Introduction

Ferroelectric crystals that exhibit reversible spontaneous polarization are important for optoelectronic applications in the fields of energy conversion1,2,3, frequency converters4, and intelligent electromechanical systems5,6,7,8. For instance, domain engineering has enabled the construction of dielectric superlattices in LiTaO39 and LiNbO310 crystals, thereby achieving efficient nonlinear harmonic generation11. By manipulating the ferroelectric domains under an external electric field, domain-wall scattering is eliminated to achieve transparent ferroelectric materials with ultrahigh piezoelectricity, thereby facilitating the development of ultrasonic imaging and electro-optic switching6. Furthermore, some unprecedented optical phenomena, such as optical solitons12, chaotic dynamics13, and diffraction-free light propagation14, have been reported in disordered ferroelectric crystals. These findings highlight new frontiers for ferroelectric applications in the fields of photonics and optoelectronics.

Benefiting from spontaneous polarization, ferroelectric crystals demonstrate a unique self-powered optical-to-electrical conversion without bias voltage. This is known as the bulk photovoltaic effect (BPVE)15. Compared with traditional p-n Schottky junctions, the BPVE allows an ultrawide voltage over the intrinsic bandgap, as well as low dark current at zero bias16. Since the discovery of the BPVE in 197715, many reports have focused on ferroelectric oxides, including BiFeO317, LiNbO318, and KTa1−xNbxO319, among others. However, these ferroelectric crystals typically possess wide bandgaps (Eg > 3 eV), which result in the limited photoactivity in the ultraviolet region, along with a relatively low responsivity20,21,22. In previous studies, some relative success to create ferroelectric oxides with low bandgap had been shown in KNN21 and BaTiO323, often at the expense of degraded ferroelectric properties. According to the BPVE photocurrent model \({j}_{{pv}}^{i}=\alpha {G}_{{ijl}}{e}_{j}{e}_{l}{I}_{0}\), the absorption coefficient α is a crucial factor in improving photocurrent generation24. To enhance α, some indirect methods have been proposed to harvest visible light via an energy transfer mechanism in hybrid plasmonic/ferroelectric structures25 and conductor/ferroelectric interfaces26. However, these materials suffer from complex fabrication processes and high costs. Therefore, enhancing the light-harvesting capability of ferroelectric crystals remains a great challenge, especially in the context of achieving full-spectrum absorption range from the ultraviolet to the mid-infrared region.

In general, black materials exhibit the strongest broadband light absorption27, as demonstrated by polycrystalline silicon solar cells and defect-induced photocatalysts. Taking TiO2 as an example, black TiO2 exhibits a reduced bandgap (3.2–1.54 eV) and can harvest infrared photons more efficiently28. Nevertheless, the creation of a black ferroelectric oxide crystal is difficult because of the intrinsic trade-off between maintaining the robust ferroelectricity and achieving the narrow bandgap for effective photovoltaic conversion29. To overcome this limitation, here we proposed a hierarchical defect engineering strategy in ferroelectric crystals to create the full spectrum optical absorption and maintain ferroelectric domains. Using transparent tungsten bronze CaxBa1-xNb2O6 (CBN) as host crystal, we introduce the synergistic Sr2+-doping and Ce3+-doping, thus extending the optical absorption to near-infrared, referring to the red Ce:CSBN. Then, via a high-temperature thermal reduction process, black Ce:CSBN is created with full spectrum absorption from ultraviolet to mid-infrared region. Their ferroelectric properties are evidenced by the P-E loops and corrosive ferroelectric domains. As a result, black Ce:CSBN exhibits a broadband self-powered responsive in 250–5000 nm, which represents the widest responsive range among all ferroelectric-based detectors. More interestingly, there is a reversible transformation between red and black samples, corresponding to the artificially controlled oxygen vacancy concentrations. These results break a long-standing BPVE spectral limitation in the wide-gap ferroelectric crystals, thus providing innovative pathways for designing broadband, high-sensitivity, and low-cost optoelectronic devices.

Results

Crystal structure characterization of the tetragonal tungsten bronze (TTB) crystal

The CBN crystal possesses a tetragonal tungsten bronze (TTB) structure (Fig. 1a) and is a typical ferroelectric crystal at room temperature. This crystal is composed of two types of NbO6 octahedra sharing common vertices, thereby forming three types of vacancies, labeled A1, A2, and C30. In the CBN crystal, Ca2+ and Ba2+ selectively occupy the A1 and A2 sites, whereas the C site remains vacant. In the high-temperature paraelectric phase, CBN possesses a centrosymmetric P4/mbm space group without spontaneous polarization. When the temperature drops below the Curie temperature (Tc), the Nb atoms at the B1 sites undergo significant displacement along the c-axis. This polar displacement not only drives the transition to the polar ferroelectric phase (P4bm space group) but also induces structural distortions of the NbO6 octahedra (Fig. 1b, c) in conjunction with the distribution of vacancies. Notably, these abundant structural vacancies are prerequisites for multiple-cation doping and defect regulation. In this study, the optical and BPVE properties of the Ce:CSBN crystals were investigated. As previously reported31, the isovalent substitution of Sr2+ ions into CBN crystals can effectively reduce Tc and create intriguing domain structures, which are conducive to investigate the optical properties of TTB crystals at room temperature. In addition, the aliovalent substitution of Ce3+ ions can bring about 4f–5d transitions and introduce additional oxygen vacancies32, both of which are beneficial for introducing defect levels and extending the optical absorption capability (Supplementary Fig. S1). Density functional theory (DFT) calculations were also performed to verify the band structure modulation potential of the Ce-doped CBN crystals (Fig. S2), and it was found that some localized defect levels emerged at the Fermi level, which can enhance optical absorption in the visible region. In a previous study33, Pankrath et al. grew Ce:SBN and observed that the absorption spectrum extended from 400 to 600 nm. Therefore, both theoretical and experimental studies have indicated that Ce3+ is an alternative candidate for cation substitution to extend the optical absorption properties of such materials.

Fig. 1: Characterization of the TTB crystals.
figure 1

a Structure projection of CSBN crystal along the c-direction. b, c Distorted NbO6 octahedra at the B1 and B2 sites. d Refined XRD pattern of the Ce:CSBN crystal. e iDPC-STEM image of the Ce:CSBN crystal along the c-direction. f, g Observed atomic arrangements of the Ce:CSBN crystal along the c- and a-directions. h iDPC-STEM image of the Ce:CSBN crystal along the a-direction. i Elemental mapping of Ce, Ca, Sr, Ba, Nb, and O in the Ce:CSBN crystal.

Full size image

In this work, both CBN and Ce:CSBN crystals were grown using the Czochralski method. The experimental details were given in Supplementary Information. The results of inductively coupled plasma (ICP) analysis (Table S1) show that the chemical formulae of the two crystals are Ca0.29Ba0.71Nb2O6 and Ce0.027:Ca0.11Sr0.32Ba0.57Nb2O6, respectively. Figures 1d and S3 depict the X-ray diffraction (XRD) patterns of the Ce:CSBN and CBN crystals, in which the sharp and symmetrical diffraction peaks indicate a good crystal quality in each case. All diffraction lines are consistent with the standard TTB phase (PDF#01-072-7029), indicating that the introduction of Ce3+ and Sr2+ does not change the crystal structure34.

To observe the microstructures of the two ferroelectric crystals, atomic-resolution characterization of the Ce:CSBN crystal was performed using aberration-corrected scanning transmission electron microscopy (AC-STEM). Under the high-angle annular dark-field (HAADF-STEM) imaging mode along the [001] zone axis (Fig. 1e, f), Ce:CSBN exhibited highly ordered tetragonal lattice arrangements with perfect periodicity of its atomic columns, thereby demonstrating its good crystallinity, possibly along with low-density defects (e.g., vacancies and dislocations). Based on the Z2-dependent HAADF imaging intensity (where Z represents the atomic number) and the atomic masses of the constituent elements, high-Z elements (i.e., Ca, Sr, Ba, and Nb) were found to display pronounced bright contrasts, whereas oxygen (low Z value) remained undetectable owing to its small scattering cross-section. In addition, along the [100] zone axis (Fig. 1g, h), a clear layered structure was observed, corresponding to the overlapping NbO6 layers and Ca2+/Sr2+/Ba2+ (alkaline-earth ion) layers. Furthermore, elemental mapping acquired via energy-dispersive X-ray spectroscopy (EDS) confirmed the homogeneous spatial distributions of all constituent elements (Ce, Ca, Sr, Ba, Nb, and O) in the Ce:CSBN crystal lattice without elemental segregation, indicating the atomically uniform incorporation of both dopants within the host (Fig. 1i). EDS mapping results show that Ce3+ ion mainly occupied at the Ba2+ site (Fig. S4). In the undoped CBN crystal, an ordered tetragonal lattice with a TTB structure was also observed.

Characterization of the dielectric and ferroelectric properties

The ferroelectricity of the TTB structure originates mainly from the distortion of the NbO6 octahedra35. In the paraelectric phase, the Nb atoms at the B1 sites exhibit no long-range ordered displacements, as can be seen in Fig. 2a. As the temperature drops to Tc, the Nb atoms undergo asymmetric displacement along the z-axis, resulting in separation of the positive and negative charge centers and the emergence of spontaneous polarization. At this time, the TTB structures undergo a paraelectric-ferroelectric phase transition, as shown in Fig. 2b. The phase transition properties of the CBN and Ce:CSBN crystals were demonstrated through temperature-dependent dielectric spectroscopy (ε-T) and differential scanning calorimetry (DSC) experiments. As shown in Fig. 2d, g, the phase-transition temperatures of CBN and Ce:CSBN were determined to be 285 and 175 °C, respectively, indicating that Sr2+ substitution reduces the Tc by ~110 °C. In addition, we have conducted a systematic analysis of the dielectric spectra characteristics of CBN and Ce:CSBN crystals, focusing on thermal hysteresis width (ΔThys), the full width at half maximum (FWHM), and the frequency-dependent dielectric shift. The results show that the ΔThys of CBN and Ce:CSBN are (33 ± 1) K and (27 ± 1) K, respectively, and the FWHM are (9 ± 1) K and (12 ± 1) K, respectively, which is consistent with the typical characteristics of first-order ferroelectric phase transitions36,37,38,39. The dielectric shift of CBN is 1 × 10−3 K decade−1, also consistent with the characteristics of a first-order ferroelectric phase transition. In comparison, the dielectric shift of Ce:CSBN increases to 5 × 10⁻³ K decade⁻¹, despite higher than that of CBN, is still two orders of magnitude lower than that of a typical SBN relaxor ferroelectric (at 0.25 K decade⁻¹) (Table S2)40. Therefore, the “relaxor-like” dielectric shift of Ce:CSBN could be assigned to polaronic effects induced by oxygen vacancies (Vo)41. In the DSC curves (Fig. S5), the observation of reversible heat flow peaks at 240–246 and 130–135 °C further confirms the thermodynamic nature of the ferroelectric transitions.

Fig. 2: Characterization of the dielectric and ferroelectric properties of the CBN and Ce:CSBN crystals.
figure 2

a, b Electron density distributions of the NbO6 octahedra at the B1 sites in the paraelectric and ferroelectric phases. c Temperature-dependent P-E loop of the Ce:CSBN crystal. d Temperature-dependent dielectric constant and loss of the CBN crystal. e P-E loop of the annealed CBN crystal at room temperature. f Temperature-dependent S-E curves of the CBN crystal. g Temperature-dependent dielectric constant and loss of the Ce:CSBN crystal. h P-E loop of the Ce:CSBN crystal at room temperature. i Temperature-dependent S-E curves of the Ce:CSBN crystal.

Full size image

The ferroelectric properties of the CBN and Ce:CSBN crystals were subsequently characterized by P-E hysteresis loops. First, we systematically investigated the dependence of the P-E hysteresis loops of CBN on the concentration of Vo. The results showed that the saturation polarization (Ps) gradually increased with decreasing Vo concentrations (Figure. S6). In the CBN sample with the lowest Vo concentration (11.87%), the maximum Ps of 15.1 μC/cm² was achieved with a coercive field (Ec) of 14.5 kV/cm (Fig. 2e), which is consistent with previous results30. In contrast, the Ce:CSBN crystal exhibits a reduced Ec of 9.2 kV/cm and an increased Ps of 27.8 μC/cm2(Fig. 2h), exceeding many conventional TTB-type ferroelectrics (typically Ps < 20 μC/cm2). The simultaneous enhancement of Ps and reduction of Ec in Ce:CSBN can be attributed to three possible mechanisms: (i) Ce3+ doping-induced lattice distortion and symmetry breaking effect42; (ii) enhanced A-B site cooperative displacements due to ionic radius mismatch43; (iii) Sr2+ substitution-mediated Tc modulation44. The temperature-dependent P-E loops of the Ce:CSBN crystal were also investigated between 40 and 180 °C at 10 Hz (Figs. 2c and S7). Remarkably, even at 180 °C, which is approaching the Tc (175 °C), distinct P-E hysteresis loops persist, providing additional evidence for retained ferroelectric orders at these elevated temperatures.

Moreover, the strain–electric field (S-E) characteristics of CBN and Ce:CSBN were determined by applying alternating electric fields over a broad temperature range. Figures 2f and 2i show that both crystals exhibit typical electrostrain responses with comparable temperature evolution patterns. At room temperature below Tc, butterfly-shaped S-E loops with pronounced negative strains appear for both the CBN and Ce:CSBN crystals. The maximum strains under positive and negative electric fields were determined to be −0.004/0.004% for CBN and −0.005/0.008% for Ce:CSBN, respectively. Upon increasing the temperature, the Ce:CSBN crystals displayed a gradual transition from the ferroelectric to the paraelectric phase, during which the butterfly loops progressively shrank with a reduced symmetry, and significantly weakened the negative strain effects. Close to Tc, the S-E curves of the Ce:CSBN exhibited an almost linear behavior, indicating the disappearance of long-range ferroelectric orders.

Modulation of the optical properties through thermal reduction/oxidation treatment

Transparent CBN is a wide-bandgap oxide (Eg = 3.2 eV) with an intrinsic absorption edge at 380 nm. Therefore, CBN can only harvest the ultraviolet photons. After Ce3+ and Sr2+ ion doping, the Ce:CSBN crystals exhibited a red color with enhanced optical absorption in the visible region, but still remained inactive at longer wavelengths. To address this challenge, it is necessary to develop an approach for precisely controlling the defect concentrations. Inspired by the thermal reduction technologies of black TiO245 and LiNbO327, a thermal reduction/oxidation strategy was employed to treat the red Ce:CSBN ferroelectric crystals. As shown in Fig. 3a, following thermal reduction at 800 °C in a 5 vol% H2/95 vol% N2 mixed atmosphere, the red Ce:CSBN sample exhibits a significant change from red to black. Notably, upon subjecting the black sample to thermal oxidation at 800 °C in a 20 vol% O2/80 vol% N2 mixed atmosphere, red color was fully restored. This reversible color transformation demonstrates that the optical properties of the Ce:CSBN crystals can be modulated by precisely controlling the parameters of the thermal treatment process. Figure 3b shows photographic images of the CBN, red Ce:CSBN, and black Ce:CSBN samples. For ease of expression, the transparent CBN, red Ce:CSBN, black Ce:CSBN, and red Ce:CSBN2 samples are denoted as T, R, B, and R2, respectively.

Fig. 3: Reversible regulation of the optical properties via thermal reduction/oxidation.
figure 3

a Schematic representation of the sample preparation process. b Photographic images of the T, R, and B samples. Crystal size: 5 × 5 × 1 mm. c Absorption spectra of the T, R, R2, and B samples. d EPR spectra of the R, B, and R2 samples. e–g XPS results for the R, B, and R2 samples. Detailed peak fitting results are given in Table S3.

Full size image

The absorption spectra presented in Figs. 3c and S8 show that B-sample exhibits significant full-spectrum absorption, and the dependence of its absorption spectrum on the wavelength is significantly weakened compared to the R-sample. Notably, the absorption range of the B-sample was extended beyond 5000 nm, indicating the possibility of an ultra-wideband photo-response from the ultraviolet to the mid-infrared region. In the case of R2-sample, the optical absorption returned to the visible region, similar to that of the original R-sample, and the ferroelectric P-E loop was also recovered (Fig. S9). To investigate the microscopic mechanisms underlying the changes in the optical properties, the evolution of the Vo concentration was considered using multiple characterization techniques. Electron paramagnetic resonance (EPR) spectroscopy revealed clear signals at g = 2.003 for all three samples, corresponding to the paramagnetic resonance response of the unpaired electrons in the Vo (Fig. 3d), consistent with the typical values reported for oxidate Vo species46,47. Quantitative analysis showed that B-sample exhibited a significantly enhanced EPR signal intensity compared to the original R-sample and the re-oxidized R2-sample, indicating a substantial increase in the Vo concentration of the B-sample. In addition, high-resolution X-ray photoelectron spectroscopy (XPS) was performed to identify the Vo concentrations in the three samples. As shown in Fig. 3e–g, through meticulous scanning and peak deconvolution of the O 1 s orbital, the Vo signature was detected at ~531.5 eV48,49. According to the peak fitting results, the Vo contents in the R- and B- samples were 11.41 and 32.81%, respectively, indicating a significant enhancement in the number of vacancy defects in the B-sample. Furthermore, in the re-oxidized R2-sample, the Vo content decreased to 12.25%, thereby confirming the dynamic reversibility of the defect regulation strategy.

In addition, to identify the Vo content at the atomic scale, atomic-resolution annular bright-field scanning transmission electron microscopy (ABF-STEM) was performed for the four crystal samples (T, R, B, and R2). ABF imaging analysis showed that the atomic column contrast strictly followed the Z1/3 dependence, where Z represents the atomic number. Consequently, the heavier elements, including Ca, Sr, Ba, and Nb, exhibited dark contrast, while the lighter O atoms exhibited characteristic bright spots50,51. Moreover, inverse fast Fourier transform (IFFT) images were recorded to provide comprehensive insights into the distribution and concentration variations of Vo in the different crystal samples52. As shown in Fig. 4a, the T-sample exhibited uniformly bright oxygen columns with perfect lattice periodicities, indicating negligible Vo concentrations. In contrast, the R-sample exhibited localized brightness variations in the oxygen columns owing to charge compensation from aliovalent Ce doping, which was accompanied by the emergence of moderate Vo concentrations (Fig. 4b). After thermal reduction, the B-sample displayed substantial contrast variations with the complete absence of oxygen columns in certain regions (Fig. 4c), thereby providing direct evidence for the induction of high-density Vo using the described defect engineering approach. Furthermore, the re-oxidized R2-sample demonstrated significant recovery of the oxygen column density, wherein the distribution characteristics were almost identical to those of the R-sample (Fig. 4d). These ABF-STEM results are consistent with the EPR and XPS data, which collectively demonstrate that thermal reduction/oxidation can effectively modulate the Vo concentration in TTB ferroelectrics and alter their optical properties. Based on this reversible defect engineering, it should be possible to design high-performance photodetectors with superior optical absorption and robust ferroelectricity.

Fig. 4: STEM-ABF and IFFT images of the different TTB samples for the revelation of the Vo concentrations.
figure 4

a CBN (T-sample). b red Ce:CSBN (R-sample). c black Ce:CSBN (B-sample). d red Ce:CSBN2 (R2-sample). In the ABF mode, heavier atoms provide a darker contrast. The solid red circles represent intact O columns, while the dashed red circles indicate columns containing Vo. The green, yellow, blue, and red balls represent the Ca, Sr/Ba/Ce, Nb, and O atoms, respectively. All images were recorded along the [001] zone axis.

Full size image

Ferroelectric self-powered photodetectors

In addition to full-spectrum optical absorption, the ferroelectric domain is an important factor in the context of self-powered photodetection. In this study, the ferroelectric properties of CBN and Ce:CSBN crystals were investigated using corrosive ferroelectric domains. As shown in Fig. 5a, the scanning electron microscopy (SEM) images of the acid-etched T, R, and B samples revealed typical prismatic ferroelectric domain structures on the (100) planes with domain walls aligned parallel to the c-axis53. This is consistent with the polarization direction of the NbO6 octahedra at the B1 sites. In addition, dense square-like domain morphologies were observed on the (001) planes (Fig. 5b), indicating the presence of rich ferroelectric 180° domain walls along the c-axis, which is beneficial for photogenerated carrier separation driven by the built-in electric field. Moreover, examination of the c-cut T-sample by polarized light microscopy also confirmed the presence of regular square domains (with 20–50 μm scales) with significant brightness contrast during rotation (Fig. S10). These results demonstrate that Sr and Ce doping did not alter the ferroelectric domain structure. Indeed, even in the high-Vo B-sample, intrinsic ferroelectricity still exists, suggesting that its self-powered detection capabilities can be retained in optoelectronic applications.

Fig. 5: Photoelectric characterization of the TTB ferroelectric crystals.
figure 5

a SEM image showing the corrosion ferroelectric domains of the CBN crystal along the a-axis. b SEM image of black Ce:CSBN observed along the c-axis. c Schematic diagram of the TTB ferroelectric crystal photodetector. d I-V relationship for B-sample at 250 nm. The inset reveals the μA-level current in this sample at 0 V, demonstrating its self-powering capability. e Broadband self-powered photocurrents in the T, R, and B samples. f Self-powered responsivities (R) of the T, R, and B samples in different spectral regions. g Self-powered detectivities (D*) of the T, R, and B samples. h Long-term stability of the self-powered photodetector based on B-sample over 50 on/off cycles at 250 nm.

Full size image

High-performance self-powered photodetectors were subsequently fabricated based on the T, R, and B samples (Fig. 5c). Based on their distinct absorption spectra, the testing wavelength ranges were carefully selected as 250 nm (UV region) for T-sample, 250–830 nm (UV-vis region) for R-sample, and 250–5000 nm (UV-vis-MIR region) for B-sample. Detailed measurement protocols and results are provided in Figs. S11–S17. The detector performances were comprehensively evaluated according to several key parameters, including the responsivity (R), specific detectivity (D*), response time (τ), and long-term stability.

The current-voltage (I-V) curves of all samples revealed the typical nonlinear I-V behavior of ferroelectric materials, in which the photocurrent shows a positive correlation with the incident optical power (Fig. 5d). In addition, a μA-level photocurrent was observed without bias voltage, which is a typical feature of the BPVE effect (inset, Fig. 5d). Remarkably, all three devices achieved self-powered operation at zero bias (Fig. 5e). Under 1 mW illumination, T-sample generated a photocurrent of 0.14 nA at 250 nm, and became inactive in the visible region. Meanwhile, the R-sample exhibited a wavelength-dependent photocurrent that decreased significantly from 0.48 nA (at 250 nm) to 4 × 10−3 nA (at 830 nm). Both T- and R-samples demonstrate no response capability beyond their absorption ranges, so the photocurrent intensity is zero. In comparison, the B-sample displayed an exceptional photoresponse across the entire spectral range from 250 to 5000 nm and maintained a stable photocurrent (tens of nA), two orders of magnitude higher than that of the R-sample, which indicates its full-spectrum responsive capability.

Compared to the T-sample and R-sample, the B-sample demonstrated significant breakthrough performance enhancement and represented an improvement of approximately three orders of magnitude in both its responsivity and detectivity. Notably, this sample achieved a record responsivity of 6.1 mA/W at 250 nm, which is comparable to those of many conventional p-n junctions and polar-driven self-powered detectors54,55,56. Moreover, B-sample exhibited an excellent wavelength-insensitive response across a wide spectral range of 250–5000 nm, maintaining a stable responsivity at the mA/W level (Fig. 5f), along with a detectivity of >108 Jones (Fig. 5g). Furthermore, upon the application of a small bias voltage of 2 V, the responsivity increased to 1 A/W at 250–5000 nm (Figure. S18), demonstrating an excellent optical-to-electric conversion capacity of black Ce:CSBN crystal. Moreover, Fig. 5h depicts the long-term performance of the self-powered detector over 50 on-off cycles and demonstrates the excellent stability retaining over 98% initial photocurrent after 500 s of continuous operation. The response times of all three samples remain at tens of milliseconds (Figs. S11, S13, S16, and Table S4). In summary, benefitting from robust ferroelectricity and excellent optical absorption capabilities, the black Ce:CSBN crystal exhibited the superior self-powered photodetection performances.

Discussion

To evaluate the physical mechanism underlying the large photocurrent and high responsivity of the TTB ferroelectric crystals, the angular dependence of the photocurrent was examined in both the R- and B- samples (Fig. 6a, b). In these cases, the photocurrent component J reaches its minimum at θ = 0° (E // a-axis) and its maximum at θ = 90° (E // c-axis). According to the BPVE theory57,58,59, the photocurrent satisfies \({J\propto \left(E\cos \theta \right)}^{2}{\sigma }_{113}+{\left(E\sin \theta \right)}^{2}{\sigma }_{333}\), where E is the electric field, θ is the angle between E and the a-axis, and σ113 and σ333 are the corresponding shift current susceptibilities. The experimental polarization ratios were 2.82 and 2.54 for the R- and B- samples at 530 nm, which is consistent with theoretical predictions (2.0–3.6, Fig. S19)60. At other light wavelengths, the polarization ratios for these samples were located between 2.13 and 2.98. Polarization-resolved transmission spectra of the R-sample at 530 nm and 630 nm exhibited the ratios of the absorption coefficient (α), \(\frac{{\alpha }_{E\parallel c}}{{\alpha }_{E\parallel a}}\), were 1.06 and 1.10, respectively (Fig. S20)61, which is not consistent with photocurrent polarization ratio. Therefore, the polarization characteristics of photocurrent predominantly originate from the second-order optical response tensor σ, rather than the anisotropic absorption coefficient α.

Fig. 6: Optical response mechanism of black Ce:CSBN crystal.
figure 6

a Polar plot of the polarization angle-resolved photocurrent for the R-sample (λ = 530 nm). b Polar plot of the polarization angle-resolved photocurrent for the B-sample (λ = 530 nm). c Temperature-dependent self-powered photocurrent of the B-sample (λ = 250 nm). d Comparison of the self-powered photocurrents between the R, B, and R2 samples. e Wavelength-dependent responsivities of the T, R, and B samples. The absorption spectra of the three samples are also plotted for comparison. f Comparison of the photoelectric response ranges of various ferroelectric-based self-powered detectors, including T-sample (CBN), KTa0.59Nb0.41O3 (KTN41)19, LiNbO366, KH2PO4 (KDP)65, Bi4Ti3O1220, BaTiO320, Fe:BaTiO367, Fe:KTa0.41Nb0.59O3 (Fe:KTN59)68, AgNbO369, BiFeO322, [KNbO3]1−x[BaNi1/2Nb1/2O3−&]x (KBNNO)21, R-sample (red Ce:CSBN), Cu:KTa0.56Nb0.44O3 (Cu:KTN)22, and B-sample (black Ce:CSBN).

Full size image

In addition, the temperature-dependent self-powered photocurrent was measured at 250 nm using an incident light intensity of 1 mW. As shown in Fig. 6c, the B-sample exhibits a unique photocurrent dependence at different temperatures, reaching a maximum at ~180 °C. In the ferroelectric phase (T <Tc), the photocurrent increased monotonically with temperature, correlating to an enhanced dielectric constant and improved carrier mobility. The photocurrent peaks at Tc and its ms-level response speed excludes significant impacts of Vo migration62. Notably, despite the metal-like optical absorption observed in the B-sample, electrical measurements showed that this sample maintained its semiconducting properties, and its resistivity gradually decreased at high temperatures (Fig. S21). Moreover, the photocurrent exhibits negligible dependence on the electrode material, which likely rules out the possible contribution from Schottky junctions to the observed photovoltaic effect (Fig. S22)63. These results indicate that the superior optoelectronic performance of black Ce:CSBN originates from its intrinsic ferroelectric domain structures, where the built-in electric fields drive carrier separation, the domain walls provide conduction channels, and critical fluctuations close to Tc boost the photoelectric conversion efficiency.

Furthermore, the reversible self-powered detection properties of the R2-sample were also investigated. As revealed in Fig. 6d, both the R- and R2-samples exhibit low photocurrent responses in the 250–830 nm spectral range, with rapid attenuation being observed upon increasing the wavelength. In contrast, the B-sample displayed an enhancement in the photocurrent intensity by approximately two orders of magnitude. As shown in Fig. 6e, the responsivity trend closely matches the optical absorption spectrum. In the B-sample, the remarkable performance improvement directly results from the increased absorption coefficient (α) over the broadband spectral range, which demonstrates that defect engineering provides an effective approach for optimizing the optoelectronic performances of ferroelectric crystals.

Finally, a comprehensive comparison was performed between the current TTB-based detectors and other ferroelectric detectors, wherein it was evident that the B-sample exhibited an excellent comprehensive performance. More specifically, its self-powered responsivity at 250 nm reached 6.1 mA/W, which is 430 and 130 times greater than those of the T- and R- samples, respectively, in addition to being six orders of magnitude higher than that of conventional BaTiO3 (Fig. S23 and Table S5)64. More importantly, in terms of the optical activity, B-sample exhibits an ultra-broadband response range of 250–5000 nm, which is significantly larger than those of other traditional ferroelectric materials, including undoped KTN (280 nm)19, BiFeO3 (450–650 nm)22, and doped Fe:BaTiO3 (400–700 nm)65, Cu:KTN (250–1030 nm)22. To the best of our knowledge, the black Ce:CSBN crystal possesses the widest optical response range among all ferroelectric oxides used for self-powered detectors (Fig. 6f), which successfully addresses the long-standing tradeoff between a wide response range and a high responsivity. As detailed in Table S6, this response range is comparable to those of commercial photodetectors, i.e. Si, Ge, and quantum dots, indicating the great potentials of black Ce:CSBN crystal for broadband photon detectors and sensors.

In conclusion, a hierarchical defect-engineering strategy was proposed for TTB ferroelectric crystals to achieve full-spectrum optical absorption and boost self-powered photodetection. By modulating the cation-doping and thermally-altered oxygen vacancies, a black Ce:CSBN crystal was successfully fabricated, which exhibited broadband photodetection over a range of 250–5000 nm with a high responsivity, representing the widest response range among all ferroelectric-based detectors. More impressively, the reversible transition between red and black samples was achieved through thermal reduction and redox processes, thereby providing an approach for the dynamic modulation of ferroelectric materials. These findings pave the way for overcoming the performance limitations of ferroelectric materials and promoting the rapid development of self-powered broadband photodetectors. Furthermore, this hierarchical defect engineering strategy can be extended to other crystals (i.e., LiNbO3, BaTiO3, SrTiO3) and functional properties (i.e., pyroelectric and optical rectification), thereby providing numerous possibilities for new applications in the scientific frontiers. At longer wavelengths beyond 5000 nm, multiphonon-assisted lattice absorption occurs in the Ce:CSBN system, indicating that some exotic elementary excitations (e.g., polaritons and polarons) could be designed and modulated using this defect engineering approach.

Methods

Crystal growth

The CBN and Ce:CSBN crystals were grown using the Czochralski method. More specifically, stoichiometric polycrystalline raw materials, which were prepared via solid-phase reaction, were melted at 1600 °C for 1 h to ensure homogenization. After controlled cooling at a rate of 0.5–2 °C/min, a (001)-oriented seed crystal was used to initiate growth, followed by necking, shouldering, and isodiametric expansion under a high-purity N2 (99.999%) environment. Post-growth, the crystals were slowly cooled at a rate of 10–20 °C/h, annealed at 1150 °C for 20–30 h (30 °C/h heating rate), and cooled at a rate of 25 °C/h to reduce stress/defects. Following annealing, the desired high-quality single crystals were obtained.

Crystal characterization

The hysteresis loops of CBN and Ce:CSBN were obtained using a ferroelectric test system (RT-Precision LC, Radiant Technology) at a test frequency of 10 Hz, and using a test temperature ranging from room temperature to Tc. The temperature-dependent dielectric constants of CBN and Ce:CSBN were surveyed using an LCR meter (ZX8528A, Zhixin Precision Electronics) with a heating rate of 5 °C/min, and test frequencies of 1, 10, 100, and 1000 kHz. The absorption spectra of CBN, red Ce:CSBN, and black Ce:CSBN were recorded using a UV-vis spectrophotometer (UV-2600i) with a step length of 0.5 nm and a test range of 200–3000 nm. EPR spectroscopy (Bruker EMX Plus) was also employed to analyze the black Ce:CSBN, red Ce:CSBN2, and red Ce:CSBN samples. Additionally, XPS was performed using an aluminum target and a monochromatic Kα source, wherein the integral area of each peak represented the content of each oxygen species in the crystal. Integrated differential phase contrast (iDPC)-STEM images were obtained for the CBN and Ce:CSBN crystals using spherical aberration-corrected TEM (Titan Cubed Themis G2300). This device was equipped with a condenser Cs corrector, to significantly enhance the coherence and spatial resolution of the probe. Prior to performing the experiment, the directionally cut wafer was finely processed using a focused ion beam to obtain an electron-transparent thin region on the sample with a thickness of <100 nm to meet the requirements of atomic resolution imaging. Details regarding all other crystal characterization details are provided in the Supporting Information.

Thermal reduction and oxidation treatment

Thermal reduction treatment was performed by placing the crystals in a high-precision tube furnace and heating to 800 °C (5 °C/min heating rate) under a 5% H2/95% N2 atmosphere. After 40 h of isothermal annealing, the samples were cooled to room temperature at a constant rate. For oxidation treatment, identical thermal parameters were applied under a 20% O2/80% N2 atmosphere.

Acid etching treatment

After grinding and polishing the CBN and Ce:CSBN plates along the (001) and (100) directions, they were placed in a solution prepared by mixing 38% concentrated nitric acid (HNO3) and hydrofluoric acid (HF) in a volume ratio of 1:1. To ensure uniform corrosion, the experiment was conducted under a constant temperature (50 ± 1 °C) for 3 h. All operations were performed in a fume hood equipped with an exhaust system.

Self-powered photoelectric detection

The [001]-oriented CBN, red Ce:CSBN, black Ce:CSBN, and red Ce:CSBN2 crystals (10 × 10 × 1 mm) were used to perform the photoelectric detection experiments. Au interdigital electrodes were sputtered using a magnetron sputtering system (KT-1650PVD) with a mask plate (200 mA, 90 s), yielding 10 nm-thick electrodes with a 0.1 mm finger width. For the purpose of the photocurrent measurements, two light sources were employed, namely a commercial LED (250–1350 nm) and a lab-built laser system (1950–5000 nm). A Keithley-2450 source meter was used to collect the signals via conductive probes that were in contact with the electrodes.