Charging of lithium cobalt oxide battery cathodes studied by means of magnetometry

Solid State Ionics 293 (2016) 64–71

Contents lists available at ScienceDirect

Solid State Ionics journal homepage: www.elsevier.com/locate/ssi

Charging of lithium cobalt oxide battery cathodes studied by means of magnetometry G. Klinser a,⁎, S. Topolovec a, H. Kren b, S. Koller b, H. Krenn c, R. Würschum a a b c

Institute of Materials Physics, Graz University of Technology, Petersgasse 16, 8010 Graz, Austria VARTA Micro Innovation GmbH, Stremayrgasse 9, 8010 Graz, Austria Institute of Physics, University of Graz, Universitätsplatz 5, 8010 Graz, Austria

a r t i c l e

i n f o

Article history: Received 23 September 2015 Received in revised form 19 May 2016 Accepted 7 June 2016 Available online 16 June 2016 Keywords: LixCoO2Oxidation state Li deintercalation SQUID magnetometry Operando technique

a b s t r a c t The variation of the structural and electronic properties of LixCoO2 upon electrochemical Li-extraction was studied over wide range concentration of 1≥ x ≥ 0.20 by means of SQUID magnetometry. The ex-situ measurements performed for 13 different compositions were supplemented by operando measurements of the magnetic moment during repetitive electrochemical in-situ cycling of the Li-concentration. From the temperature-dependent measurements an effective magnetic susceptibility with Curie–Weiss behavior and an additional temperature-independent part due to Pauli and Van Vleck magnetism is derived. The increase of the temperature-independent susceptibility with Li-extraction reflects a concomitant increase of the electronic density of states and, in addition, indicates an Anderson-type of the occurring nonmetal–metal transition. The effective magnetic moment reveals that only a fraction of 30% of the charge is transferred to Co upon Li-extraction indicating a complex oxidation behavior involving oxygen. Exposure to ambient atmosphere gives rise to a complete oxidation of Co. The results on the structural variation with Li-concentration are compared with accompanying measurements by X-ray diffraction and by our recent defect studies by positron annihilation. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Due to the high technological relevance of LiCoO2 as cathode material of Li-ion batteries, much efforts are undertaken for better understanding the processes during charging on an atomistic scale. In addition to the regime of reversible Li de-/intercalation down to Li0.50CoO2, where the exact determination of the state of charge is still a challenging issue [1], also the processes underlying the performance loss when extracting more than 50% Li are of interest. A broad variety of characterization techniques is applied for this purpose, such as X-ray diffraction [2,3], nuclear magnetic resonance [4], X-ray photoelectron [5–7], or X-ray absorption [7] to mention only a few of them. In addition to those techniques, magnetometry has proven as powerful tool since the magnetic moment of electrode materials is highly sensitive to structural phase changes, impurities, metallic/non-metallic transitions, and the oxidation state of the transition metal ions (see review by Chernova et al. [8]). In fact, the magnetic susceptibility changes by several 100% with Li de-/intercalation and, therefore, serves as sensitive fingerprint of the charging state. Over the last few years, various groups were concerned with magnetic measurements on LixCoO2 cathodes [4,7,9–22]. Studies over a certain range of concentrations ⁎ Corresponding author. E-mail address: [email protected] (G. Klinser).

http://dx.doi.org/10.1016/j.ssi.2016.06.004 0167-2738/© 2016 Elsevier B.V. All rights reserved.

were performed by extracting Li either chemically or electrochemically: chemical Li extraction by Kellerman et al. [7] (0.98 ≥ x ≥ 0.60, 6 samples), Hertz et al. [9] (1≥ x ≥ 0.51, 9 samples), or Miyoshi et al. [10] (0.99 ≥ x ≥ 0.25, 6 samples), electrochemical Li extraction by Sugiyama et al. [11] (1 ≥ x ≥ 0.70, 3 samples), Mukai et al. [12] (1 ≥ x ≥ 0.10, 8 samples) or Motohashi et al. [13] (1≥ x ≥ 0, 9 samples). In order to gain further detailed insight on the structural and electronic variation of LixCoO2 with Li-content x, the present work is devoted at in-depth study of SQUID magnetometry comprising more than a dozen different Li concentrations. Commercially applied cathode material was used for the present studies. All different sample compositions were characterized by XRD. A wide range concentration 1 ≥ x ≥0.20 was covered extending well beyond the limit of reversible charging,1 and, in addition, the effect of exposure to ambient atmosphere was addressed. Compared to previous studies, particular emphasis was laid on minimizing the background signal of the magnetometer sample holder and, moreover, applying a procedure for proper subtraction of any residual background signal so that the magnetic susceptibility exclusively from the LixCoO2 samples could be determined. The results on the structural variation with Li-concentration are compared with

1 The terminology reversibility limit used here and in the following refers to the long term reversibility, for which a Li extraction beyond x=0.50 has to be avoided.

G. Klinser et al. / Solid State Ionics 293 (2016) 64–71

65

recent defect studies by positron annihilation performed by our group on the very same type of LixCoO2-samples [23]. A particular issue of the present studies pertains the comparison with accompanying operando SQUID measurements of the susceptibility variation of Lix CoO 2 upon reversible electrochemical cycling the Li-content. Operando measurements by means of an electrochemical cell operated in a SQUID magnetometer [24] open up the possibility to monitor charging-induced variations of the magnetic moments continuously and during repetitive cycles. Initial results on this novel technique are communicated elsewhere [25]. 2. Experimental The LiCoO2 samples were prepared in the same way as commercially used ones. A mixture containing 88 wt% of LiCoO2 particles, 7 wt% carbon black (Super P) as conducting agent and 5 wt% binder (polyc vinylidene difluoride hexafluoropropylene copolymer) was dissolved in NMP (N-Methyl-2-pyrrolidone). Subsequently, the slurry was coated on aluminum foils (thickness of 0.25 mm), pre-dried at 333 K in air, followed by a 24 h heat treatment at 353 K in vacuum (10−3 mbar). On each sample foil with a diameter of 12 mm, active cathode material of about 35 mg was deposited. The lithium extraction was performed electrochemically in a Maccor Series 4000 battery tester. The LiCoO2 cathodes were mounted as working electrode into a 3-electrode test cell (Swagelok-T-cell), separated from the metallic lithium foil counter and reference electrode by a nonwoven polypropylene separator (Freudenberg FS2190). A mixture of ethylene carbonate and ethyl methyl carbonate (volume ratio EC/EMC 3:7) was used as electrolyte with 2 wt% vinylene carbonate and 1 M LiPF6 as conducting salt. After a 12 h rest, the Li ions were extracted to a predefined Li concentration x in the LixCoO2 cathodes. The chosen constant current density of 13.5 μA cm −2 corresponds to a C-rate of 0.005 (complete Li-extraction in 200 h). The charging time for a pre-defined Li concentration x in the range of 1≥x≥0.20 was determined using a theoretical specific charge capacity of 274 mAh/g for complete lithium extraction. After the lithium extraction, the cathodes were dismantled, rinsed with diethyl carbonate to remove the electrolyte, and finally dried in a vacuum (10−3 mbar) at 353 K for 24 h. In total 13 different compositions were prepared. Reproducibility was tested by comparing each two samples of identical compositions for a few selected compositions. Magnetic susceptibility measurements were performed in a temperature range between 300 and 8 K (field cooling) at a constant magnetic field of 10 kOe, using a superconducting quantum interference device (SQUID: Quantum Design MPMS-XL-7). To enable a precise determination of the magnetic susceptibility of the LixCoO2 samples, a sample holder with minimized contribution to the magnetic signal was designed. The samples were placed in the middle of a 14 cm long polyolefin tube (diameter 3.2 mm). For this purpose the LixCoO2-coated Al-foils were folded in such a way that they remained fixed at the pre-defined position. The polyolefin tube was mounted into an NMR-tube (Wilmad 505-PS-7) which was closed with a teflon plug and hemetically sealed with epoxy resin. An appropriate tube length was chosen to ensure that the tube extended well beyond the SQUID pick up coils during the measurement scan. In this way it was guaranteed that the tubes do not contribute to the measured magnetic moment, similar as described for long homogenous substrates by Manios et al. [26]. In order to subtract the magnetic signals caused by the aluminum substrate and the 12 wt% additives (carbon black and polymer binder), the temperature-dependent susceptibility was measured for both the plain Al substrate and the Al substrate coated with polymer binder and carbon in the same volume ratio as for the LiCoO2-samples. By means of these magnetic data along with the molar weights of all components, a precise correction of the background signal of each sample could be achieved. Sample preparation, handling and transfer into the SQUID device were carried out under protective Ar-atmosphere. Since the magnetic measurements were performed under He-atmosphere,

Fig. 1. Lattice parameters a and c of the three hexagonal phases deduced from XRD of LixCoO2 in dependence of Li concentration x. For Li0.45CoO2, a monoclinic phase was observed. Minor traces of the monoclinic phase were already observed for Li0.50CoO2. The data points (■) where obtained from the shoulders of the 003-peaks (see text). The dotted lines are guides for the eyes.

inert gas conditions prevail during the whole procedure up to the end of the magnetic measurements. After the magnetic measurements, the samples were structurally characterized by X-ray diffractometry using a Bruker D8 Advance diffractometer in Bragg-Brentano geometry with Cu-Kα-radiation. The experimental procedure for operando measurements of the magnetic susceptibility on identically prepared LiCoO 2 sample material is described elsewhere [25]. 3. Sample characterization The XRD measurements on LixCoO2 cathodes revealed three  rhombohedral phases (R3m) and one monoclinic phase in the studied range 1.0 ≥ x ≥ 0.40 of Li concentrations. Following the procedure in literature, the XRD data of the rhombohedral structure are analyzed according to a hexagonal phase [2,3]. In Fig. 1, the lattice parameters a and c corresponding to a hexagonal unit cell are displayed as a function of Li-content x. For the first hexagonal phase (hex I) constant lattice parameters where found in the Li concentration range of 1≥x≥0.90. For the second hexagonal phase (hex II), a slight decrease of the lattice parameter a and a linear increase of the c parameter occurs in the range 0.70 ≥x ≥0.50 which reflects an increasing Coulomb repulsion of the anions upon Li-extraction, as reported in literature [2,4]. For x = 0.90, a shoulder in the (003) diffraction peak of the dominant hexagonal-I phase indicates a minor phase fraction, the c-parameter of which nicely fits to that of the hexagonal-II phase; likewise, for x = 0.70 indication of a minor hexagonal-I phase is deduced from the (003)-peak of the dominant hexagonal-II phase (see blue squares in Fig. 1). This indicates that for 0.90 ≥ x ≥ 0.70 both hexagonal phases coexist.2 Li extraction 2 It should be noted that for LixCoO2 it is not possible to detect secondary phases at levels lower than 5% with XRD measurements [27]. Therefore the phase separation could start already at higher Li-contents as indicated by potentiostatic measurements in literature [22,27].

66

G. Klinser et al. / Solid State Ionics 293 (2016) 64–71

beyond x =0.50, for which already weak features of a monoclinic phase are visible in the XRD pattern, leads to a structural phase transition from the rhombohedral to the monoclinic phase (x = 0.45, not shown in Fig. 1) and back to a rhombohedral phase (x b 0.45) which again is displayed in Fig. 1 according to a hexagonal unit cell (hex III). In summary, the lattice parameters and crystal structures observed for the present series of samples with different Li-content (Fig. 1) agree well with reported literature data [2–4,10,13]. This shows that the present way of adjusting and determination of the lithium concentration by electrochemical extraction can be considered as reliable. The initial state of the material was further characterized by Inductive Coupled Plasma Mass Spectroscopy (ICPMS), which revealed a Li-content of 99.3% very close to a stoichiometric composition. The stoichiometric composition was further checked by the first galvanostatic charging curve, displayed in Fig. 2. According to literature, the observed characteristic shape indicates a pristine sample close to stoichiometry. Indeed, our charging curve is in full agreement with that of the really stoichiometric sample in [22]. On the one hand, an initial voltage plateau, corresponding to a coexistence of two structural hexagonal phases (see above), can be seen at the beginning of the charging curve. On the other hand, the characteristic voltage behavior around 4.2 V, which is due to the structural phase transition as observed by XRD, appears at a Licontent of x = 0.50 (see [17]).

Fig. 2. First galvanostatic charging curve of LixCoO2 cell. The discharge current was chosen to be 13.5 μA/cm 2 which corresponded to a C-rate of 0.005 (complete Li-extraction in 200 h).

4. Magnetic susceptibility measurements The variation of the molar magnetic susceptibility χmol(T) of LixCoO2 with temperature is shown in Fig. 3. A substantial variation of both the temperature dependence and of the absolute values of χ (as indicated by arrows for 130 K) can be discerned. The susceptibility curves were fitted using a modified Curie–Weiss law (cgs units)3 χ¼

NA μ 2eff C þ χ0 þ χ0 ¼ T þ TN 3kB ðT þ T N Þ

ð1Þ

with C the Curie constant, NA the Avogadro constant, μ eff the effective magnetic moment, kB the Boltzmann constant, TN the Néel temperature, and χ0 a temperature independent part. As discussed later, this temperature-independent part is assigned to Pauli and Van Vleck paramagnetism. Due to minima and step-like behavior of χ(T) in the range between 140 K and 180 K for certain Li-contents, the temperature range for fitting of Eq. (1) was restricted to 8 − 130 K. The effective magnetic moment μ eff characterizing the Curie–Weiss part of Eq. (1) is shown in Fig. 4 as a function of Li-content x. With progressing Li extraction, μ eff shows an increase between 1 ≥ x N 0.55, most pronounced for x ≥ 0.70, and a subsequent strong decrease right in the vicinity of the limit of reversible charging (0.55N x N 0.45). Upon further Li extractions beyond the reversibility limit (x ≤ 0.45), μ eff stays almost constant. In Fig. 4b, the quotient μ eff/(1 − x) is displayed which characterizes the effective moment per Co4+-ion. This is based on the simple notion that the magnetic moment arises exclusively from the Co4+-ions which are formed by oxidation of spin-0 Co3+ upon Li+-extraction (see discussion). The effective moment per Co4+-ion μ eff/(1 − x) at first decreases in the range between 1 N x ≥ 0.70 to a constant value between x = 0.7 and x = 0.55, followed by a further steplike decrease in the vicinity of the reversible charging limit (0.55 N x N 0.45), finally reaching a constant value for Li-contents below 45%. Besides eventual mass losses which may occur during handling the brittle sample material the large error bars of μ eff/(1−x) for 1≤x≤0.9 are due to uncertainties in determining the lithium content x, which affect μ eff/(1− x) particularly for high x values. 3 For comparison with literature the cgs unit emu is used for the magnetic susceptibility χ (1 emu/mol=4π10−6 m3/mol)

For the Néel temperature values in the range from 3 to 14 K were obtained (not shown), except for Li1.00CoO2 for which TN = 185 K was obtained from the fit. The temperature independent part χ0 of the susceptibility (see Eq. (1)) is shown in Fig. 4c. χ0 gradually increases from the onset of Li extraction, reaching a constant value for x ≤ 0.65 with the exception of Li0.55CoO2 and Li0.50CoO2, for which slightly reduced values are observed. The χ0-value for Li1.00CoO2 is omitted in Fig. 4c since a precise determination of this temperature-independent part is hampered by antiferromagnetic ordering which is observed below 35 K in agreement with literature [12]. A rough estimation yields a χ0 value for Li1.00CoO2 of 1.4×10−4 emu/(mol*Oe). The results of the operando measurements of the susceptibility variation of LixCoO2 upon reversible electrochemical cycling the Li-content x are shown in Fig. 5. The measurements were performed at 300 K on the same type of material as the temperature-dependent ex-situ measurements presented above. The respective χ-values for 300 K of the ex-situ measurements are included for comparison. The comparison shows that the operando measurements enable a precise and continuous monitoring of the charging-induced χ variation over various consecutive cycles. The small deviations between ex-situ and operando measurements may arise from uncertainties of the background-subtraction which in the case of the operando studies is less precise due to the Li-anode in the electrochemical cell. 4.1. Discussion of χ0: metal/insulator transition We start the discussion of the results on the temperatureindependent susceptibility χ0 with a brief comparison with literature. Fig. 6a. shows literature results from those studies which were performed in dependence of Li-concentration and which were analyzed in the same way according to Eq. (1) as done here. The present studies covers a wide range of Li concentrations, well beyond the limit of reversible charging. The higher χ0-values compared to previous studies is considered to arise from the precise background subtraction (see Section 2). The overall trend of χ0(x) with an increase with decreasing Li-concentration to a saturation value with an intermediate dip in the range of x = 0.5 is reflected in each of the studies [7,9,12,13] (Fig. 6a). The temperature independent part of the susceptibility χ0 by far dominates at 300 K as shown by the comparison with the χ-values measured at that temperature (compare Fig. 6a and Fig. 5). This temperature

G. Klinser et al. / Solid State Ionics 293 (2016) 64–71

67

Fig. 4. Effective magnetic moments μeff (a), μeff/(1−x) (b), and temperature independent part χ0 of the susceptibility (c) of LixCoO2 as a function of lithium content x. μeff/(1−x) denotes the effective moment per Co4+ ion assuming that exclusively Co is oxidzed upon Li extraction (see text). The errorbars account for uncertainties in determining the lithium content and sample mass (due to brittleness).

Fig. 3. Molar susceptibility χmol of LixCoO2 as a function of temperature T for various lithium contents x measured in field cooling mode at a constant magnetic field of 10 kOe. For clarity, the χ(T) curves are mutually shifted in vertical direction. The vertical arrow measured from the baseline of each curve marks the susceptibility at 130 K χ130.

independent part χ0 of LixCoO2 is attributed to Van Vleck and Pauli paramagnetism. Stoichiometric Li1.00CoO2 represents a band insulator [4] for which Van Vleck paramagnetism occurs exclusively. The estimated value of the temperature independent susceptibility of the starting material (Li1 CoO 2) of χ0 = 1.4 × 10−4 emu/(mol*Oe) nicely fits to the value for the Van Vleck plus diamagnetic susceptibility of 0.9 × 10 −4 emu/(mol*Oe) reported for stoichiometric Li1 CoO 2 (ref. [22]) and Na1CoO2 (ref. [28]). The Van Vleck paramagnetism is hardly affected by electron hole formation upon Li extraction, as the Van Vleck contribution to the overall magnetic signal of the Co4+ ions (energy level splitting between t2g and eg band of 2 eV [5]) can be assumed to be nearly the same as that of the Co3+ ions (energy level splitting between the t2g and eg band of 2.4 eV [5]). Therefore, the observed variation of χ0 with Li-content has to be attributed entirely to Pauli paramagnetism which emerges right at the onset of Li-extraction, as can be seen in the operando measurements

(compare Fig. 5). In fact, according to band-structure calculations by Mileweska et al. [29,30], the Fermi level is located in the bandgap for Li1CoO2, but for Li0.99CoO2 already within the electronic band.

Fig. 5. Susceptibility χmol measured operando on LixCoO2 at 300 K upon reversible electrochemical cycling the Li-content x (solid line). The respective χmol-values for 300 K of the ex-situ measurements (Fig. 3) are shown for comparison (triangles).

68

G. Klinser et al. / Solid State Ionics 293 (2016) 64–71

Using the relation between Pauli-susceptibility χ P ¼ μ 0 μ 2B Dðε F Þ;

ð2Þ

and electronic density of states D(εF) at the Fermi edge, an increase of D(εF) by 8.3/eV can be deduced from the ex-situ measured increase of χ0 between x = 0.98 and x = 0.60. From the χ-values measured operando at 300 K an increase of D(εF) by 6.8/eV is deduced for the range x = 1.00 − 0.77, taking into account that D(EF) = 0 for Li1CoO2 [29] and assuming that the contribution of localized momenta is constant within this range. The observed increase of D(εF) with Li-extraction appears to be slightly higher than the value of 4.5 states/eV which can be derived from the band-structure calculations between x = 0.99 and x = 0.60 [29]. Obviously, the χ0 variation with Li+-concentration cannot entirely be attributed to a rise of D(εF). The stronger χP-increase can well be reconciled within the notion that the nonmetal − metal transition occurring upon Li extraction is of Anderson type. Actually, such a type of transition may contribute to an enhancement of χP due to correlation effects [31]. Further support for such type of transition is derived from the fact that metallic behavior is found by conductivity measurements only for x ≤ 0.94 [4,32], rather than at the onset of Li-extraction already as derived from the observed linear increase in χ0. Because since an Anderson-transition is associated with charge carrier localization, a jump in the conductivity occurs at the nonmetal − metal transition whereas the Pauli susceptibility changes continuously (see Mott [31]). Indication of Anderson localization in LixCoO2 was also obtained from STM measurements [33]. 4.2. Discussion of μeff: oxidation behavior of Co In the range of 0.70 N x N 0.50, where the contribution of Pauli paramagnetism remains constant (see chapter 4.1), the variation of the effective magnetic moment μ eff/(1 − x) of Co4+ with Li-deintercalation (Fig. 3b) yields insight in the oxidation behavior of Co. There is clear evidence from studies of X-ray photoelectron spectroscopy [34], nuclear magnetic resonance [35], magnetic susceptibility [36], and band structure calculations [37], that in stoichiometric Li1CoO2 the Co3+ ions are in the low spin configuration with S = 0. For Co4+, a magnetic moment of 1.73 μ B is associated with the respective low spin state S=1/2 [7,10,13]. As a most striking feature, the measured effective magnetic moment μ eff/(1 − x) decreases with x and attains a value around 0.6 μ B in the range between x = 0.70 and x = 0.50, much less than 1.73 μ B (Fig. 6b). A similar reduction was reported by Kellerman et al. [7] and Hertz et al. [9]. Apparently the simple notion fails according to which exclusively localized Co is oxidized (Co3+ → Co4+) upon Li+ extraction. Indeed, an oxidation state of 3.35 rather than 4.0 is deduced from the constant measured effective magnetic moment 0.6 μ B taking into account a moment of 0 μ B for Co3+ and of 1.73 μ B for Co4+. This result from the ex-situ measurement is fully confirmed by the operando studies, from which a value of 0.55 μB is deduced from a linear fit χ − x for x ≤ 0.72 of the continuous measurement (Fig. 5) corresponding to an effective oxidation state of 3.32. The present measurements are in excellent agreement with ab-initio calculations according to which only 35% of the charge is transferred to Co ions upon intercalation [38]. This indicates a more complex oxidation behavior where oxygen is involved in addition to Co. The surface sensitive techniques XANES [39,40,7] and XPS [6] reveal that oxygen undergoes partial oxidation during extraction of Li+. The present magnetic measurement shows that this partial oxygen oxidation is not restricted to the surface region, but occurs throughout the bulk. It should be mentioned, that leaving aside the accordance with XPS and XANES measurements there might be an alternative interpretation of the observed behavior. The band structure for x b 0.70 could be such

Fig. 6. Comparison with literature. Temperature-independent part χ0 of the susceptibility (a) and effective magnetic moment μeff/(1−x) (b) as a function of Li-content x.

as D(εF) stays constant upon Co oxidation, whereas the conduction band continues to be emptied. 4.3. Discussion: limit of reversible charging and comparison with positron annihilation studies The well-known limit of reversible charging of LixCoO2 at a concentration of about 50% is considered to be due to irreversible structural transition as evidenced by numereous XRD studies, including the present accompanying XRD measurements (see above). There are only a few magnetic studies in the regime x ≤ 0.50 (see Fig. 6). The reversibility limit x=0.50 is clearly visible in the present measurements as step-like decrease of the effective magnetic moment μ eff /(1 − x) (Fig. 4b). Following Ensling et al. [5] and the line of arguments presented above, this decrease may arise from an electronic charge transfer from O-ions to Co-ions followed by a loss of oxygen. Particularly remarkable is the transitory decrease of the Tindependent part χ0 which occurs between x = 0.60 and x = 0.55 in the region where χ0 otherwise remains constant. Recent studies of positron annihilation [23] performed by our group on the very same type of LixCoO2-samples reveal in the exactly identical concentration range a decrease of the positron lifetime τ (Fig. 7) and of the line-shape parameter of positron-electron annihilation radiation (not shown). Positrons act as highly sensitive probe for atomic-sized open-volume defects. The initial increase of τ between x =1.0 and x = 0.60 (Fig. 7) reflects a formation of double vacancies and smaller vacancy agglomerates on the Li sublattice due to Li extraction [23]. The decrease of τ in the relevant range between x = 0.60 and x = 0.55 is attributed to reordering of Li-vacancy agglomerates from a two-dimensional array to onedimensional vacancy chains (see [23]). It appears obvious that the observed χ0 decrease (Fig. 7) is related to this reordering. The notion of vacancy reordering is supported by theoretical studies according to which vacancies are randomly distributed

G. Klinser et al. / Solid State Ionics 293 (2016) 64–71

69

Fig. 7. Temperature-independent part χ0 of the susceptibility in comparision to positron lifetime τ (redrawn after [23]) as a function of Li-content x. The lines are guide for the eyes.

on the Li + layers for x ≥ 0.6, whereas one-dimensional vacancy chains occur for x = 0.5 [41]. Indirect evidence is also deduced from monoclinic distortion observed by XRD (see [23] and references therein). 4.4. χ(T)-hysteresis in Li-concentration range of the hexagonal II phase A closer inspection of the χ − T curves in Fig. 3 shows kinks and minima of χ at around 140−180 K for Li-concentrations in the range 0.90 ≥ x ≥ 0.50, which actually corresponds to the range where the hexagonal II phase is stable (Fig. 1). For more detailed studies, field cooling (FC) and field warming (FW) measurements were performed as shown in Fig. 8. With decreasing Li-concentration, deviations from a monotonic χ− T-behavior were first found for Li0.90 CoO2 at about 180 K with a minor T-hysteresis between FC and FW. More pronounced temperature hystereses in this temperature range occur for x = 0.70 (not shown), x = 0.65, x = 0.55, and x = 0.50 with the exception of Li0.60CoO2, where the step is much smaller but shifted to higher temperatures of about 210 K (Fig. 8). For x = 0.45 and x = 0.40 (not shown), i.e., beyond the concentration range of the hexagonal II phase, a broad minimum around 150 K occurs. This complex χ − T-behavior is in fair well agreement with the results obtained by Motohashi et al. [13], even with respect to the particular temperature shift for x = 0.60. However, no hysteresis was found for x =0.60 by these authors [13]. Deviations from a monotonous χ − T-behavior were also observed by a number of other groups, [7,9– 12] although in detail slightly different as in the present case. According to the previous studies [7,12,13], a structural or magnetic transition in this temperature regime is excluded as origin of the hysteresis. As conductivity measurements also show kink-like behavior in these ranges of Li-concentrations and temperatures [4,7], the magnetic behavior was assigned to partial charge ordering [13]. However it is apparent from the present studies that the onset of this particular magnetic behavior appears to be related with first onset of the second hexagonal phase (Fig. 1). In addition, the hysteresis behavior might be related to reduced χ0-values in this composition range. More detailed studies are necessary in order to relate this magnetic behavior unambiguously with the electronic and structural properties. 5. Exposure to ambient atmosphere In the present work also the influence of exposing the LixCoO2samples to ambient atmosphere was studied by magnetometry.

Fig. 8. Field cooling and field warming measurements of the molar magnetic susceptibility χmol as a function of temperature T. In comparison to Li0.55CoO2 for the other figures the width of the y-scale is reduced for the sake of clarity.

A Li0.70CoO2 sample was first measured in argon atmosphere, and subsequently after storage in ambient atmosphere for 1 and 7 days. With progressing exposure to ambient atmosphere, the molar susceptibility substantially increases at low temperature (Fig. 9a), whereas the aforementioned steplike behavior at around 170 K ceases (see inset in Fig. 9a). The analysis according to Eq. (1) shows that the effective moment μeff/(1 − x) after storage at ambient atmosphere for 7 days reaches the value of 1.73 μB of Co4+, i.e., the value as expected when exclusively Co is oxidized upon Li-extraction (Fig. 9b). This behavior was verified by measuring two further Li0.70CoO2 samples after air exposure (sample 2, see double data points for 0 and 7 days; sample 3 after 38 days). In order to check that the increase of the effective moment is due to the ambient atmosphere, another sample was kept for 7 days under Ar atmosphere after preparation. Indeed, in that case, no enhanced value for μ eff/(1 − x) could be observed (Fig. 9b). In contrast to μ eff/(1 − x), the temperature-independent part χ0 is hardly affected by exposure to ambient atmosphere (Fig. 9c). The increase of the magnetic moment from about 0.6 μB towards the moment of Co4+ indicates that Co completely oxidizes upon exposure to ambient atmosphere. 6. Summary and conclusion The major novel results of the present studies performed on LixCoO2 over wide range concentration of 1 ≥ x ≥0.20 are as follows: • Minimization and correction of the background signal enabled a precise determination of magnetic susceptibility of LixCoO2. • Operando measurements enabled a monitoring of charging-induced variations of the magnetic susceptibility continuously and during repetitive cycles. • The instantaneous increase of the temperature-independent susceptibility χ0 with Li-extraction is associated with a concomitant

70

G. Klinser et al. / Solid State Ionics 293 (2016) 64–71

Acknowledgment The authors would like to thank W. Gössler (Institute of Chemistry, University of Graz) for performing the ICPMS analysis and C. Baumann (VARTA Micro Innovation GmbH) for preparing the LixCoO2 cathodes. Financial support from the Graz inter-university cooperation on natural sciences (NAWI Graz) is appreciated. References

Fig. 9. Influence of exposure to ambient atmosphere. (a) Magnetic susceptibility χmol of Li0.70CoO2 measured in Ar atmosphere (red circles) and after exposure to ambient atmosphere for 1 day (green triangles) and 7 days (blue squares). Effective magnetic moment μeff/(1− x) (b) and temperature-independent part χ0 of susceptibility (c) as a function of duration of exposure to ambient atmosphere. The values of a reference sample kept under Ar atmosphere are shown for comparison.

increase of the electronic density of states and, in addition, with an Anderson-type of the nonmetal–metal transition. The latter is also supported from the different behavior of magnetic moment and electronic conductivity at the transition. • The variation of χ0 at the limit of reversible charging appears to be related to structural reordering as observed by positron annihilation spectroscopy in the same concentration range. • The oxidation state of Co deduced from the effective magnetic moment of the Curie–Weiss part is found to be in excellent agreement with ab-initio calculations and indicates a complex oxidation behavior involving oxygen. • The effective magnetic moment reveals the oxidation to Co4+ upon exposure to ambient atmosphere. In conclusion, magnetometry reveals as sensitive tool for studying the structural and electronic properties of battery electrodes. In particular the combination of SQUID magnetometry and in-situ electrochemical charging may open attractive application potentials also for more complex battery material systems.

[1] Z. Yanhui, S. Wenji, L. Shili, L. jie, F. Ziping, J. Renewable Sustainable Energy 5 (2013) 021403, http://dx.doi.org/10.1063/1.4798430. [2] G.G. Amatucci, J.M. Tarascon, L.C. Klein, J. Electrochem. Soc. 143 (1996) 1114, http:// dx.doi.org/10.1149/1.1836594. [3] J.N. Reimers, J.R. Dahn, J. Electrochem. Soc. 139 (1992) 2091–2097, http://dx.doi.org/ 10.1149/1.2221184. [4] M. Ménétrier, I. Saadoune, S. Levasseur, C. Delmas, J. Mater. Chem. 9 (1999) 1135–1140, http://dx.doi.org/10.1039/A900016 J. [5] D. Ensling, G. Cherkashinin, S. Schmid, S. Bhuvaneswari, A. Thissen, W. Jaegermann, Chem. Mater. 26 (2014) 3948–3956, http://dx.doi.org/10.1021/cm501480b. [6] L. Dahéron, R. Dedryvére, H. Martinez, M. Ménétrier, C. Denage, C. Delmas, D. Gonbeau, Chem. Mater. 20 (2008) 583–590, http://dx.doi.org/10.1021/ cm70254 6 s. [7] D. Kellerman, V. Galakhov, A. Semenova, Y. Blinovskov, O. Leonidova, Phys. Solid State 48 (2006) 548–556, http://dx.doi.org/10.1134/S106378340603022X. [8] N.A. Chernova, G.M. Nolis, F.O. Omenya, H. Zhou, Z. Li, M.S. Whittingham, J. Mater. Chem. 21 (2011) 9865–9875, http://dx.doi.org/10.1039/C1JM00024A. [9] J.T. Hertz, Q. Huang, T. McQueen, T. Klimczuk, J.W.G. Bos, L. Viciu, R.J. Cava, Phys. Rev. B 77 (2008) 075119, http://dx.doi.org/10.1103/PhysRevB.77.075119. [10] K. Miyoshi, C. Iwai, H. Kondo, M. Miura, S. Nishigori, J. Takeuchi, Phys. Rev. B 82 (2010) 075113, http://dx.doi.org/10.1103/PhysRevB.82.075113. [11] J. Sugiyama, H. Nozaki, J.H. Brewer, E.J. Ansaldo, G.D. Morris, C. Delmas, Phys. Rev. B 72 (2005) 144424, http://dx.doi.org/10.1103/PhysRevB.72.144424. [12] K. Mukai, Y. Ikedo, H. Nozaki, J. Sugiyama, K. Nishiyama, D. Andreica, A. Amato, P.L. Russo, E.J. Ansaldo, J.H. Brewer, K.H. Chow, K. Ariyoshi, T. Ohzuku, Phys. Rev. Lett. 99 (2007) 087601, http://dx.doi.org/10.1103/PhysRevLett.99.087601. [13] T. Motohashi, T. Ono, Y. Sugimoto, Y. Masubuchi, S. Kikkawa, R. Kanno, M. Karppinen, H. Yamauchi, Phys. Rev. B 80 (2009) 165114, http://dx.doi.org/10. 1103/PhysRevB.80.165114. [14] S. Levasseur, M. Ménétrier, Y. Shao-Horn, L. Gautier, A. Audemer, G. Demazeau, A. Largeteau, C. Delmas, Chem. Mater. 15 (2003) 348–354, http://dx.doi.org/10.1021/ cm02127 9 g. [15] T. Motohashi, Y. Katsumata, T. Ono, R. Kanno, M. Karppinen, H. Yamauchi, Chem. Mater. 19 (2007) 5063–5066, http://dx.doi.org/10.1021/cm0702464. [16] V. Galakhov, V. Karelina, D. Kellerman, V. Gorshkov, N. Ovechkina, M. Neumann, Phys. Solid State 44 (2002) 266–273, http://dx.doi.org/10.1134/1.1451011. [17] K. Mukai, Y. Aoki, D. Andreica, A. Amato, I. Watanabe, S.R. Giblin, J. Sugiyama, Phys. Rev. B 89 (2014) 094406, http://dx.doi.org/10.1103/PhysRevB.89.094406. [18] K. Mukai, Y. Kishida, H. Nozaki, K. Dohmae, Chem. Mater. 25 (2013) 2828–2837, http://dx.doi.org/10.1021/cm4012348. [19] T.Y. Ou-Yang, F.-T. Huang, G.J. Shu, W.L. Lee, M.-W. Chu, H.L. Liu, F.C. Chou, Phys. Rev. B 85 (2012) 035120, http://dx.doi.org/10.1103/PhysRevB.85.035120. [20] D. Mohanty, H. Gabrisch, Solid State Ionics 194 (2011) 41–45, http://dx.doi.org/10. 1016/j.ssi.2011.05.005. [21] T. Motohashi, T. Ono, Y. Katsumata, R. Kanno, M. Karppinen, H. Yamauchi, J. Appl. Phys. 103 (2008), 07C902, http://dx.doi.org/10.1063/1.2828525. [22] M. Ménétrier, D. Carlier, M. Blangero, C. Delmas, Electrochem. Solid. St. 11 (2008) A179–A182, http://dx.doi.org/10.1149/1.2968953. [23] P. Parz, B. Fuchsbichler, S. Koller, B. Bitschnau, F.-A. Mautner, W. Puff, R. Würschum, Appl. Phys. Lett. 102 (2013) 151901, http://dx.doi.org/10.1063/1.4801998. [24] S. Topolovec, H. Krenn, R. Würschum, Rev. Sci. Instrum. 86 (2015) 063903, http://dx. doi.org/10.1063/1.4922462. [25] S. Topolovec, H. Kren, G. Klinser, S. Koller, H. Krenn, R. Würschum, J. Solid State Electrochem. 20 (2016) 1491, http://dx.doi.org/10.1007/s10008-015-3110-6. [26] E. Manios, D. Stamopoulos, N. Moutis, M. Pissas, D. Niarchos, J. Magn. Magn. Mater. 320 (2008) 3264–3271, http://dx.doi.org/10.1016/j.jmmm.2008.06.020. [27] M. Morcrette, Y. Chabre, G. Vaughan, G. Amatucci, J.-B. Leriche, S. Patoux, C. Masquelier, J.-M. Tarascon, Electrochim. Acta 47 (2002) 3137–3149, http://dx.doi. org/10.1016/S0013-4686(02)00233-5. [28] G. Lang, J. Bobroff, H. Alloul, P. Mendels, N. Blanchard, G. Collin, Phys. Rev. B 72 (2005) 094404, http://dx.doi.org/10.1103/PhysRevB.72.094404. [29] A. Milewska, K. Swierczek, J. Tobola, F. Boudoire, Y. Hu, D. Bora, B. Mun, A. Braun, J. Molenda, Solid State Ionics 263 (2014) 110, http://dx.doi.org/10.1016/j.ssi.2014.05. 011. [30] J. Molenda, D. Baster, A. Milewska, K. wierczek, D.K. Bora, A. Braun, J. Tobola, Solid State Ionics 271 (2015) 15–27, http://dx.doi.org/10.1016/j.ssi.2014.09.032. [31] N.F. Mott, Philos. Mag. 29 (1974) 59–63, http://dx.doi.org/10.1080/1478643740 8213553. [32] J. Molenda, A. Stoklosa, T. Bak, Solid State Ionics 36 (1989) 53, http://dx.doi.org/10. 1016/0167-2738(89)90058-1. [33] K. Iwaya, T. Ogawa, T. Minato, K. Miyoshi, J. Takeuchi, A. Kuwabara, H. Moriwake, Y. Kim, T. Hitosugi, Phys. Rev. Lett. 111 (2013) 126104, http://dx.doi.org/10.1103/ PhysRevLett.111.126104.

G. Klinser et al. / Solid State Ionics 293 (2016) 64–71 [34] J. van Elp, J.L. Wieland, H. Eskes, P. Kuiper, G.A. Sawatzky, F.M.F. de Groot, T.S. Turner, Phys. Rev. B 44 (1991) 6090–6103, http://dx.doi.org/10.1103/PhysRevB.44.6090. [35] M. Ménétrier, A. Rougier, C. Delmas, Solid State Commun. 90 (1994) 439–442, http://dx.doi.org/10.1016/0038-1098(94)90038-8. [36] S. Kikkawa, S. Miyazaki, M. Koizumi, J. Solid State Chem. 62 (1986) 35–39, http://dx. doi.org/10.1016/0022-4596(86)90213-6. [37] M.T. Czyżyk, R. Potze, G.A. Sawatzky, Phys. Rev. B 46 (1992) 3729–3735, http://dx. doi.org/10.1103/PhysRevB.46.3729.

71

[38] M.K. Aydinol, A.F. Kohan, G. Ceder, K. Cho, J. Joannopoulos, Phys. Rev. B 56 (1997) 1354–1365, http://dx.doi.org/10.1103/PhysRevB.56.1354. [39] W.-S. Yoon, K.-B. Kim, M.-G. Kim, M.-K. Lee, H.-J. Shin, J.-M. Lee, J.-S. Lee, C.-H. Yo, J. Phys. Chem. B 106 (2002) 2526–2532, http://dx.doi.org/10.1021/jp013735e. [40] C.-H. Chen, B.-J. Hwang, C.-Y. Chen, S.-K. Hu, J.-M. Chen, H.-S. Sheu, J.-F. Lee, J. Power Sources 174 (2007) 938–943, http://dx.doi.org/10.1016/j.jpowsour.2007.06.083. [41] A. Van der Ven, M.K. Aydinol, G. Ceder, G. Kresse, J. Hafner, Phys. Rev. B 58 (1998) 2975–2987, http://dx.doi.org/10.1103/PhysRevB.58.2975.