Silicon is widely used in photodetection due to its low purification cost, high material purity, and compatibility with complementary metal-oxide-semiconductor (CMOS) processes. However, conventional silicon-based photodetectors cannot operate in the near-infrared (NIR) wavelength band due to the 1.12 eV bandgap of crystalline silicon^[1–3]. In response to the growing demand for photodetection in the NIR wavelength band, researchers have aimed at other narrow bandgap materials, such as germanium (Ge), cadmium sulfide (CdS), and indium gallium arsenide (InGaAs). Although photodetectors made from these materials have made significant progress, their widespread use is limited by high production costs, complicated manufacturing processes, and difficulties in integrating with silicon substrates^[4–6]. Consequently, it is essential to explore novel material approaches that combine ease of preparation, superior optoelectronic properties, and CMOS process compatibility.

In recent years, “black silicon” produced through hyperdoping and surface structure modification of silicon has gained significant research attention due to its high broadband spectral absorption properties. Among various fabrication methods, femtosecond-laser (fs-laser) processing has emerged as a pivotal technique. It takes advantage of the unique ultrafast thermal melting and resolidification processes induced during the interaction between fs-laser pulses and silicon substrates, enabling effective hyperdoping and modification of surface quasi-periodic structures^[7,8]. Recently, fs-laser hyperdoped silicon photodetectors have gained significant research interest due to their remarkable sub-bandgap absorption and ultrahigh photoresponse. These properties make them suitable for applications in solar cells, photodetectors, and many other fields^[9].

However, the uneven surface morphology and high dark current limit the use of fs-laser hyperdoped silicon for obtaining arrayed devices and integrated chips. During the interaction between the fs-laser and the material, the resolidification process of the molten material is rapidly completed at the nanosecond time scale after fs-laser irradiation. This rapid process allows the dopant concentration to exceed its solid solubility limit in silicon by several orders of magnitude. However, the fast phase transition during this process causes significant lattice damage and the formation of polycrystalline phases, which can negatively impact device performance^[10–13]. Notably, ultraviolet (UV) light is absorbed by the surface dopant layer of fs-laser hyperdoped silicon, and the surface defects further reduce the detectivity of the photodetectors, particularly in the UV wavelength band. Moreover, the ultrahigh peak power of the fs-laser creates a quasi-periodic microstructure on the surface of the hyperdoped silicon. The surface features, ranging from 1 to 2 µm, lead to spatial inhomogeneity in the surface dopant layer and potential surface leakage current. Briefly, the uneven surface morphology and high dark current are key bottlenecks that limit the industrial adoption of fs-laser hyperdoped silicon.

To address the challenges of fs-laser hyperdoping and achieve more uniform hyperdoped materials, researchers have focused on optimizing both the surface structural properties of hyperdoped materials and the concentration gradient of the dopant layer. By conducting fs-laser irradiation on silicon at elevated temperatures, smoother surface microstructures can be obtained^[14]. Additionally, controlling the substrate temperature during fs-laser irradiation improves the crystallinity of surface microstructures^[15]. Uniform peak-array microstructures have also been achieved by tailoring the light-matter interaction through pulse-shaping techniques^[16,17]. Recent achievements suggest that more uniform and smooth doping concentration gradients can be achieved by varying the average number of fs-laser pulses used to irradiate the silicon material^[18,19]. While these methods improve surface quality, they fail to eliminate inherent morphological irregularities, which remains a critical bottleneck. Large surface undulations persist, degrading interfacial bonding integrity between hyperdoped silicon and metallic interconnects—particularly detrimental to back-illuminated CMOS architectures and silicon-chip integration. Consequently, reconciling atomic-scale surface flatness with superior optoelectronic functionality emerges as the pivotal challenge for enabling scalable deployment of hyperdoped silicon devices.

Herein, we report a surface-structure-free flat fs-laser hyperdoped silicon material and its application in a high-performance photodetector. Unlike conventional hyperdoped silicon with micron-scale surface features, our strategy suppresses microstructure formation by three orders of magnitude (from µm to nm scale), yielding a defect-minimized monocrystalline surface. This morphological breakthrough directly addresses interfacial incompatibility in device integration. The proposed photodetector achieves a peak responsivity of 120.07 A/W and a specific detectivity of 1.27×1014Jones (1Jones=1cmHz1/2/W) at 840 nm, the highest reported for fs-laser hyperdoped silicon photodetectors. It also enables UV-enhanced photodetection with a responsivity of 15.09 A/W at 350 nm and NIR sub-bandgap photodetection with a responsivity of 12.50 A/W at 1170 nm. The synergy of atomic-flat surfaces [root mean square (RMS) roughness <1nm, AFM-verified] and UV–NIR photodetection capabilities resolves two longstanding limitations in silicon photonics: interfacial losses in hybrid integration and spectral constraints of silicon’s native bandgap. This dual advancement positions fs-laser hyperdoped silicon as a viable platform for CMOS-compatible photodetector arrays and multispectral sensing systems.

N-type monocrystalline (100) silicon wafers with a resistivity of 3000–5000 Ω·cm and a thickness of 420 µm were used to prepare hyperdoped silicon materials and photodetectors. The wafers were cleaned using the standard Radio Corporation of America (RCA) clean method, and 200 nm thick sulfur films were deposited on them through thermal evaporation. The wafers were irradiated by a Ti: sapphire regenerative oscillator-amplifier system (Spectra-Physics), which delivers laser pulses in the fundamental Gaussian mode, with a duration of 120 fs, a central wavelength of 800 nm, and a repetition rate of 1 kHz. The surface modification and hyperdoping system is shown in Fig. 1(a). The samples were fixed on a computer-controlled three-axis translation stage, enabling the fs-laser to irradiate the samples at a scanning speed of 1 mm/s. A Pockels cell was used as the electro-optical switch, with the switching frequency set to 10 Hz to ensure that each point on the silicon surface receives an average of two fs-laser pulses. The fluence of the fs-laser pulses injected into the irradiated material surface could be controlled precisely using a combination of a half-wave plate (HWP) and a Glan-Taylor polarizer (GTP). The fluence was set to 2.2kJ/m2 to get a flat fs-laser sulfur-hyperdoped silicon (fs-laser sulfur-hyperdoped silicon subsequently referred to as FSH silicon in this study). After preparing the FSH silicon, rapid thermal annealing was performed on the flat FSH silicon to activate the dopant and improve the junction. The annealing was carried out at 600°C for 600 s. To create a flat FSH silicon photodetector, 200 nm thick aluminum (Al) electrodes were deposited in a ring-shaped pattern on the top and a filled square on the bottom of the FSH silicon by thermal evaporation [Fig. 1(b)]. This method of flat FSH silicon is compatible with the existing CMOS fabrication process. A schematic illustrating the application of flat FSH silicon in CMOS processes is shown in Fig. 1(c).

The surface structure of the flat FSH silicon was characterized using field emission scanning electron microscope (SEM) and atomic force microscope (AFM). Secondary ion mass spectrometry (SIMS) was performed to measure the sulfur concentration. A Raman spectroscope (Renishaw inVia InSpect Raman microscope) was used to analyze the lattice structure of the FSH silicon, and a UV–VIS–NIR spectrophotometer was used to measure its absorption spectrum. The light from a 250 W tungsten halogen lamp was focused onto the surface of the FSH silicon photodetector after passing through a grating monochromator with a spectral resolution of 0.2 nm, a chopper, and a high-pass filter to measure responsivity. A lock-in amplifier was used to capture the photocurrent signal at the chopping frequency. The responsivity was calculated using the standard substitution method, with a calibrated commercial Si photodetector (DET100 A/M, working from 350 to 1100 nm) and Ge photodetector (DET50 B/M, working from 700 to 1700 nm). The current–voltage (I–V) characteristics of the flat FSH silicon photodetectors were acquired with a source-measure meter (Keithley 2410).

In traditional multipulse fs-laser hyperdoping, the pulses used to irradiate silicon materials are divided into two groups: preceding pulses (≤10pulses) and succeeding pulses (>10pulses). The preceding pulses irradiate the material surface, leading to melting and generating randomly distributed surface defects. The succeeding pulses interact with initial defects, resulting in a quasi-periodic surface energy distribution. This process progressively removes material in a quasi-periodic manner, resulting in the formation of stripe-like structures. As the number of pulses increases, the energy becomes concentrated in the troughs of the stripe structure, intensifying the ablation effect and ultimately forming quasi-periodic molten peaks on the material surface^[20–22].

A low-pulse processing method was used to prepare flat hyperdoped silicon materials, fundamentally different from traditional multipulse approaches. This approach prevents the energy localization and incubation effects associated with subsequent pulses, thereby mitigating the formation of surface irregularities. The low-pulse processing method widens the fluence range between the melting and ablation thresholds, creating ideal conditions for producing flat FSH silicon and optimizing processing parameters. Furthermore, sulfur hyperdoping is achieved by fs-laser irradiation of sulfur films thermally evaporated onto the silicon surface. Compared to the traditional fs-laser hyperdoping in the SF6 atmosphere, this method offers several advantages. First, it eliminates unnecessary hyperdoping of elemental fluorine. Due to their smaller atomic size, fluorine atoms tend to occupy interstitial sites, forming deep energy-level recombination centers that degrade the performance of the resulting devices. Second, it prevents the etching effect of fluorine on the silicon surface during fs-laser irradiation, further preserving the integrity of the material^[23,24].

SEM and AFM tests were conducted to compare the surface flatness of flat FSH silicon with structured FSH silicon, which was prepared via traditional standard processes, where each surface point was irradiated by an fs-laser with an average of 200 pulses at a fluence of 1.2kJ/m2. The characterization images are shown in Figs. 2(a)–2(c). SEM images reveal that the surface of structured FSH silicon has a distinct irregular molten peak array, whereas the flat FSH silicon surface is essentially structureless. To further evaluate surface undulations, a 10 µm long region was randomly selected from both samples for AFM measurements. As illustrated in Fig. 2(c), the variation in the height of the structured FSH silicon surface, measured as the difference between the highest and lowest points, reaches 1.48 µm. In contrast, the corresponding variation in the height for flat FSH silicon is only 2.09 nm, which indicates that the surface undulations of flat FSH silicon are nearly three orders of magnitude smaller than those of structured FSH silicon. RMS is the central parameter for quantifying surface unevenness and is defined as the square root of the squared mean of all height deviations: RMS=1N∑i=1N[Δy(xi)]2.(1)

Quantitative analysis of surface roughness across identical sampling regions revealed that the structured FSH silicon has an RMS value of 0.234 µm, while the flat FSH silicon has an RMS value of only 0.485 nm, a striking three-orders-of-magnitude reduction. The ultralow RMS of flat FSH silicon approaches the intrinsic surface roughness of single-crystalline silicon, confirming its atomic-level flatness. This exceptional planarization facilitates contact resistance suppression and eliminates microstructure-induced electric field inhomogeneities, thereby enhancing the consistency of the photodetector response, and will be applicable to high-precision photodetectors and CMOS-compatible processes.

The significant undulations in structured FSH silicon are caused by strong ablation effects during fs-laser processing. In contrast, the minimal undulations in flat FSH silicon arise from surface melting and resolidification processes. Based on the combined SEM and AFM results, it can be concluded that the flat FSH silicon surface is virtually structureless. While maintaining a structureless surface, flat FSH silicon achieves effective sulfur hyperdoping, as confirmed by SIMS measurements shown in Fig. 2(d). The sulfur doping concentration in the surface layer of flat FSH silicon exceeds 1018/cm3, surpassing the solid solubility of sulfur in silicon by more than two orders of magnitude. Moreover, although both flat and structured FSH silicon samples achieve sulfur hyperdoping, the doping distribution differs significantly. In flat FSH silicon, the sulfur-doped layer exhibits a uniform in-plane doping concentration. In contrast, the sulfur-doped layer of structured FSH silicon follows the undulating molten peaks, resulting in highly uneven in-plane doping concentration.

After obtaining a flat surface, the difference in lattice structure between flat FSH silicon and structured FSH silicon is a point of interest. Raman spectroscopy was employed to compare the lattice structures of flat FSH silicon, structured FSH silicon, and the N-type silicon substrate. The normalized Raman spectra are shown in Fig. 3(a). Both flat FSH silicon and the N-type silicon substrate show the characteristic peak of monocrystalline silicon at 521cm−1. The peak width of flat FSH silicon is similar to that of the N-type silicon substrate.

In contrast, the characteristic central peak of monocrystalline silicon in structured FSH silicon exhibits a noticeable shift to 519cm−1, accompanied by significant peak broadening. These observations indicate the presence of numerous amorphous or microcrystalline states in the lattice of structured FSH silicon, leading to increased atomic disorder. Conversely, the hyperdoped layers of flat FSH silicon predominantly maintain a monocrystalline phase. This improvement is due to the use of fewer laser pulses during the preparation of flat FSH silicon, which prevents the severe ablation effects observed in structured FSH silicon. Furthermore, the resolidification of molten silicon in flat FSH silicon occurs on a smooth, structureless surface, facilitating the formation of a defect-free monocrystalline phase. The resolidification process in structured FSH silicon takes place on a quasi-periodic peak array, where multidirectional stresses make it more likely for polycrystalline silicon to form.

Additionally, intense ablation during the laser doping process contributes to the formation of amorphous silicon with more defects, especially in areas of severe ablation in structured FSH silicon. The reduced surface defect density in flat FSH silicon offers two key advantages. First, it minimizes the dark current in flat-FSH-silicon-based devices. Second, it enhances the material’s potential for UV photodetection and expands its potential applications.

The absorption spectrum of the material is a key factor in photodetector fabrication. Figure 3(b) shows the absorption spectra of flat FSH silicon, structured FSH silicon, and the N-type silicon substrate. Structured FSH silicon shows almost complete absorption in the 400–1000 nm wavelength range and over 85% absorption in the 1100–2000 nm range, exceeding the bandgap limit of monocrystalline silicon. In contrast, the absorption of flat FSH silicon in the visible wavelength band is reduced by approximately 25% compared to that of structured FSH silicon. This reduction is primarily due to the absence of peak array structures found in structured FSH silicon. These structures increase the optical path length by reflecting incident light multiple times, which lowers reflectivity and boosts absorption. Notably, the N-type silicon substrate shows a significant drop in absorption near the UV wavelength band, around 400 nm. However, both flat FSH silicon and structured FSH silicon maintain high absorption in this range without any decline, highlighting their suitability for UV photodetection.

For wavelengths beyond 1100 nm, where sub-bandgap light is dominant, the N-type silicon substrate exhibits negligible absorption. However, flat FSH silicon retains an average absorption of 35%, which is 40% lower than that of structured FSH silicon. This reduction in sub-bandgap absorption, compared to the visible wavelength band, is attributed to three factors. First, sub-bandgap light penetrates deeper into the FSH silicon, where the light-trapping effects of surface structures become evident. Second, part of the sub-bandgap light absorption of the structured FSH silicon results from structural defects caused by laser irradiation. Raman spectroscopy confirms that flat FSH silicon has fewer structural defects, resulting in fewer defect-related sub-bandgap absorptions. It is important to note that defect-related sub-bandgap absorption does not contribute to photoelectric conversion because the absorbed energy cannot generate carriers. The laser-hyperdoped sulfur plays a key role in light absorption within the sub-bandgap wavelength band. The sulfur concentration in flat FSH silicon is slightly lower than in structured FSH silicon, which further leads to a decrease in sub-bandgap absorption^[25–28]. Despite this fact, flat FSH silicon achieves an average sub-bandgap absorption of 35%, breaking the bandgap limitation of monocrystalline silicon. It demonstrates that the relatively low hyperdoping concentration required for a flat surface does not significantly affect its absorption properties. Both flat and structured FSH silicon show high absorptance across a broad spectral range, forming the basis for their use as broadband spectrum, high-optical-gain materials in silicon photodetectors. This result underscores the potential of flat FSH silicon for high-performance photodetection across a wide spectral wavelength band.

After thermal annealing and electrode evaporation, a flat FSH silicon photodetector was successfully fabricated. Although good optoelectronic properties of materials do not always result in better device performance, a common challenge encountered by researchers, we conducted comprehensive tests on the spectral responsivity, dark current, and specific detectivity (D*) of the flat FSH silicon photodetector to check its performance. The spectral responsivity was measured using the substitution method, and the results are shown in Fig. 4(a). Under a reverse bias voltage of 5 V, the flat FSH silicon photodetector achieved a peak responsivity of 120.07 A/W at 840 nm, surpassing the responsivity of the commercially available silicon photodetector Thorlabs DET100A/M by two orders of magnitude. Additionally, in the UV wavelength band, the photodetector exhibited a responsivity of 15.09 A/W at 350 nm, significantly higher than that of structured FSH silicon photodetectors. The superior UV responsivity is attributed to the unique material properties of flat FSH silicon. UV light is primarily absorbed near the material surface. In flat FSH silicon, the hyperdoped layer has few structural defects and a lower density of deep energy-level recombination centers compared to structured FSH silicon, which reduces the probability of electron-hole recombination near the surface. At the same reverse bias voltage, minority carrier holes generated due to UV light absorption in the hyperdoped layer of flat FSH silicon photodetectors are less likely to recombine. Instead, they are more likely to be captured by trap centers in the hyperdoped layer of flat FSH silicon photodetectors, contributing to photoconductive gain. Conversely, in structured FSH silicon, these holes are more frequently captured by recombination centers, diminishing device performance. Conventional CMOS devices typically exhibit low responsivity in the UV band. Improving UV responsivity involves the integration of fluorescence conversion materials or low-dimensional heterojunction structures, which complicate fabrication processes. The flat FSH silicon, with its inherent UV-enhanced photodetection properties, provides a simpler and more efficient alternative for CMOS-compatible UV photodetectors, highlighting its potential for advanced photodetection applications.

The dark current of the flat FSH silicon photodetector is 417.22 nA, shown in Fig. 4(b), at the reverse bias voltage of 5 V. The better rectification characteristics and reduced dark current of the flat FSH silicon photodetector are due to the structural differences between flat and structured FSH silicon samples. In structured FSH silicon, the undulation of the hyperdoped layer leads to an uneven junction. When a bias is applied, the current flow through the depletion layer becomes uneven, reducing rectification efficiency. The flat hyperdoped layer in the flat FSH silicon photodetector creates a well-defined and uniform junction interface, enhancing rectification performance. Moreover, the absence of surface structures in flat FSH silicon reduces surface leakage currents by eliminating the prominent surface defects found in structured FSH silicon. Therefore, the flat FSH silicon photodetector achieves significantly lower dark current levels due to its improved junction interface and reduced surface leakage.

The specific detectivity is a key figure of merit for evaluating a photodetector’s performance, as it integrates both responsivity and noise characteristics. When dark current is the primary source of noise, the specific detectivity is calculated as follows: D*=A1/2R(2eIdark)1/2.(2)Here, A is the light-sensitive area, R is the responsivity of the device, e is the charge of an electron, and Idark is the dark current of the device^[29]. By calculating D*, the performance of the flat FSH silicon photodetector can be quantitatively checked. Under a reverse bias voltage of 5 V, the specific detectivity of the flat FSH silicon photodetector reaches an impressive value of 1.27×1014 Jones at 840 nm at room temperature. This performance dramatically surpasses that of commercially available silicon photodetectors and other fs-laser hyperdoped silicon photodetectors reported in the literature, as illustrated in Fig. 4(b).

These results demonstrate the superior performance of flat FSH silicon photodetectors, making them the optimal solution for fs-laser hyperdoped photodetectors. Notably, this holds true even when surface structure requirements are not considered, further underscoring their versatility and potential for practical applications.

The flat FSH silicon photodetector demonstrates UV–NIR wide-spectrum photodetection capabilities, with a responsivity exceeding 10 A/W across a broad wavelength band from 350 to 1170 nm. Table 1 presents the responsivity and specific detectivity of the photodetector at various wavelengths. The flat FSH silicon photodetector extends the infrared photodetection range beyond the typical 1100 nm limit of silicon-based photodetectors owing to sulfur hyperdoping while simultaneously enhancing UV photodetection. In summary, the flat FSH silicon photodetector offers wide-spectrum, high-performance photodetection, and excellent compatibility, making it an ideal material for integration into photodetection chip processes.

The mechanism responsible for the enhanced photoresponse of FSH silicon photodetectors, which outperform conventional silicon-based photodetectors, remains a key focus of ongoing research. Researchers from Harvard University have attributed the external quantum efficiency (EQE) exceeding 100% to photoconductive gain. However, since photoconductive gain does not occur in an ideal PN junction, the role of the fs-laser hyperdoped junction was not discussed in detail^[30]. It has led to ambiguity regarding whether the gain originates from photoconductive effects or the breakdown of the hyperdoped junction^[31]. In vertical-structure FSH silicon devices, the coupling between the photoconductive gain of the FSH silicon layer and the influence of the dopant junction on photocurrent makes it challenging to separate and analyze their individual effects.

To address this, we designed a lateral FSHSi-Si structure to investigate the gain mechanism in FSH silicon photodetectors. The experimental system is illustrated in Fig. 5(a). An N-type silicon substrate was used to create a flat FSH silicon photodetector region, where ring-shaped aluminum (Al) electrodes were placed on the surfaces of the FSH silicon and the silicon substrate. This configuration created two circuit structures: a single flat FSH silicon (FSHSi) circuit and a flat FSH silicon-substrate silicon (FSHSi-Si) lateral junction circuit. All the parameters of the two circuits were identical except for the presence or absence of the junction. By measuring the photocurrents under identical light intensity, bias voltage, and photoconductive gain, the impact of the junction on the photocurrent was tested. The photoconductive gain (G) was calculated using the following equation: G=tlifetimettransit=tlifetime·μ·VL2.(3)Here, tlifetime represents the lifetime of minority carriers, μ is the carrier mobility, V is the bias voltage, and L is the distance between two electrodes in the circuit^[32]. During the test, the light irradiated the region within the middle ring-shaped electrode [Fig. 5(a)], ensuring that the photogenerated carriers originate from the same area. As a result, the values of tlifetime and μ are the same in both circuits. By maintaining a high electric potential on the middle ring-shaped electrode in both circuits, we measured the photocurrents in each circuit separately under the same bias, ensuring that the voltage (V) is the same in both structures. Flat FSH silicon was used because the surface peak array structure of structured FSH silicon affects the spacing between the two electrodes in the circuit, making it difficult to determine and control. The flat FSH silicon ensures precise and consistent electrode spacing in both circuits. Furthermore, the ring-shaped Al electrodes formed ohmic contacts with both flat FSH silicon and substrate silicon. Raman spectroscopy confirmed that both surfaces had similar monocrystalline structures. It minimized the surface recombination effect, allowing reliable photocurrent measurements. The results, shown in Fig. 5(b), reveal that the photocurrent of the FSHSi circuit is consistently higher than that of the FSHSi-Si circuit under identical light irradiation and reverse bias. Under a reverse bias of 10 V, the FSHSi circuit achieved a peak photocurrent of 0.98 mA at 1010 nm, while the FSHSi-Si circuit produced only 0.25 mA at 1020 nm. These results confirm that FSH silicon without any junction exhibits inherent gain. In contrast, the FSHSi-Si junction imposes a blocking effect on photocurrent, diminishing the photodetector’s photoresponse. The photocurrent gain in the vertical structure of FSH silicon photodetectors is the result of the competition between the photoconductive gain of the FSH silicon and the blocking effect of the FSHSi-Si junction under reverse bias voltage.

The photoconductive gain in FSH silicon arises from the high sulfur concentration introduced via fs-laser hyperdoping, which acts as trap centers for holes and significantly extends the lifetimes of minority carriers. The gain depends on the ratio of the lifetime of the minority carriers to the transit time of the majority carriers. The prolonged carrier lifetimes explain the huge photoconductive gain observed in FSH silicon photodetectors^[33,35]. In vertical-structure FSH silicon photodetectors, the fs-laser hyperdoping forms a graded junction. While this junction’s blocking effect on photocurrent is minimal, it effectively suppresses dark current, offering an avenue for performance optimization. By modulating the junction, it is possible to design photodetectors prioritizing higher spectral photoresponse, lower dark current, or maximum specific detectivity, depending on application requirements.

Finally, research on the relationship between the photocurrent and irradiated light intensity was carried out. The signal light from a lamp was focused on the flat FSH silicon photodetector after passing through a chopper. The chopper periodically modulated the light intensity, and an electrometer recorded the real-time variations in photocurrent within the circuit. A chopper frequency of 14 Hz was chosen to ensure that the change in the speed of light intensity was much slower than the response speed of the photodetector. This experiment ensured that the measured photocurrent accurately reflected real-time changes in light intensity. The results, depicting the relationship between photocurrent and irradiated light intensity, are shown in Fig. 6(a).

Interestingly, when the irradiated light intensity exceeds the saturation threshold, the photocurrent begins to decrease. This phenomenon can be attributed to changes in the quasi-Fermi energy levels within the hyperdoped layer of the flat FSH silicon. When light irradiates the surface of the photodetector, photogenerated electrons and holes are transiently produced. The electron quasi-Fermi level rises while the hole quasi-Fermi level falls. At lower light intensities, the hole quasi-Fermi energy level remains higher than the hole trap energy level of the sulfur dopants. Under these conditions, holes are effectively trapped near the depletion region, and electrons are quickly transported out of the device. The external circuit supplies the electrons, allowing them to cycle repeatedly, resulting in significant photoconductive gain^[36,37]. However, as the irradiated light intensity increases, the hole quasi-Fermi energy level continues to decline and eventually falls below the trap energy level. When this occurs, the probability of holes being captured by trap centers diminishes. Consequently, fewer holes are trapped, which lowers the gain and causes the photocurrent to decrease, even with a further increase in light intensity, as illustrated in Fig. 6(b)^[38,39].

This intriguing saturation phenomenon warrants further theoretical investigation and experimental validation. Understanding the underlying mechanism could pave the way for innovative applications. The saturation light intensity characteristic holds great promise for developing anti-saturation and damage-resistant photodetection devices using flat FSH silicon. Future research into this behavior could lead to advancements in robust high-performance photodetector technologies.

In this study, we fabricated a flat fs-laser sulfur-hyperdoped silicon material alongside a UV–NIR wide-spectrum, high-responsivity flat fs-laser sulfur-hyperdoped silicon photodetector. The flat fs-laser sulfur-hyperdoped silicon shows a significant reduction in the size of the irregular peak-array surface features typically seen in traditional fs-laser hyperdoped silicon, achieving a three-orders-of-magnitude decrease while maintaining its monocrystalline structure and effectively incorporating sulfur hyperdoping. This material demonstrates an average optical absorption of 35% in the 1100–2000 nm sub-bandgap wavelength band, thereby overcoming the bandgap limitations of monocrystalline silicon. The flat fs-laser sulfur-hyperdoped silicon photodetector achieves a peak spectral responsivity of 120.07 A/W at 840 nm, along with UV-enhanced photodetection of 15.09 A/W at 350 nm and an NIR sub-bandgap photodetection of 12.5 A/W at 1170 nm, at a reverse bias of 5 V. Additionally, optimizing the laser-doped junction results in a very low dark current of 417.22 nA at 5 V reverse bias, with a specific detectivity of 1.27×1014Jones. This photodetection is the highest among reported hyperdoped silicon photodetectors, to the best of our knowledge, confirming the superior photodetection performance of the flat fs-laser sulfur-hyperdoped silicon photodetector.

To understand the mechanisms behind the observed performance, we designed an FSHSi-Si lateral structure. We explored the competing theories of photoconductive gain and junction-blocking effects, providing experimental evidence to support the theory. Furthermore, we identified the anti-saturation photodetection characteristics of the flat fs-laser sulfur-hyperdoped silicon photodetector, offering a novel silicon-based material with a simple structure and compatible processing methods for anti-saturation photodetection applications.

The unprecedented optoelectronic properties of flat FSH silicon, including its atomic-scale surface planarity, broadband spectral responsivity, and ultralow dark current, position it as a disruptive enabler for next-generation silicon photonics. We anticipate that a detailed study of the flat FSH silicon will highlight its potential in monolithic multispectral focal plane arrays, high-dynamic-range CMOS image sensors, and other silicon-based devices.