirststage. In a recent survey it was reported that 88 percent of the bleach plants in Canada operate at greater than 25% substitution in the chlorination stage, and that an increasing number are operating at 100% substitution (Pryke et al. 1993). The survey found that chlorine dioxide substitution is motivated primarily by environmental concerns. The following reasons for increasing chlorine dioxide substitution were reported:  •  to decrease AOX (adsorbable organic halide), dioxins, furans, and colour in effluents;  •  to decrease dioxins, furans, and organochlorine concentrations in pulp;  •  to decrease the acute toxicity of effluent; and,  •  to improve market acceptance.  1.1. Chlorate Formation One of the few disadvantages of chorine dioxide bleaching is the formation of chlorate ion. The chlorate released in pulp mill effluent has detrimental effects on the receiving  2  Introduction environment and since it is ineffective as a bleaching chemical, chlorate formation represents a loss in bleaching power. Thus, chlorate formation has both environmental and economic implications. The chlorate ion (C10 ") is a known fungicide and herbicide and therefore presents an 3  environmental threat to the receiving environment (Germgard 1989). The chlorate contained in Swedish bleached kraft mill effluent has resulted in the disappearance of the Bladder Wrack population in areas surrounding pulp mill outfalls in the Baltic Sea (Rosmarin et al. 1986). As a result, the Swedish government is currently regulating chlorate emissions from bleached kraft mill effluent in that country.  Since chlorate is unreactive as a bleaching chemical, the formation of chlorate also represents a loss in the bleaching power charged as chlorine dioxide. Chlorate formation accounts for between 10% and 30% of the chlorine dioxide charged (Ni et al. 1993). This directly translates into higher chemical costs  Given the implications of chlorate formation, recovery of the bleaching power lost to chlorate offers both environmental and economic rewards. Since chlorine dioxide substitution in thefirstbleaching stage is increasing, there is significant incentive to recapture the bleaching power of chlorate in this stage. 1.2. Recovering the Bleaching Power of the Chlorate Residual This work is focused on the recovery of the bleaching power lost as chlorate through vanadium pentoxide catalysis. Vanadium pentoxide has been shown to catalyze the transfer of electrons from chlorate to pulp by cycling between the +5 and +4 oxidation  3  Introduction states (Deutsch et al. 1979, Rapson et al. 1959). This catalytic transfer of electrons has loosely been termed the "activation" of the chlorate residual, although it is actually vanadium that oxidizes the pulp substrate. By "activating" the chlorate residual, we hoped to address both the environmental and economic implications of chlorate formation. Previous work has shown that vanadium pentoxide can be used to catalyze the bleaching of kraft pulp with the chlorate ion. "Activating" the chlorate residual in chlorine dioxide brightening resulted in brightness gains, but only at pH values below 3 (Deutsch et al. 1979). Since the optimum pH for chlorine dioxide brightening is 4, there is little incentive to add vanadium pentoxide to the brightening stages. Furthermore, vanadium pentoxide "activation" of chlorate has also been shown to result in serious losses in pulp viscosity (Rapson et al. 1979).  This work is based on the "activation" of the chlorate residual in the delignification stage. In terms of the best site for "activating" the chlorate residual, the delignification stage has several advantages over the brightening stages. The pH used in the delignification stage is generallyfrom2 to 3, which was the pH range found to be most effective for chlorate "activation" (Deutsch et al. 1979). And since the pulp in the first stage still contains relatively high lignin concentration, there is a lower requirement for bleaching chemical selectivity than in the brightening stages. The non-selectivity of vanadium pentoxide "activated" chlorate should have a lower impact in the delignification stage than in the subsequent brightening stages.  4  Introduction 1.3. Objectives of this Work Our investigation focused on the impact of vanadium pentoxide addition to the first bleaching stage, operated at high chlorine dioxide substitution. A primary objective of the work was to choose factors and variable settings that reflect prevalent Canadian mill operating conditions, or potentially useful variations of these operating conditions. Since oxygen pre-bleaching is now widely accepted in the industry, we used an oxygen delignified pulp in our experiments. The variables included in the optimization of catalyzed chlorine dioxide delignification were the  •  Degree of substitution, pH,  •  Time,  •  Temperature, and  •  Vanadium Charge.  The degree of substitution and the pH were chosen because they are known to have strong impacts on chlorate formation (Ni et al. 1993). The retention time, temperature, and catalyst charge were included because of the potential impact on catalyst activity.  The primary goals of the work were as follows:  1.  To examine the impact of the catalyst on thefinalchlorate concentration  5  Introduction 2.  To measure the increase in bleaching power associated with catalyst addition. The bleaching efficiency was measured through brightness and kappa number determinations.  3.  To identify the impact of the catalyst addition on the physical properties of the bleached pulp. The viscosity, strength and optical properties were all measured.  6  2. LITERATURE REVIEW The following sections discuss the literature relevant to this work, including the formation and impact of residual oxidants in chlorine dioxide delignification, the aqueous chemistry and catalytic reactions of vanadium pentoxide, and the use of vanadium pentoxide in kraft pulp bleaching. A detailed review of the literature regarding chlorine dioxide bleaching is beyond the scope of this work. There are many excellent summary papers which discuss the impact of chlorine dioxide substitution on kraft pulp bleaching. The reader is referred to the papers by Ni (1992), Pryke (1992), McCubbin et al. (1991), Solomon et al. (1993), McCleay et al. (1987), and Reeve and Weishar (1991) for thorough discussions of the use of chlorine dioxide in thefirststage. 2.1. Chlorate Formation Significant progress has been made into understanding the mechanism of chlorate formation in kraft pulp bleaching, and the factors which contribute to chlorate formation. The following sections discuss the residual oxidants in chlorine dioxide delignification, the mechanism of chlorate formation, the impact of chlorate formation on the receiving environment, and the operating conditions which minimize chlorate formation.  2.1.1. Residual Oxidants in Chlorine Dioxide Bleaching When chlorine dioxide oxidizes pulp, it is reduced to lower oxidation states. These lower oxidation state chlorine compounds can further react with either the pulp or  7  Literature Review other inorganic chlorine species in solution. The inorganic chlorine species that exist following chlorine dioxide bleaching termed "residual oxidants". As chlorine dioxide bleaches pulp, it is successively reduced to chlorite (CIO2"), hypochlorous acid (HOG), and chloride (Cl). The reduction half reactions are illustrated in Equations 2.1, 2.2, and 2.3. (2.1)  C10 + e" -> C10 ' 2  2  CIO2" + 3 I f + 2e" - » HOC1 + H 0  (2.2)  HOCl + H* + 2e*  (2.3)  2  Cl" +H 0 2  The inorganic species included in equations 2.1 through 2.3 may also interact to regenerate chlorine dioxide or to form chlorate (Emmenegger and Gordon 1967). Reaction 2.5 shows that chlorite can react with molecular chlorine to regenerate chlorine dioxide. Reactions 2.4 and 2.6 show that the reaction of hypochlorous acid and chlorite can result either in chlorine dioxide regeneration or chlorate formation.  2C10 " + HOCl -> 2 C10 + Cl" +H 0  (2.4)  2C10 " + Cl  (2.5)  2  2  2  2  2C10 + 2C1"  2  2  (2.6)  HOCl + C10 - -> CIO3' + Cl" + r f 2  8  Literature Review  100  o— O — Chlorine Dioxide — X — Hypochlorous Acid —A—Chlorite — D — Chlorate  0  10  20  30  40  50  60  70  80  90  100  110  120  130  140  150  160  Time (min)  Figure 2.1: The concentration profiles for the inorganic chlorine species in pure chlorine dioxide bleaching of kraft pulp at 45°C. (Data from Ni 1992) The distribution of the inorganic species, that resultfromreactions 2.1 through 2.6, has been studied extensively (Reeve and Weishar 1991, Wartiovaara 1985). Typical concentration profiles for the inorganic species produced in chlorine dioxide bleaching are shown in Figure 2.1. The Figure shows that the primary oxidizers, chlorine dioxide and chlorite, were spent after 60 and 150 minutes, respectively. The concentration of chloride, the end product of complete reduction of chlorine dioxide, increased steadily as bleaching proceeded. The concentration of the chlorate ion increased asymptotically to approximately 20 % of the chlorine dioxide charge.  9  Literature Review 2.1.2. The mechanism of Chlorate Formation Chlorate formation resultsfromthe reaction of chlorine dioxide with pulp, followed by a series of inorganic reactions in solution. The following sections discuss the formation of chlorate with respect to both the reaction of chlorine dioxide with pulp, and the reactions of the inorganic chlorine species in the bleaching solution.  2.1.2.1. Chlorate Formation and Pulp Properties Lindgren and Nilsson (1975) studied the formation of chloratefromthe oxidation of nine lignin model compounds with chlorine dioxide. Compounds containing phenolic hydroxy! groups produced the lowest yields of chlorate: between 3 and 5 percent of the oxidizing power of the chlorine dioxide charged. Compounds containing both phenolic hydroxyl and carbonyl groups gave between 10 and 15 percent chlorate. The highest yields, between 12 and 35 percent, camefromcompounds containing an aliphatic double bond. It was also discovered through some related work that chlorine dioxide preferentially attacks phenolic hydroxyl groups over other functional groups.  Bergnor etal. (1987) summarized two pathways in which chlorine dioxide reacts with lignin: •  Path A involves the reaction of chlorine dioxide with phenolic hydroxyl groups. One-half of the chlorine dioxide reacting via this path is reduced to chlorite. In turn, small amounts of chlorite disproportionate to form chlorate, chlorine dioxide and chloride. The other half of the chlorine dioxide that reacts via path A is reduced to hypochlorous acid with no resultant chlorate.  •  Path B involves the reaction with carbonyl groups and aliphatic double bonds. Here, one half of the chlorine dioxide is reduced to  10  Literature Review hypochlorous acid with no resultant chlorate formation. The other half is directly oxidized to chlorate. Comparing paths A and B, it is evident that path B produces most of the chlorate generated from the reaction between chlorine dioxide and lignin. This two path scheme may help to explain why both oxygen and CE pre-bleached pulps produce more chlorate per mole of chlorine dioxide than kraft pulp. Since kraft pulp lignin has a larger proportion of phenolic hydroxyl groups than pre-bleached pulps, more chlorine dioxide reacts via path A, and less chlorate is formed (Fair 1992). 2.1.2.2. The Inorganic Reactions of Chlorate Formation As part of a fundamental study of chlorine dioxide bleaching, Ni et al. (1993) identified the inorganic reaction responsible for the formation of chlorate during the chlorine dioxide bleaching of kraft pulp. The three plausible inorganic reactions that result in chlorate formation are shown below (Ni et al. 1993):  1.  The disproportionation of chlorine dioxide:  C10 + 20H" -> HjO + C10 " + C10 2  2.  2  (2.7)  3  The acid catalyzed decomposition of chlorite,  4C10 " + 2H* -> Cl" + C10 2  2  +CIO3'  11  +H 0 2  (2.8)  Literature Review 3.  The reaction between chlorite and hypochlorous acid.  HC10 + C10 "-*H + Cr + C103" t  2  (2.9)  To examine the contribution of reaction 2.7 to chlorate formation, fully bleached pulp was treated with chlorine dioxide. Since no chlorate was detected, Ni et al. (1993) concluded that it is the oxidation products of chlorine dioxide that result in chlorate formation and the disproportionation of chlorine dioxide, shown in equation 2.7, is negligible. To quantify the contribution of chlorite decomposition to chlorate formation, aqueous sodium chlorite was subjected, without pulp, to a pH of 2.5 and a temperature of 45°C for up to 150 minutes. Since no chlorate was detected in the experiments, it was concluded that the chlorate formedfromreaction 2.8, the acid catalyzed decomposition of chlorite, could be neglected. This agrees with the summary of information presented by Nilsson and Sjostrom (1974) which concluded that decomposition is only significant if chlorine is present and the pH is greater than 5. To study the contribution of reaction 2.9 to chlorate formation, Ni et al. (1993) used sulfamic acid, a known hypochlorous acid scavenger (Bergnor et al. 1987), to effectively remove the hypochlorous acidfromsolution as it was produced in the bleaching reactions. Since no chlorate was detected, they concluded that the reaction between hypochlorous acid and chlorite must be responsible for chlorate formation.  12  Literature Review The mechanism of the reaction between chlorite and hypochlorous acid, shown in Figure 2.2, wasfirstproposed by Taube and Dodgen (1949) and later modified by Emmenegger and Gordon (1967). A discussion of the evidence supporting the reaction mechanism shown in Figure 2.2 is presented by Gordon et al. (1972). They conclude that the O2O2 reaction mechanism is consistent with the stoichiometry and rates reported in the literature for the reaction of C10 " with HOC1 and CI2. 2  The formation of chlorate is the result of a series of competing reactions. Both of the consecutive reactions 2.10 followed by 2.12, and 2.11 followed by 2.13, result in the formation of the dichlorine dioxide intermediate (CI2O2). The dichlorine dioxide intermediate then reacts further to form either chlorine and chlorine dioxide through reaction 2.14, or chloride and chlorate through reaction 2.15.  Ch+ClolHOCl +  (Cl  CI  ClOi--^{HO  ClOif--^(ci  (HO-Cl-ClOiY  (2.10)  -^->(C7 CI ClOi)~  -^->(C7  CI  aoi)+cr  (2.12)  CI02) + OH'  (2.13)  2(Cl ClOi)-- ^ - > C / 2 + 2C/0 (Cl-ClOi) + H20 —^cr  (2.11)  CIO2)'  +CIO3-  (2.14)  2  +2H  +  (2.15)  Figure 2.2: The mechanism of chlorate formation during chlorine dioxide bleaching.  13  Literature Review Gordon et al. (1972) examined the existing literature regarding the reactions shown in Figure 2.2 and found agreement between this reaction mechanism and the observed kinetics and stoichiometry of chlorite (CIO2) reactions. In the discussion they suggest that the following conditions will tend to favour chlorine dioxide regeneration through equation 2.14 over chlorate formation via reaction 2.15:  •  Shifting the chlorine equilibrium to favour elemental chlorine Gordon et al. (1972) report that increasing the hydrogen ion concentration or increasing the chloride ion concentration tends to favour chlorine dioxide re-generation over chlorate formation. Since, according to equation 2.16, increasing either the hydrogen or chloride ion concentration tends to favour elemental chlorine over hypochlorous acid, they conclude that shifting the chlorine equilibrium towards elemental chlorine reduces chlorate production.  Cl +H 0 -> HOCl + Cf + I f 2  (2.16)  2  14  Literature Review •  High dichlorine dioxide intermediate concentration Emmenegger and Gordon (1967) report that ki is greater than k . As a result, shifting the chlorine equilibrium shown in 2  equation 2.16 towards elemental chlorine by decreasing pH or increasing the chloride ion concentration will result in higher concentrations of the intermediate dichlorine dioxide (C1 0 ) 2  2  species. Furthermore, increasing the hydrogen ion concentration will tend to increase the rate of reaction 2.13, which will also increase the concentration of C1 0 . As a result of these 2  2  observations, Gordon et al. (1972) conclude that increasing the concentration of the C1 0 intermediate tends to decrease 2  2  chlorate formation relative to chlorine dioxide re-generation.  2.1.3. The Impact of Operating Conditions on Chlorate Formation Many studies have been performed into the impact of delignification stage operating conditions on the formation of chlorate. The existing information is summarized in sections 2.1.3.1 through 2.1.3.11.  2.1.3.1. Chlorine Dioxide Substitution As part of a practical study of the impacts of chlorination stage recycle on effluent quality, Fair (1991) examined chlorate production as a function of substitution. The results of a modelfitto Fair's data are illustrated in Figure 2.3. In general, chlorate production increased with substitution. At high levels of substitution, however, a local  15  Literature Review maximum was encountered and the chlorate production began to decrease with further increases in substitution. Farr's results are supported by the summary of existing information presented in Table 2.1. In four of the investigations reported in the Table, a maximum in chlorate production occurred between 70 and 85% substitution. In these instances, chlorate production then decreased as substitution was further increased. 2.1.3.2. Order of Oxidant Addition Figure 2.3 also illustrates the findings of Farr (1991) with respect to the impact of order of oxidant addition, or mode, on chlorate production. The Figure shows that the impact of mode on chlorate formation is a function of substitution. The DC mode gave the lowest chlorate formation up to 68 percent substitution. At substitution of greater than 68 percent, the CD mode produced the least chlorate. Simultaneous addition, C+D mode, gave a midrange chlorate production for all substitution cases.  16  Literature Review  -  (D+C)  (GO)  -  - (DC)  |  y ~~ »^  •a *  5  •2  4  I. 0)  •a O  2  * 1 0  2  0  3  **  0 '  4  0  5  0  6  0  7  0  8 0  9 0  Chlorine Dioxide Substitution (%)  Figure 2.3: The impact of substitution and mode of addition on chlorate production. (Data taken from Fair 1991) Table 2.1 shows that there is limited agreement as to the effect of mode of addition on chlorate production. The addition of chlorine dioxide before chlorate gave the lowest production in all cases except one at 90% substitution (Bergnor etal 1987).  Further evidence for the crossing of the chlorate production curves is provided by Nilsson and Sjostrom (1974). In studying the effect on chlorate formation of small amounts of chlorine added to thefirstchlorine dioxide stage, they found that less chlorate was formed when the chlorine was added before, as opposed to with, the chlorine dioxide. While not directly applicable to the chlorination stage, bleaching in a D stage with 2, 4, and 6 % chlorine substitution as studied by Nilsson and Sjostrom (1974) is somewhat analogous to bleaching in a C stage with 98, 96, and 94 % chlorine dioxide substitution. This tends to corroborate the finding that at high levels of chlorine dioxide substitution, a CD mode yields less chlorate (Bergnor et al. 1987).  17  Literature Review 2.1.3.3. Reaction Time Reeve and Wieshar (1991) found that at 50°C chlorate formation was complete after approximately 30 to 40 minutes. This is consistent with thefindingsof Ni et al. (1993) shown in Figure 2.4. Although chlorate formation at 25 °C was complete within 5 minutes, chlorate formation at 45°C required approximately 40 minutes to reach completion. The longer retention time required to reach thefinalchlorate concentration at 45°C was attributed to increased decomposition of the chlorite in solution, according to Equation 2.8. With increased decomposition of chlorite, more chlorate was formed and a longer time was required to reach the final chlorate concentration at 45°C. Further data is provided by Kramer (1972), who found that chlorate formation was independent of the reaction times used of 15, 30, and 60 minutes.  2.1.3.4. Bleaching Temperature Figure 2.4 illustrates the effect of temperature on chlorate formation (Ni et al. 1993). As the temperature was increasedfrom25°C to 45°C, the percent of the chlorine dioxide charged as chlorate approximately doubled. After 150 minutes, the chlorite accounted for 21.4 percent and 0.9 percent of the atomic chlorine in solution at 25 °C and 45°C, respectively. As discussed in 2.1.3.3, the increase in chlorate formation at 45°C is attributed to the decomposition of chlorite through Equation 2.8.  18  Literature Review  Reference  Kramer (1972) Reeve and Rapson (1980)  Unbleached kappa number  Kappa factor  Location of relative maximum  Mode  (%eq.Cl )  (ml) 32.3  0.20  C+D  -  32  0.16  CD C+D DC  , 70  b  2  70  0.20  Bergnor et al. (1987)  29.3  . Axegard (1987) Germgard and Karlsson (1988) Liebergott et  17.2  0.20  17.2  0.18  31.2  0.22  36.2  0.19  31.0  0.21  al.  C+D>CD>DC  CD  (1990) Histed et al. (1991) Farr (1991)  Relative order of chlorate production  C+D  75  DC C+D DC C+D DC  70 70 75 85  CD C+D DC DC  a  0  19  5.2  C  3.0°  CD=ODC  4.6  -  -  3.8  -  substitution < 68% CD>C+D>DC substitution > 68% DOC+D>CD  -  From graphical results at 25°C and pH2. Converted from a permanganate number of 21 ml. Pulp moisture not specified  3.2  2.2°  Table 2.1: A summary of the literature concerning chlorate formation in the chlorination stage, (expanded from Farr 1991)  b  (kg/adt) 2.5*  C+D>DC  -  CD C+D DC  substitution <60% CD>C+D>DC substitution 60-90% C+D>CD>DC substitution >90% C+D>DOCD C+D>DC  Maximum chlorate production  5.0  Literature Review  2.1.3.5. Delignification Stage pH Generally, chlorate formation is known to decrease with decreasing pH (Wartiovaara 1985). Ni et al. (1993) explained the impact of pH on chlorate production in terms of the chlorate formation mechanism. As seen in Figure 2.5, the chlorine/hypochlorous acid equilibrium is a strong function of pH. Lowering the pH favours molecular chlorine over hypochlorous acid which, according to the discussion in Section 2.1.2.2, results in reduced chlorate formation.  20  Literature Review  Figure 2.5: The aqueous chlorine equilibrium distribution (From Smook 1992)  2.1.3.6. Chlorine Contamination of the Chlorine Dioxide Charge Reeve and Weishar (1991) studied the impact of cWorine contamination on chlorate formation for pure chlorine dioxide delignification. They found that runs performed at 95 percent substitution produced the same quantity of chlorate as runs performed at 100 percent substitution and, therefore, chlorine contamination was not a concern. The results were the same for oxygen pre-bleached and unbleached kraft pulps.  Bergnor et al. (1987) investigated the impact of chlorine contamination at high levels of substitution. With 70 percent substitution and a DC mode of addition, having 10 % of the available chlorine in the chlorine dioxide solution as chlorine caused a 0.5 kg/adt decrease in chlorate formation. At 30 % percent substitution, the difference between using pure chlorine dioxide and that containing some chlorine was not significant.  21  Literature Review 2.1.3.7. Chlorination Stage Recycle As part of a larger investigation of the impact of chlorination stage recycle on effluent quality, Farr (1992) investigated the response of chlorate concentration to recycle. Fair's experiments were performedfroma practical perspective, using a laboratory scale simulation of the chlorination stage. It was found that recycle had no significant effect on the production of chlorate. An earlier study on the effect of effluent recycle on chlorate production was done by Reeve et al. (1980). Chlorination effluent recycle was simulated by addingfirststage filtratefroma pulp mill to a series of batch laboratory bleaching trials. Both C+D and DC modes of addition were investigated at 70 % substitution. Chlorate production was found to be unaffected by recycle levels of up to 10 percent total organic carbon (TOC) on pulp. Approximate calculations show that the maximum recycle of chlorinationfiltratepossible industrially corresponds to 2 percent TOC on pulp. The mill effluent was concentrated by vacuum evaporation enabling this value to be surpassed.  2.1.3.8. Pulp Consistency The production of chlorate during chlorine dioxide bleaching is decreased as consistency increases (Reeve and Weishar 1991, Ni 1992). Ni etal. (1993) explained the impact of consistency using the chlorate formation mechanism discussed in Section 2.1.2.2. As consistency is increased, the concentrations of all species, including the dichlorine dioxide intermediate, are increased. Since higher dichlorine dioxide  22  Literature Review concentrations favour chlorine dioxide regeneration over chlorate formation, less chlorate is formed at higher consistency.  2.1.3.9. Oxidant Charge The initial oxidant charge is a function of the kappa number of the incoming pulp; pulps with higher incoming kappa numbers require higher oxidant charges to reach a target bleached kappa number. Oxygen delignified pulps usually have a kappa number in the range of 16 to 19, while unbleached pulps tend to have kappa numbers closer to 30. As a result, higher oxidant charges are used for unbleached kraft pulps. Thefractionof the oxidant charge that is lost as chlorate is higher for oxygen prebleached pulps (McDonough et al. 1985). Ni et al. (1993) explained the higher fraction of chlorate formation with oxygen delignified pulps according to the chlorite/hypochlorous acid formation mechanism. At the reduced kappa number, less lignin is available for the fast reaction with the chlorine/hypochlorous acid equilibrium pair. This means that more chlorine/hypochlorous acid is available for chlorate formation, and a higherfractionof chlorine dioxide is lost as chlorate.  Naturally, unbleached kraft pulp will require a much higher oxidant charge than an oxygen delignified pulp and the chlorination stage effluent will contain higher loadings of chlorate.  2.1.3.10.Chloride Ion Concentration Several studies have reported that increasing the chloride ion concentration in the bleaching liquor reduces chlorate production (Germgard, Teder and Tormund 1981,  23  Literature Review Reeve and Weishar 1991). Ni et al. (1993) concluded that this effect is due to the shifting of the chlorine/hypochlorous acid equilibrium, shown in equation 2.16, towards elemental chlorine. Since the chlorate formation reaction is believed to result from reaction of chlorite and hypochlorous acid, this favours chlorine dioxide regeneration over chlorate formation (see section 2.1.2.2).  2.1.3.11. The Impact of Chlorate Formation on the Receiving Environment Chlorate is a known fungicide and herbicide and when contained in BKME, chlorate has the potential to seriously affect the aquatic plant life in the receiving environment (Isensee et al.. 1973). Chlorate discharges in BKME in Sweden have been linked to the disappearance of Bladder Wrack (Fucus vesiculosis), a brown algae, in the Baltic Sea (Rosmarin et al.. 1986, Germgard 1989). In one instance, the Bladder Wrack population within a 12 km area surrounding the bleached kraft mill outfall was destroyed. Since the areas 2  containing brown algae, and especially Bladder Wrack, are important breeding areas for a variety of aquatic organisms, the disappearance of the algae impacts heavily on the surrounding ecosystem.  Rosmarin et al.. (1986) studied the impact of chlorate on a bladder wrack population that was transplanted to an artificial environment. The results indicate that chlorate hinders the growth of bladder wrack at concentrations as low as 20 ug/1, which corresponds to a dilution factor in the pulp mill effluent of 1000 to 2000. Simulation of  24  Literature Review the conditions and chlorate concentrations found in Swedish BKME receiving environments showed that it takes approximately 2 months for the chlorate to disappear. The impact of chlorate onfreshwaterriversystems was studied by Perrin and Bothwell (1992). It was found that chlorate concentrations typically found in pulp mill receiving environments did not reduce the specific growth rates or the population distribution of freshwater diatoms, and that chlorate dischargesfrombleached kraft mills (BKM) would not affect the riverine algal community.  2.1.4. Reduction of Chlorate during Effluent Treatment Germgard (1989) investigated the potential of reducing chlorate produced in the bleach plant through biological or chemical treatment. Anaerobic biological reactions during secondary treatment of pulp mill effluent can reduce chlorate to chloride. Aerated lagoon treatment systems often have regions of low oxygen concentration, especially in the sediment at the bottom of the lagoon. These regions provide anaerobic zones where facultatative organisms can digest and reduce the chlorate contained in the effluent stream. If the aerated lagoon is operated with no initial aeration at a pH of 7, chlorate degradation occurs within 5 to 10 hours.  The elimination of chloratefromeffluent treated in aerated lagoon systems is supported by studies performed by Munro et al. 1990, and Pryke et al. 1993. These studies found that the chlorate levels in bleached kraft mill effluent (BKME)frommills employing aerated lagoon secondary treatment systems were below detection limits.  25  Literature Review Although, chlorate appears to be effectively treated in aerated lagoons, chlorate reduction does not occur in activated sludge systems. Activated sludge systems do not provide the facultative zones required for biological reduction of the chlorate contained in the effluent stream. As a result, effluent dischargesfromactivated sludge systems may contain chlorate. The use of chemical reducing agents is another viable treatment for chlorate. The reduction of chlorate is most easily performed using sulfur dioxide since it is readily available at bleached kraft mills. Other chemical treatments include iron chloride or iodine. The application of 3 mol of sulfur dioxide for each mol of chlorate, at pH 2 and 50 to 60°C with a retention time of at least one hour can completely eliminate the chlorate contained in the bleach plant effluent. Germgard (1989) suggests that a mill effluent pipe is a suitable reactor, provided the retention time is sufficient. 2.2. Vanadium Pentoxide Vanadium pentoxide was the catalyst used for the oxidation of the chlorate residual in the current work. The oxidation states of vanadium rangefrom-I to +V (Lee 1991). Vanadium pentoxide is in the +V oxidation state. Vanadium pentoxide is a brick red powder that is soluble in aqueous solutions at 0.07 g/1. It is completely soluble in strongly acidic and basic solutions.  26  Literature Review As shown in equation 2.17, vanadium in the +V oxidation state is a mildly good oxidizing agent with a standard reduction potential of 1.000 volt (Lee 1991).  (OH)++2H + e ' -+V0 + 3H 0 +  1  6 = 1.000 V  2+  2  0  (2.17)  When involved in oxidation reactions, vanadium is reduced to the +IV state. In aqueous solutions, vanadium in the +IV oxidation state exists as the stable vanadyl ion (V0 ). Mildly to strongly acidic solutions of the vanadyl ion are stable for months. 2+  With the standard reduction potential of 0.361 illustrated in equation 2.18, the vanadyl ion is a poor oxidizer (Lee 1991).  0  2+  +2H +e ~ ^>V* +H 0 +  8 = 0.361 V  l  2  0  (2.18)  The following sections discuss the catalytic reactions, the kinetics of vanadium oxidations, and the aqueous chemistry of the vanadium species.  2.2.1. Vanadium Pentoxide Catalysis Vanadium pentoxide is an important catalyst for industrial oxidation reactions. Among the important reactions that are catalyzed by vanadium pentoxide are the following (Lee 1991, Sidgewick 1950):  •  oxidizing sulfur dioxide to sulfur trioxide in the contact process for sulfuric acid production,  •  oxidizing naphthalene to phthalic acid,  •  oxidizing toluene to benzaldehyde,  27  Literature Review •  sulfonizing hydrocarbons and pyridine,  •  oxidizing stannous salts by stannous or chloric acid,  •  oxidizing cyclic organic compounds with hydrogen peroxide, and  •  oxidizing sugar with nitric acid.  Studies of these reactions show that pentavalent vanadium reactions predominantly proceed through 1 electron transfer reactions (Waters and Littler 1965). Thus, vanadium pentoxide catalysis results infreeradical formation.  2.2.2. Vanadium Pentoxide Oxidations of Organic Compounds Littler and Waters (1959) performed a semi-qualitative mechanistic study of the oxidation of acids, aldehydes, ketones, alcohols, phenols, ethers, olefins, and nitrogen compounds with vanadium pentoxide and acidic aqueous solutions. The induced polymerization of vinyl cyanide and the induced reduction of mercuric chloride were used to indicatefreeradical production during the oxidation reactions. In all cases there was evidence offreeradical activity. They concluded that although competing ionic reactions exist, thefreeradical mechanism predominates in the oxidation of organic compounds with vanadium pentoxide.  In a series of investigations of vanadium pentoxide oxidations of organic compounds in acetic acid solutions, Radhakrishnamurti and Devi (1975), Radhakrishnamurti and Pati (1970), and Radhakrishnamurti and Pati (1969a, 1969b) concluded that the ionic oxidation mechanism predominated.  28  Literature Review It appears that the two competing reaction mechanisms (i.e.freeradical versus ionic) are a strong function of the solution conditions. 2.2.3.Kinetics of Vanadium (V) Oxidations Several studies of the kinetics of pentavalent vanadium oxidations of organics have been performed by Radhakrishnamurti and co-workers. Most of Radhakrishnamurti's work was performed in aqueous acetic acid solutions, and so has no direct relevance to the current work. A summary of the reaction conditions and rate constants found by Radhakrishnamurti is presented in Table 2.2. The Table is intended to illustrate the wide range in rate constant values obtained for vanadium oxidations of organic compounds. The rate of vanadium oxidations of organic substrates in acetic acid solutions is a strong function of temperature, acidity, and solution composition. 2.2.4.  Aqueous Solutions of Vanadium in the +TV and +V Oxidation State  The two oxidation states of vanadium that are of interest in the current work are the +IV and the +V states. The +IV oxidation state is present in aqueous solution as the stable V 0 species, 2+  commonly known as the vanadyl ion. The V 0  2 +  exists as the VOfHjOV species (Clark 1973).  29  is soluble in aqueous solution where it  Literature Review  Reference  Substrate  Radhakrishnamurti and Devi (1975) Radhakrishnamurti and Panda (1970)  Cyclic Ketones  Radhakrishnamurti Pati (1969a) Radhakrishnamurti and Pati (1969b)  Cyclic Alcohols  Solution Composition  Temperature  Acidity  (v % Acetic Acid)  (°C)  (N-H S0 ) 1.0 2.0 1.0 0.025/0.075 0.025 0.025 0.025 4.1 4.1 5.4 8.1 10.8 8.1 8.1 8.1  60 60 30 40 30 40 40  10 30  Phenols 50 70 50  60 Halogenated Toluenes  50 70 50 70  50  70  2  4  Table 2.2: A summary of the literature concerning vanadium pentoxide oxidations of organic substrates.  a  b  c  Rate constant values shown are for phenol. Rate constant values shown are for cyclohexanol Rate constant values shown are for chlorotoluene.  30  Second Order Rate Constant (l-mor^min" ) 1  0.06-0.29 0.20-1.14 0.14-0.75 0.30/0.50* 0.53 0.59 0.84 0.13 0.42 1.79 0.003° 0.07 0.38 0.03 0.37 b  Literature Review  o  1V2O5'  1  \ \  . 1 1  V2O7 "  '  »  4  1  %  r •  HfeVioCV*  1 1 1 1 1  I  1 1 1  Polyvana dates  \  \  I  %  'r  I  %  *  1 1  '  «. VO2*  HVO4 " 2  \ ,  5  H2V04"  •  •  1 •  6  14  13  12  11  10  9  8  7  6  5  4  3  2  1  0  pH Figure 2.6: The distribution of vanadium (+5) in aqueous solution. (Data taken from Clark 1973) Figure 2.6 illustrates the complex aqueous chemistry of the +V oxidation state of vanadium. The species distribution is a complex function of the total vanadium concentration and the pH of the solution (Lee 1991, Clark 1973). At very high pH values, the vanadium is in the form of the orthovanadate ion  VO4 ". 3  As the pH is lowered, the ions  polymerize to form polyvanadates. As the pH is lowered below 3, the polymers dissolve and the pentavalent vanadium is present in the V02 form. +  2.3. Vanadium Pentoxide "Activation" of Chlorate Vanadium pentoxide has been used in previous studies to "activate" chlorate in acidic solutions.  31  Literature Review Morioko (1981) patented the use of vanadium pentoxide, alone or in combination with other metal complexes, to reduce the acidity required to produce chlorine dioxide through "activation" of the chlorate in the generator solution. Marpillero (1957) patented the use of a vanadium pentoxide "activated" chlorate bleaching process that lowered the acidity required for chlorate bleaching. The catalytic bleaching process was operated at pH 2 and a temperature of 70°C for 3 to 4 hours. Marpillero (1958) claimed that brightness gains of 5 to 15 percent over comparable chlorine and hypochlorite bleaching processes could be realized. Investigation of the pulp properties resulting from the use of "activated" chlorate in pulp bleaching showed that under practical bleaching conditions with a pH of 2, a temperature of 70°C and a four hour retention time, there was serious hydrolytic attack on cellulose (Rapson et al. 1959). The losses in cellulose integrity have precluded the use of "activated" chlorate on an industrial scale.  Deutsch et al. (1979) investigated the potential of using vanadium pentoxide to "activate" the chlorate residual in chlorine dioxide brightening of kraft pulp. They found that a vanadium pentoxide charge of approximately 0.008 weight percent on oven dry pulp provided the optimum chlorate "activation". In their patent, Deutsch and Shoemaker (1977) noted that the optimum temperature for the chlorate "activation" is between 70°C and 80°C and that brightness increases with increasing temperature, longer retention times, higher catalyst charge, and lower pH.  32  Literature Review  V * + lignin —> oxidized lignin + V * 4  6V* + CIO3- + 6H -> 6V* + CI- + 3H 0 +  2  Figure 2.7: The mechanism of the vanadium catalyzed bleaching reaction. Deutsch et al. went on to compare the vanadium catalyzed bleaching system to conventional chlorine dioxide brightening. Brightness reversion and residual chlorate concentrations were reduced at all pH values. They found that the addition of catalyst increased brightness at pH valuesfrom1 to 3, but no brightness gain was noted at higher pH values. Since the optimum pH for chlorine dioxide brightening is around 4, they concluded that there is little incentive to use vanadium pentoxide to recapture the bleaching power lost to chlorate in the brightening stages. 2.3.1. The Mechanism of the Catalytic Reaction The catalytic pulp bleaching reactions are believed to proceed through the vanadium cycling mechanism illustrated in Figure 2.7 (Deutsch et al. 1979, Rapson et al. 1959, Marpillero 1958).  The vanadium transfers electronsfromthe chlorate in solution to oxidize the pulp. It does not result in regeneration of chlorine dioxidefromchlorate; rather, the bleaching power lost to chlorate is recaptured as pentavalent vanadium oxidations of the pulp.  33  3.  EXPERIMENTAL 3.1. Experimental Design  The experiments were designed using factorial design at two levels. The entire experimental program consisted of two sets of 2 factorial designs. Thefirstset of 4  experiments focused entirely on the reduction in chlorate concentration obtained. The second experimental set was designed to examine the impact of the catalytic bleaching system on the bleaching efficiency and the pulp properties in addition to the reduction in chlorate concentration. A discussion of the choice of factor variables, the response variables used, the levels chosen for each factor variable setting, and a summary of the experimental runs performed is provided in the following sections. A detailed discussion of factorial design at two levels is beyond the scope of this report. The reader is referred to the paper in Appendix C for an overview of factorial design methods.  3.1.1. Factor Variables The operating variables which are known to impact on the quantity of chlorate formed are pH, temperature, the degree of chlorine dioxide substitution, the chloride ion concentration, the size of the oxidant charge, the consistency of the pulp suspension, and the type of pulp used. Catalytic processes are strongly affected by the concentration of both the catalyst and the substrate, and the temperature of the reaction.  34  Results and Discussion Since the pumping capabilities of most pulp mills in Canada limit the chlorination stage consistency to 3.5 % (Lee 1994), this consistency was used throughout the experimental program. The impact of chloride ion concentration on chlorate formation is small and it was not considered in the experimental design. Environmental concerns have caused a decrease in the oxidant charge used in the chlorination stage. The oxidant charge is commonly expressed as the kappa factor, which is determined according to equation 3.1.  Unbleached kappa number x kappa factor = % Equivalent Cl Applied 2  (3.1)  When the kappa factor is increased beyond 0.15 to 0.20, laboratory results have shown that the concentration of 2,3,7,8-tetra'chlorodibenzo-p-dioxin (2,3,7,8-TCDD) and 2,3,7,8tetrachlorodibenzofuran (2,3,7,8-TCDF) increase sharply (Berry etal. 1989). To simulate conditions under which the formation of 2,3,7,8-TCDD and 2,3,7,8-TCDF are minimized, a kappa factor of 0.15 was chosen for the experiments performed.  The remaining factors affecting chlorate formation and the activity of the catalyst were used in the experimental design. The first factorial design (Experimental Program 1) included pH, temperature, the degree of substitution, and the catalyst charge. The factor variables  35  Results and Discussion used in the second set of experiments (Experimental Program 2) were the degree of substitution, reaction time, catalyst charge, and temperature. Mode was not included as a factor variable because of practical difficulties. The addition of bleaching chemicals in the bag bleaching method is cumbersome, and the timing of sequential oxidant addition would be extremely difficult to control.  3.1.2. S u m m a r y of R u n s Tables 3.1 and 3.2 illustrate the factor level settings used in Experimental Programs 1 and 2, respectively. Thefirstset of factorial runs performed included the degree of chlorine dioxide substitution, pH, temperature, and catalyst charge as factor variables. The results of Experimental program 1 were used to identify the effect of each of the factors on thefinalchlorate concentration. These results were used to design Experimental Program 2. The factors included in the second design were the degree of chlorine dioxide substitution, temperature, time, and catalyst charge.  In addition to the factorial design runs outlined in the tables, blank runs were performed in order to compare the overall performance of the catalyzed bleaching runs to runs without catalyst. A summary of these runs is shown in Table 3.3. The purpose of these runs was to provide a comparison between the impacts of the uncatalyzed and catalyzed bleaching runs on the chlorate concentration, as well as the properties of the pulp produced. Two runs were also performed at 45 °C to provide a benchmark for comparing the catalyzed system with conventional delignification.  36  Results and Discussion 3.1.3. Response Variables The purpose of Experimental Program 1 was to investigate the impact of the degree of substitution, pH, temperature, and catalyst charge on thefinalchlorate concentration. Chlorate concentration was the only response variable. The intent of Experimental Program 2 was to follow up thefirstexperimental program with an investigation of the impact of the catalytic system on the bleached pulp properties. Thus, in addition to chlorate formation, the response of the brightness, kappa number, viscosity, and pulp strength and optical properties was measured for each of the runs shown in Table 3.2.  37  Results and Discussion  Run  Substitution  Catalyst  Temperature  PH  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20  0  0  0  0  -  +  -  -  +  +  +  + + +  -  0  0  0  + 0  -  + + +  -  +  + +  -  +  -  + +  -  -  +  +  -  + + +  0 +  0  0  0  +  + +  +  -  +  +  0  0  0  -  +  -  + 0  -  Table 3.1: Factor level table for Experimental Program 1.  38  -  +  -  Results and Discussion  Run  Substitution  Catalyst  Temperature  Time  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19  0  0  0  0  + +  -  +  +  -  +  -  -  -  +  -  +  +  -  +  -  -  +  +  -  -  +  0  0  0  0  +  + + + +  +  +  -  +  -  -  -. -  + +  -  -  +  -  + +  -  + + + +  +  -  0  0  0  0  -  +  +  Table 3.2: Factor level table for Experimental Program 2.  Substitution (% eq. chlorine)  Temperature (°C)  Retention Time (min)  50  40 70 80  120 120 120  75  75  120  100  40 70 80  120 120 120  Table 3.3: A summary of the uncatalyzed bleaching runs.  39  Results and Discussion 3.2. Materials and Methods 3.2.1. The Bleaching Runs The bleaching runs were performed using the bag bleaching method (Hatton 1966, Hatton 1967). A detailed discussion of the bleaching method used is provided in Appendix A. The required amounts of pulp, water, and vanadium pentoxide catalyst (added as an aqueous solution) were sealed in the bag. A clamp was placed across the bag to create an empty compartment. The chlorine and chlorine dioxide were charged, via syringe, into the empty section and the bag was resealed. The bleaching reaction was initiated by removing the clamp and thoroughly mixing the pulp and bleaching chemicals. Once mixed, the bag was placed in a constant temperature bath. The bag was removedfromthe water bath and kneaded every 20 minutes, to provide mixing of the pulp and oxidants.  3.2.2. The Pulp Source In order to simulate a contemporary bleaching reaction, an oxygen delignified pulp was used for the bleaching reactions. A never dried spruce-pine-fir furnish was provided by Cariboo Pulp. The pulp had an initial Kappa number of 17.1, which is in the normal range for an oxygen delignified pulp.  3.2.3. Vanadium Catalyst Preparation The vanadium catalyst was charged as an aqueous solution of 0.05 g/1 vanadium pentoxide.  40  Results and Discussion The solubility of vanadium pentoxide in water at 20°C is 0.08 weight percent (CRC 1982). Vanadium pentoxide is completely soluble in acid and alkaline solutions. Although the solution used in the bleaching runs was below the saturation limit of vanadium pentoxide, the dissolution of vanadium pentoxide in water did not occur at an appreciable rate. As a result, the vanadium pentoxide solution was prepared using water adjusted to pH 12 with sodium hydroxide.  3.2.4. Chlorine Solution Preparation An aqueous chlorine solution was formed by absorbing gaseous elemental chlorine in deionized water. The concentration of the chlorine solution was determined using the iodometric titration found in CPPA standard method J.22P.  3.2.5. Chlorine Dioxide Generation Chlorine dioxide was generated by acidifying sodium chlorite with ION sulfuric acid. The gaseous chlorine dioxide produced was passed through twoflaskscontaining dilute sodium chlorite solutions to convert any entrained chlorine to chlorine dioxide. The pure chlorine dioxide was absorbed in cold deionized water to produce the chlorine dioxide solution used in the bleaching runs. This solution was stored for up to one month. A detailed discussion of the chlorine dioxide generation system is presented in Appendix B.  To determine the chlorine dioxide concentration, a 10 ml aliquot of chlorine dioxide was pipetted below the surface of a mixture of 40 ml of water and 10 ml of 10% potassium iodide solution. The chlorine dioxide concentration was then determined using the  41  Results and Discussion iodometric titration found in the CPPA standard method J.14P (1966-) for chlorine dioxide plant analysis.  3.2.6. Bleached Pulp Property Testing (CPPA 1966-) The pulp property tests for the bleached pulps were performed in accordance with the CPPA standard testing methods (CPPA 1966-). All of the testing was performed at the PAPRICAN Vancouver laboratory. Table 3.4 shows the standard methods and the instruments used in the pulp and paper tests.  3.2.7. Bleaching Liquor Analysis The residual oxidants potentially present in the bleaching liquor were chlorine, hypochlorous acid (HOCl), chlorine dioxide, chlorite (C10 "), and chlorate. 2  Pulp Property Test  CPPA standard test method  Cupriethylene diamine Viscosity Kappa Number  G.24P  Brightness Tear strength  E.l D.9  Breaking Length Zero span tensile strength  D.6H Useful method D.27U  Burst strength Bulk Opacity  D.8 D.5H E.2  G.18  Instrument Used  Radiometer KTS1 Kappa number system Technibrite Micro TB-1C TMI Tear Tester 800g Pendulum Instron model 4202 Pulmac - The Troubleshooter The Mullen Tester TMI series 400 tester Technibrite Micro TB-1C  Table 3.4: A summary of the pulp testing methods used in this study.  42  Results and Discussion The reactions between pulp and chlorine, hypochlorous acid, and chlorine dioxide are known to proceed rapidly. The chlorite produced in the reactions between cellulosic pulp and chlorine dioxide is completely reacted after 2 hours at 45°C. Since temperatures higher than 70°C were used in this work, the chlorine, chlorine dioxide, chlorite and hypochlorous acid were completely consumed before the reaction was terminated. This was confirmed using the analytical methods found in Rapson and Anderson (1966).  Chlorate was the only residual oxidizing species detected in solution following the bleaching runs.  3.2.7.1. Chlorate Analysis The chlorate concentration was determined using the method given by Rapson and Anderson (1966). The accuracy of the method was verified using chlorate solutions in the concentration range found in the bleaching liquor.  At the completion of the desired reaction time, the bags were cooled under cold running tap water for at least 10 minutes. The bleaching liquor was then separatedfromthe pulp and screened. Replicate 100 millilitre samples were pipetted into 500 ml Erlenmeyerflasks.To each sample was added 10 ml of 10 % NaBr and 100 ml of concentrated hydrochloric acid (Fischer scientific reagent grade). The sample was promptly stoppered and allowed to stand for at least 10 minutes.  At the completion of the 10 minute period, the iodometric titration was performed. Due to the oxidation by air of iodide to iodine in highly acidic solutions, a small amount of sodium bicarbonate (approximately 300 mg) was added to the sample prior to iodide addition. This  43  Results and Discussion displaced the air in theflaskthrough evolution of carbon dioxide (Vogel 1978). 10 ml of 10 % potassium iodide solution was added to theflaskand the iodine liberated was titrated immediately, to the starch end point, with 0.0100 N sodium thiosulphate solution.  3.2.7.2.  The Air Oxidation of Iodide to Iodine during Chlorate Determination  A major source of error in the iodometric determination of chlorate concentration is the air oxidation of iodide to iodine (Skoog and West 1980, Vogel 1978).  The reaction proceeds as follows:  41" + 0 + 4H+ - » 2  + 2HjO  (3.2)  The rate of this reaction becomes appreciable in solutions of greater than 0.4 to 0.5 normal acid. Since the chlorate titration is performed in 8N acid, the oxidation of iodide by this mechanism is an important source of error. The oxidation of iodide is minimized through the use of the bromide intermediate in the chlorate titration. The standard half cell reduction potentials for bromine and iodine are 1.08.73 V and 0.5355 V, respectively, indicating that bromide is less easily oxidized than is iodide. As a result, the air oxidation of bromide is negligible in the determination of chlorate concentration. When the iodide is added to the solution, it is readily oxidized to iodine by the bromine in solution. This iodine must be titrated immediately to irunimize air oxidation of iodide. The oxidation of iodide by oxygen in the surrounding atmosphere during the iodine determination is minimized by adding a small amount (300 mg) of sodium bicarbonate to the  44  Results and Discussion solution at various times throughout the titration. This results in the evolution of carbon dioxide and reduces the oxygen concentration in the atmosphere over the solution.  45  4. RESULTS AND DISCUSSION The purpose of the experimental program was to investigate the impact of adding vanadium pentoxide to the delignification stage. The investigation was performed in two stages, each consisting of a 2 factorial design at two levels. The two experimental programs are 4  outlined in Table 4.1.  Thefirstset of experiments, Experimental Program 1, focused on the impact of vanadium pentoxide on the residual chlorate concentration. The factors considered were temperature, pH, substitution, and catalyst charge. As seen in Table 4.2, the results showed that maintaining high temperature and high catalyst charge was the most effective means of mimmizing the residual chlorate concentration. The starting pH and the degree of substitution had a large impact on the  Interpreting Factorial Design Data The basis of factorial experimental design is the development of a mathematical expression to fit the response of the system to changes in the factor variables. We used the following second order model to describe the system:  ;=1  The terms in the model are described as follows: Y: the measured response in the system. Xf. the factor setting for variable i. b{. the coefficient for the main effect of variable i. bib/, the coefficient for the interaction effect of variables i and j. (For a detailed discussion of factorial design, see Appendix C.) 46  Results and Discussion  Stage of the work  pH  Factor Levels (low/High) Catalyst Substitution Temperature Charge (weight %) (% Eq. Cl ) (°C) 0.001/0.01 50/100 70/80 2  Experimental program 1 Experimental program 1 Uncatalyzed runs  2/4  Time (min)  -  -  70/80  50/100  0.005/0.015  120/240  2/4  70/75/80  50/75/100  -  120  Table 4.1: A summary of the experimental programs. formation of chlorate in solution, but had little effect on thefinalchlorate concentration. Following thefindingsof Experimental Program 1, the second experimental program was designed. Experimental Program 2 was designed to optimize the reduction in chlorate concentration obtained, and to investigate the impact of chlorate "activation" on bleaching efficiency and pulp properties. Since it was shown to have little effect in thefirstphase, pH was droppedfromexperimental program 2. In its place, the impact of delignification stage retention time was added. To maximize the chlorate reductions obtained, the catalyst charge was increased for this set of experiments. The temperature could not be increased beyond 80°C due to unacceptable viscosity losses.  In many cases, the results of the catalyzed runs were compared to the resultsfromthe uncatalyzed runs. All such comparisons to catalyzed bleaching runs were made with uncatalyzed bleaching runs at the same temperature, substitution, and pH. In order to provide a reasonable comparison to standard industrial delignification, the uncatalyzed runs had retention times of 2 hours.  47  Results and Discussion A discussion of the impact of vanadium pentoxide addition on chlorate concentration, bleaching efficiency, and bleached pulp properties is presented in the following sections. The results are interpreted using factorial analysis.  4.1. The Response of the Residual Chlorate Concentration Since the chlorate concentration is the most direct measure of the catalyst activity, the final chlorate concentration was included in both factorial designs. The factor settings and the chlorate concentration responses for experimental programs 1 and 2 are summarized in Tables 4.2 and 4.3, respectively. A discussion of the impact of the factor variables on the chlorate in solution is presented in the following sections. The impact of the factor variables on the chlorate concentration is reported in two ways: the effect on thefinalchlorate concentration and the percent reduction in chlorate concentration. The percent reduction is a comparison of thefinalchlorate concentration in  Interpreting Main Effects in Factorial Design Mathematically, the main effect may be expressed as follows: Main Effect of Variable i = AY= bjkXt A large main effect shows that increasing the factor setting of variable /' has a large impact on the response of Y. If the main effect is negative, then increasing Xi tends to decrease Y;. i  For example, the data in 4.2 shows that the main effect of temperature on the residual chlorate concentration is -0.240 mM. Therefore, the statistical analysis showed that the average impact of increasing the temperaturefrom70°C to 80°C was to decrease the chlorate concentration by 0.240 mM.  48  Results and Discussion catalyzed and uncatalyzed bleaching runs performed under the same conditions. The percent reduction calculations were performed according to equation 4.1.  % Reduction = t  0  1  ^  -[dOslr, substitution, mag  substitution  x m  ( 4  1 }  [C10 ]x Substitution 3  4.1.1. The Effect of Catalyst Charge Tables 4.2 and 4.3 show that the catalyst charge has a large impact on the residual chlorate concentration..  The catalytic bleaching reactions are believed to proceed through the vanadium cycling mechanism illustrated in Figure 2.7 (Deutsch et al. 1979, Rapson et al 1959, Marpillero 1958).  Interpreting Interaction Effects in Factorial Design The effect of the interaction between two variables is calculated as follows: interaction effect =  Y +Y__ ++  2  Y _+Y_, +  2  Y++ and Y_ are the system responses with both variables at their high, and low factor settings, respectively. Y+. and Y.+ are the system responses with one variable at its high setting and the other at its low setting. If the interaction effect is 0, then the effect of the two variables is additive, and no significant interaction occurs. If the interaction effect is positive, then the two variables interact synergistically to increase the value of Y; if they interact negatively, then the variables interact to reduce the value of Y.  49  Results and Discussion Variable  Main Effects pH (at start of bleaching) Temperature  (°o Substitution (% eq. chlorine) Catalyst Charge (wt. % on oven dry pulp) Interaction Effects Temperature/pH Substitution/pH Catalyst/pH Substitution/Temperature Substitution/Catalyst Catalyst/Temperature  95% confidence interval  Chlorate Eliminated (mol %)  Factor Levels (Low/High)  Residual Chlorate (mM)  2/4  0.018  70/80  -0.240  50/100  0.019  -21.40  0.001/0.010  -0.220  52.45  -  0.020 -0.019 -0.005 0.018 0.052 0.040  3.33 21.40 -3.12 15.46 -10.12 0.09  95% confidence interval  -10.61 0.017  0.017  42.12  1.53  1.53  Table 4.2: Results of the factorial analysis of chlorate concentration for Experimental Program 1. The chlorate reduction reaction shown in Figure 2.7 is clearly a function of vanadium concentration. With increased catalyst charge, the rate of reaction increases and the final chlorate concentration is reduced.  The chlorate reduction is most efficient at catalyst charges between 0.005 % and 0.01 %. Comparing Tables 4.2 and 4.3 shows that increasing the catalyst chargefrom0.001 % to 0.01 % had a much larger effect than does the increasefrom0.005 % to 0.015 %.  This is further illustrated in Figure 4.1, which shows that comparatively small reductions in final chlorate concentration were obtained with increases in catalyst charge beyond 0.005%.  50  Results and Discussion Variable  Main Effects Time (min.) Temperature TO Substitution (% eq. Chlorine) Catalyst Charge (wt. % on o.d. pulp) Interaction Effects Temperature/Time Substitution/Time Catalyst/Time Substitution/Temperature Substitution/Catalyst Catalyst/Temperature  Chlorate 95% confidence Eliminated (mol %) interval  Factor Levels (Low/High)  Residual Chlorate (mM)  120/240  -0.071  70/80  -0.025  50/100  -0.001  -11.99  0.005/0.015  -0.015  37.82  -  -0.003 -0.001 -0.039 -0.010 -0.014 -0.010  -3.58 5.47 -15.76 -3.21 7.85 4.91  -  -  95% confidence interval  36.98 0.017  0.017  26.02  1.53  1.53  Table 4.3: Results of the factorial analysis of chlorate concentration for Experimental Program 2. 4.1.2. The Effect of Temperature Temperature had the largest main effect on chlorate reduction of any of the variables tested.  As seen in Tables 4.2 and 4.3, temperature was an important factor at all catalyst charge levels. Increasing the temperature increased the rates of reaction in solution and resulted in lowerfinalchlorate concentrations. The influence of temperature is further illustrated in Figure 4.2. Increasing the temperature from 70°C to 80°C resulted in significantly lower residual chlorate concentrations at both 50% and 100% substitution,.  51  Results and Discussion The temperature was not increased beyond 80°C because adverse effects on pulp viscosity resulted when the temperature was increased to 85°C and 90°. The impact of the catalytic system on viscosity is discussed in detail in section 4.3.1. 4.1.3. The Effect of Reaction Time Table 4.2 shows that increasing the reaction timefrom2 to 4 hours has a large main effect on the chlorate concentration. This agrees with the work of Deutsch et al. (1979), who studied vanadium pentoxide "activation" of the chlorate residual in chlorine dioxide brightening. They found that reaction times of greater than 2 to 3 hours were required for significant drops in chlorate concentration. Since chlorate formation nears completion after 30 to 40 minutes (Ni et al. 1993), the rate of chlorate formation does not account for the long retention times required for the chlorate reduction reactions. It is the rates of the cycling reactions shown in Figure 2.7 that determine the overall rate of chlorate reduction. No directly applicable quantitative data concerning the rates of reaction of V* with organics or V " with chlorate were found in the 5  1  4  literature.  4.1.4. The Effect of Chlorine Dioxide Substitution Tables 4.2 and 4.3 show the main effects of substitution on thefinalchlorate concentration and the percent reduction in chlorate concentration. While the impact of substitution on final chlorate concentration was not significant, the percent reduction in chlorate concentration was a strong function of substitution. The substitution had a large effect on the percent reduction in chlorate concentration because more chlorate was formed at 50%  52  Results and Discussion  0.000  0.002  - 7 0 d e g r e e sC / 5 0 %S u b s t i t u t i o n  7 0  d e g r e e s  C / 1 0 0 %  S u b s t i t u t i o n  - 8 0  8 0  d e g r e s s  C / 1 0 0 %  S u b s t i t u t i o n  0.012  0.014  d e g r e e s C / 5 0 % S u b s t i t u t i o n  0.004  0.006  0.008  0.010  0.016  Vanadium Pentoxide Charge (weight %)  Figure 4.1: Chlorate concentration as a function of catalyst charge, temperature, and substitution. (Bleaching conditions: final pH of 2.5, retention time of 240 minutes.) substitution than at 100% substitution in the uncatalyzed runs, while thefinalchlorate concentration in the catalyzed runs was nearly equal for both substitution levels.  As a result, the reductions in chlorate concentration were larger in the 50 percent substitution runs than in the 100 percent substitution runs. A detailed discussion of the impact of substitution on thefinalchlorate concentration and the percent reduction in chlorate concentration for the catalyzed and uncatalyzed runs is presented in sections 4.1.4.1 through 4.1.4.3.  53  Results and Discussion 4.1.4.1. Effect on Chlorate Formation in Uncatalyzed Bleaching The degree of substitution has a major impact on the concentration of chlorate in chlorine dioxide bleaching liquors (Ni et al. 1993, Reeve and Weishar 1991, Rapson and Reeve 1980). As shown in Table 4.4, the chlorate formed in the uncatalyzed bleaching runs was not a linear function of chlorine dioxide substitution. Thefinalchlorate concentration was a minimum at 100% substitution, increased to a peak value at 75%, and fell back to an intermediate value at 50% substitution.  Substitution  Chlorate concentration (mmol/1)  Fraction of C10 charge as chlorate (mol %)  50 75 100  0.661 0.917 0.460  24.1 22.3 8.4  2  Table 4.4: Chlorate production in the uncatalyzed bleaching runs. This relationship between chlorate concentration and substitution agrees with previous reports. Fair (1991) found that for simultaneous chemical addition, chlorate formation was a maximum at approximately 60% substitution; the chlorate concentration at 50% substitution was lower than at 75% substitution. These results are further supported by earlier studies which found chlorate formation to have a local maximum at 70% to 85% substitution (Reeve and Rapson 1980, Bergnor et al. 1987, Axegard 1987, Germgard and Karlsson 1988).  The chlorate concentration profile for the uncatalyzed bleaching runs can be explained using the mechanism proposed by Ni et al. 1993), in which chlorate is formedfromthe reaction of  54  Results and Discussion chlorite and hypochlorous acid. Using this mechanism, the trends found in chlorate formation can be explained as follows: •  75% substitution:  High concentrations of both hypochlorous acid and chlorite in solution results in substantial chlorate formation.  •  50% substitution:  Relatively lower chlorite concentration results in less chlorate formation than at 75% substitution  •  100% substitution:  Although the concentration of chlorite in solution is a maximum, there is little hypochlorous acid available for the chlorate formation reaction, and thefinalchlorate concentration is lower than for 50% or 75% substitution.  4.1.4.2. Effect on Chlorate Concentration in the Catalyzed Runs The effect of substitution on the chlorate concentration found in the catalyzed system can be seen in Figure 4.1 and in Tables 4.2 and 4.3. Despite the large impact of chlorine dioxide substitution on the formation of chlorate in uncatalyzed bleaching solution, thefinalchlorate concentration was not strongly dependent on substitution when vanadium pentoxide was added to the bleaching liquor. At catalyst charge levels of 0.001% and 0.005%, the main effect of substitution on final chlorate concentration was barely significant; at catalyst charges of greater than 0.005 weight percent, substitution was not a significant factor infinalchlorate concentration (see Tables 4.2 and 4.3).  55  Results and Discussion Figure 4.2 further illustrates that the chlorate concentration in the catalyzed bleaching runs was largely independent of the chlorine dioxide substitution. At catalyst charges of 0.01 % and greater, the chlorate concentration was approximately the same at 50% and 100% substitution.  40 C h l o r a t e F o r m a t i o n w i t h N o  C a t a l y s t  C h l o r a t e C o n c e n t r a t i o n w i t h V a n a d i u m  A d d e d P r e s e n t  E"  2.2 (0 K O  °  £  «  20  ~ ~ - > .  ts'  I n c r e a s i n g i n i t i a l chlorate f o n m a t b n  ^ v X  N  o o  •s 1 2 0  1 8 0  2 4 0  Time (min)  Figure 4.2: An illustration of chlorate concentration profiles in catalyzed and uncatalyzed chlorine dioxide bleaching. The small impact of substitution on thefinalchlorate concentration can be explained using Figure 4.2. Although the maximum chlorate concentration achieved is higher at 50% substitution than at 100% substitution, after 4 hours of reaction with the catalyst, the chlorate concentrations are approximately equal. The higher chlorate concentration achieved at 50% substitution results in a higher rate for the chlorate reduction reaction. As the reaction proceeds and the chlorate concentration drops, the rate of the reaction slows. After 4 hours, the rate of reaction and the chlorate concentration for the 50% substitution case matches that for the 100% substitution case.  56  Results and Discussion 4.1.4.3. Impact of Substitution on the Percent Reduction of Chlorate in Catalytic Bleaching Although thefinalchlorate concentration was not strongly affected by the degree of substitution, the percent reduction in chlorate concentration was. The large effect of substitution on the percent reduction does not indicate a significant impact on the performance of the catalyst; rather, the large reductions are the result of the impact of substitution on chlorate formation. The large percent reductions in chlorate concentration were found because  •  the amount of chlorate formed in the uncatalyzed runs varied with chlorine dioxide substitution, while  •  thefinalchlorate concentration in the catalyzed runs was similar at all substitution levels  Since the percent reduction is calculated according to equation 4.1, the impact of substitution on the amount of chlorate formed had a large impact on the percent reduction of chlorate.  4.1.5. The Effect of pH Table 4.2 shows that the impact of pH was similar to that of substitution. The effect of the starting pH of the solution on thefinalchlorate concentration was barely significant, yet pH had a large impact on percent reduction in the chlorate concentration.  The pH in the delignification stage is known to have a large impact on chlorate formation (Reeve and Weishar 1992, Wartiovaara 1985). Lowering the pH results in less chlorate formation during chlorine dioxide bleaching. Chlorate is formedfromthe reaction of  57  Results and Discussion chlorite and hypochlorous acid (Ni et al. 1993). As shown in Figure 2.5, decreasing the pH shifts the cWorine/hypochlorous acid equilibrium towards elemental chlorine, and less hypochlorous acid is available to react with chlorite.  Substitution and pH both have large impacts on chlorate formation in the uncatalyzed bleaching runs, and little impact on thefinalchlorate concentration for the catalyzed bleaching runs. As a result, the discussion of the effect of starting pH on thefinalchlorate concentration and the percent reduction in the catalytic system is analogous to the discussion presented for substitution. 4.2. The Recovery of the Bleaching Power Lost to Chlorate One of the primary objectives of the research was to investigate the impact of vanadium pentoxide addition on bleaching efficiency in the delignification stage. According to the reaction mechanism shown in Figure 2.7, adding vanadium pentoxide to the (D+C) stage should recapture the bleaching power lost as chlorate.  The response of the bleaching efficiency to catalyst addition was measured through the (D+C)E brightness and kappa number. Although the impact on brightness was significant, the kappa number was not significantly affected by adding vanadium to the (D+C) stage.  Since the brightness of the pulp is increased, the vanadium reactions must result in the oxidation of the chromophores in the lignin. The lack of impact on the kappa number of the pulp indicates that the vanadium oxidation reactions tend not to contribute to the dissolution of the lignin during vanadium catalyzed bleaching.  58  Results and Discussion The response of the kappa number and the brightness to vanadium pentoxide addition to the delignification stage is discussed in the following sections.  4.2.1. The Kappa Number Response The factorial analysis summarized in Table 4.5 shows that at 95% confidence, there was no significant decrease in the (D+C)E kappa number with vanadium pentoxide addition. Of the factors studied, substitution was the only parameter which significantly changed the kappa number. It is well known that the delignification efficiency is a function of the degree of substitution and the mode of addition of chlorine/chlorine dioxide mixtures (Hatton 1967, Hatton 1966, Munro et al. 1990, Pryke 1992, Macas and Evans 1994, Reeve and Weishar 1991, Solomon et al. 1993). In this study, the 100% substitution and the 50% substitution (D+C)E uncatalyzed bleaching sequences yielded kappa numbers of 5.6 and 6.2, respectively. Usually, delignification is found to be more efficient at 50% substitution than at 100% substitution. The disagreement found in the current work is attributed to the variable nature of pulp reactions. Changing pulp sources and reaction conditions has a large impact on the delignification efficiency and, therefore, thefinalkappa number of the bleached pulp.  59  Results and Discussion Kappa number (ml)  Brightness (ISO)  Brightness Gain (ISO)  95% Confidence Interval  0.54  0.83  0.83  Main Effects Time Temperature Substitution Catalyst Concentration  0.01 -0.09 -1.01 0.04  -0.42 -0.50 2.89 1.17  -0.42 0.90 -0.48 1.17  Interaction Effects Temperature/Time Substitution/Time Catalyst/Time Substitution/Temperature Substitution/Catalyst Catalyst/Temperature  0.21 -0.16 -0.11 0.09 0.26 -0.01  -0.75 -0.42 0.11 0.41 -0.26 0.59  -0.75 -0.42 0.82 -0.82 0.17 0.59  Variable  Table 4.5: The factorial analysis of the (D+C)E kappa number and brightness responses. Since oxygen pre-bleaching removes a large fraction of the initial lignin content of the pulp, we suspected that the impact of the vanadium oxidations on kappa number was masked by using the oxygen delignified pulp. The hypothesis was that by removing the easily oxidized lignin through oxygen pre-bleaching, the potential for catalytic delignification was masked.  To examine the impact of oxygen pre-bleacrting on the kappa number of "activated" chlorate bleached pulps, several runs from Table 4.4 were repeated using unbleached kraft pulp. As found for the oxygen pre-bleached pulp runs, the "activation" of the chlorate residual in the delignification of unbleached kraft pulp did not have a significant impact on the kappa number of the bleached pulp.  60  Results and Discussion There is no obvious explanation for the lack of kappa number reduction with vanadium pentoxide catalyzed bleaching. Since no information regarding the detailed mechanism of the reaction between pentavalent vanadium and lignin is available, it must be assumed that the oxidation reactions that occur do not promote lignin dissolution during chlorine dioxide bleaching or the subsequent caustic extraction.  4.2.2. The Brightness Response Contrary to the response of kappa number, the bleached pulp brightness did provide evidence of chlorate "activation" when vanadium was added to the chlorination stage. Table 4.5 shows that at 95% confidence, both catalyst charge and substitution have significant main effects on the brightness of the bleached pulp.  The brightness gains shown in Table 4.5 can be correlated to increased activity of the vanadium catalyst. Average brightness gains, compared to uncatalyzed bleaching runs at the same temperature, of approximately 1 percent were found for increases in temperature or catalyst charge. Since increases in temperature and catalyst charge were found to increase catalyst activity (Section 4.1), the brightness gains are attributed to "activation" of the chlorate residual. 4.2.2.1. Brightness from Catalyzed and Uncatalyzed Bleaching Runs In the previous section, we found brightness gains with catalyst addition when compared to uncatalyzed bleaching runs at the same temperature. When compared to bleaching runs at 40°C, however, there are no brightness gains associated with the "activated" chlorate system.  61  Results and Discussion Figure 4.3 shows that the brightness of the uncatalyzed bleaching runs decreased as temperature was increased. When chlorine dioxide bleaching sequences are operated at greater than 70°C, the oxidant charge may be completely consumed, and brightness losses may occur (Rapson and Strumila 1979). Since we used a relatively low charge factor of 0.15, long bleaching runs, and high temperatures, there was no residual oxidant charge remaining at the end of the bleach . The result was diminished brightness in the uncatalyzed 3  bleaching runs as the temperature was increased.  Figure 4.3 also compares the brightness attained with catalyzed bleaching runs at 70°C and 80°C with uncatalyzed bleaching runs at 40°C, 70°C, and 80°C. Although the addition of catalyst results in higher brightness at 70°C and 80°C, the brightness gains do not compensate for the losses in brightness associated with increasing the temperature from 40°C to greater than 70°C.  In summary, higher final brightnesses were obtained for uncatalyzed bleaching runs at lower temperature than for the catalyzed bleaching runs and vanadium pentoxide addition is ineffective at recovering the bleaching power lost as chlorate in the delignification stage.  N i et al. showed that at 45°C and a charge factor of 0.22, there is no residual oxidant charge after 120 minutes. Given the increase in the rate of reaction with an increase in temperature from 45°C to 70°C or 80 °C, combined with the lower charge factor of 0.15 used in this study, the oxidant charge is guaranteed to be fully consumed. a  62  Results and Discussion  70-t  Temperature (°C)  Figure 4.3: Brightness as a function of temperature and substitution.  4.3. The Impact of the Catalytic Bleaching on Pulp Properties To gain some insight into the impact of vanadium addition on the bleached pulp quality, testing of the physical properties of the catalyzed and uncatalyzed bleached pulp was performed as part of Experimental Program 2. The factorial analysis of the viscosity and the strength property responses is presented in Tables 4.6 through 4.8.  A discussion of the impact of chlorate "activation" on bleached pulp properties is presented in the following sections.  63  Results and Discussion 4.3.1. Pulp Viscosity Table 4.6 shows the results of the factorial analysis of the pulp viscosity responses to time, temperature, substitution, and catalyst charge. All of the main effects and most of the interaction effects were significant. Increases in time, temperature, and catalyst charge tended to reduce the viscosity. The viscosity was lower for the 50% runs than for the 100% runs. A discussion of the impact of each of the factor variables on the bleached pulp viscosity is presented in sections 4.3.1.1 through 4.3.1.3. 4.3.1.1. The Effect of Chlorine Dioxide Substitution Of all the variables tested, the degree of substitution had the largest impact on thefinalpulp viscosity. Substitution was expected to have a large effect on the bleached pulp viscosity because of the complex relationship that exists between the chlorine charge in thefirststage and the viscosity of the pulp produced (Macas and Evans 1994). It has long been known that the addition of chlorine dioxide to the chlorination stage offers viscosity protection results in higher -viscosity for chlorination stage pulp (Jack and Feller 1967, Hatton 1967, Hatton et al. 1966). Mill reports also show that increased substitution, even at greater than 50% substitution, can result in higher viscosities (Pryke et al. 1993) Therefore, the present findings of lower viscosity at 50% substitution than at 100% substitution are consistent with both laboratory reports and mill scale findings.  64  Results and Discussion Viscosity (cps)  Viscosity Loss (cps)  95% Confidence Interval  0.20  0.2  Main Effects Time Temperature Substitution Catalyst Concentration  -1.40 -1.87 2.23 -0.25  1.40 0.75 -1.70 0.25  Interaction Effects Temperature/Time Substitution/Time Catalyst/Time Substitution/Temperature Substitution/Catalyst Catalyst/Temperature  -0.24 -0.54 -0.40 -0.43 0.18 0.11  0.24 0.54 0.40 0.34 -0.18 -0.11  Variable  Table 4.6: The factorial analysis of the (D+C)E viscosity response. 4.3.1.2. The Effect of Catalyst Charge According to the factorial analysis, increasing the catalyst chargefrom0.005 to 0.015 weight percent had only a small impact on the pulp viscosity.  However, the effects calculated in the factorial analysis may be misleading. As shown in Figure 4.1, increases in catalyst charge beyond 0.005 weight percent on oven dry pulp had a modest impact on chlorate elimination and did not seriously increase the catalyst activity. Since the viscosity measurements were performed on the runs operated on the flat section of the catalyst activity curve, the impact of the catalyst on thefinalviscosity of the pulp was not clearly shown in this catalyst charge range.  65  Results and Discussion  35 D Without catalyst  B0.005 % vanadium  • 0.015 % vanadium  30 25  100% Substitution  50% Substitution  20 15 10 5 70  70  80  80  Temperature (°C)  Figure 4.4: The impact of vanadium addition on viscosity.  Although not clearly shown in the factorial results, the vanadium oxidation reactions did have a large impact on thefinalviscosity of the bleached pulp. As illustrated in Figure 4.4, the pulp viscosity dropped with increases in temperature and catalyst charge, and with lower chlorine dioxide substitution. Since more vanadium oxidations occur at higher temperatures, higher catalyst charges, and at lower substitution (Section 4.1), it is evident that the viscosity loss of the pulp is directly related to the catalyst activity.  The viscosity losses associated with the catalyst activity are a result of the mechanism of vanadium pentoxide oxidations. Oxidation reactions involving pentavalent vanadium in aqueous solutions are known to proceed viafreeradical mechanisms (Littler and Waters, 1959). Free radical oxidations are non selective and can result in both cleavage of cellulosic chains, and lignin oxidations. The cellulose chain cleavage results in lower pulp viscosity.  66  Results and Discussion 4.3.1.3. The Effects of Temperature and Time Since the catalytic bleaching operation must be conducted at lowfinalpH values and at elevated temperature for effective catalyst action, significant losses in viscosity due to acid hydrolysis occurred (Grace et al. 1988). Since the rate of acid hydrolysis increases with temperature, the viscosity losses were greater at 80°C than at 70°C, and the main effect of temperature on viscosity was large.  The large main effect of time on viscosity was a result of more extensive acid hydrolysis as the reaction time was increased. 4.3.2. The Response of the Physical Properties Physical testing of the bleached pulp was performed to investigate the impact of the viscosity losses on the strength properties of the pulp. The physical testing performed included the tensile index, breaking length, tear index, tear strength, burst index, burst strength, zero span index, bulk, and opacity. As seen in Table 4.7, none of the main effects of substitution, time, and catalyst charge were significant for the runs performed at 80°C.  67  Results and Discussion  Retention Substitution time  Parameter  Tensile Index Breaking length Tear Index Tear strength Burst Index Burst strength Zero Span Index Bulk Opacity  (N-m/g) (km) (mN*m /g) (rnN) (kPa-mVg) (kPa) (km) (cm /g) (ISO) 2  3  -1.29 -0.13 0.39 23.70 0.03 -1.79 -0.51 -0.03 -0.77  0.17 0.02 -0.39 -23.38 -0.07 -0.59 0.30 -0.05 -0.06  Catalyst charge  0.38 0.04 -0.38 -22.92 -0.09 -1.09 -0.25 -0.03 -0.06  95 % confidence interval 3.28 0.33 0.48 29.10 0.51 5.31 1.03 0.32 2.59  Table 4.7: Factorial analysis of the physical testing results for the runs at 80°C Furthermore, Table 4.8 shows that no significant difference was found in the strength properties of pulp bleached with and without catalyst. The factorial analysis included the effects of catalyst charge, substitution, and temperature on pulp strength. The bleached pulps used in the analysis werefromuncatalyzed runs at 40°C and runs with 0.015 weight percent catalyst charge at 80°C. These pulps represent the lowest and highest degrees of cellulose degradation found in the study; the viscosities of the catalyzed and uncatalyzed bleaching run pulps were, respectively, 18.29 and 23.75, for the 100% substitution case and 16.77 and 22.21, for the 50 % substitution case.  Despite viscosity drops on the order of 5 points, to a minimum viscosity of 16.8, there was not a significant change the physical properties of the pulp. These results are supported by recent investigations of the impact of TCF and ECF bleaching sequences on pulp strength and optical properties which show that at viscosity values as low as 13 to 14 cps, the pulp quality is not adversely affected (Gottlieb et al. 1993, Kappel, Brauer and Kittel 1993, Mokfienski 1994, and Francis ef al. 1994).  68  Results and Discussion Parameter  Tensile Index Breaking length Tear Index Tear strength Burst Index Burst strength Zero Span Index Bulk Opacity  Temperature  (N-m/g) (km) (mN*m /g) (mN) (kPa-m /g) (kPa) (km) (cm /g) (ISO) 2  2  3  -1.44 -0.15 -0.19 -11.25 • -0.24 -1.38 -1.14 -0.09 1.98  Substitution  Catalyst charge  95% confidence interval  1.62 0.16 0.23 13.64 -0.06 4.25 0.56 0.04 -2.02  -1.44 -0.15 -0.19 -11.25 -0.24 -1.38 -1.14 -0.09 1.9.8  4.64 0.48 0.69 41.16 0.72 7.51 1.46 0.46 3.66  Table 4.8: A comparison of the physical testing results for the 40°C uncatalyzed runs and the 80°C runs with a 0.015 weight percent catalyst charge. Therefore, although the "activation" of the chlorate residual does result in substantial viscosity losses, these losses are not manifested as losses in pulp strength or optical properties. 4.4. Potential Applications for Vanadium in the Delignification Stage Although the addition of vanadium pentoxide to thefirststage did not result in significant recovery of the bleaching power of chlorate, vanadium addition does have potential for reducing the chlorate concentration in the effluentfromthefirststage.  Figures 4.5 and 4.6 illustrate two conceptual operatingflowchartsusing vanadium pentoxide to reduce the chlorate dischargesfromthefirststage.  Figure 4.5, illustrates the standard operation of the chlorination stage with vanadium pentoxide added to the bleaching chemicals. A significantfractionof the chlorination stage effluent is recycled. This reduces the required vanadium charge for chlorate reduction, and minimizes the vanadium content of the bleaching stage effluent. Based on the results of this  69  Results and Discussion study, reductions in chlorate concentration of up to 95% would be expected with this system. Figure 4.6, illustrates a potential variation on the conventional chlorination stage operation, in which a vanadium stage is added to the recycle. In this scheme the vanadium could be used as an immobilized catalyst structure, perhaps as a plate or as pellets, to catalyze the reduction of the chlorate in the recycle. This would eliminate the entrainment of vanadium in bleach plant effluent. And, chlorate reduction would be accomplished without the attendant loss in pulp viscosity. The potential for this reaction scheme to reduce the chlorate concentration in the delignification stage should be investigated.  70  Results and Discussion  Wash Water  Vanadium Pentoxide Addition  C10 andCl 2  2  Brown Stock Pulp  Delignification Stage Recycle  Figure 4.5: Conceptual flow sheet for the chlorination stage operated with vanadium pentoxide addition.  71  Results and Discussion  Wash Water  Pulp to D stage • C10 andCl 2  2  Brown Stock Pulp  Delignification Stage  Wash Recycle Delignification Stage Recycle  Vanadium Pentoxide Treatment of Delignification Stage Recycle  Figure 4.6: Conceptual flow sheet for the vanadium treatment of the chlorination stage recycle.  72  5. C O N C L U S I O N S  1. The addition of vanadium pentoxide to the chlorination stage operated at high chlorine dioxide substitution can result in reductions of greater than 90% in the final chlorate concentration. 2. Temperature has a large impact on the activity of the catalyst. Temperatures of greater than 70°C are required for "activation" of the chlorate residual with chlorine dioxide. The activity of the catalyst increases dramatically with temperature; however, the serious viscosity losses are incurred at temperatures above 80°C. 3. Retention times of greater than 2 hours are required for significant reductions in the final chlorate concentration. Retention times of approximately 4 hours are required for very lowfinalchlorate concentrations. 4. The effect of the degree of chlorine dioxide substitution on thefinalchlorate concentration was barely significant for the catalyzed bleaching runs. 5. At pH values of less than 3, thefinalchlorate concentration was independent of pH.  6. Increases in catalyst charge beyond 0.005 weight percent did not substantially increase the activity of the catalyst for chlorate reduction. 7. The optimum conditions for vanadium pentoxide catalysis in the chlorination stage are temperature of 80°C, reaction time of 4 hours, and a catalyst charge between 0.005 and 0.01 weight percent on oven dry pulp.  73  Conclusions 8. The activity of the catalyst resulted in serious viscosity losses. The initial viscosities for the uncatalyzed bleaching runs at 50% and 100% chlorine dioxide substitution cases were 21 and 23, respectively. The catalyzed chlorination stage produced pulp with viscosity as low as 16 cps. 9. Despite significant losses in viscosity, the physical properties of the pulp produced using the catalytic system were not significantly differentfromthose of pulp bleached using conventional chlorination. 10. The addition of vanadium pentoxide to the chlorination stage did not result in substantial gains in bleaching power. When compared to uncatalyzed bleaching runs at the same temperature, the addition of vanadium pentoxide resulted in brightness gains of approximately 1 percent; however, thefinalbrightness of the catalytically bleached pulps was lower than pulps bleached without catalyst at 40°C. There was no significant decrease in Kappa number associated with vanadium pentoxide addition.  11. Although vanadium pentoxide catalysis of chlorine dioxide delignification is not a reasonable strategy for increasing the bleaching effectiveness, it does offer a potential method for reducing the chlorate concentration in the chlorination stage effluent.  74  6. RECOMMENDATIONS 1. More assessment of the problem of the chlorate production in North American pulp mills is required. Outside of Sweden, little attention has been given to the problem of chlorate formation. Because the bleach plants in North American pulp mills are being operated with ever increasing levels of chlorine dioxide substitution in the delignification stage, the problem of chlorate formation in the North American context must be addressed. 2. A search for other potential catalysts capable of "activating" the chlorate residual with more specific and effective oxidation of lignin should be carried out. Organic catalysts are of particular interest. An organic catalyst could be used in closed cycle mill operations where the solids contained in the bleach plant effluent are burned. Unlike vanadium pentoxide, an organic catalyst would form carbonaceous combustion products. 3. There is potential for vanadium pentoxide to be used to treat the chlorate contained in the chlorination stage recycle. The catalyst could be used in a solid form such as plates or immobilized pellets to eliminate the entraining of vanadium in the effluent stream. Conceptually, the chlorination stage recycle would contact the catalyst and the chlorate would be reduced to chloride. Thefinalchlorate concentration could potentially be reduced to very low levels without releasing vanadium to the environment. The potential for this type of treatment for chlorate elimination should be investigated further.  75  7. REFERENCES  Axegard, P., 1989. Substituting chlorine dioxide for elemental chlorine makes the bleach plant effluent less toxic. Tappi Journal 69(10):54-59.  Axegard, P., 1987. C10 substitution reduces the load of TOC1. Proceedings of the TAPPI Pulping Conference-.QVashington, DC). 105-111. 2  Bergnor, E., Germgard, U., Kolar,J.,and Lindgren, B., 1987. Formation of chlorate in chlorine dioxide bleaching. Cellulose Chemistry of Technolnology 21(3):307-314. Berry, R., Fleming, B., Voss, R., Luthe, C , and Wrist, P., 1989. Toward preventing the formation of dioxins during chemical pulp bleaching. Pulp and Paper Canada 90(8). Canadian Pulp and Paper Association. Technical Section., 1966-. Standard Testing Methods. Clark, R., 1973. Vanadium. In Comprehensive Inorganic Chemistry, vol. 3, chap. 34: 491-551. Bailar,J.C, etal., eds. Oxford: Pergamon Press.  CRC Handbook of Chemistry andPhysics, 63rd ed: 1982. Weast, R , and Astle, M., eds. Boca Raton, FL.: CRC Press.  Deutsch, H., and Shoemaker, J. Jr., and Eachus, S.W., 1979. Vanadium pentoxide catalysis of chlorine dioxide bleaching. Tappi Journal 62(12):53-55.  Deutsch, H. and Shoemaker, J., 1977. Bleaching of cellulosic pulpfiberswith chlorine dioxide in the presence of a vanadium compound. U.S. Patent No. 4,039,374.  Easty, D., Johnson, J., and Webb, A., 1986. Analysis of bleaching liquors by ion chromatography. Paperija Puu 68(5):415-417.  Emmenegger, F., and Gordon, G , 1967. The rapid interaction between sodium chlorite and dissolved chlorine. Inorganic Chemistry 6(3):633-638. 76  References  Farr, G., 1991. Chlorination effluent recycle in kraft pulp bleaching. M.A.Sc. Thesis, University of British Columbia. Francis, R, Troughton, N, Zhang, X, and Barnes, W., 1994. Caroate delignification lU. In combination with ozone. International Pulp Bleaching Conference poster session, (Vancouver, B.C.) 165-169. Germgard, U., 1989. Chlorate discharges form bleach plants - how to handle a potential environmental problem. PaperijaPuu 71(3):255-260.  Germgard, U. and Karlsson, R., 1988. Bleaching of birch kraft pulp with different fractions of prebleaching. Nordic Pulp and Paper Research Journal 3:166177. Germgard, U, Teder A, and Tormund D., 1981. Chlorate formation during chlorine dioxide bleaching of softwood kraft pulp. PaperijaPuu 63(3):127-133. Gordon, G., Kieffer, R., and Rosenblatt, D., 1972. The chemistry of chlorine dioxide. In Progress in Inorganic Chemistry, Lippard, S., ed. New York: Wiley Interscience. vol. 15:201-286.  Gottlieb, P, Nutt, W, Miller, S, and Macas, T., 1983. Mill experience in the high consistency ozone bleaching of southern pine pulps. TAPPIPulping Conference. (Houston, TX) 1183-1188.  Grace, T., Leopold, B., Malcolm, E.,and Kocurek, M., 1989. Chemical reactions of wood consituents. In Pulp and Paper Manufacture, Vol. 5 Alkaline Pulping. Montreal, PQ CPPA. 23-44. Hatton, J., 1967. Delignification of kraft pulp in thefirstbleaching stage using chlorine dioxide and chlorine. Pulp and Paper Magazine of Canada 68(4):T181T190. Hatton, J., Murray, F., and Clark, T., 1966. Studies on delignification of kraft pulp in thefirstbleaching stage using chlorine and chlorine dioxide. Pulp and Paper Magazine of Canada 67(4):T241-T248.  77  References  von Heijne, G., and Teder, A., 1973. Kinetics of the decomposition of aqueous chlorine dioxide solutions. Acta Chemica Scandinavica 27(10):4018-4019.  Histed, J, Vega Canovas, R.,and Ruscitti, G., 1991. Chlorination stage design and operation for 50% chlorine dioxide substitution. CPPA Techical Section 77th Annual Meeting Preprints: A35-A40. Isensee, A.R., Shaw, W.C., Gentner W. A., Swanson, C.R., Turner, B.C.,and Woolson, E. A., 1973. Revegitation following massive appication of selected herbicide, Weed Science. 21(5):409. Jack,W.Q.,and Feller, L.D., and Jack, W., 1967. Pulp and Paper Magazine of Canada. 68(9):T461. Kappel, J, Braeur, P,and Kittel P., 1993. High consistency ozone bleaching technology. Tappi Pulping Conference. (Atlanta, GA.): 1173-1181.  Kramer, J., 1972. Delignification of kraft pulp with chlorine, chlorine dioxide, and their mixtures. Tappi 55(6):964-971.  Lee, C-L. Personal communication, 1994. Lee, J.D., 1991. Concise Inorganic Chemistry. New York: Chapman and Hall. 697712. Liebergott, N , van Lierop, B., Kovacs, T., and Nolin, A., 1990. A comparison of the order of addition of chlorine and chlorine dioxide in the chlorination stage: Comparison at constant chemical charge. Tappi Journal 73(10) :207-213.  Lindgren, B.,and Nilsson, T.,1975. Chlorate formation during the reaction of chlorine dioxide with lignin model compounds. SvenskPapperstidning 1975(2):66-68.  Littler, J. and Waters, W., 1959. Oxidations of organic compounds with quinquevalent vanadium. Journal of Chemical .Society. (March): 1299-1305.  78  References Macas, T., and Evans, T., 1994. The effect of chlorine in the D l stage. In Bleaching.Press Anthology. (Atlanta, GA): TAPPI Press 391-395.  Marpillero, P. 1957. Canada Patent No. 550,735. December 24, 1957.  Marpillero, P., 1958. The bleaching of pulps with acitivated chlorate. Tappi 41(5):213A-216A. D. McCleay and Associates Ltd., 1987. Aquatic Toxicity ofPulp and Paper Mill Effluent: A Review. (Report EPS 4/PF/l). Ottawa, ON: Environment Canada  N. McCubbin Consultants Inc., 1992. Best Available Technologyfor the Ontario Pulp and Paper /ndws/ry.Toronto, ON: Ontario Ministry of the Environment. Water Resources Branch. McDonough, T.J., Berry, R.M., Betts, J.L., Du Manoir, J.L., Reeve, D.W., andTurnbull, J.K., 1985. Chlorine Dioxide in the chlorination stage - A summary of existing published information. International Pulp Bleaching Conference. (Quebec.PQ): 143-153. McKinnon, L., 1992. AOX as a Regulatory Parameter: A Scientific Review ofAOX Toxicity and Environmental Fate. Victoria, BC: B.C. Ministry of the Environment. Sept., 1992. Mokfienski A, Demuner B., 1994. Pilot plant experince with ozone in TCF bleaching of eucalypt pulp. International Pulp Bleaching Conference. (Vancouver,BC): 309-318. Morioko, S. 1981. Process for manufacturing chlorine dioxide. Canada Patent no. 1,102,093. Munro, F., Chandrasekaran, S., Cook, C , and Pryke, D., 1990. Impact of high chlorine dioxide substitution on oxygen delignified pulp. Tappi Journal 73 (5): 123-130. Ni, Y., 1992. A Fundamental Study of Chlorine Dioxide Bleaching of Kraft Pulp. Ph. D. Thesis, McGill University.  79  References  Ni, Y., Kubes, G.J.,and van Heiningen, A.R.P., 1993. Mechanism of chlorate formation during bleaching of Kraft pulp with chlorine dioxide. Journal of Pulp and Paper Science 19(1):J1-J6.  Nilsson, T, and Sjostrom, L., 1974. Losses in chlorine dioxide as a result of chlorate formation during bleaching. SvenskPapperstidning (17):643-647.  O'Connor, B., Kovacs, T., Voss, R., Martel, P., and van Lierop, B., 1993. A laboratory assessment of the environmental quality of alternative pulp bleaching effluents. International Environmental Symposium, (Paris.) 273297. Partridge, H. and Lai, P., 1980. Versatile process for generating chlorine dioxide. U.S. Patent No. 4,206,193. Perrin, and C, Bothwell, M., 1992. Chlorate discharges from pulp mills: an examination of effects on river algal communities. Water Pollution Research Journal of Canada 27(3):473-485. Pryke, D., 1992. Chlorine dioxide substitution: mill experience. In Tappi Bleach Plant Operations Short Course. Altanta,GA: TAPPI Press 79-90.  Pryke, D., Bourree, G , Winter, P. and Mickowski, C. 1993. The impact of chlorine dioxide delignification on pulp manufacturing and effluent characteristics at Grande Prarie, Alberta. Submitted to Pulp and Paper Canada.  Pryke, D., Dumitru, M., Cunnington, R , and Reeve, D., 1993. A Survey of Chlorine Dioxide Substitution in Bleached Pulp Mills in Canada. Montreal, PQ. CPPA Technical Section Bleaching Commitee.  Radhakrishnamurti, P., and Devi, S., 1975. Kinetics of oxidation of cyclohexanone, cyclopentanone, cyclooctanone, and cycloheptanone be V(V) in acid medium. Indian Journal of Chemistry vol 14 A, June 1976: 399-401.  Radhakrishnamurti, P., and Panda, B., 1970. Oxidation of phenols and amines by vanadium (V). Indian Journal of Chemistry \o\ 8, October 1970: 946-948.  80  References  Radhakrishnamurti, P., and Pati, S., 1969a. Oxidation of cyclanols by V(Y). Israel Journal of Chemistry vol 7:429-434.  Radhakrishnamurti, P., and Pati, S., 1969b. Kinetics of oxidation of substituted hydrocarbons by vanadium (V). Indian Journal of Chemistry vol 7, July 1969: 687-689. Rapson, W., Anderson, C , andMillen, R., 1959. Tappi 42(8):642-649. Rapson, W. and Anderson, C , 1966. Pulp and Paper Magazine of Canada. 67(1):T47-T55. Rapson, W., and Strumila, 1979. Chlorine dioxide bleaching. In The Bleaching of Pulp, chap. 6. R. Singh, ed. Atlanta:GA: TAPPI Press. 133 Reeve, D. 1992. The Principles of bleaching. In Bleach Plant Operations Short Course. Atlanta, GA: TAPPI Press. 1-12. Reeve, D., 1994. Chlorine dioxide bleaching. In Bleaching: A TAPPI Press Anthology. Atlanta,GA: TAPPI Press. 365-369.  Reeve, D., and Rapson, W., 1980. Developments in chlorine dioxide bleaching. In Proceedings TAPPI Pulping Conference. (Altanta, GA): 403-409.  Reeve, D.W. and Weishar, K.M., 1991. Chlorine dioxide delignification - process variables. Tappi Journal. 74(6): 164-167.  Rosmarin, A., Mattsson, J.,and Lehtinen, K., Notini,M., Nylen E., 1986. Ophelia (Suppl. 4):219-224.  Skoog, D.A., and West, D.M., 1980. Analytical Chemistry. Philadelphia: Saunders College. 390 - 407.  Sidgwick, N., 1950. The Chemical Elements and their Compounds. Oxford: Oxford University Press. 81  References  Smook ,G. 1992. Handbook for Pulp and Paper Technologists. 2nd ed. 171 Vancouver, BC: Angus Wilde Publications Solomon, K., Bergman, H., Hugget, R., Mackay, D., and McKague, B., 1993. A Review and Assessment of the Ecological Risks Associated with the use of Chlorine Dioxide for the Bleaching of Pulp. Erin, ON: Alliance for Environment Technology.  Taube, H., and Dodgen, H., 1949. Journal of the American Chemical Society 71:3330. Vogel, A, 1978. Quantitative Inorganic Analysis, 4th edition. New York: Longman. Wartiovaara, I., 1985. The Role ofInorganic Oxidation-Reduction Reactions in Pulp Bleaching with Chlorine Dioxide. Ph.D.: University of Helsinki.  Waters, W. and Littler, J., 1965. Oxidation by Vandium(V), Cobalt(m), and Manganese(III). In Inorganic Chemistry, Volume 5-A. Wiberg, ed. New York: Academic Press.  82  APPENDIX A  T H E DETAILED CATALYTIC BLEACHING PROCEDURE  83  CATALYTIC BLEACHING PROCEDURE  The bleaching runs were performed using the bag bleaching technique. The following steps were followed for each bleaching run:  1. The concentration of the chlorine and chlorine dioxide solutions was determined using the iodometric determination technique described in Section 3.2. From this determination, the volume of chlorine and chlorine dioxide to be used in the bleaching run was computed. 2. The pulp was weighed out to the nearest 0.0lg and added to the water required to bring thefinalbleaching solution to 3.5% consistency. 3. The required volume of catalyst was charged to the pulp and water mixture. 4. While mixing the pulp suspension thoroughly with a mechanical stirrer, the pH of the solution was adjusted to the desired level using concentrated hydrochloric acid. 5. The pulp suspension was then poured into the polyethylene bag. A clamp was placed across the bag, just above the level of the pulp suspension. The top of the bag was sealed with multiple heat treatments. 6. The chlorine and chlorine dioxide solutions were then charged via syringe from stoppered glass neckflasksto the empty upper section of the bag. The punctures in the bag were isolatedfromthe main bag section using multiple heat seals. 7. The clamp was removedfromthe bag and the water was squeezed carefully from the pulp suspension into the oxidant solution and mixed. 8. Timing was started when the diluted oxidant charge was added and kneaded thoroughly to mix with the pulp in the bag. The bag was then immersed in a constant temperature bath. 9. The pulp suspension was thoroughly mixed after 10 minutes and at 20 minute intervals thereafter for thefirsthour. The mixing was performed at  84  30 minute intervals following thefirsthour for the duration of the bleaching run. 10. The bags were removedfromthe constant temperature bath and rapidly cooled in a cold water bath; The pulp was thenfilteredfromthe bleaching liquor and washed with water using an aspirated Buchner funnel apparatus.  85  APPENDIX B T H E CHLORINE DIOXIDE GENERATION PROCEDURE  86  T H E CHLORINE DIOXIDE GENERATION PROCEDURE The apparatus used for the production of the chlorine dioxide solution is illustrated in Figure B.l. The chlorine dioxide solution was prepared by acidifying sodium chlorite. 50 g of sodium chlorite was added toflasknumber 1. 300 and 100 millilitres, respectively, of 200 g/1 sodium chlorite solution was added toflasks2 and 3. Flask number 4 was filled with 1 litre of deionized water and cooled in an ice bath. The aspirator attached to the system was then turned on. The reaction was initiated by dropping 10N sulfuric acidfromthe dropping funnel, apparatus number 7, to the concentrated chlorite solution inflasknumber 1. Thefirst5 millilitres of 10 N sulfuric acid were added very slowly and the reaction was monitored carefully. A vigourous reaction would start immediately and vent number 6 was left open to ensure adequate air flow through the generating flask. While maintaining a controlled reaction, more sulfuric acid was added until a total of 500 ml of 10 N sulfuric acid was added to the system.  Once the addition was complete, vent number 6 was closed and the chlorine dioxide/acid solution in flask 1 was stripped with air until the gaseous green colour of the chlorine dioxide entering flask number 4 was nearly gone. The result was a chlorine dioxide solution of approximately 4 to 5 g/1 with an undetectable amount of chlorine in solution.  87  88  APPENDIX C DESIGN AND ANALYSIS OF INDUSTRIAL EXPERIMENTS (MURPHY, 1977)  89  Design and analysis of industrial experiments The basic principles and practical aspects of experimental design are presented here for the various stages of an R & D investigation. Statistical data analysis techniques that can be applied with the designs are also provided. Thomas D. Murphy, Jr., American Cyanamid Co., Bound Brook, N.J.  I I Statistical design o f experiments is a proven techn i q u e that continues to show increasing use i n the c h e m i c a l process industries ( C P I ) . A s the R & D function comes under increasing pressure to produce fast, accurate results, more chemists a n d c h e m i c a l engineers are r e c o g n i z i n g the assistance that experimental design can provide. W h a t are the benefits o f this technique? Perhaps the most i m p o r t a n t one is that it c a n give more information per experiment t h a n u n p l a n n e d approaches. Chemists a n d c h e m i c a l engineers w h o practice statistical design say t h a t it reduces their lead time a n d improves their efficiency, p a r t i c u l a r l y w h e n m a n y variables are of potential i m p o r t a n c e . ' A second benefit is a n o r g a n i z e d approach toward the collection a n d analysis o f i n f o r m a t i o n . Very often, the conclusions from a statistically designed experiment are evident w i t h o u t extensive statistical analysis. A haph a z a r d a p p r o a c h , on the other h a n d , can yield results that are difficult to extract, even by a knowledgeable statistician. A n o t h e r a d v a n t a g e is an assessment of information r e l i a b i l i t y i n the light of experimental a n d analytical v a r i a t i o n . T h i s self-criticism of the results, when presented in a report, lends more c r e d i b i l i t y to the experimenter's conclusions, since nearly everyone reading the report w i l l a p p l y some o f his own j u d g m e n t of the r e l i a b i l i t y o f the data. A fourth benefit is the c a p a b i l i t y to see interactions a m o n g e x p e r i m e n t a l variables, leading to more-reliable predictions o f the response data i n areas not directly covered by e x p e r i m e n t a t i o n . M a n y investigators mistakenly persist i n the belief that experimental variables have a n a d d i t i v e effect w h i c h is often not borne out in c h e m i c a l practice.  Stages of investigation C h e m i c a l investigations pass through a well-defined series of stages, w h e t h e r . t h e y are bench studies, pilotplant programs, plant trials or plant startups. E a c h stage presents its o w n objectives and difficulties, as follows: Familiarization—In this stage the investigator is be-  c o m i n g a c q u a i n t e d w i t h a system, perhaps trying to duplicate patent results or experiments from a prior study, or learning to operate new equipment. Statistical design is generally not very fruitful here, since most experiments are c o n d u c t e d o n an i n t u i t i v e basis. Variable screening—After the goals of the investigation are better defined, the next step is to reduce the large n u m b e r of potential variables to a few effective ones. U n f o r t u n a t e l y , the i n t u i t i v e a p p r o a c h o f the previous stage is often carried a l o n g i n t o this one, investigating variables one or two at a time, a n d e n d i n g u p w i t h a large mass of disorganized data. Statistical design is very useful a n d p r o d u c t i v e at this stage, a n d imposes a considerable a m o u n t o f discipline on the experiment. Optimization—In progressing to the o p t i m i z a t i o n stage, the n u m b e r o f variables has been reduced to an effective few, but it is necessary to find their best settings in order to make the best product at the lowest cost, effluent, or energy. Statistical design can help to arrive q u i c k l y a n d efficiently at the o p t i m u m , a n d it can provide a mechanism for e v a l u a t i n g tradeoffs a m o n g product properties, costs, a n d other i m p o r t a n t attributes. Mechanistic studies—If the process or product looks technically p r o m i s i n g , a n d c o m m e r c i a l signs are favorable, then a theoretical or mechanistic m o d e l may be necessary to a i d in plant design. Statistical experimental design can assist i n choosing a m o n g c a n d i d a t e theoretical models and in estimating the system parameters most precisely. B o x a n d L u c a s [4] describe a useful procedure for selection of designs in this phase. T h e purpose of this article is to introduce the experimenter to the m a n y practical aspects o f experimental design, and to indicate where statistical methods can assist in a r r i v i n g at m e a n i n g f u l conclusions from experimental data. T h e following section covers the basic principles of statistical design for the i m p o r t a n t variable-screening and o p t i m i z a t i o n stages of the investigation. T h e final section gives some statistical-data-analysis techniques that can be a p p l i e d w i t h the designs.  Designing  experiments  T h e many elements to consider in an experimental design are shown i n T a b l e I. We w i l l define a n d discuss  90  jjjgsc elements, s h o w i n g h o w their interrelationships influence the resulting design, a n d h o w the design relates to a p a r t i c u l a r stage o f the e x p e r i m e n t a l p r o g r a m ; simple e x a m p l e w i l l be used to illustrate.  Elements of experimental design  g  Since the w o r d " e x p e r i m e n t " m a y be a m b i g u o u s , this word, as well as other terms used i n a statistical design, will be defined. A n e x p e r i m e n t a l r u n , (here shortened to tun) is s i m p l y a single experiment. A g r o u p o f runs, directed t o w a r d m e e t i n g some objective, w i l l be termed an experimental design, or s i m p l y a design. The a n a t o m y o f a r u n is p i c t u r e d i n F i g . 1. T h e outcome of a r u n is a response or o b s e r v a t i o n m a d e o n a physical e x p e r i m e n t a l u n i t . T h e v a l u e o f a response for any given r u n w i l l v a r y d e p e n d i n g o n the settings o f one or more e x p e r i m e n t a l variables, o r factors, w h i c h are under the direct c o n t r o l o f the experimenter. I n a plastics-molding e x p e r i m e n t , for e x a m p l e , the e x p e r i m e n t a l unit may be a m o l d e d - p l a s t i c specimen, the factors m a y be m o l d i n g c o n d i t i o n s , such as temperature; pressure and dwell t i m e , a n d one response m i g h t be tensile strength. Sometimes the e x p e r i m e n t a l u n i t m a y not be well defined. I n a c h e m i c a l r e a c t i o n , the e x p e r i m e n t a l u n i t might be thought o f as the a p p a r a t u s , the materials a n d the technician. Unfortunately for experimenters, this p i c t u r e o f a n experimental r u n also includes the m a n y u n c o n t r o l l e d (and even u n k n o w n ) variables that also influence response. T h e s e v a r i a b l e s give rise to b o t h systematic a n d random v a r i a t i o n s that tend to mask the true effects o f the factors o n the response. E x a m p l e s o f systematic effects are a m b i e n t c o n d i t i o n s a n d differences i n e q u i p ment, starting materials a n d technique. S o m e r a n d o m variations m a y be caused b y w e i g h i n g errors, m a t e r i a l transfers, a n d i n s t r u m e n t readings. A g o o d e x p e r i m e n tal design must consider the i m p a c t o f such v a r i a b i l i t y on the a b i l i t y to d r a w s o u n d conclusions a n d meet experimental objectives.  Response v a r i a b l e s Response variables are the d a t a from a n tal run. In most investigations there are sponses, some o f w h i c h m a y be conflicting. ency may l e a d to eventual compromises i n the final decision.  experimenseveral reT h i s tenda r r i v i n g at  It is good practice to have well-defined procedures for performing the e x p e r i m e n t a n d m e a s u r i n g the response. Too often, the experimenter m a y be u n f a m i l i a r w i t h the details of the a n a l y t i c a l m e t h o d or p h y s i c a l test. If so, a discussion w i t h the person responsible for this work would be i n order. If o n l y a p o r t i o n o f the entire experimental unit is used for testing, a n a p p r o p r i a t e s a m p l i n g procedure must also be w r i t t e n . T h i s is p a r t i c u l a r l y important i f the e x p e r i m e n t a l u n i t is nonhomogeneous, such as in the case o f a slurry or a m i x t u r e o f solids. Responses c a n be classified a c c o r d i n g to measurement scale into three m a i n types: q u a n t i t a t i v e , q u a l i t a e , or q u a n t a l . T h e q u a n t i t a t i v e or c o n t i n u o u s response, such as y i e l d , color or tensile strength, is the easiest to w o r k w i t h i n subsequent analysis. M o r e difficult to handle, b u t frequently e n c o u n t e r e d , are q u a l i t a e or categorical responses like luster, lumpiness or °dor. These are easier to evaluate if a 5 - 1 0 - v a l u e n u -  1 .  S t a t e m e n t  2 .  L i s t o f  r e s p o n s e  3 .  L i s t  f a c t o r  o f  o f  4 .  M a t h e m a t i c a l  5 .  C h o i c e  6. S  i  o f  z  e o  7 .  O r d e r  o f  8.  R  e  c  o  p r o b l e m v a r i a b l e s v a r i a b l e s m o d e l  f a c t o r f d  Table I  r  e  s  l e v e l s i  g  n  e x p e r i m e n t a t i o n r  d  i  n  g o  f d  a  t  a  m e n c a l scale can be constructed, c o n v e r t i n g the response to a semiquantitative variable. F o r e x a m p l e : 0  ±1  ±3  Equal to standard Slight difference from standard Moderate difference from standard Extreme difference from standard  T h e q u a n t a l or b i n a r y response has two values, pass o r fail, and this situation comes u p i n d r u g testing or i n destructive physical testing such as i m p a c t strength. T h i s response m a y be made semicontinuous b y p o o l i n g the results o f several identical e x p e r i m e n t a l runs into a "percentage responding" result, but special techniques are often necessary, as described by F i n n e y [ / / ] . In order to design a m e a n i n g f u l experiment, a n estimate o f the response v a r i a b i l i t y must be a v a i l a b l e . T h i s information w i l l determine the n u m b e r of runs r e q u i r e d in the design, as w i l l be discussed later. V a r i a b i l i t y is expressed i n terms o f the standard d e v i a t i o n (a). F o r those u n f a m i l i a r w i t h this concept, the s t a n d a r d d e v i a tion is roughly * the average difference between responses from duplicate experiments. T h e s t a n d a r d deviation is assumed to be constant over the range o f response values to be encountered d u r i n g e x p e r i m e n t a tion. If this can not be assumed, the f o l l o w i n g procedures can be modified to cope w i t h the p r o b l e m .  Factor variables Factors, or experimental variables, are c o n t r o l l e d by the experimenter. T h e factor level is the v a l u e or setting of a factor d u r i n g an e x p e r i m e n t a l r u n . L i k e responses, they can be classified, a c c o r d i n g to measurement scale, * M o r e precisely, o is ;  he expected difference between duplicates.  Uncontrolled or  u v  unknown  variables  tlv  Representation of the experiment  91  1 l  as q u a n t i t a t i v e or continuous (temperature, time), or as q u a l i t a t i v e o r categorical (catalyst type, solvent). T h e . latter are the most difficult to work w i t h , since the measurement scale u s u a l l y has no n a t u r a l ordering. W h e n experimenters have a choice o f available measurement scales, they must select the one most app r o p r i a t e to the technical situation. F o r example, h y drogen-ion concentration c a n also be expressed as p H . If there are two o r more factors, functional c o m b i n a tions o f factors m a y be more m e a n i n g f u l than the original ones. F o r example, instead o f m a n i p u l a t i n g the starting concentrations o f reactants A a n d B , it might be more sensible to t h i n k i n terms of the reactant ratio, A / B , a n d total reactants, A + B . T h e n u m b e r o f potential factors w i l l be greatest at the start o f a n investigation, w h e n the least is k n o w n a b o u t the system, a n d w i l l usually d r o p to a relative few as knowledge develops. A t the start o f the variablescreening phase, every effort must be m a d e to list a l l factors thought to be i n f l u e n t i a l . T h e practice o f introd u c i n g new factors i n the course of an investigation w i l l not be as efficient, a l t h o u g h it m a y sometimes be u n avoidable.  M a t h e m a t i c a l models W h e n responses a n d factors are continuous in scale, it is useful to consider the factor-response relationship i n terms o f a m a t h e m a t i c a l f u n c t i o n , or model. Interpretation o f the experimental results w i l l be more economical i n thought a n d less a m b i g u o u s w h e n considered w i t h i n the discipline o f the m a t h e m a t i c a l model. A t the start o f a n investigation, w h e n little is k n o w n about the  true relationships, a n e m p i r i c a l m o d e l , such as a first- on second-degree p o l y n o m i a l , w i l l suffice to give rough^ insights. A t the later stages, a theoretical m o d e l , derived; from first principles, m a y be necessary to give the re-' q u i r e d accuracy o f p r e d i c t i o n over a w i d e range of; conditions. ; In the case o f a single factor, the simplest empirical model is the first-order function: Y = b + b X. The model parameters, b a n d b , are termed the intercept a n d slope, respectively. T h i s m o d e l , as pictured in F i g . 2, is useful for p r e d i c t i n g a response Y over a limited range of the factor X. It can also be used i n the variable-screening stage, where the interest is centered on the factors h a v i n g largest effect o n the response Y. If i j is close to zero, we say that the factor has no significant effect o n the response. M o s t o f us, however, tend to believe that the response-factor relationship is a c o n t i n u o u s curve. The second-order function, Y = b + b^X + b X , gives a reasonably satisfactory fit i n these situations ( F i g . 3). Here we wish to determine a rough o p t i m u m , located at a factor level equal to —b /2b , where b is termed the curvature. If b is close to zero, the response is said to be approximately first-order w i t h respect to the factor, j For two or more factors, a c o m p l i c a t i n g situation k n o w n as interaction m a y exist, w h i c h means that the. factors do not operate independently o n the response. (If independent, the factors are said to be additive.) W h e n there is little o r no curvature w i t h respect to either factor, the simplest two-factor m o d e l is: Y = b + i'jA'j + b X + b X X . T h e parameters Aj and b, are the slopes corresponding to the factors X a n d A 0  0  x  x  2  0  1  Jl  xl  ll  u  0  2  2  12  1  2  1  92  2  Nomenclature ^  *.«  *« C  A d N  Y Y  Intercept term i n m o d e l Slope o f factor i i n m o d e l C u r v a t u r e o f factor i i n m o d e l Interaction o f factors i a n d j i n m o d e l N u m b e r o f centerpoints i n design M a i n effect o f d u m m y factor Number o f d u m m y variables N u m b e r o f f a c t o r i a l runs i n design N u m b e r o f factors i n design N u m b e r o f replicates o f a r u n Estimate o f response s t a n d a r d d e v i a t i o n Estimate o f response v a r i a n c e Student's t statistic T o t a l degrees o f freedom Level o f factor i Response v a l u e Average response Distance from o r i g i n o f a x i a l o r "star" point in a central composite design M i n i m u m response change o f technical interest Response s t a n d a r d d e v i a t i o n  When b , the i n t e r a c t i o n p a r a m e t e r , is zero, we have a ttrictly first-order a d d i t i v e m o d e l as p i c t u r e d i n F i g . 4 . A plot o f Y vs. X at a n y two constant values o f X w i l l be a set o f two p a r a l l e l straight lines. When b 7 ^ 0, the m o d e l is interactive. T o appreciate the effect o f the i n t e r a c t i o n terms, we c a n recast the model to emphasize the effect o f X o n the Y vs. X relationship: l2  l  2  l2  2  .;  Y=[b  0  + bx] Intercept 2  +  2  l  [b. + b ^ x , Slope  In this interactive m o d e l , the v a l u e o f X affects the ilope of the Y vs. X relationship, as well as the intercept. T h i s gives rise to two n o n p a r a l l e l lines at two constant values o f X ( F i g . 3). A l t h o u g h this model is •eoond-order i n X\ a n d X , a n y plot o f Y vs. one factor, it a constant value o f the second factor, .will be a Kraight line. 2  l  2  2  When the u n d e r l y i n g relationship between the response a n d a factor is c u r v e d , then a quadratic or curvature term is a d d e d . T h e full second-order model in two variables is: Y = b + bX 0  x  x  + bX 2  2  + ^n-^'i + b X X l2  1  2  +  b X 22  2  This k i n d o f m o d e l w i l l give an excellent description °f the response w i t h i n the region o f experimentation, ^ d can be used for interpolative purposes or to find a ugh internal o p t i m u m , if one exists. F i g . 4 is a plot of X at constant values of A ' . A n o t h e r useful plot is 'contour m a p p i n g o f the response surface Y over the factor space denned b y the values o f X^ a n d X ( F i g . 5 ) . ^•is plot indicates a m a x i m u m response near the center i h e factor space. It should be emphasized, however, second-order e m p i r i c a l models are very unreliable for prediction outside the range o f experimentation. Because o f the good (perhaps almost too good) perro  {  0  2  of  Full second-order model in two factors  Fig. 4  X , 2  1  ° Y  Y  T w o factors - 2  factorial  J  High level  X,  Run  2 +  Wsjr X , 2  <i  y,  F *  /  X,  1  Low level  3 4  +  4 f  V,  t  Mign level  L o w level  X, Three factors - 2  3  Interaction effect  {Y - Y )-(Y - Y ) (Y - Y -(^2 t  2  <V, + V ) 4  2  2  3  y  t  {Y + Y ) 2  2  3  factorial  - V , )  Run  2  X,  1  2 +  3  3 4  +  5  Choice of factors for two levels  6  +  7 8  formance o f second-order models, i t is seldom useful to consider t h i r d - or higher-order p o l y n o m i a l s . S u c h m o d els m a y predict several o p t i m a a n d m a y give the i m pression that the response behaviors are u n d u l y c o m p l i cated w i t h respect to the factors. F o r n factors, the full second-order m o d e l is  j = l  i = l j = i  which has [(n + \){n + 2)/2] parameters to be estimated. Since this n u m b e r o f parameters can become large very q u i c k l y w i t h n, a s i m p l e r m o d e l is generally assumed at the start o f the investigation (such as a first-order model). A s factors are e l i m i n a t e d by experim e n t a t i o n , the m o d e l c a n be g r a d u a l l y extended to p a r t i a l or full second-order, or it c a n be replaced by a theoretical model. F o r q u a l i t a t i v e factors, there is no continuous function l i n k i n g the response to the levels of a factor. It is necessary to think i n terms o f c o m p a r i s o n of response between two levels o f a q u a l i t a t i v e factor, a l l other factors b e i n g h e l d constant. V e r y often the best app r o a c h is to m o d e l each level o f the q u a l i t a t i v e factor separately i n terms o f the other (continuous) factors. For more than one q u a l i t a t i v e factor, each c o m b i n a t i o n of q u a l i t a t i v e factor levels must be so considered. T h i s illustrates the c o m p l i c a t i n g aspect o f using qualitative factors i n the design. A statistical consultant can be of great assistance i n these circumstances.  C h o i c e o f factor l e v e l s A n experimental design consists o f a set o f experimental runs, w i t h each r u n defined by a c o m b i n a t i o n of factor levels. T h e choice o f factor levels is influenced by the m a t h e m a t i c a l m o d e l u n d e r consideration. A n experimental design is therefore determined by the n u m ber a n d type o f factors, together w i t h the mathematical model. F o r the simplest case o f a single factor a n d first-order  +  Examples of 2" factorial designs  Fig. 7  model, o n l y two factor-levels are necessary to estimate the parameters b a n d b . I n this discussion, these two levels w i l l be coded as " l o w " a n d " h i g h " , or " — " and " + " . T h e change i n response between the levels is termed the factor m a i n effect. Since r a n d o m error can easily obscure the m a i n effect i f levels are set too close together, the m a i n effect c a n be estimated most precisely from extreme level-settings o f the factor, as shown in F i g . 5. T h e m o d e l slope parameter b is e q u a l to the m a i n effect d i v i d e d by the difference i n the a c t u a l , or uncoded, values o f the factor levels. It is more a d v a n t a geous, however, to work w i t h m a i n effects rather than slopes, since the former c a n be c a l c u l a t e d more easily. If a second-order m o d e l is under consideration, a third factor-level is r e q u i r e d to estimate the curvature parameter b . T h i s t h i r d level, n o r m a l l y r u n halfway between the extreme factor levels, is termed the centerpoint, and is coded as " m e d i u m " or " 0 " . T h e curvature effect is defined as the difference between the experimental centerpoint response a n d the centerpoint value expected from a first-order m o d e l fit to the extreme points ( F i g . 5). 0  x  x  u  M o r e than 3-5 factor levels are seldom necessary when using first- or second-order e m p i r i c a l models. If the factor is q u a l i t a t i v e , however, the n u m b e r o f factor-levels w i l l be dictated by the experimenter. W h e r e theoretical models are u n d e r consideration, more than three factor-levels may be necessary to choose the best a m o n g several candidate models. F o r more than one factor, each e x p e r i m e n t a l r u n is defined as a c o m b i n a t i o n o f factor levels. In the twofactor case, the interaction effect must be estimated in addition to the two m a i n effects. T h e m i n i m u m n u m b e r  94  D e s i g n - a  - 1  0 .  + 1  a  1 4  Three  f  a  c  t  o  r  s  X *  ,  C e n t e r p o i n t s  0  *2 0  •  1 ( d a r k  p o i n t s )  R u n  F a c t o r i a l  *3  6  I  0  —  p o i n t s  7  -  *3  -  + +  +  -  +  D e s i g n  -  -  2 ( l i g h t  R u n 2 3  + +  5 8  p o i n t s )  *2  + — — +  +  _ +  +  Half replicate of a 2 design 3  Fig. 9  factor. These a d d i t i o n a l runs (called a x i a l or " s t a r " points), combined w i t h the 2 factorial plus centerpoint, form a central composite design, as illustrated i n F i g . 8 for the two- and three-factor cases. T h e r e c o m m e n d e d level for a—the distance from the centerpoint—ranges from 1 to 2 , where n is the n u m b e r o f factors, a n d the factorial points are at unit distance i n each factor from the centerpoint. n  Star  0  0  -cc a  0  -ct  0  0  a  0  0  0  0  0  points  0  —or  0  n / 4  I  Central composite designs  of runs w i l l therefore be four, since we must estimate two slopes o f the Y — X relationship a n d compare these slopes at two values of X to check o n interaction. If the interaction effect is zero, the lines are p a r a l l e l as in Fig. 6. The four-run design described above is k n o w n as a 2 factorial design, a n d is a m e m b e r of a class of designs known as 2 factorials (n factors each r u n at two levels with all factor-level c o m b i n a t i o n s represented). These designs are useful for estimation o f m a i n effects a n d interactions. T h e factor-level c o m b i n a t i o n s o f the 2 and 2 designs (see F i g . 7) c a n be represented as the comers of a square a n d a cube, respectively. x  2  2  n  2  3  Curvature effects c o u l d be estimated by a d d i n g a third level of each factor at various c o m b i n a t i o n s of the remaining factors. In the early investigative stages, however, it is more e c o n o m i c a l to assess the overall curvature of the system. T h i s can be done by the addition of one c o m b i n a t i o n of a l l factor centerpoints, located at the centroid of the design. F o r 2 a n d 2 degns, this centerpoint r u n w o u l d be located at the centers of the square a n d cube, respectively, as shown in ^'g- 7 T h e overall curvature effect is estimated as the difference between the response at the centerpoint and e expected response, assuming a first-order model in " factors. T h i s expected response value is estimated from ihe average of the 2" points. 2  A major disadvantage o f 2" factorials is the large number of distinct runs required as n increases. T o circumvent this problem, suitable fractions, or subsets, of the entire design can be used. It w i l l be necessary, however, to then sacrifice some or a l l of the i n t e r a c t i o n estimates. Fractional 2" designs are useful i n the v a r i a ble-screening stage, and can be e x p a n d e d to larger fractions or even full 2" designs as desired. A n e x a m p l e of a half fraction of a 2 design is s h o w n i n F i g . 9. E i t h e r set of four runs can provide main-effect estimates for a first-order model in three factors. In the variable-screening stage, e x p e r i m e n t a l designs are built around full or fractional 2 designs w i t h a single centerpoint. These designs are suitable for estim a t i o n of m a i n effects a n d overall c u r v a t u r e , a n d , i n certain cases, groups of interactions. D e p e n d i n g o n the experimental results a n d the n u m b e r o f factors r e m a i n ing, these designs can be augmented to e x p a n d the range of the existing factors, a n d to estimate further second-order model parameters i n the o p t i m i z a t i o n stage. T h u s , most chemical-process investigations c a n take place in stages, w i t h the e x p e r i m e n t a l objectives changing in accordance with the results o f the previous stage. 3  n  3  u  lr|  a  If significant overall curvature is detected by the centerpoint r u n , a d d i t i o n a l runs c a n be subsequently 'dded to make separate estimates of c u r v a t u r e for each  Size of design . T h e precision of an estimate of m a i n effect, interaction, or curvature depends on the n u m b e r o f data used in the estimate, together with the m a g n i t u d e o f the response error o. In a 2" design, a l l the runs are used i n each estimate of m a i n effect a n d interaction. T h e experimenter must, therefore, set the total n u m b e r o f runs, A/, required to achieve the desired precision. To do this, the experimenter determines the m i n i m u m change in response of technical interest, denoted  95  by the s y m b o l A . T h e e x p e r i m e n t s h o u l d enable this effect to be declared statistically significant w i t h a h i g h degree o f assurance ( i f it i n d e e d exists). T h e relationship between N a n d the ratio A / o has been developed by statisticians, a n d is s h o w n i n F i g . 10. I f we wish to detect a n effect equal to twice the response error ( A / a — 2), then about 14 runs w i l l be r e q u i r e d . I n order to detect a response change o f a m a g n i t u d e e q u a l to the response error ( A / a = 1), a 5 0 - r u n experiment w i l l be necessary. If N is m u c h larger t h a n the chosen design, based on the m a t h e m a t i c a l m o d e l , then it m a y be necessary to repeat, or replicate, the entire design one or more times. T h e p r i m a r y purpose o f this r e p l i c a t i o n is to p r o v i d e sufficient precision fqr the estimates. A v a l u a b l e secondary purpose is to p r o v i d e a more-precise estimate o f o, w h i c h c a n then be used to establish the size of succeeding designs.  Order of experimentation T o g u a r d against systematic trends i n u n c o n t r o l l e d (or u n k n o w n ) variables d u r i n g execution o f the design, it is p r u d e n t to r a n d o m i z e the order i n w h i c h the runs are made. If the response-measurement process is c o n d u c t e d as a separate o p e r a t i o n , the r a n d o m i z a t i o n procedure s h o u l d be a p p l i e d here as well. T h i s w i l l , o f course, require c o o r d i n a t i o n w i t h the a n a l y t i c a l g r o u p responsible for the measurements. A n y r a n d o m process of o r d e r i n g the runs w i l l suffice, but d r a w i n g numbers out o f a container is as g o o d a m e t h o d as any. In some cases it m a y be inconvenient or i m p r a c t i c a l to r a n d o m l y v a r y the level o f one o r more factors, since this m a y require, for e x a m p l e , a lengthy temperature adjustment or a major m e c h a n i c a l change. T h i s gives rise to a special s i t u a t i o n k n o w n as split plotting (the term o r i g i n a t i n g from a g r i c u l t u r a l experiments), a n d a statistician s h o u l d assist i n the analysis o f such data. It m a y also be recognized that the experimental units are not homogeneous, but m a y differ due to some k n o w n u n c o n t r o l l e d v a r i a b l e . T h i s situation requires that the e x p e r i m e n t a l units be segregated into blocks of homogeneous e x p e r i m e n t a l units. E a c h block can be considered as a level o f a n extraneous, or b l o c k i n g , variable. E x a m p l e s o f b l o c k i n g variables are source or b a t c h o f materials, type of e q u i p m e n t , seasonality, people, or cages of animals. T h e block size is the m i n i m u m n u m b e r o f e x p e r i m e n t a l units a v a i l a b l e in a block. If block size is greater or equal to h a l f the runs in a full or fractional 2" design, then the b l o c k i n g factor can simply be i n c l u d e d as another factor in the design. If not, a statistician can be consulted to set up an appropriate b l o c k i n g pattern. It is wise not to ignore blocking v a r i a bles, for d o i n g so can result in the inflation of the error or the m i x i n g o f the block effect w i t h the other effects in the m o d e l .  R e c o r d i n g of data After the design is chosen, the coded low and h i g h levels for each factor are converted to actual factor settings a n d listed i n the e x p e r i m e n t a l instructions. These s h o u l d c o n t a i n the experimental-run procedures, s a m p l i n g procedures, a n d a n a l y t i c a l documentation. If any o f the factor levels cannot be controlled pre-  cisely, but c a n be measured, then the a c t u a l v a l u e (or average of repeated measurements) s h o u l d be taken as the factor level for statistical analysis o f the experiment. A n y u n c o n t r o l l e d variables that can be measured, a n d w h i c h c o u l d possibly influence the response, s h o u l d also be recorded. A n y u n u s u a l occurrences s h o u l d also be noted, since they c o u l d e x p l a i n possible anomalous response values.  Selection of design M a n y practical a n d statistical aspects o f experimental design have been separately discussed, a n d these elements are now b r o u g h t together to formulate a design for a specific s i t u a t i o n . T h e f o l l o w i n g i n f o r m a t i o n should be a v a i l a b l e : 1. A list of responses to be studied, together w i t h an estimate of each response error, a. 2. A list of n factors to be studied, together w i t h the ranges (low a n d h i g h levels) for each q u a n t i t a t i v e factor, and a list o f levels for each q u a l i t a t i v e factor. 3. A statement o f the p r o b l e m to be investigated, together with an a p p r o p r i a t e m a t h e m a t i c a l m o d e l . 4. T h e m i n i m u m n u m b e r o f runs (N) needed to attain the required precision for estimates of the factor effects. 5. A check to see whether all experimental units are homogeneous; if they are not, the selection o f a b l o c k i n g scheme is made. T h e situations o f v a r i a b l e screening, r o u g h o p t i m i z a tion, and final o p t i m i z a t i o n w i l l each be considered in order. D u r i n g the variable-screening phase, the p r i m a r y emphasis is on identification of the most i m p o r t a n t factors. Secondary goals w i l l be rough detection o f second-order effects (curvature a n d interaction), a n d o b t a i n i n g a better estimate of a (response error) for vise in later investigative phases. T h i s a i m is reflected i n the  96  Twelve-run screening design  R u n  X  + +  1  +  +  -  -  -  +  — + +  + + + +  1  2  3  +  climates  +  -  -  -  + -  +  -  4  5  6  7  1 3  1 2 3 4  jfmain  4 5  3 5  2 5  1 5  2 3  jflects a n d  3 6  4 6  1 6  2 6  1 4  2 4  fractions  2 7  1 7  4 7  3 7  6 7  5 7  5 6  +  X  x  1  2  2  A  8 9  +  +  1 0  -  1 1  5 6  1  1  n  3  X  7  8  X  9  x  1 0  +  -  -  +  -  +  +  -  +  +  +  -  -  -  _ +  -  +  +  +  +  +  _ -  +  + +  +  -  -  + —  — +  + —  +  _  -  -  -  7  8  15  -  +  16  +  • +  Estimates  1  2  2  .+  7  -  1  +  8  -  x  -  + +  +  1 1  +  -  14  X  -  +  13  -  +  -  +  +  -  6  +  + +  +  5  12  -  -  +  +  + +  -  +  +  +  _  -  -  +  -  +  11  -  +  +  -  -  -  _  +  +  + + . +  •  -  -  — + + + + . + + -  +  -  5  6  X  + +  -  7  +  _  -  -  -  + +  +  +  X  + +  + +  -  -  + +  +  8  X  x  X  9  _ + _  •  -  +  -  + + — + + +  + +  -  -  + + +  + + +  + _ + +  -  -  -  +  +  +  +  -  +  +  +  +  -  -  . -  +  +  +  +  +  +  -  +  -  +  +  -  -  +  4  x  1 3  + +  + + + +  + +  X  1 4  X  + +  + +  + +  + +  1 5  + + + +  -  + + + +  16  12  13  14  23  24  34  main  45  35  25  56  15  36  26  effect and  37  47  67  27  46  17  57  interactions  28  68  48  38  28  58  18  of  10  Factors can be assigned to the design c o l u m n s i n a n y convenient manner. T h e unassigned c o l u m n s , termed  X  3  9  X  4  mental r u n . E a c h design can a c c o m m o d a t e u p to k — 1 factors i n k runs, but the usual practice is to leave 2-5 unassigned columns for groups o f interactions. T h e n u m b e r o f runs i n the chosen design is c o m p a r e d w i t h A^, the n u m b e r o f runs required for precision o f the estimates. If N is m u c h larger, either the design must be replicated or a larger design must be selected from Tables I I - V I to a p p r o x i m a t e l y a t t a i n the r e q u i r e d runs. F o r example, i f n = 7, the 12-run design i n T a b l e III m i g h t be chosen. If N'— 14, this design c a n be used, but i f A ' = 23, either the 12-run design w o u l d have to be d u p l i c a t e d or the 24-run design w o u l d be selected.  n  4  10  x  3  1 2  —  1 2  -  7  x  2  Table HI  PP  - Sixteen-run screening design Run  + + +  4  ^mathematical model: Y — b + b X + • • • + b X l l + (overall c u r v a t u r e ) + (interaction). S} I n d i v i d u a l i n t e r a c t i o n a n d c u r v a t u r e effects are not "directly estimated, but their presence c a n be measured "indirectly. Some useful screening designs are given i n Tables I I - V I , w i t h r u n sizes r a n g i n g from 8 to 24 i n multiples of 4. L a r g e r - s i z e d designs, u p to 100 runs, are given by Plackett a n d B u r m a n [14~\. A t h o r o u g h discussion o f fractional 2" designs is g i v e n b y B o x a n d H u n t e r [3]. E a c h table consists o f a n array o f — a n d + values, which correspond to the low a n d h i g h levels of each factor, w h i c h a p p e a r as c o l u m n headings. E a c h row i n the array defines the factor levels for a given experi0  -  3  -  + +  +  2  + +  x  1  -  97  + + + —  -  X  11 + + — + + +  _ — -  +  the d u m m y factors, are n o t used to define the experimental c o n d i t i o n s , b u t w i l l come i n t o p l a y i n the dataanalysis stage. O n l y the c o l u m n s c o r r e s p o n d i n g to real factors are used to d e t e r m i n e the experimental runs. If q u a l i t a t i v e factors are to be inserted into the design, this c a n be done at o n l y t w o levels. Screening q u a l i t a t i v e factors at m o r e t h a n t w o levels w i l l require the help o f a statistician. W i t h q u a l i t a t i v e factors at two levels, the assignment o f — a n d + to the factor levels is arbitrary. These tables represent o n l y a s m a l l fraction o f the total n u m b e r o f factor-level c o m b i n a t i o n s that c o u l d be run. It c o u l d be possible that some o f the runs given i n the tables m i g h t be i m p r a c t i c a l to r u n , for safety or other reasons. A d d i t i o n a l l y , other runs not given i n the design m i g h t be o f interest. T o a c c o m m o d a t e the experimenter, these designs m a y be m o d i f i e d by c h a n g i n g the polarity o f one or more c o l u m n s i n the design,—i.e., s w i t c h i n g the — a n d + signs i n the c o l u m n s . T h i s way, a more desirable set o f runs c a n be attained. W h e n the design is finished, the r u n order should be r a n d o m i z e d to g u a r d against systematic errors. To estimate curvature a n d e x p e r i m e n t a l error, two or more centerpoints are a d d e d to the factorial points. T h e centerpoints s h o u l d be evenly spaced over the experim e n t a l - r u n order. F o l l o w i n g the variable-screening phase, the surviving factors are studied further i n the rough o p t i m i z a t i o n phase. I f the system turns out to be rich i n surviving factors, a statistician c a n be consulted to set up the more-complex design that w i l l be required to estimate the large n u m b e r o f interaction effects. D u r i n g the rough o p t i m i z a t i o n phase, it is advisable to estimate a l l factor main-effects a n d interactions in order to determine the d i r e c t i o n o f the o p t i m u m response. T h e estimation o f o v e r a l l c u r v a t u r e a n d experi-  m e n t a l error are secondary m a t h e m a t i c a l m o d e l applies: Y=b  0  goals. T h e interactive  + b,X + . . . +b X + b XX + ••• + b X X + (overall curvature) x  12  1  2  n  n  mn  m  n  where m = n — 1. I n d i v i d u a l curvature effects are not directly estimated. T h e proper designs for this phase are full or fractional 2" designs plus centerpoints. T a b l e II c a n be used as a full 2 design or a 1/2 fraction o f a 2 design (8 runs). T a b l e I V c a n be used as a full 2 design or a 1/2 fraction of a 2 design (16 runs). T o use these tables for this phase, the factors c a n be assigned to the columns indicated i n the legend below the design array, a n d this w i l l also give the c o l u m n s corresponding to interaction estimates. In m a n y cases the screening design c a n itself be the first rough o p t i m i z a t i o n design, a n d c a n be reanalyzed in terms of the s u r v i v i n g factors. A n e x a m p l e o f this will be given in the analysis part o f this article. If the o p t i m a l response is i n d i c a t e d to lie outside the experimental region, then a d d i t i o n a l designs at new levels w i l l have to be r u n to locate the region o f the o p t i m u m . A s this process proceeds, the c u r v a t u r e estimates will generally increase to a m a x i m u m relative to the m a i n effects, i n the o p t i m a l region. T h e final o p t i m i z a t i o n phase w i l l require a full q u a d r a t i c model to accurately predict the p o i n t o f optimal response: 3  4  4  5  Y = io + Mfj  + ••• + b X n  n  + b X\  nn  + ••• +  u  b n + *i2*i*  2  +  ••• +  b XX mn  m  n  where m = n — 1. A l l of the terms i n the m o d e l w i l l be estimated. • T h e type of design used i n this phase is the central composite design, w h i c h is basically a n a u g m e n t e d 2"  98  Twenty-four-run screening design Run X  X  1  2  Table VI  5  A  1 0  A  1 1  M 2  A  1 3  k  +  1  1 4  +  2  A  1 5  x  —  1 6  x  1 7  x  1 8  X  1 9  2 0  X  ^ 2 1  —  +  x  2 3  _ — +  3 4  * 2 2  —  5  — + +  + + +  6 7  —  8 9  —  —  —  +  +  1 0 1 1  —  +  _  1 2 1 3 1 4  +  1 5  —  1 6  +  —  _  —  1 7 1 8 1 9 2 0 2 1 2 2  —  2 3  +  _  2 4  factorial plus centerpoints. T h e a d d i t i o n a l runs are solely for the purpose o f e s t i m a t i n g the c u r v a t u r e terms for each factor. E x a m p l e s o f c e n t r a l c o m p o s i t e designs are shown i n F i g . 8 for t w o a n d three factors. A n o t h e r useful class o f designs for this purpose is described b y Box and B e h n k e n [2]._  Example An oversimplified e x a m p l e s u m m a r i z e s the experimental design process. T h e p i l o t i n g o f a new b a t c h catalytic process for c h e m i c a l Q is i n the b e g i n n i n g stages of investigation. It is assumed that the only response of interest is the reaction y i e l d , a continuous variable. It is estimated from the p r i o r bench study that 'he yield v a r i a b i l i t y , a, is a b o u t 2% y i e l d .  it is advisable to choose a design that w i l l at least detect the existence o f such interactions. Economics o f this situation dictate that a 6-8% yield increase w o u l d be of vital interest in this study. A = 6% is chosen as the response difference to be detected w i t h a high degree o f assurance. E n t e r i n g F i g . 8 w i t h a A / o value o f 6 / 2 = 3, it is seen that about N — 8 runs w i l l be required. T h e experimental unit i n this case w i l l consist o f a batch of materials in a reactor. T h e same reactor w i l l be used for all runs, a n d there is sufficient m a t e r i a l for at least 25 runs. T h e experimental units w i l l , therefore, be homogeneous, a n d no b l o c k i n g variables w i l l be used. W i t h four factors, a n eight-run design from T a b l e II  Four factors have been proposed for the variablescreening phase, together w i t h their factor-level ranges: Reaction r ~p-e ,r.a .t.u ~re Reaction , t,em Reactant ratio Reaction time Catalvst level  70-80°C  v  m  1.05-1.35 1-2 h X\  0.05-0.08%  A" four factors are q u a n t i t a t i v e , a n d the extremes o f 'hese ranges are p i c k e d as the low a n d h i g h factor levels. Since the objective is to estimate m a i n effects, the appropriate m a t h e m a t i c a l m o d e l is y  =A + bX 0  x  l  + bX 2  2  Chemical Q -factorial runs  + bX 3  3  + bX i  4  +  (overall curvature) Based on past experience w i t h such systems, the likeli-  R e a c t i o n  C a t a l y s t  t e m p g r a t u r e ,  R e a c t a n t  C  r a t i o  1  70  1.05  1  0.05 0.08  R u n  R e a c t i o n time,  h  c o n c e n t r a t i o n  t/L  2  80  1.05  1  3  70  1.35  1  4  0.08  80  1.35  1  0.05  5  70  1.05  2  6  0.08  80  1.05  2  7  0.05  70  1.35  2  8  0.05  80  1.35  2  0.08  ^°od is that these factors m a y be interactive. Therefore,  99  Chemic run-ord i  « r:TlT»  I*Tl  11 r  HI  •«  V " '  R e a c t i o n R e a c t a n t  R u n  " C  r a t i o  1  7 5  1 . 2 0  R e a c t i o n time,  c o n c e n t r a t i o n .  a/L  h  1 . 5  0 . 0 6 5  O r i g i n a l r u n  7 0  1 . 3 5  1 . 0  0 . 0 8  3  8 0  1 . 0 5  2 . 0  0 . 0 5  6  4  8 0  1 . 3 5  1 . 0  0 . 0 5  4  5  7 0  1 . 3 5  2 . 0  0 . 0 5  7  6  7 5  1 . 2 0  1 . 5  0 . 0 6 5  c p  7  7 0  1 . 0 5  2 . 0  0 . 0 8  5  8  8 0  1 . 0 5  1 . 0  0 . 0 8  2  9  8 0  1 . 3 5  2 . 0  0 . 0 8  8  1 0  7 0  1 . 0 5  1 . 0  0 . 0 5  1  1 1  7 5  1 . 2 0  1 . 5  0 . 0 6 5  c p  f r o m  F i g .  1 2 ;  c p  =  e  sents the sum over i from 1 to r. A n o t h e r f o r m u l a may  CP  2  n u m b e r  where Y — ^.YJr is the average response, a n d 2 repr .  n o *  3  r u n  Se  C a t a l y s t  t e m p e r a t u r e .  • O r i g i n a l  If K j , • • •, Y represent the response values resulting from r replicate runs, then an estimate .r o f the respop. standard deviation a can be calculated as follows: T  Table VI  be more convenient to use w i t h some calculators.  s = \/2K - rF /(r - 1)  where 2 / is the sum o f the squared responses. The n u m b e r (r — 1) is termed the degrees o f freedom of the error estimate. 2  In the case o f duplicates (r = 2), the f o r m u l a is duced to: 2  w i l l be satisfactory. A s s i g n i n g c o l u m n s X t h r o u g h X to the four factors, three c o l u m n s are left over for d u m m y factors to estimate i n t e r a c t i o n . After substituting the a c t u a l factor levels i n place o f the — a n d + values i n T a b l e I I , the deisgn layout is as shown i n T a b l e V I I . T h e decision is m a d e to r u n three centerpoints to estimate overall c u r v a t u r e a n d response error. T h e centerpoints o f the factor levels X t h r o u g h X are: A  1  Temperature Reactant ratio Reaction time Catalyst level  A  75°C 1.20 1.5 h 0.065%  T h e three centerpoint runs w i l l be m a d e at the above factor settings. T h e eight factorial runs are r a n d o m i z e d by d r a w i n g the r u n numbers 1-8 f r o m a container and r u n n i n g them i n the order d r a w n . T h e centerpoints are run as the first, m i d d l e a n d last r u n o f the design. If the factorial r u n order were r a n d o m i z e d as 3, 6, 4, 7, 5, 2, 8, 1, then the o v e r a l l design w o u l d appear as i n T a b l e V I I I . T h i s e x a m p l e w i l l be used i n the following section on analysis o f designed experiments.  Analysis o f designed  Replicates are a v a i l a b l e from two sources i n designed experiments: centerpoint r e p l i c a t i o n , a n d factorial p o i n t replication. Centerpoints can a n d s h o u l d be replicated two or more times i n each designed experiment. T h i s w i l l provide greater precision for the overall curvature estimate, as well as an estimate o f response error. F a c t o r i a l points w i l l o n l y be replicated w h e n a n entire design is replicated to p r o v i d e the r e q u i r e d precision on estimates o f m a i n effects a n d interactions. T o illustrate the use o f these formulas, suppose that as a result of this article's designed experiment, the three centerpoint yields of c h e m i c a l Q h a d been 75, 80 and 76%. T h e average a n d standard d e v i a t i o n are calculated as follows: Y = ZYi/r = (75 + 80 + 76)/3 = 77% yield s = V[(75 - 77) + (80 - 7 7 ) + 2  2  t  2  = 2(r, -  l)j2/2: . _ l ) (r  where S denotes the s u m m a t i o n over ; from 1 to k. The s y m b o l v may be used for the total degrees o f freedom. T o exemplify the p o o l i n g c a l c u l a t i o n , it is assumed that two reactions o f c h e m i c a l Q at some other fixed c o n d i t i o n had been previously r u n , w i t h yields of 529c a n d 50%. Data source  Responses  Previous Centerpoints  52, 50 75, 80, 76  .  2  = [(r, -  + (r., - l)J5]/[(r, - 1) + (r - l)j 2  = [(1)(2) + (2)(7)]/[l + 2] = 16/3 = 5.33  replicates  Response error can be estimated either directly from replicate runs, or i n d i r e c t l y using d u m m y factors.  1) = 2.6% yield  w i t h (r — 1) = (3 — 1) = 2 degrees o f freedom. E r r o r estimates from several sources may be comb i n e d , or pooled, for a firmer estimate o f error. The pooled estimate is a weighted average of the variances (or squares o f the standard deviations), weighted by their respective degrees o f freedom (df). T h e pooled variance for k separate estimates of error, s , each with r,. replicates, is: J  A n a l y s i s o f statistically designed experiments is reasonably straightforward because o f the balanced nature of the designs. T h i s part o f the article w i l l outline the calculations for e s t i m a t i o n o f response error, factor main-effects, interaction a n d overall curvature, a n d assessment o f their statistical significance. These can be easily a c c o m p l i s h e d using a calculator. Such other techniques, as regression analysis a n d contour plotting, w h i c h require the a v a i l a b i l i t y o f a computer, w i l l be m e n t i o n e d . F o r basic i n f o r m a t i o n on statistical analysis, see D i x o n and Massey [9] a n d N a t r e l l a {13}. A useful c o m p e n d i u m o f statistical tables a n d experimental designs can be found i n the CRC Statistics H a n d b o o k [/].  2  (76 - 7 7 ) ] / ( 3 -  experiments  E s t i m a t i o n o f response error using  re-  K) /2  c e n t e r p o i n t .  x  2  2  s = \/5733 - 2.3 T h e new estimate of response error is 2.3% yield with 3 100  lemical Q — caIculations an A results  SHI  ff||§f|||  fslliiiiiiill able 1  C a t a l y s t T e m p e r a t u r e  Y i e l d  R u n  r e s p o n s e  n o .  X  x  1  5 8  -  2  8 3  +  3  6 2  -  1  4  7 7  +  5  7 5  -  6  8 4  +  7  7 5  3 3 0  6 6 0  2 7 0  D i f f e r e n c e 7 5  E f f e c t  .  5  X  f a c t o r s 6  X  7  -  +  +  +  -  -  + +  -  +  + +  + — — +  — + — —  + + +  + — — — — + +  — — + + — — +  3 0 0  3 2 0  3 0 6  3 0 2  2 9 0  2 9 6  3 0 0  2 8 0  2 9 4  2 9 8  3 1 0  3 0 4  -  -  6 0  0  4 0  1 2  1 5  0  1 0  3  Estimation of main effects Tables I I - V I have the f o l l o w i n g features i n c o m m o n : 1. H a l f the runs o f each factor are at the low level, and h a l f are at the h i g h level. 2. E a c h factor is " b a l a n c e d " w i t h respect to the other factor i n that b o t h the l o w - a n d high-level runs of a n y factor are evenly split between l o w a n d h i g h levels of all other factors. R e c a l l i n g that a m a i n effect is estimated from the difference between the average h i g h - a n d low-factorlevel responses: M a i n effect o f X — i  {  X  4  -  ^degrees o f freedom—two degrees from the centerpoint ''estimate, a n d one degree f r o m the previous runs:  ^(responses at h i g h X )  x  3  + +  +  8 6  8  2+ 2-  X  2  D u m m y  c o n c e n t r a t i o n  T i m e  R a t i o  — ^(responses at low  X) i  ( h a l f the n u m b e r of factorial runs) Thus, all the d a t a i n a n experiment are used to estimate each m a i n effect, a n d each m a i n effect is estimated independently o f the other m a i n effects. T h i s feature of 2" designs is k n o w n as h i d d e n r e p l i c a t i o n , giving m a x i mum i n f o r m a t i o n per e x p e r i m e n t a l run. In poorlyplanned designs, only a subset of the runs is used for each estimate. A c a l c u l a t i o n procedure for the m a i n effects is now set up a n d illustrated, using the h y p o t h e t i c a l chemical-(? experiment. It is assumed that the yields from the factorial runs resulted in the data shown -in T a b l e I X . Note that the r u n order is the same as in T a b l e II. The c a l c u l a t i o n procedure is as follows: 1. C o p y the " — " a n d " - ) - " factor level array from the design table (Tables I I - V I ) . 2. L i s ! the responses i n a separate c o l u m n , corresponding to the run that generated it. If replicates have been r u n , list the average response in this c o l u m n . 3. Set up a d d i t i o n a l rows labeled 2 - K 2 — , Difference, a n d Effect, at the b o t t o m o f the array. 4. S u m the values i n the Response c o l u m n and enter the total i n the 2 + r o w o f this c o l u m n . D i v i d e by the number of runs i n the design table a n d enter this value  + — +  4  - 2 0 1  -  - 8 5  - 2  (which is the average response) i n the Effect row o f the Response c o l u m n . 5. F o r each factor c o l u m n : (a) S u m the responses c o r r e s p o n d i n g to the " + " entries a n d enter i n the 2 + row. (b) S u m the responses c o r r e s p o n d i n g to the " — " entries a n d enter i n the 2 — row. (c) Subtract the 2 — entry from the 2 + entry a n d enter i n the Difference row. (d) D i v i d e the difference entry b y o n e - h a l f the n u m b e r o f runs i n the design table, a n d enter this result in the Effect row. T h i s value is the m a i n effect for that factor. (e) M e n t a l l y a d d the 2 + a n d 2 — entries. T h e sum s h o u l d equal the response total under the Response column. For example, the temperature m a i n effect is the response average o f runs 2, 4, 6, 8 (high temperature) minus the response average o f runs 1, 3, 5, 7 (low temperature). T h e c a l c u l a t i o n is: Temperature effect = (83 + 77 + 84 + 86) - (58 + 62 + 75 + 75) 4  „ = 15% yield i c  Therefore, the effect of increasing the temperature from 7 0 ° C to 8 0 ° C averaged over all levels of reactant ratio, catalyst level a n d time, is to raise the yield 15%. A l s o , the time effect is 10% y i e l d , a n d the effects of the other two factors are m u c h smaller, at 0 and 3% for reactant ratio a n d catalyst level, respectively. T h e effects of the d u m m y factors c a n be similarly calculated. T h e factors are a measure of any interactions i n the system. Since the d u m m y - f a c t o r estimates are 1, —5, and — 2 % y i e l d , no o v e r w h e l m i n g interaction effects are seen, but the single —5 value alerts to the possible presence of interactions.  A s s e s s m e n t o f s i g n i f i c a n c e o f m a i n effects T h e effects just estimated are point estimates. T h e y give no idea of the r e l i a b i l i t y or precision of these estimates. T h e precision is generally stated in the form 101  XI for 95% confidence a n d v — 3 is 3.182. T o determ' the confidence intervals from the N = 8-run d e s i g ^ ts/VN/A  = ( 3 . 1 8 2 ) ( 2 . 3 ) / \ / 2 = 5.2  Therefore, the 9 5 % confidence intervals are: O r i g i n a l  d a t a  A"j temperature X reactant ratio X catalyst level X time X dummy X dummy X dummy  Y i e l d , %  T e m p e r a t u r e  T i m e  R e p l i c a t e s  A v e r a g e  7 0  1  5 8 , 6 2  6 0  8 0  1 .  8 3 , 7 7  8 0  7 0  2  7 5 , 7 5  7 5  8 0  2  8 4 , 8 6  8 5  3 i  5 6  7  C a l c u l a t i o n s Y i e l d R u n  T e m p e r a t u r e  T i m e  I n t e r a c t i o n  V a r i a n c e  r e s p o n s e , %  1  6 0  2  8 0  3  7 5  4  8 5  15 ± 0 ± 10 ± 3± 1± -5 ± -2 ±  2  5.2, 5.2, 5.2, 5.2, 5.2, 5.2, 5.2,  or or or or or or or  9.8 to 20.2 -5.2 to 5.2 4.8 to 15.2 -2.2 to 8.2 -4.2 to 6.2 -10.2 to 0.2 -7.2 to 3.2  Since a l l o f these effects b u t temperature a n d time include zero i n the confidence i n t e r v a l , it c a n be stated that the r e m a i n i n g effects are not significant at the 95% confidence level. N o t e , however, that X is almost significant, w h i c h alerts to the possibility o f either a n X X or a n X X interaction i n the c h e m i c a l Q reaction. These two interactions correspond to X i n the legend a p p e a r i n g under the design i n T a b l e I I . 6  X  2  3  i  6  2D i f f e r e n c e E f f e c t P o o l e d  7  Estimation of interaction effects  5  v a r i a n c e  T h e c a l c u l a t i o n for i n t e r a c t i o n effects follows the. same procedure as that for m a i n effects. T h e interaction effect is c a l c u l a t e d as the average response difference between one h a l f o f the factorial runs a n d the other half. T h e design table must first be a u g m e n t e d to i n clude columns for interactions. A n i n t e r a c t i o n c o l u m n is formed from the c o l u m n s o f the two factors that comprise the interaction, b y m u l t i p l y i n g the entries i n the factor columns. I n the e x a m p l e o f a 2 design given below, the X X c o l u m n is the p r o d u c t o f the X a n d X columns. 2  X  Run  2  l  X,  2  Response  X\X  2  1 2 of a confidence i n t e r v a l , w h i c h is a n interval said to i n c l u d e the " t r u e " effect at a stated confidence level. T h e most c o m m o n confidence levels are 90, 95 and 99%. T h e confidence-interval w i d t h is a function o f the response error estimate, the n u m b e r o f d a t a i n the estimate, a n d the n u m b e r o f degrees of freedom i n the response error estimate for a m a i n effect. T h e confidence interval is:  3  4  +  +  +  T h e interaction effect is c a l c u l a t e d as: [fT, + y ) - ( K , + Y )]/2 4  3  ( s e e F t g . 6)  R e t u r n i n g to the c h e m i c a l Q example, suppose it had been decided to rule out the existence o f the catalyst and reactant ratio factors, a n d reanalyze the design as a 2 factorial, replicated twice. T a b l e X shows the o r i g i n a l data a n d the calculations o f the temperature a n d time m a i n effects, together w i t h the temperature-time interaction. It can now be seen that the X\X interaction corresponds to the d u m m y v a r i a b l e X i n the screening design. 2  ( M a i n effect estimate) ± ts/ VN/4 where s = response error estimate w i t h v degrees o f freedom; N = n u m b e r o f factorial runs i n the design; a n d / = student's / statistic w i t h v degrees of freedom at stated confidence level. V a l u e s o f the m u l t i p l i e r / are given i n T a b l e X I , i n d e x e d by the confidence level and the n u m b e r o f degrees of freedom corresponding to the estimate s. If the confidence i n t e r v a l does not include zero, it can be said that the effect is significantly different from zero at the stated confidence level. T h e c h e m i c a l ( ^ e x a m p l e h a d a n estimate of s = 2.3. w i t h v = three degrees of freedom, two degrees of freedom from the three centerpoints, a n d one degree o f freedom from two previously run pilot experiments. If a 95% confidence level is desired, the l value from Table  2  6  Since the design is now considered as a 2 i n d u p l i cate, there are four sets o f " d u p l i c a t e " responses that can be used to estimate response error. F o r each temperature-time c o m b i n a t i o n , the average v a r i a n c e s can be calculated a n d shown i n the v a r i a n c e c o l u m n i n Table X . T h e average v a r i a n c e estimate is 7 (s — 2.6) with 4 degrees of freedom (one for each set o f d u p l i cates). T h i s estimate can be pooled w i t h the previous 3 degrees o f freedom estimate o f error (Table X ) to result in a n estimate o f J = 2.5% yield w i t h 7 degrees of freedom. 2  2  102  Two-sided student's f statistic D e g r e e s  Table XI  1 0 0 ( 1 — < D% c o n f i d e n c e l e v e l  o f f r e e d o m  90%  95%  99%  1  6 . 3 1 4  1 2 . 7 0 6  6 3 . 6 5 7  2  2 . 9 2 0  4 . 3 0 3  9 . 9 2 5  3  2 . 3 5 3  3 . 1 8 1  5 . 8 4 1  4  2 . 1 3 2  2 . 7 7 6  4 . 6 0 4  5  2 . 0 1 5  2 5 7 1  4 . 0 3 2  6  1 . 9 4 3  2 . 4 4 7  3 . 7 0 7  7  1 3 9 5  2 . 3 6 5  3 . 4 9 9  8  1 £ 6 0  2 . 3 0 6  3 . 3 5 5  9  1 5 3 3  2 . 2 6 2  3 . 2 5 0  1 0  1 . 8 1 2  2 . 2 2 8  3 . 1 6 9  1 1  1 . 7 9 6  2 . 2 0 1  3 . 1 0 6  1 2  1 . 7 8 2  2 . 1 7 9  3 . 0 5 5  1 3  1 . 7 7 1  2 . 1 6 0  3 . 0 1 2  1 4  1 . 7 6 1  2 . 1 4 5  2 5 7 7  1 5  1 . 7 5 3  2 . 1 3 1  2 . 9 4 7  1 6  1 . 7 4 6  2 . 1 2 0  2 . 9 2 1  1 7  1 . 7 4 0  2 . 1 1 0  2 . 8 9 8  1 8  1 . 7 3 4  2 . 1 0 1  2 . 8 7 8  1 9  1 . 7 2 9  2 . 0 9 3  2 . 8 6 1  2 0  1 . 7 2 5  2 . 0 8 3  2 . 8 4 5  2 1  1 . 7 2 1  2 . 0 8 0  2 . 8 3 1  2 2  1 . 7 1 7  2 . 0 7 4  2 5 1 9  2 3  1 . 7 1 4  2 . 0 6 9  2 . 8 0 7  2 4  1 . 7 1 1  2 . 0 6 4  2 . 7 9 7  2 5  1 . 7 0 8  2 . O 6 0  2 . 7 8 7  2 6  1 . 7 0 6  2 . 0 5 6  2 . 7 7 9  2 7  1 . 7 0 3  2 . 0 5 2  2 . 7 7 1  2 8  1 . 7 0 1  2 . 0 4 8  2 . 7 6 3  2 9  1 . 6 9 9  2 . 0 4 5  2 . 7 5 6  3 0  1 . 6 9 7  2 . 0 4 2  2 . 7 5 0  4 0  1 . 6 8 4  2 . 0 2 1  2 . 7 0 4  1 . 6 7 1  2 . 0 0 0  2 . 6 6 0  1 . 6 5 8  1 . 9 8 0  2 . 6 1 7  1 . 6 4 5  1 5 6 0  2 . 5 7 6  Chemical Q-yield vs. temperature andtime  Fig.11  :  dummy-factor effects can be used as a measure of response error. If £> represents the m a i n effect o f the i dummy factor, as calculated by the standard m e t h o d , then a n estimate of response error resulting f r o m d d u m m y factors is: l h  ;  6 0 1 2 0  oo  .  s =  \/N(2.D})/4d  where is the number o f factorial runs i n the design, a n d X is the s u m over i from 1 to d. T h i s value w i l l usually be an overestimate o f a because of the presence of interactions, a n d therefore should o n l y be used if it is the only error estimate available. It is preferable to use replicates to estimate response error. In the chemical (^example, the three d u m m y - f a c t o r m a i n effects were 1, —5 a n d —2, l e a d i n g to the estimate: s = V(8)(l + 25 + 4)/(4)(3) = \ / 2 0 = 4.5% yield T h i s estimate is almost twice the estimate obtained from pooled replicates. R e m e m b e r i n g that the second d u m m y factor was also the time-temperature interaction, it could be removed as a d u m m y factor a n d recalculated:  Confidence intervals c a n now be set up for each m a i n effect a n d interaction, a n d these are also shown i n Table X . T h e c o n c l u s i o n is that the effect o f temperature and time are statistically significant, but that their interaction is significant also. T h e y i e l d increase due to an increase in the time from 1 to 2 h is greater at 7 0 ° C than at 8 0 ° C , as can be seen g r a p h i c a l l y in F i g . 11.  Estimation of response error using d u m m y v a r i a b l e s In the absence o f interactions, the c o m p a r i s o n o f the high and low levels of d u m m y factors do not measure ny factor effects, a n d s h o u l d have an expected value o f ero. T h e y w i l l not always be equal to zero, i n any given experiment, because of response error. Therefore, a  2  j = \ / ( 8 ) ( l + 4)/(4)(2) = \/5 = 2.3% yield T h i s is much closer to the estimate that is based on replicates.  Estimation of curvature  effects  If all effects are first-order and interactive, the expected value of the response at the centroid of the design is estimated by the average o f the responses o f the factorial runs. T h e curvature effect is then estimated as the difference between the average o f the centerpoint responses and the average of the factorial points. In the chemical Q_ example, the centerpoints averaged 77% yield, and the factorial points averaged 75% y i e l d , giving a curvature effect of + 2 % yield. 103  A confidence i n t e r v a l for the curvature effect is calc u l a t e d as: . 1 1 C u r v a t u r e effect ± ts I— + — V Al C where C = n u m b e r o f centerpoints, a n d N = the n u m ber o f factorial points i n the design. I n the c h e m i c a l Q, e x a m p l e , N = 8 a n d C = 3:  , / l  +  i = (2.365)(2.5)yi  +  1 = 4.0  so that the 95% confidence interval o n c u r v a t u r e is: 2.0 ± 4.0, or —2.0 to 6.0, w h i c h is not statistically significant at 95% confidence.  Estimation for the full second-order model T h e terms i n the full second-order m o d e l are estim a t e d t h r o u g h a regression analysis p r o g r a m , w h i c h is a v a i l a b l e o n most c o m p u t e r p r o g r a m packages. In these c a l c u l a t i o n s the a c t u a l factor levels are used, rather t h a n the n o m i n a l values. T h i s is p a r t i c u l a r l y useful w h e n the a c t u a l factor levels differ from the n o m i n a l levels because o f c o n t r o l problems. A c o n t o u r - p l o t t i n g p r o g r a m can translate the m o d e l equations into m a p s of the response values over twod i m e n s i o n a l slices o f the factor space. S u c h maps can be useful for r e c o n c i l i n g two responses, b y o v e r l a y i n g o f the response surfaces. W e w i l l not be able to go i n t o the details o f these p r o g r a m s here, but w o u l d r e c o m m e n d the textbook b y D r a p e r a n d S m i t h [10] as a reference for further study.  design. F u r t h e r , i f it becomes apparent that a d e s i g ; being r u n i n the w r o n g experimental region, there is n o t h i n g to prevent its t e r m i n a t i o n . n  M a n y experimenters do not take into consideration their e x p e r i m e n t a l a n d measurement errors. T h e comm o n sin is to base a conclusion o n too few runs, only find that the result cannot be confirmed i n a later stage of the investigation. T h o s e w h o measure their response error have a good feel for the reliability o f their data T h e y k n o w w h e n a d d i t i o n a l e x p e r i m e n t a t i o n m a y be required to " n a i l d o w n " a conclusion. Use o f a m a t h e m a t i c a l m o d e l helps the experimenter to d r a w conclusions i n a m e a n i n g f u l w a y . T h e concept of factor interaction is very helpful i n u n r a v e l i n g fairly c o m p l e x relationships between the response a n d several factors. In c h e m i c a l investigations, interactions are the rule, not the exception. T h e b a l a n c e d nature o f a statistical design lends itself to an organized a p p r o a c h to e x p e r i m e n t a t i o n a n d a straightforward means o f d a t a analysis. G r a p h i c a l aids work very well w i t h results from designed experiments. E v e n w i t h these advantages, it s h o u l d be emphasized that statistical design is a tool, a means to an end. It w i l l not replace sound technical j u d g m e n t or c r e a t i v i t y in e x p e r i m e n t a l work. Nonetheless, it is a n i m p o r t a n t tool, w h i c h y o u , the experimenter, cannot afford to ignore. t 0  References 1. Bever, W . H . , " C R C H a n d b o o k o f T a b l e s for P r o b a b i l i l y a n d Statistics," Chemical 2.  A c o m p l i c a t i n g s i t u a t i o n i n experimental design arises w h e n some or a l l o f the factors are subject to constraints. A n e x a m p l e o f this is when the factors represent components o f a m i x t u r e , such as a n alloy or a b l e n d o f feedstocks. T h i s gives rise to the class of m i x ture designs described i n articles by C o r n e l l (.5), a n d Snee a n d M a r q u a r d t [J5]. These designs are useful for e s t i m a t i n g p r e d i c t i o n models for describing responses over the a l l o w a b l e c o m p o s i t i o n range. T h e r e are a n u m b e r of g r a p h i c a l techniques in use for data analysis, but a very useful one is the h a l f - n o r m a l plot, discussed by D a n i e l [7], These plots can be used for factorial designs where there is no r e p l i c a t i o n a n d no p r i o r estimate of error.  R u b b e r C o . , A k r o n , O h i o (1966).  Box, G . E . P., and B e h n k e n , D . W . , "Some New T h r e e Level Designs for the S t u d y of Q u a n t i t a t i v e V a r i a b l e s , "  Technomctrics, V o l . 2, pp. 455-475 (1960). 2 -"> F r a c t i o n a l F a c t o r i a l Designs," 0l  3. Box, G E . P., a n d H u n t e r , J . S., " T h e Tonometries, V o l . 3, pp. 311-351 (1961). 4.  Further design topics  s  Box, G . E . P., and L u c a s , H . L - , "Design of E x p e r i m e n t s in N o n - L i n e a r Situations," Biomilrika  V o l . 46, pp. 7 7 - 9 0 (1959).  5. C o r n e l l , J . A . , "Experiments w i t h M i x t u r e s — A R e v i e w , " Technomctncs. V o l . 15, p p . 4 3 7 - 4 5 6 (1973). 6. C o c h r a n . W . G . , and C o x , G . M . , " E x p e r i m e n t a l Designs," 2 n d ed., W i l e y . New Y o r k (1957). 7. D a n i e l . C , "Use of H a l f - N o r m a l Plots in Interpreting F a c t o r i a l T w o - L c v e ! Experiments." 8  Technomelna. V o l . 1, pp. 3 1 1 - 3 4 ! (1959).  Davies. O . L . , "Design and A n a l y s i s o f Industrial  Experiments."  2nd ed.,  H a f n e r . New York 0 9 5 6 ) . 9. D i x o n , W . ]., and  Massev. F. J . . "Introduction to Statistical  Analvsis.''  M c G r a w - H i l l . New York (1969). 10.  D r a p e r , N . R.. and S m i t h , H . . " A p p l i e d Regression A n a l y s i s . " W i l e y . New York (1966). '  11.  Finnev.  D . J . , 'Probit ;  Analvsis." 3rd ed., C a m b r i d g e  Press. New  York  (1971). 12. H i c k s . C . R.. " F u n d a m e n t a l ed.. H o l t . Rhinehart 13 1 4.  Concepts in the Design of Experiments.''  & W i n s t o n , New  2nd  Y o r k (1974).  N a l r r l l a . M G . , " E x p e r i m e n t a l Statistics." N a t i o n a l Bureau of Standards H a n d b o o k 91, U . S. G o v t . P r i n t i n g Office, W a s h i n g t o n , D C . (1963). Placket l . R. L . , and B u r m a n , J . P.. " T h e Design of O p t i m u m M ti hi facto rial Experiments,"  Biomclnka.  V o l . 33, pp. 305-325 (1946).  15. Snee. R . D . . and M a r q u a r d t , D . W . . "Screening Concepts and Designs for E x p e r i m e n t s with. M i x t u r e s . " Tcchnometncs, V o l . 18, pp. 19-30  Conclusion A l t h o u g h the principles of experimental design presented in this article can be put to i m m e d i a t e use, only the basics have been touched u p o n here. If y o u r location does not have access to a statistical consultant, y o u might consider attending a short course in experimental design, or reading such textbooks as C o c h r a n and C o x [5], Davies \8], or H i c k s [12]. O n e criticism o f statistical design is that the experimenter is c o m m i t t e d to a seemingly large n u m b e r of runs at the outset of an investigation. H o w e v e r , it is not u n c o m m o n for experimenters using a piecemeal approach to u l t i m a t e l y generate a larger n u m b e r of runs than w o u l d have been suggested by the experimental  The  il97tn.  author  T h o m a s D . M u r p h y , J r . , is G r o u p Leader, Statistical Analvsis, for A m e r i c a n C y a n a m i d C o . al B o u n d B r o o k , N J . He has taught courses in experiment design at Rutgers LIniycrsitv. and has had experience iri process d e v e l o p m e n t . production supervision, and pilot-plant engineering, rle holds a B.S in chemical engineering from the U n i v e r s i l v of M a r y l a n d and an M . S in statistics Irom Rutgers  104  APPENDIX D SUMMARY AND FACTORIAL ANALYSIS OF EXPERIMENTAL RESULTS  105  o o tr o_  < I-  z  LU oc  LLl D_ LL  o  CO  b < z  < <  DC O  < IL  D Z  < LO  b  z>  CO UJ DC  <  0.005  0.005  19.60  LO CD CD CD  LO  LO  LO  n  ro  |  20.45  (ml) (wt.%)  CD  20.5  Titrant  LD cn CD  CD  0.005  Thiosolfate  Charge  o o  Catalyst  o f  21.45  LL  0.34  z  -2  c  „  " S I  CO  z  o <  UJ Q U J K l > -  >- S  1  o Q_  o -  UJ f£ rZ LU O  o  X  Run Number  106  LO  r~>  CL.  Substitution  < o  (°C)  X  f£ rZ  r--  2  o z  (% eq. Cl )  or  Temperature  1  Z  LO  LO  LO  r-  co  T  LO  r--  O CXI  or  I f  o o  5  1  i?l  O  0)  or  107  o o  O  <8  I Q-l  Q) X  en <3 O  CC  108  UJ  < 5 d CM CM r— o in CM O  CM CM d d  n  +  5  °-  CO  CO  109 CO  effect  42.12  336.98  101.31 419.56 52.45  520.87  CN  00  sigma -  d  difference  CM  sigma +  ro  Catalyst  <o  effect  00  difference  -3.12 o  p  142.60  -24.95 ft-  CM  sigma -  323.56 to  CM  difference effect  298.61  ^—  CM  sigma +  d  sigma -  CO  Tenperature  p  d  sigma +  CM  479.58  -10.61  d CM O  effect  CM  -84.85  in  difference  ro  353.51  CM CM  sigma -  0.09  CM CO  pH/cat  0.73  c\i  effect  310.72  O)  268.66  sigma -  00  difference  CO  311.45  O Ui U-  o  -21.40  CM  effect  O  -171.21  O  difference  tr 396.69  sigma +  ai  225.48  o 1 £•  sigma -  LU u.  2  sigma +  \-  Substitution  eduction  z o I o <  Parameter  CTS  to to to to CM  o  CM  I  p d  o o CM r-CM CM  O  O  o  CE z  LU  CL LU  GJ u_ O  CO  to  b<  z.  <  < o c5 L? Q Z  <  CO  b CO LU  or  r -  O  Q. LU Lt r Z  LU  o o z  CO 3 Z  CC  g  O  r—  (-  O  3 O _l  6  < >  LU  or  cc  LU  hz UJ o z  >  oj cn  LU Q LL  z o  CC  O  110  cu •a "O  ID  3 h< > CC 3 O  CD  - CD  a  a  c a t_  C  a  LL  n i ii " ?i i I  cn  a  t  a  E i-  a  E a t-i  LO LO  "c  c:  c  c  •3  X  CO  CD  CD  CJ  1  3=  c <-  c r  r 'II i i 1  Tj a  c-  t  7  > HI I r T \i I 1 1"'I I I  a c  S  a  a  c  -5  (/I CD  CO  c  a  e  E  is  "  a  E  111  c E  .J T  CD CO CO CD  CD  o  CD  <D z  >s  3  S  aj r • - E t- c c  s  If  "5 C  >  "cc  r  2  0.691 0.691 0.681  1.99  sr  CO  -1.28  C\J  0.323 0.652  LO LO CJ  0.588  |  CD i  -0.179  -0.269 0.252  -1.114  0.11125  0.179  CD  0.5875 0.013  -0.0125  4.913  -0.82  7.849  CO CO CO CD CD  0.875  cn  0.2625  0.165  1  0.4125  CD  -0.2625  1  -0.263  CD  -3.206  i  ro m  CD  CO CD CD  0.013  LO  CD  2.842  CD  LO  2.423  LO  LT) LT) r~>  -0.0098  ^ LO  -0.399  CD  0.818  0.113  -0.113  |  CD  0.1125  0.163  -0.163  LO CD  -15.76  -0.239  LO  -0.00981  |  r-  -0.753  I  -1.699  n rvi  2.231 •:  1.972  CD i  -0.00085  CO  1.17  (cps)  Loss  Viscosity-  -1.869  -1.399  CJ CD i  -0.753  -0.213  0.213  I  |  (cps)  Viscosity  c n  0.621  -3.53  CJ CO T" LO  CD CD  0.038  CO  1.173  cn  .2.885  CM LO LO 1  0.513  cn  -1.013  0.549  (%)  Gain  Brightness  c n CD  CD  -0.003  CD  37.82  LO  -11.99  Tj-  0.188  CD CD  -0.0875  LO OO  CD  -0.0125  CD OO  0.0125  LO  26.02  (%)  Brightness  CD  36.98  Decrease  Kappa  CD  -0.0252  I  1.581  Kappa  CO  cn  LO CJ CD  Q  a  c  1  1 1  confidence interval  LO  -0.071  0.352  (%)  Reduction  Charge  Residual (rnmon  Chlorate concentration  CD  T fO CJ CJ  C\J  I  X CI02  co CO co  CD LO  ULO-  Chlorate  LT> LT) m  98000 0-  Variable  ID  Qi  |  Q  |  cn  Averaae  interval  confidence  95%  deviation  0.014  0.007  0.182  cn  Standard  0.183  1.581  0.735  80.121  80.000  606 08  ZD o o  0.175  g U. z o o Uo z o  79.455  UJ  0.188  o z  (%)  0.541  0.252  6.633  6.400  6.600  6.900  (ml Thio)  UJ  Number  < > rr  Reduction  0.541  0.252  -0.133  0.100  -0.100  -0.400  (ml Thio)  Reduction  o O z CO  Number  ID  0.828  0.385  47.623  47.240  47.620  48.010  (%)  Brightness  0.828  0.385  1.103  0.720  1.100  1.490  (%)  Gain  Brightness  UJ  Kappa  CL  Kappa,  o  3  0.200  0.093  20.647  20.750  20.570  20.620  (cps)  Viscosity  ID  [CI0 ]  3  Run  0.200  0.093  2.003  1.900  2.080  2.030  (cps)  Loss  Viscosity  (3 O  Q_ _J  rr  Z LU  fX  LU Q_  X  LU LL  O  CO CO  <  <  < _ J  o ho  LU  m  <  > IZ  o  Q _  Ul  LU  tr iz  <  o Q z < UJ o  UJ  > <  < rr  LU  > o o z  If)  D UJ  rr  D O  <  rr;  LU >  o uo z o  3  <  112 rr  rr  r-  >  <  IQ  o  o o  o o  LU  _l  m Q  3  < ( J  P in hin  (wt. %)  0.01  0.01  2  „ O in h-  20.10 20.50  47.62 47.24  6.60 6.40  0.175 0.183  10.80 o m CD 11.00  (cps) 20.62  (%) 48.01  |  Viscosity  6.90  |  Brightness  1 0.180  Number  Kappai  (ml Thio)  Thiosulfate  Q  tO'O  113 Q_ c  (% eq. Cl )  > (ml)  UJ N >-  Charge  ii  (hr)  Volume of  o  Time  CO  Number  X  o <  Catalyst  o_ £> . c Q. f£ ( OJ  Substitution  (D  Run  o  DC  0 _ _ J  UJ  L U 0 _  X  L U  U_  O  CO  CO  >-  <  o  2  co CO CO  in f-  m  m r--  cn  en  g o o  cc CL _J  z UJ  CC UJ Q_  X  LU U-  o CO CO  >]  <->  < z  <  <  o  12  01  at  00  o  DC O  CO  CO  IQ  o  CO  CM  d  d  i5  o  CD >  o  o  CN CSI od CO od *— CM CM  E r-  o  UJ UJ  r-  o  LU  a oo  LU Lt  o  a UJ NI  rr; < 3  to  3  UJ  rr;  a. E oj K-  O  114  o cn o  CO o  CN  CD O  o"  CD O  o Hi UUJ  < o o  £•  1 c^  E  ^  CD  3  E i-  UJ D_  co UJ  E  O  o UJ N  <  6  CO  I o  3 CO UJ  O  i  o  2  115  a o  DC CL  I CM. O  <z _J I — Z L U  DC  CM CO  CM CM  L U 0 -  X  LU UL  o  cn  CD  CO  o—  >] < z  CM  CM  d  d  Io  CD  o  < _ J  < cc o  < LL  3 H-  "S E rr •—  O LU  o p d  CD d  UJ  z g to < or  E f-  UJ  O o  N-  io  O CD o  UJ  a.  tyj LU  rr x  o o  5 Q UJ  Ni  rr < s to CO  to UJ  116  O  o  oc CL _J < IZ LU  5  Lt UJ CL  X  UJ  u_ O  CO CO >-  <  DC  o  < LL  CO  *«*  CO  r*  LO c n  cn I-  o  LU uUUJ  o CM CO CO  CM LO rco CO  o C O  cn  1  >_i < < •E  LO CM CO CO lO  < O O  a: Q o  o  117  CM hLO CO CM  CO LO CM O CO  I  I  |  4.16  389.14  -1.40  47.40  373.95  1  CO E o g I'-  118  sub/temD  effect  difference  |  |  1.43  0.23 1.32 0.17  385.21 1.32 0.16  -0.80 -2.10 -0.26  46.80 2.10 0.26  468.38  405.59  62.79  7.85  1.942  2.053  -0.110  -0.014  sinma -  difference  caftemD  effect  sinma +  sub/cat  0.01  -0.01  4.91  -0.010  0.11 0.59  effect  0.89 4.70 4.70 0.59  0.10  -0.10  39.31  -0.079  difference  383.52  -1.90  47.90  417.33  155.82  1.978  + ra E c "i?  siama -  156.71 3.24 -1.46  388.22  -1.80  47.80  456.64  I  11.19  155.55  1.55  386.53  -2.90  48.90  -3.21  -0.010  effect  |  8.28  156.98  -0.82  -0.41  0.41  0.09  difference  1.899  19.47  -0.43  -6.56  -3.32  3.30  0.70  -25.65  -0.079  C  1-4°  11.19  8.28  19.47  1.40  1.40  -3.45  1  siama -  + 0.18  8.28 11.19  157.99  4.17  387.53  -3.50  47.50  449.81  2.268  '  -2.39  384.21  -0.20  48.20  424.16  -  2.189  19.47  0.82  0.82  0.11  -0.11  -15.76  -0.039 154.54  1.40  -0.40  I  ll.19  -3.19  6.54  6.54  0.90  -0.90  -126.07  -0.310 |  8.28  157.86  -2.38  382.60  -2.30  48.30  19.47  500.02  154.67  2.618  I  2.308  cat/time  |  1.40  -0.16  5.47  -0.001  -0.54  -0.42  -0.42  effect  -3.32  -3.32  1.30 0.16  -1.30  43.76  -0.007  difference  8.28 11.19  -4.33  387.53  -2.50  48.50  415.11  158.43  2.55  2.002  19.47  154.10  -0.77  384.21  -1.20  47.20  458.87  |  -0.003  I  -0.21  -3.58  1.995  1.40  -0.24  -0.75  -0.75  -1.70  1.70 0.21  -28.65  -0.021 |  -1.91  -6.02 |  8.28 11.19  157.22  3.90 -6.02  388.88  -1.00'  47.00  451.31  19.47  2.076  155.31  -2.12  382.86  -2.70  48.70  I  422.66  2.055  siama -  sinma + I  sub/time  effect  difference  sinma -  sinma + I  temp/time  +  1  2  >  D.  ~£ Z , J=  or  "a c: 5  X a)  h-  ai  119 130.87 cn CO CO CO  B u r s t s t r e n g t h (KPa)  Zero Scan index (km)  Opacity  13.99 LO cn  13.79 2.06 93.52  1.99  Ln  B u l k (cnr'/q)  CO CO 93.27  13.01  135.63  CO cn  13.67  2.26 136.70  2.36 130.50  2.05 126.08  2.25 136.69-  CO CJ •^r  B u r s t Index (kPa.mVq)  2.05  921.25  15.35  3.33  32.60  0.015  127.43  987.18  |  1.93  CJ CD CO cn cn  • 2.27  cn CD CD CJ  2.20  15.33 919.99  14.96  3.08  0.005  30.16  |  29.80  0.015  •cr CO CD CD cn  897.83  15.23 co  15.15  2.97  3.23  31.72  3.30  29.11  I  CO  934.36  15.57  3.09  30.26  m  15.60  I  oJ  2.95 •  28.90  0.015  0.005  CD CO  LO  Tear strenolh (mN. 1 ply)  CD CD CD *r  B r e a k i n o l e n q t h (km)  I  OJ CD in  Parameter  CJ CD CD  Tesile Index (N.m/q)  CM  0.015  CD CD  0.005  o CD CD  0,005  CO  Catalvst Charqe (wt. %)  1  GO  Time (hr)  CO  Substitution (% eq. Cl )  CO OS  |  CO 05  Run N u m b e r CM OS  Factor Settinqs  CO  cn  cn •^r CD  cn  r-CO cn cn  cn  cn  cn  cn  cn  cn  •^r  cn  Cl  O  o o cr D .  _J  z L U  >: D C L U C L  X U J  O z H C O U J  I_J  s  to >X 0 _  C O  z  0J  Z>  •a  D C  1— z  O  no  CD M  o C L L U D C  h-  o  Z U J  O U J X  T3  ) a Q "O 5 M > « u >% c o co a o 5  LO co  "O ••=  ~ \  o  H O z  £  CO  < > _ l  a:  CO  C C U J  rL U  o Z L U Q L L Z  CC  1  o o cn cn U J X  ha.  LL  o  E Z E  Z  o \ZJ  O _]  <0 CL  -25 OJ  < o  121  !  c  Q. 100  LX  o o  cc a. _ i  ? z U J  LX UJ  a.  o  X  o  L U  CO  0 Z  H CO UJ  I— _ l  3  CO  >-  1  CL  o  co z r>  Lt  Q UJ  I  UJ  1 ID Z <  o  tj  t) CD  0)  LU CD  o JZ  o  o CO  O to  Q.  E  as  ID l>  (-  O  z  ZD Lt Q UJ  2 < O UJ X  hL L  O  z  o CO LX  < CL  O O _ i  < Lt  o  15  < LL  I(-  t-  t-  122  

Cite

Citation Scheme:

    

Usage Statistics

Country Views Downloads
Japan 22 0
United States 8 1
China 4 1
Mexico 3 0
Russia 1 0
City Views Downloads
Tokyo 22 0
Ashburn 6 0
Shenzhen 3 1
Unknown 3 0
San Antonio 1 0
Redmond 1 0
Monterrey 1 0
Beijing 1 0

{[{ mDataHeader[type] }]} {[{ month[type] }]} {[{ tData[type] }]}
Download Stats

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0058603/manifest

Comment

Related Items