Direct laser writing of vertical junctions in graphene oxide films for broad spectral position-sensitive detectors

2.1 Device fabrication

GO nanoplatelets (XFNANO Materials Tech) with a thickness of 0.8–1.2 nm were dissolved in deionized water, and a GO dispersion with a concentration of 1.5 mg/ml was prepared after ultrasonic treatment for more than 2 h. GO films with a thickness of approximately 1 μm measured by a step profiler were acquired by drop casting on Si/SiO₂ substrates. Such films were dried in air and a drying cabinet for 1 day, respectively. Then, Au electrodes with a thickness of 80 nm were deposited by thermal evaporation through a shadow mask. The spacing between two electrodes or the length of PSDs is 100 μm. Laser reduction was carried out using a 473 nm laser (Changchun New Industries Optoelectronics Tech. Co. Ltd), whose maximum output power is 580 mW. All devices possess a strip rGO with a width of about 30 μm.

2.2 Characterization

All measurements were performed in ambient conditions. Raman spectra of GO and rGO films were obtained by a Renishaw micro-Raman spectroscopy system with a wavelength of 633 nm. Inclined cross-sectional image and C/O ratios were determined from scanning electron microscopy (SEM; Zeiss Sigma HD) with energy-dispersive spectroscopy (EDX). More thick films were used for SEM images.

Photocurrent and electrical tests were obtained using a semiconductor parameter analyzer (Agilent B1500A). Time response was collected by a digital phosphor oscilloscope (Uni-Trend Technology Ltd.). An electromagnetic shutter was used to measure the photocurrent and temporal response. All photocurrent measurements for PSDs were carried out at 0 V unless otherwise specified. Different laser sources with wavelengths of 473, 514, 633, and 1550 nm were used to measure the lateral photovoltaic effect on the motorized stage of our Renishaw Raman system under open circuit condition. Laser powers were measured by a Newport power meter 843-R.

3.1 Laser rGO

Figure 1A is the schematic view of our laser-scribed devices and a photograph of a real device. A high-power 473 nm laser was used to reduce the solution-processed GO film with a constant sweep rate. Such GO films before laser scribing usually contain plenty of hydroxyl, epoxide, carboxyl, and carbonyl groups, as shown in the brown carbon lattice. After laser reduction, part of the oxygen-containing groups was removed (see Figure 1A, black carbon lattice). Changes can be clearly observed after laser scribing from the photograph. A narrow strip of 100×30 μm² appears, which bridges the two yellow Au electrodes on GO films.

Figure 1:

Characterization for laser rGO films.

(A) Schematic view of the laser-scribed devices. Brown hexagons decorated with oxygen-containing groups denote GO, whereas the black channel between two Au electrodes represents rGO. The inset is a photograph of a PSD device. (B) EDX spectra of GO and rGO films after reduction with different laser intensities. (C) Raman spectra of GO and rGO with different laser reduction intensities. (D) Dependence of I_D/I_G (black points) and C/O ratios (blue points) on the laser intensity. The red line is the I_D/I_G and C/O ratios of GO.

To ensure the formation of rGO, Raman spectroscopy and EDX were applied to check the reduction degree of GO films. Keeping the constant sweep rate, a C/O ratio was observed to increase with laser intensity (see Figure 1B, EDX spectra). It is natural as more energy can be absorbed for higher laser intensity to reduce the GO film with a higher degree. Figure 1C demonstrates the Raman spectra from rGO films. Two typical peaks for GO [23], [28] can be distinguished, i.e. the D band peak at 1350 cm⁻¹ attributed to structure disorder or defects and the G band peak at 1600 cm⁻¹ caused by sp² hybridization of carbon bonds. A slight red shift can be observed for both peaks when compared to the spectra of GO films (blue curve). It is notable that the D peak intensity gradually exceeds that of the G peak with the increase of laser intensity, suggesting that disorder or defect density increases with the removal of more oxygen-containing groups [29]. To demonstrate it more clearly, the ratios (I_D/I_G) between the intensities of the D and G peaks from rGO films after laser reduction with different intensities are calculated (see Figure 1D, black points). I_D/I_G ratios increase with laser intensity increasing and becoming saturated if the intensity is larger than 200 W/mm². We note that the C/O ratios (blue points) also exhibit the same tendency, suggesting that the increase of C/O ratio cannot bring about more ordered structures for the film with sp² hybridization after laser reduction. The resultant sp² clusters may still be isolated by abundant sp³ bonds in rGO. Considering that the C/O ratio obtained by EDX is an average value across the films due to a relatively large penetration depth, another possibility would be that more defects are produced as stronger laser can penetrate into the films more deeper.

3.2 C/O ratio-dependent photoresponse of rGO devices

Taking into account that the reduction degree of GO or C/O ratios will significantly influence the performance of rGO-based photodetectors, which still lacks a systematical analysis, the reduction degree dependence of various electrical properties of rGO films was investigated to find an optimized C/O value for our devices, although the ratios obtained by EDX are average values across the films. Figure 2A shows the sheet resistance of rGO films with different C/O ratios, which was measured by a four-point probe method. The sheet resistance monotonously decreases with the increase of the ratio. The value at a ratio of ~2.4 can be estimated to be about 1.9 kΩ/sq, which was five orders of magnitude smaller than that of GO with a ratio of approximately 1.4. Hence, the enhancement induced by sp² hybridization dominates the conductivity increase of rGO films.

Figure 2:

Impact of C/O ratio on the electrical and optoelectronic performance of rGO films.

(A) Sheet resistance as a function of C/O ratio. (B) Photoresponse of rGO devices with different C/O ratios. (C) Normalized photocurrent with error bars from rGO devices (blue points) with different C/O ratios. Negligible photocurrent can be obtained from GO devices (red point). (D) Rise (red) and decay time (blue) of rGO detectors with different C/O ratios.

Photoresponse from rGO films was also investigated, which shows a quite different behavior from that of dark conductivity, as shown in Figure 2B. One side of the rGO device under a bias of +1 V was illuminated by the 473 nm laser (~0.05 mW) with a spot size of ~20 μm, and the other Au electrode was grounded. To avoid further reduction of the devices, low laser powers for photoresponse measurements are employed. The photocurrent of rGO devices, in which dark current has been extracted, does not show monotonous changes with the increase of C/O ratios. Maximum photoresponse can be obtained from the device with an average C/O ratio of about 2.05. Such behavior can be reproduced in more than five samples, as shown in Figure 2C in which the photocurrent was normalized to the highest value for comparison. As expected, GO devices show a negligible photoresponse (red point) due to poor light absorption and charge mobility. With the increase of C/O ratios, larger photocurrent can be generated together with the decrease of work function usually about 4.35–5.6 eV and a narrower bandgap usually about 0.5–2.2 eV [30], [31], [32]. This could be ascribed to more efficient light absorption, which induces larger photoresponse. However, the photocurrent becomes smaller after reaching a threshold average C/O value of about 2.05. Considering that more defects are generated with the increase of C/O ratios, charge trapping will become more significant, inducing less free charges. Therefore, there would be a trade-off between more defects induced by oxygen removal and increased light absorption [32]. We note that the devices with a laser power (473 nm) of <2 mW do not exhibit degradation under normal operating conditions (see Supporting Information Figure S1), indicating that photochemical change or photoinduced thermal reduction in the devices is not significant. In addition, the dark current of the devices also keeps relatively stable in air even after 30 days, suggesting that oxygen- or moisture-induced changes in rGO films without sealing are negligible.

Rise (10–90%; red points) and decay time (90–10%; blue points) [33] of the devices with different C/O ratios were also investigated by an oscilloscope, as shown in Figure 2D. A similar trend with that of photocurrent was obtained. However, the device with an average ratio of 2.25 exhibits the longest response time with the rise time of 0.8 ms and decay time of 0.4 ms, respectively, which are the fastest values to date for pure rGO-based photodetectors [20], [34], [35], [36]. The longer response time with the increase of C/O ratio (<2.25) matches well with the fact that the trapping/detrapping effect due to more defects or lattice disorder is significant. However, decreased response time is observed at ratios of >2.25. These could be ascribed to the fact that the upper layers are closer to ideal graphene layers with decreased defects due to the annealing effect with a stronger laser, although more defects are still produced in the bottom layers, owing to a deeper laser penetration.

3.3 Position-dependent photoresponse of rGO devices

The photoresponse at other positions of rGO films with different C/O ratios was also collected under a bias of 1 V at 473 nm (200 μW), as shown in Figure 3, in which the zero point denotes the central point of the devices. It is notable that the rGO device with the average C/O ratio of 2.05 still demonstrates the largest photocurrent with the laser spot at other positions away from the central point (see Figure 3A). The photocurrent clearly exhibits a dependence on the position of laser spots. We note that the positions close to the two electrodes demonstrate the largest photocurrent with opposite polarities. However, the laser spot at the central point produces the smallest photocurrent. Figure 3B and C shows the wavelength and power dependences of the photoresponse at different positions, respectively. For comparison, responsivity R is calculated for each wavelength based on the relation R=I_ph/P_in, where I_ph is the photocurrent and P_in is the incident light power. A broad spectral response was observed for rGO devices, which covers the IR wavelength of 1550 nm in spite of a smaller responsivity (see Figure 3B). All photoresponses with different laser wavelengths exhibit the similar trend as well as those in Figure 3C with different laser powers (633 nm). We note that the photocurrent increases with the increase of laser powers at the regions away from the central point. Therefore, the position-dependent photoresponse does not show dependence on a certain laser power and wavelength. In addition, the relationship between light power and the photocurrent obeys a power law and exhibits quasi-linearity (see Supporting Information Figure S2).

Figure 3:

Photoresponse of rGO films at different positions of laser spots with (A) different C/O ratios (473 nm laser), (B) different laser wavelengths, and (C) different laser powers at 633 nm.

3.4 PSDs

Considering that laser scribing may induce the inhomogeneous reduction of GO across a thick film due to the optical filter effect of upper layers or a finite penetration length of laser beams, gradient rGO/GO junctions could be produced in which vertical built-in electric fields can be generated for charge separation and LPV. Position-dependent photovoltage was demonstrated for our rGO devices with the optimized C/O ratio under open circuit condition without a bias voltage. Figure 4A shows the LPVs of an rGO device at different positions x. For the positions below the Au electrodes (colored areas), LPV decreases with the increase of the distance. However, the self-powered device exhibits a linear relationship between the LPV and the position of laser spot (633 nm, 670 μW) in the channel, which can be explained by the carrier diffusion theory [18], [37]. For −L<x<L, LPV can be expressed as follows:

Figure 4:

Lateral photovoltaic effect for the laser-scribed rGO devices: (A) LPV at different positions of laser spots (633 nm) and (B) LPV between the two Au electrodes with different laser powers.

The inset is the sensitivity at different laser powers.

where K is the proportionality coefficient and N₀ is the density of excess charges at x. L and l denote the half-distance of the gap between the electrodes and the charge diffusion length in rGO, respectively. If x<<l, Eq. (1) can be simplified as

where slope a is usually defined as the sensitivity for PSDs, representing the displacement resolution of PSDs. Therefore, the linear relationship can be ascribed to a large l when compared to x in the devices. Based on Equations (1) and (2) with the data in Figure 4A, l can be estimated to be about 62.2 μm, which is much larger than x and naturally induces the linearity. In addition, the photocurrent from the device without applying a bias voltage also exhibits linearity (see Supporting Information Figure S3), suggesting that the nonlinearity in Figure 3 can be ascribed to the applied bias voltage.

It is notable that LPV-distance curves with different laser powers also keep the linearity, as shown in Figure 4B, in spite of different slope values (i.e. the sensitivity). The sensitivity increases with the increase of the laser powers linearly (see inset), ranging from 0.33 to 7.27 μV/μm with the laser power increasing from 33 to 670 μW. Therefore, higher sensitivity can be expected if a larger laser power is used. It is natural as a larger N₀ can be produced for higher laser powers, leading to a larger sensitivity. Another parameter used to evaluate the signal distortion of PSDs, i.e. nonlinearity, was estimated based on the following equation [3], [38]:

The maximal nonlinearity in Figure 4B can be calculated to be 5.4%, which is far less than the acceptable criteria of 15% for PSDs. In addition, rise and decay time for the self-powered PSDs were also estimated to be about 1.6 and 1.9 ms, respectively (see Supporting Information Figure S4A). It is notable that the response speed of our PSDs does not show significant dependence of the position of laser spots (see Supporting Information Figure S4B). To our best knowledge, these values are still the fastest response among pure rGO-based devices even with bias voltages. This could be ascribed to a smaller amount of defects in the upper layers of our rGO films and a short carrier transit distance, considering the longer response time of about 10 s in chemical rGO films with a similar channel distance [34].

3.5 Mechanisms for laser-scribed PSDs

Inclined cross-sectional SEM image with EDX spectra of a laser rGO film was collected to verify the existence of possible junctions (see Figure 5A). As we expected, the C/O ratios of rGO films gradually decrease from the film surface (point 1) with a value of 2.4 to the bottom (point 3) with a value of 1.54. The value at the middle point 2 is about 2.16. Hence, a gradient junction across rGO films should exist (see the energy alignment of the films in Figure 5B). Taking into account that rGO films are usually p-type semiconductors and the work function will gradually decrease with the increase of C/O ratios and a narrower bandgap, the gradient C/O ratio in rGO films will induce built-in electric fields from the upper layers toward the bottom layers, as shown in Figure 5 (red arrows). Upon light illumination, photogenerated excitons can be separated by the fields with electrons in the upper layers and holes in the bottom layers. The excess electrons in the upper rGO films can laterally diffuse to both sides of the devices, producing a lateral potential difference between illuminated and nonilluminated areas with an electron density gradient. The difference of collected electron quantity by the two electrodes thus induces the LPV.

Figure 5:

Gradient C/O ratios in the laser-scribed rGO films.

(A) Inclined cross-sectional SEM image of a thick rGO film. The scale bar is 1 μm, The ratios at points 1–3 can be determined by EDX. (B) Scheme of the energy alignment across the rGO film, in which E_c and E_v are conduction and valence bands, respectively. E_f denotes the Fermi level. Red arrows denote the built-in electric field in the junctions.

To further confirm the role of laser-scribed junctions for PSDs, thermal rGO devices with the same device structure were investigated for comparison (see Supporting Information Figure S5). It was noted that the thermally reduced films nearly exhibit a uniform reduction degree. The same laser used in Figure 4A with a light power of 670 μW was employed to study their LPV. We note that LPV from rGO films does not exhibit significant C/O ratio dependence. The signals are very weak and >16 times smaller than those from the laser-scribed devices. In addition, weak LPV does not show good linearity. These clearly indicate that LPV is mainly caused by the laser-scribed junctions as well as charge diffusion in rGO-based PSDs.