All-Printable ZnO Quantum Dots/Graphene van der Waals Heterostructures for Ultrasensitive Detection of Ultraviolet Light Maogang Gong,*,† Qingfeng Liu,*,† Brent Cook,† Bhupal Kattel,† Ti Wang,† Wai-Lun Chan,† Dan Ewing,‡ Matthew Casper,‡ Alex Stramel,‡ and Judy Z. Wu*,† †
Department of Physics and Astronomy, University of Kansas, Lawrence, Kansas 66045, United States Department of Energy’s National Security Campus, Kansas City, Missouri 64147, United States
‡
S Supporting Information *
ABSTRACT: In ZnO quantum dot/graphene heterojunction photodetectors, fabricated by printing quantum dots (QDs) directly on the graphene field-effect transistor (GFET) channel, the combination of the strong quantum confinement in ZnO QDs and the high charge mobility in graphene allows extraordinary quantum efficiency (or photoconductive gain) in visible-blind ultraviolet (UV) detection. Key to the high performance is a clean van der Waals interface to facilitate an efficient charge transfer from ZnO QDs to graphene upon UV illumination. Here, we report a robust ZnO QD surface activation process and demonstrate that a transition from zero to extraordinarily high photoresponsivity of 9.9 × 108 A/W and a photoconductive gain of 3.6 × 109 can be obtained in ZnO QDs/GFET heterojunction photodetectors, as the ZnO QDs surface is systematically engineered using this process. The high figure-ofmerit UV detectivity D* in exceeding 1 × 1014 Jones represents more than 1 order of magnitude improvement over the best reported previously on ZnO nanostructure-based UV detectors. This result not only sheds light on the critical role of the van der Waals interface in affecting the optoelectronic process in ZnO QDs/GFET heterojunction photodetectors but also demonstrates the viability of printing quantum devices of high performance and low cost. KEYWORDS: printable ultraviolet photodetectors, interface, van der Waals heterostructures, nanohybrids, zinc oxide quantum dots, graphene
O
(UV) detection with additional advantages of low cost, flexibility, and printability. ZnO is a direct bandgap semiconductor with bandgap of 3.4 eV. In the form of QDs of high crystallinity and small dimension typically 500.0 9.0 × 10−3
85.1 N/A N/A 2.5 38 66.3 3.6 × 10−3 >1000.0 1.1 × 10−2
10.0 5.6 1.0 × 10−3 8.5 5.0 1.0 1.0 20.0 20.0
9.9 × 108 1 × 107 1 × 104 13.8 ∼6.0 3.0 × 105 113.0 22.7 640.0
3.6 × 109 1 × 108 1 × 104 N/A N/A 1 × 106 385 N/A N/A
this work 12 3 29 30 8 31 32 5
Note: QDs, quantum dots; NWs, nanowires; NRs, nanorods; NPs, nanoparticles; G, graphene.
QDs to graphene. The charge-transfer process is not fully complete before the next pulse arrives at the sample, which indicates that the overall charge-transfer time is comparable to the separation between two laser pulses (40 μs). Moreover, the effects from each pulse accumulate until the saturation occurs. This indicates that (1) the free carriers generated by the charge transfer have a rather long recombination time (as shown below, the recombination time is on the order of 10 s); and (2) the charge transfer ceases after some net charges are built up in the QDs. The overall rising time of the signal is ∼0.3 ms. Therefore, the total energy density (power × rising time) needed to saturate the detector is around 40 μJ/cm2. As we will show, very similar saturation energy density is found when the detector was illuminated by an orders of magnitude less intense continuous-wave light source. Figure 4c shows two on/off cycles of the dynamic photoresponse to UV light pulses. The ratio of the photo-todark current Iphoto/Id is about 1.7. The asymmetric pattern can be clearly seen with a shorter rising time of ∼5.0 s in response to the UV light on, which is considerably shorter than the falling time of ∼85 s in response to the UV light off. The rising time measures approximately the electron−hole recombination time (or carrier lifetime to be discussed later), and the observed rising time of 5.0 s is anticipated from the strong quantum confinement in ZnO QDs. The longer falling time suggests the presence of moderate charge trapping mechanisms. Consider-
Fermi energy at the applied gate voltage. This result, therefore, provides a quantitative evaluation of the tunability of the photoresponsivity of ZnO QDs/GFET hybrid photodetectors using a gate voltage. The shift of the Dirac point observed in Figure 3b upon light illumination implies that electrons are transferred to graphene. To further understand this charge-transfer process, a femtosecond pulse laser source is used to determine the temporal response of the ZnO QDs/GFET. The sample is illuminated by a train of laser pulses with a repetition rate of 25 kHz. The wavelength is centered at 330 nm. The energy density per laser pulse is ∼5 μJ/cm2. During the measurement, the sample is connected in series with a fixed resistor. The total voltage across the sample and the resistor is fixed at 0.1 V. The photoinduced conductivity change in the sample will result in a change in the voltage across the resistor, which is measured by a 200 MHz oscilloscope. The measured voltage is shown in Figure 4a. The laser pulse-train begins to illuminate the sample at t = 0 s, and the intensity of the pulse train is illustrated in Figure 4b. After the light illumination is turned on, the voltage across the resistor drops gradually as a function of time, and the change is saturated within 1 ms. The effect originated from individual pulses can be seen in a magnified view of the curve near t = 0, which is shown as the inset in Figure 4a. The voltage change is most rapid right after the pulses strike the sample, which is presumably induced by the charge transfer from the 4119
DOI: 10.1021/acsnano.7b00805 ACS Nano 2017, 11, 4114−4123
Article
ACS Nano
Figure 5. (a) Photoresponsivity (solid with left scale) and gain (dashed with right scale) as a function of the UV (340 nm) light intensity at different Vsd of 1, 5, and 10 V, respectively. (b) Photoresponsivity (solid with left scale) and gain (dashed with right scale) as a function of the Vsd under different UV (340 nm) light intensities of 2.5, 9.0, and 15 μW/cm2, respectively. The y-axis is the logarithmic scale. (c) Gain of experimental measurement (black dots) and theoretical fitting (red curves) versus illuminating power. The inset of (c) shows the spectra of current noise density of pure GFET (black) and ZnO QDs/GFET (red) device in the logarithmic coordinates. (d) Detectivity D* as a function of the UV (340 nm) light intensity (black axis and solid curves) and Vsd (red axis and dashed curves).
ing the high-quality crystallinity of and lack of band gap states in the ZnO QDs in this work, confirmed in Figure S2e, the charge traps are most probably caused by defects on the surface of ZnO QDs, such as oxygen vacancies.18,19 It should be pointed out that the response times in our ZnO QDs/GFET heterojunction devices are consistent to the slightly shorter ones reported on single crystalline ZnO nanowire UV detectors (rising time of ∼1 s and falling time of ∼3 s) due to weaker quantum confinement effect in the nanowires with considerably larger dimension.20,21 The similar asymmetric photoresponse in the nanowire case indicates that the surface defects on ZnO nanostructures indeed play the dominant role of charge trapping. The spectral responsivity shows a consistent trend with the absorption spectrum of the ZnO QDs casted on a glass substrate in Figure 4d with a UV band edge at about 345 nm. In particular, despite the high responsivity of 1.8 × 108 A/W at 340 nm, it decreases by about 2 orders of magnitude or more at 400 nm or longer wavelengths. This confirms the absence of the band gap states in the highly crystalline ZnO QDs obtained for this work, and the ZnO QDs are responsible for the UV absorption in these ZnO QDs/GFET heterojunction devices with extraordinary spectral selectivity. The Rv value is 2.6 × 108 and 1.8 × 108 A/W under 310 and 340 nm UV illuminations, respectively, at Vsd = 10 V. As shown in Table 1 that compares the Rv values and other relevant parameters of ZnO nanostructure/graphene UV photodetectors, the Rv values of the ZnO QDs/GFET devices in this work are about 2 orders of magnitude greater than the best reported on the ZnO nanostructure/graphene devices. This difference illustrates the importance of controlling the surface of ZnO nanostructures, as we discussed in Figures 1 and 2, since the photoresponse could be sensitively affected by orders of magnitude if the charge transfer is blocked by an undesired interface layer across the
ZnO QDs/graphene heterojunction. If this layer is removed, an extraordinary Rv, anticipated from the high photoconductive gain, can be achieved in the ZnO QDs/GFET heterostructure photodetectors. The photoconductive gain is typically determined by the ratio of the carrier lifetime in QDs (Tlife) and the carrier transit time in the graphene channel (Ttransit). At a low intensity of illumination, all of the holes generated from photoinduced electron−hole pairs will occupy the surface states of ZnO QDs. These holes are trapped at the surface and complete the recombination with the negative charged oxygen ions. The photoconductive gain can be expressed as12 G=
Iphhv qiWL
= (a0qNVW )
T hv 1 = a0ϕ life qiWL Ttransit 1 + i i
n
() 0
(3)
where Iph is the photocurrent, q is the elementary charge, W is the graphene channel width, a0 is the photogenerated carriertransfer efficiency from QDs to the graphene channel, ϕ is the quantum efficiency of ZnO QDs, i0 is the saturation light intensity on the QDs when all of the surface states are filled, i is the illuminating light intensity, n is a phenomenological fitting parameter, and V = μVsdL−1 is the carrier drift velocity in graphene with the channel length L, the carrier mobility μ, and −1 source−drain voltage Vsd. The Ttransit = L2μ−1Vsd is the carrier transit time in the graphene channel. For the ZnO QDs/GFET heterostructure photodetectors, with W = 13.2 μm, L = 7.0 μm, Vsd = 10.0 V, and μ ∼ 53.6 cm2 V−1 s−1, the Ttransit is estimated to be ∼0.9 ns. The lifetime of photogenerated holes can be extracted from an approximation fitted by a biexponential function in Figure 4c, which yields Tlife ∼ 10.2 s.1,22,23 This indicates the lifetime of the holes (10.2 s) in the ZnO QDs is over 1 × 1010 orders of magnitude greater than the electron 4120
DOI: 10.1021/acsnano.7b00805 ACS Nano 2017, 11, 4114−4123
Article
ACS Nano
noise current and can be calculated from the spectra density of the noise power. The in2 of the hybrid ZnO QDs/GFET devices monotonically decreases with increasing frequency, which can be fitted by in2 ∝ 1/f in the low frequency range up to kHz, demonstrating that the 1/f noise determines the current noise behavior at low frequencies (the inset of Figure 5c). From the noise power density curves, the NEP can be calculated as ∼1.3 × 10−18 W/Hz1/2, and Figure 5d illustrates the calculated detectivity D* at different light intensities and Vsd values. At Vsd = 10.0 V, the obtained D* exceeds 1 × 1014 Jones at 340 nm at room temperature. The best D* value of 7.5 × 1014 Jones is 1 order of magnitude higher than the previously reported highest value on the ZnO/self-assembled monolayer/graphene UV photodetector12 as well as that for the PbS QD/GFET infrared detectors.1
transit time in the graphene channel (0.9 ns). This means that a very short transit time of the electrons in the GFET channel compared with the longer lifetime of the holes (Ttransit ≪ Tlife) will lead to a higher gain for the photodetector device. At the low power density limit, the a0ϕ ∼ 1,12 therefore eq 3 yields a gain ∼1 × 1010. To our knowledge, it is the highest gain that has been reported in the literature for UV detection. It represents more than an order of magnitude enhancement over the highest value recently reported on UV photodetectors using an optimized self-assembled monolayer between ZnO QDs and GFET for improvement of the interface.12 The higher gain obtained in the ZnO QDs/GFET heterostructure photodetectors suggests that the van der Waals interface can be more efficient for charge transfer. It should be noted that the photoconductive gain has contributions from both the ZnO QD surface electron depletion effect and the high mobility of graphene. In a recent work, we made a comparative study on the optoelectronic properties of ZnO nanoparticles fused through nanojunctions and the similar ZnO nanoparticles on graphene channels of dimension of 1−2 mm (width) × 0.3−0.4 mm (length).24 In both cases, the dimension of the ZnO nanoparticles is comparable to the Debye length for optimal surface electron depletion effect. Interestingly, the photoconductive gain in the former (contribution of the surface electron depletion effect) was found to be on the order of 1500, which means a higher gain exceeding 1 × 107 is attributed to the high charge mobility of the graphene in the ZnO QDs/ GFET devices. Figure 5a depicts the photoresponsivity (solid) and gain (dashed) versus light intensity curves measured on the ZnO QDs/GFET heterojunction devices at different Vsd of 1, 5, and 10 V, respectively. Figure 5b shows the photoresponsivity and gain dependent on the Vsd under different light intensities of 2.5, 9.0, and 15 μW/cm2, respectively. It should be noted that the photoresponsivity increases monotonically with the bias voltage (Vsd) of a photoconductor (Figure S3), while higher signal-to-noise ratio is observed typically at a larger bias. Figure 5c demonstrates the gain of our experimental measurements and theoretical fitting as a function of the illuminating power. A monotonic decrease of gain with increasing UV illumination power is attributed to the saturation of the surface states in QDs and reduced average lifetime of holes, and the maximum gain of 3.6 × 109 was obtained experimentally for an illuminating power of 2.3 pW at 10 V bias voltage. The red curve is the best fit for experimental gain with fitting parameters: a0ϕ =0.82, i0 = 1.5 pW, and n = 1.15. The theoretical calculation predicts the gain can be reached as high as ∼9.2 × 109 with an illuminating power of
