Boosting the power performance of multilayer graphene as lithium-ion battery anode via unconventional doping with in-situ formed Fe nanoparticles

Graphene is extensively investigated and promoted as a viable replacement for graphite, the state-of-the-art material for lithium-ion battery (LIB) anodes, although no clear evidence is available about improvements in terms of cycling stability, delithiation voltage and volumetric capacity. Here we report the microwave-assisted synthesis of a novel graphene-based material in ionic liquid (i.e., carved multilayer graphene with nested Fe3O4 nanoparticles), together with its extensive characterization via several physical and chemical techniques. When such a composite material is used as LIB anode, the carved paths traced by the Fe3O4 nanoparticles, and the unconverted metallic iron formed in-situ upon the 1st lithiation, result in enhanced rate capability and, especially at high specific currents (i.e., 5 A g−1), remarkable cycling stability (99% of specific capacity retention after 180 cycles), low average delithiation voltage (0.244 V) and a substantially increased volumetric capacity with respect to commercial graphite (58.8 Ah L−1 vs. 9.6 Ah L−1).

Scientific RepoRts | 6:23585 | DOI: 10.1038/srep23585 (e.g., Fe 3 O 4 ) reactions with lithium, resulting on higher gravimetric capacities than those of bare graphene and graphite 4 . Indeed, electroactive particles and graphene might enhance each other's performance, with the former helping to reduce the graphene restacking and the latter providing enhanced conductivity and buffering of eventual volume changes. However, as recently pointed out by Obrovac and Chevrier 15 , investigation about the density and the lithiation/delithiation voltages of such materials are essential parameters which should be provided in order to ensure the progress in the LIB field.
In this respect, we aimed to develop a composite formed by multilayer graphene and Fe 3 O 4 , because of the environmentally-friendliness of the latter compound and its expected ability to increase the rate capability by conversion to metallic iron (i.e., good electronic conductor) upon lithiation 17 . Making use of the ionic liquid microwave-assisted exfoliation method previously developed for graphene production 18 , a novel graphene-based composite 4 has been synthesized. Noticeably, the composite revealed a carved structure with nested Fe 3 O 4 nanoparticles, which has been observed and characterized for the first time. A multitude of RGO/Fe 3 O 4 compounds have been reported in the past years 19,20 , all with considerable amounts of iron oxide (usually between 30 and 70 wt.%) aiming to exploit the conversion reaction to increase the specific gravimetric capacity of the anode (causing low energy efficiency and relatively high lithiation/delithiation voltages). Differently, this novel composite, obtained by simultaneous synthesis of multilayer graphene and Fe 3 O 4 nanoparticles, only contains a minimal amount of iron oxide (i.e., <3 wt. %). However, it possesses a peculiar structure enabling superior power performance. Indeed, when the delithiation voltage is limited to 1 V vs. Li/Li + , the carved paths traced by the Fe 3 O 4 nanoparticles and the unconverted metallic iron formed in-situ upon the 1 st lithiation, incredibly enhance the anode performance at high currents without affecting the energy efficiency. According to our knowledge, such approach is reported here for the first time.

Results
Synthesis summary. The overall procedure for the synthesis of the carved multilayer graphene with nested Fe 3 O 4 nanoparticles (hereinafter called MUG-Fe 3 O 4 ) is illustrated in Fig. 1a. After 35 hours of sonication, the 1% w/w dispersion of expanded graphite (EG) in 1-ethyl-3-methylimidazolium acetate (EMIMAc) is mixed with a 4.5% w/w solution of iron (II) acetate in EMIMAc (previously prepared by 1 h of stirring), and left to stir overnight at 700 rpm. In this first step, the negatively charged thin graphite microplatelets 16 obtained during the ultrasonication process 21,22 , electronically interact with the Fe 2+ ions. Differently from the widespread procedures employing electrically insulating graphene oxide 10 as graphene precursor 20,[23][24][25][26][27] , here, by using thin graphite microplatelets, the characteristic π network of the graphitic carbon (and, consequently, the electrical conductive properties of graphite) is preserved. In the following step, the EMIMAc-based dispersion is subjected to microwave irradiation upon which the MUG-Fe 3 O 4 composite forms. Although rapid and efficient synthesis of metal oxide nanoparticles 28 with unique structures and properties were already obtained combining ionic liquids and microwaves 29,30 , to the best of our knowledge, this is the first time Fe 3 O 4 is prepared using iron acetate in combination with EMIMAc. It is furthermore worth noticing that, as previously reported and demonstrated 16,31 , the microwave heating process reduces the thin graphite microplatelets in size yielding to multilayer graphene (MUG). The composite material was recovered through vacuum filtration and thermally annealed in Ar atmosphere in order to improve its structure and morphology 32 . Finally, the complete removal of residual EMIMAc was performed via thermal decomposition resulting in the formation of benign volatile products.
Physicochemical Characterizations. As evidenced by the scanning electron microscopy (SEM) micrographs in Fig. 1b, the graphene sheets show jagged edges with irregular facets 33 along with the presence of nanoparticles. The multilayer morphology of graphene and the nanometric size of the metal oxide particles are revealed in Fig. 1c. Interestingly, a particular morphological feature is observed on the MUG surface where the iron oxide nanoparticles seem to create grooves (e.g., the ones marked by the red arrows) along various MUG edge planes 34 . It is worth noting that a similar structure, to the best of our knowledge, has been reported only twice in literature. These were, however, obtained by means of a completely different synthetic approach (i.e., catalytic hydrogenation of graphite with Nickel nanoparticles [35][36][37]. At higher magnifications (Fig. 1d,e), the details of the carved structures can be clearly seen and, the nanometric size of the Fe 3 O 4 particles is better appreciated (10-100 nm range). Apparently, the nanoparticles are able to carve different edge planes of the multilayer graphene causing, especially for the flakes with nanometric lateral sizes, a structural fragmentation on the carbonaceous matrix. When the synthesis is carried out in the absence of graphene (i.e., using only the iron precursor solution), larger Fe 3 O 4 agglomerates are formed ( Supplementary Fig. 1a,b). Moreover, if only the graphite microplatelets dispersion is processed (i.e., without the iron precursor), the obtained multilayer graphene shows a rather different morphology 16 , free of carved structures ( Supplementary Fig. 1c,d). The powder diffraction pattern in Fig. 1f allows identifying the different crystallographic features of multilayer graphene and iron oxide. The most intense reflection around 26.5 2 Theta degree is associated with the (002) diffraction plane of graphitic layers 38 . As shown in the inset in Fig. 1f, the other reflections between 35 and 80 2 Theta degree (marked with + ), can be ascribable to different graphitic crystalline planes of multilayer graphene, belonging to either hexagonal or rhombohedral phases [38][39][40] . Additionally, few low-intensity reflections (marked with §) might be due to a limited presence of non-graphitic carbon domains as carbonaceous by-products of the synthetic process 13,[40][41][42][43][44] . Minor reflections corresponding to the Fe 3 O 4 phase (marked with |) are consistent with the standard XRD data (PDF# 01-073-9877) for the cubic magnetite phase 45 ( Supplementary  Fig. 1e). As expected, the powder diffraction pattern of multilayer graphene ( Supplementary Fig. 1f ) only shows the carbon reflections but no Fe 3 O 4 signature.
Further confirmation for the MUG-Fe 3 O 4 peculiarity features comes from transmission electron microscopy (TEM) analysis. The scanning TEM (STEM) image shown in Fig. 2a highlights the z-contrast between the MUG graphitic layers and the iron oxide nanoparticles. Here, some of the carved paths traced by the Fe 3 O 4 nanoparticles (marked with red arrows) can be clearly observed. The bright field TEM micrograph (Fig. 2b) reveals the nanometric size of the iron oxide particles (between 5 and 30 nm in this specific micrograph), which are nested in the carved multilayer graphene matrix. The contrast of the dark Fe 3 O 4 nanoparticles and the bright grooves (shown by red arrows) generated by the nanoparticles, allow a deeper investigation of the composite structure. Indeed, differently carved paths (width ranging from 10 nm to 40 nm) on different edges planes of the MUG structure can be clearly noticed. In Fig. 2c, the corresponding selective area electron diffraction (SAED) pattern to Fig. 2b is shown. The d-values measured from the SAED pattern fit to the XRD reflections for the iron oxide structure, with lattice spacings of 0.295 nm, while the peaks at ~0.207 nm and ~0.125 nm fit both Fe 3 O 4 (400) and (622) as well as graphite (100) and (110) reflections. The high resolution TEM micrograph (HRTEM) in Fig. 2d distinctly shows an iron oxide nanoparticle covered by thin graphitic carbon layers (marked with red arrows). According to this micrograph, it is reasonable to assume that MUG fragments, including those generated by the carving, are wrapped around the Fe 3 O 4 nanoparticles during the composite synthesis. To the best of our knowledge, similar results were only obtained by chemical vapour deposition, as also recently reported by Chang et al. 46 However, EMIMAc can be excluded as source of carbon since it is known that it decomposes exclusively to volatile products upon thermal treatment 47,48 (Supplementary Fig. 2). To further enlighten the MUG-Fe 3 O 4 structure and composition, additional analyses were performed. The area marked by the purple rectangle in Fig. 3a,b has been investigated by STEM-EELS mapping (looking at the C K-edge, Fe L 2,3 -edge and O K-edge) in order to further confirm the elements present in the composite. In Fig. 3c-e, the elemental maps for C, Fe and O elements are shown, respectively. As expected, C clearly represents the most abundant element, while Fe and O are mostly confined where the nanoparticles are located. Fig. 3f, depicts the EEL spectra of a selected area of the composite (orange rectangle in Fig. 3b). The C K-edge EEL spectrum (see inset in Fig. 3f) shows two features for the near edge fine structure: (i) a narrow peak at around 285 eV, corresponding to the electronic transition from 1 s to π * states 49 , which is characteristic for the carbon sp 2 -hybridization and, (ii) a sequence of broad peaks, ranging from 292 eV to 311 eV, which correspond to transitions from the 1 s to the unoccupied σ * 50 . The shape and the intensity of the latter peaks fits well those of an ordered graphitic structure 51,52 , thus supporting the previous XRD analysis. At the same time, the O K-edge and Fe L 2,3 -edge EEL spectra ( Fig. 3f inset with blue and green line, respectively) confirm the presence of iron oxide 53 . It is worth noting that no significant amount of N is detected at around 450 eV, thus, confirming the complete ionic liquid removal during synthesis. In Fig. 3g, the TGA curves of both MUG and the composite show the weight loss upon heating. The multilayer graphene is stable, in oxygen, up to 450 °C, however, a loss is observable between 460 °C and 670 °C. At 1000 °C, MUG shows a weight loss of about 91.15%, which is consistent with the results obtained from the CHNS elemental analysis (Table 1) 54   carbonaceous species (i.e., those denoted as "non-graphitic carbons" in the XRD analysis which also include amorphous carbons) can be obtained, as already reported in literature for graphene and other kinds of carbon [56][57][58] . Regarding the composite, MUG-Fe 3 O 4 , its stability extends up to 500 °C, i.e., before showing a similar weight loss between 530 °C and 670 °C. Considering the TGA curves of EMIMAc and EG ( Supplementary Fig. 2), the 88.32% of MUG-Fe 3 O 4 weight loss at 1000 °C, is compatible with the presence of the Fe 3 O 4 in the composite 57 . Indeed, it is possible to calculate the Fe 3 O 4 amount as 2.73 wt. %, corresponding to a Fe content of about 1.98 wt. %, which is in the same order of magnitude of the ICP result reported in Table 1 (Fe content of about 0.25 wt. %). It is worth pointing out that no Fe was detected by ICP in the EG precursor or MUG (Table 1), thus confirming the presence of iron only in the composite. It should be furthermore noted that the low hydrogen content found in both MUG and MUG-Fe 3 O 4 is in agreement with the presence of minor non-graphitic carbon domains (as previously revealed by the XRD analysis) and dangling-bond-containing edges 44 . The influence of grooves on the porosimetric features of the composite has been also investigated by Brunauer-Emmett-Teller (BET) surface area and Barrett-Joyner-Halenda (BJH) pore size distribution analyses. From the nitrogen adsorption-desorption isotherms for MUG and MUG-Fe 3 O 4 (Fig. 3h) similar values of BET surface area are calculated (48.27 m 2 g −1 and 51.79 m 2 g −1 for multilayer graphene and composite, respectively). However, as already reported 16,59,60 , the graphene sheets' arrangement play a crucial role in this analysis as it can considerably reduce BET surface area values up to one order of magnitude. According to the BJH analysis (Fig. 3i), the presence of Fe 3 O 4 has a non-negligible effect on the pores size and volume of the composite. Up to 9-10 nm pore radius, no substantial difference in the pores volume could be noticed. However, above 10 nm the total pore volume increases reaching the final values of 0.79 cm 3 Fig. 3), showing the typical issues of a conversion material (e.g., high voltage hysteresis, capacity fading upon cycling and low coulombic efficiency), which prevent its practical application in LIB 61 .
Regarding MUG, its specific capacity, coulombic efficiency and, in particular, the voltage profiles are similar to those of graphite but, contrary to the latter, it shows acceptable lithium storage capacity even at high specific currents and low temperatures 16 . As all graphenes, MUG generally exhibits low density 4 and, consequently, its volumetric capacity is smaller than that of graphite. Surprisingly, however, the carbon density in MUG-Fe 3 O 4 is higher than that in pure MUG, which represents an interesting property.
To evaluate the electrochemical performance, MUG-Fe 3 O 4 -based electrodes were investigated in half-cell configuration and subjected to repeated galvanostatic charge/discharge cycles in the narrow voltage range comprised between 0.005 V and 1 V (Fig. 4a). In literature, negative electrodes are frequently delithiated up to 3 V. However, only a restricted operative voltage window is of practical interest, as it results in lower average lithiation/ delithiation voltages 15 (hereinafter ALV and ADV, respectively) and thus higher voltage efficiency (VE) (Supplementary Fig. 4a-c). Of course, limiting the delithiation to 1 V results on the loss of the Fe 3 O 4 conversion mechanism, which, however, contributes minimally to the composite's capacity, and only at medium-low currents ( Supplementary Fig. 4d). The results in Fig. 4a, especially the comparison with MUG and SLP30 (this latter being a commercial graphite anode material of Imerys Graphite & Carbon), indicate as the metallic Fe nanoparticles, formed during the first lithiation, as well as the carved paths, have a beneficial effect on the high rate performance (i.e., above 2 A g −1 ). During the 1 st cycle, MUG-Fe 3 O 4 obviously showed the highest lithiation capacity due to the conversion reaction of Fe 3 O 4 to Fe 0 , but, for the same reason, also the lowest CE (i.e., 63.8% vs. 76.1% and 81.1% for MUG-Fe 3 O 4 , MUG and SLP30, respectively, in Fig. 4b) (see the voltage profiles and detailed discussion in Supplementary Fig. 5). After 10 cycles, however, all active materials showed stable and similar performance, with average capacity values of about 335 mAh g −1 , 348 mAh g −1 and 370 mAh g −1 for SLP30, MUG and MUG-Fe 3 O 4 , respectively. However, as specific currents higher than 0.2 A g −1 were applied, commercial graphite suddenly lost its storage capability. The decay of electrochemical performance is even more dramatic at higher currents, e.g., 1 A g −1 , where only about 80 mAh g −1 were delivered by SLP30. In comparison, under the same current load, both MUG and MUG-Fe 3 O 4 showed a stable behaviour, providing about 340 mAh g −1 and 350 mAh g −1 at the 50 th cycle, respectively. However, when the applied current was increased to 5 A g −1 , MUG-Fe 3 O 4 substantially outperformed MUG, also. Indeed, at the 10 th cycle at 5 A g −1 , an average specific capacity of 15 mAh g −1 , 180 mAh g −1 and 265 mAh g −1 were delivered by SLP30, MUG and MUG-Fe 3 O 4 , respectively. Upon prolonged cycling at such high current (i.e., after 180 cycles), MUG-Fe 3 O 4 offered 258 mAh g −1 , i.e., about 99% capacity retention respect to the 1 st cycle performed at 5 A g −1 , compared to 165 mAh g −1 and 18 mAh g −1 of MUG and commercial graphite,   Table 1.

C, H, N, S, Fe, O contents for expanded graphite (EG), multilayer graphene (MUG) and composite (MUG-Fe 3 O 4 ). ^CHNS Elemental analysis; *ICP Analysis; # Content determined by subtraction.
respectively. The rather poor performance of the latter material is well explained by the voltage profiles in Fig. 4c, highlighting how the ohmic drop of SLP30 graphite increased dramatically with the applied current. Differently, MUG and MUG-Fe 3 O 4 showed similar behaviour up to 1 A g −1 while, at 5 A g −1 , MUG-Fe 3 O 4 achieved considerably higher specific capacity. Fig. 4d (upper part) displays the evolution of ALVs and ADVs, for the three different active material, at different currents. It should be emphasised that ADV, in particular, represents a key parameter for an anode material because it relates with the energy output of the LIB incorporating it 15 . For all the three different active materials, at the lowest current applied (i.e., 0.1 A g −1 ), the ALVs and ADVs ranged from 0.12 V-0.14 V and 0.17 V-0.2 V, respectively. However, when the specific current of 5 A g −1 was applied, the ADV of SLP30 rise up to 0.7 V while the ALV is about 0.18 V. On the contrary, MUG and MUG-Fe 3 O 4 , showed ALVs of about 0.13 V and 0.10 V and ADVs of about of 0.30 V and 0.24 V, respectively. For the same selected cycles, the assessment of the voltage efficiency, has been carried out too. VE, which is defined as the ratio of ALV and ADV under specific current conditions 62 , is an excellent tool used to estimate the electrode materials' impact on the energy efficiency of the battery (for Coulombic efficiency > 99%) 63 . As shown in Fig. 4d (lower part), the VE of SLP30, already small at 0.1 A g −1 (i.e., 64%), reach the lowest value of 26% when the highest specific current is applied. Contrarily, both MUG and MUG-Fe 3 O 4 show similar VEs of about 82%, 68% and 43% for applied current of 0.1 A g −1 , 1 A g −1 and 5 A g −1 , respectively. To understand the reason behind the lower ALV and ADV of MUG-Fe 3 O 4 (i.e., its superior ability of storing lithium ions, especially at high currents), differential capacity analysis 64 has been performed (Fig. 4e). The differential capacity plots of the selected cycles in Fig. 4c Fig. 6a). On the contrary, MUG shows a polarization similar to MUG-Fe 3 O 4 , and all the peaks are still clearly detectable. At the 250 th cycle (specific current of 5 A g −1 ), graphite shows only a flat line, thus indicating the complete loss of the staging mechanism (see also Supplementary Fig 6b) Fig. 5a highlight a similar response for the two samples. Both materials feature, indeed, two depressed semicircles at high and medium frequencies, which can be associated to the diffusion of Li + in the SEI (R SEI | CPE SEI ) and charge transfer (R ct | CPE dl ), respectively. Interestingly, the SEI formed on MUG-Fe 3 O 4 appears to be more resistive, as also confirmed by the larger activation energy (see Fig. 5b). As reported in previous studies 67,68 , the presence of Fe 0 particles might play a catalytic role for the decomposition of the electrolyte, thus, leading to the growth of a thicker, but more stable SEI. The two samples present comparable activation energies for charge transfer (see Fig. 5b), suggesting that the Fe nanoparticles do not directly affect this process. However, MUG-Fe 3 O 4 interestingly shows a reduced resistance compared to the MUG. This might be associated to the presence of the grooves in the MUG-Fe 3 O 4 . We indeed propose that, by carving the MUG surface, the nanoparticles create additional graphitic edges where Li + can easily intercalate without affecting the battery performance. This peculiar morphology of MUG-Fe 3 O 4 would also enable shorter diffusion paths for Li + ions (see Fig. 5c). The low frequency behaviour seems to support this claim. For both samples, the low frequency tail does not show the typical Warburg-like behaviour (α = 0.5), but a much steeper rise, which can be better fitted with a constant phase element (CPE), usually associated to the limit capacitance element 69 . This indicates for a very minor effect of the Li + diffusion in the material solid phase on the overall lithiation/delithiation process. With this regard, the lower slope of MUG (α = 0.85 compared to α = 0.98 for MUG-Fe 3 O 4 ) suggests higher hindrance to the solid-state diffusion. Despite the intriguing mechanism and the highly promising results in terms of Li + -ion storage behaviour, we are aware that gravimetric values have only limited relevance when it comes to practical application 15 . Therefore, taking into account the electrode densities (Table 2), specific volumetric capacities were calculated for the three materials ( Table 3). The combination of these results with the ADVs allow us to depict a more realistic scenario about their possible use in real LIB applications 15 . At the lowest specific current (i.e., 0.1 A g −1 ), although it exhibits a smaller gravimetric capacity, graphite provides the highest volumetric capacity because of its higher density. At 1 A g −1 , however, the poor performance of graphite leads to its volumetric capacity being only half that of MUG-Fe 3 O 4 (40.5 Ah L −1 vs. 79.0 Ah L −1 ). Additionally, the ADV dramatically increases for graphite (0.412 V) with respect to MUG and MUG-Fe 3 O 4 , showing rather similar values (i.e., 0.184 V and 0.189 V, respectively). At higher currents (e.g., 5 A g −1 ), MUG-Fe 3 O 4 outperforms both reference materials, providing a volumetric capacity of 58.8 Ah L −1 and an ADV of 0.244 V. On the contrary, graphite only delivered 9.6 Ah L −1 with an ADV of 0.697 V. MUG, indeed, showed also a rather low ADV (0.303 V) but, due to its lower density and storage capability than MUG-Fe 3 O 4 , also a limited volumetric capacity (33.7 Ah L −1 ).

Discussion
Carved multilayer graphene with nested Fe 3 O 4 nanoparticles has been synthesized through an innovative synthetic pathway. The characteristic structure and morphology of this material has been widely investigated proving its peculiarity in terms of structural arrangement between the Fe 3 O 4 nanoparticles and the multilayer graphene matrix. Its use as LIB anode revealed enhanced Li + storage properties of this material compared to multilayer

Synthesis of carved multilayer graphene with nested-Fe 3 O 4 nanoparticles. The EG-EMIMAc dis-
persion was prepared by mixing 1.5 g of expanded graphite (EG, SGL Carbon) and 155.5 g of 1-ethyl-3-methylimidazolium acetate (EMIMAc, Ionic Liquids Technologies GmbH) using a Hielscher Ultrasonic Processor UP400S equipped with a Sonotrode H7 probe and operating in continuous at full amplitude for 35 hours. An ice bath was used in order to avoid excessive heating of the dispersion upon ultrasonication. Afterwards, 15.0 g of the EG-EMIMAc dispersion was mixed with a 4.5% w/w solution of iron acetate-EMIMAc, and left to stir overnight at 700 rpm. The iron precursor solution was obtained by dissolving 2.6 g of iron (II) acetate (Alfa Aesar) in 55.8 g of EMIMAc under stirring condition (700 rpm) for 1 hour at 80 °C. The dispersion was placed inside a borosilicate glass reaction vials and exposed to microwave irradiation using an Anton Paar Monowave 300 microwave reactor using the procedure "heat as fast as possible" until the temperature of 230 °C was reached.  Table 3. Specific gravimetric and volumetric capacities and average delithiation voltages of graphite (SLP30), multilayer graphene (MUG) and composite (MUG-Fe 3 O 4 ) at different current rates.
drenched with 1 M LiPF 6 in EC:DMC 1:1 w/w electrolytic solution was employed as electrolyte (120 μL for each assembled cell). The electrochemical tests, galvanostatic charge/discharge cycles, were performed in two different potential ranges (from 0.005 V to 3 V vs. Li/Li + and from 0.005 V to 1 V vs. Li/Li + ) using a Maccor Battery Tester 4300 at 20 ± 1 °C. Different specific currents, ranging from 0.1 A g −1 to 50 A g −1 were applied. Cyclic voltammetry tests were carried out with a scan rate of 5 mV s −1 in a 0.005 V-3 V potential range using a VMP3 galvanostat/ potentiostat by Bio-Logic (France). The same instrumentation was used to acquire the electrochemical impedance spectra (sinus amplitude: 5 mV; frequency range: 1 MHz-10 mHz) at different temperatures (in a MK 53 climatic chamber from Binder). Prior to collecting the impedance spectra, the electrodes were subject to a charge/ discharge cycle at at 20 °C and 0.1 A g −1 (lithiation down to 5 mV and delithiation to 1 V vs. Li/Li + ) to ensure full formation, followed by reduction up to 0.11 V vs. Li/Li + and a 2 h rest step.
Electrode density calculations. The electrodes density was calculated by considering the volume of the coating layer (without the current collector) and active material mass only (no conductive additive and binder considered).
Average voltage calculations. The average lithiation and delithiation voltage was obtained by the integral of the voltage profiles (energy) divided by the specific gravimetric capacity.