3.1. Structural characterization of the pristine LFP electrode and of the Al2O3 coated electrodes
The chemical composition of the pristine LFP electrode and the same substrate coated with 5nm and 10 nm of Al2O3 by ALD was analyzed by XPS. The broad survey spectra are presented in Fig. 1. The surface composition of the pristine LFP is primarily dominated by signals from fluorine, carbon, oxygen, and phosphorus elements. The presence of iron is also detected, albeit partially embedded in the loss structures on the high binding energy (BE) side of the F1s peak. The Li1s signal is not easily discernible in the survey spectrum due to its position at 56eV, falling within the weak, low BE region of the broad spectrum. It's important to note that peaks related to the Auger transitions of electrons of O (OKLL) and F (FKLL) are also present in the survey. As expected, O1s, P2p, Fe2p, and Li1s core line signals originate from the LiFePO4 composite, while C1s and F1s signals arise from the carbon black and the polyvinylidene fluoride (PVDF) binder commonly found in commercial LFP electrodes. In contrast, the survey spectrum of LFP coated with 5nm of Al2O3 is dominated by signals of oxygen and aluminum, with significant reductions in fluorine and carbon signals, indicating the successful deposition of an alumina layer on the LFP surface (Validated with FE-SEM-EDS analysis, as shown in Fig. S.1 in the supporting information). Given the thinness of the 5nm layer and the XPS's 10nm sampling depth, the F1s peak from the underlying LFP is still observable, albeit significantly attenuated. However, the F1s peak nearly disappears in the wide spectrum acquired from the LFP sample coated with 10nm of alumina, as the coating thickness approaches the sampling depth of the technique, minimizing the contribution from the underlying LFP. Overall, Fig. 1 illustrates the successful deposition of Al2O3 on LFP substrates with varying and controlled thicknesses by ALD.
The high-resolution core lines of different elements were analyzed and deconvoluted to highlight the contribution of the various chemical states 32,33. Particularly in Fig. 2, we present detailed core lines of C1s, F1s, O1s, and Al2p for pristine LFP and LFP coated with 5nm and 10nm of Al2O3. The C1s peak acquired from pristine LFP (Fig. 2(a), green curve) exhibits a strong component at 284.5eV attributed to C-C sp2 bonds, commonly associated with carbon black in LFP electrodes. Additionally, components at 285eV (C-H bonds), 286.5eV (CF2-CH2), and 290.9eV (CF2-CH2) correspond to the PVDF binder typically added to LFP, with the 288eV component attributed to ether O-C-O or carbonyl C = O bonds from carbonaceous contaminations. In the F1s peak of the pristine LFP sample (Fig. 2(b), green curve), the sole component at 688eV corresponds to the C-F bond in the CH2-CF2 polymeric chain (PVDF). The O1s core line from pristine LFP (Fig. 2(d), green curve) displays a main component at 531.4eV attributed to lattice oxygen of LFP, with weaker components at higher BEs due to O-C-O and C = O oxidized states of carbon.
The core lines in Fig. 2 highlight the changes in LFP surface chemistry following Al2O3 deposition by ALD. As observed from the survey spectra comparison, Al2O3 coatings gradually obscure the native chemical structure of LFP as the alumina film thickness increases. This is evident in the C1s core line (red and black lines), where signals from the PVDF binder and conductive carbon black are significantly attenuated, leaving mainly C-Ox and C-H bond components, possibly from residuals of the ALD TMA reactant. Similarly, in panel Fig. 2(b), the original single C-F component in pristine LFP (green line) is greatly diminished in the coated samples (red and black lines), replaced by a lower BE component at around 685.5eV, attributed to Al-F bonds likely formed at the LFP-film interface. This contribution is notably weaker in the sample coated with 10nm of alumina (black line, panel b). The development of an Al2O3 coating on the LFP surface is further supported by Al2p and O1s core lines (Fig. 2(c and d)): deconvolution of the former acquired from coated samples reveals a main component at 75eV, the expected BE for aluminum oxide, with a minor shoulder at higher BEs due to Al-F and/or Al-OH bonds. Similarly, fitting of the latter from coated samples shows an upward shift of the main O1s component, consistent with the creation of Al-O bonds.
3.2 Electrochemical and post-mortem analyses
The cycling performance of pristine and Al2O3-coated LFP electrodes (5nm and 10nm coating) over 100 charge and discharge cycles at 1C is shown in Fig. 3(a). Initially, all three electrodes exhibit a similar capacity of approximately 103 mAh g− 1. However, after 100 cycles, the capacity of pristine LFP electrodes and 5nm and 10nm Al2O3-coated LFP electrodes decreases to around 59, 69, and 64 mAh g− 1, respectively. Notably, the capacity retention of the 5nm Al2O3-coated LFP is 66.7%, surpassing that of the other two electrodes. To investigate the origin of the improved cycling performances imparted by the Al2O3 film, top-view SEM analysis (Fig. 3(b)) was performed on the coated and uncoated electrodes in the fresh state and after cycling. After 100 cycles, the original LFP electrode exhibits agglomerates and protrusions on its surface, hindering electrolyte penetration into the internal active materials and impeding the channel for Li+ transportation. Conversely, when employing coated LFP electrodes, the morphological changes are less pronounced, indicating that the protective layer ensures a stable interface throughout the charge and discharge processes. Consequently, LFP electrodes with coated layers demonstrate both enhanced capacity and better cycling stability.
Figure 4 illustrates the rate performance of the LFP-based electrodes at different C-rates. For each cathode electrode, the discharge capacities exhibit a clear decline as the C-rate increases, a phenomenon commonly observed in LIB electrodes 34. The dominant factor influencing this behavior is the diffusion of lithium ions within the cathodes at very low rates, which plays a crucial role in the lithium insertion process 35,36. This leads to the formation of a concentration gradient for lithium ions within the cathodes during discharge. Consequently, the cathode potential rapidly declines to the cutoff potential once the cathode surface completes the discharge process. However, the central region of the cathodes may not achieve full discharge, contributing to a reduction in cathode usage efficiency. This decrease in efficiency is exacerbated when higher current densities are applied, inevitably leading to a lower discharge capacity. Here, at C-rates lower than 2C, the electrodes exhibit very similar behavior. However, at 2C and 5C, the pristine sample demonstrates a higher capacity compared to the coated electrodes. This difference may be attributed to the protective layer on the LFP surface in the case of coated electrodes, which increases resistance and enhances the transport range of Li ions to LFP electrodes. Nevertheless, the 5nm Al2O3-coated LFP electrode demonstrates better reversibility when the current rate returns to 0.1C after 50 cycles.
The market share of low-voltage LFP cathode is currently increasing due to its economical nature, high thermal stability, lack of toxicity, and safety features. Although its discharge voltage plateau is lower than that of other cathodes such as LiCoO2, the stable discharge voltage of LFP within the electrochemical stability window of existing electrolyte systems serves to prevent electrolyte decomposition and the development of a Solid Electrolyte Interphase (SEI) 37. The stability of LFP is credited to its olivine structure and secure P-O bonds 38. Despite the overall stability of low-voltage cathode materials like LFP, during overcharging the small size of LFP nanoparticles can trigger adverse reactions, leading to SEI formation and degradation in LIBs’ performance 39. This may result in local heating, causing cell destruction with potential ignition. In this work, in addition to the evaluation of the performance of both pristine and 5 nm Al2O3-coated LFP electrodes within the voltage range of 2.5 to 4.2 V, an exploratory investigation has been initiated to understand the impact of increased upper cutoff potentials, specifically set at 4.5 V, on the cyclability of LFP electrodes. As depicted in Fig. 5, the coated sample exhibits superior performance in comparison to the pristine LFP, indicating that the Al2O3 coating might efficiently shield the LFP electrodes from additional degradation.
To gain a better understanding of the electrochemical performance of the samples, we conducted EIS on both pristine and 5nm Al2O3-coated LFP electrodes. After stabilizing the cell, the electrode was completely charged and discharged at 0.1C, employing a different discharge cut-off voltage (Fig. S2). Figure 6(a, and b) display the Nyquist plot for EIS data recorded at various lithiation states of both electrodes. At intermediate frequencies, it is possible to see a depressed large semicircle which corresponds to the superposition of two semicircles related to the CEI layer formation and the charge transfer of the faradaic process, finally at low frequencies a straight line due to diffusion is observed 40. To better distinguish the semicircles, the Bode plot of cells at 50% DOD is provided as an example in the supplementary information (Fig. S3). To obtain the electrochemical parameters, the experimental data was fitted using an equivalent electric circuit (Fig. S4), in which Re, RCEI, and Rct are the resistances of the electrolyte, CEI and charge transfer, respectively, while CPE1, and CPE2 are the constant phase elements used to obtain the effective capacitance values CCEI and Cdl related to the CEI and the double layer, respectively, with the aid of the Brug equation, Eq. (1) 41. The W element represents the Warburg impedance associated with the diffusion process.
Ceff = [Q(Re−1 + R− 1)(α−1)](1/α) (1)
Our findings indicate that the overall impedance of the pristine sample is higher than that of the coated sample as the discharge process approaches its completion (Depth of Discharge, DOD of around 80%, and 100%). Notably, the CEI resistance in the pristine sample (Fig. 7(a)) exhibits more variation at different states of lithiation, whereas the coated sample maintains a RCEI of approximately 58 Ω across various discharge states (Fig. 7(b)). Table S1 and Fig. 7 present the numerical values of Re, RCEI and Rct, CCEI, Cdl, and W obtained from fitting the EIS spectra at different lithiation states.
Figure 6. Nyquist plots and fitting lines of the EIS spectra at different states of lithiation for Pristine LFP electrode (a) and 5 nm Al2O3-coated LFP electrodes (b) are shown. The discharge was performed at 0.1C after 100 cycles at room temperature.
Since the LFP electrodes used in all cell tests underwent no modification in their preparation, it is reasonable to assume that the differences in the RCEI values are related to processes on the CEI layer. Therefore, it is possible to verify that the coated electrode presents RCEI almost 10% lower than the pristine electrode, indicating a more stable CEI layer. Furthermore, it was also observed that Rct increases along with the DoD for both electrodes, indicating that the delithiated form has higher electrical conductivity than the lithiated form, consistent with previous reports for LFP-based batteries 42,43. Additionally, the coated electrode presented resistance values in the same range as the pristine, which points out that the additional non-conductive 5 nm layer of Al2O3 does not substantially hinder the electrode conductivity.
Transport properties of lithium ions play a crucial role in the performance, durability and efficiency of LIBs, if the ion mobility is hindered or slowed down, it can lead to degradation of electrode materials and loss of capacity over time. Therefore, EIS was used to estimate the diffusion coefficient of Li+ (DLi+) for both pristine and 5 nm coated electrodes to assess possible side effects from the alumina deposited film. The Warburg impedance can be used to obtain (DLi+) by means of Eq. (2) 44,45.
D Li+ = ½[(Vm/FAσW)(dE/dx)]2 (2)
Where Vm is the molar volume of LiFePO4 (43.87 cm3 mol− 1) 46, F stands for the Faraday constant, A is the surface area of the electrode (0.79 cm2), σW is the Warburg coefficient obtained by the fitting, and (dE/dx) is the slope of the discharge curves at each composition. Table S2, and Fig. S5 show the calculated DLi+ for both pristine and 5 nm Al2O3-deposited electrodes at different DoD. Based on the analysis, it is possible to verify that the more lithiated region presents higher ionic mobility, when compared to low lithiated DoD, which might be related to distortions in the lattice parameters and consequently expansion of internal channels in the LFP olivine structure when the content of Li+ increases 47. Furthermore, since the lithium insertion starts from the periphery of the electrode during the reaction, the results show that the additional 5 nm Al2O3 coating did not cause significant drawbacks in the ionic diffusion during the lithiation process 43.
In addition to analyzing the electrodes at room temperature, the cycling performance of the electrodes at elevated temperatures is considered a crucial criterion. Figure 8 illustrates the electrochemical behavior and structural changes of both uncoated and coated electrodes during cycling at 0.5C and 40°C. According to Fig. 8(a), coated LFP electrodes with 5nm and 10nm Al2O3 exhibit more stable behavior compared to the pristine electrode. In the cell with pristine LFP particles, there was a capacity fluctuation after 40 cycles of charge/discharge, likely due to the unstable reaction between the electrode and electrolyte. Consequently, uncontrollable side reactions between the electrolyte and LFP cathode led to electrolyte decomposition, resulting in an unstable interface layer. In contrast, cells fabricated using Al2O3-coated LFP electrodes exhibit stable behavior, indicating that side reactions at the interface were suppressed by Al2O3 coating. Among all cells, the LFP electrode with a 5nm Al2O3 ALD coating shows the highest capacity. The sample with a 10nm coating also demonstrates good performance but with less capacity compared to the 5nm Al2O3-coated LFP, possibly because the thicker Al2O3 coating suppressed undesirable side reactions while extending the lithium-ion transport pathway, resulting in increased resistance. Figure 8(b) illustrates the charge/discharge curves of the electrode at the 50th cycle. The overpotential of the pristine sample is higher than the other two coated electrodes, consistent with the cycling performance. Moreover, top-view SEM images of the electrodes (Fig. 8(c)) after cycling at 40°C display morphological change and crack generation on the pristine LFP surface, while the coated LFP surface undergoes minimal changes. These results further confirm that Al2O3 coating can suppress undesirable side reactions between LFP and the electrolyte, leading better cycling behavior.
XPS analyses were conducted on selected LFP samples, both pristine and coated with 5nm Al2O3 thin films by ALD, both before and after cycling at room temperature and at 40°C. These analyses proved invaluable in exploring the surface chemical changes resulting from cycling and elucidating the mechanisms influencing the varied performance of LFP cathodes coated with Al2O3 thin films compared to their pristine counterparts. In Fig. 10, the C1s, F1s, and P2p core lines (panels a, b, and c, respectively) are displayed, acquired from pristine LFP and the same samples post cycling at room temperature and at 40°C. A comparison among the various samples distinctly indicates a shift in the chemical composition of the cathode surface after cycling, as anticipated, attributable to the formation of a CEI. These chemical alterations likely stem from a combination of LFP degradation and side reactions within the electrolyte phase. Particularly in Fig. 9(a), the shape and deconvolution of the C1s core line suggest that signals related to carbon black and binder (PVDF), which constitute the fingerprint of the C1s peak on the pristine reference LFP, persist after cycling, albeit with a change in the ratio of their intensities. Conversely, all cycled samples exhibit an increase in components associated with various oxidized carbon species (COx), notably C-O or C-OH bonds around 286-287eV and O-C-O or C = O bonds around 288eV, likely resulting from electrolyte oxidation. This cycling-induced effect becomes even more pronounced when examining the F1s and P2p core lines. In Fig. 9(b), the F1s component around 688eV, originating from the C-F bonds in the PVDF binder within LFP, remains in the same position pre- and post-cycling, but with reduced intensity, while a distinct shoulder emerges around 686eV. Drawing from several published results 48–51, this shoulder band may be associated with Li-F bonds in LixPOy-1Fz + 1 (F-rich fluorophosphates) or possibly LixPFy compounds deposited on LFP due to electrolyte/salt decomposition during charging. The origin and presence of such species are also confirmed in the P2p spectrum (Fig. 9(c)). In the P2p spectrum acquired from pristine LFP, the excellent signal-to-noise ratio facilitated the fitting of the peak with two components at 133.3eV and 134.2eV. The binding energy (BE) separation of 0.9eV aligns with the spin orbit separation of P 2p3/2 and P 2p1/2 components of phosphorus P5+ in LiFePO4. Conversely, in the cycled samples, this peak is diminished in intensity, while a second peak, at higher BE, emerges. The latter, falling around 137eV, characterizes Li-F-P bonds in LixPOy-1Fz + 1 and/or LixPFy compounds deposited on the cathode, once again indicating a side reaction in the electrolyte phase during cycling 48,50,51.
All the observed changes described above reasonably correlate with the formation of a CEI on the LFP surface. The presence of an Al2O3 coating cannot mitigate side reactions involving the electrolyte phase. In fact, analogous results were obtained by XPS analyses on samples coated with 5nm Al2O3 after cycling. On the other hand, the electrochemical results obtained showed that Al2O3 ALD-coated samples demonstrated better performance under cycling, indicating that alumina coatings may have some effects in protecting the LFP material from degradation processes induced by charging cycles. To investigate this point, we focused on the Fe 2p and Fe3p core lines, considering that Fe is the only element detectable on the LFP surface that could provide insights into the LFP structure, while all the others (Li, F, P, O, C) could originate from either the LFP or the electrolyte. It's important to note that the Fe signal suffers from overlapping with the bremsstrahlung loss tail of the F1s peak for the Fe2p core line and with the Li1s for the Fe3p core line. Nevertheless, a comparative study of the two XPS regions together can provide us with some important indications regarding the degradation or preservation of LFP in the absence or presence of a protective Al2O3 layer, respectively. The Fe2p core line of pristine LFP (Fig. 10(a), black curve) exhibits two peaks at 710.2eV and 723.9eV corresponding to Fe2p3/2 and Fe2p1/2, respectively, indicating the Fe2+ oxidation state of Fe, as expected in LiFePO4 [x1, x2, x3]. Additionally, two smaller peaks at higher binding energies (BEs) are present (near 715eV and 726eV), corresponding to satellite peaks typically arising in transition metal ions with partially filled d-orbitals. In contrast, the Fe2p spectrum of cycled LFP (Fig. 10(a), red and green curves) reveals an additional signal around 713eV, attributed to a different oxidation state of iron, Fe3 + in particular. This finding suggests a degradation of the pristine LFP cathode chemical structure, leading, with cycling, to the formation of inactive Fe3+ containing species (e.g., FeOOH, Fe2O3, or FeF3). Conversely, the presence of an alumina coating on the LFP cathode surface appears to preserve the electrode from degrading reactions. This can be inferred from the fact that only the signal from the 2+ oxidation state of iron is present in the corresponding Fe2p lines (blue and purple lines) acquired on cycled alumina-coated LFP samples. A similar conclusion can be drawn by analyzing the Li1s/Fe3p region (Fig. 10(b)). The broad band deconvolution in the spectrum acquired on pristine LFP (black curve) reveals two main peaks around 54.5eV, attributed to the Li1s signal, and around 55.7eV, attributed to Fe3p (Fe2+) signal from LiFePO4. A weak shoulder at higher BEs is also present, typically attributed to Fe3p (Fe3+) from FePO4 species. Examining the charged LFP sample (red curve), a strong peak emerges after cycling at BEs higher than 58eV, attributable to highly oxidized iron, such as Fe3p (Fe3+) in FeF3 compounds. Concurrently, the Li1s signal diminishes, while a peak centered around 55.7eV persists. In this case, its origin is not solely from Fe3p (Fe2+) signal from LiFePO4, but also from Li-F bonds, which, as discussed earlier, develop during cycling as side reactions of the electrolyte. Similar components are observed in the spectrum acquired on the LFP sample cycled at high temperatures (green curve). Conversely, cycled LFP samples previously coated with 5nm of Al2O3 (blue and purple curves) exhibit a Fe3p/Li1s band where the signal from highly oxidized iron (Fe3+) is either absent or negligible. This finding aligns with our previous considerations regarding the mitigating effect of the alumina coating on the degradation of the chemical structure of LFP and the improved cycling performance demonstrated by the coated samples.