High-sensitive X-ray detection requiring a low-dose rate is of particular importance to reduce the risks of cancer caused by repeated exposure to ionizing radiation in the fields of physical examination such as medical diagnosis and security inspection1,2. Therefore, it promotes the exploration of X-ray detectors to improve the sensitivity and reduce the detection limit. High-sensitivity and low detection limit require the X-ray detectors to possess high resistivity, high attenuation coefficient, low electron–hole formation energy (εpair), and excellent charge collection ability. Here, “high resistivity” results in the selection of materials with a large bandgap to reduce the temperature-induced carrier excitation. Whereas “low εpair” needs the target materials with a small bandgap to generate more electron–hole pairs by a single X-ray photon. Therefore, a medium bandgap between 1.5 and 3.0 eV is considered appropriate to balance the εpair and resistivity3. Nowadays, excellent semiconductors such as metal halide perovskites and CZT in forms of single crystal, polycrystalline or thick film with medium bandgap have been developed for high-sensitive room temperature X-ray detection. However, they are still limited by toxicity, stability, or cost4,5,6,7,8,9.

Apart from the mentioned semiconductors, BiI3 is also an alternatively promising material with a medium bandgap. As reported, BiI3 is a 2D-layered semiconductor that belongs to R\(\bar{3}\)−148 space group with a strongly anisotropic crystal structure consists of I–Bi-I tri-layers stacked by weak van der Waals interactions10,11. Owing to the appropriate bandgap (1.67 eV), high density (5.8 g cm−3), high atomic number (ZBi = 83, ZI = 53), and high resistivity (108−1013 Ω cm), BiI3 is attractive for hard radiation detection and has achieved sensitive X-ray detection recent years12,13,14,15,16,17,18,19,20,21,22,23,24,25.

This article reports a van der Waals heterostructure of BixIy that served as a high-sensitive X-ray detector. By analyzing the samples’ cross-sectional images by aberration-corrected scanning transmission electron microscopy (ac-STEM), we found the obtained BixIy is composed of thick BiI3 layers (main) alternately stacked with new thin Bi-rich layers (minor) with a chemical formula of BiI. Namely, BixIy presents a heterostructure formed by the stackings of BiI/BiI3/BiI, which leads to a dual bandgap. Benefited from the heterostructure, BixIy exhibits a higher X-ray sensitivity than BiI3 single crystal. Moreover, like many halide perovskite single crystals, macrosize BixIy can be grown using a low-cost, handy low-temperature solution method. These advantages together with other merits such as good X-ray attenuation efficiency and charge collection ability, high resistivity, environmentally friendly, good environmental and hard radiation stability make BixIy be attractive as a potential competitor for high-sensitive room temperature X-ray detection.


BixIy grown in solution

The BixIy were grown by a low-temperature solution technique, as shown in Fig. 1a. First, Bi2O3, I2, and Au were dissolved in a mixed hydroiodic acid and ethanol solution to form precursor solutions. Excessive iodine is used to dissolve gold in hydroiodic acid fully. Then the solution refinement technique was employed to grow high-quality crystals26. Namely, the precursor solutions were pretreated by solvothermal in 1,4-butyrolactone for three times and then refined by hydrothermal for 1-time. Bi3+ was reduced by I with Au acted as a catalyst during the solution pretreatment; the chemical reactions in the solution were:

$$4{{{{{{\rm{H}}}}}}}^{+}+4{{{{{{\rm{I}}}}}}}^{-}+{{{{{{\rm{O}}}}}}}_{2}\uparrow \mathop{\to }^{\triangle }2{{{{{{\rm{I}}}}}}}_{2}\uparrow+2{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}$$
$${{{{{{\rm{I}}}}}}}_{2}+{{{{{{\rm{I}}}}}}}^{-} \mathop{\Leftrightarrow}^{ \triangle }{{{{{{\rm{I}}}}}}}_{3}^{-}$$
$${{{{{{\rm{Bi}}}}}}}^{3+}+2{{{{{{\rm{I}}}}}}}^{-} \mathop{\Leftrightarrow}^{ \triangle/{{{{{\rm{Au}}}}}}}{{{{{{\rm{Bi}}}}}}}^{+}+{{{{{{\rm{I}}}}}}}_{2}\uparrow$$
Fig. 1: Preparation and characterization of BixIy.
figure 1

a A typical growth procedure for BixIy includes solution pretreatment, solution refinement and crystal growth in room temperature water bath. b Optical image of as-grown crystals. c A typical as-grown crystal with the stripped surface. d SEM image of an exfoliated flake transferred onto a conductive tape.

The iodine diffused into 1,4-butyrolactone and made it darken. BixIy with regular hexagonal shape and size up to 6 × 6 × 1 mm3 were obtained from refined solution after 14 days of water bath growth at room temperature without any disturbance, as shown in Fig. 1b. More times pretreatment in 1,4-butyrolactone would promote massive nucleation. Smooth surfaces and Large flexible flakes of the grown BixIy can be obtained by mechanical exfoliation, as shown in Fig. 1b and Supplementary Fig. 1. The morphology image (Fig. 1d) collected by scanning electron microscopy (SEM) reveals that the studied BixIy flake possesses a high-quality exfoliated surface with no bubbles, holes, and inclusions. The morphology profile (Supplementary Fig. 1) of BixIy flakes observed by atomic force microscopy (AFM) shows a layered structure. A step height of 0.66 nm was obtained and assigned to I–Bi–Bi–I monolayer (0.65 nm obtained by ac-STEM, see below), as shown in Supplementary Fig. 1.

BixIy has the same X-ray diffraction (XRD) pattern (Supplementary Fig. 2) as BiI3, but exhibits broader peaks and preferred orientation along [001], indicates a softer nature of BixIy. The inductively coupled plasma atomic emission spectra (ICP-AES) result (35.2 wt.% Bi of as-grown BixIy) confirms the chemical composition is BiI3 (35.4 wt.% Bi in calculation). However, thermogravimetric (TG) and differential scanning calorimetry (DSC) analyses (Supplementary Fig. 3) show a different thermal behavior between BixIy and BiI3. The melting point of BixIy (408 °C) is slightly different from BiI3 (411 °C). BiI3 exhibits two decomposition temperatures above melting point at 417 °C and 431 °C, respectively. However, which are not observed in BixIy. As the temperature exceeding 411 °C, BixIy remains 46.2% of its original mass, on the other hand BiI3 lost nearly all the mass. Meanwhile, no detectable mass loss is observed in the BixIy even at a temperature of 300 °C, indicating its high thermal stability.

The stripped surface examined by XRD shows a heterostructure character of BixIy with a major diffraction of BiI3 (00k) and a regular minor secondary diffraction, as shown in Fig. 2a. The X-ray photoelectron spectroscopy (XPS) survey of Bi 4f in the freshly stripped smooth surface of BixIy shows the expected Bi3+ peaks27 at binding energies (BE) of 164.4 eV and 159.1 eV together with a distinct additional component shifted by 1.5 eV toward lower BE (Fig. 2b), assigned as Bi+ (see below). The dramatic change of valance band BE from BiI3 to BixIy indicates a significant difference in the energy band between BiI3 and the BixIy, as shown in Supplementary Fig. 4.

Fig. 2: Structure characterization of BixIy.
figure 2

a (001) surface X-ray diffraction (XRD) pattern of BixIy, the red arrows point to a minor secondary diffraction of BixIy. b X-ray photoelectron spectroscopy (XPS) of Bi 4f for BixIy. c High-angle annular dark-field scanning transmission electron microscope (HAADF-STEM) image of layered stacking of adjacent BiI3 and BiI in BixIy. d Side view of BixIy along [100] direction, where m and n are integer values. The thickness of BiI and BiI3 layers is not controlled with atomic precision. The distance of BiI3/BiI, BiI/BiI, and BiI3/BiI3 are 0.37 nm, 0.35 nm and 0.33 nm, respectively.

Then we used aberration-corrected STEM to observe the stacking sequence of BixIy by a high-angle annular dark-field (HAADF) imaging mode, which provides directly interpretable Z-contrast images at the atomic level28,29. Supplementary Fig. 5 shows several cross-sectional STEM images of flakes with different thicknesses exfoliated from a grown crystal. As seen in Supplementary Fig. 5, the flakes show alternate stacking of thin bright layers and thick dark layers. The bright layers are assigned to a Bi-rich phase due to the Z-contrast HAADF image28,29. It should be noted here that the thickness of the exfoliated BixIy is mainly dependent on the middle part of BiI3. The STEM-EDS mapping results in Supplementary Fig. 5 also confirm a higher concentration of Bi in the bright layers. Detailed Z-contrast images of Bi-rich phase and the dark layers are shown in Supplementary Fig. 6. Bi-rich phase exhibits a layered van der Waals structure built by the stacking of I–Bi–Bi–I four atomic layers with a chemical composition of BiI. On the other hand, the dark layers have a BiI3 structure characterized by the staking of I–Bi–I three atomic layers23, which is consistent with the BiI3 atomic structure model. BiI3 and BiI layers are also held together by weak van der Waals force in the BixIy structure, as shown in Fig. 2c. The STEM images clearly confirm the van der Waals heterostructure of BixIy constructed by stacking of thick BiI3 and thin BiI layers, as shown in Fig. 2d.

The BiI layers could be separated from the BixIy by mechanical exfoliation. As a result, we successfully obtained the thinnest BiI film composed of seven I–Bi–Bi–I layers, as shown in Supplementary Fig. 7. Unlike other Bi-rich bismuth iodides such as Bi4I4 with 1D structure30, the BiI we obtained is a member of 2D family and could be used to build the blocks of van der Waals heterostructures with other 2D atomic crystals. Moreover, considering the versatility of 1D Bi4I4 in thermoelectric, topological insulator, and superconductivity31, the 2D BiI may also exhibit similar properties.

Dual bandgap of BixIy

The bandgap of the BixIy was measured by UV–Vis–NIR diffuse reflectance spectroscopy (DRS) and shown in Fig. 3a. The reflectance of BixIy exhibits an indirect band nature with a sharp increase at 670–810 nm, assigned to the thick BiI3 layers, and a gentle rise after 810 nm, assigned to BiI layers. Namely, BixIy exhibits a dual bandgap. The dual bandgap of BixIy is also confirmed by the absorption spectrum (Supplementary Fig. 8) of a typical grown BixIy. Based on the DRS result, the bandgap energy (Eg) of BiI layer was obtained by the Tauc method with 0.70 eV, dramatically lower than that of BiI3 (1.67 eV) layer. The reduced bandgap of BiI promotes photo-responses up to 1800 nm in the BixIy, as shown in Supplementary Fig. 9.

Fig. 3: Band structure and charge separation kinetics of BixIy.
figure 3

a Diffuse reflectance spectra of commercial BiI3 and BixIy. Inset: calculated optical bandgaps of BiI3 layer and BiI layer in the BixIy using the Tauc method by assuming an indirect bandgap. As using the reflectance, the Tauc relation is: (F(R))1/2 = A(hυ-Eg), where A is a constant, h the Planck constant, υ the photon frequency, F(R) = (1R)2/2 R, R the reflectance. The linear extrapolation (the green lines) in the absorption edge region of (F(R))1/2 versus curve is used to determine the bandgap Eg. b UPS spectrum of BixIy, the linear extrapolation (the green line) is used to determine the cutoff energy (15.11 eV). Inset: linear extrapolation (the green lines) in the low-binding-energy region, the energy (0.31 eV) determined is used to calculate the valence band energy (Ev) as: Ev = 21.22 – 15.11 + 0.31 = 6.42 eV of BiI layer in the BixIy. c Band diagram of BixIy exhibits a dual bandgap, the values of Ec (conduction band energy) and Ev (valence band energy) of BiI3 are extracted from ref. 32. d A typical leakage current-field relation of the Ag/BixIy/Ag devices tested by probe, Ag paste was used to fabricate the device, the red and green symbols are experimental data, the solid lines are linear fitting of the experimental data. e Excitation of electron–hole pairs separated at the BiI3–BiI interface. The cross-sectional image is cut out from a STEM image and used as a schematic diagram. The red and blue circles represent electron and hole, respectively. The red and blue arrows illustrate their flow under bias.

Figure 3b shows ultraviolet photoelectron spectroscopy (UPS) of the BixIy, which exhibits different cutoff energy (15.11 eV) and the energy (0.31 eV) extracted from linear extrapolation in the low-binding-energy region from which of BiI3 (about 16.75 eV and 1.25 eV, respectively, see ref. 32 and Supplementary Fig. 4). As seen from Fig. 2d, the distance between BiI3 and BiI layer (0.37 nm) is larger than which between BiI3 layers (0.33 nm), resulting in rich of BiI in the stripped surface of BixIy. Therefore, the valence band energy (Ev) obtained by Ev = 21.22 – 15.11 + 0.31 = 6.42 eV is assigned to the BiI layer in the BixIy. The conduction band energy (Ec) of BiI layer is then calculated by Ec = Ev + Eg = 5.72 eV. The band diagram of BixIy heterostructure with a dual-bandgap can be plotted using the above data, as shown in Fig. 3c. BixIy exhibits large and anisotropy resistivities of 4.1 × 109 Ω cm for lateral (Ec) and 2.2 × 1010 Ω cm for vertical direction (Ec), respectively, as shown in Fig. 3d. The resistivity of BixIy is comparable to that of BiI3 (108−1013 Ω cm)12,13,14,15,16,17,18,19,20,21,22,23,24,25, indicating a negligible effect of BiI layer with small bandgap (0.7 eV) to the resistivity of BixIy due to its few numbers.

It can be deduced from the band diagram that electron injection from BiI3 layer to BiI layer would happen at the BiI3–BiI interface, which is confirmed by the transient absorption (TA) measurement (Supplementary Fig. 10). As seen from Supplementary Fig. 10, BiI3 exhibits an absorption bleaching at 645–675 nm, but no bleaching was observed in the BixIy. Photocarriers generated by incident laser can form full-filled electrons in the conduction band of BiI3, resulting in filled state bleaching. However, the bleaching disappeared in the BixIy due to photoexcited electron transferring from BiI3 to BiI layers at the BiI3-BiI interface instead of maintaining a filled state in the conduction band of BiI3. The electron injection is also confirmed by a much weaker fluorescence around 700 nm of BixIy than BiI3, as shown in Supplementary Fig. 11. Due to the electron injection from BiI3 layer to BiI layer, a built-in field with a direction from BiI3 to BiI emerges at the BiI3-BiI interface, which would prompt the electron–hole separation under X-ray radiation, as shown in Fig. 3e. The separation of electrons and holes into different layers would weaken their recombination and promote charge collection.

X-ray response with low detection limit

As mentioned above, the heterostructure of BixIy has high resistivity and shows benefits for charge separation. Moreover, Bi bilayer is observed in the middle of I–Bi–Bi–I four atomic layers (Supplementary Fig. 6) and further confirmed by the Raman spectrum of bismuth (Supplementary Fig. 12), which is promising for ultrafast electron transport because of the ultrahigh electron mobility of bismuth33. Therefore, BixIy is expected to have a good X-ray response with low noise.

Bulk (2.1 × 2.1 × 0.4 mm3) X-ray detectors were fabricated with a device structure of Cu/ BixIy/Cu. The copper conductive tapes (0.06 mm thickness) were pasted on both sides of the surfaces parallel or perpendicular to (001) to fabricate the vertical or lateral devices, as shown in Fig. 4a. The devices were exposed to a source with X-ray photon energy up to 70 keV. Then X-ray induced photocurrents were measured by a normal intracavity direct radiation configuration and a cavity-edge leakage radiation configuration, as shown in Fig. 4a.

Fig. 4: Room temperature device performance.
figure 4

a Illustration of X-ray detector (the Cu tapes cover the whole surface for full charge collection) and measurement configuration. b Anisotropic on/off X-ray responses of BixIy at different dose rates measured by intracavity configuration under 1 V mm−1 bias. c Anisotropic bias-dependent SNR (average from values under different dose rates with an X-ray tube current of 5–100 μA) measured by intracavity configuration. d Anisotropic bias-dependent SNR measured by cavity-edge configuration. The blue dotted line represents a SNR of 3, so the detection limits are 34 nGy s−1 for lateral and 317 nGy s−1 for vertical devices, respectively. The solid lines are smooth of the experimental data.

As seen from Fig. 4b, the photocurrents of lateral and vertical devices measured by intracavity configuration under various X-ray dose rates reveal clear on/off responses with anisotropy. A detailed on/off response of BixIy (Supplementary Fig. 13) shows rise/fall times of 49/71 ms for lateral device and 74/98 ms for the vertical device under 1 V mm−1 bias, which is comparable to the single crystal BiI3 (110/120 ms) grown by physical vapor transport (PVT)14 and the printable MAPbI3 device (less than 50 ms) for imaging34. The signal-to-noise ratios (SNR) calculated from the on/off responses (method described in ref. 26.) show stable and large average values around 2000 of the lateral device and 1600 of the vertical device, under various X-ray dose rates and bias larger than 1 V mm−1, as shown in Fig. 4c.

We also performed a leakage X-ray radiation measurement using a cavity-edge configuration (Fig. 4a) to examine the responses of the BixIy detector in a simulated radiation leakage environment. As seen in Supplementary Fig. 14, the photocurrents induced by small leakage radiations still exhibit clear on/off responses in lateral and vertical devices. According to IUPAC standard, the dose rate with an SNR value of 3 is defined as the lowest detection limit at a given electric field. The lowest detection limit, representing the minimum X-ray dose rate used for inspection, is an important parameter relevant to health risk during X-ray security examinations or X-ray medical inspections1,2. As seen from Fig. 4d, the lowest detection limit of the lateral device achieved a very small value of 34 nGy s−1, which is comparable to the excellent X-ray detectors with low detection limit4,5,6,7,8. The detection limit of the vertical device also achieves a small value of 317 nGy s−1, much lower than that required for regular medical diagnostics (5.5 μGy s–1)35.

X-ray sensitivity and stability

As mentioned above, thick BiI3 layers constitute the main body of BixIy with good X-ray radiation attenuation efficiency attributed to its high atomic number (ZBi = 83, ZI = 53), and high density (5.8 g cm−3). As seen from Fig. 5a, b, BiI3 showed a much better X-ray attenuation efficiency than MAPbBr3. For 50 keV hard X-ray, BiI3 would attenuate 99.82% of the incident photons, while MAPbBr3 88.41% at 1 mm thickness. Therefore, high attenuation efficiency enables BixIy to adequately absorb X-ray with reduced thickness, accelerating the charge collection. The charge collection ability of BixIy characterized by a μτ product, where μ is the carrier mobility and τ the carrier lifetime, is derived by fitting the photoconductivity using Hecht equation36, as shown in Supplementary Fig. 15. The BixIy exhibits anisotropic μτ products of 3.0 × 10−3 cm2 V−1 (lateral) and 4.4 × 10−5 cm2 V−1 (vertical) respectively. The corresponding lateral μτ products is comparable to that of perovskites with good X-ray detection properties5,26,37. The mobilities of BixIy measured by the space charge-limited current (SCLC) method confirmed the strong anisotropic charge transport with values of 53 cm2 V−1 s−1 (lateral) and 0.15 cm2 V−1 s−1 (vertical), respectively, as shown in Supplementary Fig. 16. According to the electronic dimensionality theory38, the electronic bands are more dispersive in the (001) plane of BixIy owing to the strong in-plane chemical bonds interaction, while more localized perpendicular to the (001) plane induced by the weak out-of-plane van der Waals interaction. Therefore, BixIy shows a better charge collection ability in the lateral direction.

Fig. 5: X-ray attenuation, sensitivity, and stability of BixIy detector at room temperature.
figure 5

a, b Attenuation efficiencies of BiI3, (NH4)3Bi2I9, Cs2AgBiBr6, MAPbBr3 and CdTe semiconductors versus thickness to 50 keV X-ray photons (a) and photon energy (b). The curves were calculated by employing the NIST XCOM photon cross-section database43. c Anisotropic X-ray photocurrent densities at different dose rates measured by cavity-edge configuration under 1 V mm−1 bias, the linear fittings (solid lines) are used to calculate the sensitivity. d Anisotropic X-ray sensitivities at different bias measured by cavity-edge configuration. e Device stability under repeated and continuous X-ray radiation. The black line is 204 repeated on/off response, followed by a response (green line) with a long continuous “on” time. The red line illustrates a stable “on” current is achieved after a repeated on/off process. The blue symbols illustrate the fluctuation of device SNR during the long-time continuous X-ray radiation. Inset: a Cu/BixIy/Cu vertical device used for radiation stability test under 120 keV/204 uGy s−1 X-ray radiation and 154 V mm−1 field. f Device stability under humidity. The specimens, after water immersion, were dried by handkerchief tissues first and then repasted the Cu tapes for the following I–V tests. Inset: the X-ray response current changed with water immersion time up to 8 h, I0 is the response current without water immersion. The absorption of the copper conductive tape was subtracted from all the data shown in (af).

Strong X-ray attenuation, high resistivity, good charge collection ability, and the resulting apparent responses at weak radiation indicate the BixIy is highly sensitive to X-ray. The sensitivity of X-ray detectors is derived from the current-dose rate relations, as shown in Fig. 5c. The lateral device has much larger sensitivity than the vertical device, as shown in Fig. 5d. The obtained sensitivity of lateral device achieved a high value of 4.3 × 104 μC Gy−1 cm−2 at 24 V mm−1 (50 V) bias, which is comparable to the newly reported high-sensitive detectors37,39,40,41,42, shown in Supplementary Fig. 17. Moreover, the sensitivity of lateral device achieves nearly one order of magnitude higher than that of the lateral BiI3 single crystal detector (0.5 × 104 μC Gy−1 cm−2 at 20 V mm−1)19, confirmed the advantage of heterostructure of BixIy for X-ray detection. However, the sensitivity of vertical device (253 μC Gy−1 cm−2 at 19.5 V mm−1) is smaller than which of the vertical BiI3 single crystal detector (660 μC Gy−1 cm−2 at 20 V mm−1)19. A larger distance between BiI3 and BiI layers than which between BiI3 layers leads to easier mechanical exfoliation of BixIy, however, harms the charge transport and then reduce the sensitivity of the vertical detector.

The BixIy detector was exposed to repeated and continuous 120 keV X-ray (used for CT) with a dose rate of 204 μGy s−1 to evaluate the anti-radiation stability. As seen from Fig. 5e, Stable X-ray photocurrent with a high SNR of around 1000 was observed after 204 circles repeated radiation and followed continuous radiation more than 4 h, confirms the highly stability of BixIy detector under high energy X-ray radiation. Moreover, nearly unchanged photocurrent intensity (Fig. 5f) of the BixIy detectors could be observed even after 8 h water (20 °C) immersion (Supplementary Fig. 18), confirms its highly environmental stability. High-sensitive and high-stable X-ray response of BixIy detector offers its great prospects in real applications.


In summary, we developed a handy and scalable solution method to first grow the macrosize van der Waals heterostructure of BixIy with regular shapes consisting of adjacent thick BiI3 (main) and thin BiI (minor) 2D layers. The BixIy heterostructure X-ray detectors exhibit stable response and anisotropic properties at different crystal orientation. The lateral device realized a high sensitivity of 4.3 × 104 μC Gy−1 cm−2 with a very low detection limit of 34 nGy s−1, meeting the demands of medical inspection to reduce the X-ray exposure to the human body. On the other hand, the BixIy photodetectors are versatile and present a photo response ranging up to 1800 nm, revealing its potential for near-infrared detection. Generally speaking, our results inspire the exploration of van der Waals heterostructure materials for high-sensitive X-ray detection.


Precursor solution preparation

All the purchased chemical reagents except Au (99.999%) were of analytical reagent grade purity and used without further purification. Solution 1 was prepared by 5.5 g Bi2O3 it was dissolved in 20 ml 55% hydroiodic acid at room temperature. Solution 2 was prepared by 2 g Au and 10 g I2. They were dissolved in 5 ml 55% hydroiodic acid for 3 days at room temperature. Solution 3 was prepared by 10% hydroiodic acid mixed with ethanol in a volume ratio of 1:1. solution 1 and solution 2 were mixed and diluted by solution 3 to 40 ml. The prepared 40 ml diluted solution was used as a precursor solution.

Solution pretreatment and refinement

The pretreatment and refinement procedures are schematically illustrated in Fig. 1a. Briefly, the obtained precursor solution was put into a 50 ml Ф20-mm conical flask. The flask was placed in a sealed beaker with 100 ml 1, 4-butyrolactone (GBL) for solvothermal treatment. Three times treatments are needed. The solvothermal treatment was performed in an 80 °C oven. After the first-time treatment (3–5 days), the solution was concentrated to 35 ml. The concentrated solution was diluted by solution 3 to 40 ml again for the second time solvothermal treatment. After the second time treatment, the solution was concentrated to 30 ml. The second time concentrated solution was diluted by ethanol to 35 ml for the third time solvothermal treatment. After the third time treatment, the solution was concentrated to 25 ml. The third time concentrated solution was then refined by hydrothermal treatment in a 60 °C oven for 3 days. Some small bulks with an irregular shape formed at the bottom of the conical flask after hydrothermal. The upper portion of the supernatant was then carefully transferred into another clear container to grow high-quality crystals.

Crystal growth

The solution, after pretreatment and refinement, was then handled by water bath (Fig. 1a) at room temperature to grow crystals. More than 7 days of growth without disturbance is needed to obtain millimeter crystals. The obtained crystals were washed by ethanol one time and dichloromethane two times followed. Bulks after washing dried naturally in the air and used for the following material characterization, device preparation, and test.


Powder X-ray diffraction was performed on a D8-DISCOVER diffractometer with Cu Kα (λ = 1.542 Å) radiation. The X-ray Photoelectron Spectroscopy (XPS) and Ultraviolet Photoelectron Spectroscopy (UPS) were performed on a Thermo Fisher ESCALAB XI + photoelectron spectrometer. Freshly exfoliated surface after 30 s Ar ion sputtering was used for XPS and UPS measurements. Thermogravimetric analysis (TGA) was carried out under continuous nitrogen flow using a NETZSCH STA 449F3 thermal gravimetric analyzer. The sample was held on a platinum pan, and heated at a rate of 5 °C min−1 up to 600 °C. AFM measurements were carried out in an Oxford Instruments Asylum Research Cypher S atomic force microscope with a contact mode. An IT500 scanning electron microscope (SEM) with a maximum 30 kV electron beam accelerating voltage was employed to observe the surface morphology of BixIy. STEM observations of the cross-section specimens were carried out in an aberration-corrected STEM microscope (Titan G2 60-300, Thermofisher equipped with a field emission gun) with 300 kV electron beam accelerating voltage. The probe convergence angle was 24.5 mrad, and the angular range of the HAADF detector was from 79.5 to 200 mrad. The cross-sectional TEM specimens were prepared by a dual-beam focused ion beam (FIB) nanofabrication platform (Helios 600i, Thermofisher). The UV–Vis–NIR diffuse reflectance spectroscopy (DRS) was measured by a HATACHI UH4150 spectrometer over the spectral range of 360−2000 nm. Room temperature photoluminescence and Raman spectra were collected by a Horiba LabRam HR Evolution microscopic confocal Raman spectrometer using a 6.8 mW, 532 nm CW Nd: YAG laser as an excitation source. The laser beam was focused to a spot size of about 0.7 μm in diameter. Transient absorption (TA) measurements were performed on a HARPIA-TA system (Light Conversion) at room temperature. A 1030 nm pulsed laser with 100 kHz repetition rate and 190 fs pulse duration was divided into two beams to generate pump laser and probe light, respectively. The pump laser of 480 nm was generated from an optical parametric amplifier system (OPA, Light Conversion) pumped by one beam of 1030 nm laser. The probe light was generated by exciting a sapphire plate by another beam of 1030 nm laser.

Photo and X-ray response measurement

The device for photodetection was fabricated on the (001) surface of a BixIy bulk, as shown in Supplementary Fig. 10. A pair of Ag electrodes with an interval of 0.5 mm was formed by painting Ag paste on a freshly exfoliated surface and then dried at 100 °C in the air. The area between Ag electrodes formed the light absorption area of the photodetector. The I–V characteristic under ambient light and infrared irradiation was measured by a KEITHLEY 2450 source meter. A YSL SC-PRO 7 supercontinuum source was used to generate CW infrared laser. Devices with Cu tape pasted on a pair of adjacent (100) or (001) surfaces formed the Cu/ BixIy/Cu structure (Fig. 3a), and were used for X-ray detection measurements. The X-ray detection performance was measured in a Pb cavity for intracavity mode and in a Ф 5  mm hole on the side of the cavity with a light-proof cover for cavity-edge mode, as shown in Supplementary Fig. 13. A commercially available MOXTEX MagPro Mini-X tube with a tungsten target and 12 W maximum power output was used as the X-ray source. The X-ray tube was operated with a constant 50 kV voltage. The total X-ray dose was modulated by changing the current of the X-ray tube. The radiation dose rate was calibrated using a Radical ion chamber dosimeter. The X-ray photocurrent was measured by a KEITHLEY 2636B source meter. For the anti-radiation test, a 150 kV HAMAMATSU/L12161-07 microfocus X-ray source with 75 W maximum power was used.