Elucidation of the Solid Electrolyte Interphase Formation Mechanism in Micro‐Mesoporous Hard‐Carbon Anodes

The microstructure of hard carbons can be designed to maximize their performance as anodes for sodium‐ion batteries. However, the nature of the electrolyte is also decisive in the capacity and long‐term stability. Here, hard carbons with a tailored bimodal pore network of internal micropores interconnected through mesopores are studied as sodium‐ion battery anodes. The evolution of their solid electrolyte interphase (SEI) is analyzed in three different electrolytes (NaPF6 in an ether‐based solvent, and NaPF6 or NaClO4 in a carbonate‐based system). Combining experiments with density functional theory calculations, it is proposed that formation of the SEI is mainly controlled by the decomposition of the salt anion. This process occurs through the intermediate functionalization of the carbon surface by the decomposed anion fragments. It is suggested that the innermost SEI sub‐layer governs the performance and long‐term stability of the anode. While the presence of a fluorine‐containing salt appears to have a determining role in the SEI stability, the electrochemical decomposition of carbonate‐based solvents is detrimental for the long‐term stability as the interfacial resistance increases. In contrast, the ether‐based system enables stable long‐term cycling as the interphase remains almost intact once the first fluorine‐rich SEI layer is formed.


Introduction
Lithium-ion batteries were commercialized in the early 1990s; [1] however, the cost of lithium has increased almost by a factor of five between 2010 and 2020, and it is expected to continue increasing with demand and further ore deployment. [2] It is therefore imperative to advance the development and understanding of alternative battery chemistries, both to offset the consumption of critical materials and the global reliance on only one type of technology. Sodium-ion batteries (NIBs) are a promising alternative for large-scale energy storage due to the natural abundance, widespread distribution, low cost, and excellent materials sustainability in comparison to lithium-reliant technologies. [3] Unlike lithium, sodium can only intercalate minimally into graphite. For NIBs, hard carbon (HC) is the preferred anode candidate, since its disordered and nongraphitizable turbostratic structure can The microstructure of hard carbons can be designed to maximize their performance as anodes for sodium-ion batteries. However, the nature of the electrolyte is also decisive in the capacity and long-term stability. Here, hard carbons with a tailored bimodal pore network of internal micropores interconnected through mesopores are studied as sodium-ion battery anodes. The evolution of their solid electrolyte interphase (SEI) is analyzed in three different electrolytes (NaPF 6 in an ether-based solvent, and NaPF 6 or NaClO 4

in a carbonate-based system). Combining experiments with density functional theory calculations, it is proposed that formation of the SEI is mainly controlled by the decomposition of the salt anion. This process occurs through the intermediate functionalization of the carbon surface by the decomposed anion fragments. It is suggested that the innermost SEI sub-layer governs the performance and long-term stability of the anode.
While the presence of a fluorine-containing salt appears to have a determining role in the SEI stability, the electrochemical decomposition of carbonate-based solvents is detrimental for the long-term stability as the interfacial resistance increases. In contrast, the ether-based system enables stable long-term cycling as the interphase remains almost intact once the first fluorine-rich SEI layer is formed. provide many sites for sodium storage. Sodium adsorption and storage occur at the randomly oriented and expanded graphene layers, edge and basal defects, and at the micropores within the graphitic fragments formed by the graphene stacks. [4] Great effort has been devoted to understand and attribute a voltage range to sodium storage within each of these microstructural features. Although still under some degree of debate, in general, the capacity of HC evolves through two distinct regimes: a high-voltage sloping region, attributed mostly to defects and intercalation between expanded layers, and a low-voltage plateau (<0.1 V vs Na + /Na) revealing sodium storage inside the native pores (micropores or nanopores) of the HC structure. Some of the processes observed within the sloping region have been regarded as irreversible or related to the formation of the solid electrolyte interphase (SEI) on the accessible surface area of the carbons during the first discharge cycle (sodium insertion). [5,6] Most synthesis strategies aim for carbons with a limited surface area but with an enhanced concentration of internal micropores, accessible solely to sodium ions, to boost the initial Coulombic efficiency and the reversible specific capacity (generally ≈300 mA h g −1 ). [7,8] Both the carbon precursor pretreatment and the consecutive annealing temperature play a crucial role in the microstructure. The pretreatment of biomass derivatives at ≈200 °C seems to enable larger micropores (and thus better interconnected) than those achieved through the direct thermal activation of the same feedstock. [9] Au et al. [10] reported a sharp decrease in gas accessible surface area with increasing annealing temperature (1000-1900 °C), and the formation of larger internal pores as a consequence of the growth of the lateral dimensions of the graphitic fragments surrounding the pores. Nevertheless, above a certain temperature (≈1500 °C), further graphitization becomes detrimental due to the reduction of the interlayer spacing and the consequent narrowing of the paths for sodium to access the pores.
In this context, maximizing the interconnectivity of closed pores for Na-ion storage by maintaining enlarged interlayer distances, number of defects, and accessible diffusion pathways enables precise tuning of the reversible capacity of HCs. In turn, this strategy produces NIBs with larger energy densities and extended service life. [11] However, tuning the texture of carbons does not suffice to maximize the performance of batteries, both in terms of the initial delivered capacity and long-term stability. Factors such as the nature of electrolyte determine the composition of the SEI, as this is formed upon the decomposition of the salt and solvent.
Excellent studies have been published regarding the role of fluorine-containing solvents or additives in the stability of the SEI for HCs [12,13] and TiO 2 anodes, [14] the stabilizing role of glymes for several types of electrodes, [15,16] the formation of ethylene dicarbonate, NaF, and Na 2 CO 3 as primary constituents of the SEI when carbonate-based electrolytes are employed, and the similarity of such mosaic-type SEI with that observed in lithium-ion batteries. [17] Nevertheless, fewer studies are found reporting the role of the salt and solvents on both the formation mechanism and properties of the SEI, [18][19][20][21] and there are even less examples where experiments and simulations are combined together to understand how exactly the SEI starts developing and what triggers stability or constant degradation.
Here, we report solvothermally prepared HCs with tailored bimodal porosities (micro-and mesopores) for Na-ion batteries.
Through the secondary mesoporous network derived from a soft template introduced during the hydrothermal treatment, we aim to enhance and widen the connectivity between the internal micropore network (Figure 1a). This strategy is expected to facilitate Na + diffusion over a larger length scale and to reach higher capacities than those observed for standard hard carbons annealed at equivalent temperatures. To understand and improve the performance of these materials, the formation of the SEI in three electrolyte systems (varying both salt and solvent) is studied and related to their electrochemical capacity and long-term Na-storage retention. The intrinsic properties of the salts and solvents (e.g., donor and acceptor numbers and redox potentials) are analyzed and the pathway of the SEI formation is proposed. Corroborated by computational studies, our findings suggest that SEI formation is mainly determined by decomposition of the salt anion and that it occurs through the intermediate functionalization of the carbon surface. The total amount of Na irreversibly consumed, and also the formation of a conductive and protective SEI layer, can be thus tailored through the choice of electrolyte mixture. In all the electrolyte systems the SEI layer presents a layered structure, consisting of an inorganic innermost layer covered by an organic-derived shell. We suggest that it is the inorganic internal SEI sub-layer that governs the performance and long-term stability of the anode. While the presence of a fluorine-containing salt appears to have a determining role in the SEI stability, the electrochemical decomposition of carbonate-based solvents can generate unwanted interfacial resistance that is eventually detrimental for the long-term stability. In contrast, in an ether-based solvent, the cyclability of the cell is enhanced and the interphase remains almost intact once the first fluorine-rich SEI layer is formed.

Soft Templated Hard Carbons: Templated Closed Porosity and Structural Evolution with Temperature
To obtain hard carbons with tailored porosities, phloroglucinol and glyoxal were dissolved in ethanol and solvothermally carbonized in the presence of a soft templating agent (Pluronic F127). Due to the equilibrium between carboxylic acid and geminal-diol groups, glyoxylic acid can both catalyze the polymerization and generate crosslinks between the phenol groups of phloroglucinol [22,23] (Figure 1a). In the presence of Pluronic F127 micelles (Figure 1a), the crosslinked resin develops around the templated domains (20-40 nm), and micro-mesoporous structures are expected once the sacrificial polymer is removed by high-temperature pyrolysis. [24,25] The resulting carbon (ST200) was therefore pyrolyzed at 1000, 1300, and 1500 °C, and labeled accordingly as ST1000, ST1300, and ST1500. The spherical particles, with diameters ranging from 2 to 4 µm ( Figure S1, Supporting Information), remained unaltered after carbonization for all temperatures.
Internal or "closed pores," inaccessible to gas molecules but accessible to Na + ions, are responsible for sodium storage in HCs at low potentials, [8,26] and their diameter can be derived from the region at 0.2 Å −1 in the total small angle neutron scattering (SANS) patterns ( Figure 1b) using the Guinier-Porod model [27,28] (Equation (S1) in the Supporting Information). Micropore diameters between 1.8 (±0.05) and 2.3 (±0.03) nm, Figure 1. a) Schematic of the bimodal porous structure of the soft templated carbons. b) Small angle neutron structure factor (Q) of the carbon precursor after solvothermal carbonization (STC) and after subsequent high-temperature pyrolysis at 1000, 1300, and 1500 °C (ST1000, ST1300, and ST1500, respectively). Inset shows the 002 and 100 peaks related to the local graphitic domains. c) Pore diameters derived from SAXS corresponding to the native HC micropores and templated mesopores; d) Relative change in the pore volume probed by N 2 gas adsorption and the intensity of the Guinier plateau in the small angle region weighted by pore volume; e) Basal domain size, stack thickness, and f) average spacing between the graphitic basal planes (002) determined from the width of the 002 and 100 peaks of the total SAXS/WAXS patterns. g) Representative SEM of ST particles after 1000 °C annealing. Scale bar is 5 µm; Additional SAXS/WAXS data and example model fits are shown in Figure SX  The mesopores, ascribed to the templating strategy, increase in diameter with increasing temperature, most significantly above 1300 °C (60%). The diameter of the micropores changes from 1.9 to 2.3 nm (40%), suggesting that micropore growth is more limited or that these combine to form significantly larger pores during the annealing process. By comparing the relative total pore volume derived from small angle X-ray scattering (SAXS; Figure 1d), a fivefold increment is observed between 1000 and 1500 °C. Notably, a substantial volume change within this temperature range is not observed for other HCs produced without a templating agent. [5] The micropore dimensions proved similar to the lateral size of the graphitic fragments ( Figure 1e) derived from wide angle X-ray scattering (WAXS; Figure S2, Supporting Information). In agreement with our previous work, [10] increasing the pyrolysis temperature within this range promotes the evolution of a greater number of slightly longer and thicker graphitic stacks ( Figure 1e) that form the walls of Na-storing closed micropores. Na-ion pore storage capacity might be increased with higher pyrolysis temperatures, provided that enlarged interlayer distances and pore access pathways are still available.
Although the evolution toward a more ordered structure is promoted with increasing temperature, between 1000 and 1300 °C there seems to be little change in the size and thickness of the graphitic fragments. Additionally, the interlayer distance ( Figure 1f,d 002 ) corresponding to the (002) diffraction of the graphitic layers at 1.8 Å −1 (inset Figure 1b) remains enlarged even at 1300 °C (3.78 Å) and only drops to 3.68 Å at 1500 °C. Such behavior can be attributed to both the presence of a templating agent and the pre-crosslinking of the glyoxalphloroglucinol resin; a "pillar effect" has been observed with HCs produced from crosslinked precursors, [9,29] preventing the shrinkage and stacking of the graphene-like layers, resulting in less dense graphitic domains with enlarged interlayer distances and larger pores once carbonized.
To further investigate porosity, gas-accessible pores (equal to or larger than 0.4 nm [30] ) were analyzed through N 2 adsorptiondesorption isotherms ( Figure S3, Supporting Information). As expected, the specific surface area (mostly constituted by mesopores, see Figure S3b in the Supporting Information) displayed a decreasing trend with the annealing temperature (181, 66, and 21 m 2 g −1 for ST100, ST1300, and ST1500, respectively). This trend opposes that observed by small-angle scattering data, suggesting that the majority of pores are closed off during pyrolysis. [8] It is worth noting that despite their considerable size (25-40 nm), the mesopores observed through neutron scattering seem to also be mostly closed, as those that are observed through N 2 -adsorption present a pore size distribution centered around 10-15 nm ( Figure S3a, Supporting Information) Nevertheless, these surface area values remain higher than those observed for nontemplated HCs annealed at equivalent temperatures, which become virtually nonporous above 1300 °C. [10,31,32] Such behavior indicates that the synthesis methodology is indeed preventing the compaction of the structure with increasing temperature.
The development of structural ordering with temperature was corroborated by transmission electron microscopy (TEM; Figure 1h-j). Interestingly, the presence of the templated micropores is not observable through TEM, as it would be expected for templated pores aligned one to another and having identical shapes. [33] Some disordered domains and short graphene fringes, typical of hard carbons, are observed at 1000 °C ( Figure 1h). Above 1300 °C, more graphitic stacks with random orientation start to be observed ( Figure S4, Supporting Information). At 1500 °C (Figure 1j), the graphitic fringes become overall more evident and aligned, suggesting the development of a higher number of more ordered and larger graphitic domains. These results are consistent with the increasing in-plane ordering implied by the Raman I D /I G ratio increment (Figure 1f), [34] indicating the reduction of defects (see Figure S5, Supporting Information, for details) and the removal of polar surface groups with temperature ( Figure S6 in the Supporting Information).

Electrochemical Performance in Na-Ion Half-Cells
The impact of the structural characteristics of the soft templated carbons on their electrochemical performance was evaluated in half-cells using sodium metal as the counter electrode and 1 m NaPF 6 in ethylene carbonate and dimethyl carbonate (EC:DMC, 1:1 v/v) as the electrolyte, between 0.005 and 3 V. The galvanostatic discharge-charge profiles at 30 mA g −1 (Figure 2a) show a common behavior: discharge curves (sodium insertion) display a sloping region above 0.1 V, predominantly associated with adsorption at edge or basal defects or insertion into enlarged interlayer spaces, and a flatter plateau region below 0.1 V, arising from Na storage within internal or "closed" pores ( Figure 2d-f). [5,10,[35][36][37] Overall, ST1500 presents the highest initial Coulombic efficiency (65%, Figure 2h) and first and second cycle capacities (523 and 354 mAh g −1 , respectively), predominantly plateau-derived, consistent with having the highest internal pore volume and the lowest surface area and number of defects. In contrast, ST1000 shows a large and mostly sloped capacity in the first cycle, which decreases substantially in the second cycle. This large and irreversible consumption of Na is consistent with the high surface area of ST1000 that is accessible for the decomposition of the electrolyte and subsequent formation of the SEI; the same phenomenon is observed, to a lesser extent, in both ST1300 and ST1500, as they have a lower presence of open porosity.
A high-voltage plateau-like region appears at around 0.4-0.6 V, even for the material annealed at 1500 °C ( Figure 2a) owing to the decomposition of ester-based solvents into double alkyl carbonates. [38] This irreversible process is also observed in the first cycle of the cyclic voltammograms (CVs, Figure S7, Supporting Information). During the first discharge step, Na is consumed irreversibly, as manifested by the broad peaks appearing at 0.3-0.8 V. In the subsequent cycles, the reduction and oxidation currents, observed closer to the lowest voltage limit and attributed to the reversible insertion and extraction of Na ions, become larger and steady only for ST1500. For ST1000 and ST1300, the Na trapping appears lessened, yet the storage of Na remains inefficient.
The specific capacity of the ST1500 carbon at the highest current density drops to a lesser extent than for the rest of carbons (Figure 2i, 57 mAh g −1 at 600 mA g −1 or 2 C). At high current density, most HCs display a near-zero capacity due to the excessive polarization of the electrodes and the restricted Na + diffusion occurring at high charging rates. [39,40] The superior capacity retention at high current densities could be related to the connectivity between the meso-and micropores, with the larger templated pores facilitating fast diffusion of Na + into the smaller pores native to the disordered HC structure.

Solid Electrolyte Interphase (SEI) Composition and Long-Term Performance
With the aim of further stabilizing the SEI through electrolyte engineering, and thus increasing the reversible capacity of the best-performing material (ST1500), two additional electrolyte systems were screened. NaPF 6 was replaced by NaClO 4 for one of the systems, keeping the same solvent (EC:DMC), and for the second, EC:DMC was replaced with bis(2-methoxyethyl) ether (diglyme). The main differences between the salts are the ion mobility (NaClO 4 > NaPF 6 ), the dissociation constant (NaClO 4 < NaPF 6 ), and the donor number (DN, DN PF6-= 2.50) < DN ClO4-= 8.4), all of which determine the ionic conductivity (in EC:DMC, σ NaClO4 = 5 S cm −1 σ NaPF6 = 6.5 S cm −1 ). [41] A higher anion DN favors ion pair interactions, which leads to lower ionic conductivities and inhibits solvation.
With regard to the choice of solvents, diglyme is an etherbased solvent (COC) while in the EC:DMC mixture, both molecules are carbonates with ester bonds (OC(O)O), where EC is cyclic and DMC is linear. According to the electron acceptor-donor power of a solvent, a higher acceptor number (AN) implies better charge-accepting ability, while a higher DN reflects better solvation of the cation. Diglyme has a relatively low AN (≈9.9) compared to the carbonates (≈18), [42] yet its higher DN (19.2 [43] vs 16.4 or 17.2 for EC and DMC, respectively) has been associated with a compact and protective solvation shell that impedes contact ion pair interactions. Glymes have thus been identified as solvents that promote a stable SEI, both in positive and negative electrodes, and that are able to stabilize high specific surface area carbons. Nevertheless, the fundamental mechanism and cause of this improvement remains unclear and experimental data have not been fully contrasted with theoretical simulations.
When cycled at 30 mA g −1 (Figure 3, and fully summarized in Table S2, Supporting Information), the NaClO 4 /EC:DMC system presents the highest first cycle capacity but the worst long-term cyclability. The first cycle capacity of the NaPF 6 salt is higher in EC:DMC compared to diglyme, although changing to diglyme improves the initial Coulombic efficiency and gives excellent capacity retention (≈70%) after 200 cycles (Figure 3a), suggesting the formation of a more stable SEI layer. [16] This lower initial capacity in diglyme compared with EC:DMC may arise from a certain degree of solvent co-intercalation within the expanded graphitic layers, limiting the overall amount of Na that the HC can accommodate. Co-intercalation of Na ions with different solvents has been widely observed in graphite, [44][45][46][47] but might be expected to a lesser extent in hard carbons because the highly crosslinked graphitic domains cannot accommodate the strain induced by the larger volume expansion. It is worth mentioning that the system employing NaClO 4 in diglyme was tested but resulted in repeated cell failure. Such behavior can be attributed to the inability of diglyme (very low AN) to solvate the high DN ClO 4 − anion, and thus, being unable to prevent its decomposition upon polarization. [48] Impedance spectra were recorded before and after the first charge/discharge cycle. From the Nyquist plots in Figure 3e,f, a higher overall resistance for both the NaPF 6 and NaClO 4 / EC:DMC systems (higher x-intercept of the semicircle) is seen. In contrast, for the diglyme-based electrolyte (Figure 3d), a flattened semi-circle (low variation of impedance with frequency) is observed, that becomes even flatter upon three galvanostatic discharge/charge cycles ( Figure S8 in the Supporting Information). This limited evolution of impedance in the NaPF 6 /diglyme , and c) ST1000. Schematic of the closed porosity evolution with temperature: d) ST1500, e) ST1300, and f) ST1000. g) First cycle total, slope, and plateau capacities, and initial Coulombic efficiency (ICE) for ST1000, ST1300, and ST1500; h) Rate performance of the discharge capacities at 30, 60, 150, 300, and 600 mA g −1 for ST1500 (orange), ST1300 (black), and ST1000 (light blue); the dotted lines delimit each of these current densities.
system corroborates the formation of a stable and conductive SEI during the first charge/discharge cycle, which remains mostly unaffected for three consecutive cycles. In contrast, the impedance of the NaClO 4 /EC:DMC system increases upon cycling, suggesting that the resistivity of the evolving SEI increases as cycling progresses. The initial evolution of the impedance could be extrapolated to understand the capacity decline observed over 200 cycles: only a low 100 mAh g −1 (44%) capacity is retained after 200 cycles at 30 mA g −1 , probably attributed to the increasingly impeded charge transfer at this constantly evolving interphase (Figure 3c). The differences at the SEI level were substantiated by analyzing the composition across the electrode/electrolyte interface with X-ray photoelectron spectroscopy (XPS; see full details of XPS characterization in the Supporting Information). The cycled electrodes were progressively Ar + etched and a compositional depth profile was obtained by determining the percentage change of the sp 2 C (285.1 eV), Na 2 CO 3 (290.2 eV), and CO (289.2 eV) bond environments within the high-resolution C 1s core-level spectrum, and the NaF (684 eV) and NaCl (200.5 eV) bonds in the F 1s and Cl 2p spectra, respectively (Figure 3g-j). The sp 2 carbon environment of the C 1s band is predominantly attributed to the pristine hard carbon which is buried underneath the SEI after cycling. With etching time, the percentage of the sp 2 contribution increases for both electrodes cycled in NaPF 6 , while it remains close to zero for the NaClO 4 system, . The contribution (%) of the different environments was extracted from the high-resolution C 1s core-level spectrum, to enable the comparison of the g) sp 2 C, h) Na 2 CO 3 , and i) CO bonds, while the j) NaF and NaCl contributions were extracted from the F 1s and Cl 2p high-resolution spectra. Extracted from the XPS analysis, a scheme of the layered nature of each of the formed SEIs when the electrodes are discharged in the different electrolyte systems: k) pristine carbon electrode before discharge, l) in 1 m NaPF 6 /diglyme, m) in 1 m NaPF 6 /EC:DMC, and n) in 1 m NaClO 4 /EC:DMC.
indicating that the surface layer in the in NaPF 6 -based electrolytes is thinner than that in the NaClO4 system.
In the case of the Na 2 CO 3 environment, its presence can be attributed to the oxidative decomposition of the electrolyte and the CO to the formation of polyesters derived from the solvent. [16,49] For the diglyme system, both bands remain extremely weak for all etching times, indicating a thin and homogenous SEI layer with a very low amount of organic electrolyte decomposition products. In the case of the systems cycled in EC:DMC, for the one incorporating NaPF 6 , the Na 2 CO 3 percentage decreases sharply after only 30 s, suggesting that Na 2 CO 3 is predominantly distributed on the outermost layer of the SEI. In the case of the CO band, the percentage increase with etching time suggests that polyester species are found mostly at the inner layer.
For the system incorporating NaClO 4 , the resulting Na 2 CO 3 layer appears to be thick enough not to change in composition with etching time and could be related to oxygen loss intrinsic to NaClO 4 decomposition. Although our findings through XPS etching are in line with the electrochemical performance, it is worth mentioning that through this technique, there is a large likelihood of introducing oxygen-containing functionalities during the etching step. Due to this limitation, further analysis is currently under development through 23 Na NMR and timeof-flight secondary ion mass spectrometry (ToF-SIMS) for our follow-up work.

Density functional theory (DFT) calculations indicate that neither the adsorption of ClO 4
− (E ads = 1.43 eV) nor PF 6 − (E ads = 1.68 eV) is energetically favored on the basal plane of the carbon, but adsorption of anion fragments is possible (see Table S4 in the Supporting Information for details of adsorption energies of all the possible fragments). Simulations of the anion fragmentation show that ClO 4 − dissociates through step-wise oxygen loss (Figure 4a) while in PF 6 − , fluorine fragments are successively detached (Figure 4b). In agreement with previous work, [50] as ClO 4 − fragments, oxygen groups can be easily incorporated into the carbon basal plane (Figure 4a-ii and -iv); for the NaClO 4 system, the steady amount of Na 2 CO 3 and CO groups across the SEI depth determined by XPS is consistent with this assertion. Furthermore, the greater thickness of the ClO 4 − -derived SEI could be attributed to the large oxygen loss that this anion undergoes as it decomposes. These oxygen moieties have been shown to subsequently promote energetically favorable Na adsorption [51] and the higher initial capacity observed in the NaClO 4 /EC:DMC system compared with NaPF 6 could therefore potentially arise from these oxygen functionalities that are introduced through anion decomposition.
The Cl and F fragments (Figure 4c,d) are both attracted to the carbon basal surface (E ads,F = −3.54 eV, E ads,Cl = −2.51 eV); in the case of F, the fragment seemingly binds to the carbon. When the carbon is fluorinated, it is observed that the FC bond puckers the basal plane (Figure 4d-i). The favorable CF bond formation is key to the passivating effect that fluorinated additives and salts have on carbonaceous anodes. For the Cl containing system, the ClC distance remains large (2.40 Å vs 1.52 Å for FC), in agreement with previous studies. [52] The adsorption of Na in the vicinity of adatoms is favorable (as compared to the pristine carbon surface), indicating that the presence of Cl, F, or oxygen close or bound to the carbon surface could be beneficial for the initial sodiation and storage, but could also lead to the irreversible entrapment of Na ions and overall capacity loss. Upon Na adsorption, the chlorine desorbs (CCl distance evolves from 2.40 to 3.24 Å) from the carbon surface, leading to the formation of a NaCl dimer; similar dimer formation is observed for NaF (Figure 4c-ii,d-ii), although showing a lower tendency to interact with the carbon surface than F (E ads,F = −3.54 eV vs E ads,NaF = −2.08 eV). The presence of NaF and NaCl (for the system using the NaClO 4 salt) bond environments in the XPS analysis (Figure 3g-j) can therefore be unequivocally attributed to the decomposition of the salt in each system. − decomposition products on carbon, b) PF 6 − decomposition products on carbon, c) Cl adsorption and Na-Cl dimer formation, and d) F adsorption on carbon surface highlighting the puckering of the carbon lattice, and subsequent Na-F dimer formation. Brown spheres are carbon, green chlorine, yellow sodium, and gray fluorine.
While the adsorption of F on the carbon surface is more favorable than Cl, the dissociation barrier for ClO 4 − is lower than that for PF 6 − (0.88 eV vs 1.23 eV, respectively). These results imply that with faster ClO 4 − breakdown and subsequent reaction with Na ions, greater irreversible consumption of Na occurs but not localized at the basal plane, consistent with the lower tendency of Cl to remain adsorbed on the carbon (E ads,Cl = −2.51 eV vs E ads,F = −3.54 eV,). Consequently, the carbon surface remains active for continuous decomposition of ClO 4 − anions, resulting in a less stable and noncompact SEI layer; thus, the greater capacity fade observed for EC:DMC/ NaClO 4 during cycling may arise from the continuous breakdown of ClO 4 − as compared to PF 6 − and the growing resistivity arising from the thick SEI, as demonstrated in the EIS before and after cycling (Figure 3d-f). Conversely, slower consumption of Na ions arising from reaction with PF 6 − decomposition fragments but occurring directly at the carbon surface, results in the formation of a thinner, more stable and protective SEI layer and therefore slower capacity loss over subsequent cycles.
Experimentally, these observations are confirmed by the evolution of the NaF band in the XPS which implies that NaF appears to be mainly present in the inner layer of the SEI, as suggested by the increase in percentage with etching time for the systems cycled with NaPF 6 . [19] For the NaClO 4 system instead, the NaCl band increases to a lesser extent than the analogous NaF for the NaPF 6 systems, suggesting a more even distribution within the SEI. The presence of NaF has been attributed to a more conductive and protective SEI, and consequently improved cycling stability. [53,54] Conversely, the appearance of NaCl has been linked to a noncompact, insulative, and unstable SEI layer that will degrade over time and expose fresh portions of electrode to the electrolyte, promoting further decomposition compounds on the newly developed interfaces. [20,48,55,56] The results from XPS are therefore consistent with the significant deterioration in capacity observed in the NaClO 4 system. The lower energy barrier to ClO 4 − breakdown also leads to oxidative decomposition at the highly reductive surface of the metallic sodium in half-cell configuration. XPS analysis of the sodium-metal counter electrode confirmed the presence of a large amount of NaCl deposits ( Figure S14, Supporting Information). Since the reversibility of metallic sodium is largely dependent on the composition of the SEI, [53,54,57] the premature failure of this cell may also be ascribed to the rapid and continuous decomposition of the electrolyte at the sodium surface. Qualitative aging experiments ( Figure S15, Supporting Information) showed that, even without any external voltage, sodium promotes the decomposition of the EC:DMC/NaClO 4 electrolyte. Indeed, when our carbon anode is assembled against a Na 3 V 2 (PO 4 ) 2 F 3 cathode (NVPF) rather than a sodium metal electrode, the capacity after 200 cycles fades to a minor extent than in the half-cell configuration ( Figure S16, Supporting Information).

Linking the SEI to the Intrinsic Properties of the Salt/Solvent System
To rationalize the SEI composition for the different solvent/salt combinations at the atomic scale, ab initio molecular dynamics (AIMD) were performed to disseminate the effect of the electrolyte composition on the sodium diffusion and mobility, separating these effects from those arising from the SEI formation and the anode material itself. In diglyme, sodium can coordinate to one, two, or three oxygens per diglyme molecule with an average NaO coordination number of 5.36 at 300 K (Figures S17 and S18 in the Supporting Information). This leads to Na being coordinated with two to four diglyme molecules, and the typical solvation shell consisting of two diglyme molecules, in good agreement with previous results. [58] Due to the relatively low polarity and high flexibility of diglyme, [59,60] only one compact coordination shell is observed (derived from AIMD pair distribution functions, see Figures S17-S19 in the Supporting Information). The PF 6 − anion remains intact during the whole simulation, only interacting with the terminal aliphatic hydrogens of diglyme. Well in agreement with previous observations, [55] this lack of interaction beyond the solvation shell for Na + confirms that diglyme can effectively keep the anion away from the cation, remaining 10.28 ± 1 Å apart on average (a plot showing the cation-anion distance is provided in Figure S20 in the Supporting Information). For the EC:DMCbased electrolytes, the cation-anion distance is longer than the corresponding distance in the diglyme electrolyte (11.92 ± 0.7 Å for Na-PF 6 , and 12.21 ± 0.8 Å for NaClO 4 ). For the EC:DMC systems, both carbonyl and ester oxygens enable a strong interaction with Na + . Additionally, due to the high polarity and relatively high rigidity of the solvent molecules, two bulky and noncompact solvation shells were identified for both NaPF 6 and NaClO 4 salts (see Tables S3 and S4 in the Supporting Information and complementary explanation). The average coordination numbers for the first NaO coordination shell at 300 K are 5.12 and 5.15 for NaPF 6 and NaClO 4 , respectively, while in the second coordination shell, the average coordination increases to 12 for both salts.
The differences in solvation for the anions (PF 6 − and ClO 4 − ) and Na + in the two solvent systems are responsible of the differences in ionic transport, diffusion coefficient (D Na and D Anion ), and sodium transport number (t) derived from our simulations ( Table 1).

NaPF 6 versus NaClO 4 in EC:DMC
The calculated D Na is similar for both the NaPF 6 and NaClO 4 salts in EC:DMC (Table 1). Thus, the diffusion of Na + through Table 1. Calculated diffusion coefficients (D) and transport number (t) for the different solvent electrolyte systems from AIMD simulations at 300 K. Previous NMR studies of 23 Na diffusion in diglyme with NaPF 6 have shown that the self-diffusion coefficient has an order of 10 −6 cm 2 s −1 , [61] and is similar to that of LiPF 6 in the same solvent. The electrolyte bulk systems were modeled in simulation cells containing 36 EC:DMC, and 36 diglyme to correspond to the experimental density and to allow for simulations of the isolated salt. the electrolyte bulk appears to have limited direct impact on the electrochemical performance observed for these systems, indicating that the difference in both the short-and long-term capacity observed between the NaPF 6 and NaClO 4 cells arises due to processes occurring at the electrode surfaces. The transport number (t), as calculated according to Equation (2) in the computational details and discussion in the Supporting Information, is a measurement of the mobility of the cation relative to the counterion, [18,61] and indicates that the movement of Na + and ClO 4 − ions are better correlated than Na + with PF 6 − ions in EC:DMC. ClO 4 − has a high donor number compared to PF 6 − (8.44 vs 2.5, respectively) closer to that of EC:DMC (16)(17). ClO 4 − can thus more effectively compete with the EC:DMC molecules to solvate Na + resulting in minimal ion separation. These results suggest that as Na + approaches the carbon electrode, far more ClO 4 − than PF 6 − anions are brought within close proximity of both the Na + and the carbon surface, promoting greater decomposition of the anion, and corroborating the experimental results above.

NaPF 6 in EC:DMC versus NaPF 6 in Diglyme
Swapping EC:DMC for diglyme while keeping the NaPF 6 salt showed an increase in the D Na (Table 1) indicating that Na diffusion in the diglyme system is more rapid. Interestingly, the diffusion of PF 6 − is also notably higher than for the other systems, [61] but the low value of (t) implies the lowest cation-anion mobility correlation in diglyme, suggesting that even if both PF 6 − and Na + are more mobile in diglyme, Na + is more protected and separated from the PF 6 − anion. Therefore, the degree of interaction between PF 6 − anions and the HC surface is even lower in diglyme compared with EC:DMC, resulting in far less decomposition and thus much less consumption of Na within the thinner SEI layer. Furthermore, PF 6 − has a higher reduction potential than diglyme [21,62] which leads to preferential decomposition of the anion before the solvent, to stabilize the surface to further decomposition. Meanwhile, comparing the reduction potentials of diglyme with EC:DMC, EC:DMC is much more easily reduced and will result in a subsequent greater organic contribution to the SEI layer, [63] as has been demonstrated by XPS.
In summary, the first cycle and long-term electrochemical behaviors of the different electrolyte systems are dependent on the interactions of the Na ions with the solvent molecules and anions (Figure 5), and the subsequent decomposition behavior at the electrode surface.
We suggest that NaClO 4 in EC:DMC ( Figure 5c) achieves a higher first cycle capacity than NaPF 6 because the incorporation of ClO 4 − decomposition products (oxygen, red spheres in Figure 5c) into the carbon basal plane can provide additional Na storage capability. However, the capacity fades quickly since unstable adsorption of other ClO 4 − fragments leaves the carbon surface available for constant decomposition and subsequent Na consumption. The SEI layer is evenly composed of NaCl and organics such as Na 2 CO 3 and CO species, and is thicker than the SEI produced in the NaPF 6 systems. NaCl leads to a nonconductive SEI, and it is this increased resistance, as well Figure 5. Schematic representation of the SEI formation in: a) 1 m NaPF 6 /diglyme, where a thin and homogenous layer of NaF formed from the decomposition of NaPF 6 serves as a Na-conductive and protective layer for the further stable cycling of the HC electrode; b) 1 m NaPF 6 /EC:DMC, where a NaF protective layer forms, however, due to worse solvation of Na + , organic compounds derived from the electrochemical decomposition of the solvent increase the interfacial resistance and in turn lower the long-term stability of the anode; c) 1 m NaClO 4 /EC:DMC, where the decomposition of ClO 4 − enables oxygen functionalization on the carbon that can irreversibly trap Na + , and formation of a nonprotective NaCl layer that contributes to the premature failure of the cell. as the continuous Na ion consumption, that results in deteriorating performance in subsequent cycles. In the NaPF 6 systems (Figure 5a,b), strong initial adsorption of F fragments results in subsequent stable adsorption of Na ions, leading to the controlled formation of a compact protective SEI. As NaF is more Na-conductive than NaCl, and deposits stably onto the carbon surface, subsequent decomposition of the anion is mostly suppressed, resulting in greater long-term cycling stability. However, changing the solvent has a great effect on both initial and long-term capacity. With diglyme (Figure 5a), the more compact solvation shell could allow a certain degree of co-intercalation, thus limiting the overall initial capacity. The significantly superior capacity retention arises from the very limited degree of solvent degradation reducing the total amount of Na irreversibly consumed, and also the formation of a less resistive SEI layer mostly formed of NaF. We suggest that it is the structure of this largely inorganic internal SEI layer that promotes excellent long-term stability in NaPF 6 /diglyme. As observed by XPS (Figure 3g-j), the structure of the SEI layer formed in EC:DMC (Figure 5b) consists of an inner NaF-rich layer too; however, in this case, the appearance of a largely organic outermost layer composed of the electrochemical reduction products of the solvent is evident and contributes to increased resistivity. This layer might arise as a consequence of the larger de-solvation energy barrier that Na + needs to overcome in the carbonate-based solvent [64][65][66] and the greater possibility of PF 6 − to penetrate into the carbonate solvation shell in comparison to the Na-diglyme system.

Conclusion
In this study, we synthesized a range of hard carbons with bimodal native and templated porosity for use as NIB electrodes. The material with the greatest closed porosity exhibited the highest capacity across a range of current densities, due to the optimized structure of enlarged interlayer distances and mesopores preserving connectivity between the native micropores. The ST1500 material was further investigated with alternative electrolyte systems and the choice solvent and salt were shown to have a significant effect on long-term cycling performance. Through a combination of XPS and theoretical modeling, we have demonstrated that the presence of F-containing species in the electrolyte is important for the formation of a compact, ion-conducting protective layer, while the solvent choice can further improve or impair electrochemical performance, depending on the interaction with both cations and anions in solution. These findings highlight the importance not only of considering active material suitability but also selecting appropriate electrolyte systems in order to obtain optimized performance for future sodium-ion batteries.

Experimental Section
Materials Synthesis: All chemicals were purchased from Sigma-Aldrich and used as received if not specified otherwise.
Soft Templated Carbons: For the synthesis of the soft templated carbons, a precursor solution was carbonized under autogenic pressure at 200 °C. In a typical experiment, 0.82 g phloroglucinol (1,3,5-benzentriol, C 6 H 6 O 3 ) and 0.61 g of glyoxylic acid monohydrate were mixed with 1.61 g Pluronic F127 triblock copolymer and dissolved in 40 mL of absolute ethanol. [22] The solution was transferred into a 100 mL Teflon autoclave liner, sealed (Parr Instruments stainless steel autoclave), and placed in an oven at 80 °C for 4 h to trigger the phloroglucinol/glyoxylic acid crosslinking reaction. Subsequently, the mixture was solvothermally carbonized at 200 °C for 12 h. After cooling, the suspension was filtered, and the solid fraction was dried under vacuum at 40 °C for 12 h. The dry material was then annealed up to 1000, 1300, and 1500 °C at a rate of 5 °C min −1 for 2 h.
Cathode for Full Cell Testing: The NVPF@C cathode material was synthesized following a hydrothermal method. [67] A typical synthesis is as follows: 0.98 g V 2 O 5 and 0.337 g oxalic acid were added into 15 mL deionized water under continuous magnetic stirring at 70 °C for 20 min. Subsequently NH 4 H 2 PO 4 (0.252 g), NaF (0.130 g) and 0.03 g of carbon black (Super P conductive, 99+%, Alfa Aesar) were added into the blue solution under stirring. The obtained mixture was transferred to 50 mL Teflon liners, which were then placed inside stainless steel autoclaves and kept at 180 °C for 12 h. After this time, the obtained gel was freezedried and subsequently annealed under argon at 475 °C for 9 h using a heating rate of 5 °C min −1 . For sake of comparison, an NVPF sample without Super P carbon was also prepared using the same methodology and annealing procedure.
Materials Characterization: Morphological investigations were carried out by using an FEI Inspect F SEM operated at 10 kV. TEM was carried out using a JEOL 2100Plus TEM at 200 kV operating voltage. N 2 and CO 2 adsorption isotherms were measured by using a Micromeritics ASAP 2020 instrument. Prior to the analysis, the materials were degassed overnight under vacuum at 200 °C. The Brunauer-Emmett-Teller surface area (S BET ) was calculated in the relative pressure range of 0.05-0.3 for N 2 adsorption and the micropore surface area (S CO2 ) was determined using the Dubinin-Radushkevich (DR) method. The pore size distribution was assessed by DFT for slit pores. Raman spectra were recorded in a Raman-Senterra II spectrometer equipped with a 100X objective and a 532 nm laser as excitation source.
SAXS/WAXS was measured on a Xenocs nano-InXider with a Cu K α source and a two-detector setup for combined SAXS/WAXS measurements. Powder samples were packed in 2 mm Kapton capillaries and measured for 10 min. The empty capillary was measured separately and used for background subtraction. Neutron total scattering was carried out at the NIMROD [68] instrument at ISIS (Harwell, UK). Each sample was measured for 2 h and data corrections and reduction were performed using Gudrun. [69] Electrochemical Characterization: Electrodes were made using an 85:5:10 weight ratio of active material, carbon black (Super P conductive, 99+%, Alfa Aesar), and carboxymethyl cellulose binder (CMC) (M w ∼ 250 000, Sigma Aldrich), respectively. A 250 µm layer of the slurry was then coated onto a carbon-coated aluminium foil (18 µm conductive coating, MTI corporation) through blade casting (doctor blade coater). Once dried at room temperature for 24 h, 13 mm disks were cut and dried at 100 °C for 2 h under vacuum. The electrolyte preparation and coin cell assembly were conducted in an argon-filled glovebox (MBraun GmbH Labstar glovebox workstation) with H 2 O < 0.5 ppm and O 2 < 0.5 ppm levels. CR2032 coin cells were assembled using sodium metal counter electrode disks (less than 0.5 mm thick, rolled, and cut from a sodium ingot, 99.8% metals basis, Alfa Aesar), the coated hard carbon as working electrodes, and a glass fiber separator (Whatman GF/B glass microfiber, 18 mm diameter) saturated with electrolyte (150 µL). For the evaluation of full cells, the NVPF@C material was coated and cut in the same manner than the hard carbons, tested first against sodium metal and subsequently assembled with our hard carbon materials as anodes. All electrolytes were prepared with a 1 m salt concentration. Either sodium hexafluorophosphate (NaPF 6 Alfa Aesar, 99+%) or sodium perchlorate (NaClO 4 , 99.9+%, Sigma Aldrich) were dissolved in the predried solvents (80 °C overnight and then placed in a bottle containing 4 Å porous molecular sieve). For the 1:1 volume mixture of ethylene carbonate (EC, anhydrous 99%, Sigma Aldrich) and dimethyl carbonate (DMC, 99.9+%, Sigma Aldrich), EC was first heated up to 45 °C inside the glovebox antechamber and then stirred with an equal volume of DMC until fully homogenous. When diethylene glycol dimethyl ether was used (diglyme, anhydrous, 99.5%, Sigma Aldrich), the salt was directly added and stirred until complete dissolution. Galvanostatic cycling tests were performed on a Basytec system in the potential range of 0.001-2.5 V versus Na+/Na and using sodiation/desodiation current rates between 30 and 600 mA g −1 , corresponding to cycle rates between 0.1 C and 2 C based on a theoretical capacity of 300 mAh g −1 . CV and electrochemical impedance spectroscopy (EIS) were performed on a Biologic VSP-300 potentiostat. CVs were recorded at a scan rate of 0.1 mV s −1 for five cycles and EIS data were collected between 200 kHz and 10 mHz with a signal amplitude of 10 mV.
Computational Details: The DFT and ab initio molecular dynamics (AIMD) simulations were conducted using the CP2K code. [70][71][72][73] The sodium adsorption on chlorinated and fluorinated carbon surfaces were modeled using a three layer graphite model (216 atoms) with a 20 Å vacuum gap. The electrolyte bulk systems were modeled in simulation cells containing 36 EC and 36 DMC molecules for EC:DMC, and 36 diglyme molecules to correspond to the experimental density. All calculations were performed at the Γ-point with a plane wave cut-off of 650 Ry, with the DZVP-SR-MOLOPT basis set, PBE functional [74,75] and DFT-D3 method with Becke-Johnson damping. [51,[76][77][78][79][80][81][82] All parameters were converged to 0.01 meV with respect to energy and 0.001 eV Å −1 with respect to forces. The AIMD simulations were conducted using an NVT ensemble at 300 K with a time step of 0.5 fs and a Nosé-Hoover thermostat. The electrolyte models were equilibrated for 10 ps, whereafter a production run of 20 ps was conducted. All analysis of the electrolyte systems were conducted using the Travis code. [83][84][85] The ionic transport properties of Na + in the different solvent salt combinations were explored through mean square displacement calculations from the molecular AIMD trajectories. The ion diffusion coefficients could be calculated from mean square displacement (MSD), which was given by [86,87] MSD 6 Na D t B = + where t is the time, D i is the self-diffusion coefficient for Na + , and B is a thermal factor associated with atomic vibrations. From the diffusion coefficients for both the anions and the sodium ion in the different solvent systems, the sodium transport number (t) could be simulated [18,61] Na Na anion t, which is a measurement of the relative mobility of the cation to its anion counterion, could be obtained from experimental NMR diffusometry measurements which could probe the individual solvent, sodium, and anion motion. Further information on the calculation of diffusion coefficients, and transport numbers are included in the Supporting Information.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.