Li dynamics in carbon-rich polymer-derived SiCN ceramics probed by NMR 4 1 Seung-Ho Baeka,∗, Lukas Mirko Reinoldb, Magdalena Graczyk-Zajacb, Ralf 0 2 Riedelb, Franziska Hammeratha, Bernd Büchnera,c, Hans-Joachim Grafea n a aIFW-Dresden, Institute for Solid State Research, PF 270116, 01171 Dresden, Germany J bTechnische Universität Darmstadt, Fachbereich Material- und Geowissenschaften, 8 Fachgebiet Disperse Feststoffe, Jovanka-Bontschits-Straße 2, 64287 Darmstadt, Germany cInstitut für Festkörperphysik, Technische Universität Dresden, 01062 Dresden, Germany ] i c s - l r t Abstract m . at We report 7Li, 29Si, and 13C NMR studies of two different carbon-rich SiCN m ceramicsSiCN-1andSiCN-3derivedfromthepreceramicpolymerspolyphenylvinylsi- - d n lylcarbodiimide and polyphenylvinylsilazane, respectively. From the spec- o c tral analysis of the three nuclei at room temperature, we find that only the [ 1 13C spectrum is strongly influenced by Li insertion/extraction, suggesting v 0 that carbon phases are the major electrochemically active sites for Li stor- 5 6 age. Temperature (T) and Larmor frequency (ω ) dependences of the 7Li 1 L 1. linewidth and spin-lattice relaxation rates T−1 are described by an activated 1 0 4 law with the activation energy E of 0.31 eV and the correlation time τ in A 0 1 : the high temperature limit of 1.3 ps. The 3/2 power law dependence of T−1 v 1 i X on ω which deviates from the standard Bloembergen, Purcell, and Pound L r a (BPP) model implies that the Li motion on the µs timescale is governed by continuum diffusion mechanism rather than jump diffusion. On the other ∗ Corresponding author. Tel:+49 (0)351 4659 801; fax: +49 (0)351 4659 313 Email address: [email protected](Seung-Ho Baek) Preprint submitted to The Journal of Power Sources January 9, 2014 hand, the rotating frame relaxation rate T−1 results suggest that the slow 1ρ motionofLionthemstimescale maybeaffectedbycomplexdiffusionand/or non-diffusion processes. Keywords: Nuclear Magnetic Resonance (NMR), 7Li dynamics, Anode, Lithium-ion battery, Silicon carbonitride, Polymer-derived ceramic 1. Introduction Polymer-derived ceramics (PDCs) that consist of Si, C, O, and N possess novel physical, chemical, and mechanical features which can be tuned by subtle changes of composition and/or microstructure as well as of processing conditions, finding applications in a variety of fields such as fibers, brakes for vehicles, sealants, coatings, and sensors [1, 2]. Among the Si-based PDCs, SiCN and SiOC are promising candidates for anodes for Li-ion batteries to replace graphite anodes. They possess a high discharge capacity compared to that of graphite, thermal and chemical stabilityagainstcorrosion,andcyclicstabilityduringcharge/dischargedueto stable 3D network structure with amorphous nature [3, 4, 5, 6, 7, 8]. Recent studies were focused on the carbon-rich SiCN ceramics due to their increased thermal stability and electrochemical performance that are attributed to the free carbon phase imbedded into their microstructure [9, 10, 11, 12, 13]. While most of these studies concentrated on the static properties of these materials, such as electrochemical performance, discharge capacity, and the relation between microstructure and precursor polymer structure, it is also very important to understand the Li ion dynamics on various timescales, characterizing the Li mobility and diffusion mechanism, and to get informa- 2 tion on the storage site of the Li. Nuclear magnetic resonance (NMR) has proven to be a powerful method for probing local structure and Li motions in numerous Li-containing ion conductors [14, 15, 16]. Although there have been a couple of solid state NMR studies performed on the SiCN and SiOC ceramics [17, 18, 19, 13], they are mostly concerned with structural aspects of these materials using the magic-angle spinning (MAS) technique, still lacking information on the Li dynamics as a function of temperature and frequency. In this paper, we report 7Li, 29Si, and 13C NMR studies of carbon-rich SiCN PDCs, adopting wide-line NMR method instead of its high-resolution solid state counterpart, providing information on the Li dynamics as well as on the Li storage site. While our data show that inserted Li ions mainly find carbons for their electrochemical storage sites, they also suggest that the mixed bond tetrahedra of Si which are formed in polysilazane-derived SiCN, act as an additional lithiation site. The spin-lattice relaxation rates as a function of temperature and frequency demonstrate that the Li mo- tion on a timescale of µs is precisely described by an activated law τ = c τ exp(E /k T). 0 A B 2. Sample preparation and experimental details Two SiCN ceramics derived frompolysilylcarbodiimide (HN1)andpolysi- lazane (HN3) (in the following denoted by SiCN-1 and SiCN-3, respectively) have been synthesized, as described in detail in Refs. [13, 20]. The elemental composition of the ceramics with regard to their carbon, nitrogen, oxygen, chlorine and hydrogen content was measured. The carbon amount was de- 3 termined by a combustion analysis with a carbon analyzer Leco C-200, the nitrogen and oxygen content by hot gas extraction with a Leco TC-436 N/O analyzer (Leco Corporation, Michigan, USA). The chlorine and the hydrogen content were measured at the Mikroanalytisches Labor Pascher (Remagen, Germany). The silicon content was calculated as the difference of the above mentioned elements to 100 %. The particle size distribution of the ceramic powders was measured with an analysette 22 COMPACT (Fritsch, Germany) which is working in a mea- surement range of 0.3 to 300 µm. The measurements were performed in ethanol at constant stirring and ultra sonic treatment. The specific surface area (SSA) and porosity of the samples were determined from the nitrogen adsorption and desorption isotherms with an Autosorb-3B (Quantachrome Instruments, USA) at 77 K using the Brunauer-Emmett-Teller equation and the Barret-Joyner-Halenda method, respectively. The ceramic powders were mixed with 7.5 wt% of CarbonBlack Super P(cid:13)R (Timcal Ltd., Switzerland) and 7.5 wt% polyvinyilidenfluoride (PVDF, SOLEF, Germany) dissolved in N-methyl-2-pyrrolidone (NMP, BASF, Ger- many). The slurry was spread on a glass plate and dried at 40 ◦C for 24 h, scratched off and ground. Approximately 100 mg of the as prepared powder was pressed uniaxially with 30 kN for 5 min to obtain pellets with a diameter of 10 mm and a thickness of roughly 0.8 mm. The pellets were dried under vacuum in a Buchi(cid:13)R Glas Oven at 80 ◦C for 24 h and transferred into a glovebox (MBraun, Germany). All together six samples have been prepared for NMR probing. One pellet of each ceramic was first fully lithiated and afterwards delithiated 4 following the later described procedure (in the following denoted as SiCN-1b andSiCN-3b). Asecond pellet ofeach ceramicwas onlyfullylithiated(inthe following denoted as SiCN-1a and SiCN-3a) and third pellet of each ceramic was prepared in the same manner as the above mentioned samples, however this samples was not tested electrochemically (in the following denoted as SiCN-1 as-prepared and SiCN-3 as-prepared). Electrochemical testing was done in a two electrode Swagelok(cid:13)R type cell with lithium foil (99.9% purity, 0.75 mm thickness, Alfa Aesar, Germany) as counter/reference electrode, 1 M LiPF in EC:DMC 1:1 wt% (LP30, Merck, 6 TM Germany) as electrolyte and a glass fiber separator (QMA, Whatmann , UK). Additionally a polypropylene separator (Celgard 2500, Celgard, USA) was placed between the glass fiber separator and the pellet to avoid con- tamination of the pellets with glass fibers. Lithiation and delithiation was performed with a VMP3 multipotentiostat (Biologic Science Instruments, France) at a current of 3.72 mA/g−1, which is equivalent to a C-rate of C/100 in terms of the theoretical capacity of graphite. Measurements were performedat25◦Candvoltage limitswere set to 0.005Vand3Vvs. Li/Li+. Afterwards the cells were disassembled in the glove box. The pellets were washed with DMC to remove the salt of the electrolyte and ground to a powder for NMR measurements. The NMR measurements were performed using Tecmag Fourier Trans- form(FT)pulse spectrometer. 7Li, 29Si, and13CNMR spectra were obtained using a spin-echo pulse sequence (π/2 τ π) which is more advantageous − − than the free induction decay (FID) method since it effectively eliminates spurious signals. For the 29Si and 13C, due to the weak signal intensity asso- 5 ciated with their low natural abundances and the long spin-lattice relaxation time T , their spectra were acquired only at roomtemperature for two phases 1 (SiCN-1 as prepared and SiCN-1a) and three phases (SiCN-3 as prepared, SiCN-3a, and SiCN-3b). For these two nuclei, the NMR spectra each were averaged more than 4000 scans, with a typical π/2 pulse length of 4 µs and a repetition time of 40 s. To probe the Li diffusion parameters, 7Li NMR spectra and relaxation rates were measured as a function of temperature in the range of 80–420 K. The spin-lattice relaxation rates T−1 in the laboratory frame and T−1 1 1ρ in the rotating frame were measured to study the Li motions on µs and ms timescales, respectively. For the T−1 measurements, the saturation re- 1 covery method was employed and T was obtained by fitting the relax- 1 ation of the nuclear magnetization M(t) to a single exponential function, 1 M(t)/M( ) = aexp( t/T ) where a is a fitting parameter. 1 − ∞ − In this study, since SiCN-1 and SiCN-3 exhibit very similar electrochemi- calperformanceaswellasNMRresultsat116.64MHz,thedetailedfrequency dependences of the linewidth and relaxation rates have been made only on SiCN-3a. 3. Experimental results and discussion 3.1. Characterization The results of elemental composition are shown in Table 1. The table also includes the amount of free carbon within the ceramics calculated according to the equation, (cid:0)x 1+ y + 3z(cid:1) M wt%free C = − 2 4 · C 100 (1) M +x M +y M +z M · Si C O N · · · 6 taken from Ref. [21] with x, y, z being taken from the empirical formula SiC O N and M , M , M and M being the molar mass of the corre- x y z C Si O N sponding elements. Both samples exhibit a high amount of free carbon and only little impurities of oxygen. The residual chlorine measured in the sam- ples isdue to endgroups ofchlorine at thesynthesized polymers andleftovers of the byproduct trimethylchlorosilane of the synthesis reaction. The evaluation of particle size distribution measurements lead to D 50 values of 11.6 µm for SiCN-1 and 10.6 µm for SiCN-3, respectively. Both SiCN ceramics demonstrate a non-porous character with a SSA lower than 10 m2g−1 and 15 m2g−1 for SiCN-1 and SiCN-3, respectively. 3.2. Electrochemical results Table 2 summarizes the lithiation and delithiation capacities of the inves- tigated samples and the coulombic efficiency η of the samples SiCN-1b and SiCN-3b calculated as the ratio of delithiation capacity to lithiation capacity times100. Thecorresponding voltageover capacityplotsareshown inFig. 1. Despite the low current applied for lithiation and delithiation, capacities for both materials are below the achieved capacities of printed electrodes. This discrepancy can be explained by the much higher thickness of the pellets needed for a suitable amount of powder for NMR probing. 3.3. NMR spectra at room temperature Figures 2, 3, and 4 show the spin-echo spectra of 7Li, 29Si, and 13C, respectively, acquired at room temperature in an external field of 7.0494 T (i.e. Larmor frequencies of 116.64 MHz, 59.624 MHz, and 75.476 MHz) for SiCN-1 and SiCN-3 samples. 7 The 7Li spectra shown in Fig. 2 reveal a narrow linewidth of 2 kHz with a small positive shift for both SiCN-1a and SiCN-3a samples. In practice, we do not observe any noticeable difference between the 7Li spectra of SiCN- 1 and SiCN-3, suggesting that the Li dynamics is weakly sensitive to the precursor polymers. We also measured the 7Li spectra in the discharged samples. Interestingly, we find that the Li ions associated with smaller shift largely remain after discharging for both samples. This finding suggests that it is more difficult to remove Li that occupies sites in isotropic surroundings with less chemical shifts. Both samples SiCN-1 and SiCN-3 yield symmetric 29Si spectra which are shown in Fig. 3 for the as-prepared and lithiated samples. It is interesting to note that for SiCN-3 the linewidth increases after charging and recovers that of the as-prepared after discharging, while there is no difference of the spectrum after charging for SiCN-1. This suggests that inserted Li ions are coupled to Si, leading to a larger distribution of the local field at the 29Si, i.e. alargerlinewidth, inSiCN-3. The factthatsuch abroadening doesnotoccur in SiCN-1 implies that local structures of Si that may act as electrochemical Li storage sites are formed in the SiCN ceramic derived from HN3 but not from HN1. While we observed a symmetric line for 29Si, the 13C spectrum is clearly resolved into two in the as-prepared SiCN sample, indicating the existence of two inequivalent carbon sites or environments as shown in Fig. 4. Since this carbon-rich SiCN ceramics contain a very high concentration of free carbon phases, we attribute these two peaks to the presence of sp2 and sp3 carbons [7, 20]. In fact, a previous high-resolution 13C NMR study in SiCN 8 revealed two groups of spectra associated with sp2 and sp3 carbons [22]. They are roughly separated by 110 ppm, which indeed accounts for the ∼ difference between the main peak at 200 ppm and the shoulder at 90 ∼ ∼ ppm. Therefore, we assign the main peak and the shoulder to sp2 and sp3 carbons, respectively. This implies that the small amount of sp3 carbons is still present even at the high pyrolysis temperature (1100 ◦C) in both SiCN-1 and SiCN-3. The insertion of Li ions affects the 13C spectrum significantly, reducing the anisotropy of the spectrum. This contrasts with the 29Si spectrum which is almost intact with Li insertion for SiCN-1. The large influence of Li inser- tion on the 13C spectrum indicates that electrochemically active sites for Li storage are mainly carbon phases. It would be interesting to recall a debate on the main Li storage in SiOC ceramics. Fukui et al. [18] suggested that the free carbon phase is the main site for Li storage, while Ahn et al. [21] proposed the mixed bond tetrahedra of Si as major Li hosting sites. It is known that HN1-derived SiCN-1 has no concentration of mixed bonds of Si unlike HN3-derived SiCN-3 [4, 17]. Thus, based on the almost identical spectra of 13C for both SiCN-1 and SiCN-3 samples after charg- ing/discharging, one can argue that carbons are the major electrochemical storage site for the Li ions, regardless of the preceramic polymers. The fact that the whole 13C spectrum, which reflects two inequivalent carbon phases, changes after charging/discharging in a reversible way implies that all carbon species participate in hosting Li ions and, once binding Li, both sp2 and sp3 carbons experience similar chemical environment surrounding them. While the 13C NMR results show the main role of free carbon phases as 9 the storage sites of Li, we note that the broadening of the 29Si spectrum with Li insertion occurs only for SiCN-3 as shown in Fig. 3. This suggests that the mixed bond tetrahedra of Si also act as an additional Li storage site, although its role is estimated to be only minor based on the fact that the 7Li spectra are almost identical for both SiCN-1 and SiCN-3 as shown in Fig. 2. 3.4. Temperature dependence of 7Li spectra and spin-spin relaxation rate Figure 5 shows the full width at half maximum (FWHM) ∆ν and the spin-spin relaxation rate T−1 of 7Li in lithiated SiCN samples as a function 2 of temperature at two Larmor frequencies. TheFWHMincreasesrapidlywithdecreasing temperatureandundergoes a crossover to a linear T behavior below 190 K. Also we find that the spin- ∼ spin relaxation rate T−1 follows the same T-dependence as the FWHM, as 2 expected in the slow motion regime where T−1 ∆ν. Although the FWHM 2 ∝ and T−1 increase linearly below 190 K instead of reaching a plateau which 2 would indicate the rigid lattice (RL) regime, it should be noted that the linear slope below 190 K for SiCN-3 is greatly reduced at a lower Larmor frequency of 33.1 MHz, while the data at T > 190 K are almost independent of frequency. Furthermore, we find that the slope for SiCN-1 and SiCN-3 at the same frequency of 116.64 MHz remains the same, as indicated by solid lines, despite their different T-dependence above 190 K. These results strongly suggest that the T-linear increase below 190 K could be an extrinsic effect being proportional to the external field strength. Therefore, we take 190 K as the onset temperature T of the RL regime, above which the onset hopping diffusion of Li ions takes place leading to motional narrowing of the 7Li linewidth. 10