Ion Potential Diagrams for Electrochromic Devices

Ion Potential Diagrams for Electrochromic Devices Francesca Varsano,6L) David Cahen$ Franco Decker," Jean François Guillemoles,0,c and Enrico Masetti*,d 6Department of Materials and Interfaces, Weizmann Institute of Science, 76100 Rehovot, Israel 6Dipartimento di Chimica, Università di Roma "La Sapienza," Rome, Italy CNRS, Laboratoire d'Electrochimie, ENSC1 Paris 75005, France 4ENEASezione Dispositivi Ottici, 00060 Rome, Italy ABSTRACT Ion potential diagrams can facilitate the description of systems in which ionic species are mobile. They depict qualitatively the spatial dependence of the potential energy f or mobile ions, somewhat akin to band diagrams for electrons. We con- struct ion potential diagrams for the mixed conducting (oxide), optically active electrodes of five-layer electrochromic devices, based on reversible Li intercalation. These serve to analyze stability problems that arise in these systems. We then use them as building blocks to arrive at ion diagrams for complete devices. This allows analyses of (dis)coloration kinetics. Infroduction The idea behind the construction of ion potential diagrams is to facilitate the description of systems in which ionic species are mobile. These diagrams depict the spatial dependence of the potential energy for mobile ions in a manner similar to band diagrams for electrons. While plots of the chemical potential of mobile ions against a spatial coordinate can be found in several places in the literature, including textbooks,1'2 the systematic use of this representation is scarce. Such plots are especially interesting when we deal with mixed ionic/electronic semiconductors, in which there is a large difference in mobility, B, between the dominant ionic and electronic carriers. This is so because in this case the concentration gradients of the relatively immobile species lead to persistent electrical potential differences in the material. Systems in which B0,, can be neglected completely are best described by electron < band diagrams. We consider situations in which Beiectn but where B0 is sufficiently large to lead to compositional changes on a time scale relevant to the experimental observation or device usage. (Opposite situations can be treated in a way that is to a large extent analogous to the one presented here.) Recently we used ion potential diagrams to explain the differences in stability between two types of solar cells containing a mobile dopant in one of the components.3 A more general description of such potential diagrams is presented elsewhere and was applied to three-layer electrochromic cells . Here we concentrate on the construction of ion potential diagrams for five-layer electrochromic devices.5 Such devices are made up of an electronically conducting contact layer, a mixed conductor A (mostly an oxide), an electronically blocking, but ion-conducting electrolyte, a second mixed conducting film B (again mostly an oxide), and another electronically conducting layer. The idea is that the system can switch from an optically transparent to optically opaque, or at least intensely colored state by an electrochemical redox process. The optical changes occur in one or both of the mixed conducting films. Most commonly those films change their optical absorption due to reversible lithium ion or proton intercalation.6 A typical device structure, shown schematically in Fig. 1 is ITO/LiWO3/Li-electrolyte/Li9CoO2/ITO parent conductor. Both solid electrolytes (e.g., LiNbO3 7) and liquid/gel ones [e.g., LiC1O4-PC/LiC1O4-PGG-PMMA 51] have been used as Li conductors [PC is propylene carbonate and PPG-PMMA is poly(propylene glycol)-poly(methyl methacrylate)]. The devices are under active study and development for use in "smart" windows and displays.5'9 Among the problems that are faced in this application are the following (i) lack of stability, for example, due to Li Electrochemical Society Active Member. 4212 briefly recalling the theoretical basis underlying ion potential energy diagrams. To analyze the devices, we first consider parts of them. Thus we look at the electrode part of the device, viz., electronic conductor + mixed conductor. This serves to analyze the problem of long-term elec- trode stability due to ion penetration into the electronic conductor, if ion mobility in the latter is significant. Subsequently we consider the mixed conductor + ion conducting electrolyte. We then use the results as building blocks to construct diagrams for the complete device, first at open circuit and then under bias, analyzing their behavior through charge and discharge cycles. Finally we consider possible evolutions of electrode materials, analyzing them from the point of view of coloring/bleaching kinetics. Theoretical Background conceptual starting point for ion diagrams is the electron band diagram in semiconductors. That diagram represents the spatial variation of potential energy levels of electrons, including the way they are changed by electric fields. It is important to realize, though, that an elecThe tron band diagram presents a mixture of quantities, as has been explained elsewhere.4 Briefly the lines drawn for the bottom of the conduction and top of the valence band correspond to [1] he° + {(—e)e} where h ji. The representation of the valence band top can also be considered as —[h + eø] [2] with h°h P°h, where e is the unit electron charge, 0 is the electrical potential, and heh are the standard enthalpies of electrons and holes. For delocalized electrons these can be Li electrolyte [A] where ITO stands for indium-tin oxide, an optically trans- * penetration into the electronic conductor (mostly ITO), (ii) slow switching from the opaque/colored to the transparent state or/and back, and (iii) parasitic reactions at the interfaces. Several of these problems also affect thin film rocking chair battery structures,1° which are very similar. We use ion potentials to analyze these systems and especially the stability problems that plague them. We start by ITO ITO =-H -Il / electrochromic oxide counter- electrode Fig. 1. Schematic view of a five-layer elecfrochromic device in its transmissive mode. ITO stands for indium-tin oxide (1n203:Sn). J. Electrochem. Soc., Vol. 145, No. 12, December 1998 The Electrochemical Society, Inc. Downloaded on 2016-02-18 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract). 4213 J. Electrochem. Soc., Vol. 145, No. 12, December 1998 The Electrochemical Society, Inc. equated with the standard partial Gibbs free energies, kh They differ in nature from the formal thermodynamic electrochemical potential = J1, + (eo) [3] which appears as the Fermi level, and from the chemical potential = + (kT) ln(n/Nc) [4] Here n is the free electron concentration and Nc is the density of states in the conduction band. When dealing with ions we are interested in the chemical3 and the electrochemical3 potentials of the mobile ion(s) and those are the quantities that we plot in ion potential diagrams. This difference from electron band diagrams stems from the much greater importance of entropy in the case of ion distributions than in that of electronic charge carriers.4 Mostly it suffices to thaw the potentials for one ion, as cases with more than one mobile ion are rare. Use of these potentials, the partial Gibbs free energies, yields the thermodynamic driving forces for ion motion (from their difference) and the fluxes of ions (which are proportional to their gradients). The use of thermodynamic potentials for ions can be understood by considering them as representing the free energy changes involved in the relevant half-reaction.1' Thus for all that matters here we are concerned with three species, viz., e, Lit, and {e, Lt}, which are connected via the half-reaction Li4 + e = Li [5] where we consider Li as a species also in the absence of metallic phase, because Li4 and e move together. Further issues, concerning the ability to define thermodynamic quantities for charged species and the supposed need for a reservoir of the relevant chemical species in the system under consideration, have been discussed elsewhere.4 Reference levels (cf. Fig. 2).—An important issue is the choice of the reference level without which we cannot r Li metal ELj " ill erence level (cf. Fig. 2 a).4 Sign conventions—In contrast with the electron band diagram, in the ion diagrams we use the same scale, irrespective of the charge of the species, i.e., larger potential energies, chemical or electrochemical ones, appear higher on the diagram.e As a result of the choice of the reference levels, = —W (the work function). Because work function values are positive, , will be negative and materials with lower work function will have a higher li The chemical potentials of neutral species will be lower (i.e., negative) than their reference levels. From the relation liL, = li + liLI the value of 1L1 is readily obtained (except for a constant). We note that this relation also defines a reference level for The way the diagrams are constructed implies that all the species (positively and negatively charged ones, as well as neutral ones) will move spontaneously from a higher energy value (higher up in the diagram) to a lower one, if a path for it exists.' This implies that a mobile ion will move from A to B if (f...11b) > 0, in accordance with the phenomenological laws where the flux of species S is pro- portional to —if. We now use these definitions and conventions to construct ion potential diagrams for the contact/mixed conductor and the mixed conductor/electrolyte parts of an electrochromic device. Diagrams for those are then combined to yield the diagram for the device. Construction of Diagrams We consider the ITO-LiWO3 and the ITO-Li,,,Co02 junctions in some detail, as those interfaces are often encountered in electrochromic devices. As is well known,7"2 one of the causes for device degradation is lithium migration into the ITO. This leads not only to a decrease in coloration efficiency (the fraction of Li inside the ITO does not contribute to the coloration), but also to a semipermanent optical (near-IR) absorption in the ITO,'3"4 due to the electrons that Li electrolyte — PLi in LixMOy I Li metal compare the potentials in two or more different materials. For the electron we use, as in electron band diagrams, the energy of the electron at infinity in vacuum, the vacuum level (cf. Fig. 2b). However for atoms and ions this is not really a practical proposition. Instead we use the energy of the discharged dopant (Li' in our case) in its thermodynamic reference state (Li metal at STP) as the common ref- Li RUU-.sxM0y) (a) accompany inserted Lit An additional reason can be ITO decomposition, because Li is reported to react with SnO2, forming an Sn-Li alloy (cf. the section on the Relative stability of the TCO layers and Ref. 15). In the following we consider not only W03 and Li,CoO2, but also ITO as a mixed electronic/Li ion conductor, to understand the factors that govern Li insertion into ITO. The diagrams are constructed using the chemical potentials for Li and electrons. Those can be calculated from literature data, or if unavailable, the necessary quantities can be determined experimentally. This is done as follows. Li chemical potential, in the oxides is deduced from emf values of electrochemical cells of the type (cf. Fig. 2a) Vacuum I Ee Work Function [B] LiMO5/electrolyte/Li We can understand what is measured here, by considering half-reaction 5. For that reaction the following relation holds 11.1+ + '1, = fle h1UxMOy (b) Fig. 2. Reference levels for ji,,, the Li chemical potential (a), and for m. the electron electrochemical potential (b). In (a) if11 (in electron volts; left side) has the same numerical value as if, the emf of the cell (in volts; right side), as shown in the box. From the relation Pt = lie + + the reference level For lu÷ is obtained. lin = liLt [6] Note that, with respect to ionic species, including defects, we adopt a convention, different from what is done in electron band diagrams. Those diagrams are used also to understand the behavior of holes, in which case the scale is reversed, i.e., higher hole energies appear lower. Sacrificing some consistency but maintaining an easy link with electron band diagrams, we keep in the ion potential diagrams this practice for holes. Thus, the electrical potential difference (eA-ea), which gives the potential energy of a positive electronic test charge, is defined as positive, if it presents a driving force for electrons to move from B to A. Downloaded on 2016-02-18 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract). 4214 J. Electrochem. Soc., Vol. 145, No. 12, December 1998 The Electrochemical Society, Inc. Because Li is supposed to be able to equilibrate freely in cell B, given sufficient time, Eq. 6 simplifies to le = P'Li + constant [7' This and Eq. 6, applied to the whole cell B gives &OL1* = 0. means that the electromotive force (emf), A1,, measures A1LI = Ap.L1 = {,1L1(LiJVIOS) — p.LI(Li)}, where p.L1(Li) is our reference. ISLI(ITO) is obtained in a similar fashion. Electron (electro)chemical potential, j.tj1J.—To construct full diagrams we need (and , if A4 0) in ITO and in LiXMOS. These we get from work function values (Fig. 2b). One should be careful here, as literature values may not be relevant to the case under consideration and/or are not very precise! This is so, because they depend strongly on the conditions under which they are measured and, in the case of a semiconductor, on its doping level. A rough estimate can be made on the basis of ioniza- tion potential, electron affinity, and bandgap values, since these data are generally easier to find in the literature. It is best, though, to determine 1W experimentally This is possible even without resorting to UHV photoemission spectroscopy (PES) techniques, for example, by measuring contact potential differences (CPS).'°"8 Both PES and CPD can be used to follow the electrical evolution of an interface, thus directly yielding values for the electrical potential difference across the interface, without the uncertainties associated with the use of individual work function values. ITO:Li Fig. 3. Ion diagram for ITO, 2.8 -3.0 lTO/LiWO3 junction (b) after electrons have equilibrated. We discuss first Li migration and diffusion processes that can occur upon contact between ITO and LiJV109, in terms of the overall thermodynamic driving force (Fig. 3 and 4).5 ITO and LiXWOS (x = 0.2) (cf. Fig. 3).—The difference between RLI in ITO (with a minimal Li concentration) and liL, in LiWO3 (x = 0.2) is around —0.8 eV. This number is calculated using a value of —2.8 to —3.0 eV for liLi inside ITO13 and of —1.8 to —2.3 eV for ILL1 inside Li02WO3. 8,19-21 Titration curves give liLt in LiWO3, for different Li concentrations, referred to lithium metal. These data are used in Fig. 3 where the thermodynamic quantities of interest are sketched for the ITO/LiWO3 junction before and after electron equilibration. At this stage we do not take into account the exact shape of tIe and Mi+ close to the interface. To have a detailed picture of the situation, data on the built-in voltage (band bending in electron band diagrams) are needed. Knowledge of the density of states for electrons/holes and the free carrier concentrations then allows an estimate for the width of the region over which tIe occurs. I— Li0 2W03 AW —1 ITO: Li ref Li Li02 W03 le 1' LiWO3 is colored. IR: interfacial region (not drawn to scale); the to the electrochemical, chemical, and electrical potentials, respec- Electronic Conductor (Contact) /Mixed Conducting Electrode Combinations /t LiXWO3 (a), and for the horizontal arrow in (b) denotes the electric field direction; W: work function; 'q, p., and e refer Discussion of Diograms 1 Li 1 Li+ 2.0 -2.2 1] le Li± R Li } Li 1 0.8 t Li R Li tively. All numbers refer to potential energies in electron volts. IR b a _____I _____ ITO:Li I lTO/LiCoO2 junction (b) after electrons have equilibrated. LiCoO2 is transparent. Symbols as in Fig. 3. LiCoO2 3t 2.8 -3.0 Fig. 4. Ion diagrams for ITO, liCoO2 (a), and for the ITO:Li ref Li LiCoO2 le 1 Li 1 LF 1 Li 1 Li R Li El Li 4) le R Li 4) R Li — 4) JR (a) (b) Downloaded on 2016-02-18 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract). 4215 J. Electrochem. Soc., Vol. 145, No. 12, December 1998 The Electrochemical Society, Inc. ITO and Li1..CoO2 (cf. Fig. 4)—Figure 4 shows p.,.. for ITO + LiCoO2, and for the ITO/LiCoO2 junction after elec- tronic equilibration. In this case Ai.t1 is less than in the drop (Aø) at the interface is equal, but opposite in sign, to Aji, the electron chemical potential difference between the implies that after two materials, i.e., = —AWB1 < B1 previous example (—0.2 to 0.4 eV) and, most importantly, has the opposite sign. This implies that PLj inside the electrochromic oxide is lower than in ITO, even when there is no measurable amount of Li in the ITO. Factors governing Li equilibration between LiXMOY and ITO—The pure thermodynamic driving force for Li redistribution, upon junction formation, is Aut1 before contact. contact the Li ions are in a nonequilibrium situation, namic driving force for Li to enter ITO, while in the latter this is not so. Note that it is Ap.Lf (= nL1), rather than &qj÷, that determines the thermodynamic driving force, because Li needs to be considered together with an electron. We stress that A(iLI does not tell us anything about the electric fields that may exist at the interfaces, even though those fields are important for the kinetics of the processes. For example, in a situation where A(it (= Aiu) > 0, between LiXMOY and ITO, as is the case for W03, the difference changes in surface state density for ITO and LiMO6) Figures 3a and 4a show how Ais for ITO/LiWO3 and ITO/Li1CoO2 differ. In the former there is a thermody- between their Fermi levels (i.e., in work functions) can retard or accelerate Li migration in the initial stages, after contact is made between the two phases. However; it cannot dictate whether the process will or will not take place. Relative stability of the TCO layers—Decomposition of the TCO can occur via reactions like15 Sn02 + 4Li -* 2Li5O + Sn if the chemical potential difference between Li in Sn02 and the Li° reference electrode is sufficiently small. As noted above, the Li chemical potential difference PLj can be translated into a cell voltage at rest (at open circuit; in = 0). From the free the short-circuited "rest" state energies of formation, the corresponding potential for tin oxide is —1.5 V vs. Li electrode (before insertion the potential is around 3 V vs. Li). As noted by Huggins15 the reaction Sn02 2Li — Li20 + SnO will occur before the abovenoted one at —1.9 V vs. Li. In the case of 1n203, a value of —1.4 V vs. Li is found for Sn formation. From the data in Table I it can be seen that use of Sn02 with the colored form of LiWO3, with NiO or with Nb205 may also lead to such undesirable reduction process. Thermodynamically ZnO is a more stable electrode (down to —1.2 V vs. Li). Therefore one can expect it to be less susceptible to decomposition due to overpotentials that may occur at this interface during operation. Depending on Li titration characteristics of ZnO and on the Li diffusion coefficient in ZnO, this compound could be another interesting candidate for the TCO layers. Kinetics of ITO/LiXMOY junction forination.—Because we the first process are considering systems where B that occurs when ITO and the LiXMO5 are brought in contact, is a flow of electrons between the two materials, which will continue until the Fermi level is constant throughout the junction. This is achieved when the electrical potential Table I. Chemical potential Electrochromic, mixed conducting oxide Li2.MoO, c Li1Nb,O, LiXV,O, a C a NiO,:Li" C Li,CoO, ITO:Li Let us consider this nonequilibrium situation for Li. Because BLI inside the mixed oxide is much larger than across the interface or inside the ITO, lithium will first of all move inside the mixed oxide, adjusting to decrease Aq1 resulting from the change in 0. We distinguish the following situations (ignoring WITO> WL1Mo1 where the Li ions will accumulate near the interface, more than they did before contact (applies to {Li}WO3) and W < Wj51 where fewer Li ions will accumulate near the interface, than before electronic equilibration (applies to Li1_CoO2). This line of reasoning can be used also to gauge how the junction will react under reverse and under forward bias. The description, given above, is simplistic as it implies a direct link between differences in work functions and interface electric fields. Especially from semiconductor science we know that this (so-called Anderson model) behavior is the exception rather than the rule!1617 Naturally the choice of mixed oxide is dictated first of all by the need to have a system that will color or discolor in tandem with the behavior of the other electrode. Additional requirements are low resistance contact and optimization of the mixed oxide/electrolyte interface. In summary, classification of different materials in terms of their stability with respect to Li migration into a transparent conducting contact, such as ITO, requires knowledge of which can be obtained from titration measurements, for example, and data for which are often available in the literature. Additional insight is gained by knowledge of the direction and magnitude of the electric field at the interface, information that can be obtained from work function data, or, preferably, work function measurements! Mixed Oxide/Electrolyte Combinations We turn now to the mixed oxide/electrolyte interface. A proper description requires definition of all chemical potentials, something that is problematic for the electrons in the electrolyte. The issue has been considered extensively in liquid electrolytes. There the redox potential of the redox couple has been identified with the Fermi level."22 We can circumvent this problem by considering the overall driving force for Li motion, as given by A(i and M+ between the two mixed oxide electrodes. Their differences will dictate where the Li will be, assuming that the system has the possibility to equilibrate freely. In addition, using the knowledge of the type of conductivity and the work function difference, we can get some of Li in possible and practical mixed conducting electrode materials, vs. Li metal, for different Li concentrations. vs. Li metal (eV) Bleached state —3.3 to —3.0 LiWO, because the electric field will have changed from that at the free oxide surface to that at the interface. —3.0—3.2 —3.6 —1.8 —3.2 —3.0 to —2.9 ,s,, vs. Li metal (eV) Colored state (x) Value Ref. —2.4 to 0.2 —3.0—1.6 —2.7 to —3.2 —3.5 0.5 9, 22, 23, 24 25, 31, 32 33, 34 —4.1 0.3-0.5 0.3 1 0.5 35 29 36 37 14 Anodic. <1. 'Cathodic. The large spread in data is due to the different techniques used for WO, sample preparation, yielding a spectrum of thin films, from amorphous to polycrystalline. Downloaded on 2016-02-18 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract). 4216 J. Electrochem. Soc., Vol. 145, No. 12, December 1998 The Electrochemical Society, Inc. idea of the electric fields and ion potential differences that Li LiCoO2 exist at the oxide/electrolyte interfaces at open circuit. Issues that can be analyzed are the potentials that need to be applied for (dis)coloration, and the ease and speed by electrolyte which the Li can be inserted and extracted. These will give an idea about the times needed for (dis)coloration, when the rate-limiting step is Li diffusion. wo3 Li Electrode/Electrolyte/Electrode System System at open circuit.—At open circuit the only relax- -j ation that is possible is that of electrons and ions in the elec- Lf1 trode material, as described above for the ITO-electrode junction, and Li ion exchange at the interface with the electrolyte. Assuming an n-type cathodic material and a p-type anodic material,h we can draw the situations for ideal inter- faces, before contact between the electrodes through the electrolyte, without (Fig. 5a) and with (Fig. 5b) surface electric fields. The surface electric fields can be due to localized electronic charge in surface states. We have assumed here that both ionic and electronic charge carriers can provide the charge exchange with the surface. It would appear to be possible to actually measure the interface electric field, as noted earlier! When the two electrode/electrolyte systems are combined, at open circuit, we get, starting from Fig. 5a, a situation as shown in Fig. 6 for the case where we neglect surface elec- tric fields. The system will evolve toward an equilibrium situation in as far as this is possible without electronic carrier equilibration. It is possible to include also surface electric fields, that result from space charge!' the interface fields are created by Li ions moving through the (ironically highly conducting) electrolyte to W03, the oxide with the lower Li ion electrochemical potential. This leads to electric fields at the electrode/electrolyte interfaces, because the electrolyte is blocking for electronic carriers. The field will Unless special measurements are made on the evolution of the field during interface formation,3 one needs to assume that the original fields are unchanged upon assembling the complete device structure. charge exchange with the electrolyte. Symbols are as in Fig. 3. extend into the materials with the lower charge carrier den- sity. In our case these will be the mixed conducting oxides, at least initially This is the situation shown in Fig. 6. System under bias—Having established the situation at open circuit, we apply bias to color the device, which means that the n-type (W03) electrode will be negative with respect to the p-type (Li1CoO2) electrode.14 Immediately after applying the bias we get the situation as shown in Fig. 7, assuming complete band flattening.' Then 1Li will have the same shape as I1Li We assume furthermore no change in the Li concentration profile, i.mtil well after the bias voltage has been imposed. This will facilitate Li ion migration into W03, and deintercalation of Li out of Li1_CoO2. The system will evolve toward steady state as bias remains applied, and electrons flow in the external circuit. While the actual positions of the electrical potential drop are not shown in Fig. 7, we note that normally most of this potential drop will be across the interfaces. If the Li concentration in the electrolyte is 1O' to 1022 cm3, the electrical potential drop across the electrolyte will be negligible. In Fig. 7 we follow what is com- W03 LiCoO2 Fig. 6. Ion diagrams for the cell LiCoO2/Li-eIectrolyte/LiWO3 (x 0) at open circuit (derived from situation depicted in Fig. 5a). Surface states are neglected and the electric field is due only to monly done in electron band diagrams, where, under the nonequilibrium condition of external bias, quasi-thermodynamic tie 1 Li potentials are used. 1lLi Li Li ® 1e 1Li Li1CoO2 Li electrolyte LiWO3 0 (a) _ W03 LiCoO2 11 Li Li - — 1 Li PLi 1L1 Li (b) Fig. 5. Ion diagrams for U_CoO2 and for LiWO3 (x = 0) (a) neglecting surface electric fields and (b) in presence of surface electric fields. Li chemical potentials are taken from Table I. Symbols as in Fig. 3. Fig. 7. Ion diagrams for the cell LiCoO2/Li elecfrolyte/LiWO3 (x = 0) under applied bias. Li,,W03 is polarized negatively with respect to LiCoO2 to have the device in the colored state. Horizontal arrows show the direction of Li flux. Symbols as in Fig. 3. Downloaded on 2016-02-18 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract). J. Electrochem. Soc., Vol. 145, No. 12, December 1998 The Electrochemical Society, Inc. Using this diagram it is also possible to examine the effect of overpotentials. In the initial state of tbe intercalation, for example, we just showed that Li ions accumulate at the interface witb W03. There might be cases where this accumulation leads to a p., value that exceeds that required for cathodic decomposition. The same way of reasoning applies to anodic decomposition reactions. Possible evolutions of electrode nsaterials.—Let us now look at some (at times) hypothetical variations of this situation. These can be classified using the electrode/electrolyte subsystem as a building block, according to the change in electronic properties of the electrode upon intercalation and deintercalation of the mobile ion. In the following we give this classification, based on the change upon Li intercalation (a) and deintercalation (b) metal (j) p-type semiconductor 5z metal (ii) n-type semiconductor t insulator (iii) insulator (iv) n-type semiconductor p-type semiconductor a p-type semiconductor y n-type semiconductor, with 4217 lematic, when the discoloration process, induced by deintercalation (b), starts. Case iv.—The initial and final situations for this case will be as described for ii, and the intermediate stages will be as described for case iii. Case v—In this case initial intercalation of Li into the p-type material will be assisted by the space-charge field that exists in it. The intercalation process will then typeconvert the material to n-type. The main problem with this case is that type conversion commonly occurs via a compensated, highly insulating state. Ignoring this problem for the moment, we can foresee a material that will be p when intercalating and switch to n when fully intercalated. To be able to facilitate intercalation and deintercalation, the system should show hysteresis in both these processes. Such hysteresis can then allow the p-state to persist sufficiently long during intercalation, to have its space-charge field help the system to reach its fully intercalated state. Upon deintercalation the system, which turned n-type upon full intercalation, should now persist in this state for sufficient time to help in the deintercalation process. Certain conducting polymers might be possible candidates for this behavior. This can be understood from looking at the recently described light-emitting electrochemical cell (LEC), where the polymer blend used by Pei et al.24 can switch from p to n passing through an intrinsic stage. In the LEC both types are created simultaneously, separated in space. When used for case v, they would need to be separated in time. Without the above mentioned hysteresis, the spacecharge fields can assist Li migration only at the beginning or at the end of the process, as for example in the LiCoO2/W03 window, studied by Goldner èt al. ,' at the end insulating (compensated) intermediate state (v) Case i applies to W03, case ii to Li1_CoO2 and case iii to V205. Vanadium pentoxide, however, due to its peculiar of the bleaching. However at the beginning of the col- the process is age, to pass through the compensating state. In principle this could be possible, given a suitable band structure with a large midgap impurity band. As shown by reference to the LEC, one way to obtain electronic band structure, transforms into an insulator upon intercalation rather than upon deintercalation, i.e., n-type semiconductor fl insulator Although cases i and ii have been at the center of our discussion, we add here their expected behavior as a result of the change in electronic properties after the initial phase of Li intercalation/deintercalation. Case i—Li intercalation (a) into W03 requires a voltage to overcome a kinetic barrier, resulting from the depletion layer in Li-free W03. As the intercalation progresses into W03, it becomes metallic. Therefore Ac across it will become negligible and, thus, there will be no longer any difference between Ap and A1LI+. This change will also de- crease the interfacial resistance and voltage drop at the interface, leaving only a voltage drop in the electrolyte. Deintercalation (b) will therefore proceed initially according to the externally imposed Ap and only at its last stages will it be subject to the influence of an internal electric field, which will then aid the process. Case u—This is similar to case i, but with the intercalation and deintercalation steps reversed. Deintercalation, after the initial stage that we described for LiCoO2, will proceed without the need to overcome any potential barrier, because Li-deficient material is degenerate. Initially, intercalation (a) will follow the applied potential, until p-type semiconductivity is regained. Then, in the final stage of intercalation, the electric field inside the mixed conductor will accelerate Li intercalation. Case iii.—Because of its special and structure, Li intercalation of V205 (a) actually fills a band and thus makes the material insulating.23 The initial situation is as that described for W03 (case i). However, as intercalation progresses, higher and higher voltages are needed to overcome the potential drop in the electrode. This is especially prob- oration process of that window, space-fields act against ion migration, as shown in Fig. 5. If type conversion could occur via a metallic intermediate state, then this would obviate the need for a high volt- material that will fit cases v is to consider composite systems. We Conclusions have shown how ion potential diagrams can give information on the possible occurrence, the direction, and the relative magnitude of particle fluxes in Li-based, fivelayer electrochromic device structures. They aid in classifying potential or actual electrode materials, with respect to Li migration into the transparent conducting contact material, and help to find suitable electrode/electrolyte combinations. The diagrams rely on experimentally accessible numerical data for Li and electron chemical potentials. The latter can also be used to predict the kinetics of Li migration. Thus, such diagrams can help selection of properties that a material should have for optimal performance in these types of devices. Extension of the ideas given here, to other systems, such as charge storage devices (batteries, capacitors) will be reported on elsewhere. Acknowledgments D.C. and J.EG. thank the EU-Israel research program via the Israel Ministry of Science (MOS) and DGXII for partial support. Israel-Italy collaboration was supported also by travel grants (to EV., ED., D.C.) from the scientific exchange program, administered by MOS (Jerusalem) and the Italian Consiglio Nazionale per le Ricerche (CNR, Rome). Manuscript submitted November 17, 1997; revised manuscript received July 30, 1998. REFERENCES 1. C. Desportes, M. Duclos, P Fabry, J. Foulétier A. Hammou, M. Kleitz, E. Siebert, and J. L. Souquet, Elec- trochimie des Solides, PUG, Grenoble (1994); H. Downloaded on 2016-02-18 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract). 4218 J. Electrochem. Soc., Vol. 145, No. 12, December 1998 The Electrochemical Society, Inc. Schmalzried, Chemical Kinetics of Solids, VCH, Weinheim (1995). 2. See, for example, J. M. Blakely, in Electrode Processes in Solid State lonics, M. Kleitz and J. Dupuy, Editors, pp. 83-118, Reidel, Dordrecht, (1976); J. Maier, Angew. Chem., 32, 313 (1993). 3. J. F Guillemoles and D. Cahen, Cryst. Res. Technol., 31, 147 (1996). 4. J. F. Guillemoles and D. Cahen, lonics, 2, 143 (1996); J. F Guillemoles and D. Cahen, To be published. 5. 5. K. Deb, Philos. Mag., 27, 801 (1973); C. G. Granqvist, in The CRC Handbook of Solid State Electrochem- istry, P. J. Gellings and H. J. M. Bouwmeester, Editors, Chap. 16, pp. 587-616, CRC Press, Boca Raton, FL (1997); Sol. Energy Mater. Sol. Cells, 39(2-4) (1995) special issue on electrochromics, B. Scrosati, Editor. 6. D. Dini, F. Decker, and B. Masetti, J. Appl. Electrochem., 26, 647 (1996). 7. R. B. Goidner, T. E. Haas, F 0. Arntz, S. Slaven, K. K. Wong, B. Wilkins, C. Shepard, and W. Lanford, Appl. Phys. Lett., 62, 1699 (1993). 8. A. M. Andersson, C. G. Grandqvist, and J. R. Stevens, Appl. Opt., 28, 3295 (1989). 9. M. Green and K. Kang, Displays, 166 (Oct 1988). 10. C. Julien, in The CRC Handbook of Solid State Electrochemistry, P. J. Gellings and H. J. M. Bouwmeester, Editors, Chap. 11, pp. 371-406, CRC Press, Boca Raton, FL (1997); Solid State Microbatteries, J. R. Akridge and M. Balkanski, Editors, NATO ASI Series B217, Plenum Press, New York (1990); C. Julien and G. A. Nazri, Solid State Batteries: Materials, Design and P. Zerigian, Solid State lonics, 70/71, 613 (1994). 15. T. Brousse, D. Defives, L. Pasquereau, S. M. lee, U. Herterich, and D. M. Schleich, Ionics, 3, 332 (1997); R. A. Huggins, lonics, 3, 245 (1997); Y. Idota, M. Mishima, M. Miyaki, T. Kubota, and T. Miyasaka, Eur. Pat. Appl., 651450 Al 950503 (1995). 16. M. Leibovitch, L. Kronick, E. Fefer, V. Korobov, and Y. Shapira, Appl. Phys. Lett., 65, 457 (1995). 17. L. Kronick, L. Burstein, M. Leibovich, Y. Shapira, D. Gal, E. Moons, J. Beier, G. Hodes, D. Cahen, D. Hariskos, R. Klenk, and H. W. Schock, Appl. Phys. Lett., 67, 1405 (1995). 18. S. K. Mohapatra and S. Wagner, J. Electrochem. Soc., 125, 1603 (1978). 19. M. Green, Thin Solid Films., 50, 145 (1978). 20. J. Nagai and T. Kamimori, Jpn. J. Appl. Phys., 22, 681 (1983). 21. C. Ho, I. D. Raistrick, and R. A. Huggins, J. Electrochem. Soc., 127, 343 (1980). 22. D. Scherson, W. Eckardt, and H. Gerischer, J. Phys. Chem., 89, 554 (1985). 23. C. G. Granqvist, Handbook of Inorganic Electrochromic Materials, pp. 310-312, Elsevier, Amsterdam (1995). 24. Q. Pei, G. Yu, C. Zhang, Y. Yang, and A. Heeger, Science, 269, 1086 (1995). 25. J-G. Zhang, C. E. Tracy, D. K. Benson, and S. K. Deb, J. Mater. Res., 8, 2649 (1993). 26. Q. Zhong, J. R. Dahn, and K. Colbow, Phys. Rev. B, 46, 2554 (1992). and Optimization, Kluwer Boston, MA (1994); Li 27. C. Julien, L. E1-Farh, M. Balkanski, 0. M. Hussain, and G. A. Nazri, Appl. Surf. Sci., 65/66, 325 (1993). Batteries, C. Pistoia, Editor; Elsevier; Oxford (1993). 11. H. Reiss, J. Phys. Chem., 89, 3783 (1985). 12. S. Hashimoto and H. Matsuoka, J. Electrochem. Soc., 138, 2403 (1991). State lonics, 86-88, 977 (1996). 29. C. G. Granqvist, Handbook of Inorganic Elect rochromic Materials, pp. 394-395, Elsevier, Amsterdam (1995). 13. S. J. Golden and B. C. H. Steele, Solid State lonics, 28-30, 1733 (1987). 14. R. B. Goidner, F. 0. Arntz, K. Dickson, M. A. Goidner, T. E. Haas, T. Y. Liu, S. Slaven, G. Wei, K. K. Wong, 28. A. Gorenstein, J. Scarminio, and A. Lourenço, Solid 30. F. Decker; B. Pileggi, S. Passerini, and B. Scrosati, J. Electrochem. Soc., 138, 3182 (1991). 31. J. Molenda, A. Stokiosa, and T. Bak, Solid State lonics, 36, 53 (1989). Downloaded on 2016-02-18 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).