Electrochemical and 119Sn Mössbauer study of sulfospinels as anode materials for lithium-ion batteries

Electrochemical and 119Sn Mössbauer study of sulfospinels as anode materials for lithium-ion batteries

Electrochimica Acta 46 (2000) 127– 135 www.elsevier.nl/locate/electacta Electrochemical and 119Sn Mo¨ssbauer study of sulfospinels as anode materials...

175KB Sizes 3 Downloads 35 Views

Electrochimica Acta 46 (2000) 127– 135 www.elsevier.nl/locate/electacta

Electrochemical and 119Sn Mo¨ssbauer study of sulfospinels as anode materials for lithium-ion batteries R. Dedryve`re a, J. Olivier-Fourcade a, J.C. Jumas a,*, S. Denis b, P. Lavela c, J.L. Tirado c a

Laboratoire des Agre´gats Mole´culaires et Mate´riaux Inorganiques (UMR place Euge`ne Bataillon, 34095 Montpellier b Laboratoire de Re´acti6ite´ et Chimie des Solides (UPRES-A 6007 CNRS), c Laboratorio de Quı´mica Inorga´nica, Facultad de Ciencias, Uni6ersidad 14004 Cordoba, Spain

5072 CNRS), Uni6ersite´ Montpellier II, CC15, Cedex, France 33, rue Saint-Leu, 80039 Amiens Cedex, France de Co´rdoba, A6da. San Alberto Magno s/n,

Received 22 March 2000; received in revised form 9 June 2000

Abstract 119

Sn Mo¨ssbauer spectroscopy, X-ray diffraction and electrochemical experiments were carried out to characterize the reaction mechanism of lithium with spinel phases of the Cu2S –In2S3 –SnS2 system. This analysis reveals an irreversible part of the mechanism linked to the reduction of the cations involving a spinel to rocksalt transformation followed by a structural breakdown, and a reversible part linked to the formation of lithium-metal alloys. The electrochemical investigation has shown interesting reversible capacities for these materials, and their possible use as anode materials in lithium-ion cells. © 2000 Elsevier Science Ltd. All rights reserved. Keywords: Spinel sulfides; Lithium-ion; Mo¨ssbauer spectroscopy; Anode materials; Cycling properties

1. Introduction Sulfide compounds with spinel-related structure have been widely studied for their properties as host materials for lithium insertion and their potential use as electrode in lithium-ion batteries [1–3]. Post-transition metals provide many examples of such compounds. Particularly, indium and tin sulfospinels like FeIn2S4 [4], CuM0.5Sn1.5S4 (M=Mn, Fe, Co, Ni) [5] and the cation deficient In2Sn0.5S4 [6,7] have been previously studied as host materials for lithium insertion. The oxidation/reduction potentials observed during electrochemical tests are lower than those observed for oxide spinels, and this is the reason why these materials are currently investigated for their possible application as * Corresponding author. Fax: +33-467-143304. E-mail address: [email protected] (J.C. Jumas).

negative electrodes for lithium-ion batteries [8]. Unfortunately, their practical use has been limited up to now by a low specific capacity and a high capacity fading with cycling, in contrast to carbon electrodes. A previous work has shown the interest of copper and indium containing sulfospinels to obtain low oxidation/reduction potentials [9]. In a recent paper [10], we have presented a study of a family of copper, indium and tin containing compounds, Cu0.5 + h In2.5 − 3h Sn2h S4 (05 h5 0.5), belonging to the spinel-related structure domain of the pseudo-ternary phase diagram Cu2S – In2S3 – SnS2 [11]. We have discussed the structural modifications induced by lithium insertion in those materials via a chemical route (n-butyl-lithium), and evidenced a phase transformation. Indeed, the spinel structure is described in the Fd3( m space group, and the unit cell contains 32 anions in the 32e sites (cubic close-packing arrangement), 8 cations in 8a sites (tetra-

0013-4686/00/$ - see front matter © 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 0 1 3 - 4 6 8 6 ( 0 0 ) 0 0 5 6 4 - 8

128

R. Dedry6e`re et al. / Electrochimica Acta 46 (2000) 127–135

hedrally coordinated by anions) and 16 cations in 16d sites (octahedrally coordinated by anions). The cation distribution in our compounds, e.g. in the sample III Cu0.75In1.75Sn0.5S4 (h=0.25), is (CuI6InIII 2 )8a[In12 SnIV ] {S } . We have shown from X-ray and neu4 16d 32 32e tron diffraction experiments and 119Sn Mo¨ssbauer spectroscopy, that lithium insertion induces a transformation from the spinel phase to a rocksalt-related phase according to the following mechanism. CuI ions are reduced into Cu0 and extruded from 8a tetrahedral sites as metallic copper, SnIV ions are reduced into SnII, LiI ions are inserted into the vacant octahedral 16c sites and InIII ions migrate from 8a to 16c sites [10], which can be represented by the equation: III IV (CuI6InIII 2 )8a[In12 Sn4 ]16d{S32}32e +14Li 0 III II “{LiI14InIII 2 }16c[In12 Sn4 ]16d{S32}32e +6Cu

In the present work, we are going to examine in more detail the mechanism of electrochemical lithium insertion in two selected compounds — CuInSnS4 (In/Sn= 50/50) and Cu0.73In1.82Sn0.45S4 (In/Sn=80/20) by galvanostatic and potentiostatic measurements, X-ray diffraction and 119Sn Mo¨ssbauer spectroscopy, and evaluate their performances for possible application as negative electrode in lithium-ion batteries by cycling tests under different conditions.

2. Experimental section The compounds were synthesized by solid state reaction. Stoichiometric amounts of the binary sulfides — Cu2S, In2S3 and SnS2 — were mixed and sealed in silica tubes under vacuum at ca. 10 − 3 Pa. The mixture was slowly heated at 1°C/min to 400°C and held for 10 h at this temperature, then kept for 10 h at 750°C, and finally for 5 days at 850°C. The samples were then slowly cooled to room temperature. Two different compositions were prepared, making the In/Sn ratio vary — CuInSnS4 (In/Sn=50/50) and Cu0.73In1.82Sn0.45S4 (In/Sn=80/20). Electrochemical lithium insertion was carried out in Swagelok™ test cells, using sulfospinel as active positive electrode material and metallic lithium as negative electrode, separated by a glass fiber separator (Whatman Glass Microfiber) soaked in a 1 M LiPF6 in EC:DMC (1:1) electrolyte solution. The positive electrode was elaborated with 6–8 mg of active material in two different ways. First (polymer technology) by preparing plastic pellets with 46% of active material, 9% carbon black, 28% DBP (dibutylphtalate) plasticizer and 17% PVDF-HFP Kynar FLEX 2801 binder (Elf Atochem), according to the Bellcore’s Plion™ technology, described elsewhere [12]. Second (powder technology) by pressing at ca. 4 tons/cm2 over a stainless steel grid a powder mixture of 80% of active material, 10%

carbon black and 10% PTFE binder. Electrochemical measurements were carried out using a multi-channel ‘Mac Pile’ system (Biologic, France) operating in galvanostatic and potentiostatic modes. Step potential electrochemical spectroscopy (SPES) curves were obtained on samples prepared as powder pellets under two different kinetic conditions — 10 mV/1 h and 10 mV/0.1 h — after allowing an initial relaxation during 10 h. Galvanostatic measurements were carried out at a rate of C/2 and C/4 for the cells prepared with the polymer technology, and at C/4 for those prepared with the powder technology, where C is defined as the insertion of one mole of lithium per mole of compound and per hour. For samples destined for X-ray diffraction and 119Sn Mo¨ssbauer characterization, the positive electrode was prepared with ca. 40 mg of active material and a slower rate of C/20 was applied (polymer technology). X-ray powder diffraction (XRD) characterization was performed on a Philips q –2q diffractometer using Cu Ka radiation and a nickel filter. Rietveld analyses of XRD patterns were carried out with the aid of the program ‘Rietveld Analysis Program DBWS-9411’ [13]. Pristine and lithiated samples were studied by 119Sn Mo¨ssbauer spectroscopy using a conventional EG&G constant-acceleration spectrometer. The g-ray source was 119mSn in a BaSnO3 matrix, working at room temperature. The velocity scale was calibrated from the magnetic sextet of a high-purity a-Fe foil absorber using 57Co(Rh) as source. Recorded spectra were fitted with Lorentzian profiles by the least-squares method using the program ISO [14], and the fit quality was controlled by the  2 test. All isomer shifts are given relative to 119Sn in BaSnO3.

3. Results and discussion The first cycle of discharge/charge of CuInSnS4 and Cu0.73In1.82Sn0.45S4 (prepared with polymer technology) is shown in Fig. 1a, and the derivative −dx/dV is represented in Fig. 1b. In this first cycle one can notice a difference of behavior between both the samples. For the compound Cu0.73In1.82Sn0.45S4, we can observe two plateaus — a first one at 1.6 V and a second one at 1.25 V. For the compound CuInSnS4, the first plateau at 1.5 V is less obvious, and the second one can be seen at 1.3 V. For Cu0.73In1.82Sn0.45S4, the first plateau can be attributed to the spinel to rocksalt transformation that takes place at the beginning of the insertion, associated to the reduction of CuI into Cu0 and SnIV into SnII, like in the case of chemical lithium insertion [10]. These two simultaneous processes can be described by the following equations:

R. Dedry6e`re et al. / Electrochimica Acta 46 (2000) 127–135

CuI0.73In1.82SnIV 0.45S4 +0.73Li 0 “ LiI0.73In1.82SnIV 0.45S4 +0.73Cu I 0.73

Li

IV 0.45 4

In1.82Sn

(45 mAh/g)

I 1.63

S +0.9Li“ Li

In1.82SnII 0.45S4

(55 mAh/g) Global: CuI0.73In1.82SnIV 0.45S4 +1.63Li 0 “ LiI1.63In1.82SnII 0.45S4 +0.73Cu

(100 mAh/g)

Thus, the end of the first plateau observed at 92 mAh/g (or 1.5 Li/mol) in Fig. 1 is close to the theoretical value 100 mAh/g (or 1.63 Li/mol). The reason why the first plateau of the compound CuInSnS4 is less pronounced results from the difference of In/Sn ratio between both the samples. Indeed, to realize the spinel to rocksalt transformation in CuInSnS4, whose cation IV distribution is (CuI8)8a[InIII 8 Sn8 ]16d{S32}32e, 24 lithium inserted are necessary to reduce 8 CuI into Cu0 and 8 SnIV into SnII, while only 16 octahedral 16c sites are available to accept lithium ions, that makes this phase transformation difficult. Thus, the first plateau observed for CuInSnS4 is less distinct and ends at a

Fig. 1. (a) First cycle of discharge/charge of both compounds CuInSnS4 and Cu0.73In1.82Sn0.45S4; (b) derivative, − dx/dV (polymer technology, C/4 rate).

129

greater lithium insertion value. This phenomenon can be seen in Fig. 2, where XRD measurements of both the compounds have been represented. Between 0 and 3 Li inserted, the replacement of the spinel phase by a good crystallized rocksalt-related phase is clear for Cu0.73In1.82Sn0.45S4 (Fig. 2b), while it is associated to an important amorphization for CuInSnS4 (Fig. 2a). The increase of the cell parameter between the pristine and the lithiated phase induces a slight shift of the peak positions to small angles (the cell parameter is a=10.60 A, for the pristine phase and a= 10.76 A, for the rocksalt-related phase in the compound Cu0.73In1.82Sn0.45S4). The theoretical end of the second plateau of Fig. 1 corresponds to the total reduction of the cations after 8.0 Li/mol inserted. Thus, the value observed for Cu0.73In1.82Sn0.45S4 at 8.2 Li/mol is in good correlation with the theoretical value. For the compound CuInSnS4, the end of the second plateau is also observed at ca. 8 Li/mol. The corresponding theoretical capacities are 504 and 490 mAh/g, respectively, for CuInSnS4 and Cu0.73In1.82Sn0.45S4, which are in good agreement with the capacity values observed in the experimental discharge curves in Fig. 1a. This reduction of all the cations is associated to a structural breakdown and an amorphization that can be observed on XRD patterns (Fig. 2). After 7 Li was inserted in Cu0.73In1.82Sn0.45S4, the rocksalt-related phase almost disappeared (Fig. 2b), and after 5 Li was inserted in CuInSnS4, the material was already amorphous (Fig. 2a). The total discharge down to 0.02 V gives capacities of ca. 960 mAh/g. If we model the low potential insertion process with the formation of Li– Sn and Li– In alloys, the formation of the lithium-rich phase Li22Sn5 corresponds to capacities of 277 and 121 mAh/ g for CuInSnS4 and Cu0.73In1.82Sn0.45S4, respectively, and the formation of the lithium-rich phase Li13In3 corresponds to capacities of 273 and 484 mAh/g, respectively, to give 550 and 605 mAh/g for this alloying process, and 1050 and 1095 mAh/g at the end of the discharge. These values are close to those observed in Fig. 1a, and show that both In and Sn are involved in this alloying process. In order to complete the study of electrochemical lithium insertion in these materials by XRD and 119Sn Mo¨ssbauer spectroscopy, a series of nine samples of each compound was prepared with, respectively, 0, 1, 3, 5, 7, 9, 11, 15 lithium inserted and a recharge up to 2.0 V (Fig. 3). The differences observed in the discharge curves in comparison to Fig. 1 are due to the great quantity of active material used to prepare the electrodes, and to the slower rate (C/20). Fig. 3 includes the 119 Sn Mo¨ssbauer spectra of these samples, in which several components can be observed. The corresponding refined hyperfine parameters are given in Tables 1 and 2. The results up to 7 Li are very similar to those

130

R. Dedry6e`re et al. / Electrochimica Acta 46 (2000) 127–135

Fig. 2. X-ray diffraction patterns of (a) CuInSnS4; (b) Cu0.73In1.82Sn0.45S4 at different lithium contents per mol of pristine compound (uCu Ka = 1.5418 A, ) (polymer technology, C/20 rate).

obtained with chemically inserted samples. The first component at ca. 1.1 mm/s corresponds to SnIV in octahedral coordination, and its contribution decreases as the lithium content increases, indicating tin reduction during the lithium insertion process. The second signal appears at ca. 3.7–3.8 mm/s, and corresponds to SnII in octahedral coordination. Its contribution increases at the beginning of the insertion, and decreases with further lithium insertion. This contribution is attributed to the formation of the rocksalt-related phase, in which tin atoms are at the oxidation state ( +II). This observation is in good agreement with the XRD results showing the progressive replacement of the spinel phase by the rocksalt-related phase. Thus, we can notice that in spectrum C (3 Li) of the compound Cu0.73In1.82Sn0.45S4, the SnIV signal has almost disappeared because the spinel phase has been replaced by the rocksalt-related phase (Fig. 2b), while in spectrum C of CuInSnS4, signals of SnIV and SnII can simultaneously be seen because both the phases coexist. The subspectra located in the interval 1.9–2.3 mm/s, that can be seen after 3 Li was inserted, can be attributed to tin atoms involved in inter-metallic bonds and/or clusters. The isomer shift values are included between those of a-Sn and b-Sn [15]. The apparition of these Mo¨ssbauer contributions and of another SnII signal at ca. 3.2 mm/s up to 9 Li correlates with the amorphization of the material that can be seen in Fig. 2. But the presence of metallic tin could not be detected by XRD, probably due to a result of the dispersion of small particles in the matrix. Thus, lithium insertion

results in a reduction of SnIV into SnII, and of SnII into Sn0, with formation of an amorphous phase containing Sn0. Discharge at lower potentials is expected to lead to the formation of Li– In and Li– Sn alloys, or mixed Li– In – Sn alloys, and from that point of view it is interesting to compare the spectra obtained for 11 and 15 Li inserted (spectra G and H) with those of Lix Sn alloys that form at potentials below 0.7 V, like Li7Sn3, Li5Sn2, Li13Sn5 or Li7Sn2 [16]. These alloys show the same isomer shift values between 1.9 and 2.2 mm/s [17,18], but in our case quadrupole splitting values up to 1.5– 1.6 mm/s were observed instead of 1.2– 1.3 mm/s at the most for Lix Sn alloys. This reflects the existence of interactions with sulfur present in the amorphous matrix, certainly due to the small size of the particles of alloys that can not be detected by XRD. During the following recharge up to 2.0 V, 6.7 Li was removed from CuInSnS4 (down to 8.3 Li) and 4.9 Li from Cu0.73In1.82Sn0.45S4 (down to 10.1 Li). The comparison between the spectra of discharge and recharge displays a reversible process and an irreversible process during the first cycle: spectrum I (8.3 Li) of CuInSnS4 at the charge is almost identical to spectrum E (7 Li) at the discharge, and spectrum I (10.1 Li) of Cu0.73In1.82Sn0.45S4 is almost identical to spectrum F (9 Li), that shows after the first charge a recovery of the spectra of the end of the 1.2– 1.3 V plateau just before the beginning of the alloy formation, which is the reversible part of the process. However, the applied cell potentials are completely different, and after the charge the materials remain amorphous, that shows that crys-

R. Dedry6e`re et al. / Electrochimica Acta 46 (2000) 127–135

tallized phases (spinel or rocksalt) can not be recovered after the structural breakdown, which constitutes the irreversible part of the process responsible for the loss of capacity of the first cycle. Step potential electrochemical spectroscopy (SPES) experiments were carried out for a more clear detection of the occurrence of multiphase processes during the redox reaction. For the slower rate (10 mV/1 h), the intensity–voltage curves are displayed in Fig. 4a and b for CuInSnS4 and Cu0.73In1.82Sn0.45S4, respectively. The first peak A (1.4 and 1.6 V, respectively) can be attributed to reduction of CuI into Cu0 and SnIV into SnII, as seen before. The second peak B (1.16 and 1.06

131

V, respectively) can be attributed to reduction of SnII into Sn0 and InIII into In0. In this way, potentiostatic measurements on samples prepared as powder pellets confirm the results obtained by galvanostatic experiments carried out on samples prepared as polymer pellets. The difference of potential between the two first peaks, A and B, implies a different energy for the two processes that may arise from a change in the lithium interaction. The first process corresponds to the occupation of geometrical empty sites by lithium ions, while the second one corresponds to a structural breakdown due to the impossibility to accommodate lithium ions in the structure. This behavior can be compared with that

Fig. 3. 119Sn Mo¨ssbauer spectra at room temperature of (a) CuInSnS4; (b) Cu0.73In1.82Sn0.45S4, at different lithium contents shown on the voltage vs. composition curve (polymer technology, C/20 rate).

R. Dedry6e`re et al. / Electrochimica Acta 46 (2000) 127–135

132

Table 1 Hyperfine parameters of refined Mo¨ssbauer spectra of CuInSnS4 before and after lithium insertion: isomer shift (l) relative to BaSnO3, quadrupole splitting (D), full-width at half-maximum (Y) and relative area (R.A.) Sample

A

B

C

D

E

F

G

H

I

Li/mol

0

1

3

5

7

9

11

15

8.3

SnIV

Sn0(1)

Sn0(2)

SnII(1)

SnII(2)

l (mm/s) D (mm/s) R.A. (%) l (mm/s) D (mm/s) R.A. (%) l (mm/s) D (mm/s) R.A. (%) l (mm/s) D (mm/s) R.A. (%) l (mm/s) D (mm/s) R.A. (%)

Y (mm/s)a Absorption (%) a b

1.12(1) 0.23(1) 100 – – 0 – – 0 – – 0 – – 0 0.86(1) 15

1.11(1) 0.29(7) 81 – – 0 2.28(3) 0.91(4) 6.3 – – 0 3.64(1) 0.44(2) 13 0.83(4) 15

1.12(1) 0.29(1) 26 2.16(6) 0.35(6) 14 2.13(6) 1.21(1) 14 3.11(8) 0.3(1) 10 3.74(1) 0.30(1) 35 0.84(1) 5.6

1.12(1) 0b 7.3 2.06(3) 0.43(4) 30 2.06(3) 1.13(5) 32 3.20(4) 0.28(6) 15 3.75(3) 0.12(6) 15 0.83(1) 5.0

– – 0 2.09(2) 0.54(2) 63 2.16(4) 1.40(5) 27 3.33(7) 0.31(1) 10 – – 0 0.88(2) 5.6

– – 0 2.20(2) 0.51(3) 54 2.28(2) 1.26(4) 40 3.4(2) 0.2(2) 6.7 – – 0 0.87(2) 4.8

– – 0 2.16(1) 0.85(2) 66 2.12(2) 1.61(3) 34 – – 0 – – 0 0.90(2) 6.0

– – 0 1.90(1) 0.65(2) 49 1.98(1) 1.44(2) 51 – – 0 – – 0 0.95(2) 6.1

– – 0 2.07(3) 0.47(3) 62 2.07(8) 1.2(1) 26 3.26(9) 0.4(1) 12 – – 0 0.81(1) 6.7

Constrained to be equal for all components. Fixed value.

Table 2 Hyperfine parameters of refined Mo¨ssbauer spectra of Cu0.73In1.82Sn0.45S4 before and after lithium insertion: isomer shift (l) relative to BaSnO3, quadrupole splitting (D), full-width at half-maximum (Y) and relative area (R.A.) Sample

A

B

C

D

E

F

G

H

I

Li/mol

0

1

3

5

7

9

11

15

10.1

SnIV

Sn0(1)

Sn0(2)

SnII(1)

SnII(2)

l (mm/s) D (mm/s) R.A. (%) l (mm/s) D (mm/s) R.A. (%) l (mm/s) D (mm/s) R.A. (%) l (mm/s) D (mm/s) R.A. (%) l (mm/s) D (mm/s) R.A. (%)

Y (mm/s)a Absorption (%) a b

1.14(1) 0.29(1) 100 – – 0 – – 0 – – 0 – – 0 0.80(1) 12

1.13(1) 0.26(2) 59 – – 0 – – 0 – – 0 3.72(1) 0.40(1) 41 0.83(1) 9.5

1.11(1) 0b 4.5 2.0(1) 0.5(1) 12 2.1(2) 0.8(3) 9 3.18(4) 0.29(7) 11 3.80(1) 0.30(1) 64 0.79(1) 8.3

– – 0 2.02(1) 0.52(2) 36 1.99(5) 1.32(9) 11 3.20(1) 0.40(1) 10 3.78(1) 0.24(2) 43 0.82(1) 5.0

– – 0 2.08(4) 0.34(8) 36 2.05(4) 0.94(7) 35 3.22(6) 0.5(1) 9.6 3.79(3) 0.21(6) 19 0.87(1) 4.0

– – 0 2.15(1) 0.53(2) 71 2.26(4) 1.30(6) 26 3.66(9) 0b 2.9 – – 0 0.90(3) 4.0

– – 0 2.24(2) 0.62(3) 68 2.30(3) 1.41(6) 32 – – 0 – – 0 0.84(3) 2.4

– – 0 1.99(2) 0.84(3) 52 2.00(2) 1.54(4) 48 – – 0 – – 0 0.87(2) 3.9

– – 0 2.11(2) 0.48(4) 64 2.08(6) 1.1(1) 27 3.43(7) 0.3(1) 9.1 – – 0 0.85(4) 4.3

Constrained to be equal for all components. Fixed values.

previously described for tin oxide, where the first discharge implies a structural breakdown accompanied by lithium oxide formation [19,20]. In the same way, the

electrochemical behavior of SnS2 in the near zero voltage range has been studied previously [21]. In this work, two overlapped broad bands appear at about 1.0

R. Dedry6e`re et al. / Electrochimica Acta 46 (2000) 127–135

V. The authors explain that these signals are due to metallic tin and Li2S formation. This effect is known to be favorable to tin dispersion in the matrix that hinders aggregation phenomena of active particles and facilitates further cell cycling. For cell potentials below 1.0 V, a set of weak bands are visible in the voltammograms of Fig. 4. These features of the curve are usually attributed to lithiummetal alloy formation, but their low intensities make difficult their assignment to either tin or indium alloys. Moreover, the studies on these alloys state the presence of several phases with different lithium compositions [22]. Thus, this series of weak bands observed below 1.0 V corresponds to the formation of several successive lithium-metal alloys, but some bands certainly overlap. The intensity relaxation curves associated with the potentiostatic discharges (not represented here) showed irregular shapes in the metal reduction potential region over 1.0 V, not fitting to common diffusion algorithm. This behavior can be ascribed to the presence of multiphase systems generated during the lithium insertion and the structural deterioration of the framework. In contrast, lithium-metal alloying processes below 1.0 V originated relaxation curves regularly shaped and short trailing currents. It displays a more homogeneous diffusion of lithium than in the previous stages.

133

The potentiostatic cycling of both samples was carried out up to five cycles at a higher rate (10 mV/0.1 h). The cell potential upper limit was fixed to 2.5 V in order to check a wide voltage region. Fig. 4c and d show the intensity– voltage curves for CuInSnS4 and Cu0.73In1.82Sn0.45S4. As expected, the first discharge is different from the successive cycles and an overlapping of the signals over 1.0 V is observed for both the spinels. For the compound CuInSnS4, the similarity of the curve shape from the second to the fifth cycle shows a good reversibility in the whole range. At this kinetic condition, the signals corresponding to each intermediate lithium alloy are not clearly separated, but their reversibility is evident by the occurrence of a broad oxidation signal at ca. 0.5 V (peak C). This observation is confirmed by the 119Sn Mo¨ssbauer results showing a recovery of the spectra of the end of the 1.2– 1.3 V plateau after the charge (spectra I), as seen before. This way, the formation of lithium-metal alloys during the discharge plays an important part in the reversibility of the system. In contrast to the results presented by Brousse et al. for SnS2 [21], where voltammograms for several cycles clearly showed no reversible reaction for high potential bands, it seems here that a partially reversible process

Fig. 4. Intensity and composition vs. voltage curves of CuInSnS4 and Cu0.73In1.82Sn0.45S4 obtained by SPES at different kinetic conditions, (a and b) 10 mV/1 h; (c and d) 10 mV/0.1 h (powder technology).

134

R. Dedry6e`re et al. / Electrochimica Acta 46 (2000) 127–135

Fig. 5. Cycling behavior of (a) CuInSnS4; (b) Cu0.73In1.82Sn0.45S4 prepared with either polymer or powder technology and used as positive electrodes in Swagelok test cells with different cell potential windows (charge and discharge capacity).

takes place above 1.0 V. This process leads to a broad and complex band at ca. 1.2 V (peak D) during the discharge and at ca. 1.8 V (peak E) during the charge. XRD measurements did not allow assigning this reversible process to a recovery of the original crystallized structure during the charge (up to 2.0 V). This process could be originated by side reactions that could yield non-crystalline products, like the oxidation of Li2S to form polysulfides Li2Sx similar to the processes observed in the sodium–sulfur battery [23], or other side reactions to re-form amorphous InIII and SnII sulfides, like e.g. in a simplified form Li2S+LiSn“ SnS+3Li+ +3e−, that could be in agreement with the presence of a SnII signal in the Mo¨ssbauer spectra after the charge (spectra I). Several authors have proposed the occurrence of aggregation phenomena of tin alloys particles along successive cycles [24]. It implies a drastic volume change, which causes electrode pulverization. The presence of this effect is detected by a progressive sharpening of the signals corresponding to alloy formation that hints at crystallization of particles. But in our experi-

ments, aggregation effects are not clearly observed, probably as a consequence of the good dispersion of metal alloy particles in the matrix. In the compound Cu0.73In1.82Sn0.45S4, similar characteristics were observed. However, an additional signal at about 0.4– 0.5 V can be observed (peak F). The assignment of this weak band may be explained by the different compositions between samples, because the higher ratio In/Sn for the compound Cu0.73In1.82Sn0.45S4 agrees with the occurrence of a Liy In phase (0.86B yB 1.2) at 0.495 V [22]. On the contrary, a progressive modification of the higher voltage signals of both charge and discharge curves is detected up to five cycles, that may affect the lithium extraction. In this way, the intensity– voltage curves (Fig. 4c and d) show a better capacity retention for CuInSnS4 than for Cu0.73In1.82Sn0.45S4. In order to complete the electrochemical behavior of these samples as electrode materials, galvanostatic cycling tests were carried out using these compounds as positive electrode and metallic lithium as negative electrode (Fig. 5). For the samples prepared with polymer technology, a rate of C/2 and a cell potential window of 0.02– 2.5 V were chosen. For the samples prepared with powder technology, a rate of C/4 and two different cell potential windows, 0.03– 1.5 and 0.03– 0.9 V, were chosen. The choice of the lower limit intends to avoid undesirable lithium electroplating on cathode surface. The choice of the upper potential limit plays a crucial role in cell capacity and capacity retention. Thus, for samples cycled below 2.0 V (powder samples), a moderate capacity is observed, although the capacity retention is better as the potential window is narrower, particularly for Cu0.73In1.82Sn0.45S4 (Fig. 5b). For samples cycled with a 2.5 V upper limit (polymer preparations), the performances are much better. The capacity decreases rapidly but remains equal to 400 mAh/g after 20 cycles, even with a higher rate of C/2 (instead of C/4). A similar behavior was observed in Cu2CoSn3S8 spinel cathodes, in which cell capacity in cycle 10 increased from ca. 100 to ca. 400 mAh/g on opening the potential window from 0 – 1 to 0 – 2.3 V [8]. Besides the potential window, the way of preparing the samples may also be a relevant factor affecting the capacity retention. The main advantage of the polymer preparation is to allow a better firmness of the electrode material during cycling.

4. Conclusion The study of the reaction mechanism of lithium with CuInSnS4 and Cu0.73In1.82Sn0.45S4 by 119Sn Mo¨ssbauer spectroscopy, X-ray diffraction and electrochemical experiments has shown the occurrence of a multi-phase reaction. Lithium insertion first induces a spinel to

R. Dedry6e`re et al. / Electrochimica Acta 46 (2000) 127–135

rocksalt transformation associated with the reduction of CuI into Cu0 and SnIV into SnII, and then a structural breakdown associated with the reduction of InIII into In0 and SnII into Sn0, that makes the material amorphous, and which constitutes the irreversible part of the lithium insertion reaction. At low potentials, the formation of lithium-metal alloys dispersed in the amorphous matrix constitutes the reversible part of this process. The existence of strong interactions between alloys particles and sulfur present in the matrix is certainly at the origin of this reversibility. The capacity retention of these materials when used in spinel/Li Swagelok cells depends markedly on the upper voltage limit. The preparation of the electrode as a polymer with a 2.5 V upper potential limit allows specific capacities of 400 mAh/g after 20 cycles. This good result, associated with the low reduction/oxidation potentials of these materials, makes them possible candidates as negative electrode in lithium-ion batteries.

Acknowledgements R.D., J.O.F. and J.C.J. are indebted to CNRS (PICS nr. 505) and P.L. and J.L.T. are indebted to CICYT (contract no. MAT99-0741) for financial support.

References [1] M. Eisenberg, J. Electrochem. Soc. 127 (1980) 2382. [2] A.C.W.P. James, B. Ellis, J.B. Goodenough, Solid State Ionics 27 (1988) 37, 45. [3] A.C.W.P. James, J.B. Goodenough, N.J. Clayden, J. Solid State Chem. 77 (1988) 356. [4] C. Pe´rez Vicente, C. Bousquet, A. Kra¨mer, J.L. Tirado, J. Olivier-Fourcade, J.C. Jumas, J. Solid State Chem. 138 (1998) 193.

.

135

[5] C. Branci, J. Sarradin, J. Olivier-Fourcade, J.C. Jumas, Chem. Mater. 11 (1999) 2846. [6] M.L. Elidrissi Moubtassim, C. Bousquet, J. Olivier-Fourcade, J.C. Jumas, J.L. Tirado, Chem. Mater. 10 (4) (1998) 968. [7] M.L. Elidrissi Moubtassim, J. Olivier-Fourcade, J.C. Jumas, J. Senegas, J. Solid State Chem. 87 (1990) 1. [8] P. Lavela, C. Pe´rez-Vicente, J.L. Tirado, C. Branci, J. Olivier-Fourcade, J.C. Jumas, Chem. Mater. 11 (1999) 2687. [9] J. Morales, J.L. Tirado, M.L. Elidrissi Moubtassim, J. Olivier-Fourcade, J.C. Jumas, J. Alloys Comp. 217 (1995) 176. [10] R. Dedryve`re, S. Denis, C. Pe´rez Vicente, J. Olivier-Fourcade, J.C. Jumas, Chem. Mater. 12 (2000) 1439. [11] T. Ohachi, B.R. Pamplin, J. Crystal Growth 42 (1977) 598. [12] S. Denis, E. Baudrin, M. Touboul, J.-M. Tarascon, J. Electrochem. Soc. 144 (12) (1997) 4099. [13] R.A. Young, A. Sakthivel, T.S. Moss, C.O. Paiva-Santos, J. Appl. Crystallogr. 28 (1995) 366. [14] W. Ku¨ndig, Nucl. Instrum. Methods 75 (1979) 336. [15] A. Svane, E. Antoncik, Phys. Rev. B 35 (1987) 4611. [16] C.J. Wen, R.A. Huggins, J. Electrochem. Soc. 128 (6) (1981) 1181. [17] J. Chouvin, J. Olivier-Fourcade, J.C. Jumas, B. Simon, O. Godiveau, Chem. Phys. Lett. 308 (1999) 413. [18] R.A. Dunlap, D.E. Small, D.D. MacNeil, M.N. Obrovac, J.R. Dahn, J. Alloys Comp. 289 (1999) 135. [19] I.A. Courtney, J.R. Dahn, J. Electrochem. Soc. 144 (1997) 2045. [20] J. Chouvin, C. Branci, J. Sarradin, J. Olivier-Fourcade, J.C. Jumas, B. Simon, P. Biensan, J. Power Sources 81 – 82 (1999) 277. [21] T. Brousse, S.M. Lee, L. Pasquereau, D. Defives, D.M. Schleich, Solid State Ionics 113– 115 (1998) 51. [22] R.A. Huggins, Electrochem. Soc. Proc. 97-18 (1997) 1. [23] H. Kawamoto, J. Electrochem. Soc. 136 (1851) 1989. [24] I.A. Courtney, W.R. McKinnon, J.R. Dahn, J. Electrochem. Soc. 146 (1999) 59.