Synthesis, First-Principle Simulation, and Application of Three-Dimensional Ceria Nanoparticles/Graphene Nanocomposite for Non-Enzymatic Hydrogen Peroxide Detection

Owing to the exceptional properties of graphene and the crucial role of substrate on the performance of electrochemical biosensors, several graphene-based hybrid structures have recently emerged, yielding improved selectivity and sensitivity. To date, most of the reported biosensors utilize solution-driven graphene ﬂakes with drawbacks of low conductivity (due to high inter-junction contact resistant) and structural fragility. Herein, we present a conductive three-dimensional CeO 2 semiconductor nanoparticles/graphene nanocomposite, as a platform for sensitive detection of hydrogen peroxide, an important molecule in fundamental biological processes. The 3D conductive graphene architecture is fabricated by chemical vapor deposition on nickel foam. The fabricated biosensor displays high sensitivity (60 μ A.mM − 1 ) at a low negative potential of − 0.25 V, a low detection limit ( < 1.0 μ M at S/N = 3), and a fast response ( < 5 s) in the range of 2.8 to 160 μ M. Furthermore, density functional theory simulations show that the improved detection is not only related to the catalytic effect of ceria nanoparticles, but also to more efﬁcient charge transfer from nanoparticles to the 3D graphene network. Moreover, it is established that the amperometric response of the biosensor is insensitive to interfering molecules such as glucose, sucrose, and potassium chloride, indicating its potential for practical applications.

Graphene, the two-dimensional (2D) form of sp 2 -bonded carbon atoms, has garnered much interest in energy storage and conversion, biosensing, and catalysis due to its outstanding electrical conductivity, an all-surface nature, and excellent electrocatalytic activity. 1 Of late, a wave of various innovative graphene-based architectures in forms of hydrogels, foams, sponges and aerogels has emerged. 2 These structures have enticed vast attention as they inherently possess structural interconnectivity and high porosity, and at the same time, all of the outstanding properties of graphene itself including low density, large surface area, unparalleled mechanical properties, biocompatibility, and excellent mass and electron transport. 3 Three-dimensional graphene (3DG) architectures can be fabricated either by self-assembly of graphene-based nanosheets (template-free methods) or template-assisted procedures such as chemical vapor deposition (CVD) and annealing. 4 While the template-free methods are technologically less complex and relatively cheaper, the obtained 3DG materials suffers from low electrical conductivity because of the low quality and high inter-junction sheets of chemically driven nanosheets. 5 Direct synthesis of 3DG networks by template-directed CVD provides large and fixable interconnected structure with high quality and electrical conductivity, which can be utilized as a platform for the fabrication of biosensors with quality sensitivity.
To enhance electrochemical performance of 3DG architectures for biosensing applications, many studies have employed nanoparticles (NPs) with different compositions, sizes and morphologies. This approach can be implemented in various ways; encapsulation of NPs by 3DG structures, 6,7 embedding catalytic NPs into 3DG via one-pot chemical synthesis methods, 8 and decoration of 3DG with NPs to form hybrid structures, 9 to name but a few. Noble metal NPs have extensively been utilized for the fabrication of electrochemical biosensors; however, high cost and scarcity of noble metals, their high working potentials, and poor selectivity are drawbacks. 10 To develop cost = These authors contributed equally to this work. a Present Address: CRANN, Trinity College Dublin, Dublin 2, Ireland. z E-mail: rezvanie@tcd.ie; simchi@sharif.edu effective substrates that can work at low negative potentials, metal oxide nanomaterials have been found of great interest during the last decade. 11 Among various metal oxides, cerium oxide, a wide bandgap semiconductor, is a promising candidate. In this oxide, Ce 3+ and Ce 4+ states can coexist on the surface. This is particularly important as it provides ceria with electrocatalytic ability (strongly related to the two oxidation states of CeO 2 , Ce 3+ and Ce 4+ ), redox activity, high oxygen storage capacity, free-radical scavenging properties, biocompatibility, and quick transition between oxidation states. [12][13][14][15][16][17] Ceria not only works as a support for catalytic reactions, but also functions as an electronic modulator for the electron transfer process in some catalytic processes; and in its nanosized form shows improved redox activity and oxygen transport properties with respect to the bulk form. 15,18,19 Moreover, the positively charged CeO 2 NPs at pH 7.0 is useful to bind negatively charged enzymes. 10,20 Considering the outstanding properties of graphene nanostructures including graphene oxide and reduced graphene oxide, efforts have been directed toward preparation of cerium oxide/graphene nanocomposites for electrochemical detection of enzymes (with low isoelectric point), uric acid, and H 2 O 2 . 21,22 Very recently, the combination of metal oxide frameworks with CeO 2 /reduced oxide graphene was also examined for the detection of uric acid at low negative potentials. 10 To the best of our knowledge, there has been no report on the combination of CeO 2 NPs with 3DG networks prepared by templatedirected CVD for detection of H 2 O 2 . Herein, we present high-quality interconnected 3DG networks comprising CeO 2 nanoparticles for non-enzymatic H 2 O 2 sensing at low potentials with satisfying sensitivity. It is important to note that oxygen vacancies in CeO 2 are crucial for sensitive and selective detection of H 2 O 2 because Ce 3+ triggers the reduction of hydrogen peroxide at less negative potentials. 17 H 2 O 2 is an important chemical product in many fields such as pharmaceutical, clinical and environmental area. 23,24 Furthermore, it is a common reactive oxygen species (ROS) and a by-product of numerous oxidative metabolic pathways. 25,26 ROS play critical roles in some signaling cascades present in living systems. At higher concentrations, ROS cause negative effects in living systems. Herein, hydrogen peroxide is important because of its stability and its links to a wide range of serious diseases such as eye and skin complications, gene mutation, and even premature aging of the human body. 24 So far, many enzymatic and non-enzymatic based electrochemical sensors have been employed for H 2 O 2 sensing. 27,28 Nevertheless, the application of enzymatic biosensors is limited by their relatively high cost, immobilization technique, poor reproducibility and the critical operating situation. 29,30 Although non-enzymatic H 2 O 2 sensors have the advantages of high stability and wide response window, enhancing the sensitivity and selectivity of the sensor is still a great challenge. It is shown that the CeO 2 /3DG composite substrate possesses better performance than the other reported structures (see Table I), due to the synergetic effect of the heterostructure.

Experimental
Template preparation.-a highly porous nickel foam (NF) (volumetric porosity >90%) with a thickness of 1.5 mm (Latech Scientific Supply, Singapore) was used as the template for CVD growth of graphene. Prior to the CVD process, the NF was ultrasonically cleaned for 5 min in acetone (Merck, Germany), isopropyl alcohol (Merck, Germany) and deionized (DI) water, respectively. All the organic solvents were of analytical grade and all the aqueous solutions were prepared with DI water.

Fabrication of biosensors from conductive CeO 2 /3DG
composite.-3D interconnected graphene structures were synthesized via CVD technique in a hot-wall quartz tube furnace (MTI Corp., USA), based on modification of the first report on CVD-grown 3DG network. 5 The NF was loaded into the furnace and ramped up to 1000 • C under an Ar/H 2 gas mixture of 450/50 sccm ratio. Then, the furnace dwelled at this temperature for 15 min to anneal the samples, reducing the oxide on the surface and modifying the grains of the nickel substrate. Graphene was then grown by passing a mixture of 10 sccm of CH 4 and an Ar/H 2 mixture of 45/5 sccm for 10 min. Subsequently, the CH 4 line was shut, the flow of Ar/H 2 was set to 30/3.5 sccm, the furnace was turned off and the quartz tube was gradually pulled out of the furnace's hot zone. The furnace was then let cool down to 60 • C and the samples were unloaded from the furnace. A colloidal suspension of cerium oxide NPs (55 nm, Tecnan, Spain) in DI water (1 mg/mL) was prepared by a probe sonicator (Ultrasonic homogenizers, Iran) for 15 min. To improve the dispersion of CeO 2 NPs in water, 100 μL of a 1M acetic acid (100% glacial acetic acid stock, Merck, Germany) was added to 20 mL of the suspension. The porous 3DG architecture was decorated with CeO 2 NPs through drop-casting of the colloidal suspension on all surfaces of the substrate, followed by drying in air overnight.
Materials characterization.-The prepared structures were characterized by Raman spectroscopy (Teksan Takram, 532 nm excitation wavelength) at up to 10 points across the samples. X-ray diffractometry (XRD) was performed in a Philips diffractometer with Cu K α radiation (λ = 0.1514178 nm). Field-emission scanning electron microscopy (FESEM; TESCAN-Mira3) was used to investigate the microstructural characteristics and growth morphological features of 3DG and the hybrid electrode. The instrument was equipped with an energy-dispersive X-ray (EDX) spectrometer which was used for compositional analysis and elemental distribution mapping in the regions of interest across the sample.
Electrochemical measurements.-Electrochemical measurements were carried out by employing an AutoLab PGSTAT302N electrochemical workstation (Eco-Chemie, Netherlands) in combination with a standard three-electrode system consisting of a working electrode, an Ag/AgCl electrode (saturated KCl) as reference electrode, and a platinum wire as counter electrode. Freshly-prepared phosphate-buffered saline (PBS, 0.1 M, pH 7.4) was used as the supporting electrolyte. All the electrochemical studies were performed under deaerated atmosphere (using high purity N 2 ). Before electrochemical measurements, the 3D substrates were treated by cyclic voltammetric scanning over 0.0 V to −1.0 V at a scan rate of 50 mV s −1 . The electrochemical activity of the CeO 2 -modified 3DG electrode was compared with NF and 3D graphene network.

Results and Discussion
Structure and composition of the 3DG hybrid architecture.- Figure 1 represents the results of the SEM investigation of microstructural and compositional characteristics of the sample, from the bare NF through the final 3D hybrid structure. As can be viewed in Fig.  1a, the Ni template is highly porous (>90%) with an interconnected pore structure and channel diameter of >500 μm. The struts have a size of about 100 μm and consist of a cellular structure with an average grain size of ∼8 μm (Figs. 1b, 1c). It is also noted that the Ni template (as-received) has a rough surface with elevated features being detectable in both the medium and high magnification SEM images (Figs. 1b, 1c). Figs. 1d-1f show the SEM images of the sample after the CVD growth process at the same magnifications as images (a-c). Most notably, it can be clearly seen that unlike the bare NF (Fig.  1b), no raised features are detectable in 3DG/NF hybrid structure and the surface is much smoother (Figs. 1e, 1f). This is attributed to the annealing step of the CVD process performed before the growth, rendering a smooth surface suitable for uniform growth of graphene. After graphene growth, equiaxed grains are covered with graphene layers, displaying a morphology that resembles a parched desert. The grown graphene is majorly continuous but of multilayer nature and has conformed to the underlying grain structure of Ni substrate. Fig.  1g shows the EDX spectrum and embedded oxygen and cerium maps of the 3DG/NF structure after drop-casting of the colloidal suspension of cerium oxide. The EDX spectrum consists of various Ni edges as the main constituent of the hybrid structure, oxygen core edge peak, and edges of cerium, though weaker in comparison to other peaks (the concentration of C, O and Ce from EDX measurement are 47.9 ± 7 at%, 12.9 ± 7.12 at% and 1.2 ± 0.2 at%, respectively). Considering that the Ni surface is covered with graphene, the observed O and Ce edges in the spectrum are ascribed to CeO 2 nanoparticles and not nickel oxide. Furthermore, the elemental distribution maps (embedded images), demonstrates the presence of CeO 2 NPs on the surface, and that Ce is uniformly distributed on the surface, with a slightly higher density of oxygen as expected in CeO 2 . XRD measurement was also carried to complement the SEM/EDX analyses. The XRD pattern (Fig. 1h)  . It is noted that the low intensity of the characteristic peaks is attributed to its low concentration and ultrafine crystal structure as well as strong and close peaks of nickel.

Electronic coupling between graphene and CeO 2 NPs.-In order
to evaluate the quality of the 3D graphene architecture and possible electronic coupling with CeO 2 NPs, Raman spectroscopy was employed (Fig. 2). Raman spectra show two prominent peaks at 1582 cm −1 (G peak) and 2706 cm −1 (2D peak) for the laser excitation wavelength used in this work. The linewidth of the 2D peak together with the I G /I 2D intensity ratio provide a fast estimation of the number of layers in graphene. 34 The full width at half maximum (FWHM) of 2D peak was measured to be ∼40 cm −1 and the I G /I 2D was ∼0.8, suggesting the presence of few-layer graphene. Interestingly, remarkable shifts occur after CeO 2 deposition. The G and 2D bands of graphene are evidently red-shifted from 1582 cm −1 to 1566 cm −1 , and from 2706 cm −1 to 2650 cm −1 , respectively. Since the Raman spectra were taken at several locations within the same and different samples, and always the same Raman behavior was observed, the lack of uniformity effect 35 on the observed significant frequency shifts is ruled out.
One of the possible reasons for these significant Raman shifts could be charge transfer between graphene and ceria NPs. It should be noted that the 2D peak position increases for p-doping, while it decreases for n-doping; and the G peak up-shifts for both types of doping. 35,36 Therefore, the strong redshift of G and 2D peaks observed in this work cannot only be explained by doping, and some other mechanism should be in play. The Raman spectrum of the hybrid electrode contains an additional feature, the F 2g band, which is the Raman peak of the bulk CeO 2 observed at 464 cm −1 . This peak is assigned to a triply degenerate first order symmetrical breathing mode of the oxygen atoms around the cerium ions. When examining the Raman spectrum of the hybrid electrode (red curve) in the region of F 2g band, due to the very low concentration of CeO 2 , as well as the high relative intensity of G and 2D peaks in comparison to F 2g peak (greater by a factor of 250), the region appears as smeared out. Thus, this peak is not very reliable and is discarded for the investigation of the properties of the system.
To investigate the possible interactions between CeO 2 NPs and the underlying graphene, a model based on density functional theory (DFT) was developed. The local spin density approximation (LSDA) as exchange correlation with Perdew-Zunger function with Double Zetta Polarized basis was used as implemented in Siesta Package. 37 In addition, to overcome the inaccuracy of DFT in modelling "f" orbitals, the Hubbard model on of Ce atoms were considered and defined by 5eV as reported before. 38 Additionally, the smallest stable structure of this group was considered as Ce 6 O 8 . An optimized supercell of graphene consisting of 72 atoms is considered and Ce 6 O 8 nanoparticle transferred on it at different possible positions and the whole new structure is optimized with 10 × 10 × 1 Monkhorst-Pack k-point sampling that defined after proper convergence tests. 39 The optimisation performed on different possible positions and the most stable one considered as the optimized structure. The mesh cut of 75Ry was defined after proper convergence test. The density of state (DOS) calculation was performed with 20 × 20 × 1 set value and the Mulliken population was studied to define the charge transfer. 40 Figs. 3a, 3b show the schematic representation and electron density of graphene with ceria NPs on its surface, respectively. The combination of orbitals between graphene and nanoparticles indicates that  there is a good bond between them, providing a platform for charge transfer. More careful analysis suggests that the orbitals near the Ce atom and graphene are combined. A slight deformation in the graphene plane due to columbic interaction of Ce 6 O 8 NPs is also noticeable. Fig.  3c shows the DOS of graphene in the presence of quantum nanoparticles. The clear shift of Fermi level toward conduction band demonstrates that there is a noticeable transfer of charge from nanoparticles to graphene which increase the number of electron carriers and lead to such shift in Fermi level. Mulliken population shows this transfer of charge to the graphene is in the order of 0.28e within the range of study, confirming the location of Fermi level. The DOS also shows the difference between spin up and down in the graphene due to Ce 6 O 8 . The few changes in the up spin compared with down one shows the potential of decorated graphene with Ce 6 O 8 NPs for spintronic applications. Based on the results of DFT calculations, the observed red-shift of the graphene Raman peaks is ascribed to the combinatorial effect of n-doping of graphene by ceria NPs, and the created (tensile) strain through coulombic forces; the principles behind the effects of strain and doping has been previously noted by Ferrari & Basko, 36 Mohiuddin et al., 41 and Ni et al. 42 Electrochemical activity of 3D interconnected CeO 2 /graphene composite was evaluated by cyclic voltammetry (CV) in PBS (2.0 mM, pH = 7.4) containing hydrogen peroxide (Fig. 4). At a constant hydrogen peroxide concentration (2.0 mM) and scan rate (100 mV s −1 ), growth of graphene significantly improves the response of electrode to oxidation of H 2 O 2 in comparison to the nickel substrate (Fig. 4a). Further improvement is attained by CeO 2 NPs deposition, owing to their active catalytic sites for electrochemical reaction. As can be seen, upon increasing the concentration of H 2 O 2 , cathodic current intensity is enhanced, and the electrode becomes more sensitive to the concentrations at low applied potentials. It is important to mention that; the developed electrode shows better responses to variation of H 2 O 2 concentrations (Fig. 4b)   species coordinated onto the surface of CeO 2 NPs. 43,44 At high H 2 O 2 concentrations (millimolar range and above), a rapid color change from yellow to orange was observed, which could be attributed to an increase in the ratio of Ce 4+ to Ce 3+ in the metal oxide, similar observations were also found by Perez et al. 45 During electrochemical reactions and applying a cathodic potential, Ce 4+ can reduce to Ce 3+ . The ability of these nanoparticles to reversibly switch from Ce 3+ to Ce 4+ is a key factor for their catalytic applications. Therefore, the active centers of CeO 2 NPs on graphene electrode could boost cathodic current. The Ce 3+ /Ce 4+ couple is regenerated during the reaction and participates in the electrochemical reaction. The suggested mechanism of the electrocatalytic reduction of H 2 O 2 is expressed as following: This catalytic reaction combines with porous and conductive structure of 3DG which provides a low resistance and promotes the electron transfer to reduce the response time. Furthermore, it has been shown 43 that the coordination number of Ce in the first coordination shell slightly increased after the addition of H 2 O 2 . Thus, peroxide species are proposed to be able to coordinate with surface Ce ions and decompose because of the intrinsic redox ability of Ce 3+ /Ce 4+ pair. Figure 4c shows CVs of the purposed electrode at different scan rates in the range of 0.05 to 0.40 V s −1 . The peak current is linearly varied with the square root of the scan rate (ν 1/2 ) (R 2 = 0.992) as Ipc(μA) = −33.4 v 1/2 (V. m −1 ) + 42.1. This observation may indicate a diffusion-controlled mechanism of the electroreduction. 46,47  We also examined the selectivity of the 3DG composite substrate through adding some interfering compounds (0.20 mM) which may have possible interference. Fig. 5c shows that the 3D hybrid electrode is only sensitive to a small amount of H 2 O 2 (0.10 mM). It is noticeable that purposed electrode possesses higher sensitivity and larger liner range toward H 2 O 2 detection in comparison to other reported structures.
In the next step, all electrodes were characterized and compared using electrochemical impedance spectroscopy (EIS). Fig. 5d shows the representative Nyquist plot of NF, 3DG/NF and CeO 2 NPs/3DG/NF in 5.0 mmol L −1 [Fe(CN) 6 ] 3−/4− solution containing 0.1 mol L −1 KCl. In these plots, the semicircle diameter implies a high charge-transfer resistance (R ct ) value. The equivalent circuit model is shown as inset in Fig. 5d. In this circuit model, the R ct is charge transfer resistance, where the diameter of the semicircle of Nyquist plot is proportional to the electron transfer resistance occurring at the electrode surface. R s is the solution resistance and C the capacitance. The evolution of the corresponding circuit model data is interesting. Upon growth of 3D graphene on NF, the charge transfer resistance of the system increases. It is postulated the lower R ct of NF alone could be due to its porosity that enhances the diffusion of the redox couple; and after the deposition of few-layer graphene on NF, the available surface for diffusion decreases or at least the diffusion path becomes more difficult, and as a result, the R ct increases.
Upon addition of CeO 2 NPs to the 3DG composite substrate, a clear decrease in R ct is observed, indicating the remarkable role of CeO 2 NPs as an electrocatalyst on the enhanced electron transfer rate. It is known that oxygen vacancies in cerium oxide (transition between Ce 4+ and Ce 3+ ) play a key role on its catalytic activity. 45,48 It has been previously reported that the activation energy for electrical conduction is lower for nanoceria. 49 To further analyze the behavior of the nanocomposite, the transmission spectrum of nickel, nickel-graphene and graphene with Ce 6 O 8 nanoparticles were studied. Although nickel and graphene show remarkable metallic behaviors, the transmission of nickel-graphene is lower than that of pristine nickel. The transmission spectrum demonstrates that the conductivity of pristine nickel is 18G 0 , whereas nickel with graphene deposited on top shows a conductivity of 16G 0 , where G 0 is quantum conductance and equal to e 2 / h . Such reduction in transmission is related to the tunneling of electron from nickel into graphene and vice versa which perturbs the basic channels inside nickel and reduces the successful transmissions. These tunneling effect are in good agreement with previous reports. 50 After decorating graphene with Ce 6 O 8 nanoparticles the conductivity is increased due to transfer of charge as discussed before.
As presented in Table I, the synergetic effect between 3DG and CeO 2 NPs yields a high sensitivity within a wide liner range toward H 2 O 2 detection.

Conclusions
The results of this work build up a comprehensive picture of the CeO 2 -modified 3D graphene nanohybrid biosensor system from experimental and theoretical standpoints. In general, the addition of CeO 2 nanoparticles lowers the charge transfer resistance of the 3D graphene/Ni foam system and provides non-enzymatic micromolar range selective detection of H 2 O 2 . When the electrochemical behavior of the Ni foam and 3DG/NF systems are compared, we can see that the charge transfer resistance of NF is lower than that of 3DG/NF. The DFT simulations suggest that the said increase in R ct could be due to the impaired electron tunneling mechanism for conductivity when graphene is deposited onto Ni foam. Further, DFT results may put forward a postulation for the observed Raman shifts in CeO 2 NPs/3DG/NF hybrid system, linking them to coulombic interaction and the resulting slight rippling of graphene. While this may be disadvantageous for some applications, one could cite the improved R ct to recommend CeO 2 nanoparticles modification for strain-independent applications.
While the sensing performance of CeO 2 NPs/3DG/NF are comparable to the common modifying entities such as Au NPs, the electrocatalytic properties of CeO 2 NPs such as oxygen storage capacity, multiple oxidation states and great electrocatalytic activity, a wider range of applications can be targeted using CeO 2 -modified 3D graphene. This means that for applications where, for instance, energy storage and conversion is important, CeO 2 NPs are the surface modifier of choice. Ultimately, the presented experimental and theoretical study of CeO 2 -modified CVD 3D graphene is only the beginning for sensing and energy applications. We hope that in light of this work the more common template-free synthesis and noble metal modifiers may be adapted to include CeO 2 NPs and its derivatives.