Powder Technology 288 (2016) 176–183

Contents lists available at ScienceDirect

Powder Technology

journal homepage: www.elsevier.com/locate/powtec

Assessing the feasibility of silica-based media for coal preparation

operations: A novel method of carbon feedstock production for the
silicon market
Seyed Hassan Amini a,⁎, Rick Honaker b, Aaron Noble c
a
Room 348 MRB, PO Box 6070, Department of Mining Engineering, West Virginia University, Morgantown, WV 26506, USA
b
504 Rose Street, 230 MMRB, Department of Mining Engineering, University of Kentucky, Lexington, KY 40506, USA
c
Room 353A MRB, PO Box 6070, Department of Mining Engineering, West Virginia University, Morgantown, WV 26506, USA

a r t i c l e i n f o a b s t r a c t

Article history: Production of silicon for solar cells and semiconductors requires the use of ultra-clean coal as the main reducing
Received 18 June 2015 agent in the silicon production process. Most coal preparation plants employ magnetite suspensions as the dense
Received in revised form 19 October 2015 medium in coal cleaning processes. Although downstream operations are designed to remove and recover magne-
Accepted 26 October 2015
tite, the unavoidable presence of residual magnetite in the final coal product dramatically decreases the purity and
Available online 27 October 2015
suitability of the coal for the silicon market. Given that silicates are the main raw materials in the silicon production,
Keywords:
a systematic study through laboratory dense-medium cyclone tests was used to evaluate fine silica-based alterna-
Silicon tive materials as a substitute for magnetite in coal cleaning operations. These alternative silica materials include
Magnetite byproducts generated during the silicon production process. The results indicate that laboratory tests with silica-
Dense medium cyclone based materials were able to achieve similar separation performance when compared to plant operations using
Silica-based material magnetite. Following the outcomes of the experimental study, a novel flowsheet arrangement was designed
using silica-based media rather than magnetite. If properly implemented, this new separation system can eliminate
iron contamination and provide coal that can be marketed as a high-quality silicon feedstock.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction The most economically viable route to produce metallurgical-grade

and electronic grade silicon is by carbothermic reduction of silica-rich
1.1. Background materials (e.g., quartz and quartzite) with carbonaceous reduction ma-
terials (e.g., coal, woodchips, charcoal, etc.) in a submerged arc furnace.
Silicon is an important raw material for various markets and applica- While the industrialized production of silicon is quite complex, the ide-
tions, including the chemical industries, alloy production, and electron- alized reaction for silicon production is expressed by [2,4,8]:
ics [1]. The end-use and the ultimate value of the silicon are largely
driven by the elemental purity as well as the composition of the con- SiO2 ðquartzÞ þ 2C ðcoal; charcoal; woodchipsÞ→Si þ 2CO: ð1Þ
taminants [2,3]. Silicon with over 95% purity is deemed metallurgical
grade silicon and is produced by reacting silica-rich materials with car- During the silicon smelting process, silica and carbon raw materials
bon-rich materials at high temperatures [4,5]. Further processing is are charged into the arc furnace, and the inside temperature is immedi-
often required to produce solar grade and electronic grade silicon ately increased to 700–1300 °C by high temperature gases rising in the
which requires over 99.9% Si purity. The cost of the final silicon product arc furnace. In this environment, different chemical reactions can occur
can be elevated by a factor of 10 by improving the purity specifications based upon the carbon coverage,1 which result in the production of free
from 0.5–100 ppm to less than 1 ppb [2,3,5]. Over the last decade, met- molten silicon, silica fume and silica slag (Fig. 1). The production of the
allurgical-grade silicon has doubled in price from $0.77 per pound in latter two materials is undesirable and results in a reduction of final sil-
2005 to a high of $3.48 per kilogram in 2011 [6,7]. Most recently, metal- icon recovery [1,9].
lurgical-grade silicon was valued at $2.67 per kilogram in 2014 [7]. The presence of metal contamination in the refined silicon product
results in poor performances and efficiency degradation in devices and
materials constructed from silicon. Of all the metal impurities, iron is
one of the most troublesome and must be minimized throughout the
⁎ Corresponding author at: 348 Mineral Resources Building, West Virginia University,
Morgantown, WV, 26506, USA.
1
E-mail addresses: [email protected] (S.H. Amini), [email protected] Carbon coverage may be defined by parameter x in the reaction between silica–rich
(R. Honaker), [email protected] (A. Noble). materials and carbon reduction agents as follows: SiO2 + xC.

http://dx.doi.org/10.1016/j.powtec.2015.10.045
0032-5910/© 2015 Elsevier B.V. All rights reserved.
 S.H. Amini et al. / Powder Technology 288 (2016) 176–183 177

Fig. 1. Stoichiometric model illustrating chemical reactions and material flowing in the submerged arc furnace (after Schei and Larsen [23]).

production process [3,10,11]. Unfortunately, iron is one of the most the overflow (Fig. 2). Traditional DMCs use a suspension of finely
abundant elements in nature and is difficult to entirely remove once it ground magnetite to control the medium density and the separation
is in the silicon production line. Due to the high boiling point tempera- cutpoint. In conventional applications, DMCs are used to remove high
ture of elemental iron, more than 99% of total iron in the raw feedstock density impurities from ROM coal particles greater than 1 mm. The sep-
material will ultimately condense into the final silicon product [12]. For aration in a DMC strongly depends on the weight distribution in the coal
electronic and semiconductor applications, iron is considered the most as a function of particle density and size, as well as the stability and rhe-
challenging impurity due to its high electrical activity. As a result, the ological properties of the medium [16,17].
maximum allowable level of iron contamination in electronic-grade sil- After the cleaning operation, the purified coal material is passed over
icon is decreasing annually as smaller electronic devices are increasingly fine drain-and-rinse screens. In this stage, water sprays are used to re-
sensitive to trace iron contamination [11,13]. Consequently, to ensure move and recover the magnetite, which is then recirculated through
the highest purity of produced silicon, iron must be completely re- the plant [18]. Despite this additional cleaning stage, small quantities
moved from the feedstock material. of ultrafine magnetite remain on the surface of final clean coal product.

1.2. Role of coal preparation

As shown in the idealized reaction (Eq. (1)), production of

metallurgical-grade silicon requires the use of ultra-clean coal. Coal is
the main reducing agent in the silicon production and accounts for
26% of raw material used in the silicon smelting process [1]. Coal prep-
aration technology plays a significant role in the silicon market by
upgrading run-of-mine (ROM) coal to satisfy the particle size and qual-
ity specifications. Modern coal preparation plants utilize a complicated
arrangement of solid–solid and solid–liquid separation processes de-
signed to improve the purity of coal thereby increasing the inherent
value and suitability for downstream usages. Coal preparation plants
can be designed and operated to serve a number of different down-
stream applications and markets, including power generation, coke pro-
duction and silicon production [14–16].
The dense-medium cyclone (DMC) is one of the most widely applied
gravity concentration units employed in the coal preparation process to
clean a major portion of the plant feed mass. The geometry of the unit
induces a centrifugal field which causes differential settling of particles
with contrasting densities. High density particles settle quickly to the
outside of the unit and are expelled through the underflow while low
density particles are retained with the air core and expelled through Fig. 2. Schematic of internal flow for a typical DMC (after Sanders [24]).
178 S.H. Amini et al. / Powder Technology 288 (2016) 176–183

While this inefficiency is acceptable for other coal usages, the unavoid- ash content of 13.22%, which was uniformly distributed in each size
able presence of magnetite in the final coal product dramatically de- fraction.
creases the purity and suitability of the coal for the silicon market. Float–sink testing was conducted according to ASTM D4371-06
using lithium metatungstate (LMT) and distilled water. Incremental rel-
ative density values of 1.25, 1.35, 1.45, 1.55, and 1.65 were used. Based
1.3. Objective
on the results of this analysis (Table 2), the coal was classified as easy-
to-clean as indicated by a cleaning index of 0.98 for the DMC feed
In an effort to improve the purity of the coal, an investigation was
(C.I. = 1.3 cumulative weight % float/1.6 cumulative weight % float).
conducted to identify a potential substitute for magnetite in the DMC
High quantities (85.48%) of the coal mass were floated in the lowest
coal cleaning process. Potential low-cost substitutes include silica-
two density fractions, and 52.00% exists in the 1.25 × 1.35 relative den-
based materials, such as fine silica sand, fine silicon and ultrafine silica
sity (RD) fraction.
fume. Contamination from inefficient drain-and-rinse operations will
not influence the suitability of these materials, as the remnant silica
2.1.2. Silica-based medium materials characteristics
will be ultimately consumed in the downstream silicon production pro-
Fine silica sand was procured from a standard dealer. Based on the
cess. Furthermore, as mentioned above, the latter two of these options
particle size distribution data in Table 3, the silica sand is significantly
are byproducts from the silicon production process and may support
coarse as indicated by 83.28% of the particles being coarser than
supplemental environmental goals by introducing a new sustainable
75 μm (200 mesh). Moreover, images from a scanning electron micro-
use for the byproduct waste.
scope (SEM) analysis shows that the material has a high population of
An experimental program was designed to address two fundamental
spherical particles (Fig. 3a).
questions associated with silica-based dense medium separations:
Ultrafine silica fume was obtained to be used as a medium stabilizer
(1) can a stable medium be generated from the alternative media and
with the silica sand. Silica fume, which is also called micro-silica, is an
(2) can the alternate media produce separation efficiencies comparable
ultrafine powder generated during the production of metallurgical-
to magnetite-based separation. Static and dynamic stability tests were
grade silicon. Microscopic analysis (Fig. 3b) shows that the particles
used to identify which silica-based materials produced stable mediums.
were mostly spherical particles, and particle size analysis confirms
The materials capable of producing a stable medium were then selected
that 96.38% of the particles are finer than 45 μm (325 mesh) (Table 3).
for separation testing in a laboratory-scale dense medium cyclone. Var-
Fine silicon material, which was purchased from a silicon production
ious indicators of separation efficiency, including final product ash, or-
facility, was the final silica-based material evaluated in this study. Unlike
ganic efficiency, probably error (Ep), and low-density bypass, were all
the other two alternatives, microscopic analysis (Fig. 3c) shows that most
used to compare the silica-based media to magnetite-based separations.
of this material had an irregular shape. Particle size analysis also indicates
The cumulative results from these tests were then used to assess the
that N95% of the material is finer than 75 μm (200 mesh) (Table 3).
suitability of silica-based media as a potential replacement for magne-
tite in DMC operations.
2.1.3. Laboratory-scale dense medium cyclone circuit
A 15 cm (6 in.) diameter laboratory-scale DMC unit was used in the
2. Experimental materials and methods separation tests. The DMC was outfitted with a 63-mm vortex and a 45-
mm diameter apex and operated at inlet pressures between 16 and
2.1. Materials 21 kPa (2.3 to 3.1-psi) which provides a total pressure head equivalent
to the industrial standard of nine cyclone diameters. As shown in Fig. 4,
2.1.1. Coal characteristics the DMC tests were performed using a closed-loop circuit with the over-
The coal sample used throughout the experimental testing was ob- flow and underflow streams of the cyclone reporting back to the feed
tained from a coal preparation plant that treats coal from the Blue sump. A feed bypass stream also reported to the sump which was
Gem seam in southeastern Kentucky. The coal is a high quality (low used to collect feed samples. The relative hardness of the coal mini-
ash and low sulfur content) resource that is categorized as high-volatile mized particle breakage during the recycling of the coal through the
bituminous B according to the American coal classification. Due to the feed sump and pump. However, the number of tests was limited for
high quality of the coal, Blue Gem coal is heavily recognized as a suitable each batch of coal to minimize particle degradation effects.
source of carbon raw material in the silicon smelting process [19].
The sampled coal was screened using 6.35 mm and 1 mm screens. 2.2. Methods
The + 6.35 mm was crushed in a laboratory jaw crusher and then
screened to obtain the 6.35 × 1 mm particle size feed material. The 2.2.1. Laboratory procedures for medium stabilization
coal sample was then mixed and split into representative 10 kg lots, To assess the potential of the silica-based materials for coal cleaning
which were used as the laboratory DMC feed material. A representative applications, an initial static stability study was conducted using fine
sample was collected for particle size and float–sink analyses. solids over a range of medium relative density values. In the absence of
The particle size analysis of the DMC feed (Table 1) shows that the coal, the alternative silica-based material was mixed with water to gener-
majority of particles (37.40%) are in the 6.35 × 4.00 mm size fraction. ate a suspension with a desired density value. The suspension was
The sample has a low inherent moisture content of 1.66% with a dry
Table 2
The float-sink data of the dense medium cyclone feed material.
Table 1
Particle size-by-size analysis of the Blue Gem dense medium cyclone feed material. Relative Individual Cumulative float Cumulative sink
density fractions
Particle size Incremental Cumulative
Float Sink Wt. (%) Ash (%) Wt. (%) Ash (%) Wt. (%) Ash (%)
(mm)
Actual Weight Moisture Ash Weight Ash
– 1.25 33.48 0.81 33.48 0.81 100.00 13.28
weight (%) (%) (%) (%) (%)
1.25 1.35 52.00 1.61 85.48 1.30 66.52 19.55
6.35 × 4.00 98.55 37.40 1.50 14.36 37.40 14.36 1.35 1.45 0.75 12.38 86.23 1.39 14.52 83.79
4.00 × 2.83 72.49 27.51 1.78 12.23 64.91 13.46 1.45 1.55 0.47 18.53 86.70 1.49 13.77 87.68
2.83 × 2.00 43.79 16.62 1.84 13.78 81.53 13.52 1.55 1.65 0.08 30.20 86.78 1.51 13.30 90.13
2.00 × 1.00 48.67 18.47 1.62 11.87 100.00 13.22 1.65 – 13.22 90.49 100.00 13.28 13.22 90.49
Total 263.50 100.00 1.66 13.22 Total 100.00 13.28
 S.H. Amini et al. / Powder Technology 288 (2016) 176–183 179

Table 3
Particle size distribution of the alternative silica-based materials.

Weight distribution (%)

Particle size (micron) Sand Silica fume Silicon

Plus 250 0.77 0.23 0.65

250 × 149 13.75 0.55 1.48
149 × 74 68.76 1.6 3.03
74 × 44 11.47 1.24 48.21
Minus 44 5.25 96.38 46.63

transferred to a 2-liter glass graduated cylinder. To determine the static

stability properties of the suspension, the mixture was then pumped
from the graduated cylinder at the one liter height line. The suspended
material removed by the pump was filtered, dried, and weighted. If the
suspension is %100 percent stable, the mass of solid in this fraction should
be exactly half of the total added solid. Therefore the ratio of suspended
mass to half of the total solid mass provides a qualitative assessment of
Fig. 4. Schematic diagram of the dense medium cyclone circuit.

the suitability of the medium with values greater than 75% being ideal
and values less than 25% being unacceptable.
Subsequently, dynamic medium stability was assessed in a 15-cm
(6-in) DMC by measuring the pulp differential density values between
the overflow and underflow streams in the absence of coal. At the begin-
ning of each stability test, the suspension density was adjusted to the
desired value in the circuit feed sump. The suspension was then
pumped to the cyclone at a desired pressure and samples were taken
from the underflow and overflow streams. The medium stability was
assessed by measuring the differential pulp RD values between the
overflow and underflow stream (i.e., ρu–ρo). The industrial operating
standard specifies that the differential density should be less than 0.4
RD units to ensure optimum separation performance [17,20]. The
most significant factors affecting the density differential include: the
shape of the solid particles, size of the solid particles, the relative medi-
um density, and the solid particle population [21,22].

2.2.2. Laboratory procedures for DMC tests

The second phase of the experimental program consisted of coal
cleaning tests using the laboratory-scale DMC unit. To prepare for
each test, the dense medium was first adjusted to the desired value
followed by the addition of the coal in sufficient quantity to provide a
medium-to-coal ratio of 4:1 on a volume basis. The cyclone was then
allowed to run in closed loop for 15 min to ensure sufficient mixing
and steady-state conditions. Simultaneous samples were then collected
from the overflow (clean coal and medium), underflow (reject and me-
dium), and bypass feed (raw coal and medium) streams. Afterward, the
samples were wet screened using a 1 mm (16 mesh) sieve to separate
medium from the solid coal material. Finally, detailed float–sink analysis
was conducted on each sample. Repeat experimental trials were con-
ducted 3 to 5 times, depending on the specific silica material. Through-
out all testing and material characterization, ash analysis was conducted
using a LECO TGA701 while particle size analysis was conducted using a
standard wet screening procedure with standard sieves.

2.2.3. Data reduction and analysis

Partition curves were used to analyze the separation efficiencies and
characterize the particle misplacement. Fig. 5 shows a typical partition
curve, illustrating the probability (y-axis value) that a particle with a
given RD (x-axis value) will report to the clean coal product. The parti-
tion curve is widely used in the coal preparation industry and provides
Fig. 3. SEM images alternative silica-based material: (a) silica sand, (b) silica fume, and several measures of efficiency including: high density bypass (R1), low-
(c) fine silicon. density bypass (R2), cutpoint (ρ50), and probable error (Ep). The first
180 S.H. Amini et al. / Powder Technology 288 (2016) 176–183

Fig. 6. Static medium stability for alternative silica-based media as determined by a static-
cylinder particle settling test.

To improve the stability of the silicon sand, silica fume was added at
different ratios during subsequent testing. Results of the additional stat-
Fig. 5. Generic partition curve for a coal cleaning operation. Symbols include: high density ic stability tests (Fig. 7) show that the mixture of sand and fume can pro-
bypass (R1), Low-density bypass (R2), Cutpoint (ρ50), and Probable error (Ep). duce stable medium mixtures over the relative density range from 1.30
to 1.50 by adding different ratios of silica fume with increased silica
fume producing more stable mediums. This trend demonstrates the
three values can be directly read from the curve and the Ep value calcu- ability of ultrafine fume particles to enhance the stability properties of
lated by: a dense medium predominantly composed of coarse sand particles
which minimizes negative medium viscosity effects on separation effi-
ρ25 −ρ75 ciency. This result also shows that at the same ratio of silica fume, a
Ep ¼ ð2Þ small increase in RD can produce a significant increase in medium
2
stability.
For dynamic medium stability, the density differential between the
where ρ25 and ρ75 are the RD values of the particles in the feed that have cyclone overflow and underflow streams is a reliable indicator of the
a 25% and 75% chance of reporting to the overflow (product) stream, re- amount of density gradation inside the separator. Figs. 8 through 10
spectively. More efficient separation processes are characterized by a show the results from the dynamic medium stabilization tests. For all
steeper slope and a lower Ep value. Under ideal conditions, the partition materials, the medium stability improved with increases in the medium
curve is vertical and the Ep value is zero. For DMC processes, Ep values in RD, as indicated by the decrease in RD differential values. This trend is
the range of 0.01 to 0.05 are typical for well-operated equipment. due to the rise in solid concentration and particle population, which hin-
In addition to the separation efficiency parameters that can be deter- ders the particle settling rate thereby providing a stable medium.
mined using partition curve, organic efficiency (OE) is widely used as an As shown in Fig. 8, the silica-sand suspension was not stable at any
indicator of the separation efficiency in coal preparation processes. The medium density fractions between 1.20 and 1.50, even at increased
OE value is calculated as a ratio of the actual clean coal energy recovery feed inlet pressures. These results confirm the expectations from the
to the theoretical maximum recovery achievable at the same ash ac- static medium tests and further show that by itself, silicon sand is not
cording to float–sink data. Since organic efficiency reflects the impact a suitable material for dense medium separations.
of both probable error and high/low-density bypass, the parameter is As indicated in Fig. 9, the fine silicon material generates a stable me-
a very effective measurement of overall process efficiency. dium over the target medium RD values (1.25 to 1.45 RD). High volu-
metric concentration (due to low particle density) combined with
3. Results and discussion

3.1. Medium stabilization

Medium stability directly influences separation performance through

the creation of density gradients within the cyclone due to the effect of
centrifugal acceleration. Consequently, static medium stability was first
assessed to qualitatively observe the stability properties of silica-based
media over a range of medium relative density values. As indicated in
Fig. 6, the fine silica sand suspension was completely unstable at the me-
dium density values between 1.20RD and 1.50RD; while the ultrafine sil-
ica fume medium was completely stable throughout the range of relative
medium density values tested. The fine silicon produced acceptable
values for RD values greater than 1.25. The instability of the silica sand
medium is likely due to the nearly spherical particles (Fig. 3a) and the
relative coarse size distribution that provided fast-settling rates in the
suspension. Conversely, the ultrafine size of the silica fume and the irreg-
ularly particle shape of fine silicon generally produces acceptable stabi- Fig. 7. Static medium stability results for composite mixtures of silica fume and silica sand
lized dense mediums. media as determined by a static-cylinder particle settling test.
 S.H. Amini et al. / Powder Technology 288 (2016) 176–183 181

Fig. 10. Dynamic medium stability results obtained for the silica sand/fume mixtures using
Fig. 8. Dynamic medium stability results for the silica sand material using the laboratory
the laboratory DMC unit at a pressure head equivalent to 9-cyclone diameters.
DMC unit at a pressure head equivalent to 9-cyclone diameters.

turbulent drag caused by the irregularly shaped particles produced a DMC circuit has been plotted alongside the laboratory results. Altogeth-
highly stable medium even at relatively low relative densities. er, this data shows that the silica-based media produce separation effi-
To overcome the instability of the fine-silica sand suspension, a bi- ciencies comparable to that of traditional magnetite-based DMCs.
modal dense medium composed of the coarse sand and the fine silica Moreover, the results indicated that a clean coal product with less
fume was generated at different proportions of silica fume to total than 1.5% ash can be produced using a medium formed from the sili-
solid concentration. As shown in Fig. 10, blends that range from 30% ca-based materials with an organic efficiency value above 98% and a
to 50% silica fume were found to provide acceptable medium stability probable error value below 0.02. There was no measurable bypass of
within the medium RD target range. The density differential decreased high density particles into the product stream or low-density particles
and thus the medium became more stabilized with an increase in the ul- into the reject stream. Testing with the sand/silica fume mixture pro-
trafine silica fume concentration. The bimodal suspension composed of duced a higher organic efficiency with a slightly higher product ash
50% silica fume and 50% fine silica sand provided acceptable medium when compared to the fine silicon and magnetite tests.
stability over the entire range of medium density values tested while a
concentration of 40% silica fume provides an ideal medium for medium 3.3. DMC circuit design
density values above 1.35.
As discussed in the previous section, the alternative materials can be
3.2. DMC separation tests utilized to replace magnetite in the DMC separation. However, a typical
DMC circuit must be modified to recover and reuse the silica-based
Based on the results of the medium stability tests, laboratory-scale dense medium in the coal preparation plant. The most significant of
DMC separation tests were conducted using the various alternative these changes involves eliminating the magnetic separator used to re-
media in the portions and suspension RD values that satisfied stability cover magnetite from a dilute slurry. This modification may consider-
requirements. In addition to the alternative medium (fine silicon and ably reduce the operating costs in coal preparation plants. However,
a 40/60 silicon fume/silicon sand mixture), plant data using magnetite additional classifying cyclones and a thickener must be installed to re-
were used as a control benchmark to determine the relative separation cover the silica-based dense medium from dilute slurry.
efficiency when compared to the industry standard. After the tests,
float–sink analysis was performed on the feed, product, and tailing sam-
ples collected from each test to construct partition curves.
Figs. 11 and 12 show the results from the laboratory-scale DMC sep-
aration tests. For comparison, data from the plant's magnetite-based

Fig. 9. Dynamic medium stability results for the fine silicon material using the laboratory Fig. 11. Typical partition curves generated from the separation achieved when using sand–
DMC unit at a pressure head equivalent to 9-cyclone diameters. fume blend, silicon, and magnetite mediums at 9-cyclone diameter inlet pressure.
182 S.H. Amini et al. / Powder Technology 288 (2016) 176–183

separation of high and low-density particles in the cyclone, the clean

coal and reject materials are passed over drain-and-rinse screens, and
then the final product is sent to the dewatering stage. Undulated
(corrected) dense medium, which is drained from the initial part of
the product and reject screens, is directly sent to the correct medium
sump and then is pumped to the main cyclone feed sump. The diluted
dense medium, which has a lower medium specific gravity, flows to
the dilute medium sump and is pumped to classifying cyclone. The
underflow material from the classifying cyclone, which has a high
solid concentration, gravitates to the over-dense medium sump, while
the overflow material, with low solid concentration, is sent to thickener.
In the thickener, high density ultrafine particle are settled and report to
the over-dense medium sump to be reused in the DMC circuit.

3.4. Discussion

Use of silica-based medium and the modified circuit design will re-
sult in the production of ultra-clean coal with no residual iron contam-
ination from magnetite. Since the value and application of final coal
product in different downstream markets is largely dictated by the qual-
ity of the clean coal, this high value ultra-clean coal product can be used
beneficially in the silicon market.
Since the dense medium can be expensive, this selection plays an in-
tegral role in determining the economics of any coal preparation plant.
In the conventional dense medium plant, 10–20% of the total operating
cost is due to dense medium loss. In general, the amount of residual
magnetite on the surface of coal product particles accounts for more
than 60% of total medium loss [18]. Considering the low cost of silica-
based materials in contrast to magnetite, utilization of silica-based me-
dium will remarkably decrease the coal preparation plant operating
cost.
One of the most effective routes for the purification of metallurgical-
grade silicon and removal of iron contamination is to pulverize the pro-
duced silicon and leach the powder with various acids including HCl, HF,
Fig. 12. Test summary showing various measures of separation efficiency for magnetite H2SO4 and HNO3. The degree of purification is largely determined by the
(plant data) and silica-based media (laboratory data).
silicon particle size, time of acid leaching and diffusion rate of iron in the
liquid state [3,5]. Since iron belongs to the fast diffusers group [11], the
Fig. 13 shows a typical magnetite based DMC circuit in contrast to a purification process requires a high amount of energy and time for
new silica-based dense medium circuit. In the new circuit, raw coal feed grinding and acid leaching of metallurgical-grade silicon. Utilization of
materials are mixed with the dense medium in the cyclone feed sump a modified dense medium circuit that employs silica-based materials
and then the mixture is pumped directly to the cyclone. After the may eliminate iron from the entire silicon production line. Therefore,

Fig. 13. Comparison between a magnetite and silica-based dense medium cyclone circuit design.
 S.H. Amini et al. / Powder Technology 288 (2016) 176–183 183

higher quality silicon may be produced and sent directly to downstream References
markets, such as solar cell and semiconductor industries with no re-
[1] V. Dosaj, M. Kroupa, R. Bittar, Silicon and Silicon Alloys, Chemical and Metallurgical,
quirement for the acid leaching purification processes. Indeed, produc- Kirk-Othmer Encyclopedia of Chemical Technology. Wiley Online Library, 2005.
tion of this highly pure silicon will save energy, time and unnecessary [2] A. Istratov, T. Buonassisi, M. Pickett, M. Heuer, E. Weber, Control of metal impurities
costs. in “dirty” multicrystalline silicon for solar cells, Mater. Sci. Eng. B 134 (2) (2006)
282–286 (Elsevier;).
[3] P. Woditsch, W. Koch, Solar grade silicon feedstock supply for PV industry, Sol.
Energy Mater. Sol. Cells 72 (1) (2002) 11–26 (Elsevier;).
4. Conclusions [4] J. Amick, J. Dismukes, R. Francis, L. Hunt, P. Ravishankar, M. Schneider, et al., Im-
proved high-purity arc-furnace silicon for solar cells, J. Electrochem. Soc. 132 (2)
(1985) 339–345 (The Electrochemical Society;).
From results of the medium stability and separation tests, three pri- [5] B. Bathey, M. Cretella, Solar-grade silicon, J. Mater. Sci. 17 (11) (1982) 3077–3096
mary conclusions are derived: (Springer;).
[6] L.A. Corathers, Silicon, mineral commodity summaries, US Geol. Surv. (2006)
144–145.
1. When used alone, standard silica sand is an unsuitable dense [7] C.A. Tuck, Silicon, mineral commodity summaries, US Geol. Surv. (2014) 144–145.
medium. The sand suspension was not stable at any of the desired [8] M. Gasik, Handbook of Ferroalloys: Theory and Technology, Butterworth-
medium densities; however, when mixed at different ratios with sil- Heinemann, 2013.
[9] A. Schei, J.K. Tuset, H. Tveit, Production of High Silicon Alloys, Tapir Trondheim,
ica fume, the mixture becomes stable over the desired operating re-
Norway, 1998.
gions. Alternatively, fine silicon is a suitable medium by itself. Many [10] S. Dubois, S. Martinuzzi, Metal impurities in multicrystalline silicon solar cells: com-
of the observations can be explained by the shape and size distribu- parison between fast and slow diffusers, Photovoltaic Specialists Conference, 2008
PVSC’08 33rd IEEE 2008, pp. 1–6.
tion of the silica materials.
[11] A. Istratov, H. Hieslmair, E. Weber, Iron contamination in silicon technology, Appl.
2. While both the fine silicon and the blend of silica sand and fume were Phys. A 70 (5) (2000) 489–534 (Springer;).
able to produce a stable medium, the sand/fume blend performed [12] E. Myrhaug, J. Tuset, H. Tveit, Reaction mechanisms of charcoal and coke in the sil-
better with respect to separation performance. This silica-based icon process, Proceedings: Tenth International Ferroalloys Congress 2004, p. 4.
[13] A. Istratov, H. Hieslmair, E. Weber, Iron and its complexes in silicon, Appl. Phys. A 69
media was capable of producing product ash less than 1.5% and or- (1) (1999) 13–44 (Springer;).
ganic efficiency values comparable to magnetite. [14] G. Luttrell, P. Bethell, R. Honaker, Designing and operating fine coal processing cir-
3. Use of silica-based media will require a redesigned medium recover cuits to meet market specifications, Int. J. Coal Prep. Util. 34 (3–4) (2014) 172–183
(Taylor \& Francis;).
circuit when implemented in a plant setting. Nevertheless, the new [15] G. Sherritt, A. Meyers, A. Bennetts, J. Graham, Delineation of large diameter dense
silica recovery circuit may reduce operating costs as fewer units are medium cyclone performance, The Australian Coal Industry's Research Program,
needed. Furthermore, high silica recoveries are not required since ACARP Report C 2010, p. 17036.
[16] H.J. Gluskoter, Meeting Projected Coal Production Demands in the USA: Upstream
any “contamination” will be readily consumed in the silicon metal Issues, Challenges, and Strategies: Sumary Brochure and Report Pdf, Virginia Center
production process. for Coal and Energy Research, Virginia Polytechnic Institute and State University,
2009.
[17] Y. He, J. Laskowski, Effect of dense medium properties on the separation perfor-
Altogether, these conclusions show that silica-based materials and mance of a dense medium cyclone, Miner. Eng. 7 (2) (1994) 209–221 (Elsevier;).
particularly byproducts from the silicon production process can be suc- [18] R. Sripriya, A. Dutta, P. Dhall, M. Narasimha, V. Kumar, B. Tiwari, An analysis of me-
cessfully utilized as alternatives to magnetite in the coal preparation dium losses in coal washing plants, Int. J. Miner. Process. 80 (2) (2006) 177–188
(Elsevier;).
processes. This substitution eliminates the iron contamination caused
[19] V. Myrvågnes, Analyses and Characterization of Fossil Carbonaceous Materials for
by residual magnetite on the clean coal product and may support sus- Silicon Production(Doctoral Dissertation) Norwegian University of Science and
tainability efforts as waste byproducts from the silicon production pro- Technology, 2008 (232 pp.).
cess can be readily used by coal preparation plants. With no iron [20] R. Honaker, C. Bimpong, Alternative materials for dense medium separations, Int. J.
Coal Prep. Util. 29 (4) (2009) 173–191 (Taylor \& Francis;).
contamination, the ultra-clean coal can be successfully marketed to [21] B. Collins, T. Napier-Munn, M. Sciarone, The production, properties, and selection of
the silicon production industry as a carbon source. ferrosilicon powders for heavy–medium separation, J. South. Afr. Inst. Min. Metall.
75 (5) (1974) 103–105.
[22] Y. He, J. Laskowski, Dense medium cyclone separation of fine particles part 2. The
effect of medium composition on dense medium cyclone performance, Coal Prep.
Acknowledgments 16 (1–2) (1995) 27–49 (Taylor \& Francis;).
[23] A. Schei, K. Larsen, A stoichiometric model of the ferrosilicon process, J. Four Electr.
The authors greatly appreciate the interest and financial support of 87 (2) (1982) 27–35.
[24] G.J. Sanders, The Principles of Coal Preparation, Australian Coal Preparation Society,
Alden Resources. The authors also acknowledge the technical contribu- Dangar, New South Wales, 2007.
tions provided by Dr. Datta Patil.