Optimization of NC-LSPR coupled MoS₂ phototransistors for high-performance broad-spectrum detection

1 Introduction

Phototransistors based on 2D materials have garnered significant attention due to their unique lattice structure, excellent photovoltaic properties, and van der Waals interfaces compatible with CMOS processes [1], [2], [3], [4]. Among them, transition metal sulfur compounds (TMDs), particularly MoS₂, are deemed to be an ideal candidate for photodetectors from the visible to near-infrared (NIR) spectra [4], [5]. This is attributed to their remarkable physicochemical stability, thickness-dependent tunability of energy bands and high photoconductive gain [4], [5], [6], [7]. However, the development of MoS₂ phototransistors has been hindered by their low detectivity (D ^*) resulting from the weak light absorption at atomic thickness [8] and a limited spectral response range (<680 nm) [9]. Current optimization efforts for MoS₂ phototransistor focused on charge-assisted transfer (CAT) strategies [10], [11], [12] or local surface plasmon resonance (LSPR) enhancement to extend the detection range [13], [14]. Nonetheless, the problematic interface of the hybrid structure often leads to the formation of leakage paths [11], [15], causing a significant increase in the dark current (I _dark) of phototransistors and limiting the D ^* to no more than 10¹² Jones [10], [11], [12], [13]. The ferroelectric negative-capacitance (NC) optimization strategy, which aimed to reduce I _dark and increase the photocurrent (I _ph), was constrained by the effective detection range.

To achieve MoS₂ photodetectors with high detectivity and a broad spectral response, we previously proposed an novel NC-LSPR coupling structure of HZO/AuNPs/Al₂O₃/MoS₂ [15], where the Au nanoparticles (AuNPs) were prepared in situ on the surface of Hf_0.5Zr_0.5O₂ (HZO) ferroelectric thin film to enhance its ferroelectric properties [15], [16]. Moreover, the AuNPs/Al₂O₃/MoS₂ structure formed by the insertion of an Al₂O₃ thin film demonstrated a non-direct-contact (NDC) LSPR effect, avoiding the complex interface of AuNPs/MoS₂. The presence of Al₂O₃ as a capacitance-matching layer (CML) also stabilized the NC effect [17], [18], [19], thereby enhancing device performance. However, as a key interlayer, it is very necessary to further investigate the impact of the Al₂O₃ thickness (T _AlO) on the NDC-LSPR effect, the tensile strain applied to MoS₂ and the capacitance matching between the HZO ferroelectric layer and the Al₂O₃ layer, and how these factors ultimately regulate the photoelectric properties of phototransistors.

Therefore, in this work, a series of Al₂O₃-interlayer thicknesses (T _AlO) are designed to observe their influence on the LSPR effect and the tensile strain of MoS₂ within the AuNPs/Al₂O₃/MoS₂ structure through optical characterization combined with COMSOL simulation. Also, the NC-LSPR coupled MoS₂ phototransistors are fabricated, and the corresponding mechanisms of capacitance matching, LSPR and strain effects induced by varying T _AlO on the optoelectronic properties of the phototransistors are investigated. It is found that the LSPR effect and strain are just slightly reduced when T _AlO is below 3 nm, and the phototransistor can achieve a SS below 60 mV/dec in two orders of output-current magnitude when T _AlO is in a 3–5 nm range. Comprehensively considering, the phototransistor could achieve an optimized optoelectronic property when T _AlO is 3 nm, which provides an in-depth insight on the effective distance range of the NDC-LSPR effect. The results show a good consistency across an array of 30 devices, offering an effective solution for large-area 2D phototransistor array with high-performance and broad-spectrum response.

2 Results and discussion

Figure 1(a) illustrates the preparation process of the NC-LSPR coupled phototransistors. Initially, a ∼9-nm HZO thin film is ALD-deposited on a heavily doped Si wafer (with a resistivity of 0.001 Ω cm) cleaned by a standard RCA (Radio Corporation of America) method. This is followed by thermal evaporation of a ∼4-nm Au thin film onto the surface of the HZO film at an evaporation rate of 0.01–0.03 nm/s. Then the Au-covered HZO film undergoes a rapid thermal annealing at 500 °C for 60s, during which the Au thin film breaks down into AuNPs and the recrystallization of the HZO film is completed [20], i.e. the post-metallization annealing (PMA) [16] enhances the ferroelectricity of the HZO film. Subsequently, 1-, 3-, 5- and 7-nm thick Al₂O₃ thin films are deposited respectively by ALD method. Next, large-area CVD-MoS₂ is transferred onto the surface of the Al₂O₃ thin film. Finally, the phototransistor array (Figure 1(b)) is prepared by PMMA-assisted transfer of the source/drain Au electrode array.

Figure 1:

Fabrication process and microstructural characterization of the NC-LSPR phototransistor: (a) the preparation process; (b) schematic diagram and (c) atomic structure of the NC-LSPR-coupled phototransistor; (d) high-resolution TEM image of cross-section of the AuNPs/Al₂O₃/MoS₂ stack; (e) line-scanning element distribution along the white dash line in (d) for Au, Mo, S, Al, Hf, Zr and Si; (f) scanning electron microscopy (SEM) image of the AuNPs, showcasing a normal distribution of diameter primarily concentrated at 15–20 nm.

The NC-LSPR coupled gate-stack structure is shown in Figure 1(c). This structure combines the NC effect from the HZO film [16], [17], [18] with the LSPR effect from the AuNPs [21], [22] to enhance the photocurrent (I _ph). The Al₂O₃ layer, located between the AuNPs and MoS₂, serves as both an interface passivation layer (IPL) and a CML to improve interfacial quality and enhance the NC effect. Additionally, the ALD-deposited Al₂O₃ thin film maintains the tensile strain across the MoS₂ film induced by the AuNPs, thereby stretching the lattice of MoS₂ and reducing its bandgap [14].

Figure 1(d) presents a high-resolution transmission electron microscopy (HRTEM) image of the NC-LSPR structure. A distinct layered structure of Si/HZO/AuNPs/Al₂O₃/MoS₂/Au and diffusion-free clean interfaces are clearly visible. The HZO layer, with its well-defined crystalline structure, is 9 nm thick. The AuNPs exhibit a dispersed and rounded granularity, and the three-layered MoS₂ has a thickness of ∼2.1 nm. Notably, the amorphous Al₂O₃ layer is visible between the MoS₂ and AuNPs, and the in-plane tensile strain of MoS₂ due to the AuNPs is highlighted within the yellow dashed box in Figure 1(d). Figure 1(e) shows the line scanning elemental distributions for Au, Mo, S, Hf, Zr and Si along the white dashed line in Figure 1(d), revealing distinct elemental peaks corresponding to MoS₂, Al₂O₃, AuNPs and HZO. Additionally, Figure 1(f) provides a scanning electron microscopy (SEM) image of the AuNPs array. The nanoparticle diameter has been analyzed using ImageJ, and the resulting normal distribution curve indicates that the diameter of the AuNPs is predominantly in a range of 15–20 nm.

To investigate the influence of T _AlO on the LSPR effect as well as the strain effect across the MoS₂ film in the AuNPs/Al₂O₃/MoS₂ structure, COMSOL simulations were performed for the stacked structure with different T _AlO values of 1 nm (a1), 3 nm (a2), 5 nm (a3) and 7 nm (a4). In the COMSOL simulation, the lateral period is set to 65 nm, with a perfect matching layer placed in the vertical direction at a distance of 760 nm from the device with a width of 380 nm. The nanoparticle diameter is 17 nm, and the mesh size is set to be 0.2 nm in both lateral and vertical directions. Also, a curved A₂O₃ layer covers the AuNPs to align the model with the experimental geometry [23] (the simulation result of the source-drain area with Au electrode is plotted in Figure S1 in Supporting Information). The electric-field distribution obtained from the simulation is shown in Figure 2(a). It is evident that the electric field corresponding to the LSPR effect decreases gradually with increasing T _AlO. Figure 2(b) and (c) display the measured and simulated light absorption spectra from pure MoS₂, AuNPs/MoS₂ and AuNPs/Al₂O₃/MoS₂ with different T _AlO, respectively. In Figure 2(b), three exciton peaks at A (653 nm), B (605 nm) and C (430 nm) correspond to pure MoS₂ [24], and the exciton peak at 550 nm corresponds to the LSPR effect induced by AuNPs [22], [25]. These four exciton peaks are also observed in Figure 2(c). Both experimental data and simulation results exhibit the same pattern: compared to pure MoS₂, the AuNPs/MoS₂ structures show distinct LSPR peaks, which decreases gradually with increasing T _AlO.

Figure 2:

Characterization of the LSPR effect and stress-induced effects in AuNPs/Al₂O₃/MoS₂ with different Al₂O₃ thickness: (a) electrical-field distributions of LSPR effect and (b) light absorption spectra simulated using COMSOL; (c) measured result of the light absorption spectra; (d) the electric field strength in the black dashed box in (a) and the LSPR peaks in the optical absorption spectra of (b) and (c) (λ = 550 nm); (e) PL and (f) Raman spectra corresponding to the tensile strain of MoS₂ induced by AuNPs and the dielectric screening effect of Al₂O₃ layer.

For comparison, the electric field strength in the black dashed box in Figure 2(a) and the LSPR peaks in Figure 2(b) and (c) are extracted as Figure 2(d). It can be seen that when T _AlO is less than 3 nm, the electric-field strength decreases at a rate of ∼2.0 × 10⁴ Vm⁻¹/nm, and when T _AlO exceeds 3 nm, the electric-field strength decreases at an increased rate of ∼5.9 × 10⁴ Vm⁻¹/nm. Similarly, the LSPR peaks obtained from COMSOL simulations and experiments exhibit a slow decreasing trend for T _AlO’s of ≤3 nm (measured results: 0.007/nm) and a rapid decreasing trend for T _AlO’s of >3 nm (measured results: 0.0245/nm). These results suggest that the NDC-LSPR effect in the AuNPs/Al₂O₃/MoS₂ stack is less affected for thin T _AlO’s (≤3 nm) and rapidly weakened for relatively thick T _AlO’s (>3 nm).

Furthermore, the tensile strain of MoS₂ induced by AuNPs can be characterized by Raman and PL spectra. The PL spectra are given in Figure 2(e), where the solid black line represents a distinct peak from pure MoS₂ at 665.5 nm. The corresponding peak of the AuNPs/MoS₂ stack undergoes a significant redshift of 22.4 nm [26], [27], whose intensity gradually weakens with increasing T _AlO, specifically a redshift of 16.5 nm at 3 nm and 2.9 nm at 7 nm. In the Raman spectra shown in Figure 2(f), two main peaks corresponding to the A_1g (404.3 cm⁻¹) and E_2g modes (386.5 cm⁻¹) can be observed in pure MoS₂ [28], in which the E_2g mode is significantly redshifted (towards lower wavenumbers) due to the addition of AuNPs [14], [27], and the redshift decreases for the AuNPs/Al₂O₃/MoS₂ stack with increasing T _AlO. These redshifts in Raman and PL spectra are originated from the bandgap reduction due to the strain effect, which enhances the long-wave absorption capability of MoS₂ [15], [26], [27]. However, the nanoscale particle morphology of AuNPs is gradually covered by the Al₂O₃ layer with increasing T _AlO (see Figure S2 in Supporting Information), leading to a significant reduction of the tensile strain across the MoS₂ film. In addition, it is worth to point out that the blueshift of PL peak [29], [30] with the increasing T _AlO is not only due to the weakening of the strain on MoS₂, but also attributed to the dielectric screening effect of Al₂O₃ layer on the Coulomb interactions either between electrons and holes or between different electrons in MoS₂ [29], [30], [31].

To illustrate the variation trend of the electrical properties of devices with Al₂O₃ thickness and optimize capacitance-matched thickness of Al₂O₃, the electrical performances of pure MoS₂, AuNPs/MoS₂ and different T _AlO’s AuNPs/Al₂O₃/MoS₂ phototransistors are measured and given in Figure 3. The polarization and current versus electric field (P–E and I–E) loops of the TiN/(HZO/AuNPs)/Au capacitor at a frequency of 1 kHz are shown in Figure 3(a). The P–E loop exhibits good ferroelectric polarization with a fixed residual polarization (P _r) of 13.1 μC/cm², and a pronounced polarization reversal current [19] is observed at a fixed coercive field (E _c) of 1.1 MV/cm. Figure 3(b) shows the transfer curves of the phototransistors in the dark state at a drain-source voltage (V _ds) of 0.1 V. The on-state current (I _on), off-state current (I _off), and current switching ratio (I _on/I _off) extracted from the transfer curves are plotted in Figure 3(c), where T _AlO = 0 nm corresponds to the transistor with a gate stack of p⁺-Si/(HZO/AuNPs)/MoS₂/Au, and “Pure MoS₂” represents the transistor with a gate stack of p⁺-Si/HZO/MoS₂/Au. Compared to the pure MoS₂ device, the electrical performance of the AuNPs/MoS₂ device is significantly degraded, mainly due to the formation of leakage paths created by the AuNPs. Fortunately, electrical properties of the AuNPs/Al₂O₃/MoS₂ device with T _AlO in a range of 0–3 nm are greatly improved due to the isolation of those leakage paths by the Al₂O₃ interlayer [15] and the enhanced NC effect resulting from the capacitance-matched ferroelectric HZO layer [17], [18]. For the device with T _AlO = 3 nm, the I _on, I _off and I _on/I _off ratio are extracted to be 2.8 × 10⁻⁷ A, 1.0 × 10⁻¹² A and 2.8 × 10⁵, respectively. After T _AlO exceeds 3 nm, I _on decreases and I _off rises slightly, resulting in a decreasing trend in I _on/I _off. At T _AlO = 7 nm, the I _on, I _off and I _on/I _off are 2.1 × 10⁻⁷ A, 2.2 × 10⁻¹² A, and 9.9 × 10⁴, respectively. This reduction in I _on/I _off is attributed to the capacitive mismatch between the Al₂O₃ and HZO layers [32]. The subthreshold swing (SS) of transistors extracted from Figure 3(b) is plotted in Figure 3(d). The SS increases significantly with the addition of AuNPs but decreases after the insertion of the Al₂O₃ interlayer. At T _AlO = 3 nm, the SS remains below 60 mV/dec from 10⁻¹² to 10⁻¹⁰ A, with a lowest SS of 25.76 mV/dec. For T _AlO’s of >3 nm, the SS gradually increases, and at T _AlO = 7 nm, the SS is below 60 mV/dec only in a small range of drain current (I _d). Further, the hysteresis of the transfer curves and the minimum SS are demonstrated in Figure 3(e). Consistent with SS variation, hysteresis increases sharply with the addition of AuNPs and achieves a minimum (348 ± 85 mV) at a T _AlO of 3 nm after the insertion of the Al₂O₃ interlayer, and then increases again when T _AlO exceeds 3 nm. The degradation of I _on/I _off, SS and hysteresis after T _AlO exceeds 3 nm can be explained by the C–V results in Figure 3(f). The C–V curves were measured on the Si/HZO/(AuNPs or AuNPs/Al₂O₃)/Au capacitor at a frequency of 5 kHz. The total capacitance C ⁻¹ = C _Fe ⁻¹ + C _MOS ⁻¹ (C _Fe is the ferroelectric capacitance of the HZO layer, C _MOS is the MOS capacitance), and the C–V curves display a sharp peak due to the NC effect of the HZO layer [33]. It can be seen that the capacitance peak is highest at a T _AlO of 3 nm, indicating that the NC-LSPR structure with a 3-nm Al₂O₃ interlayer can obtain the best capacitance matching effect [32], [34].

Figure 3:

Electrical properties of pure MoS₂, AuNPs/MoS₂ and different T _AlO’s AuNPs/Al₂O₃/MoS₂ phototransistors: (a) P–E loop and I–E loop measured on TiN/HZO/AuNPs/Au capacitor at a frequency of 1 kHz; (b) transfer curves at a V _ds of 0.1 V in the dark state; (c) variation of I _on, I _off and I _on/I _off ratio extracted from (b) with Al₂O₃ thickness, where T _AlO = 0 nm corresponds to the transistor with a gate stack of p⁺-Si/HZO/AuNPs/MoS₂/Au, and “Pure MoS₂” represents the transistor with a gate stack of p⁺-Si/HZO/MoS₂/Au; (d) SS versus I _d curves of the phototransistors; (e) the hysteresis and minimum SS for different devices with different T _AlO’s; (f) C–V curves measured on the p-Si/HZO/(AuNPs or AuNPs/Al₂O₃)/Au capacitor at a frequency of 5 kHz.

To investigate how the change in T _AlO affects the optical performance of the phototransistor, the optical performance of the NC-LSPR-coupled phototransistor is measured using the circuit connection shown in Figure 4(a) under different light intensities (P). The laser was vertically incident and the light intensity reaching the sample surface was adjusted using attenuators. The transfer curves at a V _ds of 0.1 V and the output curves at a gate-source voltage (V _gs) of 0.7 V of the phototransistor with a 3-nm T _AlO are measured in the dark state and under 528-nm illumination, as shown in Figure 4(b) and (c) respectively. The photogenerated current I _ph (I _ph = I _light − I _dark), responsivity R (R = I _ph/P _i ) and detectivity D ^∗ (D ^∗ = RA ^1/2 (2eI _dark)^−1/2) are extracted from Figure 4(c) and plotted in Figure 4(e), where I _light and I _dark is the I _d under illumination and in dark state, respectively. The incident light power P _i = AP, A is the area of the MoS₂ channel region (240 μm²). It can be observed that the R and D ^∗ of the device with a T _AlO of 3 nm are 140.1 A/W and 3.83 × 10¹⁴ Jones at P = 2.36 μW/cm², respectively. The R and D ^∗ decrease progressively with increasing P, which is consistent with the previous work [15]. Therefore, in order to better compare the optical performance of the devices, the I _ph values of these phototransistors at P = 2.36 μW/cm² and V _ds = 0.4 V are shown in Figure 4(d), from which the R and D ^∗ are calculated as a function of T _AlO, as shown in Figure 4(f). Compared with pure MoS₂, the smaller photocurrent of AuNPs/MoS₂ should be attributed to the reduced mobility by surface roughness scattering [35], [36] due to a poor interfacial quality between AuNPs and MoS₂ and the formation of source-drain leakage channels. Then, as T _AlO increases, the I _ph, R and D ^∗ exhibit an increasing and then decreasing trend, in which the device with T _AlO = 3 nm achieves the best photodetection characteristics of R = 140.1 A/W and D ^∗ = 3.83 × 10¹⁴ Jones. Additionally, I _ph of the device with T _AlO = 3 nm at different polarization angles is provided in Figure S3 in the Supporting Information.

Figure 4:

Optical performance of the phototransistor: (a) circuit schematic for the optical-performance measurement; (b) transfer curves and (c) output curves in the dark state and under 528-nm illumination for the device with a T _AlO of 3 nm; (d) I _ph versus V _ds curves of all phototransistors at P = 2.36 μW/cm²; (e) I _ph, R and D ^* under different light intensities extracted from (c), where the highest R and D ^* are obtained at P = 2.36 μW/cm²; (f) I _ph, R and D ^* for different phototransistors at V _ds = 0.4 V and P = 2.36 μW/cm².

For those phototransistors with T _AlO < 3 nm, despite a strong LSPR effect, the poor electrical properties result in a smaller photogenerated current due to interface issues caused by AuNPs and non-optimal capacitance matching. Since I ph = I light − I dark = ∫ d I d / d V gs d V th , and dI _d/dV _gs depends entirely on the SS, the relationship among ΔV _th, I _light/I _dark and SS can be expressed as lg I light / I dark = lg I light − lg I dark = ∫ S S − 1 d V th [17], indicating that a low SS can lead to a large light-to-dark current ratio. At T _AlO = 3 nm, the device has the lowest SS and a slightly weakened LSPR effect, resulting in the best light detection performance. However, at T _AlO > 3 nm, the greatly weakened LSPR effect plus the increased SS due to the capacitance mismatch from the NC effect lead to the degradation of the optical performance of the devices.

To figure out the influence of the T _AlO on tensile strain across the MoS₂, the photocurrent versus time curves of all phototransistors under 528 nm or 740 nm illumination are presented in Figure 5(a) and (b), respectively (with one light on/off cycle for each device). A significant photocurrent can be observed for all the devices as the light switches at 528 nm illumination, where the photocurrent for the device with T _AlO = 3 nm is ∼811 nA. Nevertheless, under 740 nm illumination, the photocurrent of the pure MoS₂ device is almost indistinguishable, and the photocurrent of the AuNPs/MoS₂ device is notably noisy. Following the insertion of Al₂O₃ interlayer, the photocurrent gradually increases while the noise decreases, and reaches a maximum value at T _AlO = 3 nm (∼216 pA). When T _AlO > 3 nm, the photocurrent starts to reduce and at T _AlO = 7 nm, the photocurrent exhibits significant noise again, primarily due to the attenuation of the strain effect on MoS₂ with increasing T _AlO, which aligns with the conclusions drawn from Figure 2(e) and (f). The normal distributions as well as box plots of the R and D ^∗ calculated from the photocurrents of 30 devices are plotted in Figure 5(c). Under 528 nm and 740 nm illumination, the extracted average D ^∗ values were 3.823 × 10¹⁴ Jones and 1.097 × 10¹⁴ Jones, respectively, indicating that the devices have a wide spectral response range, ultrahigh detectivity and excellent consistency.

Figure 5:

The photocurrent versus time curves of all phototransistors under (a) 528 nm and (b) 740 nm illumination; (c) the normal distributions and box plots of the R and D ^* calculated from the photocurrents of 30 devices; (d) the detectivity at different detection wavelengths for MoS₂ phototransistors optimized by different methods.

Figure 5(d) compares the detectivities at different detection wavelengths for MoS₂ phototransistors optimized by different methods [37], [38], [39], [40], [41], [42], [43] (The specific data for detectivity and wavelength from the references in Figure 5(d) can be seen in Table S1 in Supporting Information). The LSPR-enhanced devices and CTA devices can broaden the detection range, but their detectivities remained below 10¹² Jones due to increased dark current. Conversely, the NC effect-enhanced devices achieved a detectivity above 10¹⁴ Jones, far exceeding conventional MoS₂ transistors. Innovatively, this work combines the high detectivity from the NC effect with the enhanced detection range from the LSPR effect and optimize the Al₂O₃ thickness by considering its influence on the LSPR effect, the strain effect, the NC capacitance matching effect and the modulation of the optoelectronic properties. Ultimately, a high detectivity of more than 10¹⁴ Jones and excellent uniformity were achieved for the NC-LSPR coupling structure of HZO/AuNPs/3 nm Al₂O₃/MoS₂ under 740 nm illumination. This provides a theoretical basis and an experimental support for further applications of the NC-LSPR coupling structures and paves a way for the preparation of 2D high-performance photodetector array.

3 Conclusions

In summary, the influences of the thicknesses of Al₂O₃ interlayer on the LSPR effect, the tensile stress across the MoS₂, the NC effect and optoelectronic properties of the NC-LSPR coupled phototransistors with a gate stack of HZO/AuNPs/Al₂O₃/MoS₂ have been investigated. It has been found that at T _AlO < 3 nm, the strong LSPR and stress effects can be maintained, and at T _AlO of 3 nm, the NC-LSPR coupled structure could obtain the best capacitance matching effect as well as the optimized optoelectrical characteristics, i.e. a SS of 25.76 mV/dec and a detectivity of 3.823 × 10¹⁴/1.097 × 10¹⁴ Jones under 528 nm/740 nm illumination with a good uniformity. The involved mechanisms lie in a good capacitance matching and effective interface passivation provided by optimizing Al₂O₃ thickness in the gate stack of HZO/AuNPs/Al₂O₃/MoS₂ to enhance the NC-LSPR coupling effect, and results in an extremely low SS, enhanced optical absorption due to the obvious LSPR effect and an extended detection range owing to the strain in the MoS₂. This work provides an in-depth insight on the variation of optoelectrical properties of the NC-LSPR coupled devices with increasing T _AlO, and broadens the way for preparing high-performance 2D photodetector array.

4 Methods