Nanoscale differentiation of surfaces and cores for olivine phosphate particles—a key characteristic of practical battery materials

We provide a review of our recent studies on the surface chemistries and electronic structures of olivine phosphate cathode materials LiMPO4 (M = Mn, Fe, Co, Ni). Li-depletion and mixed oxidation for the transition metal ions have been detected on particle surfaces, across the family of metal phosphate cathode materials. The effects of surface Li-depletion on optical properties and electronic band structures are discussed. LiFePO4 doped with metals or ions that are reported in the literature as beneficial for rate capability enhancement show more pronounced surface Li-depletion and mixed oxidation compared to undoped material. This outcome, among others, indicates that the dopant resides predominantly on, and influences, the surface of cathode materials.


Introduction
Efficient and economic energy conversion and storage have become critical to addressing current global environmental concerns and resource shortages. Li-ion batteries (LIBs) are reversible energy storage devices, that have drawn much attention from industry and academia in recent years. Since the original work of Goodenough et al on LiFePO 4 (LFP) [1], LiMPO 4 (M = Mn, Fe, Co, Ni) materials with olivine structure have been considered as among the most promising candidates for cathode materials due to their superiority in electrochemical performance, thermal stability and environmental compatibility. However, high-rate applications of phosphate materials have been greatly inhibited by their poor electronic conductivity (10 −13 -10 −16 cm 2 s −1 ) and slow diffusion of Li + (∼10 −9 S cm −1 ) [2][3][4]. Accordingly, multiple efforts have been made to improve the electronic and ionic conductivities of LiMPO 4 , such as by applying conductive surface coatings [5][6][7][8], particle size reduction [9,10], composite electrode preparation [11][12][13] and doping [14][15][16].
As the interface to the electrolyte and the gateway for Li-ion transfer, the surfaces of the electrode materials play an important role in the performance of LIBs [17], and their detailed characterization is receiving more focused research attention. Although many efforts have been made to improve the electronic and ionic conductivities of LiMPO 4 , the surfaces of the olivine phosphates still require detailed characterization. In this paper, an overview of our recent investigations on surface characteristics and their effects on the electronic band structures for LiMPO 4 materials is presented.
This article is based on the PhD thesis work of Ms Yin Zhang, recently awarded the degree [18]. Detailed descriptions of many sections in this review have already been published [19][20][21][22][23][24] or are in preparation for separate submissions. However, the authors in consultation with the journal editors, considered appropriate to write an overview that combines and connects, in a logical, coherent fashion, various aspects of this work. We also add indicative modelling studies that suggest an effective pathway to intentional design of battery materials.

Brief background to LiMPO 4
Olivine phosphates, typified by LFP, have orthorhombic unit cell with Pnma space group (number 62). The crystal structure symmetry can be represented as shown in figure 1 [25], containing four formula units. The transition metal ions with formal valency 2+ are the electrochemically active centres, which are oxidised to valency 3+ with the extraction of Li + and removal of associated electrons [26]. In the ordered olivine structure, the oxygen (O) ions form strong covalent bonds with phosphorus (P), in a stable three-dimensional framework, which provides safety under extreme conditions [27]. However, the strong covalent O bonds also lead to low ionic diffusivity and poor electronic conductivity [28]. Computational [29,30] and experimental studies [31] on LFP have suggested that the more favourable diffusion path for Li is along the b axis. This favoured diffusion path is a slightly curved, one dimensional chain, that can be easily blocked by impurity atoms [30].
Among the family of LiMPO 4 , LFP and LFP-based LIBs have been applied in industry for over 15 years [32]. LFP has a theoretical capacity of 170 mAh g −1 with a flat operating voltage at 3.45 V vs Li + /Li. The stability of the de-lithiated compound, FePO 4 , allows for the full withdrawal of Li ions [1]. The other olivine phosphates exhibit higher operating voltage (4.1 V for LiMnPO 4 (LMP) [33], 4.8 V for LiCoPO 4 (LCP) [34] and 5.1 V for LiNiPO 4 (LNP) [35]) making them potential cathode materials for advanced LIBs, once issues with electrolyte stability at higher voltage plateaus are addressed. However, the extremely low intrinsic electronic conductivities of these materials greatly inhibit their electrochemical performance. Therefore, it is critical to understand the transport mechanism in order to improve the rate capability of olivine phosphate materials.

Electronic band structure and transport mechanism
Polaron hopping, which is a thermally activated process, not directly associated with the electronic band structure, has been reported as the major transport mechanism for LFP [26]. However, improvements to transport properties, for example via doping, may modify the electronic band structure and enhance band contributions to the transport behaviour. It is therefore of practical interest to gather knowledge on the electronic band structure of LFP and related phases, even if only as a benchmark for subsequent modification of the structure.
An important means to evaluate electronic conductivity of semiconductors includes the electronic band gap; for the same reason, band gaps should also be investigated for battery materials. The band gap value is also a convenient parameter for comparisons between experimentally determined and calculated values; thus, validating source calculations. The experimental band gaps reported in the literature for a selection of LMPO 4 materials are summarized in table 1. There is large disagreement on the experimental band gap values for LFP and for its de-lithiated phase.  Note: UV-Vis = ultraviolet-visible spectroscopy, XAS = x-ray absorption spectroscopy, RIXS = resonant inelastic x-ray scattering, XES = x-ray emission spectroscopy. Adapted from Zhang et al, RSC Advances [19], published by the Royal Society of Chemistry. materials have not been determined and require further investigation. Therefore, we have made systematic efforts to gain further knowledge on the electronic band gap of various transition metal olivine phosphates [19,21,23]. Investigations on LFP and its de-lithiated phase have been revisited [19] and Tauc plots from this work are shown in figure 2. Unlike the results reported in previous literature [36,37], which appear to be measured on LFP samples synthesized using a solution or hydrothermal method [19], for nanoparticles with lower carbon content the absorbance rises gradually in the range of 1.7 eV-5 eV, with no sharp absorption edges in the measured energy range. The large Urbach tails shift greatly with the presence of small nanoscale amounts of additional carbon, which makes the large Urbach tails look like a surface related absorption from comparable nanoscale surface dimensions. These observations are in contrast to previous studies using alternative preparation methods for LFP. Therefore, a detailed investigation of the surface chemistry was undertaken using x-ray photoelectron spectroscopy (XPS) and Raman spectroscopy. As illustrated in figure 2, a mode related to the antisymmetric stretching of PO 3− 4 anion due to Li deficiency and a substantial fraction of Fe 3+ (where nominally there should only be Fe 2+ ) has been detected using both forms of spectroscopy. These data confirm the existence of Li depletion and Fe 3+ on the surface of LFP.
These types of investigations have been extended to the family of LiMPO 4 [21,23]. The presence of both surface Li-depletion and TM 3+ have been further confirmed with synchrotron-based soft x-ray absorption spectroscopy (sXAS), as shown in figures 3(a)-(d). The total electron yield (TEY), partial electron yield (PEY) and total fluorescence yield (TFY) modes have been used to collect the near-edge x-ray absorption fine structures (NEXAFS) spectra. The TFY mode collects signal from the fluorescent x-rays with an escape depth of ∼3000 Å, whereas the EY modes collect signal from the Auger electrons with an escape depth of 50 Å. Consequently, the TFY mode gives information about the bulk, while the EY mode yields information near the surface [42]. The PEY mode filters Auger electrons with higher energy, which generally originate from the very surface of the particles, which makes PEY mode even more surface-sensitive compared with the TEY mode [43]. Therefore, both surface and bulk information can be obtained by simultaneously acquiring the PEY, TEY and TFY signals. Since the technique is depth sensitive [42,43], the line shape difference within different modes indicates the difference in oxidation state of a TM ion with respect to the detection depth.
The effect of surface chemistry on the electronic structure has also been investigated by comparing the diffusive reflectance spectra of the pristine and ground LiMPO 4 samples. As shown in figure 3(e), after grinding, the intermediate absorptions in the visible energy range are weakened significantly and the main absorption edges are shifted to higher energy ranges for all the LiMPO 4 samples. After all the evidence is taken into consideration and following a similar perspective to that for LFP, band gaps of 4.6 eV for LCP and 5.1 eV for LNP can be determined.

Experimental validation of EBS calculations
With the experimental band gaps obtained in preliminary studies, the functionals within density functional theory (DFT) have been validated. The full comparison of GGA, GGA + U, HSE06 and sX-LDA have been reported in separate publications [19,23]. In this brief review, a comparison of HSE06 and sX-LDA, two recommended hybrid functionals for more accurate band gap calculations [44,45], is highlighted. The estimated electronic band structures and density of states with antiferromagnetic configurations are shown in figure 4.
Band gap estimations with the sX-LDA functional show a better agreement with the experimental optical gaps for all the olivine phosphates, except for LNP, which appears to have reported crystal information files (.cif) with small differences between reported values and slightly large weighted profile refinement R-factor [46][47][48]. Evaluating the structure with DFT, it also appears that the atoms represented by the .cif file have large residual non-equilibrium forces. A calculated band gap value for FePO 4 using sX-LDA (3.3 eV) matches the sharp optical absorption edge better than previously reported calculations [19]. Calculated values for LFP using sX-LDA (6.2 eV) also matches electron energy loss spectroscopy results very well. Remarkably, ultraviolet photoemission spectroscopy results, which allow for ionization potential and work function estimations, indicate that LFP has a negative electron affinity; that is, its vacuum level is below the conduction band minimum. In other words, the ionization potential is smaller than the band gap. A  Reprinted (adapted) with permission from [23]. Copyright (2020) American Chemical Society. measured ionization potential of 5.98 eV further confirms that the calculated and experimentally determined band-gap value of 6.2-6.3 eV (>5.98 eV) is likely to be a more correct determination of LFP properties [19].
Since Li extraction is accompanied by the removal of electrons from the valence band maximum (VBM), a study of the VBM is also important. The VBM of all olivine phosphates is dominated by TM-3d states hybridizing with O-2p states with HSE06 functional calculations, while O-2p states dominate the VBM of LNP calculated with the sX-LDA functional. This apparent dichotomy may also arise from the abovementioned LNP .cif files, with large residual non-equilibrium forces, affecting the electron density distributions.
Besides the electronic band structure alignment, Li intercalation voltages have been used for further comparison and validation of these approaches to delineate electronic behaviour of olivine phosphates. The estimated Li intercalation voltages of LiMPO 4 are compared with experimental values in figure 5. Overall, calculated results obtained with the sX-LDA functional show the best accuracy for the estimated Li intercalation voltages of LMP, LFP and LCP, except for LNP. Figure 6. Schematic of the core-shell structure and its effect on the electronic band structure. Reprinted (adapted) with permission from [21]. Copyright (2020) American Chemical Society.

Surface chemistry, optical absorption and electronic band structure
As mentioned above, surface chemistry has a profound influence on the optical absorption properties of LiMPO 4 . Hence, this chemistry also affects the electronic band structure; an understanding of which is critical to detailed knowledge of electronic transport mechanisms and of band alignment between cathodes and electrolytes. A schematic showing this influence is illustrated in figure 6. As shown in the estimated electronic band structure, LiMPO 4 materials with perfect crystal structure are expected to have clear band gaps, that give sharp, well-defined absorption edges in optical absorption spectra. Intervalence charge transfer has been confirmed experimentally and explained theoretically in the literature [ ) PO 4 solid-solution, as the Fe 3+ -3d states fall in the gap between Fe 2+ -3d states. A similar story can also be expected for the mixed oxidation states of TM on the surface, resulting in localized TM-3d impurity states inside the band gap of bulk LiMPO 4 . This condition would lead to an intermediate absorption peak value and large Urbach tails in the optical absorption spectra of these samples.
In earlier investigations [26] the ionic and electronic conduction mechanisms of LiMPO 4 are considered to be via diffusion of Li vacancies and hopping of small polarons. The small polaron in olivine phosphates can be considered as a d-hole on TM 3+ that can hop onto a neighbouring TM 2+ (and turn it into TM 3+ leaving behind TM 2+ ). Therefore, surface chemistry, such as the concentration of Li vacancies and the ratio of TM 2+ /TM 3+ , must have a significant impact on the surface electronic and ionic conduction of LiMPO 4 , and ultimately on bulk electrochemical performance.

Preferential surface doping of LFP
After re-evaluating the properties and characteristics of pristine materials, a series of cation doped LFP, which was previously reported as beneficial to the rate capability, has also been investigated [22]. The surface chemistry has been characterized with XPS and sXAS as shown in figure 7. The concentration of Fe 3+ on the sample surfaces increases with the addition of cation dopants, while the cores remain similar to the original stoichiometric samples. This relationship suggests that the effect of cation doping is more significant on the surface of the LFP than within the bulk. The LiMPO 4 olivine structure has been reported as having no tolerance for aliovalent doping on either Li or TM sites due to high solution energy [49], although there are numerous reports on experimental doping success, resulting in improved rate capability [14,[50][51][52][53][54][55][56]. We suggest it is likely that the cation dopants are pushed to the particle surfaces during phase formation, which intensifies the surface distortion of LFP particles. Similar surface accumulation of dopants has been reported in Fe substituted LiCoPO 4 samples [57]. Uniform distribution of the dopants can only be achieved when the sample is fast annealed (heating up to 650 • C in 3 min and cooling down to room temperature in 20 min) [57]. Furthermore, significant drops in charge transfer resistance and polarization have been found for the doped LFP samples in our study. These effects suggest a more favourable surface for charge transfer has been obtained [20] due to surface dopant segregation.

Junction effects on particle surfaces
As mentioned above, DFT calculations of the electronic band structures for LFP and FePO4 using the sX-LDA functional closely match experimentally measured values for their respective band gaps. This convergence of calculation and experiment also applies to results on work functions and ionization potentials for these compounds [58]. Based on these results, the schematic in figure 8 summarizes the approximate relative positions of energy levels for the core and the surface of LFP particles: (a) prior to, and (b) after, placing them in contact. The FePO 4 band structure has empty levels just above the Fermi level, due to an odd d 5 configuration of the Fe-ions [19]. Some electrons from LFP will be transferred to the FePO 4 , resulting in an effective p-doping and n-doping of LFP and FP, respectively. Some possible p-n junction effects in LFP battery materials have been discussed in an earlier publication [59].
Carbon coating is also part of the surface nanostructure of LFP materials [24]. The work function of carbon is reported to be very dependent on the specific type of carbon. For example, the work function can vary within a wide range covering <1 eV and about 5-6 eV [60]. In particular, it has been shown that doping of carbon itself with other ions is able to reduce the value of the work function [61]. Bulk analyses of LFP materials typically undertaken using inductively coupled plasma optical emission spectroscopy analyses, often show a slight excess of Li [20,23]. Since this excess Li is not in the core LFP structure, let alone on the de-lithiated surface (as shown above), it is likely that small amounts are doped into the amorphous carbon coating. Preliminary DFT calculations on work functions for different carbons indicate values around 5 eV for either graphitic or amorphous carbon without doping. Li-doping appears to reduce the calculated work function value to ∼3.5 eV. Such work functions, if equal or under 3.5 eV, will position the Fermi level of the amorphous conducting carbon ideally for good ohmic contact [62]. Good ohmic contact is no doubt an essential condition for improved electrochemical performance. More detailed theoretical and experimental investigations on work functions of amorphous carbon are currently underway and will be published separately.

Conclusions and perspectives
The effects of surface chemistry on the electronic and ionic conductivities of LiMPO 4 compounds are real, detectable and calculable. Thus, surface nanostructures and their optimization should be considered at every level of material design, characterization, modification and testing. A more recent study on the solid-solution Li 0.5 FePO 4 has also confirmed the importance of particle surfaces in electrochemical processes. Li ions have been found to migrate along the solid/liquid interface, without leaving the particle surfaces [63]. This mechanism appears to take place during both lithiation and de-lithiation and to control the phase transformation rate in Li x FePO 4 .
Since surface differentiation from the core has been observed in this research, junction effects between the surface and core of particles, and even the conductive carbon surface coating, may be expected. Thorough and quantitative understanding of the contributions from all the components and interfaces on the electronic structure and transport mechanisms, including all the electronic band structure, work function and ionization potential alignments, is required.
Our preliminary studies on lithium nickel manganese cobalt oxides (NMC) show that similar differential nanostructures from surface to bulk are also present [18]. Therefore, investigations on the surface properties and the effect on the electrochemical performance of these compounds should be extended to NMC and, quite probably, to other electrode materials. Overall, differentiated surfaces from the cores appear to be a significant characteristic of cathode materials that influences measured properties. Thus, inclusive design of cathode surfaces can provide new ideas for optimization of performance and cyclability of battery materials.