Self-filtering monolithic organic/PbS quantum dot photodetector for visible and short-wave infrared selective vision in low-light Authors: Seunghyun Oh, Byung Ku Jung, Hyun Woo Ko, Tae Hyuk Kim, Young Kyun Choi, Suyeon Jo, Gyeong Min Lee, Hyung Jin Choi, Ohhyun Kwon, Chan Hyeong Im, Sunghyun Han, Paul Hongsuck Seo, Gayoung Ham, Hyojung Cha, Donghee Park, Jisoo Hong, Han Jung Park, Sang-Jun Kim, Seon Kyu Yoon, Min-Chul Park, Soong Ju Oh, Jae Won Shim Published: 2026 ABSTRACT AbstractDual-mode photodetectors with vertically stacked photoactive layers enable bias-controlled, band-selective extraction from their constituent photoactive layers. While their applications ranging from non-invasive diagnostics to optical communication demand high-fidelity detection of faint light, performance is often limited by interlayer-derived structural complexity and noise. Here, we present an interlayer-free monolithic organic/PbS photodetector that achieves visible and short-wave infrared dual-mode operation with low noise and crosstalk. Inducing vertical and lateral phase-separation in the organic layer facilitates charge carrier dynamics that yield high specific detectivity without auxiliary interlayers, enabling simple device structures that detect small signals with high precision. This strategy can be utilized in a variety of photoactive layer combinations spectrally targeted for application-specific devices. Its practicality is demonstrated through single-pixel imaging, which reconstructs high-fidelity images across visible and SWIR bands even under light attenuation. Furthermore, its dual-mode capability enables silicon alignment through registration of front- and back-side features.Subject terms: Sensors and biosensors, Electronic devices Introduction Vertical stacking of spectrally complementary photoactive layers has emerged as a promising platform for multispectral photodetection1,2. Unlike conventional broadband photodetectors (PDs) that rely on external bandpass filters to isolate target bands, this monolithic architecture enhances photon utilization efficiency by avoiding filter-induced optical losses and supports denser pixel configurations for higher spatial resolution3. Moreover, its high degree of spectral flexibility through the selection of photoactive materials with tunable absorption profiles enables devices responsive at application-optimal wavelengths (λs). A representative class within this framework is the dual-mode PD that, within a single pixel, distinguishes between two spectral bands (e.g., visible and short-wave infrared (SWIR); near-infrared (NIR) I and NIR II) by using bias polarity to selectively extract charge-carriers photogenerated by each band4–8. Whereas their applications (e.g., noninvasive diagnostics, optical communication, semiconductor processing) require high-fidelity photodetection of even the fainter of lights, dual-mode PDs often face difficulties owing to limited photosensitivity and high intrinsic noise levels. Typical dual-mode PDs rely on an intermediate buffer/transport layer to stack photoactive layers, deplete optical signals, and, more importantly, direct photocarrier pathways, thereby suppressing interband crosstalk9,10. As spectral isolation is central to operation, this interlayer is often considered an essential component of the device architecture. Nevertheless, its incorporation not only imposes strenuous processing steps and restricts architectural versatility, but also introduces noise-generating interfacial defects detrimental to small-signal detection11,12. Although several studies have developed single-photoactive-layer designs that circumvent the incorporation of the interlayer, their dependency on film thickness inevitably exhibits a strong trade-off between spectral selectivity and quantum efficiency. Accordingly, an architectural strategy that concurrently overcomes these barriers is required for practical deployment. In this regard, organic materials are compelling photoactive components for dual-mode PDs. Their form variability, intrinsically low noise levels, and compatibility with spectrum-tunable counterparts (i.e., perovskite and quantum dot (QD)) could unlock a wide array of high-performance hybrid device architectures13–18. In particular, integrating QD materials, such as PbS, enables the hybrid device to detect longer wavelengths, where organic materials are typically insensitive19. In addition, the relatively low charge mobility of the organic layers helps suppress the high dark current characteristics of QDs, further enhancing the synergy between the two materials20. Nonetheless, one of the defining attributes of organic blends lies in their ability to undergo donor–acceptor morphology evolution during film formation to produce nanoscale phase arrangements with tailored domain size and vertical composition21–27. These structures yield distinctive charge-carrier dynamics, defining exciton dissociation sites and selective percolation pathways, thus offering a materials-level route to spectral discrimination without architecture-related penalties28. Yet, morphology-guided design remains underexplored in dual-mode PDs. Herein, we report an interlayer-free monolithic organic/PbS PD that achieves visible/SWIR dual-mode detection with high photoresponse, suppressed noise, and strong spectral isolation through organic morphology control. The device incorporates vertically and laterally phase-separated (VLPS) donor–acceptor domains that enforce vertical selectivity, suppress undesired carrier generation–recombination, and constrain bias-opposite carrier transfer at the organic/QD junction, thereby limiting interband coupling and attaining ultralow noise currents in both channels. Comprehensive comparison to a well-mixed (WM) analogue through meticulous characterizations highlights the optical and electrical self-filtering efficacies of the VLPS stack for dual-mode operation. Selection flexibility of organic donors and subsequent photoactive materials preserves spectral tunability, extending this strategy applicable to a wide array of applications. Significantly, the faint-signal detection capability of the optimized device enables sensing for highly demanding deployments. To validate the efficacy of the morphology-driven strategy in practical scenarios, single-pixel imaging was performed, in which the VLPS device yielded high-contrast RGB and SWIR images, demonstrating superior resolution and color balance even under low-illumination conditions. Additionally, its dual-mode utility was showcased via a through-silicon alignment demonstration, where bias polarity was utilized to selectively extract and overlay alignment keys from opposing sides of a wafer stack without the need for mechanical optical adjustments. Results Film morphology of the organic photoactive layers consisting of poly[4,8-bis(5-(2-ethylhexyl)thiophen-2-yl)benzo[1,2-b;4,5-b']dithiophene-2,6-diyl-alt-(4-(2-ethylhexyl)-3-fluorothieno[3,4-b]thiophene-)-2-carboxylate-2-6-diyl] (PTB7-Th) and [6,6]-phenyl-C71-butyric acid methyl ester (PC71BM) were controlled to form WM or VLPS using a combination of reported methodologies (Supplementary Fig. 1)29–31. Full dual-channel PD devices were fabricated based on these films in the structure of glass/ITO/ZnO/PTB7-Th:PC71BM/PbS-PbX2(X=Br,I)/ZnO/Al, as presented in Fig. 1a. The thickness of the organic films, determined by cross-sectional scanning electron microscopy, was optimized to 1.1 μm, which quenches visible transmission to the QD layer while avoiding performance penalties caused by excessive thickness (Fig. 1b, Supplementary Figs. 2 and 3)18. For the SWIR photoactive layer (PbS-PbX2), PbS QDs with an average diameter of 5.5 nm were synthesized and tuned via a solution-phase ligand exchange process with PbX2, then vertically stacked on the organic films (Supplementary Fig. 4). Energy level diagram of the materials is shown in Supplementary Fig. 5. Detailed fabrication process of the devices is provided in the “Methods” section. Visible and SWIR dual-mode PDs. a Schematic of monolithic organic–QD device. b Cross-sectional image of device captured using scanning electron microscopy. The scale bar represents 500 nm. c Atomic force microscopy phase images of WM (left) and VLPS (right) films at their surfaces. The scale bars represent 1 µm. d Time-of-flight secondary ion mass spectrometry (ToF-SIMS) measurements of WM (top) and VLPS (bottom) films. Device operation mechanism schematics of the ideal devices in e visible-mode and f SWIR-mode. To probe phase separation in the organic films, atomic force microscopy (AFM) phase topographies were acquired (Fig. 1c). The VLPS film exhibits larger domains, implying enhanced lateral segregation of the organic components and, consequently, a reduced donor–acceptor interfacial area. Normalized phase pixel distributions further show that the WM film surface is richer in PC71BM relative to PTB7-Th by a factor of 1.13, whereas the VLPS surface is richer in PTB7-Th by a factor of 1.38 (Supplementary Fig. 6). The pronounced enrichment of PTB7-Th at the organic/PbS junction (1.56× higher in VLPS), indicative of vertical stratification, would bifunctionally serve as an electron-blocking interlayer that suppresses undesired charge leakage across the photoactive layers. Vertical phase separation of donor and acceptor molecules was verified with time-of-flight secondary ion mass spectrometry, as presented in Fig. 1d. F− and C70− ions were counted to trace PTB7-Th and PC71BM, respectively. Whereas these ions are evenly dispersed throughout the WM film, a gradual distribution is evident in the VLPS film, confirming higher donor concentration at the film surface and higher PC71BM concentration near the substrate. Based on these organic morphologies, the operation mechanisms of the dual-mode PDs are illustrated (Fig. 1e, f). Under a forward bias (V), the device operates in visible-mode in which the front PTB7-Th:PC71BM layer serves as the detecting element, while the rear PbS/ZnO heterojunction serves as a photo-inactive charge-transport pathway via the Shockley–Read–Hall recombination process. Visible excitation in the organic film generates electrons that percolate through PC71BM to ITO and holes that traverse the hybrid stack towards Al. On the other hand, SWIR-generated electrons in the PbS layer face unfavorable transfer across the QD/organic interface set by the PTB7-Th lowest unoccupied molecular orbital barrier and poor transport through the thick organic, where they bimolecularly recombine, thus preserving spectral isolation32. When a reverse V is applied, the device operates in SWIR-mode and the roles interchange; SWIR is detected by the rear QD, whereas visible-generated carriers in the organic are readily suppressed33,34. The extent to which electron transfer at the organic/QD junction is impeded and recombination in the organic is efficient depends on donor–acceptor morphology. These mechanistic insights are corroborated and examined through subsequent characterizations. Figure 2a, b display the external quantum efficiency (EQE) spectra of the WM and VLPS devices under various V, respectively. Corresponding responsivity (R) spectra are provided in Supplementary Fig. 7. While their EQE in the visible λs at an identical forward bias is similar (e.g., ≈43% at λ = 635 nm, V = +1 V), the two devices exhibit distinct spectral profiles, reflecting their respective donor–acceptor distributions. At V = 0 V, diffusion-driven extraction favors electrons photogenerated in the vicinity of ITO35. Consequently, uniform WM yields a relatively flat spectral shape, whereas the acceptor-rich composition near ITO in the VLPS device results in a short-λ EQE enhancement. Forward bias extends charge collection depths into regions that absorb longer λs due to exponential optical density profiles consistent with the Beer–Lambert Law36. Accordingly, the EQE increase at longer λs is modest in WM but more pronounced in the VLPS due to its donor-rich upper layer. No SWIR signals are extracted in the visible mode, reflecting complete electrical filtering of the photogenerated electrons by the organic layers. Under reverse bias, both devices exhibit similar responses in the SWIR wavelengths (≈20% at λ = 1310 nm, V = −1 V), while an undesired, off-channel visible signal leakage is observed exclusively in the WM device. This is attributable to the visible-generated photoelectrons leaking at the PC71BM/PbS junction towards the Al electrode. Fringe patterns in the spectra originate from Fabry–Pérot interference, which modulates absorption in the PbS layer (Supplementary Fig. 8). Spectrum-selective photoresponse. EQE spectra at varying applied voltages of a device incorporating a WM or b VLPS layer. c Spectral crosstalk of the devices between operation at +1 V and −1 V. Rise time (τrise) and fall time (τfall) of the dual-mode PDs under d vis-mode (λ = 635 nm) and e SWIR-mode (λ = 1310 nm). f Electrochemical impedance spectroscopy (EIS) Nyquist plots of organic-only devices. To quantitatively assess spectrum-selective detection capabilities at opposing biases, the key feature of dual-mode PDs, the spectral crosstalk (CV1V2(λ)) was evaluated as: where V1 and V2 are representative V for each mode: +1 V for the visible mode and −1 V for the SWIR. As plotted in Fig. 2c, CV1=1VV2=−1V(λ) reaches record-low values of ≈−80 dB in the visible λs and ≈−60 dB in the SWIR λs for the VLPS device, enabled by effective signal isolation and high R in both channels. In contrast, the WM device exhibits a significant degradation in visible-regime crosstalk, reaching only ≈−40 dB. This disparity arises because the WM device parasitically extracts photocarriers generated by the visible-λs even when operating in the SWIR-mode, as shown in the EQE and R spectra. The evolution of these spectrum-selective characteristics was further examined at elevated optical power (Φopt), at which the contrast between the two device architectures becomes more pronounced. Visible illumination using a λ = 635 nm laser diode at Φopt = 2.3 mW (cf. sub-μW for R) markedly increases photocurrent leakage in the WM device under SWIR-mode operation, while the VLPS device maintains robust spectral rejection (Supplementary Fig. 9). In the WM device, abundant donor–acceptor interfaces near the organic/QD boundary generate a high density of free carriers, which fill charge-transfer and tail states and increase local photoconductivity37. This, together with the acceptor-rich organic/QD interface, establishes percolation pathways for electrons into the PbS, allowing visible-photogenerated electrons to be extracted at the top electrode. In contrast, the VLPS film characterized with a low exciton dissociation rate limit and high electron-transfer barriers preserves charge-selectivity even at high Φopt, suppressing illumination-induced leakage. These results underscore that morphology-engineered transport is key to low-leakage, dual-spectral operation. Photoresponse speeds of dual-mode PDs are critical factors for low-latency detection required in targeted applications. Under visible-mode, the VLPS organic layer exhibits faster rise and fall times (τrise and τfall; 57 and 41 μs) than the WM (163 and 102 μs) as PC71BM-aggregated domains near ITO guide electrons vertically into selective contacts, while the lower density of donor–acceptor interfaces and traps incur charge storage and slow de-trapping. (Fig. 2d). On the other hand, during SWIR operation, the slower τrise of the WM device could be ascribed to stronger hole-accumulation at the PC71BM-abundant organic/QD interface, which enhances Maxwell–Wagner polarization and transiently screens the field in the PbS, consistent with the slower τfall relative to its counterpart (Fig. 2e)38. Corresponding −3 dB frequency (f−3 dB) plots are presented in Supplementary Fig. 10. The deviation of measured τrise from calculated values (τrise ≈ 0.35/f−3 dB) indicates that additional dynamics beyond single-pole RC contribute to the transient response. The stronger charge-recombination behavior in the WM film was ascertained with electrochemical impedance spectroscopy (Fig. 2f). Conducted using organic-only devices in the structure of ITO/ZnO/PTB7-Th:PC71BM/MoOx/Ag, the recombination resistance of the VLPS device was substantially larger than that of the WM device. Figure 3a presents the dark current densities (JD) in relation to V. The VLPS device exhibits enhanced JD suppression in both operational modes, effectively inhibiting undesirable charge-carrier leakage. Specifically at V = +1 V, the JD values are 15.2 nA cm−2 for the WM device and 2.18 nA cm−2 for the VLPS device, whereas at V = −1 V, JD is 101.1 nA cm−2 for WM and 18.2 nA cm−2 for VLPS. The higher JD observed in the WM device can be attributed to (1) large donor–acceptor interfacial area promoting intra-gap trap states-assisted charge-carrier thermal generation, trapping/de-trapping, and Poole–Frenkel effects39,40, and (2) percolating donor–acceptor networks synergistically exacerbating current leakage. Additionally, the higher dark current under SWIR-mode operation relative to visible-mode operation originates from the narrower bandgap of the QD layer, which facilitates thermally activated carrier generation and thereby increases leakage current (Supplementary Fig. 11). Larger current flows under voltage bias can, in principle, degrade long-term stability by facilitating Joule heating that promotes interfacial reactions and trap generation over time (Supplementary Fig. 12). Photodetection characteristics at varying optical power. a Voltage (V)-swept dark current densities (JD). Device area is 0.1 cm2. b Schematic of the equipment setup for photocurrent measurement at varying optical power. c Current transients (I(tj)) of the WM- (left) and VLPS-based device (right) measured in their visible-mode (V = +1 V) in dark and under 635 nm illumination, and d corresponding root-mean-square noise currents (Irms; Irms = I(tj) − I(t) where I(t) is the average dark current). e Multiple Irms measurements and their distributions. The center line represents the median value; the whiskers, the min and max values; and the box limits, the lower and upper quartiles. f Signal-to-noise ratio (SNR) of the devices in their visible and infrared channels at low optical power. g Linear dynamic range of the PDs. h Specific detectivity (D*) spectra calculated using the median Irms and wavelength-specific D* values accurately obtained using noise equivalent powers (NEPs). Next, current transients (I(tj)) were measured in the dark and under illumination at a bandwidth (Δf) of 1 Hz to assess electronic noise and photo-characteristics41,42. Φopt of the laser diodes (635 or 1310 nm) were precisely controlled using the measurement setup shown in Fig. 3b. I(tj) acquired during stabilization in the dark was fitted to a biexponential decay function to obtain the average current (I(t)), which was subtracted from I(tj) to yield root mean square (rms) noise currents (Irms). Figure 3c, d representatively show the measured I(tj) and corresponding Irms of the devices in visible-mode operation (V = +1 V), respectively. Irms measurements were performed using multiple devices to account for device variability (Fig. 3e). A five-fold difference in the median Irms was observed between the WM (385.8 fA) and VLPS devices (74.8 fA) in their visible-mode, while the discrepancy was smaller in their SWIR-mode (WM: 980.1 fA, VLPS: 156.2 fA), consistent with their respective rms shot-noise currents (Irms,shot; Irms,shot ≈ (2q · JD · A · Δf)0.5, where q is electric charge, and A is device area) (Supplementary Fig. 13)43. Since the PDs were evaluated at a flicker-dominant frequency, measured Irms values reflect carrier number fluctuations from trapping/de-trapping processes at electronic trap sites44–46. Likewise, noise measurements in the frequency domain offer coherent analysis (Supplementary Fig. 14). Hooge’s law, which relates 1/f noise to trap density, is given by: where SG represents power spectral density, ID is dark current, α is the Hooge coefficient (a measure of energetic disorder), f is frequency, and N is the number of free charge carriers47,48. Under visible-mode operation, flicker-noise curves fitted to Hooge’s law indicate a greater effective Hooge coefficient in the WM morphology, indicative of prevalent capture–release events at the larger donor–acceptor interfaces. On the other hand, the curves attained from the devices under SWIR-mode operation exhibit a minuscule difference, implying near-identical trap states within the PbS films. Noise measurements of a VLPS device incorporating a MoOx interlayer between the organic and QD films are presented in Supplementary Fig. 15. A pronounced noise at the flicker-frequencies is observed, particularly in visible-mode, directly indicating a high density of trap sites induced by the introduction of the interlayer49. Photocurrent (Iph(Φopt); I(tj) under illumination subtracted by ID) was meticulously measured, particularly at low Φopt, where R typically deviates from linear dynamic range (LDR) linearity. These data enabled the derivation of the noise equivalent power (NEP), the Φopt at which a PD achieves an SNR (SNR = Iph(Φopt)/Irms) of 1, by interpolation of the SNR values at their unity crossing as shown in Fig. 3f. In visible operation, the low noise level of the VLPS device promoted a small NEP of 733 fW, compared to 5.73 pW of the WM device50. In the SWIR-mode, the attained NEP values were 48.6 and 13.2 pW for the WM and VLPS devices, respectively. Iph measured over a higher Φopt provided the LDR for each mode (Fig. 3g and Supplementary Fig. 16). Under organic layer operation, the limited donor–acceptor interfacial area in the VLPS film constrains efficient exciton dissociation at high generation rates; however, the stratified morphology provides efficient charge transport pathways at low generation rates, extending the lower bound of linearity. As a result, VLPS achieves a wider LDR (90 dB) than WM (85 dB), which is critically restrained at low Φopt by signal loss via geminate recombination and noise floor limits51. On the other hand, LDR under PbS operation was similar for both devices at ≈40 dB, generally unaffected by organic morphology. Specific detectivity (D*), an indicator of the ability of a PD to detect faint light and thereby image objects in adverse conditions with high fidelity, of the devices is calculated using the equations: Discrete D* points derived from NEP and full D* spectra from median Irms and R(λ) are presented in Fig. 3h. Notably, the champion VLPS device attained a peak D* of 4.3 × 1012 Jones at λ = 710 nm and 6.4 × 1011 Jones at λ = 1300 nm, underscoring VLPS as a high-performance platform for dual-mode PDs (Supplementary Fig. 17). Yet, it is important to note that the widespread method of extrapolating R measured within the LDR to Φopt = NEP overestimates D*, as evidenced by the discrepancy between NEP-based D* values and those yielded from Irms. Accurate, NEP-based D* values were 5.52 × 1010 Jones and 4.32 × 1011 Jones at λ = 635 nm for WM and VLPS devices, respectively, and 6.51 × 109 Jones and 2.39 × 1010 Jones at λ = 1310 nm. The high D* of the VLPS-based device in both modes indicates a strong aptitude for desired applications. A table comparing key metrics of this work to those of state-of-the-art dual-mode PDs is presented in Supplementary Table 1. Scene-level imaging across the visible and SWIR spectra was implemented under varying illumination to verify the high performance of the VLPS device when operating in a practical setting. In particular, the single-pixel imaging technique was adopted to directly translate the inherent capacities of the PDs, particularly Irms, into spatially resolved image quality. Detailed optical and mechanical layout of the imaging setup for each spectrum is illustrated in Supplementary Figs. 18 and 19, respectively. Figure 4a shows the subject used to obtain red, green, and blue (RGB) images in the visible mode (V = +1 V). RGB channels were processed independently by deriving linear mappings from spatial current histograms of each channel, then combined to form the complete images presented in Fig. 4b (Supplementary Figs. 20 and 21). Processing the three channels separately improves color balance and reduces artificial color bias from channel brightness or filter throughput52,53. Under high illumination, both the WM and VLPS devices deliver the contour of the scene; however, the VLPS image provides superior contrast and clarity with a flatter background and sharper edges. As the illumination is reduced to 0.1×, the WM device develops pronounced row-wise fluctuation and channel-dependent banding that washes out fine texture and locally disrupts color balance, whereas the VLPS device preserves object boundaries and overall color balance. Next, a United States Air Force resolution test target mask was used to image SWIR (1310 nm) signals under reverse bias (V = −1 V) (Fig. 4c, d, and Supplementary Fig. 22). At high illumination, although both devices reconstruct the bar patterns, the WM device displays background-level drift that softens bar edges and local contrast, while the VLPS device maintains more abrupt transitions. At reduced light intensities of 0.1×, contrast in the WM device rapidly collapses, resulting in loss of resolvable features; on the other hand, the VLPS device maintains image clarity. These results clearly demonstrate the high signal-to-noise ratio of the VLPS device, attributed to its low Irms and extended LDR into the lower Φopt, which enables robust signal extraction under identical optical conditions, particularly in low-light environments. These results align with previously observed photodetection performance and underscore the advantages of morphology-guided carrier routing in the VLPS film for low-light detection. Visible and SWIR single-pixel imaging. a Object used for visible imaging. The white box indicates the scanned field of view. b Obtained single-pixel RGB images reconstructed with independent R/G/B acquisitions from a WM (left) and VLPS (right) device under high (top) and low (bottom) illumination. c USAF resolution test target mask used for SWIR (~1310 nm) imaging. The scale bar represents 10 mm. d Acquired single-pixel SWIR images using a WM (left) and VLPS (right) device under high (top) and low (bottom) illumination. The scale bars represent 10 mm. For high-density packaging or three-dimensional integration in semiconductor processing, the accurate alignment of multiple wafers is essential to realize through-silicon vertical interconnect access (TSV). In practice, alignment inspection often relies on multiple sensors and complex optical instruments, which can introduce inter-channel parallax calibration burdens. To explore the feasibility of a simpler approach, the monolithic organic/PbS PD was used to capture both front- and back-side alignment marks simply by switching the applied bias. The proof-of-concept setup used to verify the competence of the VLPS device for such an application is illustrated in Fig. 5a. Dot-patterns emulating alignment keys were placed in front of and behind a stack of six wafers. While visible and SWIR light were delivered to the patterns using a beamsplitter, a forward bias (V = +1 V) or reverse bias (V = −1 V) was applied to selectively extract the front-side reflection or back-side transmission images, respectively. Figure 5b summarizes the alignment procedure. Dot-anchor features corresponding to each other in the visible and SWIR images were first identified, from which the transformation matrix and correction offset for physical alignment were estimated. To ensure an unbiased assessment, the detected points were partitioned into a fitting subset for transformation estimation and an independent hold-out set for validation54. Details on the hold-out-based alignment and evaluation procedure, and a process-level view of the workflow are presented in Supplementary Figs. 23 and 2455. In the final physically aligned state, the residual misalignment was quantified from the re-acquired visible image with respect to the reference SWIR image, while the overlapped SWIR image was used for qualitative visualization (Supplementary Fig. 25). Silicon alignment using a spectrum-selective monolithic organic/PbS PD. a Optical setup for through-silicon SWIR imaging (transmissive) and front-side visible imaging (reflective) of alignment patterns. The inset illustrates image acquisition and dot-anchor feature matching processes. The source model of the XY linear motorized stage (VT-80) is used with permission from Physik Instrumente. b Silicon alignment workflow consisting of image acquisition, feature matching, and physical alignment. Fitting points are used to estimate the transformation matrix M and correction offsets (Δθ, Δx, Δy). Residual misalignment after correction is quantitatively evaluated from the re-acquired visible image, while the overlapped image provides qualitative visualization of the final alignment. The dimension of the pattern mask is 35 mm × 35 mm, the diameter of each dot is <1.5 mm, and each pixel captures 81 μm × 81 μm. Following a physical transformation comprising a −9.43° rotation and a sub-millimeter-scale x, y translation, the alignment marks achieved seamless spatial convergence. The resulting mean residual misalignments were quantified at −5.63 μm in the x and −4.79 μm in the y. Considering the dimensions of the pattern and pixel, these minuscule residuals hint at the high image fidelities of the VLPS device, attributed to its high D*. These results demonstrate that molecular strategy-driven dual-mode PDs could potentially obviate the need for complex configurations in TSV-like alignment tasks, mitigating parallax and fixture-change errors. Discussion We demonstrate that organic morphology can be used as a primary design variable to realize an interlayer-free, dual-mode photodetector with low noise and minimal crosstalk. Inducing VLPS donor–acceptor domains reduces the donor–acceptor interfacial area, suppressing charge trapping–detrapping processes and yielding nearly an order reduction in Irms across both the visible and SWIR bands. Consequently, the VLPS device attains a superior D* of 4.32 × 1011 Jones at λ = 635 nm and 2.39 × 1010 Jones at λ = 1310 nm, enabling precise photodetection even under low-light conditions. Moreover, VLPS enhances spectral isolation by establishing vertical selectivity through suppressed percolative leakage and constrained bias-opposite transfer across the photoactive layers, maintaining robust operation even under high illumination. This morphological contrast leads to a pronounced improvement in CV1V2(λ), particularly in the visible range, where the VLPS device exhibits ≈−80 dB, twice of that of its counterpart. These comparative analyses confirm that morphology-guided carrier-routing enhances dual-mode integrity without the need for auxiliary interlayers. We effectively translated these device performances into practical applications, demonstrating that the VLPS architecture retains high-fidelity image reconstruction even under attenuated lighting conditions where the WM morphologies fail. Furthermore, the practical utility of this visible and SWIR spectrum-selectivity was validated through a silicon alignment process, achieving precise registration of front and back side features via simple bias switching. This strategy is compatible with diverse organic donors and subsequent photoactive layers, hence providing a simple, universal framework for application-specific spectral bands. Methods Materials PTB7-Th and PC71BM were purchased from Luminescence Technology Co., respectively, whereas chloroform (99%), chlorobenzene (99.8%), 2-methoxymethanol (99.9%), and ethanolamine (98%) were purchased from Sigma-Aldrich. Lead acetate trihydrate (Pb(Ac)2·3H2O, 99.999%), oleic acid (OA, 90%), 1-octadecene (ODE, 90%), bis(trimethylsilylmethyl) sulfide ((TMS)2S, 98%), dimethylformamide (DMF), butylamine (BTA, 99.5%), dimethylsulfoxide (DMSO, 99.9%), toluene (99.8%), ethanol (99.5%), octane (99%), acetone (99.8%), cesium bromide (CsBr, 99.9%), lead bromide (PbBr2, 98%), zinc acetate dihydrate (ZnAc2·2H2O, 99.99%), methanol (99.8%), potassium hydroxide (KOH, 90%), and lead iodide (PbI2, 99.999%) were purchased from Sigma-Aldrich. ZnO preparation ZnO(sol-gel) used as the bottom electron transport layer was prepared by stirring 1640 mg of zinc acetate dihydrate in 10 ml of 2-methoxyethanol and 0.5 ml of ethanolamine overnight in ambient conditions. ZnO(nanoparticles(NPs)) were synthesized by placing zinc acetate dihydrate (2.95 g) and methanol (125 mL) into a 250 mL three-neck flask and heating the solution to 65 °C. Separately, KOH (1.48 g) was dissolved in methanol (65 mL), then slowly added to the heated flask. The mixture was maintained at 65 °C for 10 h, then cooled to 25 °C via natural cooling. The resulting ZnO(NPs) was purified by centrifugation at 4530 × g for 5 min, repeated three times with methanol. Finally, the purified ZnO(NPs) was dispersed in a chloroform/methanol mixture at a concentration of 75 mg mL−1. Colloidal quantum dot (CQD) synthesis and CQD ink preparation A modified hot-injection method was employed to synthesize the PbS-CQDs; first, Pb(Ac)2·3H2O (1.53 g), OA (22 mL), and ODE (20 mL) were loaded into a 250 mL three-necked flask and degassed at 120 °C for 2 h56,57. After degassing, under N2 atmosphere, the sulfur precursor of (TMS)2S (0.4 mL) in degassed ODE (20 mL) was swiftly injected into the flask, and the reaction mixture was allowed to cool naturally to 25 °C. 45 mL of acetone was added to the crude solution of PbS-CQDs, followed by centrifugation at 5600 × g for 3 min. The supernatant was discarded, and the precipitate was redispersed in toluene (15 mL), then purified twice by adding ethanol (10 mL) as an antisolvent. The final precipitate was dried, redispersed in octane, and stored in a glovebox. The solution phase ligand exchange method was utilized to exchange the surface of PbS-CQDs from OA to PbX2 (X=Br, I)58. PbI2 (0.33 g), PbBr2 (0.11 g), and CsBr (0.022 g) were dissolved in 10 mL of DMF. Subsequently, 20 mL of the synthesized QDs (2.5 mg mL−1) and the DMF solution were vigorously mixed for 1 min, after which the CQDs were completely transferred to the DMF phase. The DMF solution was then washed two times using octane. After the exchange process, CQDs were then separated by centrifugation at 5600 × g for 3 min, followed by a vacuum-drying process. Precipitated CQDs powder was dispersed in a mixed solvent of 4:1:0.2 (volume ratio) BTA: DMF: DMSO = (2 g mL−1). Finally, this high-concentration solution was redispersed in BTA (500 mg mL−1). Fabrication of dual-mode photodetectors ITO-covered glass substrates were cleansed in detergent-containing deionized water (DIW), DIW, acetone, and isopropanol, respectively for 20 min each, using an ultrasonic bath. ZnO(sol-gel) solution was dispensed through a 0.45 μm polytetrafluoroethylene (PTFE) filter and spin-coated onto the substrates at 5000 rpm for 60 s, then thermally annealed at 200 °C for 30 min. Next, the organic photoactive layer of PTB7-Th:PC71BM (1:1.5, 75 mg/ml; stirred at 55 °C for 4 h) was spin-coated in a nitrogen-filled environment. To control morphology via solvent-dependent solidification pathways, two distinct processes were employed. WM films used a chloroform:chlorobenzene (9:1) mixture spin-coated at 1500 rpm for 10 s, followed by an inverted off-center spinning at 2000 rpm for 30 s. The high volatility of chloroform and the external centrifugal/gravitational forces maintain a uniform vertical composition30. The VLPS film used chlorobenzene spin-coated at 800 rpm for 30 s followed by 2 min of ambient drying. The slower evaporation of a lower vapor-pressure solvent prolongs the wet-film state, allowing PC71BM to diffuse toward the high surface-energy ZnO interface and coalesce into larger nanoscopic domains for an energetically stable, phase-separated state. Both films were subsequently thermally annealed at 100 °C for 10 min. Both films were formed with an identical thickness of 1.1 μm. CQD ink was deposited on the organic-covered substrates via dynamic spin-coating at 8000 rpm for 30 s for an optimum thickness (≈450 nm), followed by thermal annealing at 80 °C for 5 min to remove solvent residues. After cooling down the substrate, ZnO(NPs) were dynamically spin-coated at 8000 rpm for 30 s. Finally, an aluminum electrode of 100 nm was deposited through a shadow mask using thermal deposition under high vacuum pressure of ~10−8 hPa. Device characterization AFM measurements were carried out using an XE-7 system (Park Systems). Phase imaging was performed in tapping mode to probe variations in local mechanical and adhesive properties across the sample surfaces. The cantilever oscillation phase shift was recorded directly in degree units. Spectral EQE and R were measured at 50 Hz using a xenon lamp-based spectral measurement system (K3100 IQX, McScience), optical chopper (MC2000B-EC, Thorlabs Inc.), lock-in amplifier (SR830 DSP, Stanford Research Systems), low noise current preamplifier (SR570, Stanford Research Systems), and source measurement unit (SMU; Keithley 2601A, Tektronix). A Si PD (S1337-1010BQ, Hamamatsu) and a InGaAs PD (G12180-050A, Hamamatsu) were used as references for the incident optical power. 121 discrete points were obtained for ~30 min for accurate characterization. Electrochemical impedance spectroscopy (EIS) measurements were performed using an SMU (Keithley 2400, Tektronix) and an LCR meter (WK4100, Wayne Kerr). Response times and –3 dB frequency values were obtained using a digital storage oscilloscope (DSOX1204G, Keysight) connected to a 50 Ω parallel resistor. Voltage-swept dark current was measured with an SMU (Keithley 2636B, Tektronix). 171 points were obtained across the swept voltage range for 171 s to account for transient charge behaviors59. Transient time current, I(tj), was typically measured for 1.5 h at a Δf of 1 Hz using an SMU (Keithley 2636B, Tektronix). For photocurrent measurements, the optical powers of 635 nm and 1310 nm-illuminating laser diodes (DJ532-40 and L1330P5DFB, Thorlabs Inc.) were controlled with a current controller (LDC205C, Thorlabs Inc.), and the temperatures were maintained at 298 K with a temperature controller (TED200C, Thorlabs Inc.). Neutral-density filters were used to vary input optical densities. For precise measurements, the Si and InGaAs PDs and optical power sensor (S130VC, Thorlabs Inc.) were used for calibration. Noise current spectra of the PDs in the frequency domain were measured by amplifying signals using the low-noise preamplifier and recording them at desired frequencies with the lock-in amplifier. Imaging setup Single-pixel imaging was performed using a custom scanning platform supporting both visible reflection and SWIR transmission. For visible imaging, illumination was provided by two LED panels (LUXPAD22H, Nanguang) with sequential bandpass filters at R (610 ± 30.5 nm), G (525 ± 30.5 nm), and B (450 ± 40 nm). A 50 mm lens (AF-S NIKKOR, Nikon) defined an object–lens distance of 150 mm and a lens–sensor distance of 75 mm (≈1/2 magnification). The PD was translated on an XY stage with a 0.1 μm step, and collected three measurements per color for averaging. The low condition (0.1×) used an absorptive ND filter (NE10A, Thorlabs Inc.) in front of the PD, while the high (1×) condition omitted it. Bias applied to the organic–QD PDs switched between +1 V and −1 V to select visible or SWIR channel. For silicon alignment imaging, a 1300 nm LED source (M1300L4, Thorlabs Inc.) was conditioned by a 100 µm pinhole and a plano-convex lens (LA1027-C-ML; f = 35 mm, BBAR 1050–1700 nm); the object–lens distance was set to 105 mm, and an object–sensor distance, 52.5 mm (~1/×2 magnification). SWIR transmission datasets were acquired from a silicon stack carrying a custom 35 mm × 35 mm dot-pattern transmission mask that emulates an alignment key. Low-light inspection conditions were implemented by driving the visible and 1300 nm light sources with a programmable LED driver (Thorlabs DC4100/UPLED). In both visible and SWIR measurements, the same VLPS PD was used, with photocurrent measured using an SMU (Keithley 2636B) and sample motion controlled by a stepper motor controller (C-663, PI). Supplementary information Supplementary Information Transparent Peer Review File Source data Source data Acknowledgements Dr. Min-Chul Park acknowledges Professor Hiroshi Harashima (The University of Tokyo) for insightful discussions and valuable comments on this work. Author contributions S.O., B.K.J. and H.W.K. contributed equally to this work. S.O. conceptualized this study, performed experiments, measurements, data curation, investigation, and wrote the manuscript. B.K.J. performed experiments, measurements, investigation, and drafted the manuscript. H.W.K. designed the imaging setup, performed imaging, and drafted the manuscript. T.H.K. and Y.K.C. provided feedback on PD measurements and the manuscript. G.M.L. and H.J.C. conducted optical analysis. S.J., C.H.I., S.H., P.H.S., D.P., J.H., S.-J.K. and S.K.Y. assisted with imaging and its analysis. O.K., G.H., H.C. and H.J.P. assisted with organic film analysis. M.-C.P., S.J.O. and J.W.S. supervised this project. Peer review Peer review information Nature Communications thanks Furong Zhu and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Funding This work was supported by the Korea Institute of Science and Technology (KIST) Institution Program (no. 26E0002), by the Technology Innovation Program (or the Industrial Strategic Technology Development Program) grant funded by the Ministry of Trade, Industry & Energy (MOTIE, Korea) (no. 02316245), and by the National Research Foundation of Korea (NRF) grants funded by the Korean Government (MSIT) (no. RS-2022-NR070314, no. RS-2024-00438059, no. RS-2026-25468502, and no. RS-2026-25471259). Data availability The data generated in this study are provided in the Source data file and are also available from the corresponding authors upon request. Source data are provided with this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Seunghyun Oh, Byung Ku Jung, Hyun Woo Ko. Contributor Information Min-Chul Park, Email: minchul@kist.re.kr. Soong Ju Oh, Email: sjoh1982@korea.ac.kr. Jae Won Shim, Email: jwshim19@korea.ac.kr. Supplementary information The online version contains Supplementary material available at 10.1038/s41467-026-74407-z. References Associated Data Supplementary Materials Supplementary Information Transparent Peer Review File Source data Data Availability Statement The data generated in this study are provided in the Source data file and are also available from the corresponding authors upon request. Source data are provided with this paper.