Probing the Importance of Charge Balance and Noise Current in WSe2/WS2/MoS2 van der Waals Heterojunction Phototransistors by Selective Electrostatic Doping

Abstract Heterojunction structures using 2D materials are promising building blocks for electronic and optoelectronic devices. The limitations of conventional silicon photodetectors and energy devices are able to be overcome by exploiting quantum tunneling and adjusting charge balance in 2D p–n and n–n junctions. Enhanced photoresponsivity in 2D heterojunction devices can be obtained with WSe2 and BP as p‐type semiconductors and MoS2 and WS2 as n‐type semiconductors. In this study, the relationship between photocurrent and the charge balance of electrons and holes in van der Waals heterojunctions is investigated. To observe this phenomenon, a p‐WSe2/n‐WS2/n‐MoS2 heterojunction device with both p–n and n–n junctions is fabricated. The device can modulate the charge carrier balance between heterojunction layers to generate photocurrent upon illumination by selectively applying electrostatic doping to a specific layer. Using photocurrent mapping, the operating transition zones for the device is demonstrated, allowing to accurately identify the locations where photocurrent generates. Finally, the origins of flicker and shot noise at the different semiconductor interfaces are analyzed to understand their effect on the photoresponsivity and detectivity of unit active area (2.5 µm2, λ = 405 nm) in the p‐WSe2/n‐WS2/n‐MoS2 heterojunction device.


Methods
Device fabrication. We prepared polyvinyl alcohol (PVA) as water-soluble layer and poly (methyl methacrylate) (PMMA, 950K) thin film as acetone soluble layer on SiO 2 wafer for 2D TMDCs (HQ graphene) transfer. Homemade optical microscopy equipped with ultra-long working distance lens (ULWD, Motic) and heating plate (130 °C) was used to align 2D materials. The fine electrodes on 2D materials were fabricated using an acceleration voltage of 20 keV from Raith 150 TWO EBL system. Free patterned gate-and source-drain electrodes were fabricated as the thickness of Titanium (5 nm)-Gold (20 nm), Titanium (10 nm)-Gold (40 nm) using electron beam evaporation system (SORONA), respectively. 2D material active area was proceeded using reactive ion etching (RIE) system as O 2 and CHF 3 gases with RF 80W condition.
Electrical and scanning photocurrent mapping measurement. Homemade LabVIEW program was used to control three probe transistor measurement set-up using Keithley 2636B (GPIB cable) and scanning photocurrent mapping system (RS232 cable). Laser (405 nm, BLM405TA-80R) scanning system is operated by a single protected silver mirror (450~1000 nm) with a motorized x-y scan controller. Laser scanner resolution is < 20 nm. 100X objective ULWD lens was used for in-situ current probing. All measurements were performed in the darkroom. All mapping data was precisely analyzed using the origin plot program with interval 150 step color. Time-reolved photocurrent response (TRPR) was collected using precise time-interval measurement system (buffer time measurement) and laser transistor-transistor logic (TTL) modulation through function generator Hz controller.
Noise current measurement. LabVIEW program (buffer time measurement) to operate the Keithley 2636B was designed for 1000 Hz sampling and 5 second current tracing. For the accurate current measurement, the current level (nA, A) in LabVIEW was fixed to minimize noise. All measurements were conducted with RF current measurement cable and ground handing in measurement condition. We calculated detectivity through flicker noise fitting (1/f) α . All noise power density was extracted for each Hz using fast Fourier transform (FFT) from raw noise current tracing data. The noise current ( ) can be 3 obtained by taking root in the noise power density. The responsivity of 1 Hz modulation was considered to calculate detectivity using the noise current (A). [1,2] Kelvin probe force microscope (KPFM) measurement. We used model NX 10 from the Park system company. KPFM measurement was conducted in non-contact mode/EFM mode. We used the NSC36/Cr-Au tip as an AFM probe. Keithley 2410 was used for external gate voltage source and ground unification with NX 10, and work function was measured through a KPFM tip. HOPG was used to calibrate the work function between the KPFM tip and the material. We acquired images through line scan method for precise work function measurement, and conducted more than 20 samplings.
Supporting section 1: 2D thickness measurement using AFM Figure S1. a) is optical microscope image of our multifunctional 2D heterojunction phototransistor for checking AFM area (1-3 section). b) is AFM mapping image at each section with WS 2 , MoS 2 , WSe 2 , and h-BN. c) is AFM line profile at each 2D materials.
We chose bulk TMDCs of around 20 nm thickness to enhance absorption property. In additionally, we referred the C.H. Lee et al report external quantum efficiency of 2 % (monolayer junction) and 50 % (multilayer > 9 nm junction) in MoS 2 -WSe 2 . [3] h-BN of 22 nm thickness was used for efficiently current (A) blocking as the gate dielectric layer. [3,4] Supporting section 2: WSe 2 annealing to enhance p-type property

Hole enhancement of WSe 2 through annealing process
We annealed to enhance the hole properties of WSe 2 . Annealing 200 o C in atmosphere can control the surface acceptor by forming a self-limiting oxide layer (WO x ) of WSe 2 on the surface. We found that the annealing time of about 1 hours reached the limit of the oxide layer and the characteristics were not changed anymore. Furthermore, the stronger hole properties can be obtained in the ozone O 3 and 100 °C temperature environment, but the annealing conditions described above were used to evaluate the device characteristics in the air atmosphere. [5,6] We observed that the hole characteristics of WSe 2 sufficiently strengthened after annealing improved the rectification characteristics (1 order magnitude current from 10 -8 to 10 -9 A) by forming a sufficient depletion layer thickness at the interface with WS 2 . The photovoltaic effect also lowers the rectified current characteristics, and we can see that a more efficient p-n junction is formed due to the increased amount of photocurrent and lower rectified current level from 10 -12 to 10 -13 A. On the other hand, the second diode rectification characteristic formed at the n-n junction tends to decrease (1 order magnitude current from 10 -11 to 10 -10 A).
Tunable carrier (electron) concentration of WS 2 as a function of a gate bias. We measured the WS 2 FETs of 16 and 6 nm, respectively, in order to investigate the basic transport properties and the change of carrier concentration as a function of a gate bias. First, the mobility of WS 2 can be calculated by the following equation: Where and is length and width of channel, is gradient ( layer, is drain-source voltage bias. WS 2 of 16 nm has 70 cm 2 V -1 S -1 and WS 2 of 6 nm has 38 cm 2 V -1 S -1 . The on/off ratio of WS 2 was measured to be ~10 7 , and S.S. was 80 mV dec -1 , which was confirmed to be a very good device considering the limit of 60 mV dec -1 at room temperature. Based on the mobility of the device, we calculated the carrier concentration through the following equation: Where is conductivity, is electron (coulomb), and is drain-source current.
The WS 2 channel carrier concentration of 16 nm was measured to be from 4.8×10 10 cm -2 at -0.5 V G to 1.6×10 13 cm 2 at 4 V G . The WS 2 channel carrier concentration of 6 nm was measured to be fro m 2×10 10 cm -2 at -0.5 V G to 6×10 13 cm -2 at 3 V G . Two WS 2 channels of different thickness have a suffi ciently wide bandgap, so the carrier concentration is determined by the capacitance of the h-BN dielect ric. The evidence is that the same carrier concentration ratio is observed as like 200 order magnitude. [3,7,8]

Kelvin probe force microscope (KPFM) measurement to understand the work function modulation as a function of gate bias
We measured KPFM to demonstrate the electrical field effect applied only to WS 2 . As shown in Figure S2.3a, we line scan the central area of the MoS 2 /WS 2 /WSe 2 devices to increase the reliability of the KPFM and observe modulation in work function. We matched the common ground in the measurement equipment Keithley to NX10 (KPFM) ground, and the field effect was applied to the sample by connecting the Keithley voltage terminal to the gate pad of the sample. We observed the work function modulation through the reading probe (after HOPG reference) on the surface of the 8 sample, which is receiving the overall field effect. Figure S2.3c shows the result of observing the modulation in the work function at each gate bias by performing line scan more than 20 times. We can clearly observe the mapping result that the work function modulates according to each gate bias. The energy results obtained from the line scan accumulated more than 20 times are averaged and shown in the following table S1 and S2. Table S1. TMDCs work function modulation history by the electrical field effect.
The difference between the work function of WS 2 and MoS 2 is 80 meV, and the work function of WS 2 is lower than that of MoS 2 , and it is observed to be similar to 94.82 meV of the literature value. [9] The difference between the work function of WS 2 and WSe 2 was measured to be 220 meV, and the work function of WSe 2 is lower than that of WS 2 , and this trend is similar to the literature value of 113 meV. [10] As in the cascade band alignment shown in our main Figure 2c, the work function of MoS 2 , WS 2 , and WSe 2 were measured in order of magnitude. It can be seen that the difference between our measurement work function and the literature value comes from the thickness of TMDCs, and the difference can be more than 80 meV. [11]   In this study, we demonstrate the characteristics of the photocurrent and photoelectric properties of the two-dimensional heterojunction properties of charge balance by controlling the Fermi level of very thin 2D materials using the gate field effect. The thickness and the ratio of the depletion layer generated in the p-n junction structure applied in the silicon semiconductor will be briefly described using the following equation. [12]

= = √[ ]
Where is the hole concentration of WSe 2 , is the electron concentration of WS 2 , is the dielectric constant, and is junction voltage in depletion region. Assuming that the value of WSe 2 is unchanged, the thickness of the total depletion layer narrowed by increased due to the gate field effect in WS 2 . The thickness of the depletion layer formed on each bonding surface can be derived by the following equation. [12] = √[ ], When the value of WS 2 is increased, the thickness of the entire depletion layer is decreased and the ratio of WS 2 is also decreased. In addition, the value of the field effect ( ) generated inside the junction by the depleted depletion layer can be expressed by the following equation. [12] = When the thickness of the depletion layer is narrowed, the tunneling effect is maximized because the field effect is maximized. The tunneling effect removes the rectification effect from the junction and can explain the phenomenon that the origin position of the photocurrent is shifted to the cause of the single layer WS 2 . (Figures 3e, f,  A depletion layer is formed in the area band represented by the red line when the charge density ratio is the same, and the depletion layer ratio formed in each 2D material is also divided into 1:1. In the gate bias where the charge ratio is broken ( ), the overall thickness reduction of the depletion layer and the ratio of the depletion layer are expressed as shown in Figure S7.

Supporting section 10: Flicker Noise
Noise consist of thermal noise (Johnson noise) as a function of temperature, Shot noise as a function of applied voltage and flicker noise as a function of various variables under 100 Hz, as mentioned in the main manuscript. Johnson noise and Shot noise exist in the form of a floor, regardless of the frequency band. Therefore, Johnson noise can be calculated by the following formula; [13] = √ Where is Boltzmann's constant in joules per kelvin, T is absolute temperature, is electrical bandwidth, R is resistance of materials in device. Noise generated when an external voltage is applied to the photodetector has shot noise and flicker noise. Shot noise refers to the current noise that causes a difference in the rate at which an electron enters a material through an electrode. This current noise is created by locally obstructing the current flow due to the contact resistance between the material and the metal electrode and the defects, which occur carrier trap in the material. shot noise can be calculated from the DC signal by the following formula; [13,14] = √ Where is electron, is dark current formed by applied DC voltage, and is bandwidth. Using fast Fourier transform (FFT) algorithm obtained by finite sampling of the dark current of photodetector, the method of measuring flicker noise can obtain the noise-frequency spectrum of a sample, which is calculated by the following formula; The noise frequency obtained by the FFT can be sampled by setting the rated current level of the Keithley 2636B precision instrument and up to 1000 samples per second (1000 Hz). [15][16][17] We calculated the Noise Power Density ( ) of the Figure 4a, b through the following current sampling data of Figure S10.

Discussion of gate functional hysteresis
As shown in the Figure S11a

Supporting section 13: The perspective of junction capacitance
The junction capacitance determines the driving speed in the van der Waals heterojunction device.
The junction capacitor is formed by the field effect generated in the space charge region (SCR) around the depletion layer generated in the van der Waals heterojunction. We can express the junction capacitor with the following formula. [12] = | | Where is the difference value of charge across SCR, and is applied voltage bias at source-drain electrodes. This value is determined by the thickness of the depletion layer and the externally applied voltage. From the viewpoint of the charge concentration deeply related to the depletion layer formation, the junction capacitor is expressed by the following equation. [12] = √ Then, the conversion into depletion layer thickness can be summarized by the following formula. []

=
Assuming that the characteristics of the van der Waals heterojunction structure is a comprehensive RC substitution circuit, the response speed can be approximated by an RC time constant ( ). [18] The TRPR following Figure S12.1 and S12.2 shows the analysis at the p-n junction and the n-n junction controlled by external voltage application and gate. Figure S12.2 adjusted by external voltage shows the difference in rising photocurrent tendency, which is one-step photocurrent rising (-1 V D ) and twostep photocurrent rising (0 V D ). The thickness of the depletion layer at 0 V D is smaller than the depletion layer thickness at the reverse of -1V D , and thus shows the difference in observed different photocurrent rising tendencies.
Furthermore, the rising tendency of the photocurrent at the n-n junction at 1 V D shows a different phenomenon from the plane secondary photocurrent rise as observed at 0 V D . Assuming that a very narrow depletion layer has been formed as described in the main menu script with the Schottcky junction of the n-n junction, the consistency of our TRPR results is fully understood. Considering that the TRPR of the 2D material vertical heterojunction reported previously is at the picosecond level, observations at the level of microseconds in our lateral three-stage heterojunction device show that the distance (L) required to separate and harvest the excitons is also a very important factor. More precise analysis of this is needed.