All-solid-state Li-ion batteries having incombustible solid electrolytes are promising energy storage devices because they have significant advantages in terms of safety, lifetime and energy density. Electrochemical reactions, namely, Li-ion insertion/extraction reactions, commonly occur around the nanometer-scale interfaces between the electrodes and solid electrolytes. Thus, transmission electron microscopy (TEM) is an appropriate technique to directly observe such reactions, providing important information for understanding the fundamental solid-state electrochemistry and improving battery performance. In this review, we introduce two types of TEM techniques for operando observations of battery reactions, spatially resolved electron energy-loss spectroscopy in a TEM mode for direct detection of the Li concentration profiles and electron holography for observing the electric potential changes due to Li-ion insertion/extraction reactions. We visually show how Li-ion insertion/extractions affect the crystal structures, electronic structures, and local electric potential during the charge–discharge processes in these batteries.
Rechargeable Li-ion batteries (LIBs)  are essential for not only portable electronic devices but also power sources of recent electric and hybrid vehicles [2,3]. Moreover, LIBs have begun to be used as energy-storage devices for renewable energy, such as wind and solar power. Such systems with LIBs drastically reduce fossil-fuel consumption and CO2 and/or toxic-gas emissions. However, there are still problems to be solved, i.e. those regarding safety, cost, lifetime and energy density. All-solid-state LIBs with nonflammable solid-electrolytes may likely solve these problems. Concerns over leakage of flammable liquid electrolytes and an additional circuit for protection against burning are not necessary, improving safety and reducing cost. Moreover, all-solid-state LIBs can reduce the superfluous reaction around the electrode/electrolyte interfaces, e.g. gas emission and/or solid-electrolyte-interface (SEI) formation resulted from liquid electrolytes, and the LIBs can use Li metal as an anode electrode, improving lifetime (battery cyclability) and energy density . A serious problem preventing the practicality of all-solid-state LIBs is low power density caused by the large interfacial resistance of the Li-ion transfer at the electrode/solid-electrolyte interfaces. Many techniques have been proposed to improve the Li-ion transfer at these interfaces [5–9]. One effective technique is the in situ-formation of electrode active materials from solid electrolytes [5,10–13]. Such in situ-formed electrodes were discovered in Li2O–Al2O3–TiO2–P2O5 (LATP  [http://www.ohara-inc.co.jp/en/product/electronics/licgc.html])-based solid electrolytes [10,11]. The electrode active materials are irreversibly formed by decomposing the negative-side solid electrolyte with the Li-ion insertion reaction. The electrode and solid electrolytes are connected on an atomic scale, resulting in drastic improvement of the Li-ion transfer . However, it was not clear how the electrodes were formed during the charge–discharge processes and how Li-ion insertion reaction affected the crystal structure, electronic structure and local electric potential around the formed electrodes. These electrochemical reactions usually take place on a nanometer scale; thus, dynamic observation under the reaction condition (operando observation) by transmission electron microscopy (TEM) would be useful not only for understanding the fundamental electrochemistry but also developing high-performance batteries.
To observe the electrochemical formation of in situ-formed electrodes, we used two specialized TEM techniques for this study, spatially resolved electron energy-loss spectroscopy in a TEM mode (SR-TEM-EELS) [15–17] and electron holography (EH) [18–20]. The SR-TEM-EELS technique is useful to evaluate battery materials because it can simultaneously detect the Li distribution and accompanying electronic-structure changes in other elements. The EH technique is also unique and has been commonly used for observing potential distribution in semiconductor devices [21–24] and magnetic flux distribution in magnetic materials [25–29]. In 2010, we applied EH to battery materials and succeeded in visualizing the potential changes during battery reactions by using a transmission electron microscope . Because Li-ion insertion/extraction reactions change the local electrode potential in the electrodes, EH can provide local potential profiles due to Li-ion movement. Such potential profiles around the electrode/liquid-electrolyte interfaces can be theoretically predicted [31–33], but have never been observed. Through operando observations using the above two TEM techniques, we visually show how electrochemical reactions occur in all-solid-state LIBs.
Materials and methods
TEM sample preparation of all-solid-state LIBs for operando observations
Figure 1a shows a schematic illustration of the prepared LIB sample. A 90-µm-thick LATP sheet with Si and Ge doping (LASGTP) was used as the solid electrolyte. This LASGTP solid electrolyte is a nonflammable ceramic with ion conductivity of 10–4 S cm−1 at room temperature (now commercially available from OHARA Inc., Japan) [http://www.ohara-inc.co.jp/en/product/electronics/licgc.html]. As a positive electrode, an 800-nm-thick film of LiCoO2 was deposited on one side of the sheet by pulsed laser deposition. Then, a Au film was deposited as a positive-side current collector on top of the LiCoO2 electrode using a sputtering method. On the other bare side of the sheet, a Pt film was directly deposited as a negative-side current collector. When the charging voltage is applied between the Pt and Au current collectors, the Li-ions extracted from the LiCoO2 are transferred and charged around the negative-side Pt/LASGTP interfaces. The excess Li-ions cause irreversible decomposition of the LASGTP at the interfaces; consequently, the in situ-formed negative electrodes are formed, as shown in Fig. 1a . We measured the interfacial resistance at the negative-electrode/LASGTP interface using an AC impedance method, it was about 100 Ω cm2, which was 40 times lower than that of the LiCoO2/LASGTP interfaces (4000 Ω cm2) .
To operate the battery reaction in a transmission electron microscope, copper plates were electrically connected to the LIB bulk sample with a silver paste, and the sample was fixed between the two electrodes of a biasing TEM holder, as illustrated in Fig. 1b. Then, one part of the negative side (observation area) of the sample was thinned using a focused ion beam (FIB). This holder has a dedicated mechanism to avoid exposing the sample to air. The head part of the holder is mechanically slid into the holder cylinder while transferring from the FIB system to the transmission electron microscope. Although the LASGTP solid electrolyte is water proof, it would be better not to expose the battery sample to air because the sample surface of the observation area might react in a moist environment. Sample protection using a glove bag with an Ar gas is somewhat effective if an air-protection TEM holder is not available.
Spatially resolved TEM EELS for operando observation
Electron energy-loss spectroscopy has been used to directly detect Li signals on a nanometer scale [34–36]. Although EELS using a scanning transmission electron microscope (STEM-EELS) is a common technique to provide a two-dimensional Li map, the strong convergent electron beam easily damages battery materials compared to the parallel electron beam, particularly regarding solid electrolytes. Moreover, the beam deflection for scanning and scan noise sometimes deteriorate the reliability of chemical shifts of the EELS spectra. In this study, we adopted spatially resolved (SR-)TEM-EELS that does not require scanning of convergent electron beams.
Figure 2a illustrates the experimental set-up of SR-TEM-EELS [37,38]. The battery sample connected with a voltammeter is uniformly illuminated by collimated electrons under a TEM lens mode. Thus, the common TEM image can be seen at the bottom of the TEM column. The region of interest, e.g. the negative-electrode/solid-electrolyte interfaces, is selected using a rectangular slit installed at the TEM column. The electrons passing through the slit are energy-dispersed using a magnetic prism in the EELS system. Then, by adjusting the EELS lens system, the energy-dispersed plane is line-focused along the y-direction on a two-dimensional charge-coupled-device (CCD) camera. Figure 2b illustrates a typical EELS image. The horizontal and vertical axes indicate the loss energy and sample position in the y-direction, respectively. The EELS signals along the x-direction of the sample are integrated in the EELS image when the electrons are line-focused in the EELS system. Thus, the interfaces in the TEM image need to be parallel to the x-direction by adjusting the TEM lenses. The signals of zero-loss peak (ZLP) and other elemental EELS peaks are seen along the y-direction as illustrated in Fig. 2b. When the Li concentration changes during the charge–discharge reactions, the signals of the Li K-edge should change around the energy loss of 60 eV. If we compare it with the EELS peaks of other important elements, e.g. Ti L-edge and O K-edge, we can simultaneously evaluate their electronic-structure changes due to Li-ion insertion/extraction reactions. Even if the ZLP is randomly shifted by some external noise during the acquisition time of EELS images, whole EELS spectra are shifted together with the same shift value. This means that the chemical shifts of the spectra are relatively maintained in the EELS image. Therefore, SR-TEM-EELS enables the detection of subtle chemical shifts with about 0.1-eV precision .
Electron holography for operando observation
Electron holography requires two imaging processes to display the electromagnetic fields . The first process is to record an electron interference fringe pattern (hologram) using a transmission electron microscope equipped with an electron biprism, and the second is to numerically reconstruct the modulated wave function from the hologram. Figure 3a shows a schematic illustration to record the hologram by using the microscope. The sample connected with a voltammeter is illuminated by coherent electron waves. The waves are split into two parts by using the electron biprism, one is the object wave modulated by the potential in the battery and the other is the reference wave passing through the vacuum. The hologram is commonly recorded with a CCD camera. Figure 3b shows the reconstruction procedure. The hologram is processed by two-dimensional Fourier transformation (FT), then one of the side-bands, corresponding to the FT of the object wave, is selected using a spatial frequency filter and shifted to the center of the image. Then the side-band is applied by inverse FT, finally the reconstructed object wave is obtained as a complex number matrix .
The phase of the reconstructed wave includes the information of inner potential of the materials, electron charging to solid electrolytes by incident electrons, and electrode potential changes due to Li-ion insertion/extraction reactions. To obtain the information of the electrochemical reactions due to Li, we numerically subtracted the phase values of the initial state (under short-circuit) from that of the voltage-applied state because the inner potential of the materials and the electron charging do not change by biasing the sample. The obtained phase value is proportional to the electrode potential changes caused by Li-ion insertion/extraction reactions. To obtain the conversion rate from the phase to the potential changes, we experimentally measured the total phase changes between the positive and negative electrodes by EH [30,40], then by dividing the applied voltage value by the total phase value, we obtained a conversion rate of 0.10 V/rad.
Results and discussion
Crystal-structure changes due to Li-ion insertion reactions
Before operando observations, we evaluated the crystal-structure changes around the in situ-formed-negative-electrode/LASGTP interfaces by conducting a common TEM analysis . The LIB bulk sample was cycled 50 times in a vacuum with a cyclic voltammetry mode. The applied voltage was swept between 0.0 and 2.0 V at a rate of 40 mV/min. Immediately after the cycles, a small piece around the negative-side interfaces was lifted out from the bulk using a micro-sampling method in an FIB system and thinned to a thickness of about 100 nm. Figure 4a shows a TEM image around the interfaces. Black, gray, and white contrast regions were observed in the parent LASGTP solid electrolyte, which are Li1+xAlxGeyTi2−x−yP3O12 (main-phase), Li1+x+3zAlx(Ge,Ti)2−x(SizPO4)3 (sub-phase), and AlPO4, respectively. The former two phases have Li-ion-conductive structures. We found a different TEM-contrast layer of about a 400-nm region near the Pt current collector, as shown in Fig. 4a. Figure 4b and c shows the electron diffraction patterns from the area (circle ‘B’ in Fig. 4a) and from the parent LASGTP (circle ‘C’ in Fig. 4a), respectively. Although the diffraction spots belonging to the basic LiTi2(PO4)3 crystal of the LASGTP were clearly observed in circle ‘C’, no spots were observed in circle ‘B’. This means that the 400-nm region is an amorphous layer. Because we did not see such layers before battery cycling, this region is the insitu-formed negative electrodes where the excess Li-ions were irreversibly inserted. The negative electrodes are united with the LASGTP solid electrolytes around the interfaces, which is the reason the interfacial resistance of Li-ion transfer was low. In the negative electrode layer, however, some cracks were also observed, as indicated in Fig. 4a. We found from an elemental mapping by energy dispersive X-ray spectroscopy (EDS) (for more detail, see Ref. ) that these cracks were formed near the non-Li-ion conductive grains, i.e. AlPO4. These grains seem to be affected by physical pressure from the surrounding grains expanded by the Li-ions. If the fabrication processes of the LASGTP materials are improved so as not to form AlPO4 grains, the interfacial resistance would further decrease.
Li profiles and electronic structure changes detected by SR-TEM-EELS
We carried out SR-TEM-EELS measurement on the same TEM sample shown in Fig. 4a to observe the Li profiles and accompanying electronic-structure changes of the other important elements, Ti and O . The region surrounded with a dashed line in Fig. 4a was selected using the rectangular slit, then the SR-TEM-EELS images were recorded around the Li K-, Ti L- and O K-edges. Figure 5a shows the selected region and Fig. 5b–d shows the EELS images of Li, Ti and O, respectively. The background intensity of each spectrum image was subtracted by a common power-law fitting. The Li signals were detected in the negative-electrode region, as shown in Fig. 5b. Figure 5e shows the relative Li concentration profile along ‘A’–‘B’ direction surrounded by broken lines in Fig. 5b, where the average Li signals in the parent LASGTP region was set to 1. The Li concentration gradually increased toward the Pt/negative-electrode interfaces. The maximum Li concentration was about eight times higher than that in the original LASGTP. The basic LiTi2(PO4)3 crystals of LASGTP can accept up to three times higher Li concentration, i.e. Li3Ti2(PO4)3. Therefore, it is easily understandable how the excess Li-ions amorphized the negative-side LASGTP crystals and why the formation of this negative electrode proceeded irreversibly.
In the Ti L-edge image (Fig. 5c), a clear chemical shift was observed in the negative electrodes. Figure 5g shows the EELS spectra at lines ‘C’–‘D’ and ‘E’–‘F’ in Fig. 5c. The spectrum in line ‘C’–‘D’ (negative-electrode region) shifted to the lower energy-loss direction compared with that in line ‘E’–‘F’ (LASGTP solid electrolyte region). Furthermore, the spectrum in the negative electrode did not have shoulder structures, while the spectrum in the LASGTP had shoulders, as indicated by the arrows. This feature shows that the Ti in the negative electrode was reduced from the original state of Ti4+ to Ti3+ by the Li-ion insertion reaction . The Ti4+ and Ti3+ regions are designated on the Li profile in Fig. 5e. In the Ti3+ region, the Li signal is significantly larger than that in the Ti4+ region. It turns out that there is a threshold level of Li concentration to reduce Ti. This is surprising because other elements need to compensate the additional charge due to Li-ion even if the Li concentration is lower than the threshold level. As we describe in the next paragraph, O plays a significant role to compensate the charge instead of Ti under the threshold level.
In the O K-edge image (Fig. 5(d)), we can see two main peaks ‘a’ and ‘b’. Both peaks are shifted to the lower energy-loss direction at the negative-electrode/LASGTP interface indicated with a light-blue dashed line. This means that the spectrum shifts are related to the Li-ion insertion reaction. Figure 5h shows the spectra along lines ‘G’–‘H’ (Ti3+ region in the negative electrode), ‘I’–‘J’ (Ti4+ region in the negative electrode), and ‘K’–‘L’ (Ti4+ in the LASGTP solid electrolyte). Peak ‘a’ reflects the electron transitions from the O 1s-state to 2p-state by the incident electrons, and it is well known that the O 2p-state is hybridized with the transition-metal Ti 3d-state . From some reports [43,44] using X-ray absorption spectroscopy and first-principle calculation for LiTi2O4 and (Li, La) TiO3 battery materials, the additional electrons due to the Li-ion insertion occupied the Ti- and O-hybridized orbit and suggest that O also contributed to the charge compensation for Li-ion insertion/extraction reactions. Therefore, the spectrum shift of peak ‘a’ in Fig. 5d and h shows evidence that O in the negative electrode compensated for the charge balance for Li-ion insertion. More significantly, the intensity of peak ‘a’ in ‘I’–‘J’ was slightly higher than that in ‘G’–‘H’, as shown in Fig. 5d and h. Spectrum ‘I’–‘J’ was taken from the Ti4+ region in the negative electrode where some of the Li-ions and electrons were extracted at the discharged state. These electrons were partially extracted from the Ti and O hybridization orbit, so that the unoccupied state increased in the orbit. This is the reason that the intensity of peak ‘a’ in ‘I’–‘J’ became higher. Consequently, SR-TEM-EELS can provide important electrochemistry information, not only on Li distribution but also on electronic-structure changes where the electrons are inserted or extracted during the reactions.
Furthermore, other interesting information can be obtained from peak ‘b’ in Fig. 5d and h. Peak ‘b’ resulted from the resonance scattering of ejected core electrons between neighboring O atoms. In the previous reports [42,45,46], there is one simple relation:
Operando SR-TEM-EELS measurement to observe Li-movement
We carried out operando observation using SR-TEM-EELS to observe how the Li-ions are inserted and trapped to form negative electrodes . The sample was loaded on the air-protection biasing holder (Fig. 1b), and charged and discharged in the microscope. In order to detect enough changes of Li signals during the reactions, we adopted constant-current (CC) and constant-voltage (CV) modes that allow evaluation of Li concentration changes in a long time scale. Figure 6a shows the battery cell voltage as a function of the CC/CV charge–discharge time, and Fig. 6b shows the charge–discharge current (left vertical axis) and total charge (right vertical axis) as a function of time. The sample was first charged with a CC of 80 nA. When the cell voltage reached 2.0 V, the sample was further charged with a CV of 2.0 V for 7.5 h until the current value was almost steady (equilibrium state). Although the charge current decreased, the total charge gradually increased. This means that the Li-ions continued charging around the negative-side LASGTP. Third, the sample was discharged with a CC of −80 nA until the cell voltage decreased to 0.0 V. This took about 0.5 h. Finally, the sample was short-circuited (CV of 0.0 V) for 11 h and fully discharged.
The SR-TEM-EELS images around the negative-side interface (Fig. 7a) were recorded at states (B)-(G) indicated in Fig. 6a. Figure 7b–g shows the EELS images of the Li K-edge (after background subtraction) measured at states (B)–(G), respectively. The vertical and horizontal axes of the EELS images represent the loss energy of electron and sample position from the interface, respectively. Figure 7b′–g′ shows the horizontal profiles of the Li K-edge signals integrated in the full energy range shown in Fig. 7b–g, revealing the Li-movement during the charge–discharge processes. We defined the relative value of 1 in the vertical unit as the average Li signals in the parent LAGSTP solid electrolyte. During charging (Figs. 7b–d and 7b′–d′), the Li signals increased from the Pt/LAGSTP interface and gradually spread around the 400-nm region. The Li profiles shown in Fig. 7d′ and e′ are similar to that in Fig. 5e, although the Fig. 5e data were taken immediately after 50th discharge in the cyclic voltammetry mode. At the discharging states (Figs. 7e–g and 7e′–g′), the Li signals gradually decreased and diffused such that the distribution was almost flat in the 400-nm region, as shown in the profile of Fig. 7g′. Because the negative electrode was irreversibly formed, some Li-ions remained even after being fully discharged. We also recorded the Ti L-edge EELS images during the reactions and found that the position of the Ti4+/Ti3+ transition also moved as the Li concentration changed, as indicated by the arrows in Fig. 7b′–g′ (for more detail, see Ref. ). Operando SR-TEM-EELS is an effective technique to directly understand the Li distribution and its effect on other elements.
Operando EH measurement to observe local electric potential
The Li-ion insertion reaction in electrodes generally decreases the potential because the reduction of Li takes place at the ‘standard electrode potential’ of −3.05 V, which is commonly defined as the difference from the redox potential of hydrogen. The in situ-formed negative electrodes are irreversibly formed due to the Li-ion insertion reaction, as shown in Fig. 1a. Thus, we can detect the potential changes due to the reaction by EH . Recently we have pointed out the influence of potential leakage of applied voltage on EH measurement . When the distance between two electrodes is as small as several micrometer, a strong electric field spreads three-dimensionally around the observation area and affects the phase of electron waves, and therefore the potential leakage needs to be considered to interpret the phase image. In this battery sample, however, the distance between the electrodes is about 90 µm, it is much longer than the observation scale. The electric field is not so strong to affect the electron waves, and therefore the influence on the EH measurement would be almost negligible.
Figure 8a shows the TEM image around the negative-side LASGTP region. The thickness of the TEM sample was about 60 nm. We applied a voltage between the Au and Pt current collectors to operate the battery reaction in the microscope. Figure 8b shows the macroscopic curve of the initial cyclic voltammetry, where the sweep rate was 40 mV/min. The standard electrode potentials of the LiCoO2 positive electrodes and insitu-formed negative electrodes were 0.90  and −0.75 V , respectively. Thus, the charge and discharge peaks appeared around 1.6 V, which corresponds to the difference in those electrode potentials. We recorded the holograms around the negative-side region at the charge–discharge states (C)−(J) indicated in Fig. 8b, then obtained the potential profiles using the conversion rate of 0.10 V/rad. Figure 8c–j shows the potential profiles along line ‘A’–‘B’ (Fig. 8a) at different voltages (C)−(J) in Fig. 4b, respectively. The horizontal axis indicates the distance from the Pt/LASGTP interface and the vertical axis shows the relative potential changes against the flat potential level in the LASGTP solid electrolyte (set to 0 V).
At 0.42 V (Fig. 4c), a no-potential profile was observed because the charge current was almost zero and the Li-ions did not move to the observation area. As the applied voltage increased, the Li-ions started to charge around the region; as a result, the local potential gradually decreased toward the Pt/LASGTP interface (Fig. 8d−f). When the applied voltage was reversed for the discharging mode, the gentle slopes on the right side from the red arrows gradually flattened (Fig. 8g−j). However, the lower potential remained on the left side of the red arrows. This potential profile is similar to the Li profile in Fig. 5e, although the sign of the vertical axis is opposite, as described above. Therefore, this region is the insitu-formed negative electrode, where the Li-ions were trapped in the negative-side LASGTP. Moreover, on the right side of the arrows in the discharging states (Fig. 8g−i), slightly lower potentials (0.06−0.08 V) were observed in the 1600-nm region, and the region moved several hundred nanometers as the discharge proceeded. This region might be a much lower Li concentration distribution that is probably undetectable by operando EELS considering the signal-to-noise ratio in Fig. 7. We believe that EH is more sensitive in detecting Li profiles than EELS because the electrode potential commonly changes a large enough signal for EH with a small amount of Li-ion insertion/extraction. To prove this scientifically, further experiments and comparison of EH and EELS are necessary.
We introduced two types of specialized TEM techniques, SR-TEM-EELS and EH, to reveal how the solid-state electrochemical reactions proceed around the in situ-formed-electrode/solid-electrolyte interfaces during the charge–discharge processes. The insitu-formed negative electrodes had amorphous structures resulting from the highly concentrated Li-ion insertion. This amorphous electrode continuously connected with the solid electrolyte by changing the Li concentration. This is the reason that the interfacial resistance became lower around the interfaces. In addition, SR-TEM-EELS showed the electronic-structure changes, not only the Ti4+/Ti3+ transition but also what orbits the additional electrons were inserted to or extracted from. The picometer-scale expansion of O–O distances due to the Li-ion insertion was also clearly detected. Operando observations by SR-TEM-EELS and EH clearly showed how the negative electrodes were formed from the parent solid electrolytes by displaying the Li profiles and local electric potential profiles. Their combination provided some convincing information. Both techniques are capable of applying any other all-solid-state batteries. Therefore, in the near future, SR-TEM-EELS and EH analyses would contribute to advances in electrochemistry and innovation regarding high-performance batteries.
We would like to thank Prof. Z. Ogumi, Prof. T. Abe, Prof. Y. Uchimoto, Prof. Y. Koyama, Prof. I. Tanaka, Prof. Y. Ukyo, Dr A. Shimoyamada, Dr T. Sato, Dr K. Kimoto, Dr J. Kikkawa and Mr. H. Matsumoto for their valuable discussion and advice on the SR-TEM-EELS experiments. We also thank Dr T. Kato, Mr. R. Yoshida, Mr H. Kurobe and Mr T. Hamanaka for preparing the battery sample and TEM analysis. We would like to acknowledge Dr Y. Sugita, Mr K. Nonaka, Mr. K. Miyahara, Dr H. Fujita, Dr T. Asaka, Dr C.A.J. Fisher, Dr H. Moriwake, Dr. A. Kuwabara and Dr R. Huang for their valuable comments regarding the EH results. We are grateful to OHARA Inc. for supplying the LASGTP sheet used as the solid electrolyte.
One part of this work regarding the SR-TEM-EELS technique was supported by the RISING project of the New Energy and Industrial Technology Development Organization (NEDO) in Japan. The other part of the work regarding the EH technique was financially supported by Chubu Electric Power Co., Inc.