Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Acoustic emission from the crystallization of potassium bromide doped with lead Cook, Adrian P. 1993-08-08

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

Item Metadata

Download

Media
831-ubc_1993_spring_cook_adrian.pdf [ 5.38MB ]
Metadata
JSON: 831-1.0061775.json
JSON-LD: 831-1.0061775-ld.json
RDF/XML (Pretty): 831-1.0061775-rdf.xml
RDF/JSON: 831-1.0061775-rdf.json
Turtle: 831-1.0061775-turtle.txt
N-Triples: 831-1.0061775-rdf-ntriples.txt
Original Record: 831-1.0061775-source.json
Full Text
831-1.0061775-fulltext.txt
Citation
831-1.0061775.ris

Full Text

ACOUSTIC EMISSION FROM THE CRYSTALLIZATIONOF POTASSIUM BROMIDE DOPED WITH LEADbyADRIAN P. COOKB.Sc., The University of Nottingham, U.K., 1989A THESIS SUBMITTED IN PARTIAL FULFILMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinTHE FACULTY OF GRADUATE STUDIESDEPARTMENT OF CHEMISTRYWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAJANUARY 1993© Adrian Cook, 1993In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)Department of  ChemistryThe University of British ColumbiaVancouver, Canada .Date 16 February 1993DE-6 (2/88)iiAbstractSimultaneous optical and acoustic information was used to elucidate themechanism that causes acoustic emission (AE) during growth of potassium bromide (KBr)crystals. The generation mechanism postulated here is one of inter-crystal interaction.AE was used to monitor the growth of different crystal structures of KBr grown withvarying degrees of lead (Pb) dopant (0-32000 ppm). Digital images were collectedsimultaneously at magnifications up to x75. Integrated AE profiles and image whiteningcurves were calculated. Best agreement between these was found for the crystal growthof the dendritic form of KBr (4080-8120 ppm Pb dopant), and the doped forms of12210 ppm Pb and above. The dendritic form of KBr observed between 4080 and8120 ppm Pb consisted of finger-like needles that grew outward from the crystallizingsolution. Scanning electron microscopy showed these fingers to consist of contiguouschains of tiny (< 25 gm dia.) octahedra. This microscopic structure provided manypotential sources of emission. The total AE observed per gram of crystals producedduring the growth of the different KBr morphologies was found to vary in a reproduciblemanner over 60 experiments. The dendritic form of KBr exhibited 10 times the acousticemission per gram of crystals produced compared with the undoped (cubic) form. Themass of crystals grown for individual experiments was 0.004-0.212 g. AE was able toreliably follow the growth of these very small masses of crystals whose morphologiesconsisted of clusters of crystals in intimate contact. A confirmatory mechanistic study wascarried out using ammonium chloride (NH4C1) which is known to produce classicdendritic growth. In contrast to the clusters of tiny microcrystals observed for thedendritic form of KBr, one would not expect the growth of the single crystal branches ofNH4C1 to exhibit acoustic emission, as there is no means to generate AE by theinter-crystal interaction mechanism. Single branch fracture is possible but rare.Experimental results showed very limited emission from NH 4C1, and thus supported the111hypothesis of inter-crystal interaction. Primary nucleation could not be detectedacoustically. The end of crystal growth was acoustically determinable. The AE associatedwith the growth of the dendritic form of KBr finished abruptly when the maximum imagewhiteness was reached. The acoustic waveforms detected during the growth of the KBrmorphologies were compared with artificially produced signals of bulk fracture and crystalimpact. Acoustic waveform analysis showed these classes of signals to be distinguishable.ivCONTENTSAbstract^ iiList of Tables viiList of Figures^ viiiI Introduction^1.1^Acoustic Methods in Chemistry^ 11.2^Acoustic Emission^ 11.2.1 Acoustic Emission Experimental Techniques^21.3^Acoustic Emission Data Analysis^ 41.3.1 Acoustic Waveform Analysis 41.3.2 Principal Components Analysis ^ 61.4^Crystallization Processes^ 71.4.1 Nucleation 71.4.2 Growth^ 81.4.3 Growth Forms 81.5^Effect of Impurities on Crystal Growth^ 101.6^Crystal Growth Monitoring^ 111.7^Acoustics and Crystallization 121.8^Crystal Growth of KBr^ 141.8.1 Effect of Lead Impurity 161.9^Purpose of the Present Work^ 18VII Experimental^2.1^Reagents^ 192.1.2 Solution Preparation^ 192.2^Apparatus^ 202.2.1 Image Acquisition Apparatus^ 202.2.2 Software Development for Image Acquisition ^222.2.3 Apparatus for Acoustic Emission Detection 222.2.4 Modes of Acoustic Acquisition^ 232.2.5 Acoustic Emission Integration 232.2.6 Individual Acoustic Signal Detection and Collection^242.2.7 Ex-situ Analysis Methods^ 272.3^Experimental Procedure^ 282.3.1 General Procedure for KBr Morphological Studies^282.3.2 Preliminary Experiments^ 292.3.3 Complete Morphological Study 292.3.3.1 Acoustic Integration and Optical Imaging^292.3.3.2 Capture of Acoustic WaveformsArising During Crystal Growth^322.3.4 Artificially Produced Acoustic Signalsfrom Crystallization Processes^322.3.5 Primary Nucleation Study ^ 322.3.6 Dendritic Growth Study 342.4^Data Analysis Methods^ 342.4.1 Integrated Acoustic Emission andImage Whitening Analysis^ 342.4.2 Acoustic Waveform Analysis 35viHI Results and Discussion3.1^Findings of Preliminary Morphological Study of KBr ^383.1.1 Trends in Integrated Acoustic Emission 383.2 Comprehensive Morphological Study of KBr ^403.2.1 Crystallization Sequences at a Magnification of x20^403.2.2 Studies at Magnification x75^ 483.2.3 EDX and XRD Analysis of Needles 483.2.4 Trends in Integrated Acoustic Emission ^533.3^Source of the Acoustic Emission^ 553.3.1 Comparison of the Rate of Acoustic Emission andRate of Crystallization^ 553.3.2 Integrated Acoustic Emission andImage Whitening Rise Times^ 583.3.3 Dimensionality of the Dendritic Growth Form ^613.3.4 Fracture Related Damage at High Dopant Levels^653.4^Characterization of Crystallization Processes byAcoustic Waveform Analysis^ 653.4.1 Generation of Different Signal Classes^653.4.2 Signal Intensities^ 663.4.3 Signal Class Separation by PCA^ 703.5^Classic Dendritic Growth^ 783.6^Primary Nucleation 80IV Conclusions^ 81V Further Work 83Literature Cited^ 85viiLIST OF TABLESTable I:^Descriptors used to mathematically characterize^5each acoustic signal.Table II:^Average values of weight and atomic percentages^49of K, Pb, Br determined by Energy DispersiveX-ray Analysis for needle crystals shown in fig. 15a.Table III:^Average values of weight and atomic percentages ^49of K, Pb, Br determined by Energy DispersiveX-ray Analysis in the octahedral matrix shown in fig. 15a.Table IV:^Powder X-ray diffraction peak positions for K2PbBr4 and KBr.^51Table V:^Descriptors that showed the highest resolution for^71the 6 different classes of crystallization signals.Table VI:^Descriptors that showed the highest resolution for^73the pairwise separation of artificial signals of bulkfracture and crystal impact, from the cubic (undoped)form of KBr, and the dendritic (doped) form of KBr.viiiLIST OF FIGURESFigure 1.^Growth of crystal faces. Adapted from reference [60].^9Figure 2.^Unit cell of potassium bromide.^ 14Figure 3.^Classic dendritic growth. 15Figure 4.^Schematic representation of the change in habit of KBr^17from cubic to octahedral on the addition of lead impurity.Adapted from reference [61].Figure 5.^Apparatus schematic.^ 25Figure 6.^(a) Transducer's frequency response curve,^26(b) Background frequency spectrum of blank.Figure 7.^Arrangement used for crystal growth experiments. ^30Figure 8.^Typical image of cubic KBr crystals on the surface of ^31the transducer. Image viewed through a continuouslyfocusable microscope with a magnification of x20.Figure 9.^Experimental arrangement for KBr whisker formation. ^33Figure 10.^Data analysis strategy for acoustic waveforms collected ^37during crystallization processes.Figure 11.^Dendritic form of KBr crystals observed with 4923 ppm^38Pb dopant. Growth of fingers on the transducer surface.Figure 12.^Variation of total acoustic emission per gram of KBr crystals^39grown during preliminary study. Crystal morphologyinfluenced by Pb doping.Figure 13.^Scanning electron micrographs of (a) cubic KBr at 0 ppm ^43Pb dopant, and (b) cubic to octahedral change of KBr at1990 ppm Pb dopant.Figure 14.^Scanning electron micrographs of (a) central islands of KBr ^44observed in droplet at 4080 ppm Pb dopant, showing theclustering of the octahedra, and (b) dendritic fingers of KBrwith 6070 ppm Pb dopant.ixFigure 15.^Scanning electron micrographs of (a) K2PbBr4 needles in ^45an octahedral matrix of KBr, observed at 24840 ppm Pbdopant, and (b) enlargement of the central portion of fig. 15a,showing the interpenetration of a needle with the octahedralmatrix.Integrated acoustic emission and image whitening curveswith time, for the growth of KBr crystals viewed at a x20magnification with varying amounts of lead dopant,(a) 0 ppm Pb, (b) 1990 ppm Pb, (c) 4080 ppm Pb,and (d) 22060 ppm Pb.Integrated acoustic emission and image whitening curveswith time, for the growth of the dendritic form of KBr crystalswith 8120 ppm Pb. Two different magnifications shown,(a) x20, and (b) x75.Powder X-ray diffraction patterns for (a) needle crystalscollected during the growth of KBr crystals from a solutiondoped with 29790 ppm Pb, and (b) analytical grade KBr.Total acoustic emission per gram of KBr crystals grownwith varying amounts of lead dopant. 3 experimental seriesare shown.Figure 16.Figure 17.Figure 18.Figure 19.46475254Figure 20.^Integrated acoustic emission profiles observed during the ^59growth of KBr crystals with varying degrees of lead dopant.Figure 21.^Comparison of acoustic integration and image whitening rise ^60times as the KBr crystal morphology is changed by increasingthe lead dopant.Figure 22.^Total acoustic emission per gram of crystals produced during ^62the growth of 5 replicate samples of KBr doped with8120 ppm Pb.Figure 23.^Effect of vapour pressure on the total acoustic emission ^64per gram of crystals produced during the growth of thedendritic form of KBr with 8120 ppm Pb. Numbers 1-5refer to replicates of fig. 22.Figure 24.^Typical acoustic signals and associated power spectra for^68KBr crystals grown with (a) 0 ppm, (b) 7620 ppm and(c) 24920 ppm Pb dopant, (d) bulk fracture of fingers ofKBr grown with 7620 ppm Pb dopant, and (e) impact ofKBr crystals on the transducer surface.Figure 25.^Average power spectra for acoustic signals collected during ^69the growth of KBr crystals with (a) 0 ppm, (b) 7620 ppm,(c) 24920 ppm Pb dopant, and artificially produced signals of(d) bulk KBr dendrite fracture, (e) KBr crystal impacts on thetransducer.Figure 26.^Principal components analysis showing the separation of^72different crystallization processes. Only 18 of 32 descriptorswere used (see table V).Figure 27.^Principal components analysis showing the separation of^75acoustic signals that occurred during crystal growth fromartificially generated signals mimicking secondary nucleationprocesses.Figure 28.^Principal components analysis of acoustic signals collected^76from both artificially produced crystallization processes, andactual growth processes of different KBr crystal morphologies.Figure 29.^Principal components analysis of acoustic signals collected^77with a 2.5 MHz bandwidth during the growth of different KBrcrystal morphologies. Only 18 of 32 descriptors were used(see table V).Figure 30.^Integrated acoustic emission profile for the dendritic growth^79of ammonium chloride, showing the infrequent signals, leadingto a stepped plot.Figure 31.^Scanning electron micrographs of (a) dendritic NaCI,^84(b) granular NaCl. Reproduced from reference [68].xiACKNOWLEDGEMENTSMany thanks are due to all my lab mates both past and present who sufferedthrough this thesis, but particularly to Larry, Oliver, Helen and Kevin. Special thanks goto Heather and Jim for keeping me sane, as well as to my parents for much needed advice.For tuition and the explanation of many queries concerning scanning electronmicroscopy thank you to Mary Mager, and similarly to Norman Osborne for XRD tuition.Thanks must also be extended to those at Infrascan Inc. (Richmond) for the help with theproduction of figure 8.Finally, thanks to my supervisor Adrian Wade, and to the following agencies formy support; the Canadian National Networks of Centres of Excellence, the Institute forChemical Science and Technology and the National Science and Engineering ResearchCouncil of Canada.xii"Where the telescope ends, the microscope begins.Which of the two has the grander view?"Victor Hugo : Les Miserables III. iii.1I INTRODUCTION1.1 Acoustic Methods in ChemistryThere are three broad categories of acoustic methods within chemistry. Firstly, anexternal sound source (ultrasonic) can be used to enhance chemical reactions; this isreferred to as sonochemistry [1]. Secondly, external acoustic sources may be used asanalytical probes, such as in acoustic microscopy [2], ultrasonic imaging and surfaceacoustic wave (SAW) devices [3]. Thirdly, one may detect acoustic waves producedwithin the material which accompany certain chemical and physical changes of thematerial. Photoacoustic spectroscopy whereby acoustic waves are generated within asample due to the absorption of optical radiation [4, 5] and Acoustic Emission (AE) fallinto this third category. The latter is the method employed in this thesis.1.2 Acoustic EmissionAcoustic emissions are transient elastic waves resulting from a rapid release ofenergy within a material [6]. The acoustic emission takes the form of many bursts of highfrequency (50kHz-5MHz) acoustic energy. These bursts are generated by the rapid massmotion or physical transformation of large collective groups of atoms. A number ofphysical processes have been cited as sources of acoustic emission, such as bubblebursting [7, 8], the motion of dislocations [9], propagation of cracks and crystal fracture[9, 10], and rapid volume changes that can occur during phase transitions andpolymorphic transformations [11, 12].21.2.1 Acoustic Emission Experimental ArrangementsAcoustic emission signals can be readily detected using a piezoelectric transducer,which converts the acoustic pressure wave to an electrical impulse. The duration andfrequency content of the acoustic wave produced are indicative of the underlying physicalmechanism that produced the acoustic burst. Indeed, signals from different physicalprocesses such as crystal fracture and bubble formation have been distinguished based onthe acoustic characteristics of their pressure waves [13].Unfortunately, the original acoustic wave and that detected by the transducer willnot be identical. The final signal detected will be modified by its transmission through thereacting medium, and across any container and transducer interfaces. Despite this,sufficent information is retained in the signal to study and sometimes identify the physicalmechanism that produced it. In order to ensure good acoustic transmission across thetransducer reaction vessel interface an acoustic couplant is used between the reactionvessel and the transducer surface. This couplant is designed to minimize attenuation at theinterface and is usually an exceptionally viscous liquid such as petroleum jelly [14]. Theseproblems of attenuation and propagation of the acoustic wave across all the interfaces in astandard experimental arrangement have detracted somewhat from the advantage of AE asa non-invasive, non-destructive analytical technique [15, 16]. One approach to thisproblem is to remove as many interfaces as possible by conducting experiments directly onthe transducer surface, so increasing the experimental reproducibilty.An experimental arrangement for AE monitoring would typically consist of apiezoelectric transducer, an amplifier and a recording device. The recording devicechosen depends on the information sought. The simplest recording device is a chartrecorder. The direct current (d.c.) output from commercially available acoustic emission3amplifiers provides peak level (signal intensity) information. The chart recorder traceshows the number of acoustic signals (as individual spikes) and the duration of the AE.Event counters, or acoustic power integrators can also be used to follow the kinetics ofthe process emitting AE by monitoring the d.c. output. A high speed digitizer can be usedfor individual acoustic signal capture if frequency signatures of the process are required.This is achieved by connecting the digitizer to the alternating current (a.c.) output fromthe amplifier. Acoustic signals from the process being monitored are distinguished frombackground noise by selecting a trigger level voltage which is greater than the peakamplitude of the background acoustic signals. Most processes monitored by AE in thislaboratory emit in the ultrasonic region (>20kHz), and so a high-pass filter may be used toeliminate any ambient noise in the audible region (3-16kHz).Many commercially available broad-band or resonant piezoelectric transducers aresuitable for the detection of AE. The commonest design consists of a small (approxlmm2) piezoelectric crystal of a matrial such as lead zirconate titanate, protected by a thinmetal case (commonly stainless steel). Such sensors have bandwidths within the range2kHz-4MHz. The frequency response of these transducers is unfortunately not flat overthis bandwidth, and substantial variability is often encountered even between the responsesof transducers of the same model. This transducer variability, together with the problemsassociated with the attenuation and propagation of the acoustic wave through theexperimental apparatus, means that for comparisons between experiments to be valid thesame experimental arrangement and transducer must be used throughout a series ofexperiments.41.3 Acoustic Emission Data AnalysisThere are two main methods of acoustic emission data analysis. The first simplyinvolves the integration of the d.c. signal with time, enabling the rate and amount ofacoustic emission produced to be followed. The second method involves the waveformanalysis of the acoustic signals, and is more complex.1.3.1 Acoustic Waveform AnalysisTypically 1000's of signals may be collected in a single experiment. Computeraided statistical pattern recognition techniques are used to look for differences andsimilarities in the acoustic signals. In order for the computer to compare the acousticsignals each signal may first be mathematically represented by a number of time domainand frequency domain parameters, e.g. the frequency which has the highest intensity in thepower spectrum. These descriptors are listed in table I and fully described elsewhere [17].Once each signal has been reduced to its mathematical descriptors a variety of differentpattern recognition methods may be employed. One such method used in this work isPrincipal Components Analysis.The frequency comparison of different acoustic signals can also be achieved byconverting the time domain signals collected to power spectra in the frequency domain byusing the Fourier Transform. The resulting power spectra of individual signals may thusbe compared. Average power spectra for many acoustic signals can also be calculatedenabling a comparison in the frequency domain of different acoustic processes.5Table L Descriptors used to mathematically characterise each acoustic signal.Time Domain DescriptorsRMS^Root mean square voltagePEAK^Maximum voltage amplitude (absolute)AREA^Integration of the absolute voltages within a signal^•^CREST^Ratio of peak voltage to root mean square voltageKURTOSIS^4th statistical moment (deviation from a Gaussiandistribution)T@AREA/2^Time to half area (signal decay measurement)50-CROSS^Number of times signal crosses ± 50% of maximium voltage25-CROSS^Number of times signal crosses ± 25% of maximum voltage10-CROSS^Number of times signal crosses ± 10% of maximum voltage0-CROSS^Number of times signal crosses 0 volts8 TIME OCTILES^Normalized time octiles of root mean square voltage1/8T....8/8TFrequency DomainDescriptorsFRQ MED^Frequency at the mid-area of the integrated frequencyintensities.FRQ MEAN^Frequency equal to the summation of the intensity-weightedfrequencies divided by the total intensityF-CREST^Ratio of the maximum power to the root mean squarepowerFBW>15%^Bandwidth of frequencies having intensities > 15% ofmaximum intensityFQRTLBW^Bandwidth of frequencies between the second and thirdintensity integrated quartiles8 FREQUENCY OCTILES Normalized area in each power spectrum octileDFB1....DFB861.3.2 Principal Components AnalysisPrincipal Components Analysis (PCA) is a method of finding the set of orthogonalaxes which represent the greatest variance in a multidimensional data set. Consider asingle acoustic waveform represented by n descriptors. This waveform can be consideredas one point in n-space. Now consider many different acoustic waveforms eachrepresented as a single point in n-space. PCA rotates the axes of n-space until a set ofaxes (principal components) is found that displays the optimal variance in the data set.Display of data using the two or three most important of these orthogonal components,allows data to be easily visualized [18]. The PCA method involves expressing data aslinear combinations of the independent contributions, in this case the 32 descriptors. PCAthen derives a set of orthogonal basis vectors called principal components, such that eachsuccessive principal component describes the maximum amount of variance possible in thedata not accounted for by previous principal component's [19]. If two or more acousticmechanisms with different acoustic signatures exist in the data set being analyzed by PCA,and the bulk of the variance can be accounted for by just a few principal components, thesignals of like mechanisms should cluster together on the plane of the data set beingdisplayed (e.g. Principal Component #1 vs Principal Component #2).71.4 Crystallization ProcessesCrystal growth from solution requires the attainment of supersaturation. Once thishas been achieved, crystallization can be divided into two main processes, nucleation andgrowth [20].1.4.1 NucleationFor a crystal to grow from solution a stable crystal nuclei must first be present inthe solution to act as a growth centre. This stable nuclei must be of such a size that it willnot redissolve. This is called the critical sized nucleus.There are two methods for producing critical sized nuclei. Firstly, the process ofspontaneous (primary) nucleation which involves the rapid "snap" change of a portion ofthe solution to form an ordered (crystalline) structure from the solute particles. Theproduction of critical sized nuclei has been studied by Garten and Head usingcrystalloluminescence [21]. The speed of the volume change within the solution toproduce a primary nucleus is less than 5 ns [22], and it has been postulated that thisvolume change may produce an acoustic wave, in a similar manner to the acousticemission generated during the martensitic transformation in steel [11]. The anomalousbehaviour of the velocity of ultrasonic waves near the freezing point of a substance due tothese density fluctuations has been used to calculate the size of the elementary nucleusnecessary for the beginning of crystallisation (-4 , 1µm dia.) [23]. However, this result is100 times larger than that obtained by Garten and Head using crystalloluminescence [21].Indeed the smaller value is more likely.8The second process involves the production of critical sized nuclei from crystalsalready present in the solution. Crystals suspended in the solution may impact with thevessel walls or with each other and small pieces of crystalline material can be broken off.These tiny crystalline fragments may be larger than the critical size and thus secondarynuclei are produced. Secondary nuclei may also be produced by the fracturing of anyoutgrowths of the crystals due to shear forces acting on the crystal produced by agitationof the solution. These processes are termed "secondary nucleation". The study of thecollision breeding of MgSO4 crystals by Mason and Strickland-Constable [24] is thedefinitive work in this area, although more recently significant contributions have beenmade by Chernov et al [25].1.4.2 GrowthAs soon as stable nuclei have formed in a supersaturated or supercooled systemthey begin to grow. The nuclei grow by adsorbing solute material from the surroundingsolution onto the surface of the stable nuclei. Any perturbation in the processes of solutematerial to the growing crystal, or the interference with the adsorption of material onto thefaces of the growing crystal can thus have a profound effect on the growth of a crystal.1.4.3 Growth FormsThe shape of a growing crystal is determined by the existence of or lack ofdominant growth directions. Different crystallographic faces under identical growthconditions may grow at different rates. It is the relative rates of growth of the crystalfaces that determine the final crystal habit [26]. Consider the section of three adjacentcrystal faces overleafb---a, NAxis to Face 1Axis to Face 2Axis to Face 39Fig 1. Growth of crystal faces. Adapted from reference [60].The normal growth rates of the crystal faces are denoted by v, w, and x such thatin unit time the faces are displaced to position a. The crystal shapes at each stage of thegrowth remain similar (position a) providing that v/w = K 1 and x/w = K2 (K1 and K2 areconstants). If the rate of growth of face 1 and face 3 increase such that v > K 1w andx > K2w (position b) then the size of face 2 increases during the growth of the crystal,thus the slowest growing face determines the crystal morphology [27]. If the rate ofgrowth of faces 1 and 3 decreases with respect to face 2, then the size of face 2 decreases(position c). One common method of anisotropically affecting growth rates of crystalfaces is the introduction of impurities into the system [28].101.5 Effect of Impurities on Crystal GrowthAny substance other than the material being crystallized can be considered animpurity. Some impurities can suppress growth entirely, some enhance growth, whileothers produce a highly selective effect, acting on certain crystallographic faces only, andthereby modifying the crystal habit (as above). The quantity of impurity necessary toinfluence crystal growth varies from system to system, with some impurities exerting aninfluence at concentrations as low as fractions of ppm (such as is the case for theferrocyanide ion in changing the habit of sodium chloride [29]), while others require muchlarger concentrations of up to 1000's of ppm.Impurities influence the growth of crystals by a variety of methods. Theintroduction of an impurity can change the properties of the solution (sometimesstructurally) [27], or they can alter the adsorption layer at the crystal solution interface[26]. They may be selectively adsorbed onto certain crystal faces exerting a blockingeffect, or be incorporated into the lattice [30]. Some impurities may interact chemicallywith the crystal and selectively alter the surface energies of different faces. Recentresearch has focussed on this last mechanism producing a number of 'tailor made' organicadditives that are designed to selectively adsorb onto certain surface sites of growingorganic crystals and thus retard growth selectively in the x, y or z direction. One suchsystem where this control is possible is for the influence of L-glutamic acid on the growthof L-asparagine [31]. Even the solvent of the crystallizing system is strictly an impurity[28], and indeed changing the solvent frequently results in a change of crystal habit andgrowth kinetics [32, 33].The modification of crystal habit is of great importance in industry, and not just ofacademic curiosity. Certain crystal habits are desired because of their anti-caking11characteristics, and good flow capabilities. General handling and packaging of materialscan also be aided by the correct choice of crystal habit [34, 35]. For certain applicationsthe rate of dissolution of the crystal is crucial (e.g. washing powders, pharmaceuticals);and this is an area where the habit of the crystal and specifically the surface area to volumeratio can again have a significant effect.1.6 Crystal Growth MonitoringMany methods have been used to follow crystal growth. The methodologiesused to follow crystal growth are diverse, as it only requires the measurement of a singlecomponent related to the growing crystal [36]. Direct optical measurements of particularface growth rates [37-39] are used to follow the growth of different single crystal habits,while crystal size distribution measurements have been used to follow batchcrystallizations [40]. Indirect methods such as raman microscopy [41], refractometry[42], light scattering [43] and potentiometry [36] have also been employed to monitor theuptake of solute material from solution. Synchrotron radiation is now being used to probethe crystal solution interface in-situ. This work has been pioneered by K. J. Roberts [44].With the advent of more powerful personal computers in the laboratory,automated image collection techniques are superseding the manual observations of facegrowth rates. Slow crystal growth techniques such as vapour diffusion lend themselves tothis type of monitoring [45]. Time lapse photography has been used by McPherson et alto follow the growth of lysozyme [46]; digital image capture methods have beendeveloped by Rosenblum et al [47] for following the growth of lysozyme aboard theNASA space shuttle.121.7 Acoustics and CrystallizationSound has been used as an external structural probe for single crystals [48-50] andsolution composition near the freezing point [23], but acoustic emission has largely beenoverlooked for monitoring crystallization. Betteridge et al first reported the emission ofacoustic signals from the recrystallization of KCI in 1981 [51], but little work has beendone since. A study of bulk precipitation of dichloro(pyrazine)zinc(II), a crystallineinorganic polymeric material, from solution was carried out by Munro [53]. Munroattributed the source of acoustic emission purely to the gross fracture of crystalssuspended in solution due to inter-particle collision, or contact with the vessel walls: inother words, the very processes of secondary nucleation. Bulk recrystallizations haveattracted some interest from acoustic emission researchers, as the processes of secondarynucleation should be amenable to acoustic monitoring [52]. The non-invasive nature ofAE would be a great advantage when following the kinetics of secondary nucleation inindustrial crystallizers, as at present they have to be regularly sampled, and crystal sizedistributions calculated.Several workers have reported acoustic emission from the crystallization ofpolymers from the melt [54-56]. They identify the source of the acoustic signals as theabrupt negative pressure release in regions of the melt occluded by spherulites duringcrystallization. This mechanism involves the release of stress from within a solid matrix, acommon source of acoustic emission. The crystallization of metallic glasses from the melthas also been studied, and a similar mechanism proposed [57].Lube et al have used acoustic emission to study the perfection of single crystalsduring melt growth [9]. They attributed the emission to a variety of solid state processesincluding, dislocation pile-ups, slipping, twinning, inclusion movement and cracking.13While these emissions gave an indication of the ideality of the crystal gown, the numberand regularity of the acoustic signals were not sufficent to provide an acoustic methodwhich could monitor the rate of growth of a single crystal.The first attempt to detect acoustic signals from morphological changes of crystalsdue to the addition of impurities was made by Delly [58]. He attempted to acousticallydetect the change of habit of sodium chloride on the addition of urea, but without success.The bandwidth used however, was limited by the equipment available to a very lowfrequency range (3-16kHz, audible frequency range). Differences in acoustic signals fromgrowth of various morphologies of crystals were briefly mentioned by Sawada et al [6]who noted that formation of plural crystals during the precipitation of sodium thiosulphategenerated the most intense acoustic signals.141.8 Crystal Growth of KBrThe ionic alkali metal halide KBr, under normal conditions, has a face-centredcubic unit cell according to the simple electrostatic model.0 Potassium Ion Bromide IonFig 2. Unit cell of potassium bromide.The potassium ions and bromide ions are both in face-centred cubic arrangements so thatthe lattice consists of two interpenetrating face-centred cubic lattices [59]. Theco-ordination number of both ions in the lattice is six. The habit of KBr crystals growingfrom a pure solution is thus hexahedral, with dominant bounding faces of {100}.15There are many methods whereby KBr can be persuaded to crystallize in formsother than cubic. If the evaporation rate is too fast the growth layers follow each other soquickly that the central parts of the crystal are not filled, and the result is termed "skeletalgrowth" [60]. This can be produced by changing the solvent from pure water to a mixtureof water and alcohol. A further increase in evaporation rate can result in the dendriticgrowth of KBr [60].Classic dendritic growth consists of a main stem from which primary andsecondary branches grow. Many compounds exhibit this type of growth (e.g. NH4C1[60]). It usually occurs in one plane, with very regular spacing of the branches. Thedendrites are single crystals with the branches following definite crystallographicdirections. This mode of growth is still not well understood.Primary BranchingsISecondary BranchingsI.^I1^1^1C^7Main StemFig 3. Classic dendritic growth.The growth of KBr whiskers, microfine (1-10 pm) single crystals which grow inone direction only, is also possible [60]. Polyvinyl alcohol has been reported as animpurity that enhances the formation of KBr whiskers from evaporating solutions,16whereas lead impurities such as lead nitrate inhibit the formation of KBr whiskers [60].Whiskers can also be produced by mechanical agitation of the crystallizing vessel as asaturated solution of KBr cools while sealed. An "avalanche" of tiny crystals is seenwithin the solution [60]. Mechanical agitation is a common method of initiating primarynucleation.Changing the evaporation rate to produce different KBr morphologies is not easilycontrollable. However, the doping of KBr crystallizing solutions with varying amounts ofimpurity can give fine control of the crystal habit.1.8.1 Effect of Lead ImpurityThe cubic to octahedral change of KBr on the addition of Pb impurity has longbeen known. A recent study by Agarwhal et al showed that as the lead content of acrystallizing solution of KBr is increased from 0-4000 ppm, { 111 } faces appear, togetherwith the {100}, until eventually they become the dominant bounding faces [61]. Thecrystal habit of KBr has thus been changed to octahedral.The mechanism by which the lead impurity affects the growth of KBr is one ofselective adsorption. This is explained using the model proposed by Sears [62]. The leadimpurity is adsorbed on the growth steps of the { 111 } face. The growth rate from a stepis then controlled by the equilibrium concentration of single unfilled sites in the line ofadsorbed atoms. As the concentration of the lead increases the number of vacant sitesalong each step decreases and thus the rate of step motion is itself decreased. As indicatedpreviously, it is the slowest growing face that dominates the crystal habit. The retardationof the {111} face thus produces an octahedral habit from the original cubic habit.17Fig 4. Schematic representation of the change in habit of KBrfrom cubic to octahedral on the addition of lead impurity.Adapted from reference [61].A further increase in the level of Pb impurity to rz 1% led to the observation of afinger-like growth of KBr [60]. This was termed the growth of KBr 'dendrites' by theauthors, although the growth observed was not classic dendritic growth. Indeed theindividual dendritic branches were seen to run forwards from the evaporating droplet,sometimes rising above the surface by 1-2mm. The branches were observed to beirregularly spaced and twisted in many different directions. The crystal morphologies ofKBr grown from solutions containing more than 1% lead dopant have not been reported.181.9 Purpose of the Present WorkThe goals of the present research have been three-fold. Firstly, to betterunderstand the origins of AE during crystal growth. To facilitate this the investigation ofdifferent modes of growth of crystals by using simultaneous optical and acousticmonitoring was carried out. By comparing the rate of acoustic emission to the rate ofcrystallization information about the acoustic mechanism was sought. Secondly, toascertain if there were any differences in the acoustic activity and signal characteristics ofacoustic waveforms emitted during the crystallization of these different crystalmorphologies. To date a successful study linking acoustic activity and crystal morphologyhas not been completed. Thirdly, to compare the signals emitted during crystal growthwith signals designed to mimic processes that may occur during secondary nucleation, andthereby assess the practicability of monitoring crystallization processes by acousticemission.Primarily, one system was used for this study: KBr grown from solution byevaporation. The KBr solutions were doped with varying degrees of lead impurity thatenabled many different crystal morphologies to be grown. A second system, NH4C1grown by evaporation, produced classic dendritic growth. These experiments were donefor comparison with the finger-like growth of KBr dendrites [60]. Primary nucleation wasstudied by producing copious numbers of KBr whiskers by mechanical agitation. Thissought to determine if the generation of the critical sized nucleus was acousticallyemissive.19II EXPERIMENTAL2.1 ReagentsAll the chemicals used were analytical grade unless otherwise stated. Potassiumbromide (KBr, 5 ppm lead impurity max) was obtained from BDH Chemicals (Toronto,ON). Lead nitrate (Pb(NO3)2, 99%+) and ammonium chloride (NH 4C1, 99%+) wereobtained from the Aldrich Chemical Company (Milwaukee, WI).2.1.2 Solution PreparationThe solutions were prepared saturated at 20°C, just below room temperature, toensure a short evaporation time at ambient temperature before the solution becamesupersaturated and crystals began to form. Three separate series of KBr solutions dopedwith varying degrees of lead impurity were prepared by weight (± 0.0005 g) using ananalytical balance (model H10, Mettler, Greifensee, Zurich).KBr Series #1:^6 solutions of KBr with 0 ppm, 600 ppm, 1010 ppm, 2080 ppm,2750 ppm, and 4920 ppm Pb dopant.KBr Series #2:^14 solutions of KBr with 0 ppm, 670 ppm, 1050 ppm, 1990 ppm,4080 ppm, 6070 ppm, 8020 ppm, 12210 ppm, 16000 ppm,19960 ppm, 22060 ppm, 24840 ppm, 27790 ppm, and 31960 ppmPb dopant.20KBr Series #3:^11 solutions of KBr with 0 ppm, 770 ppm, 2070 ppm, 3980 ppm,5900 ppm, 7620 ppm, 11930 ppm, 15980 ppm, 20090 ppm,24920 ppm, and 29790 ppm Pb dopant.Further Solutions for KBr whisker growth and the dendritic growth of NH4C1were prepared.Solution #1:^KBr, undoped, saturated at 70°CSolution #2:^NH4C1, undoped, saturated at 20°CAll solutions were sealed and stored in the dark for the duration of each individual study.2.2 ApparatusThe various experimental apparatus and instrumentation used are outlined in thefollowing sections (2.2.1-2.2.7). A schematic of the combined apparatus for imageacquisition and acoustic emission acquisition is shown in fig. 5.2.2.1 Image Acquisition ApparatusA continuously focusable microscope (CFM) (Infinity Photo-Optical Company,Boulder, CO) with a camera (VK-C 150, Hitachi) attached was used to collect the images.The CFM focuses continuously from infinity down to 6 mm. The primary magnification isdependent on the distance of the lens from the sample. The camera is based on a solidstate image sensor (8.8 x 6.6 mm 2, 11 mm diagonal), and outputs a standard video signal.The video signal output was fed into a frame grabber card (PCVisionplus, ImagingTechnology Inc., Woburn, MA) in a host microcomputer (PC/AT) and digitized. Thedigitized image, 640 x 480 pixels, 256 grey scale (8-bit), was then displayed on a21monochrome monitor (330mm diagonal, Sync Master II, Datatrain, Roland DG CanadaInc.). This image was updated according to a time delay specified by the operator (seesection 2.2.2) Samples were illuminated with an optic fibre light source (CambridgeInstruments Inc., Buffalo, NY) which imparted very little heat to the sample. Indirectpartial dark field illumination was used to prevent any intense white spots being created byreflections from the sample surface.The total magnification produced by the CFM was calculated as follows:Total Magnification = Primary Magnification x Aspect Ratio(solid state sensor to monitor)For the system used here the aspect ratio is x30 (330 mm / 11 mm), and the primarymagnification varies in a linear manner with the distance of the lens from the sample, withvalues of x0.2 at 152 mm, to x9 at 6 mm. This scale is not linear, with the greatest rate ofchange at the shorter distances, and so the magnifications quoted in this thesis werecalculated directly by measuring the dimensions of the image from an object of knownsize. When the method of calculation outlined above was used it varied by ± 10%(relative) from the value calculated by measuring the dimensions.The video signal output from the camera was also simultaneously fed into a colourmonitor (KV-2064R, Trinitron, Sony Corp.) via a standard VHS VCR unit(HR-D670U, NC). The colour monitor displayed the real-time image, so that theoperator could easily follow the crystallization, and the VCR enabled subsequent review(see fig. 5).222.2.2 Software Development for Image AcquisitionA routine was written in the 'C' programming language to control the framegrabber, allowing the user to acquire a specified number of images (maximum of 1000)with a specified time delay between each image (minimum of 3 s). This routine alsoperformed a near real-time pixel summation of the grey level values across the entireimage to quantify the whiteness of the image. This value was then stored along withsuitable image identifiers such as the image number and time of acquisition as an ASCIIfile. As each image required approximately 307kB of memory for storage only selectedimages (e.g every 50th) were permanently recorded on disk for later visual analysis anddisplay. The routine allowed for a baseline correction to ensure that the grey scale wasnot saturated by the whiteness of the sample (the maximum whiteness of a singlepixel = 255). The whitening of the image was used as an independent method of followingthe crystallization. This enabled the acoustic activity to be compared to a reference.2.2.3 Apparatus for Acoustic Emission DetectionThe acoustic signal detector used for all the experiments was a broad-bandpiezoelectric transducer (model 8312, serial# 1381660, Bruel and Kjaer, Naerum,Denmark). This transducer contained a piezoelectric crystal sensor (k11.5 mm 2) and aninternal preamplifier of 34 decibel (dB) gain. The manufacturer's frequency responsecurve for the transducer, and a background blank recorded in the laboratory are given infig. 6. For the background blank a trigger level of 0 V was set and signals were recordedevery 2 s for a period of 5 min. The transducer was mounted on a piece of foam matting(25 mm) to damp out unwanted laboratory vibrations.23The output from the transducer was directed to a conditioning amplifier (model2638, serial# 1283189, Bruel and Kjaer) with a variable gain of 0-60 dB, and a bandpassfilter of 50kHz-2MHz. The conditioning amplifier had two outputs, an a.c. output whichproduced the conditioned transducer signal (a representation of the acoustic waveform)and a d.c. output which gave the damped peak amplitude of the transducer signal. Allacoustic data that were collected were normalized to a total gain of 64 dB (34 dBpreamplification, 30 dB main amplification) so that quantitative comparisons could bemade.2.2.4 Modes of Acoustic AcquisitionTwo methods of acoustic monitoring and quantitation were used in this study, thusutilizing both the a.c. and d.c. outputs of the conditioning amplifier.2.2.5 Acoustic Emission IntegrationThis method involved the collection of the peak d.c. amplitude output via a 12-bitanalog to digital convertor (model RTI-815F, Analog Devices, Norwood, MA) in asecond host microcomputer (Compaq model II (PC/XT), Compaq Computer Corp.,Houston, TX). Software developed within this laboratory samples this d.c. output andintegrates it over time. This has been reported previously [63], so only a brief descriptionwill be given here.The software developed enabled the peak d.c. amplitude to be collected every40 ms and then averaged every 20 points. This produced one data point every 0.8 s,corresponding to the acoustic emission intensity observed in that 0.8 s time window. Thedata provided a method of following the rate of acoustic emission and gave a measure of24the total acoustic emission collected during the experiment. Before each experiment abaseline was measured over a period of 10 min. The baseline level was automaticallysubtracted from the levels obtained for the real signals. Real signals were then collectedby means of a trigger level that was set at three times the standard deviation of thebaseline readings.2.2.6 Individual Acoustic Signal Detection and CollectionAcoustic waveform data were collected from the a.c. signal output of theconditioning amplifier using a fast digitizer (model SDA2000, Soltec Inc., San Fernando,CA). Each signal was represented by 1024 points, 992 points for the signal plus 32pre-trigger points, with 12-bit resolution. Two sampling rates were used 2.5 MHz and5 MHz, and the trigger level was set to 1.5 times the background noise level. Thebackground level was determined by finding the trigger level voltage at which thetransducer barely detected acoustic signals in the open laboratory over a period of 10 min.The background was determined prior to every experiment. The digitized signal was thentransferred to a host microcomputer (PC/AT) via an IEEE-488 interface (model PC IIA,National Instruments, Austin, TX) and stored on the PC/AT hard drive for later analysis.The data acquisition program used was developed in this laboratory and has been reportedelsewhere [64]. Simultaneous acoustic waveform analysis and image acquisition wasimpossible using the experimental arrangement described, as the same PC/AT was usedfor both the data acquisition from the fast digitizer and image acquisition using the framegabber card (see fig. 5).Real TimeColourMonitor^,25.Frame Grabber640 x 480NMonochromeStillImageMonitorCameraVCR jLightpeak levelsignalData Logger(PC / AT)Acoustic Integrator(PC / XT)FastDigitisera. c. signalCond'tioningAmplifierMicroscopeLight[ Sample] ZPiezoelectricTransducer^,■^I(exploded in fig 7)Fig 5. Apparatus schematic.26908 5(f)> 807570656055100 200 300 400 500 600 700 800 900 1000Frequency (kHz)0.0050.0040.0030.00247;0.0010.0000^250^500^750^1000^1250Frequency (kHz)Fig 6. (a) Transducer's frequency response curve,(b) Background frequency spectrum of blank.272.2.7 Ex-situ Analysis MethodsEx-situ analysis of the crystal morphologies grown were investigated usingscanning electron microscopy (SEM) (0.01 gm resolution) (S2300, Hitachi). Anaccelerating voltage of 20 kV was used with the detector at a working distance of 15 mm.Energy dispersive X-ray analysis (EDX) (S570, Hitachi) was used to determine any spatialvariations (1 resolution) in chemical composition for the morphologies grown. Theworking distance of the detector was 35 mm with an identical accelerating voltage as theSEM settings.Samples for SEM and EDX analysis were prepared as follows. The KBr crystalsdoped with Pb were carefully removed from the transducer surface and mounted onaluminium holders using adhesive applied from a removeable label. The samples weregold coated to provide an electrically conductive surface to facilitate analysis by SEM andEDX. The gold coat was deposited under a reduced pressure (100-200 mtorr) argonatmosphere, with a direct current of 20 mA. The exposure time used was 4 minutes,resulting in a 100 A thick coating.Structural features of the morphologies grown were identified using a magnifyingglass and removed from the transducer with tweezers. These crystal samples were thenground, mounted on a microscope slide using vaseline, and analyzed by powder X-raydiffraction (XRD) (Rotating Anode machine, Cu target, Kai = 1.5418 A; Rigaku USAInc., Danvers, MA). The X-ray tube current and tube voltage used were 150mA and60 kV respectively. The XRD served to confirm variations in chemical composition of thedifferent structural features indicated by EDX analysis.282.3 Experimental ProcedureThe following sections give the experimental arrangements and procedures usedfor the different crystallization studies.2.3.1 General Procedure for KBr Morphological StudiesA droplet of the KBr solutions doped with lead impurity was placed on the upperstainless steel surface of the transducer and allowed to evaporate at ambient temperature(23 ± 2°C) (see fig. 7). Solutions were selected in no particular order until one trial hadbeen completed for each solution in the given series. This process was repeated until atleast 3 replicate experiments of each solution had been completed. This procedure wasadopted to minimize variations across the series due to changes in ambient conditions.The growth of the crystals was monitored acoustically using the integrationmethod (section 2.2.5) for a period of 4 hrs. The total acoustic emission was thenrecorded, normalized to the mass of crystals produced, and an AE vs. time trace obtained.The results for replicates were averaged. Prior to each crystallization experiment thetransducer was thoroughly cleaned with de-ionized water to remove all residual crystallinematerial which might form unwanted sites for nucleation in future experiments. Acetonewas used to remove any grease build-up and ensure comparable wetting of the transducersurface for each experiment.292.3.2 Preliminary ExperimentsInitially the KBr / Pb(NO3)2 system was investigated over the dopant range of0-4920 ppm Pb using the solutions of series #1. The procedure was as outlined above.No microscopic imaging or ex-situ analysis was carried out. A large range of droplet sizeswas tried, 50-2502.3.3 Complete Morphological StudyA comprehensive study of the KBr / Pb(NO3)2 crystallization was carried out overthe dopant range of 0-31960 ppm Pb using the solutions of series #2 and series #3.2.3.3.1 Acoustic Integration and Optical ImagingSimultaneous microscopic observation with image collection and acousticmonitoring (see general procedure) was carried out (see figs. 5, 7). A typical image seenthrough the microscope is shown in fig. 8. The crystallizing solutions of series #2 wereused. Studies were completed with magnifications of x20 and x75. Droplet volumes ofbetween 4-100 p.1, were used depending on the degree of magnification and the amount ofspreading expected during the growth of each different crystal morphology. These weredispensed via a precision syringe (± 1p1) (Hamilton Co. Inc., Whittier, CA). Exactvolumes are discussed later. Comparisons were then made of the acoustic emissionintegration trace with time vs. the image whitening trace with time to gain an insight intothe acoustic emission generation mechanism.The ex-situ analysis of each morpholgy grown was carried out using the techniquesoutlined in section 2.2.7.(to frame grabber)Piezoelectric Crystal SensorFibre OpticLight SourceBroadbandPiezoelectricTransducer25mm1.5-11cm^ 41mm^---Foam MatImageHitachi VK-C150CameraContinuouslyFocusableMicroscope(CFM)30Acoustic EmissionSignal(to conditioning amplifier)Fig 7. Arrangement used for crystal growth experiments.SiFig 8. Typical image of cubic 10r crystals on the surface of the transducer.Image viewed through a continuously focusable microscope with amagnification of x20.322.3.3.2 Capture of Acoustic Waveforms Arising During Crystal GrowthThe solutions of series #3, similar in composition to series #2, were used as asource of acoustic waveform data during growth. These experiments were done withoutsimultaneous image acquisition. Experiments were of 4 hrs duration, and the apparatusused was as outlined in section 2.2.6. Two frequency bandwidths were used for signalcollection and analysis, 1.25 MHz and 2.5 MHz. The signals collected during growthwere then compared by using PCA to determine if there were any differences between thegrowth signals of different KBr morphologies.2.3.4 Artificially Produced Acoustic Signals from Crystallization ProcessesArtificial signals of bulk fracture and impact were produced to mimic the processesof secondary nucleation. Bulk fracture signals were produced as follows; the fingers ofthe dendritic form of KBr obtained with 7620 ppm Pb dopant were fractured with the aidof a syringe needle. Impact signals were produced by dropping a variety of different sizedKBr particles (0.1-0.5 mm) from heights of between 2-3 cm onto the surface of thetransducer. These signals were compared with the growth signals using PCA. Afrequency bandwidth of 1.25 MHz was used.2.3.5 Primary Nucleation StudyA primary nucleation study was carried out by producing a large number of tinyKBr whiskers using mechanical agitation. The method used was as follows. A 50 mlgraduated cylinder was filled to 2/3 of its height with solution #1. A thermometer wasplaced in the graduated cylinder and the tube sealed using parafilm to prevent the entry ofdust particles and limit evaporation. The solution was then heated to 80°C to fully33dissolve any KBr, and then mounted on the transducer surface using apiezon grease(Apiezon Products Ltd., London, U.K.) as an acoustic couplant (fig 9). When thetemperature had reached 68°C the tube was tapped with a wooden block to induce massnucleation. This primary nucleation was monitored acoustically as outlined in section2.2.6.Parafilm Seal ThermometerGraduated CylinderSolution #1Acoustic Emission^_s--.42Signal(to conditioningamplifier)Foam mattingTransducerApiezon GreaseFig 9. Experimental arrangement for KBr whisker formation.342.3.6 Dendritic Growth StudyClassic dendritic growth was produced using a solution of NH4C1, solution #2(section 2.1.2). The general procedure outlined in section 2.3.1 was used to monitor thisdendritic growth.2.4 Data Analysis MethodsThe different methods of data analysis used during this study are outlined in thefollowing sections.2.4.1 Integrated Acoustic Emission and Image Whitening AnalysisThe integrated acoustic emission profiles were first converted to an ASCII format.The traces of integrated acoustic emission with time and image whitening with time werethen compared directly on the same plot using a commercially available graphing package(Sigmaplot, version 5.0, Jandel Corporation, Sausalito, CA). The time taken for thecurves to rise from 10% to 90% of the final value were calculated and used to comparethe rise times of the AE profile and the image whitening profile.It is pertinent to mention here the units that are used to represent the acousticemission data in this thesis. There is no standard reference measure (calibrant) foracoustic emission, and SI units of acoustic emissivity do not exist [64]. This has led toacoustic emission being measured using a variety of different methods, with the signalsobserved strongly dependent on the equipment used. The acoustic emission data in thisstudy has thus been given arbitrary units (AU), with all scales comparable.352.4.2 Acoustic Waveform AnalysisAll the software for signal processing and analysis was written in this laboratoryand has been reported elsewhere [65]. Before any generation of descriptor information orfrequency power spectra, all over-ranged and under-triggered signals were removed fromthe data set. Over-ranged signals have amplitudes that exceed the voltage range selectedfor collection, while under-triggered signals are a result of trigger level instability. Neitherof these signal types are desirable. Power spectra were calculated by the using a fastFourier Transform to convert the time domain signal to the frequency domain. Theaverage power spectrum was then calculated; this is an average of the individual frequencyintensities of each signal. Descriptor information was calculated for each signal collectedin order to characterise the signals. These descriptor values were then used for PCA.Prior to PCA, one of each pair of descriptors that was found to correlate with eachother at a level greater than 90% was removed. Obviously, descriptors that correlatehighly with each other will not add significant information to the variance within the dataset. Those that were found to have a poor resolution across the different classes of signalsunder study, i.e. descriptors that have a poor separation capability, were also removedprior to PCA. Only 18 of the 32 descriptors were thus used (see table V) to separate thedifferent signal classes. Assignments of signal classes to the different data sets will begiven later.The descriptors were then autoscaled and link-scaled [18] to correct for changesin variance in the time octiles and frequency octiles (see table I). A plot of any 2 of theprincipal components generated could then be used to view signal clustering within thedata set (see fig. 10).36For improved separation of the different classes of signals the three descriptorswith the highest resolution across the data set were determined. The data set was thenre-analyzed by PCA. The removal of all the independent contributions (descriptors)except three decreases the total variance across the signals of individual classes within thedata set, and proved adequate for separation in most cases.37( Generate and Detect Acoustic Signals )(Collect Data Set of 800-2500 Signals)Remove Bad Signals from Data SetGenerate Frequency Power Spectra(Display and Compare Frequency SpectraCalculate Descriptor Values )Remove Descriptors thatCorrelate at a Level > 90%^ I[Determine Descriptors withHighest Resolution( Scale Descriptors)( Generate of Principal ComponentsFig 10. Data analysis strategy for acoustic waveforms collectedduring crystallization processes.38III RESULTS AND DISCUSSION3.1 Findings of Preliminary Morphological Study of KBrThe KBr crystal forms grown during the preliminary study (solutions of series #1)were visually seen to change from a hexahedral habit with 0 ppm Pb through an octahedralhabit, until at the highest dopant level examined, 4920 ppm Pb, the growth of irregulartwisted fingers was observed. These fingers spread out rapidly from the droplet until theentire surface, and in some cases even the sides, of the transducer were covered. Some ofthe fingers rose off the surface of the droplet by as much as 0.5 cm, in agreement withpreviously reported work [60]. The crystals of the cubic and octahedral morphologies ofKBr grown stayed within the boundaries of the initial droplet.KBr finger-like growths...""t-.., \\'^<>..- ,1\ \ \ ,1 , \ \\^s^\ '-`\ k^\ ---•.::: '-^:.^\k- r^lo.^'''.ti..":*....^'.5::".7.•:!..^..ffl':'.-TransducerFig 11. Dendritic form of KBr crystals observed with 4923 ppm Pb dopant.Growth of fingers on the transducer surface.3.1.1 Trends in Integrated Acoustic EmissionThe total acoustic emission (normalized to the mass of crystals grown) asmeasured during the growth of each of these different forms of KBr was found to increasewith Pb dopant (see fig. 12). This trend was observed to varying degrees for each3940 COL._' 45 030cpO cn•Fr) . c- _- 20EL._U 0 100r+6 0 0^1000 2000 3000 4000 5000Degree of Lead Doping (ppm)Key:—0^03 Replicate SeriesAverage of the 3 seriesFig 12. Variation of total acoustic emission per gram ofKBr crystals grown during preliminary study.Crystal morphology influenced by Pb doping.40repeated trial across the solutions of series #1. The sample size in the majority of caseswas such that between 0.1-0.2 g of crystals were produced.It was postulated that the trend in AE per gram of crystals grown (fig. 12) was dueto the change in the KBr crystal morphology because of the influence of the lead dopant.A more comprehensive study of the effect of the crystal morphology grown on the AEobserved was thus undertaken. The findings are reported in the following sections.3.2 Comprehensive Morphological Study of KBrA comprehensive study of the different growth morphologies of KBr doped withPb and the associated acoustic emission produced was undertaken using the solutions ofseries #2 and series #3.Solutions of series #2 were used to compare the rate of acoustic emission and rateof image whitening for the different KBr morphologies grown. Solutions of series #3 thatproduced the same KBr morphologies as series #2 provided acoustic waveform data forcrystal growth.3.2.1 Crystallization Sequences at a Magnification of x20As the solutions of series #2 were allowed to evaporate the KBr crystallized indistinct morphologies, which depended on the level of lead dopant within the evaporatingsolution. Scanning electron micrographs for 0-29000 ppm Pb dopant are shown infigs. 13, 14, 15. The total mass of crystals grown was between 0.06-0.12 g for theexperiments carried out.41The following crystallization sequences were observed. With 0 ppm Pb doping,individual cubes of KBr were seen to nucleate and grow within the drop (fig. 13a), untileventually when almost all the solution had evaporated the individual cubes began totouch, and grow together. No acoustic emission accompanied the nucleation and growthof the solitary KBr cubes. However, acoustic signals were observed as soon as the cubesbegan to touch and grow into each other. This can be clearly seen in fig. 16a where theacoustic integration trace and the image whitening curve are offset in time, with theacoustic integration lagging behind the image whitening. In fact the majority of theacoustic emission was observed after the crystals had visually "stopped growing". For adiscussion of the source of this acoustic emission see section 3.3.As the dopant level of Pb increased from 0 to 1990 ppm the KBr crystal habitchanged from cubic to octahedral (fig. 13b). The crystals showed increased tendency toclump together during the initial nucleation and growth phases, so that groups of two orthree crystals were observed together. The acoustic integration trace showed a distinctivetwo stage rise (fig. 16b), with the first stage following the image whitening exactly. Thesecond rise, which occurred after the image whitening had maximized, indicated again thatmuch emission comes after the crystals had "stopped growing", as was seen with thesolutions at 0 ppm Pb dopant.At dopant levels of between 1990-4080 ppm Pb the KBr crystals were alloctahedral, and began to cluster even more during the initial stages of evaporation, withgroups of four or more tiny octahedra growing together (fig. 14a). This is the beginningof the growth of the dendritic form of KBr. At 4080 ppm Pb the acoustic integrationtrace very closely paralleled the image whitening (fig. 16c), without the second rise seen infig. 16b for crystals grown with 1990 ppm Pb dopant. Less acoustic emission is now seenafter the maximum image whiteness has been reached. A decrease in the size of the42octahedra is seen at dopant levels of between 1990-4080 ppm when compared with thoseproduced at dopant levels of between 0-1990 ppm (see figs. 13b, 14a).At dopant levels of between 4080-8020 ppm Pb, fingers of KBr [60] were seen togrow outwards from the edge of the drop, with islands of crystals observed in the centreof the droplet. Less spreading of the finger-like growth of KBr from the boundaries of thedroplet was observed compared with the preliminary experiments. This was due to thesmaller droplet size used (100 gl instead of up to 250 gl). The growing "fingers" werethus able to be kept within the field of view of the microscope. Microscopic analysis ofthese KBr fingers by SEM showed that they were comprised of long chains of many tinycoupled octahedra 25 gm or less in size (fig. 14b), while the central islands were found tobe clusters consisting of up to 10 tiny octahedra of similar dimensions. These tinyoctahedra were smaller than those observed in the dopant range of 1990-4080 ppm. Thisdendritic form of KBr gave excellent agreement between the acoustic integration and theimage whitening traces, and little or no acoustic emission was observed after the crystalsstopped growing (figs. 17a).At dopant levels of 12210 ppm Pb and above, needle crystals were observed withinan octahedral matrix (figs. 15a, 15b) that consisted of a mat of many octahedra of varyingsizes. These octahedra were larger than those produced during the growth of thedendritic form of KBr. This mat grew with much interaction of the crystals unlike theindividual crystal growth seen for the cubic form. Good agreement between the imagewhitening trace and the acoustic integration trace start and finish times was observed, butin 10 of 21 cases with lead dopant above 12210 ppm some emission was observed afterthe maximum image whiteness was reached. This is indicated by the second acoustic risein fig. 16d which occurred about one hour after the crystals had visibly "stopped"growing.13(a)(b)Fig 13. Scanning electron micrographs of (a) cubic KBr at 0 ppm Pb dopant,and (b) cubic to octahedral change of KBr at 1990 ppm Pb dopant.yH(a)(b)Fig 14. Scanning electron micrographs of (a) central islands of KBrobserved in droplet at 4080 ppm Pb dopant, showing the clusteringof the octahedra, and (b) dendritic fingers of KBr at 6070 ppm Pbdopant.(a)(b)Fig 15. Scanning electron micrographs of (a) K2PbBr4 needles in anoctahedral matrix of KBr, observed at 24840 ppm Pb dopant, and(b) enlargement of the central portion of fig. 15a, showing theinterpenetration of a needle with the octahedral matrix.0 4000 8000 12000 16000Time (secs)4000 Lo3.50(.7 'E3000 cl" --(1)D3.252000 c D,3.001000 3(7) .0^(n5.^2.75466000^4.005000^3.752.500^4000 8000 12000 16000Time (secs)2500020000CD15000 cl10000 050003cncn03.503.2501Ca)3.00a>010E2.752.5035000^3.50300005--25000 7^3.25LOC.]^0-)20000 7a. a)1.E15000 2^3.0010000 Fi  `,25000^2.75E .02.504.003.753.50:E 3.2503.002.752.50 10000080000rnIn60000 040000U)20000 rn3(n .0^00^3000 6000 9000 12000^0^4000 8000 12000 16000Time (secs)^ Time (secs)Integrated Acoustic Emission (Arbitrary Units (AU))Image Whitening (Grey level sum (*10e7))Fig 16. Integrated acoustic emission and image whitening curveswith time, for the growth of KBr crystals viewed at ax20 magnification with varying amounts of lead dopant,(a) 0 ppm Pb, (b) 1990 ppm Pb, (c) 4080 ppm Pb,and (d) 22060 ppm Pb.Magnification x7550000^3.5040000 CDtip030000 a)C120000 co10000 13(7) .0^53.252.75472.500^1000 2000 3000(b )3.503.25*E3.00a)0-)E2.752.500^4000 8000 12000 160001500CD1000 c00CD0500 0ccC)o 3(.7)'0u) .Time (secs)^ Time (secs)Integrated acoustic emission (Arbitrary Units (AU))Image whitening (Grey level sum (*10e7))Fig 17. Integrated acoustic emission and image whitening curveswith time, for the growth of the dendritic form of KBrcrystals with 8120 ppm Pb. Two different magnificationsshown, (a) x20, and (b) x75.483.2.2 Studies at Magnification x75Studies of all the different morphologies grown were also carried out at x75magnification. The same trends were observed between the acoustic integration trace andthe image whitening curves for all the morphologies. This included excellent agreementagain for the dendritic form of KBr (4080-8020 ppm) described in the preceding section,where the acoustic integration trace was found to parallel the image whitening tracealmost exactly (fig 17b). This is even more impressive when one considers that anexceptionally small mass of crystals grown was grown when using this magnification so asto keep all the growing crystals within the field of view of the microscope. The dropletsize used for these increased magnification experiments was 4 p.1 which produced a massof crystals on the order of 0.004 g.Total acoustic emission per gram of crystal produced was not calculated for thestudies carried out at increased magnification due to the exceptionally small mass ofcrystals grown.3.2.3 EDX and XRD analysis of NeedlesThe needles observed at dopant levels of 12210 ppm and above were identified byEDX analysis and XRD as K2PbBr4. When analyzed by EDX the needles showed a largeweight percentage of lead, 32%, relative to the octahedral matrix for which no lead wasdetected (table II and III). This amount of lead was consistent with the formula K 2PbBr4.However, given the roughness of the sample (non-ideal, in terms of the geometricassumptions made for truely quantitative EDX analysis), the EDX analysis was purelyused as an indicator of the relative amounts of each element in the different growth formsobserved, and not as a method of stoichiometric identification.49Table II.Average values of weight and atomic percentages of K, Br, and Pb determined by EnergyDispersive X-ray (EDX) analysis for 3 different needles shown in fig. 15a. Theoreticalvalues for K2PbBr4 are given for comparison. The range of experimental values waswithin ±10% (absolute).Experimental TheoreticalElement Weight % (±10) Atomic % (±10) Weight % Atomic %K 16 34 12.92 28.57Br 52 53 52.83 57.14Pb 32 13 34.24 14.29Table EllAverage values of weight and atomic percentages of K, Br, and Pb determined by EnergyDispersive X-ray analysis (EDX) at three different positions in the octahedral matrixshown in fig. 15a. Theoretical values for an octahedral crystal of KBr containing29790 ppm are given for comparison. The range of experimental values was within ±10%(absolute).Experimental TheoreticalElement Weight % (±10) Atomic % (±10) Weight % Atomic %K 37 54 29.65 48.51Br 63 46 60.60 48.51Pb 0 0 9.60 2.98Table II does indicate though a good agreement between the theoretical atomicand weight percentages of K, Br, and Pb in K2PbBr4 and those observed in the 3 needlecrystals grown at 29790 ppm Pb dopant. Table III showed that no lead was detected byEDX analysis in the octahedral matrix, whereas one might expect evenly distributed leadat a level of 29790 ppm to be detected. However, given that much lead is present asK2PbBr4 needles, the actual lead content of the octahedral matrix must have beensignificantly reduced. This is in keeping with the results observed.50A number of these needles (ft 20) were collected by hand using tweezers and amagnifying glass, and ground to a fine homogeneous powder using pestle and mortar.Powder X-ray diffraction patterns were then obtained and compared with referencespectra for KBr, K2PbBr4, KPbBr3, and KPb2Br5. The best spectral match was found tobe from K2PbBr4 (table IV, fig. 18).The crystal structure and lattice parameters of K 2PbBr4.H20 have been reportedpreviously [66]. Comparison of the experimentally obtained XRD pattern with standardpatterns from the International Centre for Diffraction Data confirmed the stoichiometry ofthe needles. The first two XRD peaks at positions of 20 = 12.4° and 13.6° were the maindiagnostic peaks used (see table IV). Two other possible stoichiometric combinations ofK, Pb and Br exist, KPbBr 3 and KPb2Br5. The XRD pattern for KPbBr3 does not showany peaks below 20=21.6° however, and KPb 2Br5 while having peaks at a similar positionto K2PbBr4 also has a major peak at 20=9.5° and numerous peaks between 20=13.5° and20° where the observed XRD pattern had none. Obviously, one would expect a little pureKBr from the octahedral matrix (fig. 15a) to be present as the needle separation wascarried out manually. The XRD pattern observed (fig. 18a) showed a combination ofpeaks from KBr and K2PbBr4, and thus did confirm the preliminary compositional analysisof the needles determined by EDX.The white spots observed on the needles in figs. 15a, 15b were examined forcompositional variation with respect to the bulk needle by EDX. No compositionaldifference was found. It is possible that the white spots are evidence of internal damage ofthese needles. There is some evidence for this as most of the white spots occur along theedges of the needles and at points of interaction with the octahedral mat.Peak Positions ofK2PbBr4^Peak Positions of KBr20 (degrees)^20 (degrees)12.42213.63319.93621.18821.82122.26223.58024.16425.35326.42727.59327.76829.06329.81732.20934.20834.70134.96735.40935.84738.18539.89240.35941.62544.83423.33126.98338.52845.52147.68455.66361.14062.92269.73774.63682.62687.33688.90995.14299.872101.431107.902112.901114.597121.696127.380138.037148.46751Table IV.Powder X-ray diffraction (XRD) peak positions for K2PbBr4 and KBr obtained from theInternational Centre for Diffraction Data, Powder Diffraction File #2. Compare withfig. 18.56 648^16^24^32^40^4820 (degrees)523000^ 26) (degrees)T1 )-1^200010000Fig 18. Powder X—Ray diffraction patterns for (a) needle crystalscollected during the growth of KBr crystals from a solutiondoped with 29790 ppm Pb, and (b) analytical grade KBr.5000>, 4000—cna) 300010000700060007t72 5000ccn4000.- - 20^40^60;^80^100 1208^16^24^32^4028 (degrees)48^56^64(b)533.2.4 Trends in Integrated Acoustic EmissionThe total acoustic emission was recorded for each replicate of the different KBrmorphologies and normalised to the mass of crystals produced (< 0.12g). The plot infig. 19 shows the total acoustic emission variation with the degree of lead doping due tothe changes in morphology of the KBr crystals. A large increase (10 fold) in totalemission can be seen as the level of lead dopant is increased from 0 ppm to 8020 ppm.Thereafter the total emission remains high, but is variable. The range of the results is suchthat the unequivocal confirmation of a second maximum at 24840 ppm Pb dopant was notpossible (see fig. 19).The lower variability in the total acoustic emission observed for the replicates ofseries #2 compared with series #1 is attributed to the difference in sample size. The largersample size used for series #1 produced a greater spreading of the crystals outwards fromthe piezoelectric crystal sensor. The sensitivity of the transducer decreases with distancefrom the central piezoelectric crystal sensor and so some signals may no longer beobserved. This limitation was confirmed during the course of the impact experiments(section 2.3.3.3).16.0co0(f)^14.04-6 (I) 12.0oo cy) 10.0c^8.00 EE iD 6.00 0L._4.0o00 2.00F-0.00^5000^10000^15000^20000^25000^30000^35000Degree of Lead Doping (ppm)Fig 19. Total acoustic emission per gram of KBr crystals grown with varying amounts oflead dopant. 3 experimental series are shown.553.3 Source of the Acoustic EmissionThe proposed mechanism for acoustic emission generation during thecrystallization of the different KBr morphologies is one of inter-crystal interaction. This isthe first time such a mechanism has been postulated, and evidence for it will be discussedin the following sections.3.3.1 Comparison of the Rate of Acoustic Emission to the Rate of CrystallizationTwo different stages of acoustic emission activity occur during the recrystallizationof potassium bromide, corresponding to two distinct rises in the acoustic integrationprofile. Firstly, there is the acoustic activity which for certain KBr morphologiesparalleled the image whitening curve exactly, as in the case for KBr crystals grown with8020 ppm Pb dopant (see fig. 17a, 17b). Then there is the acoustic activity which wasobserved after the crystals have visually stopped growing, such as in the case for the KBrcrystals grown with 0 ppm lead dopant (see fig. 16a). In between these two dopant levelsthere are a series of two stage curves (see fig. 20), with the first stage following the imagewhitening curve exactly and then a second stage of activity seen after the crystals havestopped visually growing. As the dopant level increased from 0 ppm to 8020 ppm thesecond stage acoustic emission rise decreased until eventually at dopant levels above 4080ppm Pb it disappeared completely, and only the first stage acoustic curve was evident. Asecond stage acoustic rise reappeared at dopant levels of 12210 ppm and above in 10 of21 cases. This second stage acoustic rise again came after the maximum image whitenesshad been reached.The integrated acoustic emission profile evolution during the growth of KBrcrystals with varying levels of lead dopant is shown in fig. 20. The first stage acoustic rise56closely paralleled the image whitening curves (figs. 15b, 15c, 16a, 16b). The evolution ofthis first stage acoustic rise can be seen across figs. 20a-20g. This evolution correspondedto the increased clustering observed for the KBr morphologies with increasing Pb dopant,until the dendritic form of KBr (which consisted of contiguous chains of tiny octahedra)was formed at dopant levels of between 4080-8020 ppm (fig. 20). The integrated acousticemission profiles of the dendritic form of KBr shown in fig. 20 gave the best agreementwith the image whitening curves. No acoustic emission was observed after the maximumimage whiteness was reached for these forms (figs. 16c, 17a, 17b).The acoustic emission activity which was seen to accompany the growth of theKBr morphologies between 4080-8020 ppm gave a reliable indication of the rate ofgrowth of this particular morphology. Simultaneous acoustic emission with imagewhitening only appeared when the growth forms were such that many microcrystalsnucleated and grew together (see figs 14a, 14b, 15a-octahedral mat), so that muchinter-microcrystal interaction could occur.Consider the solution in contact with the surface of the finger-like KBr structuresproduced with between 4080-8020 ppm Pb. As it evaporates, any tiny octahedralmicrocrystal that nucleates would exert a force on the neighbouring microcrystals withinthe structure. The force exerted by each microcrystal causes tiny cracks to develop in thebulk structure of the finger, leading to acoustic emission. Since many tiny microcrystalsmake up these structures, numerous potential sources of acoustic emission exist.Therefore for crystal morphologies consisting of these clusters of tiny microcrystalsacoustic emission provides a very reliable method of monitoring the overall growth rate ofthe structure. The growth of the dendritic form of KBr which was observed between4080-8020 ppm is perhaps the best evidence for the inter-crystal interaction mechanism.57The greatest number of possible acoustic sources for the inter-crystal interactionmechanism are present within the structure of dendritic KBr. This would account for theincrease in the total amount of emission seen in fig. 19 over the dopant level range of0-8020 ppm. The initial increase seen to occur between 0-4080 ppm could be accountedfor by the increased clustering observed, before eventually leading to the dendritic form.It is postulated that the acoustic activity which was observed after the crystals hadstopped growing was in part due to the release of stresses which had developed as thecrystals pushed against each other during their growth. This is primarily thought to occurby the movement of dislocations within the crystals and microcracking in and betweencrystals. Consider the small spaces between two cubic crystals. As they begin to touchany crystalline material that nucleates and grows between the two cubic crystals is goingto induce some stress in the matrix, and thus provide a potential site for acoustic emission.The acoustic emission after the crystals had stopped growing is indicated by the secondstage rise in the integrated acoustic emission profiles of fig. 20. It should be noted that theintegrated acoustic emission profile for the undoped (0 ppm Pb) KBr (figs. 16a, 20) is asingle stage curve. However, this curve consists entirely of what has been termed here thesecond stage acoustic rise, i.e. acoustic emission that occurred after the maximum imagewhiteness had been reached (see fig. 16a).The plateauing seen in fig. 19 for those crystals grown with dopant levels greaterthan 12210 ppm can be explained by considering the size and number of octahedra whichare present in the mat. The octahedral mat of these systems does not have as manyinter-crystal interactions as the dendritic growth form observed at 8020 ppm due to thelarger size of the octahedral crystals in the mat (100 gm) compared with those observed inthe fingers of KBr at 8020 ppm Pb (<25 gm). There are thus less potential sites for thegeneration of acoustic emission. The numerical values for the total acoustic emission per58gram had a great variation for dopant levels of 12210 ppm and higher. This variation wasattributed the growth and fracture of needles of K2PbBr4 (see section 3.3.4).3.3.2 Integrated Acoustic Emission and Image Whitening Rise TimesA comparison of the rise times for both the integrated AE profiles and the imagewhitening curves for all levels of Pb dopant is illustrated in fig. 21. As can be seen theimage whitening rise time is very similar across all levels of Pb dopant. This suggests thatthe relative saturations of the starting solutions and the ambient conditions in thelaboratory were such that the rates of crystal growth were similar for all solutions studied.This is important as the speed of the many physical phenomena has a direct bearing on theamount and rate of acoustic emission generated [10].The rise times were calculated by considering the time taken for the curves to risefrom 10% to 90% of the final value. The agreement between the average total AE risetime and the average image whitening rise time (at one dopant level) increased as thedopant was increased from 0-8020 ppm. Beyond this point it showed a greater degree ofvariability. The best agreement between the rise time of the two profiles is observed forthe dopant range of 4080-8020 ppm where the greatest clustering of micro-crystals isobserved. This is as expected from figs. 14a, 14b. The AE rise time for the signalscollected during the growth of KBr crystals with 0 ppm dopant is much larger than theimage whitening rise time. This showed that much acoustic emission occured after thecrystals had visually stopped growing. The variation between the average total AE risetime and the average image whitening rise time at dopant levels above 12210 ppm was dueto the second stage rise (see figs. 16d, 20). When one considers just the average AE risetime for the first stage acoustic rise only, then the agreement with the average imagewhitening rise time at dopant levels of 12210 ppm and above is excellent (see fig. 21).8120 ppm J 4080 ppm111111J 6070 ppm111[124840 ppm 27790 ppm 31960 ppm111^11111Time (5000s per div.)A59J 1050 ppmI^I1990 ppm1 1 1 110 ppm1 11 1 1670 ppmQ)00CN1C0—(/)_0Q)C—0cY)(3)0(1)C7)0-C(1)s_(1)012210 ppm - 16050 ppm^i19960 ppmi 22060 ppm1111 1Fig 20. Integrated acoustic emission profiles observed duringthe growth of KBr crystals with varying degrees oflead dopont.60I^I^1Needles in an octahedral matrix1000080006000400020000^7000^14000^21000^28000^35000Degree of Lead Doping (ppm)Key:—■ ■^Average image rise time0^0^Average of first stage AE rise time.L^.L^Average of total AE rise time (1st and 2nd stages)Fig 21. Comparison of acoustic integration and image whiteningrise times as the KBr crystal morphology is changedby increasing the lead dopant.613.3.3 Dimensionality of the Dendritic Growth FormThe dendritic growth form of KBr occuring between 4080-8020 ppm wasobserved to grow outwards from the solution droplet to cover the surface of thetransducer. The "dendrites" produced were irregularly shaped and twisted in all directionswith further branchings on each outgrowth observed. These finger-like structures wereobserved in most cases to rise from the surface of the transducer by as much as 1 cm toproduce a tree-like structure in 3 dimensions (fig. 11). Occasionally however, the growthobserved was confined to 2 dimensions on the surface of the transducer. These fingerswere observed to be much thinner (less clustering) and spread further.A comparison of the two forms showed that the total emission of the 2D growthwas lower (reduced by half or more) per mass of crystal than the 3D form (see fig 22).This result can be rationalised by considering the inter-crystal interaction mechanismproposed earlier. During the growth of the 2D form one degree of interaction has beenremoved, and indeed the fingers observed were thinner due to less clustering of themicrocrystals in the structure. Again this result is in keeping with the proposedmechanism.Partial 3Dgrowth ofKBr fingers _ 2D growthof KBrfingers623D growth of KBr fingers140)co 12o a)0cy) 10a)o_. _• cn. 0ELtJao-+;(/) 11- SU08^21^2^3^4^5^6Replicate #Fig 22. Total acoustic emission per gram of crystals proucedduring the growth of 5 replicate samples of KBrdoped with 8120 ppm Pb.63It was postulated that the change in dimensionality observed for this growth formwas related to the rate of evaporation of the droplet. The ambient temperature did notchange significantly during these replicates, but on examining climatic data for the days onwhich the crystals were grown the relative humidity was found to have changedsignificantly. It was therefore appropriate to determine if a correlation existed.Meteorological data collected on an hourly basis were obtained from the University ofBritish Columbia's Climate Station. The relative humidity (R.H.) measurements yieldedthe saturated water vapour pressure of the air by using Teten's empirical formula (1). Therelative humidity relationship (2) was then used to calculate the actual water vapourpressure at any hour:e' = (0.6108) antilog10 [7.5 T / (T + 237.3)]^(1)R.H. = 100 e / e'^ (2)where e' = saturation vapour pressure (kPa), e = vapour pressure (kPa) andT = temperature (°C).Given that the chemistry department is vented directly to atmosphere with no artificialmethods of humidification it might be expected that the meteorological vapour pressuretrend would approximate that inside the laboratory. A comparison of the total acousticemission per gram of crystals grown with the initial vapour pressure showed no trend (seefig 23). The 5 replicate experiments considered produced both 2D and 3D dendritic formsof KBr (8020 ppm Pb). Indeed no trend was observed between the vapour pressure andtotal AE across the replicates of any of the KBr morphologies grown. These comparisonswere made subsequent to completion of the crystal growth experiments.64cn 14cf)(i) 12o0o1 0LQ)0 8-c—r)^(I) EE60 402H-0.8^1.0^1.2^1 . 4Vapour Pressure (kPa)0 1^ 4 00 25 00 3Fig 23. Effect of vapour pressure on the totalacoustic emission per gram of crystalsproduced during the growth of thedendritic form of KBr with 8120 ppm Pb.Numbers 1 —5 refer to replicates of fig. 22.653.3.4 Fracture Related Damage at Dopant Levels Above 12000 ppm PbThe needles observed at dopant levels of 12210 ppm and above grow into theedges of the octahedral crystals which form the mat. Fig. 15b shows such an interaction.There is damage at the point of contact between the needle and the octahedra, withassociated debris at that location (LHS, fig. 15b). Complete fracture of these extremelyfine needles may also occur. A close examination of fig. 15a indicates some misalignmentof the needles where they encounter the edges of octahedra within the mat. The damagedue to the interaction of the needles with the edges of the octahedra (flaking of debris) andthe complete fracture of the needles, would be possible sources for acoustic emission.Fracture would be an exceptionally "loud" event (discussed below). Variation in thenumber of needle breakages would account for the larger degree of scatter in fig. 19, atdopant levels of 12210 ppm and above.3.4 Characterization of Crystallization Processes by Acoustic Waveform AnalysisThe signals collected were classified prior to being analyzed in both the time andfrequency domain.3.4.1 Generation of Different Signal ClassesAcoustic frequency data were collected for each different KBr morphology using abandwidth of 1.25 MHz. Each individual signal was described by 32 mathematicaldescriptors (see table I). These descriptors, their mathematical meaning, and theirinterpretations have been fully described elswhere [17, 67]. From the 1000's of signalscollected for each KBr morphology a random (computer random number generator)sample set of 50 signals was selected. Signal sets collected during the growth of 666different KBr morphologies were combined and classified to form a single set of 300signals. The signals collected for the dendritic form of KBr at 5900 ppm Pb dopant and7620 ppm Pb dopant were treated as one class (0). Signals from the KBr morphologiesobserved at 24920 ppm Pb dopant, and 29790 ppm lead dopant were treated as a secondclass (A). Signals from the KBr morphologies observed at 0 ppm (0) and 15980 ppm(V) were each treated as separate classes. The reduction in the number of signals, fromthe 1000's collected to 50 signals per class, decreased the computational time forsubsequent PCA calculations. The resulting principal components plots also remainedrelatively uncluttered, enabling the different signal classes to be readily identified.This set of 300 signals was then compared with the artificially generated signalsdesigned to mimic secondary nucleation phenomena (see section 2.3.3.3)3.4.2 Signal IntensitiesTypical acoustic signals for these 6 classes were selected by comparing eachsignal's descriptor values with the average descriptor values for the class. The signal withthe lowest variance (from the mean) for each data set, over all the descriptors used wasthus selected as typical, so avoiding any bias. These signals and their power spectra areshown in fig. 24.It should be noted that the intensity of all the signals is at least ten times larger thanthe background blank (compare with fig. 6). Further, that the intensity of the signal due toartificial fracture is much greater than that observed during the growth of the differentKBr morphologies. The intensity of the signals observed during the growth of thedendritic form of KBr (at 4080-8020 ppm), a plural crystal, showed no intensity differencefrom those signals collected at 0 ppm Pb. This is contrary to what Sawada et al observed67during the precipitation of sodium thiosulphate when the production of plural crystals wasaccompanied by acoustic signals of the greatest intensity [6].Although the artificially generated signals do not mimic the processes of secondarynucleation exactly, the large difference in intensity of the signals of bulk fracture (fig. 24d)compared with the signals observed during growth (figs. 24a-24c) has significantramifications. AE is of interest for non-invasive monitoring of industrial crystallizers.Care would be needed if detection of growth signals were to be a desired analyticalparameter, since it is likely that any growth signals will be below the background level setfor any industrial monitoring.The average power spectra (AVP) for the same processes indicated in fig. 24 areshown in fig. 25. The AVP's clearly demonstrate the similarity of the signals observedduring the growth of the different KBr morphologies (figs. 25a-25c). It can also be clearlyseen from fig. 25 that the AVP's observed during growth are much more akin to theartificial fracture signal (fig. 25d) than to the impact signal (fig. 25e).It should further be noted, that during the bulk fracture of the dendritic fingers ofKBr, any crack would propagate around the edges of all the small microcrystals. Thepower spectra of individual signals of this bulk fracture and the AVP calculated across allthe signals of bulk fracture are thus almost identical (compare figs. 24d, 25d). Theindividual signals arising during the formation of the dendritic KBr morphology resultfrom the microfracture around just one microcrystal. The resulting power spectrum of anindividual signal in this case is much less complex than that produced during the bulkdendritic finger fracture (compare figs. 24b, 24d). The average power spectra of thesetwo processes are however very similar (compare 25b, 25d). This is further evidence forthe inter-crystal interaction mechanism.680.500.250.00- 0.25- 0.500.500.25- 0.0073,> -0.25- 0.500.500.250.00E -0.25- 0.506.003.00crC75^0.00-3.00-6.0010.005.00i^(a)l'I Ili^+#104ritioomkor-I(b) --f1__,..4.,„._.....-4.4.....14.-P ......Aitt,....&■(c)__,.......4,,All(d) -0.1'11i_I)16, A OLAJa,„___ir4' I\4vV(e)-j\Iii" _,_,i^I^iiiiiU444....._A I0.12^0.24^0.36 0^300 600 900 1200^Time (ms) Frequency (kHz)0.00-5.00-10.000.000.030.020.010.000.030.020.010.00 CD00.03^r-1 -CD0.020.01 CCD0.00 C0.600.400.200.001.200.800.400.00Fig 24. Typical acoustic signals and associated power spectrafor KBr crystals grown with (a) 0 ppm, (b) 7620 ppmand (c) 24920 ppm Pb dopant, (d) bulk fracture offingers of KBr grown with 7620 ppm Pb dopant, and(e) impact of KBr crystals on the transducer surface.iI^i^i^i^i^I_ A (a)^-111 1250^500^7500 1000 1250i^ i 1i^69(b)^_0.0100.0080.0050.0030.0000.0160.0120.0080.0040.0000.0100.0080.0050.0030.0000.2000.1500.1000.0500.0001.5001.0000.5000.000Frequency (kHz)Fig 25. Average power spectra for acoustic signals collectedduring the growth of KBr crystals with (a) 0 ppm,(b) 7620 ppm, (c) 24920 ppm of Pb dopant, andartificially produced signals of (d) bulk KBr dendritefracture, (e) KBr crystal impacts on the transducer.703.4.3 Signal Class Separation by PCAPrincipal Components Analysis (PCA) was used to ascertain if there were anydifferences in the acoustic signatures from the 6 different classes of acoustic crystallizationsignals. The number of descriptors used for PCA was reduced from 32 to 18 using themethod outlined in section 2.4.2 (see table V).A plot of the first two principal components is shown in fig. 26. Two readilydistinguishable sets of signals can be identified. One group consisted almost exclusively ofacoustic signals generated by the impact of KBr crystals on the transducer surface (0);while the other group consisted of a tight cluster of acoustic signals from the crystalgrowth of the higher doped KBr crystal morphologies, (5900 and 7620 (0), 15980 (V),24920 and 29790 ppm Pb (s)), together with a secondary cluster of acoustic signalsobserved during the artificial fracture of KBr dendrites (•). The acoustic signals observedafter the KBr crystals doped with 0 ppm Pb had stopped growing (0) were scatteredthroughout this second group, with the characteristics of some signals observed at 0 ppmPb dopant being very similar to the impact signals that were artificially produced.In order to obtain a better separation of the signal classes, the artificially producedsignals of KBr dendrite fracture and impact were then compared with those signalsgenerated during the growth of two distinctly different KBr morphologies, the dendriticform of KBr (5900 and 7620 ppm Pb), and the undoped cubic form of KBr (0 ppm Pb)again using PCA. These four classes of acoustic crystallization signals were thencompared pair-wise.71Table V.Descriptors that showed the highest resolution for the 6 different classes of crystalli7ationsignals.Time Domain DescriptorsRMSCRESTKURTOSIST@AREA/210-CROSSTIME OCTILES1/8, 5/8, 8/8TRoot mean square voltageRatio of peak voltage to root mean square voltage4th statistical moment (deviation from a Gaussiandistribution)Time to half area (signal decay measurement)Number of times signal crosses +/- 10% of maximumvoltageNormalized time octiles of root mean square voltageFrequency DomainDescriptorsFRQ@MAXFRQ MEANF-CRESTFBW> 15%FQRTLBWFREQUENCY OCTILESDFB1, 2, 5, 6, 7Frequency of highest intensity in the power spectrumFrequency equal to the summation of the intensity-weightedfrequencies divided by the total intensity.Ratio of the maximum power to the root mean squarepowerBandwidth of frequencies having intensities > 15% ofmaximum intensityBandwidth of frequencies between the second and thirdintensity integrated quartilesNormalized area in each power spectrum octile0—264cv2—4—672—6^—4^—2^0^2^4^6^8^10Principal Component #1Key:-KBr crystal growth signals with:^Artificially produced signals:o 0 ppm Pb dopant^ • KBr dendrite fracture^ 15980 ppm Pb dopant^■ KBr impact,n, 24920 and 29790 ppm Pb dopant^ 7620 and 5900 ppm Pb dopantFig 26. Principal components analysis showing the separationof different crystallization processes. Only 18 of 32descriptors were used (see table V).RMS25-CROSS10-CROS S0-CROSS5/8T8/8 TFRQMEANFRQMEDDFB1DFB4Root mean square voltageNumber of times signal crosses ± 25% of maximum voltageNumber of times signal crosses ± 10% of maximum voltageNumber of times signal crosses 0 voltsNormalized time octile of root mean square voltage0.205-0.256 msNormalized time octile of root mean square voltage0.359-0.410 msFrequency equal to the summation of the intensity-weightedfrequencies divided by the total intensityFrequency at the mid-area of the integrated frequencyintensitiesNormalized for area power spectrum octile 0-156 kHzNormalized for area power spectrum octile 625-781 kHz73For each pair of classes selected the highly correlated (>90%) descriptors wereagain discarded, and the remaining descriptor differences between the two classes wereanalyzed. The three descriptors which showed the highest resolution for each pair werethen used for PCA. The descriptors that were found to have the highest resolution fromall the different pairs are shown in table VI.Table VLDescriptors that showed the highest resolution for the pairwise separation of artificialsignals of bulk fracture and crystal impact from the cubic (undoped) form of KBr, and thedendritic (4080-8120 ppm Pb) form of KBr.Descriptor NameThe PCA plots in fig. 27, clearly show the ease of separation of the KBr impactsignals from the acoustic signals collected during the growth of both morphologies of KBr(fig. 27a, 27d). Note that the separation obtained is much cleaner than might have beenexpected from fig. 26 alone. The KBr impact signals were also readily separated from theKBr dendrite fracture signals (fig. 28b). The signals obtained from artificially produced74KBr dendrite fracture could be easily distinguished from the KBr crystal growth signalsfor 0 ppm Pb dopant (fig. 27c) but less easily from the KBr crystal growth signals for7620 and 5900 ppm Pb dopant (fig. 27b). The KBr crystal growth signals of the twodifferent morphologies could not be distinguished (fig. 28a).Signals from the six different morphologies of KBr crystal growth (excluding thebulk fracture and impact signals) were also collected using a bandwidth of 2.5 MHz. Thesubsequent PCA (fig. 29) did not show any increased separation of these classes whencompared with the analysis conducted using signals collected with a bandwidth of1.25 MHz (fig. 26). This lack of separation between the individual classes of signalscollected during the crystal growth of different KBr morphologies shows that the twostages of acoustic emission, the first that parallels the image whitening and the secondobserved after the crystals have visually stopped growing, could not be distinguished bytheir acoustic signatures using the present descriptors.This analysis demonstrates that the signals collected during the growth of thedendritic form of KBr (fig. 14b) were more similar to the artificially produced fracturesignals than the impact signals. This lends further support to the inter-crystal interactionmechanism which involves the inter-microcrystal fracture as the mechanism for acousticsignal generation. This analysis also suggests that the signals which arise during crystalgrowth might be readily discerned from bulk fracture signals (due to secondary nucleation)and signals from impact of crystals onto a sensor or waveguide that would occur in anindustrial crystallizer [53].5 4 3 2 1 0 1 2 30-2-3 -2 1 0^1^2^3 -3 -2 -1 0 1^2 3CN^212NJca)c0a1I^I_^(d)IE0 0UTD0_- 1 05C E0_ -2 I^I I075(N^3."1-c.'^2a)coa 1E00 00a (7.)^-1C Ea_ -2-3-2-1 0 1 2 3 4 5CN^32.4"E'a)c0^1aE^00U--6^1a(.7) -2Ca_- -3Principal Component #1^Principal Component #1Principal Component #1 Principal Component #1Key:-Artificially produced signals: KBr crystal growth signals with:• KBr dendrite fracture^o 0 ppm Pb dopant■ KBr Impact^0 7620 and 5900 ppm Pb dopantFig 27. Principal components analysis showing the separationof acoustic signals that occurred during crystal growthfrom artificially generated signals mimicking secondarynucleation processes. Descriptors used for separationswere (a) RMS, 0-Cross, Frqmean, (b) RMS, 25-Cross,Frqmed, (c) RMS, 10-Cross, 8/81, (d) RMS, 0-Cross,DFB1 (see table VI).b()   it,• F—■ Iv dpi•%c••*IS?" 1■ i■••^•    • •• • • ■•• ■■■32CN1E' 1C00._0Eo00°- —1:)C ,T_a_ —2—32—3CV^10—1—276—2 —1 0 1 2 3 4 5^—3 —2 —1 0 1 2 3 4Principal Component #1 Principal Component #1Key:-Artificially produced signals:^KBr crystal growth signals with:• KBr dendrite fracture^o 0 ppm Pb dopant■ KBr impact^ o 7620 and 5900 ppm Pb dopantFig 28. Principal components analysis of acoustic signalscollected from both artificially produced crystallizationprocesses, and actual growth processes of differentKBr crystal morphologies. Descriptors used forseparations were (a) 10—Cross, 5/81, DFB4, and(b) RMS, 5/8T, 10—Cross (see table VI).^,^—rr^••06GL o • 0.•,.01■ v.A^ V 0^^°1^C)^C.0 A CA v ■ O .70, • 4 .0 A g 07^9y, ■^7 C.t. II Ilb  A A A v 07 it v-... cm Zw ' .S:70., sr.^•^0■ 0 ■^ ■ 0• ^0 •• 0^• 0 v v,o, v^w .111^6 A ^ ; 7^■ IV. ■ ■7■ AY7^•CZ! OAL lit ll ° •o 7 13 0 7 111 •".•■ o^•—A^7 c-f?3 0^Q ° % ■ .n,oci 'I;'^C V " [Yci °^ 7 v aF o#^o v^• ^ °■ 0•OO•••0o •0 0O86—4—6420—277—10 —8^—6^—4^—2^0^2^4^6Principal Component #1Key:—KBr crystal growth signals with:• Oppm Pb dopant o 15980 ppm Pb dopant^ 3980 ppm Pb dopant ° 24920 ppm Pb dopant■ With 7620 ppm Pb dopant v 29790 ppm Pb dopantFig 29. Principal components analysis of acoustic signalscollected with a 2.5 MHz bandwidth during thegrowth of different KBr crystal morphologies. Only18 of 32 descriptors were used (see table V).783.5 Classic Dendritic GrowthThe acoustic activity of the dendritic form of KBr observed with 4080-8120 ppmPb dopant was compared with the classic dendritic growth produced upon evaporation ofNH4C1. A typical integrated acoustic emission trace observed during the dendritic growthof NH4C1 is shown in fig. 30. Compare this very stepped curve with the smooth curveobserved during the growth of the dendritic form of KBr (figs. 17a, 17b). The totalemission observed for the dendritic growth of NH 4C1 is negligible compared with thatobserved during the growth of the dendritic form of KBr.These observations were in agreement with predictions based on the nature of thetwo growth forms. The dendritic growth of NI-1 4C1 involves the growth of single crystalstems oriented in particular directions (see fig. 3 on p. 15), i.e. no inter-crystal interactionis expected. The source of emissions observed during the growth of this form of NH4C1 isattributed to the occassional bulk fracture of these stems, an event that does not mimic thegrowth rate of this form. The resulting stepped curve is indicative of the infrequentfractures of these stems or dislocation movements within the stem. Each step is from adiscrete event, whereas the smooth curve observed for the dendritic form of KBr isindicative of many sources of acoustic emission resulting from the interaction of the tinymicrocrystals that make up the structure. Obviously, with a small enough time intervalbetween each data point the smooth AE profile observed for the dendritic form of KBrwould also show discrete events.60500cn(7)< 40U1 Q1000^1000^2000^3000^4000 500079Time (secs)Fig 30. Integrated acoustic emission profile for thedendritic growth of ammonium chloride, showingthe infrequent signals, leading to a stepped plot.85Literature Cited(1) J. Berlan and T. J. Mason, Ultrasonics, 30(4), (1992), 203-212.(2) M. A. Bukhnii, L. A. Chernosatonskii and R. G. Maev, J. Microsc., 160(3),(1990), 299-313.(3) E. T. Zellers, R. M. White and S. M. Rappaport, Anal. Chem., 62, (1990),1222-1227.(4) H. Coufal and J. F. McClelland, J. Molec. Struc., 173, (1988), 129-140.(5) K. Marr and K. S. Peters, Biochemistry, 30, (1991), 1254-1258.(6) T. Sawada, Y. Gohshi, C. Abe and K. Furuya, Anal. Chem., 57, (1985),1743-1745.(7) S. D. Lubetkin, J. ADD!. Electrochem., 19, (1989) 668-676.(8) P. D. Wentzell, S. J. Vanslyke and K. P. Bateman, Anal. Chim. Ada, 246,(1991), 43-53.(9) E. L. Lube and A. T. Zlatkin, J. Crystal Growth, 98, (1989), 817-826.(10) R. M. Belchamber, D. Betteridge, M. P. Collins, T. Lilley, C. Z. Marczewski andA. P. Wade, Anal. Chem., 58(8), (1986), 1873-1877.(11) J. A. Simmons and H. N. G. Wadley, J. Research (NIST), 89(1), (1984), 55-64.(12) 0. Lee, Y. Koga and A. P. Wade, Talanta, 37(9), (1990), 861-873.(13) P. D. Wentzell and A. P. Wade, Anal. Chem., 61, (1989), 2638-2642.(14) A. G. Beattie, J. Acoustic Emission, 2(1-2), (1983), 67.(15) P. A. Gaydecki, F. M. Burdekin, W. Damaj, D. G. Johns and P. A. Payne,Meas. Sci. Technol., 3, (1992), 126-134.(16) Y. Higo and H. Inaba, J. Acoustic Emission, 8(1/2), (1989), 7-24.(17) P. D. Wentzell, 0. Lee and A. P. Wade, J. Chemom., 5, (1991), 389-403.(18) R. G. Brereton, "Chemometrics: Applications of Mathematics and Statistics toLaboratory Systems", Ellis Horwood Ltd., Chichester, U.K., (1990).803.6 Primary NucleationOn tapping the graduated cylinder when the temperature had reached 68°C (2°C ofundercooling) a few hexahedral crystals were observed within the solution. After a shortperiod of time (<3 mins) the production of a large number "avalanche" of tiny KBrwhiskers was observed in the cylinder. The crystals produced fell through the solution toproduce a layer 1-2 cm deep.The mass primary nucleation observed in the solution produced no detectableacoustic signals. This is in keeping with the crystallization sequences observed for KBrwith Pb dopant where the nucleation of individual hexahedral crystals produced noaccompanying acoustic emission. The difficulty in separating the onset of acousticemission from the nucleation of KBr crystals at higher levels of Pb dopant is due purely tothe initial clustering of the tiny micro-crystals. The resulting acoustic emission isproduced by the inter-crystal interaction mechanism. The whiskers of KBr observed wereso small that even their impact on the bottom of the graduated cylinder was notacoustically detectable with the apparatus used. However, this was a function of theattenuation produced by the experimental arrangement81IV CONCLUSIONSThe combination of in-situ optical imaging with acoustic emission and ex-situmorphological analysis has led to the first clear understanding of the mechanism ofacoustic signal generation during crystallization. It has been proven that the growth ofcrystal structures in which many microcrystals are in intimate contact can be followedreliably using acoustic emission. The dendritic form of KBr (4080-8020 ppm Pb dopant)and the octahedral crystal mat (> 12000 ppm Pb dopant) are two such forms. Thereproducibility of acoustic emission and optical measurements was surprisingly good,especially when one considers that only a small mass of crystals (0.004-0.12g) was grown.The importance of determining the exact mechanism of acoustic signal generationduring crystallization processes cannot be stressed enough; at present there are no modelsfor these processes. To within the detection limits of our apparatus the growth of a singleoctahedral microcrystal was not acoustically emissive. However, in all cases interaction ofmicrocrystals during the formation of larger crystal structures produced much emission.For example the dendritic form of KBr fingers that was found to consist of chains of manytiny coupled octahedra produced copious amounts of emission during growth.The inter-crystal interaction mechanism postulated in this thesis explains all of theobservations not only at the microscopic level for the dendritic form of KBr but also at themacroscopic level for the cubic form of KBr, when the initially discrete crystals begin totouch and interact. However it is the interactions of microcrystals during the formation ofthe dendritic morphology and octahedral mat that make acoustic emission particularlyvaluable for monitoring the growth of these larger structures.82Primary nucleation of crystals was not detected acoustically during growth of thedifferent KBr morphologies. Indeed even the massive primary nucleation of KBr whiskerswas not detectable. From this one may conclude that the onset of crystallization cannot bedetected acoustically for these crystal forms. Completion of growth of the dendritic formsof KBr was accompanied by cessation of the observed acoustic emission. For structureswhere the inter-crystal interaction mechanism is the primary source of acoustic signalgeneration, acoustic emission monitoring is shown to provide a method of following theprogress and determining the end of crystal growth.The acoustic signals generated by growth of these KBr morphologies werecharacteristic of the process and could be readily distinguished from acoustic signalsgenerated by artificially produced dendrite fracture and impact of KBr crystals on thetransducer surface. This study provides the basis for the non-invasive monitoring ofcrystallization whereby acoustic detectors may be used to follow the growth and theprocesses of secondary nucleation within a solution. The ablity to distinguish betweensignals arising during growth, those produced by impact of crystals suspended in solution,and those produced during bulk fracture will be enhanced if the mechanisms of acousticgeneration are known. During this study it was proved possible to separate these differentsignal classes using a priori knowledge. However, the use of acoustic emission forfollowing crystallization processes will only reach its maximum potential when thesesignals can be unambiguously classified on a routine basis.83V FURTHER WORKThe dendritic form of KBr that produced the most interesting results in this studyis a morphology of KBr that occurs when the crystals are grown from solution byevaporation. Industry however uses undercooling (at several degrees below theequilibrium saturation temperature) as the method to produce supersaturated solutions forcrystallization. Further work should focus on other dendritic forms of crystals that wouldbe produced in solution. One such system is the growth of sodium chloride doped withsodium ferrocyanide [68]. The dendritic form of NaC1 consists of tiny cubelets clusteredtogether to form a larger hexahedral structure (fig. 31a). This is much like the dendriticform of KBr, although the outward appearance of the structure is more ordered. Thegreater order of this structure may decrease the number of inter-crystal interactions andthus lessen the effectiveness of AE for monitoring of this form. A granular form of NaClcan also be produced that appears to consist of a single crystal central core with apolycrystalline outer layer of tiny cubelets arranged to produce a sphere (fig. 31b). Thissecond form might provide more sites for AE due to the less ordered nature of thestructure. The study of the NaC1 system should provide further insight into themechanism of acoustic signal generation during crystal growth.811^ ilrr 1411 111*.(a)(b)Fig 31. Scanning electron micrographs of, (a) dendritic NaC1,(b) granular NaCl. Reproduced from reference [68].86(19) R. E. Aries, D. P. Lidiard and R. A. Spragg, Chem. Brit., 27(9), (1991),821-824.(20) J. W. Mullin, "Crystallisation", Butterworth & Co. Ltd., London, U.K., (1972),p. 136.(21) V. A. Garten and R. B. Head, Phil. Mae., 8, (1963), 1793-1803.(22) V. A. Garten and R. B. Head, Phil. Mae., 14(132), (1966), 1243-1253.(23) N. F. Oputshchennikov, Soy. Phirs.-Crystallogranhy, 22), (1962), 237-239.(24) D. P. Lal, R. E. Mason and R. F. Strickland-Constable, J. Crystal Growth, 5,(1968), 1-8.(25) A. N. Chernov, N. P. Zaitseva and L. N. Rashkovich, J. Crystal Growth, 102,(1990), 793-800.(26) R. F. Strickland-Constable, "Kinetics and Mechanism of Crystallization",Academic Press, U.K., (1968), p. 281.(27) J. W. Mullin, "Crystallisation", Butterworth & Co. Ltd., London, U.K., (1972),p.170.(28) idem, ibid, p.196.(29) E. G. Cooke, Krist. Tech, 1, (1966), 119.(30) P. R. Collins and W. J. Fredericks, J. Crystal Growth, 71, (1985), 739-743.(31) S. N. Black, R. J. Davey and M. Halcrow, J. Crystal Growth, 79, (1986),765-774.(32) R. J. Davey, J. Crystal Growth, 76, (1986), 637-644.(33) J. R. Bourne and R. J. Davey, J. Crystal Growth, 44, (1978), 613-614.(34) H. Svanoe, Chem. Ene. Proe., 55(5), (1959), 47-54.(35) L. Phoenix, Brit. Chem. Enz., 11(1), (1966), 34-38.(36) F. Grases and J. G. March, TrAC, Trends in Analytical Chem., 10(6), (1991),190-195.(37) J. W. Mullin and A. Amatavivadhana, J. Ana Chem., 17, (1967), 151-156.87(38) R. J. Davey, R. I. Ristic and B. Zizic, J. Crystal Growth, 47, (1979), 1-4.(39) R. C. Mattel and R. S. Feigelson, J. Crystal Growth, 97, (1989), 333-336.(40) B. Liang and R. W. Hartel, J. Crystal Growth, 108, (1991), 129-142.(41) G. A. Hussman, K. A. Berglund and M. A. Larson, ADDI. Spec., 39(3), 1985,560-562.(42) H. Takubo and H. Makita, J. Crystal Growth, 94, (1989), 469-474.(43) J. H. Bilgram and R. Steininger, J. Crystal Growth, 99, (1990), 30-37.(44) D. Cunnigham, R. J. Davey, K. J. Roberts, J. N. Sherwood and T. Shripathi,J. Crystal Growth, 99, (1990), 1065-1069.(45) W. M. Zuk and K. B. Ward, J. Crystal Growth, 110, (1991), 148-155.(46) S. Kozelak, D. Martin, J. Ng and A. McPherson, J. Crystal Growth, 110, (1991),177-181.(47) W. M. Rosenblum, J. P. Kennedy and B. Bishop, J. Crystal Growth, 110, (1991),171-176.(48) D.L. Portgal and E. Burstein, Phvs. Rev., 170(3), (1968), 673-678.(49) A. S. Pine, Phvs. Rev. B, 2L61, (1970), 2049-2054.(50) C. Lin, F. Tao, Z. Shen and W. Ma, Phys. Rev. Lett., 58(20), (1987),2095-2098.(51) D. Betteridge, M. T. Joslin and T. Lilley, Anal. Chem., 53, (1981), 1064-1073.(52) B. F. Munro, M.Sc. Thesis, University of British Columbia, (1991).(53) J. G. Bouchard, M. J. Beesley and J. A. Salkfeld, 3rd International Symposium onAnalytical Techniques for Process Control, Atlanta, April 1992, Abstract L59.(54) A. Galeski, L. Koenczoel, E. Piorkowska and E. Baer, Nature, 325, (1987),40-41.(55) A. Galeski, L. Koenczoel, E. Piorkowska and E. Baer, J. Polymer Sci. Part B,28, (1990), 1171-1186.88(56) J. Shen, D. Xu, R. Yao, R. Liu and D. Li, Chin. J. Polymer Sci., KU, (1990),69-74.(57) 0. Hunderi and K. Lonvik, Proc. 3rd Int. Conf. "Rapidly Quenched Metals",Brighton, U.K., 3-7th July 1978, B. Cantor (ed.), Volume 1, pp. 375-82,Met. Soc.: London, U.K., July 1978.(58) J. G. Deily, Micro-Notes, 6.(11, 1979.(59) J. E.Huheey, Inorganic Chemistry, 3rd edn., Harper and Row, New York, p. 56.(60) I. Tarjan and M. Matrai (eds.), "Laboratory Manual on Crystal Growth",Akademiai Kiado, Budapest, (1972), pp. 44-58.(61) B. P. Agarwal, Indian. J. Phys., 56A, (1982), 312-322.(62) G. W. Sears, J. Chem. Phys., 29(5), (1958), 1045-1048.(63) P. D. Wentzell, S. J. Vanslyke and A. P. Wade, TrAC, Trends  in Analytical Chem., an, (1990), 3-8.(64) A. A. Pollock, J. Acoustic Emission, Mb 1990, 140-141.(65) I. H. Brock, 0. Lee, K. A. Soulsbury, P. D. Wentzell, D. Sibbald and A. P. Wade,Chemom. Intell. Lab. Sys., 12, (1992), 271-290.(66) H. L. Keller, J. Less Comm. Met., 109, (1985), 19-29.(67) K. A. Soulsbury, A. P. Wade and D. Sibbald, Chemom. Intell. Lab. Sys., 15,(1992), 87.(68) R. J. Davey, M. M. Harding and R. J. Rule, J. Crystal Growth, 114, (1991),7-12.

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

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.831.1-0061775/manifest

Comment

Related Items