Voltage controlled polarity switching of photoresponse in graphene oxide-based memristor†

Soma Saha^a, Anindya Datta^ab and Tapanendu Kundu*^ac
aCentre for Research in Nanotechnology & Science (CRNTS), Indian Institute of Technology Bombay, Mumbai 400076, India
^bDepartment of Chemistry, Indian Institute of Technology Bombay, Mumbai 400076, India
^cDepartment of Physics, Indian Institute of Technology Bombay, Mumbai 400076, India. E-mail: tkundu@phy.iitb.ac.in

First published on 28th August 2024

---

---

1. Introduction

Graphene oxide (GO) has been intensively focused in the last few decades due to its unique conversion property from an electrically nonconductive state to a semi-metal state, known as graphene. The current–voltage (I–V) characteristics of GO thin film show nonlinear behavior, which deviates from the ohmic nature of pure graphene. Various conduction mechanisms play pivotal roles in determining the nonlinearity of the observed I–V curve.^1,2 This deviation originates due to the generation of defect states, various trap sites, and other charge carriers arising from different terminal groups present in the GO thin film. Hence, the extent of nonlinear characteristics depends explicitly on the degree of reduction, method of fabrication, and junction characteristics of GO-electrode configuration. According to the design of electrodes, different conduction processes of charge carriers are established based on the applied electric field across the thin film. In addition, upon illumination of the biased GO thin film, the photogenerated charge carriers also participate in the conduction process. This provides further opportunities to modulate the (I–V) characteristics of the GO-based device for different electronics and optoelectronics applications.

Besides nonlinearity, in general, GO thin film shows a hysteresis under the application of a sweeping voltage across the device. The hysteresis in the I–V plane, consisting of a low resistance state (LRS) and a high resistance state (HRS), is the signature of a memristor device. The memristor concept was proposed in 1971 as the fourth essential circuit element required to solve six mathematical equations relating to four fundamental electrical variables, i.e., current, charge, voltage, and magnetic flux.^3,4 However, it received much attention in 2008 when HP Lab fabricated the first memristor in a titanium oxide (TiO₂) thin film.⁵ Since then, many metal oxides⁶ have been employed to fabricate memristors. The resistive switching between HRS and LRS is the fundamental feature of developing memristor-based resistive random access memory (RRAM) devices.⁷ RRAM based on the metal/insulator/metal (MIM) configuration is one of the most prevalent substitutes in the present technology in terms of excellent performance and compliance with the standard microelectronic industry.^8,9 The charge transport processes in RRAM devices are heavily influenced not just by the choice of electrodes but also by the material selected as an active layer for the MIM structure. Recently, the attractive properties of GO, such as cost-effectiveness, easy fabrication process, and tunable film thickness, make it an ideal choice as an active layer for MIM-based devices.^10–14 GO sheets are highly oxygenated compared to graphene, with carboxyl and carbonyl groups at the sheet edges with epoxide and hydroxyl functional groups on their basal planes.^15,16 Additionally, GO sheets are extremely hydrophilic due to their surface functional groups, enabling them to disperse easily in water.^11,17 GO thin film-based MIM devices are generally fabricated in vertical and planar configurations. The vertical configuration has been widely attended, where the GO thin film is sandwiched between the top and bottom metal electrodes.^{10,12,18–26} Various charge transport mechanisms have been proposed to explain the observed resistive switching phenomenon in this kind of MIM structure. Jeong et al.²⁷ in 2010 reported the resistive switching in a GO-based symmetric MIM structure where GO is taken as an active layer between two aluminium electrodes. It was proposed that the observed switching of initial HRS to the LRS and returning back to the original state was due to the local filaments forming/decomposing in the thin insulating layer at the interface between top Al electrodes and the GO film. Such filaments were created by transporting oxygen ions between an insulating barrier (AlO_x) at the top electrode and the GO layer interface. It was also demonstrated that oxygen ions drifting and altering the energy barrier in the bulk GO might also be responsible for resistive switching.²⁸ The other mechanism was proposed where the metal electrode ions contribute to resistive switching, and the GO thin film only acts as an insulator medium.²⁹ Due to the strong electric field generated inside the device by applying a positive bias voltage to the top electrode, metal ions migrate through the GO layer from the top electrode to the bottom electrode. The resistive switching from HRS to LRS happens when a conducting filament is formed due to the accumulation of these metallic ions through the GO layer.^6,22,30–32 The filament ruptures when the voltage is withdrawn, and the system returns to the initial HRS state. In the case of the oxidation–reduction mechanism, the resistive switching in GO is caused by the drift of functional groups and its impact on the ensuing distinct sp³ and sp² domains throughout GO thin film. Due to the partial reduction of GO sheets, sp² nanoislands are generated across the sp³ domain of GO flakes.^18,33 When voltage is applied, these nanodomains expand, which increases the probability of tunneling between the sp² domains.³⁴ In the top and bottom MIM configuration, the voltage is applied across the stacked GO thin film, which is of the order of a few nm. At the device operating voltage, the magnitude of the electric field reaches almost the onset of the electric field required for reduction. Hence, there is always a probability of the nonreversible reduction of graphene oxide domains, which degrades the stability of the device.

Apart from these resistive switching characteristics, GO thin film also shows photoresponse when exposed to visible light. Jian et al.³⁵ showed that the overall current decreases upon exposure to light in the GO thin film on a planar electrode configuration. However, in the case of reduced graphene oxide (rGO), it has been observed that the current increases when it is exposed to light.^36,37 This is interesting in the sense that the photocurrent is subtractive rather than additive in the case of GO. It was proposed that the photogenerated excitons dissociate to free charge carriers, and the possibility of crossing the Schottky barrier at the GO/−ve electrode interface by the holes results in excess electrons in GO film. These offset the dark current; hence, the device current negatively responds to light exposure. The above-mentioned studies raised mainly two issues to be addressed. Firstly, understanding of the charge conduction mechanism in GO thin film with planar electrode configuration for memristor applications. Secondly, the interplay between the injected and photogenerated charge carriers establishes the time response of device photocurrent under different bias conditions.

In the vertical configuration, GO flakes typically align parallel to the surface, causing the current to flow perpendicular to the plane of these flakes. In such a configuration, devices with thinner GO films of thickness around 100 to 200 nm showed abrupt transitions between the HRS and LRS, forming a conducting pathway (filament) between the top and bottom electrodes.³⁸ In contrast, planar devices feature a current path that runs parallel to the GO flakes. Planar devices with thicker film thickness result in resistive switching with a smooth transition between the HRS and LRS states. Since the planar structure provides the opportunity to observe the conduction mechanism through the GO thin film under a low electric field environment,^33,34,39 optoelectrical characteristics of GO thin film in the planar MIM configuration have been investigated and presented to get insight into the above points.

In this work, a detailed study has been carried out on the I–V characteristics of dropcasted GO thin film in planar configuration under both dark and illumination conditions. A nonlinear, nonzero crossing hysteresis loop was observed in the current–voltage space when a sweeping voltage of ±5 V was applied. The formation of this hysteresis has been studied in detail to explain the observed phenomena, taking note of various charge conduction mechanisms. The GO-based memristor characteristics have also been extensively examined. The charge conduction under different bias voltages has been depicted in a schematic representation to explain the observed hysteresis in I–V space. Further, the photoresponse of GO thin film under illumination was also thoroughly examined and explored its dependency on various parameters, such as applied bias, sweeping direction, and illumination power. An interesting phenomenon of bias voltage-dependent polarity switching of photocurrent was observed under illumination. The observation of an exponentially growing photocurrent with a characteristic time constant of approximately 2 s at bias voltage +1 V suggests that the device exhibits capacitive behavior in response to light illumination, indicating the storage and release of charge carriers within the two planar electrode junctions. This switching of photoresponse was analyzed based on the interplay between the injected and photogenerated charge carriers. The time response of the device was also examined to understand the performance capability of the device for application purposes. The analytics are presented here to explain the time behavior characteristics of the GO thin film under various illumination conditions. The significance of this work lies in achieving a deeper understanding of the voltage-dependent photoresponse switching behavior in planar GO-based memristors. The insight into this time dynamics on current–voltage space will have an important impact on designing and engineering the pattern-based smart photoelectronic GO thin film platforms for various device applications. In essence, the originality of this work lies in bridging the gap between the observed nonlinear, nonzero hysteresis in current–voltage space and voltage-dependent photoswitching to explore the intriguing connections that offer insights into fundamental material behaviors and potential technological applications.

2. Materials and methods

Preparation of graphene oxide thin film

Water dispersion of graphene oxide (concentration of 500 mg L⁻¹, flake size of 0.3 to 0.7 microns, and thickness of one atomic layer) was commercially procured from Graphene Laboratories Inc. (USA). Before making a GO thin film, the GO solution was sonicated for 15 minutes to obtain a homogeneous GO stock solution. For the characterization, 20 μL of the GO solution was dropcasted on the piranha-cleaned quartz wafer/slide and left overnight to dry for making a GO thin film.

Experimental section

The thickness of the film was obtained using field emission scanning electron microscopy (FESEM). The SEM micrograph was taken with a JEOL JSM-IT800 microscope with an acceleration voltage of 10 kV and an emission current of 32.4 μA. The Raman spectrum of the film was acquired using the HR800-UV confocal micro-Raman Spectroscopy system (Horiba Jobin Yvon, France) of wavelength 532 nm. UV-Vis absorption spectrum was obtained on a Lambda 950 spectrometer (PerkinElmer, US). XPS analysis was carried out using a Kratos Analytical, AXIS Supra instrument.

Current–voltage characteristics

For current–voltage (I–V) characterization, 20 μL of the GO solution was dropcasted on fabricated planar Au/Cr electrodes on a quartz wafer to make GO thin film. To measure the I–V characteristics of the GO device under dark and illumination conditions, the Keithley 2450 source meter was connected in two probe configurations. Photoresponse of the device was carried out under illumination of Argon-ion laser (Steller-Pro, serial no-ML0947PR011AC0) of wavelength 514 nm of beam diameter 0.065 cm. During exposure, the effective area has been calculated, and the effective power on the film has been calibrated. An Arduino-based shutter controller was used for laser illumination under continuous and chopped conditions. All the measurements were carried out at room temperature with an average humidity of 44%.

3. Results and discussion

Characterization of the film

The Au/Cr electrodes were fabricated using a photolithographic process. The microscopic image of the fabricated Au/GO/Au device is shown in Fig. 1(a). The cross-sectional FESEM image of the GO thin film is shown in Fig. 1(b). The effective area of the GO thin film between the electrodes is ∼0.03 cm², and the average thickness of the dropcasted GO thin film is ∼13.55 μm. The absorption spectrum of the pristine GO thin film, as shown in Fig. 1(c), consists of two characteristic peaks at 230 nm and 300 nm. The origin of the 230 nm is due to the π–π* transition of the sp² carbon network arising from nanodomain sp² clusters of CC bonds, and the 300 nm peak is due to the n–π* transition of CO bonds.^36,37 The Raman spectrum of the GO thin film (Fig. 1(d)) shows the characteristics of D and G bands at ∼1355 cm⁻¹ and ∼1587 cm⁻¹ respectively. The C–C stretching gives rise to the G band and mainly occurs due to the in-plane dispersion of an E_2g phonon.^40,41 The ratio of the intensity of the D and G band I_D/I_G value has been calculated from the Gaussian fitted spectrum, as shown in Fig. 1(d). The I_D/I_G value from the Gaussian fitting was obtained as 1.02. The appearance of a 2D band at ∼2719 cm⁻¹ shows the presence of carbon sp² rings. The low intensity of this band is due to the presence of the oxygen groups in GO that significantly decreases the quantity of sp² hybridized carbon network. The combined band (D + G) is observed at ∼2937 cm⁻¹. The chemical bonds between carbon and oxygen in the fabricated GO thin film were analyzed using X-ray photoelectron spectroscopy (XPS). Fig. 1(e) shows high-resolution XPS spectra of the C1s region. The deconvoluted peak at a binding energy of 284.73 eV is associated with the C–C bond.⁴² The deconvoluted C1s peaks show peak binding energies at 286.9 eV, 287.62 eV, and 228.74 eV, which correspond to C–O, CO, and COOH molecular bonds.^42–44 Fig. 1(f) represents the high-resolution XPS spectra of the O1s region. The deconvoluted O1s peaks at binding energies 531.37 eV, 532.74 eV, and 533.29 eV represent CO, C–OH, and C–O–C bonds, respectively.^45,46 Table 1 represents the peak area (A) ratios of oxygen-containing bonds to CC bonds as determined by XPS, along with the I_D/I_G values obtained from Raman analysis.

---

Charge conduction mechanism of graphene oxide thin film

To get insight into the charge conduction mechanism in pristine graphene oxide thin film under dark conditions, the I–V characteristics of the prepared Au/GO/Au device (Fig. 1(a)) have been carried out with a scan rate of 20 mV s⁻¹. As shown in Fig. 2(a), a hysteresis loop has been observed in the I–V characteristics for forward (−5 V to +5 V) voltage sweep following the path ABCD and reverse (+5 V to −5 V) voltage sweep through the path DEFG. The observed hysteresis is almost symmetric in nature, and this infers that the barriers at both the Au/GO interfaces formed during the fabrication process were almost equivalent. The initial dark current of −3.96 × 10⁻⁷ Amp was observed at −5 V, which changed to 2.90 × 10⁻⁷ Amp at +5 V. During the reverse voltage sweep, the curve crossed at −3.83 V with the forward sweep curve at a current value of −2.24 × 10⁻⁷ Amp. This kind of nonzero crossing symmetric as well as asymmetric hysteresis loops in the current–voltage plane have been observed in various metal–oxides films. The observed nonzero current at zero volt generally occurs due to the capacitive as well as nano battery effect.⁴⁷ The dependency of scan rate and the magnitude of voltage sweep on this hysteresis loop has been examined. The effect of the magnitude of the voltage scan for ± 1 V to ± 5 V is depicted in Fig. 2(b). It has been observed that the overall hysteresis increases with increasing the magnitude of sweeping voltage. Fig. 2(c) shows the I–V curves of sweep voltage ±5 V for scan rates of 10 mV s⁻¹, 20 mV s⁻¹, and 50 mV s⁻¹. The variation observed in these curves could be related to the transit time of charge carriers governed by the various trap sites present in the GO thin film. The difference in currents at zero voltage (ΔI)_V=0 between the forward sweep at point C and reverse voltage sweep at point F is 5.1 × 10⁻⁸ Amp. The voltage at which the current is minimum is known as open circuit voltage or built-in potential (V_bip). The difference in V_bip between point B in the forward voltage sweep and point E in the reverse voltage sweep (ΔV_bip)_I=0 is 1.149 V. (ΔV_bip)_I=0 and (ΔI)_V=0 for different voltage sweep is plotted in Fig. 2(d). Both (ΔV_bip)_I=0 and (ΔI)_V=0 increase with Increasing the magnitude of the voltage sweep. The observed systematic changes can be attributed to the result of various charge conduction mechanisms across the film under the applied electric field.

To understand the mechanisms that govern the current at different voltage regions, the I–V curve of the first quadrant of Fig. 2(a) was obtained in a double logarithmic scale for further analysis, as given in Fig. 3(a). Based on the linearity, this curve has been divided into five regions and fitted with linear equation Y = nX + c, where n is the slope of the curve. A summary of the fitted parameters is listed in Table 2. From the slope of the curve, the processes of mobility of charge carriers in the thin film under the applied electric field could be envisaged. The voltage between 0.0195 V to 2.1795 V has been divided into two regions. Region I has a slope value of 0.142, and Region II has a slope value of 0.583. To get insight into these non-Ohmic regions, both regions I and II are plotted in lnI ∼ V^0.5 space as shown in Fig. 3(b), and a linear relationship between them has been obtained with a slope value of 0.50. This type of current–voltage relation implies that the conduction mechanism in these two regions could be due to the thermionic emission type process.^48,49 The transportation of these free carriers happened via Schottky emission. Here, charge injection dominates the conduction process. The logarithmic I–V relationship in region III with voltage ranging from 2.19 V to 4.99 V showed Ohmic conduction behavior with a slope of 1.18. This Ohmic conduction accomplishes that there is a formation of a conduction path between the electrodes through the GO film. In the reverse voltage sweep path (5 V to 0 V), the shape of the curve shows that the charge transport behaviour follows the trap-controlled space charge limited conduction (SCLC) mechanism.^50–52 Region IV (4.99 V to 1.04 V) shows linearity with a slope value of 1.54. This region shows a transition from Ohmic to the nonlinear dependence of the voltage. The region V (1.019 V to 0.736) has a slope value of 2.25, where the charge transport mechanism is driven by Child's law (I ∝ V²). It is to be noted that similar situations like region IV and V appear when the sweep voltage starts from −5 V. Therefore, the Au/GO/Au device shows different conduction mechanisms when the sweep voltage starts from −5 V. Since both the electrode junctions are similar, irrespective of the polarity, if the starting applied voltage is of the order of 5 V, this Au/GO device establishes a conduction path. As the magnitude of the voltage decreases, the Ohmic conduction changes to a trap-controlled charge transport mechanism. When the polarity of the voltage changes, the conduction changes to thermionic emission-controlled conduction, and finally, near the +5 V, the conduction enters the Ohmic region again. The resistance of regions I, II, and III are lower than regions IV and V, so the forward direction can be termed as the low resistance state (LRS), and the reverse sweep can be termed as the high resistance state (HRS). This kind of resistive switching between LRS and HRS is an essential feature of a memristor. The impact of various bias voltage scan rates on the performance of the GO-based device was investigated. When the voltage scan rates increased from 10 mV s⁻¹ to 50 mV s⁻¹ at bias voltage 1.2 V, it was observed that the resistance of LRS was almost unchanged while the resistance of HRS decreased (Fig. 3(c)). The resistance ratio of LRS/HRS achieves the maximum value when the voltage scan rate is 50 mV s⁻¹, as shown in the inset of Fig. 3(c). This behavior can be related to the transit time of injected electrons, depending on the trap sites present in the GO thin film and the GO interfaces. The sp² nanodomains in the bed of sp³ which act as trap sites, the injected electrons undergo trapping and de-trapping processes in the bulk GO film. At low bias voltage scan rates, injected electrons have sufficient time to undergo the trapping and de-trapping processes. In contrast, at high scan rates, the electrons lack enough time to complete these processes, resulting in a decrease in the resistance of the HRS. This variation in resistance is influenced by how the bias voltage affects the motion of charge carriers.⁵³

---

From the I–V curve of the device under different voltage sweeping windows, the resistance of HRS increases, and the resistance of LRS decreases with increasing voltage sweep shown in Fig. 3(d). This indicates that the voltage amplitude has a smaller impact on LRS compared to HRS.⁵³ It has been observed that the LRS/HRS resistance ratio is different for different voltage windows, and the LRS/HRS resistance ratio decreases with increasing voltage windows. Under the application of ±2 V, the LRS/HRS resistance ratio reaches a maximum of 0.32. It can be concluded that the performance of the GO-based memristor device can be tuned by applying different sweeping voltage windows and different voltage scan rates to the device. To evaluate the reproducibility and stability of the device, the endurance and retention characteristics of the device were analyzed for ±5 V. The endurance performance of the device was carried out for 60 cycles at an operating voltage of 1.2 V, shown in Fig. 3(e). The HRS and LRS behavior of the device stayed stable throughout the process. In Fig. 3(f), the retention property of the device is shown, and the device stayed stable and showed excellent retention performance over several cycles. Therefore, the device exhibits excellent stability and repeatability for device applications. A summary of the reported GO-based memristor, both in planar and vertical configurations, is tabulated in Table S1 (in the ESI†). It has been observed that the on/off current ratio varies from 7.20 to 10⁴ in vertical (top-bottom) configurations. The on/off current ratio for our fabricated device in planar configuration is found to be 21 at 0.7 V. It should be noted that apart from the device configuration, this value depends on many other factors, such as the type of electrodes, GO film thickness, electrode separation, and environmental conditions.

The above observations of the charge dynamics based on the applied voltage could be summarized in the following schematics as given in Fig. 4. The graphene oxide film can be considered as sp³ nature in bulk, but there are sp² nanodomain islands randomly distributed across the film. These sp² nanodomains can expand by increasing the applied electric field, which induces the conversion of the sp³ to sp² reduction process. If the electric field is reduced, these domains shrink by changing the sp² to sp³ oxidation process in the presence of ambient air or oxygen. It was observed that the rate of change of the current with respect to voltage is higher in air than in vacuum. These low bandgap sp² domains can act as the trap sites of the charged particles. The observed total current could be the result of tunneling through these nanoislands, which formed a conducting path between the electrodes, in addition to contributions from other sources, such as migration of defects in the material (oxygen vacancies).³⁸ When −5 V is applied initially, due to the high electric field, all the injected electrons reach the anode at the same time. This potential is enough to overcome the barrier of the trap sites, as depicted in Fig. 4(a). This results in a high current, as observed initially. When the magnitude of the voltage is gradually reduced, the trap sites start getting filled, and the magnitude of the current starts reducing drastically, as seen in between the voltage range −5 V to −0.573 V. This is the onset of the trap sites-controlled conduction mechanism. With more reduction of this voltage, the system thus gets into HRS. The observed zero current, not at the zero voltage, indicates that there is a generation of some current opposite to the bias current, as seen in Fig. 4(b). Due to the dielectric nature of the medium, the space charge generation may happen near the Au/GO junction, and this can generate the opposite electric field, reducing the net electric field inside the dielectric medium. At V_bip, these opposite currents balance the applied bias current, resulting in zero net currents. There are also possibilities of the charges being trapped at the junction barrier. When the bias voltage reaches zero volt, as presented in Fig. 4(c), the observed nonzero current is due to the electric field of this charge accumulation, as observed at point C in Fig. 2(a). By changing the polarity of the voltage, the applied electric field is now in parallel with the emf current. Due to this reason, the slope of the curve slowly changes as the magnitude of the bias voltage increases. Here, the system goes to LRS. With a further increase in voltage, the nanoislands further expand, and the probability of tunneling through these islands increases. Finally, beyond this particular voltage (2.1795 V), a continuous conduction path is established between the electrodes, and the conduction reaches the Ohmic region, as presented in Fig. 4(d). Similar to −5 V, at +5 V, the potential is enough to dislodge charge carriers from the trap sites, and the magnitude of the current further increases, as shown in Fig. 4(e). In the reverse sweep, by decreasing the magnitude of the voltage, the conduction goes again to the trap-controlled tunneling mechanism, and the system behaves like HRS.

Apart from the above conduction of charges, the photogenerated charge carriers also participate actively under the applied electric field. To understand the effect of these photogenerated charge carriers, the Au/GO/Au device has been investigated using a 514 nm CW Argon-ion laser of four different illumination powers. The I–V characteristics of this device under 1.6 mW, 3.3 mW, 5 mW, and 6.6 mW illumination power are shown in Fig. 5(a). It can be seen that the shape of the hysteresis loop is monotonously changed when the power is increased. The (ΔI)_V=0 and (ΔV_bip)_I=0 for different powers are shown in Fig. 5(b). The (ΔI)_V=0 initially increased and then continuously decreased when the power was increased. On the other hand, (ΔV_bip)_I=0 is linearly increased and reaches saturation when the power is 6.6 mW. A closed look at the I–V as depicted in the inset of Fig. 5(c), shows that during reverse voltage sweep at V = 1.3 V, there is no change in dark current under illumination. The current under illumination decreases when the bias voltage is more than 1.3 V. On the other hand, the illumination current increases by applying the bias voltage less than this voltage. A similar crossing point was also observed in the forward sweep at −1.96 V. This point in this I–V space is important in the sense that the device current under illumination can be increased or decreased in reference to the dark current based on the bias voltage. This point is the switching voltage point (V_sp), where illumination does not have any effect on the net current. At any bias voltage, the photocurrent (I_ph) can be expressed as I_ph = I_l − I_d, where I_l and I_d represent the illumination and dark currents, respectively. The dark current can be represented as V_bias/R_bias, where V_bias and R_bias are the device's bias voltage and corresponding resistance. Upon illumination, the voltage generated by the light can be depicted as V_ph, and the change of the resistance under illumination can be termed as R_ph. So, the illumination current can be written as (V_bias + V_ph)/(R_bias + R_ph). Thus, the photocurrent (I_ph) can be expressed as

	(1)

The above observation in Fig. 5(c) can be explained using eqn (1) by depicting three situations before switching voltage point (V_sp), after V_sp, and at the V_sp. Since the observed photocurrent (I_ph) is positive below V_sp (V < V_sp), the condition V_phR_bias > V_biasR_ph should be satisfied. The observed photocurrent is negative at bias voltage more than V_sp (V > V_sp); therefore, the condition V_phR_bias < V_biasR_ph will be satisfied. At V_sp, I_ph has zero value and can be written as

	(2)

The above condition shows that, since V_ph and R_ph depend on the intensity of light, the switching voltage point can be tuned by changing the illumination power. Fig. 5(d) shows the variation of V_sp under different illumination conditions where V_sp increases with increasing the laser power.

The observed voltage-dependent switching of photoresponse in GO thin film is an interesting phenomenon, and to carry out further investigation, the I–V characteristics with a sweeping voltage from 0 V to +5 V under dark and illumination conditions have been carried out with a freshly prepared sample and presented in Fig. 6(a). Similar to the previous observation, the switching voltage point occurs at 1.86 V. The bias voltage below this switching voltage point, the current increases upon illumination. However, when the bias voltage exceeds the switching voltage point, the photocurrent decreases, and the magnitude of this negative photocurrent increases as the bias voltage further increases. To understand this phenomenon further, the time response of the Au/GO device has been carried out under different applied voltages with a chopped (10 s) 514 nm Argon-ion laser. Fig. 6(b) shows the time response of the device under various bias voltages, keeping laser power constant. It is seen that before the switching voltage (at bias +0.5 V and +1 V), the photoresponse starts increasing exponentially when the illumination is on and reaches a maximum value. The device current returns back to the dark current upon turning off the illumination. Close to the switching voltage, at bias voltage +2 V, the magnitude of this exponentially growing photoresponse initially increases and then starts to decrease, reaching almost to the dark current when the device is still illuminated with the same power. After crossing the switching voltage, the nature of the curve completely changes to a fast initial growth followed by an exponential decay under illumination. Upon turning off the illumination, the nature of the curve changes to an initial decay, then followed by exponential growth. This resembles the differentiating nature of an R–C circuit. Thus, a simple way to represent this device as a parallel capacitive memristor device. When the applied voltage is less than the switching voltage, the resistance of the device is much higher and fulfills the condition V_phR_bias > V_biasR_ph, as mentioned previously. Below V_sp, the response time of this device is thus governed by the capacitive nature of the GO thin film. When the bias voltage is more than V_sp, the resistance of the device decreases, and thus, it behaves like a differentiating circuit. However, a closed look at the nature of this curve reveals that the device current at bias voltages +3 V and +4 V reaches below the dark current even when the illumination is still on. This can’t be explained by a simple parallel R–C model and needs further look into the charge dynamics, as mentioned earlier. This kind of offset of the dark current is also observed by Jian et al. at +5 V.³⁵ Table S2 (in the ESI†) includes the observation of negative photoresponse in GO-based devices reported earlier.⁵⁴ The aforementioned phenomenon can be understood by considering the interplay between the two counter-interacting illumination currents. The forward current consists of a bias current along with a photocurrent, and the reverse current can be thought of as a bias voltage-dependent current generated only under illumination. There could be several possibilities for developing this reverse current, such as tunneling of holes through the Schottky barrier, recombination of charges, and expansion and reduction of trap sites. Further, the generation of trap sites may be possible by GO reduction under illumination. However, this kind of reduction is reversible since the observed photocurrent returns back to the initial dark current when the illumination is off.

Thus, an empirical formalism of the time dynamics for the device photoresponse I_L(t) can be written as-

	(3)

where I_bias is the dark current, I_ph is the illumination current and a function of the intensity of illuminating light, I_bias,l is a constant, function of bias voltage along with the intensity of illumination. I_bias,l is considered to be zero under the absence of illumination and no bias conditions. The first two terms of eqn (3) constitute the forward current having the time constant τ₁. The third part of this equation represents the reverse current with a time constant τ₂. When the applied voltage is below the switching voltage point (V < V_sp), as shown in Fig. 6(b), I_bias,l is very small, and the second term of the equation dominates with respect to the third term under illumination, and the resultant photocurrent is additive as observed. Thus, the exponential rise in the current profile below the switching voltage point can be very well explained with the growth term in eqn (3). The baseline subtracted curves Fig. 6(b) were fitted with the growth term in eqn (3), taking dark current zero, and the fitted values are given in Table 3. As the bias voltage increased from +0.5 V to +1 V, the rise time decreased from 2.29 s to 1.22 s under illumination. When the applied voltage is greater than the switching voltage point (V > V_sp), both second and third terms could be comparable based on the bias voltage as well as the power of the illuminating light. Therefore, the envelope of the current profile would thus be the result of these two terms in eqn (3). The shape of the curve depends on the magnitudes as well as the time constants of the forward and reverse currents. The fitted values are presented in Table 3. As seen, the contribution of the photocurrent increases as the bias voltage increases. However, the rise time of this forward current reaches a limited value of ∼ 0.5 s close to the mechanical shutter response. The reverse current also increases with the bias voltage. However, the time constant decreases from 9.23 s to 5.07 s. So, as the bias voltage increases, the time constant of this reverse current becomes comparable to the forward current and contributes to the shape of the net device current. If the intensity of light increases, at bias voltage +1 V, the photoresponse increases, as seen in Fig. 6(c). The nature of the envelope within this exposure didn’t change. This infers that the forward current still dominates over the reverse current under the application of this bias voltage. It is observed that, after V_sp, at bias voltage +4 V (Fig. 6(d)), when the power is low (1.6 mW), the forward current still dominates within the exposure time of 10 s. As power increases, the reverse current contributes and surpasses the forward current. This results in exhibiting the negative photoresponse offsetting the dark current. Hence, the observed time dynamics can very well be controlled by both bias voltage as well as the intensity of light, and the change in polarity in photoresponse can very well be achieved in this kind of GO thin film device for optoelectronic applications.

---

4. Conclusions

GO thin film with Au/GO/Au planar configuration investigated in this work showed a nonzero crossing hysteresis I–V characteristics under dual sweeping from −5 V to +5 V. This nonpinched hysteresis characteristic of the I–V curve is the signature of a memristor device. The schematic analysis of this I–V curve reveals that depending on the bias voltage, various charge dynamics, the presence of trap sites, and the Au/GO interface barrier play a pivotal role in determining the extent of this hysteresis loop. Apart from that, the pseudo-reduction of GO sp² nanoislands within the sp³ domain contributes to the charge conduction mechanism. The reproducibility of these I–V characteristics infers that the change of these nanodomains under various bias voltages is reversible. Moreover, the photogenerated charge carriers also contribute to establishing the total device current. A remarkable switching in the photoresponse was observed when the bias voltage varied from 0 V to +5 V. This photoswitching occurs through the switching voltage point (∼2 V), where the illumination does not have any effect on the net device current. The reason for this observation is that the photocurrent is additive when the bias voltage is less than the switching voltage point and subtractive when the bias voltage is greater than that. Since the reduced graphene oxide always shows additive photoresponse, the observed photoswitching in our device depends on the degree of nonlinearity of the hysteresis. Our understanding throws light on the insight of the overall device characteristics based on the nonzero, nonlinear hysteresis with voltage-dependent photoswitching. Below the switching voltage point, the time dynamics upon illumination showed exponential positive growth, adding to the bias current, and the device behaves like a capacitive memristor. On the other hand, when a bias voltage is greater than the switching voltage point, a fast growth followed by a long decay was seen in the current profile, which can be envisaged as the interplay between the two counter-interacting currents and the time profile is the result of the difference in time constants between these two currents. Since the obtained device's current profile upon exposure to light depends on both the bias voltage and intensity of illumination, this time response provides an opportunity to employ it judiciously when conceptualizing various GO-based memory devices and sensing applications.

Author contributions

Soma Saha: device fabrication, all experimental measurements, data analysis, investigation, validation, writing – original draft. Anindya Datta: supervision, writing – review & editing. Tapanendu Kundu: conceptualization, supervision, writing – review & editing.

Data availability

Data used in the manuscript are available from the corresponding author on request.

Conflicts of interest

The authors declare no competing financial interest.

Acknowledgements

This work was supported by The University Grants Commission (UGC), India. The authors would like to thank Prof. George Mathew, IIT Bombay, for providing help with field emission scanning electron microscopy (FESEM) measurements. The authors would like to acknowledge SAIF, IIT Bombay, for allowing the usage of the Raman facility. The cleanroom facilities of the Centre for Excellence in Nanoelectronics (CEN), IIT Bombay, were used for device fabrication. The authors acknowledge the Central Surface Analytical Facility (ESCA) of IIT Bombay for XPS measurements.

References

1. E. W. Lim, M. T. Ahmadi and R. Ismail, J. Comput. Electron., 2016, 15, 602–610 CrossRef CAS .
2. E. W. Lim and R. Ismail, Electronics, 2015, 4, 586–613 CrossRef CAS .
3. L. Chua, IEEE Trans. Circuit Theory, 1971, C, 507–519 Search PubMed .
4. A. Vercik and L. C. De Oliveira Vercik, Semicond. Sci. Technol., 2020, 35, 055020 CrossRef CAS .
5. D. B. Strukov, G. S. Snider, D. R. Stewart and R. S. Williams, Nature, 2008, 453, 80–83 CrossRef CAS .
6. B. Mohammad, M. A. Jaoude, V. Kumar, D. M. Al Homouz, H. A. Nahla, M. Al-Qutayri and N. Christoforou, Nanotechnol. Rev., 2016, 5, 311–329 CAS .
7. L. Chua, Appl. Phys. A: Mater. Sci. Process., 2011, 102, 765–783 CrossRef CAS .
8. L. Zhu, J. Zhou, Z. Guo and Z. Sun, J. Materiomics, 2015, 1, 285–295 CrossRef .
9. P. La Torraca, F. M. Puglisi, A. Padovani and L. Larcher, Materials, 2019, 12, 3461 CrossRef CAS PubMed .
10. F. Zhuge, B. Hu, C. He, X. Zhou, Z. Liu and R. W. Li, Carbon, 2011, 49, 3796–3802 CrossRef CAS .
11. S. Porro, E. Accornero, C. Fabrizio Pirri and C. Ricciardi, Carbon, 2015, 85, 383–396,  DOI:10.1016/j.carbon.2015.01.011 .
12. S. Mahaboob Jilani, T. D. Gamot, P. Banerji and S. Chakraborty, Carbon, 2013, 64, 187–196,  DOI:10.1016/j.carbon.2013.07.051 .
13. I. Banerjee, P. Harris, A. Salimian and A. K. Ray, IET Circuits Devices Syst., 2015, 9, 428–433 CrossRef .
14. R. Singh, R. Kumar, A. Kumar, R. Kashyap, M. Kumar and D. Kumar, AIP Conf. Proc., 2018, 1953, 050053 CrossRef .
15. A. Lerf, H. He, M. Forster and J. Klinowski, J. Phys. Chem. B, 1998, 102, 4477–4482 CrossRef CAS .
16. H. He, J. Klinowski, M. Forster and A. Lerf, Chem. Phys. Lett., 1998, 287, 53–56 CrossRef CAS .
17. J. P. Rourke, P. A. Pandey, J. J. Moore, M. Bates, I. A. Kinloch, R. J. Young and N. R. Wilson, Angew. Chem., Int. Ed., 2011, 50, 3173–3177 CrossRef CAS PubMed .
18. G. Khurana, P. Misra and R. S. Katiyar, J. Appl. Phys., 2013, 114, 124508 CrossRef .
19. F. Hui, E. Grustan-Gutierrez, S. Long, Q. Liu, A. K. Ott, A. C. Ferrari and M. Lanza, Adv. Electron. Mater., 2017, 3, 1600195 CrossRef .
20. L. H. Wang, W. Yang, Q. Q. Sun, P. Zhou, H. L. Lu, S. J. Ding and D. Wei Zhang, Appl. Phys. Lett., 2012, 100, 63509 CrossRef .
21. S. Ki Hong, J. Eun Kim, S. O. Kim and B. Jin Cho, J. Appl. Phys., 2011, 110, 44506 CrossRef .
22. C. L. He, F. Zhuge, X. F. Zhou, M. Li, G. C. Zhou, Y. W. Liu, J. Z. Wang, B. Chen, W. J. Su, Z. P. Liu, Y. H. Wu, P. Cui and R. W. Li, Appl. Phys. Lett., 2009, 95, 232101 CrossRef .
23. S. K. Hong, J. E. Kim, S. O. Kim, S. Y. Choi and B. J. Cho, IEEE Electron Device Lett., 2010, 31, 1005–1007 CAS .
24. M. K. Yadav, S. S. Kundale, S. S. Sutar, T. D. Dongale, P. Kumar and N. Panwar, J. Appl. Phys., 2023, 134, 105101 CrossRef CAS .
25. L. Valentini, M. Cardinali, E. Fortunati and J. M. Kenny, Appl. Phys. Lett., 2014, 105, 153111 CrossRef .
26. S. Ren, Z. Li, L. Tang, X. Su, H. Zhang, G. Zhang, H. Zhang, G. Cao and J. Zhao, Adv. Electron. Mater., 2020, 6, 2000151 CrossRef CAS .
27. H. Young Jeong, O. Jong Yun Kim, J. Won Kim, J. Ok Hwang, J.-E. Kim, J. Yong Lee, T. Hyun Yoon, B. Jin Cho, S. Ouk Kim, R. S. Ruoff and S.-Y. Choi, Nano Lett., 2010, 10, 48 Search PubMed .
28. H. W. Nho, J. Y. Kim, J. Wang, H. J. Shin, S. Y. Choi and T. H. Yoon, J. Synchrotron Radiat., 2014, 21, 170–176 CrossRef CAS PubMed .
29. F. J. Romero, A. Toral, A. Medina-Rull, C. L. Moraila-Martinez, D. P. Morales, A. Ohata, A. Godoy, F. G. Ruiz and N. Rodriguez, Front. Mater., 2020, 7, 1–5 CrossRef .
30. J. Van Den Hurk, S. Menzel, R. Waser and I. Valov, J. Phys. Chem. C, 2015, 119, 18678–18685 CrossRef CAS .
31. C. Wang, H. Wu, B. Gao, T. Zhang, Y. Yang and H. Qian, Microelectron. Eng., 2018, 187–188, 121–133 CAS .
32. M. Sparvoli, M. A. Gazziro, J. S. Marma and G. Zucchi, 2019 IEEE 10th Latin American Symposium on Circuits and Systems, LASCAS 2019 - Proceedings, 2019, 57–60.
33. M. Abujabal, H. Abunahla, B. Mohammad and A. Alazzam, Nanomaterials, 2022, 12, 1812 CrossRef CAS PubMed .
34. H. Q. Wei, P. Zhou, Q. Q. Sun, L. H. Wang, Y. Geng, D. W. Zhang and X. B. Wang, 2012 IEEE Nanotechnology Materials and Devices Conference, IEEE NMDC 2012, 2012, 54–57.
35. S. K. Chang-Jian, J. R. Ho, J. W. John Cheng and Y. P. Hsieh, AIP Adv., 2012, 2, 022104 CrossRef .
36. S. Saha, S. Das, A. Datta and T. Kundu, 2021 Conference on Lasers and Electro-Optics, CLEO 2021 - Proceedings, 2021, 1–2.
37. S. Saha, S. Das, A. Datta and T. Kundu, 2022 Workshop on Recent Advances in Photonics, WRAP 2022, 2022, 2, 2022–2023.
38. A. H. Jaafar and N. T. Kemp, Carbon, 2019, 153, 81–88 CrossRef CAS .
39. G. N. Panin, O. O. Kapitanova, S. W. Lee, A. N. Baranov and T. W. Kang, Jpn. J. Appl. Phys., 2011, 50, 070110 CrossRef .
40. J. Hong, M. K. Park, E. J. Lee, D. Lee, D. S. Hwang and S. Ryu, Sci. Rep., 2013, 3, 2700 CrossRef PubMed .
41. D. López-Díaz, J. A. Delgado-Notario, V. Clericò, E. Diez, M. D. Merchán and M. M. Velázquez, Coatings, 2020, 10, 524 CrossRef .
42. Y. Matsumoto, M. Koinuma, S. Y. Kim, Y. Watanabe, T. Taniguchi, K. Hatakeyama, H. Tateishi and S. Ida, ACS Appl. Mater. Interfaces, 2010, 2, 3461–3466 CrossRef CAS PubMed .
43. S. Stankovich, D. A. Dikin, R. D. Piner, K. A. Kohlhaas, A. Kleinhammes, Y. Jia, Y. Wu, S. T. Nguyen and R. S. Ruoff, Carbon, 2007, 45, 1558–1565 CrossRef CAS .
44. R. Y. N. Gengler, D. S. Badali, D. Zhang, K. Dimos, K. Spyrou, D. Gournis and R. J. D. Miller, Nat. Commun., 2013, 4, 2560 CrossRef PubMed .
45. R. Al-Gaashani, A. Najjar, Y. Zakaria, S. Mansour and M. A. Atieh, Ceram. Int., 2019, 45, 14439–14448 CrossRef CAS .
46. L. Torrisi, L. Silipigni, M. Cutroneo and A. Torrisi, Vacuum, 2020, 173, 109175 CrossRef CAS .
47. Y. Ye, J. Zhao, L. Xiao, B. Cheng, Y. Xiao and S. Lei, ACS Appl. Mater. Interfaces, 2018, 10, 19002–19009 CrossRef CAS PubMed .
48. C. Ye, Q. Peng, M. Li, J. Luo, Z. Tang, J. Pei, J. Chen, Z. Shuai, L. Jiang and Y. Song, J. Am. Chem. Soc., 2012, 134, 20053–20059 CrossRef CAS PubMed .
49. Z. Ma, C. Wu, D. U. Lee, F. Li and T. W. Kim, Org. Electron., 2016, 28, 20–24 CrossRef CAS .
50. Y. Kim, C. Lee, I. Shim, D. Wang and J. Cho, Adv. Mater., 2010, 22, 5140–5144 CrossRef CAS PubMed .
51. Y. C. Yang, F. Pan, Q. Liu, M. Liu and F. Zeng, Nano Lett., 2009, 9, 1636–1643 CrossRef CAS PubMed .
52. Y. C. Chang and Y. H. Wang, ACS Appl. Mater. Interfaces, 2014, 6, 5413–5421 CrossRef CAS PubMed .
53. J. Wang, B. Sun, G. Zhou, S. Zhu, C. Yang, C. Ke, Y. Zhao and H. Wang, J. Alloys Compd., 2023, 939, 168761 CrossRef CAS .
54. C. Chen, W. Cai, M. Long, B. Zhou, Y. Wu, D. Wu and Y. Feng, ACS Nano, 2010, 4, 6425–6432 CrossRef CAS PubMed .

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4tc02812k

This journal is © The Royal Society of Chemistry 2024