YU Ling-xiao, LU Rui-tao,2,*
(1. State Key Laboratory of New Ceramics and Fine Processing, School of Materials Science and Engineering, Tsinghua University, Beijing 100084, China;2. Key Laboratory of Advanced Materials (MOE), School of Materials Science and Engineering, Tsinghua University, Beijing 100084, China)
Abstract: Surface-enhanced Raman scattering (SERS) has been regarded as an attractive technique for efficient molecular sensing because of its nondestructive detection, fast response and high sensitivity. However, the majority of studies on SERS are still based on noble metals (e.g. Au, Ag), which suffer from the drawbacks of high-cost, low uniformity and poor stability, thus limiting their widespread use. Graphene shows an efficient SERS performance because of its two-dimensional (2D) atomically flat surface,large specific surface area, high stability and unique electronic/optical properties, which open up new avenues for SERS research. In recent years, other 2D inorganic layer materials, such as transition metal dichalcogenides (TMDCs), hexagonal boron nitride (h-BN),black phosphorus (BP), and MXenes, have also attracted increasing research attention. We summarize the SERS mechanisms and state-of-the-art progress on substrates based on 2D materials, including graphene and other 2D inorganic layer materials. The challenges and prospects for future research on high-performance SERS substrates are considered.
Key words: Graphene;SERS;Enhancement mechanism;Molecular sensing
Since C.V. Raman reported the scattering phenomenon in 1928, Raman scattering has been regarded as a promising analytical technology, however the poor intensity of Raman signal caused by the lack of inelastic scattered photons restricts its wide applications[1]. In 1974, the discovery of surface-enhanced Raman scattering (SERS) on rough silver surface by M. Fleishmanet al.contributed to enhancing scattering intensity several orders of magnitude[2], remarkably broadening its utilizations on molecular sensing,such as environmental monitoring[3], food safety[4],disease diagnosis[5], biosensing[6], and molecular fingerprint recognition[7]. Besides, because of the shorter vibrational relaxation time than that of electron, there are higher emitted rate of Raman photons by a molecule than fluorescence photons, which makes it promising for single molecular detection with high sensitivity for SERS[8]. In virtue of noncontact,nondestructive detection, rapid response and high sensitivity, SERS has attracted numerous efforts of researchers on developing novel SERS substrates with higher sensitivity, better uniformity and reproducibility[9–12]. To date, noble metals, such as Au and Ag, are main materials served as SERS platforms with high performance, which however suffer from high-cost,low surface uniformity, poor stability, strong spectral background and even arising side catalytic reactions for too strong interaction with probe molecules sometimes, thus hindering their further scale-up applications[13]. In this context, it is crucial to develop noblemetal-free SERS-active substrates for future widespread applications.
Recently, graphene and other 2D inorganic layered materials, such as transition metal dichalcogenides (TMDCs), hexagonal boron nitride (h-BN),black phosphorus (BP), MXenes,et al., have been investigated as possible candidates for SERS substrates for molecular sensing to overcome the drawbacks of noble metals with large specific surface areas, unique electronic and optical properties, atomic uniformity,biological compatibility and high stability (Fig.1)[14].However, there are still few studies on 2D inorganic layered nanomaterials served as alternative SERS substrates compared with noble metals and less experience has been accumulated to guide further development. In this article, we will focus on the recent development and enhancement mechanism of 2D inorganic layered materials for molecular sensing. Then we will emphasize the factors on SERS performance in details, which will be insightful for tuning the properties of substrate materials. Finally, we propose the challenges ahead based on current advances of 2D inorganic layered materials and perspectives for future development.
Fig. 1 Various inorganic 2D layered materials for surface-enhanced Raman scattering (SERS). Here h-BN denotes hexagonal boron nitride. TMDCs and BP denote transition metal dichalcogenides and black phosphorus, respectively.
Raman signal arises from the inelastic scattered photons (i.e. the Stokes scattering and Anti-Stokes scattering) when the incident photons irradiates on the samples. As shown in Fig. 2a and b, it can be roughly divided into two steps: a few incident photons gain (or lose) energy when they interact with samples and then scatter with more (or less) energy, in which the change of the energy is called as Raman shift[10,15]. As for molecular sensing, the incident light will induce a dipole moment (μ) into the probe molecule, of which the square is proportional to the Raman intensity (I),expressed as[11]:
Thus, the Raman signal can be amplified through enhancing the dipole moments of the molecules, such as increasing the incident electromagnetic field surrounding the molecules and promoting the polarization of the molecules. Up to now, there are two mechanisms of enhancement for Raman scattering accepted generally, the electromagnetic mechanism (EM)and the chemical mechanism (CM)[14].
Fig. 2 Schematic illustrations of (a) SERS, (b) mechanism of Raman scattering (ν0 and Δν represent the frequency of incident light and the change of frequencies after scattering, respectively), (c) electromagnetic mechanism (EM) and (d) chemical mechanism (CM) for SERS (Here E0 is the incident electromagnetic field, μmol is the induced dipole moment of probe molecule, r is the radius of the metal spheres and d is the distance between the probe molecule and the surface of metal sphere, CT represents the charge transfer, HOMO and LUMO denote the highest occupied molecular orbital and the lowest unoccupied molecular orbital, respectively).
It can be simply described that metal nanospheres are exposed in the applied electromagnetic field and the oscillating electromagnetic field will excite the electrons in the metal spheres leading to polarizations of charges and inducing dipole moments[15].When the angular frequency of the induced dipole moments is equal to that of the incident light, there will be a significant enhancement for local electromagnetic field on the surface caused by local resonance effect, called localized surface plasmon resonance (LSPR)[17]. Therefore, while the molecules approach the surface of the metal spheres, they can be polarized by a stronger local electromagnetic field(Eloc), which can be expressed as[15]:
whereISERSrepresents the intensity of the Raman signal,EincandEscaare regarded as the fields from the incident light and the scattering light respectively.Therefore, the electromagnetic enhancement factor(EEF) is approximately proportional to the fourth power of the ratio of the local electromagnetic field to the incident electromagnetic field, which can be expressed as:
Generally, the orders of magnitude of the enhancement factor can be up to 8 or even larger[18,19].Indispensably, it is worthy to note that the most intense scattering arises not from the metal particle itself but in the gaps between the surrounding particles to bring large enhancement in most cases, which are usually called as “hot spots”[15].
Chemical mechanism (CM) is more ambiguous compared to electromagnetic mechanism, there is still no specific and concrete consensus to comprehensively explain the process of CM. Encouragingly, with the further studies on plasmon-free SERS materials and fast advances of computer simulation technology,chemical mechanism has been elucidated gradually[9,10,19]. Chemical mechanism for SERS is dependent on the electronic structure of the interface between the probe molecules and the substrates, thus the probe molecule layer closest by the substrate is the key factor for CM, which is regarded as “first layer effect”generally[11]. There are three forms of chemical mechanism: (1) the probe molecules interact with the substrate by chemical bonds. (2) the probe molecules form coordination compounds with the SERS substrate. (3) charge transfer occurs between the probe molecules and the SERS substrate. Among them, it is notable that the charge transfer (CT) mechanism tends to occur between the adsorbed molecules and 2D inorganic layered material SERS substrates. According to frontier molecular orbital theory, the electrons in molecules can be divided into different molecular orbital energy levels for different energies[20], among which the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) are the keys to the interaction with other matters. Fig. 2d illustrates the charge transfer mechanism for SERS,the electrons in the HOMO levels of the probe molecules can be excited to the LUMO levels by the incident light (Fig. 2d(i)), then the electrons may migrate from the valence band (VB) or Fermi level of substrates to the HOMO levels of the probe molecules and recombine with remaining holes(Fig. 2d(ii))[21], in which according to the Fermi’s golden rule, the electron transition probability ratewlkcan be given as[22]:
whereg(Ek) is the density of the states (DOS) andrepresents the matrix element for a transition from LUMO to HOMO. Therefore, the Raman enhancement is determined by the DOS and the matrix element of different substrate materials, which can be tailored by the position of the energy levels and the interfacial interaction[21]. Besides, the Raman signal can be also dramatically enhanced by charge transfer resonance when the energy gap between the energy level of the substrates (taking the Fermi level as an example in Fig. 2d(iii)) and LUMO or HOMO levels of the probe molecule is close to the energy of the laser[23]. In addition, the interfacial dipole-dipole interaction can also improve the polarization of the probe molecules to increase the final Raman signal. Thus,the total dipole moments of the molecule in the chemical mechanism can be simply expressed as[24]:
Hereμ0andμlocare the dipole moments induced by incident light and substrates, respectively. In general, the enhancement factor of chemical mechanism is about 100–102, which is much weaker than that of electromagnetic mechanism. But inspiringly, there are also many works on 2D layered material SERS substrates verified to have ultra-high sensitivities and low limit of detection recently by tuning the electronic structures and properties with various methodologies.
Graphene, the star of 2D inorganic layered materials, is a monolayer sp2-hybridized carbon atoms arranged into honeycomb-like skeletons, which has attracted vast attentions of researchers over the past decades[25]. Since it was discovered in 2004[26], graphene has been employed in various fields due to its large specific surface area, atomic uniformity, excellent mechanical capacities, high stability and good biocompatibility, as well as unique and tunable electronic and optical properties[14,27–28]. Recently,graphene has also been demonstrated as an active SERS substrate with an efficient fluorescence quenching ability and enhancement for Raman signal of some dye molecules, such as rhodamine B (RhB), crystal violet (CV), methylene blue (MB), phthalocyanine (Pc)and so on, by chemical mechanism, which is also termed as graphene-enhanced Raman scattering(GERS) effect[29,30]. So far, there are two models of charge transfer mechanisms for chemical enhancement, the ground-state and the excited-state mechanisms. In order to investigate the type of charge transfer mechanism in GERS, Linget al.[31]tested the Raman spectra of CuPc under different laser wavelengths and they found that there was no chargetransfer resonance peak at about 1.9 eV, which is not consistent with the excited-state model, identifying a ground-state charge transfer mechanism in GERS. For further understanding of GERS, the factors responsible for enhancement factor of GERS can be considered from three aspects: graphene, the probe molecule and the incident laser.
Fig. 3 (a) Electrical field modulation on Fermi level for graphene-enhanced Raman scattering (GERS). (i) Schematic diagram of the device on tuning the graphene Fermi level by electrical field effect. (ii) The energy level alignment between the molecular energy level and modulated graphene Fermi level for charge transfer resonance[23] (reprinted with permission from American Chemical Society). (b) Defect engineering for GERS. (i) Schematic illustrations of the Fermi level modulation by O2 plasma treatment[33] (reprinted with permission from Elsevier). (ii) the dI/dV curves (averaged over 9 points spectra taken in 1×1 nm2 area) measured on N2AA dopants and undoped graphene[25] (reprinted with permission from Spring Nature). The density of states (DOS) for Rhodamine B (RhB) on (iii) the pristine graphene (PG) and (iv) N-doped graphene (NG), EF represents the Fermi level[14] (reprinted with permission from AAAS),(v) Schematic of SERS on Si-doped graphene[32] (reprinted with permission from John Wiley and Sons). (c) Thickness-dependence of GERS. (i-ii) Raman spectra of copper phthalocyanine (CuPc) molecule and (iii) schematic illustration of the energy band structure of mono- and bilayer graphene under low and high concentration molecular solution[36] (reprinted with permission from American Chemical Society).
Graphene is a zero-band-gap semiconductor and its Fermi level tends to locate between the LUMO and HOMO levels of dye molecules generally, which can be tunedviavarious strategies, such as electrical field effect (EFE), chemical doping, defect engineering and so on[23,32–35]. For example, Xuet al.[23]demonstrated an electrical field modulation method to tune the Fermi level of graphene by electrical field effect and they found that the Fermi level could be down-shifted by a negative gate voltage due to the EFE-induced hole doping, whereas up-shifted due to the electron doping, as shown in Fig. 3a. It was reported that the downshifted Fermi level made the energy gap between the LUMO level of the CoPc molecule and the Fermi level of graphene close to the laser energy,thus leading to a charge transfer resonance process,which efficiently enhanced the Raman intensity. Recently, plasma-based technique has been proved as an elaborate route to decorate graphene. There exist strong local dipoles with oxygenated species on treated surface and down-shifted Fermi level of graphene by an O2plasma treatment (Fig. 3b(i)). The local dipoles may promote the polarization of the probe molecule for strong interaction with substrate.And the down-shifted Fermi level can cause a charge transfer resonance with laser energy, thus resulting in a large chemical enhancement. Indeed, the limit of detection (LOD) of RhB can be reduced to 10−7mol L−1after the treatment with O2plasma for graphene,greatly amplified compared with that of pristine graphene[33]. More importantly, substitutional doping is also a very promising way to tailor the electrical and chemical properties of graphene. Our group investigated a remarkable enhanced Raman scattering with an effective fluorescence quenching effect on Ndoped graphene (NG) substrate, mainly due to the higher density of states (DOS) and more suitable tailored energy gaps in NG than the pristine graphene(PG) (Fig. 3b(ii-iv)), thus facilitating charge transfer resonance in the system and resulting in a low detection level of 5×10−11mol L−1for RhB[14,25]. Furthermore, we found that Si-doped graphene reached an even better SERS performance than NG and PG for detecting different dyes, including CV, RhB and MB.Si-dopants can not only change the DOS and Fermi level of graphene like NG but also induce local curvature around the substitutional sites for much larger volume of silicon atom than carbon (Fig. 3b(v)),thus promoting the interaction between the probe molecules and graphene substrates[32].
In fact, it is not easy to obtain “true” graphene with just one layer, it is common to get bilayer and multilayer graphene sheets in most cases. Linget al.[36]found that graphene was a thickness-dependent SERS substrate and determined by concentration of molecular solutionviasolution soaking method. They demonstrated that the enhancement factor generally decreased with the increase of the layer number in most cases. However, it is interesting in Fig. 3c (i) and(ii) that bilayer graphene showed a stronger Raman intensity than monolayer graphene under low concentration solutions (4×10−5mol L−1) for CuPc detection,
whereas there was little difference between them in high concentration (4×10−4mol L−1). It can be explained thatp-doping is induced in the graphene with a downshifted Fermi level when the CuPc molecule attached the graphene and there was similar change in Fermi level positions of mono- and bilayer graphene at high concentration with similar energy gaps for charge transfer. They also found that monolayer graphene is more sensitive to dope than bilayer graphene, thus resulting in a larger decline of Fermi level at low concentration. So that, the Fermi level of the monolayer graphene will be much lower than the HOMO of the CuPc molecule (Fig. 3c (iii)), thus the charge transfer needs extra energy to emerge, while there is less reduction of the Fermi level in bilayer graphene and the Fermi level still keep higher than the HOMO level of the molecule, thus facilitating charge transfer without extra energy. Besides, the stacking configuration of bilayer graphene can also affect the enhancement effect for molecule sensing, it was indicated that rotated graphene could detect the MB under low concentrations about 10−6mol L−1whereas AB stacking seemed to be insensitive for MB molecule[37].In general, high quality and clean surface are beneficial for graphene to gain superior sp2carbon domain,which greatly dominates the fluorescence quenching effect[38]. With quantum confinement and edge effect,high-quality graphene quantum dots prepared by plasma-enhanced chemical vapor deposition (P-GQD)demonstrated a high enhancement for SERS of sensitivity down to 1×10−9mol L−1rhodamine with good fluorescence quenching and its enhancement factors were up to 437 (1 648 cm−1Rhodamine 6G (R6G)),about 7 times higher than that on CVD graphene(Fig. 4a). It is clarified by DFT calculation that the band gap increases with the decreasing of the diameter of the GQDs and there is a more sufficient charge transfer between the molecule and P-GQD than on perfect graphene with higher DOS, especially when the size is about 6.2 nm, leading to a larger polarizability and the higher Raman scattering cross-section for great enhancement effect[39].
Fig. 4 (a) The influence of size of graphene quantum dots on GERS. The calculated charge transfer integrals (I) of rhodamine 6G (R6G) with (i) pristine graphene and high-quality graphene quantum dots prepared by plasma-enhanced chemical vapor deposition (P-GQDs) of (ii) 2.2 nm and (iii) 6.2 nm sizes[39](reprinted with permission from Spring Nature). (b) Impact of molecule and laser energy. (i) Schematic illustration of the influence of molecule and laser on GERS (here M1–M4 represent four different molecules). Raman spectra of CuPc, zinc phthalocyanine (ZnPc), and copper(II)1,2,3,4,8,9,10,11,15,16,17,18,22,23,24,25-hexadecafluoro-29H,31H-phthalocyanine (F16CuPc) under (ii) 532 nm and (iii) 633 nm laser. (iv) Raman spectra of tetrathienophenazine(TTP), tris(4-carbazoyl-9-ylphenyl) amine (TCTA), and 2,2′,7,7′-tetra(N-phenyl-1-naphthyl-amine)-9,9′-spirobifluorene (sp2-NPB) on graphene[40] (reprinted with permission from American Chemical Society).
In essence, the enhancement of Raman intensity mainly relies on the relation between the Fermi level of the graphene and the HOMO or LUMO levels of the probe molecules, therefore it is also important to take the properties of the probe molecules into consideration for GERS performance. There are two key factors of molecules carried out to have influence on the final GERS effect: (1) the distinct alignments of HOMO and LUMO energy levels and (2) various molecular structures of various probe molecules. It can be explained by the time-dependent perturbation theory that the enhancement effect required the HOMO and LUMO levels to be suitable range with respect to the Fermi level of graphene, which determined the emergence of SERS. Besides, the similar symmetry of the molecules and graphene (D6h) is more likely to yield a better GERS performance. We can inferred from Fig. 4b(iv) that tetrathienophenazine (TTP) (D2h)exhibited about 3.4 times and 5.4 times higher GERS enhancement factor than that of tris(4-carbazoyl-9-ylphenyl) amine (TCTA) (C3) and 2,2′,7,7′-tetra(Nphenyl-1-naphthyl-amine)-9,9′-spirobifluorene (sp2-NPB) (S4), respectively, with the similar HOMO/LUMO energy arrangements[40]. In addition, due to the large delocalized π electron cloud of graphene, when the π orbitals in molecule couple with graphene at a certain orientation, there may be larger interfacial dipoles and electron transition relying on the π-π interaction, thus definitely facilitating stronger GERS intensity[41].
Moreover, laser energy will also have a remarkable influence on GERS. In our previous work, we found that different dye molecules had different resonant laser lines to achieve the best enhancement result, for instance, 514.5 nm laser line for CV and RhB and 647 nm laser line for MB[32]. And we believed that the strongest Raman signal could be achieved when the laser energy is close to the HOMO-LUMO gap of the molecule[14]. Similarly, Huanget al.[40]made comparisons between 633 nm and 532 nm laser energies on different probe molecules (Fig. 4b(i-iii)) and they demonstrated that strong GERS enhancement generated when the laser energy is close to the HOMOLUMO gap or the energy gap between the Fermi level of graphene and the HOMO or LUMO levels of the probe molecules due to higher Raman scattering efficiencies according to the perturbation theory and Fermi’s golden rule.
The successful exploration on graphene served as active-SERS substrates has opened up new avenues for molecular sensing on manifest 2D inorganic layered materials beyond graphene, including TMDCs, h-BN, BP, MXenes and so on, which will be discussed in the following sections.
Transition metal dichalcogenides, with sandwich structures, have been also demonstrated as attractive platforms for SERS due to the large specific surface area, tunable electron structure and extraordinary optical properties. In general, the transition metal atoms(M) always arrange between the two layers of chalcogen atoms (X), usually denoted as MX2, which could accelerate charge transfer process through M-X bonds and cause large interfacial polarization due to the difference in electronegativities of atoms[42–44]. Therefore,the enhancement of Raman signal can be attributed to the large interaction with probe molecules through not only charge transfer but also interfacial dipole-dipole interactions. For example, our group synthesized single crystal NbS2with atomically flat surface controllably by chemical vapor deposition (CVD) method and the as-grown NbS2demonstrated much higher sensitivity for probing trace than that of graphene[45].
Different TMDCs have been studied on SERS in the past years and most of them have been confirmed to be practicable substrates with impressive SERS performance. For example, Taoet al.[46]employed the 1T’-WTe2on R6G detection and the substrate exhibited a high sensitivity with the EF of 4.4×1010and the LOD of 4×10−15mol L−1, better than most of graphene-based substrates ascribed to the higher DOS and more favorable charge transfer with the probe molecule, which can be seen in Fig. 5a and b. Additionally, they also tested the sensing performance of 1T’-WTe2, the substrates exhibited ultra-high sensitivity as well with low LOD of 4×10−14mol L−1R6G and EF of 6.2×109. Likewise, 1T’-MoTe2films, preparedviaCVD possessed superior capacities of molecular detection for biosensing with a LOD of 10−9mol L−1and an enhancement factor of about 104onβ-sitosterol. It is noteworthy that the MoTe2films also demonstrated an outstanding linear correlation between the intensity (at the peak of 1 668 cm−1) and concentration logarithm with a coefficient of 0.998 8, showing a promising SERS substrate for quantitative biosensing[6].
Besides 1T’-phase, there are always three other main types: hexagonal phase (2H), trigonal phase (1T)and rhombohedral phase (3R)[47–49]. Different phases tend to have different electron structures, thus affecting SERS results. Yinet al.[21]investigated the difference between semiconducting 2H-MoX2and metallic 1T-MoX2(X= S, Se) and pointed out that the Raman enhancement effect could be increasedviaa phase transition from 2H- to 1T-phase (Fig. 5c). The energy levels can be represented as Fermi levels and valence bands (VB) and conduction bands (CB) in 1T- and 2H-phase, respectively, out of the different electronic properties. In virtue of the higher position of Fermi level than that of the top of VB (TVB), it is easier for electrons in 1T-phase to transfer to the HOMO level of the probe molecule (Fig. 5d). Besides, as for the case of the excited state, in which the electrons are probably excited from the HOMO level of the molecule to the CB/Fermi level of the substrate. As shown in Fig. 5e, smaller energy gap between Fermi level of 1T’-phase and HOMO level of the MB molecule contributes to more effective electron migration with lower excited energy than that of 2Hphase[50]. Hence, metallic TMDCs have tendency to obtain higher Raman intensity because of its more suitable energy levels. However, some novel semiconducting TMDCs still perform excellent enhancement effect. SnSe2, a semiconductor with a bulk bandgap of about 1.05 eV, was claimed to detect R6G with the ultralow concentration of 10−17mol L−1due to the matched energy level (−5.05 eV of VB and −4.0 eV of CB) with the HOMO/LUMO levels of the probe molecule to drive charge transfer, especially the unusual high VB level (Fig. 5f)[51,52].
Fig. 5 (a) The electron density difference isosurfaces for R6G on graphene and 1T’-WTe2. (b) DOS of 1T’-WTe2 and graphene [46] (reprinted with permission from American Chemical Society). (c) SERS spectra and (d) schematic illustration of charge transfer mechanism of CuPc on different substrates[21] (reprinted with permission from John Wiley and Sons). (e) Schematic illustration of energy level and charge transfer resonance process for SERS of MB on 1T’- and 2H MoTe2[50] (reprinted with permission from IOP). (f) Schematic illustration of energy level and charge transfer process for SERS of crystal violet (CV) on SnSe2[52] (reprinted with permission from John Wiley and Sons).
The bulk TMDCs generally have indirect band gap, which can be transformed into direct band gap when reduced to monolayer, thus most of TMDCs display thickness-dependent SERS performance like graphene[43]. For example, Menget al.[53]prepared WS2ranging from monolayer to bulk-layer by CVD.They found that the single layer WS2offered the strongest Raman intensity of R6G and the Raman enhancement effect decreased with the increase of the layer numbers due to the indirect relaxation process.Moreover, the metallic 2D NbS2monolayer synthesized by Songet al.[54]demonstrated a detection limit down to 10−14mol L−1, much lower than that of 1TMoS2and graphene, because the high DOS led to an increased intermolecular charge transfer probability. It was reported that the amount of the 2H phase, which was verified higher DOS than that of 3R phase, increased with the decrease of the layer number, hence NbS2with fewer layers exhibited larger Raman enhancement. Besides, ReS2always crystallizes in distorted octahedral (1T’) with in-plane anisotropic optoelectronic properties, its Raman enhancement effect is still sensitive to the thickness with the direct-indirect bandgap transformation and the single-layer ReS2showed a detection limit as low as 10−9mol L−1. Nevertheless, as shown in Fig. 6a, the fluorescence background signal will also increase with the decrease of the layer number, hindering its further applications[55].Interestingly, it was reported by Wanget al.[56]that the underlying substrate for ReS2films might play roles on fluorescence quenching effect. They found that monolayer ReS2on mica expressed obvious fluorescence quenching compared to ReS2on SiO2and quartz (Fig. 6b and c). Electrons tend to transfer from the SiO2substrate to ReS2, resulting in electron doping to restrict the electron transformation from LUMO level of the molecule to CB of ReS2, which is fundamental to the fluorescence quenching, while no electron doping exists on mica for its inert surface. Furthermore, due to the high flexibility, the ReS2films on mica showed a good robustness after bending for 1 000 times, which is beneficial for their applications in flexible devices. Our group elucidated the influence of thickness on Raman enhancement for NbSe2. Largearea NbSe2flakes ranging from multilayer to monolayer was synthesized controllably by ambient pressure CVD. Different from most of researches, the NbSe2with 6 layers was demonstrated a five orders of magnitude lower LOD of 5×10−16mol L−1than that of monolayer NbSe2(Fig. 6d and e). The DFT calculation indicates that the as-grown NbSe2flakes could contribute to strong interaction with R6G and the 6LNbSe2possessed the highest DOS value at the Fermi level, giving rise to the most efficient charge transfer process for SERS by chemical mechanism (Fig. 6f)[57].Indeed, it can be found that the electron structure,such as energy levels and DOS, is the main factors of the thickness-dependence phenomenon. Defect engineering is also a promising strategy to tailor electron structure of materials. Liuet al.[44]tuned the atomic ratio of WSe2monolayers by Au ion beam to obtain an optimized SERS substrate, which can be seen in Fig. 7a. And the atomic ratio (Se:W) can be tailored from 1.92 to 2.00 by different controlled dose/fluence(S1=1012ions cm−2,S2=1013ions cm−2,S3=1014ions cm−2) of ion beam (Fig. 7b). They found that the EF was increased to more than 40 times on WSe2monolayer with the atomic ratio of 1.96 than the pristine WSe2(Fig. 7c). According to fs optical pump-probe spectroscopy (Fig. 7d), the quantity of the exciton increased by about 9 times, which is beneficial to the enhancement effect. And it was confirmed by DFT that the total density of states (TDOS) increased with the decrease of atomic ratio, as shown in Fig. 7e.
Fig. 6 (a) SERS spectra of 10−6 mol L−1 R6G on different layer number ReS2 films[55] (reprinted with permission from John Wiley and Sons). (b-c) SERS spectra and normalized fluorescence of R6G on monolayer ReS2 with different underlying substrates[56] (reprinted with permission from Elsevier). SERS spectra of R6G (d) with 5×10−6 mol L−1 on NbSe2 with different layer numbers and (e) on 6L-NbSe2 with different R6G concentrations. (f) DOS of the outermost layer of 2H-NbSe2 with different layers in the A’B stacking mode[57] (reprinted with permission from Royal Society of Chemistry).
There are also other strategies to further improve the SERS performance for TMDCs. Chenet al.[58]fabricated vertically-aligned MoS2and WS2nanosheets on metal foils with mixed 1T and 2H phase by a hydrothermal reaction. The SERS measurements showed that the detection limits of both MoS2and WS2can attain 5×10−8mol L−1for R6G, which is attributed to the high content of 1T-phase for the charge transfer and abundant edges to interact with the molecule. Likewise, Majeeet al.[59]synthesized an interconnected network of vertically oriented multilayer MoS2nanosheets on SiO2by CVD and the MoS2nanosheets showed sensitive detection of 10−10mol L−1concentration of R6G and methyl orange (MO). The plenty of exposed edges enhanced light absorption and dye adsorption, thus further facilitating the interaction between MoS2and dyes and leading to a charge transfer resonance, resulting in the EF of 8.6×104for R6G at the peak of 1 362 cm−1and 5.8×104for MO at 1 180 cm−1. Moreover, Liet al.[60]compared different MoS2nanosheets with different interlayer distances,which were obtained by using the intercalated molecules with different sizes (L-cysteine, thiourea and glucose)viahydrothermal method. The MoS2nanosheets with 0.62 nm (termed as MoS2−0.62),0.87 nm (termed as MoS2−0.87) and 1.12 nm (termed as MoS2−1.12) interlayer distances could be synthesized by intercalating L-cysteine, thiourea and glucose,respectively. It can be found that the EF could be up to 5.31×105for MoS2−0.62, 1.44×105for MoS2−0.87 and 7.75×104for MoS2−1.12 on 4-mercaptobenzoic acid (4-MBA) probe molecule. And the SERS results of other two molecules, i.e. 4-aminothiophenol (4-ATP) and 4-mercaptopyridine (4-MPy) also exhibited that the MoS2with smaller interlayer distances possessed better SERS performance. According to the DFT calculation, the S-Mo bonds could serve as electron transfer paths between MoS2and the probe molecules and more efficient interfacial charge transfer process occurred from MoS2nanosheets with smaller interlayer distance, which accelerated moredz2orbitals of Mo elements to transform into the valence band maximum of the MoS2/probe molecule composite,thus enhancing the Raman signal remarkably.
Fig. 7 (a) Schematic diagram of Au ion irradiation. (b) Evolution of atomic ratio of WSe2 with the influence of incident ions. (c) The enhancement factor (EF)of SERS for CuPc on pristine (S0) and irradiated WSe2 monolayers by different controlled dose/fluence (S1=1012 ions cm−2, S2=1013 ions cm−2, S3=1014 ions cm−2)of ion beam. (d) The intensity of fs optical pump-probe spectroscopy at the wavelengths of 510 nm (A’), 596 nm (B) and 744 nm (A) at a time of 0 ps. (e) The total density of states (TDOS) of WSe2 with various atomic ratios[44] (reprinted with permission from John Wiley and Sons).
Besides graphene and TMDCs, some other 2D layered materials have also been demonstrated to be efficient SERS substrates with unique electron structures, such as BN, BP, MXenes.
h-BN is similar to graphene for the hexagonal structure and the bond length (0.144 nm of B―N and 0.142 nm of C―C) with white color so that they are called as “white graphene” sometimes. Meanwhile, it has many distinctive properties from graphene by replacing the carbon atoms in graphene with boron and nitrogen atoms[61]. The large difference in electronegativity between boron and nitrogen atoms can induce strong interfacial dipole-dipole interaction between the probe molecules and h-BN substrate, which is beneficial for SERS effect. Linget al.[22]studied Raman enhancement effect on graphene, h-BN and MoS2for CuPc detection (Fig. 8a), they pointed out that the enhancement mechanism on h-BN was strong dipole-dipole interaction to increase matrix elementvialocal symmetry-related perturbation but the charge transfer,which is different from graphene mainly by charge transfer and TMDCs by both charge transfer and weak interfacial dipole-dipole interaction. It can be explained that h-BN possesses insulating performance with a large band gap of about 5.97 eV[62], thus the electron DOS has less influence on electron transition probability for charge transfer mechanism. Therefore,h-BN is a thickness independent SERS substrate because of the similar dipole interaction for various layer numbers (Fig. 8b). Linget al.also found that h-BN tended to enhance lower frequency phonon modes of CuPc (like 682, 749, 1 142 and 1 185 cm−1) while higher frequency phonon modes (such as 1 342,1 452, 1 531 cm−1) were increased more on graphene than on h-BN (Fig. 8c)[22]. The most attractive property of h-BN is the high thermal stability, the single layer h-BN can stabilize at more than 800 ℃ in air,which definitely shows potential to be served as reusable substrates for SERS.
Fig. 8 (a) Schematic illustration of graphene, h-BN and MoS2 served as SERS substrates. SERS spectra of CuPc on (b) different substrates and (c) h-BN flakes with various thicknesses[22] (reprinted with permission from American Chemical Society). (d) Charge distribution of valence band for CuPc/black phosphorus(BP) (here AC denotes armchair direction). (e) SERS spectra of CuPc on SiO2/Si with BP[65] (reprinted with permission from American Chemical Society).Schematic illustration of charge transfer on (f) methylene blue (MB)/Ti3C2-Al(OH)4 and (g) MB/Ti3C2-OH/F[70] (reprinted with permission from American Chemical Society). (h) Schematic of the adsorption and intercalation of MB in the MXene nanosheets[71] (reprinted with permission from AIP).
Layered black phosphorus has aroused increasing interests in virtue of its high electron mobility,high biocompatibility and outstanding electronic properties[63]. It exhibits tunable bandgap dominated by thickness like graphene[64]. The most intriguing performance of BP is angle-dependent properties, the electron mobility is much higher along armchair direction than the zigzag direction owing to lower effective mass of electrons for faster charge carriers, leading to different Raman enhancement results with different orientation, as shown in Fig. 8e. When contacted with CuPc molecule, the charge near the Fermi level of BP redistributed to 1D chains along armchair direction (Fig. 8d). Besides, the photo-excited excitons in BP also tend to follow armchair direction. So that the strongest charge interaction appears only when the primary axis of the probe molecule follows armchair direction, thus leading to the largest matrix element of electron transition for the strongest Raman signal[65]. It was reported that inducing nano-void array on layered BP flake by low power focused laser irradiation could further improve the Raman enhancement (about 30%) to achieve a LOD of about 10 nmol L−1[66]. The in-plane ferroelectric properties of BP favor the enhanced local electric field near the nano-voids.
At present, MXenes, a new family member of 2D materials consisting of transition metal carbides, nitrides and carbonitrides with large specific surface area and excellent electrical conductivity, have been also declared to have chemical SERS effect. MXenes are always expressed as Mn+1XnTx(n=1, 2, 3), where M is transition metal, X denotes carbon or nitrogen and Txrefers to surface terminated functional group like ―F, ―OH and so on[67]. A simple paper supporting Ti2N MXene SERS substrate was reported a Raman enhancement factor of 1012for R6G detection,and the high enhancement of Raman signal may be attributed to the high electron density on the N atoms transferred from the Ti atoms[68]. Elumalaiet al.[69]discovered a MXenes-enhanced resonance Raman scattering on titanium carbide film for CV molecule detection. Because of the fitted energy levels of the sensing system, the laser energy is close to the energy gap between the HOMO level of CV and the Fermi level of the MXene for electron transition (~1.9 eV), and the charge transfer resonance resulted in a large EF of 3.42×109. Furthermore, the surface metal atoms on the MXenes often show high chemical reactivities to modify with various surface functionalities on surface during the preparation, which can be applied as a convenient way to tailor the surface structure of MXenes.Ti3C2sheets with Al(OH)4ligands (Ti3C2-Al(OH)4)and OH/F ligands (Ti3C2-OH/F) can be obtained by the Al-extraction reaction in acid and base medium,respectively[70]. It was characterized by scanning Kelvin probe microscopy that the charge distribution on the Ti3C2-Al(OH)4sheets was more uniform than that on Ti3C2-OH/F. In general, the OH ligands would take tight interaction with N atoms on the probe molecule for hydrogen bonding, whereas the F ligands tend to show large repelling force against the atoms with high electronegativities like N and S, which locate in most dye molecules (Fig. 8f and g). Therefore,dye molecules like MB could lie flat on the Ti3C2-Al(OH)4sheets with strong interaction, conducive to the subsequent charge transfer. Meanwhile, the oxygen species on the surface of Ti3C2-Al(OH)4can protect the samples from further oxidation. Integrating the above advantages, the Ti3C2-Al(OH)4sheets exhibit excellent SERS effect with the LOD of pmol L−1level and high stability in air. Importantly, same as graphene, MXenes were also demonstrated as thickness-dependent SERS substrates[71]. The higher position of Fermi level of Ti3C2TxMXene than that of LUMO level of MB molecule benefits for charge transfer and the relatively large interlayer space of Ti3C2TxMXene nanosheets caused by coproducts from etching process enables the intercalation of MB to interact with all individual MXene layers (Fig. 8h),manifesting the increase of Raman enhancement with the thickness ranging from 5 to 120 nm.
Each material has its unique advantages, on the contrary also has its own disadvantages, designing heterostructure or hybrid structure has thus been regarded as a feasible strategy to make up the weaknesses of the individual component due to the synergistic effect. Tanet al.[72]stacked monolayer WSe2and graphene to form different heterostructure with different stacking sequence as SERS substrates, including G/W, W/G, G/W/G/W and W/G/G/W(the material in the left is close to the probe molecules). The heterostructure G/W exhibited a higher Raman intensity than pristine graphene and WSe2for detecting CuPc. According to the density function theory (DFT)calculation, wave functions lied in both graphene and WSe2for the electronic tunneling, thus increasing the DOS of the graphene for SERS. Besides, the W/G heterostructure possessed a weaker SERS than that of G/W because the WSe2on the top led to lower electronic transition and weaker interfacial interaction.Critically, the first layer effect for CM on 2D materials relatively limits the further improvement of SERS performance, we can find in Fig. 9a-c that the G/W/G/W and W/G/G/W showed similar SERS effect to G/W and W/G, respectively. So, it is confined to tailor the electronic properties further by stacking types.Importantly, defect engineering is also a potential strategy to tune the properties of heterostructures. Seoet al.[73]synthesized a graphene/ReOxSyvertical heterostructure as ultrasensitive substrate. The oxygen,located in the lattice structure, could induce dipole moment on the surface for complementary resonances and dipole-dipole interactions with the probe molecules (Fig. 9d). The degree of defect could be adjusted by the temperature of sulfurization for ReO2during the process. Together with high DOS, elaborate gap between energy levels and the interlayer coupling effect between ReOxSyand graphene, the vertical ReOxSy/graphene heterostructure demonstrated a femtomolar level Raman enhancement effect as shown in Fig. 9e. Moreover, in virtue of the excellent mechanical flexibility of the vertical heterostructure,the SERS substrate can keep high sensitivity after 1 000 cycles of bending test.
Fig. 9 (a-c) SERS spectra of CuPc on different heterostucture substrates[72] (reprinted with permission from American Chemical Society). (d) Schematic of SERS effect on graphene/ReOxSy heterostructure and bending test. (e) SERS spectra of R6G on graphene/ReOxSy vertical heterostructure[73](reprinted with permission from American Chemical Society).
More lately, 2D material coupled with noble metal nanoparticles (NPs) has been constructed as an optimized strategy for SERS, such as AuNPs/WS2/graphene[74], AuNPs/MoS2/graphene[75], AuNPs/GaTe[76]and so on. It was demonstrated that metal NPs decorated graphene-family nanomaterial would show pM level LOD by both CM from strong interfacial polarization and effective charge transfer and EM from metallic LSPR[77]. Besides, thanks to the negatively charged surface groups on MXenes, it is easy to absorb Au nanorods with positive charge uniformly by electrostatic self-assembly (Fig. 10a) and the SERS substrate could lead to a low LOD of 10−10mol L−1thiram and 10−8mol L−1diquat, both meet requirements of the U.S. Environmental Protection Agency,expanding the application in food safety monitoring[4].Moreover, as shown in Fig. 10b, metal nanoparticles covered by h-BN layers could obtain a high stability in air with ultra-sensitive SERS detection[78,79], which can still work on 2D material layers[80]. Raniet al.[81]induced artificial edges in monolayer MoS2by lowpower focused laser-cutting for following deposition of AuNPs to obtain AuNPs/MoS2hybrid structure(Fig. 10c). The AuNPs tended to accumulate along with the edges and induced intensified plasmonic effects as hot spots, where a LOD of about 10−10mol L−1for RhB could be detected, better than that of each single material, which can be seen in Fig. 10d. In addition, miRNA, a kind of typical cancer biomarkers,may often be detected by probing DNA with generated Cyanine 5 (Cy5). Therefore, it is important to support enough DNA with high sensitivity of Cy5 for miRNA detection. Through synthesizing MXene/MoS2@AuNPs hybrid structure, a high SERS effect for miRNA detection was achieved by Liuet al.[82]. As shown in Fig. 10e, the flower-like MoS2vertically adhering on the MXene layers provided large area to support Au nanoparticles, of which plenty of probe DNA could be adsorbed by Au‒S bonds. The substrate exhibited a LOD of 10−9mol L−1of Cy5 with a EF of 4.8×108as benchmark of substrate characteristic Raman peaks (382 cm−1and 402 cm−1of MoS2and 611 cm−1of MXene). According to the DFT calculation, the gap could be a container for holes to sustain charge transfer between MXene and MoS2and the Fermi level of MoS2increased by MXene layers.Thus, both the redistributed Fermi level of the composite, which increased the electron transition probabilities due to the enlarged DOS, and the uniform distributed AuNPs “hot spots” contributed to Raman enhancement effect by both CM and EM. Especially, for some specific morphologies, TMDCs can also induce LSPRviaphotodoping, the nanostructured TMDCs may undergo photoinduced carrier doping to enhance the Raman signalviaEM as the free carriers (both electrons and holes as shown in Fig. 10f) are excited to the CB by laser, which may also be facilitated by the dipole-dipole interaction and charge transfer in the TMDC nanodomes/graphene vdW heterostructure(Fig. 10g). Since both CM (viadipole-dipole interaction and charge transfer) and EM (viaphotoexcited LSPR) contributed to SERS effect, the SERS effect of R6G on TMDC nanodomes/graphene vdW heterostructure could be down to 5×10−12mol L−1at the peak of 613 cm−1, which is 4-5 orders of magnitude higher than that of single Mo(W)S2(5×10−6mol L−1) and graphene substrate (5×10−7mol L−1)[83]. Furthermore,the intermixed WS2+MoS2nanodisks/graphene vdW heterostructure even exhibited a lower LOD of 5×10−13mol L−1at 613 cm−1peak of R6G by the superposition of two types of LSPR effect from different plasmonic nanodisks[84]. Additionally, strain engineering is also a feasible method for 2D materials to further improve the SERS performance and the outstanding mechanical properties of 2D materials are beneficial to induce high strains. For example, Chenet al.[85]synthesized a wrinkled graphene/Au nanoparticles hybrid platform with a low detection level of 10−9mol L−1for R6G. They found that the morphology of the wrinkled graphene and the gap distance of the Au nanoparticles could be well tuned by different strains. The highest Raman intensity of 612 cm−1could be obtained when the substrate was stretched by 10% tensile train. Moreover, the properties of 2D TMDCs can also be tailored by inducing strain[86,87].Hwanget al.[88]obtained multilayer MoS2nanoscrolls decorated with noble metal nanoparticles (Ag and Au NPs) with a high local strain, which demonstrated a high SERS enhancement factor of ~107. The local bending strain can not only facilitate the plasma resonance of noble metal nanoparticles but also promote the 2H to 1T phase transition of MoS2, indicating a powerful potential of strain engineering on SERS effect.
In this article, we reviewed a series of 2D inorganic layered material for SERS effect, including graphene and TMDCs, h-BN, BP and MXenes, etc(Table 1). The main SERS enhancement mechanisms consist of electromagnetic mechanism and chemical mechanism, of which the former is usually for rough metal surface with LSPR and the latter is mainly for 2D material-based substrates with charge transfer and dipole interaction. Besides, some key factors impacting the SERS effect for 2D materials are summarized,including the Fermi level, thickness, size, defect,stacking mode, orientation and so on. Based on the enhancement mechanisms, the factors can be sum up in three aspects: Fermi level, density of states and polarized interaction. Therefore, the modulation of SERS performance can be considered based on the following aspects: (1) tune the Fermi level to match with the HOMO or LUMO energy levels of the probe molecules as fixed laser energy. (2) increase the DOS of the substrates to gain large electron transition probability. (3) facilitate strong dipole-dipole interfacial interaction for large polarization of the probe molecules. Moreover, constructing heterostructure or hybrid structure has been reported to exhibit superior SERS effect due to the efficient charge transfer between interlayers and synergistic effect with different materials. Especially when coupled with noble metal, the SERS substrates can enhance Raman signal by coupling both EM and CM. As mentioned above, 2D materials have been revealed as promising SERS substrates due to their unique electrical and optical properties with various captivating advantages and achieved ultra-high sensitivity with femtomolar level detection or even lower. However, 2D materials served as SERS platforms still face some challenges due to the relatively few researches. There is a long way to go for their practical applications and some aspects need to be well addressed in the future as below:
(1) Deep understanding of CM. 2D inorganic materials as novel SERS substrates mainly by chemical mechanisms are less studied compared to the abundant studies on noble metal rough surface as SERS substrates since its discovery in 1974, which leads to less theoretical researches on chemical mechanism for SERS. So far, there are two main models for CM, i.e.charge transfer and dipole interaction, generally supported by DFT calculation with some parameters like bandgap and DOS, which however still stays in the ambiguously qualitative interpretation and has no specific sequence of factors and concrete expression like EM. The intricate interaction among platform, probe molecule and laser energy retards to clarify inherent principles of CM-based SERS effect. Last but not the least, the variety of 2D materials with various electron structures may increase the difficulty on mechanism study, but it can also create more new opportunities to make in-depth exploration in the future.
Table 1 SERS performances and related mechanisms of various 2D materials.
(2) Accuracy of quantitative detection. Though more and more 2D materials have demonstrated ultrahigh sensitivities with pmol L−1or even fmol L−1level detection, only a few studies test the performance quantitatively, among which few works show superior linearity with coefficient close to 0.999 between the peak intensity and the molecule concentration logarithm. In order to overcome the bottleneck in quantitative detection, it is of great importance to develop novel methods for robust quantified analysis technology with high precision, for instance, improve the uniformity of the as-prepared SERS substrate materials,including dispersion, size, thickness, position and even structures, so as to adsorb the probe molecules more uniformly. Critically, based on the thermodynamics and kinetics, combined with experimental and theoretical calculations, it is also necessary to reveal the adsorption process of molecules on two-dimensional material surfaces, thus guiding the design of SERS substrates for accurate quantitative detection.
(3) Exploring more probe molecule varieties.
At present, the probe molecules in most researches always tend to be dye molecules for 2D material-based SERS detection, such as R6G, MB, CuPc, CV and so on, which obviously limits wider application works in other fields like biology and medicine. The low-cost,biocompatibility and flexibility capabilities of 2D material are beneficial for disposable applications like food safety, biosensor and medical diagnosis with fast response and real-time monitoring. Thus, more works on various practical probe molecules need to be devoted on 2D material-based SERS effect for widespread applications in the future.
We expect that this review will give insights on advanced and in-depth exploration on high-performance SERS substrates in the future.
Acknowledgements
This work was supported by the National Natural Science Foundation of China (Grant No.51972191).