Ultralong free-standing single crystal perovskite microwires with extremely low dark current

— Super long perovskite microwires (PMWs) are in a great demand in many fields such as low-loss microcables and integrated optical waveguide. Despite decades of research into PMWs, single crystal PMWs with several centimeters long have not been obtained. Here, ultralong (up to 7.6 centimeters) monoclinic crystal structure CH 3 NH 3 PbI 3 ·DMF PMWs have been synthesized. The high-quality microwire exhibits long carrier lifetime of 1775.7 ns. The as-prepared free-standing PMWs can be integrated to any arbitrary substrate and 808 nm near-infrared photodetectors have been successfully demonstrated. The fabricated device shows a high light on/off ratio of 1.79×10 6 and an extremely low dark current of 2.5 fA at 1 V bias. This work provides a strategy for the solution growth of ultralong microwires.


Introduction
Since the perovskite has been widely used in solar cells 1 , there has been a great interest in the production and processing of lead halide perovskites due to their large absorption coefficients 2 , high carrier mobilities and carrier diffusion lengths [3][4][5] . In general, single crystal perovskite nano/microwires with well-defined crystal structures have very low defect levels and unique photoelectricity properties, which are superior to their large single crystal and quantum dot counterparts 6 . For specific applications, the optical and electrical properties of nano/microwires can only be harnessed if long continuous nano/microwires can be synthesized. For example, ultralong microwires can be used as ultralow-loss conducting microcables or the integrated optical waveguide [7][8][9] . Until now, the growth of perovskite nano/microwires is reported by tremendous substrate-dependent methods. Spin/dip-coating precursor on a substrate and then to realize the growth of nano/microwires by anti-solvent or solvent evaporation methods are the most common approaches 10 . However, the unavoidable contamination and unevenness of the substrate can cause excess nucleation sites, which make the growth lack enough driving force and hamper the formation of ultralong perovskite microwires (PMWs) 11 . Besides, templateguided 12 , inkjet printing and roll-to-roll methods may be the promising technologies to achieve the design of ultralong microwires at the desired locations 13,14 , these materials often suffer from low crystal quality, which limited the device performance and the preparation technology is rather complicated. To overcome the above challenges, we introduce a facile substrate-free and low temperature crystallization approach to synthesizing high quality and ultralong (centimeter) CH3NH3PbI3 PMWs, which show a long carrier lifetime of 1775.7 ns. Interestingly, the fabricated photodetection devices made from microwires have an ultralow dark current (2.5 fA) under 1 V bias and a high light on/off ratio of 1.79×10 6 .
Almost all the previous growth methods of single crystal PMWs depend on the substrate by solvent evaporation or antisolvent crystallization ( Table 1). These methods result in the formation of a supersaturated solution, which leads to more crystal nucleus on the substrate and the growth orientation and location of the microwires are always stochastic in nature 15 . Cooling adopted and substrate-free method, by contrast, show a great success in creating ultralong and freestanding CH3NH3PbI3·DMF PMWs, as shown in Fig. 1a. In this synthesis route, the use of the polar aprotic solvent N, N-dimethylformamide (DMF) can effectively promote the anisotropic crystal growth to form a one-dimensional structure 16 . There are three regions about the relationship between supersaturation and spontaneous crystallization, more specifically, undersaturated zone, metastable zone and oversaturated zone. The key of crystal growth is to control the growth temperature and precursor concentration. Thus, the concentration of CH3NH3PbI3 perovskite solution concentration is controlled at 2.2 mol/L, and the growth temperature is controlled at room temperature (RT) to make the solution system enter the oversaturated region (see the supplementary materials for detailed synthesis method). In the oversaturated region, nucleation will spontaneously occur and seed crystals deposit to the bottom of the bottle under gravity. In order to make the growth details more vivid, 2 mL solution is used to synthesize the microwires (Fig. S1). The growth mechanism of the microwires is shown in Fig. 1b. Due to the strong coordination effects between perovskite precursors and solvents, DMF molecules combine with PbI2 to form one-dimensional (1D) structures of PbI2·DMF 17 . Subsequently, PbI2·DMF reacts with CH3NH3I to form interphase-crystals CH3NH3PbI3·DMF 18 , which serve as seed crystals. In the metastable solution system, competitive growth will occur among precipitates when the seed crystals with various sizes are dispersed. This growth derives from the concentration gradients around the crystal particles caused by the thermodynamic behavior, which makes the whole system meets the surface energy is lowered as much as possible 19 . The longer PMWs grow upward at the cost of the dissolvent of smaller ones. To achieve the best crystallinity, there should be no interruption in the process of crystallization. From Fig. 1c, we observed the length of the microwire can be up to 7.6 cm. Time course images are shown in Fig. S2. The growth of ultralong CH3NH3PbI3·DMF microwires can reflect the high crystal quality, because if more defects are generated in the growth process, the periodicity of the lattice structure will be destroyed and the continuous growth will be terminated. Fig. 1d is the photo image of the as-grown single crystal microwires removed from the solution and the CH3NH3PbI3·DMF is degraded into black CH3NH3PbI3 at the surface in air. Subsequently, we take out 50 perovskite microwires and calculated their average diameter and length as 90.08 μm and 78.33 mm, respectively ( Fig. S3 and Fig. S4).
The single-crystal sample only has one set of diffraction pattern. Single crystal X-ray diffraction (XRD) shows that the white-microwires in the bottom have high crystal quality and belong to monoclinic system (Fig. 2a). The total molecular formula of the microwires is C4H13I3N2OPb (CH3NH3PbI3·DMF) and DMF molecule is intercalated between [Pb2I6] 2layer and the CH3NH3 + cation (Fig. S5). When we removed the microwires from DMF solvent, the DMF solvent molecules volatilize from the sample and transform into CH3NH3PbI3.
Transmission electron micrograph image reveals a tetragonal crystal structure of the degraded microwires. The interplanar crystal distances measured from the high-resolution TEM image in  23,24 . The long lifetime component is generally assigned as intrinsic recombination, up to 1.8 μs, indicating very low density of traps, as signposted by a longer lifetime 25 . Subsequently, the lifetime and length of as-prepared microwires and those of previously reported halide perovskite are shown in Fig. 2f (The performance parameters are also listed in Table S1 in the Supporting Information). Notably, the length and lifetime of asprepared microwires are about an order of magnitude larger than those of previously reported results. The longest carrier lifetime demonstrates that the obtained PMWs possess the least nonradiative recombination centers (text S1).
The schematic of the device structure and the crystal structure of the CH3NH3PbI3 are shown in Fig. 3a. For device fabrication, the middle segments of microwires was used and conducting silver paste was coated onto both ends of the PMWs (length: 1.93 mm), constructing an ohmic-type device 26 . Fig. 3b depicts the current versus voltage (I-V) characteristics of the CH3NH3PbI3 PMWs-based device under dark and 808 nm light illumination from 45 μW/cm 2 to 88 mW/cm 2 , respectively. Dark current, which reflects the noise characteristics, is caused by surface or internal defects of the devices as well as harmful impurities (text S2). In this PMWsbased device, the measured dark current is just 2.5 fA at the bias of 1 V, which guarantees the device to operate under very weak optical signals by setting the low noise level of the photodetector. The dark current closely related to the defect state is called trap-assistant-tunnel current and the ultra-low dark current can be attributed to the high crystal quality (text S3) 27 .
Subsequently, visible light of the microscope is applied to verify the stability of the device.
The dynamic stability experiment shows that the photocurrent did not decline significantly at the bias of 10 V (Fig. 3c), which shows relatively high repeatability under repeated light on/off switching cycles. The light illuminated continuously on the device for more than 10 minutes and then was withdrawn. The measurement results are shown in the Fig. 3d. The photocurrent is quite stable and then drop sharply as the light is turned off. It indicates a quick recombination of the carriers inside the materials and no remarkable accumulation of the carriers occurs. We monitored the degradation of the device by recording the photocurrent up to one month and found the photocurrent nearly kept constant in vacuum (Fig. 3e). Self-assembling test system is used for estimating the response speed as shown in the inset picture of Fig. 3f (see the supplementary materials for experimental details). Due to the fast charge generation and recombination, the rise and fall times can be calculated to be 0.18 ms and 0.28 ms, respectively.
The response time is comparable to the value of previous CH3NH3PbI3 perovskite network photodetector and much lower than the audiovisual reaction time (around 100 ms) 28 .
Next, light to dark current ratio is extracted from the I-V curves at the bias of 1 V, as shown in Fig. 3g. Increasing ratios arise from the fact that the conductance of the microwires increases with the increased light intensity. At the light intensity of 88 mW/cm 2 , the on/off ratio can go up to 1.79×10 6 , which is superior to that of the previously reported devices (See Table S2). We also calculate the figures of merit of responsivity (R) and detectivity (D*). Fig. 3h shows the values of R and D* versus the incident light intensity ranging from 45 μW/cm 2 to 88 mW/cm 2 at 1 V bias. Increasing the illumination intensity induces a decreased value, which is related to the increased possibility of carrier scattering and recombination 29 . Significantly, the calculated maximum R and D * can up to 1.94×10 -4 A/W and 4.85×10 10 Jones at the bias of 1 V, respectively.
The relative low responsivity can be attributed to the intrinsic characteristics of the microwires.
In addition, the linear dynamic range (LDR) is calculated to be 125.03 dB, which is comparable with that of Si photodetectors (120 dB) (Fig. S8). Detailed calculations for these parameters are given in the text S4. Fig. 3i shows that as-prepared device has the extremely low dark current among previously reported results. All the data presented here are obtained at the same bias of 1 V (see Table S2 for details). To characterize the device performance of the same microwire at different segments, the start, middle and end segments of the microwire (~ 2 mm) typically used for device fabrication and device performance test are shown is Fig. S9. There were no significant differences of the devices fabricated by choosing different segments. The reasonable differences of device performance may be attributed to electrode preparation and length error caused by interception.
We further demonstrate the uniformity of the devices. Five microwires were taken out from the bottle and bonded to the printed circuit board (PCB) for fabricating the devices. The length is controlled to be approximately 1.24 cm as shown in Fig. 4a. Under visible light illumination (2.33 mW/cm 2 ), all devices exhibit similar photocurrent on the order of 10 -9 A (Fig.   4b). This result confirms that the devices constructed from CH3NH3PbI3 microwires have a satisfactory uniformity. Slight dissimilitude of photocurrent of different devices can be attributed to the electrode preparation and diameter of microwires. In order to promote its flexible applications, another microwire (~1.5 cm) is made into a detector and sealed with PDMS (Fig. 4c). The photocurrent is measured without bending and after several cycles of bending to a curvature radius of 5 mm. Remarkably, it was observed that the device maintains operation status even after bend. The photocurrent of the device has a slightly changing after 300 cycles of bending (Fig. 4d), (see the supplementary text for experimental details) suggesting the flexibility of the device. The decline is mainly from fatigue process of the microwire caused by stress during cyclic loading and the damage once cracks form 30 .
In summary, we developed a unique strategy to synthesize up to 7.6 cm long free-standing perovskite microwires, which indicates the possibility of achieving high-quality PMWs breaking the previous length limitation. The device made from freestanding microwires has an ultralow dark current of 2.5 fA, which reflects the intrinsic characteristics of perovskite. As a result, our work has a significant potential impact on a wide range of electrical and photonic applications including lab-on-a-chip systems, micro cables and microprocessor.

Experimental Section
Materials. CH3NH3I (≥99.5%), PbI2 (>99.99%) and N, N-Dimethylformamide with molecular sieves (DMF, ≥99.9%) were purchased from Xi'an Polymer Light Technology Corp., After the growth completed, the microwires were removed from the bottle. The devices were fabricated with silver paste at both ends of the microwires.

Materials characterization.
To determine the lattice structure, the CH3NH3PbI3 PMWs samples were prepared firstly by using the focused ion beam (FIB) method. The equipment used was FIB precision sample preparation system (FEI Helios) and the prepared FIB slice was perpendicular to the growth direction. Subsequently, the transmission electron microscope (TEM) images and selected area electron diffraction (SAED) patterns were acquired from a TEM analysis system (main unit: FEI Tecnai F20) from FMMedia Technology Co., LTD.. The scanning electron microscopy (SEM) images and energy-dispersive X-ray (EDS) analysis images were obtained based on SEM analysis system (Zeiss Merlin) for observing the morphology and determining elemental composition of the samples, respectively.
Photoluminescence (PL) spectra and time-resolved photoluminescence (TRPL) decays were performed on a microscopic fluorimeter system, where the hybrid perovskite was excited by a 405 nm picosecond diode laser (62.5kHz, pulse duration 50 ps). PL spectra were collected via a spectrometer (Ocean Optics, QE-Pro). TRPL was acquired via a time-correlated single-photon counting (TCSPC) module (PicoHarp 300). Crystal structure of the CH3NH3PbI3·DMF microwire was directly solved by single crystal X-ray diffraction (XtaLAB PRO 007HF) and the software (CRYSALIS pro ). To avoid degradation in air, the white microwires was taken out from the solution rapidly and cleaning the mother liquid with dimethicone. And then the sample with dimethicone is fixed on the test bench with an annular sleeve.
Photo-electrical measurements. The current-voltage and current-time curves were measured using a semiconductor parameter analyzer system (Agilent B1500A and Keithley 4200-SCS) except the dark I-V curve was directly measured using semiconductor parameter analyzer (FS-Pro) from Platform Design Automation. The 808 nm LED was used as the illumination source, and the light power density was measured with a light power meter (Thorlabs PM100D).
The response speed is an important figure of merit. In order to test the response speed of the device, two 4 mm microwires were connected in series with the battery pack (8.4 V), one of which acts as a series resistance and the other as a detection device. The 808 nm LED was connected with the signal generator (SP F20) to provide 1 kHz pulse light for the device. At the same time, the digital oscilloscope (Agilent DSO-X 3024A) was used to record the highfrequency photoelectric signal of the device.
To test the flexibility, the device was encapsulated with PDMS and the part of silver paste was exposed as the electrode. The probe was electrically connected with the electrode, and the photocurrent was measured by the Keithley 2636B. The number of bends was manually controlled with the frequency less than 1 Hz. To ensure that each bend has the same radius of curvature (5 mm), both ends of PDMS are slight contact (Fig. S10).

Supplementary Text
Text S1. Defect properties The defect properties of perovskites analysed by a fluence-dependent steady-state PL intensity measurement is shown in Fig. S11. The spot size is larger than the diameter of microwires. PL quantum yield of halide perovskites depends closely on the competition between radiative recombination (bimolecular) and nonradiative recombination (defect assisted and Auger). That is, with increasing excitation intensity, PL quantum yield increases in the low excitation regime.
We measured the PL intensity as a function of excitation fluence. The PL intensity exhibits a super-linear dependence with increasing excitation fluence in double-logarithmic coordinates, which is ascribed to the increased weight of bimolecular radiative recombination with increasing fluence.
Such a featured super-linear PL is interpreted using the theory of defect state filling effect. PL intensity can be expressed as the total excitation fluence minus nonradiative recombination, as Auger recombination is negligible at the regime of low excitation. The incremental in PL intensity with increasing excitation fluence can be expressed as 1 : where , , are the excitation fluence, detection efficiency and absorption cross section of the samples, respectively; C is a constant and is the filling factor of the defects, which is correlated with the excitation fluence and relaxation rates of the defect states.
The slope of 1.41 is obtained, significantly smaller than that previously obtained in MAPbI3 (e.g., 1.8) and MAPbI3 by CVD (relative lower defect density of 1.6), indicating low defect trapping in the microwires 2,3 .
To further validate the TRPL results, we measured excitation-fluence dependent PL decays of the PMWs in supplement fluence range (9-90 mW/cm 2 ). As shown in Fig. S12, with increasing excitation fluence, the sample exhibits an evident decrease of PL lifetime, which is the dynamic process that defect saturates at higher excitation laser fluence, and bimolecular recombination of charge carriers significantly increases. Note that the defects are easily saturated at this range of low excitation power, therefore further confirmed the low trap densities observed.
Text S2. Noise current The total current noise of the device includes the 1/f noise, shot noise, and thermal noise. The 1/f noise, which mainly origin from the process of carrier capture and de-capture, can be expressed as the following equation: where Nt, If, q, Kb, A, k and T are the trap density, the 1/f noise, the elementary charge, Boltzmann's constant, the device area, the tunneling transmission coefficient, and the device temperature. An effective way to reduce the noise is to improve the crystal quality and thus to decrease the trap density. The shot noise can be expressed as: is=(2qIdB) 1/2 , where Id is the dark current and B is the bandwidth. Besides, the thermal noise is given by it=(4KbTB/R) 1/2 , where R is the resistance of the device. We found that the total noise current can be reduced by increasing the device resistance, reducing the crystal defect and decreasing the device area.
Text S3. Dark current The dark current mainly includes diffusion current, generation-recombination (G-R) current,

trap-assistant-tunnel (T-A-T) current and band-to-band (B-T-B) current, the expression of T-
A-T current is as follows: where Nt is the trap density, me * is the electron effective mass, Eg is the bandgap energy, P is the Kane matrix element, M is the transition matrix element between the trap level and conduction band. Therefore, dark current is intimately correlated to the defect density in halide perovskites. where Ip is the photocurrent, Id is the dark current. At the bias of 1 V, LDR=20×log (4.48×10 -9 /2.51×10 -15 ) =125.03 dB.
Response speed, which including rise time and fall time, is another import figure of merit. Rise time is defined as the time experienced from 10% to 90% peak value, while fall time is defined as the time experienced from 90% to 10% peak value. In this work, the rise time and fall time are calculated to be 0.18 ms and 0.28 ms, respectively. Figure S1. Growth mechanism of free-standing perovskite microwires. a, The front view and b, upward view of the material growth indicates that the microwires begin to grow at the position of the seed crystals. In order to study the growth mechanism of PMWs, 2 mL CH3NH3PbI3 perovskite solution was dropped into the bottle and additional slight shaking can cause the perovskite solution stick to the wall of the bottle. After 3 days, many seeds can be seen at the bottom of the bottle. The nucleation rate can be expressed as follow: where Kb is Boltzmann constant, ΔG is the activation free energy, b is a constant and T is the growth temperature 4 .