Synthesis and Characterisation of Reduced Graphene Oxide/Bismuth Composite for Electrodes in Electrochemical Energy Storage Devices

Abstract A reduced graphene oxide/bismuth (rGO/Bi) composite was synthesized for the first time using a polyol process at a low reaction temperature and with a short reaction time (60 °C and 3 hours, respectively). The as‐prepared sample is structured with 20–50 nm diameter bismuth particles distributed on the rGO sheets. The rGO/Bi composite displays a combination of capacitive and battery‐like charge storage, achieving a specific capacity value of 773 C g−1 at a current density of 0.2 A g−1 when charged to 1 V. The material not only has good power density but also shows moderate stability in cycling tests with current densities as high as 5 A g−1. The relatively high abundance and low price of bismuth make this rGO/Bi material a promising candidate for use in electrode materials in future energy storage devices.


Introduction
The efficient storageo fe nergy is ak ey challenge in the adoption of renewable energy sourcesa nd the deployment of clean power technologies. In recent years, supercapacitors have been considered as promising candidates for the next generation of energy storage devices. [1] Compared with batteries, supercapacitors have higher power density and better cycle life. [1b, 2] However,t heir low energy density and small potential window limit their applications. [3] It has been suggestedt hat this problem may be addressed by hybrid systemst hat merge the advantages of supercapacitors and batteries, [3] often termed" supercapatteries". Such hybrid systems may involve an asymmetricc ell structure of ac apacitive and ab attery electrode [4] or,a sr eported here, composite electrodes that display elements of both supercapacitive and battery behaviours. [5] There have been an umber of previous studies in which composites of carbon and metal oxide were synthesized and their electrochemical properties analysed. [6] However,f ew studies have been reported on metal and metal composite materials. [7] Ag/C electrodes prepared through af acile hydrothermal methodf ollowed by ac alcination step achieved ac apacity value of 211mAhg À1 , [8] whereas Ru/mesoporous carbon compositess ynthesized by am icrowave-assisted methodr eached as pecific capacitance value of 287 Fg À1 . [7b] Ru/carbon nanocomposites preparedb yapolyol process at 170 8Ch ave Ru particlesa ttached to the carbon surface [7c] and with 60 %R u loading, this composite achieved as pecific capacitance of 549 Fg À1 . [7c] Some other metal nanoparticles, such as Au and Ag, have also been considered in electrodes and achieved acapacitance of 70 Fg À1 . [7d] However, the relativelyl ow abundance and high cost of these noblem etals limit their commercial applications.W ith increasing global concern regarding energy saving, environmental protection and CO 2 emissions, the searchf or al ow-costa nd environmentally friendly materialf or electrodes in energy storaged evices is important.
Bismuth, as one of the post-transition metals, has as table + 3o xidation state and can also exist at a + 5o xidation state. Its good electrochemical properties and environmentally friendly naturem ake bismuth an excellent candidate for use in electrode materials. [9] Recently,b ismuth was reviewed as one of the most extensively studied elements in solid-state physics because of its electronic properties, such as along Fermi wavelength (around 30 nm [10] )a nd high Hall coefficient. [10a] Ap articularly attractive feature of bismuth is that, in spite of its heavy metal status,i ti sc onsidered as afe andn on-toxic material. [11] Ar educed graphene oxide/bismuth (rGO/Bi)c omposite was synthesized for the first time using ap olyol processa talow reactiont emperature and with as hort reaction time (60 8Ca nd 3hours, respectively). The as-prepareds ample is structured with 20-50 nm diameter bismuth particles distributed on the rGO sheets. The rGO/Bi composite displays ac ombinationo f capacitive and battery-like charges torage,a chieving as pecific capacityv alue of 773 Cg À1 at ac urrent density of 0.2Ag À1 when charged to 1V .T he materialn ot only has good power density but also shows moderate stabilityi nc ycling tests with currentd ensities as high as 5Ag À1 .T he relatively high abundance and low price of bismuth make this rGO/Bi material ap romising candidatef or use in electrode materials in future energy storagedevices.
Moreover,alarge amount of bismuth is produced as ab yproduct of the coppera nd tin refining industry. [11] All these attributesm ake bismuth apromisingcandidate forelectrochemical energy storagematerials.
Here, we report on an ovel material, ar educed graphene oxide/bismuth composite (rGO/Bi). This composite material was prepared by am odified low-temperature polyolp rocess, in which hydrazine was used as the reducing agent [12] while ethylene glycol (EG) wasu sed as both solvent and reducing agent. An intermediate complex is formed by EG and the metal ions absorbed on the rGO surfacep roducing nano-sized particles and preventing aggregation. [13] Bismuth particles, which are oxidised and reduced during electrochemical cycling, are formed with an approximate lateral size of 20 to 50 nm and attach to the rGO sheets. Assembly of graphene into three-dimensional structures has the potential of creating electrodes with extremely large (and accessible) specific surface areas coupledw ith good electrical conductivity,w hiche nables fast electron transfer. [14] The decoration of such structures with faradaic charges toragem aterials can create composite electrodes that maximize electrode capacity beyond that offered by the theoretical upper limit of 550 Fg À1 (550 Cg À1 at 1V)i nc arbon-based materials. [14] Twoc omposite materials, similart ot hat presented herein, were the subjecto fp reviousi nvestigation.W ang et al. [15] investigated the electrochemical charge-storage behaviour of amorphous carbon-bismuth oxide composites with Bi 2 O 3 contents of between~14 and 33 %, which they incorrectly characterised as pseudocapacitive. It is important to differentiate between the specific capacitance and the specific capacity of an electrode. [16] The former refers to the capacitance per unit mass and is only applicable to charge storagethat is (pseudo)capacitive in nature-i.e.,d emonstrates an almost rectangular cyclic voltammogram (CV) and linear galvanostaticc harge/discharge (GCD) characteristics. Materials displaying non-capacitive faradaic charge storage (batterym aterials), whichp ossess peaks in CVs and plateau regions in GCD curves should be characterised in terms of the second quantity,t he total charge stored per unit mass. From the GCD data presented by Wang et al. [15] it is possible to derive as pecific capacity for their amorphous carbon/Bi 2 O 3 composite of~333 Cg À1 at 1Ag À1 .
The electrochemical behaviour of ar GO/Bi 2 O 3 composite containing 23.85 wt %B i 2 O 3 was also studied. [9d] Once more, this material was wrongly described as pseudocapacitive, the GCD data showing battery-like behaviour.F rom the GCD curve presentedi nt hat work it is possible to derive as pecific capacity for the rGO/Bi 2 O 3 composite of 204 Cg À1 at 1Ag À1 .H ere we report the structure,c ompositiona nd electrochemical performance of ar GO/Bi composite with as pecific capacity of 460 Cg À1 at 1.2 Ag À1 ,w hich is substantially larger than that of the previously reported materials, and reaches 773 Cg À1 at 0.2 Ag À1 .W es uggest that the improved specific capacity of the composite detailed in this work arises from the excellent electricalc onductivity afforded by the rGO backbone, the good electrical contact with the bismuth particles, which are initially deposited in metallicform, and ahigh utilization of bis-muth during charge/discharge, which is related to the microstructure of the composite.

Results and Discussion
An X-ray powder diffractogram (XRD) of the as-prepared rGO/ Bi composite is showni nF igure 1. The strongest three peaks appear at 27.06, 37.80 and 39.468,w hichc orrespond to the (012), (104) and (110) reflectionso fb ismuth, respectively (Natl. Bur.Stand.,U.S.), and therefore confirm the dominantp resence of bismuth metal on the graphene surface. The weak peak that appearsa t1 2.648 indicates an interlayer spacingo f 0.7 nm, which could be relatedt og raphene oxide (GO). [17] The small hump around 258 is caused by the disordered stacking of layers of rGO. [18] Peaks with positions at 30.02 and 32.668 cannotb ei ndexed with the crystal structure of bismuth, but are in agreement with the (103) and (110) crystal planes of bismuth subcarbonate (Natl. Bur.S tand.,U .S.). Both GO and EG, which were used as starting materials, could be the carbon source for the Bi 2 O 2 CO 3 observed. The absence of peaks related to bismuth oxidesi nt he diffractogram of the as-prepared composite indicates that the startingmaterialprimarily consists of rGO and metallic bismuth. It has been observedp reviously that bismuth metal nanostructures, such as nanowires or nanoparticles, readily oxidize when exposed to air at atmospheric pressure. [19] Metallic bismuth wires typically have an oxide layer 1nmt hick after 4h exposure to air. [19b] After 48 he xposure, the thickness of the oxide layer is~4nm. [19b] High temperature hydrogen and ammonia environments were found to reduce the oxide withoutd amaging the bismuth metal after as ufficient amount of time, but the oxide was found to reform in less than 1min of exposure to air. [19c] We note that graphene sheets act as impermeable atomic membranes to many gases [20] and therefore it is likely that the absence of significant bismuth oxidation observed in the as-prepared materials is related to ar etardation of this process though protection of bismuth by rGO.
In the Fourier Transform Infra-Red (FTIR) spectroscopy data from GO, Figure 2a,ab road peak is present between 800 to 1400 cm À1 which can be assigned to in phase CÀCÀOs tretching (800-1000cm À1 ), out of phase CÀCÀOs tretching (1000-1260 cm À1 )a nd CÀOÀHb ending (1200-1430 cm À1 )m odes. [21]  The peaks observed at around1 600 and 1720 cm À1 are attributed to the skeletal vibration from unoxidized graphiticd omains and the C=Os tretching of unsaturated carbonyl groups, respectively. [21,22] The broad peak appearing at 3200-3600cm À1 originates from the hydrogen bonded OH stretching vibration. [21,23] In the FTIR spectrumf rom rGO/Bi, the peak at 424 cm À1 mainly arises from the displacement of oxygen atoms with respect to bismuth causing BiÀOb ond elongation. [24] The peak that appears at 675 cm À1 results from BiÀO bonds of different lengths in distorted BiO 6 units. [25] The broad peak at around8 45 cm À1 can be attributed to the antisymmetric stretching of CO 3 groups. [26] Compared with the FTIR result from GO, rGO/Bi has fewer peaks in the range from 1200 to 2000 cm À1 and from 3200 to 3600 cm À1 ,w hich indicates the successful removal of oxygen functional groupsf rom the surface of GO.
Raman spectroscopy was used to comparet he density of defects in GO, rGO/Bi and rGO ( Figure 2b). Twoo bviousp eaks, which appear at around1 580 cm À1 (G band) and 1350 cm À1 (D band), were observed in all three materials. The peak at 1580 cm À1 is causedb yt he in-phase vibration of the sp 2 graphite lattice whereas the peak at 1350 cm À1 results from structurald efects and disorder. [27] Thei ntensity ratio of the Dand G-bandp eaks (I D /I G )c hanges from 0.90 in GO to 1.17 in rGO/Bi and 1.29 in rGO, indicating ad ecreasei nt he average size of the sp 2 domains. Similar results have been reported in the literature [12] and explained in terms of the creation of new graphitic domains upon reduction of GO to rGO, which are smaller in size but larger in quantity compared with those in the starting material.
The layered substance shown in the Scanning Electron Microscopy (SEM) images in Figure 3, with dimensions larger than 1 mm, can be identified as rGO. [28] Therefore, the particles with laterals izes in the range of 20 to 50 nm attached to the rGO layers are considered to be bismuth (see also discussion below). In some parts of the rGO/Bi samples bismuth particles are seen to have agglomerated and formed clusters with sizes larger than 500 nm, as shown in Figure 3d.
Transmission Electron Microscopy (TEM) images of rGO and rGO/Bi are showni nF igure 4a and b, respectively.A gglomeration is observed to occur in isolated regions of the sample, forming bismuth aggregates with lateral sizes larger than 200 nm, as seen in Figure4c. As elected area electron diffraction (SAED) pattern ( Figure 4d)o fo ne such particlei nF igure 4c confirms the crystal structure of metallic bismuth. Three rings are observedi nt his diffractionp attern, which correspond to reflections from the (012), (110) and (300)p lanes of bismuth metal (Natl.Bur.Stand.,U .S.). The SAED pattern is in good agreement with the strong peaks associated with metallic bismuth observed in XRD ( Figure 1).
An additional crystalline structure ( Figure 4e)w as observed in some locations in the sample. The atomic structure shown in the TEM image could be indexedw ith the (101) and (011) crystal lattice planeso fb ismuth subcarbonate. This resultc onfirms the existence of smallq uantities of bismuth subcarbonate as impurities, again in agreement with the XRD resultsp resentedi nF igure 1. Energy dispersive X-ray spectroscopy (EDS) from the rGO/Bi composite (Figure 4f)d isplays strong bismuth, carbon and copperp eaks. Bismuth peaks originate from bismuth particlesa nd bismuth subcarbonatea nd the carbon peak could contain contributionsf rom both rGO and Bi 2 O 2 CO 3 . The copper peaks are ar esult from the copper TEM support grid. The low carbon peak intensity compared with the high bismuth peak intensity suggestst hat the amount of bismuth subcarbonate is not great. Figure 4g and is how TEMi mages of the rGO/Bi composite after cycling. Agglomerationi so bserved to occur,f orming particles with sizes from 100 to 200 nm. Both bismuth (Figure 4h)a nd bismuth subcarbonate (Figure 4j)w ere observed in the SAED patterns obtaineda fter electrochemical cycling.
The microstructure and pore-size distribution of rGO and rGO/Bi were determinedf rom N 2 adsorption-desorption isotherms, Figure 5a and b, respectively.B oth isotherms can be classified as type Ii sotherms for microporous solids. [29] The assynthesized rGO/Bi is found to have as pecific surfacea rea of 10.55 m 2 g À1 with pore-size diameters in the range from 2-8nmw hereas the rGO has as pecific surfacea rea of 23 m 2 g À1 with pore-size diameters of 1-3 nm. Compared with rGO, the rGO/Bic omposite has al arger pore size, which may originate from the insertiono fb ismuth nanoparticles into the material and can facilitatem ore ready penetration of ions into the composite electrode, increasing surface accessibility.G iven that ap-  proximately half the weight of the rGO/Bi composite consists of rGO the reduction in specific surface area by af actor of~2 suggestst hat the incorporationo fb ismuth has not significantly changed the total surface area offered by the rGO component.
X-ray photoelectrons pectroscopy (XPS) was performed on rGO/Bi composites 27 monthsa fter fabrication. Figure 6s hows as urvey spectrumo btained from the rGO/Bi composite. There are strongp eaks associated withb ismuth, oxygen and carbon. As mall signal from nitrogen is also presentc orrespondingt o ac oncentration of < 2at% which,i nt he absence of any signal from Bi(NO 3 ) 3 (see below) is likely to originate from nitrogen inclusion in the rGO resulting from hydrazine treatment, as previously observed by Park et al. [30] No other elementsc an be detected.
Ah igh resolution XP spectrum of the Bi 4f lines is presented in Figure 7a along with the associated fit. Three components    [31] and Bi 2 O 3 . [32] The doublet located at 156.70 AE 0.04 eV (4f 7/2 )a nd 162.03 AE 0.04 eV (4f 5/2 )i sd ue to metallicb ismuth, [33] whilst the third doublet, located at1 57.8 AE 0.1 eV (4f 7/2 )a nd 163.1 AE 0.1 eV (4f 5/2 ), which must be included to ensure an appropriate fit, can be attributed to bismuth suboxides, such as BiO. [34] There is no evidence for Bi 4f components associatedw ith residual Bi(NO 3 ) 3 [33] or Bi in the + 5oxidation state in the XP spectrum. The relative strengtho ft he Bi IIIrelated doublet in comparison with that of the metal is explained by the surface sensitivity of XPS:u sing the approach of Tanuma, Powell and Penn [35] we determine the electron inelastic mean-free path for the Bi 4f lines to be~3nm. Hence, at hin oxide layer present on bismuth particles at the surface of the composite would be expected to dominate the XPS signal. Indeed,t he presence of a4 fc omponent associated with metallicb ismuth demonstrates that the surface oxide layer is no more than af ew nanometres in thickness.
Ah igh resolution XP spectrum of the C1s region is shown in Figure 7b.T he signal is dominated by an asymmetric graphitic line (Doniach-Šunjić line-shape, a = 0.14) with ab inding energy of 284.40 AE 0.05 eV,c onsistent with graphitic materials. Small peaks (< 5% of total C1s intensity) associated with CÀOH, C=O and O=CÀOH are located at 286.1 AE 0.5, 287.3 AE 0.5 and 288.7 AE 0.5eVb inding energy,r espectively,r eflecting residual oxygen containing groups on the rGO surface. [36] The fit component associated with CÀOH is the largest of these, consistent with previous observations that residual ÀOH groups are the most prevalent oxygen containing groups in rGO after hydrazine treatment [36] (although there may also be acontribution to this component from carbon bound to nitrogen [12] ). To obtain ag ood fit, it was also necessaryt oi nclude am inor peak (< 10 %o ft otal C1s intensity)a t2 85.2 AE 0.2 eV,w hich has previously been associated with sp 3 -hybridised defects within nanostructured carbons [37] suggesting that residual disorder remains in the rGO when oxygen-containing groupsa re removed.
The composition of the rGO/Bi composite was determined from the XP spectra by standard approaches [38] using photoelectron cross-sections calculated by Yeha nd Lindau [39] and inelastic mean-free paths determined as above. [35] The composite was found to contain carbon, oxygen and bismuth in the (atomic) ratio 0.78:0.18:0.03 (with an estimated error of AE 0.02 for each species).
Differential thermala nalysis( DTA) and thermo-gravimetric analysis( TGA) curves of rGO/Bi and rGO are presentedi n Figure 8. The DTAd ata from the rGO/Bi composite (Figure8a) show two broad exothermic peaks, P1 and P2, and two very small endothermic peaks, P3 and P4. P1 can be attributed to the adsorption of oxygen at the surfaceofr GO in the presence of bismuth at low temperatures (175-250 8C). The small endothermicp eak P3, at about 275 8C, is associated with the melting of metallic bismuth. The exothermic peak P2 is very broad and represents an overlapo fd ifferent exothermic processes: oxidation of bismuth between 325-375 8Cw hich involves am ass increase of 2-3 wt %a nd carbon combustion between 355-525 8C, accompanied by am ass loss of 18-20 wt %. The small endothermic peak, P4, at 730 8Chas no mass variation associated with it and probablyc orresponds to meltingo fb ismuth oxide, with the melting peak shifted to low temperatures because of the nanometer-scale dimensions of the particles. The broad peak labelled P5 in the TGA data from rGO/Bi, showingamass increase of around 5wt%,p robably corresponds to the combined effects of the processes described by the exothermic peak P1 and part of P2 in the DTAc urve,a ttributed to bismuth oxidation.
The DTAc urve of rGO powder (Figure 8b)e xhibits only one broad exothermic peak with an onset temperature of about 400 8C. This exothermic behaviouri sa ttributed to carbon combustion in air and takes place with am ass loss of 74 wt %. In the absence of bismuth the rGO combustion peak is shifted to higher temperatures. Over the temperature 450 to 800 8Cr GO/ Bi has as maller weightl oss compared with rGO. This might be because by adding the bismuth nanoparticles, the rGO/Bi has better graphitizationa nd de-oxygenation with enhanced van der Waals interactions between layers. [40] Based on the 5.13 %weightgain (peak P5), bismuth and rGO have aw eight ratio of 0.44:0.56. This is in good agreement with the atomic ratio given by XPS (bismuth and carbon were found by XPS to have an atomic ratio of 0.03:0.78, as discussed above,w hich corresponds to aweightratio of 0.4:0.6).  Electrochemical properties of the as-prepared rGO/Bi, rGO and Ni foam were analysed by CV under different scanning rates, as shown in Figure9.C Vm easured at different scanning rates, 20 and 50 mV s À1 ,p resented similar shaped curves. Voltages from 0.2 to À0.8 Vv ersus standard hydrogen electrode (SHE) were applied. Three clear peaks (A, Ba nd C) and as mall plateau (D) were observed in the CV experiments.P eak A, which appears at around À0.7 V, is associatedw ith the reduction of bismuth from the + 3o xidation state to the metallic state (0 oxidation state). [41] Peaks Ba nd C, which appear at À0.5 and À0.3 V, represent the formation of BiO 2 À and Bi(OH) 3 during the oxidation of bismuth from metal to the + 3o xidation state. [41] The surface layer of bismuth was partially dissolved in the KOH electrolyte and forms BiO 2 À in the first reduction reaction. [42] It is possible that the plateau Dm ay be due to the oxidation of un-transformed bismuth. [41] Ap revious study has shownt hat this plateau becomes dominant in bismuth films as the film thickness is reduced. [41] The plateau has only been observed in thin bismuth (metal) films with highly rough surfaces. [41] In Ref. [41] it is suggested that very high bismuth oxidation states of + 4o r+ 5m ight occur because of the hypothetical formation of gel like electrolyte (when bismuth metal is rough) and these oxidation states are responsible for the observed plateau through faradaic processes. However,C Vcurveso f rGO/Bi in the range from À0.2 to 0.24 V ( Figure 9c and d) display ar ectangularshape almostidenticalto that of rGO when scaled for the rGO mass content. It is therefore more likely that the constant capacitance value in this potentialw indow indicates that, over this range of potential, the rGO/Bi displays an electrical double layer (EDL) capacitance originating primarily from the rGO component in the composite and that the composite electrode therefore demonstrates both supercapactive and battery chracteristics.
During the oxidation and reduction processes intermediate products, whichi nclude Bi(OH) 3 , BiOOH and BiO 2 À ,m ay be formed as follows: [41,43] Peak A Peaks Ba nd C CVs of pure Ni foam and rGO were also measured and are presented in Figure 9a and bf or comparison. The CV of rGO measured at both 20 and 50 mV s À1 showr ectangulars hapes without any noticeable peaks, which indicates that the capacitance of rGO only arises from the EDL capacitance. [44] Chronopotentiometry wasu sed to study the charge/dischargeb ehaviours of the as-prepared rGO/Bi and rGO materials. The GCD curvesw ere measured at different current densities, rangingf rom 0.2 to 1.2 Ag À1 (Figure 10). The GCD curves of rGO/Bi show similar behaviour at different current densities.  In the enlarged discharge curve of rGO/Bi, as in Figure10c, both slope and plateau wereobserved.
The quasi-linear behaviour at the beginning of the discharge curve indicates ac ontribution from capacitor-like behaviour. This originates from charges tored electrostatically [16b] on the surfaceo fr GO, as described above.T he plateaui ndicates material undergoing ap haset ransformation during the redox reaction, [6b] as described by peak Ai nF igure 9a.T he charge/discharge characteristics of rGO, presented in Figure 10 b, do not contain any obvious peaks, in agreement with the CV results of Figure 9b.S ince the energy storage mechanism of rGO/Bi includes as ignificant non-capacitive faradaic or battery-like contribution, the appropriate wayt om easure the amount of charge stored in the electrode is the specific capacity (C s ) using C s = IDt/m, [3,45] where C s is the specific capacity (C g À1 ), i/m is the current density employed in the measurement (A g À1 )a nd Dt is the discharge time in seconds. [45] Figure 11 shows the specific capacity of the rGO/Bi composites calculated from the GCD curves. Composite samples achieved as pecific capacity value as high as 773 Cg À1 at ac urrent density of 0.2 Ag À1 .T he specific capacity is seen to decrease as the current density increases, which can be attributedt oi ncomplete utilization of the active material at high current densities. [9c] When ah igh current density is used, the redox reaction only occursa tt he surface of active materials. [9b] However, the rate at which the specific capacity drops decreases with increasingc urrentd ensity,i ndicating that the electrode material can still show good capacity even at high current density. When the current density reaches the range of 0.4-1.2 Ag À1 , the specific capacity maintains almostaconstant value, in the range of 587-494 Cg À1 .T he specific capacitance of the pure rGO was found to be 283 Fg À1 at ac urrent density of 0.2 Ag À1 , which is comparable to the value of 205 Fg À1 foundf or gasphase reduced rGO. [46] At ac urrent density of 1.2 Ag À1 ,t he specific capacitance of rGO was found to decrease to 125 Fg À1 , which, at ap otentialo f1V, stores approximately aq uartero f the capacity value found forthe rGO/Bi material.
From the bismuth content of the rGO/Bi composite it is possible to calculate the maximum theoretical contribution to the total specific capacity of the electrodef rom this component of the material. The specific capacity associated with oxidation of bismuth is 1385 Cg À1 resulting in ac ontribution to the electrode materialo f6 10 AE 20 Cg À1 (170 AE 6mAhg À1 ). If the specific capacitance of the rGO in the composite is unaltered, we would therefore expect ac ontribution to the specific capacity of the electrode of 160 AE 3Cg À1 when the voltage range of the galvanic discharge curve is 1V (as used in our experiments). Hence, we would expect at heoretical specific capacity of 770 AE 20 Cg À1 for the composite overapotential of 1Vif all the bismuth present participates in electrochemical storage, which is remarkably close to the 773 Cg À1 measured at ad ischarge currento f0 .2 Ag À1 .T his result suggestsh igh accessibility of the bismuth within the rGO/Bi composite, reflecting the larger pore size of the composite material, compared with rGO, as described above.
Electrochemical impedance spectroscopy (EIS)w as performed on the rGO/Bi composite in af requency range from 10 mHz to 10 kHz using an alternating current( AC) amplitude of 5mV, Figure 12. As emicircle is observed in the high frequencyr egion of the plot (inset of Figure 12) corresponding to the faradaicp rocesses, while the linear part in the low frequency region corresponds to ion diffusion capacitive behaviours. [47] The solution, or series, resistance (R s )a nd the charget ransfer resistance (R ct )c an be estimated from the intercepts of the semicircle on the real axis, [15] which are 0.3 and 10 W,r espectively.T he low value of R s ,w hich originates from the resistance of the electrolyte and the internal resistance of the electrode, [48] indicates that the rGO composite is highly conductive, facilitating rapid charge transport. Thes mall value of R ct suggests that the electroactive bismuth particles are well coupled to the rGOs upport, whichm ight arise from the metallicn ature of the particles in the as-prepared material.
Cycling performance was determined by repeating the charge/discharge test 800 times at ac urrent density of 5Ag À1 (Figure 13). This sample achieves as pecific capacity of 235 Cg À1 at the start of cycling, which gradually decreases to 175 Cg À1 after 800 cycles. Hence, 74.5 %o ft he specific capacity was maintained after 800 cycles. The graduald ecrease of capacity during cycling may be owed to degradation of the active material, bismuth. [49] In addition, the relatively faster de-

Conclusions
Ar educed graphene oxide/bismuth (rGO/Bi)c omposite, in which the rGO inhibits atmospheric bismuth oxidation, has been synthesized the first time through ap olyol process in which ethylene glycol was used as both the solvent andr educing agent. The lowr eactiont emperature, short reactiont ime and low cost of starting materials make this synthesis procedure appropriate for large-scale application. The composite materiali sf ound to consist of bismuth nanoparticles with lateral sizes between 20 and 50 nm supported by rGO. The asprepared rGO/Bi composites displayed specific capacity values as high as 773 Cg À1 at ac urrent density of 0.2 Ag À1 .T he capacity of the rGO/Bi composite described in this work can be attributed to the excellent accessibility of the bismuth and the efficiency of electrochemical reaction resultingf rom high electrode conductivity and good contact between the bismuth nanoparticles and rGO. Since the electrochemical behaviouro f the composite shows contributionsf rom the electrical double layer capacitance of the rGO and faradaic charges toragea ssociated with bismuth, it is reasonable to describe rGO/Bi as a" supercapattery" material. This materialh as am oderate stability in cycling tests even at current densities as high as 5Ag À1 .T he excellent electrochemical properties of the rGO/Bi composite, simplicity of production and low cost indicate that this material is ap romising candidate as an electrode material in electrochemical energy storaged evices.
GO was prepared by am odified Hummers method. [50] Graphite (3 g) and KMnO 4 (8 g) were weighed and added into am ixture of H 2 SO 4 (100 mL) and H 3 PO 3 (20 mL). This suspension was kept at room temperature for three days while stirring continuously.H 2 O 2 was added into this mixture until it turned ab right yellow colour. This mixture was washed and filtered using 5% HCl and followed by deionised (DI) water (18 MW cm À1 resistivity) for several times until ap Ho f7w as achieved. GO was obtained after drying the deposit in an oven at 60 8Co vernight. Bismuth nitrate (0.3 mmol) and GO (0.03 g) were dispersed into am ixture of EG (23 mL) and nitric acid (2 mL). The suspension was sonicated to reach ahomogeneous dispersion. This suspension was transferred into ar ound bottom flask. Hydrazine (5 mL) was added into this suspension while stirring vigorously.T his reaction was held at 60 8Cf or 3hours. The synthesized material was collected in as mall sample vial after being washed with DI water several times and dried in air overnight. Undoped rGO was synthesized by the same approach to act as acontrol. . Ap otentiostat (Bio-logic Science instruments) was used to analyse the electrochemical behaviour of the composites, using CV and measurement of the charge/discharge behaviours. CV results were used to study the mechanism of the reaction taking place during the faradaic redox reaction of bismuth. As mall amount (9 mg) of sample was dispersed in DI water.P TFE (10 mg mL À1 )w as added as ab inding agent with as ample to PTFE weight ratio of 9:1. After obtaining ah omogeneous suspension by sonication, some drops were applied to an ickel foam substrate used as the current collector,w orking electrode. At hree-electrode system was used for the electrochemical properties test. 2.48 mg mixture of rGO/Bi and PTFE pressed on Ni foam was used as working electrode. AH gO/ Hg electrode was used as the reference electrode. AP tw ire was used as the counter electrode and aK OH (6 m) solution was used as the electrolyte. Current densities are quoted in Ag À1 as the true surface area of the electrodes is difficult to determine.
Samples for SEM imaging were prepared so that as mall amount of as ample was dispersed in absolute ethanol. This mixture was sonicated until ah omogeneous suspension was achieved. One drop of the suspension was cast on an SEM sample holder and dried in air.
The surface area of the samples was determined from N 2 adsorption isotherms using aS urfer system (Thermo Scientific). The samples were pre-degassed for 4hours at 10 À2 torr (1.333 Pa) before analysis. The surface area was calculated by measuring the amount of adsorbed nitrogen gas in ar elative vapour pressure of 0.05-0.3 at 77 Kb yB runauer-Emmett-Teller analysis.
XP spectra were measured with aK ratos Axis Ultra spectrometer, using monochromated AlK a X-rays (hn = 1486.6 eV) in normal emission geometry.H igh resolution XPS data were fitted using UNIFIT2007 [51] employing aS hirley-type background and peaks defined by ac onvolution between Gaussian-Lorentzian lineshapes, with the exception of the main C1sl ine, which is fitted with the asymmetric Doniach-Šunjić lineshape characteristic of graphitic materials. [52] It was not possible to determine aunique value for asymmetry parameter of the Doniach-Šunjić line and hence av alue of [37] However,s imilar results were obtained using asymmetries characteristic of bulk graphite. [52] The accuracy of resulting fits was attested to by reduced c 2 values close to 1a nd minimal systematic variation in the fit residuals.
DTAa nd TGA measurements were carried out in air with ah eating rate of 10 8Cmin À1 from 100 to 800 8C. Pure alumina was used as the reference material. The accuracy of these analyses is about 1-2%.