A Yolk-Shell Design for Stabilized and Scalable Li-Ion Battery Alloy
Anodes
Nian Liu,
†,⊥
Hui Wu,
‡,⊥
Matthew T. McDowell,
‡
Yan Yao,
‡
Chongmin Wang,
§
and Yi Cui*
,‡,∥
†
Department of Chemistry and
‡
Department of Materials Science and Engineering, Stanford University, Stanford, California 94305,
United States
§
Environmental Molecular Sciences Laboratory, Pacific Northwest National Laboratory, Richland, Washington 99352, United States
∥
Stanford Institute for Materials and Energy Sciences, SLAC National Accelerator Laboratory, 2575 Sand Hill Road, Menlo Park,
California 94025, United States
*
S
Supporting Information
ABSTRACT: Silicon is regarded as one of the most
promising anode materials for next generation lithium-ion
batteries. For use in practical applications, a Si electrode must
have high capacity, long cycle life, high efficiency, and the
fabrication must be industrially scalable. Here, we design and
fabricate a yolk-shell structure to meet all these needs. The
fabrication is carried out without special equipment and mostly
at room temperature. Commercially available Si nanoparticles
are completely sealed inside conformal, thin, self-supporting
carbon shells, with rationally designed void space in between
the particles and the shell. The well-defined void space allows
the Si particles to expand freely without breaking the outer carbon shell, therefore stabilizing the solid-electrolyte interphase on
the shell surface. High capacity (∼2800 mAh/g at C/10), long cycle life (1000 cycles with 74% capacity retention), and high
Coulombic efficiency (99.84%) have been realized in this yolk-shell structured Si electrode.
KEYWORDS: Silicon nanoparticle, Li-ion battery, anode, yolk-shell, solid-electrolyte interphase, in situ TEM
E
lectrochemical energy storage h as become a critical
technology for a variety of applications, including grid
storage, electric vehicles, and portable electronic devices. The
lithium-ion battery (LIB) is an attractive energy storage device
because of its relatively high energy density and good rate
capability. To further increase the energy density for more
demanding applications, however, new electrode materials with
higher specific and volumetric capacity are required. Since the
initial commercialization of the LIB two decades ago, there has
been little progress in commercializing new electrode materials
with significantly higher capacity.
1
To meet the increasing
demand for energy storage capability, novel electrode materials
with higher capacity, low cost, and the ability to be produced at
large scale are of great interest.
Alloy-type anodes (Si, Ge, Sn, Al, Sb, etc.) have much higher
Li storage capacity than the intercalation-type graphite anode
that is currently used in Li-ion batteries.
2
Among all the alloy
anodes, silicon has the highest specific capacity: Experiments
have demonstrated an initial specific capacity of >3500 mAh/g,
which is 10 times the capacity of graphite.
3
In addition, silicon
is the second most abundant element in the earth’s crust (28%
by mass), indicating its potential to be utilized in large
quantities at low cost.
4
A further benefit is that mass production
of elemental silicon is already a mature technology in the
semiconductor industry. Despite these advantages, graphite
anodes still dominate the marketplace due to the fact that alloy
anodes have two major challenges that have prevented their
widespread use. First, alloy anodes undergo significant volume
expansion and contraction during Li insertion/extraction.
2
This
volume change (∼300% for Si) can result in pulverization of
the initial particle morphology and causes the loss of electrical
contact between active materials and the electrode framework.
Second, due to the low electrochemical potential of Li
insertion/extraction (<0.5 V vs Li
+
/Li), the anode surface
becomes covered by a solid-electrolyte interphase (SEI) film,
which forms due to the reductive decomposition of the organic
electrolyte.
5
In graphite anodes, a thin passivating SEI forms
during the first few cycles, and its further formation is
terminated due to the electronically insulating nature of the
SEI.
6
In alloy anodes, however, the SEI will rupture due to the
volume change during cycling, causing the electrode surface to
be cyclically exposed to the electrolyte. This results in continual
formation of very thick SEI films, which causes the electrolyte
to be continually consumed during cycling. The formation of
SEI is further complicated by particle fracture, since fracture
creates new active surfaces for SEI growth. The excessive
growth of SEI causes low Coulombic efficiency, higher
resistance to ionic transport, and low electronic conductivity
Received: April 20, 2012
Letter
pubs.acs.org/NanoLett
© XXXX American Chemical Society A dx.doi.org/10.1021/nl3014814 | Nano Lett. XXXX, XXX, XXX−XXX
of the whole electrode, and it will eventually result in the
exhaustion of the electrolyte and dry-out of the cell.
7
Recent work has shown that reducing the size of bulk silicon
to the submicrometer scale in at least one dimension can
effectively avoid fracture and therefore improve the cycling
performance.
3,8−12
Most of these nanostructures are fabricated
by chemical vapor deposition (CVD) from a silane gas
precursor, which is expensive and difficult to scale up. Silicon
nanoparticles (SiNPs), on the other hand, are a promising
candidate because they are commercially available, industrially
scalable, and compati ble with th e current slurry coating
manufacturing process for LIB electrodes. However, conven-
tional polyvinylidene fluoride (PVDF) binder does not connect
the SiNPs well because of the dramatic volume changes and
displacement of particles during cycling. To overcome this
problem, several novel binders have been reported, such as
sodium alginate, poly(acrylic acid) (PAA), sodium carbox-
ymethyl cellulose (CMC), and conductive binder.
13−16
Even
though these binders result in well-connected electrodes and
minimal loss of active materials, the surfaces of silicon particles
are still directly exposed to the electrolyte and unstable SEI
formation remains a problem.
To tackle this problem, conformal coatings on Si anode
structures have been explored.
17,18
Though the low working
potential of the anode limits the choice of coating materials,
some coatings, including amorphous carbon and metallic
coatings, have shown good chemical stability. However, upon
the volume expansion of Si, these coatings will fracture, and the
Si surface will still be exposed to electrolyte.
19
Very recently, Si nanotubes confined in SiO
x
outer shells
(termed the double-walled Si-SiO
x
nanotube anode) have
demonstrated excellent electrochemical performance.
20
The
void space in the center and the outer SiO
x
clamping layer force
the Si tube to expand inward during lithiation. Therefore, the
SEI formed on the outer surface of the SiO
x
shell remains intact
during cycling and does not continually grow, resulting in thin
SEI films and stable cycling for thousands of cycles. Even
though the fabrication process still utilizes silane CVD, this
structure proves that this is an effective way to control the SEI
growth on Si anodes. In addition, SiNPs embedded in tubular
carbon structures have been demonstrated.
21,22
The empty
space around the SiNPs allows the SiNPs to expand without
rupturing the carbon tubes. Therefore, a thin and stable SEI on
the surfaces of the carbon tubes is maintained, which results in
stable cycling for over 200 cycles. Unfortunately, long tubes are
needed to prevent electrolyte ingress into the two ends, which
makes this initial demonstration not fully compatible with
current slurry coating manufacturing process. Also, the
distribution of the SiNPs inside the carbon tubes is not well-
defined.
Intentionally controlling the size and distribution of porosity
inside the Si NP-based electro des to allow for volume
expansion/contraction has been a challenge. The slurry drying
process generates pores inside the electrode, but they are
randomly distributed and randomly sized (Figure 1A). In other
words, even though the total empty space is enough for the Si
volume expansion, the local porosity around each individual
particle might not be sufficient to accommodate volume
expansion. Therefore, a successful design must take into
account all of the above considerations: nanostructuring of Si,
formation of a stable SEI, well-controlled pore space, and
scalable fabrication.
Here, we design a “yolk-shell” structure for a stabilized and
scalable Si anode. The structure has SiNPs (∼100 nm) as the
“yolk” and amorphous carbon (5−10 nm thick) as the “ shell”
(Figure 1B). Each SiNP is attached to one side of the carbon
shell, while there is an 80−100 nm void space on the other side.
This yolk-shell structure has several advantages for LIB alloy
anodes. First, the carbon shell is a self-supporting framework,
and the well-controlled void space between the SiNPs and the
carbon shell allows for the SiNPs to expand upon lithiation
without breaking the carbon (Figure 1C). This in turn allows
for the growth of a stable SEI on the static surface of the carbon
shell and prevents the continual rupturing and reformation of
the SEI. Second, the carbon shell is uniform and mostly free of
pinholes, which prevents the electrolyte from reaching the
SiNP surface inside the shell. Lithiation of the Si occurs by Li
diffusion through the carbon shell into the Si core. Even if there
are some minor imperfections or pinholes in the carbon shell
initially, the SEI formed on the carbon shell will fill the holes
and isolate the inside of the shell from the electrolyte with
cycling. Third, the carbon shell is both electronically and
ionically conducting, which allows for good kinetics. Fourth,
Figure 1. Schematic of the materials design. (A) A conventional slurry
coated SiNP electrode. SEI on the surface of the SiNPs ruptures and
reforms upon each SiNP during cycling, which causes the excessive
growth of SEI and failure of the battery. The expansion of each SiNP
also disrupts the microstructure of the electrode. (B) A novel Si@
void@C electrode. The void space between each SiNP and the carbon
coating layer allows the Si to expand without rupturing the coating
layer, which ensures that a stable and thin SEI layer forms on the outer
surface of the carbon. Also, the volume change of the SiNPs is
accommodated in the void space and does not change the
microstructure of the electrode. (C) A magnified schematic of an
individual Si@void@C particle showing that the SiNP expands
without breaking the carbon coating or disrupting the SEI layer on
the outer surface.
Nano Letters Letter
dx.doi.org/10.1021/nl3014814 | Nano Lett. XXXX, XXX, XXX−XXXB
unlike high-aspect-ratio nanotubes or nanowires, the yolk-shell
nanostructure is fully compatible with current slurry coating
technology. Fifth, unlike traditional slurry coated electrodes,
our Si@void@C electrode has a well-defined void space around
every Si particle, which allows for each particle to expand upon
lithiation without deforming the electrode microstructure.
We have developed a room temperature solution method to
conformally coat SiNPs first with a SiO
2
sacrificial layer and
then with a polydopamine layer, which is subsequently
carbonized to form a nitrogen-doped carbon coating. After
selectively removing the SiO
2
sacrificial layer by hydrofluoric
acid (HF) treatment, the yolk-shell Si@void@C structure is
obtained (Figure 2A,B). The whole fabrication is scalable, and
the powder-like product is fully compatible with current slurry
coating technology (Figure S1, Supporting Information). The
solution coating method was chosen instead of a solid or gas
phase method because it produces a conformal, homogeneous
coating on every individual particle (Figure S2, Supporting
Information). The homogeneity of the coating is crucial to a
successful yolk-shell design because ideally the shell should
prevent electrolyte ingress so the SEI is formed only on the
outside of the shell. While we were preparing this manuscript, a
paper was published in which a similar structure was studied,
but a gas-phase carbon coating method was used, which results
in a nonconformal coating. Uncoated (bare) Si surface is
evident from the X-ray photoelectron spectroscopy (XPS)
analysis, in which the Si2p signal is still strong in the coated
sample.
23
In addition, the cycling performance of the coated Si
is not much different than that of bare SiNPs. Therefore, a
conformal and homogeneous coating is crucial for good
electrochemical performance of the yolk-shell structure.
We have used a sol−gel method to conformally coat
amorphous SiO
2
onto the SiNPs. Because the SiNPs have
native oxide on the surface, the decomposition of tetraethyl
orthosilicate (TEOS) in ammonia solution occurs preferentially
on the SiNP surfaces.
24
By controlling the TEOS concen-
tration, pH, and coating time, the thickness of the SiO
2
coating
and therefore the void space size can be easily controlled
(Figure S3, Supporting Information).
25,26
The SiNPs utilized
for this study have an average diameter around 100 nm. Smaller
particles might yield better battery performance, but they are
more costly as well. Therefore we have chosen the
commercially available ∼100 nm particles. Assuming a 300−
400% volume expansion of crystalline Si upon complete
lithiation,
27
the sacrificial coating thickness should be >30 nm
to allow an individual SiNP to expand without rupturing the
shell. Taking into account the size variation of the SiNPs, we
control the thickness of the SiO
2
layer to be about 40−50 nm
in further experiments.
The second coating of polydopamine was conducted by the
self-polymerization of dopamine in a solution of pH 8.5 in the
presence of oxygen. This coating method has been reported to
give a highly conformal polydopamine coating with a thickness
that can be controlled by varying the coating time.
28
The silica
sol is stable in a pH window between 8 and 11, so the SiO
2
layer is stable during the polydopamine coating process.
29
The
carbon precursor has an important effect on the morphology
and properties of the resulting carbon coating. Polydopamine
has been reported to yield highly conformal and effective
carbon coatings.
30
The carbon coating obtained here is 5−10
nm in thickness (see the TEM image in Figure 2B). Its
presence is also evident from the enhanced XPS C1s peak
(Figure 2C), and it is mostly amorphous according to selected-
area electron diffraction (SAED) (Figure 2B, inset) and X-ray
diffraction (XRD) results (Figure 2E). Moreover, XPS results
show that the carbon layer is nitrogen-doped (Figure 2C); the
nitrogen originates from the dopamine precursor, and it
comprises 2.2% of the elements on the electrode surface.
This probably benefits the electrochemical performance, since
nitrogen doping has been reported to facilitate the electronic
Figure 2. Characterization of Si@void@C material and electrodes. (A) SEM and (B) TEM images of synthesized Si@void@C powder. The inset of
(A) is a magni fied SEM image of one yolk-shell particle. The inset of (B) is the SAED pattern with the diffraction rings indexed. (C) XPS spectra of
the electrode surface with active materials consisting of SiNPs and Si@void@C, respectively. Both electrodes were made under the same conditions
with carbon additive and alginate binder. (D) High-resolution XPS spectra of the Si2p peaks for the same electrodes. The Si signal is significantly
reduced in the Si@void@C electrode. In (C) and (D), both XPS spectra were collected under the same conditions without further processing. (E)
XRD pattern of Si@void@C powder. The peaks are all from crystalline Si.
Nano Letters Letter
dx.doi.org/10.1021/nl3014814 | Nano Lett. XXXX, XXX, XXX−XXXC
conductivity of the carbon layer and the charge transfer at the
interface.
31
The rigidity of the thin carbon coating allows it to
form a self-supporting shell outside the SiNP core (Figure
2A,B) after the removal of the SiO
2
sacrificial layer. Some
carbon shells contain more than one SiNP, which is due to
slight aggregation during SiO
2
coating or polydopamine
coating. It is important to note that most individual SiNPs
contact the outer shell in at least one location in the final
product, which allows for Li transport from the carbon shell
into the SiNPs during lithiation. All these fabrication steps are
carried out at room temperature except for carbonization. No
silane precursor is used, and the whole process is easily scalable.
As discussed, the key of the yolk-shell design is the self-
supporting and conformal shell that separates the SiNPs from
the electrolyte. To determine if our synthesized Si@void@C
material has a high-quality yolk-shell structure, we conducted
surface sensitive high-resolution XPS on Si2p peaks for SiNP
and Si@void@C electrodes (Figure 2D). The Si@void@C
electrode shows a negligible Si signal compared to the SiNP
electrode. The XPS elemental analysis shows that the surface
atomic percentage of Si decreases from 15.9% in the bare SiNP
electrode to 1.4% in Si@void@C electrode, while the
percentage of carbo n increases from 32.3% to 85.6%.
Therefore, in our synthesized Si@void@C material, the
SiNPs appear to be completely sealed inside the hollow carbon
shell, though there might be some nanoscale pores in the
carbon walls since HF can di ffuse into the structure to etch
away the SiO
2
. We hypothesize that this nearly completely
sealed structure is crucial for obtaining excellent electro-
chemical cycling performance. The nanoscale pores will be
blocked by the SEI formation after initial battery cycling, which
will prevent the direct contact of SiNPs with the electrolyte.
The recent application of in situ TEM to study electro-
chemical reactions has provided a powerful way to monitor the
structural changes of materials during electrochemical pro-
cesses.
32,33
Here, we use this in situ TEM technique to observe
the deformation and structural changes during lithiation of the
Si@void@C material to gain insight into the volume change
process. The in situ electrochemical cell is shown schematically
in Figure 3A. Si@void@C particles are first drop-cast onto Si
nanowires (NWs) grown on a flat Si substrate. The NWs have a
thin layer of copper thermally evaporated onto their surface to
improve electrical conductivity in the device.
19
A few NWs with
the attached Si@void@C particles are then transferred to a
metallic probe on a specialized dual-probe biasing TEM holder,
as shown at the bottom of the schematic. LiCoO
2
particles are
attached to the other metallic probe (the top of the schematic),
and a drop of ionic liquid electrolyte is placed on this electrode.
The ionic liquid has extremely low vapor pressure and therefore
does not evaporate when exposed to the high vacuum of the
TEM column. Inside the TEM, the NW is positioned so that
the NW tip is immersed in the ionic liquid. In this way, a
nanoscale electrochemical device is formed where the NW/Si@
void@C electrode is the working electrode and the LiCoO
2
is
the Li-containing counter electrode. By applying a −4 V bias to
the NW side, Li
+
is reduced and diffuses into the NW. Since the
Si@void@C particles are physically attached to the NW, the
particles also become lithiated due to Li diffusion between the
structures. While this experimental geometry is obviously
different than in an actual battery, the deformation character-
istics are expected to be similar, especially since fast surface
diffusion causes lithiation to proceed relatively uniformly in the
Si particles.
9
Figure 3B shows a series of images taken from a movie of the
in situ lithiation of Si@void@C particles (movie S1, Supporting
Information). In the first image (0 s), pristine Si nanoparticles
are visible within the surrounding C shell. The already fully
lithiated NW that is in physical contact with the Si@void@C
particles is also seen in the frame (the NW has the Li
15
Si
4
structure). In subsequent frames, the Si particles expand in
volume as Li diffuses through the carbon coating and reacts
with the Si particles. The volume expansion is most evident in
the set of particles at the bottom of the structure. In the frame
labeled “105 s”, the particles are partially lithiated, and an
amorphous Li
x
Si shell/crystalline Si core structure is readily
discernible in the largest particle (the crystalline core is the
region with dark contrast). This two-phase reaction mechanism
is well-known to occur during the lithiation of crystalline Si.
34
After complete lithiation, the diameter of the largest particle
increases from ∼185 to ∼300 nm. Fracture was not observed in
Figure 3. In situ TEM characterization of Si@void@C expansion during electrochemical lithiation. (A) Schematic of the in situ TEM device. (B) In
situ TEM image series captured from movie S1, Supporting Information. In this series of images, the silicon particles are observed to expand within
the outer carbon shell. The entire volume expansion is accommodated within the available void space, and the shell does not rupture. In addition,
these data indicate that Li transport through the carbon layer is sufficient for good rate capability. Scale bar: 200 nm.
Nano Letters Letter
dx.doi.org/10.1021/nl3014814 | Nano Lett. XXXX, XXX, XXX−XXXD
these particles. In addition, the carbon coating also becomes
lithiated; the thickness of the carbon increases from 5 to 10 to
∼20 nm after lithiation without noticeably increasing the size of
the shell. Some of the observed increase in the thickness of the
carbon shell could also be due to build-up of a thin layer of
ionic liquid electrolyte at the surface.
This in situ TEM experiment reveals important details
related to the volume changes in this structure. First, in this
configuration, Li must diffuse from the lithiated NW through
the carbon shell into the individual particles for them to
become lithiated. All the Si particles in the observation window
were lithiated (Figure 3B and Movie S1 in the Supporting
Information). The successful lithiation of the Si particles
indicates good contact between the carbon shell and Si
particles. In addition, the rate of Li di ffusion through the carbon
shell (5−10 nm) seems to be fast enough for lithiation to occur
in a reasonable time (in this experiment the particles are fully
lithiated in ∼340 s). This indicates that even though the silicon
and carbon only contact at a small area, reasonable current
densities could be supported. Second, it is clear that the carbon
coating remains intact after Si expansion even though the
expansion causes the Si particles to impinge upon the carbon
coating. This could be attributed to (i) the plastic flow of
lithiated Si so that it expands into the void space away from the
carbon shell
35
and (ii) to the fact that there is enough void
space to accommodate the full expansion of each Si particle.
Because the overall shape of the yolk-shell structure does not
change appreciably upon lithiation, it is expected that a battery
electrode made of this Si@void@C structure will undergo
minimal microstructural damage upon cycling, in contrast to an
electrode made of bare SiNPs. Finally, the void space is
carefully designed so that the Si particles occupy almost all the
void space after full lithiation, which maximizes the volumetric
energy density.
The successful design and fabrication of the yolk-shell
structure for a stabilized anode is evident from the excellent
electrochemical behavior (Figure 4). The specific capacity
values reported are calculated on the basis of the total weight of
the Si@void@C material, in which silicon comprises ∼71% of
the mass determined by thermogravimetric analysis (TGA,
Figure S4, Supporting Information). It should be noted that the
carbon shell is also lithiated, as observed in the in situ TEM
experiments (Figure 3B). Therefore, both the Si particles and
the carbon shell contribute to the capacity. Upon deep
galvanostatic cycling between 0.01 and 1 V, the reversible
capacity reaches 2833 mAh/g for the first cycle at C/10 and
stabilizes at ∼1500 mAh/g for later cycles at 1C (Figure 4A).
The specific capacity of our yolk-shell structure is much higher
Figure 4. Electrochemical cycling results for Si@void@C electrodes. (A) Delithiation capacity and CE of the first 1000 galvanostatic cycles between
0.01−1 V (alginate binder). The rate was C/10 for one cycle, then C/3 for 10 cycles, and 1C for the later cycles. (B) Voltage profiles plotted for the
first, 250th, 500th, 750th, and 1000th cycles. (C) Galvanostatic cycling of different silicon nanostructures (PVDF binder). All samples were cycled at
C/50 for the first cycle, C/20 for the second cycle, and C/10 for the later cycles. (D) SEM image of Si@void@C electrode after 800 electrochemical
cycles, showing the Si−C yolk-shell structure coated with a uniform thin SEI layer. (E) Delithiation capacity of Si@void@C with alginate binder
cycled at various rates from C/10 to 4C.
Nano Letters Letter
dx.doi.org/10.1021/nl3014814 | Nano Lett. XXXX, XXX, XXX−XXXE
