Alkali treatments of Cu(In,Ga)Se2 thin‐film absorbers and their impact on transport barriers

We study the impact of different alkali post‐deposition treatments by thermal admittance spectroscopy and temperature‐dependent current‐voltage (IVT) characteristics of high‐efficiency Cu(In,Ga)Se2 thin‐film solar cells fabricated from low‐temperature and high‐temperature co‐evaporated absorbers. Capacitance steps observed by admittance spectroscopy for all samples agree with the widely observed N1 signature and show a clear correlation to a transport barrier evident from IVT characteristics measured in the dark, indicating that defects are likely not responsible for these capacitance steps. Activation energies extracted from capacitance spectra and IVT characteristics vary considerably between different samples but show no concise correlation to the alkali species used in the post‐deposition treatments. Numerical device simulations show that the transport barrier in our devices might be related to conduction band offsets in the absorber/buffer/window stack.

behave very differently from individual layers due to modifications occurring during further processing steps, electrical interactions, or inter-diffusion between adjacent layers. 5 Furthermore, applied bias and illumination during realistic device operation needs to be accounted for.
Despite the importance of electrical device characterization, the standard interpretation particularly of the voltage-dependent, frequency-dependent, and temperature-dependent capacitance has frequently been challenged for thin-film solar cells. One particular capacitance step, termed "N1", 6 is commonly observed for all CIGS thin-film solar cells, and is also dominating the frequency-response of our devices (see Section 3). The origin of the N1 signature has been discussed controversially for already 2 decades. Initially, this signature was attributed to continuously distributed defects at the CIGS/buffer interface, as the activation energy was found to be sensitive to oxidation or air annealing. 6,7 By contrast, the N1 signature was later also identified in drive-level capacitance profiling of the CIGS bulk and was thus reevaluated as an acceptor-like bulk defect. 8  is a combination of interface states and interface-near deep defects in the bulk. Also using DLTS, Zabierowski et al 10 found that the N1 signal consists of 4 distinct components and identified In Cu -related metastable defects as potential origin of the N1 signature. Independent of the proposed nature or location of these defects, a defect response appears to be the most widely used interpretation of the N1 signal, and in fact any capacitance step above a potential freezeout in admittance spectroscopy. There have been an increasing number of publications which provide alternative explanations for the N1 level, most linked to the transport characteristics of the solar cell.
Reislöhner et al argue that the N1 signature could be explained by a mobility freeze-out due to hopping conduction, 11 or by percolative charge transport due to spatial inhomogeneities in the absorber. 12 Lauwaert et al 13 showed that DLTS signals recorded for different pulse directions are not compatible with standard models of carrier capture and emission by defect states in the device. Instead, R-C-like non-ideal contacts or interlayers in series with the main junction were shown to result in capacitance steps and DLTS peaks which could explain the N1 signature. [13][14][15][16] With the main junction located at the front CIGS/buffer interface in the device, the R-C-like barrier or counter-diode in series with the main junction has been attributed to the back contact in these studies. Experimentally, this is supported by the observation of a phototransistor effect in the bulk of CIGS solar cells, [17][18][19] which indeed suggests the presence of a Schottky contact at the back of the device. Furthermore, Eisenbarth et al 20 reported that changes in the space charge region width around the N1 capacitance step do not directly correlate with the thickness of the buffer layer and might thus be more likely caused by the back contact. By contrast, Igalson et al 9 reported on a correlation between the N1 signature and blocking of the diode current in forward bias, which they attributed to Fermi level pinning at the CIGS/buffer interface at the front of the device. Recently, we demonstrated that excessive RbF treatment leads to modifications of the CIGS/buffer interface, which affect the N1 signature observed in admittance spectroscopy. 21 We further demonstrated that bias-dependent and illuminationdependent impedance spectra of CIGS solar cells suggest a depleted buffer layer but are difficult to reconcile with deep defects as origin of the N1 signature. 22 However, both different back electrodes 23 and different buffer layer stacks 24 at the front of the device were reported to modify the admittance spectrum, and the main capacitance step of all devices was found to agree with the N1 signature. Differences in experimental observations and interpretation were also proposed to be a consequence of a multitude of signals potentially coexisting in CIGS devices. 25 Note, that signatures comparable to N1 were also observed in photo-induced current transient spectroscopy (PICTS) measurements 26,27 on CIGS absorbers, which probe the bulk of the absorber and where no n-type front layers are deposited on the CIGS.
The preceding discussion illustrates that different potential explanations for the N1 signature in CIGS solar cells exist, notably bulk or interface defects, or transport barriers at the front or rear of the absorber layer. While these previous studies show that transport barriers can indeed be a potential explanation for the N1 signature, this is challenging to verify experimentally using only capacitancebased techniques, because defects and transport barriers lead to comparable frequency-responses in ac techniques. On the other hand, the dc current transport characteristics of a solar cell device in strong forward bias will be sensitive to transport barriers at different locations of the device. 28 Nevertheless, the complicated device structure of the CIGS hetero-junction solar cells suggests that simple analytical transport models might be insufficient to adequately explain experimental data, and that numerical device simulations could provide more detailed insight concerning the location of dominant transport barriers within the device. In this paper, we link experimental ac and dc transport characteristics of many solar cells to explore the importance of transport barriers in the device. We compare our data to numerical device simulations of the temperature-dependent current-voltage characteristics in order to identify the dominant interface barrier in the device. We use photoluminescence (PL) to verify that deep defects indeed do not play a significant role in our devices.
Due to the wide spread of results in literature even for untreated solar cells, studies of the impact of alkali PDTs on CIGS solar cells will always suffer from ambiguities in the characterization methods if only individual or few measurements are taken into account. We present a systematic study of the impact of different alkali PDTs on the electrical device properties of solar cells fabricated from state-of-the-art high-efficiency absorbers. Because likely several possible origins of the N1 signature coexist, this choice of samples ensures that our measurements are representative of current record-efficiency devices. We also develop further insight into the interpretation of capacitance steps observed in TAS based on the large set of data on highlyefficient CIGS solar cells presented here. In Section 3, we discuss temperature-dependent capacitance spectra and show that the frequency response of our devices is dominated by 1 or 2 capacitance steps, which would be identified as N1 signatures according to common practice. Photoluminescence experiments in Section 4 demonstrate that the high-efficiency absorbers show no significant defect luminescence at defect energies compatible with the N1 signature observed in admittance spectroscopy. This indicates that defect concentrations in the relevant energy range are low and should not contribute significantly to the capacitance spectrum. Instead, we link the main capacitance step to a transport barrier evident from temperature-dependent current-voltage characteristics in Section 5 and show in Section 6 that such transport barriers are likely related to the conduction band offsets at the front of the device.

| SAMPLE PREPARATION AND CHARACTERIZATION
We compare 2 different sets of Cu(Ga,In)Se 2 (CIGS) thin-film solar cells fabricated at the Center for Solar Energy and Hydrogen Research (ZSW) and the Swiss Federal Laboratories for Materials Science and Technology (Empa), respectively. Details of the processing conditions are similar to those in Chirilă et al, 2 Jackson et al, 3 and Chirilă et al. 29 All absorbers are grown by co-evaporation of Cu, In, Ga, and Se in a multistage vacuum process. At ZSW, sputtered Mo on SLG is used as substrate, and the co-evaporation process is performed at standard elevated temperatures. The Empa absorbers are designed for the deposition onto flexible polyimide foils, which requires substrate temperatures below 450°C at all times during the deposition process. However, to simplify handling, the Empa absorbers are deposited onto 1-mm-thick SLG substrates in this study. A silicon oxide (SiO x ) layer is introduced between SLG and Mo back contact to suppress Na diffusion from the SLG, which would also not be present if polyimide foil was used instead.
Different alkali post-deposition treatments (PDT) are performed by in-situ co-evaporation of alkali-fluorides in Se atmosphere after the CIGS deposition. The different PDTs employ KF, RbF, and CsF for ZSW absorbers, and NaF, NaF + KF, and NaF + RbF for Empa absorbers. Additional solar cells without PDT are studied as well. Note that all ZSW samples contain Na, and in lower concentrations K, due to diffusion from the Na-containing and K-containing SLG substrate.
In the case of Empa absorbers, a NaF-PDT is always applied before any KF-or RbF-PDT in all samples, because the SiO x barrier layer and reduced substrate temperature suppress alkali diffusion from the SLG substrate into the CIGS absorber. Any further reference to KFor RbF-PDTs on Empa samples thus implies a previous NaF-PDT. the range of 13% to 14% for the alkali-free low-temperature absorbers, and 16% to 20% for absorbers containing alkali elements (up to 21% with ARC). Table 1 shows a summary of the designated area cell efficiencies η for the different PDTs and absorbers used in this study. is set once prior to the measurement to yield the correct previously measured short-circuit current density at a temperature of 300 K.
The temperature-dependent dark and light current-voltage characteristics (IVT) are recorded while cooling down the sample in steps of 10 K (set temperature) in a range between 320 and 20 K, allowing sufficient time for temperature stabilization before each measurement. Note that the measured sample temperature might deviate significantly from the setpoint due to the low thermal conductivity of the glass substrate and typically only reaches 40 to 50 K at the lowest temperature setting of T set = 20 K. After IVT measurements, the sample is heated to 300 K and kept in the dark for at least 12 hours to ensure a sufficient relaxation of any photo-induced instabilities. This is verified by ensuring a constant capacitance reading in short-circuit conditions. The admittance spectrum is recorded in the same temperature range while cooling down in a frequency range of f = 100 Hz to 1 MHz at a dc bias voltage of 0 V and an ac amplitude of 30 mV rms. A parallel equivalent circuit model is used to separate the conductance G and capacitance C. Note that some measurements show an extreme capacitance dispersion at high frequencies above a few 100 kHz, which does not appear to follow any consistent temperature dependence. As this feature also may differ between measurements on the same sample, the most likely origin is related to the external contacts to the sample.
Thus, features at the highest frequencies are discarded during analysis.

| CAPACITANCE STEPS IN ADMITTANCE SPECTROSCOPY
All samples with both Empa and ZSW absorbers qualitatively show the same features in their admittance spectra. A representative example of such a spectrum is shown in Figure 1 for a RbF-treated ZSW absorber. Some of the typical features of the capacitance spectra are more pronounced for certain samples, and we present the capacitance spectrum of a KF-treated Empa absorber in Figure 2 for comparison (note the different scale on the y-axis). From high to low measurement temperatures in the admittance spectra, we observe the following features, which are marked by arrows in Figures 1 and 2 30 This feature appears to be enhanced for some of the samples only containing Na, but at this point in time provides no further insight and hence will not be covered in the present paper.

A well-defined capacitance step at intermediate temperatures,
typically in the range of 100 to 250 K (corresponding to a capacitance drop from 38 to 26 nF/cm 2 for the example shown in Figure 1). We denominate this signature the "main" capacitance step, and it will be discussed in more detail below.
3. There appears to be a second capacitance step just below the main step, which is visible as a shoulder at a frequency range of 10 to  Figure 2, this step is unambiguously related to a conductivity freeze-out of the absorber, as the capacitance drops to the geometrical capacitance C geo = ε r ε 0 /d ≈ 5 nF/cm 2 , which indicates that the absorbers become insulating at these temperatures and frequencies. Although the attribution of this capacitance step to a freeze-out is only possible for some samples, we assume that the nature of this step is most likely the same in all samples due to the similar temperature range.
The activation energy of a process responsible for a capacitance step can be obtained from the temperature dependence of the inflection frequency f t , or, more commonly, the angular inflection frequency ω t = 2πf t of the respective capacitance step. We determine the inflection frequency experimentally by finding the maxima in the derivative of the capacitance as a function of the logarithm of frequency, dC/dln( f ). The simplest form of a thermally activated inflection frequency is given by the equation where X 0 is a constant. Accordingly, the activation energy is then obtained from a linear fit of ln(ω t ) vs inverse temperature 1/T. One complication in the analysis arises if the prefactor X 0 is in fact temperature dependent. For a capacitance step due to a defect response, it has been shown that X 0 has a weak ( ∝ T 2 ) temperature dependence due to the temperature-dependent thermal velocity and effective density of states. The inflection frequency is thus given by the equation 4 where v th is the thermal velocity, N C,V is the effective density of states in the conduction or valence band, σ n,p is the electron or hole capture cross section of the defect, and the prefactor ξ 0 is now assumed to be independent of temperature. The activation energy E a is then obtained from an Arrhenius plot of ln(ω t /T 2 ) vs inverse temperature 1/T.
with the characteristic energy E char and a prefactor ξ 00 . 31 It is apparent from Figure 3 that the freeze-out step, as expected, is separate from the other steps and has no relation to the N1 and N2 signatures. Furthermore, the freeze-out, if resolvable from the admittance spectra, always appears at an energy of 55 ± 10 meV. Such an activation energy would be consistent both with a shallow acceptor level typically observed at 40 to 60 meV, 35 and with a mobility freeze-out due to transport barriers of approximately 60 mV at the grain boundaries. 5,35 It is worth noting that we can identify the inflection frequencies of the freeze-out steps only for some of the low-temperature absorbers, not for any of the high-temperature absorbers. The details of the freeze-out mechanism depend on the bulk doping and transport properties across grain boundaries and are thus likely to vary between different absorber processing conditions.
The inflection frequencies of the main capacitance step observed in this study appear to follow the trend of the N1 signature reasonably well, although the scatter in our data is substantial. Such pronounced scatter around the reference line plotted in Figure 3 Figure 4 shows the PL spectrum at a temperature of T = 10 K for an untreated lowtemperature absorber. We observe a single emission peak in the energy range of 0.92 to 1.24 eV. This peak is fairly broad, as expected for a Cu-poor absorber, 41 and the "true" peak shape is obscured by interference effects ("dips" in the PL peak). 42 The PL spectrum clearly demonstrates that no significant PL emission beyond background noise is recorded for energies below the main PL peak, which corresponds to photon energies E < 0.92 eV (left dashed line in Figure 4). From the absence of PL emission, we can conclude that the defect densities in this energy range must be fairly small.
In   Despite qualitatively similar ac admittance spectra for all samples, see Section 3, we observe 2 different types of IVT behavior: in some samples, the forward current contribution of the main junction, ie, an exponential current increase in forward bias, is largely suppressed at low temperatures [termed "type A", compare Figure 5A], while other samples show a diode-like forward current over the full temperature range ["type B", compare Figure 5B]. Accordingly, the low-temperature IV curves for "type A" samples are dominated by the shunt current.
Below a certain temperature, the dark IV curves of these devices no longer exhibit a realistic diode-like behavior: the diode current either disappears completely compared with the shunt current, as exemplified in Figure 5A, or the diode ideality factor increases to unrealistic values well above 10. This transition occurs in a temperature range of 100 to 150 K for our samples, which agrees remarkably well with the lowest temperature where the main capacitance step in the admittance spectra is observed.
We do not find any obvious correlation between IVT behavior (A or B) and absorber growth temperature, buffer layer type, or alkali species. Instead, as discussed in more detail at the end of this section, we find that samples with large activation energies in the capacitance spectrum, independent of other sample properties, predominantly show type A behavior with strongly suppressed diode current. Note that the RbF-treated high-temperature absorber representing type A behavior in Figure 5A is the same sample shown in Figure 1 for the admittance spectra.
Both graphs in Figure 5  For a more quantitative analysis, we look at the dark IV curves in high forward bias, where shunt currents are negligible at room temperature. We find that the forward current cannot be adequately described over the full voltage range by a simple analytical model, assuming just a junction diode and ohmic series and shunt resistances.
More elaborate device models accounting for transport barriers can certainly be devised and applied to these measurements, but they would necessarily be more complex and contain many free parameters and would thus lead to unreliable fitting results. Accordingly, we do not attempt to extract the fundamental device parameters by fitting the data for all temperatures. We rather discuss the temperature evolution of the shape of the IVT curves and compare our observations with numerical device simulations in Section 6. We estimate the blocking of the diode current from the temperature-dependent dark current density at a fixed reference voltage of V ref = 0.9 V. This voltage was chosen high enough to ensure a dominant diode current at high temperatures, but low enough to limit the impact of current saturation at high bias. Our choice of reference voltage is ultimately arbitrary, but the impact on the extracted activation energies appears to be fairly small. For example, even a choice of 1.2 V as reference voltage yields comparable activation energies within a range of ±15%. The insets in Figure 5 show the natural logarithm of the corresponding dark current density, ln (J 0.9V ), as a function of inverse temperature 1000/T. The current is evidently thermally activated, and the activation energy (circles in Figure 6) is obtained from a linear fit to the data shown in the insets. Data at low temperatures where the forward current starts to level off with temperature, presumably due to the impact of shunt resistance, is excluded from the fit.
An alternative approach to describe the current blocking is by means of the effective series resistance, defined as the inverse slope of the IV curve, R s = dV/dJ, at high forward bias (squares in Figure 6). Here, we again chose a fixed reference voltage, at 1.2 V, in a voltage range where resistive effects clearly limit the diode current. This approach provides a figurative model for the drop in diode current and would be straightforward to implement in a numerical diode model. It is worth pointing out, however, that this effective series resistance is just a model parameter to describe the behavior of the diode current and is different from the "true" ohmic series resistance of the device. This differentiation is obvious from the admittance spectra, where the ohmic series resistance is given by the high-frequency limit of the real impedance. The high-frequency impedance only shows a weak temperature dependence and is below 5 Ωcm 2 for all measurements, much lower than effective series resistance values up to 5 kΩcm 2 for some devices at low temperatures.
This clearly indicates that the effective series resistance, as defined above, must have a capacitive component and thus might be related to a barrier or space charge region connected in series with the main hetero-junction. Figure 6 shows the correlation between the activation energies of the main capacitance step (x-axis) obtained by ac admittance spectroscopy, of the effective dc series resistance at a forward voltage of 1.2 V (y-axis, squares), and of the dark dc current density at a reference voltage of 0.9 V (y-axis, circles) for samples with different PDTs on both low-temperature and high-temperature absorbers. Colors in Figure 6 represent the alkali species. Most samples exhibit a reasonably good correlation between all 3 activation energies, indicated by the blue line in Figure 6 for a 1:1 correlation. This relation supports our hypothesis that the main admittance step in these samples is indeed related to the transport properties of the solar cell, rather than deep defects in the absorber. Such a suppression of the diode current and a drastic increase in effective series resistance at low temperatures might, for example, be caused by unfavorable band offsets or non-ohmic contacts, which could impede current flow in parts of the device.
The impact of the exact nature and location of a transport barrier on the magnitude and voltage dependence of the forward current will be addressed by numerical device simulations in Section 6.
We find that the magnitude of the activation energy determined from admittance or IVT fully determines the type of IVT behavior, as indicated by the red dotted line in Figure 6: devices with activation Despite the good agreement between all 3 activation energies for many samples, we observe deviations for 2 types of samples: • For 2 samples marked with a blue asterisk in Figure 6, the activation energies of series resistance R s and diode current J d agree remarkably well but are significantly lower than the activation energy of the main capacitance step. One of these samples is likely damaged (zero-bias shunt resistance below 130 Ωcm 2 ) and thus will be disregarded. For the other sample, the correlation is nearly perfect if the second capacitance step is taken as reference instead of the main capacitance step (activation energies approximately: main and second capacitance steps-200 and 80 meV, series resistance-90 meV, diode current-100 meV, respectively).
• For most samples with type B IVT behavior, ie, for capacitance steps with activation energies below 150 meV, the diode current follows the activation energy of the capacitance step, while the activation energy of the series resistance (squares in Figure 6) is significantly lower. Note that this behavior is not universal, and the activation energy of the effective series resistance is only reduced for low-temperature absorbers and a CsF-treated hightemperature absorber, not for any of the other high-temperature absorbers. These deviations suggest that transport properties differ between devices. For example, in low-temperature devices with moderate transport barriers in the buffer/window stack, ie, with low activation energy of the capacitance step, the series The deviations discussed earlier suggest that different types of barriers might be present in the devices based on processing conditions and alkali species. We also find that the absolute value of the activation energy differs significantly between samples from different fabrication runs, even with nominally the same alkali PDT. For example, using admittance spectroscopy, we obtain activation energies of 80 and 160 meV, respectively, for 2 RbF-treated high-temperature absorbers. These values again indicate that the transport barrier in these devices is fairly sensitive to the processing conditions during device fabrication. Accordingly, our study does not reveal any reliable, systematic dependence on alkali PDT, because variations between samples overshadow any differences due to different alkali PDTs.
Although we cannot discern trends between different alkali species, the activation energies found in this study suggest that the transport barrier is largest without any alkali species present and appears to be reduced to varying extent by any alkali PDT.

| DEVICE SIMULATIONS
In an attempt to localize the observed transport barrier within the device, we perform numerical device simulations of the currentvoltage characteristics of a typical CIGS thin-film solar cell using Synopsys Sentaurus-Tcad. In a first step, we vary the conduction band offsets at the absorber/buffer (ΔE A/B , "A/B") and buffer/window (ΔE B/W , "B/W") hetero-interfaces in order to discriminate between the effects of the respective interface on the IV characteristics. In these simulations, current transport is modeled with drift/diffusion equations, and the back contact is assumed to be ohmic. A constant distributed series resistance of 0.5 Ωcm 2 is used at the back contact.
In a second step, we employ different transport models at the front and rear of the cell: thermionic emission for electrons across the barrier at the window/buffer interface, and a Schottky contact for holes at the rear contact, respectively. Further details of these simulations are presented elsewhere. 45 The conduction band edge in the window layer is initially assumed to be continuous, which is representative of a standard ZnO window stack (ZnO:Al + i-ZnO). We then introduce a conduction band discontinuity also within the window stack in order to describe devices with a (Zn,Mg)O layer instead of the standard i-ZnO. Note that our simulation are only exemplary and aim to qualitatively assess trends in the effect of different interfaces on the current transport through the device. Due to the large spread of experimental activation energies, indicating that exact interface properties are likely sensitive to processing conditions, and due to the importance of choosing a suitable transport model, see discussion below, we do not attempt to pinpoint the "correct" quantitative conduction band offsets within a specific real device. Figure 7A shows the simulated dark IV characteristics for a "spike"   Figure 7B]. This feature causes the diode current to saturate in forward bias, which results in a significant suppression of the forward current even for small conduction band offsets at the front-side of the device. Note that experimentally we do not observe a constant saturation current density in forward bias, and a Schottky contact at the back thus cannot be the sole explanation for the observed transport barrier.
In the preceding simulations, we have assumed a window layer  The preceding discussion clearly demonstrates that various types of transport barriers can limit the forward current, and their impact on the current-voltage characteristics depends critically on the exact conduction band offsets and transport model employed in the simulation. Accordingly, small variations in the exact band alignment or material properties, for example caused by differences in composition or intermixing at the interfaces, 5   In both cases, the transport barrier is located within the buffer/ window stack and thus formed after CIGS growth and alkali PDT. On one hand, any variation or treatment affecting the absorber bulk or surface might still somewhat influence the subsequent growth of the buffer/window stack, and thus modify its actual transport properties.
The specific IVT characteristics of the device might thus also depend on absorber properties, even if the limiting interface is entirely located within the buffer/window stack. On the other hand, modifications of the absorber surface due to alkali PDT will only indirectly affect the dominant transport barrier, and small variations in the buffer/window stack itself, unrelated to the alkali PDT, could obscure any influence of the alkali elements. This prediction is in agreement with our experimental results in Section 5, where we did not observe any obvious relation between alkali treatments and experimental activation energies obtained from admittance and IVT measurements, corroborating that the transport characteristics of the device are mainly sensitive to processing parameters unrelated to the choice of alkali PDT.

| CONCLUSIONS
The activation energies of capacitance steps observed in admittance spectroscopy were compared for a wide range of high-efficiency absorbers subjected to different alkali PDTs. The activation energy of a freeze-out at temperatures below 100 K was found to be 55 ± 10 meV, which is consistent with typical values of acceptor levels (carrier freeze-out) or grain-boundary barrier heights (mobility freezeout). The activation energies and thermal prefactors of capacitance steps occurring at higher temperatures were found to agree reasonably well with the N1 signature established in literature. Remarkably, this agreement was observed even for 2 separate capacitance steps resolved in the same admittance spectra, and thus cannot be attributed to a common defect signature. Photoluminescence spectra recorded at 10 K do not show any significant PL signal at transition energies corresponding to deep defects, and thus confirm that deep defects play a minor role in these devices. Because deep defects are negligible also in untreated absorbers, alkali PDTs do not appear to alter the deep defect concentration significantly.
Temperature-dependent current-voltage characteristics were recorded to study transport barriers in the devices, which revealed a suppression of the forward current at low temperatures in all devices.
In many cases, this suppression is drastic, and the IV characteristics no longer show an exponential diode-like behavior at low temperatures.
The activation energy of the temperature-dependent dark current density at fixed forward bias was shown to correlate with the activation energy of the main capacitance step in admittance spectroscopy. Combining all evidence, we find that the main capacitance step observed in admittance spectroscopy for all samples cannot be attributed to a defect but appears to be caused by an electron injection barrier at the front of the device. Due to the reasonable agreement of this Our results highlight that the electronic effects of alkali treatments of CIGS thin-film absorbers-at the moment-are not readily discernable by standard electrical measurements. We find that such measurements are heavily influenced by the buffer/window stack, which modifies or obscures any contribution from the absorber bulk or surface. Accordingly, prevailing models of un-treated and alkalitreated CIGS absorbers need to be carefully reconsidered, taking into account conduction band offsets at the front of the device.