Photothermal and Joule-Heating-Induced Negative-Photoconductivity-Based Ultraresponsive and Near-Zero-Biased Copper Selenide Photodetectors

Abstract

The development of a highly responsive, near-zero-biased broadband photo and thermal detector is required for self-powered night vision security, imaging, remote sensing, and space applications. Photothermal-effect-based photodetectors operate on the principle of photothermal heating and can sense radiation from the UV to IR spectral region for broadband photo and thermal detection. This type of photodetector is highly desirable, but few materials have been shown to meet the stringent requirements including broadband optical/thermal absorption with high absorption coefficients, low thermal conductivity, and a large Seebeck coefficient. Here, we demonstrate ultraresponsive, near-zero-biased photodetectors made of mass-producible Cu_2±xSe nanomaterials. Our photodetectors are fabricated with powder pressing and operate on the principle of negative photoconductivity that utilizes the Seebeck effect under the combined effects of Joule and photothermal heating to detect extremely low levels of broadband optical radiation. We show that copper-deficient Cu_1.8Se and selenium-deficient Cu_2.5Se copper selenide materials have negative photoconductivity. However, stochiometric Cu₂Se copper selenide shows positive photoconductivity. We demonstrate that a photodetector made from the Ag:n⁺-Cu_1.8Se:p-Ag:n⁺ system has the best photoresponse and generates a 520 mA/mm negative photocurrent and a high responsivity of 621 A/W under low bias.

Introduction

The conversion of photon energy to an electrical signal is the fundamental mechanism for a wide range of technological applications such as photodetection,^1,2 optical and thermal imaging,³⁻⁵ biomedical sensing and scanning,^6,7 and environmental monitoring.^8,9 Despite recent advancements in high-performance photoresponsive materials and advanced and sophisticated fabrication technologies for photodetection, there is still a lack of low-cost mass-producible materials capable of broad-wavelength detection, a superior optoelectronic response, and a higher speed of operation.¹ Light to electrical energy conversion generally depends on two mechanisms, positive photoconductivity (PPC) and negative photoconductivity (NPC). In conventional PPC-based photodetectors, photocurrent is generated by photons having energy larger than the band gap (ΔE_g) of semiconductor materials.⁹ Therefore, PPC-based photodetectors can detect photons having wavelengths of λ ≤ ΔE_g/hc, where h is Planck’s constant. The PPC photodetectors generally have narrow optical bandwidths, a large dark current, an inability to operate at higher ambient temperature, and a slower operating speed because of their intrinsic operating principle. In contrast to the minority-carrier-operated PPC, less explored NPC is a majority-carrier-dominated photodetection phenomenon in which hot carriers are either trapped by self-assembled monolayers or their mobility is decreased as a result of photothermal-effect-induced phonon scattering.^10,11 In previous NPC demonstrations, hot carriers are generated by exciting light having a wavelength either close to the surface plasmon resonance (SPR) or near the band edge of the semiconductors, thus having narrow bandwidths.^10,11

Very recently, photothermal or Seebeck-effect-induced NPC was observed in two-dimensional materials such as black phosphorus¹² and graphene,¹³⁻¹⁷ where majority charge carriers available in the channel region are depleted by a light-induced temperature gradient. Because of the photothermal dependence, Seebeck-effect-based NPC photodetectors can detect broadband optical and thermal radiation if the photodetectors are fabricated with a material having (i) a high absorption coefficient in a broad spectral range, (ii) a low specific heat, and (iii) a large Seebeck coefficient.^1,18 However, the 2D materials demonstrated in previous NPC have very limited use because of their low absorption coefficient, narrow spectral absorption response, low Seebeck coefficient (tens of μV/K),¹⁸⁻²⁰ unreliability and difficulty in fabrication, and high fabrication costs. Moreover, the extremely small effective active area of 2D materials requires microscopic illumination with precise and costly microfabrication techniques to observe a photoresponse. The material and fabrication limitations render NPC detectors with subpar performance in which they are capable of detecting only high-intensity laser light.¹³⁻²⁰ The search for mass-producible photodetector material with a high photothermal response in a broad spectral range that is also suitable for simple, cost-effective fabrication technology is required for the production of large-active-area photodetectors at low cost.

Recently, copper selenide (Cu_2–xSe) nanomaterials have shown a wide range of applications in thermoelectric and photovoltaic energy harvesting,²¹⁻²³ energy savings,^24,25 light-induced repairable battery electrodes,²⁶ and photoablation therapy.²⁷ Because of their metal-like behavior, copper selenide nanocrystals (NCs) have strong localized surface plasmon resonance (LSPR) absorption tunable by their size, shape, and chemical composition.^28,29 Broad spectral absorption from deep UV to mid-IR, a high Seebeck coefficient, the compositional tunability of electronic, optical, and thermoelectric properties,^30,31 and scalable production methods make copper selenide an ideal candidate for photothermal-effect-induced NPC photodetectors.

Here, we demonstrate a high-performance NPC-based photodetector made of Cu_2–xSe nanomaterial that produces a negative photocurrent of 520 mA/mm with an ultrahigh responsivity of 621 A/W under a low bias. Our NPC photodetector works on the principle of charge carrier depletion from the channel region under the influence of a temperature gradient generated from the combined effects of Joule and photothermal heating. The NPC photodetector produces negative photocurrents from ∼30 to ∼60 μA and from ∼3 to ∼4 μA under bias voltages of 50 and 5 mV s, respectively. Photodetectors made of nonstochiometric copper selenide nanomaterials (Cu_1.8Se and Cu_2.5Se) show an NPC response, while the one made with stochiometric copper selenide (Cu₂Se) shows PPC. This work presents the first demonstration of a Joule- and photothermal-heating-induced NPC-based photodetector using copper chalcogenide nanomaterials. Owing to its high absorption coefficient from the deep UV to far IR spectral range, a high Seebeck coefficient, and efficient thermal/photothermal to electricity conversion under near-zero-bias voltage, copper selenide-based NPC photodetectors can be used as self-powered night vision security and imaging systems, with the detection of radiation from the deep UV to terahertz region for a range of applications.

Results and Discussion

Figure 1a–f illustrates schematics and experimental curves that describe the working principle of photothermal- and Joule-heating-effect-induced NPC photodetection. Figure 1a presents the charge carrier distribution at time t = 0, when there is no thermal or optical radiation (temperature difference ΔT = 0). Under the dark condition (Figure 1a–c), the applied bias voltage (V_bias) Joule thermal effect, Q_JH = (V_bias²/R_eff)t, where R_eff is effective resistance between contact points and t is time, heats the channel region to produce a temperature gradient. Thus, the produced temperature gradient moves the charge carrier from the hotter region to the relatively colder surfaces, which causes a decrease in charge carrier density in the channel region, an increase in resistivity, and hence a decrease in current with time. The temperature gradient creates a charge carrier density gradient of −∂n_c/∂z that produces a thermoelectric field E_TE = −∂V_TE/∂z = S∂T/∂z, where V_TE is the thermoelectric potential and S is the Seebeck coefficient. Under thermal equilibrium (Figure 1b) at t = t_∞, E_TE acts as a retarding field, E_ret, that opposes further flow of charge carriers from the hot side to the cold side, thus setting up a constant current. Figure 1c shows experimental curves for an increase in surface temperature due to Joule heating (blue line) and a decrease in dark current (black curve) with time under V_bias = 50 mV. Figure 1d,e present the charge distribution due to Joule heating and combined effects of the Joule and photothermal heating when the light is turned off and turned on, respectively. When light of power P is turned on, it photothermally heats the sample surface to further increase the temperature gradient and hence decrease the transient current due to additional carrier depletion from the channel region (Figure 1e). However, when the light is turned off (Figure 1d), it relatively cools and thus restores carriers in the channel region, resulting in an increase in the overall current. Figure 1f shows the decrease and increase in the total current when the light is turned on and turned off, respectively. Experimental data presented in Figure 1c,f are from a photodetector made of copper-deficient (Cu_1.8Se) copper selenide nanocrystals.

Figure 1.

Schematics for the decrease in conductivity of the channel region due to (a, b) Joule heating and (d, e) the combined effects of Joule and photothermal heating. (c) Experimental curves for the increase in surface temperature due to Joule heating and the corresponding decrease in the current in the channel region (inset: equivalent resistive circuit of the device) of the Cu_1.8Se photodetector. (d, e) Schematic representation of carrier depletion in light-on and carrier restoration in light-off states. (f) Experimental observation of the decrease and increase in current when the light is turned on and off, respectively, from the Cu_1.8Se photodetector.

To realize a photothermal- and Joule-heating-effect-induced NPC photodetector, we hydrothermally produced copper selenide nanocrystals (Cu_2–xSe) with three different compositions of copper and selenium. The reaction temperature was tuned from 200 to 260 °C at a given concentrations of reactants and time for the reaction to tune copper to selenium (Cu/Se) ratios from 1.8 to 2.5 following our previous work.²⁶ SEM images and EDAX spectra of three samples produced at 200, 230, and 260 °C reaction temperatures are presented in Figure 2a–c. Reaction temperature controls the size, shape, morphology, and composition of as-produced nanocrystals. On the basis of EDAX measurements, one can conclude that nanomaterials produced at 200, 230, and 260 °C reaction temperatures are Cu_1.8Se, Cu₂Se, and Cu_2.5Se, respectively. Here, one can see that nearly stoichiometric copper selenide NCs are produced at a 230 °C reaction temperature while cation- and anion-deficient NCs are produced at 200 and 260 °C reaction temperatures, respectively. Copper selenide (Cu_2±xSe) nanocrystalline powder samples produced under different reaction temperatures are in the α crystalline phase (Figure 2d (JCPDS 47-1448)). The diffraction peak corresponding to the (541) crystalline plane of α-Cu_2±xSe shifts toward a smaller 2θ value with the increase in reaction temperature that demonstrates and supports our EDAX result of an increasing copper to selenium ratio at higher reaction temperatures.²⁶ Diffraction peaks of nanocrystals produced at 230 °C reaction temperature are very close to the positions of stoichiometric copper selenide (Cu₂Se) with a minimum distortion in the lattice (JCPDS 47-1448); however, for higher (260 °C) and lower (200 °C) reaction temperatures, cation to anion ratios are respectively larger and smaller relative to the stoichiometric value. A higher intensity of diffraction peaks of stoichiometric copper selenide NCs (Cu₂Se: red curve) over nonstoichiometric copper selenide (Cu_1.8Se, black curve; Cu_2.5Se, blue curve) shows higher crystallinity and possibly larger electronic and thermal conductivities of Cu₂Se over those of the other two samples. Transmission electron microscopy (TEM) images of Cu_1.8Se, Cu₂Se, and Cu_2.5Se nanocrystals are presented in Figure 2e–g. These images show that copper-deficient (Cu_1.8Se) and stoichiometric (Cu₂Se) copper selenide samples have an abundance of large (>500 nm) single crystals with a high degree of crystallinity. On the other hand, TEM images of Cu_2.5 Se NPs show that larger crystals are made of the self-assembly of a large number of smaller (5–20 nm) particles. The electron diffraction pattern in the inset of Figure 2f shows that the nanocrystal is polycrystalline in nature.

Figure 2.

SEM and EDAX images of (a) Cu_1.8Se, (b) Cu₂Se, and (c) Cu_2.5Se nanopowders. (d) XRD spectra of different as-produced samples. TEM images of (e) Cu1.8Se, (f) Cu2Se, and (g) Cu2.5Se nanocrystals. .

UV–visible–IR absorption spectra of copper selenide samples dispersed in toluene (Figure S1) show absorption maxima near 500 and 1300 nm corresponding to an interband transition and surface plasmon resonance absorption, respectively.^26,29 Absorption maxima near 2200 nm may be due to the phonon vibration mode. A hemispherical absorption spectrum of 100-μm-thick copper selenide powder, cast on the glass slide, is measured from deep-UV to long-wavelength infrared (LWIR) regions (250 nm to 16 μm) showing >90% absorbance in the solar spectral window (Figure 3) with peaks at ∼400 nm and ∼830 nm. Strong broadband absorption in the solid powder may be due to the hybridization of SPR modes and different phonon vibrational modes. The absorption spectrum in the spectral region of 2.7 to 16 μm with >78% absorption and peak maxima in the atmospheric window demonstrates the application of copper selenide nanomaterials in environmental monitoring and thermal sensing.¹⁸

Figure 3.

UV–visible–IR absorption spectrum of Cu_1.8Se powder in solid form from the deep-UV to long-wavelength infrared (LWIR) spectral region. Shaded regions show absorbance in solar (left) and atmospheric (right) windows.

As produced copper selenide nanopowders were pressed at room temperature to make 5 mm × 5 mm × 0.3 mm slabs followed by writing two silver contacts. A schematic of the device and its equivalent circuit is presented in Figure 4a,b. The device was first biased with different biasing voltages, and the temperatures of top and bottom surfaces were measured with two thermocouples. Time graphs for the top and bottom surface temperatures for the Cu_1.8Se sample due to Joule heating under the turn on and turn off of different bias voltages are shown in Figure 4c. Thermal images, recorded with an IR camera, of samples at different times after the application of V_bias = 0.6 V are presented in Figure 4d. Turning the bias voltage on and off correspondingly increases and decreases the temperatures of the top and the bottom surfaces of the sample with time following equations T(t) = T_max – ΔTe^–t/τ and T(t) = T_amb + ΔTe^–t/τ, where T_max is the maximum temperature, ΔT = T_max – T_amb, T_amb is the ambient temperature, and τ is the response time. These equations are used to fit temperature versus time data for the top surface of a device to obtain response times for Joule thermal heating and cooling (Figure S3). The values of response times for heating (τ_heat) and cooling (τ_cool) are 15.7, 11.6, and 11.3 s and 16.5, 12.2, and 10.6 s, respectively, for bias voltages 0.4, 0.5, and 0.6 V, respectively.

Figure 4.

(a) Schematic of a device made of metal (Ag:n⁺)–semiconductor (Cu_2±xSe:n or p)–metal (Ag:n⁺) junctions with (b) a symbolic representation of back-to-back connected Schottky diodes. (c) Increase in the temperature of the device with time due to Joule heating under bias voltages of 0.4, 0.5, and 0.6 V. Temperature with V_bias = 0.4 V measured again to show the reproducibility of the data. (d) IR images for the increasing temperature of the device with time for V_bias = 0.6 V. (e) Thermoelectric potential difference (ΔV_z) generated with time under bias voltages of 0.4 (black), 0.5 (red), and 0.6 V (green) representing the depletion of charge carrier from the channel region and accumulation at the opposite surface due to a temperature gradient (scattered data) and exponential fits for the rise and decay of the potential (solid line) to get response times for charge depletion and restoration. (f) V_bias² versus ΔT (left) and ΔV_z versus ΔT graphs for the determination of thermal conductivity and the Seebeck coefficient of copper selenide nanomaterials. (Inset: schematics for setup to measure the temperatures on the front and back surface of device and the voltage difference.)

The temperature gradient normal to the sample surface pushes charge carriers toward the colder (bottom) surface and generates a potential gradient ΔV_z (Figure 4e) that is proportional to the temperature difference (ΔT). The slopes of linear fits of V_bias² versus ΔT and ΔV_z versus ΔT result for the values of thermal conductivity (k) of 30.15 W K^–1 m^–1 and a Seebeck coefficient (S) of 4.43 ± 0.26 mV K^–1 (Figure 4f and Figure S4) at room temperature. The value of the Seebeck coefficient (4.43 ± 0.26 mV/K) obtained for our copper selenide sample is 10² times larger than the Seebeck coefficients (tens of μV/K) of 2D materials.^19,20 Rise and decay portions of ΔV_z versus time curves are fitted (Figure 4e, solid lines) using equations and , respectively, where (ΔV_z)_max is the maximum value of the potential gradient produced at a given bias voltage and τ_dep and τ_res are time constants for carrier depletion from and carrier restoration into the channel region, respectively. Time constants (response times) for carrier depletion from the channel region are ∼1.2, ∼0.95, and ∼0.76 s for bias voltages of 0.4, 0.5, and 0.6 V, respectively, while carrier restoration (recovery times) times in the channel region are ∼0.99 s for all bias voltages.

To understand the effects of Joule and photothermal heating on the surface temperature of the device and the associated change in conductance of the channel region, a sweeping bias voltage was applied across the contact points. Because Joule heating depends on the value of the applied bias voltage and duration for the application of a bias voltage, different sweeping rates result in different maximum surface temperatures. Bias voltage is swept from +1 to −1 V with different sweeping rates and temperatures for the top surface of the devices made with Cu_1.8Se, and Cu_2.5Se copper selenide samples are measured in the dark (Figure 5a) and under light irradiation (Figure 5b) for three cycles (Figures S5.1–S5.3). At slower sweeping rates, the device surface temperature reaches higher values because of the longer duration of the application of bias voltage (Figure 5c for the dark and Figure 5d under light irradiation). For instance, at 200 mV/s, it takes only 10 s to sweep the voltage from +1 to −1 V, while at a sweeping rate of 25 mV/s the surface will be Joule heated for 80 s. With the irradiation of light (84 mW cm^–2), the surface of the device gets photothermally heated (Figure 5d) and the application of varying bias voltage further increases the surface temperature through Joule heating on top of a photothermally heated background. The I–V measurements were made upon sweeping the bias voltage from +1 to −1 V with a sweeping rate of 25 mV/s in the dark (Figure 5e) and under light irradiation (Figure 5f), and the corresponding change in the device surface temperature is measured (Figure 5g,h). The Joule heating increases the average surface temperature (ΔT_JH) by 1.2 °C (Figure 5g), while photothermal heating under 84 mW/cm² irradiation raises the surface temperature (ΔT_PH) by 1.5 °C over the ambient temperature (Figure 5h). Similar measurements were made for a photodetector made of Cu_2.5Se (Figures S5.1 and S5.2) where ΔT_JH ≈ 0.75 and ΔT_PH ≈ 1.0 °C. The lower values of ΔT_JH and ΔT_PH are due to the higher resistance or lower conductance and lower optical absorbance (Figure S2) of Cu_2.5Se as compared to those of Cu_1.8Se nanomaterials. Moreover, the next cycle in the I–V curve in the dark as well as under light irradiation follows its previous cycle more closely for the Cu_1.8Se sample (Figure S4.3) as compared to Cu_2.5Se, demonstrating its faster recovery time and better stability.

Figure 5.

Schematics for measurements of the cyclic I–V curve and temperature (a) in the dark and (b) under broadband light irradiation. Increase in device temperature with time due to (c) Joule heating and (d) the combined effects of Joule and photothermal heating under the sweeping bias voltage from +1 to −1 with different sweeping rates for three cycles. Time diagram for the applied sweeping bias voltage from +1 to −1 V (sweeping rate 25 mV/s) and corresponding current (e) in the dark and (f) under broadband light irradiation of 84 mW/cm² and (g, h) corresponding changes in device temperature. Black dotted vertical lines show the time for cycle completion (T, 2T, and 3T), and blue dotted vertical lines present times when the current becomes zero, in a positive half cycle, and when the temperature attains its minimum. ΔT_JH and ΔT_Ph represent temperature increase due to Joule and thermal heating, respectively.

Figure 6a–c presents I–V data from photodetectors made of Cu_1.8Se, Cu₂Se and Cu_2.5Se copper selenide nanomaterials in the dark and under a broadband light irradiation of 84 mW/cm². Here, metal (Ag:n⁺)–semiconductor (Cu_2±xSe:n or p)–metal (Ag:n⁺) junctions make back-to-back connected Schottky diodes with n⁺–p–n⁺ or n⁺–n–n⁺ configurations (Figure 4a,b). By engineering the copper to selenium ratio in the Cu_2±xSe nanomaterial, one can easily tune the nature and density of its charge carriers. For example, bulk and stoichiometric copper selenide (Cu₂Se) is a gapless material with metal-like behavior; therefore, it should have high electrical and thermal conductivities with n-type charge carriers.²⁷ However, copper-deficient Cu_2–xSe is an intrinsic p-type semiconductor with direct and indirect band gap energies in the ranges of 2.1–2.3 and 1.2–1.4 eV, respectively.^27,29 Therefore, Cu_1.8Se and Cu_2.5Se should be p-type and n-type in nature with lower electronic and thermal conductivities as compared to the stochiometric copper selenide, Cu₂Se. A photodetector fabricated with the Ag–Cu_1.8Se–Ag system is equivalent to back-to-back connected n⁺–p and p–n⁺ Schottky diodes, while the device made with the Ag–Cu_2.5Se–Ag system is equivalent to n⁺–n and n–n⁺ Schottky diodes. I–V curves of the photodetector (Figure 6a–c) fabricated from nonstochiometric copper selenides (Cu_1.8Se and Cu_2.5Se) show a nonohmic p–n junction characteristic with a zero knee voltage (bottom-right inset Figure 6a, Figure S3). These measurements demonstrate the capability of the ultralow-biased operation of these devices and the Schottky nature of the metal–semiconductor junction. Under light irradiation of 84 mW cm^–2 from a broadband white light source, the device shows a decrease in the overall photocurrent in forward and reverse biases with a >150% increase in device resistance (32.9 Ω/mm in the dark versus 82.4 Ω/mm in the light, Figure S3). A nonlinear increase in the dark (I_D) and the photocurrent (I_P) with the bias voltage shows the generation of secondary charge carriers through avalanche ionization (Figure 6a). In contrast to the photodetectors made of nonstochiometric copper selenides, a device made with Cu₂Se has an ohmic nature (Figure 6b), metallic behavior, of an I–V curve with a small positive photoresponse. Very similar to the photodetector made with copper-deficient (Cu_1.8Se) copper selenide (Figure 6a), a device that is made with selenium-deficient (Cu_2.5Se) copper selenide also shows negative photoconductivity (Figure 6c). Similar conductance values (dI/dV_bias) of different samples in the dark and under light irradiation are determined from the linear fit of 100 mV sections of I–V curves (Figure 6d). For Cu_1.8Se, the conductance and hence carrier density in the channel region increase linearly with bias voltage, demonstrating that the rate of carrier generation through avalanche ionization is proportional to the applied electric field. The photocurrent (ΔI_Ph = I_photo – I_dark) and responsivity (R = ΔI_Ph/(AP), where A is the effective working area and P is the power of the light) of photodetectors made of different copper selenide nanomaterials and their dependence on bias voltage are shown in Figure 6e,f, respectively. At a given light power, the photocurrent and hence responsivity increases with the bias voltage. For the n⁺–p–n⁺ (Ag–Cu_1.8Se–Ag) system, the responsivity increases linearly with bias voltage and attains a value of as high as 618 A/W at V_bias = 0.88 V. However, the corresponding responsivity for the n⁺–n–n⁺ (Ag–Cu_2.5Se–Ag) system is 45 A/W. The higher responsivity of a photodetector made of Cu_1.8Se over one made of Cu_2.5Se may be due to its higher thermoelectric coefficient and higher optical absorbance. The responsivity of both of these devices are much higher than in previously reported works based on nanowires and 2D materials.^11,13−16 Photodetectors made with nonstochiometric copper selenide samples (Cu_1.8Se and Cu_2.5Se) have negative photoconductivity due to their high Seebeck coefficient and hence large thermoelectric effect. However, a device made with Cu₂Se shows positive photoconductivity because of the dominance of the photovoltaic (PV) coefficient over the thermoelectric electric (TE) coefficient.

Figure 6.

I–V curves for (a) Cu_1.8Se, (b) Cu₂Se, and (c) Cu_2.5Se in the dark and under 84 mW/cm² light irradiation. (d) Variation in the conductance of different samples with V_bias under dark and light irradiation. Variation in (e) the photocurrent and (f) responsivity of photodetectors made of different copper selenide samples with applied bias voltage.

Figure 7 represents the photoswitching characteristics of the photodetector made of the Ag–Cu_1.8Se–Ag system under a given light intensity of 84 mW/cm² at different bias voltages (Figure 7a,b and Figure S5) and under varying light intensity at a given V_bias of 100 mV (Figure 7d). At each bias voltage, the total current is the sum of the dark current (I_dark) and the photocurrent (I_photo). Because of the Joule heating, Q_JH = (V_bias²/R_eff)t, of the device surface and associated depletion of the charge carriers from the channel region, I_dark decreases almost linearly with time. However, at a given time, the value of I_dark increases almost linearly with the bias voltage because of the generation of additional charge carriers at higher temperatures through thermal excitation. When light is turned on along with the generation of hot photocarriers through the Joule thermal effect, light also photothermally heats the channel region and further depletes the charge carrier to decrease the overall current. However, when the light is turned off, the channel cools, which restores charge carriers back into the channel region to increase the overall current. Photoswitching curves for the Cu_1.8Se–Ag photodetector at different bias voltages are fitted (Figure S5) following the equations I(t) = I₀ exp(−t/τ_res) – I₀ and I(t) = I₀ – I₀ exp(−t/τ_rec), where I₀ is the value of current at the instant when light is turned on/turned off and τ_res and τ_rec are the response and recovery times of the photodetector. A combination of these two equations fairly fits the photoswitching curves (Figure 8 and Figure S5) at different bias voltages and results in response and recovery times of the photodetector (Table S1). Values of the response (turn on) and recovery (turn off) times are ∼2.92, ∼2.74, and ∼1.68 s and ∼2.69, ∼2.58, and ∼2.14 s under 10, 50, and 100 mV bias voltages, respectively.

Figure 7.

Negative photoswitching characteristic of a photodetector made of Cu_1.8Se with (a) V_bias = 5, 10 mV s and (b) V_bias = 50, 100 mVs under light irradiation of 84 mW/cm². The shaded region presents a time window when the light is turned on. (c) Variations in photocurrent and gain with bias voltage. (d) Photoswitching characteristic of photodetector made of Cu_1.8Se at V_bias = 100 mV with varying light intensity from a broadband light source in the range of 75 to 230 mW/cm², with corresponding (e) photocurrent and responsivity and (f) per watt percentage gain and increase in the resistance of the channel region.

Figure 8.

I–t curve of the photodetector made of Cu_1.8Se with exponential fits for the current increase and decrease when the light is turned off (red curve) and turned on (green curve), respectively, under (a) V_bias = 50 mV s and (b) V_bias = 100 mV. The Intensity of broadband light source is 84 mW/cm².

The photocurrent increases almost linearly with the bias voltage and has values of ∼28 and ∼575 μA/mm per watt at bias voltages of 5 and 100 mV s, respectively (Figure 7c). In the previous NPC work on black phosphorus, the current decreases from 50.5 to 21.2 μA/μm under 14.9 W/cm² laser illumination that results in a <2 μA/μm per watt current change.¹² The device gain, defined as % gain = (ΔI_Ph/I_dark)/PA, is presented in Figure 7c with a blue curve. At higher bias voltages, the device temperature gets increased by Joule heating that causes the thermal generation of additional charge carriers and an increase in the dark current. Though ΔI_Ph and hence responsivity increase with the bias voltage, gain decreases with V_bias > 50 mV due to an increase in the dark current at higher bias voltages. At V_bias = 5 mV, the value of ΔI_Ph is ∼28 μA/mm per watt with a per watt percentage gain of ∼2700%. The maximum observed gain is ∼2800% per watt at V_bias = 50 mV. The responsivity of the photodetector can be increased under the larger bias voltage operation at the expense of the gain. There is a light-intensity-dependent change in the photoconductivity, where light generates additional hot carriers to increase the conductivity through the PV effect on one hand. However, on the other hand, it photothermally heats the surface to either deplete the charge carriers from the channel region (Seebeck effect) or reduce the carrier mobility by enhanced phonon scattering. These competing mechanisms decide the positive or negative nature of the photoresponse of the semiconductor materials. In the case of the photothermal-effect-induced NPC, the carrier depletion mechanism dominates the carrier generation. Figure 7d shows light-intensity-dependent photoswitching at V_bias = 100 mV. At a light intensity of 230 mW/cm², the total current changes from 25.93 to 25.88 mA, giving ΔI_Ph ≈ 531 μA/mm. A linear fit of ΔI_Ph versus light intensity data shows that the photocurrent increases linearly at a rate of 4.36 ± 0.28 (μA mm^–1/mW cm^–2) for light intensity <126 mW/cm² and the rate decreases to 1.32 ± 0.0.065 for higher light intensities (Figure 7e). The responsivity and gain of the photodetector have maximum values of 0.32 A/W and 1225% per watt for a light intensity of 126 mW/cm² (Figure 7e,f). Varying the light intensity from 75 to 230 mW/cm² increases the channel resistance from 22 to 73 mΩ/mm (Figure 7f). Figure 8 shows the fitting of the I(t) versus time curve that results in response and recovery times of ∼2.74 and ∼2.58 s for a 50 mV bias voltage and ∼1.68 and ∼2.14 s for a 100 mV bias voltage. For the detection of weak light, the responsivity of the photodetector can be increased by increasing the bias voltage; however, for the detection of intense light the device should operate at a low bias voltage to avoid the saturation effect. The device responsivity and gain can be further improved by proper thermal management on the back side of device to increase the temperature gradient.

Conclusions

We have demonstrated a highly responsive, per watt high-gain, ultra-low-biased broadband photodetector made of mass-produced copper selenide nanomaterials. The device works on the principle of Joule- and photothermal-heating-induced negative photoconductivity, where intrinsic and photothermally produced charge carriers are depleted from the channel region by a thermoelectric field. The copper to selenium ratio in copper selenide nanomaterials was tuned to control the value and nature of the photoconductivity. Devices made of the Ag:n⁺–Cu_1.8Se:p–Ag:n⁺ system and the Ag:n⁺–Cu_2.5Se:n–Ag:n⁺ system demonstrate excellent negative photoconductance, while that made of the Ag:n⁺–Cu₂Se:p–Ag:n⁺ system shows a very small positive photoresponse. The Ag:n⁺–Cu_1.8Se:p–Ag:n⁺ system showed the best photoresponse and generates a 520 mA/mm negative photocurrent under the light exposure with high responsivity of 621 A/W at low bias. At a bias voltage of as low as 5 mV, the Cu_1.8Se system results in a negative photocurrent of ∼30 μA/mm with a corresponding per watt percentage gain of ∼2675%. Changing the light intensity in the range of 75–230 mW/cm² at a given bias voltage of 100 mV produces negative photocurrent in the range of 180–530 μA/mm, increases the resistance of the channel region from 22 to 73 mΩ/mm, and demonstrates a maximum per watt gain of ∼1225% at 126 mW/cm² light intensity. The generation of the voltage difference (ΔV_z) across the height of the photodetector (Figure 4e) and its correspondence with a time-dependent increase in the temperature difference (Figure 4c) and a decrease in the current through the channel region (Figure 1c) directly demonstrate that negative photoconductivity is due to carrier depletion under the combined effect of Joule and photothermal heating. This work provides a new class of photo and thermal detectors that can be self-powered by ambient energy and continuously detect photo and thermal signals. This work provides further insights to understand the influence of bias voltage and voltage sweeps on the Joule heating of semiconductors for a wealth of energy applications including wearable heaters,³² water desalination,³³ and hybrid Joule and solar thermal water heating and solar and thermal energy storage.

Experimental Section

Synthesis Method

Copper selenide nanomaterials with different compositions of copper and selenium were hydrothermally synthesized following our previous work.²⁶ In a typical synthesis procedure, 1.6 g of CuO, 1.11 g of SeO₂, and 0.5 g of poly(vinylpyrrolidone) (PVP) were added to 36 mL of EG and homogenized by ultrasonic dispersion for 30 min to make the first solution. In a separate glass vessel, 0.8 g of NaOH was dissolved in 100 mL of deionized water to make a 0.2 M solution. Both solutions were transferred to a 150 mL Teflon-lined stainless-steel autoclave and maintained at a constant reaction temperature (200–260 °C) for 24 h, followed by natural cooling. Products were separated by centrifugation, washed two to three times sequentially with water and ethyl alcohol, dried at 60 °C in an air oven, and finally stored in dried and cleaned glass vials for further characterization and applications.

Characterization

As-obtained power was characterized for X-ray diffraction using a PANalytical X-ray diffractometer with a λ = 1.5406 Å line from a Cu Kα source, while the optical absorbance of doubly distilled water-dispersed power was measured over the 200–2500 nm spectral range using a PerkinElmer Lambda-900 double-beam spectrophotometer. A Zeiss Auriga scanning electron microscope was used for surface morphology measurements, while an energy-dispersive X-ray absorption (EDAX) spectrometer attached with SEM was used for the elemental analysis of nanoparticles. Three different spots on the same particle and five different particles of the same sample were used to get the average elemental composition of each sample. A Kratos Ultra DLD X-ray photoelectron spectrometer (XPS) was used to measure the oxidation states of copper in copper selenide.

Device Fabrication and Testing

Devices were fabricated by room-temperature pressing of as-produced copper selenide powers into 5 mm × 5 mm × 0.32 mm tablets followed by making a pair of parallel contacts of ∼1 mm length separated by a 0.1 mm gap on the top surface using silver paste. A back contact was also prepared by making a circular point of ∼0.5 mm diameter to measure the thermoelectric voltage generated by the temperature gradient. Devices were placed on a metallic copper stage for heat dissipation from the back side. Steady as well as sweeping bias voltage was applied, and the current was measured between contact points using a Keithley 4200 source meter. Two K-type thermocouples, connected to a computer through a TC08 (Omega Engineering) data logger, were used to measure the temperatures of top and bottom surfaces of the device with time. Light from a broadband white light source (xenon lamp) with peak intensity at around 400 nm was used to irradiate the photodetector. The intensity of light was measured with a calibrated PV solar cell (PV Measurements Inc., USA).

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsaelm.9b00174.

• Optical absorption spectra of copper selenide nanomaterials in UV–vis-NIR region, X-ray photoelectron spectroscopic measurements, thermal response and I–V curves of the device, temperature measurements with voltage sweeping, cyclic I–V measurements, fitting of the photoswitch curves at different bias voltages, and variations in current and temperature with time under different bias voltages (PDF)

The authors declare no competing financial interest.