Galvanostatic Rejuvenation of Electrochromic WO3 ... - ACS Publications

May 9, 2017 - Electrochromic (EC) smart windows are able to decrease our energy footprint while enhancing indoor comfort and convenience. However, the...
0 downloads 3 Views 3MB Size
Research Article www.acsami.org

Galvanostatic Rejuvenation of Electrochromic WO3 Thin Films: Ion Trapping and Detrapping Observed by Optical Measurements and by Time-of-Flight Secondary Ion Mass Spectrometry Bill Baloukas,*,† Miguel A. Arvizu,‡ Rui-Tao Wen,‡ Gunnar A. Niklasson,‡ Claes G. Granqvist,‡ Richard Vernhes,† Jolanta E. Klemberg-Sapieha,† and Ludvik Martinu† †

Department of Engineering Physics, Polytechnique Montréal, Montreal, Quebec H3C 3A7, Canada Department of Engineering Sciences, The Ångström Laboratory, Uppsala University, SE-75121 Uppsala, Sweden



S Supporting Information *

ABSTRACT: Electrochromic (EC) smart windows are able to decrease our energy footprint while enhancing indoor comfort and convenience. However, the limited durability of these windows, as well as their cost, result in hampered market introduction. Here, we investigate thin films of the most widely studied EC material, WO3 . Specifically, we combine optical measurements (using spectrophotometry in conjunction with variable-angle spectroscopic ellipsometry) with time-of-flight secondary ion mass spectrometry and atomic force microscopy. Data were taken on films in their as-deposited state, after immersion in a Li-ionconducting electrolyte, after severe degradation by harsh voltammetric cycling and after galvanostatic rejuvenation to regain the original EC performance. Unambiguous evidence was found for the trapping and detrapping of Li ions in the films, along with a thickness increase or decrease during degradation and rejuvenation, respectively. It was discovered that (i) the trapped ions exhibited a depth gradient; (ii) following the rejuvenation procedure, a small fraction of the Li ions remained trapped in the film and gave rise to a weak short-wavelength residual absorption; and (iii) the surface roughness of the film was larger in the degraded state than in its virgin and rejuvenated states. These data provide important insights into the degradation mechanisms of EC devices and into means of achieving improved durability. KEYWORDS: electrochromism, ellipsometry, ion trapping, smart windows, rejuvenation, time-of-flight secondary ion mass spectrometry, ToF-SIMS, WO3

1. INTRODUCTION Devices relying on repeated ion insertion and extraction tend to lose their performance upon long-term use, as is well-known for battery technology1,2 and electrochromics.3−5 This phenomenon limits the devices’ operating lifetime and remains one of the most-critical issues for electrochromic (EC) smart windows. Recently, we discovered that the degradation of EC thin films of WO3 is dependent on the number of insertion−extraction cycles and, very importantly, that this degradation can be eliminated through a simple galvanostatic treatment so that the EC films regain their original properties.6 This rejuvenation process was found to be sustainable, i.e., it could be applied repeatedly.7 Furthermore, we demonstrated that the reason for the degradation was irreversible ion incorporation in the WO3 (ion trapping) and that rejuvenation was connected with voltage- and current-induced removal of the same ions (detrapping).6,8 Ion trapping and detrapping are in line with recent notions regarding various types of intercalation sites: a network of connected sites with low intersite barriers and allowing fast ion diffusion throughout the host material and other sites with high energy barriers, which are able to trap the diffusing ions.9−11 © XXXX American Chemical Society

Earlier work of ours found evidence for ion trapping and detrapping by use of time-of-flight elastic recoil detection analysis (ToF-ERDA).8 The present work provides an in-depth analysis of an analogous thin film system wherein we combine precise optical characterization, based on spectrophotometric absorption measurements (earlier work of ours relied on transmission measurements), with variable-angle spectroscopic ellipsometry (VASE) to obtain the intrinsic optical properties of the material (optical constants) as well as additional information pertaining to film thickness, surface roughness, and the possible presence of gradients in Li concentration. Time-of-flight secondary ion mass spectrometry (ToF-SIMS) and elastic recoil detection analysis (ERDA) measurements were then used to offer unambiguous evidence for the cycledependent trapping of Li+ ions in WO3 thin films. Specifically, our results give clear evidence in favor of the trapped ions exhibiting a depth-dependent distribution. Furthermore, films with large amounts of trapped ions have much-rougher surfaces Received: January 24, 2017 Accepted: May 1, 2017

A

DOI: 10.1021/acsami.7b01260 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Research Article

ACS Applied Materials & Interfaces

of 6°. Spectral absorptance A(λ) was then obtained from A(λ) = 1 − T(λ) − R(λ). Additional optical measurements were performed in the 200 < λ < 1700 nm range by use of a RC2 variable-angle spectroscopic ellipsometer from J.A. Woollam Co. All data analyses were performed with WVASE32 software, also from J.A. Woollam Co., using the Ψ and Δ ellipsometric data obtained at 45°, 55°, 65°, and 75° in conjunction with T(λ) and R(λ) obtained with the UMS. A pair of model windows were implemented for each data set to allow for the thickness of the film to vary in the model (as both measurements may not have been taken exactly at the same spot on the sample). The optical properties of the glass substrate were obtained after the removal of the ITO coating by use of a radiofrequency plasma in a mixture of Ar and O2. The ITO itself was modeled using a Tauc−Lorentz and a Gaussian oscillator for the near-ultraviolet (near-UV) absorption and a Drude oscillator to account for ITO’s conductivity and consequent near-infrared reflection and absorption. As is often the case, the resistivity ρ of the Drude oscillator was found to be depth-dependent in the film; it varied from 1.37 × 10−4 Ω cm at the bottom to 1.93 × 10−4 Ω cm at the surface. With a thickness of 26 nm, the sheet resistance was calculated to be, on average, 64 Ω/□, which is close to the nominal value. The WO3 films were also modeled using a Tauc−Lorentz oscillator for the absorption in the near-UV, as well as by the addition of a top surface roughness layer, whose effective dielectric function was represented by a Bruggeman effective-medium approximation. To better account for the residual absorption of the rejuvenated film, two Gaussian oscillators were also added in the lower part of the visible spectrum (λ < 550 nm), and, due to the fairly low level of absorption in the 500 nm range, more weight was put on R(λ) and T(λ) during the modeling. The obtained results offer insight into the intrinsic optical properties of the coatings (its optical constants n and k) as well as information on the thickness and surface roughness evolution and the possible presence of gradients; the latter feature, as we will show below, might be linked to the presence of compositional gradients due to trapped Li+ ions. Finally, to confirm the obtained thickness values, SEM images of the cross-sections of some of the films were observed using a JEOL JSM-7600F field-emission microscope at an acceleration voltage of 10 kV. ToF-SIMS data were acquired on a ToF-SIMS IV instrument (ION-TOF GmbH, Münster, Germany) at a pressure below 5 × 10−7 Pa. Analyses were performed in bunch mode with a pulse width of 19.1 ns using Bi+ primary ions at an energy of 25 keV. The primary ion current of 1.3 pA was rastered over an area of 50 μm × 50 μm. Depth profiles of positive ions were obtained in the interlaced mode using O2+ ions at 3 keV rastered over areas of 500 μm × 500 μm for sputtering. Charge compensation was obtained using an electron flood gun. A reflectron time-of-flight analyzer and a multichannel plate along with a time-to-digital converter were used to detect secondary ions. Spectral analysis and peak identification were performed using SurfaceLab 6 software. The ion intensities reported in the depth profiles correspond to the Poisson corrected area of the peak identified at the mass of the corresponding ion. For comparison, lithium levels were also assessed by ERDA using a 6 MV tandem accelerator with a 40 MeV beam of Co8+ ions at a scattering angle of 30°. The elemental compositions of the films were calculated using in-house software. To observe changes in surface roughness, AFM measurements were performed using a Dimension ICON instrument

than in their initial and rejuvenated states, as found by ex situ atomic force microscopy (AFM) measurements, and trapped ions also lead to increased film thickness, as confirmed by scanning electron microscopy (SEM). In addition, our ToFSIMS and optical absorption results indicate that a very small amount of permanently trapped ions yields some residual optical absorption at wavelengths below 550 nm. Finally, it is important to note that all the above-mentioned observations are in accordance with the information derived from the ellipsometric modeling.

2. SAMPLE PREPARATION AND CHARACTERIZATION TECHNIQUES Thin films of WO3 were prepared by reactive DC magnetron sputtering in a deposition system based on a Balzers UTT 400 unit. The sputter target was a 5 cm diameter disk of 99.95% pure tungsten. The deposition chamber was first evacuated to a base pressure of ∼6 × 10−5 Pa. A mixture of argon and oxygen, both 99.998% pure, was then introduced through mass-flowcontrolled inlets; the O2-to-Ar gas-flow ratio was set to 0.13, and the total gas pressure was set to ∼4 Pa. Presputtering was conducted for 5 min at a power of 200 W, and depositions under the same conditions were then performed onto unheated 1.1 mm-thick glass substrates precoated with layers of transparent and electrically conducting In2O3:Sn (indium−tin oxide, ITO) with a sheet resistance of 40 Ω/□. Film thicknesses were 300 ± 10 nm, as recorded by surface profilometry using a DektakXT instrument. X-ray diffraction measurements showed the films to be amorphous. Electrochemical measurements by cyclic voltammetry (CV) were carried out in a three-electrode arrangement by using a computer-controlled ECO Chemie Autolab/GPES Interface. The WO3 film was taken as working electrode, and Li foil was used as counter and reference electrodes. It is natural to normalize the potential versus Li/Li+, and the counter electrode allows for the opposite reaction, i.e., Li → Li+, to take place. The electrolyte was dry 1 M LiClO4 dissolved in propylene carbonate (PC). CV data were recorded for up to 20 cycles between 1.5 and 4.0 V versus Li/Li+ at a scan rate of 10 mV s−1. Films were thus only aged by CV cycling and were not submitted to long-time saturation injection. Rejuvenation of degraded WO3 films can be performed by applying a constant current or, alternatively, by applying a constant potential.12 In the present study, we used a current density of 10 μA cm−2 in the direction opposite to that leading to Li+ ion intercalation and coloration; this magnitude of the current was chosen as it is of the order of the peak value observed under potentiostatic detrapping. All electrochemical experiments were performed inside a glovebox with water content below ∼0.5 ppm. Spectral optical transmittance T(λ) was determined in situ in the 400 < λ < 800 nm wavelength interval, encompassing visible light, by the use of an Ocean Optics fiber-optical spectrophotometer. The electrochemical cell was mounted between a tungsten halogen lamp and the detector; the 100% level corresponded to the transmittance recorded prior to immersion of the sample in the electrolyte. Ex situ measurements of optical absorption were performed using a double-beam Cary 7000 universal measurement spectrophotometer (UMS) from Agilent Technologies. This system allows for precise characterization of spectral reflectance R(λ) and transmittance as the detector is moved around the thin-film sample, while the sampling beam remains unchanged. Measurements were performed for p-polarized light at an angle B

DOI: 10.1021/acsami.7b01260 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Research Article

ACS Applied Materials & Interfaces from Bruker, operating with NanoScope 9.1 software, in the peak force quantitative nanoscale mechanical (QNM) imaging mode across a 5 μm × 5 μm surface area. The data were subsequently analyzed using NanoScope Analysis software.

3. EFFECTS OF ELECTROCHEMICAL CYCLING The first step in the experiment consisted in coloring the sample using cyclic voltammetry. Figure 1 shows typical cyclic

Figure 2. Spectral transmittance for a ∼300 nm thick WO3 film immersed in 1 M LiClO4 in PC and studied in its as-deposited state, after the indicated number of voltammetric cycles, and after galvanostatic rejuvenation.

in Figure 2) have been associated with “shallow” and “deep” lithium trap sites, respectively.12 The most-interesting data in Figure 2 are those recorded after galvanostatic treatment of the film subsequent to its degradation upon 20 CV cycles (green curve in Figure 2). Now, the spectral transmittance in the bleached state clearly shows that the properties are rejuvenated and very close to those of the undegraded film. The one minor difference between the pristine and rejuvenated film is in the somewhat lowered optical transmittance at λ < 550 nm for the latter one. This discrepancy is due to a minor fraction of Li+ ions, which is permanently incorporated in the WO3 film, as demonstrated below; the same conclusion was drawn from transmission measurements in our previous study and was related to “irreversible” Li trap sites.12

Figure 1. Cyclic voltammograms at 10 mV s−1 for a ∼300 nm thick WO3 film immersed in 1 M LiClO4 in PC and studied after the indicated number of voltammetric cycles.

voltammograms for a WO3 film; their overall shapes are consistent with literature data.3,6,8 The CV curves changed rapidly as the number of cycles was increased. The charge capacity Q, defined as Q = ∫ (j dV)/s, where j is current density (mA cm−2), V is voltage (in V), and s is voltage scan rate (V s−1), went from ∼60 mC cm−2 during the first CV cycle to ∼3 mC cm−2 after 20 cycles. These results are not surprising because a coloration voltage of 1.5 V at a scan rate of 10 mV s−1 has been shown to lead to severe degradation and ion trapping during lithium intercalation.12 Figure 2 reports in situ spectral transmittance for a WO3 film with CV data such as the ones shown above (Figure 1). Data taken after the first cycle shows that the transmittance is almost the same for the as-deposited and the bleached film; the transmittance lies in the 85−98% interval, and oscillations indicate optical interference. The colored film has a transmittance of ∼17% at λ = 400 nm, which decreased monotonically to ∼3% at λ = 800 nm. Clearly, the initial film displays strong electrochromism due to polaronic absorption with a maximum in the near-infrared (around 1000 nm) and which is the result of the insertion of electrons and Li ions. As the film continues to be cycled at a high negative voltage, the amount of trapped Li ions continuously increases after each cycle so that, upon reaching the 20th cycle, a large amount of Li is trapped in the film and limits the insertion of additional ions. The properties are then radically different: the bleached-state transmittance is now ∼20% at λ = 400 nm and rises to ∼70% at λ = 800 nm. Furthermore, the EC effect is miniscule, which is consistent with the corresponding very small magnitude of Q. The observed increase in colored-state transmission (low polaronic absorption: blue curve in Figure 2) and decrease in bleached-state transmission (residual absorption: orange curve

4. OPTICAL CHARACTERIZATION We now focus on the bleached-state optical properties and discuss WO3 films in their as-deposited state, after immersion in a Li ion conducting electrolyte, after severe degradation by harsh voltammetric cycling, and after galvanostatic rejuvenation to regain the original EC performance. The ex situ optical absorptance of the films was assessed by precisely measuring the reflectance and transmittance at 6° in p-polarized light (Figure 3a). For reference purposes, the absorptance of the base ITO-covered substrate was also recorded (black curve in Figure 3a). One can note that the as-deposited film displays negligible absorption as the resulting absorptance is comparable to that of ITO; in fact, the obtained values are slightly lower in the 425 < λ < 800 nm range. This can be explained by contemplating that the WO3 film displays minima and maxima in R(λ) due to destructive and constructive interference (Figure S1), which will correspondingly increase and decrease the amount of light being transmitted into the absorbing ITO film, respectively, and result in some slight oscillations in the absorptance; in fact, the lowest absorption, at λ ≈ 525 nm, corresponds to a maximum in reflection (Figure S1). The film that had been immersed in the electrolyte is almost on par with the as-deposited film, albeit with an absorptance value ∼0.09% higher at λ = 550 nm. Following immersion in the electrolyte, the film was subjected to 20 harsh CV cycles to purposely C

DOI: 10.1021/acsami.7b01260 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Research Article

ACS Applied Materials & Interfaces

accordance with the in situ transmission data (Figure 2), the absorption is higher in the blue region of the spectrum and monotonically decreases toward the red part. Finally, for the film subjected to galvanostatic rejuvenation, the absorption essentially disappears for wavelengths above ∼550 nm, once again in line with the in situ data. Additional information on the optical characteristics was obtained through the use of spectroscopic ellipsometry (Table 1). To ensure accuracy of the resulting models, the transmission and reflection data from the UMS were also included during modeling (see Figures S1−S4 for the resulting fits). The thickness of the as-deposited film was found to be 313 nm, i.e., near the expected 300 ± 10 nm range. The refractive index of 1.972 at λ = 550 nm corresponds to a packing density of approximately 77% (5.5 g cm−3) using a Lorentz−Lorenz effective-medium approximation13 and supposing a bulk refractive index of 2.50.14 This value is in accordance with our previous work, in which Rutherford backscattering spectrometry showed that films produced under identical deposition conditions had a density of ∼5.5 g cm−3.6 The extinction coefficient k at λ = 550 nm is negligible, and the relatively high surface roughness is not surprising considering that the film is porous. The immersed film exhibits a slight decrease of its refractive index, while all other parameters remain unchanged. This slight decrease in n could be the result of some Li+ insertion because the refractive index is known to decrease upon ion intercalation.15 The film’s characteristics change significantly once it is CV cycled as the refractive index further decreases. However, in this case, the thickness is also shown to increase by 14%, the surface roughness more than triples, and the extinction coefficient is now 0.027. For the first approximation, simply using the Beer− Lambert law, one can estimate the absorption of the cycled sample to be around 20%. The UMS-measured value was 21.3% and is indeed very close because it includes the 1.5% absorption of the ITO. This consistency clearly demonstrates that the ellipsometric model is representative of the actual sample’s properties. In the present case, and contrary to results for the other samples, the model can be refined, and the corresponding mean square error between the experimental and modeled data further decreased by ∼12% by including a gradient in the refractive index versus thickness. The obtained value of −4.8% indicates that the refractive index decreases toward the surface of the sample and can be reconciled with the presence of a gradient in the concentration of the trapped lithium ions deep below the surface, as will be discussed below. It is important to add that, although the addition of a gradient did help, there are still some discrepancies between the measured ellipsometric data (Ψ and Δ) and the model, most noticeably for

Figure 3. Absorptance (A) spectra obtained by measuring transmittance (T) and reflectance (R) at 6° in p-polarized light (A = 1 − R − T) for ∼300 nm thick WO3 films subjected to the indicated treatments. Data are shown in panel a for films in different states: asdeposited, immersed in 1 M LiClO4 in PC, cycled 20 times as indicated in Figure 1, and following galvanostatic rejuvenation. The black curve corresponds to the absorptance of the ITO-covered glass substrate. The surface roughness of the samples is much smaller than the wavelength of visible light, and hence, surface scattering is taken to be negligible in comparison to absorption. Panel b shows extinction coefficients obtained through the combined use of spectroscopic ellipsometry and the data in panel a.

degrade it and incorporate a large amount of trapped Li+ ions; this incorporation clearly takes place as the absorption is much higher (56.3% at λ = 400 nm and 21.3% at λ = 550 nm). In

Table 1. Main Results of the Ellipsometric Study of the Base ITO electrode and of the WO3 Films Subjected to the Shown Proceduresa sample ITO as-deposited immersed cycled rejuvenated

thickness (nm) 26 313 315 360 331

n (550 nm)

k (550 nm)

b

± ± ± ±

2 5 13 5

b

0.008