Facile synthesis of iron oxides/reduced graphene oxide composites: application for electromagnetic wave absorption at high temperature

Iron oxides/reduced graphene oxide composites were synthesized by facile thermochemical reactions of graphite oxide and FeSO4·7H2O. By adjusting reaction temperature, α-Fe2O3/reduced graphene oxide and Fe3O4/reduced graphene oxide composites can be obtained conveniently. Graphene oxide and reduced graphene oxide sheets were demonstrated to regulate the phase transition from α-Fe2O3 to Fe3O4 via γ-Fe2O3, which was reported for the first time. The hydroxyl groups attached on the graphene oxide sheets and H2 gas generated during the annealing of graphene oxide are believed to play an important role during these phase transformations. These samples showed good electromagnetic wave absorption performance due to their electromagnetic complementary effect. These samples possess much better electromagnetic wave absorption properties than the mixture of separately prepared Fe3O4 with rGO, suggesting the crucial role of synthetic method in determining the product properties. Also, these samples perform much better than commercial absorbers. Most importantly, the great stability of these composites is highly advantageous for applications as electromagnetic wave absorption materials at high temperatures.

complicated and time-consuming procedures, which greatly limit their potential large scale application. The development of a facile, cost-effective and scalable method to synthesize iron oxides/rGO composite with high EM wave absorbing performance is highly desired.
A huge disadvantage of the conventional magnetic absorbing materials is the loss magnetic properties and consequently EM wave absorbing properties under high temperatures 10 . In fact, the temperature increment due to the conversion of electromagnetic energy into heat may cause serious damage to magnetic absorbing materials and related 11 . This is especially true for military stealth materials for radar cross section (RCS) reduction. The heat generated on the surface of hyper-velocity missiles, bombers, rockets, aircrafts and spacecrafts due to friction can result in high temperatures (600-800uC), leading to composition changes and even destruction the EM wave absorbing materials.
Herein, we report a simple, efficient and scalable procedure for the synthesis of iron oxides/rGO composite from thermochemical oxidation of FeSO 4 ?7H 2 O and reduction of graphite oxide. Interestingly, the initially formed a-Fe 2 O 3 /rGO composite was converted to Fe 3 O 4 /rGO composite via c-Fe 2 O 3 /rGO intermediate when the temperature was increased from 500uC to 800uC. In contrast, treating FeSO 4 ?7H 2 O powders without the aid of rGO under the same conditions generated a-Fe 2 O 3 at every temperature point. This is the first example of phase transition between a-Fe 2 O 3 , c-Fe 2 O 3 and Fe 3 O 4 regulated by rGO. All three composites showed great EM wave absorption abilities. More importantly, the thermal stability of the Fe 3 O 4 /rGO composite at up to 800uC may open up a whole new field for high temperature application of carbon-based composite materials as EM wave absorption materials.

Results
Experimental Procedure and Samples Labeling. Experimental details can be found in Methods section. The starting material S90 was obtained by drying FeSO 4 ?7H 2 O and GO at 90uC in air for 24 h. After heating, the products were grounded in an agate mortar, annealed at a certain temperature and labeled as SX, where X denotes the treating temperature. The control samples were prepared by treating FeSO 4 ?7H 2 O powders (without GO) under the same conditions, and labeled as S 0 X, where X denotes the treating temperature.
X-ray Diffraction (XRD) and Raman Analysis. XRD patterns of GO and S90 are shown in Figure 1a. The characteristic peak at 2h 5 10.2u is indexed to the (001) plane of GO, indicating the oxidation of graphite to graphite oxide. The disappearance of the peak at 2h 5 10.2u in the XRD patterns of S90 (inset in Figure 1a) is due to the interruption of the GO layered structures with the intercalation of Fe 21 and SO 4 22 ions. All the peaks for S90 are well indexed to szomolnokite (FeSO 4 ?H 2 O, JCPDS 45-1365) and butlerite (Fe(OH)SO 4 ?7H 2 O, JCPDS 25-0409), resulting from the loss of water molecules and the partial oxidation of FeSO 4 during the drying process at 90uC 12,13 . XRD patterns of control samples are shown in Supplementary Figure S1 in Supplementary Information. Figure 1b shows the XRD patterns of samples S500, S600, S700 and S800. For S500, the appearance of iron sulfate (FeSO 4 , JCPDS 17-0783 and JCPDS 33-0682) can be attributed to the loss of water from FeSO 4 ?7H 2 O at high temperature 13,14 , and the thermal decomposition of FeSO 4 led to the formation of hematite (a-Fe 2 O 3 , JCPDS 33-0664) 12 . At 600uC, parts of hematite are transformed into maghemite (c-Fe 2 O 3 , JCPDS 39-1346). Based on TG/DTG analysis and the XRD pattern of sample S540 (a-Fe 2 O 3 , Supplementary Figure S1), a mixture of a-Fe 2 O 3 and c-Fe 2 O 3 was observed at the temperatures between 540uC and 700uC. At 700uC, the product was dominated by magnetite (Fe 3 O 4 , JCPDS 19-0629) with a little bit of hematite. At 800uC, only magnetite phase was observed. Based on the above results, a clear phase transition sequence between the three iron oxides can be summarized: It is usually difficult to distinguish c-Fe 2 O 3 from Fe 3 O 4 based on XRD analysis due to their analogous inverse spinel structures. Therefore, Raman analysis was carried out, which showed that sample S500 is mainly a-Fe 2 O 3 , S600 is the mixture of a-Fe 2 O 3 and c-Fe 2 O 3 , S700 and S800 are Fe 3 O 4 (Supplementary Figure S2). These results are fully consistent with the XRD analysis.
Transmission Electron Microcopy (TEM) Analysis. Figure 2 shows the TEM images, HR-TEM images and SAED patterns of S600, S700 and S800. Many thin sheets of rGO decorated with iron oxides particles were detected, suggesting that they are easily exfoliated by sonication during the TEM test, since aggregation of rGO layers was observed in SEM images (Supplementary Figure S3 (Figure 2 b, c, f, g, j and k) suggests the existence of only Fe 3 O 4 phase. All of these results are consistent with XRD and Raman analysis. In addition, no significant morphological difference was observed among these three oxides.
Fourier Transformed Infrared (FT-IR) and Thermogravimetric/ Differential Thermogravimetry (TG/DTG) Analysis. In order to understand the mechanism of phase transitions of iron oxides, FT-IR and TG/DTG measurements were carried out and the results are shown in Supplementary Figure S4 and S5. The characteristic peaks of various carbon-oxygen functional groups from GO disappeared for S700 and S800, indicating the complete removal of epoxide and hydroxyl groups. For S800, only peaks of Fe 3 O 4 are detected (Supplementary Figure S4). Compared with pure FeSO 4 ?7H 2 O, the introduction of graphite oxide into S90 greatly reduces the starting decomposition temperature and the maximum weight loss temperature (510uC vs 430uC, 578uC vs 538uC). In addition, the endothermic peak at 538uC was identified as the decomposition temperature of FeSO 4 to a-Fe 2 O 3 (Supplementary Figure S5).

Discussion
Generally, c-Fe 2 O 3 tends to convert to a-Fe 2 O 3 under high temperatures 15 or other extreme conditions 16  Four main types of functional groups are believed to exist at the edges and on the surface of graphene oxide sheets: epoxide, carbonyl, carboxyl and hydroxyl 22 . Epoxide and hydroxyl groups are more stable than the other two kinds of functional groups 23,24 . This was confirmed in our FT-IR study, which showed that only epoxide and hydroxyl groups remained after annealing at 600uC. Based on the above analysis, a possible mechanism for the transformation from a-Fe 2 O 3 to c-Fe 2 O 3 in our system is proposed. From 540uC to 700uC, the temperature field provides energy for the rearrangement of O 22 from hexagonal closed-packing to cubic closed-packing. The hydroxyl groups and H 2 gas generated from the thermal reduction of graphite oxide are believed to play an important role in the transformation considering the similarities in conditions of our system with previously reported reaction systems. For example, the adsorbed hydroxyl groups have been reported to be crucial to the structural transformation from a-Fe 2 O 3 into c-Fe 2 O 3 , and the formation of c-Fe 2 O 3 phase started after the hydroxyl groups were exhausted 25 . Recently, ab initio calculations showed that the adsorption of H 2 on the Fe-terminate a-Fe 2 O 3 surface gained adsorption energy of 20.7 eV per H 2 molecule and the angles and bond lengths of Fe-O bond were changed 21 . The chemical bonds in a-Fe 2 O 3 have a tendency to those of c-Fe 2 O 3 , implying that hydrogen plays a key factor to the phase transformation. Considering the complexity of our sample, the exact transformation mechanism is to be explored.
The transition from c-Fe 2 O 3 to Fe 3 O 4 was believed to be facilitated by the porosity of the sample, which efficiently traps the reducing gases from the thermal reduction of graphite oxide. Nitrogen adsorptiondesorption isotherm analysis was carried out for sample S800 (Supplementary Figure S6). The calculated specific surface area and the most probable pore diameter are 20.1 m 2 /g and 4 nm. It was shown before that the thermal reduction of graphite oxide could generate reducing gases including CO, CH 4 and H 2 26,27 . Because of its porous structure, sample S800 could adsorb and confine these reducing gases especially when the sample is dense solid, thus providing reducing atmosphere for the reduction of c-Fe 2 O 3 to Fe 3 O 4 .
In addition, this mechanism was supported by the fact that phase transition initially takes place in the inner part of samples with unchanged a-Fe 2 O 3 shells covering the sample surface (Supplementary Figure S7 and Supplementary Figure S8). The reducing agents CH 4 , CO and H 2 first reacted with the particles inner surface (Supplementary Figure S9a), removing the oxygen ions at the surface layers and transforming Fe 31 ions into Fe 21 ions. The diffusion of the initially formed Fe 21 ions from the surface into the lattice generates the corresponding positive charges. Therefore, electron hopping process took place from the ferrous ions at the surface to the ferric ions inside the lattice. Consequently, Fe 31 ions are regenerated at the surface and the reduction process continues. The concentration gradient difference of iron/oxygen ratio between the surface and the inner part of the particles is responsible for the diffusion process. Considering the dimensions of the particles, smaller grain size could facilitate the diffusion process. With the aid of rGO sheets, electron hoping between Fe 21 and Fe 31 ions in Fe 3 O 4 may also take place between neighboring Fe 3 O 4 particles 28 .
In support of the proposed mechanism, the grinding of S90 into powders is harmful to the a-Fe 2 O 3 to Fe 3 O 4 transition. As mentioned above, S800 contains only Fe 3 O 4 phase. In contrast, annealing of S90 powders at 800uC afforded ground sample containing mainly a-Fe 2 O 3 (nonmagnetic) phase as indicated by XRD analysis (Supplementary Figure S9b, S9c). M-H hysteresis measure at 300 K indicates that the saturated magnetization of the ground sample and S800 are 1.93 and 80.4 emu/g, respectively (Supplementary Figure  S9d). The small magnetization of the ground sample suggests the presence of a small amount of magnetic iron oxides, which might be covered by the a-Fe 2 O 3 phase and could not be detected by X-ray analysis due to limited penetration. The crumples of GO sheets (Supplementary Figure S3f) and loose accumulation of FeSO 4 ?7H 2 O during the drying process are believed to confine the reducing gases and retard their effusion, facilitating the phase formation of a-Fe 2 O 3 .
As was shown above, the introduction of rGO sheets is beneficial to the stability of Fe 3 O 4 phase at high temperature (800uC), which is highly desired for radar absorbing materials, especially for hypervelocity missiles and spacecrafts. Therefore, the microwave absorption properties of sample S600, S700 and S800 were studied. The RL values were calculated using the relative complex permittivity and permeability at a given frequency and layer thickness based on the following equation (transmit line theory) 29 , where f is the microwave frequency, d is the thickness of the absorber, c is the velocity of light, Z 0 is the impedance of air, and Z in is the input impedance of the absorber. Figure 3 shows the relationship between the RL and frequency for sample S600, S700, S800 and the sample with 10% GO annealed at 700uC at different thickness. The minimum RL were found moving toward the lower frequency region with increasing thickness. Sample S600 with 4 mm in thickness shows the strongest absorption peak at 3.76 GHz with a minimum RL of 213.6 dB and a broad peak at 14.4 GHz with a minimum RL of 26.6 dB. Sample S700 and S800 with 3.5 mm in thickness have strong peaks at 6.72 GHz and 7.68 GHz with the minimum RL of 217.1 dB and 215.2 dB, respectively. The bandwidths of RL values below 210 dB are 0.96 GHz, 1.84 GHz and 2.35 GHz for S600, S700 and S800, respectively. Clearly, 3.5 mm is the optimal microwave absorption thickness for S700 and S800, which can be explained by the relationship between matching frequency and thickness 30  Where l, m r , and e r are the wavelength in the materials, complex permeability at f m , and complex permittivity at f m , f m and t m are the peak frequency and the matching thickness of maximum microwave absorptions, and c is the velocity of light. According to the eq. 3, the improved m r and e r of the materials are necessary to obtain small matching thickness t m in lower frequency. These results are of importance since the absorption frequency ranges can be tuned easily by changing the thickness of the absorbers, and broadband absorption can be achieved by multilayered absorbing structure 31 .
When the GO dosage in the starting materials is increased from 5% to 10%, the microwave absorption performance of the sample annealed at 700uC is further improved, with RL value of 220.8 d B at 6 GHz (Figure 3d).
The microwave absorption properties of materials are related with the impedance matching between dielectric loss and magnetic loss. The imaginary parts of complex permittivity and complex permeability symbolize the loss of electrical and magnetic energy, respectively. Loss tangents (tand E and tand M ) represent the loss properties of incident electromagnetic wave in the microwave absorbent 32 . Higher values of e0, m0, tand E and tand M imply better performance of microwave absorption 6,33 . To investigate the possible mechanism of microwave absorption of the above three samples, the complex permittivity, complex permeability and corresponding dielectric and magnetic tangent loss of the samples with the thickness of 3 mm were measured at room temperature and the results are shown in Figure 4 and Table 1. Figure 4a and 4b display the frequency dependence of the real part (e9) and imaginary part (e0) of the complex permittivity, which represent the energy storage and inner dissipation abilities, respectively. It can be seen that the e9 of S600 first decreases from 24 to 17 in the range of 2-12 GHz, then increases and fluctuates and finally approaches a constant. On the other hand, the e9 of S700 and S800 are about 8 and 11 at 2 GHz, respectively and stay constant up to 8 GHz and show a small fluctuation between 8 and 18 GHz, indicating a resonance behavior. The e0 of S600 decrease first and then increase to 4.8 with a minor fluctuation in the 11.7-18 GHz range. For S700 and S800, the e0 are about 0.5 and 1.5 at 2 GHz, respectively and stay constant up to 8 GHz and show a small fluctuation between 8 and 18 GHz. It is well known that the resonance behavior of the permittivity originates from the electron polarization, ion polarization, space charge polarization, dipole polarization, and interfacial polarization 34 . The ion polarization and electron polarization often take place in the range of THz and PHz 35 , the resonance of permittivity are believed to be from the space charge polarization, dipole polarization, and interfacial polarization. For Sample S700, a layer of covered dielectric a-Fe 2 O 3 effectively decreases the electric conductivity and enhances the space charge polarization. Meanwhile, the dielectric a-Fe 2 O 3 layer also introduces additional interfacial polarization charges 36 . In addition, with the increase of frequency, the dipole polarization was reported to be the dominant factor, resulting in the fluctuation of complex permittivity 37 , which is consistent with our experimental results (Figure 4b). For the existence of dielectric a-Fe 2 O 3 , it is reasonable to deduce that the dipole polarization of S700 is larger than that of S800. Based on the above discussion, the e0 of S700 is higher than those of S800. The e0 valley of S600 is supposed to be from the c-Fe 2 O 3 phase. The slightly lower m0 value of S700 than S800 in the range of 2-16 GHz means a smaller magnetic loss. In general, magnetic loss is reported to be dependent of magnetic hysteresis, domain-wall resonance, eddy current effect, exchange resonance and natural resonance. Considering the weakly applied magnetization field during the electromagnetic measurement 38 , magnetic hysteresis could be neglected. The domain-wall resonance is reported to takes place in the MHz range, and can also be excluded 39 . For S700, the eddy current effect is effectively suppressed and can also be excluded due to the existence of the insulative a-Fe 2 O 3 layer. Therefore, the magnetic loss of S700 is mainly from the exchange resonance and natural resonance. Based on the reported exchange resonance frequency at about 12 GHz for cobalt 36 , the peak of m0 around 17 GHz can be attributed to the exchange resonance based on the Aharoni theory 40 . Figure 4e and 4f show the dielectric and magnetic tangent loss for the three samples, respectively. For S600, in the range of 10-15 GHz, the tand E is low; while in other frequency range, it is high. On the other hand, the tand M shows an opposite trend to that of tand E . These results suggest that in the low frequency range, the microwave absorption mainly results from the dielectric loss of a-Fe 2 O 3 , while in the high frequency range, magnetic loss of c-Fe 2 O 3 dominates. For both S700 and S800, tand E values are relatively low in the low frequency range (2-10 GHz) and then increase in the range of 10-18 GHz with a fluctuation, while the tand M of both samples are high in the low frequency (2-10 GHz) range and then decrease with the increasing frequency. Noteworthily, in the range of 16-18 GHz, S700 shows an abrupt increase, similar to the m0. Although the magnetic loss of S700 is lower than that of S800, the effective complementarity of the complex permittivity and permeability is more important to improve the microwave absorption property than merely a high magnetic loss, which can explain the microwave absorption of S700 is superior to S800.
Moreover, the EM wave absorption performance of the mixture of separately prepared Fe 3 O 4 powders and rGO sheets were studied. The Fe 3 O 4 powders was fabricated by one-pot co-precipitation method based on ref 41. The XRD pattern, Raman spectrum and SEM images of Fe 3 O 4 powders and Fe 3 O 4 powders/rGO can be found in Supplementary Figure S10 and S11. Figure 5a shows the RL of Fe 3 O 4 /rGO mixture. Compared with our samples, the RL loss is only 26 dB at 3 mm thickness. This clearly demonstrates the importance of synthetic procedure in the control of the EM wave absorption properties. Moreover, commercial absorber: [Fe,Ni] (Beijing Reintech Electronics Technologies Co. Ltd, China) was selected for the comparative study in the same frequency range. Figure 5b and c shows the RL of [Fe,Ni] before and after heating. The RL of the as-obtained [Fe,Ni] is only 27 dB at 3.5 mm thickness. After being heated at 800uC for 1 hour in Ar, the RL was decreased to 25 dB, both of which are much inferior to our samples. In contrast, the RL for sample S700 under the same conditions is 217 dB at 3.5 mm thickness. These results clearly demonstrate the advantages of the synthetic strategy, the great EM wave absorption properties of the asprepared samples, and the excellent thermal stability of these samples.  In summary, we reported the synthesis of iron oxides/rGO composites via a facile thermochemical process using graphite oxide and FeSO 4 ?7H 2 O. Through the regulation of the preparation temperature, we can obtain a-Fe 2 O 3 /rGO, c-Fe 2 O 3 /rGO, and Fe 3 O 4 /rGO composites conveniently. Interestingly, graphite oxide and rGO sheets were shown to regulate the phase transition from a-Fe 2 O 3 to Fe 3 O 4 via c-Fe 2 O 3 . The hydroxyl groups in graphene oxide sheets and the H 2 gas generated from the thermal reduction of graphene oxide were believed to be responsible for the transformation from a- Electromagnetic wave absorption studies indicated that these samples showed great EM wave absorption performances. With the increase of GO dosage, the absorption performances can be greatly improved. These samples perform much better than the mixture of rGO with Fe 3 O 4 , as well as commercial absorbers. Most importantly, our samples possess advantages for high temperature applications due to its excellent stability in high temperature (up to 800uC).

Methods
Synthesis of iron oxides/graphene composites. In a typical synthesis: 5.04 g (0.018 mol) FeSO 4 ?7H 2 O was dissolved into 20 mL distilled water with stirring for 30 min. Graphene oxide was synthesized by a modified Hummers method 42 . The resulted concentration of GO aqueous solution was 4.0 mg mL 21 .
The preparation of the iron oxides/rGO composites was based on ref 43. 63 mL GO aqueous solution was mixed with 20 mL FeSO 4 ?7H 2 O aqueous solution with ultrasonication for 30 min and vigorous stirring for 1 h. Then, the mixed suspension was dried at 90uC for 24 h, and the resulted black solid bulk was put in a quartz boat and placed in the middle part of a quartz tube, which was mounted horizontally inside a furnace. A protective gas of high purity argon (99.999%) was passed through the quartz tube at a rate of 500 standard cubic centimeters per minute (sccm) for 10 min to purge the air in the tube. The system was then heated to expected temperatures at a heating rate of 3uC per minute and held at that temperature for 1 hour before it cooled down to room temperature in the protective gas. The exhaust was imported into high concentration of NaOH solution in order to absorb SO 2 and SO 3 gases produced from the reaction process. The control samples were prepared by directly heating FeSO 4 ?7H 2 O powders in furnace with the same procedure without GO. Figure 6 shows the schematic diagram of sample preparation process. Graphene oxide is negatively charged due to the rich electronegative oxygen species (epoxy, carboxyl, carbonyl and hydroxyl) on the surface and edges 44 . FeSO 4 ?7H 2 O is easily dissolved in water and ionized into Fe 21 and SO 4 22 ions. After mixing, the positively charged Fe 21 will be adsorbed on the GO sheets due to the electrostatic interaction, leading to the intercalation of Fe 21 and SO 4 22 ions into GO layers 45  Electromagnetic wave absorption tests. The samples used for microwave absorption test were prepared by homogeneously mixing the composites with paraffin and pressed into a toroid with an outer diameter of 7.0 mm and inner diameter of 3.04 mm. The relative complex permittivity and permeability of the paraffin composites containing 50 wt% of the samples were determined by using a coaxial method with the vector network analyzer (Agilent E5071C, Agilent, USA) in the frequency range of 2-18 GHz. Finally, the microwave absorption properties were evaluated by the transmission line theory.
Characterization. The structures, microstructures and morphologies of the asobtained samples were characterized by X-ray powder diffraction (XRD Bruker D8-ADVANCE) using an 18 kW advanced X-ray diffractometer with Cu Ka radiation (l 5 1.54056 Å ), Raman spectroscopy (inVia-Reflex, Renishaw, UK), Field-emission scanning electron microscopy (FE-SEM, S4800, Hitachi, Japan), Transmission electron microscopy (TEM, JEM-2100, JEOL, Japan), Fourier transform infrared microscopy (VERTEX 801HYPERION2000, Bruker Optics, Germany), Simultaneous thermal analyzer ((DSC/DTA-T))STA 449 F3 JupiterH, Netzsch, Germany). Magnetic properties were measured by Physical Property Measurement System (PPMS-9T (EC-II), Quantum Design, USA). Barrett-Emmett-Tellter (BET) method in the relative pressure P/P 0 range of 0.01-0.20 was applied for the calculation of the pore volume, which was determined from the adsorption branch of the N 2 isotherm curve at the P/P 0 5 0.97 signal point. The pore diameter was derived from the maximum of the pore size distribution curve obtained using Barrett-Joyner-Halenda (BJH) method based on the adsorption branch of the N 2 isotherm curve.