As a typical ultra-wide bandgap semiconductor material, β-Ga₂O₃ has a large band width (4.4−4.9 eV)^[1−4], critical breakdown electric field strength (8 MV/cm) and high Baliga figure of merit (3444)^[5−7]. Due to these excellent properties, it has the broad application prospects in power electronic devices and solar-blind ultraviolet (UV) detectors^[8−12]. Compared to the bulk materials, β-Ga₂O₃ nanobelt has the larger surface volume and the higher surface density of states. Therefore, it can carry carriers more effectively and improve the performance of the device^[13−15].

At present, β-Ga₂O₃ nanobelts are synthesized by the various methods including chemical vapor deposition (CVD)^[16], gas-liquid-solid method^[17], gas phase transmission^[18]. Feng etal.^[19] grew β-Ga₂O₃ nanobelts by directly evaporating Ga in a controlled environment. Most of the prepared nanobelts were tens of nanometers in diameter and tens of micrometers in length, and a solar blind UV detector was prepared on a single nanowire which dark current reached the order of pA. Chen etal.^[20] prepared Ga/β-Ga₂O₃ nanowire structure with CVD and deposited gold electrode on it to prepare Au/β-Ga₂O₃ nanobelt Schottky-junction photodetector. The descent speed of the detector reached 64 ns. Kim etal.^[21] developed an ozone-treated β-Ga₂O₃ nanobelt field-effect tube photodetector with a rejection ratio of more than 8 orders of magnitude.

A large number of studies have shown that Ga₂O₃ nanobelts have good optical and electrical properties, but up to now the width of the grown β-Ga₂O₃ is mostly only a few microns and the length direction is limited, which greatly increases the difficulty of device’s preparation and hinders the application process of nanobelts^{[22, 23]}. At the same time, the mainstream junction-type optoelectronic devices of nanobelts are mainly concentrated in Schottky devices with the fast response speed, but these devices also own the large dark-current problem due to the simple structure. Therefore, fabricating large Ga₂O₃ nanobelts and extending the device to more complex structures should utilize the excellent potential of the nanobelts.

In this paper, ultra-long and ultra-wide Ga₂O₃ nanobelts were prepared by carbothermal reduction method without catalyst. The width of the grown Ga₂O₃ nanobelts can reach more than 100 μm. The surface morphology and crystal quality of Ga₂O₃ materials were analyzed and characterized by X-ray diffraction (XRD), field emission scanning electron microscopy (SEM) and transmission electron microscopy (TEM). In addition, the dual-Schottky-junctions coupling device (DSCD) was fabricated by using the synthesized nanobelt as the N-type channel of the device and Ni/Au as the electrode. The electrical performance of the device was tested, and the effects of different UV light intensities on the device was explored.

β-Ga₂O₃ nanobelt was obtained by catalyst-free carbothermal reduction. During the growth process which is following the vapor-solid (VS) growth mechanism, the generated Ga₂O₃ vapor could be directly deposited on the substrate for nuclear growth. Firstly, the corundum boat and plate were ultrasonically oscillated in organic solvent for 20 min, and then the β-Ga₂O₃ powder with the purity of 99.9% and the carbon nanotubes with the purity of 95% were fully mixed at the ratio of 1 : 1.5. The surface of the boat was covered with three pieces of plate. Nitrogen was introduced into the furnace as as a protective gas. When the temperature of the furnace rose to 900 ℃, the corundum boat was pushed into the central temperature zone of the diffusion furnace and kept for 90 min until the temperature dropped to room temperature. Afterwards, the corundum boat was taken out and the synthesized β-Ga₂O₃ nanobelts would attach to the corundum plate^[23]. SEM, XRD and TEM were used to analyze the surface morphology and crystal quality of the generated β-Ga₂O₃ nanobelts. Then the nanobelts on the corundum plate were transferred to a beaker containing absolute ethanol, and the beaker was sonicated until the nanobelts were fully dispersed. The dispersed solution was transferred to the SiO₂ substrate with a rubber dropper. After lithography and electron beam evaporation processes, Ti/Au (100/50 nm) and Ni/Au (100/50 nm) electrodes were deposited on the β-Ga₂O₃ nanobelts and then DSCD was prepared. The current−voltage (I−V) and current−time (I−T) characteristics of the device were measured by Keysight B1505 semiconductor characterization system under dark condition and 254 nm UV light with different power densities.

As can be seen from Fig. 1(a), a large number of nanobelts with the length of about 2−3 mm can be clearly observed. Fig. 1(b) is an enlargement of the yellow dotted box of Fig. 1(a). It shows that these grown nanobelts range from tens of microns to 132 μm. The thickness of these nanobelts is about 100−200 nm. And their surface are very smooth and uniform. Fig. 1(c) shows that the XRD patterns of the obtained nanobelt is corresponding to the β-Ga₂O₃ standard mode (JCPDS card number: 43−1012) at the peak position. There are no other crystalline phases or impurities in the generated Ga₂O₃, which indicates that the synthesized nanobelt is monoclinal β-Ga₂O₃ nanobelt^[24]. It can be seen that the grown nanobelt has obvious single-crystal diffraction spots and also presents superstructures from Fig. 1(d). Fig. 1(e) shows the TEM low-magnification image of the nanobelt. It can be observed that the edge of the nanobelt is straight. The surface of the grown nanobelt is neatly arranged as can be seen in Fig. 1(f). The lattice space is 0.286 and 0.293 nm, which corresponds to plane (400) and plane (201) in Fig. 1(c) respectively, and the growth direction is consistent with the theoretical value. No obvious lattice defects and amorphous layers were found in Fig. 1(f), indicating that the grown nanobelt had excellent crystallinity.

In order to study the electrical properties of grown nanobelt, DSCD was fabricated based on β-Ga₂O₃ nanobelt. Fig. 2(a) shows the schematic diagram of the device structure. Ohmic contacts and Schottky contacts were fabricated on one nanobelt^[25]. S₁ and S₂ are Ni/Au electrodes, which can form Schottky-junctions with β-Ga₂O₃ nanobelt. O₁ and O₂ are Ti/Au Ohmic contacts and used to verify whether the Schottky-junction characteristics of S₁ and S₂ are worked. Although the width and length of the grown nanobelts are relatively large, but they are easy to be broken by ultrasonic vibration, so the width of the nanobelts used for the final preparation of the device is slightly smaller. The final length and width of the nanobelt in the actual fabricated device were 600 and 12 μm, respectively. The space between S₁ and S₂ was 500 μm. The distance between the two adjacent electrodes of Ni/Au and Ti/Au was 6 μm, and the width of the Ni/Au electrode on the nanobelt was 6.5 μm.

To verify the formation of good Schottky-junctions between Ni/Au electrodes (S₁ and S₂) and Ga₂O₃, the forward biases were applied to S₁−O₁ and S₂−O₂ electrode pairs to test their I−V curves.

Equations for the current I of Schottky diode is:

1I=A⋅J0[eqV/nkT−1]=I0[e(EFN−EFM)/nkT−1],

2J0=A*⋅T2e−qφB0/kT,

where A is the area through the current flows, J₀ is the reverse saturation current density and I₀ is the reverse saturation current. V is the applied voltage. n is the ideal factor. A* and T are the ideal Richardson constant and temperature, respectively. In the Eq. (1), I₀ is mainly affected by the barrier height qφ_B0. In Fig. 2(c), when the forward bias is applied, the electron barrier height on the Ga₂O₃ side decreases, so the electrons in the Ga₂O₃ flood into the positive electrode, and the current will increase exponentially. Fig. 2(d) shows the experimental I−V curves between electrodes S₁−O₁ and S₂−O₂ at room temperature. The two I−V curves present the consistent Schottky diode characteristics. Under the forward bias, the current of the device reaches about 2.3 nA at 10 V. The small current indicates that the electron density N_d of the nanobelt is very low, which is caused by less oxygen vacancies and also verifies the good quality of the Ga₂O₃ nanobelt further^{[26, 27]}.

Since S₁ and S₂ electrodes form two Schottky-junctions with Ga₂O₃ nanobelts, they can constitute a DSCD. In a sense, DSCD is like a bipolar transistor with the floating base. Here, S₁ and S₂ can be the emission region or collector region, and Ga₂O₃ nanobelt serves as the base region. Then S₁ and S₂ can be biased at V_S1 and V_S2 respectively to drive the device operation. Fig. 3(a) shows the experimental I_S2−V_S2 curve of DSCD under the condition of grounded S₁. I_S2 which is the current at S₂ begins to increase rapidly when V_S2 is larger than 1 V (A−B stage). Once V_S2 is beyond 2.5 V, I_S2 enters the saturation state (B−C stage). DSCD shows obvious transisitor-like current saturation characteristics, and its saturation current is maintained at about 10 pA. DSCD presents the I_S2 curve very different from I_S2O2 curve of a single Schottky-junction device, and I_S2 is much smaller than I_S2O2 as shown in inset of Fig. 3(a). The physical mechanism of above phenomena is shown in Fig. 3(b). When no bias is applied to S₂ (V_S2 = 0 V), its Fermi level E_FM2 is equal to the Fermi levels of S₁ and Ga₂O₃, E_FM1 and E_FN, respectively. In this case there should be no current flowing in the device as shown in the upper part of Fig. 3(b). When V_S2 > 0 V, E_FM2 decreases, as shown in the bottom part of Fig. 3(b). This results in a high E_FN relative to E_FM2, which positively biases the junction S₂, so that electrons are rapidly extracted from Ga₂O₃ into S₂ to form I_S2. As the electrons in the Ga₂O₃ region are transported to the S₂ region, the energy band E_FN decreases. This results in E_FM1 being higher than E_FN and then the formation of their differernce ΔE₁. This time, the Schottky-junction formed at S₁ is in the reversed biased and electrons enters into the Ga₂O₃ region to form I_S1. According to Eq. (1), regardless of the direction of current, I_S1 and I_S2 at the two junctions are as follows:

3IS1=I0[1−e(EFN−EFM1)/nkT]=I0[1−e−ΔE1/nkT],

4IS2=I0[e(EFN−EFM2)/nkT−1]=I0[eΔE2/nkT−1].

Based on the principle of continuity of current, I_S1 = I_S2:

5I0(1−e−ΔE1/nkT)=I0(eΔE2/nkT−1).

According to Eq. (5), as ΔE₁ increases to 60n meV, e−ΔE2/nkT≪1, ΔE₂ will reach the maximum value ΔE_2max, thus:

6ΔE2max=nkTln2,

7VS2=(ΔE1+ΔE2)/q.

This means that as V_S2 continues to increase, the base conduction band and E_FN near S₂ are lower, so ΔE₁ continues to increase, while ΔE₂ will remain constant after increasing to the maximum ΔE_2max.

Based on Eqs. (3), (4), and (7), I_S2 is as:

8IS2=I0sh(qVS22nkT)/ch(qVS22nkT).

According to Eq. (8), the A−B stage of I_S2 in Fig. 3(a) begins to increase with the increase of V_S2. After ΔE₁ increases to 60n meV, I_S2 changes little with the increase of V_S2 and enters a saturated state. The saturation value of I_S2 is I₀ which is about 10 pA, as shown in the B−C stage in Fig. 3(a). Theoretically, the reverse current I_S1O1 of S₁ junction in the inset of Fig. 2(d) should be basically I₀. Compared with I_S2 of DSCD which rapidly enters the saturation state, the current I_S2O2 of single Schottky-junction increases exponentially with the positive V_S2 as shown in Eq. (1). As a result, I_S2 of DSCD in the semilogarithmic coordinates in Fig. 3(a) is more than two orders of magnitude smaller than I_S2O2 of single Schottky-junction at 10 V. To verify the above mechanism, the bias-setting conditions of S₁ and S₂ are exchanged. Namely, S₂ is grounded and V_S1 is scanned. So the I_S1−V_S1 curve should then take on a similar shape as the I_S2−V_S2 curve. Fig. 3(c) shows that I_S1 really increases with the increase of V_S1, and finally enters the saturation state with magnitude of 11 pA. Due to the saturation current of DSCD in foward V_S2 is only several pA and the off current is about 10⁻¹⁴ A, so the current ratio of on/off is about 10². Further, according to the above mechanism, if S₁ is grounded and a negative bias is applied to S₂ (V_S2 < 0 V), electrons should flow from S₂ to Ga₂O₃, which causing the Fermi level E_FN to rise as shown in Fig. 3(d). So the S₁ junction is positively biased, which makes the electrons flow into the S₁ region from Ga₂O₃ region. And then I_S2 should be negative. If the above inference is correct, the current direction is opposite and the I_S2 curve should also be similar to the case of V_S2 > 0 V. Fig. 3(e) exactly shows that with the increase of |V_S2|, the current |I_S2| increases rapidly and then enters saturation. Additionally, the current of DSCD is much smaller than that composed of single Schottky-junction, which may be due to the coupling between two Schottky-junctions and the floating Ga₂O₃ base region.

As mentioned above, the floating Ga₂O₃ base region plays a special role in modulating the current. So when the device is under the solar-blind UV light, the photogenerated electrons in the Ga₂O₃ base region must cause the change of the energy band. If S₁ is ground and V_S2 = 0 V, E_FM1 and E_FM2 are equal and there should be almost no dark current in the device^[28]. However, the accumulation of photo-generated electrons Δn_g0 in the base region during 254 nm UV light illumination will definitely cause the entire energy band to shift upwards, and then induce ΔE_FN which is the change of E_FN. The relationships between them are as follows:

9Δng0=G⋅τn,

10ΔEFN∝Δng0,

where G is the generation rate of photo-generated electrons and τ_n is the lifetime of excess electrons.

From Eq. (10), as Δn_g0 increases, E_FN will be higher than E_FM2, which makes both S₁ and S₂ Schottky-junctions form a positive bias effect. As shown in Fig. 4(a), these photogenerated electrons will cross the potential barrier and enter the S₂ region to form I_S2^[29]. Therefore, the measurement of I−V characteristics can directly focus on the change of current caused by photogenerated electrons generated by UV light in the base region. Fig. 4(b) shows the experimental I_S2−T curve when UV light power densities P is 709 μW/cm². It can be seen that I_S2 has a good consistency. The current is approximately 2 × 10⁻¹⁴ A when UV light is off, and the photocurrent increases to 2 × 10⁻¹³ A when UV light is on. This result is basically consistent with the above hypothesis, which means that the photogenerated electrons can indeed interact with the floating base region. Furthermore, when the bias is applied to V_S2, the photogenerated electrons will be coupled with V_S2, and affect the current of the device more significantly^[30]. Fig. 4(c) shows the I_S2−V_S2 curve under different power densities of solar blind UV light. Compared with the dark condition, the I_S2 increased significantly with the increase of P from 405, 709, 1023 to 1820 μW/cm². This result is attributed to the positive bias applied to V_S2, which decreases E_FN as shown in Fig. 4(d) and thus causes the S₁ junction to reverse bias to form a depletion region with width of W:

11W={2ε(Vbi+ΔE1/q)/qND}1/2,

where N_D is the electron concentration of Ga₂O₃ nanobelts and V_bi is the self-established potential.

UV light produces a large number of photogenerated electrons in W. At the same time, the neutral region also produces some photogenerated electrons. These two photo-generated electrons flow towards S₂ to form a photo-current I_ph:

12Iph=qA∫0WGdx+qAGLp=qAG(W+Lp),

13G∝αPhν.

In Eq. (12) the first term is the photocurrent generated in the depletion region, and the second term is the photocurrent generated in the neutral region. Lp is the diffusion constant of the hole. Where α is the UV light absorption coefficient of Ga₂O₃, h is Planck's constant, and ν is the UV wavelength.

Since I_S2 = I_S1 + I_ph and I_S1≪I_ph, I_S2 can be derived:

14IS2∝αPhν⋅qA⋅(W+Lp).

As can be seen from Eq. (14), at the same light intensity, the depletion region W increases with the increase of V_S2, so these photogenerated electrons are rapidly pushed to S₂, making I_S2 increase rapidly, as shown in Fig. 4(c). According to Eq. (13), G increases with the increase of P, so it appears in Fig. 4(c) that I_S2 increases with P. When P is very large, the photogenerated electrons Δn_g is very large, which makes I_S2 rise faster. Especially when V_S2 enters the high voltage stage, W will become larger, which leads to a large number of photogenerated electrons Δn_g accumulation in the conduction band valley near S₂. Therefore, S₂ becomes positively biased and the recombination effect near S₂ becomes significant, where the recombination rate R_e of residual carriers is^[31]:

15Re=Δng/τn.

According to Eq. (15), the increase of R_e induced by quite a lot of Δn_g hinders the increase of G and promotes it to stabilize. So I_ph reaches saturation after a rapid increase, which in turn causes I_S2 to enter a saturated state. This result can be seen in Fig. 4(c) that I_S2 enters a saturated state at around 7.5 V when P is 1820 μW/cm².

In order to further verify the mechanism of current generation caused by photogenerated electrons, Fig. 4(e) shows the comparison between the current of S₁−O₁ single Schottky-junction at reverse bias mode and S₂−O₂ single Schottky-junction at forward bias mode under 1820 μW/cm² UV light. It can be seen that the I_S1O1 is very small, which further validates the I_S1r≪I_ph condition in Eq. (12). Although I_S2O2 is close to I_S2, but it is still smaller than I_S2. The reason is that the DSCD also has the depletion region W of the S₁ junction compared with the single Schottky-junction device, so more electrons are generated and the current is larger. The difference between the two currents ΔI in Fig. 4(c) reflects the contribution of W to the current. In addition, under the UV irradiation of 1820 μW/cm² in Fig. 4(c), the photo-to-dark current ratio (PDCR) of the device at 10 V exceeds 10⁴, which is far higher than PDCR of 10² of the single Schottky-junction I_S2O2 before and after illumination, indicating that DSCD has good UV response characteristics. Fig. 4(f) shows the relationship between the power densities of UV light P and I_S2 at different voltages. It shows that the larger P is, the greater the change of I_S2 under different bias voltages (V_S2 = 6, 8, 10 V). At the same UV power density, the higher V_S2 is, the greater the change of I_S2 after irradiation is.

Figs. 5(a) and 5(b) show responsivity R (0.74, 0.96, 1.34, 3.31 A/W), PDCR (5.48 × 10², 1.097 × 10³, 2.196 × 10³, 1 × 10⁴), External quantum efficiency EQE (3.6 × 10²%, 4.7 × 10²%, 6.5 × 10²%, 1.6 × 10³%), Detection rate D* (1.67 × 10¹², 2.15 × 10¹², 2.99 × 10¹², 7.42 × 10¹² Jones) under the different P from 405, 709, 1023 to 1820 μW/cm² at V₂ = 10 V. It can be seen that R, PDCR, EQE and D* increase with the increase of UV light intensity^{[32, 33]}.

Fig. 6(a) shows the I_S2−T curves with different V_S2. Note that the larger the positive and negative bias are, the larger the photocurrent I_S2 is. The rise and decay times are the critical indexes for the photoelectric detector^{[34, 35]}. Fig. 6(b) shows the rise and decay times of I_S2 with different forward biases. The rise time under the V_S2 positive bias of 2, 4, 5 and 10 V ranges from 36 to 52 ms, and the decay time ranges from 43 to 55 ms. Fig. 6(c) shows the rise time of 30−35 ms and the decay time of 25−28 ms under V_S2 negative bias from −2, −5, −7 to −10 V. The rise and fall times are very short in both cases, which indicates that the device has excellent solar blind UV light response characteristics^[36]. From Fig. 6(d), it is found that when the UV light is turned off, I_S2-off does not return to the value of dark current I_S2-dark. This is because of the large Δn_g0 generated in on-state, which rapidly undergoes recombination process when UV light is turned off. However, During R_e process, the photogenerated electrons does not completely disappear, and Δn_g1 still existed, so I_S2-off with a certain value would still be remained, as shown in Fig. 6(e). Moreover, when V_S2 is larger, Δn_g1 is larger, so the I_S2-off in Fig. 6(a) is larger. Fig. 6 (f) extracts the I_S2-on and I_S2-off from Fig. 6(a). It shows that both of I_S2-on and I_S2-off have a nearly linear relationship with V_S2:

16IS2−on=aVS2+b,

17IS2−off=cVS2+d.

Here, a and c are the slopes of I_S2-on and I_S2-off lines. They are 1.186 × 10⁻¹⁰ and 6.231 × 10⁻¹¹, respectively. ΔI_S2, the corresponding difference between I_S2-on and I_S2-off, also increases linearly with V_S2, and its slope is 5.63 × 10⁻¹¹.

In this paper, high quality and ultra-wide and ultra-long Ga₂O₃ single crystal nanobelt were prepared by catalyst-free carbothermal reduction method. The results of SEM, XRD and TEM show that the prepared nanobelts are β-phase single crystals. The DSCD was fabricated based on the grown ultra-long nanobelt. Compared with the single Schottky-junction device the current of which increases exponentially with the increase of voltage, DSCD quickly enters the saturation state with the increase of applied bias, and the saturation current is only 10 pA, which has good output current characteristics. When the device is irradiated by UV light, the Ga₂O₃ region generates photogenerated electron−hole pairs and accumulates electrons. This leads to the upward shift of the floating base region E_FN and then reduces the height of the Schottky barrier where the electrons reach the S₂ region, and then the output current increases. As the power density of UV light increases, the current increases and becomes saturated soon. DSCD has a rise and decay time of 30−50 ms under different V_S2, and has good solar-blind UV light response characteristics. Under 1820 μW/cm² UV light, the PDCR of the device is more than 10⁴, R is 3.31 A/W, EQE is 1.6 × 10³%, and D* is 7.5 × 10¹² Jones at V_S2 = 10 V. This paper should be beneficial to the preparation of devices and related chips based on β-Ga₂O₃ nanobelts.