1. Introduction

The infrared region beyond the Si bandgap (over 1100 nm), known as short-wavelength infrared (SWIR), has attracted significant attention for its optical communication, spectroscopy, and imaging applications [1–5]. The SWIR photodetectors currently available in the market are mainly based on III-V materials such as InGaAs and InGaAsP. The major obstacles for a wide spread of the SWIR technology are the expensive epitaxial growth cost of III-V materials and the high integration cost of the III-V materials with the Si read-out integrated circuits (ROICs) [6–8]. Therefore, the requirements for the SWIR technology are process compatibility with Si ROICs and a low production cost. The lead sulfide (PbS) semiconductor nanocrystals, known as colloidal quantum dots (QDs), are under intense study as an alternative satisfying those criteria by their bandgap tunability in the SWIR region, high absorption coefficient, low fabrication cost, and the solution processability on Si ROICs [1,5].

The patterning QD is essential for fabricating the arrays of QD photoelectronic devices in the applications of the QD-based image sensors or QD-LED displays [9–14]. Various innovative approaches have been investigated for patterning QDs, including photoactive cross-linker, direct optical lithography of functional inorganic nanomaterials (DOLFIN), transfer-printing, and inkjet-printing [10,15–18]. Those innovative methods have resolution limitations or require particular ligands. The resolution, as refer as pixel dimensions, can be easily improved with the photolithography technique; however, the direct casting photoresist (PR) on top of a QD film has limited because of solvent resistance of the QD films [19], PR and lift-off resist (LOR) [20]. The dissolution of PR and LOR prevents the patterning QD layers with a layer-by-layer ligand exchange procedure, limiting the electrical properties and the thickness of QD layers in photolithography procedures.

In this study, we demonstrated PbS QD-based SWIR array image sensors using conventional lithography techniques using the methanol and TBAI ligands, which retains the patterned LOR from the dissolution. With the customized photolithographic process, thickness could be precisely controlled, and size uniformity of each pixel could be achieved. The fabrication method was reproducible, so every pixel in the 1 ${\times} $ 10 photodiode array was functional and showed the standard deviation under 10% in device performance such as responsivity. The performance of the pixels was evaluated by extracting the responsivity and bandwidth. A prototype 1.3 µm (O-band) image sensor array was demonstrated as an example of practical applications. This stackable QD patterning process could pave the way to a cost-effective SWIR imager.

2. Experimental and methods

2.1. Materials

Lead oxide (PbO, Aldrich, 99.9%), oleic acid (OA, Aldrich, 90%), 1-octadecene (1-ODE, Aldrich, 90%), hexamethydisilathiane ((TMS)₂S, Aldrich), tetrabutylammonium iodide (TBAI, Aldrich, 99.0%), zinc acetate dehydrate (Aldrich, ≥ 98%), 2-methoxyethanol (Aldrich, 99.8%), ethanolamine (Aldrich, ≥ 99.5%), LOR 3A (Microchem), GXR601 (AZ Electronic Materials).

2.2. Synthesis

PbS QD was synthesized by a previously reported method with some modification [21]. 4 mmol of PbO, 13 ml of OA, and 26 ml of 1-ODE were mixed in a 3-neck round bottom flask. After degassing for 15 min at 100 °C, 2 mmol of (TMS)₂S in ODE was swiftly injected into the mother solution at 120 °C. The temperature of the solution was maintained for 2 min for QD growth. After the growth, heating mantle was removed and purification with acetone was done at least three times. Absorbance and photoluminescence spectra of the synthesized PbS QDs are shown in Fig. S1.

Nanocrystalline(nc)-ZnO solution was prepared by sol-gel synthesis [22,23]. 0.5 M of zinc acetate dehydrate with ethanolamine was mixed in with 2-methoxyethanol under a nitrogen atmosphere. After stirring at 80 °C for 2 h, the solution was cooled to room temperature.

2.3. Device fabrication

ITO layer was deposited with an RF sputter on a glass substrate and patterned with HCl etching. After e-beam deposition of Y₂O₃ layer, it was patterned using photolithography. Then, the ZnO solution was spin-coated at 4000 rpm for 30 s and dried at 240 °C for 5 h. The PbS QD solution (30 mg/ml in octane) was spin-cast and TBAI ligand exchange (10 mg/ml in methanol) was performed to form the PbS QD layer. The PbS QD layers were stacked 6 times, layer by layer (LbL), to obtain the desired thickness. The whole PbS QD layer was patterned with PR (GXR 601) and LOR (LOR 3A). Afterward, spin-coating of poly-TPD (8 mg/ml in chlorobenzene) at 2000rpm for 30 s and drying at 100 °C for 30 min was repeated twice. Ag layer was deposited by an e-beam evaporator. The size of each pixel was 200 µm × 200 µm with a pitch of 500 µm. The schematics for PbS QD patterning and TBAI ligand exchange during device fabrication are shown in Fig. 1(a).

 

Fig. 1. Schematic diagram of (a) PbS QD patterning and TBAI ligand exchange, (b) a single pixel structure of photodiode, and (c) the photodiode array. The light source (SWIR, 1310 nm) is incident on the transparent bottom (ITO/glass) side. (d) Cross section of SEM image,the single pixel of PbS QD PD array. The scale bar is 200 nm. (e) Optical images of patterned LOR resist (left) and patterned PbS QD (right). Scale bars are 20 µm.

Download Full Size | PDF

2.4. Characterization

The current-voltage curve was characterized by a semiconductor parameter analyzer (4156C, Agilent). The responsivity was measured with a lock-in amplifier (SR830, SRS), 1310 nm LD (ML725B8F, Thorlabs), LD driver (LDC205C, Thorlabs), a function generator (HP 33120A, Keysight), and a low-noise current preamplifier (SR570, SRS). The temporal response measurement was carried out using the lock-in amplifier and an oscilloscope (TDS2024B, Tektronix).

Ultraviolet photoelectron spectroscopy (UPS) measurement was performed using Nexsa(Thermo Fisher Scientific). He-I line (21.2 eV) was used as the light source for UPS. The work function and the highest occupied molecular orbital (HOMO) level were obtained by analyzing the energy of electrons emitted into the vacuum. The lowest occupied molecular orbital (LUMO) level was obtained by combining the UPS results with absorption and PL spectra.

2.5. Imaging

An imaging system consisting of a 1310 nm LD, a LD driver, a low-noise current preamplifier, a function generator, a switch system (3706A, Keithley), a lock-in amplifier and a photodiode array has been constructed as illustrated in Fig. S2. Digital Micromirror Device (DMD) was used as the subject for imaging.

The PD array was illuminated with a 1310 nm laser reflected from the DMD pattern. The laser operation frequency was 20 Hz and 100 Hz. A current amplifier and a lock-in amplifier were used to measure the photocurrent. A switching system was used to select the PD pixels in sequence. Five measurements were taken per pixel and the mean value of the measured intensity was used. By reading the photocurrents from all the pixels, data for a portion of the DMD pattern was obtained. In order to obtain a 2D image of 10×10 pixels, a line (10 pixels) of the image from the DMD was vertically shifted after each scan.

3. Results and discussion

Figure 1(a) shows a schematic diagram of the lift-off process of PbS QD PD array composed of 1 × 10 pixels. First, LOR and PR were spin-coated on the patterned nc-ZnO/ITO/glass. After UV exposure and development, PR was removed with acetone, but LOR was not. Then, PbS QD solution was spin-cast, and the native OA ligands were exchanged to TBAI ligands as described in the Experimental section. The PbS QD film was immersed with the TBAI solution in MeOH for 30 s, followed by washing three times. LOR did not dissolve in the solvent in this step due to its insolubility. Since the ligand exchanged QD is no longer soluble in the original solvent (octane in our case), another QD layer can be deposited on the ligand exchanged QD film. Through the repetition of this QD deposition and ligand exchange procedures, QD layers were stacked to the desired thickness. The thickness of 1 layer was about ∼10 nm at the center of the pattern. The edge of the QD pattern is thicker due to QDs accumulating at the edge of the LOR during spin-casting due to the meniscus effect (Fig. S3). Finally, the coated QD film was lifted off by LOR. This QD patterning process shows that vertically stacking QD films with conventional PR and LOR can be easily achieved using the insolubility of LOR against MeOH. Using a thin layer of QDs would allow only a small portion of the incident light to be absorbed. However, our method allows multiple layers to be stacked for the desired thickness. A cross-sectional SEM image of PbS QD PD is shown in Fig. 1(d). The pixel consists of ITO/nc-ZnO/PbS QD/poly-TPD/Ag structure. To reduce the leakage current and improve the photocurrent, nc-ZnO was adopted as the electron transport layer/hole blocking layer and poly-TPD as the hole transport layer/electron blocking layer [22,23]. Ultraviolet photoelectron spectroscopy (UPS) measurement was performed to understand the flow of charges generated by light. The electronic band diagram of each layer is described in Fig. S4.

As shown in Fig. 1(e), a smaller pattern was also fabricated to demonstrate high-resolution pattering ability of our method. The patterned QD film with a length of 20 µm was homogenous and crack-free, suggesting that this method can be applied for high-resolution. Inkjet printing with QDs requires ink banks to achieve a narrow pattern, which hinders narrowing pixel size and pitch [15]. However, the patterning process with LOR is able to achieve narrow patterning even without a bank.

To evaluate the photodiodes, the current-voltage (I-V) characteristics under a 1310-nm laser illumination were measured as shown in Fig. 2(a). The photodiode showed rectifying behavior in dark conditions, and the photocurrent and the photovoltage were generated under illumination, as shown in Fig. 2(a). For the incident light intensity of 1.2 µW, the photocurrent increased by about an order of magnitude compared to the dark current at -1 V.

 

Fig. 2. (a) Voltage-current characteristics of the PbS QD photodiode as a function of illumination power illuminated by 1310 nm laser. (b) Responsivity of all photodiode pixels at 0 V. (c) Average responsivity of the pixels at 0 V and -1 V. The error bars correspond to standard deviation. (d) Temporal response of the devices with different applied voltage at 20 Hz.

Download Full Size | PDF

Figure 2(b) shows the responsivity of the photodiode from the photocurrent at different illumination power. Responsivity was extracted from the following equation: R = I_photo /P_in, where I_photo is the photocurrent and P_in is the input optical power. The responsivity decreased as the optical power increased, opposite to the behavior of the photocurrent. This is attributed to the photo-generated carriers in the thin QD film being saturated at high optical power [24,25]. Figure 2(c) shows average responsivity from 10 pixels of the photodiode array with 20 Hz square wave. Under 12-nW illumination, the average responsivity of the pixels at 0 and -1 V were 5.54 ${\times} $ 10⁻³ and 1.20 ${\times} $ 10⁻² A/W with a standard deviation of 9.44% and 9.98%. This confirms the uniformity between each pixel and the reproducibility of the fabrication process. We also calculated the specific detectivity of the device from the responsivity and the noise characteristics, which was 4.75×10⁸ jones (Fig. S5). Moreover, the linear dynamic range (LDR) of the diode is 30 dB, which is calculated from the equation as follows: LDR = 20·log(P_max/P_min), where P_max is the maximum optical power when the photodetector showed a non-linear response and P_min is the minimum optical power that the photodiode can detect (Fig. S6).

To characterize device speed, the temporal response of the photodiode was measured as shown in Fig. 2(d). The rise time (t_rise) was 2.20 ± 0.13 ms, and the fall time (t_fall) was 3.56 ± 0.27 ms using a 10–90% confidence interval. Using this result, the bandwidth calculated from the relation: BW = 0.35/t_r, where BW is the bandwidth, was 159.1 Hz [26]. This was consistent with the 3dB-BW of 165 Hz measured from the frequency response (Fig. S7). It confirms that the PbS QD PD array has enough speed for SWIR imaging [2,27].

To test the potential as an imaging sensor, an imaging setup was constructed as illustrated in Fig. 3(a). A more details are included in the Experimental section and the supporting information (Fig. S2). A collimated 1310-nm laser was used to illuminate the DMD, and the reflected beam from the DMD pattern was projected onto the PD array through a 4-f lens system. Several 10 ${\times} $ 10 pixelated alphabets were displayed on the DMD The photocurrents were measured using a lock-in technique at frequencies of 20 and 100 Hz to exclude the effects of stray light and environmental noise. In addition, the diode array was measured after 5 months from the fabrication, showing good air stability due to the encapsulated structure with poly-TPD/Ag and TBAI ligands passivating the surface of PbS QD with iodine(I^-) ligand, which prevents further oxidation [28].

 

Fig. 3. (a) Schematic diagram of the imaging setup. (b) Target images and imaging results at -1 V. Red circles indicate (3,4), (3,7) pixel, respectively.

Download Full Size | PDF

As shown in Fig. 3(b), all the images were successfully obtained. Successful reconstruction of the images at 100 Hz indicates that this photodiode array is suitable for conventional imaging applications [2,27]. It is worth mentioning that (3, 4) pixel and (3, 7) pixel in the letter K showed clear contrast to neighboring pixels. This result indicates no electrical crosstalk between the pixels and also implies that the QD pixels were well isolated through our patterning process [29].

4. Conclusion

We have fabricated PbS QD photodiode pixel arrays with a patterning method using LOR and TBAI ligand. With this method, a 1 × 10 photodiode array with pixel size of 200 µm × 200 µm and pitch of 500 µm was demonstrated. All the pixels showed uniform responsivity and response time. We demonstrated a prototype photodiode image sensor array to prove its potential as a practical SWIR imaging device. In addition, 20 µm × 20 µm sized QD pattern was fabricated to demonstrate the suitability of our patterning method for high-resolution patterning. We hope that our method can contribute to the development of commercial PbS QD image sensor arrays.