This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 10885--10894 10885

Cite this: Phys. Chem.Chem.Phys.,2013,15, 10885

Predictions of particle size and lattice diffusionpathway requirements for sodium-ion anodes usingg-Cu6Sn5 thin films as a model system†

Loıc Baggetto,*a Jean-Claude Jumas,b Joanna Gorka,c Craig A. Bridgesc andGabriel M. Veith*a

Geometrically well-defined Cu6Sn5 thin films were used as a model system to estimate the diffusion

depth and diffusion pathway requirements of Na ions in alloy anodes. Cu6Sn5 anodes have an initial

reversible capacity towards Li of 545 mA h g�1 (Li3.96Sn or 19.8 Li/Cu6Sn5), close to the theoretical

586 mA h g�1 (Li4.26Sn), and a very low initial irreversible capacity of 1.6 Li/Cu6Sn5 (Li0.32Sn). In contrast,

the reaction with Na is limited with a reversible capacity of 160 mA h g�1 compared to the expected

516 mA h g�1 (Na3.75Sn). X-ray diffraction and 119Sn-Mossbauer spectroscopy measurements show that

this limited capacity likely results from the restricted diffusion of Na into the anode nanoparticles and

not the formation of a low Na-content phase. Moreover, our results suggest that the Z-Cu6Sn5 alloy

should have optimized particle sizes of nearly 10 nm diameter to increase the Na capacity significantly.

An alternative system consisting of a two-phase mixture of Cu6Sn5 and Sn of nominal composition

‘Cu6Sn10’ has been studied and is able to deliver a larger initial reversible storage capacity of up to

400 mA h g�1. Finally, we have demonstrated that the presence of Cu in Cu6Sn5 and ‘Cu6Sn10’

suppresses the anomalous electrolyte decomposition normally observed for pure Sn.

Introduction

Recently there has been a renaissance of interest in manyelectrochemical systems other than lithium-ion batteries (LIBs)due to the well-known materials, energy density and safetylimitations of LIBs.1 Alternative technologies such as lithium–sulfur, lithium–air, and sodium-ion batteries are of interest dueto the availability of materials, i.e. S, O2, Na, and the muchhigher energy densities available for the S and O2 systems.1 Thesodium-ion concept is particularly drawing a great deal ofattention thanks to the low cost and natural abundance ofNa, as well as an attractive Na/Na+ redox potential only 0.3 Vabove that of Li/Li+. Compared to Li-ion, Li–S and Li–air, there arerelatively few studies on the fundamental properties of Na-ioncathodes and anodes.2–4 A few notable exceptions on the anode

side have been reported, such as for pure Sb (B0.65 V,660 mA h g�1),5 pure Sn (B0.3 V, 847 mA h g�1)6,7 or Cu2Sb(B0.55 V, 250 mA h g�1).8

Metallic Sn has large Na and Li storage capacities6,7 butsuffers from structural fatigue due to repeated large volumechanges during cycling. Intermetallic M–Sn materials are alternativeanodes with potentially high storage capacity and low voltageoperation similar to Sn. Moreover, due to the presence of inactivemetal, the volume expansion is expected to be reduced as comparedto pure Sn. Previously, Cu6Sn5 has been explored in detail for Li-ionbatteries using electrochemical methods, X-ray diffraction (XRD),119Sn Mossbauer spectroscopy, X-ray photoelectron spectroscopy(XPS) and transmission electron microscopy (TEM) combinedwith electron diffraction (ED), and its reaction mechanismwas found to proceed via Li-ion insertion/Cu displacement/extrusion.9–17 Assuming full conversion into the end-memberLi21+5/16Sn5 (equal to Li4.2625Sn, abbreviated as Li4.26Sn),18

Cu6Sn5 anodes have a high theoretical capacity of 586 mA h g�1.As Sn can electrochemically alloy with Na up to a composition ofNa15Sn4,6,7 the theoretical capacity of Cu6Sn5 for Na-ion batteries is516 mA h g�1.

In this work we have investigated the differences in Li-ionand Na-ion reactions in metallic copper tin (Cu6Sn5) electrodes

a Materials Science and Technology Division, Oak Ridge National Laboratory,

Oak Ridge, TN 37831, USA. E-mail: baggettol@ornl.gov, veithgm@ornl.govb Institut Charles Gerhardt, Universite Montpellier II, 34095 Montpellier Cedex 5,

Francec Chemical Sciences Division, Oak Ridge National Laboratory, Oak Ridge,

TN 37831, USA

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c3cp51657a

Received 18th April 2013,Accepted 24th April 2013

DOI: 10.1039/c3cp51657a

www.rsc.org/pccp

PCCP

PAPER

Publ

ishe

d on

02

May

201

3. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Irvi

ne o

n 31

/10/

2014

00:

20:0

3.

View Article OnlineView Journal | View Issue

10886 Phys. Chem. Chem. Phys., 2013, 15, 10885--10894 This journal is c the Owner Societies 2013

in order to develop a predictive model for electrode geometriesneeded to obtain high capacity and reversible anodes. One ofthe critical challenges for Na-ion intercalation systems isdesigning electrodes which will facilitate the rapid and reversibletransfer of the larger Na-ion (IV-coordinate ionic radius B 0.99 Å)compared to the much smaller Li-ion (IV-coordinate ionic radiusB 0.59 Å).19 To study the potential of Cu6Sn5 as the Na-ionanode, we have fabricated highly crystalline Cu6Sn5 thin films bymagnetron sputtering to serve as a model system. The presentarticle reports the reaction mechanism of the electrode materialCu6Sn5 during cycling in Li-ion and Na-ion cells, as studied bymeans of galvanostatic electrochemical measurements, ex situXRD and 119Sn Mossbauer spectroscopy. Moreover, the apparentdiffusion lengths and the diffusion pathways for Li and Na ionsinside Z-Cu6Sn5 structure are investigated by detailed structureanalyses and predictions of the utilized electrode volume.Finally, we present experimental evidence for the design ofanodes with a higher storage capacity relying on Sn–Cu6Sn5

mixtures.

Results and discussion

The morphology and XRD data collected for the startingZ-Cu6Sn5 thin films are presented in Fig. 1. As evidenced bySEM inspection (Fig. 1a) the films are composed of secondary

particles on the order of 200 nm in size, which are in turncomposed of agglomerated smaller nanocrystallites below100 nm. XRD data show that apart from reflections of the Cufoil, which has a strong (022) preferred orientation (B741 2y),the diffraction peaks are attributed to Z-Cu6Sn5 (P63/mmc),indicating that sputtering results in the formation of the hightemperature Z-Cu6Sn5 phase (transition from Z0 to Z at189 1C20,21). The formation of the Z-Cu6Sn5 phase is clearlyevidenced by the absence of additional diffraction lines normallyobserved for the monoclinic distortion Z0-Cu6Sn5 (compareFig. S1 and S2, ESI†).

The (111) Cu and (012) Cu6Sn5 diffraction peaks partiallyoverlap at 43.51 2y; however, due to the strong (022) orientationof Cu, the estimated contribution of the (111) Cu peak is onlyabout 25% of the total peak intensity. Due to the air sensitivenature of the lithiated and sodiated phases a protective Kaptontape was used for the ex situ XRD measurements. As illustratedhere on the pristine material (Fig. 1b), the Kapton tape causes a50% reduction in the measured intensity and produces a broadhump at around 201 2y (described in more detail later).

Refining the XRD pattern of the pristine film (Fig. S2, ESI†)yields lattice parameters of a = b = 4.2033 (�0.0002) Å and c =5.0914 (�0.0007) Å for Z-Cu6Sn5, in good agreement withexpected values.21 There is a small contribution from pureb-Sn, as also found in the target material (Fig. S1, ESI†), whichis estimated to be about 2.5 wt% based on the Rietveld analysis(Fig. S2, ESI†). Using the Scherrer equation,22 we calculate theCu6Sn5 crystalline domain size to be nearly 50 nm in diameter,in good agreement with the SEM data, Fig. 1. It is interesting tohighlight that post-annealing or substrate biasing were notnecessary to obtain highly crystalline films. This is in directcontrast to earlier reports for sputtered13 and co-sputtered15

thin films of Cu6Sn5. The starting target material, substratematerial, target–substrate distance, deposition pressure andpower all importantly influence the resulting film properties,which are found here to be very good from the point of view ofcrystallinity.

The Cu6Sn5 films were cycled against Li and compared toreports in the literature to gauge the suitability of this materialfor electrochemical studies (Fig. 2). The room temperaturepotential profiles for the Li-ion reaction are very similar to thosereported previously, which consists of conjugated plateaus at0.4/0.8 and 0.1/0.4 V during discharge–charge.9–17 The initialdischarge capacity is equal to 589 mA h g�1, which is very closeto the theoretical value (586 mA h g�1) expected for the formationof Li4.26Sn. The reversible (charge) capacity equals 545 mA h g�1,equivalent to Li3.96Sn, and is much higher than for previousreports. Indeed, the initial irreversible capacity is only 1.6 Li/Cu6Sn5 (Li0.32Sn), which is significantly lower than the best valueof about 6 Li/Cu6Sn5 previously measured on composite powderelectrodes.12,14 This large improvement is attributed to fasterkinetics as well as better mechanical properties of the filmnanoparticle electrodes with respect to the composite micron-sized powder electrodes.

During the second cycle, the discharge capacity remainsnearly unchanged (584 mA h g�1), whereas the charge capacity

Fig. 1 Characterization of as-prepared Z-Cu6Sn5 thin films: (a) SEM photographand (b) XRD pattern. (*) and (#) indicate Cu and b-Sn respectively. All other peaksresult from Z-Cu6Sn5.

Paper PCCP

Publ

ishe

d on

02

May

201

3. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Irvi

ne o

n 31

/10/

2014

00:

20:0

3.

View Article Online

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 10885--10894 10887

slightly decreases to 526 mA h g�1. The reversible capacity forLi-ion reaction (Fig. 2, inset) fades rapidly after 10 cycles. Theloss of capacity originates from the large volume expansion/contraction during cycling from 0 to 2 V, as generally measuredfor Li–Cu6Sn5 when cycling from 0 to 1.2 V.11 The capacityretention can be significantly improved when cycling is limitedbetween 0.2 to 1 V.11,15

The local structure of Sn atoms has been studied by means of119Sn Mossbauer spectroscopy (Fig. 3) in the fully (de)lithiatedstates. The pristine material has a similar signature to that ofCu6Sn5 powder12,23 characterized by a doublet with an IS of2.22 mm s�1 and a quadrupole splitting (QS) of 0.61 mm s�1

(w2 = 0.48), confirming the structure identified by XRD (Fig. 1b).The Mossbauer spectra for fully discharged (0 V) and fully charged(2 V) Li–Cu6Sn5 electrodes are in excellent agreement with formerreports.12,23 The fully lithiated electrode is represented by a singletwith an IS of 1.86 mm s�1 and a QS of 0.40 mm s�1 (w2 = 0.61)which are very close to the hyperfine parameters obtained forfully lithiated Cu6Sn5 powder electrodes.12 These values alsomatch very well those of the Li–Sn end member compound ofcomposition Li4.26Sn, which is characterized by an IS of 1.83 mm s�1

and a QS of 0.31 mm s�1.24 At full charge (2 V), the spectrum ischaracterized by an IS of 2.20 mm s�1 and a QS of 0.57 mm s�1 (w2 =0.57). The values obtained for the Cu6Sn5 powder electrodes at 1 or2.6 V12 indicated an IS of 2.12 mm s�1 and a QS of 0.65 mm s�1,which can be indicative of remaining Li in the powder electrode andthe absence of the full formation of Cu6Sn5. Here in the case of ourthin films, the measured IS is much closer to that of the pristinematerial (Fig. 3). These Mossbauer spectroscopy results therebyfurther confirm that these Cu6Sn5 thin films, which have a muchlower initial irreversible capacity (Fig. 2) than the powder samples,12

are capable of fully reversibly reacting with Li.This improvement is likely related to better reaction

kinetics, especially solid-state diffusion of Li, and mechanicalproperties as compared to composite micron-sized powder

electrodes. Furthermore, nearly the entire electrode is electro-chemically active, indicating that under the experimental conditionsLi ions can diffuse through the bulk of the polycrystalline film.The XRD, Mossbauer spectroscopy, and electrochemical resultsdemonstrate that these thin films are an excellent model systemfor Z-Cu6Sn5.

Cu6Sn5 thin film electrodes, from the same batch of samplesused for the Li experiments described above, were tested versusthe Na-ion reaction (Fig. 4). The room temperature dischargeprofile is dominated by a few small plateaus at around 0.8–0.9 Vand a slope leading to a more pronounced plateau near 0.45 V,followed by a slope and a plateau at low voltages of around0.05 V. During charge (desodiation), several plateaus areobserved with a wide plateau near 0.19 V and a narrow plateauat 0.25 V, followed by plateaus around 0.58 V and near 1.9 V.For pure Sn electrodes, the reaction with Na is mainly repre-sented during discharge by plateaus near 0.45, 0.19, 0.07and 0.04 V, and during charge by plateaus at 0.13, 0.26, 0.53,0.62 and 1.85 V.6 The similar plateau positions for Cu6Sn5 andSn suggest that Sn from Cu6Sn5 can alloy with Na. It isworthwhile to emphasize that the anomalous decompositionof the electrolyte observed for pure Sn during the seconddischarge as a plateau at 1.2 V (ref. 6) is completely suppressedhere when Cu is present, similarly to what was reported earlierfor Li–Cu6Sn5.25

It is clear that these Cu6Sn5 films do not exhibit a largereversible storage capacity towards Na (180 mA h g�1) com-pared to the theoretical capacity (516 mA h g�1) even whenemploying low currents (7.9 mA cm�2 equivalent to C/20).

Fig. 2 Electrochemical properties of Z-Cu6Sn5 thin film electrodes. First (black)and second (grey) electrochemical potential profiles for Li-ion reaction of 1 mm thickelectrodes, with cycle life plots as insets. Current is 7.9 mA cm�2, corresponding toabout C/50, respectively.

Fig. 3 119Sn Mossbauer spectroscopy data for Li-Cu6Sn5 pristine, fully discharged(0 V) and fully charged (2 V) electrodes. The hyperfine parameters (isomer shift (IS),quadrupole splitting (QS) and line width (LW)) are given in the figures.

PCCP Paper

Publ

ishe

d on

02

May

201

3. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Irvi

ne o

n 31

/10/

2014

00:

20:0

3.

View Article Online

10888 Phys. Chem. Chem. Phys., 2013, 15, 10885--10894 This journal is c the Owner Societies 2013

Furthermore, cycling experiments at 1.6 mA cm�2 (equivalent toC/100, cf. Fig. S3, ESI†) did not yield a higher reversiblecapacity, and using thinner films (200 nm, not presented) alsodid not help to increase the storage capacity. These results arelikely indicative that limited diffusion within the primaryelectrode particle is the main hindering factor. The smallquantity of b-Sn present in the starting films does not signifi-cantly contribute to the capacity; nonetheless, due to the highergravimetric storage capacity of Sn compared to Cu6Sn5, it canbe calculated that the intrinsic gravimetric capacity for Cu6Sn5

towards Na is equal here to 160 mA h g�1.To explore the origin of this low Na storage capacity,

changes in crystallographic structure have been investigatedusing XRD during the first cycle (Fig. 5). For convenience, thestarting material covered with Kapton (Fig. 1b) is included hereas reference.

During discharge (sodiation, cf. Fig. 5a) the intensity of thediffraction peaks for Sn and Cu6Sn5 decreases. Remarkably, Snis still present at the low potential of 0.17 V and the signal ofCu6Sn5 has decreased but has not vanished at 0 V. Thepersistence of Cu6Sn5 at 0 V has been measured for cellsdischarged at the very low current of about 3.9 mA cm�2 (C/40)and also for cells short circuited over 40 h, and corroboratesthe relatively low storage capacity discussed earlier. XRD datacollected for the 0 V electrodes clearly evidence the formation ofNa15Sn4, which indicates that Sn from Cu6Sn5 can alloy with Nato form the end-member of the Na–Sn system.6 During charging(Fig. 5b) diffraction lines for Sn become visible at 0.8 V, as alsofound for Sn electrodes at this potential.6 The weight ratiobetween Sn and Cu6Sn5 obtained from Rietveld refinements ofthe XRD pattern at 2 V is much larger (0.6 : 5) than for thepristine electrode (0.15 : 5), but lower than the expected 1.2 : 5ratio if 31% (160 mAh g�1) of Cu6Sn5 would fully convert into Snirreversibly. These results thus indicate that Cu6Sn5 can bepartially reconstructed from Cu and Sn.

The changes in lattice parameters of the Z-Cu6Sn5 structureduring discharge with Na, obtained using Rietveld refinementsof the XRD data, are presented in Fig. 6. No substantial changesare observed across the range of composition for the c axis, andonly a slight trend towards an increase of a = b is apparentlyobtained. Given the experimental resolution, we remain verycautious as to whether the apparent slight increase of the a = blattice parameter is real, and these results may instead indicatethat the bulk of the Cu6Sn5 is not reacting with Na.

119Sn-Mossbauer measurements were performed to corroboratethe XRD results. To allow the reaction to proceed to its maximumextent at room temperature, six Na–Cu6Sn5 electrodes were shortcircuited for 72 hours to ensure maximum sodiation, and three of

Fig. 4 Electrochemical properties of Z-Cu6Sn5 thin film electrodes. First (black)and second (grey) electrochemical potential profiles for Na-ion reaction of 1 mmthick electrodes, with cycle life plots as insets. Current is 7.9 mA cm�2, corres-ponding to about C/20.

Fig. 5 Changes in crystallographic structure characterized during the firstNa-ion electrochemical cycle using XRD. (*), (#), (^) represent Cu, Sn and Na15Sn4,respectively. All other peaks result from Z-Cu6Sn5.

Fig. 6 Lattice parameters of the Z-Cu6Sn5 (P63/mmc) phase obtained fromRietveld refinements of the XRD patterns (Fig. 5).

Paper PCCP

Publ

ishe

d on

02

May

201

3. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Irvi

ne o

n 31

/10/

2014

00:

20:0

3.

View Article Online

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 10885--10894 10889

the cells were subsequently charged to 2 V (Fig. 7). To ensure enoughabsorption during the Mossbauer spectroscopy measurements, eachset of three identical electrodes was measured simultaneously. Notethat the low absorption for the spectrum obtained on the pristinematerial originates from the measurement of a single electrode(see the Experimental section). The short-circuit electrode materialcan be represented by a doublet with an IS of 2.15 mm s�1 and a QSof 0.69 mm s�1 (w2 = 0.58, Fig. S4, ESI†).

Based on the XRD results (Fig. 5), fitting the spectrum witha singlet and a doublet is most reasonable to account for theco-existence of Na15Sn4 and Cu6Sn5, respectively (Fig. 7). Indeedthe insertion of Na ions into Sn and the partial reaction of Naions with Sn resulting from Cu6Sn5 are expected to lead to theformation of Na15Sn4, which is normally characterized by asinglet with an IS of 2.14 mm s�1.6 Moreover, the presence of arelatively large peak splitting confirms the presence of Cu6Sn5,as also evidenced by XRD (Fig. 5a). The corresponding fitincluding the co-existence of Na15Sn4 (IS of 2.15 mm s�1) andCu6Sn5 (IS of 2.15 mm s�1 with a QS of 0.69 mm s�1) yields abetter fit (w2 = 0.53). Several alternative fits were tried on thespectra; one of the fits, including the presence of b-Sn, isreported in Fig. S5 (ESI†). This fit accounts for the minorfraction of b-Sn in the pristine material and provides a slightlybetter fit (w2 = 0.47 against 0.48), as suggested by the smallintensity asymmetry visible as a slightly more intense peak atabout 2.55 mm s�1. In the case of the discharged electrode, theaddition of b-Sn to the aforementioned signals of Na15Sn4 and

Cu6Sn5 does not improve the fit (w2 = 0.54) and yields an IS of2.65 mm s�1 which is substantially larger than the valueexpected for b-Sn.6,24

The spectrum of the charged electrode (2 V) is characterizedby a broad peak properly described by a singlet with an IS of2.20 mm s�1 and a QS of 0.63 mm s�1 (w2 = 0.76), and matcheswell the hyperfine parameters of Cu6Sn5. Although evidencedby XRD (Fig. 5), no b-Sn could be fitted for the spectrum of thecharged electrode (Fig. S5, ESI†) as the spectrum does not showthe presence of an additional signal near 2.55 mm s�1. Sincethe amount of b-Sn is expected to be larger in the chargedelectrode than in the pristine material (see discussion of Fig. 5),and given the much lower Lamb–Mossbauer fraction of b-Sncompared to Cu6Sn5 at room temperature,23,24 the presence ofb-Sn in the alternative fits shown in Fig. S5 (ESI†) has not beenincluded in Fig. 7. In order to determine more accuratelythe materials Mossbauer hyperfine parameters, using lowertemperatures to increase the recoil-free absorption and/orusing a larger amount of active material would be necessary.

Given the large duration of time during which the electrodeswere short-circuited, the amount of unreacted Cu6Sn5 in theelectrodes is expected to be the same. Consequently, comparingthe response attributed to Cu6Sn5 at 0 V and at 2 V (aftercharging) can allow quantification if some Cu6Sn5 was formedduring Na-ion removal. Given the much larger maximumabsorption of the charged electrodes (6.5%) compared to thedischarged electrodes (3%), we suspect that the partial reformationof Cu6Sn5 from Cu and Sn occurs during charge (Na-ion removal),consistent with the XRD results. The reaction mechanism ofCu6Sn5 with Li10,11 was reported as a multi-step intercalation/Cudisplacement/extrusion reaction, according to:

10Li + Cu6Sn5 - 5Li2CuSn + Cu (1)

xLi + Li2CuSn-Li2+xCu1�ySn + yCu (2)

with 0 o x o 2.26 and 0 o y o 1.The first step was also proposed as the sum of two steps,17

according to:

x0Li + Cu6Sn5 - Lix0Cu6Sn5 (10)

10 � x0Li + Lix0Cu6Sn5 - 5Li2CuSn + 6Cu (10)

The formation of Li2CuSn was evidenced based on thepresence of several broad peaks matching the strongest diffrac-tion lines of Li2CuSn10 or electron diffraction (ED) data.17 Theformation of LixCu6Sn5 solid solution is supported by the EDdata17 and is also suggested by Mossbauer spectroscopy data23

showing no significant changes for x0 = 4.5.For the reaction towards Na, no additional XRD peaks

representing an intermediate phase (Fig. 5) have been measured,which indicates that the potential intermediate NaxCu6�ySn5

phase is amorphous or is immediately converted into Cu andNaxSn. XRD patterns collected at 0.7 and 0.46 V before the reactionof Sn with Na6 do not indicate the formation of a new phase(Fig. 5) and also do not reveal substantial changes in the latticeparameters of Cu6Sn5 (see for low Na/Cu6Sn5 ratios in Fig. 6).

Fig. 7 119Sn Mossbauer spectroscopy data for Na-Cu6Sn5 pristine, fully dis-charged (short-circuited, 0 V) and fully charged (2 V) electrodes. The hyperfineparameters (isomer shift (IS), quadrupole splitting (QS) and line width (LW)) aregiven in the figures.

PCCP Paper

Publ

ishe

d on

02

May

201

3. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Irvi

ne o

n 31

/10/

2014

00:

20:0

3.

View Article Online

10890 Phys. Chem. Chem. Phys., 2013, 15, 10885--10894 This journal is c the Owner Societies 2013

As the Cu6Sn5 lattice parameters remain fairly constant (Fig. 6),we assume that the crystallites are composed of an unreactedcore surrounded by a fully reacted shell. Indeed, the sodiationreaction down to 0.46 V consumes only 2 Na/Cu6Sn5 and thefollowing sodiation reaction proceeds at potentials much lower(Fig. 4) than for pure Sn.6 In addition, as the overall storagecapacity is only 160 mA h g�1 instead of the expected 516 mA h g�1,even when using lower currents (Fig. S3, ESI†) or thinner layers(not presented), we suspect that the solid-state diffusion of Nawithin Cu6Sn5 is fairly slow.

The capacity of 160 mA h g�1 corresponds to the reactionof 31% of the electrode volume, which is equivalent to thereaction of the outer 3 nm of 25 nm radius spherical crystal-lites, based on the average domain size estimated from the XRDdata. Thus, the NaxCu6�ySn5 domains possibly formed from theouter 3 nm Cu6Sn5 may not be simply detected by XRD due totheir small expected domain size and low crystallinity. More-over, one can expect a strong concentration gradient of Na fromthe surface to the core of the particles due to sluggish diffusion.A strong concentration at the surface exceeding the critical Naconcentration proceeding with the formation of NaxSn and Cumay also explain the absence of intermediate crystalline phases(Fig. 5) and the absence of substantial changes in the latticeparameters of Cu6Sn5 in the XRD results (Fig. 6). In summary,the results may indicate that the reaction proceeds on the outershell of the grains whereas the large majority of the grains coreis unreacted. This is supported by the persistence of the Cu6Sn5

XRD (Fig. 5) and Mossbauer spectroscopy (Fig. 7) signals at 0 Vand the absence of other diffraction lines resulting fromintermediate phases. The material on the outer shell may firstconvert into an amorphous/nanocrystalline product by a combinedintercalation/Cu extrusion reaction until all Cu–Sn bonds arebroken and only Na–Sn and Cu–Cu bonds are present.

The structure of Z-Cu6Sn5 is composed of Sn atoms having atrigonal prismatic coordination while Cu atoms are coordinated inoctahedral sites.9,10 A qualitative estimate of the most favorablediffusion pathways is highlighted in Fig. 8. For the sake ofcomparison between Na and Li, the following discussion is basedupon an analysis of void spaces present in the pristine Z-Cu6Sn5

material, before possible intercalation and expansion of the lattice.Void spaces are considered here to be the space between occupiedSn and Cu sites along the Na diffusion pathways. Na and Li areexpected to diffuse through the largest void spaces, which connectin the structure along two potential diffusion pathways: as zigzagsalong the c axis (Fig. 8a) or in the form of tunnels (Fig. 8b), with thelatter likely being more favorable. Assuming that a solid-solution ofLixCu6Sn5 forms,17 Li is expected to occupy the sites depicted inFig. 8b and a, with the latter corresponding to the site described inref. 17. For both diffusion pathways, the closest distance betweenthe Sn sites is 3.51 Å and is 4.19 Å between Cu sites. Consideringionic radii of 0.99 Å or 1.02 Å for Na+ in IV-fold or VI-foldcoordinated configurations,19 respectively, it is clear that whilethe volume of the void available for diffusion between the occupiedSn–Sn sites may be sufficient for Li, the available void space isrelatively limited for Na ion diffusion. Indeed Li has much smallerionic radii of 0.59 Å or 0.76 Å for IV-fold or VI-fold coordinated

configurations,19 respectively. In turn, Li should possess muchfaster kinetics and the storage capacity obtained with Li can beexpected to be much higher, as evidenced earlier. For Na, however,we suspect that Na ion diffusion is much more severely limitedand may induce enough strain to effectively block the diffusionchannels. As a result one could expect a fairly small diffusion depthand low storage capacity, as measured here.

Although the two diffusion pathways discussed above bothpossess the same void restriction imposed by Sn–Sn atomsseparated by 3.51 Å, the zigzag configuration is expected topossess a higher energy barrier for diffusion. In this configu-ration, atom hopping occurs through the center of rhombi of2.733 Å side length, 3.510 Å Sn–Sn diagonal and 4.190 Å Cu–Cudiagonal depicted in Fig. 9 in the [100] and [430] projections.The higher energy barrier results from the longer jump distancebetween sites, which is approximately 4.445 Å (Fig. 9). Thisdistance can be measured when projecting the structure alongthe [110] direction, which is shown here for convenience with aslight tilt leading to a [430] projection. On the other hand, thetunnels formed by Sn atoms (Fig. 8b), also visible in the [111]projection in Fig. 9, offer a shorter jump distance of about3.295 Å, as visible in the [1%10] projection.

The comparison between Li and Na reactions in Z-Cu6Sn5

helps to understand which structures would be more favorableas Na-ion insertion materials. These materials should ideallycombine 2D or 3D diffusion pathways, voids which are sufficientlylarge to allow Na ion diffusion (quantifiable for example by a large

Fig. 8 Projections of Z-Cu6Sn5 P63/mmc hexagonal structure along (a) the [110]direction with a slight rotation around the c axis and (b) along the [111] direction.Cu and Sn atoms are represented in orange and grey, respectively. The dashedarrows in (a) depict the zigzag diffusion pathways along the c axis. The circleplaced in the center of a channel in (b) represents an atom inside the 1D tunneldiffusion pathway. For simplicity, the Sn vacancies (1/6) are not represented.

Paper PCCP

Publ

ishe

d on

02

May

201

3. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Irvi

ne o

n 31

/10/

2014

00:

20:0

3.

View Article Online

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 10885--10894 10891

ratio of the void volume to alkali atom volume) and/or a structurethat can expand upon Na-ion insertion to favor Na-ion transport.These beneficial properties are likely to be found in cubic or layeredstructures. A counter example of expected undesired structure isthe Z0-Cu6Sn5 monoclinic C2/c parent discussed previously. Thisstructure, although applicable for Li-ion applications,11,14,23

is composed of open and blocked tunnels (Fig. 10). The closestSn–Sn distances in the open tunnels are between 3.35 and 3.40 Å,which represents a significant decrease in volume for the voidcompared to Z-Cu6Sn5.

Assuming that grain boundary diffusion is significantlygreater than intra-particle diffusion, and based upon the50 nm estimate for the average crystalline domain size from XRDdata, on average the Li ions diffuse through at least 25 nm of Cu6Sn5.Oppositely, the estimated penetration depth of Na ions insideCu6Sn5 crystallites is approximately 3 nm. In turn, we have calculatedthe evolution of the expected Na-ion capacity as a function of thecrystallite radius for spherical particles of Cu6Sn5 (Fig. 11). In orderto highlight the beneficial improvements of better solid-state diffu-sion of Na ions, we have also included the predicted active materialvolume ratios for larger diffusion depths of up to 30 nm. It is clearthat a restricted diffusion depth (e.g. here 3 nm) can severely limitthe practical storage capacity (160 mA h g�1), which falls off veryquickly for larger crystallite sizes. However, particles composed ofsmaller crystallites would significantly increase the reversible storagecapacity when assuming the same solid-state diffusion kinetics.

For example, an advantageous 350 mA h g�1 capacity for acrystallite radius of 10 nm could be achieved (Fig. 11). Anotherapproach to increase the storage capacity is to select chemistrieswhich allow faster Na-ion diffusion kinetics. It can be seen thatsuch materials having, for instance, 2 to 3 times higher diffusiondepths, 6 or 9 nm, respectively, could allow accessing 55 to 70%of the volume of 25 nm radius crystallites. Larger improvementscould be obtained if the crystallite radius would be decreased to

Fig. 9 Projections of Z-Cu6Sn5 P63/mmc hexagonal structure along the [100], [430], [111] and [1 %10] directions. Cu and Sn atoms are represented in orange and grey,respectively. The dashed arrows depict the diffusion pathways between potential sites highlighted by dashed open circles and separated by the distances indicated inthe figures.

Fig. 10 Projection of the Z0-Cu6Sn5 C2/c monoclinic structure along the [110]direction. Cu and Sn atoms are represented in orange and grey, respectively. Thecircle placed in the center of a channel represents an atom inside the 1D tunneldiffusion pathway. Note that some channels (for example neighboring the circle)are partly blocked by Cu atoms.

PCCP Paper

Publ

ishe

d on

02

May

201

3. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Irvi

ne o

n 31

/10/

2014

00:

20:0

3.

View Article Online

10892 Phys. Chem. Chem. Phys., 2013, 15, 10885--10894 This journal is c the Owner Societies 2013

10–15 nm, thereby leading to an expected 85 to 100% utilizationof the volume of the electrode material.

In order to increase the storage capacity of the Cu–Sn films,another strategy was investigated experimentally, whichconsists of using thin films composed of a mixture of Cu6Sn5

and Sn of nominal composition ‘Cu6Sn10’ corresponding toCu6Sn5 + 5Sn. The theoretical capacities of this composition are728 mA h g�1 for Li and 641 mA h g�1 for Na, assuming fullconversion into Li4.26Sn and Na3.75Sn, respectively. The XRDpatterns for the starting target and resulting sputtered thinfilm diffraction patterns (Fig. S6, ESI†) clearly evidence theco-existence of Cu6Sn5 and Sn domains within the materialswith Cu6Sn5 having similar crystallite sizes to the materialdiscussed earlier. The potential profiles of the corresponding‘Cu6Sn10’ thin films are presented in Fig. 12 for a charge cut-offpotential of (a) 2 V and (b) 1 V vs. Na/Na+. The potential profilesare dominated by features of Sn6 superimposed onto thefeatures observed for Cu6Sn5 (Fig. 4). The initial dischargecapacities are about 600 mA h g�1 with reversible capacitiesof about 400 mA h g�1 and 350 mA h g�1 when charged at 2 and1 V, respectively. The second cycle reversible capacity is equal to290 and 330 mA h g�1 when charged up to 2 and 1 V,respectively. It is thus clear that charging up to 2 V leads to amore rapid fading of the capacity compared to charging up to1 V (27% vs. 6% losses in one cycle), which is likely dueto mechanical degradation accompanied by full contractionduring desodiation at 2 V.11,15

Rietveld refinements of the thin film XRD patterns (Fig. S7,ESI†) indicate a mixture of Cu6Sn5 and Sn with about 66 and34 wt%, respectively. Therefore, it can be calculated that thereversible capacity of 400 mA h g�1 corresponds to the reactionof both Sn and Cu6Sn5 with full participation of Sn and partialparticipation of Cu6Sn5 to about 32% of its theoretical capacity,similarly to what is measured for the pure Cu6Sn5 films (31%).The absence of Cu6Sn5 full reaction may be advantageousto maintain the electrode structural integrity and provide

electronic pathways for Na-ion anodes with significantly higherstorage capacities than that of pure Cu6Sn5. Although thecapacity retention and the irreversible losses due to the electrolytedecomposition need to be improved for practical applicationpurposes, for example by using carbon conductive additives,an inert ceramic binder material and electrolyte additives, thepresent results highlight the good potential of Cu6Sn5–Snmixtures as a possible anode material. Alternative electrodecomposite mixtures could be made of, non-exhaustively, physicalmixtures of Cu and Sn nanoparticles or embedded Sn particlesinside Cu nano-architectures.

ExperimentalSample preparation

Cu6Sn5 and ‘Cu6Sn10’ (Cu6Sn5 + 5Sn) targets were preparedfrom stoichiometric mixtures of Sn (99.8%, Alfa Aesar) and Cu(99.5%, Sigma-Aldrich) ball milled with yttria-stabilized

Fig. 11 Predicted capacity and active material volume ratio for Na-ion diffusiondepths of 3, 6, 9, 15, 21 and 30 nm inside spherical crystallites as a function of thecrystallite radius. The dashed line represents the current situation with 25 nmcrystallite radius and an estimated diffusion depth of 3 nm.

Fig. 12 Electrochemical properties of 1 mm thick ‘Cu6Sn10’ thin film electrodesversus Na-ion reaction. First (black) and second (grey) electrochemical potentialprofiles with a charge cut-off of (a) 1 V and (b) 2 V. Current is 7.9 mA cm�2 in bothcases, corresponding to about C/70 and C/50 discharge and charge currents,respectively.

Paper PCCP

Publ

ishe

d on

02

May

201

3. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Irvi

ne o

n 31

/10/

2014

00:

20:0

3.

View Article Online

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 10885--10894 10893

zirconia balls for 15 min to intimately mix the particles. Theresulting powder was pressed into a 200 diameter disk, annealedat 200 1C for 20 hours under Ar and slowly cooled down to roomtemperature. The resulting Cu6Sn5 material was analyzed byXRD and corresponds to the expected Z0-Cu6Sn5 (C2/c) mono-clinic phase formed at low temperatures with lattice parametersa = 11.0262 (�0.0014) Å, b = 7.2745 (�0.0012) Å and c = 9.846(�0.0021) Å. This is very close to previously reported latticeparameters.20 A small fraction of pure Sn was found in theCu6Sn5 target at a concentration of about 3 wt%, see Fig. S1(ESI†). The material of nominal composition ‘Cu6Sn10’ is amixture of Cu6Sn5 and 5Sn with nominal weight ratios ofCu6Sn5 and Sn equal to 62.15 and 37.85 wt%, respectively,which leads to a theoretical capacity of 641 mA h g�1.

Thin film deposition was carried out using DC magnetronsputtering in Ar plasma. Sputtering was conducted at 30 Wpower, 15 mTorr pressure with a 5 cm target–substratedistance, resulting in a deposition rate of about 0.9 nm s�1,and yielding the formation of thin films with the structure ofthe high temperature hexagonal Z-Cu6Sn5 (P63/mmc) phase (seeResults and discussion). Film thickness was varied between0.2 and 4 mm. The weight of the films was measured on aMettler MT5 balance with 1 mg precision. After preparation, thesamples were stored inside an Ar-filled glovebox.

Characterization

Electrochemical characterization was conducted at 25 1C with2-electrode 2032 coin cells prepared inside an Ar-filled glove-box. For Na-ion cells, pure Na metal was used as a counterelectrode, 1 M NaClO4 (Sigma-Aldrich) dissolved in propylenecarbonate (PC, Sigma-Aldrich) was used as the electrolyte, andtwo disks of glass fiber separator were used. For Li-ion cells,pure Li (Alfa Aesar), 1 M LiPF6 in ethylene carbonate/dimethylcarbonate (EC/DMC) (Novolyte) and Celgard 2325 separatorswere used. Galvanostatic cycling was performed using a Maccor4000. For ex situ characterizations, specimens were extractedfrom coin cells inside an Ar-filled glovebox and pressed onto afiber paper to remove the excess electrolyte.

A Scintag PDS 2000 diffractometer and a Siemens D5005diffractometer equipped with Cu Ka radiation and Ni filter wereused for XRD collection. For ex situ XRD measurements, cycledthin film electrodes were sealed with Kapton tape onto a glassslide and measured from 15 to 801 2y with 0.051 step size and4 s dwell time. Rietveld refinements were performed using thePANalytical HighScore Plus software. SEM photographs wereacquired using a JEOL JSM-6500F Field Emission microscope.119Sn transmission Mossbauer Spectroscopy data were acquiredat room temperature in the constant acceleration mode usingcomponents manufactured by ORTEC and WissEl. The sourceused for these experiments was 119mSn embedded in a CaSnO3

matrix. The velocity scale was calibrated with the magneticsextet of a high-purity iron foil as the reference absorber, and57Co (Rh) as the source. The spectra were fitted to Gauss-Lorentzian profiles by the least-squares method. All isomershifts (IS) are given with respect to the room temperaturespectrum of BaSnO3. The maximum experimental error on

hyperfine parameters is estimated to be �0.05 mm s�1. A singleelectrode was used for the measurement on the pristine material,for Na–Cu6Sn5 electrodes three identical samples were measuredsimultaneously whereas for Li–Cu6Sn5 electrodes two identicalsamples were measured simultaneously.

Conclusions

Using electrochemical, XRD and 119Sn-Mossbauer measure-ments we estimate that Na diffuses on average approximately3 nm into the Cu6Sn5 alloy structure. The Na that incorporates thestructure eventually fully converts Cu6Sn5 into Cu and Na15Sn4,which is the expected phase assuming full Na incorporation. Thelack of full reaction in the core of the nanoparticles is attributed tothe steric hindrance of Na ion diffusion as compared to Li ionsinside the Z-Cu6Sn5 structure. Moreover, the results also indicatethat the Cu6Sn5 alloy should have a particle size near 10 nm toutilize most of the reversible capacity provided by Sn. Furthermore,requirements on the geometry and size of the voids for Na+

diffusion inside the lattice have been given after detailed analysesof the diffusion pathways of Z-Cu6Sn5. Finally, a mixture ofCu6Sn5 + 5Sn of nominal composition ‘Cu6Sn10’ is proposed as away to increase the reversible storage capacity up to 400 mA h g�1.

Acknowledgements

This work was supported by the U.S. Department of Energy(DOE), Basic Energy Sciences (BES), Materials Sciences andEngineering Division. Research supported by ORNL’s SharedResearch Equipment (ShaRE) User Program (Microscopy) issponsored by DOE-BES. JCJ gratefully acknowledges RegionLanguedoc-Roussillon (France) for the financial support to the‘‘X-rays and gamma-rays platform’’ of Universite Montpellier IIin relation with Mossbauer spectroscopy experiments.

Notes and references

1 M. Armand and J.-M. Tarascon, Nature, 2008, 451, 652–657.2 S.-W. Kim, D.-H. Seo, X. Ma, G. Ma and K. Kang, Adv. Energy

Mater., 2012, 2, 710–721.3 B. L. Ellis and L. F. Nazar, Curr. Opin. Solid State Mater. Sci.,

2012, 16, 168–177.4 N. Yabuuchi, M. Kajiyama, J. Iwatate, H. Nishikawa,

S. Hitomi, R. Okuyama, R. Yasuhiro and S. Komaba, Nat.Mater., 2012, 11, 512–517.

5 A. Darwiche, C. Marino, M. T. Sougrati, B. Fraisse,L. Stievano and L. Monconduit, J. Am. Chem. Soc., 2012,134, 20805–20811.

6 L. Baggetto, P. Ganesh, R. P. Meisner, R. R. Unocic,J.-C. Jumas, C. A. Bridges and G. M. Veith, J. Power Sources,2013, 234, 48–59.

7 S. Komaba, Y. Matsuura, T. Ishikawa, N. Yabuuchi,W. Murata and S. Kuze, Electrochem. Commun., 2012, 21,65–68.

8 L. Baggetto, E. Allcorn, A. Manthiram and G. M. Veith,Electrochem. Commun., 2013, 27, 168–171.

PCCP Paper

Publ

ishe

d on

02

May

201

3. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Irvi

ne o

n 31

/10/

2014

00:

20:0

3.

View Article Online

10894 Phys. Chem. Chem. Phys., 2013, 15, 10885--10894 This journal is c the Owner Societies 2013

9 M. M. Thackeray, J. T. Vaughey, A. J. Kahaian, K. D. Keplerand R. Benedek, Electrochem. Commun., 1999, 1, 111–115.

10 K. D. Kepler, J. T. Vaughey and M. M. Thackeray, J. PowerSources, 1999, 81–82, 383–387.

11 D. Larcher, L. Y. Beaulieu, D. D. MacNeil and J. R. Dahn,J. Electrochem. Soc., 2000, 147, 1658–1662.

12 L. Fransson, E. Nordstrom, K. Edstrom, L. Haggstrom,J. T. Vaughey and M. M. Thackeray, J. Electrochem. Soc.,2002, 149, A736–A742.

13 K.-F. Chiu, K. M. Lin, H. C. Lin, W. Y. Chen and D. T. Shieh,J. Electrochem. Soc., 2007, 154, A433–A437.

14 S. Naille, R. Dedryvere, H. Martinez, S. Leroy, P.-E. Lippens, J.-C.Jumas and D. Gonbeau, J. Power Sources, 2007, 174, 1086–1090.

15 I.-S. Kim, J. T. Vaughey and O. Auciello, J. Electrochem. Soc.,2008, 155, A448–A451.

16 L. Trahey, J. T. Vaughey, H. Kung and M. M. Thackeray,J. Electrochem. Soc., 2009, 156, A385–A389.

17 W. Choi, J. Y. Lee and H. S. Lim, Electrochem. Commun.,2004, 6, 816–820.

18 G. R. Goward, N. J. Taylor, D. C. S. Souza and L. F. Nazar,J. Alloys Compd., 2001, 329, 82–91.

19 R. D. Shannon, Acta Crystallogr., Sect. A: Cryst. Phys., Diffr.,Theor. Gen. Cryst., 1976, 32, 751–767.

20 A.-K. Larsson, L. Stenberg and S. Lidin, Acta Crystallogr.,Sect. B: Struct. Crystallogr. Cryst. Chem., 1994, 50, 636–643.

21 A. Gangulee, G. C. Das and M. B. Bever, Metall. Trans., 1973,4, 2063–2066.

22 A. Patterson, Phys. Rev., 1939, 56, 978–982.23 S. Naille, PhD thesis, Universite Montpellier II, 2008.24 F. Robert, P. E. Lippens, J. Olivier-Fourcade, J.-C. Jumas,

F. Gillot, M. Morcrette and J.-M. Tarascon, J. Solid StateChem., 2007, 180, 413–422.

25 S. D. Beattie, T. Hatchard, A. Bonakdarpour, K. C. Hewittand J. R. Dahn, J. Electrochem. Soc., 2003, 150, A701–A705.

Paper PCCP

Publ

ishe

d on

02

May

201

3. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

Irvi

ne o

n 31

/10/

2014

00:

20:0

3.

View Article Online