Cu-doping SnO 2 -NiO p-n Heterostructure for Significant Raman Enhancement with EF > 10 10 : Toward Ultrasensitive VOCs Sensing

Surface enhanced Raman scattering (SERS) based on chemical mechanism (CM) has attracted tremendous attention for its great selectivity and stability. However, extremely low enhancement factor (EF) limits its practical applications for trace detection. Here, a novel sponge-like Cu-doping SnO 2 -NiO p-n semiconductor heterostructure (SnO 2 -NiOx/Cu), was first designed and created as a CM-based SERS substrate with a significant EF of 1.66 × 10 10 . This remarkable EF was mainly attributed to the enhanced charge-separation efficacy of p-n heterojunction and charge transfer resonance resulted from Cu doping. Moreover, the porous sponge structure enriched the probe molecules, resulting in further SERS signals magnification. By immobilizing copper phthalocyanine as an inner-reference element, SnO 2 -NiOx/Cu was developed as a SERS nose for selective recognition of multiple lung cancer related-volatile 2 and holes from SnO 2 to NiOx/Cu, further facilitating the separation of photogenerated electron-hole pairs 11 . Fur-thermore, the narrower bandgap of NiOx/Cu also realized CT resonance to promote the CT process. The enhanced charge-separation efficacy of type-II p-n heterojunction and CT resonance resulted from Cu doping led to incredible SERS performance improvement on SnO 2 -NiOx/Cu with an EF 10 7 times higher than that on SnO 2 -NiO. This CT process was verified by density functional theory (DFT) calculation and time-resolved diffuse reflectance (TDR) spectroscopic measurements 12 . The present high-EF SnO 2 -NiOx/Cu substrate was then developed as a SERS nose (SnO 2 -NiOx/Cu-CuPc) through immobilization of copper phthalocyanine (CuPc) as an inner-reference element for volatile organic compounds (VOCs) sensing. Exploiting the target molecular vibrational fingerprint and the the well matching degree of energy level between target molecules and SnO 2 -NiOx/Cu, our SERS nose showed excellent selectivity to ‘sniff out’ and simultaneously quantify multiple toxic VOCs in the exhaled breath (EB) of early stage lung cancer patients at  ppb level, which was 10 times lower than those obtained by gas chromatography-mass spectrometry (GC-MS) 13 and fluorescence spectroscopy 14 , 100 times lower than those obtained by electrochemical methods 15 . A visual diagnostic report on the concentration of toxic VOCs was obtained by SERS mapping and their corresponding barcodes. This new avenue shows great potential for developing accurate, fast, easy and cost-effective diagnosis instruments for screening of early-stage lung cancer. heterostructure (SnO 2 -NiOx/Cu) for Raman enhancement. SnO 2 -NiOx/Cu nanosponge was obtained using a hydrothermal method combined with a subsequent calci-nation. It was designed and created as a CM-based SERS substrate with extremely high Raman enhancement (EF=1.66 × 10 10 ) using CuPc as a prototype probe molecule under 785 nm laser excitation. value for CM-based SERS substrates reported by far. Both experimental results and theoretical calculations revealed that the SERS enhancement was mainly stemmed from efficient charge separation by SnO 2 -NiOx/Cu p-n heterojunction and charge transfer resonance caused by Cu doping, which enhanced charge transfer efficiency between SnO 2 -NiOx/Cu substrate and probe molecules. Besides, enrichment of probe molecules from porous sponge structure further magnified SERS signals. Then, by immobilizing of CuPc molecule as an inner-reference element, the SnO 2 -NiOx/Cu substrate was developed as a SERS nose (SnO 2 -NiOx/Cu-CuPc) for VOCs analysis with high sensitivity, selectivity, and accuracy. The developed SERS nose was able to ‘sniff out’ and simultaneously quantify multiple toxic VOCs in the EB of early stage lung cancer patients at ppb level, which was 10 times lower than those obtained by GC-MS and fluorescence spectroscopy, 100 times lower than that obtained by electrochemical method. Unique barcodes denoting the VOCs concentrations were obtained by integrating R, G, and B values of the Raman intensity images, where these barcodes can be easily read by electronic devices, such as smartphones and handheld scanners. This work has provided a methodology for designing and synthesizing p-n heterostructures to enhance Raman signals and has demonstrated the practical potential of a desktop SERS device for biomarker screening. This established method can also be extended to monitor the levels of other VOCs, which are related to diabetes, advanced liver disease, kidney failure, and lung abscess.

Surface-enhanced Raman scattering (SERS) has been widely utilized as a promising analytical technique for trace detection 1 .
Each breakthrough in the SERS field is inseparable from the development of the enhancement substrates. Conventional SERS substrates based on electromagnetic mechanism (EM) 2 , such as Au and Ag, exhibit ultrahigh Raman enhancement with enhancement factor (EF) up to 10 12 . However, the EM-based SERS effect has no selectivity for Raman reporters, resulting in complicated signal outputs 3 . Another SERS mechanism is chemical mechanism (CM), which relies on the charge transfer (CT) processes between adsorbed molecules and SERS substrate 4 . Today, due to the high molecule selectivity and signal reproducibility, the development of CM-based SERS substrates, such as metal oxides 5 , graphene 6 , and other semiconductors 7 has attracted more attention. Unfortunately, the CM-based SERS effect shows relatively low EFs (10 3 -10 5 ) since the CT process is a short-range one 8 . The poor charge separation in CM-based SERS substrates is a major limitation, hindering further enhancement of their SERS performance 9 .
Here, a sponge like Cu-doping NiO (NiOx/Cu)-SnO2 p-n semiconductor heterostructure (SnO2-NiOx/Cu) was designed and created as a CM-based SERS substrate with extremely high Raman enhancement (EF=1.66 × 10 10 ). The SnO2-NiOx/Cu heterostructure was confirmed at the atomic resolution scale by aberration-corrected scanning transmission electron microscopy (ACSTEM). This notable EF value was mainly attributed to a CT process with high efficiency generated by the integration of p-n heterojunction and Cu doping, and enrichment of probe molecules by the porous nanosponge structure of SnO2-NiOx/Cu, as illustrated in Fig. 1. The selective Cu doing on NiO narrowed the band gap through Fermi-level alignment, leading to the conversion from type-I SnO2-NiO heterojunction into type-II SnO2-NiOx/Cu heterojunction 10 . This structure ensured the delivery of photo-generated electrons from NiOx/Cu to SnO2 and holes from SnO2 to NiOx/Cu, further facilitating the separation of photogenerated electron-hole pairs 11 . Furthermore, the narrower bandgap of NiOx/Cu also realized CT resonance to promote the CT process. The enhanced charge-separation efficacy of type-II p-n heterojunction and CT resonance resulted from Cu doping led to incredible SERS performance improvement on SnO2-NiOx/Cu with an EF 10 7 times higher than that on SnO2-NiO. This CT process was verified by density functional theory (DFT) calculation and time-resolved diffuse reflectance (TDR) spectroscopic measurements 12 . The present high-EF SnO2-NiOx/Cu substrate was then developed as a SERS nose (SnO2-NiOx/Cu-CuPc) through immobilization of copper phthalocyanine (CuPc) as an inner-reference element for volatile organic compounds (VOCs) sensing. Exploiting the target molecular vibrational fingerprint and the the well matching degree of energy level between target molecules and SnO2-NiOx/Cu, our SERS nose showed excellent selectivity to 'sniff out' and simultaneously quantify multiple toxic VOCs in the exhaled breath (EB) of early stage lung cancer patients at ppb level, which was 10 times lower than those obtained by gas chromatography-mass spectrometry (GC-MS) 13 and fluorescence spectroscopy 14 , 100 times lower than those obtained by electrochemical methods 15 . A visual diagnostic report on the concentration of toxic VOCs was obtained by SERS mapping and their corresponding barcodes. This new avenue shows great potential for developing accurate, fast, easy and cost-effective diagnosis instruments for screening of early-stage lung cancer.

Characterization and SERS Performance of SnO2-NiOx/Cu
The sponge like SnO2-NiOx/Cu p-n heterostructure was first designed and synthesized through a hydrothermal method combined with a subsequent calcination. As shown in scanning electron microscopic (SEM) image (Fig. 2a), the as-prepared porous SnO2-NiOx/Cu nanosponges were highly uniform with dimeter of 1.5 ± 0.1 µm (n=15) (Supplementary Fig. 1a). The SERS activity of SnO2-NiOx/Cu was evaluated using CuPc as a prototype probe molecule under 785 nm laser excitation. Remarkably, the prominent signal peak at 1530 cm -1 corresponding to the in-plane ring symmetric nonmetal bound N-C stretch of CuPc (10 -5 M) 16 was detected on SnO2-NiOx/Cu, while on SnO2-NiO, SnO2 (for details, see Supplementary Fig. 2 and 3) and bare Si substrates, the signals were almost undetectable (EF < 2.15× 10 3 ) ( Fig. 2b and 2c). The EF of SnO2-NiOx/Cu for CuPc was calculated to be as high as 1.66 × 10 10 (for calculation details, see Supplementary Methods), which was the highest EF of CM-based substrates reported to date (Supplementary Table 1).
This remarkable EF was deduced to be mainly ascribed to p-n junction of SnO2-NiOx/Cu formed by p-type NiOx/Cu and n-type SnO2. In the Mott-Schottky plot, the p-n junction characteristic was observed with an inverted "V-shape" (Supplementary Fig. 5a) 17 .
The precise p-n heterostructure of SnO2-NiOx/Cu was further explored by ACSTEM high angle annular dark field imaging (HAADF) ( Fig. 2e-g). The lattice fringes corresponding to (110) plane of tetragonal phase SnO2 18 and (111) plane of NiOx/Cu 19 were clearly  Fig. 5b) and X-ray diffraction (XRD) pattern (Fig. 2d). The corresponding ACSTEM energy dispersive spectroscopy (EDS) mapping illustrated that Sn atoms were uniformly distributed across one side of heterojunction boundary while Ni and Cu atoms were distributed over the other side ( Fig. 2g (ii)), indicating that Cu was only doped into NiO instead of SnO2.
Next, X-ray photoelectron spectroscopy (XPS) was performed to elucidate the chemical bonding in SnO2-NiOx/Cu nanosponge.
The results suggested that Cu doping in SnO2-NiOx/Cu altered the Ni surrounding chemical environments. XPS data of Cu showed the binding energy of Cu 2p3/2 for SnO2-NiOx/Cu at 932.5 and 933.3 eV, which were consistent with the divalent states of Cu + and Cu 2+ (Fig. 2j) 22 . From O 1s XPS spectrum of SnO2-NiOx/Cu, three peaks were clearly observed, in which O1s peak at 532.2 eV was ascribed to the lattice oxygen (O 2-), while the high binding energy (533.3 eV) was assigned to the absorbed oxygen ions (O 2and O -) on the surface 23 , which were consistent with those of SnO2-NiO. In addition, another peak at the lower energy (531.0 eV) was attibuted to the vacant oxygen (OV) 24 , indicating that Cu doping induced a large number of OV in SnO2-NiOx/Cu. These obvious changes indicated the existence of interaction between n-type SnO2 and p-type NiOx/Cu, resulting in an efficient CT process between Raman reporter and SnO2-NiOx/Cu SERS substrate.

Mechanism of Raman enhancement in SnO2-NiOx/Cu
The CT process between SnO2-NiOx/Cu and CuPc was then investigated through TDR spectroscopic measurements. Excited by 785 nm laser, the absorption bands centered at 530 nm were observed for CuPc-SnO2, CuPc-SnO2-NiO and CuPc-SnO2-NiOx/Cu systems, which were assigned to trapped electrons ( Supplementary Fig. 7). Time profiles of the transient absorption were fitted by two-exponential functions (Fig. 3a). The decay lifetime of electrons in CuPc-SnO2-NiOx/Cu (1236.39 ps) was much longer than that in CuPc-SnO2 (433.82 ps) and CuPc-SnO2-NiO (325.37 ps), suggesting that p-n junction dopped by Cu at SnO2-NiOx/Cu interface acted as an electron trap site and led to efficient CT. To further explain TDR results, we investigated the electronic structure and the coupling of CuPc and SnO2-NiOx/Cu as well as the consequent CT using DFT calculations. As shown in Fig. 3b, the amount of electron transfer from CuPc to SnO2-NiOx/Cu (1.32 e -) was much more than that on SnO2-NiO (0.68e -) or SnO2 (0.87 e -). Both TDR results and DFT calculations provided strong evidence for effective CT between SnO2-NiOx/Cu and CuPc, resulting in significant enhancement of Raman activity for CuPc on SnO2-NiOx/Cu substrate.
As well known, the CM-based SERS effect strongly depends on the matching degree of the energy level between molecules and substrates 7 .To further elucidate the CT process, we determined CB and VB of SnO2, NiO and NiOx/Cu, and highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) levels of CuPc using DFT calculations. As shown in Fig. 3c, the calculation result was in a good agreement with that of the standard Tauc plot method (Supplementary Fig. 8). Without Cu doping, the energy barriers between VB of NiO or SnO2 substrates and HOMO of CuPc were 2.2 and 2.1 eV, which were much greater than the energy of 785-nm laser (1.58 eV). Thus, the photon energy of 785 nm laser failed to excite the electrons at VB of the substrate into HOMO of CuPc and hard to separate electron-hole pairs effectively, thus leading to very weak Raman signals. However, the introduction of OV created by Cu doping shifted the Fermi level of NiO toward CB and narrowed the band gap 13 . Under this situation, electrons at VB of NiOx/Cu substrate were excited and transferred into HOMO of CuPc successfully. Meanwhile, the energy difference between VB of NiOx/Cu and SnO2 in CuPc-SnO2-NiOx/Cu nanosystem was 1.5 eV, which was close to 1.58 eV. In this case, CT resonance occurred, leading to an obvious Raman enhancement on SnO2-NiOx/Cu (Fig. 3d).
Significantly, the selective Cu doing on NiO narrowed the band gap through Fermi-level alignment, resulting in the conversion from type-I SnO2-NiO heterojunction into type-II SnO2-NiOx/Cu heterojunction. As for type I SnO2-NiO semiconductor, the band of SnO2 was embedded within that of NiO so that photogenerated electrons and holes were both concentrated in SnO2. After Cu doping, the Feimi-level of NiOx/Cu lay closer to CB (Fig. 3c). Electrons in n-type SnO2 filled some of the holes in p-type NiOx/Cu because these holes were available at lower-energy states. This left positively charged holes in n-type SnO2 and extra negatively charged electrons in p-type NiOx/Cu 25 . Thus, an inner electric field was formed at the equilibrium (Fig. 3c). Driven by the internal field, holes at VB of SnO2 were excited into VB of NiOx/Cu, and electrons were accelerated from CB of NiOx/Cu into CB of SnO2. In this case, photogenerated electron-hole pairs were effectively separated by type II SnO2-NiOx/Cu p-n junction between NiOx/Cu and SnO2, realizing efficient spatial charge separation and prolongating the lifetime of charge carriers 26 , and further enhancing the SERS activity.
In order to confirm the effect of p-n junction on Raman enhancement for SnO2-NiOx/Cu, SnO2-In2S3 porous nanocomposite with nonp-n junction was synthesized, in which CB and VB levels of n-type In2S3 (-3.7 eV and -5.7 eV) were close to that of p-type NiOx/Cu. TDR results showed that electrons decay lifetime in CuPc-SnO2-In2S3 (588.65 ps) was much shorter than that in CuPc-SnO2-NiOx/Cu, indicating that p-n junction in SnO2-NiOx/Cu acted as an electron trap site and led to more efficient CT (Supplementary Fig. 9a and   b). Thus, compared with SnO2-NiOx/Cu, obvious decline (30%) in SERS intensity of CuPc was observed on SnO2-In2S3 porous nanocomposites ( Supplementary Fig. 9c). Consequently, such synergetic promotion of CT resonance and type II p-n heterojunction formation in SnO2-NiOx/Cu led to a high CT efficiency, further resulting in a remarkable SERS enhancement.
The above theoretical and experimental results revealed that the ultrahigh Raman enhancement on SnO2-NiOx/Cu was mainly derived from CT mechanism. To clarify whether EM mechanism can be observed in SnO2-NiOx/Cu substrate, finite difference time domain method (FDTD) was used to simulate and calculate the EM field distribution under 785-nm excitation for SnO2, SnO2-NiO and SnO2-NiOx/Cu. As shown in Fig. 3e, SnO2-NiOx/Cu obtained stronger electric near-field than SnO2 or SnO2-NiO. The reason was that Cu doping led to the generation of a high density of holes in VB of NiOx/Cu, sustaining an localized surface plasmon response (LSPR) in the near infrared (NIR) region. However, the EM effect on SERS enhancement was very limited, which was 8 orders of magnitude lower than the CM effect. (Supplementary Fig. 10).
Moreover, the Brunauer-Emmett-Teller surface area and pore size of as-prepared SnO2-NiOx/Cu nanosponge were calculated to be as high as 26.16 m 2 g −1 and 3.8±0.3 nm ( Supplementary Fig. 11), which was similar to that obtained from SEM image. Owing to CuPc enrichment by this porous structure, the SERS effect of as-prepared SnO2-NiOx/Cu for CuPc was 1.8 times higher than that of SnO2-NiOx/Cu with compact surface (Fig. 3f). Consequently, the remarkable EF of SnO2-NiOx/Cu should be mainly ascribed to 3 factors: efficient CT process attributed to (i) type II p-n junction in SnO2-NiOx/Cu for charge collection and separation, (ii) CT resonance under 785-nm laser excitation, and as well as (iii) the molecule enrichment by porous SnO2-NiOx/Cu nanosponge.
increasing concentration of these VOCs. Then, the VOC concentration information was automatically transformed into an optical, machine readable barcode by integrating R, G, B values of Raman images and reading concentrations based on linear curves (Supplementary Fig. 14). Hence, the composition and concentration information on VOCs were integrated to yield barcodes that can be read rapidly and conveniently by a barcode scanner.Meanwhile, the SERS nose showed a good stability within 3 months (< 4.21 %) and a satisfying reproducibility with RSDs below 3.18% (n=5) (Supplementary Fig. 15). These results demonstrated that the present SERS nose provided good sensitivity, specificity and stability for multiple VOC detection at ppb levels.
Furthermore, we took advantage of the developed SERS nose to demonstrate the simultaneous detection of PYR, 2-NT and EBZA in the EB of early stage lung cancer patients (Fig. 5a). Fig. 5b and 5d show SERS spectra and corresponding barcodes of EB samples for 10 lung cancer patients and 5 healthy people. A regression analysis between the proposed SERS nose and GC-MS was conducted to demonstrate its utility ( Supplementary Fig. 16). Compared with single-VOC tests, the simultaneous detection of three VOCs improved the reliability and avoid false positive diagnoses (Fig.5c). With the aid of multiple VOC detection based on SnO2-NiOx/Cu-CuPc SERS nose, the diagnosis and screening of lung cancer can be conducted in a single fast, easy and cost-effective assay, thus affording an ideal test method for point of care detection.

Discussion
In this work, a novel sponge-like SnO2-NiOx/Cu-CuPc SERS nose was first developed for ultrasensitive and selective recognition of multiple VOCs down to ppb level. First of all, the sponge like SnO2-NiOx/Cu type II p-n semiconductor heterostructure, which was verified by ACSTEM and Mott-Schottky plots, was synthesized using a hydrothermal method combined with a subsequent calcination. The EF value of SnO2-NiOx/Cu for CuPc was achieved up to 1.66 × 10 10 , the highest EF value for CM-based SERS substrates reported by far. Both experimental results and theoretical calculations revealed that the SERS enhancement was mainly stemmed from efficient charge separation by SnO2-NiOx/Cu p-n heterojunction and charge transfer resonance caused by Cu doping, which enhanced charge transfer efficiency between SnO2-NiOx/Cu substrate and probe molecules. Besides, enrichment of probe molecules from porous sponge structure further magnified SERS signals. Then, by immobilizing of CuPc molecule as an innerreference element, the SnO2-NiOx/Cu substrate was developed as a SERS nose (SnO2-NiOx/Cu-CuPc) for VOCs analysis with high sensitivity, selectivity, and accuracy. The developed SERS nose was able to 'sniff out' and simultaneously quantify multiple toxic VOCs in the EB of early stage lung cancer patients at ppb level, which was 10 times lower than those obtained by GC-MS and fluorescence spectroscopy, 100 times lower than that obtained by electrochemical method. Unique barcodes denoting the VOCs concentrations were obtained by integrating R, G, and B values of the Raman intensity images, where these barcodes can be easily read by electronic devices, such as smartphones and handheld scanners. This work has provided a methodology for designing and synthesizing p-n heterostructures to enhance Raman signals and has demonstrated the practical potential of a desktop SERS device for biomarker screening. This established method can also be extended to monitor the levels of other VOCs, which are related to diabetes, advanced liver disease, kidney failure, and lung abscess.

Methods
Methods are included in the Supplementary Information.

Data availability
The data that support the plots within this paper and other findings of this study are available from the corresponding author on reasonable request.