The emerging organic–inorganic hybrid perovskites, being regarded as one of the most promising semiconductor materials, appear to possess advantages of both inorganic and organic semiconductors [1–9]. The hybrid components enable the perovskites to exhibit excellent electronic characteristics, such as high carrier mobility, long carrier lifetime, and low trap density, just like inorganic semiconductors [10–12]. In addition, hybrid perovskite semiconductors also display such features as adjustable bandgap and low-temperature processing, similar to their organic counterparts [13–17]. Ascribed to the unique features of hybridization, perovskites have already shown bright prospects in the fields of solar cells [18–27], light-emitting diodes (LEDs) [28–31], and lasers [32–35]. Along the same lines, research on perovskite photodetectors has also attracted widespread interest [2,36–40]. Various perovskites photodetectors including single-crystal, thin-film [41–46], and quantum dot [47–52] perovskite photodetectors have been successfully demonstrated. The figure of merit that characterizes a photodetector’s ability to detect weak-light irradiation is the specific detectivity (D*). Only high-quality semiconductor crystals can suppress the dark current, resulting in high-D* photodetection performances.

Perovskite photodetectors can be categorized into different groups according to the crystalline properties of the perovskite layer. Generally, single-crystalline perovskite bulk material, wafers, and films are superior to their polycrystalline counterparts in terms of lower carrier recombination loss and longer carrier lifetime because of a reduced density of grain boundaries and defects. Therefore, single-crystalline-based perovskite photodetectors show advantages of lower dark current, promoting the D* as well as the linear dynamic range (LDR), which is crucial for weak-signal photodetection [53–58]. Differing from the frequently investigated bulk materials or wafers, low-dimensional perovskite nanostructures such as nanowires and nanoplatelets allow the manipulation of light at the nanoscale via naturally formed boundaries [59–62], and they also meet the need for good flexibility [63–69]. A series of different solution-processing techniques has been applied for preparing single-crystalline perovskite nanowires, based on which high-performance photodetectors have been demonstrated [63,64,67–80]. In 2016, by the blade-coating method, Jie’s group achieved ordered MAPbI3 wires of the micrometer scale and demonstrated photodetectors with a D* of 5×1012Jones [67]. Later, they demonstrated a type of MAPbI3 nanowire photodetector with MAPbI3 wires grown via the saturated-vapor-assisted solution-growth method, and obtained a D* of 2.6×1013Jones [76].

Although the state-of-the-art MAPbI3 nanowire photodetectors have achieved tremendous success, most reported MAPbI3 nanowires show obvious morphological imperfections, e.g., rough surface [67,69,81] and irregular facets [64,73,76], associated with unintentional surface defects. It is expected that these surface defects will certainly induce inevitable carrier recombination loss and unfavorable scattering loss, shortening the carrier lifetime as well as deteriorating the dark current. It is apparent that reducing the density of surface defects in the perovskite crystal is a promising route to lower the dark current and promote the specific D* of perovskite photodetectors [1]. With this aim, Tang et al. immersed the solution-grown MAPbI3 nanowires in oleic acid, bringing forward a reduced dark current as well as an increased response under illumination, so that the D* was improved by an order of magnitude (reaching 2×1013Jones), and the corresponding LDR reached 140 dB, which is the largest reported [73]. It is noticed that Song et al. proposed a so-called surface-initiated solution-growth strategy to produce single-crystalline MAPbI3 nanowires [16,60,82], which could act as high-performance laser cavities with a quality factor as high as 3600, outperforming all the other reported MAPbI3 nanocavities [83–86]. The surfaces of their nanowire products displayed an RMS roughness value as low as 4 nm, corresponding to a relatively low density of surface defects. It is expected that this type of MAPbI3 nanowire with fewer surface defects may serve as a promising platform for developing photodetectors with high D*.

In this work, based on the surface-initiated solution-growth strategy, we produced high-quality, single-crystalline MAPbI3 nanowires with a carrier lifetime as long as 112.9 ns, which is around twice that of oleic acid passivated MAPbI3 nanowires (51.1 ns) [73]. The atomic force microscopy (AFM) characterization indicates that the produced nanowires display atomically smooth surfaces with an RMS roughness down to 0.27 nm, an order of magnitude lower than the best result in the literature [60]. Such a smooth morphology of the as-grown MAPbI3 nanowires indicates significantly minimized surface defects, accounting for the obtained long carrier lifetime. Through a dry-contact technique, we realize ohmic contacts between MAPbI3 nanowires and Au electrodes, and further demonstrate high-performance photodetectors that can respond linearly to illumination with the power density varying from 5.5nW/cm2 to 370mW/cm2, corresponding to an LDR as large as 157 dB. At the lowest detectable illumination (Plow) of 5.5nW/cm2, its responsivity (R) is as high as 8.52×103A/W, and its D* derived from the measured noise current reaches 1.2×1014Jones. To the best of our knowledge, our photodetector presents both the largest LDR and the highest D* among all MAPbI3 nanowire photodetector devices. Such superior photodetector performance metrics are attributed to, first, the defects-suppressed property of the as-grown MAPbI3 nanowires, which leads to a quite low noise current in the dark, and second, the ohmic contact between MAPbI3 and Au interfaces, which gives rise to an improved R compared with the Schottky contact counterpart. Our work contributes to the development of high-performance, next-generation, low-cost photodetectors.

The single-crystalline MAPbI3 nanowires are grown by the optimized surface-initiated solution-growth strategy as shown in Fig. 5 (Appendix A). When the coated transparent lead acetate (PbAc2) solid film is immersed into the methyl ammonium iodide (MAI) solution, the PbAc2 film immediately turns brown, due to the production of a thin layer of polycrystalline MAPbI3 on its surface. This layer serves as a seed layer to initiate the crystal growth; hence, the strategy is referred to as the “surface-initiated solution-growth” strategy. After the reaction is finished, the product of the single-crystalline MAPbI3 nanowires is harvested with an additional process of isopropanol (IPA) wash for removing the residual MAI. Our previous work concluded that the length and amount of the produced nanowires can be well controlled by the precursor concentration and the growth temperature [87]. In this work, we also investigate the influence of environmental humidity on the reaction and find that a humidity of 30%±2% is the optimal condition for obtaining the MAPbI3 nanowire products with the highest density and best crystallinity, as shown in Fig. 6 (Appendix B). Because IPA can take up moisture from air due to its hygroscopic property, we suspect that water molecules can interact with the polycrystalline perovskite surface and localize electrons close to its surface [88], which might act as a driving force for initiating the growth of the MAPbI3 nanowires. With the increase in humidity from 20% to 30%, there would be more water molecules in the IPA solution, resulting in more sites with localized electrons at the polycrystalline perovskite surface, thereby producing more MAPbI3 nanowires. In comparison, at a high level of humidity (e.g., 35%), water molecules interact with each other and form a hydrogen bonded network [88]. Thus the localization of electrons at the polycrystalline perovskite surface cannot be produced efficiently, which means the favorable factor for initiating the growth of MAPbI3 nanowires is weakened. As a result, the density of the MAPbI3 nanowire product decreases at a high humidity of 35%.

Figure 1(a) presents a typical optical image of the as-grown MAPbI3 nanowires accompanied by a few nanoplatelets on a glass substrate. It can be seen that the prepared perovskite nanowires are rich in color because the thicknesses of the nanowires are different, and interference occurs between the beams reflected from the top and bottom surfaces of the nanowires [89]. These nanowires have flat rectangular end facets with dimensions of a few hundred nanometers as indicated by the scanning electron microscopy (SEM) images in Fig. 1(b). The transmission electron microscopy (TEM) analysis of a single MAPbI3 nanowire was carried out as displayed in Fig. 1(c). A high-resolution TEM image of a single MAPbI3 nanowire with the spacing of the lattice fringes measured to be ∼0.31nm is shown in Fig. 1(d), and the fringes are indexed as (004) or (220) of the tetragonal MAPbI3 phase [73]. The inset of Fig. 1(d) presents the corresponding selected-area electron diffraction and fast Fourier transform (FFT) patterns. The sharp diffraction spots indicate that the synthesized MAPbI3 nanowire has a tetragonal crystal structure with zone axes of [110] or [001].

The X-ray diffraction (XRD) patterns of the prepared MAPbI3 nanowire are characterized as shown in Fig. 1(g), in good agreement with the corresponding simulated tetragonal-phase MAPbI3 XRD pattern shown in Fig. 7 (Appendix B). Apparently, the pattern shows multiple strong diffraction peaks at [002], [110], [220], [004], etc., which can be assigned to the tetragonal phase with the space group of I4/mcm (a=8.8743 Å and c=12.6708 Å). Additionally, the split of the [220] and [004] peaks due to the ordering of MA+ ions is clearly observed, further confirming that the as-grown MAPbI3 product is of the tetragonal phase. The stronger diffraction peaks from high-index lattice planes (e.g., [130], [224], [134], [404]) further validate the success of the growth of high-quality single-crystalline MAPbI3. Here, it is noted that the IPA wash process is critical for producing a pure MAPbI3 single-crystalline product (see Fig. 8 in Appendix B). To examine the composition uniformity of the as-grown MAPbI3 nanowires, energy-dispersive X-ray spectroscopy (EDS) mapping was performed on the nanowire product, as shown in Fig. 9 (Appendix B). The distribution of elements indicates that Pb and I are homogeneously distributed in the individual MAPbI3 nanowires and the atomic ratio of Pb to I is approximately 1∶3.

An AFM analysis on a single MAPbI3 nanowire on top of the PEDOT:PSS film was also carried out to further characterize the surface morphologies, as shown in Fig. 1(e). Apparently, the MAPbI3 nanowire displays a steep wall, in agreement with the results of the SEM image in Fig. 1(a). Figure 1(f) indicates that the average RMS roughness of the selected nanowire region is measured to be 0.27 nm, reaching the atomically smooth level. For comparison, the AFM image of the PEDOT:PSS surface was also measured, as displayed in Fig. 10 in Appendix B (RMS: 0.16 nm), which is only a bit lower than that of the MAPbI3 nanowire. Compared with the RMS value (4 nm) in a previous report [60], the present nanowire shows a much smoother surface, which should benefit from the systematic control of the reaction procedures. By fitting the transient photoluminescence (PL) spectrum of the MAPbI3 nanowire [see Fig. 1(h)] with a double exponential function, which describes a fast decay resulting from bimolecular recombination and a long decay resulting from recombination of free carriers in the radiative channel whose decay time is called the carrier lifetime, we know that the carrier lifetime of the prepared MAPbI3 nanowires is 112.9 ns. Notably, the carrier lifetime of the present MAPbI3 nanowires without any surface passivation is around twice that with oleic acid treatment (51 ns) [73]. Such an excellent feature of long carrier lifetime is an indicator of the significantly suppressed defects in the as-grown atomically smooth MAPbI3 nanowires, enabling the products to serve as ideal candidate constituents for photodetectors with high D*.

To construct photodetectors, high-quality, single-crystalline, defects-suppressed MAPbI3 nanowires are transferred onto a glass substrate patterned with interdigitated metal electrodes through the dry-contact technique (see Appendix A for more details). To validate the photodetector performances, more than 10 photodetectors were fabricated and then analyzed for the light detection performance. The demonstrated photodetector devices are of the planar metal–semiconductor–metal (MSM) type (see Fig. 11 in Appendix C). Figure 2(a) depicts the logarithmic current–voltage (I-V) characteristics of the MAPbI3 nanowire photodetector in the dark and under 343-nm, 505-nm, and 660-nm light illumination with a power density of 10.19mW/cm2. One sees that the photo current increases significantly with the increase in applied bias, and is much greater than the dark current. The dark current at 1 V is as low as 1×10−10A; the photo current under 660-nm illumination reaches 1.47×10−7A at 1 V, corresponding to a light-to-dark current ratio of 1.47×103. Under 660-nm illumination, as the light intensity increases, the photocurrent increases, as shown by the linear I-V plots in Fig. 2(b).

To analyze the contact properties, the double-logarithmic I-V curve at a power density of 10.19mW/cm2 under 660-nm illumination is plotted in Fig. 2(c). It is found that under higher bias, the I-V characteristic rises sharply, and the fitting slopes before and after the turning voltage (VT) are 1 and 2.3, respectively. This trend coincides with the space-charge-limited current with trap filling [90,91]. At low bias, the I-V current follows the ohmic response, indicating ohmic contacts are formed at the MAPbI3/Au interface. After entering the space-charge-limited-current regime, the current I is approximately proportional to V2+n, with n denoting the enhancement factor due to trap filling. Figure 12 (Appendix C) shows the dark and photo I-V characteristic of a photodetector prepared by evaporating the electrodes on top of the as-grown MAPbI3 wires. It reflects that at high bias, a saturation effect of the photocurrent takes place due to the Schottky contact formed between MAPbI3 nanowires and Au electrodes [57,60,87]. Differently, the suppression of photocurrent responses due to the Schottky contact can be fully eliminated in our ohmic contact MSM devices. It means that, compared with the Schottky contact, the ohmic contact between the MAPbI3 and the Au enables maximization of the photocurrent responses. Moreover, it is noticed that during the process of evaporating electrodes, the MAPbI3 nanowires degraded in the high temperature chamber, which also deteriorated the device stability very severely.

Next, we carried out the characterization of the LDR of the prepared MAPbI3 nanowire photodetector. The LDR is defined as 20log(Psat/Plow), where Psat and Plow represent the highest and lowest detectable illumination power densities, respectively. Outside this range, the photocurrent begins to deviate from the linear regime, and thus the light signal cannot be accurately detected. Here, we used a 532-nm laser to characterize the LDR under bias of 1 V. The power-density-dependent photocurrent curve of the demonstrated photodetector is displayed in Fig. 3(a), in which Psat=370mW/cm2 and Plow=5.5nW/cm2, corresponding to an LDR of 157 dB. This LDR value exceeds that of the conventional Si-based photodetector (120 dB) [2]. These demonstrated photodetectors that can detect incident light with power density varying over such a wide range are quite promising for application as power meters and imaging sensors. Our device performance is also compared with a few representative results of perovskite nanowire photodetectors from the literature, as compiled in Table 1 (Appendix C). Obviously, the LDR demonstrated in this work is up to 17 dB larger than the LDR (140 dB) of the oleic-acid-passivated MAPbI3 nanowires photodetector [73]. Note that our Plow is two orders of magnitude lower than the previous record for all-inorganic perovskite nanowire photodetectors (10−4mW/cm2) [94] and is comparable to that of the best quasi two-dimensional perovskite nanowire photodetectors (10−6mW/cm2) [78].

From the obtained I-V measurement results, the external quantum efficiency (EQE) and R values of photodetectors under various illumination power densities are derived based on Eqs. (D1) and (D2) in Appendix D and are plotted in Figs. 2(b) and 2(c), respectively. In MSM photodetectors, it is common to see photocurrent amplification or EQE>100% [see Eq. (D3) in Appendix D]. In this case, each photogenerated electron may induce more than one hole sweeping across the device. As observed in Fig. 3(b), at power densities from 5.5nW/cm2 to 370mW/cm2, all EQE values are greater than 100%. It is also clear that EQE and R increase with decreasing power density of the incident light, consistent with the results in the literature [95–99]. This trend is ascribed to the electron traps existing within the MAPbI3 nanowires [57]. Under bias, most of the electron traps are quickly filled by the injected electrons; thus, only a small fraction of residual electron traps remain unfilled, offering opportunities for the photogenerated electrons to be trapped. With the reduction in illumination power density, due to the limited number of residual electron traps, the number of photogenerated trapped electrons per unit photon increases, and hence, EQE and R rise gradually. Consequently, at the lowest detectable power density of 5.5nW/cm2, one obtains the highest EQE of 2.0×106% and the highest R of 8.5×103A/W. Here, the obtained highest R is greater than the results of most of the MAPbI3 nanowire photodetectors reported in the literature, as indicated by Table 1.

Although the previously reported MAPbI3 nanowire photodetector with a record R (12.5×103A/W) displayed an expected Plow of 10−4mW/cm2 and a measured Plow of 10−2mW/cm2 [64], our device exhibits a superior ability to detect weak light with four orders of magnitude lower Plow in comparison, due to a significantly reduced surface defect density. In our work, the as-grown single-crystalline MAPbI3 nanowires have atomically smooth surfaces and regular facets, which significantly reduces the number of defects as well as the scattering loss, thereby guaranteeing the ability to distinguish very weak optical signals. In order to further confirm that our photodetectors made of atomically smooth MAPbI3 nanowires are ideal candidates for detecting weak light, the noise currents are measured; hence, the device D* can be calculated. As shown in Fig. 13 (Appendix C), the measured average noise current of the MAPbI3 nanowire photodetector is ∼7.2×10−14A·Hz−1/2, which is very close to the reported noise current of the oleic-acid-passivated MAPbI3 nanowires [73]. This coincides with the agreement of the lowest detectable power densities of our device and the oleic-acid-passivated one. Such a low noise current should have originated from the suppressed surface defects produced by the atomically smooth surface of the MAPbI3 nanowires. One sees that the noise current does not change with frequency. This reflects that the noise of this device is not dominated by the 1/f noise because the MAPbI3 nanowires are of single-crystalline form, which has a near absence of grain boundaries [100–102]. For our device, the shot noise (in,s) and thermal noise (in,t) are the two main noise sources and can be derived from the dark current [see Eqs. (D4) and (D5) in Appendix D]. As shown in Fig. 2(g), the shot noise and thermal noise limits are derived to be 1.9×10−15A·Hz−1/2 and 4.3×10−16A·Hz−1/2, respectively. Taking the square root of these two noise currents, we evaluate the theoretical total noise current limit to be 1.9×10−15A·Hz−1/2, which is a bit lower than the measured noise current of 5.7×10−14A·Hz−1/2. Thus, there is still some room for improvement in the crystalline quality and morphologies of the MAPbI3 nanowires to reach the theoretical noise limit, which could be achieved by passivating the grain-boundary defect of the perovskite, interface engineering to avoid leakage currents in the detectors, etc. From the measured noise current and R, the D* of the photodetector can be deduced [see Eq. (D7) in Appendix D]. Figure 3(d) shows that the D* increases with decreasing power density of the incident light, and all the D* values under illumination with the power densities from 5.5nW/cm2 to 370mW/cm2 are greater than 1010Jones. At the lowest detectable illumination power density of 5.5nW/cm2, we obtain the highest D* of 1.2×1014Jones, outperforming all reported values of MAPbI3 nanowire photodetectors, as shown in Table 1. Because the shot noise of our photodetector is one order greater than its thermal noise, the D* can also be derived considering only the shot noise [see Eq. (D8) in Appendix D], which indicates an estimated D* of 2.3×1014 Jones at the illumination power density of 5.5nW/cm2, comparable to that deduced from the noise current. Here, the achieved record-high D* is attributed to the simultaneous realization of low noise current and high R, which are guaranteed by the atomically smooth surface of the as-grown MAPbI3 nanowires and the long carrier lifetime, respectively, assuming ohmic contacts are formed at the MAPbI3/Au interfaces.

From the I-V characterizations (see Fig. 14 in Appendix C), the wavelength-dependent EQE spectrum of the demonstrated photodetector is derived, as shown in Fig. 2(d). It can be seen that the EQE values over a broad spectral range from 300 nm to 750 nm are relatively high, in accordance with the device’s broadband absorption characteristic. The rise and fall times of the MAPbI3 nanowire photodetector were evaluated as 350 μs and 670 μs, respectively, corresponding to the −3-dB cutoff frequency of 2.6 kHz; see Figs. 4(a)–4(c). Here, the rise time (Trise) and fall time (Tfall), defined as the time for the photocurrent to rise from 10% to 90% (fall from 90% to 10%) during the on and off cycles of illumination, are measured to be 350 μs and 670 μs, respectively, as shown in Fig. 4(b). The sum of the rise time and fall time, counted as the response time, equals 1020 μs, the same level as those measured values of MAPbI3 nanowire photodetectors in literature [92]. Figure 4(c) shows a −3-dB cutoff frequency of 2.6 kHz, which is also comparable with the performance of other MAPbI3 nanowire photodetectors [64]. Stability tests were taken every five days under 660-nm, 10.19-mW/cm2 illumination. As displayed in Fig. 4(d), in 20 days, the dark current changed negligibly, while the photocurrent decreased to approximately 70% of its original value in the first five days and was maintained afterwards. This reveals good stability of the natural single-crystalline MAPbI3 nanowire photodetectors without any passivation or capsulation processes. Moreover, the stability test results shown in Fig. 4(d) reveal that the present single-crystalline MAPbI3 nanowire photodetector without any passivation or capsulation processes can maintain quite good detection ability after 20 days of storage.

In conclusion, high-quality single-crystalline MAPbI3 nanowires with atomically smooth faces and regular facets were prepared by the surface-initiated solution-growth strategy and exhibit a carrier lifetime as long as 112.9 ns. Based on the as-grown MAPbI3 nanowires, photodetectors in the MSM configuration were fabricated that exhibit excellent abilities to distinguish photo signals. The space-charge-limited-current I-V characteristic indicates that ohmic contacts are formed between the MAPbI3 nanowire and Au electrodes, yielding a light-to-dark current ratio of 1470 under 660-nm illumination at a power density of 10.19mW/cm2. The prepared device can linearly sense light with the power density varying from 5.5×10−6mW/cm2 to 3.7×102mW/cm2, corresponding to an LDR up to 157 dB, outperforming all reported MAPbI3 nanowire photodetectors. Due to the long carrier lifetime of the MAPbI3 nanowires, the photo response exhibits strong amplification with the EQE exceeding 100% at all detectable power densities. When the illumination power density diminishes to 5.5×10−6mW/cm2, the photocurrent amplification effect is maximized with the external quantum efficiency and R reaching 2.0×106%, and 8.5×103A/W, respectively. Benefiting from the significantly minimized surface defects of the MAPbI3 nanowires, the measured noise current is extremely low, reaching a level of 10−14A·Hz−1/2, and, based on that, the D* under the illumination power density of 5.5×10−6mW/cm2 is derived to be 1.2×1014Jones, outperforming all reported values of MAPbI3 nanowire photodetectors. Distinctly, the achieved record-high D* results from the simultaneous realization of high R and low noise current, produced by the long carrier lifetime and suppressed defects of the as-grown MAPbI3 nanowires, respectively, assuming that the MAPbI3/Au contacts are of the ohmic type rather than the Schottky type. Such an outstanding performance at weak-light detection is superior to that of all reported MAPbI3 nanowire photodetectors. The proposed approach can also be applied to suppress the unintentional defects of other all-inorganic perovskite nanowires or two-dimensional perovskite nanowires, which are expected to have better stabilities than their hybrid three-dimensional perovskite counterparts. Our work contributes to the development of high-performance and low-cost photodetector devices for applications in bio-imaging sensing or long-range light communication for which weak-light signal detection is demanded.