Introduction

Electrochromic materials are able to change their optical properties under the action of an electric field1,2. This unique property gives electrochromic devices a wide range of applications in information displays, adjustable mirrors, variable emittance surfaces and smart windows3,4,5, and the latter is especially attractive in decarbonlization due to the combined advantages of energy-saving in buildings while maintaining indoor comfort6,7,8,9,10. Amorphous tungsten oxide (a-WO3) is the most studied electrochromic material, and almost all oxide-based electrochromic devices employ WO31,11,12,13,14. As many other electrochromic oxides possess similar coloration dynamics, WO3 represents a widely used model for exploring the electrochromic processes1,2,15,16,17. Small polaron hopping is the most accepted theory1,15,18,19,20, which asserts that as an ion and an electron are simultaneously inserted, the electron is localized around a W6+ site and forms a W5+ one. The localized electron is named a polaron and optical absorption then arises from polaron hopping between the W5+ site and a nearby W6+ site through15,21:

$${{{{{{{\rm{W}}}}}}}_{({{{{{\rm{a}}}}}})}}^{5+}+{{{{{{{\rm{W}}}}}}}_{({{{{{\rm{b}}}}}})}}^{6+}\mathop{\leftrightarrow }\limits^{{{hv}}_{1}}{{{{{{{\rm{W}}}}}}}_{({{{{{\rm{a}}}}}})}}^{6+}+{{{{{{{\rm{W}}}}}}}_{({{{{{\rm{a}}}}}})}}^{5+}$$
(1)

Despite the fact that the phenomenon of electrochromism has been discovered for over 50 years22,23,24, and associated research as well as commercialization also keep growing with enhanced durability of WO325,26,27, performance degradation upon cycling is inevitable and its origin is still far from clear24,28,29,30,31,32, which largely limits the development of superior devices. The ion trapping/detrapping phenomenon reported a few years ago provides a new vista to reframe the electrochromism field and also pave the way to extend the lifetime of WO3 thin films28,33,34,35. It has been found that failure to extract the inserted ions in a single cycle leads to ion accumulation in the electrodes as cycling proceeds16,17,28,30,35,36. The ion accumulation is termed “ion trapping” and erodes electrochromic performance. Specifically, three different types of traps are found to degrade the electrochromic performance28,33: (i) shallow traps which degrade the colored state only; (ii) deep traps that degrade both colored and bleached states; and (iii) irreversible traps, in which the resided ions cannot be released any longer, degrade the short-wavelength region of optical transmittance and are considered to be a minor effect. Although extensive studies have been made33,37,38,39,40,41,42 and the ion detrapping procedure is shown to be effective to regain initial electrochromic performance, the origin of the traps and associated dynamics remain unknown which largely limits the understanding of traps and further attempts for trap suppression or even elimination.

Here we show direct experimental evidence of origins of all three kinds of traps that are present during ion trapping and detrapping processes. Shallow traps are due to a phase change from amorphous tungsten bronze (LixWO3) to orthorhombic lithium tungstate (Li2WO4) which largely suppresses the hopping from W5+ ↔ W6+ sites. Compare to WO3, the extra oxygen in an orthorhombic Li2WO4 molecular is from the electrolyte. Details of dynamics for deep traps have also been unveiled, specifically, a rejuvenation from deep traps is a multiple-step-determined process and the ion-releasing sequence starts from decomposition of coupled W4+-Li2O, amorphous Li2WO4 and coupled W4+-Li2WO4, revealing a complex structure reconfiguration in the host matrix. Irreversible traps are coupled W4+-Li2WO4 and are non-decomposable. In addition to the small polaron hoping between W5+ and W6+ sites, optical absorption in the short-wavelength region is due to bipolaron hopping between W4+ and W6+ sites, after formation of W4+ upon reversible intercalation or ion trapping process. Three forms of W4+ sites, reversible and irreversible, are also identified. Overall, our work unveils the dynamics of ion trapping and detrapping, and provides a general picture of electrochromism based on polaron hopping, in the representative electrochromic oxide, WO3. Electrochromism of the cathodic oxides (such as WO3, MoO3, Nb2O5, Ta2O5 and TiO2) is based on small ion intercalation and considered to obey polaron hopping1,15,43,44,45,46. Meanwhile, ion-trapping induced degradation is a ubiquitous phenomenon in cathodic electrochromic oxides16,17,28,46. For example, in addition to the model electrochromic material, WO328,33, ion trapping and detrapping also prevail in the rest of cathodic electrochromic oxides, i.e., MoO3, TiO2, Nb2O5 and Ta2O5. Therefore, our findings here not only pave the way to develop superior electrochromic devices based on WO3, but also serves as the guideline to explore the degradation dynamics for other cathodic electrochromic oxides. The methodologies of a combination of spectroelectrochemistry and other spectro- and microscopic characterizations can be well expanded to other electrochromic materials and ion intercalated systems.

Results and discussion

Performance degradation and rejuvenation

We start from the most used potential window for WO3, i.e., 2.0–4.0 V vs Li25,26,27. Figure 1a shows cyclic voltammetry (CV) curves of WO3, indicating a typical amorphous structure. The envelope area of the CV curves, from which inserted and extracted charge in each cycle can be calculated, are reducing as cycling proceeds. The inserted charge is larger than the extracted one, and this leads to ion accumulation in the electrode and decreases the reversibility (the difference between inserted and extracted charge in each cycle is shown in Supplementary Fig. 1). Open circuit potential (OCP) drops from 3.3 V at the pristine state to 3.1 V after 1000 cycles and the associated electrochromic performance (i.e., optical modulation, ∆T, defined as the transmittance in the bleached state (Tb) minus the one in the colored state (Tc).) exhibits a clear degradation (Fig. 1c). It can also be noted from the in-situ optical transmittance that degradation of ∆T only arises from a degraded colored state, whereas the bleached state maintains the same transmittance as the pristine state (Fig. 1c for the wavelength 550 nm and Fig. 1g for the full spectrum from 380 nm to 2250 nm; transmittances at other wavelengths can be seen in Supplementary Fig. 2). The coloration is due to typical polaron hopping between W5+ and W6+ sites when electrons are injected as described in Eq. 1 and the optical transmittance is basically flat except for a broad peak at the short-wavelength region. Bear in mind that the origin of the degradation of the colored state is as ascribed to shallow traps28,33 which evenly lift-up the full spectrum at the colored state upon cycling. This indicates that resided ions in shallow traps do not contribute to optical absorption28,32,33 but only “suppress” the W6+ ↔ W5+ hopping.

Fig. 1: Spectroelectrochemistry of WO3 during ion trapping and detrapping process.
figure 1

a, b CV data for various cycle numbers at 2.0–4.0 V (20 mV s−1) and 1.5–4.0 V (10 mV s−1), respectively. c In-situ optical transmittance at 550 nm upon various ion exchange processes. The sequence is 1000 CVs at 2.0–4.0 V (20 mV s−1), galvanostatic detrapping for 20 h, 10 CVs at 2.0–4.0 V (20 mV s−1), 20 CVs at 1.5–4.0 V (10 mV s−1) and galvanostatic detrapping for 20 h. Optical modulation, ∆T, defined as the transmittance in the bleached state (Tb) minus the one in the colored state (Tc). d, e Charge capacity variation at 2.0–4.0 V (20 mV s−1), and 1.5–4.0 V (10 mV s−1), respectively. f Enlarged view of Fig. 1c for 10 CVs in the range of 2.0–4.0 V (20 mV s−1) after galvanostatic degrapping and rapid degradation in the following 20 CVs in the range of 1.5–4.0 V (10 mV s−1), as pointed out by the pink arrow. g, h Full transmittance profiles (350–2250 nm) at 2.0–4.0 V (20 mV s−1) and 1.5–4.0 V (10 mV s−1), respectively. i Potential profiles during galvanostatic detrappingfor the two trapping states. j Current profiles during potentiostatic detrapping for the two trapping states.

Full size image

A so-called galvanostatic “detrapping” procedure was conducted after 1000 cycles, namely, a constant current of 3.0 μA cm−2 was applied to the degraded film aiming to regain the initial electrochromic performance (schematic diagram of detrapping, and other constant currents used for detrapping were also compared, as shown in Supplementary Fig. 3). Optical transmittance in the bleached state remains constant throughout the entire “detrapping” process. Performing subsequent 10 CV cycles after “detrapping” clearly demonstrates that both charge capacity and optical modulation are completely rejuvenated (Fig. 1c and g). Moreover, the detrapping procedure from shallow traps is a continuous process33 and is qualitatively different from the situation for deep traps, as will be shown later.

The rejuvenated film was then subjected to cycling in the range 1.5–4.0 V (Fig. 1b) with a purpose of studying on the practicability to extend the potential window of WO3 and the case of significant ion trapping. Charge capacity (Fig. 1e and Supplementary Fig. 1b) as well as optical transmittance (Fig. 1c and enlarged view in Fig. 1f) degrade more rapidly than for the case between 2.0 and 4.0 V, indicating severe ion trapping in the electrode and eroded electrochromic activity. The OCP drops from 3.3 V to 2.4 V after 20 cycles, and more importantly, optical transmittances at both colored and bleached states degrade. The origin of degradation of both colored and bleached states is ascribed to deep traps. Clearly, ions residing in deep traps are responding in a different manner as compared to shallow traps. Moreover, the broad transmittance peak at the short-wavelength region becomes flat (Fig. 1h, 1st colored, switching process can be seen from Supplementary Movie 1) or even shows a valley (Supplementary Fig. 4) while the long-wavelength region increases. This shows that the polaron hopping mode in the electrode has also changed in the range 1.5–4.0 V.

An analogous “detrapping” procedure was conducted to release the ions from deep traps. Three distinct steps can be noted from optical transmittance profiles (green area in Fig. 1c): step one, full optical spectrum slightly lifts up at the very beginning of the detrapping process; step two, full spectrum maintains as detrapping proceeds; step three, a sharp increase of the transmittance reaching to its pristine state, except in the short-wavelength region. The variation of transmittance in the whole spectral range (Supplementary Fig. 5) is consistent with the single-wavelength (550 nm) data except for the un-recovered part in the short wavelength region (350 nm–550 nm), whose origin is ascribed to irreversible traps. Here, we confirm that ions residing in deep traps can also be released whereas ions in irreversible traps become permanently immobile. Moreover, the potential profiles during galvanostatic detrapping processes (Fig. 1i), also indicate that ions released from shallow and deep traps are distinctly different. Consistent results can also be found when detrapping is carried out through a potentiostatic process (Fig. 1j), where the current density shows a peak, analogous to the apparent valley of the potential profile during galvanostatic detrapping, coincident with the occurrence of the sharp increase of optical transmittance33. It should be pointed out that, for a-WO3 samples, the deposition parameters were chosen in order to study the degradation and rejuvenation on a reasonable time scale. It shows that a-WO3 thin films with low porosity possessed much slower degradation rate. As revealed in Supplementary Fig. 6, colored states of a denser WO3 film degraded by only ~14% after 3000 cycles, where the a-WO3 films used in this paper degrade by ~30% after 1000 cycles. The characteristics of ion trapping/detrapping processes for WO3 thin films with different porosities are fully consistent. Therefore, it is conclusive that, ion trapping takes place in both porous and dense WO3 films. In fact, the ion trapping and detrapping are found to prevail in all other cathodic electrochromic oxides as well, i.e., MoO347, TiO246, Nb2O5 and Ta2O5 as revealed in Supplementary Fig. 7.

Origins of shallow traps

We combine X-ray photoelectron spectroscopy (XPS), Raman spectroscopy and transmission electron microscopy (TEM) to depict the entire picture of ion trapping and detrapping processes. Evaluation of valence states by XPS is shown in Fig. 2a. a-WO3 at the pristine state is first studied as a reference. Four characteristic peaks are noted from the W-4f spectra at this state (Fig. 2a, pristine state). Considering the spin-orbit splitting, the two paired peaks are assigned to W6+ and W5+ states with an absolute W6+ dominance37,40,48 (Fig. 2b and Supplementary Table 1 present the proportions of each W state), which is in good agreements with the observations from others41. The O-1s spectra show two contributions with peaks at 530.5 and 532.1 eV, respectively. In addition to the main peak at 530.5 eV originating from matrix oxygen, the peak at 532.1 eV is assigned to oxygen atoms bonding to adsorbed impurities49,50. Naturally, no characteristic peak of Li-1s was detected in the pristine film.

Fig. 2: Composition and structural characterization of shallow traps induced degradation (cycling in the range of 2.0–4.0 V).
figure 2

a XPS characterization for WO3 at various states and Pristine state is shown as a reference. Yellow circles represent the raw data, purple and blue lines stand for the signals from W6+ and W5+. The shaded regions by the lines indicate the portion of marked valency. b the proportions of each W valency. c Raman characterization for WO3 at various states. Indium tin oxide, denotes as ITO, served as the substrate for film deposition. The two yellow dash lines represent the Raman signals from ITO. The blue dash line indicates the peak position for W+6 = O stretching. Shaded green and yellow regions illustrate the W+5-O and W+5 = O, as well as W+6-O stretching. d–f HR-TEM images of WO3 film at pristine state, 1000th bleached state and after detrapping, showing the formation of orthorhombic Li2WO4 upon long-term cycling and its release after detrapping. Insets: associated FFT patterns.

Full size image

We consider films working in the range of 2.0–4.0 V first, where shallow trap is the main cause for performance degradation. As Li ions are inserted (Fig. 2a, 5th colored state) to achieve dark blue, an obvious Li-1s signal can be seen at ~55 eV47,49,50,51,52. Simultaneously, W6+ are reduced to W5+ as a result of electron insertion from the external circuit, and optical absorption is achieved due to small polaron hopping between W5+ ↔ W6+ sites15,21, as illustrated in Eq. 1. Because of the low photoionization cross-section of Li in XPS, the x values in LixWO3 were calculated here by different proportions of W valences, instead of deriving it directly from Li-1s signals. x is found to be 0.43 at this stage, this is slightly larger than 0.35 which is believed to serve as a ‘rule of thumb’ for safe long-term device operation6. Once Li ions are released through a reverse scan (i.e., from 2.0 V to 4.0 V), optical transmittance returns to its initial state and all XPS spectra are identical to the reference (Fig. 2a, 5th bleached state). At this stage, the ion trapping effect is not noticeable because of the limited number of cycles (only 5 CVs).

At the colored state of the 1000th cycle, the Li-1s signal increases (Fig. 2a, 1000th colored state). Note that the contribution to the Li-1s signal here is both from inserted Li ions in the 1000th cycle and accumulated Li ions in shallow traps over the past 1000 cycles. This statement is reasonable because the intensity of the W5+ signal becomes weaker compared to the one at the 5th cycle, which is due to the lower amount of inserted Li ions at the 1000th cycle, as confirmed by the decreased reversible charge capacity in Fig. 1d. Additional evidence is that the intensity of Li-1s also largely decreases at the bleached state of the 1000th cycle while W-4f returns to its initial state (Fig. 2a, 1000th bleached state), indicating that the weak Li-1s intensity only arises from Li ions in shallow traps. Since the amount of W5+ decreases, the efficiency of small polaron hopping between W5+ ↔ W6+ sites drops (Supplementary Fig. 8), and thus optical modulation, i.e., ∆T, is significantly reduced. The results shown here emphasize that accumulated ions in shallow traps associated with pairing electrons do not reduce W6+ to W5+ or even W4+ in the host matrix, suggesting the non-coloration nature of the shallow traps, therefore optical transmittance at the bleached state can always reach its initial value (Fig. 1c, g). In addition, no signal from Cl is detected (Supplementary Fig. 9) which means it is not involved into trapping processes. One may note that it shows a nonnegligible binding energy position shift of both W6+ and W5+ (as well as W4+ in the next part in Fig. 3a) which was also observed from others37,53. We temporarily ascribe this slight shift to coordination environment variation upon ion insertion and extraction.

Fig. 3: Trapping and detrapping dynamics for WO3 at 1.5–4.0 V.
figure 3

a XPS data for WO3 at various trapping/detrapping states. Yellow circles represent the raw data. Purple, blue and black lines, with associated shaded regions, stand for the signals from W6+, W5+ and W4+, respectively. b the proportions of various W valences. c Raman data for WO3 at various trapping/detrapping states. Indium tin oxide, denotes as ITO, served as the substrate for film deposition. The two yellow dash lines represent the Raman signals from ITO. The red, purple and blue dash lines indicate the peak position for W+4-O stretching, Li2WO4 vibrating and W+6 = O stretching, respectively. Shaded green and yellow regions illustrate the W+5-O and W+5 = O, as well as W+6-O stretching. d, e HRTEM images of WO3 at 3rd colored state and 20th bleached state, indicating the amorphous characteristics of Li2WO4 in these conditions. Insets: associated FFT patterns.

Full size image

Raman spectra are shown in Fig. 2c, for a pristine film on ITO/glass (Fig. 2c, pristine state). Five broad peaks are noted whereas the peaks centered at ~560 cm−1 and ~1100 cm−1 are assigned to ITO/glass (Supplementary Fig. 10). Two peaks centered at ~665 cm−1 and ~775 cm−1 are assigned to W+6-O stretching modes of WO352,54 (also see Supplementary Fig. 11 for the Raman spectrum of commercial WO3 powders with a grain size of ~20 nm and in hexagonal phase). The peak centered at ~950 cm−1 is assigned to the stretching mode of W+6 = O, which is characteristic for a-WO3, arising from terminal oxygen atoms on the surfaces of clusters and/or microvoid structures55,56. When ions are inserted, the intensity of stretching modes due to W+6-O and W+6 = O clearly decreases (Fig. 2c, 5th colored state) while an enhanced intensity from ~330 cm−1 to ~450 cm−1 is noted because of the emerging stretching modes of W+5-O and W+5 = O57. This demonstrates that some of the W+6-O and W+6 = O bonds are broken and new W+5-O and W+5 = O bonds are formed due to the inserted Li ions. As expected, this is consistent with the XPS results that insertion of an ion-electron pair leads to a reduction of W6+ to W5+ in tungsten bronze (LixWO3). One may note that the signal between ~330 cm−1 and ~450 cm−1 here is not sharp, this is because the increased fluorescence due to coloration37. The amorphous nature of the WO3 electrode may also decrease peak intensity (Supplementary Fig. 12 for Raman measurements of WO3 on W/Si substrates to distinguish this disturbance). As ion/electron pairs are extracted, the features of the Raman spectra return to the ones analogous to the pristine state (Fig. 2c, 5th bleached state). After 1000 cycles, the intensity of W+6-O and W+6 = O stretching modes are enormously reduced at the colored state (Fig. 2c, 1000th colored state). Even after bleaching (Fig. 2c, 1000th bleached state), peak intensities are only partly recovered, suggesting that ions residing in shallow traps suppress the stretching of W+6-O and W+6 = O modes rather than reduce W6+ to W5+.

High-resolution TEM (HRTEM) and fast Fourier transform (FFT) patterns of a pristine sample indicate an amorphous nature of the as-deposited samples (Fig. 2d). At the bleached state of the 1000th cycle (Fig. 2e), nano crystalline grains embedded in the amorphous host are found and FFT patterns of these nanograins are assigned to orthorhombic Li2WO4 (JCPDS No: 28-0596). We investigated more than 10 samples after long-term cycling and the results are consistent and showing indisputable evidence that orthorhombic Li2WO4 were formed. Some representatives are provided in Supplementary Fig. 13. Grazing incidence XRD has also been performed on a severely aged WO3 film (cycled in the range of 2.0–4.0 V) and confirmed the formation of orthorhombic Li2WO4 (Supplementary Fig. 14). After detrapping, these nanograins vanished as validated by both HR-TEM and the FFT pattern, leaving the amorphous phase of the electrode (Fig. 2f). Thicker WO3 films (i.e., 800 nm) were also deposited and subjected to long term cycling (2000 CV cycles). TEM results show that the concentration of orthorhombic Li2WO4 nanograins has an obvious incremental gradient distribution from ITO substrate to film surface where the formed orthorhombic Li2WO4 nanograins are very rare near ITO substrate (Supplementary Fig. 15), suggesting orthorhombic Li2WO4 is more easily formed near the film surface.

As shown by spectroelectrochemistry in Fig. 1c, XPS in Fig. 2a and Raman spectroscopy in Fig. 2c, the initial performance is completely regained after detrapping. This suggests that orthorhombic Li2WO4 was effectively decomposed and Li ions were fully extracted from the host matrix. Bearing in mind that the W6+ valency is maintained in orthorhombic Li2WO4, therefore the bleached state can always reach its initial value in every cycle upon long-term cycling (Fig. 1c, g). It should be pointed out that no signal was detected for orthorhombic Li2WO4 from Raman measurements (Fig. 2c), which is consistent with the observation reported in ref. 37. From the TEM analysis that the formed orthorhombic Li2WO4 nanograins were found small in quantity and rather dispersed in the amorphous host, as well as the amorphous host gives a strong background in Raman measurements, leading to the un-observed signal. However, the formed orthorhombic Li2WO4 accounts for the gradually degraded colored state due to: (i) formation of a Li2WO4 molecule is equivalent to consume a WO6 octahedron which was used to achieve reversible coloration; (ii) dispersed Li2WO4 nanograins in an amorphous host, especially relatively higher concentration near the surface, also block the migration of Li-ions. A combination of the two contributions leads to a degradation of electrochromic performance.

So far, it can be concluded that formation of orthorhombic Li2WO4 during long-term cycling is the origin of shallow traps in a-WO3 films. It should be noted that, in addition to the inserted Li ions from electrolyte, there is an extra oxygen atom in an orthorhombic Li2WO4 molecular, as compared to the molecular of a WO3. Due to the incremental gradient distribution of the formed orthorhombic Li2WO4 nanograins, it suggests the extra oxygen is from the electrolyte (i.e., LiClO4-PC). To validate this hypothesis, LiClO4-PC was replaced by LiPF6-PC and an accelerated degradation was observed (Supplementary Fig. 16a). XPS measurements showed gradient F signal was found for samples at the bleached state after long term cycling (Supplementary Fig. 16b, c). It is known that F− has a larger electronegativity than O2−, refs. 58,59, which indicates F− can be easily incorporated into WO3 than O2− and results in the accelerated degradation. Analogous to the films cycling in LiClO4-PC that no-detected Cl signal, P signal was either found for films cycled in LiPF6-PC (Supplementary Fig. 16d). The solvent of PC has also been replaced to EC/DEC (i.e., the new electrolyte is LiClO4-EC/DEC) and yielded a consistent result as compared to LiClO4-PC (Supplementary Fig. 17), suggesting the external oxygen is not from solvent. The above results reveal that, to build WO3 based electrochromic devices, the selection of salt for the electrolyte should be more seriously considered than the solvent. The extra oxygen from ITO substrate was also excluded, as shown in Supplementary Fig. 18. Moreover, oxygen deficient tungsten oxide (WO3-z) thin films, which appeared to be bluish in the pristine state due to the large number of oxygen vacancies induced W5+ (Supplementary Fig. 19), were subjected to the same cycling as done for a-WO3. The gradually increased transmittance at bleached states upon cycling was due to the oxidization from W5+ to W6+, which has to be achieved through oxygen incorporation from electrolyte. Finally, the depth profile of oxygen (Supplementary Fig. 20) showed obvious decrease from film surface to ITO, indicating that oxygen is also from the electrolyte, and is in good accordance with, and also well explains, the TEM results which shows orthorhombic Li2WO4 nanograins concentrate in near the surface.

Origin of deep and irreversible traps

Next, we consider the deep and irreversible traps, i.e., films cycled in the range 1.5–4.0 V (Fig. 3). At colored state of the 3rd cycle (Fig. 3a, 3rd colored state), a large amount of Li ions is inserted as indicated by the strong Li-1s signal. From W-4f spectra, three paired peaks are noted and this is different from the case of cycling in the range 2.0–4.0 V. The newly emerged pair of peaks are assigned to the spin-orbit splitting of W4+ 40,48. The formed W4+ largely decreases the optical transmittance at the short-wavelength region as presented in Fig. 1h (see also Supplementary Fig. 4). This is in contrast to small polaron hopping of W5+ ↔ W6+ transitions (i.e., Eq. 1), from which a broad transmittance peak in the short-wavelength region is shown at the colored state. Based on the W proportions, x in LixWO3 at this stage is calculated to be 0.56, larger than 0.43 at the 5th colored state in 2.0–4.0 V; this is in accordance with the larger charge capacity and stronger Li-1s signal at this state.

When a positive scan is applied for bleaching (Fig. 3a, 3rd bleached state), the valence states in the W-4f spectra return to their initial state and the Li-1s signal also disappears, suggesting a reversible intercalation at the current stage (i.e., ion trapping, if it occurs, is insignificant). At the 6th cycle bleached state (Fig. 3a, 6th bleached state), substantial ion trapping can be noted from the Li-1s spectra. The inserted electrons also reduce W6+ to W5+ and W4+, as seen from the W-4f spectra. Meanwhile, a strong absorption can be observed in the short-wavelength region due to the strong increase of W4+ sites (Supplementary Fig. 4, 1.5–4.0 V, 6th bleached). This is in striking contrast to the shallow traps (Fig. 2a, 1000th bleached state), where resided ions/electrons do not change the valence state of tungsten and thus the optical transmittance can always be identical to the initial value.

As shown in Fig. 1h, optical modulation at the 20th cycle is extremely faint and the intensity of the W4+ and Li signals are further enhanced at both colored and bleached states (Fig. 3a, 20th colored state and 20th bleached state). In addition, a peak at 528.3 eV appears in the O-1s spectra, suggesting that formation of Li2O50,60 takes place as cycling proceeds. In fact, a trace of Li2O is already noticed at the bleached state of the 6th cycle (Fig. 3a, 6th bleached state).

The increased proportion of W4+ at these states (Fig. 3b and Supplementary Table 2) yields an enhanced optical absorption in the short-wavelength region, thus confirming that optical absorption at short wavelengths originates from bipolaron hopping between W4+ ↔ W6+ sites. Here, small polaron hopping between W4+ ↔ W5+ sites is unlikely, because W5+ gradually vanishes during continuous ion insertion (Fig. 3b and Supplementary Table 2), while an increased/decreased optical absorption at short-/long-wavelength regions in the full spectrum is simultaneously observed (Supplementary Fig. 4). This is typical for a combination of decreased/increased hopping by W5+ ↔ W6+ and W4+ ↔ W6+ transitions (switching of the two modes can be seen from Supplementary Movie 1). The optical absorption in the short-wavelength region also suggests that a higher energy is needed to activate W4+ ↔ W6+ hopping. Therefore, in addition to the polaron hopping mode as described in Eq. 1, a new hopping mode between W4+ ↔ W6+ also exist and can be written as:

$${{{{{{{\rm{W}}}}}}}_{({{{{{\rm{a}}}}}})}}^{4+}+{{{{{{{\rm{W}}}}}}}_{({{{{{\rm{b}}}}}})}}^{6+}\mathop{\leftrightarrow }\limits^{{{hv}}_{2}}{{{{{{{\rm{W}}}}}}}_{({{{{{\rm{b}}}}}})}}^{6+}+{{{{{{{\rm{W}}}}}}}_{({{{{{\rm{b}}}}}})}}^{4+}$$
(2)

When the severely aged film has gone through a galvanostatic detrapping process, XPS signals for both Li and Li2O disappear (Fig. 3a, after detrapping). However, a weak signal for W4+ remains which is consistent with the observation of irreversible traps that are not recoverable and result in optical absorption at short wavelengths as presented in Fig. 1h. It should be noted that W4+ sites cannot be formed without inserted Li ions, thus the un-observed Li signal here is only due to the low concentration of Li ions as well as the detection limit for light elements in XPS.

As illustrated in Fig. 3c of the Raman spectra, the stretching modes of W+6-O and W+6 = O in a-WO3 are already severely inhibited at the colored state in the 3rd cycle (Fig. 3c, 3rd colored state). This is due to the larger number of inserted ions/electrons, as compared to the case of 2.0–4.0 V (see Fig. 1d, e). More importantly, two more peaks emerge: one centered at ~220 cm−1 due to stretching modes of W+4-O56 which is well consistent with the observation of W4+ signals from the XPS measurements (Fig. 3a, 3rd colored state); the other centered at ~282 cm−1 and due to Li2WO437,61. The HR-TEM images and the diffuse rings in the FFT pattern (Fig. 3d) demonstrate that the structure is still discorded at the colored state of the 3rd cycle, indicating that the formed Li2WO4 is amorphous and mixed with the a-WO3 host matrix. As the film is bleached (Fig. 3c, 3rd bleached state), peaks for stretching modes of W+6-O and W+6 = O are recovered, while the peaks assigned to W+4-O and amorphous Li2WO4 vanished, suggesting that both W4+ and Li2WO4 are reversible at this stage because of the limited number of cycles. Here the appearance and disappearance of W+4-O and Li2WO4 are concomitant, suggesting a coupling between W4+ and Li2WO4. This is rational, because there is an extra oxygen atom in Li2WO4 as compared to the initial WO3, and one of the three oxygen atoms which originally binds to W(a)6+ migrates to a nearby W(b)O3 unit and forms Li2W(b)O4, leaving the W(a) atom to be reduced to W(a)4+ (Fig. 4a, blue dash-line square). This configuration is in agreement with our experimental observations of the W4+-Li2WO4 coupling and explains well the origin of the extra oxygen atom in coupled W4+-Li2WO4. As noted, this is different from the case for long-term cycling in the range of 2.0–4.0 V, where the external oxygen is gradually incorporated from electrolyte.

Fig. 4: Various states of WO3 during Li ion intercalation, trapping and detrapping.
figure 4

a Schematic illustration of the formation of independent W5+ and W4+, coupled W4+-Li2WO4 and coupled W4+-Li2O in amorphous WO3 host matrix. The arrows illustrate the possible migrate paths for oxygen. The white cubes, denote as Vo0, indicate the oxygen vacancies. b In-situ optical transmittance during the ion trapping and detrapping process. The current density profile upon potentiostatic detrapping shows a peak when the sharp increase of the optical transmittance occurs. Insets mark the 5 selected points to be tested by XPS and Raman spectroscopy. c XPS results for the selected points as well as during potentiostatic ion insertion. Yellow circles represent the raw data. Purple, blue and black lines, with associated shaded regions, stand for the signals from W6+, W5+ and W4+, respectively. d Proportions of each W state of the selected points. e Raman results for the selected points as well as during potentiostatic ion insertion. Indium tin oxide, denotes as ITO, served as the substrate for film deposition. The two yellow dash lines represent the Raman signals from ITO; the red, purple and blue dash lines indicate the peak position for W+4-O stretching, Li2WO4 vibrating and W+6 = O stretching, respectively. Shaded green and yellow regions illustrate the W+5-O and W+5 = O, as well as W+6-O stretching.

Full size image

The bleached state at the 6th cycle, as well as colored and bleached states at the 20th cycle are discussed together, since little difference is noted among these three states (Fig. 3c, 6th bleached, 20th colored and 20th bleached). In these states, peaks for W+6-O and W+6 = O stretching modes are enormously inhibited while peaks for W4+-O and amorphous Li2WO4 are rather intense. This indicates that films at these states are severely degraded and a limited number of active sites are available for coloration, thus the films almost lose their optical modulation, as confirmed by the in-situ spectroelectrochemistry results in Fig. 1e, h. On the other hand, no peak of Li2O in the Raman spectra is detected, suggesting the formation of Li2O is limited to the film surface and is small in quantity, which is also confirmed by XPS depth profile (Supplementary Fig. 21). The Li2O on the surface is also in accordance with its sequence of formation and decomposition, as will be shown later.

Once the detrapping is conducted, peaks for the stretching modes of W+6-O and W+6 = O are approximately recovered, suggesting a quasi-complete rejuvenation of the film. Besides, signals at ~220 cm−1 and ~282 cm−1, although weak, are still present, indicating that part of the coupled W4+-Li2WO4 can’t be decomposed any longer. The associated optical transmittance in the short-wavelength region is not completely recovered (Fig. 1h). We therefore conclude that the origin of irreversible traps is non-decomposed W4+-Li2WO4. In the XPS spectra, a trace of Li2O is noticed at the 6th cycle and this intensity strongly increases at the 20th cycle. After detrapping, the signal assigned to Li2O vanished first. In fact, formation and decomposition of Li2O is coupled to W4+, which will be justified in the next section in order to clarify the origin of deep traps.

Capturing the detrapping dynamics of deep traps

So far, the detrapping dynamics for the deep traps is still unknown, and this will be revealed herein. For severely aged films in the range of 1.5–4.0 V, the optical transmittance shows a three-step characteristic upon galvanostatic detrapping. Five points are selected, as marked in Fig. 4b, to elaborately uncover the associated dynamics. Point I was selected at the middle of the first milder transmittance increase. A slight drop of W4+ and increase of W5+ intensity is noted (Fig. 4c, Point I state) when comparing to the deep trapped state (proportions of each W state can be seen in Fig. 4d and Supplementary Table 3). This indicates that the slightly increased optical transmittance is due to partial oxidation of W4+ upon detrapping. Thus, the small polaron hopping efficiency involving W4+ is slightly declined, yielding an increased optical transmittance. On the other hand, the peak intensity for amorphous Li2WO4 in the Raman spectra is hardly varied (Fig. 4e, Point I state), suggesting that the present release of W4+ is independent of Li2WO4. From the XPS spectra, it can be found that the signal assigned to Li2O vanished at stage II (Fig. 4c, Point II state), and this indicates that the W4+ released here is coupled with Li2O. In fact, the formation Li2O consumes an oxygen atom from a WO6 octahedron and induces a W4+ formation simultaneously as shown in Fig. 4a (purple dash-line square).

Points II, III and IV are selected to be just after the first milder transmittance increase, at the middle of the plateau, and at the turning point prior to the second transmittance increase, respectively. From point II to IV, the intensity of W-4f peaks show almost no variation whereas Li-1s shows a consecutive reduction until it vanished. Note that the vanished Li-1s signal here is due to the low concentration of Li ions at this state, considering the detection limit for light elements of XPS. Consistently, Raman spectra (Fig. 4e, point II, III and IV states) show that stretching modes of W+6-O and W+6 = O are steadily enhanced because of ion release. Therefore, the combination of un-varied W-4f signal, the decreased intensity of Li-1s in XPS and the recovered stretching modes of W+6-O and W+6 = O in Raman spectra suggest that the ion release is from decomposition of amorphous Li2WO4 without varying the valence state of W. Besides, when cycled at 2.0–4.0 V after point IV, both CV curves and optical modulations show notable rejuvenation (Supplementary Fig. 22), again providing a strong proof for this claim.

Point V was selected to be at the middle of second transmittance increase where the transmittance is largely recovered. It can be noted that the peak intensity for W4+ is greatly reduced in both XPS and Raman spectra, indicating a decomposition of the coupled W4+-Li2WO4. This process is very rapid and different from the situation for shallow traps. Because of the rapid decomposition of coupled W4+-Li2WO4, the current density shows a peak simultaneously with the occurrence of the sharply increased transmittance (Fig. 4b). As observed in Fig. 4e (Point V state), when the detrapping process is finished, the Raman spectra return to the pristine state, expect for the coupled W4+-Li2WO4 from irreversible traps. One may notice that the peak at ~282 cm−1 seems to be increased, but this is because of vanished fluorescence vanishes due to the bleaching of the film after detrapping, thus reducing the background intensity around this range (Supplementary Fig. 12 for Raman measurements of WO3 on W/Si substrates to distinguish this disturbance).

Now, we discuss the different forms of W4+ sites. When conducting potentiostatic insertion of Li ions at 2.0 V for one minute, W4+ is formed (Fig. 4c, e, 2.0 V, 1.0 min state). However, no signal from Li2WO4 or Li2O is found in XPS and Raman spectra, suggesting that W4+ forms in the same fashion as in tungsten bronze (LixWO3), as shown in Fig. 4a (pink dash-line square). We term the current W4+ sites as type I. Type I forms first and, more importantly, they are reversible, as proved by a complete ion extraction at 4.0 V, as shown in Supplementary Fig. 23. As Li ions are continuously inserted at 2.0 V or by expanding the low potential limit to 1.5 V, the host matrix gets reconfigured to form coupled W4+-Li2WO4 (type II), as indicated in Fig. 4c, e (2.0 V, 20 h) and Fig. 3c. Type II can be reversible, releasable or irreversible as we discussed in the context of Fig. 3, and this is likely dependent on the associated configurations. Finally, coupled W4+-Li2O (type III) is formed after the generation of W4+-Li2WO4, and decomposed at the very beginning of the detrapping process. In fact, Li2O can even be decomposed at 4.0 V, as confirmed by the release of coupled W4+-Li2O after potentiostatic detrapping at 4.0 V (Supplementary Fig. 24). Overall, W4+, in any form, leads to optical absorption in the short-wavelength region due to the emergence of W4+ ↔ W6+ hopping. The spectroelectrochemistry, XPS, and Raman results support the fact that decomposition of these three forms of W4+ is just the opposite process to their formation. However, a small quantity of coupled W4+-Li2WO4 units are not decomposable are described as irreversible traps.

Based on the above discussion and experimental evidence, Fig. 5 provides a general picture of the ion intercalation, trapping and detrapping in a-WO3. Optical absorption in tungsten bronze originates from small polaron hopping W5+ ↔ W6+ transitions, which results in optical absorption in the whole range, especially in the long-wavelength region (i.e., 550 nm–2250 nm). W4+ can also form in tungsten bronze depending on the degree of ion insertion, and associated W4+ ↔ W6+ transitions lead to an optical absorption in the short-wavelength region. After long-term cycling, orthorhombic Li2WO4 is formed in which W6+ sites are maintained, but the reversible sites for intercalation are largely reduced. Therefore, polaron hopping efficiency between W5+ ↔ W6+ drops, yielding a diminished optical modulation. Coupled W4+-Li2WO4, amorphous Li2WO4 and W4+-Li2O can also form and their reversibility depends on the potential window as well as the extent of ion trapping. Some coupled W4+-Li2WO4 could even not be decomposed and are denoted as irreversible traps. The origin of deep traps is complex, where ion trapping is a multiple-step-determined process, and is composed of the formation of W4+-Li2WO4, amorphous Li2WO4 and W4+-Li2O. The ions residing in shallow and deep traps can be fully released through either a potentiostatic or galvanostatic detrapping process, therefore initial electrochromic performance can be completely regained. Another difference between shallow and deep traps is that, disproportionation reaction takes places to form coupled W4+-Li2WO4 in deep traps and extra oxygen is from the adjacent oxygen in the host matrix rather than electrolyte.

Fig. 5: A general picture of the ion intercalation, trapping and detrapping in a-WO3.
figure 5

It shows the formed species, reversibility and how it varies the optical modulation.

Full size image

It should be pointed out that the origin of shallow traps was never touched upon, due to the unvarying bleached state. Some hints of deep traps were indicated in previous research work. Hashimoto et al.32 subjected a-WO3 to heavy lithiation in an electrolyte containing 3000 ppm water, and observed that 7Li2WO4∙4H2O was formed in the colored state where the film showed a black-brown color. It was proposed that W4+ was responsible for the black-brown color. Indeed, W4+ is the main cause of this color, i.e., by the strong absorption at the short-wavelength region, but it bonds to a more complex structure as we revealed above. Takayanagi et al.37 conducted in-situ hard XPS also on heavily lithiated a-WO3 and proposed that formation of Li2WO4 eroded the electrochromic performance. However, direct identification of Li2WO4 was missing. Cao et al.39 directly observed formation of Li2WO4 through in-situ TEM in heavily lithiated hexagonal WO3 nanowires. However, no long-term cycling was conducted in both Takayanagi’s and Cao’s work, which makes it hard to connect the formed Li2WO4 to various traps and the associated electrochromic performance.

With a combination of in-situ spectroelectrochemistry and other ex-situ spectroscopic techniques, the origin of the three different types of traps are demonstrated, and the ion trapping and detrapping dynamics as well as coloration mechanisms in a-WO3 thin films is provided. Ion trapping is also found to take place in films with low porosity, although the degradation rate is relatively slower. The origin of shallow traps is orthorhombic Li2WO4 which formed during long-term cycling in the range 2.0–4.0 V. Shallow traps degrade the colored state by immobilizing reversible intercalation sites and block migration channels for Li-ions. As the potential window is expanded to 1.5–4.0 V, deep traps formed which degraded both the bleached and colored states. Importantly, deep traps are found to be more complex, composed of coupled W4+-Li2WO4, amorphous Li2WO4 and coupled W4+-Li2O. The detrapping process shows that ion release from deep traps is sequential, where decomposition of W4+-Li2O occurs first, followed by decomposition of amorphous Li2WO4. Decomposition of coupled W4+-Li2WO4 takes place last, and part of it is unable to be decomposed which accounts for the irreversible traps. In addition, three forms of W4+ are found in a-WO3 whose reversibility is associated with the coupled species. We also confirm that polaron hopping due to W6+ ↔ W4+ transitions results in optical absorption at the short-wavelength region. Our findings demonstrated in this paper provide a general picture of ion reversible intercalation, trapping and detrapping in a model electrochromic material, WO3, which is consultative to other materials and devices of ion intercalation based. For example, ion trapping induced degradation was also found in other cathodic oxides (i.e., Nb2O5, Ta2O5, MoO3 and TiO2) whereas ion detrapping was proved to be effective. With the understanding of trapping/detrapping dynamics, this work may pave a way for future attempts to suppress or even eliminate traps with the aim to develop superior electrochromic devices. Moreover, the methodologies employed here can be well expanded to other electrochromic materials.

Methods

Materials

All the materials are available commercially and used as received. The target was a 3-inch-diameter plate of metallic tungsten/niobium/tantalum (99.999%). Argon and oxygen gas are both with 99.998% purity. Clean In2O3: Sn (ITO) coating glasses with a resistance of 60–80 Ω per square are used as the substrate. Lithium foil (Alfa Aesar, 99.9%), propylene carbonate (PC, Sigma-Aldrich, 99.7%), Diethyl carbonate (DEC, Aladdin, 99%), Ethylene carbonate (EC, Aladdin, 99%), lithium perchlorate (LiClO4, Aladdin, 99.99%), lithium hexafluorophosphate (LiPF6, RHAWN, 99.5%) were utilized without initial treatments.

Thin film deposition

Thin films of WO3, Nb2O5 and Ta2O5 were deposited by reactive DC magnetron sputtering. The distance between target and substrate is ~10 cm. The substrates were not intentionally heated during the film deposition. The deposition chamber was first evacuated to ∼6 × 10−5 Pa. Pre-sputtering took place in argon (99.998%) and oxygen (99.998%) for 5 min prior to sample deposition. The power to the target was set at 200 W. For WO3, the total pressure during sputtering was maintained at 4.0 Pa, O2/Ar gas-flow was kept at a constant value of 5/35 sccm; for Nb2O5, the total pressure was 1.0 Pa, O2/Ar gas-flow was 5/45 sccm; for Ta2O5, the total pressure was 4.0 Pa, O2/Ar gas-flow was 5/30 sccm. The present conditions were chosen in order to produce porous films and to study degradation and rejuvenation on a reasonable time scale. As for the dense WO3 films, the pressure was 1.0 Pa, with all the other parameters unchanged. The substrate is rotated during sputtering to ensure uniform coating. The thickness of the prepared films we used in this paper are about 300 ± 30 nm, which can be varied by changing the deposition time.

Electrochromic and electrochemical measurements

The electrochemical and electrochromic measurements were carried out in an argon glove box with water/ oxygen content below ∼0.01 ppm., using a custom-made spectroelectrochemical cell in the three-electrode configuration. The WO3, Nb2O5 and Ta2O5 film served as the working electrode and was electrochemically cycled in 1 M LiClO4 dissolved in propylene carbonate (or 1 M LiPF6 in PC, or 1 M LiClO4 in EC/DEC). Both the counter and reference electrodes were Li foils. Cycle voltammetry (CV), chronoamperometry (CA) and chronopotentiometry (CP) were conducted by an electrochemical workstation (Bio-logic, model VSP). For the CV measurement, the sweep rate is 20 mV s-1 within 2.0–4.0 V to yield shallow trapped state, and 10 mV s-1 within 1.5–4.0 V to yield deep trapped state, respectively. Charge capacity Q (in units of mC cm–2) was determined from cyclic voltammetry data by

$$Q=\int \frac{{jdV}}{s}$$

where j is current density (in mA cm–2), s is the sweep rate (in V s–1), and V is the voltage. Both CA and CP techniques were used to extract Li+ ions from the films; specifically, a constant potential of 5.8 V, or a constant current of 3 μA cm−2 was applied in the direction opposite to the one yielding Li+-ion insertion in the host material. The corresponding potential level (5.5 V or more) will lead to electrolyte degradation, which accounts for a large amount of the charge passed. Depending on the purpose of the measurement, a sequence of CV, CA or CP with certain orders were programed to operate. In situ optical transmittance was recorded by using a fiber-optical instrument from Ocean Optics (QEpro, Ocean Optics, USA). The spectral range was 350–2250 nm, and optical data were recorded simultaneously with the electrochemical measurement. The sample is positioned in a quartz electrochemical cell, between a tungsten halogen lamp and the detector. The 100% level was taken as the transmittance recorded before immersion of the sample into the quartz cell with electrolyte. Both single wavelength data and full range (350–2250 nm) spectra were recorded.

Materials characterization

Thickness was measured by using a AlphaStep D-300 from KLA Corporation, USA. X-ray diffraction (XRD) was performed on an X-ray diffractometer (Rigaku Smartlab) with the incident radiation at 40 kV and 200 mA using Cu Kα (λ = 1.5418 Å) radiation. The Selected area electron diffraction (SAED) pattern of the films was examined using a transmission electron microscope (TEM; Tecnai F30, FEI, USA) operated at 300 kV. The TEM samples were prepared using focused ion beam (FIB; Dual Beam Heilos Nanolab 600i, FEI) milling to be 50–100 nm in thickness. Carbon is coated on the sample to protect sample before milling. XPS analysis was conducted using a Escalab Xi+ from Thermo Fisher Scientific with Al Kα radiation, hυ = 1486.7 eV. Raman spectroscopy was conducted using a LabRAM HR spectrometer. Spectra were acquired using a 532 nm laser at constant power (25%). All the tested samples were taken out of the electrolyte cell, and washed with dimethyl carbonate (DMC, Sigma-Aldrich, >99.5%) to rinse off the residual electrolyte on the surface. The rinsing and transfer process were conducted in the glove box with oxygen and water both less than 0.01 ppm. The sample was transferred to the XPS device through the vacuum transfer chamber, during this process, water and oxygen will not interfere with the sample.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.