Letter Cite This: J. Phys. Chem. Lett. 2017, 8, 5621-5625
pubs.acs.org/JPCL
Fast Organic Near-Infrared Photodetectors Based on Charge-Transfer Absorption Sascha Ullbrich,* Bernhard Siegmund, Andreas Mischok, Andreas Hofacker, Johannes Benduhn, Donato Spoltore, and Koen Vandewal* Dresden Integrated Center for Applied Physics and Photonic Materials (IAPP) and Institute for Applied Phyics, Technische Universität Dresden, Nöthnitzer Straße 61, 01187 Dresden, Germany S Supporting Information *
ABSTRACT: We present organic near-infrared photodetectors based on the absorption of charge-transfer (CT) states at the zinc-phthalocyanine−C60 interface. By using a resonant optical cavity device architecture, we achieve a narrowband detection, centered around 1060 nm and well below (>200 nm) the optical gap of the neat materials. We measure transient photocurrent responses at wavelengths of 532 and 1064 nm, exciting dominantly the neat materials or the CT state, respectively, and obtain rise and fall times of a few nanoseconds at short circuit, independent of the excitation wavelength. The current transients are modeled with time-dependent drift-diffusion simulations of electrons and holes which reconstruct the photocurrent signal, including capacitance and series resistance effects. The hole mobility of the donor material is identified as the limiting factor for the high-frequency response. With this knowledge, we demonstrate a new device concept, which balances hole and electron extraction times and achieves a cutoff frequency of 68 MHz upon 1064 nm CT excitation.
I
limited regime and to reach cutoff frequencies of up to 24 MHz at zero bias voltage, independent of the excitation wavelength. Moreover, we identify two decay channels in the transient signal, attributed to electron and hole extraction, with the latter being the limiting factor for the high-frequency response. On the basis of these findings, we design planar-mixed heterojunction (PMHJ) detectors with optimized travel distances for electrons and holes and increase the cutoff frequency up to 68 MHz. With this new device concept we demonstrate that fast light detection using organic photodetectors based on direct CT state excitation is possible and does not necessarily require materials with high or well-balanced charge carrier mobilities. Nanosecond NIR Detection Using CT Excitation. The organic photodetectors investigated in this work comprise three important components: the active layer, electron and hole transport layers, and electrodes (Figure 1A). The active layer is a blend comprising the fullerene C60 as electron acceptor and zinc-phthalocyanine (ZnPc) as electron donor. While the two neat absorber materials have negligible absorption at wavelengths greater than 850 nm, the intermolecular ZnPc-C60 CTstate has a significantly red-shifted, but weak absorption.8 Transparent doped layers allow selective transport for either holes or electrons and are furthermore used as optical spacers to achieve wavelength selectivity as described below. Finally, the outer layers, consisting of a semitransparent and a fully reflective silver layer, fulfill two purposes: first, they allow to
nfrared photodetectors are omnipresent and used, for instance, for quality and moisture control in food industries or in biomedical applications for cancer diagnosis or to monitor hemodynamics.1,2 Especially the latter application shows growing interest in wearable solutions. Here, the potentially flexible and lightweight design of near-infrared (NIR) detectors based on organic materials is very promising.3 Besides the synthesis of new molecules with absorption in the NIR,4,5 high detectivity, filterless narrowband organic NIR detectors rely on active layers with thicknesses on the order of micrometers.6,7 Because of the comparably low carrier mobilities of organic materials, such detectors suffer from rather slow response times.7 Potentially faster response times are expected for a recently introduced device concept which is based on the absorption of the intermolecular charge-transfer (CT) state by enhancing its naturally weak absorption8 in a microcavity design.9 When the layer thicknesses are adjusted, the detection wavelength can be gradually changed to achieve a miniature near-infrared spectrometer.10 Knowledge of factors determining the response time of such cavity-enhanced devices is important for possible future applications. Moreover, given reports on the possible role of excess energy on the free charge carrier generation11 and extraction rates,12 the question arises whether there are fundamental differences in the photocurrent transients for above- and subgap excitation. In this work, we study the transient photocurrent signal of cavity-enhanced NIR detectors based on ZnPc and C60 blends by exciting the device with a 25 ps laser pulse, exciting both the CT state and the neat material. Using small area devices and 150 nm thick active layers allows us to measure the fall time of the detectors in the non-RC© XXXX American Chemical Society
Received: September 28, 2017 Accepted: November 2, 2017 Published: November 2, 2017 5621
DOI: 10.1021/acs.jpclett.7b02571 J. Phys. Chem. Lett. 2017, 8, 5621−5625
Letter
The Journal of Physical Chemistry Letters
approximated by a plate capacitor model (see Supporting Information II). We overcome RC-limitations of the measurement setup, using devices with ZnPc:C60 blend layer thickness of 150 nm and an active area of 0.04 mm2. As an example for a non RC-limited device, Figure 2 shows a photocurrent transient for an excitation wavelength of 1064
Figure 1. (A) Simplified schematic of the layer sequence of the cavityenhanced organic BHJ detector. Thicknesses of the hole (HTL) and electron transport layer (ETL) are optimized to enhance the absorption in the NIR. (B) The blue solid line presents the external quantum efficiency (EQE) of the bulk heterojunction (BHJ) photodetector with optimized layer thicknesses to achieve a resonance wavelength at the CT state energy. The dark and light gray lines show the emission [from electroluminescence (EL)] and EQE of a ZnPc:C60 device without cavity-enhanced absorption; dashed lines correspond to fits.14 The energy of the relaxed intermolecular CT state is 1.17 eV, elucidating that the resonance wavelength of the detector is very close to this energy. Vertical red and green solid lines illustrate the excitation wavelengths used for response time measurements at 1064 and 532 nm.
Figure 2. (A) Transient photocurrent (TPC) response of the BHJ detector with a fall time of 8.1 ns upon excitation at 1064 nm. The highlighted area below the curve corresponds to the charge carrier density used to quantify the illumination intensity. (B) Fall time as a function of generated charge carriers upon excitation of the CT-state (red circles: 1064 nm) and above gap excitation (green squares: 532 nm). The filled diamond denotes the measurement of the plot in panel A. The inset shows the circuitry used for drift-diffusion simulations, with RS as effective series resistance and the corresponding countervoltage VS. The solid black line shows the drift-diffusion simulation of the fall time for RS = 200 Ω; dashed black line for RS = 0.
contact the device electrically, and importantly, form a microcavity. Constructive interference for a specific wavelength λcav occurs to a good approximation, if the condition λcav = 2n*dcav/m is fulfilled, with n* being the effective refractive index, dcav the thickness of the material between the reflective electrodes (see Figure 1 A)), and m a natural number describing the resonance order (being 1 in this work). With the aid of optical transfer-matrix simulations13 the layer thicknesses are optimized for maximum enhancement of the CT absorption in the near-infrared. The resulting detectors are selective for a narrow wavelength range, centered around λcav,10 which makes them particularly suitable for applications in which a high wavelength selectivity is required. However, given the different nature of the photogeneration process, with much less excess energy involved, it is unclear whether such detectors based on direct CT state excitation can reach sufficiently high response times. To investigate this in detail, we build a device with a resonance wavelength that is close to the energy of the relaxed ZnPc:C60 CT state of 1.17 eV (see Figure 1B). The energy of the CT state is hereby fitted as described by Vandewal et al.14 and approximately corresponds to the crossing point of the CT emission and absorption spectrum of the blend. The resulting cavity-enhanced device reaches an external quantum efficiency (EQE) of about 4% at 1055 nm, corresponding to a 16-fold enhancement as compared to a device with minimal interference. Note that in the spectral region of stronger CT absorption, EQEs above 20% have been achieved using this concept.9 This device is exposed to 25 ps light pulses at 1064 and 532 nm, exciting exclusively the CT state and neat ZnPc or C60, respectively. The transient photocurrent response is measured with an oscilloscope by connecting the device to its 50 Ω internal resistance. No external bias voltage is applied. By varying both the active area A in the range of 6.44 mm2 down to 2.5 × 10−3 mm2 and the ZnPc:C60 layer thickness d from 20 to 150 nm, we reduce the capacitance, which can be
nm. For applications, the fall time is often reported, being the time needed for the current signal to decrease from 90% to 10% of its maximum value. For the device in Figure 2, the fall time is 8.1 ns. The charge carrier density, generated by the laser pulse, is simply related to the integral of the photocurrent signal.15 We perform transient photocurrent measurements at different intensities and plot the photogenerated charge carrier density versus the fall time, as depicted in Figure 2B. We find a decreasing fall time toward lower carrier densities, saturating at a value slightly below 10 ns. The intensity-dependent fall time is caused by a counter-voltage produced by the photocurrent flowing through the device and measurement resistances. At high intensities, this voltage is sufficiently large to lower the internal field and hamper the carrier extraction, thus slowing the response time.16,17 We find the same trend and similar fall times for an excitation wavelength of 532 nm (see Figure 2B). This result shows the independence of the photocurrent transient on above- and subgap excitation and emphasizes the efficient and fast generation of free charges produced from relaxed CT states.18 Carrier Mobility Imbalance Limits the Response Time. A semilogarithmic plot of the transient signal of the ZnPc:C60 detector (Figure 2A) is shown in Figure 3A. It reveals two distinct decay channels as illustrated by the purple and orange dashed lines. We identify them with the extraction of the two carrier types: holes and electrons. The mobilities of donor and acceptor are very often imbalanced,19,20 causing the carrier types to get extracted within different transit times tr.21 In a simplified picture (following Figure 3C), the carrier distribution is modeled to be spatially homogeneous within the active layer at t = 0. Because of the built-in field, the carriers move toward the electrodes, resulting in a current. In a first approximation, 5622
DOI: 10.1021/acs.jpclett.7b02571 J. Phys. Chem. Lett. 2017, 8, 5621−5625
Letter
The Journal of Physical Chemistry Letters
confirms that the device is neither influenced by the RC time nor by the counter-voltage, and that the mobilities used in the simulations are reasonable. The cutoff frequency fc, defined as the point at which the signal is attenuated by −3 dB, is determined via a Fourier transformation of the temporal current signal (Figure 3B). Here, we achieve a fc of 24 MHz for both direct CT state and above-gap photoexcitation, which is among the fastest for organic photodetectors.23−26 However, these findings also clearly identify the comparably slow extraction of holes to be the limiting parameter for faster response times of this CT-detector.27 In fact, when looking at the cutoff frequency of the device (red line) as compared to the Fourier transform of the electron extraction signal (purple dashed line) (Figure 3), we conclude that the slow hole extraction leads to a loss of about 50% in cutoff frequency. This implies for fast organic detectors, materials with high and balanced electron and hole mobilities are required. While reducing the active layer thickness d of the devices can also lead to higher cutoff frequencies, it would also make them more prone to RC-time limitations. Because of their higher capacitance they would require smaller electrode areas. Detectors with Balanced Travel Distances. A more elegant way to increase the detector speed is therefore to balance the extraction times of both carrier types. Extraction times for electrons and holes can be balanced by shortening the travel distance for the slower carrier. In order to prove this concept, we put a comparably thin ZnPc:C60 blend layer (4 nm) between the neat ZnPc and C60 layers. Upon CT state excitation, charge generation occurs only in this thin blend layer. We then vary the thicknesses of the neat C60 (de) and ZnPc (dh) layers, while keeping the overall thickness of the active layer constant (de + dh = const). The remaining transport layers are adjusted by optical transfer matrix simulations to match the resonance wavelength of 1064 nm. A scheme of the device architecture and the corresponding frequency plot are shown in Figure 4 (details on layer thicknesses and EQE can be found in Supporting Information I). The device with the thickest ZnPc layer (device ①) has the slowest response and lowest cutoff frequency (23 MHz). Moving the charge
Figure 3. (A) Transient current response of the BHJ detector upon excitation at 1064 nm. The semilog plot reveals two distinct decay channels, which are attributed to hole (pink dashed line) and electron (orange dashed line) extraction. The sum of the two reassembles the total current transient (green dashed line). (B) Fourier transform (FT) of the transient current signal (red line) of the BHJ. The −3 dB cutoff frequency is 24 MHz. The dashed lines represent the FT of the single exponential fits for electrons (purple) and holes (orange), with the green solid line as the sum of the two carrier types (doubleexponential fit). (C) Schematic representation of the extraction of electrons and holes and the resulting shape of the current−time signal.
neglecting a spatial broadening of the carrier packages, the current transients will have a triangular shape. Different electron and hole mobilities will cause a delayed triangular current transient for one carrier type.22 The ZnPc:C60 detector indeed shows such a transient signal, and the integral of the separate decay signals yields the carrier densities of holes p and electrons n, respectively. Indeed, we find n and p to be very similar in magnitude, as expected for photogenerated charges, where n = p (see Supporting Information III). Using one-dimensional time-dependent drift-diffusion simulations for electrons and holes we further simulate the fall times dependent on the generation rate. Simulation parameters can be found in Supporting Information V. The counter-voltage Vs over the external resistance Rs of the circuit is determined in an iterative way using the Newton root-finding method to solve Vs + RsI = 0. As shown in Figure 2B (black solid line), there is a very good agreement between the experimental and simulated fall times as a function of carrier density. The simulation reproduces the two decay channels due to the mobility imbalance, which disappears in the case μe = μh (see Supporting Information V).22 At low illumination intensities, i.e., at small carrier densities (
