Investigation of capacitance transients in CuInS2 due to ionic migration

ARTICLE IN PRESS Solar Energy Materials & Solar Cells 92 (2008) 1579–1585

Contents lists available at ScienceDirect

Solar Energy Materials & Solar Cells journal homepage: www.elsevier.com/locate/solmat

Investigation of capacitance transients in CuInS2 due to ionic migration Ruben Loef a,,1, Joop Schoonman b, Albert Goossens a,1 a b

Opto-Electronic Materials, Delft University of Technology, Julianalaan 136, 2628 BL Delft, The Netherlands Materials for Energy Conversion and Storage, Delft University of Technology, Julianalaan 136, 2628 BL Delft, The Netherlands

a r t i c l e in fo

abstract

Article history: Received 6 March 2008 Received in revised form 24 June 2008 Accepted 13 July 2008 Available online 19 August 2008

Transport of mobile ions in n-TiO2/n-CuInS2/p-CuInS2 thin-film devices is studied with the transient ion-drift (TID) method. In contrast to the normal TID method, a mobile ion profile is created in the CuInS2 layer, which can be described by the Gouy–Chapman theory. Activation energies for diffusion of 0.5 and 1.0 eV are found. We postulate that these activation energies are related to the associated defect, x (Cu00In Indd Cu ) , which introduces a deep electronic state inside the bandgap of CuInS2. This defect can accept or release an electron and drift out of the depletion region. This will lower the concentration of recombination centers in the depletion region, which explains the self-healing property of CuInS2. & 2008 Elsevier B.V. All rights reserved.

Keywords: CuInS2 TID DLTS Self-healing

1. Introduction Chalcopyrite semiconductors are promising candidates to be used in thin-film solar cells [1–4]. Although impressive efficiencies up to 19% are reached for Cu(In,Ga)Se2 [5] and 12.5% for CuInS2 [3], further improvement of the device performance can be accomplished. The mobility of copper (Cu) ions in chalcopyrites is a well-known issue in these solar cells [6–11], but also indium (In) mobility has been reported [12]. Although ion-mobility can give chalcopyrite-based solar cells self-healing properties, and probably plays a role in the light soaking effect [13], it is also shown that under certain conditions Cu can physically segregate from the chalcopyrite layer [7]. Under specific circumstances Cu-transport can be used to change the semiconductor type. In this way p–n homojunctions or nano-scaled transistors are created in chalcopyrites [14–20]. Furthermore, it has been shown that Cu(In,Ga)Se2based solar cells also recover after a decrease in efficiency caused by electron or proton irradiation [21–23]. The nature of the instability of p-type CIGS, on long and short range, is nicely reviewed and discussed by Guillemoles et al. [24]. Long-range effects are assigned to Cu mobility. Short-range effects are ascribed to donor defects that capture electrons or are transformed to acceptors via defect chemical reactions. In the latter case, both isolated single defects and defect complexes can be involved.  Corresponding author. Fax: +31182787421.

E-mail address: [email protected] (R. Loef). Also at Delft Institute for Sustainable Energy, Delft University of Technology, Julianalaan 136, 2628 BL Delft, The Netherlands. 1

0927-0248/$ - see front matter & 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.solmat.2008.07.005

Here, we present a detailed transient ion-drift (TID) study of TiO2/CuInS2 heterojunctions. With this method the mobile ion transport in thin-film TiO2/CuInS2 devices can be determined, leading to better understanding of the chemical stability. In previous investigations we have shown that a p–n homojunction is present in CuInS2 in TiO2/CuInS2 heterojunctions [25]. Fig. 1 shows the band diagrams of these devices at three potentials. A brief description of the TID method is given below. In the past, this method has been applied on single-crystalline CuInSe2, but never on thin films. Diffusion coefficients of 1012–1013 cm2/s have been found, which were assigned to Cu-migration [26]. Much larger diffusion coefficients in single crystalline and polycrystalline CuInSe2 and CuInS2 are reported using electron beam-induced current (EBIC) and point electrode techniques, i.e., 105–1010 and 106–109 cm2/s, respectively [6,7,10,11]. The origin of the Cu-transport in chalcopyrites is not well understood but is believed to be related to mobile copper interstitials (Cudi ) or copper vacancies (V00Cu ). In the present TID measurements smaller diffusion coefficients are found, and are probably related to the migration of In instead of Cu. Moreover, we postulate a relation between this In-related defect and the earlier reported, but yet unassigned, defect around 1 eV above the valence band [27–29].

2. TID theory The first application of the TID method has been reported by Heiser and Mesli in 1993 [30,31]. Later, Lyubomirsky et al. [26] applied this method on single-crystalline CuInSe2. The TID method is similar to deep-level transient spectroscopy (DLTS)

ARTICLE IN PRESS 1580

R. Loef et al. / Solar Energy Materials & Solar Cells 92 (2008) 1579–1585

Fig. 1. Band diagrams for p-CuInS2/n-CuInS2/TiO2 heterojunctions at the three potential regions as concluded in our previous study [25]. (I) The band diagram at reverse potentials, when n- and p-type CuInS2 are both in full depletion. (II) The band diagram around 0 V when only p-CuInS2 is in full depletion. (III) The situation close to the flat band potential when neither the n-type nor p-type CuInS2 is in full depletion.

[32], although in TID the observed transients are caused by ionic transport instead of trapping of electrons. Mobile ions drift under the influence of the electric field in the depletion layer. After applying a potential step in reverse direction, the mobile ions move in the same direction as the electronic charges, i.e., mobile acceptors are negatively charged and follow the electrons, and mobile donors are positively charged and follow the holes. After some time all mobile ions are removed from the depletion layer and accumulate at the depletion layer edge. Subsequently, the concentration profile of mobile ions will flatten due to thermal diffusion to the bulk. The following assumptions are made: (i) drift of mobile ions is much faster than thermal diffusion, (ii) the mobile ion concentration is much lower than the doping level (i.e., non-mobile donors and acceptors), which implies that the space-charge region is not affected noticeably by the redistribution of the mobile ions, and (iii) the background doping is uniform. Under these conditions a capacitance transient,

DCðtÞ ¼ ½Cð0Þ  Cð1Þ½1  et=t 

(1)

can be expected. In general, the transient time t is longer for TID than those obtained with DLTS, because ionic diffusion is usually a slower process than trapping and detrapping of electrons. Similar to DLTS, there are basically two transient times. The equivalent of the trap filling time is the ion ‘flattening time’

tf 

w2 D

(2)

and the equivalent of the carrier emission time, after the reverse voltage step, the ion ‘accumulation time’

ta ¼

kT  w2 kT  q2 DN qDfr

(3)

Here, w is the depletion width, q is the elementary charge, D is the ion-diffusion coefficient, N is the non-mobile donor or acceptor concentration, k is Boltzmann constant, e is the dielectric constant, and fr is the reverse potential. Although TID and DLTS are almost similar, there is an important difference. For DLTS the filling time of traps is much shorter than the carrier emission time, while the opposite is true for their TID equivalents, i.e., tf/ta41. The ion-diffusion coefficient from Eq. (3) can be written as D ¼ D0 eEA =kT

(4)

in which D0 is the pre-exponential factor and EA the activation energy. The activation energy can be calculated from the slope of a ln(ta/T) vs. 1/T plot.

3. Experimental aspects Flat films of TiO2 and CuInS2 are obtained from Advanced Surface Technologies in Bleiswijk. The samples are made with spray deposition on transparent conductive oxide (TCO) glass (SnO2:F) as described elsewhere [33–35]. Carbon spots (diameter 2.3 mm) are used as back contact (Graphite conductive adhesive, aqueous-based, Alfa Aesar). Impedance and CV measurements are described in our previous work [25]. Both are recorded with a potentiostat (PAR283) in combination with a frequency response analyzer (Schlumberger FRA 1255). A dc voltage is aplied over the samples with the potentiostat. An ac voltage of 10 mV with a frequency of 1 MHz is superimposed on this voltage by the FRA. The CV measurements are recorded at a scan rate of 50 mV/s. TID measurements are obtained with the same set-up. First, the samples are brought to a reverse potential of 1 V for 100 s. Subsequently, the potential is stepped to the flattening potential of 0 V for 300 s and finally back to the accumulation potential of 1 V. The capacitance response is monitored on a computer using Labview with a rate of 2 points/s. The TiO2/CuInS2 samples are mounted in a cryostat (Oxford Instruments Optistat DN). TID and impedance spectra are recorded in the temperature range of 300–500 K under nitrogen atmosphere to prevent oxidation of CuInS2. All measurements are duplicated on another sample from another deposition with the same parameters. Similar results are found for both samples. Raman measurements are performed using a home-built setup. A Nd:YVO4 laser, operating at a wavelength of 532 nm, is used as the excitation source (SpectraPhysics Millennia). Neutral density filters are used to adjust the power of the laser. Light detection occurs with a liquid-nitrogen-cooled CCD camera (Princeton Instruments LN/CCD-1100-PB) connected to an Acton Spectra Pro 2500i monochromator.

4. Results and discussion 4.1. Impedance and Mott–Schottky measurements In a previous paper, we have discussed the impedance and Mott–Schottky behavior of the samples in detail [25]. The main results and conclusions of that work are necessary to clarify the present TID results and are summarized here. The impedance spectra in Fig. 2 show that, at frequencies above 1 MHz, a resistor and capacitor in series describe the equivalent circuit for the samples at both flattening (0 V) and accumulation (1 V) potential for all applied temperatures. The capacitance represents the capacitance of the depletion region, which can be obtained from

ARTICLE IN PRESS R. Loef et al. / Solar Energy Materials & Solar Cells 92 (2008) 1579–1585

1581

4.2. TID results

Fig. 2. Impedance spectra of a TiO2/CuInS2 heterojunction at different temperatures at the accumulation (1 V) and flattening (0 V) potentials.

the imaginary part of the impedance. The resistor is the combined resistance of semiconductor layers and measurement equipment. Fig. 3 shows Mott–Schottky plots recorded at different temperatures. For all temperatures the inflection potential bordering situations II and III, i.e., the potential at which p-CuInS2 is just out of full depletion (Fig. 1), is above the flattening potential of 0 V. This means that the p-CuInS2 is always in full depletion during the TID measurements. For To370 K the I–II inflection potential, i.e., the potential at which n-CuInS2 is just out of full depletion, is close to the flattening potential (0 V). At higher temperatures this inflection potential moves to more reverse (negative) potentials. At the accumulation potential (1 V) both CuInS2 layers are in full depletion for all applied temperatures. From the Mott–Schottky plots, we derived that the system under investigation is described as a p–n–n junction (p-CuInS2/ n-CuInS2/n-TiO2). A 40 nm thin n-CuInS2 layer with a donor density of 2  1017 cm3 at 400 K is present between the n-TiO2 and the p-CuInS2. The p-CuInS2 layer has a thickness of 90 nm and an acceptor density of 4  1016 cm3 at 400 K. Band diagrams for this p–n–n junction are shown in Fig. 1. We have shown that at the accumulation potential of 1 V both CuInS2 layers, n- and p-type, are in full depletion, irrespective of the temperature (Fig. 1-I). When going to forward potentials, first n-CuInS2 loses its full depletion (Fig. 1-II) and later p-CuInS2 (Fig. 1-III).

As explained above, capacitance transients can be related to electronic (DLTS) and ionic (TID) effects. Fig. 4 shows a transient curve at 300 K. We find that the transient time at the flattening potential of 0 V is higher than the transient at the accumulation potential of –1 V, i.e. tf/ta41. Therefore, we conclude that capacitance transients are likely to be explained by TID, rather than DLTS at this temperature. The same is true for all applied temperatures. Nevertheless, the tf/ta is smaller than expected, probably due to the fact that there is no bulk region present in the CuInS2 layers. Fig. 5 shows capacitance transients at the accumulation potential. The reversibility of the experiments was checked by repeating the experiments a number of times with intervals of 1 up to 3 days. In the applied temperature range, the reversibility was good. Only at temperatures above 470 K irreversible changes occur. At temperatures below 360 K a capacitance growth after the instantaneous capacitance drop is found, while at temperatures above 360 K a capacitance decay is observed after the potential step. The capacitance transients are fitted to Eq. (1). In some cases two exponents are needed to obtain good fits. The transient times are used to calculate diffusion coefficients using Eq. (3) . Relatively small diffusion coefficients are found—e.g., depending on whether NA;p-CuInS2 or ND;n-CuInS2 is used, at 350 K diffusion coefficients of 2  1013 and 5  1014 cm2/s, respectively. This cannot be related to the transport of Cu, which is much faster [6,7,10,11,36]. The time it takes for Cu-ions to move through the complete CuInS2 layer is too fast to be detected in our set-up. A drift mobility of 6.6  1015 cm2/V s is reported for In-ions in CuInSe2 [12], which matches with the values found here. Therefore, we conclude that the ionic drift is related to In migration. The In-related defects in intrinsic CuInS2 are Frenkel defects (V00 In and Inddd ) and anti-site i ¨ ger–Vink notation is used. Cu defects Cu00In or Indd Cu . Here, the Kro and In are the metal ions, V is a vacancy and i an interstitial position. The normal letters refer to the type of atom, while the subscripts refer to the lattice position of this atom. Negative, positive and neutral charges, relative to the lattice sides, are given by the superscripts 0 , d, and x, respectively. Fig. 6 shows plots for ln(t/T) vs. 1/T, from which we determine the activation energies for diffusion according to Eqs. (3) and (4). Generally, four slopes can be observed, A–D. From each slope an activation energy is calculated. The activation energies are summarized in Table 1. Essentially, two activation energies are found, i.e., one around 0.5 eV and one around 1 eV. The latter is found with three different pre-exponential factors.

4.3. Discussion The system under investigation is a thin film p–n–n heterojunction instead of a several micrometers thick single-layer device. In the original TID theory, accumulation of mobile ions at the depletion layer edge is assumed. In reality, a concentration profile will be present at the depletion layer edge according to the Gouy–Chapman theory. This theory describes the formation of an electrical double layer that forms at the interface of an object when it is placed in a liquid. The double layer consists of two ionic layers, a surface charge at the surface of the object and a diffuse layer in the liquid. The ion concentration in the diffuse layer follows an exponential decay away from the interface. The behavior of the mobile ions in CuInS2 will be similar to that of the ions in the diffuse layer [37]. The electric field in the depletion layer forces the concentration of mobile defects to follow an exponential profile, Nmobile ¼ Nmobile;0 ezqf=kT

(5)

ARTICLE IN PRESS 1582

R. Loef et al. / Solar Energy Materials & Solar Cells 92 (2008) 1579–1585

Fig. 3. C2–f plots of a TiO2/CuInS2 heterojunction at different temperatures recorded at 1 MHz. The indicated regions I, II, and III correspond to the three situations explained in Fig. 1.

Fig. 4. TID signal at 300 K. Between 0 and 500 s the sample is at the flattening potential of 0 V. After 500 s the potential is stepped to the accumulation potential of 1 V. The first decay is due to the flattening of the mobile ion profile. After the potential step, an almost instantaneous decrease of the capacitance is observed due to the increase of space charge region. Subsequently a slow increase is observed, related to the movements of mobile ions. The t’s are the time constants obtained from fitting the exponential decay and increase, respectively.

Here, Nmobile,0 is the bulk concentration of mobile ions and z is the ionic charge. In thick layers it is allowed to describe this profile as a small accumulation layer. However, in our thin films this profile is distributed over a substantial part of the film, as shown in Fig. 7, where the Gouy–Chapman theory is used to model the profile of mobile donors and acceptors at the accumulation and flattening potentials at two temperatures. It is shown that there are significant differences in donor and acceptor profiles at the accumulation and flattening potentials, especially at high temperatures,

when the mobile acceptor profile even crosses the p-CuInS2/ n-CuInS2 interface, leading to changes in both the acceptor- and donor-densities in CuInS2. These profile changes may therefore be the main reason for the observed capacitance transients, as proposed by Igalson and Zabierowski [38]. In their work they also suggest a second explanation, in which activated shallow defects are converted into deep, non-activated, neutral midgap states. However, we do not believe that physical changes of defects occur in our samples, because it is unlikely that defect chemical reactions are completely reversible, as must be the case when degradation is absent. Although the activation energies found above are related to ionic phenomena and excitation of electrons over the bandgap is an electronic phenomenon, it is remarkable that both observed activation energies of the transport of mobile defects almost add to the bandgap energy of 1.5 eV. Furthermore, the activation energy of 1 eV matches well with earlier reported deep-trap states about 1 eV above the valence band [27–29]. Therefore, we believe that there is a relation between the ionic drift discussed above and electronic processes. The movement of F-centers due to capturing and releasing of electrons at different anion vacancies [39] inspired us and we postulate that the activation energies found in the present study are related to charging of the 1 eV electronic state. Fig. 6 shows that for all observed activation energies the transient times decrease with increasing temperatures, which implies charging of traps in all processes. More precisely, both the capture of an electron (EA ¼ 1 eV) and the release of an electron (EA ¼ 0.5 eV) result in charging of the defect (see Fig. 8). The above discussion is in line with earlier reported behavior of Cu(In,Ga)Se2 after proton or electron irradiation. Electron irradiation lowers the effective acceptor density of Cu(In,Ga)Se2 [40]. The efficiency of the ZnO/ CdSe/Cu(In,Ga)Se2 decreases after proton and electron irradiation. However, after thermal or light annealing the solar cell recovers [21–23].

ARTICLE IN PRESS R. Loef et al. / Solar Energy Materials & Solar Cells 92 (2008) 1579–1585

1583

Fig. 5. TID signals of a p-CuInS2/n-CuInS2/TiO2 heterojunction at different temperatures. At t ¼ 0 a potential step from 0 to 1 V reverse bias voltage is applied. An instantaneous decrease of capacitance is observed due to the increase of the depletion region in the heterojunctions. Consecutively, a slow increase (To360 K) or decrease (T4360 K) is observed due to ionic mobility.

Fig. 6. Temperature dependence of the TID transient times. Closed symbols are related to exponential growth, open symbols to decay. Four slopes are observed, A–D. EA is calculated from these slopes according to Eqs. (3) and (4). Calculated activation energies are summarized in Table 1.

Table 1 Summary of found transients and activation energies Slope

Tmin (K)

Tmax (K)

Transient type

Prefactor

EA (meV)

A B C D

300 340 340 420

330 380 420 460

Growth Decay Decay Decay

37.070.7 17.171.9 30.471.4 31.674.9

920720 465760 955750 10657190

A–D refer to the slopes in Fig. 6.

The nature of this neutral defect may be found in an associated x lattice defect. With the In-related defects in mind, (Cu00In Indd Cu ) is the most likely candidate. This associated defect is related to the well-known Cu–Au lattice disorder in CuInS2, where Cu and In are exchanged [41]. The Raman spectrum in Fig. 9 illustrates that indeed Cu–Au disorder is present in our samples. The peak around 295 cm1 indicates the chalcopyrite structure. The broad peak around 304 cm1 is related to the Cu–Au lattice defects [42–44]. The above findings are in line with current photoluminescence and transient absorption studies [45], in which it is shown that the associated lattice defect introduces two energy levels around

Fig. 7. Mobile donor and acceptor profiles in CuInS2 at 350 and 450 K at the accumulation potential (1 V, solid lines) and the flattening potential (0 V, dotted lines), modelled following the Gouy–Chapman theory.

0.25 and 1.1 eV above the valence band for Cu00In and Indd Cu , respectively. However, Cu00In and Indd Cu will remain associated, due to their strong Coulomb interaction. The rate-determining process in this ambipolar diffusion process will be the slowest component, being Indd Cu .

ARTICLE IN PRESS 1584

R. Loef et al. / Solar Energy Materials & Solar Cells 92 (2008) 1579–1585

clarity effective charges of the point defects are not given. After (6a)–(6c) are completed, atom B on an atom A position has moved to a new lattice site: BA ð2Þ ! BA ð1Þ

Fig. 8. Schematic representation of the defect state in CuInS2. In the ground state the defect is neutral (a). After capturing an electron the defect will be negatively charged, leaving a free hole in the valence band (b). After releasing an electron to the conduction band the defect will be positively charged (c).

(6d)

This process also takes place with element A on a B position, i.e. AB. Dissociation of the defects is considered, but is very unlikely due to the reversibility of the processes. Because In-ions diffuse slower than Cu-ions, In diffusion determines the transient times. The activation energy depends only on whether an electron is captured or released. Nevertheless, the pre-exponential factor, D0, in Eq. (4) depends on several factors (e.g., the number of electrons that are captured or released, or whether the associated defect is located in the n- or p-type CuInS2). Herewith, we can explain the observation of one activation energy with different pre-exponential factors in Fig. 6. Transport of associated Cu–Au defects towards the depletion layer edge reduces the concentration of deep recombination centers in CuInS2 inside the depletion region. Ion migration due to the change of the internal field at the heterojunction region upon applying light or an external voltage explains the self-healing properties of CuInS2, which is in line with the propositions of Guillemoles et al. [24].

5. Conclusions With TID measurements diffusion coefficients between 1016–1015 cm2/s at 300 K, with activation energies of 0.5 and 1 eV, have been found. It is postulated that these activation energies are related to charging of a neutral defect located 1 eV above the valence band, probably the associated Cu–Au disorder x defect, (Cu00In Indd Cu ) . Upon charging this defect complex it migrates to the depletion layer edge, which lowers the concentration of recombination centers. This explains the self-annealing properties of CuInS2.

Acknowledgement Advance Surface Technologies (Bleiswijk, The Netherlands) is acknowledged for supplying the samples. References Fig. 9. Raman spectrum of a TiO2/n-CuInS2/p-CuInS2 heterojunction. The dashed lines show the deconvolution of the spectrum into two Gaussian curves, resulting in peaks around 295 and 304 cm1.

When charged, the associated defect will move through the layers under influence of the electric field. Because the concentrations of Cu- and In-vacancies are relatively small, it is most likely that the movement of Cu- and In-ions will occur via swapping of lattice sites. A possible path for the transport of Cu00In x x and Indd Cu is via CuCu and InIn lattice sites, respectively, according to AA ð1Þ ! Ai þ VA ð1Þ

(6a)

BA ð2Þ þ VA ð1Þ ! BA ð1Þ þ VA ð2Þ

(6b)

Ai þ VA ð2Þ ! AA ð2Þ

(6c)

A and B are metal ions Cu or In, V is a vacancy and i an interstitial position. (1) and (2) are different lattice sites. For the sake of

[1] J.A.M. AbuShama, S. Johnston, T. Moriarty, G. Teeter, K. Ramanathan, R. Noufi, Properties of ZnO/CdS/CuInSe2 solar cells with improved performance, Prog. Photovolt. 12 (2004) 39. [2] T. Glatzel, D.F. Marron, T. Schedel-Niedrig, S. Sadewasser, M.C. Lux-Steiner, CuGaSe2 solar cell cross section studied by Kelvin probe force microscopy in ultrahigh vacuum, Appl. Phys. Lett. 81 (2002) 2017. [3] R. Scheer, R. Klenk, J. Klaer, I. Luck, CuInS2 based thin film photovoltaics, Sol. Energy 77 (2004) 777. [4] K. Ramanathan, M.A. Contreras, C.L. Perkins, S. Asher, F.S. Hasoon, J. Keane, D. Young, M. Romero, W. Metzger, R. Noufi, J. Ward, A. Duda, Properties of 19.2% efficiency ZnO/CdS/CuInGaSe2 thin-film solar cells, Prog. Photovolt. 11 (2003) 225. [5] M.A. Contreras, B. Egaas, K. Ramanathan, J. Hiltner, A. Swartzlander, F. Hasoon, R. Noufi, Progress toward 20% efficiency in Cu(In,Ga)Se2 polycrystalline thinfilm solar cells, Prog. Photovolt. 7 (1999) 311. [6] M. Kleinfeld, H.D. Wiemhofer, Chemical diffusion-coefficients and stability of CuInS2 and CuInSe2 from polarization measurements with point electrodes, Solid State Ion. 28–30 (1988) 1111. [7] M. Kleinfeld, H.D. Wiemhofer, Chemical diffusion in CuInS2 in the temperature-range of 20 1C–100 1C, Ber. Bunsenges.Ges.-Phys. Chem. Chem. Phys. 90 (1986) 711. [8] K. Gartsman, L. Chernyak, V. Lyahovitskaya, D. Cahen, V. Didik, V. Kozlovsky, R. Malkovich, E. Skoryatina, V. Usacheva, Direct evidence for diffusion and electromigration of Cu in CuInSe2, J. Appl. Phys. 82 (1997) 4282.

ARTICLE IN PRESS R. Loef et al. / Solar Energy Materials & Solar Cells 92 (2008) 1579–1585

[9] R. Herberholz, U. Rau, H.W. Schock, T. Haalboom, T. Godecke, F. Ernst, C. Beilharz, K.W. Benz, D. Cahen, Phase segregation, Cu migration and junction formation in Cu(In, Ga)Se2, Eur. Phys. J. Appl. Phys. 6 (1999) 131. [10] G. Dagan, T.F. Ciszek, D. Cahen, Ion migration in chalcopyrite semiconductors, J. Phys. Chem. 96 (1992) 11009. [11] D. Soltz, G. Dagan, D. Cahen, Ionic mobility and electronic junction movement in CuInSe2, Solid State Ion. 28 (1988) 1105. [12] K.V. Reddy, A.P. Kumar, Drift velocity measurements in CuInSe2 thin films, Proc. SPIE 2733 (1996) 403. [13] J.F. Guillemoles, U. Rau, L. Kronik, H.W. Schock, D. Cahen, Cu(In,Ga)Se2 solar cells: device stability based on chemical flexibility, Adv. Mater. 11 (1999) 957. [14] D. Schmid, M. Ruckh, H.W. Schock, Photoemission studies on Cu(In,Ga)Se2 thin films and related binary selenides, Appl. Surf. Sci. 103 (1996) 409. [15] D. Schmid, M. Ruckh, F. Grunwald, H.W. Schock, Chalcopyrite defect chalcopyrite heterojunctions on the basis of CuInSe2, J. Appl. Phys. 73 (1993) 2902. [16] Y. Yan, K.M. Jones, J. Abushama, M. Young, S. Asher, M.M. Al Jassim, R. Noufi, Microstructure of surface layers in Cu(In,Ga)Se-2 thin films, Appl. Phys. Lett. 81 (2002) 1008. [17] D. Cahen, J.M. Gilet, C. Schmitz, L. Chernyak, K. Gartsman, A. Jakubowicz, Room-temperature, electric-field induced creation of stable devices in CuInSe2 crystals, Science 258 (1992) 271. [18] S. Richter, Y. Manassen, D. Cahen, Bulk changes in semiconductors using scanning probe microscopy: nm-size fabricated structures, Phys. Rev. B 59 (1999) 10877. [19] S. Richter, D. Cahen, S.R. Cohen, K. Gartsman, V. Lyakhovitskaya, Y. Manassen, Fabrication of sub-mm m bipolar transistor structures by scanning probe microscopy, Appl. Phys. Lett. 73 (1998) 1868. [20] I.M. Ko¨tschau, H.W. Schock, Depth profile of the lattice constant of the Cu-poor surface layer in (Cu2Se)1x(In2Se3)x, evidenced by grazing incidence X-ray diffraction, J. Phys. Chem. Solids 64 (2003) 1559. [21] A. Jasenek, H.W. Schock, J.H. Werner, U. Rau, Defect annealing in Cu(In,Ga)Se2 heterojunction solar cells after high-energy electron irradiation, Appl. Phys. Lett. 79 (2001) 2922. [22] U. Rau, M. Turcu, A. Jasenek, Time constants of open circuit voltage relaxation in Cu(In,Ga)Se2 solar cells, Thin Solid Films 515 (2007) 6243. [23] A. Boden, D. Braunig, J. Klaer, F.H. Karg, B. Hosselbarth, G. La Roche, Protonirradiation of Cu(In,Ga)Se2 and CuInS2 thin-film solar cells, in: Photovoltaic Specialists Conference, 2000, Conference Record of the Twenty-Eighth IEEE, 2000, p. 1038. [24] J.F. Guillemoles, L. Kronik, D. Cahen, U. Rau, A. Jasenek, H.W. Schock, Stability issues of Cu(In,Ga)Se2-based solar cells, J. Phys. Chem. B 104 (2000) 4849. [25] R. Loef, J. Schoonman, A. Goossens, Elucidation of homojunction formation in CuInS2 with impedance spectroscopy, J. Appl. Phys. 102 (2007) 024512. [26] I. Lyubomirsky, M.K. Rabinal, D. Cahen, Room-temperature detection of mobile impurities in compound semiconductors by transient ion drift, J. Appl. Phys. 81 (1997) 6684.

1585

[27] J. Krustok, J.H. Schon, H. Collan, M. Yakushev, J. Madasson, E. Bucher, Origin of the deep center photoluminescence in CuGaSe2 and CuInS2 crystals, J. Appl. Phys. 86 (1999) 364. [28] T. Onishi, K. Abe, Y. Miyoshi, K. Wakita, N. Sato, K. Mochizuki, Study of deep photoluminescence levels in CuInS2 crystals, J. Phys. Chem. Solids 66 (2005) 1947. [29] E. Rudigier, T. Enzenhofer, R. Scheer, Determination of the quality of CuInS2based solar cells combining Raman and photoluminescence spectroscopy, Thin Solid Films 480 (2005) 327. [30] T. Heiser, A. Mesli, Determination of the copper diffusion-coefficient in silicon from transient ion-drift, Appl. Phys. A Mater. Sci. Process. 57 (1993) 325. [31] A. Zamouche, T. Heiser, A. Mesli, Investigation of fast diffusing impurities in silicon by a transient ion drift method, Appl. Phys. Lett. 66 (1995) 631. [32] D.V. Lang, Deep-level transient spectroscopy—new method to characterize traps in semiconductors, J. Appl. Phys. 45 (1974) 3023. [33] M. Nanu, J. Schoonman, A. Goossens, Nanocomposite three-dimensional solar cells obtained by chemical spray deposition, Nano Lett. 5 (2005) 1716. [34] R. O’Hayre, M. Nanu, J. Schoonman, A. Goossens, A parametric study of TiO2/ CuInS2 nanocomposite solar cells: how cell thickness, buffer later thickness, and TiO2 particle size affect performance, Nanotechnology 18 (2007) 055702. [35] R. O’Hayre, M. Nanu, J. Schoonman, A. Goossens, Q. Wang, M. Graetzel, The influence of TiO2 particle size in TiO2/CuInS2 nanocomposite solar cells, Adv. Funct. Mater. 16 (2006) 1566. [36] J. Hofhuis, J. Schoonman, A. Goossens, Time-of-flight studies on TiO2/CuInS2 heterojunctions, J. Appl. Phys. 103 (2008) 014503. [37] A.J. Bard, L.R. Faulkner, Electrochemical Methods; Fundamentals and Applications, Wiley, New York, 1980. [38] M. Igalson, P. Zabierowski, Transient capacitance spectroscopy of defect levels in CIGS devices, Thin Solid Films 361 (2000) 371. [39] R.W. Pohl, Electron conductivity and photochemical processes in alkali halide crystals, Proc. Phys. Soc. London 49 (1937) 3. [40] A. Jasenek, U. Rau, Defect generation in Cu(In,Ga)Se2 heterojunction solar cells by high-energy electron and proton irradiation, J. Appl. Phys. 90 (2001) 650. [41] D.S. Su, S.H. Wei, Transmission electron microscopy investigation and firstprinciples calculation of the phase stability in epitaxial CuInS2 and CuGaSe2 films, Appl. Phys. Lett. 74 (1999) 2483. [42] J. Alvarez-Garcia, A. Perez-Rodriguez, A. Romano-Rodriguez, T. Jawhari, J.R. Morante, R. Scheer, W. Calvet, Raman scattering structural evaluation of CuInS2 thin films, Thin Solid Films 387 (2001) 216. [43] J. Alvarez-Garcia, J. Marcos-Ruzafa, A. Perez-Rodriguez, A. Romano-Rodriguez, J.R. Morante, R. Scheer, MicroRaman scattering from polycrystalline CuInS2 films: structural analysis, Thin Solid Films 361 (2000) 208. [44] M. Nanu, J. Schoonman, A. Goossens, Raman and PL study of defect-ordering in CulnS2 thin films, Thin Solid Films 451–52 (2004) 193. [45] J. Hofhuis, J. Schoonman, A. Goossens, Elucidation of the excited-state dynamics in CuInS2 thin films, unpublished work.