In situ electrochemical high-energy X-ray diffraction using a capillary working electrode cell geometry

The ability to generate new electrochemically active materials for energy generation and storage with improved properties will likely be derived from an understanding of atomic-scale structure/function relationships during electrochemical events. Here, the design and implementation of a new capillary electrochemical cell designed specifically for in situ high-energy X-ray diffraction measurements is described. By increasing the amount of electrochemically active material in the X-ray path while implementing low-Z cell materials with anisotropic scattering profiles, an order of magnitude enhancement in diffracted X-ray signal over traditional cell geometries for multiple electrochemically active materials is demonstrated. This signal improvement is crucial for high-energy X-ray diffraction measurements and subsequent Fourier transformation into atomic pair distribution functions for atomic-scale structural analysis. As an example, clear structural changes in LiCoO₂ under reductive and oxidative conditions using the capillary cell are demonstrated, which agree with prior studies. Accurate modeling of the LiCoO₂ diffraction data using reverse Monte Carlo simulations further verifies accurate background subtraction and strong signal from the electrochemically active material, enabled by the capillary working electrode geometry.

1. Introduction

Electrochemically active materials are poised to enable significant technological advancements in energy conversion and energy storage through the development of catalysts and electrode layers in fuel cells, batteries and supercapacitors (Winter & Brodd, 2004; Nie et al., 2015; Goriparti et al., 2014; Roy & Srivastava, 2015; Zhi et al., 2013). While development to date is encouraging, new materials with improved properties are needed to compete with and eventually supplant current technologies. Rigorous structure/function relationship development is expected to enable rational design of materials with improved electrochemical properties. In particular, understanding atomic-scale structural changes during electrochemical events can help establish a mechanistic understanding of electrocatalytic reactions and/or charge storage phenomena, as well as provide insights into material longevity (Young et al., 2015, 2016). Recent developments using in situ liquid cell transmission electron microscopy (TEM) have demonstrated structural, chemical and morphological changes of active materials under applied bias (Gu et al., 2013; Abellan, Mehdi et al., 2014; Zhu et al., 2014). Though these studies are insightful, electron microscopy is statistically limited by both the number of measurements that can reasonably be performed and the amount of material subsequently characterized, and thus may not represent the materials properties observed on an ensemble average collection of material. Additionally, electron-beam-induced processes are often observed and can complicate subsequent analysis (Abellan, Woehl et al., 2014; Woehl et al., 2013).

To obtain a more statistically accurate representation of atomic-scale structure of a material during electrochemical events, in situ X-ray based characterization methods have proven to be greatly beneficial. In particular, in situ electrochemical X-ray absorption spectroscopy (XAS) has been used to correlate changes in local atomic coordination and bond lengths to electrochemically driven transformations (Gorlin et al., 2013; Nowack et al., 2016; Becknell et al., 2015). However, XAS can only provide structural information within the first coordination sphere of the probed element, potentially complicating subsequent structural modeling. For example, face-centered cubic and hexagonal close-packed lattices exhibit a first coordination sphere of 12 and thus are indistinguishable with XAS. High-energy X-ray diffraction (HE-XRD) coupled with atomic pair distribution function (PDF) analysis can provide much longer-range (>10 nm) (Proffen et al., 2003) structural information than XAS through the Fourier transformation of diffraction data taken at sufficiently high reciprocal space vectors into a function of real-space atomic pair distances. Atomic modeling of PDFs derived from HE-XRD patterns can then be used to assess structure/function relationships (Prasai et al., 2015; Petkov et al., 2008; Bedford et al., 2015; Doan-Nguyen et al., 2014). PDF analysis can be performed on any material irrespective of the level of long-range structural coherence as both diffuse and Bragg components of the diffraction pattern are considered (Billinge & Kanatzidis, 2004; Petkov, 2008). This is particularly advantageous for nanoscale materials, which often exhibit both Bragg-like and diffuse features in their corresponding diffraction patterns due to the lack of long-range order which is inherent to nanomaterials.

Despite the valuable information that can be gained from HE-XRD measurements, fewer in situ electrochemical studies have been performed using HE-XRD than XAS. This can be partially explained by the challenge of designing in situ cells for HE-XRD measurements versus for XAS measurements. XAS is element-specific, and the background signal from the cell can be minimized by appropriately choosing the cell material and electrolyte with little need to optimize the cell geometry. In contrast, during HE-XRD measurements, everything in the path of the incident X-rays (both Bragg and diffuse scattering materials) contributes to the raw diffraction pattern. As such, the signal from the electrochemically active material over the background scales with the relative quantity of the active material and its relative X-ray scattering factor compared with all the components (electrolyte, substrate, electrical contacts, etc.) present in the X-ray path through the electrochemical cell. Reducing diffraction from these other components complicates in situ cell design and subsequent experimentation.

In situ electrochemical PDF analysis has been reported, however, most notably from in situ battery cycling experiments from Sector 11 at the Advanced Photon Source (Hu et al., 2013; Jung et al., 2015; Chupas et al., 2008; Borkiewicz et al., 2015) and recent in situ electrocatalytic experimentation reported by Professor Valeri Petkov and colleagues (Petkov et al., 2016; Tuaev et al., 2013; Wu et al., 2015). While these reports illustrate the utility and importance of atomic PDF analysis for in situ electrochemical experimentation, readily available in situ cells (Figs. 1a, 1b) have not been optimized for HE-XRD measurements, and thus the material systems investigated have been limited to those with sufficient scattering to obtain usable HE-XRD data for PDF analysis. Depending on the particular application and active material, experimentation with existing in situ electrochemical cells may not be feasible. In addition, attempts to increase the amount of active material can result in poor electrical connectivity to the active material and/or a larger fraction of electrochemically inactive materials, making the interpretation of in situ data sets more problematic.

Figure 1 Prototypical in situ electrochemical cells: (a) a three-electrode half-cell configuration; and (b) coin cell configuration for battery cycling experiments.

In response to the aforementioned issues associated with in situ electrochemical HE-XRD measurements, we report on a new in situ electrochemical cell design based on an aligned carbon fiber capillary working electrode. This geometry consists of a three-electrode half-cell and improves sample scattering by maximizing sample amount in a capillary environment, similar to commonly performed ex situ HE-XRD measurements. Electrolyte volume is minimized to reduce background scattering and constantly recirculated to ensure constant replenishment of reactive species. In addition, the use of an aligned carbon fiber capillary working electrode takes advantage of tight manufacturing tolerances to provide a consistent background between measurements, and results in a largely oriented background scattering that can be readily masked out of the diffraction pattern during image processing without losing structural information at equivalent q-space values. Using various examples, we illustrate the improvement in signal to noise using this capillary cell versus comparable traditional in situ cells, and demonstrate the ability to probe electrochemically active materials in situ using PDF analysis.

2. Experimental

2.1. X-ray measurements

HE-XRD data reported in this paper were collected at 11-ID-B, 11-ID-C and 6-ID-D at the Advanced Photon Source (APS), Argonne National Laboratory, using 58.65, 105 and 100 keV irradiation, respectively. All experiments were performed in a transmission geometry with a sample-to-detector distance to maximize the sampled q-range. Where applicable, HE-XRD data were background-corrected, converted into total structure functions and Fourier transformed into their corresponding atomic PDFs using PDFgetX3 (Juhás et al., 2013). Reverse Monte Carlo (RMC) calculations were performed to analyze the PDFs collected for LiCoO₂ using the fullrmc Python package (Aoun, 2016).

2.2. Materials

Fe–Ni layered double hydroxide (LDH) aqueous nanoparticles suspensions were synthesized using co-precipitation (Cavani et al., 1991) of FeCl₃ (Spectrum; minimum 98%) and NiCl₂·6H₂O (Amresco; high purity) in 0.15 M NaOH (EMD Millipore; ACS grade) under a continuous argon purge (AirGas; UHP), resulting in a nanoparticle concentration of ∼1 wt%. For use in the half-cell design in Fig. 1(a), Fe–Ni LDH nanoparticles were dropcast onto carbon paper (see below). For use in the capillary cell in Fig. 2, Fe–Ni LDH nanoparticles were lyophilized and crushed to obtain a powder. Pt nanoparticles on multi-walled carbon nanotubes (MWCNTs) or graphene sheets were synthesized by chemical co-reduction of potassium tetra­chloro­platinate(II) (K₂PtCl₄, ≥99.9%, Sigma Aldrich) along with graphene oxide (GO) sheets or surface-oxidized MWCNTs in aqueous ethyl­ene glycol (EG) solutions (Li et al., 2010; Hu et al., 2012). Here, the pristine GO nanosheets were synthesized using the well-established modified Hummers' methods (Xue et al., 2012; Kovtyukhova et al., 1999). The surface-oxidized MWCNTs were synthesized via direct oxidation with HNO₃ (Xing, 2004). In brief, the uniformly dispersed K₂PtCl₄/GO/EG or K₂PtCl₄/surface-oxidized MWCNT/EG solutions with an adjusted pH of 13 were heated at 120°C for 4–12 h under magnetic stirring. After the reduction reaction, the final Pt–graphene or Pt–MWCNT composites were filtered and thoroughly washed with ethanol and de-ionized water, and further dried in a vacuum desiccator. LiCoO₂ powder (97%) was purchased from Alfa Aesar and used as received.

Figure 2 Schematic of the capillary in situ electrochemical cell, including: (a) an overall schematic of the cell, containing the ep­oxy-aligned carbon fiber capillary working electrode connected to an electrolyte reservoir containing the counter electrode, and reference electrode. Electrolyte is recirculated via peristaltic pump to replenish the electrolyte species; (b) the components connecting the working electrode to the electrolyte reservoir; and (c) schematic of the interior of the working electrode, containing amorphous carbon powder support for the electrochemically active material.

2.3. Electrochemical measurement

During in situ electrochemical HE-XRD experimentation, electrochemical measurements were performed in a three-electrode configuration using a SP-300 potentiostat (Biologic), controlled with EC-Lab software (Biologic). A saturated Ag/AgCl reference electrode (BASi) and graphite rod counter electrode (Princeton Applied Research) were employed in both the half-cell and capillary working electrode cells described below. Unless otherwise indicated, a 1 M NaOH electrolyte was used for electrochemical measurements, and a continuous oxygen purge was maintained throughout operation.

3. Cell design and operation

3.1. Traditional cell designs

As described in the Introduction, many of the existing cells used to collect in situ electrochemical HE-XRD data have not been optimized for X-ray diffraction. For instance, the three-electrode half-cell depicted in Fig. 1(a) was designed for XAS measurements (McBreen et al., 1987). In this cell, the active material is commonly dropcast or pressed into a low-Z substrate resulting in a total active material thickness of the order of 10–100 µm. Here, we employ a non-woven carbon fiber paper (Fuel Cell Earth; Toray carbon paper, thickness ∼300 µm) substrate. This carbon fiber paper is compressed against a coiled Au wire by bolting Kapton-covered aluminium windows on either side of the cell. For fluorescent XAS measurements, the active material is placed adjacent to a large window in the aluminium plate, and the cell is angled between the X-ray beam and the fluorescent detector. This cell is well designed for this purpose, and allows for convenient operation and good signal in fluorescent XAS measurements. However, when used for X-ray diffraction measurements, the half-cell's limited capacity for active material and comparatively long X-ray path length through the electrolyte and carbon fiber paper produce low signal with diffuse noise arising from the electrolyte. Furthermore, the Kapton windows in the cell are flexible, and conform to the sample, making it difficult to obtain a consistent background between X-ray diffraction measurements (Borkiewicz et al., 2012). These aspects of the design are inconsequential for fluorescent XAS measurements, but make high-resolution HE-XRD measurements challenging, and in many cases preclude the use of PDF analysis.

Another common implementation configures the electrochemical cell geometry such that the active component under investigation is stacked with the remaining device materials, such as a membrane electrode assembly for electrocatalyst testing (Sasaki et al., 2016; McBreen et al., 1987) or coin cell for battery/supercapacitor testing (Richard et al., 1997). The coin cell design, as depicted in Fig. 1(b), emerged from the prevalence of coin cells in lithium ion battery research (Richard et al., 1997) and was also not designed to optimize the X-ray diffraction signal. This cell uses conventional coin cell hardware, and adds low-Z windows (e.g. Kapton) to the coin cell casing, and center holes to the current collectors to allow for X-ray transmission. For the battery community, this design allows for convenient construction of an in situ X-ray diffraction cell. However, this cell geometry also limits the amount of active material which can be loaded into the cell, and the flexible Kapton windows lead to the same difficulties for obtaining a consistent background between measurements as discussed above (Borkiewicz et al., 2012). This design also requires that the diffraction signal from the counter electrode be subtracted for the measurements, or that a window be cut through the counter electrode. While many of these issues have been addressed with recent improvements to this cell design (Borkiewicz et al., 2012; Wilson et al., 2014; Roberts & Stewart, 2004; Braun et al., 2003), the coin cell geometry limits the possible electrochemical systems which can be studied. The coin cell geometry does not conveniently allow for the use of a reference electrode, and requires that the process taking place on the working electrode be balanced by an equivalent process on the counter electrode.

3.2. Capillary working electrode cell

Here, we report a new electrochemical cell with a capillary working electrode designed specifically for in situ X-ray diffraction measurements. The working electrode design is inspired by the capillaries commonly used in ex situ HE-XRD measurements, and allows for a higher mass of active material and minimal electrolyte volume in the X-ray path. This provides dramatic signal enhancement over existing cell designs and enables PDF analysis for electrochemical systems where previous cell designs have proven to be problematic. A schematic of the cell is presented in Fig. 2. Fig. 2(a) shows the overall layout of the cell, with a counter and reference electrode in an electrolyte reservoir, and the capillary working electrode at the base of the reservoir. Fresh electrolyte is continually supplied to the working electrode by pulling electrolyte from the reservoir through the working electrode using a peristaltic pump, as depicted in Fig. 2(a). Recirculation ensures that a continuous supply of electrolyte is available for lengthy measurements. The electrolyte may also be collected for analysis rather than recirculated.

An exploded view of the working electrode is shown in Fig 2(b). The outer wall of the working electrode is a low-Z electrically conductive capillary. Here, we employ an ep­oxy-aligned carbon fiber capillary (0.125" OD; ACP Composites). Tight manufacturing tolerances [here 0.125" + 0.000"/0.003" OD and 0.072" ± 0.003" ID (ACP Composites Inc., 2016)] on this and equivalent carbon fiber capillaries provide a consistent background between measurements, while the aligned carbon fibers result in a largely anisotropic background, which is readily masked out during processing of the diffraction data. The capillary working electrode is attached to the electrolyte reservoir with a compression fitting. The compression fitting allows for rapid swapping of capillaries, enabling more efficient operation when analyzing an array of samples. The carbon fiber capillary is electrically isolated from the compression fitting using an electrically insulating O-ring. The active material in the carbon fiber capillary rests on a porous frit, here a polyethyl­ene frit with 20 µm porosity, which is held in place using fitted rubber tubing. The specific technique for loading the capillary can be tuned for a given material system. Here, we chose to add conductive granular activated carbon (Norit GAC 400M-1746) to the carbon fiber capillary, and then added active material on top of the carbon as depicted in Fig. 2(c). The granular activated carbon provides a porous network for the electrolyte to flow through the carbon fiber capillary, and also provides electrical connectivity to the active material. Additional granular activated carbon contributes minimal background to the resulting HE-XRD pattern. An electrical connection was made to the working electrode by attaching an alligator clip to the exterior of the carbon fiber capillary just below the compression fitting.

In order to optimize the diffraction signal from the active material against the electrical connectivity from the carbon powder, we aligned the X-ray beam just below the interface between the carbon powder and active material as depicted in Fig. 2(c). A photograph of a prototype capillary working electrode cell is depicted in Fig. 3. Alignment was performed by scanning the cell in the vertical direction (parallel to the axis of the capillary tube) while using a photodiode detector as shown in Fig. 3 to measure the direct X-ray beam intensity (i.e. unscattered X-rays). The photodiode showed a high X-ray intensity when scanning the X-ray beam over the low-Z granular activated carbon, and a drop in X-ray intensity when the X-ray beam was positioned over the higher-Z active material. The vertical cell position was adjusted to the inflection point in the photodiode signal to ensure that a blend of active material and conductive carbon support were in the X-ray path. Finally, the cell position was fine-tuned with the photodiode removed to optimize the scattering from the active material as measured directly from the X-ray area detector.

Figure 3 Photograph of a prototype in situ electrochemical cell with capillary working electrode in operation at 6-ID-D beamline at APS, showing (1) two-dimensional X-ray area detector, (2) photodiode, (3) carbon fiber capillary working electrode, (4) compression fitting, (5) electrolyte reservoir, and (6) ports with counter electrode, reference electrode and tubing for electrolyte recirculation.

4. Results and discussion

To demonstrate the enhancement in signal from the electrochemically active materials and the ability to perform in situ electrochemical HE-XRD measurements to obtain subsequent atomic PDFs using the capillary working electrode cell, we perform HE-XRD measurements on a series of example materials. We contrast the X-ray diffraction signal obtained with our capillary working electrode cell geometry (Figs. 2 and 3) against the signal obtained using the traditional half-cell geometry (Fig. 1a) for lower-Z transition-metal hydroxide nanoparticles (Fe–Ni LDHs) and high-Z platinum nanoparticles on nanocarbon supports. While these two material systems serve the purpose of contrasting the diffraction signal over the background between the cell geometries, atomic structure changes during electrochemical operation have not been well characterized for these materials. In order to verify the ability to accurately determine structural changes under electrochemical operation using the capillary working electrode cell, we perform HE-XRD measurements on commercial LiCoO₂ powder using the capillary working electrode cell and compare the structural changes we observe with prior studies on LiCoO₂.

In general, LDH nanoparticles have shown promise for a range of applications including energy conversion (Sels et al., 1999; Song & Hu, 2014), energy storage (Chen et al., 2014; Gong et al., 2014) and drug delivery (Choy et al., 2000). Here, electrocatalytic Fe–Ni LDH nanoparticles (Long et al., 2014, 2015) are used to showcase the improvement of active material scattering visually from the transmitted X-ray diffraction patterns in Fig. 4. For use in a traditional half-cell geometry (Fig. 1a), Fe–Ni LDH nanoparticles were deposited onto a non-woven carbon fiber working electrode for in situ electrochemical experiments. The diffraction patterns for a bare carbon paper, and carbon paper with Fe–Ni LDH nanoparticles in the assembled half-cell geometry with electrolyte are shown in Figs. 4(a) and 4(b), respectively. The white object oriented diagonally in Figs. 4(a) and 4(b) is a beam-stop. Only a faint diffraction ring arising from the Fe–Ni LDH nanoparticles can be seen around the center black circle in Fig. 4(b). This diffraction ring has an intensity which is only 2.7% larger than the background signal. The minimal differences observed in the diffraction patterns in Figs. 4(a) and 4(b) highlight the poor signal from the active material in the traditional half-cell geometry. Given the relative amount of Fe–Ni LDH nanoparticles in comparison with the background materials (Kapton windows, carbon fiber working electrode and electrolyte), the small signal from the active material is to be expected. Only when nanomaterials containing sufficiently high-Z elements are used will the diffraction signal be measureable above the background scattering of the cell (such as Pt, see below). This severely limits the use of this traditional configuration to study materials based on first-row transition metals, which is a staggering limitation given the earth abundance of first-row transition metals, and their corresponding prevalence in emerging electrochemical research (Cottineau et al., 2006; Poizot et al., 2000; Burke et al., 2015).

Figure 4 HE-XRD patterns for Fe–Ni LDH nanoparticles using two different in situ cell configurations: (a) conventional three-electrode half-cell without Fe–Ni LDH nanoparticles; (b) conventional three-electrode half-cell with Fe–Ni LDH nanoparticles; (c) aligned carbon fiber in situ cell without Fe–Ni LDH nanoparticles; and (d) aligned carbon fiber in situ cell with Fe–Ni LDH nanoparticles. HE-XRD from (a) and (b) were obtained at 11-ID-B (58.65 keV) while (c) and (d) were obtained at 6-ID-D (100 keV).

To evaluate the signal improvement with the capillary working electrode geometry, Fe–Ni LDH nanoparticles were loaded into an aligned carbon fiber capillary with granular carbon and inserted into the capillary working electrode cell, as described above. The resulting diffraction patterns for an empty carbon capillary and a capillary with Fe–Ni LDH nanoparticles in the assembled capillary working electrode cell geometry with electrolyte are shown in Figs. 4(c) and 4(d), respectively. The capillary working electrode geometry yielded dramatically improved scattering from the nanoparticles over the background of the cell. Multiple diffraction rings from the Fe–Ni LDH nanoparticles are evident in Fig. 4(d). The primary diffraction ring at 2.4 Å⁻¹ has an intensity which is 38% larger than the background signal, corresponding to a 14-fold signal improvement over the half-cell geometry. The capillary working electrode geometry results in minimal background scattering while increasing the path length of active material probed by the X-rays, such that sufficient signal can be extracted from the background for subsequent PDF analysis. Furthermore, the Bragg diffraction from the carbon fiber working electrode in Fig. 4(c) is oriented due to the alignment of carbon fibers in the as-processed capillaries. The most visible component of this Bragg diffraction is two vertically elongated black ovals seen in the horizontal diffraction direction, which are symmetric about the center of the pattern. The oriented background from the aligned carbon fibers is also visible in Fig. 4(d). This largely oriented background can easily be masked out of the diffraction pattern during image processing, allowing for improved background subtraction, which is particularly useful when examining materials with similarly positioned diffraction peaks to carbon fibers.

Examination of Pt nanoparticle scattering from both cell geometries in Fig. 5 further demonstrates the improved scattering achievable with the capillary working electrode geometry. The diffraction patterns in Fig. 5 are integrated via image processing from diffraction patterns analogous to those presented in Fig. 4. Pt nanoparticles are established electrocatalysts (Curnick et al., 2012; Lee et al., 2006) and serve as a good comparison material between the electrochemical cells given that Pt is a high-Z material that strongly scatters X-rays. In Fig. 5(a), the scattering intensity signal is shown to be equivalent for Pt on MWCNTs and graphene when measured ex situ in Kapton capillaries with equivalent amounts of probed material. Using a three-electrode half-cell geometry (Fig. 1a), scattering from Pt nanoparticles deposited on MWCNTs results in very small Bragg peaks that at most are only ∼1% above the background scattering (Fig. 5b). Minor changes in HE-XRD patterns are noted in the Pt nanoparticles when cycled in 1 M NaOH at potentials between 0.1 V and 0.7 V under continuous oxygen purge, yet the scattering intensity remains sufficiently low to preclude PDF analysis. In contrast, Pt nanoparticle deposited on graphene nanosheets clearly exhibit a much higher intensity HE-XRD pattern as compared with the background as depicted in Figs. 5(c) and 5(d) when using the capillary working electrode geometry (Fig. 2). The primary diffraction peak at 2.74 Å⁻¹ has an intensity which is ∼17% above the background, an order of magnitude signal enhancement versus the conventional half-cell geometry. Such improvement in intensity over the background provides for a more accurate background subtraction for subsequent PDF processing and analysis. Modeling of in situ Pt nanoparticle PDFs with RMC simulations is expected to obtain atomic-scale structural differences across the length of the nanoparticle during electrochemical cycling (Petkov et al., 2016), and is the subject of future studies.

Figure 5 Raw HE-XRD patterns for Pt nanoparticles (a) in ex situ Kapton capillaries, (b) in a three-electrode half-cell geometry at open circuit voltage (OCV) (red line), cycled for 30 min (green line) and 60 min (blue line) as compared with a background MWCNT pattern, and (c) in an aligned carbon fiber working electrode geometry at OCV (red line) compared with the background graphene pattern (black line). HE-XRD patterns for the ex situ Kapton capillaries were recorded at the 11-ID-C station (105 keV) at APS. HE-XRD patterns for the three-electrode hall-cell geometry were recorded at the 11-ID-B station (58.65 keV) at APS. HE-XRD patterns from the aligned carbon fiber working electrode geometry were recorded at the 6-ID-D station (100 keV) at APS.

While specific structural changes for the above systems will be discussed in future reports where adequate context can be given, here we demonstrate the electrochemical operation of the capillary working electrode cell geometry using LiCoO₂, a common cathode material for lithium ion batteries. The electrochemical behavior of LiCoO₂ has been studied using in situ XRD (Reimers & Dahn, 1992), TEM (Wang et al., 1999) and electrochemical quartz crystal microbalance (Choi et al., 1998) techniques. The operation of LiCoO₂ as an electrode for lithium ion batteries has been explored in both non-aqueous (Plichta et al., 1989) and aqueous (Ruffo et al., 2009) electrolytes. At potentials more positive than 3.9 V versus Li/Li⁺ (0.7 V versus Ag/AgCl), Li is expelled from the LiCoO₂ structure, forming Li_1–xCoO₂ and leading to structural differences which are discernable by X-ray diffraction (Ruffo et al., 2009; Plichta et al., 1989; Reimers & Dahn, 1992). On reverse polarization, the Li⁺ incorporates back into the Li_1–xCoO₂ structure, reforming LiCoO₂.

During in situ measurements on LiCoO₂ in this work, the LiCoO₂-containing working electrode was repeatedly cycled between +1 V (oxidation) and −1 V (reduction) versus Ag/AgCl for 8 min intervals in an aqueous 0.1 M LiCl electrolyte under continuous argon purge. The PDFs for LiCoO₂ under oxidation and reduction conditions averaged over three electrochemical cycles are shown in Fig. 6(a). We note that the short discharge time of 8 min is expected to result in the removal of only a fraction of the Li⁺ from the LiCoO₂ structure. Also, the bulk LiCoO₂ powder used here is expected to exhibit modest structural changes compared with nanoscale electrode materials of interest to next-generation batteries. Nonetheless, we do observe statistically significant structural differences (exceeding two standard deviations) in the PDFs under different applied biases (see below for discussion).

Figure 6 (a) Pair distribution function calculated from in situ HE-XRD data for LiCoO2 in an aligned carbon fiber working electrode geometry for measurements taken under oxidizing and reducing potentials averaged over three cycles. (b) RMC fit of the reduction data in (a) using a bulk LiCoO2 starting structure. (c) A zoomed window of (a) highlighting changes measured under applied bias. (d) A species decomposed intensity from the reduction potential fit in (b) over the region highlighted in (c). Experiments were performed at the 6-ID-D beamline at APS.

We also performed RMC modeling on the PDF data as shown in Fig. 6(b) using a pristine LiCoO₂ starting structure (Aoun, 2016). This modeling consisted of >10⁷ RMC steps, including atomic translations, Li-Co swaps and removal of Li. Despite the impurities and defects expected to be present in commercially available bulk LiCoO₂ powder, we see mostly good agreement between the experimental data and model based on pristine LiCoO₂. We note that the RMC modeling does not fully capture the experimental PDF at pair distances between 2 and 5 Å. This pair distance range is expected to correspond to first and second coordination sphere metal–oxygen and metal–metal pairs, as supported by the partial G(r) traces in Fig. 6(d). We expect that local variations in the atomic structure and composition in the commercial LiCoO₂ used here give rise to the disparity between the experimental PDF and the RMC model. Overall close agreement of experimental data with the LiCoO₂ model indicates that a sufficiently large active material diffraction signal and reproducible background can be obtained using the capillary cell to provide accurate structural information. We note that uncertainty in the RMC modeled structure points to an incomplete description of the exact atomic structure, but still allows for a qualitative interpretation of the statistically significant structural changes observed experimentally.

Prior work has suggested that lithium removal from LiCoO₂ leads to structural differences in the region surrounding an atomic pair distance of ∼4.7 Å (Reimers & Dahn, 1992). This region of Fig. 6(a) is highlighted in Fig. 6(c), and we observe structural differences as a function of applied bias using the capillary working electrode cell geometry which are in agreement with these previous results. Furthermore, the pair-decomposed partial G(r) traces calculated using the Interactive Structure Analysis of Amorphous and Crystalline Systems (ISAACS) program (Le Roux & Petkov, 2010) as shown in Fig. 6(d) suggest that the changes to the PDF in Fig. 6(c) arise primarily from changing Co–O distances, which is consistent with prior reports (Wang et al., 1999; Reimers & Dahn, 1992). These structural shifts have been attributed to a canting of the c-axis in the hexagonal LiCoO₂ structure (Reimers & Dahn, 1992) and the formation of exchange defects between Co and Li (Wang et al., 1999). While further analysis and interpretation of the structural changes is outside the scope of this paper, the promise of new structural information from in situ PDF analysis enabled by our capillary working electrode cell geometry is demonstrated with the LiCoO₂ data we present here.

5. Conclusion

The described capillary working electrode electrochemical cell geometry was designed to address issues with scattering intensity and reproducible background signal for in situ HE-XRD measurements. The use of a capillary working electrode is beneficial for three reasons: (i) it allows for easy loading of appreciable amounts of material to provide sufficient scattering over the background signal; (ii) the background signal from the carbon capillary has low variability, arising from tight manufacturing tolerances; (iii) as the capillary consists of aligned carbon fiber, the majority of the scatter background is anisotropic, which can readily be masked out of the two-dimensional diffraction pattern during image processing. Using a series of examples, we have demonstrated the utility of this cell for enhancing the HE-XRD signal by an order of magnitude for in situ electrochemical measurements on various material systems. The improved resolution in HE-XRD data afforded by the capillary cell geometry will enable PDF analysis for a range of electrochemical systems where this has not been possible before. PDF analysis on these systems will help to improve understanding of structural changes during electrochemical events and to establish structure/function relationships for electrochemical technologies, driving a new age of material innovation for electrochemical applications.

Footnotes

‡Present address: Energy Systems, Argonne National Laboratory, Argonne, IL 60439, USA.

§Present address: Materials and Manufacturing Directorate, Air Force Research Laboratory, Wright-Patterson AFB, OH 45433, USA.