Structure and electrochemistry of LixWO2

Solid State Ionics 45 ( 199 1) 67-75 North-Holland

Structure AK

and electrochemistry

Sleigh and W.R. McKinnon

Solid State Chemistry, Received

6 August

Division of Chemistry,

1990; accepted

of Li,W02

’ National Research Council of Canada, Ottawa, Canada KIA OR9

for publication

1 October

1990

Using in situ X-ray diffraction and electrochemical techniques the structure and phase diagram of Li,WO, for 0
1. Introduction WOZ is one of the many rutile-related transition metal oxides which can be intercalated with lithium [ 11. Because the voltage at which lithium intercalates is low (below 1.2 V versus Li metal), Li,W02 is a possible replacement for Li metal anodes. The feasibility of Li,W02 anodes has been demonstrated with cathodes of LiCoO, and TiS2 [ 2,3 1. But when cycled over the full range of x (0
should be addressed.

these two oxides are almost indistinguishable below 20=45”. In both structures the rutile chains surround tunnels into which the lithium atoms fit. The structure and electrochemistry of Li,MoO, have been studied in detail [ 6 1. This material too does not cycle well between x=0.5 and x= 1.0. Through this range of x the structure changes from orthorhombic to monoclinic, although it is not known whether this change is related to the poor cycling. Other than this change in symmetry, the main effect of the intercalated lithium is to expand the structure in the directions normal to the rutile chains. Like LiMoO,, LiWOz has monoclinic symmetry, and again lithium insertion causes expansion normal to the rutile chains

111. We have used electrochemical and in situ X-ray diffraction techniques to study Li,W02. We discuss our experimental methods in section 2 and give the lattice parameters in section 3. In section 4 we describe the cycling characteristics of Li,W02 cells. We discuss the results in section 5.

2. Experimental The WOz used for most of this work was prepared by reacting stoichiometric quantities of W03 and W in a sealed quartz tube at 1050 oC for 75 h. The monoclinic cell parameters of this powder (table 1) agree

l/$03.50 0 199 1 - Elsevier Science Publishers B.V. ( North-Holland

)

68

A.K. Sleigh, W.R. McKinnon /Structure and electrochemistry of Li, W02

with previous values ( [ 51 and [ 71); the largest discrepancy is an increase of 0.0 1 A in the u parameter. This material, which was used for all the results reported here, was contaminated by a few percent of W03. We also made WOz by reducing W03 in hydrogen. This was done in a commercial thermogravimetric analysis (TGA) system, by heating 40 mg of WOx in a mixture of 5% H2 in argon at a rate of 2’ /min to 800°C. No W03 could be seen in this material by X-ray diffraction or in electrochemical cells, but it otherwise behaved the same as the other WOz. Electrodes for the electrochemical cells were made from 90% WOz and 10% carbon. The mixture was ground in propylene glycol, spread on a precleaned Ni sheet, and dried. Cells [ 81 of Li/Li, WOz were assembled and cycled at constant current to measure V(x) [ 91. For the X-ray diffraction work, a WOz slurry with 15% C was spread onto a Be sheet to make one electrode of the cell, the other being a sheet of Li as usual. The cell was mounted in a powder diffractometer, the Be window allowing X-ray diffraction patterns to be measured (with Cu radiation) as the cell was charged or discharged. We measured complete diffraction patterns (8” <28< 100’) for WOz, Li0,5W02 and Li,,0W02 (the latter two cells were discharged and then held at constant voltage). We then scanned three regions of 20 continuously while the cell charged or discharged at constant current. Initially these regions were 23-27”, 33-40” and 52.4-68”, but later the upper range was reduced to 54.4-60.8’. (It took 0.7 1 h to do one set of scans over these three regions with the reduced upper range.) The lattice parameters were calculated from least-squares refinements of peaks in these three regions. Scans were done beginning or ending with the cell in one of three states: fully charged, fully discharged, or halfway through a charge or discharge. The values of x in these three states were assumed to be 0, 1, and 0.5, respectively. Values of x at intermediate points were calculated from the elapsed time and current, then scaled to match the beginning and end points. The lattice parameters obtained from a cell discharging or charging at constant current are actually an average over the range of x (0.04) covered during a scan over the three regions. In addition to this uncertainty in x, there is an added uncertainty in determining exactly where a two-phase region begins

and ends, because when one phase is present in only small quantities it is hard to see the peaks. Thus the position of the phase boundaries quoted below are uncertain to about Ax= 0.05.

3. Lattice parameters for Li,W02 Curves of voltage IJ’against x for cells cycled betweenO
1 on discharge

Between 0
I

1.6-

-

I

I

1st charge

-

2nd discharge

0

0.2

0.4

Apparent

0.6

0.8

1

x in Li,wo,

Fig. 1. Voltage versus x in Li,W02 for two cells cycled at 20 h rates. The upper curve has been offset by +0.3 V for clarity.

A.K. Sleigh, W.R. McKinnon /Structure and electrochemistry ofLix WO,

larger region of two-phase coexistence, where the M, phase coexists with a third monoclinic phase, M,. The M, phase extends over a narrow range of x near x= 1.0. Fig. 2 shows the (0, 1, 1) Bragg peak as the cell is discharged. The small peak at 26” is due to remnant particles of disconnected WOz which do not intercalate lithium. The Bragg peaks clearly shift during intercalation. Fig. 3 and table 1 give the monoclinic lattice parameters determined by least-squares refinement. For structures near x=0 several of the Bragg peaks overlap, so only 10 peaks could be used in the refinement. For larger x not in a two phase region, a good fit was often obtained using 15 peaks. The a parameter decreases slightly as Li is added, corresponding to a slight contraction along the W chains, but the lattice expands normal to the chains in both the b and c sin j? directions. The volume of Li,W02 increases by 5.6% between x=0 and x=0.5, similar to the increase observed for Li,MoOz for the same region. The monoclinic angle changes by about 0.1’. Through the larger two-phase region between x=0.56 and x=0.98, the volume increases by about the same amount, 6.3%, but most of the increase is due to c sin p; the b parameter changes much less. (This causes several pairs of peaks to switch relative positions and hindered our original refinement of the lattice parameters.)

0

0.2

69

0.4 x

I. 0.6

1. 0.2

1

in Li,WO,

Fig. 3. (a) V(x); (b) unit cell volume; (c) b, (solid circles) and c, sin(P) and (d) a,,, versus x in Li,WOz. The shaded areas indicate the 2-phase regions.

3.2. O
0.02 0 23

I

I

24

25

Scattering

26

27

Angle / deg

Fig. 2. The (0, 1, 1 ) Bragg peak of Li,WO, measured on discharge at the values of x shown in the figure. The small peak remaining at 26” is the residual (0, 1, 1) WO, peak.

on charge

The structural changes proceed differently on charge than on discharge over the range 0 X> 0.88 the entire compound is a single monoclinic phase (M,). But as x is reduced further, some parts of the cathode behave differently than others. In some particles or regions the M, phase coexists with and then replaces the M, phase. In other parts, however, an orthorhombic phase appears, which we label as 0, because it appears near x=0.5 (as M, does). As the cell is charged past x=0.5, the parts of the

70

A.K. Sleigh, W.R. McKinnon /Structure and electrochemistry of L&W02

Table 1 Lattice parameters

of Li,WO,

Monoclinic

on discharge.

phase X

Voltage ( V) 1.400 1.210 1.093 1.030 0.970 0.904 0.861 0.817 0.801 0.797 0.782 0.729 0.750 0.659 0.602 0.557 0.432 0.376 0.400

0.000 0.020 0.060 0.100 0.140 0.180 0.220 0.260 0.300 0.340 0.380 0.460 0.476 0.500 0.524 0.571 0.952 1.ooo 1.ooo

8

aM

h.,

CM

Volume

(A)

(A)

(A)

(AX)

5.5734 5.5716 5.5673 5.5636 5.5575 5.5476 5.5459 5.5412 5.5398 5.5373 5.5348 5.5239 5.5303 5.5265 5.5261 5.5258 5.5271 5.5266 5.5342

4.8961 4.9023 4.9163 4.9343 4.9485 4.9573 4.9732 4.9838 4.9927 5.0026 5.0064 5.0614 5.0764 5.0772 5.0787 5.0793 5.0715 5.0706 5.0739

5.6606 5.6633 5.6662 5.6761 5.6828 5.6937 5.7079 5.7 152 5.7267 5.7378 5.7434 5.8003 5.8068 5.8073 5.8111 5.8135 6.0162 6.0218 6.032 1

132.829 133.046 133.440 134.016 134.359 134.696 135.467 135.823 136.336 136.837 136.974 139.607 140.311 140.280 140.396 140.470 148.594 148.578 149.186

P

120.69 120.67 120.64 120.68 120.72 120.66 120.63 120.62 120.60 120.58 120.61 120.58 120.60 120.58 120.59 120.58 118.22 118.30 118.26

electrode in the M, phase convert to the M, phase, and the parts in the 0, phase also convert to the M, phase. Determining x values in this region is difficult. There are peaks present from three different structures and we do not know what proportion of the lithium is going into the monoclinic or orthorhombic phase at any time. We do know that the M, to M, transition occurs at a lower voltage and that by about x=0.18 the electrode is single phase M,. The x values shown in fig. 5 and later in fig. 9 are calculated assuming uniform lithium transfer to all particles at all times. 33

34

35

Scattering

36

37

38

39

40

Fig. 4. Bragg peaks of LiXWOz in the second sampling region at four values ofx shown in the figure. d indicates that the measurement was made on discharge and c on charge. The dotted vertical lines are guides to the eye to indicate the extra structure present at x= 0.5 on charge. The Bragg peaks have been identified by letters as follows: for the monoclinic structure U= (0, 2, 0), w= (0, 0,2),x=(-Z,O,Z),y=(-2, 1, I),z=(Z,O,O);fortheorthorhombic structurep=(O, 2,0), q= (2,0,0), r= (0, 1, l), s= (1, 0, 1).

3.3. o
Angle / deg

If the cell is cycled between x=0 and 0.5, the structural changes are the same on charge as on discharge. After the discharge to x=0.5 only the M, phase is present. On charge the structure converts back to M, through a two-phase region, and then remains in M, back to x= 0.0. In contrast, during a full charge from x= 1, both phases M, and 0, are present at x=0.5. The presence or absence of the phase 0, at x= 0.5

71

A.K. Sleigh, W.R. McKinnon /Structure and electrochemistry of Li, WO, Table 2 Lattice parameters

of Li,W02

Monoclinic Voltage

on charge.

phase x

( V) 0.400 0.494 0.602 0.632 0.657 0.669 0.770 0.922 0.945 1.001 I .094 1.400

1.000 0.984 0.953 0.922 0.860 0.798 0.549 0.269 0.207 0.145 0.083 0.000

Orthorhombic Voltage

b, (A)

CM (A)

Volume

5.5342 5.5351 5.5187 5.5233 5.5260 5.5200 5.5258 5.5296 5.5422 5.5419 5.5507 5.5734

5.0739 5.0700 5.0660 5.0679 5.0640 5.0617 5.0793 4.9575 4.9652 4.945 1 4.924 1 4.896 1

6.032 I 6.0245 6.0173 6.0225 5.9855 5.9717 5.8135 5.6939 5.6988 5.6795 5.6685 5.6606

149.186 148.965 147.989 148.23 1 147.674 147.166 140.470 134.250 134.969 133.881 133.237 132.829

a0 (A)

bo (A)

CO (A)

Volume

5.1992 5.1582 5.1185 5.1156 5.1086

5.0686 5.0382 5.0212 5.025 1 5.0173

2.7678 2.7667 2.7664 2.7689 2.7686

72.9376 71.9014 71.0985 71.1787 70.9653

x

0.642 0.580 0.518 0.456 0.394

can be easily seen from ax/a V. Fig. 6a shows ax/a V versus V for a cell charged from 0.68 V and one from 0.4 V. The charge from 0.68 V shows a single peak at 0.9 V, corresponding to the conversion of MB to M,. The charge from 0.4 V shows three peaks. The one at the lowest V corresponding to the plateau at 0.6 V in fig. 5. Of the two peaks near 0.9 V, one is associated with the monoclinic phase (MB to M,), and occurs at the same voltage as for the charge from 0.8 V; the other corresponds to the orthorhombic phase (0, to M,). Note that the peak corresponding to the M, phase is at lower voltages than that for O,, so the M, phase disappears first, in agreement with the X-ray diffraction results. 3.4. 0.5
P

(A3) 118.26 118.22 118.40 118.44 118.16 118.22 120.58 120.67 120.61 120.67 120.69 120.69

phase

(v 0.705 0.743 0.809 0.862 0.881

ahl (A)

1.0

The discharge from x=0.5 to x= 1 depends on whether the cell had previously reached x=0.5 by charge or discharge. After a charge from x= 1 to x=0.5, both the M, and 0, phases are present; on

(A’)

the other hand, after a discharge from x=0 only M, is present. The different starting configurations produce different ax/al’ curves (fig. 6b). In addition, on cycles over the range 0.5
4. Cell cycling The first discharge of our WOz+C cells is different from the subsequent discharges (such as those shown in figs. 1 and 6), with extra peaks in ax/al/. Fig. 7 compares ax/a V for cells of WOz + C, W03 + C, and C alone on first discharge, showing that these extra peaks are not associated with the WOz. The first extra peak, at about 1.2 V, is associated with the carbon; the second, at about 0.87 V, is associated with impurities of W03 in the WOz. The carbon contribution can shift in voltage and split into several

A.K. Sleigh, W.R. McKinnon /Structure and electrochemistry of Li, WOz I

I

’

mixed

’

I

’

I

phase region

4

0

4

0

.

-I

Fig. 6. ax/aV against V (a) for cells charged from 0.4 V (solid curve) and 0.68 V (dashed curve) and (b) for cells discharged from 1.4 V (solid curve) and 0.77 V check (dashed curve). . .

0

0.2

0.4 x

Fig 5. (a) V(x); (b) bols), a, and c, sin(p) d along the metal atom parameters and squares

0.0

0.8

1

in Li,WO,

unit cell volume; (c) b,, b, (solid symand (d) twice the average W-W distance chains. Circles show the monoclinic cell the orthorhombic phase.

peaks. It is probably due to reaction of the electrolyte with the carbon, because it appears only on the first discharge and does not lead to any capacity at higher voltages. The peak from W03 at 0.87 V is also only seen on the first discharge, but subsequent charges and discharges show broad features in ax/Cl V at higher voltages, between 1.5 and 2.5 V. Such behaviour, a plateau that leads to capacity at higher voltage on subsequent cycles, suggests a severe change in the W03, either a rearrangement of the structure or a breakdown of the structure to other oxides of W and Li; analogous changes are seen in sulfides and selenides [ lo]. If W03 is discharged to 0.05 V there is another plateau in V(x), and in the subsequent

0.4

0.6

0.8

1

1.2

1.4

v/v

Fig. 7. ax/aV against V for the first discharge of three cells with cathodes as shown. The curve for the W03+C cell has been scaled by I/2.

charge dx/aV shows a broad peak centered around 1.4 V, suggesting further changes, perhaps the production of an intimate mixture of W and LizO. If a WOz cell is discharged past 0.4 V on its first discharge there is a long voltage plateau between 0.25 and 0.05 V which consumes Ax= 3.5 worth of Li in

A.K. Sleigh, WR. McKinnon /Structure and electrochemistry ofLi, W02

Li , +&WOZ, as shown in fig. 8. The subsequent charge to 3.0 V shows a broad peak at about 1.4 V in ax/ aP’, corresponding to the reaction of about 4 Li per W. As in W03, the change in ax/aP’after this plateau at 0.15 V suggests a breakdown of the Li,W02 into mixtures of tungsten and lithium oxides. Although this plateau and associated breakdown of WOZ appears below 0.25 V on the first discharge of a cell, it appears at higher voltages in cells that have been cycled. In cells cycled only to 0.4 V, ax/aVshows an upturn near 0.4 V on discharge after about the fourth cycle. Subsequent charges show a greater proportion of M, phase at x=0.5. In cells cycled between 0.4 and 1.4 V after a deep discharge, this breakdown in itself does not reduce the cells apparent capacity because the capacity from the decomposition products roughly compensates for the capacity loss from WOz. Subsequent cycling of the decomposition products does, however, show a fast capacity fade. The amount of 0, phase present at x=0.5 depends on the depth of discharge. We discharged several new cells to different lower trip points on their first discharges, and then charged them back to above 1.O V. We estimated the relative amounts of O8 and M, phase at x=0.5 from the relative heights of the two peaks in dx/aV near 0.9 V on charge (see fig. 6a). After a discharge to between 0.4 V and 0.5 V, the peak associated with the 0, phase is approximately twice the height of that associated with the M, phase. Discharging to 0.25 V, however, gives

- - - 2nd discharge

Fig. 8. Voltage versus x in Li,W02 for a cell cycled between 3.0 and 0.05 V. The first discharge is shown together with the subsequent charge and second discharge.

73

mostly M, phase, and to 0.15 V gives all M, phase. Furthermore, the disappearance of the 0, phase after a deep discharge is irreversible. A fresh cell discharged to 0.15 V, charged, and then discharged to only 0.4 V showed no 0, phase on either the first or the second charge. This cannot be accounted for purely by decomposition of the cathode material because the size of the monoclinic peak increases at the expense of the orthorhombic. In other words, some parts of the cathode which would have undergone an orthorhombic phase transition follow the monoclinic path after a deep discharge.

5. Discussion In contradiction to the results of ref. [ 41, we see clear shifts in the lattice parameters as x varies in Li,W02. We suggest that the chemical method of lithiation (the use of Li-naphthalide) used in this reference has the same effect as a voltage in electrochemical cells of about 0.15 V. At that voltage, Li,W02 decomposes into some mixture of oxides. The decomposition products are poorly crystalline, and are not seen in powder X-ray diffraction. Thus the samples prepared in ref. [4] consisted of the starting phase of WOZ and these decomposition products, and the powder diffraction patterns showed only the crystalline starting material. Murphy and Christian [ 111 estimate the redox potential of naphthalide to be roughly equivalent to the lower end of the cycling limits of WOZ. Fig. 9 summarizes the room temperature phase diagram of Li,W02. As discussed above, the error in the positions of the phase boundaries is about 0.05 in x. The diagram for charge is divided into two portions, to distinguish those particles or regions that form the 0, phase on charge from those that do not. The phase diagram is similar to that of Li,MoO, determined by Dahn and McKinnon [ 61, but there are significant differences. In Li,MoOz, the single-phase region from x=0 to x= 0.5 is not interrupted by a two-phase region, even though ax/aV shows a peak; in Li,WO* the peak does correspond to a two-phase region. Consistent with this difference, the curves of V(x) for Li,MoOz, 0
A.K. Sleigh, W.R. McKinnon /Structure and electrochemistry of Li, WO,

74

DISCHARGE I

I

’

II

’

MY Mp + M

7 -i

0.0

0.0

I

x in Li,WO,

I

CHARGE ’ 1

I

I

0

O+M,

MY

C

MY 0

0.2

0.4 x

0.6

0.0

in Li,WO,

Fig. 9. The phase diagram of LixW02 at room temperature for discharge from x=0 and charge from x= I. M,, M, M, are three different monoclinic phases, 0 is the orthorhombic phase (see text ).

whether or not the 0, phase is formed. In Li,MoO*, the 0, phase forms on discharge near x=0.5, but with no feature in ax/a V. Presumably the coexisting monoclinic and orthorhombic phases have almost the same Li content, so the region of coexistence is small. In Li,WO*, the 0, phase forms on charge, but only after the M, phase has been formed on discharge, and only if the M, phase has not been fully lithiated. Even then not all of the material in the cathode forms the 0, phase. The ratio of 0 to W in the host W02 is fixed, so Li,WO* can be regarded as a pseudobinary phase. In a binary system in equilibrium two phases can coexist only at a single composition at a given temperature. Thus the coexistence of two phases (M, and 08) over a range of Li composition implies that the

system is not in true equilibrium. This emphasizes that there must be some kinetic barrier to the conversion of 0, to M,. Cycling through the transition to the M, phase leads to a rapid loss of capacity in Li/LiXWOZ cells. When cycled over only the higher voltage plateau the cells cycle reasonably well, and lose only 1% of their capacity per cycle. In cells cycled between 0.4~ V< 1.4 over both plateaus, however, the voltage curves show considerable hysteresis (fig. 1) and there is about 5% loss in capacity per cycle. There is no obvious reason for the capacity loss. The lattice parameters change about as much from x=0.5 to x= 1 as from x= 0.0 to x= 0.5, so there is no reason to blame structural changes in the electrode. The voltage is only 0.24 V lower on the second plateau than on the first, so it seems unlikely that any decomposition of the electrolyte would suddenly become worse on the lower plateau. Understanding the reason for this capacity loss is important, because the usefulness of WOZ is questionable if it can only be used over 0
References [ 11 D.W. Murphy,

F.J. Di Salvo, J.N. Carides Waszczak, Mater. Res. Bull. 13 ( 1978) 1395. 121J.J. Aubom and Y.L. Barberio, J. Electrochem. ( 1987) 638.

and

J.V.

Sot.

134

A.K. Sleigh, W.R. McKinnon /Structure and electrochemistry of Li, WOz [ 31 M. Lazzari and B. Scrosati, J. Electrochem.

Sot. 127 (1980) 773. [4] K.M. Abraham, D.M. Pasquariello and E.B. Willstaedt, J. Electrochem. Sot. 137 ( 1990) 743. [ 5] D.B. Rogers, R.D. Shannon, A.W. Sleight and J.L. Gillson, Inorg. Chem. 8 ( 1969) 84 1. [6] J.R. Dahn and W.R. McKinnon, Solid State Ionics 23 (1987) 1. [7] D.J. Palmer and P.G. Dickens, Acta Cryst. B 35 (1979) 2199.

75

[8] D.C. Dahn and R.R. Haering, Solid State Commun. 44 (1982) 29. [9] J.R. Dahn and W.R. McKinnon, J. Electrochem. Sot. 131 (1984) 1823. [lo] L.S. Selwyn, W.R. McKinnon, U. von Sacken and CA. Jones, Solid State lonics 22 (1987) 337. [ 111 D.W. Murphy and P.A. Christian, Science 205 ( 1979) 65 1.