1. Introduction

Wide band gap (WBG) materials such as ZnO (∼3.37 eV) and Al_xGa_1-xN (∼3.4 at x = 0 to ∼6.2 eV at x = 1) have been highlighted in recent decades for their high responsivity and solar blind ultraviolet (UV) light photodetector (PD) applications [1–3]. Moreover, due to their excellent chemical and thermal stability, they could be one of potential choices for the future applications, such as early detection of flame [4] and UV fluorescence tissue diagnostics in-vivo for human tissues [5]. Utilization of the AlGaN/GaN heterojunction with two dimensional electron gas (2-DEG) at the hetero-interface has especially leaded to high-performance UV PDs with cut-off wavelength down to 360 nm using a gateless high electron mobility transistor (HEMT) structure [6,7]. However, these HEMT-based PD structures could not exhibit desirable on/off current ratio (photocurrent/dark-current) and sensitivity despite of their impressive device performance of high responsivity. For example, UV PDs with ZnO nanorod (NR) gated AlGaN/GaN HEMT structure showed an extremely high responsivity of ∼10⁵ A/W, but they revealed fairly low on/off current ratio of ∼4 when the conventional mesa isolation scheme was used [8].

It has been shown that the UV photoresponse characteristics of ZnO NR-gated HEMTs can be greatly influenced by the dimensional characteristics, such as gate length, gate width, gate capacitance, and channel geometry [8,9]. In this type of device, the photonic response especially depends on the chemisorption kinetics of the oxygen ion (O₂⁻) at the surface of UV-absorbing ZnO nanostructures [10]. Chemisorption physics of O₂⁻ ions at the ZnO surface was also extensively examined by many earlier studies [11,12], and it was known to be affected by a variety of factors, such as free electron concentration and the defect density near the surface of ZnO with native n-type doping [13–15]. Not many research efforts have been made to date on investigating the effect of ZnO NR surface modification on the photoresponse characteristics of AlGaN/GaN HEMT-based PDs to promote the structural quality of ZnO surface. For instance, surface modification methods for the ZnO thin films or nanostructures grown on the planar devices based on the diode or homojunction structures have shown to be effective in terms of various aspects of device performance, such as responsivity, response speed, and sensitivity when plasma treated by O₂ or NH₃ and post-annealed in modified ambient conditions [2,16–20].

Solution-based hydrothermal method [21] used in this work to grow ZnO NR crystals is known to deliver various kinds of extrinsic and intrinsic defects [22]. In the recent decades, much effort has been made on improving the crystalline quality of the ZnO NRs by plasma treatments using conventional gases, such as H₂, N₂, NH₃, and O_2, for various ZnO passive-type PDs [19,23–28]. N₂O plasma treatment was also reported in thin film transistors (TFTs) of the bottom-gate configuration to suppress the intrinsic defects, such as oxygen vacancies of ZnO [29]. Rapid thermal annealing and subsequent N₂O plasma treatment increased the carrier mobility and on/off current ratio by two order of the magnitude with an improvement in crystalline quality of the thin film channel.

In this work, the influence of plasma treatment on nanocrystalline structure, photo-emission properties, intrinsic defects, and stoichiometry in ZnO crystals were examined using a variety of surface characterizations. We also investigated the effect of N₂O plasma treatment on ZnO NRs grown atop the gate area (2×100 µm²) of AlGaN/GaN HEMT PDs. The performance parameters of the PDs fabricated with NRs plasma-treated under different conditions were discussed with our material characterizations.

2. Experiments

Figure 1(a) shows the schematic illustrations of the HEMT-based PD structures fabricated in this work. For this, epitaxial structure of the GaN buffer/3000-4000 nm GaN channel/20 nm Al_0.25Ga_0.75N barrier/1.25 nm GaN cap layer from the bottom were grown on a (111) Si substrate by using the metal organic chemical vapor deposition. Active mesa region was defined by an optical lithography pattern of a 100×40 µm² dimension. The mesa region was then isolated by etching a depth of 100 nm from the AlGaN/GaN epitaxial layer in a reactive ion chamber (RIE) using BCl₃ and Cl₂ gases. The ohmic contacts were achieved by the subsequent deposition of Ti/Al/Ni/Au layers (30/180/40/150 nm) followed by a rapid thermal annealing at 900 °C for 35 s in nitrogen ambient. 100 nm thick silicon nitride (Si₃N₄) passivation layer was then deposited by using a plasma-enhanced chemical vapor deposition (PECVD). After the completion of gateless HEMT structure to this point, ZnO NRs were grown on the gate region. For this patterned growth, the passivation layer was first selectively etched from the gate area (2×100 µm²) using CF₄ plasma in our RIE system. For the ZnO NRs growth, a seed layer (SL) solution was prepared by 0.2 M zinc acetate dehydrate in 30 ml 1-propanol. The SL solution was then spun onto the gate area at 3000 rpm, and this spin coating step was repeated 15 times with the subsequent prebake (120 °C, 60 s) for each iteration to ensure achieving a final SL thickness of 20 nm. After the deposition of SLs, all the samples were annealed at 300 °C for 1 hr on a hotplate in air ambient [30].

 

Fig. 1. (a) Schematic illustrations (untreated and plasma-treated), (b) top-view optical image, and (c) top-view SEM micrograph of the fabricated ZnO NR-gated AlGaN/GaN HEMT. (d) Top-view (left), expanded top-view (top-right), and cross-sectional SEM view (bottom right) of as-grown ZnO NRs.

Download Full Size | PDF

In the next step, the NRs were grown in an equimolar (25 mM) aqueous solution of zinc nitrate hexahydrate (Zn(NO₃)₂·6H₂O, 99%) and hexamethylenetetramine (HMTA, C₆H₁₂N₄, 99.5%). After the aqueous solution was prepared in sealable glass beaker, the samples were immersed upside down into the surface of the growth solution. The beaker was then sealed with the aluminum foil and placed on a hotplate at 90 °C for 6 hr. The samples were finally washed by de-ionized water and dried at room temperature. We synthesized the NRs prepared under four different conditions including a control sample (CS) and N₂O plasma-treated samples for three different time intervals of 1, 3, and 6 min in a reactive ion plasma chamber at 50 mTorr base pressure, N₂O gas flow rate of 100 sccm, and a RF power of 100 W.

Top-view morphologies of the ZnO NRs and fabricated PDs were characterized by optical microscopy and field emission scanning electron microscopy (FE-SEM, S-4800, Hitachi). The crystalline structures and optical properties of the NR samples were respectively characterized by X-ray diffraction (XRD, D8Advance spectrometer of Bruker AXS with Cu Kα 1.540 Å radiation) and room temperature (RT) photoluminescence (PL, RPM 2000 Accent). The oxygen stoichiometry in the chemical bonding for each NR sample was investigated by X-ray photoelectron spectroscopy (XPS, Thermo-fisher scientific, NEXSA) spectrometer with an X-ray source gun of 400 µm spot size. I-V characteristics of the devices were measured by a Keithely SMU under both dark and UV illumination conditions. The spectral response of the HEMT-based PDs were measured by using a monochromatic light source with a wavelength range of 290-700 nm. A broad band Xenon lamp (200-1100 nm, 300 W) light was filtered and chopped into monochromatic light using a computer program controlled diffraction grating. Uniform monochromatic light was passed through a frequency chopper wheel at 30 Hz and illuminated on to the sample. The frequency of monochromatic light was fed to a lock-in amplifier, and the photocurrents from the devices were recorded at each wavelength (290-700 nm) and analyzed.

3. Results and discussion

As shown in Figs. 1(b) and 1(c), ZnO NRs were grown on 2×100 µm² gate area of the gateless AlGaN/GaN HEMT structure fabricated with source and drain electrodes partially covered by Si₃N₄ passivation. Top inset of Fig. 1(c) demonstrates the growth profile of NRs in an expanded view with a defined gate length of ∼2 µm. Shown in Fig. 1(d) are the vertically aligned growths of NRs with diameters ranging from 70 to 80 nm (see top right inset) and an average length of ∼1.4 µm (see bottom right inset).

UV photo-detection of ZnO based PDs is associated with the desorption of oxygen (O₂) molecules which are chemisorbed at the surface of ZnO under dark condition [9,15]. As illustrated in Fig. 2(a), O₂ molecules are adsorbed near the surface of the ZnO NRs by scavenging the free electrons from the surface volume of NRs and thereby forming O₂⁻ ions under dark condition. Due to the adsorption of O₂⁻ ions, the surface depletion region is expanded, and it produces a negative potential in the gate region of the AlGaN/GaN HEMT. The development of this negative potential at the surface of ZnO NRs (UV light-adsorbing structure) pushes the electron in 2-DEG deep into the GaN channel layer and reduces the free electron concentration. This process decreases the channel conductance and the drain current of the HEMT-based PD structure under dark condition as shown in Fig. 2(a). Upon the irradiation of UV light, electron-hole pairs are generated near the surface and in the bulk of ZnO NRs. While the holes recombine with the adsorbed O₂⁻ ions, desorbed O₂ molecules migrate from the NR surface into the atmosphere. Consequently, the negative gate potential at the surface of UV absorbing structure is reduced, and it increases the carrier concentration in 2-DEG channel as shown in Fig. 2(b). This reaction will eventually increase the drain current under the UV illumination.

 

Fig. 2. Illustration of O₂⁻ (a) adsorption in the dark condition (b) and desorption under UV illumination taking place on the ZnO NR surface. Free electron concentration in two-dimensional electron gas (2-DEG) of HEMT is influenced by this surface adsorption-desorption process.

Download Full Size | PDF

Because of this sensitive dependence of channel conductance on the ZnO surface state, to control the defect states of NR surface will be one of the most crucial parameters affecting the performance of ZnO NRs gated HEMT-based PDs. Broadband (400-750 nm) visible PL emissions shown in Fig. 3 are generally associated with the intrinsic defect states such as, oxygen vacancies (V_O), zinc vacancy (V_Zn), oxygen interstitial (O_i), zinc interstitial (Zn_i), oxygen anti-site (O_Zn), zinc anti-site (Zn_O), and native defect clusters such as V_OZn_i [22]. Experimental observation and its theoretical interpretation performed by Deng et al. [31] revealed that ZnO nanostructures have the highest concentration of V_O at the surface, and these vacancies act as the trap centers for the n-type carriers in the conduction band (free electron). Since the atmospheric O₂ molecules are highly electronegative, they can take the free electrons easily from conduction band near the surface of ZnO NRs thereby forming the negatively charged O₂⁻ ions [32]. Therefore, suppressing the defect density can give rise to the increases in free electron concentration and chemisorbed O₂⁻ ions at the surface by extending the depletion region even further. A. R. Jason [33] studied the roles of absorbed O₂⁻ ions and V_O on the green emission of the ZnO NRs and suggested that the green emission is associated with the trap-assisted recombination of the electrons taken by the adsorbed O₂⁻ ions [14,33]. Extensive spectral analyses performed by earlier studies suggested that the green (∼510 nm) and yellow/orange emissions can arise from optical transitions in the single ionized V_O (∼2.3 eV) and the single negatively charged O_i (∼2.0 eV) [34] inside the ZnO crystals. The formation of these defects are typically associated with either oxygen deficiency or excess oxygen in the ZnO crystals. On the other hand, unlike other emissions, yellow emissions are not significantly influenced by the surface state. It was also found that these various visible emissions from the ZnO crystals are strongly dependent on the surface reactivity and chemical composition at the surface. All these research efforts indicate the same direction that any slight modification in the ZnO surface state can have significant effect on non-radiative transitions at various deep level defects and luminescence emission properties.

 

Fig. 3. PL spectra of (a) CS and N₂O plasma-treated ZnO NRs for (b) 1, (c) 3, and (d) 6 min with deconvoluted visible emission spectra. (e) Comparison of PL emissions from CS and N₂O plasma-treated samples for 1, 3, and 6 min. (f) Energy band diagrams for the ZnO NR crystals in the cases of (left) CS and (right) plasma-treated samples.

Download Full Size | PDF

In our PL analysis, the most significant reduction was observed in green emission when the samples were exposed under N₂O plasma as shown in Fig. 3. On the contrary, UV emission was significantly increased by the plasma treatment, which suggest that the defect concentration near the ZnO surface were reduced consequently.

Underlying effect of plasma treatment can be explained by the chemistry of N₂O plasma. N₂O molecules are very stable at RT; however, N-O bonds of N­₂O in cold plasma, which are relatively weaker than N-N bonds, can be broken easily by producing various plasma radicals such as NO*, N*, and O* [35]. These plasma radicals are highly reactive, and they can have enough energy to initiate reactions with the active sites (V_O, V_Zn, O_i) present in ZnO NRs at low temperature. The origin of these defects can be associated to the growth mechanism of ZnO NRs. Chemical reactions associated with the formation of ZnO crystallites are given by the followings.

In general, ZnO NRs grown by hydrothermal methods tend to have many carbon functional group complexes near the surface, such as hydrocarbons (COOH groups) and carbon oxides (COOR). These impurities have strong electron scavenging properties, therefore they can lead to serious degradation in the near band edge emission (NBE) and increase the trap-assisted emission in each respective wavelength range [22,33]. Since the N₂O plasma radicals can provide enough enthalpy for the reaction to reduce the COOH and COOR groups, the following reaction can take place during the N₂O plasma treatment process.

Furthermore, it can be assumed that many O* radicals in the plasma can reduce the concentration of V_O defects. The activation energy of these defects near the surface is ∼2.3 eV [34] which can easily be overcome by the high-energy plasma radicals. As demonstrated in Figs. 3(a)–3(d), the suppression of visible emission defects is evident as revealed in PL spectra with the reduction of defects causing the radiative recombination with oxygen adsorbates after the plasma treatment. On the other hand, the NBE was significantly increased with the plasma treatment. This UV emission is associated with band-to-band excitonic recombination in ZnO, and its intensity is closely related to the crystalline quality. The reason behind this evolutionary change in this NBE is that shallow-level defects and non-radiative defects near band edge were passivated by the plasma treatment.

Shown in Fig. 4 are the 2θ scan XRD spectra measured from the samples prepared under different plasma treatment conditions. The most intense peak intensities were obtained from (002) reflections (2θ = 34.5°) in every samples, while the negligible or no reflection from (100), (101), and (102) planes were shown. This indicates that the ZnO nano-crystallites have a strong (002) preferred orientation aligned to the vertical direction to the substrate surface. One interesting observation is a prominent reflection from (100) β-phase Zn(OH)₂ at 2θ = 32.8° according to JCPDS files (card No. 24-1444) as shown in logarithmic-scale plot of the CS. In solution-based growth process, a pivotal reaction of $Z{n^{2 + }} + \textrm{ }2O{H^ - } \leftrightarrow \textrm{ }Zn{({OH} )_2} \leftrightarrow ZnO(s )\textrm{ } + \textrm{ }{H_2}O$ can take place from the principal aqueous species of $Z{n^{2 + }}({aq} )$ during the formation of ZnO NRs [21]. Once theses hydroxides are formed, they can be either decomposed into the solution or consumed for the ZnO formation; but, a small amount of colloidal clusters Zn(OH)₂ can remain as precipitates inside the crystals. When the samples were exposed to plasma treatment, the peak intensity associated with hydroxides started to decrease with the increase of (002) intensity of ZnO. This suggests that the residual Zn (OH)₂ starts to dissociate into ZnO crystals in the presence of plasma radicals. The degree of preferred orientation F(hkl) of (hkl) plane can be defined by $F({hkl} )= [{P({hkl} )- {P_0}({hkl} )} ]/[{1 - {P_0}({hkl} )} ]$, where $P({hkl} )= I({hkl} )/\Sigma I({hkl} )$, ${P_0}({hkl} )= {I_0}({hkl} )/\Sigma {I_0}({hkl} )$, $I({hkl} )$ is the intensity of peak measured from (hkl) plane, and ${I_0}({hkl} )$ is the reference peak intensity of (hkl) plane given by JCPDS card number 36-1451 [36–39]. As shown in Table 1, the highest F(002) of ∼0.996 was obtained from the samples plasma-treated for 6 min. This remarkable improvement in F(002) suggests that our plasma treatment promoted the most stable growth mode of ZnO crystal along (002) direction by effectively suppressing the impurities and intrinsic defects present in significant amount caused by the solution-based synthesis process.

 

Fig. 4. XRD 2-theta scan patterns in (a) logarithmic scale and (b) linear scale in intensity (theta = θ).

Download Full Size | PDF

Table 1. Parameters extracted from XRD analysis (PT: plasma treatment).

View Table | View all tables in this article

Shown in Fig. 5(a) is the XPS spectra in a wide binding energy (BE) range of 0-1200 eV measured from each sample. In these spectra, the photoemissions from each core level, spin orbital splittings, and Auger-electron peaks from key elements of Zn and O were observed, while traces of carbon were also found in each sample due to the carbon rich solution-based growth of ZnO NRs. Expanded views of the emission spectra in a BE range of 280-295 eV were shown in Fig. 5(b). Two peaks from C1s were shown at ∼284.5 (major) and ∼288.5 eV (minor), where each peak corresponds to the BE of C-C bonds and oxygen containing functional groups such as COC epoxy, COOH hydroxyl, and COOR groups [40]. A clear trend in decreasing peak intensity with the increase of plasma time was shown from C-1s spectra; therefore, it can be assumed that carbon-related impurities in various forms present near the surface were effectively suppressed by the plasma treatment. As discussed earlier with the chemical reactions in Eqs. (8) and (9), the reduction in carbon-related impurities can be due to surface reaction of N₂O plasma radicals with these impurities. As shown in Table 2, the atomic percentage of C was initially ∼5% in the CS, however it was greatly reduced to ∼2% after the N₂O plasma treatment for 6 min. Figure 5(c) presents the BE spectra of the Zn core electrons from the CS as well as plasma-treated samples. As observed in the spectra, spin-orbital splittings of Zn2p_1/2 (∼1044.28 eV) and Zn2p_3/2 (∼1021.18 eV) were shown with a doublet splitting of 23.1 eV, which represent the different chemical state of Zn as Zn²⁺. More careful observation on the positions of each peak listed in Table 2 reveals that both peaks were shifted toward higher BE by ∼0.2 eV after the plasma treatment for 6 min. These chemical shifts in BE indicate that the interaction of Zn interstitials and stoichiometric Zn with O* radicals produced by N₂O plasma can take place by producing the higher charged states of Zn atoms as reported in the previous studies [20].

 

Fig. 5. XPS spectra of each sample in BE ranges of (a) 0-1200, (b) 280-295, (c) 1020-1050, (d) 534-527 eV (inset shows a magnified view in a BE range of 533.5-529.5 eV). Deconvoluted O1s peaks into three satellites of O_a, O_b, and O_c extracted from (e) CS and N₂O plasma-treated samples for (f) 1, (g) 3, and (h) 6 min. (i) Ratios of ∫O_b/∫O_a and ∫O_c/∫O_a in deconvoluted satellite spectra of O1s versus plasma treatment time.

Download Full Size | PDF

Table 2. Binding energies (BE) for Zn2p doublet, atomic percentage contributions of the key elements, and O_a/Zn ratio in ZnO NRs (PT: plasma treatment).

View Table | View all tables in this article

Figures 5(d)–5(h) illustrate the core levels of O1s in an expanded spectra including three Gaussian deconvoluted peaks of O_a, O_b, and O_c. The main peak O_a (529.88 eV) originates from the binding state of O²⁺ with metal ions such as Zn²⁺ in the ZnO crystals [41]. Because the amount of O_a emission is a good measure of stoichiometric oxygen present in the crystal, we estimated the O stoichiometry by calculating the atomic percentage contribution of O_a in O­1s (%O_a= %O1s× [∫O_a/(∫O_a + ∫O_b + ∫O_c)]) from each sample as summarized in Table 2, where %O1s is the O1s curve integration divided by total curve integration, and ∫O_i (i = a, b, c) is the curve integration of each satellite peak. It was shown that the stoichiometric oxygen percentage was increased to 40.1% in the case of N₂O plasma-treated sample for 6 min compared to 35.7% in the CS. At the same time, the atomic percentage of Zn was also increased from 40.7% of the CS to 41.8% with much improved stoichiometric ratio of O_a to Zn (∼0.96) in the case of the sample plasma-treated for 6 min. This improvement can be attributed to the reduction of non-stoichiometric oxygen including the V_O defects in plasma treated samples caused by the interaction with plasma radicals.

Table 2 also summarized the percentage contribution from each satellite peak of O_b and O_c. O_b is centered at ∼531 eV and known to arise from O₂⁻ ions in the oxygen-deficient regions (where oxygen vacancies are present) of the ZnO matrix [42]. %O_b also exhibited a significant reduction from ∼9.5% of the CS to 4.0% in the case of the sample plasma-treated for 6 min. The third component of O1s peak at higher BE (∼532 eV) expressed by O_c was reported to be associated with either chemisorbed oxygen entity or the species of OH⁻ during the fabrication process in organic solutions [43]. This phenomenon was not fully understood, however it is thought that the increase of %O_c under plasma exposure for 1-3 min increased the chemisorbed oxygen on the ZnO surface under an environment ambient in O-radicals, which can eventually give rise to more band bending and increased surface potential at the surface of ZnO. However, further plasma exposure can activate the surface and suppress the hydroxyl groups present near the surface with the reaction with radicals. A quantitative analysis is also shown in Fig. 5(i) to present the change of our deconvoluted O1s peaks with respect to plasma treatment time. As shown in the plot, the area ratio ∫O_b/∫O_a was significantly decreased with the increase of plasma treatment time from ∼0.28 (in the case of CS) to ∼0.12 (in the sample plasma-treated for 6 min). This trend showed a good agreement with our PL observation, where the green emission was reduced with the plasma treatment, while the band edge emission was enhanced on the contrary. The suppression of green emission can be most likely caused by the reduction of V_O defects on the NR surface [34] due to the irradiation of high energy O* radicals in the plasma. The area ratio of ∫O_c/∫O_a showed an opposite trend with plasma exposure time, which can be due to the enhanced adsorption of oxygen related species as discussed earlier.

Shown in Fig. 6 are the measured photoresponse characteristics of the HEMT-based PDs fabricated with ZnO NRs prepared under different treatment conditions. Figure 6(a) presents drain current (I_ds) versus drain voltage (V_ds) in the dark (I_{ds_DARK}) and UV exposure (I_{ds_PHOTO}) conditions. Measured values of the I_{ds_DARK}, I_{ds_PHOTO}, and on/off ratio (I_{ds_PHOTO}/I_{ds_DARK}) at a V_ds of 5 V were summarized in Table 3. As discussed in the sensing mechanism, the I_ds depends upon the negative surface potential built in the gate area by the adsorption of O₂⁻ ions at the ZnO surface. As discussed in our material characterizations, evolutionary change in surface state and crystalline quality of ZnO NRs was made after plasma treatment with further promoted adsorption of O₂⁻ ions at the surface in the dark condition. For this reason, measured I_{ds_DARK} showed the lowest value of ∼0.59 mA/mm after plasma treatment for 6 min, whereas much higher I_{ds_DARK} of ∼1.91 mA/mm was measured from the CS. The correlation of I_{ds­_DARK} with plasma exposure time can be attributed to the improved surface crystalline state of ZnO NRs, which promotes the adsorption of O₂⁻ ions with the increase in negative surface potential on the NRs. The negative surface potential pushes 2-DEG electrons (present at the interface of GaN/AlGaN) deep into the GaN channel, thereby reducing the concentration of 2-DEG electrons in the channel and resulting in lower I_{ds_DARK} after the plasma treatment. Similarly, I_{ds_PHOTO} was also increased by almost 4 times from the PDs with NRs plasma-treated for 6 min compared to that in the case of untreated NRs, and the improvement is closely related to the improved crystalline quality near the surface of ZnO NRs as revealed in our PL analysis. This supports our assumption that the plasma-treated NRs will have smaller numbers of surface traps contributing to the negative potential barrier in the gate region near the surface [14]. However, as shown in Fig. 6, the effect of plasma treatment for 1 min did not make clear effect on both I_{ds_DARK} and I_{ds_PHOTO} values due to insufficient plasma exposure for the change in NR crystal surface quality. In consequence, lower trap density leads to more reduced negative surface potential with plasma treatment even after the desorption of O₂⁻ ions in the presence of UV light and gives rise to higher I_{ds_PHOTO} as shown in Table  3. On/off ratios measured from the PDs at a V_ds of 5 V were summarized in Table  3. The PDs with NRs plasma-treated for 6 min showed the highest on/off ratio of ∼39, while the PD with untreated NRs exhibited the lowest value of ∼5.

 

Fig. 6. (a) Measured I_ds versus V_ds, (b) I_{ds_PHOTO} versus P_i measured at V_ds of 5 V, (c) spectral responsivity, (d) specific detectivity as functions of radiant light wavelength measured from the PDs, (e) time resolved photoresponse characteristics at V_ds = 8 V, and (f) magnified views with τ_r (rise time) and τ_f (fall time) of transient characteristics upon UV on and off.

Download Full Size | PDF

Table 3. Summary of I_{ds_PHOTO}, I_{ds_DARK}, on/off ratio and g_m,photo measured at V_ds of 5 V from HEMT-based PDs fabricated with untreated and plasma-treated ZnO NRs (1, 3, 6 min) (PT: plasma treatment).

View Table | View all tables in this article

Spectral responsivity R and specific detectivity D* are the vital figures of merit for the PDs as defined in the following equations [1,8,9]

The real time transient characteristics of the PDs were measured at a V_ds of 8 V by using the UV light source of a peak wavelength at 350 nm and a programmable light shutter, as shown in Figs. 6(e) and 6(f). The photocurrents measured from every device examined in this study showed very fast rise-ups upon UV exposure and relatively slower fall-offs with UV turn-off. The photocurrent oscillations showed a fairly good repeatability over 100 cycles from each device. We obtained the rise time (the time for photocurrent to rise up to 90% of maximum saturation value under UV turn-on) of ∼0.7 s and the fall time (the time for photocurrent to fall off by 90% from maximum value under UV turn-off) of ∼3.3 s from the reference device (no plasma-treated), as shown in Fig. 6(f). After 6 min plasma treatment, the rise time was slightly reduced to 0.5 s, however the fall time was increased to 6.2 s. These measured response speeds of our HEMT-based PDs were far faster than those (few tens of seconds) of the ZnO NR-based passive detectors reported to date [45,46].

4. Conclusions

Visible-blind UV PDs were fabricated using the gateless AlGaN/GaN HEMTs with ZnO NRs of which surface structures were modified by N₂O plasma treatment. The best photoresponse characteristics in terms of photocurrent (22.6 mA/mm), dark current (0.59 mA/mm), on/off current ratio (∼38.6), responsivity (∼1.5×10⁵ A/W), and detectivity (D*, ∼ 4.7×10¹³ Jones) were achieved when the NRs were plasma-treated for 6 min. The PL analysis showed the highest intensity in near edge emission and the greatest suppression in defect-related visible emissions upon plasma treatment. The stoichiometry in ZnO crystals also showed a significant improvement after plasma treatment as an oxygen to zinc atomic ratio of ∼0.96 was measured by XPS analysis. The highest (002) degree of orientation (∼0.996) were demonstrated by the XRD investigation from the NRs plasma-treated for 6 min.