Energy & Fuels 1992,6, 743-752
743
An Investigation into the Process of Centrifugal Sink/Float Separation of Micronized Coals. 2. Multiple Fractionation of Single Coal Samples Gary R. Dyrkacz' and Ljiljana Ruscic Chemistry Division, Argonne National Laboratory, 9700 South Cass Avenue, Argonne, Illinois 60439 Received March 5, 1992. Revised Manuscript Received July 8, 1992
Multiple sink/float density fractionations were done on three different coal samples. Density gradient centrifugation methods were used to monitor the results of the separations. The purity of a float or sink phase was directly tied to the density distribution of the coal. The size of the density interval chosen for fractionation was shown to also be important in determining the purity. Sink fractions were consistently more impure than the float material. The main reason appeared to be unavoidable mixing of float material into the sink phase during the centrifugation cycle. However, with appropriate handling very pure maceral fractions were obtained in many cases.
Introduction In part 1of this series we explored some of the protocols needed for successful centrifugal sink/float (S/F) separations of finely ground coal maceral particles.' From simpletheoreticalconsiderations,we found that separation purity should be dependent on the density distribution of the original coal. The experimental findings supported the calculations, but there was an additional unexpected contamination mechanism. To some extent, float-phase particles mixed back into the solution, due to either changing forces or thermal variations during the centrifugation process. Several solvent media were also investigated for their separating ability; two were excellent for coal separations: aqueous CsCl/Brij-35 and aqueous Ca(NO&/Brij-35. The previous results were based on single centrifugal S/F density cuts on the parent coal. However, we often wish to examine the chemical and physical interrelationships of intimately associated macerals within nearly the same coal-formingenvironment. Thus, in a more realistic maceral separation,a single coal sample would be subjected to a series of density cuts to produce a series of fractions. In addition, multiple fractionation should be an especially important consideration in light of recent work using density gradient separation techniques. Such separations show that broad variations in density exist for nearly every coal and even extend to the macerals within a single maceral group. This density variability is also an indicator of chemical The work reported here examines the multiple S/F fractionation of single coals. The methods used in this work are a direct application of what was found in part 1. However, we have shown using continuous flow (l)Dyrkacz, G. R.; Ruecic, L. R.; Fredericks, J. R. Energy Fuels, preceding paper in this h u e . (2) Dyrkacz, G. R.; Bloomquist, C. A. A.; Ruacic, L. Fuel 1984, 63,
---- --.-.
1 1M-11'7X
(3) Karas, J.; Pugmire, R. J.; Woolfenden, W. R.; Grant, D. M.; Blair, S.Int. J. Coal Geoi. 1986,5, 315-338. (4) Crelling, J. C. Roc., 1988 Ironmaking Conf. 1988,47, 351-356. (5) Taulbee, D.; Poe, S. H.; Robl, T.; Keogh, B. Energy Fuels 1989,3, 662-670. (6) Dyrkacz, G. R.; Bloomquist, C. A. A.; Ruecic, L.; Crelling, J. C. Energy Fuels 1991,5,155-163.
0887-0624/92/2506-0743w3.00l0
centrifugation separation of coals that multiple density fractionation places additional constraints on the purity of separated phases.' We will show that the same factors must be considered in simple S/F centrifugal separations as well. Experimental Section The experimental procedures used for the sinklfloat separations have been detailed in part 1of this series.' There were no significant deviations from these routines. Fifty-milliliter nonscrew-capcentrifuge tubes were used for all the separations. The centrifugeused was a B e c k " 52-21 refrigerated centrifuge with a JS 13.1swinging bucket rotor. Except where noted, all the S/F separations were centrifuged for 1 h at speed. A deceleration program was used to stop the rotor in 30 min from lo00 rpm. At the start of each separation, the coal was added to a 42-45-mL solution at the appropriate density. The solution was then ultrasonically treated to disperse the coal material. For intermediate separations, the phase of the coal was adjusted to the appropriate new density, and solution of the same density added to bring the total volume to -45 mL. S/F phases were analyzed by analytical density gradient centrifugation (ADGC). Because of concerns about extremely fine particles in some phases, several ADGC separations were centrifugedfor longer periods (4-6 h) than the typical 1h. Within experimental error, all the ADGC separations were the same. Purities of the phases were determined by integration of the correspondingADGC patterns. The float-phase integration was done on the normalized density distributions, from the first data point to the nominal solution density point. This value was then divided by the entire peak area to find the percent purity. The sink phase was handled similarly. In a few separation cases reported in this work, the yield and purity data produce composite overall yields which are greater than possible. We believe this is due to random errors in both the purity and weight (yield) determination procedures.
General S/F Separations There are four mechanisms that we have identified so far in centrifugal sink/flat separations that have a direct impact on the purity of separated phases: (1)the density chosen for a separation relative to the density distribution (7) Dyrkacz, G. R.; Bloomquist, C. A. A. Energy Fuels 1992,6,357374.
0 1992 American Chemical Society
Dyrkacz and Ruscic
744 Energy & Fuels, Vol. 6, No. 6,1992
of the coal sample; (2) direct mixing of float-phase material into the sink phase due to either mechanical or thermal problems; (3) the size of the density interval between the successive fractions; (4) aggregation of the particles in the media used for separation. From part 1 of this series, neither aqueous Ca(NOd2 nor aqueous CsCl in the presence of Brij-35 appears to induce much, if any, coal agglomeration. Because these will be the only media used for the separations described here, the fourth mechanism will not be considered. The first mechanism limiting purity was dealt with in part 1.1 The dependency of the purity upon density is a direct consequence of all particles not having sufficient time to reach their respective phases. Once the mechanical parameters of the separation, such as separation time and rotor speed are chosen, then only the particle size distribution of the coal should dictate the separation efficiency. This efficiency, together with the density distribution of the coal, will dictate the ultimate purity of a sink or float phase. In part 1 we used both the experimental particle size distributions and the density distributions for each coal to calculate the expected purity for any density cut. We found that the experimental data was in qualitative, but not quantitative, correspondence with the calculations. There were two additional contamination processes identified in part 1that could affect the separations. Both processes resulted in contamination of the sink phase by float-phase material. The first was the obvious case where float material inadvertently contaminated the sink phase during removal of the float phase. Although such contamination did occur in a few cases, careful handling of the centrifuge tubes minimized this problem. A second contamination mechanism was inferred from the behavior of polystyrene latex particles under separating conditions similar to that used for the coal samples. In this case, contamination occurred because of changing centrifugal forces and/or thermal variations during the deceleration phase of the separation. With simple centrifuge tubes, it was not possible to overcome these problems in our centrifuge system. As we suggested in part 1,the amount of contamination from phase mixing is not based solely on the original coal distribution but may be a function of particle size and density distribution. What is clear from all our previous experiments is that there can be substantial contamination of the sink phase by float material, beyond that predicted from simple dynamical considerations. The third mechanism that controls purity is the density interval between successive cuts. We first recognized this limit on purity in our investigations of the continuous flow separation of ma~erals.~ Figure 1shows schematically how narrowing the density separation interval of a coal can lead to low purities of material. Even if the amount of contamination is constant, as the separation interval decreases we must expect that the purities will be lower. Because each intermediate density fraction results from two individual separations, the effect of this problem can be particularly pronounced at some densities. Figure 2 presents the effect of density interval on purity for the real case of APCS 7 coal. To calculate the displayed data, the experimental coal density distribution was divided up into a series of constant width density intervals. The calculated separation efficiency curves from Appendix A in part 1 were used to obtain the expected density distribution of each fraction. Appropriate integration of
A
:
10%
ri
i
Cont.
m
;
i,'
> t
s
10% Cont.
j
80%
80%
b
Pure A
, Density
Density
+
+
Figure 1. Idealized densitydistributionshowing how the density fraction intervalwill affect the purity of a separation. The dashed lines bracket the constant amount of contamination in both cases. (A) is the desired phase.
L 0.8
"." , 1.1
0.0 1.2
1.3
1.4
1.5
Density (g c m 3 )
Figure 2. Calculation of the purity at various fractionation intervals for APCS 7. The numbers in the legend represent the density size interval of the appropriatecurve in g ~ m - ~The . data
used for this calculation can be found in ref 1: the experimental particle size data were from ref 1, Figure 2, and the calculated purities are from ref 1, Figure 23. Starting density was 1.10 g ~ m - Each ~ . horizontalline represents the density fractionwidth. The data were calculatedassuming 50 % of the liquid was removed as float phase. The centrifuge conditions were assumed to be 10 OOO rpm for 1 h. each fraction's density curve then provided the purity of the fraction. Two important conclusions can be derived from data such as is shown in Figure 2. First, the purity curves for multiple fractionations somewhat mirror the density distribution of the coal. This meam that the higher the concentration of a particular maceral or maceral group in the original coal, the higher the achievable purity will be for that maceral. Second, the narrower the density interval for fractionation, the lower will be the purity of the final fractions. Even for the idealized separation cases displayed in Figure 2, the data indicate we should expect substantial reduction in purity for density intervals even as large as 0.1 g cm-3. The overall behavior of purity with fractionation is easy to understand. Every fraction, except the first and last, is the product of two separations, each with its own purity limit. For separations near a large band, there is a proportionally greater concentration of particles from that
Energy & Fuels, Vol. 6, No. 6, 1992 748
Separation of Micronized Coals. 2 Table I. Elemental Analyses (wt %, daf) of Coals Used in This Study coal C H N Ob Sbt asha 1.5 14.2 0.82 18.8 APCS 7 78.6 4.9 0.35 1.5 12.6 0.82 APCS 7 (demin) 80.1 5.0 80.8 4.79 1.4 11.6 0.85 16.5 PSOC-732 9.7 0.74 0.33 1.4 PSOC-732 (demin) 83.4 4.7 7.22 1.3 15.9 3.4 APCS 3 74.3 5.1 3.87 1.5 11.8 5.1 APCS3 (demin) 76.2 5.4 By difference. On a dry basis. Table 11. coal Sp. APCS7 10.2 PSOC-732 10.2 APCS3 3.5
Maceral Analyses (vol %) of Coals. Cut. Res. Bit. Vit. S-Fus. Fus. Ind. 0.3 1.0 0.0 72.8 10.9 1.9 1.7 0.0 0.4 1.5 50.4 27.9 2.8 3.7 0.0 0.6 0.4 90.4 2.8 1.0 1.1
Mi. 1.1
3.1 0.2
Sp = sporinite;Cut. = cutinite;Res. = resinite;Bit. = bituminite; Vit. = vitrinite; S-Fus.= Semi-fusinite; Fus. = fusinite, Ind. = inertodetrinite; Mi. = micrinite. Analyses done in both blue light and white light. Two samples, 500 counta each. a
band which will contaminate the intended fraction. The contamination occurs because the undesirable particles are either too close to the solution density or too small and, therefore, move more slowly. These particles will not all be cleared from the desired phase. Thus, fractionations near large bands will generally have lower purities, as displayed in Figure 2.
Results Coal Samples and General Separation. Analytical data for the three coals used in this study are shown in Tables I and 11. These coals were the same as used in the previous study. APCS 7 and APCS 3 were obtained from the Argonne Premium Coal Sample Program. APCS 7 coal is from the Lewiston-Stockton Seam in West Virginia, and APCS 3 coal is from the Illinois No. 6 Seamas PSOC732,from the Upper Kittanning Seam in West Virginia, was obtained from the Pennsylvania State University Coal Bank. All three coals are of high-volatile bituminous rank. The coal samples were ground using a fluid energy mill and chemically demineralized according to established procedures.9 The detailed maceral density distributions for the three demineralized coal samples were presented earlier in part 1. Bothaqueous CsCVBrij-35 and Ca(NO&/Brij-35media were used for the S/F maceral separations of the different coals. Fifty percent of the upper phase was removed as the float phase, though this is probably not the optimum split for obtaining pure sink-phasematerials. In our earlier work, we found that there was a likelihood for a portion of the float phase to mix into the sink phase, causing lower than anticipated purities. However, the separations need not be optimum to understand the process. In fact, if the separations were run under very efficient conditions, we would not have been able to accurately observe changes in purity. Even with the conditions used here, at some densities, the purities were high enough that the expected variations with density were within experimental error. Multiple Fractionation of APCS 7. This coal was fractionated at four different densities. Figure 3 shows the ADGC density gradient distribution of the starting coal and where the density cuts where made. Two (8) Vorree, K.S.Energy Fuels 1990,4, 420-426. (9) Dyrkacz, G. R.; Horwitz, E. P. Fuel 1982, 61, 3-12.
(10)Dyrkacz, G. R.; Bloomquist, C. A. A. Energy Fuels 1992,6,374-
386.
1.075
1.175
1.275
Density
(g
1.375
1.475
~ m - ~ )
Figure 3. Analytical density gradient separation of APCS 7 showing the position of the density cuts used for the S/F separations. ~1.240 + 1.330
I= 1.240
1.270
33.8
92.9
Figure 4. Results of L:H multiple S/F separation of APCS 7. The left number in each box is the float-phase purity after the S/F separation; the right number is the corresponding sink-phase purity. For recycles, the top number in a box is the float-phase purity, and the bottom number the sink-phase purity. Densities are displayed between the boxes. [Coal] = 50g/L; rpm = 10 OOO; centrifugation time = 1 h.
separation sequences were done: The first started at the lowest desired density and proceeded to higher density fractionations (L:H sequence). The second started at the highest desired density and proceeded to lower density fractionations (H:L sequence). In the L H separation sequence, the sink fraction from the preceding S/F separation became the feed for the next S/F cycle. Figure 4 shows the progress of the separation in terms of the overall purity of the float or sink phase. The data in the horizontal elongated boxes represent the primary separations, while the vertically elongated boxes represent recycles. Leftmost or topmost (for recycles) areas within each box represent float data. The right or bottom area in each box represents the corresponding sinkphase purities. All the purities were determined by integrating the ADGC data for each separated phase. Arrows track which fraction was used for either a subsequent primary separation or a recycle separation. Table I11provides a summary of each fractionation’s purity and yield. The yields for the present work serve mainly as cross-check on the purity information. The float phases for all the fractionations were quite pure. However, as discussed in part 1, high float-phase purities would be expected for some separations, even if there was no separation at all. The initial purities can be seen from the data in Table IV, which shows the apportionment of material at the separation densities, before
Dyrkacz and Ruscic
746 Energy & Fuela, Vol. 6, No. 6, 1992 Table 111. APCS 7 Separation Results in CsCl/Brij-36 float phase sink phase density (g ~ m - ~ ) %pure %yield %pure %yield low to high density (LH) 1.240 1.240RFb 1.270 1.310 1.310RF 1.330 1.3301.330RF 1.330RFRPP 1.31ORFRFc high to low density (HL) 1.330 1.310 1.330RSf 1.270 1.240 1.240RFd
89.3 89.4 96.4 97.5 95.0 92.8 79.9 92.7 94.3 96.0
7.9 75.6 19.6 70.8 98.5 11.0 40.8 61.3 89.2 49.5
92.8 39.4 59.2 83.3 20.1 78.0 92.9 33.8 51.2 64.7
75.4 24.4 68.0 29.2 1.5 89.0 59.2 38.7 10.8
97.1 90.1 98.3 95.7 91.5 98.9
53.7 95.2 74.2 30.5 26.7 89.8
86.5 43.0 43.4 43.3 84.9 77.7
12.1 4.8 25.8 69.5 73.3 10.1
50.5
a Starting concentration of coal was 50 g/L. * RF = recycle of previous float phase at specified density; Rs = recycle of previous sink phase. RFRF = recycled float phase of the f i t recycled float phase. 6 h centrifugation. e This ia a separation at 1.310 g cm-3 of (previous the float material from the second recycle at 1.330 g sink fraction was recycled at 1.330 g entry). f The 1.310 g
Table IV. Percent of Float and Sink Fraction Expected Prom APCS 7 at Various Densities theor purity (wt % ) density amt of phase expected (wt % ) (g cm-3) float phase sink phase float phase sink p h 90.4 88.0 99.8 9.6 1.240 96.2 98.0 43.5 56.5 1.270 99.6 80.0 20.0 96.5 1.310 99.7 97.1 83.8 16.2 1.330 ~~~
~~
p=1.240-1.270
~
a This ia the purity of each phase calculated by the method in part 1 for an ideal separation system with 1 h of centrifugation at 10 OOO rpm.
any separation. For example, the sink material for the 1.240 g cm-3 separation starts at 90.4% pure. Similarly, the float phases to the high-density (inertinite) side of the main band all start with a high purity. In all cases, when we compare this baseline data with the data in Table 111, the float materials have definitely been enriched. In part 1,we developed a separation efficiency index to circumvent this varying baseline purity problem. However, in the present case of multiple separations, we will not present the data this way. Many fractions are the product of more than one separation. This multiplicity complicates the concept of the separation index, because the nature of a particular separation will depend on previous separations. The separation efficiency would have to be based on the density distribution of the previous sink or float phase, and not the original coal. Even though the data were available to generate the separation efficiencies, we do not believe it would be useful in the present case. Our purpose here is not strictly comparisons of different coals or different separation conditions. Table IV also shows the expected purity of the fractionations, if each was based on single-densitycuts on the starting coal. The expected purities were derived by the calculation method used in part 1. On comparison of the calculated float-phase purity data with the experimental float-phase purity data in Table 111,the phases show fair agreement. Not only is the expected drop in purity observed in the lowest density fraction, but even the magnitudes are similar. Admittedly, caution needs to be
1
1.2
1.3
1.4
5
Density (g ~ m - ~ )
Figure 5. Density distributions of the final fractions obtained from L H multiple separation of APCS 7. The vertical lines represent the poaitions of the cut densities. Table V. Final Purity Data for Separated Density Fractions of APCS 7. density (g cm-3)
a.240 1.240-1.270 1.270-1.310 1.310-1.330 >1.33
separation sequence (g cm-3) % pure 1.24 to 1.33 (LH) 1.33 to 1.24 (HL) (theor) 89.3 98.9 88.0 81.3 74.2 89.5 47.6 39.4 92.4 58.5 27.4 63.0 78.0 86.5 97.1
a The purity data for the intermediate fractions is derived from two integrations: one at the nominal cut density and a second at the previous cut density. The purity data for the previous cut is not displayed here.
used in these comparisons, because some fractions are the products of previous separations. Comparing the sets of sink-phase data, we find very little agreement with the calculated information. The lack of correlation occurs because float-phase material mixes into the sink phase during centrifugation, as discussed earlier. As indicated in the General S/F Separations section, we do not know how to calculate for this mechanical mixing. Nevertheless, all the sink phases are still substantially enriched by the separations. Each fraction other than the first and last is the product of two separations. The purity of each intermediata fraction must therefore be lower than any of the single separation purities. Figure 5 and Table V show the final purities of the desired fractions for the L:H density separation sequence; the purities of many fractions are indeed low. Since the separations were not all done at uniform density intervals, Table V also provides the maximum expected fraction purities based solely on the theoretically derived curves. This data was generated in a manner similar to that shown in Figure 2. Except for the first two low-density fractions, the L:H results do not match wellwith theory. The reason is that the high-density sink-phase purities are quite low in the individual separations due to the phase mixing phenomenon. For example, in Figure 5 the fraction distribution in the density
Separation of Micronized Coals. 2
Energy & Fuels, Vol. 6, No.6,1992 747
p = 1.330
+
1.240
JAPCS7J
11.330 1.310
Figure 6. Results of H:Lmultiple S/F separation of APCS 7. See Figure 4 for details.
interval 1.310-1.330 g ~ m shows - ~ a large amount of material to the low-density side of the 1.310 g cm-3 cut. This material was derived from the sink fraction at the 1.310 g ~ m separation. - ~ (Figure 5 is also a good example of the utility of ADGC for this type of work. The source and type of contamination is immediately clear.) As an alternate sequence we also examined the separation starting at the high density of 1.330 g ~ m and - ~ fractionated at lower densities ( H L series). This separation scheme and the corresponding phase purities are displayed in Figure 6. Here the float phase of the preceding separation was used for the next density cut. Data for the individual fractions is reported in Table 111. When we compare the float-phase data for the low- to high-density separation and high- to low-density separation, there are no significant differences. The sink-phase purities also seem to match quite well. However, the sink-phase data are more scattered than for the float phases due to mixing of float-phase material into the sink phase. There is one exception in the comparisons: The sink purities at 1.31 g cm-3 are quite different. We suspect that the very low sink-phase purity at 1.310 g cm", in the H L separation, may have been due to mechanical contamination during removal of the float phase from the tube. The final fraction purities for the H L separations are shown in Figure 7 and Table V. Note that there is a fair correspondence between the fractions from Figures 5 and 7. The final purities for both the H:L and L H separation sequences show similar general patterns, but some of the differences are outside the expected ADGC analysis error (-5% ). However, as already indicated, the differences were primarily due to variations in the sink phases. We conclude that even though the data are scattered due to the nonpredictable float-phase contamination of the sink phase, there appears to be no clear advantage using a H L or L:H separation sequence. When both separation sequences are compared to the theoretically predicted purity values in Table V, there is a fair, but not a close, correlation. Particularly, the results for the 1.270-1.310 g cmm3fraction diverge from the expected purity. We have already suggested that the mechanical float contamination of the sink phase is probably density dependent, with more float contamination occurring when the density cut is just to the high-density side of a large band. We surmise that this is what causes the abnormally low purity of this fraction. In spite of these masking problems, we believe there is sufficient similarity in the experimental and theoretical data to suggest our simple separation model is on the correct track.
I
1.1
p-1.270-1.240
1.2
1.3
1.4
1
1.6
Density (g ~ m - ~ )
Figure 7. Density distributions of the final fractions obtained from H L multiple separation of APCS 7. The vertical lies represent the positions of the cut densities.
Several fractions were recycled in each sequence to see if any improvement could be made in the purities. In part 1 we discussed how recycling can improve the purity through two factors: the first is the obvious removal of large contaminating particles due to any mechanical contamination processes; the second factor is a dilution effect. Fine particles, or particles that are very close to the solution density, have very low velocities and will act as nearly a homogeneous solute. The purity improves because the large particles of the desired phase move rapidly to their respective phase, leaving behind the slower particles. We then remove the undesirable phase. However, half the undesirable material will remain in the desirable phase because the particles move too slowly. Thus, the operation dilutes the undesirable materials. In cases where the float phases were recycled,differences in purity were only noted when the S/F separation time was increased to longer times, e.g., 6 h. Even then, the initial high purities of the float phases were so high that the changes were small and usually within experimental error. Only one sink phase was recycled. The 1.330 g cm-3 sink phase in the L H sequence showed marked improvement upon recycling, going from 78 to 93 % purity. Such pronounced changes in purity are to be expected, if the proposed mechanisms for contamination of the sink phase are correct. We also tried reseparatingrecycled materiale at densities different from the original recycle (the 1.310RFRF entry in Table 111). The resulta show that there is no inherent difficulty in any kind of recycling. However, in any recycling, it is advisable to spin the samples for longer periods of time or at higher centrifugal speeds, to substantially improve the quality. Multiple Fractionationof PSOC-732.This coal was used extensively in part 1in the development of our S/F centrifugal separation techniques. In the present work, limited use was made of this coal. Three densities were used to fractionate this coal: 1.240, 1.295, and 1.330 g
Dyrkact and Ruscic
748 Energy & Fuels, Vol. 6, No.6, 1992 1.240
Table VI. Percent of Float and Sink Expected from PSOC-732 at Various Densities
1.2,96 l.3,30 I
,
0
,
density (g cm-9
1.240 1.295 1.330
0.3
1
1
o.n ,
1.1
1.3
1.2
1.4
1.6
Denslty (g ~ m - ~ ) Figure 8. Analytical density gradient separation of PSOC-732 showing the position of the density cuts used for the S/F separations. p = 1.24
+
1.33
wt%
float phase 8.3 53.4 71.5
sink phase 91.7 46.6 28.5
theor purity (wt % ) float phase sink phase 89.0 99.9 98.3 98.4 99.0 98.2
Table VII. Multiple Fractionation of PSOC-732 float phase sink phase density cut %pure wt% %pure wt% Ca(NO~)dBrij-35~ low to high (LH)b 1.240 93.0 4.5 98.2 85.0 1.294 86.1 17.6 86.3 62.8 1.294RFc 89.0 88.5 52.9 11.5 1.330 96.9 33.3 78.3 45.7 high to low (H:L)b 98.3 55.0 70.7 45.0 1.330 1.296 92.2 43.8 50.7 56.2 C~Cl/Brij-35~ low to high (L:H) 85.9 7.0 98.8 1.240 93.5 1.296 88.7 23.0 85.0 63.6 1.330 93.3 24.4 92.6 45.4
*
a All ADGC samples were dried before analysis. [Coal] = 10 g/L. RF = recycle of previous float phase at specified density. ADGC samples were run from a wet condition. [Coal] = 50 g/L. S/F centrifuge conditions: 10 OOO rpm for 1 h.
PSOC-732
Figure 9. Results of L H and H:Lmultiple S/Fseparations of PSOC-732. [Coal] = 60 g/L; rpm = 10 OOO; centrifugation time = 1 h. See Figure 4 for key. cm-3. Figure 8 shows the position of these densities relative to the density distribution of this coal. The low-density separation mimics a liptinite separation, and the highdensity separation mimics an inertinite separation. In addition, the separations were done in two media: aqueous CsCl/Brij-35 and aqueous Ca(NO&/Brij-35. The latter medium was shown previously in part 1 to be just as effective as the CsCl/Brij-35 combination for separations . this point the viscosity rises less than 1.4 g ~ m - ~(Above rapidly.) Figure 9 schematically displays the course of maceral fractionation in Ca(N03)2/Brij-35. As in the previous coal, the separations were again run in low to high (L:H) and high to low (H:L) sequences. Table VI shows the relative percent of mass expected at each density cut, if single cuts are made. The calculated single-cut purities are also given in Table VI. Note that the expected fraction purities for the sink phases are higher than for APCS 7 coal (Table IV). This occurs because of the larger amount of inertinite material present in PSOC-732. All other factors being roughly equal, the relative proportion of float contamination in the sink phases must then be less than in APCS 7 coal. The experimental data for all the separations are shown in Table VII.
The individual density cute using Ca(N03)2/Brij-35in the L:H sequence show very high purities for all the floatphase separations. As with the previous coal, the sinkphase purities are more variable and show a lower purity at higher densities. However, for this coal, neither the float data nor the sink data suggest whether the predicted purities in Table VI are sensible. Besides the main sequence resulte, the float phase from the 1.295g ~ mwas - ~recycled at the same density and with the same centrifuge conditions (Figure 9). Some improvement in the float-phase purity was observed, but as in the previous coal, the change is not large. Partly, this is due to the fact we did not recycle under more severe centrifugation conditions. Also, the original material was quite pure to begin with. In the H:L sequence, only the two high-density fractionations were done. The data for this sequence (Table VII) compares quite well with the L:H data. Separation data using CsCl/Brij-35 as the separation medium are also displayed in Table VII. This was an L:H sequence performed at the same densities displayed in Figure 9, but without any recycling. The float-phase data in this case show behavior more like what we theoretically expected in terms of the pattern of separation. Still, the magnitude of the purity is typically lower than predicted. The sink data again suggest that the separations are somewhat decoupled from simple Stokes law dynamics. The density distribution of the final fractions, and their corresponding purities, are given in Figures 10 and 11. Table VI11 contains the summary of the displayed data. Comparing the theoretical with the experimental data, it seems that the pattern of separation is consistent. In fact, the data here are more closely allied with the predictions than those found for APCS 7. One likely reason for this agreement is the higher concentration of high-density inertinite material in this coal. Also note that the Ca(N03)2/Brij-35 data, except for the first fraction, are lower than the CsCl/Brij-35 data.
Energy & Fuels, Vol. 6, No.6,1992 749
Separation of Micronized Coals. 2 3.0
3
1.284 1.309
1.0
0.5
:
0.0 - - , 1.16
~4.294-1.330
1.25
1.35
145
Density (g ~ m - ~ )
Figure 12. Analytical density gradient separation of APCS 3 showing the position of the density cuts used for the S/F separations.
1.1
1.2
1.3
1.4
1.5
Density (g ~ m - ~ )
Figure 10. Density distributions of the final fractions obtained from multiple (L:H) separation of PSOC-732 in Ca(N03)~.The vertical lines represent the positions of the solution or cut densities.
I
PSOC-732
11
1.2
I
A
13
14
15
Density (g ~ m - ~ )
Figure 11. Density distributions of the fiial fractions obtained from multiple (H:L) separation of PSOC-732 in Ca(N03)~.The vertical lines represent the positions of the solution or cut densities. Table VIII. Purity Data for Final Density Fractions of PSOC-732
C1.24 1.24+1.295 1.2961.330 >1.33
93.0 79.4 58.4 78.3
46.2 70.7
85.9 85.6 70.6 92.6
89.0 95.7 84.8 98.2
This trend can also been seen in the single-cycle separations as well. There are three possible reasons for this: (1)The coal density patterns may be slightly shifted in Ca(N03)~
compared to CsCl solutions,so that the purities are lower. (2) The ADGC patterns for the Ca(N03)~fractions were run from dried samples. It is possible that there is a small shift upon drying, which we have not previously noticed. (3) There is a greater amount of aggregation occurring in the Ca(N03)2solutions than in the CsCl solutions. Which of the cases is most reasonable cannot be decided based on the present data. From Table VI1 and Figures 10 and 11,the sink contamination is again the main source of the lower purity of the final fractions. Whatever the cause of the discrepancy between the media, the effect appears to be subtle relative to problems due to mechanical phase mixing phenomena. Multiple Fractionationof APCS 3. The APCS 3 coal was a source of problems for us in part 1. Large and enigmatic changes in density occurred under different analysis conditions. Worse, there was a time component present in the problems. In spite of this, the single-stage S/F separations did show quite reasonable results. Thus, we decided to do some limited multiple separations of this coal. Figure 12 shows the two densities used for the separation. Both separation densities were just to the high-density side of the main vitrinite density band. The densities were chosen because this is the area with the worst expected separation purities. This also represents an incompletesequence. When the separation information detailed below was generated, we became convinced that little would be gained by further fractionation of this coal. Figure 13shows the L H and H L separation sequences. In the L H sequence, the final 1.309 g ~ m sink - ~fraction was recycled; in the H:L sequence, the fiia11.284 g ~ m - ~ float-phase fraction was recycled. The detailed data are presented in Table IX. Table X shows the amount of material expected, and the predicted purity for single cuts at the two densities used for the separation sequences. In both sequences the float fractions were quite pure. From the data in Table X this is expected, since the separations were to the high-density side of the vitrinite band. Because of the already high purity, recycling the fiial float phase in the H L sequence did not show any improvement. The corresponding sink-phase purities, on the other hand, were very low. Recycling the final sink fraction in the L:H sequencemarkedly improved the purity of the sink-phase material, from 47 to 98%. One further L H sequence of multiple separations was done on this coal and is displayed in Figure 14. The initial stages of this sequence were the same as for the L H
760 Energy & Fuels, Vol. 6,No.6,1992 p = 1.309 -+ 1.284
1-
1.309
1.284
99.3 41.4 = 1.284 + 1.309 APCS 3 1.284
98.0 87.6
[ 96.7 [ 28.4 1 1.309
Figure 13. Results of L:H and H:L multiple S/Fseparations of APCS 3. [Coal] = 20 g/L; rpm = 10 OOO; centrifugation time = 1 h. See Figure 4 for key. ~~~~~~~~
81.4 55.8
Table IX. Multide Fractionation of APCS 3. float phase sink phase density cut %pure wt% %pure wt% low to high (LH) 28.4 50.6 96.7 42.4 1.284 47.4 32.6 99.9 67.4 1.309 25.0 98.2 66.3 1.309RS 92.3 high to low (H:L) 49.0 18.7 72.6 1.309 97.0 12.8 55.2 94.9 40.5 1.284 5.1 43.1 88.7 57.7 1.284RF low to high (LH) 1.285 99.3 41.4 1.310 98.0 87.6 1.28SRCb 81.4 55.8 ~
a All sample purities me based on ADGC analysis of samples that were not dried prior to analysis. b Feed sample w a ~the combined sink and float from previous separation.
Table X. Percent of Float and Sink Expected from APCS 3 at Various Densities density (g cm-3) 1.285 1.310
wt%
float phaee 70.2 89.5
sink phase 29.8 10.5
theor % purity float phase sink phase 98.6 95.1 99.8 93.0
sequence shown in Figure 13. Although the float data between the two are in fair agreement, the sink data exhibit a broad range. However, after the final separation at 1.310 g cm-3, the float and sink phases were recombined and reseparated at 1.285 g ~ m - ~If.there were no separation occurring, we would expect to find 41.4% pure sink phase, ae in the fmt separation. The purity of the sink phase increased to 56.8% on this second cycle. However, we found that when we combined the sink and float phases from this last 1.285g ~ mseparation, - ~ the purity was 58% rather than the expected 41% for the sink phase. This information suggests that the density distribution of this coal wae W i n g to higher density during the sequence of separations. The reason for this shifting behavior remains uncertain. The purity of the fmal fractions for APCS 3 is presented in Table XI and Figures 15 and 16. Agreement between the L:Hand H L series is remarkably close. Unfortunately, the separations are very poor for this coal. Once again, there ie qualitative, but not quantitative, correspondence between the experimentaland theoretical data. The worst
1.1
1.2
1.3
1.4
5
Density (g ~ m - ~ )
Figure 15. Final density distributions for the L:H multiple S/F separation of APCS 3. Table XI. Final Purity of Fractions for APCS 3 exptl % purity density (g cm-3) LH HL theor % purity a.284 96.7 94.9 98.6 1.284-1.310 9.5 9.3 84.6 >1.310 47.4 49.0 93.0
separation is to be found for the density fraction closest to the main band (1.284-1.309g ~ m - ~ From ) . the figures, there is little eeparation evident. The purity in this region is below what we would have expected if we did not separate
Energy & Fuels, Vol. 6, No. 6,1992 751
Separation of Micronized Coals. 2
1
APCS-3
p