Letter pubs.acs.org/NanoLett
In Situ Study of Silicon Electrode Lithiation with X‑ray Reflectivity Chuntian Cao,†,‡ Hans-Georg Steinrück,† Badri Shyam,† Kevin H. Stone,† and Michael F. Toney*,† †
SSRL Materials Science Division, SLAC National Accelerator Laboratory, Menlo Park, California 94025, United States Department of Materials Science and Engineering, Stanford University, Stanford, California 94305, United States
‡
S Supporting Information *
ABSTRACT: Surface sensitive X-ray reflectivity (XRR) measurements were performed to investigate the electrochemical lithiation of a native oxide terminated single crystalline silicon (100) electrode in real time during the first galvanostatic discharge cycle. This allows us to gain nanoscale, mechanistic insight into the lithiation of Si and the formation of the solid electrolyte interphase (SEI). We describe an electrochemistry cell specifically designed for in situ XRR studies and have determined the evolution of the electron density profile of the lithiated Si layer (LixSi) and the SEI layer with subnanometer resolution. We propose a threestage lithiation mechanism with a reaction limited, layer-bylayer lithiation of the Si at the LixSi/Si interface. KEYWORDS: Li-ion battery, Si anode, X-ray reflectivity, in situ, solid electrolyte interphase ince the commercialization of the first rechargeable Li-ion battery (LIB) in 1991,1 LIBs have become key components of portable electronic devices, as well as in electric vehicles, military and medical equipment, backup power supplies, and even grid storage.2,3 However, the energy storage capacity of current LIBs is still too low to meet the increasing demand of key markets, such as the need for long-distance transportation via electric vehicles. As a consequence, researchers are trying to build new generations of LIBs using a variety of different materials.4 Among those materials, silicon (Si) is one of the most promising anodes for LIBs due to its high specific capacity (4200 mAh/g, corresponding to Li4.4Si),5 which is more than 10 times that of commercially used graphite (372 mAh/g). Moreover, Si has a low discharge potential versus Li/Li+ and is earth-abundant.6 Despite large research efforts, there still exist challenges to overcome for large-scale use of Si as an anode material. In particular, when Si is lithiated it undergoes a volume expansion of ca. 300%,7,8 which can cause cracking and pulverization of the Si anode, resulting in the loss of mechanical/electrical contact and subsequent capacity fading.9,10 Extensive work has been carried out to understand the Si (de)lithiation process11,12 and to improve Si electrode’s performance.13−16 Previous studies have shown that a well-defined and sharp interface between unlithiated crystalline silicon and lithiated silicon (LixSi) exists.12,17 These experiments have also highlighted that interfacial reactions and phase transformations play a significant role in the lithiation process. Some fundamental questions, however, have not been much explored so far. These include the following: How does Li+ react with Si, break up the Si lattice, and form a LixSi alloy? Is the lithiation process diffusion or reaction-rate limited? Does the lithiation front propagate
S
© XXXX American Chemical Society
into the Si lattice layer-by-layer or in a nonuniform fashion? Therefore, one of our goals is to study the lithiation reaction at the interface level in real-time with subnanometer (subnm) resolution and under realistic charge/discharge rates. A better understanding of the structural aspects of the lithiation in Si can provide fresh insights to mitigate the large volume change and loss of cyclability. Another phenomenon closely related to battery performance, irreversible capacity loss, and cyclability is the solid-electrolyte interphase (SEI) layer found at the surface of electrodes. The SEI is a chemically complex, often inhomogeneous region at the surface of electrodes, which is formed by the decomposition of the electrolyte.18 The ideal SEI layer allows lithium ion transport, stops further electrolyte decomposition on the electrode surface, and enhances the battery’s stability and cyclability.19 Accordingly, the SEI on Si has been studied using various techniques20−33 including, among others, electron microscopy, X-ray and infrared spectroscopy, as well as neutron reflectivity. However, these results are often inconsistent, particularly in terms of the SEI thickness and composition, and continue to be debated in the literature. For example, a recent SIMS study28 has shown that the SEI thickness is different for different surface conditions and under various cycling methods and can vary from 2 to 170 nm. The reported SEI layer thickness on amorphous Si varies from 40 to 240 nm.31,32 Thus, a better fundamental understanding of the SEI structure and formation as well as its properties is still missing. Received: July 14, 2016 Revised: October 13, 2016 Published: October 26, 2016 A
DOI: 10.1021/acs.nanolett.6b02926 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
Figure 1. Schematic of experimental setup. (a) XRR kinematics specific to an SEI terminated Si wafer electrode immersed in electrolyte. (b) Exemplary surface normal EDP. (c) Corresponding calculated XRR. (d) Schematic of the electrochemical cell, containing the following parts: (1) Li foil counter electrode, (2) Si wafer working electrode, (3) stainless steel screw that fixes the Si wafer and is used as the current collector, (4) PEEK frame, (5) gasket, (6) Kapton window, and (7) stainless steel frame. The inset shows the details of the electrical connection between the Si wafer and the current collector as well as the details of how the Si wafer is secured in the cell. (e) Goniometer mount and Helium chamber, in which the electrochemical cell is mounted. (f) Electrochemistry of the Si wafer electrode cycled at 25 μA/cm2. The inset shows a magnified view of the 0.05− 0.11 V region.
more detailed mechanistic model of the lithiation process in crystalline Si. During the first lithiation of Si under galvanostatic conditions, we find that the LixSi layer has a constant electron density on the lithiation plateau, and its thickness increases linearly with the charge passed; the inorganic-SEI layer starts to form between 0.8 and 0.6 V, and its thickness increases from 10 to 80 Å during lithiation. On the basis of our data, we propose a three-stage lithiation mechanism where the lithiation at the LixSi/Si interface is surface reaction limited and the growth propagates layer-by-layer. Experimental Section. X-ray Reflectivity (XRR). X-ray reflectivity (XRR) measures the intensity fraction of a monochromatic incoming X-ray beam that is specularly reflected from a sample.34−37 Specular reflection is observed within the plane of incidence at an exit angle β equal to the incoming angle α. In this setup, the scattering vector has a component solely perpendicular to the surface (z-direction), qz = 4π/λ sin α. The simplest example for the reflection of X-rays is an ideally sharp interface between two media of different electron densities. For simplicity from a vapor−medium interface and neglecting absorption, the reflected signal is the Fresnel reflectivity
We believe that the practical applications of Si anodes can greatly benefit from a better understanding of the SEI layer. In particular in situ studies can help resolve some of this disagreement, as some of the controversy may originate in the post-mortem treatment and the corresponding exposure to ambient conditions for electrodes investigated ex situ. Here, we measured the SEI growth in situ using X-ray reflectivity (XRR) and explored how SEI density and thickness change during lithiation. We also aim to understand the effect of the native oxide layer on the SEI during lithiation. We note that it has recently been suggested29,30 that the SEI consists of two subregions, a surface adjacent “inorganic” SEI and an outer “organic” SEI. The mesoporous “organic” SEI has a similar electron density as the electrolyte, while the electron density of “inorganic” SEI is higher than that of the electrolyte because of the presence of inorganic constituents such as Li2CO3, Li2O, and LiF.29 XRR is predominantly sensitive to the inorganic portion of the SEI. In this paper, we utilize in situ XRR to investigate the structural aspects of the lithiation in a single crystalline native oxide terminated Si (100) wafer. This well-defined model system is well suited to spatially track the reaction front and phase boundary between Si and LixSi and to obtain a better understanding of the LixSi layer and the SEI that form during lithiation. XRR is a surface and interface sensitive technique and can be carried out under realistic electrochemical conditions with a time resolution of minutes. This yields important insights into Si lithiation process. To our knowledge, no in situ XRR studies probing the lithiation of Si have been published. We have designed an X-ray transparent cell to perform the proposed experiments in real time during galvanostatic cycling. XRR yields subnm resolution structural insights into the surface normal electron density profile (EDP), allowing us to develop a
RF(qz) =
qz −
qz2 − qc2
qz +
qz2 − qc2
2
where qc is the critical scattering vector of total external reflection. For a flat sample, a surface film of thickness d (e.g., the SEI) will result in characteristic oscillatory reflection pattern with a period of Δqz = 2π/d called Kiessig-fringes.38 Figure 1a shows the XRR kinematics specific to our experiment. Guided by literature,29,30 we assume a system B
DOI: 10.1021/acs.nanolett.6b02926 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
Figure 2. Measured XRR data, fits and corresponding EDPs: (a,c) Measured, Fresnel-normalized (R/RF), XRR (symbols) of lithiated Si for various charge passed, and the corresponding model fits (solid lines); (b,d), the fit-derived EDPs. All curves are vertically offset for clarity. In the scheme at the bottom of (b), the colors mark the individual sublayers: light blue, electrolyte; dark blue, inorganic-SEI; brown, lithiated Si; magenta, “dense-Si”; purple, silicon. z = 0 is set at the crystalline Si and lithiated Si interface.
0.2) before mounting it onto the beamline diffractometer. Inert helium gas is flushed through the chamber to drive out Ar gas and to protect the cell from air and moisture. The X-ray beam path through the electrolyte is about 3 mm, corresponding to a transmissivity of ∼47% for 12 keV X-rays. We used a half-cell configuration with a Si wafer as the working electrode and Li metal as the counter and reference electrode. The working electrode is a (100) n-type arsenicdoped Si wafer with a resistivity between 0.001 and 0.005 Ω· cm. The wafer is 500 μm thick and is cleaved into 2 × 14 mm2 pieces. A 100 nm thick Cu film was evaporated on the back side of the wafer to ensure good contact and a uniform electric field. A 2 nm Ti adhesion layer was predeposited to improve the adhesion between the Cu film and the Si wafer. Prior to electrochemical cycling, the Si wafer was ultrasonicated in acetone and then in isopropyl alcohol for 20 min. The electrolyte used is 1 M LiPF6 in a nonaqueous solution of ethylene carbonate and dimethyl carbonate with a weight ratio of 1:1. The cell was controlled via a Bio-Logic SP150 potentiostat and cycled in galvanostatic mode at 25 μA/cm2. Results and Discussion. Electrochemistry. The Si wafer electrode was cycled at 25 μA/cm2. For a 120 nm film or 2.8 × 10−5 g/cm3 loading Si particles, this corresponds to about C/4. Electrochemistry and reflectivity data were simultaneously measured during the first lithiation, corresponding to the first discharge cycle. Figure 1f shows the potential versus charge passed profile obtained through our in situ XRR experiment. The voltage drops quickly in the beginning, followed by a small plateau at about 0.6 V. Subsequently, a gently sloping plateau, corresponding to the lithiation of Si11, is present between 0.09 and 0.066 V during 4 h. Data were taken until the film approached a thickness close to the resolvable limit for our XRR setup. XRR. In the following, we qualitatively explain the XRR data sets, present our qualitative fitting results, and then describe the details of our EDP model. Figure 2a,c shows the in situ Fresnelnormalized XRR (R/RF) of the Si wafer as a function of
consisting of Si electrode, inner (inorganic) SEI layer, outer (organic) SEI layer, and the electrolyte. An exemplary EDP profile is illustrated in Figure 1b. The minimal electron density difference between the outer SEI and the electrolyte means that XRR is only sensitive to the inner, higher density, SEI. The corresponding reflectivity is then calculated recursively via the Parratt formalism,39−41 and is shown in Figure 1c. In practice, an EDP, ρe(z), is built up by a certain number of physically meaningful layers. Each layer has a uniform but variable, electron density, thickness, and roughness. Smearing of the interfaces due to interfacial roughness in real systems is 2 2
described by a Debye−Waller-like factor of the form e−qzσj /2, where σj represents a Gaussian roughness of the jth layer.42 The calculated reflectivity is then compared with the experimental data. In an iterative process the parameters of the slabs are varied until the calculated and experimental curves match.43 In Situ Electrochemical Cell and Electrochemical Data Acquisition. We have designed a cell that can be used in an electrochemical half-cell configuration in order to perform in situ synchrotron XRR measurements of the lithiation of crystalline Si. The cell drawing is shown in Figure 1d. The positive electrode (Si wafer) is fastened on the bottom of the polyether ether ketone (PEEK) frame (3 mm thickness) by a nut and PEEK washer on the stainless steel screw, which also acts as the current collector. The screw is fastened onto a piece of copper foil, which ensures electrical connection to the Cucoated back side of the silicon wafer. The negative electrode (Li metal) is placed parallel to the Si wafer inside the cell (distance ∼3 mm) and is connected to a stainless steel screw current collector through a slit in the PEEK frame. The stainless steel frames, the Kapton windows, the fluorosilicone gasket sheets and the PEEK part are sandwiched together and fastened by screws. Subsequently, electrolyte is filled into the cell. The cell is mounted in a sealed chamber shown in Figure 1e. The chamber is mounted onto the diffractometer through its goniometer mount. We assemble and seal the cell in the chamber under argon atmosphere in a glovebox (ppm (O2) < C
DOI: 10.1021/acs.nanolett.6b02926 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
ρcharge ≳ 30 μAh/cm2 and use as an example ρcharge = 76 μAh/ cm2 to show that this model appears valid. Figure 3a shows the R/RF (symbols) and four different model we evaluated (solid lines) to describe the data. The corresponding electron densities are shown in Figure 3b in the same color.
lithiation. RF is defined as the reflectivity from an ideal Si/ electrolyte interface. We indicate the amount of charge passed at the left side of the respective data points. Figure 2b,d shows the fit-derived EDP of each data curve, in which the interface between crystalline-Si (c-Si) and amorphous-LixSi (a-LixSi) is set to z = 0, thus highlighting the lithiation front. Each XRR curve yields the EDP of the Si electrode at the time when the data was taken. From this we derive the surface layer(s) thickness, electron density and roughness. Figure 2 shows oscillatory XRR patterns and the periodicity of the fringes is mainly related to the inverse thickness of surface layer(s) as d = 2π/Δqz. A qualitative look at the obtained XRR curves in Figure 2a,c suggests the following: The first XRR data was taken at open circuit voltage and is nearly featureless, exhibiting only a very shallow maximum at ∼qz = 0.3−0.4 Å−1, indicating the presence of the native silicon oxide.44 By fitting the XRR data to a recently proposed Si/SiO2 interface model,45 we obtained a SiO2 thickness of about 16 Å. The second curve was taken when the discharge just started and is similar to the first profile. A pronounced minimum is observed in the third curve (ρcharge = 7 μAh/cm2) at 0.6 V, indicating the formation of an interfacial layer, which is likely due to the onset of the SEI formation at Ewe = 0.8−0.6 V. From ρcharge = 11−27 μAh/cm2, we observe the evolution of nonuniformly shaped oscillations in the XRR. The minima shift toward lower qz, corresponding to an increase in layer thickness, and become more pronounced with increased charge passed. Starting at approximately ρcharge = 30 μAh/cm2 (0.08 V, where the large capacity plateau starts), a new trend is apparent in the XRR. A set of evenly spaced and well-defined oscillations appear. The oscillation period decreases and the amplitude decreases slightly with charge passed until the last measurement. This shows the growth of surface layer(s) on the Si electrode with a roughness that increases slightly with lithiation. We show below that this layer is LixSi, as expected for the potential ( 0.1 V), the inorganicSEI thickness (Figure 4a) is 10−15 Å, and the electron density (Figure 4b) increases from 0.6 to 0.7 e/Å3. On the lithiation plateau (Ewe < 0.08 V), the inorganic-SEI thickness increases and saturates at ∼80 Å. The electron density decreases and finally plateaus at around 0.43 e/Å3, a value slightly higher than the electron density of the electrolyte (0.40 e/Å3). We believe E
DOI: 10.1021/acs.nanolett.6b02926 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
part of the SEI, which by definition contains everything that is between the electrode (Si/LixSi) and the electrolyte, appearing in the EDP as an increase in the SEI electron density, which is seen as a bump in EDP, indicated with arrow (1) in Figure 2a. As the inorganic−SEI layer continues to grow, and the relative concentration of LiySiOz becomes smaller in the SEI, the inorganic−SEI layer thickness increases and the electron density decreases. The inorganic components in SEI include Li2O, LiF, Li2CO3, and LiOH. A low-electron-density layer (LixSi) formed by the lithiation of native SiO255 emerges beneath the inorganic−SEI layer, the electron density of which is lower than inorganic−SEI layer and decreases with lithiation, indicated by arrow (2) in Figure 2a. The thickness of this new layer increases from 2.6 to 4.5 nm from 11 to 27 μAh/cm2. The final thickness of LixSi layer formed in this step is 4.5 nm, corresponding to a 3-fold increase of the thickness of native oxide (1.6 nm). This expansion agrees well with a previous TEM study.46 Another phenomenon appearing in Step 2 is the initial Li+ diffusion into bulk Si, which is indicated by the small density drop from 0.71 e/Å3 (c-Si electron density) to 0.59 e/Å3 in the top 10−20 nm range in bulk Si. We speculate that Li+ ions diffuse into Si and cause volume expansion, forming a less dense LixSi layer which we denote as initial-LixSi layer (Figure 5b). This is called “initial lithiation” as compared to the lithiation process in Step 3, because the electron density of the fully lithiated LixSi formed on the plateau (Step 3) is much smaller at 0.36 e/Å3. Note that the voltage in step 2 has not reached the lithiation plateau yet. The electron density of initial-LixSi layer is larger than that of the LixSi layer formed during lithiation plateau. After the native oxide is lithiated, the bulk Si starts to fully lithiate (Figure 5c) at 0.08−0.09 V. Step 3 starts from ρcharge = 34 μAh/cm2 (Ewe = 0.08 V). In Step 3, the thickness of LixSi increases linearly with charge, while the electron density remains constant, slightly larger than that of Li15Si4. The inorganic-SEI continues to grow, but at ρcharge > 80 μAh/cm2 (Ewe = 0.07 V), the inorganic−SEI thickness appears to plateau (Figure 4a). Furthermore, we observe a dense-Si layer at the junction between Si and LixSi in which a small amount of Li ions have diffused into the Si host lattice, suggesting the existence of a Li0.17Si prealloy. A similar interface layer between crystalline Si and amorphous LixSi has been observed in previous TEM results.12,17 While we have broken the lithiation process into three separate steps, this is somewhat of a simplification and there is likely a more continuous changes in the Si(001) structure during lithiation. On the basis of our results, we propose the following model for the native oxide terminated Si lithiation process: On the lithiation plateau, when Li+ diffuses to the Si/LixSi interface, Li+ first intercalates into several Si layers forming a dense-Si layer. As more Li+ diffuses into the dense-Si, the Si lattice starts to distort; Si−Si bonds break, Li−Si bonds form, and the product LixSi forms. The thickness of dense-Si layer is about 4 nm, and thus only 7−8 crystalline Si layers participate in this process. Moreover, the LixSi electron density is very close to the fully lithiated c-Li15Si4, so the Si/LixSi reaction front moves forward only when those reacting layers are nearly fully lithiated. Therefore, we propose the lithiation of Si on the plateau to be a layer-by-layer process, consistent with a previous TEM result.12 The lithiation front is remarkably uniform at the LixSi/Si interface, and the roughness of dense-Si layer is sharp (
