Paper based electronics have gained popularity to meet the requirements of the next generation of smart instruments, especially disposable electronic devices¹. Paper is commonly used in daily life, it is manufactured at very large scale and at low-cost and there are widespread policies and industrial processes for recycling it^2-4. The price of common paper is ~0.1 € m⁻², far less than that of crystalline silicon wafers and plastic substrates, thus making paper an appealing candidate as an alternative and low-cost substrate for the construction of functional electronics⁵. Furthermore, unlike flexible polymer substrates such as polyimide (PI), polyethylene terephthalate (PET), and polydimethylsiloxane (PDMS)^6-9, the paper substrate is biodegradable^{10, 11}. The recyclable and biodegradable features of paper substrates provide a great potential to alleviate the electronic waste issue. Benefiting from these superior characteristics, various novel paper-based electronic devices such as sensors^{12, 13}, transistors^{14, 15}, triboelectric nanogenerators (TENG)^{16, 17}, solar cells^{18, 19}, supercapacitors^{20, 21}, and photodetectors, have been developed.

As one of the representative paper electronics, paper-based photodetectors are capable of converting optical pulses into electrical signals effectively, therefore could be used in many fields spanning optical communication, imaging, biological detection, and environmental monitoring^22-27. Layered transition metal dichalcogenides (TMDCs) (e.g., WS₂ and MoS₂ ), hold high carrier mobility, a sizeable band gap of 1–2 eV, and strong light-matter interaction^{28, 29}, which make these materials very prospective for optoelectronic devices³⁰. The weak van der Waals (vdW) bonds among the adjacent layers allow bulk TMDCs crystals to be effectively exfoliated, enabling direct deposition of such layered crystals by direct abrasion against the substrate^{5, 31-36}. In addition, the high surface roughness and porous nature of cellulose fibers not only aid the adhesion of the deposited material but also provides a larger photoactive area for optoelectronic devices as compared to conventional planar substrates³⁷.

Standard printer paper is typically made up of interconnected cellulose fibers with diameters of 20−40 μm and lengths up to 2−5 mm³⁸. This structure, very different from that of conventional micro-electronic substrates, makes it necessary to develop specialized manufacturing processes to integrate electronic materials onto paper substrates. Over the past decades, the majority of efforts to integrate vdW materials onto paper substrates have been focused on printing dispersions of liquid-phase exfoliated (LPE) nanosheets to form networks/ films^{39, 40}. However, high-temperature thermal treatment (which can damage the paper substrates or the vdW films) is needed to evaporate the solvents used during liquid phase exfoliation^{41, 42}. On the other hand, the remaining residual surfactant molecules hamper the intimate contact between flakes of vdW materials, thus reducing the performance of the device.⁴³ In general, the responsivity of the photodetectors based on LPE nanosheets is in the order of 10−1000 μA W⁻¹, which limits their practical application^{44, 45}. Hence, it remains a challenge to establish an efficient deposition on paper for the mass production of thin films with high purity³¹. Our recent work has illustrated an improved all-dry abrasion method to deposit a wide variety of vdW materials on standard (untreated) office paper⁴⁶, making it possible to fabricate optoelectronic devices without the utilization of solvents.

This study aimed to integrate high-performance semiconductor photodetection devices on common paper substrate through the direct abrasion of photoactive WS₂ crystals. The WS₂ devices fabricated on paper exhibit a remarkable photoresponse behavior with responsivity values in the order of ~10 mA W⁻¹ at a bias voltage of 10 V over a broad spectral range from ultraviolet (365 nm) to near-infrared (940 nm). As the applied bias voltage is increased, the responsivity is dramatically enhanced and reaches a maximum value of ~270 mA W⁻¹ at a bias voltage of 35 V. Moreover, we demonstrate that the atmospheric oxygen molecules have a negative impact on the electrical conductivity and photoresponse performance of paper-based WS₂ photodetectors. In fact, the performance of the fabricated paper-supported WS₂ photodetectors is optimized when operated in vacuum and we thus propose to explore encapsulation techniques in future works. In addition, a WS₂ photodetector with the narrow channel distance is fabricated on paper using interdigitated Au electrodes, achieving a responsivity of ~200 mA W⁻¹ at 5 V bias. Finally, we construct an optical spectrometer using a paper-based WS₂ device to demonstrate its potential application in sensitive photodetection components.

Common office copy paper (Winner Paper Co Ltd, 80 g m⁻²) was used as a low-cost substrate without any pre-treatment. Micronized WS₂ (tungsten disulfide) powder (0.6 micron APS) was purchased from Hagen Automation Ltd to serve as the photosensitive material. Graphite pencil (4B grade, ~80% graphite content) commercially available (Faber Castell) was utilized to deposit conductive electrodes for interfacing with readout electronics.

The paper-based WS₂ devices with graphite electrodes were fabricated as described in detail in Fig. 1(a). Using this all-dry abrasion method, the WS₂ film, with a thickness of ~20±5 μm⁴⁶ and a width of 2 mm, was prepared as the photosensitive channel. After drawing of the graphite electrodes, the length of the WS₂ channel could be reduced to ~250−500 μm using the assistance of a glass slide. For the paper-based WS₂ devices with interdigitated Au electrodes, the Au layer (100 nm) was thermally evaporated on top of the large-area deposited WS₂ film through a patterned shadow mask (Ossila, E323) with a channel length (the space between two fingers) of 45 μm.

The prepared WS₂ devices were placed in a homebuilt probe station with adjustable pressure and temperature. Details about the vacuum probe station system can be found in ref.⁴⁷. High-power fiber-coupled LEDs (Thorlabs) with different wavelengths were employed as the light source⁴⁸ to produce a light spot with a diameter of 3 mm by connecting the free-end of the multimode fiber (Thorlabs, M28L05) to a collimator. The power intensity of the output light can be regulated by a computed programmable DC power supply (Tenma 72-2710) and determined using a power meter (Thorlabs, PM100D). A Keithley-2450 source measure unit was used to measure the current vs. voltage characteristics and the temporal current response.

The schematic diagram of the fabrication process of paper-based WS₂ photodetectors is shown in Fig. 1(a). The first step in device preparation is to print out the outlines of the channel and electrode on a standard copy paper with a commercial laser printer. The interior outline is then enclosed by attaching conventional masking tape to shape a rectangular mask. A continuous film is deposited onto the unmasked paper surface by simply abrading the fresh WS₂ fine crystals with a cotton swab. After the masking tape is removed, the electrode region within the exterior outline is filled by drawing with a graphite pencil of 4B grade, leaving a narrow channel for exposing the photoactive WS₂ to light illumination. Previously reported transfer length measurements showed that graphite electrodes yield low contact resistances to abraded WS₂ channels, in fact negligible in comparison to the channel resistance⁴⁶. We address the reader to a video that shows the overall fabrication process (see Video S1 in Supplementary information). A paper-supported 3 × 3 photodetector array, as seen from the photograph in Fig. 1(b), is easily prepared through this all-dry abrasion fabrication route. This fabrication route is advantageous in avoiding the use of solvents and the high-temperature heating process. Figure 1(b) also illustrates the flexibility of paper substrates. We have recently studied similar devices to those shown in Fig. 1(b) under uniaxial tension and compression (ref.³⁴) finding that the resistance of the semiconductor channel strongly depends on the applied strain: increasing with tensile strain and decreasing with compressive strain. We attribute this strain-dependent resistance to a simple phenomenon: tensile strain yields reduced overlapping area between adjacent flakes while compressive strain would increase the overlapping area. This change in flake-to-flake overlap would lead to a substantial change in the resistance of the channel. But more importantly we found that the strain-induced changes in the electrical properties of the device were reproducible and reversible, most likely due to the easy flake-to-flake sliding for van der Waals materials. This illustrates the advantageous use of paper substrate in combination with van der Waals materials for the fabrication of flexible devices.

Figure 1(c) displays an optical micrograph of one of the prepared devices, showing the compact and continuous nature of the graphite and WS₂ films. This could be attributed to the fact that the friction force generated in the process of abrading and drawing is capable of cleaving the layered vdW crystals by breaking the weak interlayer vdW interactions, thereby forming a homogeneous coating with interconnected platelets. Interestingly, the WS₂ photoactive channel can be made as narrow as ~300 μm despite utilizing this manual “abrading/drawing” method. A recent combined scanning electron microscopy and atomic force microscopy study of abrasion-induced deposited WS₂ on polycarbonate showed that the abrasion forces experienced during the rubbing process lead to a reduction of the thickness of the WS₂ deposited flakes, from ~100−200 nm in the as-received powder to ~20−80 nm for the abraded WS₂ films³⁵. We refer readers to our previous work for further characterizations on morphology and structure⁴⁶.

The photodetection performance of the paper-based WS₂ device was thoroughly investigated by measuring their electrical characteristics under dark and upon light illumination. We found that the exposure to atmospheric conditions can hamper the electrical and optoelectronic performance of the fabricated photodetectors and thus, in order to fully optimize their performance, we have studied them under high vacuum conditions. We address the reader to the Supplementary information Fig. S1 where the current is recorded during the pumping down finding a large increase of the dark current (×5) while reducing the pressure. This suggests that the ambient atmosphere may affect the electrical properties of WS₂ devices. More specifically, we found that the presence of atmospheric oxygen is the main cause of this degradation of the WS₂ performance upon air exposure (see Supplementary information Fig. S2).

To further explore the impact of ambient atmosphere on the photoresponse performance, the paper-based WS₂ device was measured under vacuum and air conditions, respectively. We subjected the WS₂ device to a periodic ON/OFF switching of light illumination at a fixed incident power density of 35 mW cm⁻² over 100 min. As shown in Fig. 2(a) and 2(b), the current increases when the device is illuminated and then returns to the initial dark current value once the illumination is shut off. Moreover, the traces of current recorded in both vacuum and air conditions exhibit reproducible switching under 150 repeated illumination ON/OFF cycles. The significant difference is that the current trace in the vacuum condition is more stable showing only an overall smooth drift while the trace in air shows jumps. From the zoomed-in profile in Fig. 2(b), one can also observe the increment of current with lower signal-to-noise for the WS₂ device tested in vacuum as compared to that in air. These results demonstrate that the WS₂ device has better photoresponse performance in vacuum.

Response time is another essential parameter for assessing the performance of photodetection devices. The rise time at light excitation denotes the time needed for the current to climb from 10% to 90% of the photocurrent whereas the fall time in light outage represents the time taken for the current to drop from 90% to 10% of the photocurrent⁴⁹. For the paper-based WS₂ photodetector (device A), the rise and fall time are calculated as 7.31 s and 6.58 s in vacuum, whereas in air they are 1.36 s and 1.54 s respectively.

We further determined the dependence of the photocurrent on the incident power to get a deeper insight into the photocurrent generation mechanism of the paper-based WS₂ photodetector (device A) in vacuum and air conditions. The photocurrent can be calculated by subtracting the dark current from the current under illumination condition. Fig. 2(c) plots the photocurrent of the WS₂ device as a function of time while the illumination is switched ON/OFF at increasingly high incident power intensity from 1.1 mW cm⁻² to 35 mW cm⁻². Higher photocurrent is naturally obtained at higher illumination power intensities (both in air and in vacuum conditions) since higher power intensity provides a larger number of photons, enabling the formation of a larger number of electron-hole pairs⁵⁰. Interestingly, the relationship between the photocurrent and the illumination power strongly differs when the device is measured in air and in vacuum. The relationship between the photocurrent and the incident power intensity (Fig. 2(d)) can be fitted by a power-law equation of I_ph ∝ P^α, where the exponent α specifies the photocurrent response to the incident power^51-53. While the device tested in air show an exponent α_Air = 0.69, the device in vacuum presents an exponent value of α_Vacuum = 0.93. An ideal exponent of α = 1 is expected for photoconductive photodetectors where the number of photogenerated carriers is trivially proportional to the number of incident photons. On the other hand, the value of α less than 1 is typically observed in photodetectors, where photogating or strong carrier-dependent recombination processes play a significant role in the photocurrent generation mechanism⁵⁴. The α_Vacuum value which is very close to α = 1 points out the nearly linear dependency of the photocurrent on the incident power intensity, confirming the photoconductive nature of the all-dry deposited WS₂ photodetectors operated in vacuum.

The effect of the bias voltage on the photocurrent was further investigated by applying different bias voltages to the paper-based WS₂ photodetector (device B) for the same power-dependent measurements as mentioned above. Figure 3(a) and Fig. S3 show photocurrent vs. time at different bias voltages. The device shows a distinct power-dependent photoresponse behavior when applying the bias voltages ranging from 1 V to 35 V, which is also verified with IV curves in Fig. S4. The photocurrents as a function of incident power under various bias voltages are plotted in Fig. 3(b). The photocurrent maintains generally a near-linear relationship with the incident power even when subjected to a high voltage of 35 V. More importantly, one can observe the remarkable enhancement in photocurrent at higher voltages, which could be attributed to the increase of carrier drift velocity with stronger electrical field generated in the channel. The transit time F(θ)=∑n=0NAnej[(n−1)2πλdsinθ−ϕn],, where μ is the carrier mobility, V is the bias voltage, and l is the distance between the source and drain electrodes⁵⁵. The stronger electric field favors the separation of photoexcited electron−hole pairs and shortens the transit time, thus accelerating the charge accumulation at electrodes and increasing the photocurrent⁵⁶.

The responsivity (R) is a typical detector parameter used to compare the performance of different photodetection devices, which can be calculated as^{5, 57}:

1H(f)=ΓK[∑k=0K−1ake−jφke−j2πfkτ],

where I_ph is the generated photocurrent, P is the incident power density and S_device is the illuminated active channel area of the device. As plotted in Fig. 3(c), the calculated responsivity is nearly independent of the incident power (as expected as α_Vacuum ~1). We quantitatively compared the photocurrent and responsivity at a selected power intensity of 35 mW cm⁻² collected at various bias voltages (Fig. 3(d)). The photocurrent (responsivity) of the paper-based WS₂ photodetector (device B) is 0.26 μA (1.2 mA W⁻¹) at 1 V and dramatically raises as the applied voltage increases, reaching a maximum of 56.5 μA (268.7 mA W⁻¹) at 35 V. On the other hand, we constructed the WS₂ device with a narrower channel length of 45 μm on paper by thermally evaporating the interdigitated Au electrodes (100 nm in thickness) and carried out identical power-dependent measurements (Fig. S5). At a bias voltage of 5 V, the WS₂ device with Au electrodes delivers a significantly enhanced responsivity of 193.9 mA W⁻¹ under the same power intensity (35 mW cm⁻²) compared with the WS₂ device with graphite electrodes (6.4 mA W⁻¹). These responsivity values measured are superior to that of other TMDCs-based photodetectors (0.018−124 mA W⁻¹) fabricated by solvent-involved approaches^{37, 56, 58-60}, and that of the WS₂ photodetectors with atomically thin layers (5×10⁻³−12.5 mA W⁻¹) obtained through CVD^61-63, magnetron sputtering^{64, 65}, and mechanical exfoliation methods⁶⁶. Table 1 compares their typical device characteristics in further detail.

In order to illustrate the reproducibility of the devices fabricated by this abrasion method we measured 10 WS₂ devices with graphite electrodes at 10 V of bias voltage. The statistical results are summarized in Fig. S6. Under the illumination with a power intensity of 35 mW cm⁻², the photocurrent of the as-prepared WS₂ devices shows a median value of 2.6 μA and a low deviation with 50% of the devices scattering less than 1 μA. Accordingly, the responsivity of the devices presents a median of 12.7 mA W⁻¹ and even reaches a maximum of ~30 mA W⁻¹. The device-to-device variation for the paper-based WS₂ photodetectors could be attributed to the percolative character of the WS₂ film since the conduction paths of charge carriers are randomly distributed over the percolative network of interconnected WS₂ flakes.

We also studied the role of temperature in the performance of the WS₂ photodetectors. As indicated in Fig. S7, the photocurrent and responsivity both increase upon temperature increase, yielding slopes of 7.2 nA °C⁻¹ and 34.3 μA W⁻¹ °C⁻¹, respectively. In general, the conductivity of 2D materials is increased with higher temperature. Structural defects such as grain boundaries would also lead to large density of localized gap state. It is very likely that the mobility of WS₂ increased during heating up at high temperature⁷⁰. Both Mott variable range hopping, and an Arrhenius-type activated transport could contribute to the conductivity of WS₂⁷¹. A more thorough characterization on the temperature effects of WS₂ photodetectors is beyond the scope of this paper.

The spectral responsiveness of the paper-based WS₂ photodetection device was evaluated at a fixed power intensity of 13 mW cm⁻² using 17 high-power fiber-coupled LED sources with different wavelengths. Figure 4(a) summarizes the time-resolved photocurrent of the WS₂ photodetector (device B) in response to light stimulation with various wavelengths ranging from 365 nm (ultraviolet) to 940 nm (near-infrared). The fabricated paper-based WS₂ photodetector exhibits a broad spectral response, which could be attributed to the photon absorption capability of active WS₂ ultrathin sheets over a vast spectrum range⁷². The extracted responsivity values over different spectra ranges are shown in Fig. 4(b). The WS₂ photodetector delivers a maximum responsivity value of 14.4 mA W⁻¹ at 660 nm and reduces slightly to 11.2 mA W⁻¹ when the illumination wavelength exceeds 850 nm. This spectral response above 900 nm could be due to thermally activated indirect band gap absorption as multilayer WS₂ indirect gap is 1.32 eV⁷³. Another plausible explanation for the broadband spectral response and the lack of strong excitonic features in the photocurrent spectrum is the presence of strong bolometric effect, i.e. change in the channel resistance due to the increase of temperature induced by light absorption, in the photocurrent generation. The linear power dependence and the slower response time of the devices in vacuum further support this scenario where light would be absorbed by the graphite electrodes, even for photons with energies lower than the WS₂ bandgap, and because of the low thermal conductivity of paper substrate the WS₂ film would start to increase its temperature that would lead to a decrease of its resistance³² and thus an increase of the current flowing through the channel.

The broad-band photodetection properties of the paper-supported WS₂ photodetector, along with the linear dependence of the photocurrent upon incident power, motivated us to explore its application in optical spectrometers. We constructed a proof-of-concept spectrometer as shown in Fig. 5(a), containing a light source, light-scattering optical element, and detection element. In this spectrometer system, a diffractive grating scatters the incoming light beam passing through a collimator into multiple wavelengths, and thereafter the light with a specific wavelength is collected with the other collimator by changing the angle of the diffractive grating. The fabricated paper-supported WS₂ device is then used to detect this selected outgoing light. Fig. 5(b) and 5(c) show the comparison of the spectra measured using a commercial silicon photodiode (Thorlabs, PM100D) and with our paper-supported WS₂ photodetector (device C). We have used a supercontinuum laser and different spectral filters (long-pass or band-pass) to generate the test spectra to be measured with our proof-of-concept spectrometer. The remarkable agreement between the spectra measured with the commercial photodetector and our paper-supported WS₂ device suggest that our WS₂ device performs well in the quantitative detection of spectra.

In summary, we developed a paper-based WS₂ photodetector through a facile all-dry deposition strategy that involved the abrasion of the WS₂ photoactive channel and pencil drawing of graphite electrodes. We demonstrated that the weak van der Waals interactions between WS₂ layers allow the friction force produced during the abrasion process to cleave the layered WS₂ crystals, resulting in the formation of a uniform WS₂ film with interlinked platelets. The high purity of such an all-dry deposited WS₂ film leads to highly effective electron-hole separation and hampers the recombination. Therefore, the fabricated WS₂ photodetector shows high photoresponsivity to the incident light in a broad spectral range from ultraviolet to near-infrared. We investigated the photodetection performance of the paper-based WS₂ device in vacuum and air conditions, confirming the significant effect of oxygen molecules decreasing the photoresponse. In addition, the responsivity can be markedly improved by applying the higher biasing voltage, achieving a maximum of ~270 mA W⁻¹ at a voltage of 35 V. Finally, we demonstrated the potential application of the paper-based WS₂ photodetector on spectrometers. This work paves the way for large-scale fabrication of other cost-effective electronic/optoelectronic devices.