Graphitic microstructure and performance of carbon fibre Li-ion structural battery electrodes

Three types of PAN-based CFs were used; the IM fibres T800 and IMS65 and the HM fibre M60J (table 1). The choice to study IM and HM fibres is motivated by the substantial difference in their turbostratic graphitic structures. The HM fibre shows higher lattice-dependent properties, i.e. tensile modulus and electrical and thermal conductivity, but lower strength and strain to failure than the IM fibres. IMS65 and T800 IM fibres were studied as they have the highest electrochemical capacity of the CFs studied to date [6–8]. The carbon content in the fibre is controlled by the final heat treatment temperature (HTT) [15]. As the same final HTT interval is reported for the two IM fibres, the carbon content of IMS65 is assumed to be the same as for T800, i.e. 96%.

The basic electrochemical properties of these CFs, average delithiation capacities and coulombic efficiencies, were reported by Hagberg et al [8] (table 2). The low capacity M60J HM fibre was chosen in order to elucidate how its structure differs from a high capacity fibre.

Previously, Jacques et al [16] measured the maximal free expansion in the axial fibre direction due to lithiation for T800 and IMS65 cycled at 0.11 C and 0.08 C, respectively. Electrochemical cycling at 1 C refers to full charge in one hour. The results showed that the fibres underwent longitudinal free expansion of 1.05% and 1.03%, respectively, while no data have so far been reported for M60J.

Here we therefore report on M60J using the same method as in [16]; measuring the change in reaction force on a CF tow at constant strain during electrochemical cycling, in the knowledge that the tensile stiffness remains unchanged during electrochemical cycling [17]. Tensile specimens were manufactured from dry M60J 6 K fibre tows used as supplied, according to the method detailed in [16] (figure 1(a)). Three pristine specimens of M60J were used for reference tests, which were repeated after electrochemical cycling. Each test was carried out at a displacement rate of 0.1 mm min⁻¹, with a sampling rate of 10 Hz, up to a maximum load of 250 N. The stiffness was extracted from the linear section of the measured force-displacement curves using a least-squares fit.

Zoom In Zoom Out Reset image size

The M60J specimens were also used as electrodes in Li-ion battery pouch cells, made according to [16] but with 250 μl rather than 150 μl of LP40 electrolyte (figure 1(b)). Electrochemical cycling of these pouch cells was monitored using a BioLogic VSP potentiostat controlled with BioLogic EC-Lab software. Each cell was subjected to 5 galvanostatic cycles, 0.002–1.5 V versus Li⁺/Li° at an applied current of 0.1686 mA, corresponding to a 0.1 C charge rate based on the theoretical maximum capacity for graphite of 372 mAh g⁻¹, and m_CF = 4.53 mg. The HM M60J fibres are significantly stiffer than the IM fibres used in [16], and this combined with the load restriction of 250 N on the microtester meant that the pouch cell bag had to be repeatedly re-strained in order to relax sufficiently to allow the entire amplitude of intercalation strain to be captured. A reference cell was made in order to quantify the force relaxation of the pouch cell bag, with the CFs cut prior to sealing the pouch. This cell was then subjected to a constant strain equivalent to that experienced by the CFs in a typical expansion test, while the force relaxation was measured.

To measure if the CFs were under tension at maximum intercalation strain the piezo-electrochemical transducer (PECT) effect described in [18] was utilised. When fully intercalated a cyclic force of amplitude ΔF = 20 N was applied and the open circuit potential (OCP) measured. The CFs were confirmed to be in tension if the OCP responded linearly to the cyclic force. In order to calculate the intercalation strains the force/time relaxation curves of the pouch cells were corrected for the reference cell, and the force changes ΔF of the lithiation expansion (positive) and delithiation contraction (negative) were obtained. The intercalation strains ε were calculated using equation (1), where k is the stiffness of the fibre tow.

$$\begin{eqnarray}&&\varepsilon =\displaystyle \frac{{\rm{\Delta }}F}{k}\end{eqnarray} \tag{ 1 }$$

After each expansion test the specimens were removed from the pouch cells and subjected to a tensile test using the methodology described above in order to verify the stiffness measurement.

A Titan 80–300 TEM, operated at 300 kV, was used to analyse the carbonaceous structure in the cross-sections of the CFs. In order to prepare TEM specimens with a thickness of less than 100 nm, the in situ lift-out technique [19] was employed using an FEI Versa 3D workstation, a combined focused ion beam microscope and SEM (FIB/SEM), equipped with an Omniprobe micromanipulator. To protect the very surface of the CF from ion damaging during the specimen preparation procedure, a layer of Pt was deposited with the aid of the gas injection system in the workstation (figure 2(a)). Subsequently, a small piece was cut out using the FIB, transferred to a 3 mm TEM grid using the micromanipulator, and finally ion polished down to less than 100 nm thickness with the FIB (figure 2(b)).

Zoom In Zoom Out Reset image size

2.4.1. Ex situ Raman spectroscopy

All ex situ Raman spectra were obtained using a Dilor Labram micro-Raman spectrometer equipped with a confocal microscope (Olympus BX40) and an objective (Olympus LMPlan FI) with a 50× magnification and a 0.5 numerical aperture. All the measurements were performed at 23 °C by employing a HeNe laser (632.82 nm) as excitation source. The ca. 5 μm diameter laser focus allows for analysis of lateral surfaces and cross-sections of single CFs.

Data were collected using a 100 μm slit, 1800 lines/mm grating and no filter was applied. Spectra were acquired in back scattering mode with a charge-coupled device (CCD) collector for both first- and second-order regions: 900–1760 cm⁻¹ and 2320–3000 cm⁻¹, respectively. To increase the signal-to-noise ratio (S/N), all spectra contained 40 acquisitions with an exposition time of 30 s each.

All spectra were analysed using the commercial software PeakFit v.4, by Systat Software Inc. The spectra were corrected through the subtraction of a linear baseline to eliminate the fluorescence of the background and a Savitzky-Golay smoothing was applied to further improve the S/N. Peaks were fitted for Raman shift, intensity, area and width using a Voigt function. For comparison purposes, the intensity of each spectrum was normalised to the band of highest intensity and the spectra were shifted arbitrarily along the vertical axis to give a clearer representation. Since the maximum experimental window for a spectrum is 900 cm⁻¹, four windows were partially overlapped to obtain the full spectrum of the same spot on the surface.

The ex situ Raman spectroscopy of single virgin CFs monitored the shifts of the D, G and G' bands for IMS65 and M60J as a result of macroscopic axial tensile strain—to allow us to exclude shifts caused by free expansion of the CF in the in situ tests. The single CFs were fixated on an aluminium foil frame with a cyanoacrylate adhesive (Loctite Superglue) and the frame mounted in a tensile rig operated manually by screws (figure 3)—all put in the translational stage of the Raman spectrometer. A Raman spectrum was first acquired on the unloaded fibre and subsequently in intervals on the strained fibre whereby the positions of the Raman bands could be analysed versus the applied strain and the strain sensitivity of the Raman shift determined by a linear fit. To avoid any interference from the clamping forces all spectra were acquired from areas centrally on the CFs, away from the mounting points.

Zoom In Zoom Out Reset image size

2.4.2. In situ Raman microscopy

In situ Raman microscopy was performed in order to follow the changes in the Raman spectra as a function of the lithium insertion and extraction in IM T800 and IMS65 as well as HM M60J, ultimately providing information on both the insertion mechanism and the electrode state during the charge/discharge process.

A small CF bundle (ca. 5 mg) was dried overnight in a vacuum oven at 80 °C. Subsequently a spectro-electrochemical cell, described in detail in a previous work [20], and derived from a previous cell used for x-ray diffraction [21] and x-ray absorption spectroscopy [22] was assembled inside an Ar-filled glovebox. During this procedure the dry CFs were spread on a copper (Cu) foil current collector, putting the foil at the top of the cell in contact with the Raman-transparent quartz window (figure 4). Small holes were made in the Cu foil to allow the laser beam to reach the CF electrode. A glass fibre fabric was used as separator (Whatman GF/A glass fibre filter) and Li-metal as the counter electrode. An LP30 liquid electrolyte, 1 M LiPF₆ in ethylene carbonate and dimethyl carbonate (EC:DMC, 1:1) was allowed to soak the electrodes and the separator. Due to problems of luminescence at low potentials, 1 M LiPF₆ in propylene carbonate (PC) was used for a few M60J tests, but as no appreciable improvements were observed, the majority of the tests used LP30 (table 3).

Zoom In Zoom Out Reset image size

Table 3.  Selected in situ Raman spectroscopy sample and test data.

CF	mCF (mg)	Electrolyte	Electrolyte volume (μl)	Applied current (μA)
T800	6.2	LP30	60	115
IMS65	4.0	LP30	75	50
M60J	7.4	LiPF6 in PC	60	138
M60J	7.7	LP30	60	143

The voltage between the two electrodes was measured under OCP until it stabilised and an equilibrium potential was reached. Electrochemical cycling was performed under galvanostatic conditions using a potentiostat/galvanostat (Gamry Instruments) between 0.002–1.5 V vs Li⁺/Li° at a rate of C/20 with respect to the theoretical capacity of graphite (i.e. 372 mAh g⁻¹). The Raman spectra were acquired at different voltages using an Ar laser with a wavelength of 514 nm and an initial power of 50 mW as excitation source (the power measured on the sample was less than 10 mW). Each spectrum was acquired in the range 1000–3300 cm⁻¹ for 20 min, which is a short time compared to the total duration of the charge/discharge process. The red- and blue-shifts discussed are all with respect to the Raman shifts. The Raman confocal microscopy set-up was a laser system (BeamLok 2060, Spectra-Physics) and a spectrometer (XY 800, Dilor), equipped with an optical microscope (BH2, Olympus) and a CCD camera (Symphony, Horiba Jobin Yvon).

The results from the virgin M60J CF specimens are used as a baseline for evaluation of the influence of electrochemical cycling as it is known that the stiffness remains unchanged during electrochemical cycling. This was confirmed with the post-cycling tensile tests that showed no measureable stiffness changes after electrochemical cycling.

Moving to the role of electrochemical cycling for M60J CF expansion, the combination of measurements of potential as a function of capacity, i.e. lithiation/delithiation, (figure 5(a)), and force response upon cycling (figure 5(b)) shows the highest intercalation strain, $\varepsilon =0.13 \% ,$ to occur for the first charge/discharge cycle $({\rm{\Delta }}F\approx 59\,{\rm{N}}).$ A lower capacity (<80 mAh g⁻¹, figure 5(a)) than previously reported for M60J (154 mAh g⁻¹, [8]) was obtained for all cells tensioned prior to cycling. It is suspected that this is due to the altered cell format which includes end tabbing of the CF tow, as well as possible transport issues caused by a large number of CF filaments in the tow [7]. Previous work has however shown that the intercalation strain is linear with capacity [16], and so a likely maximum strain can be extrapolated from these results.

Zoom In Zoom Out Reset image size

By using high-resolution transmission electron microscopy (HR-TEM) micrographs of the longitudinal cross-sections of the M60J, IMS65, and T800 CFs (figures 6(a)–(c)) including their convergent beam electron diffraction patterns (insets) the graphitic microstructural ordering can be analysed in detail. For M60J the size of the graphite crystals is relatively large, >300 Å along the longitudinal direction and stacking to >100 Å in thickness, with an average interlayer spacing of the (002) layers of 3.47 Å. The size of the electron beam focus is only a few nanometres, which means that the convergent diffraction patterns reflect the local crystallographic information from a volume of a few nanometres in width and approximately 100 nm in depth—and for M60J the patterns are sharp, small, and round disks, hence matching well with the HR-TEM data. In contrast both IMS65 and T800 show highly disordered structures, with a large number of very small crystals—for IMS65 with average length, L_a = 19 Å, and thickness, L_c = 28 Å, while for T800 the crystals are 18 Å both in length and thickness. The average (002) interlayer spacing is 3.49 Å for IMS65, close to the value for M60J, while a much larger spacing of 3.70 Å is found for T800. Here the (002) diffraction patterns appear as more diffuse arcs, which implies that within the small information collecting volume the orientation of the graphene layers varies significantly. All these data will serve as a basis for the discussions below of structure-property relationships, especially with respect to the role of microstructure for electrochemical capacity and Li-ion intercalation mechanisms.

Zoom In Zoom Out Reset image size

The basis for CF stress and strain analysis at the molecular level by the resulting Raman shifts is the anharmonicity of the carbon—carbon sp² bond and that any macroscopic deformation also involves stretching or compression of these bonds [23, 24]. For CFs a tensile stress applied in the direction of the fibre axis has been shown to result in a Raman bands shift to lower wavenumbers, while the opposite occurs under compression, and the stress/strain sensitivity to be proportional to the fibre modulus—a combined effect with layer slippage and crystal rotation, which does not affect the position of the sp²-induced Raman bands. The contribution of bond deformation to the total strain increases with crystal orientation along the fibre axis [25], which also leads to higher tensile modulus [12].

The Raman spectra analysis of the virgin T800, IMS65, and M60J CFs focus on the two main bands typical of carbonaceous materials with sp² hybridised carbon atoms: the D band (due to the ring breathing mode of A_1g symmetry) at ca. 1330 cm⁻¹ and the G band (due to in-plane bond-stretching) at ca. 1580 cm⁻¹. Very similar spectra are obtained for the IM CFs T800 and IMS65, while the spectrum from the HM CF M60J is distinctly different (figure 7). The Raman spectra are all in line with the literature [8], but for M60J all bands are sharper, indicative of a higher degree of order [26], and especially it renders the D' band clearly detectable at ca. 1620 cm⁻¹ as a feature together with the G band. Furthermore, the second-order region (>2000 cm⁻¹) is for M60J dominated by the presence of the G' peak at ca. 2659 cm⁻¹, both for the cross section (not shown) and the lateral surface, while there is no clear peak in this region for T800 or IMS65.

Zoom In Zoom Out Reset image size

The first-order Raman spectra acquired from the cross section and the lateral surface of the HM M60J show them to be remarkably different; the ratio between the D and the G bands (R = I_D/I_G) increases from 0.8 for the lateral fibre surface spectrum to 1.56 for the cross section (figure 8, table 4). No such changes are found for the IM fibres, for which R is practically constant.

Zoom In Zoom Out Reset image size

By using two different approaches, the Tunistra and Koenig relationship [27] and the measure suggested by Mallet-Ladeira et al [28], the average crystal length (L_a) can be estimated in two ways. The former approach uses the ratio R, while the latter uses the full width at half maximum for the G band (fwhm_G). For the cross section of M60J the two obtained L_a differ significantly, almost by 100%, while for all other spectra the average difference in L_a is less than 8 Å, <25% (table 4). In addition, for the IM T800 and IMS65 the fibre cross sections and the lateral fibre surface have more or less the same L_a, while for the HM M60J there is an ambiguity created by the two measures.

Having established a baseline for the Raman spectroscopy analysis by the measurements on virgin fibres above, the first-order region spectra for IMS65 and M60J CFs under tension show all bands to shift towards lower wavenumbers with applied strain (figure 9, table 5). While only the spectra at zero and maximum strains are reported for clarity the shifts are approximately linear versus applied strain, and in line with the literature for other CFs [24, 25], while no strain effects were observed for the bandwidths or the ratio R.

Zoom In Zoom Out Reset image size

Armed with these two sets of Raman data, for virgin and strained fibres, especially the strain sensitivity can now be compared to shifts resulting from electrochemical cycling—analysing molecular changes due to lithiation and how this affects the Raman spectra—and if/how it differs from the macroscopic tensile strain effects.

Prior to the in situ spectro-electrochemical Raman studies we need to establish the capacities for the different CFs—using the pouch cells created and noting that for the half-cells versus Li-metal charge and discharge refers to Li-ion (de-)insertion/lithiation into/of the CF and is associated with a change in the electrode voltage (or potential). The capacities for the first charge, i.e. the first delithiation: 231 mAh g⁻¹ for T800, 252 mAh g⁻¹ for IMS65, and 146 mAh g⁻¹ for M60J, all compare reasonably well to the literature (table 2, [8]).

The corresponding series of Raman spectra for the first discharge of the IMS65, T800 and M60J CFs initially show all the features found for the virgin CFs. For IMS65 (figure 10(a)) the D and G bands are found at ca. 1365 and 1591 cm⁻¹, respectively, broad and partially overlapping. For T800 the D and G bands are found at ca. 1366 and 1591 cm⁻¹, respectively, broad and partially overlapping, while the second-order region is almost featureless, and for M60J more distinct bands are present at 1352 cm⁻¹ (D), 1586 cm⁻¹ (G), and 2709 cm⁻¹ (G') (figures 10(b) and 11). Data from the discharge cycles of are provided in Tables as supporting information. The IM CFs T800 and IMS65 Raman spectra changed very similarly; upon initial discharge/lithium insertion the G bands red-shift by ca. 30 cm⁻¹ and 17 cm⁻¹ for T800 and IMS65, respectively, while the D bands do not follow any particular trend, and R increases at lower potentials.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In contrast, the HM M60J G band is initially observed to blue-shift from 1586 cm⁻¹ at OCP to 1605 cm⁻¹ at 145 mV and to broaden and have its intensity reduced. As the potential is further decreased the band continues to broaden and moves slightly back towards lower wavenumbers.

The D band is weak already in the OCP spectrum and upon discharge it broadens as its intensity is further decreased, while its position is constant. The ratio R has no clear trend with potential; from 450 to 150 mV it decreases, but this is merely a consequence of the strong reduction in intensity of the D band.

Finally, the G' band continuously broadens and red-shifts, from 2709 cm⁻¹ to 2665 cm⁻¹ upon discharge—hence in the opposite direction to the G band. Similar to the D band the G' band disappears at ca. 150 mV.

All the band shifts observed during the first lithiation (discharge) were almost completely recovered during the first delithiation (charge), and reappearing again during a second discharge cycle, which indicates a nicely reversible process.