Use of Laser Diffraction Techniques for Alumina

Gerald Roach and Nicholas Pearson

Research and Development,

Alcoa of Australia Limited,

Cockburn Road, Kwinana, 6167

ABSTRACT

Laser diffraction particle sizing is widely used in the Bayer process, especially for quantifying the -20m m material in product alumina. There are many issues in the analysis such as what is actually measured, the errors in sub-sampling within the equipment and the algorithms. Data based on twelve years experience at Alcoa of Australia Limited using the second generation laser sizing equipment is presented. Sub-sampling within the equipment is not necessarily as accurate or as precise as that required for the alumina industry for –20 m m measurement.

Comparisons between dry and wet sizing indicated little difference in sizing provided due attention is paid to potential attrition during dry sample presentation. Advantage was made of the dry feeder’s propensity to attrite by using it as an attrition tester. Attrition index information could be obtained in a matter of minutes and for different levels of attrition.

The previously proposed use of the light intensity information to obtain fractals, and thereby surface morphology information (AQW Darwin, 1996), was tested using hydrates with extreme morphology differences. Surface morphology differences were not distinguishable using the proposed technique. However the use of fractals to obtain shape information, especially on the fines, is possible (the fines must be measured separately). Consequently, for –20 m m analysis, the potential exists of determining a conversion factor based on the fractal derived shape factor to convert the laser sizing data to equivalent wet sieve or spherical volume size information.

 

Use of Laser Diffraction Techniques for Alumina

Gerald Roach and Nicholas Pearson

1.0 INTRODUCTION

Laser diffraction has become the sizing analysis of choice in the Bayer industry for process control. In particular the new generation of laser sizers became capable of measuring the -20m m fraction in typical smelter grade alumina samples with the desired degree of accuracy and precision on the total sample (Roach and Scott, 1990). Since that time the quality of alumina in terms of fines content has improved with often the finest fraction (generally the electrostatic precipitator dust, ESP) no longer being included in product alumina. This has reduced the -20m m fraction from previous values of 3 to 5% to around the 1% mark. Significant benefits in terms of flow properties, reduction in segregation and dusting have resulted at the smelters and thus -20m m values of this level are becoming the desired level. The reduction to this lower level imposes an extra demand on the measurement equipment.

There have also been various developments in the laser particle sizing equipment, not the least being there are now literally tens of manufacturers whereas previously there were only a few. Alcoa has continued to monitor the development of such particle sizers and to assess whether they can meet the increased levels of sensitivity and precision now being sought. Some of the issues with the modern equipment will be discussed. Unfortunately not all the equipment is suitable. This is not always a fault of the equipment but results from the very specific requirement of our industry.

Prior to any discussion it is important to assess what information is being sought and what a laser particle analyser measures; that will be first discussed.

Another issue is whether dry or wet sizing is the appropriate measure for the alumina industry. Dry sizing (sieves) is the standard above 45m m whereas wet sizing is used below that level. In the cement industry dry sizing throughout the size range is the norm. As alumina is handled dry and it is the handling properties that are most affected by sizing, dry sizing at all sizes might be the most appropriate measure. Data for wet and dry sizing will be compared.

Another area that sizing is used is in attrition testing. A dry powder feeder has been modified such that rapid attrition measurements can be obtained with remarkably good precision.

At the last Alumina Quality Workshop Cleaver and Amal (1996) presented a paper indicating that laser sizers could give morphology information by determining fractal information from the light intensity data. Testwork to assess whether such information can be obtained was undertaken and is presented.

2.0 LASER PARTICLE SIZERS

2.1 Laser Particle Sizers - what do they measure

Laser particle sizers infer a volume distribution related to size from a light intensity spectrum related to the diffraction angle. Proprietary algorithms are used to convert the light intensity to this size distribution by assuming homogeneous spherical particles – ie the particles are assumed to have the same refractive index (important at the finest sizes) and all to have the same density. For example a hydrate particle and a typical alumina particle of the same size will give the same volume reading when analysed by a laser sizer but their weights are different as their respective bulk densities are different. This is a consequence of the alumina particles being porous. At fine sizes, < 5m m, where some of the laser light can be transmitted through the particle rather than diffracted by the particle, Mie theory rather than Fraunhoffer theory is frequently used. For such calculations the refractive index of the material is required. Often that for alumina is used which is that normally for alpha alumina (the one most readily available). Most of the particles are intermediate alumina phases and sometimes the fines can contain hydrate. The intermediate alumina phase particles are porous and the refractive index of a porous particle is totally different to that for a solid particle. Further the particles are not spherical but can range from plates through to strange shaped agglomerates, some of which can resemble barbels. These effects are most pronounced at the fine end of the measurement range that unfortunately is where laser sizing is most used for smelter grade alumina. For example if ESP dust is included in a product alumina such product contains hydrate, intermediate aluminas and alpha alumina. Particle shapes can vary from plates which are chips off larger crystals, agglomerate type shapes, prismatic shapes akin to single crystals and perfect spheres caused by continuous attrition in calcination. Hence the measured value on a weight basis is very sensitive to the nature of the fines. A change in the process to alter the shape, phase composition etc. will affect the result obtained. This is shown in Table 1 where 3% ESP dust ground and unground has been added to a fines free smelter grade alumina sample; different values are obtained. Also, the measured value is not the same as the true value because of the issues discussed above. Significantly greater differences have been obtained using a wider variety of fines. Because of all of these issues, standardisation is the key to laser sizing if comparisons are to be made, and this would be greatly assisted by standard samples relevant to the alumina industry.

Table 1

The effect of morphology on sizing.

Addition of 3wt% ESP dust to non-fines alumina

Fines type – 3wt% (all –20m m)

% -20m m measured

ESP standard

5.2

ESP ground

4.7

 

2.2 The Algorithm

Ideally all laser sizers should give the same size distribution for the same light intensity distribution pattern. The algorithms are one of the most protected parts of the equipment. There may be improvements in terms of the number of detectors or improvements in the detectors themselves such that the light intensity distribution is improved, however the algorithms should not be altered. Unfortunately in our experience that has not been the case. Running light intensity data from one model on a newer version of the same instrument has resulted in different results. For example a sample having 0.3% -20m m as measured on the old software gave a value of 1.0% with the new software. This was reported to result from use of a "very polydisperse" versus a "polydisperse" type model. Changing the model type certainly improved the situation but did not completely resolve the issue.

 

2.3 Repeatability

The repeatability of laser equipment has always been excellent. For smelter grade alumina samples it is the finest end of the distribution, representing 1 to 5% of the sample, which needs to be quantified. The precision that has been requested by the refineries for control is 0.2% which, for a 1% fines material, appears a relatively high 20%. However, as the total sample is being measured, the actual precision required is 0.2%. This is far greater than virtually any other standard analytical procedure such as chemical analyses, screening etc. With over eight years of running control charts the precision of laser equipment has generally been around 1.3% ( 3s limits) which in itself is excellent. The latest machines have further improved and are now close to attaining the desired 0.2%.

2.4 Accuracy

The major issue in accuracy is not related to the measurement of the particle size or the algorithms discussed above which as much relate to how particle size is defined. The issue is related to subsampling of the sample within the instrument. Laser sizers are designed to measure the size of a wide variety of materials with both a narrow and wide particle size distributions. For this they are excellent and describe probably greater than 95% of the size distribution both accurately and precisely for a wide range of materials. For smelter grade alumina only the finest few percent of the sample needs to be quantified. The first step for most laser sizers (especially those sizing in water) is to sub-sample. The initial sample is mixed, agitated in water (with or without dispersant and ultrasonics) and a subsample pumped through the light cell. All the possible errors associated with dispersion, attrition from agitation etc. will be ignored; generally nowadays they are small (they were quite significant for the old generation equipment with their rotary pumps etc. and a few sizers on the market do still give the particles quite a ‘bashing’, which for some industries is desired).

The major issue is the subsampling from the agitated bath. In Table 2 data are given for a smelter grade alumina sample measured with different sample presentation units on otherwise identical equipment from the same manufacturer. The introduction of a ‘new improved design’ sample presentation unit gave highly biased readings for the fines. The equipment had to be redesigned and it took two redesigns and over twelve months before unbiased subsampling was achieved such that the new laser equipment could be installed for routine analysis. Equipment from some other manufacturers has also proved to be completely unsuitable because of such sub-sampling issues. The message here is that the equipment had not been designed specifically for the alumina industries needs. The subsampling is acceptable for most industries where the whole distribution is being measured and generally described. However for the alumina industry there is a specific requirement. Hence the issue is not with the manufacturer but rather it is one where the industry as a whole needs to communicate its needs to the manufacturer (the alumina industry purchases relatively few laser sizers).

Table 2

Sizing data obtained from different sample presentation units on the same laser sizer

-20m m

-45m m

"old unit"

2.3

7.9

"new unit"

4.0

12.5

Redesigned unit

1.9

7.2

A similar issue arises for measurement at the coarse tail of the distribution.

 

2.5 Testing for Sub-sampling

Two methods were used to demonstrate the bias in the sub-sampling. First standard additions of fines can be made. This can be by standard addition of ideally spherical particles to minimise other effects. However, in practice it is sufficient to use an easy to disperse material. The second method is to repeat the measurements continuously but allowing the recycle stream to be discharged and topping up with water. The obscuration decreases as progressively less sample is analysed on each cycle, however the size distribution should remain the same (This is akin to repeatedly taking samples of alumina out of a jar. Normally it is only in the last few spoonfuls that the bias becomes obvious with the sample invariably getting finer.).

Testing of the newest presentation unit via these various methods has shown it to be effectively free from sub-sampling bias. Also the repeatability of the measurement is much improved over the previous models (and it is more user-friendly).

2.6 Wet versus Dry Sizing

Alumina is used in the dry state and the sizing control on alumina is nowadays primarily related to its handling characteristics and dusting potential. Alumina when handled can generate high electrostatic charge. When coarse alumina particles are examined, fine alumina particles can be observed ‘clinging’ to the particle surface. Such particles are readily washed off when the particles are immersed in water. Such fines will contribute to the measured -20m m and may influence the handling properties in a way different to free fines (in some analyses they can also be double accounted). A dry particle feeder was used to compare dry and wet sizing data. The initial results are shown in Table 3. The dry sizing was significantly finer than the wet sizing.

Measurement of the product from the dry sizer (a once through system) by the wet sizer gave the same results as the dry sized material. This indicated that not only was attrition occurring in the dry sample feeder, but also that the wet and dry sizers gave the same result as there is no attrition after the light cell on the dry sizer. Following modification to the feed system to minimise attrition, the dry sizer gave identical results to the wet sizer, see Table 3, suggesting that fines clinging to coarse particles represented little of the fines. That was verified by dry screening at 75m m and washing the +75m m product. The mass of fines washed off contributed only 0.1wt% of the total sample.

Table 3

Comparison of dry and wet sizing data

-20m m

-45m m

Wet sizing

3.4

11.9

Initial dry sizing

18.2

32.3

Dry sizing following modification

3.5

12.0

3.0 ATTRITION TESTER

The attrition noted in the dry sizer suggested that it could possibly be used as an attrition tester. To obtain readings similar to those for wet sizing the air flow had been reduced to below the minimum recommended value (some physical changes to the equipment had to be made). The dry feeder is primarily used in the cement and pharmaceutical industries where the powder product is to be physically dispersed. Consequently there is a need for the fairly severe pre-analysis treatment. Sizing results for the –20m m size fraction are presented in Figure 1 at various air flow rates. Increasing the flow rate dramatically increases the level of breakdown. A sample can be run and a sizing obtained at the minimum air flow. The air flow is then increased and a new reading taken and subsequently an attrition value can be calculated. That can be done in a matter of minutes. The product can be collected and the nature of the fines examined - they are not too dissimilar to the breakdown product seen in fluid bed calciners.

Image273b.GIF (2411 bytes)

Figure 1

Weight % -20m m versus air flow for a dry powder feeder

4.0 MORPHOLOGY ANALYSIS OF HYDRATE UTILISNG FRACTALS

At the AQW in Darwin a technique was reported (Cleaver and Amal, 1996) for quantifying the structure of hydrate using the light intensity data from laser light scattering equipment. The light energies are converted to intensities and the log of the intensity is plotted against the log of the scattering angle. The absolute value of the slope of the linear portion of the plot is the fractal dimension of the sample. The fractal dimension is a measure of the degree of occupation of space by an object. It was claimed that the fractal dimension could be related to other properties of the material such as attrition index. The work reported by Cleaver and Amal was for unsized material. This was surprising as the information at the high angle dispersions where fractal information is obtained relates to the surface morphology (protrusions) of the coarser particles, the amount of fine material present and the shape of the fine material.

An unanswered question had always been whether laser light scattering from a rough particle gave both a signal related to the main bulk of the sample and also to minor protrusions from the surface, the latter being equivalent to fine particulates attached to the surface. To answer this, and also to determine whether the work of Cleaver and Amal had any veracity, sized hydrate samples with quite different morphologies were analysed. The three different morphologies examined are shown in Figure 2; clearly they are extreme examples in terms of morphology. Also various sized fractions from a typical calciner feed sample were analysed. Plots of Log Intensity versus Log Angle are shown in Figures 3 and 4 for the two suites of samples. (Malvern Instruments supplied the necessary information to enable such plots to be calculated.) The slope of the relevant portion of the plots are almost identical for each sample despite the significantly different structures. The fractal dimension is approximately two for all samples (|slope| approximately = |-2| =2). Vertical offsets relate to the amount of material in the measurement and do not affect the slope. Only the finest fraction of the calciner feed sample showed ‘good’ fractal behaviour. Differences of the fractal dimension would have been expected with size as the particles become more spherical at the coarser size fractions.

 

(a) "Spiky" hydrate

(b) "Chunky" hydrate

(c) "Round" hydrate

Figure 2

Samples of hydrate with widely differing morphologies, used to appraise the determination of fractal analysis

Figure 3

Log Intensity versus Log Angle for hydrates shown in Figure 2

 

Figure 4

Log Intensity versus Log Angle for sized fractions from typical calciner feed

Taking as an example the particles in Figure 2b (chunky), the protrusions represent less than 8% of the total volume. The proportion of that volume which would end up giving recorded intensity signals at high dispersion angles would be less than a tenth of this. Hence, even for sized particles, there would be little light intensity recorded which would relate to the "surface roughness" of the particle. For unsized particles, as in the original work, there would be the contribution to the light intensity from fines that needs to be removed prior to teasing out such information. If such fines have a different morphology then that is not readily possible. Also the algorithms assume spherical particles; if the particles deviate far from spherical the fractal dimension will relate to both shape and surface texture. Possibly the reason for the correlations found in the original paper was that the fractal dimension was primarily describing the shape of the particles. Agglomerated particles tend to be less spherical than growth type particles and also their attrition behaviour is different.

The idea of utilising the light intensity signal to obtain useful morphology information is certainly a novel one, but it would appear that improved resolution of intensity data is required together with better models before such information can be reliably obtained. Ideally information on both protrusions and the general shape of the particles could be obtained. The greatest, and perhaps the most significant, potential is that of obtaining fractal information from the sizing data related to shape for the fines fraction. The finest fraction of the calciner feed sample showed good fractal behaviour and a reliable fractal dimension was obtained. Even though there are a variety of particle shapes in such fines, the fractal dimension will give a composite value that could be used for conversion of the laser sizing to other size parameters. For a smelter grade alumina sample, by screening and sizing the fines (say -45m m), their fractal dimension could be obtained. A conversion factor based on that could be calculated which would enable the laser -20m m value (obtained on the total sample) to be converted to either a wet screen (second longest dimension) or spherical particle equivalent. This could overcome many of the issues related to -20m m measurement discussed in Section 2.1.

5.0 CONCLUSIONS

Laser particle sizers have not been specifically tailored to the needs of the alumina industry. Consequently, when measuring parameters such as the –20m m fraction, it is necessary to ensure that there are no potential biases introduced from the equipment, especially in sub-sampling.

There is no significant difference between wet and dry particle sizing.

Dry feeders can be used as a quick attrition tester.

Surface morphology analysis via fractal dimensions is not possible with the current equipment even for sized samples. However the potential exists to obtain shape information via such fractals.

With such fractal based shape information on fines, the -20m m information could be converted to either a screen equivalent or spherical volume value thus leading to the potential for improved quantitative analysis of the -20m m fraction.

ACKNOWLEDGMENTS

The assistance from Maurice Wedd of Malvern Instruments Limited in supplying the ‘magic numbers’ for the fractal work and for general technical advice over the years is gratefully acknowledged. Mr Sean Collier (Alcoa) was responsible for the installation, control work and generating much of the data on the laser equipment.

REFERENCES

Roach G.I.D and Scott M. 1990 "Laser Sizing in the Alumina Industry", Alumina Quality Workshop, Perth.

Cleaver J. and Amal R. 1996 "Fractal Analysis of Alumina Trihydrate" Alumina Quality Workshop, Darwin.