Studies of ternary Li2SO4–Li2O–P2O5 glasses

Studies of ternary Li2SO4±Li2O±P2O5 glasses

Munia Ganguli, K.J. Rao *

Solid State and Structural Chemistry Unit, Indian Institute of Science, Bangalore 560 012, India

Received 8 April 1998; received in revised form 12 August 1998

Abstract

Glasses in a wide range of compositions in the ternary system xLi2SO4±yLi2O±zP2O5 where x ranges from 0 to 30

mol%, y ranges from 35 to 55 mol% and z ranges from 25 to 50 mol% have been prepared and their properties measured

using infra-red, Raman, and 31P magic angle spinning nuclear magnetic resonance spectroscopic techniques. We

conclude that a random close packing of phosphate and sulphate ions which also leads to formation of connected voids

in the structure is consistent with our data. There is also evidence for formation of condensed sulphate±phosphate

species in the liquid which may be retained in the glass structure. Ó 1999 Elsevier Science B.V. All rights reserved.

PACS: 61.43.F

1. Introduction

Lithium ion conducting glasses have attractedattention in recent times because of the possibilityof their application as electrolytes in lithium bat-teries [1±5]. An understanding of lithium ion mo-tion in disordered lattices is still lacking, and suchan understanding would lay foundations for fur-ther development and design of glassy solidelectrolytes. Lithium ions are conventionally in-troduced into glasses as Li2O which also acts as amodi®er in typical network glasses such as SiO2,P2O5, and B2O3. Glasses with such modi®ed net-works containing Li� ions are ionic conductors[6,7]. The activation barriers for conduction gen-erally decrease with increasing concentration ofthe modifying oxide. Ion transport is a�ected whenmore than one type of cation or anion is present

(mixed ion e�ects) in a glass [8]. When a cationsuch as Li� is present in a mixed anion environ-ment, the conductivity is observed to be largerthan when it is present in the environment of asingle type of anion [9].

It is known that in metaphosphate glasses, theanionic skeleton is quite like an organic polymerwith chains of connected PO4 tetrahedra andtherefore the non-bridging oxygen ions which arecharged are attached to all the phosphorus atomsin the backbone [10,11]. Further addition of alkalioxide decreases the connectivity of the networkstructure and leads to the formation of shorterphosphate chains and ultimately to pyrophosphateunits [10,11]. Modi®ed network glasses can be`doped' with other lithium salts such as LiI whichcontains the same cation (Li�) and discrete anions(Iÿ). Such modi®ed and `doped' glasses have largerionic conductivities. Their structures are howevercomplex. It is of interest to investigate the salt`doped' amorphous systems to understand andfurther develop glassy electrolytes.

Journal of Non-Crystalline Solids 243 (1999) 251±267

* Corresponding author. Tel.: +91-80 309 2583; fax: +91-80

334 1683, e-mail: [email protected].

0022-3093/99/$ ± see front matter Ó 1999 Elsevier Science B.V. All rights reserved.

PII: S 0 0 2 2 - 3 0 9 3 ( 9 8 ) 0 0 8 3 2 - 1

In this paper we report our investigations internary Li2SO4±Li2O±P2O5 glass system in whichLi� ions are present in an anion matrix of complexphosphate and discrete sulphate ions. We expect itto be a good glassy ionic conductor. We have ex-amined the glass formation and glass transitionover a range of compositions in this system andhave investigated the structures of these glassesusing Raman, infra-red (IR) and high-resolutionmagic angle spinning nuclear magnetic resonance(HR-MASNMR) spectroscopic techniques. Theseinvestigations have been used to model a possiblestructure of these glasses and correlate the ob-served properties to the structure.

2. Experimental

Glasses were prepared by the conventional meltquenching method. The starting materials,Li2SO4 á H2O (BDH laboratory reagent), Li2CO3

(Qualigens), and (NH4)2HPO4 (BDH laboratoryreagent)(all of Analar Grade), were mixed bygrinding in appropriate quantities to constitute a5±10 g batch. The ground mixtures were heated inporcelain crucibles at 573 K for about 5 h in amu�e furnace to decompose Li2CO3 and

(NH4)2HPO4 to Li2O and P2O5, respectively, andalso to remove water from Li2SO4 á H2O. Thebatches were then melted at 1123 K for about 10min, stirred to improve homogeneity, and quen-ched between stainless steel plates at room tem-perature. For the purpose of making IRspectroscopic studies, ®lms of glasses were pre-pared by blowing bulbs from the melt held at thetip of a tapered quartz tube. The nominal compo-sitions of the glasses prepared in this manner andtheir designations are listed in Table 1. There ap-pears to be negligible weight loss during melting(0.005 g in a 10 g batch) as seen by weighing theinitial reaction mixture and the ®nal product aftermelting and cooling. Thus the actual glass com-positions are likely to be same as the nominalcompositions mentioned in Table 1. The glasseswere all transparent and colourless. Since they wereslightly hygroscopic, both the glass wafers and the®lms were stored in desiccators and taken out onlyat the time of measurement of their properties.

The amorphous state of the samples was con-®rmed using powder X-ray di�raction (JEOLJOX-8P X-ray di�ractometer). The glass transitiontemperatures (Tg) and heat capacities (Cp) weredetermined using a di�erential scanning calorime-ter (Perkin Elmer DSC 2). Dry nitrogen was used

Table 1

Compositions of the glasses prepared, along with codes of designation, densities, molar volumes and relative amounts of the phosphate

units present

Code Composition (mol%) Density g/cm3

(�0.001g/cm3)

Molar volume

cm3(�0.02 cm3)

Crystalline molar volume

cm3

P1 + P2

Li2SO4 Li2O P2O5

CLSP1 30 35 35 2.330 39.97 40.57 70P2

CLSP2 30 39 31 2.340 37.94 38.75 46P2 + 16P1

CLSP3 30 42 48 2.340 36.44 37.39 28P2 + 28P1

CLSP4 30 45 25 2.340 35.01 36.04 10P2 + 40P1

CLOP1 10 45 45 2.340 37.74 38.01 90P2

CLOP2 15 45 40 2.340 37.06 37.52 70P2 + 10P1

CLOP3 25 45 30 2.350 35.54 36.53 30P2 + 30P1

CLOP4 30 45 25 2.340 35.01 36.04 10P2 + 40P1

CP1 10 55 35 2.360 32.67 33.48 30P2 + 40P1

CP2 15 50 35 2.350 34.51 35.26 40P2 + 30P1

CP3 25 40 35 2.340 38.09 38.8 60P2 + 10P1

CP4 30 35 35 2.330 39.97 40.57 70P2

CLP1 0 50 50 2.350 36.56 36.74 100P2

CLP2 10 45 45 2.340 37.74 38.01 90P2

CLP3 20 40 40 2.340 38.72 39.29 80P2

CLP4 30 35 35 2.330 39.97 40.57 70P2

252 M. Ganguli, K.J. Rao / Journal of Non-Crystalline Solids 243 (1999) 251±267

as a purge gas. For speci®c heat measurements,gold pans were used in which glass wafers weigh-ing about 25 mg were placed. The samples were®rst annealed at temperatures about 20 K belowTg (an estimate of Tg was obtained from a prior 20K/min scan) and DSC scans for speci®c heat atconstant pressure, Cp, measurements were carriedout at a heating rate of 10 K/min. Single crystallinealumina (sapphire) was used as the calibrationstandard for calculating Cp. Tg were determined asthe intersection of the extrapolated linear portionsaround the glass transition region in the Cp vs Tplots. The di�erences in heat capacities of thesamples and the supercooled liquids at Tg (DCp)were determined by suitable extrapolation of therespective Cp at temperatures Tg � 20 K.

Densities (q) of annealed glass bits free of airbubbles and cracks (determined by visual exami-nation) were measured by standard Archimedian(apparent weight loss) method with xylene as thereference ¯uid. The weights of the glass bits usedwere about 0.3 to 1 g. The molar volume of eachcomposition was calculated as Vm � M=q where Mis the molecular (formula) weight of the glass (themolecular unit is taken as xLi2SO4±yLi2O±zP2O5

with x� y � z � 1).IR transmission measurements from 4000 to

400 cmÿ1 using a double beam IR spectrometer(PE 580, with an instrument resolution of 1 cmÿ1)were made on blown ®lms. Fourier Transform IR(FTIR) spectra of selected samples were recordedin the far IR region (between 250 and 550 cmÿ1)using a FTIR spectrometer (Bruker IFS 113).Glass discs of approximately 1 cm2 area and about0.5 mm thickness were used for recording theFTIR spectra.

Unpolarized Raman spectra were recorded (inthe range of 100 to 1500 cmÿ1) on a Ramanspectrometer (SPEX 1403) making use of the ex-citation wavelength of 514.5 nm from an Argonion laser. Glass pieces of approximately 1 cm2 areaand 0.8 to 1 mm in thickness were used for theRaman measurements.

31P HR-MASNMR spectra of powdered glasssamples were recorded on a solid state high reso-lution spectrometer (Bruker DXL 300, magnetic®eld� 7.05 T) operating at 121.494 MHz. 90°pulses of 5 ls duration were employed with a delay

of 5 s between the pulses. A cylindrical zirconiarotor was used to spin the samples with spinningspeeds of 3±7 kHz. The spinning sidebands wereidenti®ed by virtue of their drifts when sampleswere spun at di�erent speeds. The chemical shiftswere measured with respect to 31P resonance in an85% solution of H3PO4. Static 7Li NMR spectrawere also recorded for selected samples on thesame spectrometer operating at 116.642 MHz.

3. Results

3.1. Molar volume

The densities and molar volumes of all thesamples are listed in Table 1. The sum of the molarvolumes of the components of the glass in theircrystalline state have also been calculated andlisted in the table. The molar volume of lithiummetaphosphate in its crystalline state is 34.91 cm3

which is less than the weighted sum of the molarvolumes of Li2O and P2O5 (36.74 cm3). Suchnegative volume changes con®rm chemical inter-actions. Molar volume of lithium metaphosphateglass (CLP1) is equal to 36.56 cm3 and is largerthan the crystalline volume by about 4.7% which isof the same order as observed in several otherglasses [12]. The glass molar volumes are alsouniformly less than the weighted linear sums of thevolumes of the component oxides, although to amuch smaller fraction (�1.5%). The variation ofexperimentally observed molar volumes and thosecalculated on the basis of the proposed structuralmodel (see Section 4) is shown in Fig. 1.

3.2. Thermal properties

The glass transition temperatures (Tg) of thevarious samples, the heat capacities at (Tg)20)K,the change in heat capacity at the glass transitiontemperature, and the Dulong Petit heat capacitiesare listed in Table 2. The Tgs lie in a range from540 to 590 K despite the range of their composi-tions. The highest glass transition temperature ismeasured on the pure metaphosphate sample. In-creasing Li2SO4 as in the CLP series seems to a�ectTg which decreases with increasing Li2SO4. Hold-

M. Ganguli, K.J. Rao / Journal of Non-Crystalline Solids 243 (1999) 251±267 253

ing Li2SO4 concentration a constant and varyingLi2O/P2O5 ratio as in CLSP series of glasses doesnot a�ect the Tg although the variation of theLi2O/P2O5 alters the chemistry of the glass. In theCP and CLOP series, Li2O/P2O5 ratios vary inopposite direction but the trend in Tg is the same

and follows the trend for changes in the concen-tration of Li2SO4. It is also noted that Tg is notcorrelated with total lithium ion concentrationwhich is the total positive charge concentration inthe glass. Tg does not vary, within errors of mea-surement, in the CLSP series where Li� ion con-centration increases, and decreases in the CP serieswhere Li� concentration is held constant. There-fore the glass transition temperature is in¯uencedmost by changes in Li2SO4 concentration, and thisis evident in Fig. 2.

The heat capacities of all samples (Table 2) areless than the corresponding Dulong Petit heat ca-pacities, although the di�erences are less in thesamples containing large amounts of Li2SO4. Theheat capacity changes at the glass transition tem-peratures were step-like in all cases. Relaxationalhumps in heat capacity were also evident in theplots as normally observed in ionic glasses (seeFig. 3). DCp ranged between 38 and 67 J Kÿ1

molÿ1. DCp seem to be dependent on the phos-phate content of the glasses as shown in Fig. 4.

3.3. Infrared spectra

Infrared spectroscopy has been used in the midIR region to identify the anionic species present inthese samples and the spectra are shown in Fig. 5.The absence of structural or free water in the

Table 2

Thermal properties of the Li2SO4±Li2O±P2O5 glasses

Sample Tg K (�2 K) Cp (at Tg)2 K) J Kÿ1 molÿ1

(�0.5 J Kÿ1 molÿ1)

3 nR J Kÿ1 molÿ1 DCp J Kÿ1 molÿ1

(�1 J Kÿ1 molÿ1)

CLSP1 540 136.3 139.7 48

CLSP2 540 129.2 135.9 38

CLSP3 542 130.4 132.7 42

CLSP4 548 125.9 129.7 )CLOP1 576 102.0 129.7 57

CLOP2 561 110.5 129.7 55

CLOP3 553 102.9 129.7 42

CLOP4 548 125.9 129.7 )CP1 569 104.0 119.7 55

CP2 563 108.1 124.7 50

CP3 555 124.6 134.7 42

CP4 540 136.3 139.7 48

CLP1 591 108.9 124.7 67

CLP2 576 102.0 129.7 57

CLP3 565 118.5 134.7 58

CLP4 540 136.3 139.7 48

Fig. 1. Variation of experimentally observed molar volumes

with calculated molar volumes (based on the model) for all

compositions (line drawn as a guide to the eye, denotes the

error bar of data points magni®ed 10 times).


samples was veri®ed by the total absence of IRpeaks at both �1670 cmÿ1 and �3600 cmÿ1. Thesamples contain anionic species such as SO2ÿ

4 andphosphate ions of as yet undesignated type. Thereare only two modes of the SO2ÿ

4 ions which areinfrared active, m3 and m4, which are the asymmetricstretching (1135 cmÿ1) and the bending (645 cmÿ1)

modes respectively [13]. These modes are present inall the samples except in CLP1 in which Li2SO4 wasabsent. The absence is more obvious in the bendingmode region than in the asymmetric stretchingmode because phosphates also produce absorptionin the latter region. The stretching modes of thephosphate groups (with one or more non-bridgingoxygens) all occur in the region of�1100 cmÿ1 [14±16]. Similarly P±O±P stretching modes are seen at�890 cmÿ1 (asymmetric) and �700 cmÿ1 (sym-metric) respectively and their maximum amplitudesvary with the Li2O/P2O5 ratio [14±16]. In sampleswith larger Li2O/P2O5 ratios noticable absorptionoccurs around 970 and 1030 cmÿ1 and this ab-sorption is particularly evident in all the glass se-ries. However the presence of these features in thespectra of CLP samples seems to be surprising sincethe Li2O/P2O5 ratio is just unity. Absorptionsaround 600 and 470 cmÿ1 in the spectra can beassociated with P±O±P bending vibrations [14±16].These bands are present in all the samples either asa distinct or a shoulder-like feature at these fre-quencies. We suggest that the presence of only ashoulder in the spectrum of CLP1 (around 600cmÿ1) shows that this region of the spectra isdominated by the bending vibrations of the SO2ÿ

4

ions. However, the 470 cmÿ1 absorption has a

Fig. 2. Variation of glass transition temperature (Tg) with vol-

ume fraction of Li2SO4. (line drawn as a guide to the eye,

denotes the error bar of data points).

Fig. 3. Variation of Cp with temperature for the sample CP1.

(line drawn as a guide to the eye).

Fig. 4. Variation of DCp with the phosphate content of all

glasses. Experimental values are denoted by n while those

calculated on the basis of the model are denoted by s. de-

notes the error bar of data points.


Fig. 5. Infra-red spectra of (a) constant Li2SO4 (CLSP) glass series (b) constant Li2O (CLOP) glass series (c) constant P2O5 (CP) glass

series and (d) constant Li2O/P2O5 glass series in the glass system Li2SO4±Li2O±P2O5.


larger width (see the spectra of CLP1, CLP2, CP2,CP3, CLOP2 and CLSP1). There is an absorptionat 1240 cmÿ1 [14] in all the samples due to thepresence of P@O. This feature is however broaderin Li2SO4 rich samples which may be due to thee�ect of larger concentrations of mobile Li� ionson the general pro®le of the vibrational potential ofthe P@O group. This e�ect is evident in the clearlycontrasting pro®les of 1240 cmÿ1 bands in thespectra of Li2SO4-free CLP1 and the Li2SO4 richCLP4. The dominance of the P@O stretching bandis also a�ected by the increase in the Li2O/P2O5

ratio. It becomes less de®ned as the ratio of Li2O/P2O5 increases. This decrease is evident in CLSP,CLOP and CP series. That the P@O frequenciespersist even to a Li2O/P2O5 ratio of about 1.5 maybe compared with the observation of Brow et al.[17] who noted that P@O frequencies vanish forratios >R2O/P2O5 � 0.43, where R�Na. This ef-fect points to the fact that lithium ions may resideclose to the non-bridging oxygens so that the res-onance between a P@O and a P±Oÿ is suppressed.The role played by Li2SO4 itself in the retention ofP@O bond in the structure is however unknown.

FTIR spectra for some selected compositions inthe far IR region are shown in Fig. 6. The spectrashow the presence of a band at �450 cmÿ1 alongwith several other absorptions with smaller am-plitudes in the lower frequency region of thespectrum. These bands arise from vibrations oflithium ions in oxygen ion cages [18]. The band at�450 cmÿ1 has a small width (FWHM� 15±25cmÿ1) compared to bands due to lithium ion vi-brations in pure lithium metaphosphate glasses[18]. This width is probably because in thesemodi®ed glasses, due to presence of large concen-trations (up to 30 mol%) of Li2SO4, there is easyformation of oxygen coordination polyhedraaround lithium ions since the oxygens can arisefrom both phosphate and sulphate units.

3.4. Raman spectra

Raman spectra of the glasses are given in Fig. 7.The bands in the Raman spectra can all be asso-ciated with the anionic species in these glasses.Raman active SO2ÿ

4 vibrations occur in the regionof 465 cmÿ1 (m2), 645 cmÿ1 (m4), 1010 cmÿ1 (m1) and

1135 cmÿ1 (m3) [13]. The presence of bands around465 and 645 cmÿ1 is evident in the spectra of all thesamples except CLP1 in which Li2SO4 is absent.The band at �1010 cmÿ1 is also present in all thesamples except CLP1. The 1135 cmÿ1 band, how-ever, appears to have merged into the �1175 cmÿ1

band which is the symmetric stretching mode ofthe metaphosphate groups [19±21]. They are notobserved as separate and well de®ned bands. ARaman band around 1280 cmÿ1 [19±21] present inthe spectra whose maximum amplitude is consis-tently smaller than that of the 1175 cmÿ1 band canalso be associated with the metaphosphate group.The P±O±P stretching at around 700 cmÿ1 [21] ispresent in all the glasses. It is interesting to notethat in the CLSP series where Li2SO4 proportion isheld constant and Li2O/P2O5 ratio is increased, the

Fig. 6. FTIR spectra of some selected glasses from the CLSP

and CLP series.


Fig. 7. Raman spectra of (a) constant Li2SO4 (CLSP) glass series (b) constant Li2O (CLOP) glass series (c) constant P2O5 (CP) glass

series and (d) constant Li2O/P2O5 glass series in the glass system Li2SO4±Li2O±P2O5.


maximum amplitudes of the bands at �1175 and1280 cmÿ1 decrease (accompanied by a shift tolower energies) but the �1010 cmÿ1 band increasesin amplitude. There is also the emergence andgrowth of bands at �750 and �1043 cmÿ1. Sincethe SO2ÿ

4 spectral features are not expected to in-crease in amplitude, the bands at 750 and 1043cmÿ1 and their growing amplitudes are attributedto increasing modi®cation of the phosphate chainsby the increased proportion of Li2O [21]. This ef-fect is borne out in the spectra of CLOP series ofsamples also. In CP series an opposite trend isobserved by the increasing amplitudes of 1175 and1280 cmÿ1 bands and the disappearance of the 750and 1015 cmÿ1 bands which is consistent with thedecreasing Li2O/P2O5 ratio. In the CLP series, theamplitudes of the peaks due to metaphosphategroups decrease but less than that in the CLSPseries because of the relatively smaller decrease intheir proportion (see Table 1). But there is an in-crease in the intensity of the 1016 cmÿ1 band whichwe attribute to an increase in SO2ÿ

4 ions. The 750cmÿ1 band is absent in these samples. In sampleswith larger Li2O/P2O5 ratios, the band at 1015cmÿ1 splits. This split is particularly evident in thespectra of samples CLP4, CLSP4 and CP1.

The absorption in the 330 cmÿ1 region is seen inall the samples. The symmetric vibrations of [LiOn]polyhedra may be responsible for the broad fea-ture at 330 cmÿ1 [22].

3.5. 31P HR-MASNMR and 7Li static NMRspectra

The MASNMR spectra of the 31P nucleus weremeasured for all the samples and the spectra areshown in Fig. 8. The 31P resonances have beenidenti®ed and all the chemical shifts are listed inTable 3. Only two 31P resonances are observed,and by reference to the literature [23±25] they areassociated with 31P in metaphosphate (chemicalshift �)22 ppm) and pyrophosphate (chemicalshift � )3ppm). The relative intensities of the tworesonances measured as areas under the peaks arealso listed in Table 3.

A few 7Li NMR spectra obtained as simplepowder patterns are shown in Fig. 9. The shapesof the resonance peaks indicate lithium ion mo-

bility because the widths of resonance peaks aresmaller than that expected for a powder pattern[26] (FWHM� 5.33 kHz for the signal with thegreatest width).

4. Discussion

4.1. Structural model

To understand the variation of the physicalproperties and the spectroscopic features presentedin the previous section as a function of composi-tion, we ®rst develop a possible model for thesesamples. We note here that the glass consists of longphosphate chains only in the metaphosphate com-position (where Li2O/P2O5� 1). Metaphosphatechains can also give rise to closed rings, the presenceof which has been noted in chromatography [27]and other experiments [10,28]. But in the otherthree series of samples considered in this work(CLSP, CLOP and CP) the Li2O/P2O5 ratio is notheld at unity but is uniformly larger such that theratio lies between 1 and 2. In these cases thereforethe network (linear metaphosphate chains or rings)is further modi®ed to various extents dependingupon the Li2O/P2O5 ratio. Hence truncated chainsin which a fraction of the phosphate tetrahedrapossess three unshared oxygen corners produce thestructure [10,11]. We designate the tetrahedralspecies as P2 and P1 where P2� [POO2=2O]ÿ andP1� [POO1=2O2]2ÿ. The subscripts 2 and 1 corres-pond to the number of bridging oxygens (BO)present in the phosphate tetrahedron (the numberof non-bridging oxygens NBO would be 4 - BO butthe coulombic charge on the phosphate groupwould be 3 - BO). In any given composition, theproportions of P2 and P1 can be calculated in thefollowing way. Consider the CLSP2 sample whichhas 39 mol% of Li2O and 31 mol% of P2O5. Thestructure of a unit P2O5 in the glass is

and can be written as 2[POO3=2]� 2P3. Modi®ca-tion implies breaking of P±O±P bonds by theadded O2ÿ ions as


Fig. 8. 31P MAS-NMR spectra of (a) constant Li2SO4 (CLSP) glass series (b) constant Li2O (CLOP) glass series (c) constant P2O5 (CP)

glass series and (d) constant Li2O/P2O5 glass series in the glass system Li2SO4±Li2O±P2O5. The spectra were recorded at a spinning

speed of 7 kHz.


or

2�POO3=2�0 � Li2O! 2�POO2=2O�ÿ � 2Li�

or

P2O5 � Li2O! 2P2 � 2Li�:

We assume that in the ®rst step all P2O5 (P3 units)are converted into P2 units which is followed bythe next stage of modi®cation

or

2P2 � Li2O! 2P1 � 2Li�:

Hence in CLSP2, the ®rst stage is formation of P2

units using the required quantity of P2O5

31Li2O� 31P2O5 � 62P2:

This formation is followed by the formation of P1

by the reaction of remaining Li2O with P2

62P2 � 8Li2O � 16P1 � 46P2:

The numbers of P1 and P2 units present in eachcomposition are listed in Table 1. The importanceof these structural units will be discussed while ra-tionalizing the observed properties of these glasses.

4.2. Molar volume

It is instructive to see if the molar volume oflithium metaphosphate can be compared with the

Fig. 9. 7Li static NMR spectra for some selected samples in the

Li2SO4±Li2O±P2O5 glass system.

Table 331P chemical shifts of the Li2SO4±Li2O±P2O5 glasses and the

relative proportions of P1 and P2 species as determined by 31P

HR-MASNMR

Sample Chemical shifts ppm

(�0.01 ppm)

A1/A2 (�0.01)

CLSP1 )2.99, )21.35 0.75

CLSP2 )3.02, )20.27 0.9

CLSP3 )2.41, )19.132 1.32

CLSP4 )2.8, )18.1 1.99

CLOP1 )4.37, )22.71 0.66

CLOP2 )4, )21 0.82

CLOP3 )3.9, )20.3 1.31

CLOP4 )2.8, )18.1 1.99

CP1 )2.43, )18.16 1.37

CP2 )2.82, )19.58 1.07

CP3 )2.92, )20.19 0.93

CP4 )2.99, )21.35 0.75

CLP1 )21.74 0.0

CLP2 )4.37, )22.71 0.66

CLP3 )3.29, )21.31 0.71

CLP4 )2.99, )21.35 0.75


random close packed (RCP) volume of only the 3moles of oxide ions present in them (oxide ionradius is 1:36 �A). Li� and P5� ions may be con-sidered as simply embedded in the voids of theRCP assembly of O2ÿ ions. Since RCP results inthe formation of tetrahedral voids almost exclu-sively, the radius of the embedded ion can be atmost 0:3 �A�0:22� 1:36 �A� and can therefore ac-commodate the P5� ions (radius 0:35 �A). However,Li� ion (of radius 0:68 �A) can be expected to ex-pand the random close packed assembly of oxideions. We assume that the four surrounding ionsmove outwards isotropically. To a ®rst approxi-mation this expansion due to the insertion of onemole of Li� ions can be treated as a volume of

�4p=3��3p

a=2� 0:38� �3

ÿ��3p

a=2� �3

� �� 10ÿ24�3N=4� cm3 � 7:42 cm3;

where��2p

a is the inter oxygen ion distance and N isthe Avogadro number. The last factor of (3N/4) isintroduced because not all oxygen cages are inde-pendent. Therefore on a per oxygen basis the in-crease in volume will be �4p=3�� 3p a=2� 0:38�3ÿ� ��3p a=2�3� � 10ÿ24=4 cm3 and for the 3N oxy-gens, it is multiplied by 3N. Thus the RCP volumeof the LiPO3 glass (with packing fraction of 0.63)should be

3� �4p=3�N � �1:36�3 � 10ÿ24=0:63� 7:42 cm3

which is equal to 37.64 cm3 and is close to theobserved molar volume of 35.76 cm3. We note herethat the glass volumes calculated on the basis ofclose packing of N phosphate (PO4) groups withspherical shapes corresponding to circumscribedspheres (corrected for the shared oxide ion vol-ume) leads to unphysically large molar volumes ofthe glass (67 cm3).

We suggest that molar volumes indicate that thelong chains of metaphosphate are intertwined soas to give rise to smaller distances between oxygenions, although the large sizes of Li� ions (largerthan the voids), P±O±P angles (with a distributionbetween 120° and 180° [29]), and the Oÿ±Oÿ

coulombic repulsions prevent the metaphosphate`spaghetti' from getting to be as dense as the RCPstructure of only O2ÿ ions (30.22 cm3). The glass

structure therefore has a built-in excess of struc-tural volume which can assist lithium ion motion.The volume of oxide ions alone is 30.22 cm3 andthe excess volume due to increased size of Li� ionoccupied voids is �4p=3�N � 10ÿ24�0:68ÿ 0:3�3� 0:14 cm3. Thus in the actual glass volume (36.56cm3) the structural `excess' volume is about�36:56ÿ 30:22ÿ 0:14�cm3 � 6:2 cm3. We assumethis excess volume to be randomly distributed inthe structure as random voids. Although there is adistribution of void sizes we examine the conse-quences of assuming a single size for the voids. Ifthe voids are all of the size of oxide ions, 6.2 cm3 ofexcess volume constitutes a volume fraction (6.2/30.22)� 0.2 of the RCP volume of 3 moles of ox-ide ions. Therefore it is equivalent to adding0.2 ´ 3� 0.6 moles of the voids of the same size asoxide ions to the system. We suggest that withvolume fraction being 0.2 these voids are alreadyin the percolation regime ( P 0.16)(connected voidpaths). This connectivity is likely to be the reasonfor lithium ion conductivity in these glasses whichwe have investigated.

In the CLP series of samples up to 30 mol% ofLi2SO4 has been added with an attendant increasein the molar volumes. But this increase in molarvolume (3.41 cm3) is actually less than the increasecalculated by taking a linear sum of the volumes ofLiPO3 and Li2SO4 (assuming that glassy Li2SO4

has the same volume as crystalline Li2SO4) by 0.48cm3. It is therefore seen that the added lithiumsulphate exploits, to some extent, the availablestructural void volume in the host lithium phos-phate. We may now recalculate the volume frac-tion of (oxide ion size) voids in CLP4 which is a�0:7� 3N � 0:3� 4N ��3:3N oxide ion assembly(oxide ions of the sulphate are treated in the samemanner as oxide ions of phosphates in computingthe total number of oxide ions). The RCP volumeof oxide ions is 33.24 cm3, experimental molarvolume is 39.97 cm3 and Li� ion size related vol-ume increase is �0:14� 1:3� � 0:18 cm3. Thus thestructural excess volume is �39:97ÿ 33:24ÿ 0:18�� 6:55 cm3. This volume immediately results in avolume fraction of 6.55/39.97� 0.16 equal to thevolume of 0.53 moles of O2ÿ ion sized voids. 0.16 islarger than the percolation limit for the voids inthese systems [30]. It is evident that in CLP2 and in


CLP3 the volume distributions are those betweenthe extremes of CLP1 and CLP4.

The exploitation of excess structural volume bylithium sulphate in the glass structure togetherwith the fact that the molecular weights and molarvolumes of Li2SO4 are both �28% and �35%larger than that of LiPO3 make the glass densitiesalmost constant for the whole series up to 30 mol%substitution by Li2SO4.

We may also note that when a bridging oxygen(BO) is converted to a non-bridging oxygen (NBO)in a phosphate tetrahedron, the average P±O dis-tance increases. The speci®c volume of P1 istherefore larger than that of P2. The volume in-crease is, in part, due to P1 which has a volumecomponent corresponding to the excess half atomof oxygen in it compared to P2. In the two series ofsamples (CLOP and CP) in which proportion ofLi2SO4 is not held a constant, the molar volume ofLi2SO4 is also a determining factor. Since themolar volume of Li2SO4 is larger than the molarvolumes of both P1 and P2 (P1 and P2 are chargedand associated with Li� ions) we would expect thatthe increase in Li2SO4 in a sample would increasethe molar volume. To comprehend the totality ofthe in¯uences we have evaluated the molar vol-umes on the basis of molar volumes of P1, P2 andLi2SO4 assuming the molar volume of Li2SO4

(49.5 cm3) to be the same as in its crystalline state.The model based molar volume is calculated asN1V1 � N2V2 � NsVs where N1, N2 and Ns are thenumber of moles of P1, P2 and Li2SO4, respec-tively. In Fig. 1 the measured molar volumes areplotted against the model volumes for all theglasses investigated in this work. The line drawn inthe ®gure corresponds to a slope of �1 and themodel seems to be satisfactory.

4.3. Thermal properties

In a naive approach, glass transition tempera-ture can be thought to mark the initiation of liq-uid-like ¯ow. In the absence of, or even in thepresence of very small amounts of P2O5 ¯owshould be associated with breakdown of theframework of sulphate anions. But a glassy stateof pure Li2SO4 is not easily realized and glasses

form only in the presence of substantial propor-tions of P1 and P2 units from phosphates. A matrixof only P1 and P2 may break up at temperatureshigher than that for a SO2ÿ

4 ion matrix because oftheir polymeric state. However when up to 30mol% of Li2SO4 is present in a glass, the break-down of the anionic framework can still occurwhen SO2ÿ

4 ion matrix breaks up. Tg should,therefore, be similar for the samples since thebreakdown of the structure is dependent only onLi2SO4. The presence of P1 and P2 (particularly P2)in the composition increases Tg. But all the com-positions having more than 20 mol% Li2SO4 seemto have a Tg of 560 K or less.

However, as phosphatic species in the structurealso exhibit motion above Tg, we speculate on themotions of phosphate units in¯uenced by the sul-phate ions. In the presence of sulphate ions, thereis a dynamical exchange of negative charge, andhence of the oxide ions, leading to several transientspecies:

2SO2ÿ4 � Pÿ2 � P3ÿ

0 � S2O2ÿ7 ;

2SO2ÿ4 � 2Pÿ2 � 2P2ÿ

1 � S2O2ÿ7 ;

SO2ÿ4 � Pÿ2 � SPO3ÿ

7 :

It is seen (in Table 4) that the molecular electro-negativities (v) of the SO2ÿ

4 , P2 and the SPO3ÿ7 are

respectively 2.08, 2.39 and 2.246, suggestive of apossible reaction [31] leading to the formation ofSPO3ÿ

7 (henceforth referred to as dithiophosphatespecies or DTP). (Molecular electronegativities ofionic groups are calculated by the procedureof Sanderson [32] which is based on the concept ofelectronegativity equalization.) Although the re-actions above appear to be of second order andhence Tg should depend on the concentrations of

Table 4

Group electronegativites of the various structural species pres-

ent in Li2SO4±Li2O±P2O5 glasses along with the partial charges

on their constituent atoms

Species v dO dS dP

SO2ÿ4 2.08 )0.45 )0.2 )

[POO2=2O]ÿ 2.39 )0.36 ) 0.09

[POO1=2O2]2ÿ 1.89 )0.54 ) )0.13

SPO3ÿ7 2.25 )0.41 )0.13 0.02


both sulphate and phosphate ions, the higherproportions of phosphate in the glass compared toSO2ÿ

4 ions make them pseudo-®rst order reactionswith apparent dependence of Tg on only SO2ÿ

4 ionconcentration. The variation of Tg as a function ofvolume fraction of lithium sulphate is plotted inFig. 2 and is consistent with this proposal.

The variation of the glass transition tempera-ture as a function of Li2SO4 concentration is alsoconsistent with the free volume model of glasstransition [33]. The free volume available in themolten phase is more e�ectively utilized fortransport by discrete anions along with the cat-ions. The excess free volume frozen into the glasswould be reduced by the presence of discrete an-ions. Thus the Tg would be close to the thermo-dynamic limits. Shown in Fig. 10 is the variation

of glass transition temperature as a function ofmole fraction of sulphate ions (total anions areconsidered as the sum (S + P1 + P2) where S is thenumber of SO2ÿ

4 ). The almost linear decrease of Tg

is consistent with such a conjecture. In fact thehypothetical Li2SO4 glass should have a Tg ofabout 456 K. (Tg/Tm) of Li2SO4 would thereforebe (456/1032)� 0.44 which is less than 2/3. Thismagnitude is consistent with the fact that Li2SO4 isnot a good glass former.

An examination of the heat capacities of thesesamples shows that in the CLSP series the mea-sured heat capacities are close to but less than theDulong Petit heat capacities. As we mentioned inthe earlier paragraph, in these Li2SO4 rich samplesthe glass transition temperatures should be close tothe ideal glass transition temperatures. Thiscloseness is a known property of ionic glasses inwhich there is very little frozen entropy or con-®gurational heat capacity [12,34]. Generally allvibrational modes are expected to be excited belowTg. But in the polymeric chains of metaphosphatespresent in these samples, excitation of vibrationalmodes should continue to occur even above theglass transition temperature [35] which would re-sult in the measured heat capacities being less thanthe corresponding 3nR heat capacities. Althoughpresence of polymeric P2 groups can increase vis-cosity of the liquids, Tg, and frozen entropy orcon®gurational heat capacity, the observed heatcapacities are consistently less than the 3nR heatcapacities and indicate that the vibrational modesin P2 and P1 are not fully excited.

We note that in pure metaphosphate glass(CLP1) DCp is largest. To examine if there is anycorrelation between the concentration of phos-phatic species and DCp we have plotted in Fig. 4DCp versus fraction of phosphatic species com-puted as [(P1 + P2)/(S + P1 + P2)] where S is theconcentration of SO2ÿ

4 ions. DCp shows a linearincrease with the fraction of phosphate units. TheDCps evaluated on the basis of DCps in CLP1 (puremetaphosphate) where S� 0 and the ratio is 1 isalso shown for comparison. This observation im-plies that the con®gurational heat capacity whichis associated with DCp at Tg is largely due to thephosphate in the composition. This association isunderstandable since the rotations on P±O±P

Fig. 10. Tg as a function of the fraction of sulphate units (ex-

pressed as S/(S + P1 + P2)). Line drawn shows the best least

squares ®t of the data points to the equation Tg�A + B [S/

(S + P1 + P2)], where A� 586.505 � 4.697, B�)120.523 �

18.279. Correlation coe�cient R2� 0.81. denotes the error

bar of data points, magni®ed 10 times.


bonds in P1 and P2 groups together with theirproportions in the present glasses ([P1 + P2] � S)appear as responsible for the correlation.

4.4. Infrared and Raman spectra

We suggested above that the sulphate±phos-phate and sulphate±sulphate interactions give riseto P±O±S and S±O±S linkages in transient inter-mediate species. We expect such species to bepresent in the samples and to have vibrationalmodes in the region of 733 and 802 cmÿ1 due to S±O±S and S±O±P stretching modes [36] particularlyin glass compositions rich in P1. The 730 to 750cmÿ1 features observed in P1 rich samples such asCP1, CP2, CLOP3, CLOP4, CLSP3 and CLSP4arise from such structural species. In fact forma-tion of DTP dimers decreases the symmetry of thesulphate species from Td to C3v and should makethe modes both Raman and IR active. However inthe Raman spectra of these glasses there is nofeature which can be attributed with con®dence toS±O±P or S±O±S linkages as in the IR spectra.

4.5. 31P HR-MASNMR and 7Li static NMRspectra

In the 31P MASNMR spectrum of CLP1, thereis a single resonance observed at a chemical shift of)21.7 ppm. This shift is consistent with the factthat only metaphosphate (P2 units) are expected tobe present. The surprising observation in theMASNMR spectra of the CLP series is the pres-ence of additional resonances in the pyrophos-phate regime with shifts of )3 to )4 ppm. Theamplitude of the peak at )3 to )4 ppm also in-creases with the mole fraction of Li2SO4. We notethis observation as consistent with the conjecturemade above that SO2ÿ

4 and P2 species may con-dense to produce DTP structures whose concen-trations we expect to increase with theconcentration of Li2SO4 and Li2O/P2O5 ratio. Infact, the formation of DTP species which in e�ectis a pyrophosphate half substituted by sulphateshould be common in all Li2SO4 containingglasses. If we designate the areas under P1 and P2

as A1 and A2 respectively then we expect

A1 / �P1� � c�P2��SO2ÿ4 �

or

A1=A2 / �P1� �P2�= � c�SO2ÿ4 �;

c being a constant. We have ignored the decreasein concentration of P2 due to the formation ofDTP. In the CLP series it is possible to plot A1/A2

(calculated by taking the ratio of the weights of theareas under the curves corresponding to P1 and P2

NMR signals) as a function of [SO2ÿ4 ] and deter-

mine c since �P1� �P2�= is zero in this series. We usethis c and calculate [P1]/[P2] + c[SO2ÿ

4 ] for all glasscompositions. The plot shown in Fig. 11 shows alinear increase of A1/A2 with [P1]/[P2] + c[SO2ÿ

4 ]lending strength to the proposition that DTP isformed in the structure. (The correlation coe�-cient is 0.79.)

From the 7Li static NMR spectra it is seen thatincrease in Li2SO4 is associated with a decrease inFWHM or dipolar and other broadening e�ects.This decrease can be readily associated with themotion of Li� ions which, as in a liquid, eliminatesdipolar broadening. In CLP4, the FWHM is 2.33kHz. The lithium ion motion is evidently in¯u-enced by the concentration of SO2ÿ

4 ions. Motion

Fig. 11. Fraction of P1 and P2 as calculated from 31P MAS-

NMR resonances as a function of the structural species present

in the glass. Line drawn shows the best least squares ®t of the

data points to the equation A1/A2�M + N [P1/P2 + cSO2ÿ4 ],

where M� 0.564 � 0.095, N� 0.355 � 0.057. Correlation co-

e�cient R2� 0.79. denotes the error bar of data points,

magni®ed 10 times.


of Li� ions is also in¯uenced by the P1 units in anopposite manner. This e�ect is likely to be aconsequence of a more e�ective packing of the[POO1=2O2]2ÿ units. This can be seen by compar-ing the FWHM of 7Li signals in CLP4(FWHM� 2.33 � 0.01 kHz) and CLSP4 (FWHM� 2.99 � 0.01 kHz), where there is a de®nite indi-cation that formation of pyrophosphate units tendto decrease the mobility of lithium ions. The mo-bility can decrease due to a deepening of the po-tential well when pyrophosphate ions are presentin the lithium coordination sphere because theaverage (formal) charge of the oxygen associatedwith a pyrophosphate group is 4/7 while an oxygenassociated with sulphate ion carries a formalcharge of only 4/8.

5. Conclusions

We have shown that in Li2SO4±Li2O±P2O5

glasses, phosphate and sulphate ions are in ran-dom close packing but with structural voids in-creasing in proportion to the phosphate. Voidsmay be of approximately the size of oxide ions andform a percolating network.

There is evidence for the formation of dithio-phosphate type of units in the structure and theymay play a role in both glass structure and trans-port properties. NMR spectroscopy con®rms sucha structural model and also suggests that lithiumion motion is coupled to the proportion of sul-phate ions in the glass.

Acknowledgements

The authors are thankful to the Commission ofthe European Communities for ®nancial support.One of the authors (M.G.) is grateful to theCouncil for Scienti®c and Industrial Research(CSIR), India for a senior research fellowship.

References

[1] H.L. Tuller, D.P. Button, D.R. Uhlmann, J. Non-Cryst.

Solids 40 (1980) 93.

[2] C.A. Angell, Solid State Ionics 9&10 (1983) 6.

[3] C.A. Angell, Chem. Rev. 90 (1990) 523.

[4] A.R. Kulkarni, H.S. Maiti, A. Paul, Bull. Mater. Sci. 6

(1984) 201.

[5] K.J. Rao, M. Ganguli, Handbook of Solid State Batteries

and Capacitors, World Scienti®c, Singapore, 1995.

[6] A. Levasseur, B. Cales, J.M. Reau, P. Hagenmuller, Mater.

Res. Bull. 14 (1979) 921.

[7] J.P. Malugani, G. Robert, Mater. Res. Bull. 14 (1979)

1075.

[8] D.E. Day, J. Non-Cryst. Solids 21 (1976) 343.

[9] B. Carette, M. Ribes, J.L. Souquet, Solid State Ionics 9&10

(1983) 735.

[10] J.R. Van Wazer, Phosphorus and its Compounds, vol. I,

Wiley/Interscience, New York, 1958.

[11] S.W. Martin, Eur. J. Solid State Inorg. Chem. 28 (1991)

163.

[12] P.S.L. Narsimhan, K.J. Rao, J. Non-Cryst. Solids 27

(1978) 225.

[13] E.I. Kamitsos, M.A. Karakassides, G.D. Chryssikos, J.

Phys. Chem. 90 (1986) 4528.

[14] D.E.C. Corbridge, E.J. Lowe, J. Chem. Soc. (1954) 493.

[15] R.M. Almeida, J.D. Mackenzie, J. Non-Cryst. Solids 40

(1980) 535.

[16] J.J. Hudgens, S.W. Martin, J. Am. Ceram. Soc. 76 (7)

(1993) 1691.

[17] R.K. Brow, D.R. Tallant, J.J. Hudgens, S.W. Martin, A.D.

Irwin, J. Non-Cryst. Solids 177 (1994) 221.

[18] G.J. Exarhos, P.J. Miller, W.M. Risen Jr., J. Chem. Phys.

60 (11) (1974) 4145.

[19] S. Krimi, A. El Jazouli, L. Rabardel, M. Couzi, I.

Mansouri, G. El Flem, J. Solid State Chem. 102 (1993)

400.

[20] C. Nelson, D.R. Tallant, Phys. Chem. Glasses 25 (2) (1984)

31.

[21] M. Tatsumisago, Y. Kowada, T. Minami, Phys. Chem.

Glasses 29 (2) (1988) 63.

[22] P. Tarte, Spectrochim. Acta 20 (1964) 238.

[23] R.K. Brow, R.J. Kilpatrick, G.L. Turner, J. Non-Cryst.

Solids 116 (1990) 39.

[24] M. Villa, K.R. Carduner, G. Chiodelli, J. Solid State

Chem. 69 (1987) 19.

[25] S. Prabakar, K.J. Rao, C.N.R. Rao, Chem. Phys. Lett. 139

(1987) 96.

[26] W. Muller-Warmuth, H. Eckert, Phys. Reports 88 (1982)

91.

[27] B.C. Sales, J.O. Ramey, L.A. Boatner, J.C. McCallum,

Phys. Rev. Lett. 62 (1989) 1138.

[28] S. Ananthraj, K.B.R. Varma, K.J. Rao, Mater. Res. Bull.

21 (1986) 1369.

[29] D.E.C. Corbridge, The Structural Chemistry of Phospho-

rus, Elsevier, Amsterdam, 1974.

[30] R. Zallen, The Physics of Amorphous Solids, Wiley, New

York, 1983.

[31] S. Muthupari, K.J. Rao, J. Phys. Chem. 98 (1994) 2646.

[32] R.T. Sanderson, Polar Covalence, Academic Press, New

York, 1983.


[33] M.H. Cohen, D. Turnbull, J. Chem. Phys. 31 (1959) 1164.

[34] H.G. K. Sundar, K.J. Rao, J. Chem. Soc. Farday Trans. I

76 (1980) 1617.

[35] U. Selvaraj, K.J. Rao, J. Non-Cryst. Solids 104 (1988) 300.

[36] R.G. Brown, S.D. Ross, Spectrochim. Acta 28A (1972)

1263.


Documents

Studies of ternary Li2SO4–Li2O–P2O5 glasses