1
Scientific RepoRts | 7:42734 | DOI: 10.1038/srep42734
www.nature.com/scientificreports
Beads-Milling of Waste Si Sawdust
into High-Performance Nanoakes
for Lithium-Ion Batteries
Takatoshi Kasukabe
1
, Hirotomo Nishihara
1,2
, Katsuya Kimura
3
, Taketoshi Matsumoto
3
,
Hikaru Kobayashi
3
, Makoto Okai
4
& Takashi Kyotani
1
Nowadays, ca. 176,640 tons/year of silicon (Si) (>4N) is manufactured for Si wafers used for
semiconductor industry. The production of the highly pure Si wafers inevitably includes very high-
temperature steps at 1400–2000 °C, which is energy-consuming and environmentally unfriendly.
Ineciently, ca. 45–55% of such costly Si is lost simply as sawdust in the cutting process. In this work,
we develop a cost-eective way to recycle Si sawdust as a high-performance anode material for lithium-
ion batteries. By a beads-milling process, nanoakes with extremely small thickness (15–17 nm)
and large diameter (0.2–1 µm) are obtained. The nanoake framework is transformed into a high-
performance porous structure, named wrinkled structure, through a self-organization induced by
lithiation/delithiation cycling. Under capacity restriction up to 1200 mAh g
−1
, the best sample can retain
the constant capacity over 800 cycles with a reasonably high coulombic eciency (98–99.8%).
e world production of silicon (Si) metal in 2014 was about 1,766,400 tons
1
, and about 10% of them is high-purity
grade (> 4N) for Si wafers which are used for semiconductor industry including integrated circuits and photovol-
taic cells. Prior to the production of Si wafers, Si is rst prepared by the reduction of SiO
2
at a very high temper-
ature (> 1900 °C) with the presence of reducing agent such as charcoal and coal. e raw Si thus obtained is then
reacted with HCl to become SiHCl
3
, which is further rectied into high-purity Si metal. Aerwards, the Si metal
is molten at a high temperature above its melting point (> 1414 °C), and a large ingot of single-crystalline or poly-
crystalline Si is prepared typically by the Czochralski or oating zone technique with taking more than several
tens of hours. us, the production of Si ingots is extremely energy-consuming, i.e., environmentally unfriendly,
and such a high-temperature process inevitably increases the price of Si. en, the ingots are nally cut into Si
wafers and they are used for a variety of applications. What is awfully inecient is the fact that ca. 45–55% of
the high-quality Si is lost simply as sawdust in the cutting process
2
. is means that a large amount of energy
used for the production of the Si ingots is idly abandoned. Hence, the reuse of the Si sawdust is highly desired
from the sustainable point of view. One of the potential applications of the Si sawdust could be a high-capacity
anode material for lithium-ion batteries (LIBs). Si has a very high theoretical capacity (3572 mAh g
−1
)
3
compared to that of conventional graphite (372 mAh g
−1
), and there have been a great deal of researches report-
ing the development of Si-based high-capacity materials (800–3000 mAh g
−1
) for the LIB application as found in
comprehensive review papers
4,5
. In 2015, the world demand for graphite in all batteries is estimated to be 125,000
tons
6
. If graphite is replaced by Si, the necessary amount of Si is estimated to be ca. 15,000–58,000 tons, considering
its high capacity. Accordingly, the amount of Si sawdust (ca. 88,320 tons) meets the demand of anode materials
for LIBs.
However, there are several technical hurdles to realize the reuse of Si sawdust for LIBs. First, Si sawdust is
obtained as sludge mixture containing other contaminants, and therefore, a purication process is necessary. For
the manufacture of Si wafers, a Si ingot is sliced by a wire saw, and during this cutting process, a large amount of
coolant is used and thus it remains in the sawdust. At the same time, metal impurities derived from the wire saw
are mixed in the sawdust. Additionally, the cutting process usually needs to use a substrate such as graphite, alu-
mina, and polymers, and sawdust of the substrate is also mixed. Hence, these impurities should be removed prior
1
Institute of Multidisciplinary Research for Advanced Materials, Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai
980-8577, Japan.
2
PRESTO, The Japan Science and Technology Agency (JST), 4-1-8 Honcho, Kawaguchi 332-0012,
Japan.
3
The Institute of Scientic and Industrial Research, Osaka University, 8-1 Mihogaoka, Ibaraki, Osaka 567-0047,
Japan.
4
Center for Technology Innovation-Materials, Hitachi Ltd., Hitachi, Ibaraki 319-1292, Japan. Correspondence
and requests for materials should be addressed to H.N. (email: nisihara@tagen.tohoku.ac.jp)
received: 14 October 2016
accepted: 12 January 2017
Published: 20 February 2017
OPEN
www.nature.com/scientificreports/
2
Scientific RepoRts | 7:42734 | DOI: 10.1038/srep42734
to use for LIBs. In this work, we recover Si sawdust whose quality is high enough for LIB application, based on
our proposed technique on the production of Si nanoparticles from the Si sawdust
7
. e second problem is about
how to mold Si sawdust into an appropriate nanostructure which is the key factor to achieve a high performance
and a long-term durability as an anode for LIBs. Si has intrinsic problems of (i) low conductivity, (ii) low reaction
rate with lithium, and (iii) large volume change upon lithiation/delithiation. To overcome these problems, it is
generally anticipated that the following means are of importance: (I) the domain size of Si should be less than
a few hundred nanometers, (II) Si is necessarily mixed with conductive additives, and (III) buer nanospace
should be placed around Si
8
. us far, high performance has been achieved in a variety of Si-based nanomaterials
such as Si nanowires
9–11
, Si nanotubes
12,13
, porous Si particles
14
, carbon inverse-opal decorated with Si nano-
layer/nanoparticles
15,16
, C/Si or Si/C core-shell nanowires
11,17
. However, most of them are produced by multistep,
costly, and environmentally unfriendly processes. In addition, it is impossible to mold Si sawdust into either of
these specic nanostructures. us, Si nanoparticles have been prepared from Si sawdust for LIBs, by means
of plasma jet treatment
18
, ultrasonic spray-draying
19
, and high energy ball milling
20
. ey show high capacity
(1000–2000 mAh g
−1
) at early cycles, but the capacity is faded by 50–150 cycles
18–20
, and the cyclability should be
improved much more for practical application. e reason for the insucient cyclability could be the morphol-
ogy of the materials prepared from Si sawdust, i.e., discrete Si nanoparticles. One can be aware that most of the
afore-mentioned high-performance Si-based materials possess a continuous Si framework like bers in contrast
to isolated and discrete Si particles. is is because such continuous Si frameworks show specic structure change
during charge/discharge cycling
21–25
. By repeating volume expansion by lithiation and contraction by delithiation,
the continuous Si frameworks turn into porous framework resembling wrinkled papers, in which a high perfor-
mance can be retained. We have indeed demonstrated that Si nanoparticles which are tightly connected each
other to have a continuous structure can be readily changed to high-performance ‘wrinkled’ Si framework upon
the repetition of lithiation/delithiation
21,22
. Later on, almost the same phenomenon has been reported also by
Park et al.
23
. Moreover, we have reported the detailed structure of the wrinkled Si framework to elucidate its high
performance
21
. e formation of similar structures to the wrinkled structure have been reported in the Si nanow-
ires
24
and carbon nanotube covered by Si nanolayer
25
, and also in Ge
26–34
and SnO
2
35
, which involve large volume
expansion/contraction upon lithiation/delithiation. In these examples, the presence of a continuous framework
consisting of a nano-sized active material (Si, Ge, or SnO
2
) is the key factor for the formation of the wrinkled
structure, thereby exhibiting a high performance. By contrast, discrete Si nanoparticles prepared by the practical
milling process never show such structure transformation, and their performance is quite low as a result
22
. us,
it is a challenge to realize the wrinkled structure in discrete Si nanoparticles prepared by milling process which
is practically sound.
In this work, we demonstrate that an appropriate milling process can convert Si sawdust into nanoakes hav-
ing a very thin thickness (ca. 16 nm) and a large domain size (0.2–1 µ m), which functions as a continuous frame-
work. As a result, the nanoakes are transformed into the wrinkled structure during charge/discharge cycling by
self-organization, and exhibit better performance than discrete and spherical Si nanoparticles. Moreover, we have
examined the eect of carbon-coating over the Si nanoakes, and also several dierent cell-assembling methods
according to literature, to further enhance the performance of the Si nanoakes. e obtained results demonstrate
an excellent feasibility of the recycling of Si sawdust as a high-performance anode material for LIBs.
Results and Discussion
Preparation and Characterization of Si nanoparticles. Si sawdust (Si(sd); about 0.2–4 µ m in particle
size) was produced from real Si-sawdust sludge derived from phosphorus-doped n-type crystalline Si ingots with
a resistivity of 1–2 cm. Shortly, Si(sd) was prepared by washing the waste sludge with acetone to remove the
coolant. Si(sd) thus obtained contains 4 wt% of graphite sawdust which is derived from a graphite substrate used
for the cutting process of the Si ingots. Fortunately, graphite functions as an anode of LIBs and it is not necessary
to be removed. For comparison, commercial Si powder (Si(com); purity 99.9%, average particle size 1–5 µ m,
Alfa Aesar) was used. Each of the Si powdery samples was milled with the presence of water or isopropanol
down to nanoscale by the following two methods: (1) ball-milling with a Pulverisette 7 premium line apparatus
(Fritsch Co., Ltd.)
22
, and (2) beads-milling with a LMZ015 apparatus (Ashizawa Fintech Ltd.). en the sample
was washed with 1 wt% HF solution to remove a surface SiO
2
layer. e more detail information is shown in
SupplementaryInformation. e Si nanoparticles thus obtained are referred to as Si(X)-Y(Z) where X expresses
the raw Si powder (sd or com, corresponding to sawdust and commercial Si powders, respectively), Y stands for
the type of the milling apparatus (ba or be, corresponding to ball-milling and beads-milling, respectively), and Z
stands for the solvent (w or ipa, corresponding to water and isopropanol, respectively).
Figure1a–f show photographs and scanning electron microscopy (SEM) images of the Si samples, together
with their symbolic illustrations. Si(com) (Fig.1a) is crushed particles with a diameter of 1–5 µ m. By contrast,
Si(sd) (Fig.1b and c) has a aky shape since it is produced by a wire saw, and this process results in the inclu-
sion of graphite impurity (4.0 wt%). As described above, graphite is the conventional anode material for LIBs,
and therefore, the presence of such a small amount of graphite does not seriously lower the performance of the
Si materials. e elemental mapping of Si(sd) revealed a homogeneous distribution of carbon in this sample
(SupplementaryFig.S1a–c), and we cannot distinguish graphite particles from the Si particles, suggesting that
graphite exists as a very ne powder. Despite the very dierent morphologies between Si(com) and Si(sd), their
milled products, Si(com)-ba(w) (Fig.1d) and Si(sd)-ba(w) (Fig.1e), have almost the same shape, i.e., spherical
particles with a diameter of less than 200 nm. us, the initial Si shape does not aect the resulting morphol-
ogy of this ball milling process using water. On the other hand, the beads-milling process with the presence
of isopropanol yields nanoake morphology as is found in Si(sd)-be(ipa) (Fig.1f) and also in Si(com)-be(ipa)
(SupplementaryFig.S2a). ese ndings indicate that the particle shape of the precursor does not aect the nal
product shape in both milling processes. To examine the sole eect of a liquid medium, the ball-milling process
www.nature.com/scientificreports/
3
Scientific RepoRts | 7:42734 | DOI: 10.1038/srep42734
was applied to Si(com) with the presence of isopropanol instead of water. e resulting sample, Si(com)-ba(ipa)
has intermediate morphology between nanosphere (Fig.1d,e) and nanoake (Fig.1f and SupplementaryFig.S2a),
as is found in SupplementaryFig.S2b. Thus, isopropanol makes the resulting morphology more like flake.
Considering the fact that there is no great dierence between viscosities of isopropanol (1.77 mPa·s at 30 °C) and
water (0.78 mPa·s at 30 °C), the eect of the liquid media is ascribed to their dierent chemical properties rather
than the rheological properties. Single Si crystal has several cleavage planes such as (111), (100), and (110), in
which the crystal tends to be smoothly cut. Si(sd) is derived from a single crystal ingot, and therefore, it is no
wonder that Si(sd) is crushed into ake-like shape rather than spherical. is could be the case of isopropanol. On
the other hand, in the case of water, it intensively and immediately oxidize fresh Si surfaces that are generated by
milling. Indeed, when ball-milled Si(sd) with water was treated by HF to remove the surface SiO
2
layer, the orig-
inal sample weight was decreased down to approximately 35%. e weight loss occurs not only by SiO
2
removal
but also by technical issues such as ltration, but in the case of isopropanol, the same HF treatment (including
the same ltration process) achieves a recovery of 64%. ese results clearly indicate that water tends to oxidize
Si during the milling process much more than isopropanol, which generally works as a reducing agent. us, the
intense surface oxidation by water may hinder cleavage along the crystal planes, resulting in the formation of
spherical nanoparticles. Moreover, the type of the milling apparatus is also a key factor. e beads-milling appa-
ratus used in this work enables to apply greater shear stress than the case of the ball-milling apparatus.
e color change of the samples is worth noting. As is found from their photographs, both Si(com) and Si(sd)
are dark grey powder (Fig.1a,c). By the ball-milling treatment on Si(com), its color turned to be brown (Fig.1d),
which is generally seen in Si nanoparticles
8
. ough the particle size of Si(sd)-ba(w) is almost the same as that
of Si(com)-ba(w), the color of the former is dark grey (Fig.1e). is is owing to the presence of graphite in this
sample (4.8 wt%). It is noteworthy that a simple mixing of Si(com)-ba(w) with 5.0 wt% of graphite powder with a
motor does not change the brown color at all (SupplementaryFig.S3). e dark grey color of Si(sd)-ba(w) thus
suggests a homogeneous dispersion of the graphite. Indeed, the elemental mapping with energy dispersive X-ray
spectrometer (EDX) proved the very uniform distribution of carbon in this sample (SupplementaryFig.S1d–f).
Similarly, Si(sd)-be(ipa) has also grey color (Fig.1f) and 5.4 wt% of graphite is uniformly distributed also in this
sample (SupplementaryFig.S1g–i).
e specic surface areas of Si(com)-ba(w), Si(sd)-ba(w), and Si(sd)-be(ipa) are 57, 58, and 59 m
2
g
−1
, respec-
tively. By the assumption of spherical particle shape, their average particle sizes can be calculated as 45, 44, and
44 nm. e former two values almost agree with the sizes observed in their SEM images (Fig.1d,e), since they
have spherical morphology. On the other hand, the nanoake diameter of Si(sd)-be(ipa) is estimated to be about
0.2–1 µ m from its SEM image (Fig.1f). With the specic surface area and the nanoake diameter, the thickness of
the nanoake is therefore estimated as 15–17 nm, which is extremely thin. e actual particle-size distributions
of these samples were analysed by a laser diraction technique, as shown in Fig.2a. Generally, nanoparticles are
likely to aggregate and form secondary particles. Si(com)-ba(w) and Si(sd)-ba(w) thus have a major peak around
0.1–0.3 µ m, corresponding to the secondary particle sizes. On the other hand, Si(sd)-be(ipa) has a major peak
around 0.1–0.3 µ m with a shoulder around 0.3–1.0 µ m. Since Si(sd)-be(ipa) has the nanoake morphology, each
of nanoakes should be easily stacked to form thicker secondary akes, while the stacking does not increase the
diameter of the akes. Indeed, the range of the shoulder well accords to the diameter of the sample observed by
SEM (Fig.1f). erefore, the major peak and the shoulder can be ascribed to the thickness and the diameter of
Figure 1. Photographs, SEM images, and symbolic illustrations of (a) Si(com), (b,c) Si(sd), (d) Si(com)-ba(w),
(e) Si(sd)-ba(w), and (f) Si(sd)-be(ipa). e photographs and the symbolic illustrations are shown as upper and
lower insets, respectively, in (a,c–f).
www.nature.com/scientificreports/
4
Scientific RepoRts | 7:42734 | DOI: 10.1038/srep42734
the stacked nanoakes, respectively. us, Si(com)-ba(w) and Si(sd)-ba(w) have the secondary particle size of ca.
0.1–0.3 µ m by weak physical aggregation of spherical nanoparticles, while Si(sd)-be(ipa) has a similar dimension
as a continuous framework of nanoake. As shown later, the latter exhibits better performance.
XRD patterns of Si(sd) and its milled samples are shown in Fig.2b. Si(sd) shows sharp peaks of Si, together
with weak and broad peaks around 25–35° and 45–60°, which are not observed in Si(com)
22
. e broad peaks
suggest the presence of very small crystalline Si in Si(sd). Additionally, there is also a peak of graphite (002), but
no other crystalline impurities are detected. us, the present washing process for the raw Si-sawdust sludge suc-
cessfully removes inorganic impurities except for graphite. In the milled samples, the sharp peaks of Si become
much broader, which accords to the very small primary particle sizes of these samples estimated from the SEM
images and the specic surface areas. e crystallite size was calculated for each of the Si peaks by the Scherrer
equation, and the result is shown in SupplementaryTableS1. e crystallite size of Si(sd)-be(ipa) is in the range
of 10–30 nm, and this is much smaller than the nanoake diameter which is estimated by specic surface area and
SEM. us, the nanoake is found to be a polycrystalline solid. In Si(sd)-ba(w), the graphite (002) peak is almost
lost, indicating the breaking down of the graphite crystallite. On the other hand, Si(sd)-be(ipa) retains the sharp
peak of graphite. As shown before, the beads milling with isopropanol yields aky Si particles, and therefore,
aky graphite would not be broken down by this type of milling. e presence of the well-dispersed graphite
(SupplementaryFig.S1i) is advantageous to achieve a lower inner resistance for LIB application.
Charge/discharge performance of Si nanoparticles. Charge/discharge performance of Si nanoparti-
cles were examined by a 2032-type coin cell. A counter electrode was Li foil, and an electrolyte was 1 M LiPF
6
in
a mixture of ethylene carbonate and diethyl carbonate (1:1 by volume ratio). See “Method I” in the experimental
section as for the details. Figure3 shows charge/discharge (delithiation/lithiation) capacities and coulombic e-
ciencies of Si nanoparticles during 100 cycles without and with capacity restriction (see SupplementaryFig.S4a–c
for the original charge/discharge curves of the samples). All of the curves are basically similar to those in the
typical Si-nanoparticle materials: lithiation occurs at 0–0.3 V, and delithiation at 0.2–0.6 V
14,21,22,36
. Si(sd)-be(ipa)
Figure 2. (a) Particle-size distributions of the milled samples measured by the laser diraction technique.
(b) XRD patterns of Si(sd) and its milled samples.
Figure 3. Charge/discharge capacities (open/solid symbols) and coulombic eciency (solid lines) vs. cycle
number. (a) Without and (b) with discharge-capacity restriction up to 1500 mAh g
−1
measured in 1 M LiPF
6
in
a mixture of ethylene carbonate and diethyl carbonate (1:1 by volume).
www.nature.com/scientificreports/
5
Scientific RepoRts | 7:42734 | DOI: 10.1038/srep42734
exhibits superior cyclability and rate capability to the other two samples both in Fig.3a,b. is could be ascribed
to the presence of the highly dispersed aky graphite and/or the nanoake morphology of Si. To examine which
factor is important, a counterpart sample which has the same nanoake morphology but does not contain graph-
ite was prepared by the beads-milling of Si(com) (Si(com)-be(ipa) in SupplementaryFig.S2a). e charge/dis-
charge performance of Si(com)-be(ipa) is shown in SupplementaryFig.S5, which exhibits good performance
similar to that of Si(sd)-be(ipa) even without graphite, indicating that the nanoake morphology is more signif-
icant to achieve a high performance. e performance of Si(com)-be(ipa) is comparable to that of CVD-derived
Si (Si(CVD))
21
. e morphologies of Si(CVD), Si(sd)-ba(w), and Si(sd)-be(ipa), are illustrated in Fig.4a–c,
respectively. ey are carbon-black-like networking nanoparticles (Fig.4a), discrete and spherical nanoparticles
(Fig.4b), and nanoakes (Fig.4c). Si(CVD) consists of nanoparticles with the diameter of ca. 80 nm, and the
nanoparticles are tightly connected to form a continuous network structure (length is ca. 1–10 µ m)
21,22
. Such
a structure, i.e., nanosized framework with a long continuous structure, has been proved to give better perfor-
mance than that of discrete nanoparticles shown in Fig.4b
22
. As found in Fig.4c, Si(sd)-be(ipa) has a continuous
framework whose dimension is comparable to Si(CVD), and the high-performance of the former material shown
in Fig.3 can be thus rationally understood. On the other hand, weak physical aggregation found in the discrete
nanoparticles (Fig.4b) is not eective to enhance the performance of Si.
Even not being critical, the presence of highly dispersed aky graphite in Si(sd)-be(ipa) improves its cyclabil-
ity, as is found its better capacity retention aer the 80th cycle than Si(com)-be(ipa) (see SupplementaryFig.S5b).
Figure 4. e relationship between the initial structures of Si (a–c) and the corresponding ones aer 100
charge/discharge cycles (d–f) under a capacity restriction of 1500 mAh g
−1
. In (a–c), D, L, and D
sec
are particle
diameter (a,b) or ake thickness (c), a length of a continuous structure, and the size of secondary aggregation
estimated by the laser diraction, respectively. (d–f) are TEM images of Si(CVD), Si(sd)-ba(w), and Si(sd)-
be(ipa), respectively, aer 100 charge/discharge cycles. Insets represent their structures: wrinkled structure
(d,f) and aggregated lump (e).