Self-powered UV-visible photodetector with fast response and high photosensitivity employing an Fe:TiO 2 /n-Si heterojunction †

An ultrasensitive, fast response and self-powered photodetector would be preferable in practical applications. In this paper, we report fast response, high photosensitivity and self-powered Fe doped TiO 2 (Fe:TiO 2 )/n-Si UV-visible detectors via a facile solution process. As-fabricated devices exhibited excellent photoresponse properties, including high responsivity of 46 mA W (cid:1) 1 (350 nm) and 60 mA W (cid:1) 1 (600 nm) with 0.5 mW cm (cid:1) 2 light irradiation under zero bias, as well as an ultrasensitive (on/o ﬀ ratio up to 10 3 ), fast (rise/decay time of <10/15 ms), and broad-band (UV-visible) photodetection with no or low external energy supply. Furthermore, the quantum e ﬃ ciency of the heterojunction rose up beyond 100% with a broad wavelength range at a small reverse bias of (cid:1) 0.5 V. The self-power originated from the existence of a built-in electric ﬁ eld between Fe:TiO 2 and Si that helps facilitate the separation of photogenerated electron – hole pairs and regulate the electron transport. Capacitance – voltage ( C – V ) measurements of Fe:TiO 2 /n-Si devices also con ﬁ rmed the existence of the built-in electric ﬁ eld. Such Fe:TiO 2 /n-Si heterojunction photodetectors might be potentially useful for relative applications with weak-signal fast detection in the UV-visible band.


Introduction
In recent years, functional nanosystems must not only be sensitive, controllable and multifunctional, but must also have a low power consumption. Self-driven or self-sufficient nanodevices and nanosystems have recently attracted a lot of attention due to their various superior performance advantages. 1,2 Among them, many reports about self-powered photodetectors (PDs) have proved that they can operate independently and sustainably. [3][4][5][6][7][8][9][10][11] This not only enhances the adaptability of the device, but also greatly reduces the size and cost of the system. In practical applications including imaging techniques, lightwave communication, and environmental sensing, ultrasensitive and fast response self-powered PDs would be more urgently needed. However, small photocurrent response, slow temporal recovery, narrow range of detectable light wavelengths and high detectable light intensities are the limitations for self-powered PDs. In addition, photoelectrochemical (PEC) self-powered systems have the problems of electrolyte toxicity and volatilization, and most of the PEC PDs only respond to the ultraviolet band region. 5,6 Some routes introducing plasmonic materials, 7 ferroelectric materials, 8 monolayer graphene lm as transparent electrode, 9 high-quality single crystals obtained via molecular beam epitaxy 10 have been developed to further improve device performance. However, additional complexity of the processing and the high cost limit their practical applications. Therefore, high performance and reliable self-powered PDs fabricated by a facile and low cost way is necessary.
Beneting from the mature silicon (Si)-based process in the microelectronics industry, semiconductor/Si vertical heterojunction PDs, with a planar device structure, have attractive applications in the optoelectronic integrated circuits. [12][13][14][15][16][17] The construction of semiconductor nanostructure materials/Si heterojunction can not only compensate for the limited spectral bandwidth of Si due to the extensive absorption spectra of different semiconductor materials (especially for UV applications), but also promote the separation and transport of photoinduced carriers. Among them, wide-bandgap semiconductor materials ZnO and TiO 2 has been widely investigated. In contrast, the chemical stability of TiO 2 is more stable than ZnO. And it has been proved that the optical sensitivity and activity of TiO 2 can be improved by ion doping. Ion doping of TiO 2 either introduce impurity energy level(s) or adding energy states to band edge, as a consequence, spectral response range can be broadened. [18][19][20][21][22][23] In Fe-doped TiO 2 (Fe:TiO 2 ), Fe 3+ ions tend to substitute for Ti 4+ ions and take their lattice sites. Fe 3+ cations can act as shallow traps to decrease of the recombination rate (see Fig. S4, ESI †). 17 Some good self-powered PDs based on TiO 2 have been reported, such as TiO 2 -SnO 2 , 6 TiO 2 -PANI, 24 TiO 2 -CH 3 NH 3 PbI 3 (ref. 25) and so on. But as far as we know, there are few reports on photodetection with doping of TiO 2 . Herein, we designed a novel self-powered UV-visible PD employing heterojunction between the Fe:TiO 2 and Si. Such Fe:TiO 2 /n-Si heterojunction fabricated by a facile and low cost way, with high performance and reliable PDs, it can be used for broad-band (UV-visible) photodetection in a self-powered mode operation. The built-in electric eld formed between the interfaces creates the driving force for the fast and efficient separation of the photogenerated carriers. A fast response speed of 10/15 ms were achieved in the fabricated PDs. Notably, under zero bias, the device exhibited high responsivity of 46 mA W À1 (350 nm) and 60 mA W À1 (600 nm) with a 0.5 mW cm À2 light irradiation. At a small reverse bias of À0.5 V, the quantum efficiency of the device rose up beyond 100% with a UV-visible wavelength range. The novel device design described in this paper may point out a promising new direction for manufacturing advanced selfpowered PDs.

Materials synthesis and characterization
3 mL of tetrabutyl titanate (C 16 H 36 O 4 Ti) solution was stirred with 10 mL of ethanol solution overnight, and 0.02 M anhydrous ferric chloride (FeCl 3 ) ethanol solution was stirred overnight. Two sols were mixed with magnetic stirring until homogeneous. The above process avoided water involvement. Light yellow sol was spin-coated at n-Si (111) (8-13 U cm) wafer with 1000 rpm for 10 s. Subsequently annealed under an air atmosphere in muffle furnace at 400 C for 1 hour and spontaneously cooled to room temperature. Namely, Fe:TiO 2 /n-Si heterojunction nanostructure samples were obtained.
Morphology of Si/TiO 2 heterojunction were determined by a eld emission scanning electron microscope (FESEM, Hitachi, S-4800). Powder X-ray diffraction (XRD) of the samples was conducted by a D/max-2550VB+/PC X-ray diffractometer (Rigaku, Japan) equipped with a rotating anode and a Cu Ka radiation source (l ¼ 1.54056Å). Raman spectra were recorded at room temperature on an inVia-Reex micro-Raman spectroscopy system from Renishaw with 532 nm radiation. Room temperature UV-vis absorption spectroscopy was conducted on a Shimadzu UV3600 UV-vis-NIR spectrophotometer. XPS data were obtained using an Escalab 250Xi (Thermo Scientic) spectrometer with an excitation source of Al-Ka radiation. Photoluminescence (PL) measurements were carried out at room temperature using uorescence spectroscopy (PTI Quan-taMaster) with a xenon lamp as the excitation source (l ex ¼ 370 nm).

Photodetector fabrication and measurement
The top electrode is Ag-planted digitated nger shape with total surface area of 25 mm 2 . The bottom of the Si wafer was coated with an Ag layer as a bottom electrode. To illustrate the reliability of the manufacturing process, sample A, B, C and D were prepared under the same conditions, and the following test data was based on sample A if there are no special instructions. The current-voltage (I-V) curves and capacitance-voltage (C-V) characteristic of the device were measured by using the Keithley-4200 semiconductor characterization system (SCS) on a probe station at room temperature. Monochromatic light with wavelength ranging from 350 to 600 nm was achieved from a 300 W Xenon lamp with monochromators.

Results and discussion
Characterization SEM image showing the morphology and structure of asfabricated heterojunctions consisting of TiO 2 and Fe:TiO 2 thin lm/Si wafer via a solution process of spin coating ( Fig. 1a and b). As can be seen from the cross-section of heterojunction, the nanoscale particles formed a dense lm with a thickness of about 200 nm. The pure and Fe-doped TiO 2 were characterized by XRD. As shown in Fig. 1c, all XRD peaks are indexed to the TiO 2 anatase phase (JCPDS card no. . No feature of Fe 2 O 3 was detected for Fe-doped sample and all peaks were similar for both doped and undoped samples. It indicated that the TiO 2 crystalline structure did not noticeably inuenced by low-doping Fe. Raman spectra was used to further investigate the structural changes of TiO 2 aer doping. All peaks from Raman spectra can be assigned to E g (147 cm À1 , 199 cm À1 and 639 cm À1 ), B 1g (398 cm À1 ) and A 1g + B 2g (517 cm À1 ) mode of anatase TiO 2 , which almost have not change in comparison with that of undoped TiO 2 due to the low-dose implant for TiO 2 (Fig. 1b). XPS analysis was used to investigate the change of surface chemical bonds of the TiO 2 nanoparticles induced by iron as shown in Fig. 1e-g. Fig. 1e (Fig. 1f), respectively, which are in agreement with the values of Ti 4+ in TiO 2 . For the low doping level, the signals of Fe are weak (Fig. 1g). The binding energy at 711.27 eV and 724.47 eV should be assigned to 2p 3/2 and 2p 1/2 of Fe 3+ , respectively. These data exhibited a positive shi compared to those in Fe 2 O 3 (2p 3/2 and 2p 1/2 of Fe 3+ at 710.7 eV and 724.3 eV, respectively), 26 probably indicative of more positively charged surface Fe 3+ . The slight enhancement of Fe 2p level binding energy can be attributed to the diffusion of Fe 3+ into TiO 2 lattice and the formation of Fe-O-Ti bond in the samples. [27][28][29][30] Photoresponse of Fe:TiO 2 /n-Si heterojunction photodetector We measured the light intensity dependence of the Fe:TiO 2 /n-Si heterojunction PD. Fig. 2a shows the current-voltage (I-V) curves of the Fe:TiO 2 /n-Si heterojunction device under 365 nm light with increasing incident light powers from 0.2 to 1.4 mW cm À2 . The I-V curves of the photodetector exhibited a good rectifying behavior. When the light illuminates on the device, the negative biased current increased with increasing the incident light power. With the reverse voltage increasing, the I ph (photocurrent) rose sharply and subsequently reached saturation rapidly. The increasing electric eld can separate electronhole pairs sufficiently and promptly, giving rise to the increase of the photocurrent. But when the bias is sufficiently large, owing to the available number of charge carriers excited by the xed illumination power, the photocurrent saturates. 28 Another interesting nding is that the forward current keeps almost constant regardless of the power density. While in the forward bias, the barrier height and width enlarged, and electrons and holes could not tunnel from it, leading to no variation of forward current. 27 Fig. 2b presents the dependence of photocurrent on the power density at different reverse biases. For lower reverse bias, the photocurrent increased nonlinearly with increasing the power density, while for higher reverse bias, the photocurrent was linearly related to the power density (see Fig. 2b). In general, the dependence of photocurrent on light intensity satises the relationship I ph f P q , where P is the light intensity and q is an empirical value related to the recombination processes of the photoexcited carriers. 33 By tting the curves in Fig. 2b, we obtained q ¼ 0.97 at reverse bias of À1 V, indicating that low trap states in Fe:TiO 2 /n-Si heterojunction photodetectors. 14,33,34 Fig. 2c shows the photovoltage responsivity as a function of the incident light power. The absolute device responsivity R V (¼ V OC /P, V OC is the open-circuit voltage) exceeded 10 2 V W À1 at incident light powers from 0.2 to 1.4 mW cm À2 . It can be as another important parameter for weak signal detection.
Responsivity (R) and external quantum efficiency (EQE) are important quality factors for photodetectors, which can be given by the following formulas: where the I ph , I d , P, S, h, c, e, l are the photocurrent, the dark current, the light power density, the effective area under irradiation, Planck's constant, the velocity of light, the electronic charge, and the wavelength of illuminated light, respectively. In Fig. 2d, R value was as high as 60 mA W À1 at zero bias under a relatively low light intensity of 0.2 mW cm À2 . Notably, R value tended to decrease at high light intensity, which may be attributed to the increased carrier recombination with increasing light intensity. 31 Furthermore, EQE of the Fe:TiO 2 /Si heterostructure detector was determined to be 20% at lower illumination intensity of 0.2 mW cm À2 , which indicated photocarriers can be effectively generated in the heterostructure.
To further investigate the wavelength-dependent characteristics of Fe:TiO 2 /n-Si heterojunction, I-V curves were recorded when the wavelength of the incident radiation was switched from 350 to 600 nm at a constant intensity of 0.5 mW cm À2 (Fig. 3). The photocurrent of Fe:TiO 2 /n-Si heterojunction was found to depend on the wavelength of the incident radiation. As shown in Fig. 3a, the photocurrent of the device increased quickly from À5.77 mA at 350 nm to À7.67 mA at the wavelength of 600 nm. The wavelength dependence of both photocurrent responsivity and EQE of the device are presented in Fig. 3b and c, respectively. Fig. 3d shows the photocurrent responsivity and EQE with different incident light wavelengths at zero bias. Notably, the EQE exceeded 10% when the wavelength of the incident radiation was switched from 350 to 600 nm at a constant intensity of 0.5 mW cm À2 . Furthermore, EQE of the heterojunction rose up beyond 100% with a broad wavelength range at a small reverse bias of À0.5 V (Fig. 3c). Fig. S1 † shows photoresponse characteristics of sample B, C and D. The results indicate that the as-fabricated Fe:TiO 2 /Si PDs have excellent reproducibility, suggesting a promising direction for manufacturing advanced self-powered PDs. Fig. 4a is one cycle of the time-resolved photoresponse of the device at zero bias upon 365 nm light illumination at light intensity of 1.0 mW cm À2 . Photoresponse time is dened as the time of switching duration from 10% to 90% of the pulse peak. From rising and falling edges, we can get rising time t r ¼ 10 ms, falling time t f ¼ 15 ms, respectively (detectable limit of our instrument). In Fig. 4b, the two on/off operational measurements interval one month maintained almost the same light response behaviors. Furthermore, our PDs showed strong response at zero bias voltage toward UV-visible light (0.5 mW cm À2 , l ¼ 365, 400, 450, 500, 550 and 600 nm) as shown in Fig. 4c, suggesting a broadband detection spectrum. We collected photocurrent under different wavelengths with the same light power, and found that the linear correlation between photocurrent and the number of incident photons (see Fig. S2, ESI †). The biasdependent switching characteristics of the photodetector were investigated by applying different bias voltages of 0, À0.5, À1.0, À1.5, and À2.0 V to the device is shown in Fig. 4d. As the bias voltage increases, the photocurrent increases signicantly and eventually reaches a substantially constant. This result provides the possibility of adjusting the optical response of the device by applying an appropriate range of bias voltages.
Excellent photoresponse of the Fe:TiO 2 /Si heterojunction device can be explained by the processes of photo-excitation and carrier transport in the junction illustrated in energy band diagram (see Fig. 5). The Fe doped TiO 2 layer could absorb UV light and part of short-wavelength visible light (see Fig. 5b and c). The photogenerated electron-hole pairs are only created inside TiO 2 side under UV light and part of short-wavelength visible light, while the electron-hole pairs are only created inside Si side under visible light. TiO 2 and Si form a space charge region as shown in Fig. 5b where part of the photogenerated electrons diffuse from TiO 2 towards the junction and are swept into Si, while the photo-generated holes are conned in the Si side because of the large valence band offset. Since the presence of an internal electric eld (at space charge region) facilitates the separation of electron-hole pairs under UV-visible light irradiation, which signicantly reduces the electron-hole recombination ratio, resulting in much higher   photocurrent. In the case of the reverse bias (positive bias is applied on Si sides shown in Fig. 5c), with the same direction of the internal electrical eld, could promote the separation of the electron-hole pairs. Meanwhile, the width of the depletion layer narrows (the height of the barrier becomes low), and the generated electrons in the space charge region of TiO 2 could transport through the depletion layer driven by the applied bias. 35 High depletion region of heterostructure holds a powerful built-in electric eld that rapidly separates photon-generating carriers, leading high quantum efficiency and fast response speed during thermal equilibrium. This make our PDs show self-powered properties with very fast temporal response, as well as excellent EQE and responsivity in UV-visible region. In addition, the vertical conductive distance between TiO 2 lm and the electrodes prevents unnecessary recombination, thereby further improving the response speed. 31,32,36,37 In Fig. S3, † compared to TiO 2 thin lm PD in photoconductive mode, heterostructure PD has higher responsivity and faster response speed. Fig. S5 † shows the PL spectra of pure TiO 2 lm and Fe-doped TiO 2 lm. The intensity increase of the PL from the Fe-doped TiO 2 , compared to pure TiO 2 , which indicated higher carrier concentration. Furthermore, a strong accelerating electric eld is easily formed on thin lm at the applied reverse bias, which could facilitate carrier separation and produce external gain. 38 Moreover, the increased donor density of Fe-doped sample is expected to shi the Fermi level of TiO 2 toward the conduction band. The upward shi of the Fermi level facilitates the charge separation at the interface by increasing the degree of band bending at the TiO 2 surface. The enhanced charge separation and transportation are the major reasons for the observed better performance enhancement.
To verify the presence of built-in electric elds, the capacitance-voltage (C-V) measurements were conducted. As shown in Fig. 6, the C-V curves were performed at 10 kHz in dark condition (black curve line). Further analysis about the interface could be given by Mott-Schottky plot (red curve line). Based on the equation: 39 where q is the elementary charge, 3 s is the dielectric constant of Si, 3 0 is the permittivity of a vacuum, N d is carrier density of n-Si, V is the applied potential, and V bi is the built-in potential. The C-V characteristic of the photodetector shows a typical p-n junction behavior. The capacitance decreased quickly as the reverse voltage increasing, due to the increase in the depletion layer width of the heterojunction. The increased depletion width would benecial to the improvement of the photosensitivity and enhance the response speed. According to the 1/C 2 versus V plot shown in Fig. 6 (red curve line), the value of V bi À kT/q was determined to be z0.12 V by extending the straight line to the voltage axis (kT/q is equal to 0.026 V at the temperature of 300 K). So the potential voltage caused by the internal electrical eld (V bi ) was deduced as 0.15 V.

Conclusions
In summary, self-powered UV-visible PDs employing heterojunction between the Fe:TiO 2 and Si was fabricated via a facile solution process. The existence of built-in electric eld between Fe:TiO 2 and Si help facilitate the separation of photo-generated electron-hole pairs and regulate the electron transport. A fast response speed of 10/15 ms were achieved. Under zero bias, the device exhibited high responsivity of 46 mA W À1 (350 nm) and 60 mA W À1 (600 nm) with a 0.5 mW cm À2 light irradiation. At a small reverse bias of À0.5 V, the EQE of the PD rose up beyond 100% with a broad wavelength range. The exploring of Fe:TiO 2 / n-Si heterojunction PD demonstrates an ultrasensitive, fast, and broad-band (UV-visible) photodetection with no or low external energy supply. We expect a productive future for the selfpowered PDs integrated on exible platforms, even multifunctional and smart systems.

Conflicts of interest
There are no conicts to declare.