III. RESULTS AND DISCUSSION

The Li transport and lithiation of carbons were confirmed by real-time EELS analysis in the following two different experimental setups [Figs. 1(a) and 3(a)]. Both sets of EELS data (similar to soft-x-ray absorption near-edge spectroscopy, XANES) were acquired in the image mode from the am-C film near a Li₄SiO₄ particle, the acquisition areas defined by the EELS aperture are the same, and the current density of the electron beam was similar.

FIG. 1.

(a) Schematic of experimental Setup I. (b) A series of real-time EELS. The inset shows the background subtracted Li K-edge. (c) Integrated intensity of the Li K-edge, with the decomposed components of Li–O and Li–C.

In Setup I, as shown in Fig. 1(a), both the Li₄SiO₄ particle and the am-C film, where the EELS spectra were acquired, were illuminated simultaneously during data acquisition. Some selected time-resolved EELS spectra are shown in Fig. 1(b). (A full time series is given in S.I.) It is seen that there are significant changes in the Li K-edge spectra. With an increased exposure, the Li K-edge changes gradually from two-peak to one-peak structure. Most importantly, the intensity of the Li K-edge drops sharply at the beginning of the exposure. This is seen more clearly in the profile of integrated intensities, which are normalized to the intensity of the zero-loss peak [Fig. 1(c)]. (The spectra were deconvoluted from the plural scattering and so reflect Li composition change.^{Reference Egerton23}) Initially, the amount of Li within the area defined by the EELS aperture decreased almost linearly with exposure until about 160–180 s of irradiation. The continued exposure produced little change in the Li content.

It is noticeable that the two-peak fine structure of the Li K-edge in Fig. 1(b) is slightly different from the reference Li K-edge of the Li₂O crystal.^{Reference Jiang and Spence24} The difference is attributed to the mixture of two types of bonding states, i.e., Li–O and Li–C, respectively. The Li K-edge in Li–O exhibits a characteristic two-peak structure, while it is a single peak in Li–C, and its threshold shifts about 2 eV to higher energy (Fig. 2). This assignment has been confirmed from both the real-time EELS analysis and simulations detailed in S.I. Thus, the Li K-edge in Fig. 1(b) can be decomposed into contributions from Li–O and Li–C, respectively, using the reference spectra in Fig. 2 (see S.I.). The decomposed intensities are also plotted in Fig. 1(c). The results suggest that only a small portion (about one fifth) of Li is involved in forming Li–C bonds, i.e., lithiation of carbons, while the majority of the remainder is oxidized by the O released from the decomposed Li₄SiO₄. It is noticeable that the Li–O component drops rapidly to almost zero after 200 s of exposure. By contrast, the Li–C component remains constant throughout.

FIG. 2.

Extracted spectra of Li–O and Li–C bonding states.

To understand the decrease in the Li–O intensity, we designed another set of experiments, Setup II [Fig. 3(a)]. In this experiment, only the am-C film was under electron illumination during real-time data acquisition, after the Li₄SiO₄ particle was preilluminated and decomposed. Contrary to Setup I, the Li K-edge EELS spectra show almost no change [Fig. 3(b)]. Most significantly, the decreased rate of integrated intensities [Fig. 3(c)], normalized to the intensity of the zero-loss peak, is now much slower compared to that observed in Setup I and is negligible.

FIG. 3.

(a) A schematic of experimental Setup II. (b) A series of real-time EELS. The inset shows the background subtracted Li K-edge. (c) Integrated intensity of the Li K-edge. Dotted line is a guide.

It is worth noting that in both setups, the Li₄SiO₄ particles have been decomposed before in situ EELS data were acquired, and the only difference between these two experiments was that the decomposed particle was still under electron beam irradiation in Setup I, but it was not in Setup II. The large difference in the rate of decrease between Setup I and Setup II indicates that knock-on sputtering or mechanisms involving radiolysis could not be responsible for the rapid loss of Li⁺ observed in Setup I. Otherwise similar decrease rates would be observed, since knock-on sputtering and radiolysis are due to elastic or inelastic collisions between energetic electrons and atoms. Therefore, the rapid decrease of the Li intensity must be associated with the coillumination of the decomposed Li₄SiO₄ particle nearby, which is the only significant difference between the two experiments.

The decomposition of Li₄SiO₄ under electron irradiation is caused by an induced electric field due to ionization.^{Reference Jiang22} When excited by energetic incident electrons, the ionization process ejects Auger and secondary electrons near the surface of the Li₄SiO₄ particle into a vacuum. Since Li₄SiO₄ is a good insulator for electrons, the charge balance cannot be immediately restored by electrons from the surroundings through the supporting am-C film.^{Reference Jiang22,Reference Cazaux25} A positively charged shell is thus formed around the Li₄SiO₄ particle, resulting in highly mobile Li⁺ ions flooding the supporting am-C film.

Suppose that I _E is the current of emitted electrons flowing out of the particle and I _A is the current of electrons flowing into the particle from the supporting am-C film, in compensation. The charge build-up rate in the particle is approximately^{Reference Jiang22}

(1)$${{{\rm{d}}Q} \mathord{\left/ {\vphantom {{{\rm{d}}Q} {{\rm{d}}t}}} \right. \kern-\nulldelimiterspace} {{\rm{d}}t}} = {I_{\rm{E}}} - {I_{\rm{A}}}\quad .$$

Beam electrons do not contribute to the charge accumulation directly, because the specimen is thin enough to be transparent to the incident electrons. Since the positive potential on the surface suppresses the yields of emissions,^{Reference Cazaux25}I _E decreases with exposure time due to a positive charge accumulation in the particle. By contrast, the charge accumulation enhances the attraction of surrounding electrons into the particle, so I _A increases. Thus, at the beginning of irradiation, I _E(0) has its maximum value but then drops with time, while I _A(0) = 0 but increases gradually. At a certain time t ₀, I _E(t ₀) and I _A(t ₀) reach balance, so that

(2)$${\left( {{{{\rm{d}}Q} \mathord{\left/ {\vphantom {{{\rm{d}}Q} {{\rm{d}}t}}} \right. \kern-\nulldelimiterspace} {{\rm{d}}t}}} \right)_{t = {t_0}}} = 0\quad .$$

Meanwhile, the accumulation of charge also reaches its maximum at time t ₀, i.e.,

(3)$${Q_{{\rm{max}}}} = \int_0^{{t_0}} {\left[ {{I_{\rm{E}}}(t) - {I_{\rm{A}}}(t)} \right]{\rm{d}}t} \quad .$$

In theory, Q _max is determined not only by the intrinsic physical and chemical properties of Li₄SiO₄, but also by the illumination conditions, the beam current density, and physical contact between the particle and am-C film.^{Reference Jiang22}

Further irradiation (i.e., t > t ₀) should not change Q _max as long as the particle is exposed to the same electron beam, because ${I_{\rm{E}}}({t_0}) = {I_{\rm{A}}}({t_0}) \ne 0$, unless the particle is decomposed. Intrinsically, this change is mainly caused by the change in the resistivity of the new phase.^{Reference Jiang22} Amorphous SiO₂ has a band gap exceeding 9.0 eV,^{Reference Garvie, Rez, Alvarezb and Buseck26} which is larger than the calculated 5.24 eV band gap of Li₄SiO₄ (Ref. Reference Duan and Parlinski27) so that decomposition may increase the particle's resistivity. Therefore, in Setup I, the charge remains as long as am-SiO₂ is continuously exposed to the same electron beam. As a result, the electric field induced by the fixed charges in the am-SiO₂ particle drives away all the mobile Li⁺ ions within the range in which the electric field is strong enough to overcome the binding energy of a Li⁺ ion to am-C. By contrast, in Setup II, the charges were neutralized by electrons from the surroundings, since the beam was off the decomposed particle. Therefore, the dispersed Li ions on the am-C around the particle remain. There is no driving force to cause Li transport when the induced electric field is turned off.

We conclude that the decrease of the Li content observed in Setup I is due to Li transport on the am-C, driven by the induced electric field on nearby decomposed am-SiO₂. Apparently, the mobile Li is in ionic form, bonded to O. On the am-C, these mobile Li⁺ ions are in the form of ultrathin layers or clusters. This conclusion results from an analysis of plasmon excitations. As shown in Fig. 4, the plasmon peak for Li-covered am-C is significantly different from that of am-C and Li₂O nanoparticles. In Li₂O, the bulk plasmon has a sharp peak at about 18.4 eV, while am-C has a broad peak around 25.0 eV. The plasmon peak of Li₂O is not visible in the Li covered am-C. The weighted sums of Li₂O and C are also compared in Fig. 4. This clearly shows that the bulk plasmon peak of Li₂O was strongly suppressed in the Li-covered am-C, indicating that the Li⁺ layers on am-C film are ultrathin.^{Reference Jiang, Su, Spence, Zhou and Qiu28}

FIG. 4.

Comparison of the EELS spectrum recorded from the Li covered am-C with reference spectra of Li₂O and am-C, and their weighted sums. The ratios of the weighted sum are listed.

It follows that Li⁺ ions in the Li–O form are responsible for Li transport driven by an induced electric field. By contrast, Li⁺ ions in the Li–C form are stable under the same strength of electric field. Since the Li–O species are absorbed on the surface of the am-C and the Li–O bond is ionic (polarized by electrostatic attraction between a positive Li⁺ ion and a negative O²⁻ ion), Li–O can be easily decoupled by the induced electric field, resulting in Li⁺ transport and O²⁻ desorption.

While the Li–C bond is less polarized than Li–O, it could be relatively stable under an electric field. Since disordered carbons show a similar short-range structure to graphite or graphene,^{Reference Claye and Fischer29} the intercalated Li could either be inserted between layers of graphite segments^{Reference Sato, Noguchi, Demachi, Oki and Endo30} or “coat” both sides of graphene sheets,^{Reference Dahn, Zheng, Liu and Xue31} forming local clusters resembling the structure of LiC₆.^{Reference Kganyago and Ngoepe32} This was confirmed by the similarity of the Li K-edge EELS between the lithiated am-C and crystal LiC₆, since EELS is more sensitive to the local structure around the absorbing atom rather than long-range order (see S.I.).