KINETICS AND DYNAMICS OF ION INSERTION INTO HOST ELECTRODES FOR RECHARGEABLE Li AND Mg BATTERIES

M.D. Levi, Y. Gofer, E. Levi and D. Aurbach

Department of Chemistry, Bar-IlanUniversity, Ramat-Gan52900, Israel

Abstract

Using cyclic voltammetry and electrochemical impedance techniques, Li and Mg ions insertion into the Chevrel phase MxMo6S8 (M = Li, Mg) has been studied. The drastic difference, both in the cyclic voltammetric responses, as well as in the impedance, is ascribed to a relatively fast kinetics of Li ion insertion. This comes in contrast to very slow initial insertion kinetics (and final deinsertion) of Mg ions due to trapping effect. When the kinetics of Mg ion insertion is improved drastically, it comes with an increase in the intercalation level. The model of ion insertion, considering two non-equivalent lattice sites with two different energy barriers describes semi-quantitatively the measured impedance characteristics (semicircle at low-frequencies and Gerischer-type impedance at high frequencies). Li ion insertion is not inhibited by ion trapping and is better reproduced by a single-site model (a particular case of two-site model). Qualitative arguments referring to the unusual crystallographic structure of Chevrel phase were used to substantiate the application of the two-sites model for Mg ions insertion. Understanding the origin of multicharged ions trapping and the conditions at which this effect generates are of crucial importance in developing rechargeable Mg battery.

Introduction

The operation of high-energy density, rechargeable Li and Mg batteries is based on the phenomenon of electrochemical insertion of Li and Mg ions into appropriate host of the cathode materials (positive electrode). for anode, Li metal can be replaced by some appropriate hosts capable of Li-ion insertion at a potential not far from the equilibrium potential of Li electrode, e.g. graphite and some other carbonaceous materials. Until now, there were no reports in literature concerning Mg-ion insertion anodes, thus Mg metal remains the single anode to Mg-ion insertion cathode.

Li-ion intercalation (deintercalation) processes occurring in a rechargeable batteries can be understood by considering the following reactions. Fully lithiated (fully charged) graphite is a stoichiometric compound LixC6, which release Li-ions during charge (i.e. applying anodic current):

LixC66C + xLi+ +xe- (A)

The released Li-ions and electrons are inserted into the positive electrode host (usually, a transition metal (M) oxide or sulfide, Li1-xMX2, M = O or S) according to the following reaction:

xLi+ + xe- + Li1-xMO2→LiMO2 (B)

In case of Mg battery, Mg ions are discharged to the solution by Mg metal electrodissolution:

xMgxMg2++ 2xe- (C)

which, together with the charge-balancing electrons, are inserted into the positive electrode (e.g. Mo cluster chalkogenide, the so-called Chevrel phase,Mg2-xMo6Y8(Y=S,Se)):

xMg2+ + 2xe- + Mg2-xMo6S8 → Mg2Mo6S8 (D)

High specific capacity of Li or Mg ion insertion electrode implies both high value of x and a relatively low molecular weight of the host material. A search for a proper candidate host material should normally include good knowledge of their crystallographic features and electronic structure as well as the character of ion-electron interactions.

High-energy density of the battery necessitates, in addition to high specific capacity, that the difference in equilibrium potentials between positive and negative electrodes to be large. This potential difference should be, of course, within the limits of electrolyte solution stability (both with respect to oxidative and reductive decomposition).

One more important characteristic of a battery based on the ion-insertion electrodes, is its rate capability, i.e. the rate at which battery charge can be consumed or recharged by an external device. The rate capability of practical batteries can be substantially improved by proper design, e.g. (i) using highly porous electrode masses (e.g. porous powdery composite electrode instead of a compact one), (ii) good impregnation with the electrolyte solutions, (iii) use of low-resistance contact between the active electrode mass and current collector, and low ion-resistance membrane, separating the two electrodes, keeping them as close as possible. All these requirements become understandable, considering reactions A - D above, and keeping in mind that low-resistance paths should be ensured for both ionic and electronic fluxes from one electrode to the other. Rechargeability of the battery means that reactions A – D may occur in backward direction, i.e. from right to left.

However, in addition to the above factors i-iii, determining an improved rate capability of the battery as a whole, one should consider the dynamics of electrons and ions movements within the bulk of the electrode particles, e.g. their electronic and ionic conductivities, determining, in turn, the related mobilities and diffusion coefficients.

Recently, we reported that a reversible cathode can be realized with the Chevrel phase compound (reaction D) using simple Mg2+-containing salts in organic carbonate solutions (1). Further studies resulted in an increase of the specific capacity of the cathode owing to a proper choice for the preparation of the initial Chevrel compound (leaching procedure (2)). Furthermore, we have shown improvement of the electrode cycleability by replacing Mg(ClO4)2/propylene carbonate electrolyte solution for alkylmagnesium chlorides of the general formula Mg(AlCl(4–n)Rn)2 (with R = alkyl group) possessing a relatively high anodic stability. Their preparation and properties were already reported (3). Mg(AlCl2-EtBu)2, abbreviated here as DCC, was used as 0.25 M solution in THF. Note also that DCC/THF is the only electrolyte solution, in which Mg anode behaves reversibly (reaction C)

First studies of Mg ions insertion into Chevrel compound, when compared with that for Li ions insertion, revealed their specific features that are probably typical for the intercalation of multicharged ions in host materials: (i) The mobility of Mg ions appeared to be extremely low at ambient temperatures at the beginning of the insertion process, and at the end of the deinsertion. (ii) The ions mobility drastically increases from certain intercalation level. (iii) Estimation of the coulombic efficiency of Mg insertion revealed a considerable difference between the first and the subsequent cycles: about 15 – 20% of the ions inserted into the host in the course of the first discharge cannot be withdrawn from it at ambient temperatures at the same rate of charge. They remain in the host as trapped ions during the subsequent discharges. All these features were not observed for Li ion insertion into the same Chevrel phase electrodes, which exhibited the characteristic of highly reversible process (4). Taking into account that the host possesses high electronic conductivity, the drastic difference in the electrode behavior could be unequivocally ascribed to the difference in the ionic dynamics of the single and double-charged ions as they enter the electrode bulk.

In this paper, we summarize our early reports (4-8) concerning the origin of the different kinetics of Li and Mg ions during insertion into Chevrel phase electrode. Without proper understanding the nature of the above difference, one can hardly avoid the effect of trapping Mg ions in the host electrodes.. Special attention is paid throughout this work to the impedance characterizations of Li and Mg ion insertion systems (6-8), which can be translated into the difference of the related dynamics of these ions.

Results and Discussion

The voltammograms of Li and Mg insertion into Mo6S8 Chevrel phase (a and b, respectively) are presented in Fig. 1a and b, respectively. It is seen from this figure that whereas intercalation of Li ions occurs reversibly via three first-order phase transitions (note 3 reversible redox-peaks a/a', b/b' and c/c' in Fig. 1a), Mg ions intercalation was found to occur via two first order phase transitions (two redox-peaks a/a' and b/b' in Fig. 1b). The coulometric analysis for the first cycle shows that the initial magnesiation (and final demagnesiation) involves considerable charge-trapping at ambient temperature (25°C). During the subsequent cycling, although the amount of the trapped ions does not change, the voltammetric behavior is typical of very slow insertion kinetics. The reversibility of Mg insertion, however, drastically increases at higher intercalation levels, so that the second Mg-ions transfer (i.e. the transfer from Mg1Mo6S8 to Mg2Mo6S8) proceeds under quasi-equilibrium conditions. We earlier suggested (6-8) that the slow initial kinetics of the Mg-ion insertion is due to trapping of the divalent cations whithin the host crystal.

The structure of this phase can be presented as a three-dimensional array of Mo6S8 units, consisting of distorted Mo6-octahedra, which, in turn, are surrounded by S8 cubes (see Fig. 2) (9). Intersecting, three-dimensional channels crossing the crystal structure of the Chevrel phase make possible the transport of the guest ions to the vacant sites during intercalation. It is known that small guest ions (below 1 Å) do not occupy the exact center of the S8 – vacant site, but rather shift away from it (9). Moreover, due to a characteristic thermal motion of these ions in site, small cations are delocalized. Li ion has a radius close to 0.68 Å, and thus belong to the class of small cations. They were proven (9) to be distributed randomly within the 6 inner and 6 outer ring-like sites (see Fig. 2). The stoichiometry of the Li-insertion reaction in the chevrel phase has been previously studied by Schollhorn (10) using chronopotentiometry and 7Li NMR. Combination of these techniques allowed for a reasonable conclusion with regard to the nature of the sites for Li-accommodation and the character of their bonding in the host as a function of the intercalation level.Neutron diffraction analysis of lithiated Mo6S8, revealed (11) that there are two types of sites for small ions accommodation in this host, namely, a six-sites inner and six-sites outer rings.

Mg-ion has the radius of 0.67 Å, similar to that of Li-ion. It has completely filled 2s22p6 sublevels, rendering a small cation that exhibit high charge density since it is divalent. This high charge density of Mg ion, in combination with relatively low polarizability of the sulfide-anion framework is assumed to be the major reason for the high diffusion barriers during the first stage of Mg ions intercalation (occupation of the inner-ring site). Each of the two rings containing 6 sites is capable to host a single Mg ion (totally 2 Mg ions per a Mo6S8 unit). As intercalation level x in MgxMo6S8 increases, one should expect appearance of strong repulsion between neighboring double-charged Mg ions within each site. Due to this repulsion, the first intercalated Mg ion switches position from the inner ring to the outer. Since the diameter of the outer ring is larger, the distance between two neighboring sites in the later one is larger than that in the inner ring, thus relaxing some of the repulsion. Furthermore, the distance between two outer rings from two neighboring unit-sites is small enough to provide easy path for transport of the divalent Mg ions. Therefore, the above scenario predicts an increase in the Mg ions mobility with an increase in the intercalation level. the lower charge density of the monovalent Li ions, in comparison with that of Mg ions, makes their transition from inner to outer ring sites easier (lower energy barrier), so that Li ions can be completely removed from the inner-ring sites via transitional outer-ring sites.

Thus, the difference between the dynamics of Li and Mg ions transport in the Chevrel phase can be formulated as follows: For Mg ion insertion one should distinguish between two types of sites in the host, namely, between the shallow and the deep sites with different diffusion barriers. For Li ion insertion, the difference in site energy and the related diffusion barriers is not as large as that for Mg ion. Electrochemical impedance spectroscopy (EIS) study, combined with modeling of ion insertion into a host with the energetically non-equivalent sites would be the best diagnostic tool for revealing the difference of dynamics of both ions and their rational explanation.

Fig. 3 shows typical Nyquist plots related to Li insertion into LixMo6S8 (0 < x < 1) Chevrel phase electrodes. The spectra comprise high frequency semicircle (HFS) which relates to charge transfer (no surface films or other complicated surface phenomena in this case), and a low-frequency Warburg-type element which reflects the solid-state diffusion. Finally, at the very low frequencies, the -Z'' vs. Z' plots become steep and reflect the differential intercalation capacitance (Cdif = -1/ Z'',  →0). These spectra reflect the rapid intercalation of Li ions into this host, which occurs with no complications as charge trapping, slow diffusion or pronounced surface resistance.

At the beginning of Mg-ions intercalation (see the relevant Nyquist plots for Chevrel phase electrode at the potentials 1.4 and 1.3 V in Fig. 4) the diameter of the LFS is so large that practically only an arc can be obtained down to the lowest frequency of 5 mHz (for details see ref. (6)). While the initial Mg intercalation involves the full capacity of the host (2 Mg ions per Mo6S8 unit), the reversible intercalation level of Mo6S8 with Mg-ions at ambient temperature (25º C) is around x = 1.6 – 1.7 (4), i.e. by 15 – 20 % less than the theoretical value x = 2. At 1.25 V the host accommodates x = 0.6 – 0.7 of Mg-ions (4), so that starting from this potential only, a well-developed LFS is formed, which transforms into a purely capacitive vertical line in the limit of the lower frequencies (see Fig. 4). A characteristic feature is observed as the frequency increases: the LFS transforms rather abruptly to a line, which can be approximately described by a semi-infinite Warburg element (see Fig. 4 and ref. (6-8) for details). Thus, two characteristic frequencies (shown in Fig. 4 as solid circles), one related to the above border between the LFS and the Warburg domain, and the other (at lower frequencies) located between the LFS and the limiting capacitive line) can be experimentally distinguished.

Further decrease in potential (i.e. an increase in the intercalation level) results in a continuous decrease in the LFS diameter, while the higher frequency domain becomes more and more Gerischer-type like impedance (see Fig. 4 dotted curly brace and refs. (6-8)),. The Gerischer domain is located between the HFS and the LFS. The former semicircle relates to the interfacial charge transfer, complicated by the chemical equilibria in the DCC/THF solution and possible involvement of adsorption processes. It will be further ignored because herein we focus on the solid-state diffusion and ions entrapment). It is of great interest that as the potential approaches the value of 0.72 V, which is close to the limit of Mg-ions insertion into the Chevrel phase electrode at ambient temperature (x = 1.6 – 1.7), the high-frequency limit of the impedance does not include the semicircles anymore (neither the LFS nor the curved Gerischer-type impedance): the impedance behavior is rather similar to a semi-infinite Warburg line, transforming into a vertical capacitive line at lower frequencies (Fig. 4). Within this intercalation domain, the impedance spectra for Mg ions insertion become qualitatively similar to that for Li (compare with Fig. 3).

The above features of the impedance behavior for Mg-ions insertion into Mo6S8 Chevrel phase electrodes can be rationalized and semi-quantitatively described by a two-state model elaborated recently by Bisquert and Vikhrenko (12). Originally, this model considers two energetically different types of sites: deep (or hollow) and shallow ones. At the beginning of the intercalation, the hollow sites with limited ion mobility (because of the higher diffusion barrier) are occupied. As intercalation level increases, the shallow sites (with a lower diffusion barrier) start to fill with the guest ions. It is easy to see that such two-state model can be applied for description of the Mg ions intercalation into Mo6S8 taking into account the two types of occupation sites for the divalent ion insertion as outlined above. In this case, the overall complex plane impedance of the intercalation system containing two types of sites (Z(s)) can be conveniently presented as a function of the angular frequency () in the form (see ref. 12):

, (1)

where , and the diffusion resistance Ro (related to the motion of the intercalated ions between the shallow sites) relates to the corresponding diffusion coefficient, Do, through the intercalation capacitance of the shallow sites, Co, and the characteristic frequency, o:

. (2)

Here L is the electrode thickness. According to the two state model (12), the intercalated ions move by ordinary diffusion mechanism between the shallow sites, interacting, at the same time, with the deep sites and becoming immobilized for a lapse of time. The characteristic frequency related to ion trapping and release at the deep sites, t,, defines an effective, frequency dependent diffusion coefficient, , through the ratio of the intercalation capacitances describing the occupation of the hollow and the shallow sites, Ctrap and Co, respectively (12):

. (3)

Note also that the diffusion motion coupled with ions trapping in the deep sites can be presented by a characteristic frequency in a conventional form:

, (4)

with obtained from Eq. 3. Thus introduced should be inserted in Eq. 1 for final calculation of the electrode impedance in case of the diffusion coupled with ion trapping.

The impedance obtained with the use of Eq. 1 can be better understood in terms of the transmission line representation (see Fig. 5). Here, the two transverse branches in parallel, comprising the intercalation capacitance of the shallow sites, Co, and the intercalation capacitance of the deep sites, Ctrap, in series with the relevant kinetic parameter, namely, the trapping resistance, Rtrap. An obvious equation expresses the additive (cumulative) character of the total capacitance observed at very low frequencies:

Ctot= Co + Ctrap. (5)

Application of Eqs. 1 - 5 for the modeling of the impedance spectra of an intercalation electrode with two different energy sites require four independent parameters: o, t, C trap, andCo. One of the essential problems of the modeling is the knowledge of the dependencies of these parameters on the electrode potential (or intercalation level, x). A general property of the two states models (e.g. considered by Bisquert and Vikhrenko (12), or Chvoj et al (13)) is that the chemical diffusion coefficient (D) of the inserted ions increases sharply at a certain intercalation level related to the end of the occupation of the deep sites and the beginning of occupation of the shallow ones. We assume that the impedance spectrum measured at 1.25 V (see Fig. 4) approximately fulfills this condition. In order to reproduce the impedance spectra in the vicinity of this transition, the only parameter, which changes with potential, is believed to be the characteristic frequency of the trapping process, t. This is because Mg-ions are doubly-charged, and as a result, an increase in the intercalation level beyond the above limit promotes ion-hopping from the deep (inner ring) to the shallow site (outer ring) owing to a drastically increasing electrostatic repulsion between the Mg ions. For the purpose of a qualitative comparison of the model with the experimental curves, we assumed an augmentation in the t by more than 3 orders of magnitude as the intercalation level increases (see captions to Fig. 4). Notice that the intercalation of singly charged Li-ions into the same Chevrel phase electrode did not reveal ion trapping (see Fig. 3) and should be described by a single-energy state model.