A colloidal quantum dot infrared photodetector and its use for intraband detection

Wavefunction engineering using intraband transition is the most versatile strategy for the design of infrared devices. To date, this strategy is nevertheless limited to epitaxially grown semiconductors, which lead to prohibitive costs for many applications. Meanwhile, colloidal nanocrystals have gained a high level of maturity from a material perspective and now achieve a broad spectral tunability. Here, we demonstrate that the energy landscape of quantum well and quantum dot infrared photodetectors can be mimicked from a mixture of mercury selenide and mercury telluride nanocrystals. This metamaterial combines intraband absorption with enhanced transport properties (i.e. low dark current, fast time response and large thermal activation energy). We also integrate this material into a photodiode with the highest infrared detection performances reported for an intraband-based nanocrystal device. This work demonstrates that the concept of wavefunction engineering at the device scale can now be applied for the design of complex colloidal nanocrystal-based devices.


Supplementary Figures
Supplementary Figure 1 Transmission Electron Microscopy (TEM) images of the nanomaterials used in this study. a. HgTe 6k nanocrystals, with bandgap around 6000 cm -1 (740 meV). Scale bar: 50 nm. b. HgTe 4k nanocrystals, with bandgap around 4000 cm -1 (500 meV). Scale bar: 20 nm c. HgTe 3k nanocrystals, with bandgap around 3000 cm -1 (370 meV). Scale bar: 20 nm. d. HgSe nanocrystals with intraband absorption around 2500 cm -1 (300 meV). Scale bar: 10 nm. x HgSe ) propagation. c. Absorbance ratio between HgTe 4k (interband, taken around 4000 cm -1 ) and HgSe intraband (taken around 2500 cm -1 ) as a function of the NC ratio Sample are kept in a cryostat and illumination is provided by a 4.4 µm QCL. For the determination of I photo /I dark , the Quantum Cascade Laser (QCL) is chopped at 1 Hz and current is directly acquired through the source-meter (left graph). For the photoresponse dynamics, QCL is chopped at 100 Hz and the current is acquired with an oscilloscope through a transimpedance amplifier (right graph). b. Control experiment on pure HgTe 4k sample, showing no response to mid-infrared illumination compared to a 35% HgSe hydrid device. c. I photo /I dark ratios for several HgSe / HgTe 6k fractions at 50 K, showing a reduction of the modulation for low HgSe fractions due to low absorption (i.e. the loss of absorption is no longer balanced by the decrease of dark current).

Supplementary Figure 5
Field Effect Transistor (FET) data for several HgSe ratios. a. Scheme of the electrolytic FET geometry. CQD: Colloidal Quantum Dots. b-e. FET transfer curves for pure HgSe (b), 90% HgSe (c), 35% HgSe (d) and pure HgTe 4k (e) films. All films are ligand-exchanged toward EDT and measurement is performed in air. Gate leak currents are given in black. f. Hole/electron mobility ratio (black) and neutral point (gate voltage corresponding to the minimum drain current, in red) for all the studied HgSe fractions. Error bars are determined by the hysteresis of the measurement and variations on several cycles. Figure 6 Thickness dependence of the photoconductive device's performances. a. I-V characteristics of three devices with increasing thickness at room temperature. b. Normalized cooling curves of the three devices with associated Arrhenius fits. c. Extracted device resistances (under 1V bias, at 300K) and activation energies. Vertical error bars are set to 30 meV, the typical repeatability of the measurement. d. Photoresponse of the three devices to a 2.9 mW Quantum Cascade Laser (QCL) illumination chopped at 97 Hz. Devices are cooled at 120 K. e. Spectral current density of the three devices at 120 K, in the dark. f. Evolution of the photoresponse (under 2.9 mW QCL illumination) and detectivity with temperature, for the three devices.

Supplementary Note 1: X-Ray Photoemission
XPS measurements have been conducted on the TEMPO beamline of Synchrotron SOLEIL, under ultra-high vacuum. The formula we used to determine binding energies (BE) is ℎ exp = + where KE is the kinetic energy of the electron measured at the analyser, BE is its binding energy towards Fermi level and ℎ exp is the photon energy.

Sample preparation.
A silicon wafer is cleaned before 5 nm of chromium and 80 nm of gold are deposited by thermal evaporation. 8 mm x 8 mm pieces are cut and each substrate is drop-cast with HgTe/HgSe NCs in hexane/octane solution. Ligand exchange is performed using the same procedure as for transport devices, then the sample is degassed under ultra-high vacuum (10 -10 mbar). The sample is introduced in the analysis chamber, where photoelectrons are detected at normal emission from the polarization vector E.
Photon energy. Photon energy is set to 150 eV, but ℎ has to be precisely measured. To do so, we look at the Hg 4f core levels of our material at the first and second order. Photon energy is then determined as ℎ = 2 − 1 where KE2nd is the kinetic energy of Hg4f7/2 core level measured at second order and KE1st is the kinetic energy of Hg4f7/2 core level measured at first order.

Spectra calibration.
We use a monocrystalline gold sample to calibrate the analyser, since its work function changes with the pass energy used. We measure the Fermi level of the gold sample as well as the photon energy for all the pass energies needed. We then set this Fermi level to zero as the sample is metallic and get the correction value to be applied to every spectrum.
Overview spectra. We take large scale spectra (overviews) in order to confirm the presence of mercury, selenium and tellurium in our samples. Spectra are acquired under 150 eV illumination with a step of 10 meV. All the observed peaks can be attributed to Hg, Se and Te core levels, see Supplementary Figure 2a. The measured kinetic energy is converted into binding energy using: = ℎ − .

Valence band measurement.
We determine the value of VB -EF (energy difference from VB to Fermi level) by using 150 eV photon energy and looking at high kinetic energies electrons. We determine ℎ using the procedure described above. We measure KEVB as the highest kinetic energy available and use the following formula to extract VB-EF : The measured value is a mean value of Fermi level to valence band energy distance in our effective medium, see Supplementary Figure 2b where we reconstruct the energy diagram of each material using XPS measurement in combination with optical spectroscopy.

Supplementary Note 2: Field Effect Transistor measurements
Electrolyte preparation: The electrolyte is prepared in a nitrogen-free glove-box. 0.5 g of LiClO4 are mixed with 2.3 g of PEG (MW = 6 kg.mol −1 ). The resulting mix is heated at 170 °C for 2 h until the solution turns clear, then cooled down overnight before use.

Supplementary Note 3: Thickness dependence of photoconductive devices
Samples have been prepared according to the procedure we report in the paper (deposition + solid state ligand exchange toward EDT), with 2 layers for the thin device (10-20 nm), 4 layers for the intermediate one (50-100 nm) and 7 layers for the thick one (150-200 nm). We use a 4.4 µm QCL to measure the devices responsivities.
In Supplementary Figure 6a and b we provided the I-V characteristics of the three samples at room temperature as well as their cooling curve. Device resistance and activation energy are extracted form S14 previous data and are presented in Supplementary Figure 6c. The device resistance increases with thickness, which is expected for transport in disordered arrays of nanocrystals since thicker films come with a greater number of 3D percolation paths. Activation energy, on the other hand, is rather similar for the three devices as it only depends on the HgSe proportion in the initial solution. The responsivity of the three devices have then been measured between 80 K and 120 K using a 4.4 µm QCL chopped at 100 Hz, as well as the associated noise spectral density, see S15 Supplementary Figure 6d and e. There is a clear increase in responsivity with thickness (mostly due to higher absorption), when the noise level of the detector stays mostly unaffected as long as the NC film is of good quality (i.e. no cracks, no rough spot).
Supplementary Figure 6f presents the evolution of responsivity and detectivity with temperature.
However, when it seems clear that film thickness should be maximized to get the highest detectivity in our devices, one should remember that when embedding such a film into a vertical device (such as a diode), the film quality is of utmost importance. Indeed, in order to prevent shortcuts and/or excessive noise, the NC film have to be extremely smooth, homogeneous over a large area and pinholefree. Hence, there is a trade-off between thickness and film quality.

Supplementary Note 4: Al grid electrode
Sapphire substrates (2x2 cm²) are sonicated in acetone and cleaned with acetone and isopropanol before being exposed to an oxygen plasma for 5 min. Adhesion promoter (TI Prime) is spin coated on the substrate and annealed at 120 °C for 2 min. Then, the UV sensitive photoresist is spin coated on the substrate and is baked at 110 °C for 90 s, then a MJB4 mask aligner is used to expose the slides to UV light through a quartz mask (1.5 s exposition). After a 125 °C bake for 2 min and a 40 s flood exposure, substrates are developed using AZ726 developer for 20 s and rinsed with distilled water. After a 5 min plasma cleaning, 80 nm of aluminum is deposited using a thermal evaporator. Liftoff is conducted overnight in an acetone bath. The electrode is a U-shaped, 1 mm thick line patterned with a grid to allow light transmission, see Supplementary Figure 7. The overall transmission of this electrode is around 70% in the target range of wavelengths.

Supplementary Note 5: Light polarization effect
We investigated the effect of polarization on the diode's responsivity, see Supplementary  Figure 8. A horizontally-polarized 1.55 µm laser diode is used to illuminate the sample kept in a cryostat at 80 K. A /2 waveplate is used to rotate the light polarization and the photocurrent is measured using a Keithley K2634b source-meter. Illumination power has been measured for each polarization to ensure that any observed effect is not due to a change of incident power. Power variations stay below 3%.
While there is a slight effect of polarization orientation on the device responsivity the relative variation of around 5% is well below what is expected for traditional epitaxial QWIP devices, where the selection rules strongly depend on the incident light polarization. Since the illumination is provided from the back of the diode (through the Al mesh electrode), we believe that those slight variations might be due to diffraction effects through the fine digits of the Al electrode, see Supplementary Figure  7.

Supplementary Note 6: Limitations of the HgTe 6k barrier and device stability under bias
With a 200 meV offset between HgTe 4k and HgTe 6k and assuming a 30 nm-thick barrier, the critical electrical field for the barrier to become triangular is = 70 kV/cm. Assuming a 250 nm total thickness (HgTe 6k layer + HgSe/HgTe 4k layer), this critical field is reached under 1.75 V bias. To prevent any damage to the diodes and to ensure the correct operation of the HgTe 6k unipolar barrier, those didoes need to be operated in the -1 to 1 V ranges of biases. Bias voltages above 2 V will eventually lead to electrical shorts.

Supplementary Note 7: Detectivity measurements
The device is enclosed in a closed-cycle cryostat and cooled down to the desired temperature. Supplementary Figure 9 provides the experiments schematics. Illumination is provided from the backside of the device through two ZnSe windows (one on the outer cryostat enclosure and one on the shield) and through a hole in the sample holder. The source is an Omega BB-4A calibrated blackbody operating at 650 °C and placed 30 cm away from the sample. The flux coming from the blackbody is filtered with a Ge window (1.9 µm cut-off high pass filter) and chopped at 1 kHz. Photocurrent from the sample is measured using a Zurich Instruments MFLI lock-in amplifier which also serves as a bias source.
Noise measurement: device is biased using a 1.5 V battery and a potentiometer. Current from the device is amplified by a Femto DLPCA-200, then fed into a SRS SR780 signal analyzer. A homemade LabView program is used to measure the spectral current density as a function of signal frequency. Value of the current spectral density at 1 kHz is used for the detectivity calculation.

Supplementary Note 8: Estimation of the intraband diode NETD
Because ultimately this type of detector is designed to make thermal imaging, we have tested the potential of our diode for detection of a cold object (i.e. a hand which surface is around 32.5 °C) in the room environment (T=20 °C).
As can be seen on Supplementary Figure 10a, the blackbody emission of a 32 °C object is mostly negligible in the interband part of the device spectrum and only the intraband part of the spectrum shows a relevant contribution. Supplementary Figure 10b shows the effect of this hand (cold black body) in front of the device, producing a photocurrent of ℎ = 0.4 nA. Assuming that the half-angle of view of the detector is defined by the cryostat windows, we have = 14°. Hence the blackbody formula gives: Where is the transmission factor through the cryostat windows (55%), the detector area, and ( , ) the spectra radiance of the black body at 32.5 °C. The associated responsivity is then = ℎ ⁄ = 5.2 mA/W Which is a decent fraction of the responsivity measured with the broadband, 650°C blackbody.
The same way, one can compute the NETD of the detector at 80 K with this experiment. Assuming that this photocurrent corresponds to a temperature difference Δ = 12.5 (32.5 °C hand vs 20 °C background) we have: Where Δ is the temperature difference, Δ ℎ the associated photocurrent signal and the RMS noise of the device. We have Δ = 12.5 and Δ ℎ = 0.4 , and the noise spectral density is measured to be 30 fA/Hz 1/2 under 1 V at 80 K and 1 kHz. Hence with a 1 kHz bandwidth, we have = 0.9 and = 35 .

Supplementary Note 9: MHz photoresponse of the diode device
The PV device enclosed in a cryostat kept at 80 K, and illumination is provided by a 4.4 µm pulsed QCL. The duration of the QCL pulse is set from 500 ns to 5 µs while keeping the duty cycle to 10%. Bias is provided by a 1.5 V battery equipped with a potentiometer, and current is amplified by a Femto DUPVA 70 GHz amplifier over a 1 kΩ load. Temporal trace is acquired on a Tektronix MDO3102 S18 oscilloscope. Supplementary Figure 11 shows a scheme of the setup as well as temporal traces of MHz QCL pulses resolved by the diode device at 80 K.

Supplementary Note 10: Stability of the diode device
Devices are kept under air-free atmosphere and exposed to air only when being transferred into a cryostat. I-V curves acquired in the dark on a 3-month old sample in the dark shows noticeable change, see Supplementary Figure 12a, mostly a decrease of dark current on the negative bias side.
The photocurrent spectrum of the aged device under zero bias and 80 K (Supplementary Figure 12b) still shows photocurrent contributions from HgTe 6k, HgTe 4k (though fairly reduced compared to older measurements, see Figure 5a) and, more importantly, HgSe in the mid-infrared.