A gate-tunable symmetric bipolar junction transistor fabricated via femtosecond laser processing

Two-dimensional (2D) bipolar junction transistors (BJTs) with van der Waals heterostructures play an important role in the development of future nanoelectronics. Herein, a convenient method is introduced for fabricating a symmetric bipolar junction transistor (SBJT), constructed from black phosphorus and MoS2, with femtosecond laser processing. This SBJT exhibits good bidirectional current amplification owing to its symmetric structure. We placed a top gate on one side of the SBJT to change the difference in the major carrier concentration between the emitter and collector in order to further investigate the effects of electrostatic doping on the device performance. The SBJT can also act as a gate-tunable phototransistor with good photodetectivity and photocurrent gain of β = ∼21. Scanning photocurrent images were used to determine the mechanism governing photocurrent amplification in the phototransistor. These results promote the development of the applications of multifunctional nanoelectronics based on 2D materials.


Introduction
The development of microelectronics technology is inseparable from the 1951 invention of bipolar junction transistors (BJTs), which has helped to produce the digital revolution over the past half-century. 1,2 A BJT, as a three-terminal (emitter, base and collector) device, is the fundamental building block of modern electronic devices, and its main feature is signal gain. 3 These devices have been used in high-power ampliers, high-frequency switches, analog circuits, and radio frequency (RF) systems, [4][5][6] making them widely used in consumer electronics (e.g., communication products), computers, audio-visual systems, and sound equipment. 7 Besides, low power BJTs have also been widely used for the small-signal applications of ampliers, controllers, oscillators and switches, such as electronic ballasts and mobile phone chargers. 8 With the progress of the modern industry and the development of nanotechnology, the applications of nano-dimensional integrated high-performance devices have become the trend of future electronics. As a result, BJTs based on traditional bulk materials, including silicon/silicon-germanium alloys, aluminum gallium arsenide/ gallium arsenide, and indium phosphide/indium gallium arsenide, [9][10][11][12][13] will be not able to meet the demand due to their difficult fabrication procedures and limited performance.
Two-dimensional (2D) materials, including graphene, 14,15 semiconductors (e.g., transition metal dichalcogenides (TMDs), 16,17 and black phosphorus (BP) 18 ), and insulators (e.g., hexagonal boron nitride (hBN) 19 ) have been extensively studied due to their splendid electrical, thermal, optoelectrical, and mechanical properties. [20][21][22][23][24] One of the most signicant investigations in 2D materials is in van der Waals heterostructure devices. 14 Multifunctional p-n diodes, 25,26 ultrasensitive photodetectors, 27,28 high-performance memories, 29,30 light-emitting diodes, 31 and bipolar junction transistors 32 have been fabricated from these materials, showing their potential application in future nanoelectronics. Although several studies on BJTs based on 2D van der Waals heterostructures can be found in the literature, [33][34][35][36][37][38] these structures have an intricate growth procedure 33,34 and tedious multistep transfer process, [35][36][37] making them difficult to fabricate. More importantly, one of the critical factors that inuences BJT performance is the difference in the major carrier concentration between the emitter and collector, which has never been explored until now. A BJT can also operate as a phototransistor (where the base is le oating). 33 Photo FETs based on other TMDs have shown large responsivity and gains, which are mainly operated in photoconductive mode or photovoltaic mode. However, the mechanism of photo-induced current gain in bipolar phototransistors is different from those photodetectors or phototransistors based on two-dimensional materials. In a bipolar phototransistor, the inherent photodiode produces a photocurrent that is subsequently amplied by the internal electrical gain in the BJT. As such, bipolar junction phototransistors could present ultrasensitive photoelectric response in weak light intensity scenarios, due to their additional internal gain, to amplify the weak photoinduced current. Besides, the bipolar transistors can also achieve wellrounded performance in terms of gain, bandwidth, responsivity, quantum efficiency, and dark current noise. [39][40][41][42] Therefore, it is signicant to fabricate high-performance van der Waals heterostructure BJTs in a convenient way.
In this paper, we present a PNP symmetric bipolar junction transistor (SBJT) fabricated with p-type black phosphorus (BP) and n-type MoS 2 with femtosecond laser processing (FSLP). [43][44][45] Compared with other intricate growth procedures and tedious transfer processes, we can produce devices with a single stacking step using FSLP. First, this SBJT exhibits a p-n junction rectication ratio of 10 3 and photoresponsivity of 2.2 A W À1 . This SBJT exhibits bidirectional electrical amplication output thanks to its symmetric energy band structure since the emitter and collector have the same thickness and crystal orientation. We placed a top gate on one side of the SBJT to change the difference in the major carrier concentration between the emitter and collector in order to further investigate the effect of electrostatic doping on the device's performance. Moreover, this SBJT also acts as a kind of phototransistor and exhibits maximum photoresponsivity of R ¼ 151 mA W À1 and maximum photocurrent gain of b $ 21. Scanning photocurrent images (SPI) were used to determine the mechanism governing photocurrent amplication in the SBJT, which was reported in our previous studies. [45][46][47] These results illustrate a novel, convenient method for fabricating multifunctional heterostructure devices.

Results and discussion
The fabrication procedure for a BP/MoS 2 /BP SBJT with femtosecond laser processing (FSLP) and conventional processing is schematically illustrated in Fig. 1. First, a thin BP ake of uniform thickness was mechanically exfoliated using adhesive tape from a bulk BP crystal and placed on a 285 nm SiO 2 /p +doped Si substrate. We used FSLP (800 nm, 35 fs, and 30 mW) to cut the BP into two pieces. More details about FSLP are presented in Fig. S1. † Next, the MoS 2 ake was used to bridge the two BP pieces using a dry-transfer technique, which was reported in our previous studies. [45][46][47] Subsequently, the hBN ake was transferred onto the le BP ake in the same way as the top gate dielectric. Finally, 50 nm thick Au electrodes were patterned and deposited using photolithography and magnetron sputtering, respectively. Fig. S2 † shows optical microscope images of the SBJT aer each fabrication step.
In contrast to FSLP, conventional processing is more complicated and difficult. This method not only includes multistep transfer but also needs to guarantee that the top BP sample cannot contact the bottom BP sample during the transfer. In particular, it is hard to keep the thickness and crystal orientation of the top BP sample consistent with that of the bottom BP sample. Thus, it becomes unpractical to further investigate the effect of electrostatic doping (i.e. the difference in the major carrier concentration between the emitter and collector) on the device's performance. Fig. 2a shows a schematic of the SBJT. To ensure that our FSLP can completely separate a BP ake, we rst fabricated an individual BP device to investigate its optical and electrical properties before and aer FSLP. Fig. 2b shows an optical microscope image of the BP device before FSLP. The scanning electron microscope (SEM) images in Fig. 2c and d show that the minimum linewidth can be reduced to 1 mm, which is conned to our facility and process approach, and we can clearly see that the central region of the BP sample vanished aer FSLP. Fig. 2e shows I-V curves for such a device. Before FSLP, the device exhibited typical ohmic contact behavior (black line). Aer FSLP, the device was no longer conductive (red line), indicating that the pristine BP ake was cut into two pieces. An optical microscope image of the SBJT is shown in Fig. 2f. The le and right BP akes are outlined in blue, MoS 2 is outlined in red, and hBN is outlined in green. The thickness of the BP, MoS 2 , and hBN ake, were measured to be $7, $13, and $14 nm, respectively, using atomic force microscopy Fig. 1 Schematic of the fabrication procedure for femtosecond laser processing (FSLP) and conventional processing. measurements, as presented in Fig. S3. † We can also see that the BP ake was well preserved and there were no obvious oxidized spots. In addition, considering the thickness dependence of MoS 2 (base) for the performance of BJT, there are two reasons for choosing this kind of MoS 2 ake as the base layer. First, the thinner base layer is susceptible to electrical breakdown with higher input current or bias voltages; besides, when the device operates in the forward active region, the collector current I C would better remain constant. If the base is too thin, the holes from the emitter may overcome the barrier height and contribute to I C . Second, if the base layer is too thick, which means the increase in the concentration of the majority carrier (electrons) in the base layer, this results in more carriers (injected from the emitter) being recombined with the carriers of opposite polarity in the base layer so that the current amplication gain will be decreased. The similar thicknesses of BP, MoS 2 and hBN akes were used for another sample. Fig. 2g shows the Raman spectra from BP, MoS 2 , the hBN ake, and the heterojunction. Raman peaks observed at $361 cm À1 , $439 cm À1 , and $466 cm À1 correspond to Ag 1 , B 2g , and A 2 g phonon modes in BP, 35,47 respectively. The peaks observed at $383 cm À1 and $408 cm À1 correspond to the E 1 2g and A 1g phonon modes of MoS 2 , respectively, 35,36 and the peaks observed at $1366 cm À1 correspond to the E 2g phonon modes in hBN. 19 Raman modes from the overlapped region of the stacked hBN/BP (le), BP (le)/MoS 2 , and BP (right)/MoS 2 layers correspond to peaks for each ake, which indicates good lm quality in the junction region aer exfoliation and dry-transfer. The band structure of the (p-type) BP/(n-type) MoS 2 /(p-type) BP SBJT are shown in Fig. 2h. The le and right BP regions have the same thickness and crystal orientation, and the major carrier concentration (holes) and band gap in the le BP are the same as those in the right BP, forming a symmetric structure. On this basis, we placed the hBN as the top gate dielectric to modulate the hole concentration in the le BP and break the symmetry and further study the effect of electrostatic doping on the device's performance.
Next, we investigated the electrical characteristics of the SBJT. All the electrical and optoelectrical measurements were performed at room temperature and in ambient conditions within several hours. First, the basis to ensure the BJT's performances is the two p-n junctions between the emitter-base and collector-base. I-V curves from the two p-n junctions (i.e., the le and right BP/MoS 2 regions) of the SBJT are shown in Fig. 3a and b, respectively. The insets show these I-V curves on a log scale. The forward current could be $10 À7 A, while the reverse current is very small, $10 À10 A. The p-n junctions exhibit rectication ratios of $127 and $4760 at V ds ¼ À2 and +2 V, respectively. The ideality factor was calculated with the following equation: where I is current through the diode, V is the voltage across the diode, I 0 is the dark saturation current, n is the ideality factor, k is the Boltzmann constant, and T is the absolute temperature. 48 The ideality factors for the le and right BP/MoS 2 diodes were calculated to be 2.18 and 1.76, respectively, which are similar to This journal is © The Royal Society of Chemistry 2020 Nanoscale Adv., 2020, 2, 1733-1740 | 1735 Paper Nanoscale Advances the values found in previous studies. 25 These results demonstrate that the two p-n junctions with excellent rectifying behaviors were formed in the BJT. The reason that the right BP/ MoS 2 diode has a lower ideality factor than the le BP/MoS 2 diode may be due to the smaller contact resistance between the right BP and MoS 2 samples. Fig. 3c and d show I-V curves from the le and right p-n junctions in the SBJT while illuminated with light from a 532 nm laser with various incident powers (the insets show the photoresponsivity (R) at V ds ¼ 2 V), respectively. The photocurrent increases as the incident light power increases. The photoresponsivity R is dened as I ph /P laser , where I ph is dened as I illumination -I dark , and I illumination and I dark are respectively I ds with and without illumination, and P laser is the incident laser power. Every p-n junction exhibits R values up to 348 mA W À1 and 2.2 A W À1 , respectively. Fig. S4 † shows I-V curves from the le BP/MoS 2 p-n junction under illumination by a 532 nm laser with various incident powers and top gate bias, illustrating its tunable photodetectivity, which will be discussed later.
Under the common-emitter conguration, this SBJT exhibits bidirectional electrical output thanks to its symmetric energy band structure since the emitter and collector have the same thickness and crystal orientation, as shown in Fig. 4. Fig. 4a demonstrates the I C -V CE characterizations of the SBJT at various injection currents (I B ). The inset shows the common-emitter conguration, i.e. the le BP acts as the emitter (ground), MoS 2 acts as the base, and the right BP acts as the collector. It is worth mentioning that when the device is operating in the small V CE , it would be in the saturation region, the base-emitter junction and base-collector junction are both in the forward bias, leading to majority carriers in the emitter being injected into the base, and the current increase with the increase in V CE .
The collector current did not highly depend on the base current. When the V CE is large enough, it will be in the forward active region; the curves exhibit a slight ascendant trend, where I C increases with V CE . Similar results can be seen when the le BP acts as the collector, MoS 2 acts as the base, and the right BP acts as the emitter, as shown in Fig. 4b. In the saturation region, the current of the collector would not be controlled by the current of the base-emitter junction. It is worth noting that when V CE increases to $À1.8 V, it would be in the forward active region, and the base-collector junction would transform to reverse bias. The current of the collector I C would remain almost constant, and only has a relationship with the injection current (I B ). These results indicate that our device works under ideal conditions, as shown in Fig. 4b. The common-emitter current gain can be dened as b ¼ I C /I B . The common-emitter current gain b versus the collector-emitter voltage (V CE ) curves at various injection currents (I B ) corresponding to Fig. 4a and b are demonstrated in Fig. 4c and d, respectively. The device shows good current gain behaviors and the maximum b were calculated to be $6 and $3, respectively. The current gain b of $6 is not too big but it is comparable to previously reported results. 32,34,35 There are two main reasons for the relatively low current gain. First, the gap size made by laser processing between the emitter and collector may be still large. Second, the doping concentration of the carrier between the p-type BP and n-type MoS 2 may not be properly matched. We believe that the performance could be further improved by optimizing the construction of the device, such as shortening the gap width between the emitter and collector. As the gap width becomes narrow, the transit time of the minority hole in the MoS 2 (base) decreases, and the probabilities of the recombination of the minority holes and majority electrons are also reduced. As a result, the current gain increases and the device shows better performance. However, there needs to be further investigation on related experiments. We noticed that when the le BP acts as the emitter, the output amplication abilities are better as compared to when the right BP acts as the emitter, which may also be attributed to the smaller contact resistance between the right BP and MoS 2 samples. To verify the stability of the devices, we performed similar electrical output measurements aer two months as shown in Fig. S5. † During this time, we kept this device in a glove box lled with N 2 . We observed that the two p-n junctions still showed good rectifying behaviors, as demonstrated in Fig. S5a and b. † The bidirectional electrical amplication output characterizations under common-emitter conguration are also exhibited in Fig. S5c and d † and these results indicate that our device has good stability.
Usually, as the doping concentration in the emitter increases, the output current increases. 9-13 Therefore, we placed a top gate on one side of the SBJT to further investigate the effect of electrostatic doping on the device's performance. In our device, the top gate applied on the le black phosphorus (emitter) is a high k hBN dielectric with a thickness of $14 nm, therefore, the gate-tunable behavior is more effective and comprehensive. Fig. S6 † shows the gate-tunable I-V curves and corresponding band diagrams when the SBJT is operated in the forward active mode. The common-base conguration is shown This journal is © The Royal Society of Chemistry 2020 Nanoscale Advances Paper in Fig. S6a. † The le BP acts as the emitter, MoS 2 acts as the base (ground), and the right BP acts as the collector. The top gate was placed on the le BP (emitter). We observed that the collector current I C decreased regardless of the gate voltage decrease or increase (Fig. S6b and c †), which is not consistent with our expectations. Fig. S6d-f † shows a schematic and band diagrams of the gate-tunable SBJT in the forward active mode with zero, negative, and positive gate modulation, respectively. We can explain this kind of current transport in the SBJT with the energy band models. Under negative gate modulation, although the concentration of holes in the emitter increases, the barrier's height between the emitter/base increases and thus few holes can cross the le BP/MoS 2 junction and be collected by the right BP, and the output current decreases. Under positive gate modulation, the height of the barrier decreases. However, the concentration of holes in the emitter decreases and thus few holes can cross the le BP/MoS 2 junction and be collected by the right BP, yielding a smaller current. More details are presented in the ESI (Fig. S6 †). Similar results were obtained in another SBJT. Fig. S7a † shows an optical microscope image of another SBJT. Fig. S7b and c † show I-V curves from the le and right BP/MoS 2 p-n junctions in another SBJT (the insets show I-V curves on a log scale), respectively. Typical rectifying behaviors in the p-n junction were observed at the two junctions with rectication ratios of $35 and $37 at V ds ¼ À2 and +2 V, respectively. For another SBJT, the le BP acts as the emitter, MoS 2 acts as the base (ground), and the right BP acts as the collector, as was the case in the SBJT discussed in the ESI. † Fig. S8 † shows the gate-tunable I-V curves of this second SBJT. The output current decreases when the gate is positively or negatively modulated. Therefore, there is something meaningful in the collector current phenomenon I C and we can conclude that the top gate on the emitter affects the carrier concentration in the emitter and modulates the band alignment between the emitter and base. To get better performance, optimized devices and experiments should be implemented.
Our SBJT can also operate as a phototransistor with gatetunable photocurrent amplication when the base is le oating. Two important parameters for evaluating Fig. 5 Photoelectric properties of the gate-tunable phototransistor. (a) I-V curves from the gate-tunable phototransistor while illuminated with 532 nm laser light at various incident powers and V g ¼ 0 V. Inset: schematic of the gate-tunable phototransistor and photoresponsivity (R) of the device at various incident laser powers with V ds ¼ 2 V. R ¼ 151 mA W À1 at P ¼ 100 nW. (b) Gate-tunable R with various incident powers at V ds ¼ 2 V. The maximum photoresponsivity value is R $151 mA W À1 at V g ¼ 0 V. (c) Gate-tunable optical gain b with varying V ds at P ¼ 100 nW. The maximum optical gain is b ¼ 21 at V g ¼ 0 V. Scanning photocurrent images at different biases with zero gate voltage and corresponding schematic diagram of the device illuminated with 532 nm laser light; This journal is © The Royal Society of Chemistry 2020 Nanoscale Adv., 2020, 2, 1733-1740 | 1737

Paper
Nanoscale Advances a phototransistor are its photoresponsivity (R) and gain (b). 42 b is dened as b photo ¼ I ds /I pn for a given level of illumination, 33,42 when the base is le oating; I ds is the two-terminal current measured between the le and right BP regions. The gain here is in reference to the photocurrent generated in the le BP/MoS 2 p-n diode (I pn ), which we measured previously. The gate-tunable photoelectric properties of the phototransistor are shown in Fig. 5. Fig. 5a shows I-V curves of the phototransistor illumination with a 532 nm laser at various incident powers with V g ¼ 0 V. The photocurrent increases as the input illumination increases, which is attributed to the increased photoinduced carrier concentration. The insets show a schematic of the gate-tunable phototransistor and the photoresponsivity (R) of the device at various incident laser powers with V ds ¼ 2 V and V g ¼ 0 V. The value R ¼ 151 mA W À1 was calculated at P ¼ 100 nW. The gate-tunable I-V curves from the illuminated phototransistor can be seen in Fig. S9. † Fig. 5b shows the gate-tunable photoresponsivity R at various incident powers with V ds ¼ 2 V; a maximum photoresponsivity of R ¼ 151 mA W À1 was observed at zero gate bias. Similar to the case demonstrated in Fig. S6, † the photocurrent decreases when the gate voltages decrease due to higher barrier at the emitter-base junction, while increasing the gate voltage decreases the holes' concentration, which also decreases the photocurrent. The gatetunable optical gain b with varying V ds is shown in Fig. 5c, and maximum gain of b $ 21 was found at V g ¼ 0 V; although this result is not very outstanding, it is better than that of other bipolar phototransistors, 40,42,49,50 as shown in Table 1. The maximum b value was also observed at zero gate bias for the same reason stated above. Scanning photocurrent images (SPI) were gathered at zero gate bias and while the device was illuminated with 532 nm laser light in order to determine the mechanism governing the photocurrent amplication. Fig. 5d shows the SPI and corresponding schematic diagram of the device at V ds ¼ 2 V. Apparent photocurrent signals can be observed at the interface between the le BP region and MoS 2 , as shown in Fig. 5d(i). Under forward bias, holes were driven from the le BP region to right BP region by the electric eld, forming a forward current. When the le BP/MoS 2 interface region was illuminated, photoinduced electron-hole pairs will be separated by the electric eld at the junction, and holes ow into the base and increase the photocurrent, as shown in Fig. 5d(ii). On the other hand, although there are photogenerated electron-hole pairs at the right BP/MoS 2 junction when this region is excited, the separated electrons will recombine with holes from the base, which does not affect photocurrent amplication, as shown in Fig. 5d(iii). Fig. 5e shows the SPI and corresponding schematic diagram of the device at zero bias. Due to the photovoltaic effect, the photogenerated electron-hole pairs will separate into free charges at the p-n interface and form a photocurrent driven by the built-in electric eld. The two p-n junctions have mirror symmetry and thus, two opposite photocurrents can be observed at the le and right BP/MoS 2 interface regions. Fig. 5f(i) shows the SPI from the device at V ds ¼ À2 V. One can see that the strongest photoresponse occurs at the right BP/MoS 2 interface, which is attributed to photoinduced carriers at the junction. The symmetric structure of our phototransistor can be explained as illustrated in Fig. 5d. The photogenerated electron-hole pairs will be separated by the electric eld at the right BP/MoS 2 junction when this region is excited; holes move into the base and amplify the backward photocurrent, as shown in Fig. 5f(ii). When the le BP/MoS 2 interface region is illuminated, photoinduced electron-hole pairs will be separated by the electric eld at the junction; separated electrons will recombine with holes in the base, leading to very low photocurrent ( Fig. 5f(iii)). The gate-tunable SPI from the phototransistor are shown in Fig. S10. † In spite of the various gate bias values, the photocurrent is primarily concentrated in the same location for a given bias. The strongest photocurrent was measured when the gate was held at 0 V, and the photoresponse decreased gradually when the gate bias was non-zero. These phenomena also occur due to the broken symmetry at non-zero gate bias.

Conclusions
In summary, we presented a symmetric PNP SBJT fabricated from p-type BP and n-type MoS 2 with FSLP. Compared with other intricate growth procedures and tedious transfer processes, we can produce BJT with a single stacking step with the advantage of femtosecond laser processing. On this basis, femtosecond laser processing allows us to reduce the cost and time of the fabrication process, improve the success rate of the device fabrication. This SBJT exhibits a rectication ratio of 10 3 and photoresponsivity of 2.2 A W À1 . This SBJT exhibits bidirectional electrical output due to its symmetric band structure since the emitter and collector have the same thickness and crystal orientation. Moreover, this SBJT can act as a kind of phototransistor with maximum photoresponsivity of R ¼ 151 mA W À1 and maximum photocurrent gain of b $ 21.
Better performance could be further achieved by optimizing the construction of the device, such as shortening the gap width between the emitter and collector. Scanning photocurrent images (SPI) were used to determine the mechanism governing photocurrent amplication in the SBJT. These results illustrate a novel, convenient method for fabricating multifunctional heterostructure devices.

Fabrication of symmetric bipolar junction transistor
Black phosphorus thin akes of uniform thickness were mechanically exfoliated using adhesive tape (3M Scotch) from bulk BP crystals (XFNANO, Inc) on a clean SiO 2 /Si substrate. The thickness of SiO 2 on the p-doped Si substrate was $285 nm. FSLP (800 nm, 35 fs, and 30 mW) was used to cut the BP ake into two pieces. More details regarding FSLP are given in Fig. S1. † A MoS 2 ake was bridged onto the two BP pieces using a dry-transfer technique with a micromanipulator and an optical microscope. Subsequently, an hBN ake was transferred onto the le BP ake as the top gate dielectric in the same way. Aerwards, 50 nm Au electrodes were patterned using photolithography with the femtosecond laser (800 nm, 35 fs) and deposited using magnetron sputtering (JZCK-465 of Sky Technology Development). Fig. S2 † shows the corresponding optical microscope images of the SBJT aer each fabrication step. Finally, vacuum thermal annealing (320 C for 1 h) was performed to achieve good contact.

Characterizations
Raman spectroscopy and AFM were measured using a confocal Raman/AFM system (Alpha 300 R, WITec). Raman spectroscopy was conducted using a 532 nm laser with power of 3 mW (integration time of 30 s and number of accumulations of 5), and the resolution of the ne Raman spectrum was above 0.1 cm À1 .

Electrical and optoelectrical measurements
All the electrical and photoresponse measurements were gathered under ambient conditions and at room temperature. The electrical and optoelectronic characteristics were measured with a semiconductor parameter analyzer (Keithley 4200A), and the SPI were gathered with digital source-measure units (Keithley 2400 and 2450).

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