# Fileset

[s41467-026-75365-2 (1).pdf](https://mdr.nims.go.jp/filesets/f9865b08-ab5f-4349-bb1d-1e1e6c4af48c/download)

## Creator

Yunqing Kang, Olga Guselnikova, Lei Fu, Norman C.-R. Chen, José Julio Gutiérrez Moreno, [Joel Henzie](https://orcid.org/0000-0002-9190-2645), Yuqi Guo, Xiangyang Liu, Yingji Zhao, Wei-Hsiang Huang, Min-Hsin Yeh, Yagmur Tanis, Markus Valtiner, Stanislav Ježek, Jun Zhou, Yusuke Asakura, Kevin C.-W. Wu, Yusuke Yamauchi

## Rights

[Creative Commons BY-NC-ND Attribution-NonCommercial-NoDerivs 4.0 International](https://creativecommons.org/licenses/by-nc-nd/4.0/)

## Other metadata

[Low curvature combined with high amorphization improves SERS sensitivity of semiconducting Rh–Se mesoporous nanospheres toward non-adsorbing resistant pollutants](https://mdr.nims.go.jp/datasets/068ddb0c-6cef-4545-a565-c266b6d80d0f)

## Fulltext

Article https://doi.org/10.1038/s41467-026-75365-2Low curvature combined with highamorphization improves SERS sensitivity ofsemiconducting Rh–Se mesoporousnanospheres towardnon-adsorbing resistantpollutantsA list of authors and their affiliations appears at the end of the paperAmorphous metal–selenium (M–Se) alloys hold promise for applications suchas catalysis, electronics, and surface-enhanced Raman spectroscopy (SERS)sensing due to their surface reactivity, isotropic properties, and tunablebandgap. However, controlling the synthesis of nanostructured noble-metal-based M–Se materials is difficult, as the strong metallic bonding and lowchemical reactivity of noblemetals hinder homogeneous selenization.Hereweshow a one-step micellar self-assembly reduction method to produce amor-phous semiconducting Rh–Se mesoporous nanospheres (a-Rh–Se MNs). Wesystematically investigate how composition, surface curvature, and crystal-linity affect SERS performance. a-Rh–Se MNs with low curvature and highamorphization achieve enhanced sensitivity toward typically non-adsorbing,low-polarizability analytes, including hexachlorobenzene, anthracene, andpolytetrafluoroethylene microplastics (even in wastewater). Combinedexperimental and theoretical analyses reveal that co-modulating curvatureand amorphization tunes key electronic descriptors (e.g., d-band center andwork function), thereby promoting resonant charge transfer. This findingoverturns the conventional paradigm that prioritizes high-surface-area sub-strates. Our work establishes mesostructured a-Rh–Se for SERS applicationsand provides insights into M–Se surface chemistry for designing advancedSERS materials.Amorphous materials offer distinct advantages, including enhancedsurface reactivity, isotropic properties, and the absence of grainboundaries, which can be advantageous for applications in catalysis,sensing, and electronics1–3, although their intrinsic structural dis-order may also introduce defect states and reduced charge carriermobility. Conventional routes to amorphous phases, such asquenching rapidly from the molten state, are well established4.However, these methods face limitations when applied to close-packed noble metals like Ag, Au, Ir, Pt, and Rh due to their inherentpropensity for rapid crystallization5. Although incorporating non-metallic (e.g., P, Se) or metalloid elements (e.g., B, Te) via wet-chemical reduction effectively suppresses crystallization and pro-motes amorphous alloy formation6–8, precise atomic-level control ofamorphous nanostructures remains a challenge.Received: 19 September 2025Accepted: 26 June 2026Check for updatese-mail: yunqing.kang@uq.edu.au; olga.guselnikova@tuwien.ac.at; y.yamauchi@uq.edu.auNature Communications |         (2026) 17:9592 11234567890():,;1234567890():,;http://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-75365-2&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-75365-2&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-75365-2&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-75365-2&domain=pdfmailto:yunqing.kang@uq.edu.aumailto:olga.guselnikova@tuwien.ac.atmailto:y.yamauchi@uq.edu.auwww.nature.com/naturecommunicationsMetal–selenium (M–Se) amorphous alloys (e.g., Rh–Se, Pt–Se)have emerged as a compelling class of metal chalcogenide semi-conducting materials exhibiting advantageous properties9–13. Metalchalcogenides have shown promising performance in areas such asthe electrocatalytic nitrogen-reduction reaction14 and surface-enhanced Raman scattering (SERS)15,16. Despite their potential, sev-eral challenges hinder the rational design and application of amor-phous M–Se materials. First, the intrinsic long-range atomic disorderof amorphous structures makes precise control over nanoscale mor-phologydifficult5. Amorphousmetal selenides are typically obtained inthe formof nanoparticles14,17, rather than in well-defined architectures,suchasmesoporous structures,which areparticularly attractive due totheir tunable porosity and broad application potential18. Second, theformation of amorphous phases is highly sensitive to composition19,yet the relationship between the M/Se ratio and phase evolutionremains poorly understood. Third, a systematic understanding of howdifferent degrees of crystallinity influence structural and chemicalfunctional properties is still lacking. Fourth, the surface and interfacechemistry of amorphous M–Se materials remains insufficiently con-trolled. These knowledge gaps limit advances in the controlledsynthesis and structure–property correlation of amorphous materials.Recently, the deliberate manipulation of nanoscale curvature hasemerged as a strategy for tailoring material properties20–23. Suchgeometry-dependent effects have also been observed in relatednanostructured systems, where size or curvature influences surfacechemistry and interfacial behavior24. For example, nanostructures withsmaller surface curvature radii resulted in higher-concentration iondistributions for enhanced capacitive deionization performance20.Chen et al.21 reported that the curvature engineering of Pt nanoconescan greatly promote the H3O+ supply rate during electrocatalytichydrogen evolution reaction. In the context of SERS, manipulatingcurvature alters the interaction between adsorbates and the semi-conductor surface in many different aspects (surface area, specificadsorption sites, electronic structure, band alignment, etc.), therebyinfluencing chemical enhancement25. Inspired by these insights, wehypothesize that integrating amorphization with curvature modula-tion in Rh–Semesoporous nanostructures offers an effective platformfor tuning semiconductor SERS activity. This dual approach mayenable detection beyond model analytes, such as dyes or thiols, butthe detection of non-adsorbing and low-polarizable analytes remains along-term challenge, limiting semiconductor applications in SERS. Thecommonly used semiconductors, such as Cu2O26, ZnO27, TiO228,MoTe229, ZnSe30, and CdSe31, show limited adsorption toward non-adsorbing and low-polarizable analytes, leading to poor electroniccoupling32.Here, we present a facile, one-step synthesis of amorphous Rh–Semesoporous nanospheres (a-Rh–Se MNs) with tunable curvature viamicellar self-assembly chemical reduction. We systematically investi-gate the influence of composition and thermal treatment temperatureon the resulting crystallinity and, critically, on their performance inSERS. We find that a-Rh–Se MNs with lower surface curvature andreduced surfacearea exhibit SERSactivity, contrary to the intuitive andcommonly observed high SERS sensitivity of high surface area mate-rials. Unlike the commonly reported SERS sensing of highly polarizablemolecules—which readily adsorbonto and interactwith the substrate—we focused on the more challenging detection of low-polarizability,surface-resistant pollutants. The lowest limit of detection (LOD) forhexachlorobenzene (HCB) and polytetrafluoroethylene (PTFE) micro-plastics (MPs) was achieved with low curvature and highly amorphousa-Rh–Se MNs. We attribute this performance to the synergy betweenthe amorphous structure (providing optimum electronic features andabundant active sites) and nanoscale curvature modulation (tuningsurface energy levels, thereby improving electronic coupling andorbital alignment with the analyte). This work provides deeper insightsinto the synthetic chemistry of amorphous nanomaterials andsignificantly advances the understanding of the structure–propertyrelationships in these complex systems. Our results challenge con-ventional SERS design paradigms and offer insight intosemiconductor–analyte interactions critical for real-world pollutantsensing.ResultsSynthesis of a-Rh–Se MNs with various compositionsThe amphiphilic poly(ethylene oxide)-b-poly(methyl methacrylate)(PEO-b-PMMA) diblock copolymer served as a dynamic soft templateto engineer tunable compositions and mesoporosity in a-Rh–Se MNsvia a one-pot micelle-mediated reduction (Fig. 1a). Upon introducingRh/Se precursors (Na3RhCl6/H2SeO3) and water into the tetra-hydrofuran (THF)-polymer solution (see Methods), PEO coronasselectively sequestered metal ions around PMMA cores, initiatinguniform micelle nucleation (validated by the Tyndall effect in Sup-plementary Fig. 1 and Supplementary Movie 1). Subsequently, thesolution was left at 80 °C for 12 h, during which formic-acid-drivenreduction and Rh–Se coalescence occurred within the micellar scaf-fold. Subsequent template removal unveils an interconnected porousframework of typical a-Rh–Se MNs with relatively large mesopores (a-Rh–Se-L MNs), as observed in scanning electron microscopy (SEM)images (Supplementary Fig. 2), bright-field transmission electronmicroscopy (TEM) image (Fig. 1b, c and Supplementary Fig. 3a), andscanning transmission electron microscopy (STEM) images (Fig. 1dSupplementary Fig. 3b, c). The network of mesopores is absent in thenonporous control sample (i.e., a-Rh–Se-np, Supplementary Fig. 4).The composition ratio of a-Rh–Se MNs can be precisely tuned bycontrolling the Rh/Se precursor ratio (see “Methods”), with the finalalloy composition following the same trend as theprecursor feed ratio,as confirmed by inductively coupled plasma optical emission spec-trometry (ICP-OES) and SEM energy-dispersive X-ray spectroscopy(EDS) (Supplementary Table 1 and Supplementary Figs. 5–8). PowderX-raydiffraction (XRD) analysis was employed to probe the correlationbetween crystalline structure and Rh/Se stoichiometry. We observedthat as the Rh content in Rh–Se alloys increased, the mesostructureremained (see SEM images in Supplementary Figs. 6 and 7); however,at a critical Rh composition of ≈67.8 at.% (i.e., feeding molar ratio ofRh/Se = 3:1), the face-centered cubic (fcc) Rh crystalline phase becamepredominant (Fig. 1e). Conversely, progressive reduction in the Rhratio (i.e., increased Se content) led to a gradual predominance ofamorphous character. This inverse relationship implies that Se acts as apivotal amorphous-phase inducer in these alloy systems. We probeddifferent Rhx–Sey compositions towards HCB sensing via Ramanspectroscopy with 532nm excitation wavelength. HCB (10−6 M) waschosen as a probe molecule, and the peak at 375 cm−1 (C–Cl stretchingvibration) was used as a characteristic for the enhancement factor (EF)analysis because this mode provides a strong and spectrally isolatedsignal with minimal overlap from substrate or background contribu-tions, enabling reliable quantitative comparison across samples. Wefound that a-Rh–Se-L MNs with a feeding molar ratio of Rh/Se = 1:3provide the highest EF > 4 × 103, and a further increase of Rh percen-tage does not lead to an increase of EF (Supplementary Fig. 9).Therefore, further studies focused on a-Rh–Se-LMNswith a final Rh:Seatomic ratio of 20:80 (obtained by ICP-OES, Supplementary Table 1).The amorphous nature of the typical a-Rh–Se-L MNs was alsocarefully studied by using characterization techniques of high-resolution TEM (HRTEM) and selected-area electron diffraction(SAED). The absence of distinct lattice fringes in the HRTEM image(Fig. 1c) and the broad diffraction rings in the SAED pattern (inset ofFig. 1c) collectively confirmed the amorphous nature of the a-Rh–Se-LMNs, as also evidenced by diffuse fast Fourier transform (FFT) rings(Supplementary Fig. 10). STEM-EDS maps show a uniform distributionof Rh and Se throughout a single MN (Fig. 1d and Supplemen-tary Fig. 11).Article https://doi.org/10.1038/s41467-026-75365-2Nature Communications |         (2026) 17:9592 2www.nature.com/naturecommunicationsTo investigate the chemical states and local coordinationenvironments of Rh and Se in the a-Rh–Se-L MNs, X-ray absorptionnear-edge structure (XANES) and extended X-ray absorption fine-structure (EXAFS) measurements were carried out at both the Rh andSeK-edges.Metallic Rh foil and Rh2O3 were used as references for Rh,while Se foil and SeO2 served as references for Se. Analysis of the RhK-edge XANES spectra revealed that the first-derivativemaximum forthe a-Rh–Se-L MNs appeared at ≈23222.6 eV (SupplementaryFig. 12a), positioned between that of Rh foil (23219.3 eV) and Rh2O3(23226.4 eV). Based on a linear correlation between oxidation stateand the edge position (Supplementary Fig. 12b), the average valenceof Rh in the a-Rh–Se-L MNs was estimated to be ≈+1.4 (estimatedfrom the calibration). The small amount of the oxidized species mayoriginate from partial oxidation upon exposure to air. In the SeXANES spectrum (Supplementary Fig. 13), the Se in Rh–Se is mainlypresent in a near-zero oxidation state, similar to that of the Sereference, rather than SeO2. The surface chemical state of a-Rh–Se-LMNs was also probed by X-ray photoelectron spectroscopy (XPS)(Supplementary Fig. 14). The results indicate that surface Rh exists ina coexistence of metallic and oxidized states, while Se is primarilypresent in its zerovalent state.The EXAFS analysis at the Rh K-edge of a-Rh–Se-L MNs (Fig. 1f)reveals a local structure in Rh–Se, characterized by a dominant coor-dination peak at ≈2.38 Å, which could be assigned to Rh–Se bonding,matching well with other reported metal-Se compounds33. This con-trasts sharply with reference Rh foil and Rh2O3 spectra, where distinctRh–Rh (≈2.73 Å) and Rh–O (≈1.84 Å) scattering paths typically appearat longer and shorter distances, respectively. To elucidate the localcoordination environment and atomic distribution of Rh in a-Rh–Se-LMNs, wavelet-transform analysis of k3-weighted EXAFS spectrarevealed a dominant Rh–Se bonding feature in contrast to the char-acteristic Rh–Rh metallic bonding observed in Rh foil (Fig. 1g).Curvature modulation of a-Rh–Se MNsSurface curvature was tuned by adjusting the pore size. Duringchemical reduction synthesis, PEO-b-PMMA served as both surfac-tant and pore-forming template, with the PMMA molecular weightdictating pore diameter—higher PMMA molecular weights producedlarger pores, as confirmed by SEM (Supplementary Fig. 15), high-angle annular dark-field scanning transmission electron microscopy(HAADF-STEM) images and STEM tomography renderings (Fig. 2a–c,Supplementary Figs. 3, 11, and 16). The amorphous structures ofRh foilRh–RhR(Å)a-Rh–Se-LR(Å)k (Å−1)b50 nme f gc10 nm10 nm−120 nmd Rh+SeSeRhIntensityRh fcc2θ (°)   R (Å)|X(R)| (Å-3)a-Rh–Se-LRh foilRh2O320 40 60 80 0 2 4 6 8 0 5 10 1512312301.4750.73800.7640.382Rh0.5Se3.5Rh1.0Se3.0Rh1.5Se2.5Rh2.0Se2.0Rh3.0Se1.0SeRhDisordereda-Rh–Se MNCo-assemblyTHF+precursorsReductionMicelleTemplate removalaFig. 1 | Schematic of material synthesis and characterization of a-Rh–Se-Lmesoporous nanospheres (MNs). a Illustration of the synthesis route of micelle-induced wet-chemical reduction synthesis of a-Rh–Se MNs. The dark regions in a-Rh–SeMN represent surface voids, corresponding to pore sites in themesoporousRh–Se framework. b Bright-field TEM, c HRTEM, d HAADF-STEM and correspond-ing EDSmappingsof a-Rh–Se-LMNs. The inset of (c) shows the SAEDpattern fromasingle a-Rh–Se-L MN. The experiment in (b–d) was independently repeated threetimes with similar results; representative results are shown. e Powder XRDpatternsof a-Rh–Se with various compositions (fcc: face-centered cubic). f Fourier trans-form extended X-ray absorption fine structure of a-Rh–Se-L MNs with standardreferences. g EXAFS wavelet-transform spectra of Rh foil (top) and Rh–Se-L MNs(bottom) providing a high resolution in both k and R spaces. Source data areprovided as a Source data file.Article https://doi.org/10.1038/s41467-026-75365-2Nature Communications |         (2026) 17:9592 3www.nature.com/naturecommunicationsa-Rh–Se-S (S: small pore of ≈8 nm), a-Rh–Se-M (M: middle pore of≈20 nm), and a-Rh–Se-L (L: large pore of ≈26 nm) were confirmed byXRD patterns (Supplementary Fig. 17). Samples a-Rh–Se-S anda-Rh–Se-M exhibit surface chemical states similar to those of samplea-Rh–Se-L (Supplementary Fig. 14). The high-resolution O 1s and C 1sanalyses further indicate that the Raman enhancement is notdependent on uncontrolled C/O contributions. Small-angle X-rayscattering (SAXS) measurements indicate the presence of a uni-formly distributed porous structure within the porous a-Rh–Seensemble (Supplementary Fig. 18). Among them, theBrunauer–Emmett–Teller specific surface area obtained by N2adsorption–desorption measurements follows the order: a-Rh–Se-S(41.1m2 g−1) > a-Rh–Se-M (37.2m2 g−1) > a-Rh–Se-L (29.1m2 g−1). How-ever, subsequent SERS measurements (further discussed later)revealed that the a-Rh–Se-L MNs delivered stronger enhancementthan a-Rh–Se-S MNs and a-Rh–Se-M MNs (the sample with middlepore size), indicating that SERS activity does not correlate directlywith specific surface area (Supplementary Fig. 19).Here, we therefore proposed a conceptual concave porousstructure to investigate the effect of curvature engineering (Fig. 2d),where the Gaussian curvature is expressed in Eq. 1:K = k1k2 ð1Þwhere k1 and k2 are the principal curvatures at a given point, (repre-senting the maximum and minimum curvature, respectively)34. In thiswork, since the pore-forming agents (i.e., PEO-b-PMMA micelles) arespherical35, the resulting mesopores are approximated as sphericalstructures for the convenience of quantitative analysis. In such a case,any point on the inner surface of a spherical pore has two equalprincipal curvatures. For a concave spherical surface, both principalcurvatures are negative and equal to the inverse of the pore radius,Eq. 2:k1 = k2 = � 1að2Þwhere a is the pore radius.As a result, the Gaussian curvature was obtained by followingEq. 3:K = k1k2 = � 1a� �� 1a� �=1a2 ð3ÞFor a-Rh–Se-S, -M, and -L with approximate pore diameters of 8,20, and 26 nm, respectively, the correspondingK are 0.063, 0.010, and0.0059nm−2. The curvature engineering induced changes in the sur-face structure. Valence-band XPS analysis revealed that the d-bandcenter εd shifts progressively upward with decreasing curvature (cur-vature degree order: a-Rh–Se-S >M> L) (Supplementary Fig. 20). Toensure the reliability of these subtle electronic shifts, each measure-ment was repeated in triplicate; the resulting standard deviation for εdis less than 0.025 eV, which is well within the statistical confidenceinterval for trend analysis. This shift is a consequence of enhancedsurface strain and reduced atomic coordination at smaller curvature,which modulates the d-electron distribution. Potential artifacts fromsurface charging or contamination were excluded by employing anelectron flood gun and rigorous C 1s calibration (284.8 eV) across allsamples. A related geometry-dependent effect has been reported innanostructured systems24. Concurrent atomic forcemicroscope (AFM)measurements of the work function (Φ) showed the sequence a-Rh–Se-S (4.723 eV) > a-Rh–Se-M (4.713 eV) > a-Rh–Se-L (4.702 eV)(Supplementary Figs. 21 and 22), with the lowest Φ in a-Rh–Se-L facil-itating more efficient electron emission and metal–molecule chargeinjection. Together, these findings indicate that as curvature decrea-ses, the lower d-band center affects energy-level alignment for thechemical enhancement mechanism (CEM) of SERS, while the reducedΦ promotes electron transfer from the metal to the adsorbedmolecules.Crystallinity modulation of a-Rh–Se-L MNsConsidering the temperature sensitivity of amorphous materials,which tend to crystallize upon heating, this study systematicallyinvestigated the structural evolution of typical a-Rh–Se-LMNs throughcontrolled thermal treatments under an inert atmosphere and theirstructure-activity relationship in SERS performance, as will be dis-cussed subsequently. Figure 3a illustrates the schematic structuraltransformation scheme during annealing, along with the correspond-ing XRD patterns documenting the crystallization process (Supple-mentary Fig. 23). The thermally treated samples were designated asRh–Se-x, where x denotes the annealing temperature (in °C). Theresults demonstrate that the as-prepared a-Rh–Se-L MNs exhibit anamorphous structure, showing slight crystallization and a smallincrease in crystallite size (from 1.54 to 2.00 nm) after treatment at200 °C (i.e., Rh–Se-200 MNs) (Supplementary Table 2). However,annealing at 400 °C induced crystallization, yielding crystallites withan average size of 7.27 nm at 2θ = 32.9°, consistent with the RhSe2+ xphase (PDF#19-1047). Upon further heating to 600 °C, the crystallitesize of Rh–Se-600 MNs increased substantially to 15.3 nm, accom-panied by distinct phase separation into a mixture of Rh3Se4 (PDF#19-1051) and fcc Rh (PDF#05-0685). This structural evolution indicateselemental rearrangement within the Rh–Se system during thermalprocessing.To further investigate this transformation, the samples werecharacterized using TEM, STEM, SAED, and HRTEM analysis. Post-annealing materials retained a well-defined mesoporous structure(Fig. 3b–d and Supplementary Fig. 24). Enhanced crystallinity wasconfirmed by SAED patterns, with HRTEM images providing corro-borative evidence (Fig. 3b–d). The compositional analysis revealed aprogressive decrease in the relative Se content and a correspondingincrease in Rh concentration (Rh/Se = 21/79, 33/67, and 72/28 forRh–Se-200, Rh–Se-400, and Rh–Se-600 MNs, respectively) within theRh–SeMNs as the annealing temperature rose (SupplementaryFig. 25).aCurvaturee−Flat surface Porous surfacea b cd Rh SeFig. 2 | Structural characterization with different curvatures. HAADF-STEMimages and STEM tomography renderings of a a-Rh–Se-S (yellow), b a-Rh–Se-M(green), and c a-Rh–Se-L (purple). Scale bars: 20nm. The experiments in (a–c) wasindependently repeated three times with similar results; representative results areshown. d The schematic illustration of curvature variations in porous surface withenhanced charge transfer.Article https://doi.org/10.1038/s41467-026-75365-2Nature Communications |         (2026) 17:9592 4www.nature.com/naturecommunicationsThis process destabilizes the metastable coordination environment ofRh atoms previously bound to Se, consequently promoting the for-mation of Rh–Rh metallic bonds.SERS performance of various Rh–Se samplesIn this work, we aimed to sense resistant pollutants like chlorinatedaromatics and polyfluorinated molecules that accumulate in theenvironment because of resistance to degradation36,37. While semi-conductor sensors can detect certain resistant gaseous pollutants,such as CH4 and CF4, via adsorption-induced changes in electricalconductivity38, they lack themajor advantages of SERS sensing, such asselective molecular fingerprint identification under ambient condi-tions. However, potential difficulty in their detection usingsemiconductor-based SERS arises from their resistance to adsorptionand surface interactions, which prevents polarization or dipolegeneration32. This is because common semiconductors shaped inlayered films or nanoparticles (NPs), such as Cu2O, ZnO, TiO2, MoTe2,ZnSe, CdSe, have mainly been demonstrated using readily adsorbingmodel analytes, whereas the detection of weakly adsorbing analytesremains challenging39,40. This is contrasting with the sensing of com-monly reported model dye Rhodamine 6G, methylene blue (MB), and4-mercaptopyridine on semiconductors, where analytes are easy topolarize. In addition to CEM, defect-induced localized charge accu-mulation in semiconductors can generate nano-localized electric fieldsunder optical excitation, contributing to Raman enhancement even innon-plasmonic systems41. However, this pathway is not considered indetail within this study. We started with HCB having a polarizabilityvolume of ≈6 × 10−24 cm3, which is 5–6 times lower than commonlydetected dyes, such as R6G and MB (≈30–40× 10−24 cm3).Surface-curvature effectFirst, we investigated the SERS detection of HCB (10−6 M) on non-porous and curvature-modulated amorphous a-Rh–SeMNs. The testedsubstrates included three curvatures: high (a-Rh–Se-S, 0.063 nm−2),medium (a-Rh–Se-M, 0.010 nm−2), and low (a-Rh–Se-L, 0.0059 nm−2)degrees. SERS spectra of HCB show peaks at 220, 376, 1231 and1523 cm−1 similar to Raman spectra (Supplementary Fig. 26) and a fewnew peaks at 247, 320, 620, 945, and 990 cm−1 potentially assigned tosemiconductor–analyte complex (Fig. 4a). The increase in pore sizeand decrease in curvature degree led to a 20.7-fold increase inadsorption capacity of HCB measured in aqueous solution afterovernight premixing (Fig. 4b). Typically, an increase of surface areaand adsorption capacity lead to the increase of SERS intensity becausemore molecules of analyte are in proximity to active spots formolecule–substrate interactions40. For example, Zhang and colleaguesreported ZnO coating by a higher surface area layer for enrichment ofnon-absorbingmolecules by ZIF-8 shells, where the SERS performanceis achieved by adsorption control32. However, in our case, a-Rh–Se-Lwith the lowest surface area provided the highest EF of >4 × 103 com-pared to a-Rh–Se-M and a-Rh–Se-S. Wemeasured HCB concentrationsin the range of 10−9–10−4 M on a-Rh–Se-S, -M, and -L substrates todirectly compare substrate-dependent sensing performance. For eachvariant, in addition to EF, we evaluated standard deviation (SD), andsignal-to-noise ratio (SNR) to assessnot only amplification strengthbutalso signal reproducibility and analytical reliability (SupplementaryFig. 27 and Supplementary Table 3). Signal variability remains mod-erate across the series (SD = 8.8–11.8%), while SNR ranges between8.47and 11.44. This indicates that key parameters for efficient photo-induced charge transfer (PICT) within CEM for a-Rh–Se are not justlimited to surface area.Therefore, we investigated the relationship between the electro-nic structure of curvature a-Rh–Se and its EF HCB performance. Spe-cifically, the work function (Φ), measured by Kelvin probe forcemicroscopy (KPFM; Supplementary Fig. 22) and the d-band center (εd),derived from XPS (Supplementary Fig. 20), are correlated with EF HCB(Fig. 4c and Supplementary Fig. 28). Previously, CEM performance wasRh Se50 nm 50 nm 50 nm5 nm−12 nm 2 nm2 nmab c d5 nm−1 5 nm−1Fig. 3 | Preparation and characterization of Rh–Se MNs with different crystal-linity. a Schematic diagram of the sample preparation with increasing annealingtemperatures. TEM images, SAED patterns, and HRTEM images of b Rh–Se-200,c Rh–Se-400, and d Rh–Se-600 MNs. The experiment in (b–d) was independentlyrepeated three times with similar results; representative results are shown.Article https://doi.org/10.1038/s41467-026-75365-2Nature Communications |         (2026) 17:9592 5www.nature.com/naturecommunicationscorrelated with electronic structure in the context of aligning thesubstrate’s Fermi level (EF) (conduction band (CB) and valence band(VB) positions) with the analyte’s HOMO or LUMO (highest occupied/lowest unoccupied molecular orbital). However, εd and Φ are over-looked,whichare critical for understanding charge transfer andorbitalhybridization. As shown in Supplementary Fig. 20, εd shifts far awayfrom 3.588 eV (relative to EF) in the a-Rh–Se-S to 3.670 eV in thea-Rh–Se-L. The trend for Φ is opposite, a-Rh–Se-L with a lower Φ of4.702 eV, resulting in the highest EF HCB. This trend can be explainedby the d-band theory for catalysis developed by Nørskov42, whichstates that transition metals’ partially filled d orbitals govern adsorp-tion strength and catalytic activity by tuning d-band position andwidth. Similarly, εd position shift in a-Rh–Se-L optimizes the alignmentof Rh d-electron states available for excitation byRaman laser (Fig. 4d).This enhances the probability of forming hybridized electronic statesbetween a-Rh–Se-L and HOMO/LUMO of HCB. This orbital overlapfacilitates resonant charge transfer and strengthens the Ramanpolarizability of HCB, thus amplifying its Raman signal.Previously, Lin et al. showed that the lower Φ enabled more effi-cient interfacial charge transfer between semiconducting Cu2O andanalytes43. We confirmed this trend on a-Rh–Se-L, where large poreshave lower Φ and upward-shifted EF relative to the vacuum level,reducing the energy barrier for carrier injection into the LUMOofHCB.This improved energy alignment makes carrier transfer to HCB ener-getically favorable under 532 nm excitation, further reinforcing theCEM (Fig. 4d).Therefore, we demonstrate that electronic interactions dominatesurface area effects in semiconducting a-Rh–Se. Despite reduced sur-face area at larger pore sizes, the increased density of low-curvaturedefects and low-coordinated Rh–Se sites leads to a lower Φ and adownshifted εd relative to EF. These interface transformations elevatethe EF,synergistically optimizing the energy alignment between theelectronic structure of a-Rh–Se-L and the frontier orbitals of HCB,enabling more efficient PICT under illumination.Crystallinity modulationThe crystallinity of a-Rh–Se-L was varied via post-annealing tempera-ture (25–600 °C) as confirmed by XRD (Supplementary Fig. 23), TEMand SAED patterns (Fig. 3b–d). As a descriptor of crystallinity, the fullwidth at half maximum (FWHM) of the XRD signal at 2θ = 33.0°, 33.0°,33.4°, and 32.9° for a-Rh–Se-L, Rh–Se-200, Rh–Se-400, andRh–Se-600was chosen, respectively (Supplementary Table 2 and Fig. 4e). SmallerFWHM indicates more crystalline material commonly obtained athigher temperature, while larger FWHM indicates the amorphousstates of material. It has been recently reported that the amorphousengineering of semiconductors has a significant effect on enhancingSERS activity, which provides opportunities to bypass the bottleneckof the low SERS activity of crystalline semiconductors40. However, theeffect of this amorphousness remains poorly understood, represent-ing a gap in current knowledge regarding the SERS performance ofsemiconductors.We found that increasing the annealing temperature of a-Rh–Se-Land higher crystallinity decreased the HCB EF with negligible Ramanactivity of the crystalline sample (Fig. 4e). This confirms that amor-phous, defect-rich a-Rh–Se is crucial for enabling efficient chargetransfer toHCB, likely due to thepresence of disorder-inducedbindingsites and electronic states that facilitate easier carrier excitation.Importantly, this contrastswith noblemetal SERS,where crystallinity isoften favored because of spectrally sharper plasmon resonances andstronger field localization on crystalline material44.To rationalize observed effects of amorphousness on SERS per-formance, Rh20Se80 structures with different degrees ofRaman shift (cm−1)a0123453.43.53.63.74.704.724.74EF(103)Curvature (nm −2)Φ (eV)εd (eV)cbeCrystallineAmorphousΦEFCBVB532nmLUMOHOMOa-Rh–Se-La-Rh–Se-Ma-Rh–Se-SεdμEXEvacd02460 200 400 60001234EF(103)Annealing temperature (°C)FWHMat2θ≈33°012340.010.020.030.04EF(103)q t(mgmg−1)Pore size (nm)4 8 16 24 0.07 0.06 0.01 0a-Rh–Se-npa-Rh–Se-Sa-Rh–Se-Ma-Rh–Se-L500 1000 1500Intensity Rh SeFig. 4 | Probing SERS sensing ofHCBona-Rh–Se-S/M/Lwith different curvaturedegrees. a Raman spectra of hexachlorobenzene (HCB) adsorbed on a-Rh–Se-S, a-Rh–Se-M, a-Rh–Se-L, and a-Rh–Se-np. b Correlation between pore size, HCBenhancement factor (EF) and HCB adsorption capacity (q). c Correlation betweencurvature, HCB EF and d-band center (εd) and work function (Φ). d Schematic illus-tration of the HCB CEM on a-Rh–Se, highlighting the role of electronic structuredescriptors. a-Rh–Se-S/M/L exhibit varying valence band (VB), CB, EF,Φ, and εd (PICT:photo-induced charge transfer; EX: electronic excitation; Evac: energy of the vacuumlevel). e EvolutionofHCBSERS EFwith variation of annealing temperature/crystallinityof a-Rh–Se (XRD patterns and schematic local structures). Data in (b, c, e) are pre-sented as both individual data points and mean± standard deviation (SD) (n=3independent measurements). Source data are provided as a Source data file.Article https://doi.org/10.1038/s41467-026-75365-2Nature Communications |         (2026) 17:9592 6www.nature.com/naturecommunicationsamorphousness were modelled using classical molecular dynamics(MD) simulations. The bulk periodic structures were generated using a200-atom Rh20Se80 model with randomized atomic positions, andannealed at various temperatures viaMD, using theuniversal CHGNet45force field to obtain two structures with different crystallinity (Sup-plementary Fig. 29). It is worth noting that the selection of annealingtemperatures is not aimed at simulating specific experimental proto-cols, but rather at generating a series of structural models exhibitingsystematically varied degrees of amorphousness. These disorderedmodels reflect the experimentally generated amorphous structures ofporous Rh20Se80.The average radial distribution functions (RDFs) for Rh–Rh,Se–Se, and Se–Rh pairs reveal clear structural evolution of Rh and Sespecies with increasing simulated annealing temperature (Supple-mentary Fig. 30). At lower temperatures (thicker lines), sharp andwell-defined peaks indicate strong short-range order, characteristic ofamorphous materials with local coordination. As the temperatureincreases (thinner lines), the RDF peaks broaden and weaken, reflect-ing enhanced atomic mobility, partial bond disruption, and loss ofcoordination order. This trend signifies a transition toward a moredisordered system with decreased crystallinity. The first Se–Rh peak(≈2.4 Å) remains dominant, whereas the Rh–Rh peakis less prominent,highlighting the persistence of local chemical bonding despite thegrowing disorder at longer distances. The presence of distinct sec-ondary peaks is only observed in the low-temperature model. Wechose two model surfaces with different degrees of amorphousnessand further named them as a-Rh–Se (low amorph) and a-Rh–Se (highamorph). The adsorption energies are strongly favorable and sponta-neous in all cases examined. The most stable adsorption for HCBoccurs inside a surface binding pocket of a-Rh–Se (low amorph),exhibiting a very strong interaction of −8.7 eV (Fig. 5a). For adsorptionat high amorph, the binding energy is −3.3 eV (Fig. 5b). It should benoted that while the low-amorphous model exhibits a higher absolutebinding energy, this likely corresponds to a specific, high-coordinationbinding pocket identified during the site-selection process. Thislocalized energyminimum is attributed to themore compact nature ofthe low-amorphous surface, which allows for increased atomic coor-dination around the analyte. However, we should note that adsorptionis favorable and spontaneous in both models. Thus, the experimentalobservation that the highly amorphous sample delivers superior SERSenhancement suggests that Raman activity is not solely governed bythe thermodynamic stability of a single binding site, but rather by thecollective electronic environment of the disordered surface, includingits lower work function and optimized d-band alignment, which facil-itate broader charge-transfer efficiency across the substrate. In addi-tion, to probe different types of resistant analytes, additionalmolecules were also modelled in the expected most favourable con-figuration to determine if the strong binding trends remained con-sistent across different aromatic adsorbates (Supplementary Fig. 31).In fact, comparisonof ourmodelswith anthracene shows a very similaradsorption behaviour, which is also strong and spontaneous withadsorption energies of −9.1 eV for low amorph and −2.6 eV for highamorph models. The adsorption behaviour of highly symmetric aro-matic adsorbates, such as HCB and anthracene, has also been com-pared with MB (Supplementary Fig. 32), a molecule that has beenbroadly studied and is well known to be easily adsorbed28,46. In thiscase, the adsorption is also found to be very strong across all models(−12.1 eV in low and −4.8 eV in highly amorphous), which is in line withthe values found for the other compounds.This confirms that local disorder enhances HCB–substrate inter-actions, in agreement with experimental trends: a stronger SERS signalcorrelates with lower work function and a highly amorphous surface.The relatively low polarizability of HCB (6 × 10−24 cm3) indicates thatefficient charge transfer is essential for inducing chemical enhance-ment. Among the configurations, the strongest adsorption on theamorphous surface likely represents the most favorable geometriesfor PICT and thus the highest SERS activity. To further quantify theinterfacial interaction and electron transfer upon HCB adsorption,Mulliken population analysis47, which estimates electron density dis-tribution among atoms, was performed (Supplementary Table 4). InHCB on Rh–Se exp.a532 nmbceAbsorbanceWavelength (nm)a-Rh–Se-La-Rh–Se-L+HCBHCBRaman shift (cm−1)IntensityHCB on Rh–Se sim.HCB exp.500 1000 1500400 600 8000.00.30.60.9CClSeRhda-Rh–Se-LEFCBLUMO−4.7−2.1−4.8E (eV)−2.70.95VBHOMO532 nmEg↓2.3 Fig. 5 | DFT adsorption models of HCB at a-Rh–Se. a Rh–Se coordinated (lowamorph), b Rh–Se (highly amorph), c UV–vis spectra of a-Rh–Se-L, HCB and mix-ture a-Rh–Se-L+HCB in methanol and d tentative mechanism of the SERSenhancement of HCB on a-Rh–Se substrates via multiple charge transfer pathwaysthat collectively contribute to the CEM. e Comparison of experimental (exp.)Raman spectrum of bulk HCB and SERS spectrum of HCB adsorbed on RhSe,together with corresponding DFT simulation (sim.) spectra of free HCB andHCB@RhSe (insert: optimized DFT cluster models of HCB adsorbed on RhSe,illustrating interfacial binding geometry used for vibrational analysis). Source dataare provided as a Source data file.Article https://doi.org/10.1038/s41467-026-75365-2Nature Communications |         (2026) 17:9592 7www.nature.com/naturecommunicationsthe isolated state, HCB exhibits a neutral-like distribution with Clatoms at +0.07 e. Upon adsorption, a significant shift in charge isobserved across both models, confirming substantial electron dona-tion from the a-Rh–Se surface to the analyte’s C–Cl bonds. Specifically,the average charge of Cl atoms increases to +0.18 e and +0.12 e for thelow- and high-amorphous models, respectively. This substantialcharge redistribution confirms the formation of a strongsemiconductor–analyte complex, providing the necessary ground-state coupling for the CEM.Tentative mechanismThe absorbance suppression and the appearance of a new peak at290–300nm observed in ultraviolet–visible spectroscopy (UV–vis)spectra after mixing HCB and a-Rh–Se suggests a change in electronicstructure (Fig. 5c). This spectral evolution indicates the formation ofmolecule–substrate interfacial states resulting from adsorption-induced electronic coupling. Upon adsorption of HCB onto the a-Rh–Se-L surface, interfacial interactions lead to an energetic alignmentof theHCBHOMO(−2.7 eV relative to EF) andLUMO(0.95 eV relative toEF) with the substrate, resulting in a reduced energy bandgap (Eg) from3.6 to 2.3 eV (Supplementary Table 5). This alignment enables PICT,whereby the HCB molecular orbitals accept electrons from the pho-toexcited Rh–Se CB or valence d-states. This leads to a depletion ofavailable photoactive states in Rh–Se, causing an overall reduction inabsorbance at the visible range. The experimental Eg of a-Rh–Se wasdetermined to be 2.67 eV from Tauc plot analysis. The EF location wasdetermined atΦ = 4.702 eV (Supplementary Fig. 22), and the VB onsetwas measured by XPS to be 0.1 eV below EF. This places the VB max-imum at –4.81 eV and the CB minimum at –2.14 eV with respect to thevacuum level. The density functional theory (DFT) LUMO level of HCBafter adsorption is ≈–1.5 eV, placing it energetically accessible forelectronpopulation from substrate states under 2.33 eV excitation andenabling efficient electron transfer from the substrate to the moleculeunder 2.33 eV laser excitation (Supplementary Table 5).To further clarify the role of excitation wavelength, Raman spec-tra obtained under 473, 532, and 633 nm excitation show unchangedRaman band positions but strongly wavelength-dependent signalintensities and relative mode ratios (Supplementary Fig. 33 and Sup-plementary Table 6). For example, the normalized intensity ratioI1228=I376 changes from 0.995 at 473 nm to 0.731 at 532 nm and furtherto 0.669 at 633 nm excitation, where such wavelength-dependentvariations in relative band intensities indicate excitation-dependentmodification of vibrational polarizability via charge-transfer transi-tions rather than uniform electromagnetic amplification. The stron-gest enhancement and highest signal-to-noise ratio are obtainedunder532nm excitation, while signals decrease significantly at 633 nm, andspectra measured at 473 nm resemble the bulk Raman responsewithout pronounced mode selectivity. Such excitation-dependentenhancement indicates that Raman amplification efficiency is gov-erned by photon-energy matching with interfacial electronic transi-tions rather than purely electromagnetic field enhancement.Consequently, selective enhancement of specific vibrational modesarise from excitation-dependent modulation of molecular polariz-ability via charge-transfer transitions.The observed SERS enhancement of HCB on a-Rh–Se-L substratescan be attributed to multiple charge transfer (CT) pathways that col-lectively contribute to the CEM (Fig. 5d). First, interfacial ground-statecharge transfer (μGSCT) occurs via static orbital hybridization betweenthe HCB HOMO and low-lying unoccupied states of Rh–Se. Mullikenpopulation analysis confirms this interaction, with Cl atoms showing acharge increase upon adsorption, indicating electron donation fromthe substrate. Second, interfacial photo-induced charge transfer(μPICT) is enabled by the energetic alignment between the CB of Rh–Seand the HOMO of HCB (–0.8 eV, DFT data), allowing excited electronsto transiently populate the analyte’s antibondingorbitalsunder 2.33 eVlaser excitation. This process is facilitated by the elevated εd and lowerΦ of the more disordered, porous Rh–Se surfaces. Lastly, intramole-cular electronic excitation (μEX) within HCB itself may also contribute,as its HOMO-LUMO gap closely matches the laser energy, furtherenhancing the polarizability via resonance. These synergistic CTinteractions, both static and dynamic, explain the pronounced SERSsignal amplification observed on highly amorphous Rh–Se substrates.To elucidate the origin of chemical enhancement, we comparedHCB-derived signals on Raman spectra, SERS on a-Rh–Se and simu-lated using hybrid-DFT calculations (Fig. 5e and SupplementaryFig. 34). DFT vibrational spectra were computed using both the iso-lated molecule and adsorption models on non-periodic cluster struc-tures. The comparison between experiment and simulation revealsadsorption-induced vibrational modifications consistent with inter-facial electronic coupling. The C–Cl shear mode appears additionallyat 247 cm−1 upon adsorption (in addition to 220 cm–1 assigned to bulkHCB), indicating vibrational constraint. The 376 cm−1 band shows nomeasurable frequency shift between bulk Raman and SERS, yet itsrelative intensity decreases in SERS. The aromatic ring-breathingmodeshifts from 1229 to 1216 cm–1, reflecting perturbation of the π-electronsystem. Importantly, a new band near 880 cm–1 appears exclusively foradsorbed HCB. The adsorption-specific mode is attributed to sym-metry breaking at the Rh–Se interface. Free HCB possesses inversionsymmetry, but adsorption introduces an asymmetric surface–vacuumenvironment, relaxing Raman selection rules and enabling previouslyforbidden or weakmodes to gain intensity. Simultaneously, interfacialcharge-transfer coupling modifies the molecular polarizability tensor,selectively amplifying vibrational modes under excitation energiesresonant with interfacial electronic transitions. The strong agreementbetween experimental and simulated spectral evolution thereforeprovides direct spectroscopic evidence that SERS enhancement in a-Rh–Se originates from charge-transfer-driven chemical enhancementrather than purely electromagnetic amplification. Such enhancementis consistent with the charge-transfermechanismobserved in previoussemiconductor-based SERS studies48,49.To understand how the electronic landscape supports thismechanism, we examined the projected density of states (PDOS).PDOS analysis reveals that the Rh 4d-band, centered ≈−1 to −3 eV,shows only minor depletion with increasing amorphousness (Supple-mentary Fig. 35), suggesting that bulk electronic structure is largelypreserved. Upon HCB adsorption, Cl 3p and C 2p states broaden andhybridize with Rh–Se orbitals, producing interfacial states close to EFthat could facilitate excitation-dependent electron transfer pathways.We should note that while for Rh–Se our DFT calculations showmetallic behavior for all atoms, with a localminimum located at EF, theexperimental outcomes suggest a semiconducting behavior arisingprobably from dynamic effects. This might be due to charge localiza-tion via polaron formation, where electron-phonon coupling in thedisordered network could open an optical gap despite the metallicband structure. Despite this discrepancy, the atomic-level geometriesand adsorptionmodels remain representative of the short-range orderwithin the material. Upon HCB adsorption, distinct Cl 3p and C 2pstates hybridize with RhSe orbitals, especially on a-Rh–Se (highamorph). This hybridization, accompanied by a pronounced reductionof the molecular HOMO-LUMO gap (from 3.63 to 2.31 eV), supports aCEM via PICT. Unlike conventionald-band theory systems (e.g., Pt, Pd),where εd governs reactivity, here the disorder-induced local states atthe interface appear more relevant to SERS activity.Quantitative detection for HCBThe quantitative SERS performance was evaluated over an HCB con-centration range of 10−4–10−10 M50. The band at 375 cm−1 (C–Cl) wasused as the analytical marker to construct a calibration curve(Fig. 6a, b), which exhibited a linear response over four orders ofmagnitude. The LOD, calculated based on calibration curve (usingArticle https://doi.org/10.1038/s41467-026-75365-2Nature Communications |         (2026) 17:9592 8www.nature.com/naturecommunicationsresidual standard deviation of a regression line)51, was determined tobe 6.2 × 10−9 M, confirming the high sensitivity of the a-Rh–Se-Ltowards non-resonant, low-polarizability analyte in trace-level con-centrations. To assess real-world applicability, HCB was detected in awastewater treatment (WWT) sample collected from the TATRAKopřivnice wastewater treatment facility, Czechia (SupplementaryTable 7). This WWT sample was characterized by a chemical oxygendemand (COD) of 153mg L−1 and a biological oxygen demand (BOD5)of 44.8mgL−1. The relative high COD/BOD5 ratio (≈3.4) indicates asubstantial fraction of poorly biodegradable matter52 (SupplementaryTable 7). This makes the sample highly relevant for detecting persis-tent organic pollutants, such as HCB. Despite matrix complexity, theSERS signal of HCB remained clearly detectable.To assess long-term sensing stability and reproducibility, weevaluated the SERS response of a-Rh–Se-L substrates for HCB detec-tion inWWT across five independently prepared batches over a periodof 90 days (Fig. 6c). The SERS intensity remained highly stable, with arelative standard deviation (RSD) of 2.55%, confirming both temporaland inter-batch reproducibility. Such matrices present a challenginganalytical environment because of (i) background of organic mattercompeting with HCB for adsorption, (ii) potential fluorescence orRaman signal suppression from humic-like substances in WWTmatrix53,54. Unlike noblemetal SERS substrates,which often suffer fromsignal degradation due to surface oxidation/contamination andnanoparticle aggregation55,56, a-Rh–Se-L maintained stable perfor-mance for long-term monitoring of trace-level pollutants in complexenvironmental matrices.To properly compare the performance of a-Rh–Se and AuNPs (Sup-plementary Fig. 36), we also performed HCB sensing on AuNPs (Sup-plementary Fig. 37). The resulting EF of 34,702wasmuch higher than fora-Rh–Se (up to >4× 103). However, a-Rh–Se samples possess lower signalvariability and higher SNR compared to AuNPs (Supplementary Table 3).Thus, while AuNPs provide stronger absolute EM amplification for smallmolecules, such as HCB, the a-Rh–Se substrates deliver more uniformenhancement and superior analytical stability. This result demonstratesthat SERS performance in this system is governed primarily by substrateelectronic structure and charge-transfer efficiency rather than poreinclusion effects or purely hotspot-driven EM mechanisms.Anthracene sensing in industrial wastewaterTo further demonstrate applicability beyond chlorinated aromatics,sensing experiments were extended to polycyclic aromatic hydro-carbons (PAHs), using anthracene as a representative pollutant withpolarizability volume of 2.67 × 10−23 cm3. PAHs can occur in industrial1 2 3 4 5yadegarotSBatch number350037003900410043004500HCB intensity1315306090025050075010001250PTFE Int025050075010001250PTFE IntRSD8.7%PTFE MPsAuNPsa-Rh–Se-LRSD18.7%IntensityRaman shift (cm−1)Intensity(103,arb.u.)CPTFE (109 MPs mL−1) Raman shift (cm−1)Intensity(103,arb.u.)X (μm)15 3045515Y(μm)01250X (μm)15 304501250515Y(μm)DIWWTHCBa b cd e fIntensityRaman shift (cm−1) Log (CHCB M−1)Intensity(103,arb.u.)Intensity(103,arb.u.)Raman shift (cm−1)1315306090StoragedaysBach number1 2 3 4 53.54.5103HCBintensity300 600 1200 −8 −6 −408162432600 1500300 400 1200900010203040400 8001200 1800 000.3 0.6 0.90.51.01.52.02.50123456300 600 900 1200 1500Fig. 6 | SERS performance of a-Rh–Se-L substrates for HCB and PTFE micro-plastics (MPs) detection in a real wastewater (WWT) sample. a QuantitativeSERS sensing of HCB using concentrations 10−4 to 10−10M. b The correspondingcharacteristic HCB peak intensities at 375 cm−1 as a function of the concentration(CHCB: the concentration of HCB). c SERS spectra of HCB (10−6M) detected indeionized (DI) water and WWT matrix; inset shows long-term storage stability(90 days, 5 batches), confirming reproducibility (RSD= 2.55%). dQuantitative SERSsensing of PTFE MPs using concentrations 0.01 to 109 particles mL−1. e Thecorresponding characteristic PTFEMPs peak intensities at 734 cm−1 as a function ofthe concentration. f SERS spectra of PTFE MPs (0.5 × 109 particles mL−1 in WWT) ona-Rh–Se-L versus AuNPs substrates. Insets show SERSmaps of intensity at 734 cm−1across a 20× 50 µm2 area, with calculated relative standard deviation (RSD) valuesof 8.7% (a-Rh–Se-L) and 18.7% (AuNPs), indicating higher signal reproducibility on a-Rh–Se-L. Data in (b, e) are presented as both individual data points and mean ± SD(n = 3 independent measurements). Source data are provided as a Source data file.Article https://doi.org/10.1038/s41467-026-75365-2Nature Communications |         (2026) 17:9592 9www.nature.com/naturecommunicationswastewater as a consequence of combustion-related and otherindustrial activities57,58. As persistent hydrophobic pollutants withlimited surface affinity, PAHs, such as anthracene, are typically difficultto detect using conventional semiconductor SERS substrates59,60.SERS spectra of a-Rh–Se substrates show characteristic anthra-cene bands at 395 and 1402 cm−1 (Supplementary Fig. 38), enablingquantitative detection across concentrations from 10−9 to 10−4 M.Measurements performed in industrial wastewater obtained from aspecialty glass manufacturing facility (Supplementary Table 8) showthat anthracene signals remain clearly detectable despite matrixbackground contributions. Additional Raman bands observed in thewastewatermatrix appear at 600–620, 780–820, and 1000–1100 cm−1,associated with silicate network vibrations from glass components, aswell as at 1060, 1130, 1295–1305, 1375, and 1440–1450 cm−1 originatingfrom hydrocarbon residues of oils and lubricants, while bands near1600 cm−1 arise from aromatic residues and 1640 cm−1 correspond towater bending modes (Supplementary Table 9). These backgroundfeatures do not obscure anthracene marker peaks. These results con-firm that the a-Rh–Se platform enables the detection of chemicallyresistant pollutants in realistic industrial water matrices.Persistent microplastic PTFE sensingBuilding on the CEM-driven detection of HCB, we extended theRh20Se80 SERS platform to non-adsorbing PTFE MPs with a diameter≈250 nm. SEM imaging combined with EDS elemental mapping (Sup-plementary Fig. 39) confirms that PTFE particles are deposited on theexternal mesostructured a-Rh–Se-L surface, with homogeneous fluor-ine distribution co-localized with the selenium framework. PTFE MPsare a growing concern due to their environmental persistence andpotential release from consumer and industrial materials61–63. PTFEpresents a challenging target due to its low Raman cross-section andlow polarizability volume of ≈5 × 10−24 cm3, even lower than that ofHCB. Conventional EM-driven and semiconducting substrates oftenfail to detect PTFE because (i) minimal affinity to noblemetal surfaces,and (ii) there is a generalmismatch between the large (hundreds of nmto µm) size ofMPs and plasmonic hot spots. We recently reported thatsuch a mismatch limits effective EM field confinement and signalenhancement, necessitating alternative strategies that enable analytecapture and SERS activation63.The quantitative SERS detectionof PTFEMPs consistently showedthe characteristic vibrational bands of PTFE at 734 and 1382 cm−1(Supplementary Fig. 40) across a wide concentration range(0.24–5 × 109 particles mL−1) (Fig. 6d, e)64. The tested concentrationrange was selected to evaluate the analytical performance towardsubmicrometric PTFE particles65,66. The saturation trend was observedat concentrations above 1 × 109 particles mL−1 because of surface cov-erage limitations. At high PTFE loading, the substrate becomes satu-rated with PTFE particles, leading to crowding, partial signalquenching, and reduced access to enhancement sites. This saturationbehaviour likely arises from surface coverage limitations: the a-Rh–Se-L becomes saturated with PTFE, leading to crowding, partial signalquenching, and reduced access to enhancement sites. However, suchconcentrated MP suspensions could be easily detected by analyticalmicroscopy tools67. The calibration curve obtained from sensing in anenvironmentally relevant concentration range is 0.24–1 × 109 particlesmL−1, showing a strong linear relationship (R2 = 0.996), indicating reli-able quantification capability.Finally, the SERS performance of a-Rh–Se-L and common 13 nmAuNPs (Supplementary Fig. 36) was compared using 0.5 × 109 parti-cles mL−1 PTFE MPs suspension in WWT (Fig. 6f). Each spectrumrepresents the average acquired across a 20 × 50 µm2 area usingintensity at 734 cm−1. The PTFE vibrational mode observed at 756 cm−1on AuNPs appears at 734 cm−1 on a-Rh–Se-L, corresponding to a22 cm−1 red shift. On AuNPs in the area of 300–400 cm−1, PTFEmodesappear as single, symmetric features consistent with EM-dominatedenhancement. In contrast, a-Rh–Se-L exhibits peak broadening andmode splitting, particularly near 300 cm−1, along with a 5–7 cm−1 redshift of the 400 cm−1 mode. Variations in vibrational features areattributed to interfacial electronic coupling, supporting a significantCEM contribution on a-Rh–Se-L. CEM can selectively amplify modescoupled to interfacial charge-transfer transitions and may inducesmall frequency shifts due to surface electronic coupling. The aver-age intensity of PTFE is ×1.5 higher for a-Rh–Se-L than for AuNPs,despite the EM field mechanism of SERS enhancement for noblemetals commonly providing higher Raman signal enhancementthan CEM.Within the SERS map, up to ≈20,000 PTFE particles (250nm)could theoretically fit under densemonolayer conditions. However, at0.5 × 109 particlesmL−1, the estimated number of particles is negligible,indicating that SERS signals depend primarily on particle adsorptiononto the substrate rather than on random diffusion through theprobing region. Comparative measurements revealed that Rh20Se80exhibited higher and more uniform SERS signals, with a RSD of 8.7%,compared to 18.7% for AuNPs (Fig. 6f). This improved reproducibilitymay reflect a higher surface affinity of PTFE for a-Rh–Se-L, likely due tohigher adsorption energy or stronger hydrophobic interactions. Incontrast, the hydrophilic and negatively charged AuNPs may repel thePTFE particles, leading to lower adsorption density and greater signalvariability.Overall, amorphous metal–selenide semiconductors, such as a-Rh–Se-L, offer a complementary materials route for SERS sensing ofchemically resistant pollutants by emphasizing substrate–moleculeelectronic coupling rather than hotspot-dominated electromagneticamplification. Prior studies have demonstrated that microplastics68and other weakly interacting pollutants69,70 can be detected usingplasmonic SERS platforms, including macroporous Ag foams com-bined with machine-learning spectral analysis for complex plasticmixtures63 and anisotropic Au nanostar substrates enabling detec-tion of submicron and nanoplastic particles at low massconcentrations71. However, detection of hydrophobic and resistantcontaminants on plasmonic SERS substrates often requires addi-tional interfacial engineering, such as linker-induced nanoparticlejunction formation to create adsorption-active hot spots69 or in situextraction and self-assembly strategies using ligand-modified Aunanoparticles to overcome weak molecular affinity and matrixcomplexity70. In this context, amorphous metal selenides provideseveral general advantages for SERS beyond specific noble-metalsystems: (i) improved long-term stability (resistance to oxidation andaggregation), (ii) uniform, substrate-wide enhancement via CEM,driven by high density of accessible low-curvature sites and danglingbonds, (iii) efficient detection of low-polarizability molecules viaorbital alignment and PICT, (iv) higher reproducibility and signalconsistency in complex matrices like wastewater, (v) strongerparticle–substrate affinity, further enhanced by geometric bindingpockets that stabilize analyte-substrate interactions, enabling con-sistent detection of large-size pollutants like microplastics.DiscussionIn this work, we developed a-Rh–Se MNs using a chemical reductionmethod. The SERS performance of these MNs is significantly influ-enced by both their surface curvature and crystallinity. Our resultsshow that a-Rh–Se-L with low curvature reduces the energy barrierfor charge transfer, promoting efficient energy alignment and carrierinjection. This effect is coupled to the material’s amorphous char-acter, which introduces disorder-induced electronic states thatenhance carrier excitation. Together, reduced surface curvature andamorphization reshape the electronic structure, the d-band centerand the work function, to promote resonant charge transfer. Con-sequently, this strategy enables semiconductor-based SERS to detectchemically challenging analytes, including HCB, anthracene andArticle https://doi.org/10.1038/s41467-026-75365-2Nature Communications |         (2026) 17:9592 10www.nature.com/naturecommunicationsPTFE MPs, extending SERS capabilities beyond conventional dyesand thiolated probes. Our findings indicate that the surface curva-ture, together with amorphization, plays a critical role in governingthe SERS activity, overturning the traditional SERS assumption thatfavors high-surface-area substrates. This work deepens our under-standing of how a semiconductor’s structural parameters govern itsSERS activity.MethodsMaterialsAll reagents were used as purchased without further purification.Sodium hexachlororhodate (III) (Na3RhCl6), selenous acid (H2SeO3,98%), gold (III) chloride trihydrate (HAuCl4·3H2O, ≥99.9% tracemetals basis), sodium citrate tribasic dihydrate (Na3CA, ≥99.0%) andformic acid (HCOOH, ≥96%) were purchased from Sigma Aldrich.PEO-b-PMMA with different molecular weights of PMMA blocks,including PEO3500-b-PMMA1700, PEO10000-b-PMMA5500, and PEO10500-b-PMMA18000 (the subscript numbers indicate the average molecularweight (gmol−1) of the corresponding blocks), were obtained fromPolymer Source. THF (99.5%), ethanol (99.5%), methanol (99.5 %),and acetone (≥99.0%) were purchased from FUJIFILM Wako PureChemical Corporation.Synthesis of a-Rh–Se-L MNsThe a-Rh–Se-L MNs were synthesized via a wet-chemical reductionapproach combined with diblock copolymer micelle assembly. In atypical procedure, 4mg of PEO10500-b-PMMA18000 was first dissolvedcompletely in 0.4mL of THF at 70 °C in a vial, followed by cooling toambient temperature (≈22 °C). Subsequently, 1.6mL of deionizedwater, 4mL of an aqueous solution containing Na3RhCl6 (40mM)and H2SeO3 (40mM) at predetermined ratios (typical feedingvolume is Na3RhCl6:H2SeO3 = 1:3), along with 1mL of HCOOH, wereadded dropwise to the THF solution under gentle stirring (200 rpm).After 5min of mixing, the reaction system was transferred to an oilbath and maintained at 80 °C for 12 h under static conditions. Theresulting products were then collected by centrifugation at18,000 × g, thoroughly washed with a 2:1 acetone-water mixture(6 × 30mL), and finally vacuum-dried. The nonporous a-Rh–Sereference was prepared similarly to a-Rh–Se-L MNs, except omittingthe diblock copolymer.Synthesis of Rh–Se MNs with different crystallinityThe Rh–Se MN samples with different crystallinities were obtained byannealing the as-synthesized a-RhSe-L at different temperatures (200,400, and 600 °C) under an inert argon atmosphere. The temperaturewas increased at a rate of 2 °Cmin−1 to the target temperature andmaintained for 3 h. The resulting products at 200, 400, and 600 °Cwere designated as Rh–Se-200, Rh–Se-400, and Rh–Se-600,respectively.Synthesis of a-Rh–Se-MThe synthesis of a-Rh–Se-M was consistent with the process describedfor sample a-Rh–Se-L MNs above, with the only difference being thereplacement of the polymer (PEO10500-b-PMMA18000) with PEO10000-b-PMMA5500. The feeding amounts of Na3RhCl6 and H2SeO3 were 1 and3mL, respectively.Synthesis of a-Rh–Se-SThe synthesis of a-Rh–Se-S was consistent with the process describedfor sample a-Rh–Se-L MNs, with the only difference being the repla-cement of the polymer (PEO10500-b-PMMA18000) with PEO3500-b-PMMA1700 and the volume of deionized water was changed to 3.2mL.The feeding amounts of Na3RhCl6 and H2SeO3 were 1 and 3mL,respectively.Synthesis of AuNPsAuNPs were prepared by a modified Turkevich method72. In a 250mLround-bottom flask, 1mL of HAuCl4·3H2O (100mg in 10mL) and2.5mL of Na3CA (284.9mg in 25mL) were added to 100mL of boilingwater. Heating continued for 10min during which time the solutionbecame red. After cooling to room temperature, AuNPs were purifiedfrom Na3CA by centrifugation and washing with a 1:3 water-methanolmixture (7050× g, 30–60min). After each centrifugation cycle, AuNPswere sonicated for 1–2min. The resulting precipitate of AuNPs wasresuspended in 50μL of methanol, which was deposited on a silicaslide (0.5 × 0.5 cm2) by drop-casting for further Raman study.CharacterizationsThe elemental composition of a-Rh–Se MNs was analyzed by ICP-OESusing an Agilent 5800 spectrometer. Morphological characterizationwas conducted with a Zeiss Gemini SEM 560 scanning electronmicroscope operated at 10 kV, equippedwith a Bruker 5060F flat quadEDS detector for elemental mapping. Surface potential measurementswere performed using KPFM on a Bruker Dimension Icon system. TEMand HAADF-STEM with EDS were performed on a Talos F200X G2(Thermo Fisher Scientific) operated at 200 kV in nanoprobe mode,with a beam convergence angle of 20mrad and a HAADF detectorcollection range of 60–200mrad, equipped with a Super-X EDSdetector. For STEM tomography, tilt series were acquired over a rangeof −65° to +65° in 2° increments and reconstructed using the simul-taneous iterative reconstruction technique (SIRT, 10 iterations) inTomviz. Segmentation and 3D rendering were performed in 3D Slicer,and the resulting models were processed in Autodesk Fusion 360 forvisualization. XPS measurements were conducted on a Thermo Sci-entific K-Alpha+ system with a monochromatic Al Kα X-ray source(1486.6 eV), and binding energies were calibrated against the C 1s peakat 284.8 eV. All XPS samples underwent identical preparation andmeasurement protocols. Crystalline structure was examined byXRD on a Rigaku SmartLab diffractometer using Cu Kα radiation(40 kV, 30mA), with data collected from 20° to 90° (2θ) at a step sizeof 0.02°. SAXS measurements (Rigaku NANO-Viewer) were employedto determine the pore-to-pore distance. Additionally, nitrogenadsorption–desorption isotherms were obtained at 77 K using aBELSORP-mini (BEL, Japan), and pore size distributions were derivedfrom the Barrett–Joyner–Halenda model. XAS measurements wereconducted in beamline TPS 44A, quick-scanning X-ray absorptionspectroscopy, NSRRC, Taiwan. The synchrotron X-ray energy rangesfrom 4.5 to 34 keV (depending on the target central elements). Thefocused spot size is 10 × 10μm2. All the powder samples were groundinto fine powder, concealedwith Kapton tape, and fixed on the sampleholder.SERS measurementsPreparation of SERS substrate and analyte deposition. The sus-pension (2mgmL−1) of powder of a-Rh–Se was prepared in methanoland deposited onto a silicon wafer (0.5 × 0.5 cm2) by drop casting. Foranalyte deposition to the SERS substrate, a-Rh–Se suspension (0.1mL)was premixed with HCB solution or a PTFE MP dispersed solution(0.9mL) in specific concentrations for 1 h and then centrifuged toconcentrate it to 0.1mL of a-Rh–Se. The sample was prepared by dropcasting onto a silicon wafer (0.5 × 0.5 cm2).Raman spectroscopy measurements. Raman spectra of the MPswere collected on an inVia™ confocal Raman microscope (Renishaw,United Kingdom) with 532 nm (6mW, 10 s for each exposure) and aresolution of 2 cm−1 (600 grooves mm−1 grating) spanning 2000 to200 cm−1. The spectra were acquired using a 20× objective and a 40 scollection time. In mapping mode, the objective was parked over thesubstrate and then scanned, with a collection time of 40 s per point.Article https://doi.org/10.1038/s41467-026-75365-2Nature Communications |         (2026) 17:9592 11www.nature.com/naturecommunicationsAdditional measurements have been done using LabRAM ARAMIS(Horiba, Japan) with red (633 nm) and blue (473 nm) lasers at a powerdensity of 14 and 20 kWcm−2, respectively. The spectra wereacquired using a 20× objective and a 40 s collection time. For thecalculations of EFs, three independent measurements were acquiredand intensities were used with SD calculated from N spectraaccording to the relation SD=ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPNi = 1ðxi�xÞ2Nr, where x is the intensity ofthe Raman signal, �x is the mean of the intensities, and N is thenumber of spectra used.EF calculation. Calculation of the SERS EF was performed according tothe standard relation: EF = ISERS=CSERSIRS=CRS, where the ISERS and IRS represent theRaman scattering intensities on SERS-active and reference silicon sur-faces, and CSERS and CRS are the corresponding concentrations of ana-lyte. For measuring the IRS on silicon, 10−1M of HCB and PTFE MPsaqueous solution was used. To accurately estimate the number ofmolecules excited by the confocalmicroscope, a 20× objective lenswitha 1.1 µmbeamwaist was illuminated through a container with a depth of3 µm. Assuming the Gaussian beam excites a volume that is roughly theshape of a cylinder, we can estimate the volume excited by the beamusingV =0.5 ×D×π×h, whereD is the beamwaist, andh is the height ofthe container. By multiplying the concentration of the solution (inmolecules µm−2) by the beam volume, we can estimate the number ofmolecules excited in the Raman reference measurement (CRS).Calculation of LOD. The calibration curve was plotted using analyticalsolutions of HCB and PTFE in different concentrations in a 3:1 water-methanol mixture. The spectra were measured with a 40 s accumula-tion time for eachmeasurement using a 532 nm laser (6mW). The LODwas calculated using a calibration curve (using residual standarddeviation of a regression line). The residual standard deviation of aregression line or the standard deviation of the y-intercepts ofregression lines was used as the standard deviation. The RSD of thesignal was calculated according to relation RSD calculationRSD= ðSDx Þ× 100%, from three spectra according to the relation, where�x is the mean of the intensities of the Raman signal (376 and 736 cm–1for HCB and PTFE MPs, respectively).Preparation of PTFE MP solutions in WWT. HCB was dissolved in theWWTmatrix, whereas PTFEMPsweredispersed to obtain the specifiedparticle concentration. The water was collected on 20.03.2025, andanalyzed by ALS Czech Republic, s.r.o. under order numberPR2561392. The matrices were raw water from the TATRA Kopřivnicesite. The analysis was conducted in accordance with the internalquality control standards of ALS ČR, following ČSN EN ISO/IEC17025:2018. The composition of the sample is given in SupplementaryTable 7.Preparation of anthracene solutions in industrial wastewater.Anthracene stock solutionswereprepared in ethanol (analytical grade)and subsequently diluted to the desired concentrations (10−9–10−4 M)using industrial wastewater. The wastewater sample was collected on20.03.2025 and analyzed by ALS Czech Republic, s.r.o. under ordernumber PR2561378. The water was collected from SCHOTT AG was-tewater treatment facility. The analysis was performed in accordancewith internal quality control standards of ALS ČR following ČSN ENISO/IEC 17025:2018. The compositions of sample were given in Sup-plementary Table 8.Storage stability test. The stability of SERS substrates was estimatedunder the normal storage conditions in air at 23–25 °C. The a-Rh–Sewas used for HCB (10−6 M) detection, and then the sample was storedup to 3 months after preparation and measured by Raman in experi-mental conditions as described above.Computational methods. Bulk amorphous RhSe, containing 80 at.%Se and 20 at.% Rh (henceforth a-Rh–Se), was prepared using classicalMD simulations. The LAMMPS package was used to run therelaxations73. The atomic interactions were obtained from the uni-versal CHGNET version 0.3 force field45, which is trained on data fromdifferent RhSe binary polymorphs. The different amorphous struc-tures were produced by annealing a 200-atom (160 Se and 40 Rh)system with randomly distributed in a periodic crystalline cubic box.The Nosé–Hoover thermostat in its isothermal-isobaric form (NPT)was set to a constant pressure of 1 atm, with a time step of 1 fs. Tointentionally produce structural models with different degrees ofamorphousness, a common initial system was annealed at tempera-tures of 100 and 2000K and equilibrated for 50ps, until reachingstructural equilibration of the models at their final stages. This pro-cedure leads to two equilibrated cubic systems, in which the densitiesand crystallinity are inversely proportional to the annealing tempera-ture. The pair distribution function was obtained from sampling every10 steps over the final 5000 steps.The surface models were then cleaved from the bulk at a planeparallel to one of the box sides, where a large vacuum layer is intro-duced. Whenever possible, and with the objective of creating slabsfeaturing binding pockets for the molecules, the surfaces were inten-tionally cleaved at regions next to voids formed in the periodicstructures. We expect these adsorption models to be more relevantand comparable with the experimental measurements.The slabswere relaxed using the SIESTADFT code (v5.2)74,75. Initialrelaxations of the slabs were directly performed via DFT due to theforce field’s lower reliability in vacuum systems. Hexachlorobenzeneand MB were then individually adsorbed onto the RhSe surface, withvarious orientations and locations explored to find the most energe-tically favoured conformation for further analysis. Models were initi-ally relaxed using a 500Ry mesh cutoff and a 0.02 Ry energy shift,while the most stable candidates were refined at 800 and 0.005Ry,respectively. All calculations employed double-zeta polarised basissets, the PBEsol exchange-correlation functional76, and Pseudodojopseudopotentials77. Long-range interactions were accounted for usingGrimme’s D3 method78. Electronic convergence criteria were set at10−6 eV for energy differences and 0.02 eVÅ−1 for forces. In-plane lat-ticeparameterswerefixed,while the atomic coodinates and theout-of-plane cell dimension were allowed to relax. A 2 × 2 × 1 Monkhorst-Packk-point mesh was used, with no replication along the vertical axis.Atomic figures and visualizations were generated using the Quantu-mATK builder package79.To investigate the vibrational properties of the Rh–Se systemwithhighprecision, a two-stagemodeling approachwas employed. Initially,Rh5Se20 clusters were relaxed within a periodic framework usingSIESTA74,75 to establish stable local geometries within a bulk/surfacecontext. These pre-relaxed structures were then extracted and treatedas isolated clusters for further relaxation and vibrational analysis. Thiscluster-based transition allows for the deployment of hybrid func-tionals and triple-zeta basis sets, which provide the necessary accuracyfor Raman intensity simulations that are otherwise computationallyprohibitive in large systems with periodic boundary conditions.All non-periodic calculations were performed using the ORCAv6 suite80. Geometry optimizations and frequency analyses were con-ducted using the B3LYP hybrid functional81,82. The electronic environ-ment was described using the def2-TZVP basis set. For rhodium(Z = 45), scalar relativistic effects were accounted for via the def2-ECPeffective core potential. Se, C, and Cl atoms were treated using an all-electron description. To accurately capture the non-covalent interac-tions between the HCB and the cluster, Grimme’s D3 dispersion cor-rection with Becke–Johnson damping was applied78,83. To ensure high-fidelity Raman spectra, the numerical differentiation increment for theHessian was tightened to 0.001 Bohr. Molecular geometries andvibrational modes were visualized using the Avogadro2 softwareArticle https://doi.org/10.1038/s41467-026-75365-2Nature Communications |         (2026) 17:9592 12www.nature.com/naturecommunicationspackage84, applying a broadening to bridge the gap between discreteDFT peaks and experimental observations.Reporting summaryFurther information on research design is available in the NaturePortfolio Reporting Summary linked to this article.Data availabilityThe rawdata generated fromthe atomistic simulations in this study arepublicly available via NOMAD85. Additional data are available from thecorresponding authors upon reasonable request. Source data areprovided with this paper.References1. Kang, J. et al. How amorphous nanomaterials enhanced electro-catalytic, SERS, and mechanical properties. JACS Au 3,2660–2676 (2023).2. Kang, J. et al. Recent progress of amorphous nanomaterials.Chem.Rev. 123, 8859–8941 (2023).3. Kang, Y. et al. Porous nanoarchitectures of nonprecious metalborides: from controlled synthesis to heterogeneous catalystapplications. ACS Catal. 12, 14773–14793 (2022).4. Jiang, H. et al. Structures and functional properties of amorphousalloys. Small Struct. 2, 2000057 (2021).5. Wu, G. et al. A general synthesis approach for amorphous noblemetal nanosheets. Nat. Commun. 10, 4855 (2019).6. He, Y. et al. Amorphizing noble metal chalcogenide catalysts at thesingle-layer limit towards hydrogen production. Nat. Catal. 5,212–221 (2022).7. Jiang, B. et al. Noble-metal–metalloid alloy architectures: meso-porous amorphous iridium–tellurium alloy for electrochemical N2reduction. J. Am. Chem. Soc. 145, 6079–6086 (2023).8. Kang, Y. et al. Mesoporous amorphous non-noble metals as versa-tile substrates for high loading and uniform dispersion of Pt-groupsingle atoms. Sci. Adv. 10, eado2442 (2024).9. Neergat, M., Gunasekar, V. & Singh, R. K. Oxygen reduction reactionand peroxide generation on Ir, Rh, and their selenides—a comparisonwith Pt and RuSe. J. Electrochem. Soc. 158, B1060–B1066 (2011).10. Yang, Y. et al. A new generation of alloyed/multimetal chalcogen-ide nanowires by chemical transformation. Sci. Adv. 1,e1500714 (2015).11. Lu, T., Dong, S., Zhang, C., Zhang, L. & Cui, G. Fabrication of tran-sition metal selenides and their applications in energy storage.Coord. Chem. Rev. 332, 75–99 (2017).12. Park, S. H., Choi, J. Y., Lee, Y. H., Park, J. T. & Song, H. Formation ofmetal selenide and metal–selenium nanoparticles using distinctreactivity between selenium and noble metals. Chem. Asian J. 10,1452–1456 (2015).13. Jain, V. K. & Kedarnath, G. Applications of metal selenium/telluriumcompounds inmaterials science. Phys. Sci. Rev. 4, 20170127 (2019).14. Yang, C. et al. A generalized surface chalcogenation strategy forboosting the electrochemical N2 fixation of metal nanocrystals.Adv. Mater. 32, 2001267 (2020).15. Li, Z. et al. 1T′-transition metal dichalcogenide monolayers stabi-lized on 4H-Au nanowires for ultrasensitive SERS detection. Nat.Mater. 23, 1355–1362 (2024).16. Li, A., Lin, J., Huang, Z., Wang, X. & Guo, L. Surface-enhancedRaman spectroscopy on amorphous semiconducting rhodium sul-fide microbowl substrates. iScience 10, 1–10 (2018).17. Yu, Z. et al. Selective and durable H2O2 electrosynthesis catalyst inacid by selenization induced straining and phasing. Nat. Commun.15, 9346 (2024).18. Eguchi, M. et al. Materials space-tectonics: atomic-level composi-tional and spatial control methodologies for synthesis of futurematerials. Angew. Chem. Int. Ed. 62, e202307615 (2023).19. Li, M.-X. et al. Data-driven discovery of a universal indicator formetallic glass forming ability. Nat. Mater. 21, 165–172 (2022).20. Xiang, L. et al. Curvature-induced ion docking effect in capacitivedeionization. Nat. Water 2, 1195–1206 (2024).21. Chen, T. et al. 3D-porous electrocatalyst with tip-enhanced electricfield effect enables high-performance proton exchangemembranewater electrolyzer. Adv. Mater. 37, 2418527 (2025).22. Ma, Y. et al. Streamlined mesoporous silica nanoparticles withtunable curvature from interfacial dynamic-migration strategy fornanomotors. Nano Lett. 21, 6071–6079 (2021).23. Zhang, C. et al. Highly ordered arrays of hat-shaped hierarchicalnanostructures with different curvatures for sensitive SERS andplasmon-driven catalysis. Nanophotonics 11, 33–44 (2022).24. Bashouti, M. Y. et al. Role of silicon nanowire diameter for alkyl(chain lengths C1–C18) passivation efficiency through Si–C bonds.Langmuir 31, 2430–2437 (2015).25. Deriu, C. & Fabris, L. A surface chemistry perspective on SERS:revisiting the basics to push the field forward. Chem. Soc. Rev. 54,5224–5247 (2025).26. Lin, J. et al. Ultrasensitive SERS detection by defect engineering onsingle Cu2O superstructure particle. Adv. Mater. 29, 1604797(2017).27. Liu, Q., Jiang, L. & Guo, L. Precursor-directed self-assembly ofporous ZnO nanosheets as high-performance surface-enhancedRaman scattering substrate. Small 10, 48–51 (2014).28. Qi, D., Lu, L., Wang, L. & Zhang, J. Improved SERS sensitivity onplasmon-free TiO2 photonic microarray by enhancing light-mattercoupling. J. Am. Chem. Soc. 136, 9886–9889 (2014).29. Fraser, J. P. et al. Application of a 2Dmolybdenum telluride in SERSdetection of biorelevant molecules. ACS Appl. Mater. Interfaces 12,47774–47783 (2020).30. Islam, S. K., Tamargo, M., Moug, R. & Lombardi, J. R. Surface-enhanced Raman scattering on a chemically etched ZnSe surface.J. Phys. Chem. C 117, 23372–23377 (2013).31. Livingstone, R. et al. Surface enhanced Raman spectroscopy ofpyridine onCdSe/ZnBeSe quantumdots grown bymolecular beamepitaxy. J. Phys. Chem. C 114, 17460–17464 (2010).32. Liu, X. et al. Boosting electromagnetic enhancement for detectionof non-adsorbing analytes on semiconductor SERS substrates.Chem 9, 1464–1476 (2023).33. Wang, K. et al. Cu-doped heterointerfaced Ru/RuSe2 nanosheetswith optimized H and H2O adsorption boost hydrogen evolutioncatalysis. Adv. Mater. 35, 2300980 (2023).34. Han, L. & Che, S. An overview of materials with triply periodicminimal surfaces and related geometry: from biological structuresto self-assembled systems. Adv. Mater. 30, 1705708 (2018).35. Kang, Y. et al.Mesoporousmultimetallic nanosphereswith exposedhighly entropic alloy sites. Nat. Commun. 14, 4182 (2023).36. Carvalho, F. P. Pesticides, environment, and food safety. FoodEnergy Secur. 6, 48–60 (2017).37. Jiménez-Skrzypek, G. et al. Microplastic-adsorbed organic con-taminants: analytical methods and occurrence. TrAC Trends Anal.Chem. 136, 116186 (2021).38. Raju, P. & Li, Q. Review—semiconductor materials and devices forgas sensors. J. Electrochem. Soc. 169, 057518 (2022).39. Guselnikova, O. et al. New trends in nanoarchitectured SERS sub-strates: nanospaces, 2D materials, and organic heterostructures.Small 18, 2107182 (2022).40. Wang, X. & Guo, L. SERS activity of semiconductors: crystalline andamorphous nanomaterials. Angew. Chem. Int. Ed. 59,4231–4239 (2020).41. Sadhujan, S. et al. p-type surface defects on n-GaN nanorods.NanoLett. 25, 9118–9124 (2025).42. Hammer, B. & Nørskov, J. K. Electronic factors determining thereactivity of metal surfaces. Surf. Sci. 343, 211–220 (1995).Article https://doi.org/10.1038/s41467-026-75365-2Nature Communications |         (2026) 17:9592 13www.nature.com/naturecommunications43. Lin, J. et al. Direct experimental observation of facet-dependentSERS of Cu2O polyhedra. Small 14, 1703274 (2018).44. Sweedan, A. O. et al. Evolutionary optimized, monocrystalline golddouble wire gratings as a novel SERS sensing platform. Small 20,2311937 (2024).45. Deng, B. et al. CHGNet as a pretrained universal neural networkpotential for charge-informed atomistic modelling. Nat. Mach.Intell. 5, 1031–1041 (2023).46. Chen, B. et al. Au nanoparticles decorated β-Bi2O3 as highly-sensitive SERS substrate for detection of methylene blue andmethyl orange. Analyst 149, 4283–4294 (2024).47. Vivoni, A., Birke, R. L., Foucault, R. & Lombardi, J. R. Ab initio fre-quency calculations of pyridine adsorbed on an adatommodel of aSERS active site of a silver surface. J. Phys. Chem. B 107,5547–5557 (2003).48. Chen, L., Merino, J. P., Torrent-Sucarrat, M., Hou, H.-L. & Prato, M.Charge transfer-induced SERS enhancement of MoS2/dopantsdependent on their interaction difference. Adv. Mater. Interfaces 11,2400272 (2024).49. Cong, S., Liu, X., Jiang, Y., Zhang, W. & Zhao, Z. Surface enhancedRaman scattering revealed by interfacial charge-transfer transi-tions. Innovation 1, 100051 (2020).50. Mrema, E. J. et al. Persistent organochlorinated pesticides andmechanisms of their toxicity. Toxicology 307, 74–88 (2013).51. ICH. Validation of analytical procedures: text andmethodology,Q2(R1), ICH Harmonised Tripartite Guideline (2005).52. Lee, J., Lee, S., Yu, S. & Rhew, D. Relationships between waterquality parameters in rivers and lakes: BOD5, COD, NBOPs, andTOC. Environ. Monit. Assess. 188, 252 (2016).53. Richards, D. A. & deMello, A. J. The matrix reiterated. ACS Sens. 8,3986–3987 (2023).54. Masson, J.-F. Consideration of sample matrix effects and “biologi-cal” noise in optimizing the limit of detection of biosensors. ACSSens. 5, 3290–3292 (2020).55. Khabarov, K. et al. Effect of Au nanoparticle agglomeration on SERSsignal amplification. Nanomaterials 13, 812 (2023).56. Bachenheimer, L. et al. Degradationmechanism of Ag nanorods forsurface enhanced Raman spectroscopy. Sci. Rep. 7, 16282 (2017).57. Wang, L.-C.,Wang, I. C., Chang, J.-E., Lai, S.-O. &Chang-Chien,G.-P.Emission of polycyclic aromatic hydrocarbons (PAHs) from theliquid injection incineration of petrochemical industrial wastewater.J. Hazard. Mater. 148, 296–302 (2007).58. Salihoglu, N. K. et al. Comparison of polycyclic aromatic hydro-carbons levels in sludges frommunicipal and industrial wastewatertreatment plants. Arch. Environ. Contam. Toxicol. 58,523–534 (2010).59. Eremina, O. E. et al. Dual-purpose SERS sensor for selective deter-mination of polycyclic aromatic compounds via electrondonor–acceptor traps. ACS Sens. 6, 1057–1066 (2021).60. Xie, Y. et al. Selective SERS detection of each polycyclic aromatichydrocarbon (PAH) in a mixture of five kinds of PAHs. J. RamanSpectrosc. 42, 945–950 (2011).61. Cole, M., Gomiero, A., Jaén-Gil, A., Haave, M. & Lusher, A. Micro-plastic and PTFE contamination of food from cookware. Sci. TotalEnviron. 929, 172577 (2024).62. Karpenko, A. A., Odintsov, V. S. & Istomina, A. A. Micro-nano-sizedpolytetrafluoroethylene (Teflon) particles as a model of plasticpollution detection in living organisms. Environ. Sci. Pollut. Res. 29,11281–11290 (2022).63. Guselnikova, O. et al. Pretreatment-free SERS sensing of micro-plastics using a self-attention-based neural network on hier-archically porous Ag foams. Nat. Commun. 15, 4351 (2024).64. Koenig, J. L. & Boerio, F. J. Raman scattering and band assignmentsin polytetrafluoroethylene. J. Chem. Phys. 50, 2823–2829 (1969).65. Alaraby, M., Abass, D., Velázquez, A., Hernández, A. & Marcos,R. Polytetrafluoroethylene microplastic properties, pollution,toxicity and analysis: a review. Environ. Chem. Lett. 24,27–59 (2026).66. Lohmann, R. & Letcher, R. J. The universe of fluorinated poly-mers and polymeric substances and potential environmentalimpacts and concerns. Curr. Opin. Green Sustain. Chem. 41,100795 (2023).67. Nguyen, B. et al. Separation and analysis of microplastics andnanoplastics in complex environmental samples. Acc. Chem. Res.52, 858–866 (2019).68. Chen, Q. et al. A review of recent progress in the application ofRaman spectroscopy and SERS detection of microplastics andderivatives. Microchim. Acta 190, 465 (2023).69. Kubackova, J., Fabriciova, G., Miskovsky, P., Jancura, D. & Sanchez-Cortes, S. Sensitive surface-enhanced Raman spectroscopy (SERS)detection of organochlorine pesticides by alkyl dithiol-functionalized metal nanoparticles-induced plasmonic hot spots.Anal. Chem. 87, 663–669 (2015).70. Li, R. et al. Simultaneous in situ extraction and self-assembly ofplasmonic colloidal gold superparticles for SERS detection oforganochlorine pesticides in water. Anal. Chem. 93,4657–4665 (2021).71. Caldwell, J., Taladriz-Blanco, P., Rodriguez-Lorenzo, L., Rothen-Rutishauser, B. & Petri-Fink, A. Submicron- and nanoplastic detec-tion at lowmicro- to nanogramconcentrations using gold nanostar-based surface-enhanced Raman scattering (SERS) substrates.Environ. Sci. Nano 11, 1000–1011 (2024).72. Kimling, J. et al. Turkevich method for gold nanoparticle synthesisrevisited. J. Phys. Chem. B 110, 15700–15707 (2006).73. Plimpton, S. Fast parallel algorithms for short-range moleculardynamics. J. Comput. Phys. 117, 1–19 (1995).74. Soler, J. M. et al. The SIESTA method for ab initio order-N materialssimulation. J. Phys. Condens. Matter 14, 2745 (2002).75. García, A. et al. SIESTA: recent developments and applications. J.Chem. Phys. 152, 204108 (2020).76. Perdew, J. P. et al. Restoring the density-gradient expansion forexchange in solids and surfaces. Phys. Rev. Lett. 100,136406 (2008).77. van Setten, M. J. et al. The PseudoDojo: training and grading a 85element optimized norm-conserving pseudopotential table. Com-put. Phys. Commun. 226, 39–54 (2018).78. Grimme, S., Antony, J., Ehrlich, S. & Krieg, H. A consistent andaccurate ab initio parametrization of density functional dispersioncorrection (DFT-D) for the 94 elements H-Pu. J. Chem. Phys. 132,154104 (2010).79. Smidstrup, S. et al. QuantumATK: an integrated platform of elec-tronic and atomic-scale modelling tools. J. Phys. Condens. Matter32, 015901 (2020).80. Neese, F. Software update: the ORCA program system—version6.0. WIREs Comput. Mol. Sci. 15, e70019 (2025).81. Becke, A. D. Density-functional thermochemistry. III. The role ofexact exchange. J. Chem. Phys. 98, 5648–5652 (1993).82. Stephens, P. J., Devlin, F. J., Chabalowski, C. F. & Frisch, M. J. Abinitio calculation of vibrational absorption and circular dichroismspectra using density functional force fields. J. Phys. Chem. 98,11623–11627 (1994).83. Grimme, S., Ehrlich, S. & Goerigk, L. Effect of the damping functionin dispersion corrected density functional theory. J. Comput. Chem.32, 1456–1465 (2011).84. Hanwell, M. D. et al. Avogadro: an advanced semantic chemicaleditor, visualization, and analysis platform. J. Cheminform4, 17 (2012).85. Gutiérrez Moreno, J. J. RhSe dataset, NOMAD, https://doi.org/10.17172/nomad.467q-zm1m (2026).Article https://doi.org/10.1038/s41467-026-75365-2Nature Communications |         (2026) 17:9592 14https://doi.org/10.17172/nomad.467q-zm1mhttps://doi.org/10.17172/nomad.467q-zm1mwww.nature.com/naturecommunicationsAuthor contributionsY.K. and O.G. co-supervised the project, performed the sample synth-esis, structural characterization, SERS measurement, and drafted thepaper. Y.T. did part of the SEM and Raman experiments. L.F. and J.Z.carried out the AFM measurements. N.C.-R.C., W.-H.H., M.-H.Y., andK.C.-W.W. performed the XAFS measurements and fitted the data.J.G.M. carried out the atomistic simulations. S.J. performed SERS mea-surement. J.H. performed the TEM, STEM-EDS, and STEM tomographymeasurements. Y.G., X.L., and Y.Z. carried out the XPS and BET mea-surements. Y.A. and M.V. revised the manuscript. Y.Y. conceived theidea, supervised the project, discussed the experimental results, ana-lyzed the data, and drafted the manuscript.FundingThe authors gratefully acknowledge the support of the JST-ERATOYamauchi Materials Space-Tectonics Project (JPMJER2003). Y.Y. disclosessupport for the research of this work from the Australian ResearchCouncil(ARC) (FL230100095). Y.K. discloses support for publication of this workfrom ARC (DE260101548) and Japan Society for the Promotion of Science(26K17641). O.G. acknowledges the FFG FO999923096 project and Aus-trian Science Fund (FWF) (10.55776/COE5) (Cluster of Excellence MECS).J.G.M. acknowledges the AI4S fellowship within the “Generación D”initiative by Red.es, Ministerio para la Transformación Digital y de la Fun-ción Pública, for talent attraction (C005/24-ED CV1), funded by Next-GenerationEU through the Recovery, Transformation and Resilience Plans(PRTR), the MLALH project from the EIG CONCERT-JAPAN program andthe MaX Center of Excellence, supported by the European Union and theSpanish State Research Agency with project no. PCI2023-143426 (MCIN/AEI/10.13039/ 501100011033), and EU grant agreement 101093374and PCI2022-135083-5602, respectively. The electron microscopy workwas supported by the Advanced Research Infrastructure for Materials andNanotechnology in Japan (ARIM) of the Ministry of Education, Culture,Sports, Science and Technology (MEXT) of Japan, proposal numberJPMXP1226NM5072. This work used the Queensland node of the NCRIS-enabled Australian National Fabrication Facility (ANFF).Competing interestsThe authors declare no competing interests.Additional informationSupplementary information The online version contains Supplemen-tary material available athttps://doi.org/10.1038/s41467-026-75365-2.Correspondence and requests for materials should be addressed toYunqing Kang, Olga Guselnikova or Yusuke Yamauchi.Peer review information Nature Communications thanks MuhammadBashouti and Francois Lagugne-Labarthet for their contribution to thepeer review of this work. A peer review file is available.Reprints and permissions information is available athttp://www.nature.com/reprintsPublisher’s note Springer Nature remains neutral with regard tojurisdictional claims in published maps and institutional affiliations.Open Access This article is licensed under a Creative CommonsAttribution-NonCommercial-NoDerivatives 4.0 International License,which permits any non-commercial use, sharing, distribution andreproduction in any medium or format, as long as you give appropriatecredit to the original author(s) and the source, provide a link to theCreative Commons licence, and indicate if you modified the licensedmaterial. Youdonot havepermissionunder this licence toshare adaptedmaterial derived from this article or parts of it. The images or other thirdparty material in this article are included in the article’s CreativeCommons licence, unless indicated otherwise in a credit line to thematerial. If material is not included in the article’s Creative Commonslicence and your intended use is not permitted by statutory regulation orexceeds the permitted use, you will need to obtain permission directlyfrom the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.© The Author(s) 2026Yunqing Kang 1,2,16 , Olga Guselnikova 3,16 , Lei Fu 1,4, Norman C.-R. Chen1,5, José Julio Gutiérrez Moreno6,Joel Henzie 7, Yuqi Guo1, Xiangyang Liu1, Yingji Zhao1, Wei-Hsiang Huang 8,9, Min-Hsin Yeh 9,10, Yagmur Tanis 3,Markus Valtiner 3, Stanislav Ježek11, Jun Zhou 4, Yusuke Asakura1, Kevin C.-W. Wu5,12,13,14 & Yusuke Yamauchi 1,2,151Department of Materials Process Engineering, Graduate School of Engineering, Nagoya University, Nagoya, Japan. 2Australian Institute for Bioengineeringand Nanotechnology (AIBN), The University of Queensland, Brisbane, QLD, Australia. 3Institute of Applied Physics, Vienna University of Technology,Vienna, Austria. 4Center of Nanomaterials for Renewable Energy, StateKey Laboratory of Electrical Insulation and Power Equipment, Xi’an JiaotongUniversity,Xi’an, PR China. 5International Graduate Program of Molecular Science and Technology (NTU-MST), National Taiwan University, Taipei, Taiwan, ROC. 6Bar-celona Supercomputing Center, Barcelona, Spain. 7Research Center for Materials Nanoarchitectonics, National Institute for Materials Science (NIMS),Tsukuba, Japan. 8National Synchrotron Radiation Research Center (NSRRC), Hsinchu, Taiwan, ROC. 9Sustainable Electrochemical Energy Development(SEED) Center, National Taiwan University of Science and Technology, Taipei, Taiwan, ROC. 10Department of Chemical Engineering, National Taiwan Uni-versity of Science and Technology, Taipei, Taiwan, ROC. 11ENVITES, spol. s r.o., Brno, Czechia. 12Center of Atomic Initiative for New Materials (AI-MAT)/Department of Chemical Engineering, National Taiwan University, Taiwan, ROCTaipei. 13Faculty of Engineering, Chulalongkorn University, Bangkok, Thailand.14Department of Chemical Engineering and Materials Science, Yuan Ze University, Taoyuan, Taiwan, ROC. 15Department of Convergent Biotechnology &Advanced Materials Science, Kyung Hee University, Yongin-si, South Korea. 16These authors contributed equally: Yunqing Kang, Olga Guselnikova.e-mail: yunqing.kang@uq.edu.au; olga.guselnikova@tuwien.ac.at; y.yamauchi@uq.edu.auArticle https://doi.org/10.1038/s41467-026-75365-2Nature Communications |         (2026) 17:9592 15https://doi.org/10.1038/s41467-026-75365-2http://www.nature.com/reprintshttp://creativecommons.org/licenses/by-nc-nd/4.0/http://creativecommons.org/licenses/by-nc-nd/4.0/http://orcid.org/0000-0001-8719-5098http://orcid.org/0000-0001-8719-5098http://orcid.org/0000-0001-8719-5098http://orcid.org/0000-0001-8719-5098http://orcid.org/0000-0001-8719-5098http://orcid.org/0000-0002-2594-9605http://orcid.org/0000-0002-2594-9605http://orcid.org/0000-0002-2594-9605http://orcid.org/0000-0002-2594-9605http://orcid.org/0000-0002-2594-9605http://orcid.org/0000-0002-0575-6128http://orcid.org/0000-0002-0575-6128http://orcid.org/0000-0002-0575-6128http://orcid.org/0000-0002-0575-6128http://orcid.org/0000-0002-0575-6128http://orcid.org/0000-0002-9190-2645http://orcid.org/0000-0002-9190-2645http://orcid.org/0000-0002-9190-2645http://orcid.org/0000-0002-9190-2645http://orcid.org/0000-0002-9190-2645http://orcid.org/0000-0001-9503-0373http://orcid.org/0000-0001-9503-0373http://orcid.org/0000-0001-9503-0373http://orcid.org/0000-0001-9503-0373http://orcid.org/0000-0001-9503-0373http://orcid.org/0000-0002-6150-4750http://orcid.org/0000-0002-6150-4750http://orcid.org/0000-0002-6150-4750http://orcid.org/0000-0002-6150-4750http://orcid.org/0000-0002-6150-4750http://orcid.org/0009-0000-3690-3314http://orcid.org/0009-0000-3690-3314http://orcid.org/0009-0000-3690-3314http://orcid.org/0009-0000-3690-3314http://orcid.org/0009-0000-3690-3314http://orcid.org/0000-0001-5410-1067http://orcid.org/0000-0001-5410-1067http://orcid.org/0000-0001-5410-1067http://orcid.org/0000-0001-5410-1067http://orcid.org/0000-0001-5410-1067http://orcid.org/0000-0001-5729-7261http://orcid.org/0000-0001-5729-7261http://orcid.org/0000-0001-5729-7261http://orcid.org/0000-0001-5729-7261http://orcid.org/0000-0001-5729-7261http://orcid.org/0000-0001-7854-927Xhttp://orcid.org/0000-0001-7854-927Xhttp://orcid.org/0000-0001-7854-927Xhttp://orcid.org/0000-0001-7854-927Xhttp://orcid.org/0000-0001-7854-927Xmailto:yunqing.kang@uq.edu.aumailto:olga.guselnikova@tuwien.ac.atmailto:y.yamauchi@uq.edu.auwww.nature.com/naturecommunications Low curvature combined with high amorphization improves SERS sensitivity of semiconducting Rh–Se mesoporous nanospheres toward non-adsorbing resistant pollutants Results Synthesis of a-Rh–Se MNs with various compositions Curvature modulation of a-Rh–Se MNs Crystallinity modulation of a-Rh–Se-L MNs SERS performance of various Rh–Se samples Surface-curvature effect Crystallinity modulation Tentative mechanism Quantitative detection for HCB Anthracene sensing in industrial wastewater Persistent microplastic PTFE sensing Discussion Methods Materials Synthesis of a-Rh–Se-L MNs Synthesis of Rh–Se MNs with different crystallinity Synthesis of a-Rh–Se-M Synthesis of a-Rh–Se-S Synthesis of AuNPs Characterizations SERS measurements Preparation of SERS substrate and analyte deposition Raman spectroscopy measurements EF calculation Calculation of LOD Preparation of PTFE MP solutions in WWT Preparation of anthracene solutions in industrial wastewater Storage stability test Computational methods Reporting summary Data availability References Author contributions Funding Competing interests Additional information