Nomenclature

Surface plasmon resonance (SPR) sensing methods can realize analyses of biomolecular interactions with high sensitivity, rapid response, and no labeling [1–3]. Optical fiber-based SPR effect can be regarded as the nonlinear multiple integral of the prism-based SPR effect at the incident angle and the incident plane. Optical fiber SPR sensors possess the advantages of anti-electromagnetic interference, corrosion resistance, small volume, and great flexibility, and they can easily form a sensor network through cascading or other manners. This allow us to achieve the real-time, online, remote, and multi-parameter detection of target analytes [4,5]. Therefore, the research on fiber SPR sensors has made significant progress in the field of biochemical sensing. However, on the one hand, the size of viruses, cells, proteins, nucleic acids, and some other biological macromolecules is in the order of micrometers or even millimeters and is much larger than the penetration depth of surface plasmon polaritons (SPPs), which is on a scale of hundreds of nanometers [6,7]. This indicates that the above substances cannot be detected sensitively via conventional SPR sensors. On the other hand, the full width at half-maximum (FWHM) of conventional SPR sensors is relatively large, which exceeds 100 nm, and will result in a low detection accuracy (DA). Furthermore, the detection repeatability and the anti-interference capability of conventional SPR sensors need to be improved. Therefore, performance improvement approaches for fiber SPR sensors have been widely studied and are mainly categorized into the following three branches.

Reasonable and flexible design of fiber substrates by means of cascading, polishing, and other methods can increase the evanescent field leakage and can introduce the in situ interference self-compensation function. This will promote the capability of sensors in detecting the variety and the intensity of external factors.

Adjusting intrinsic layers (metal layers) of fiber SPR sensors can modulate the original light field, which makes sensors suitable for different application scenarios. Attenuated total reflection (ATR)-based long range SPR (LRSPR) [8,9], coupled plasmon waveguide resonance (CPWR) [10,11], nearly guided wave SPR (NGWSPR) [11], and waveguide coupled SPR (WCSPR) [12] can be considered as optimization parameters to improve the performance of conventional SPR sensors from different aspects.

The modification of various-dimensionality nanomaterials with high refractive index (RI) on the sensor surfaces by physical sputtering or chemical bond coupling can cause large disturbance of the evanescent field [9,13], can strengthen the confined electric field intensity [11], and can promote the capability of sensors for detecting small changes in the ambient environment.

According to above discussions, an overview of performance improvement approaches for fiber SPR sensors and their sensing applications is presented. There have been some reviews on SPR, but they are different from our report in terms of research scope and focus. In 2011, Shalabney et al. [11] classified sensitivity-enhancement techniques focused on the prism coupling-type SPR sensors. Fiber SPR sensors can be regarded as one more dimension than prism-based SPR sensors due to the multiple incident angles of light and the curved incident surface. Therefore, the fiber SPR effect focused in our paper is obviously different from the reported prism SPR effect in Ref. [11], and more performance indicators are involved in our review paper.

Optical fiber SPR sensors based on commercial fibers, customized fiber structures, and metastructure platform have been outlined in Refs. [14–19], but they are more focused on the classification of fiber SPR sensors. Meanwhile, applications of fiber SPR sensors in various scenarios (e.g.,  environmental monitoring, food safety detection, and biological immunoassay) have also been involved in Refs. [15,16,20–23]. Both classification and application scenarios are obviously different from the performance improvement approaches discussed in our paper. In addition, the sensing mechanism and applications of specific types of fiber SPR sensors (e.g., D-type fiber SPR sensors, hybrid plasmonic fiber sensors, molecular imprinting improved fiber SPR sensors, and surface plasmon cavities) have been systematically reviewed in Refs. [24–30]. The focuses of these reviews are particular, which are different from the broader focuses of our paper. We hope that this review can provide an insightful overview of fiber SPR sensors and inspirations for the development of SPR sensors.

The disassembly of cylindrical surfaces in the optical fiber allows the SPR phenomenon at each point to be explained by a three-layer structure. As shown in Fig. 1(a), the layer configuration of a traditional SPR sensor consists of a substrate layer, an intrinsic layer, and a dielectric layer. The substrate can employ commercial fibers (e.g.,  multi-mode fiber (MMF) [32], photonic crystal fiber (PCF) [33], hollow-core fiber (HCF) [34]), customized fiber such as suspended-core fiber [35], and the fiber grating (e.g., tilted fiber Bragg grating (TFBG) [5], long period fiber grating (LPFG) [36]). The intrinsic layer utilizes materials, with negative real-part and positive imaginary-part in dielectric constants, such as mono/bimetallic film with high reflectivities and low absorption rates (e.g.,  Au [37], Ag [38], Cu [39], Al [40]). The dielectric layer refers to the analyte layer.

The total reflection will occur when the transverse magnetic (TM) polarized light is projected on the substrate-metal layer interface at an angle larger than the critical value. The resulted evanescent wave (EW) enters the metal layer, and its energy decreases exponentially. This excites SPPs to propagate along the upper surface of the metal layer [7,11]. When wave vector components of the EW and SPPs on the interface are equal, SPPs will be strongly coupled with the EW. At this point, most of the energy of the incident light is coupled into SPPs [14] while a resonance dip appears in the transmission spectrum, as shown in Fig. 1(b). This phenomenon is called SPR. The SPP mode, which is strongly coupled to the EW, is called the resonance SPP mode, i.e., SPR mode. The SPP mode that is weakly coupled to the evanescent wave is called the non-resonance SPP mode. The wavelength of the incident light, which corresponds to the occurrence of the strong coupling, is defined as the resonance wavelength (λres). When the RI of the dielectric on the metal layer surface changes, a new resonance will occur for the incident light at another angular frequency. This can be macroscopically manifested in the change of the resonance wavelength. According to the change of the resonance wavelength, we can measure the change in the average RI of the dielectric on the metal layer surface, which is caused by the change of a merging of parameters belonging to bulk/volume sensing and others belonging to surface sensing, to realize the detection of the analyte.

The performance improvement of fiber SPR sensors mainly is targeted at some indicators, and this section introduces the typical performance indicators of the sensor and their influencing factors. Since it is not convenient to adjust the incident angle at the fiber core-cladding interface, most fiber SPR sensors are designed based on the wavelength scanning. The depth of the resonance dip is determined by the energy coupling between the resonance SPP modes and the evanescent wave. The stronger the coupling is, the deeper the resonance dip will be. The FWHM of resonance dip is determined by the difference between the resonance SPP modes-evanescent wave coupling and the non-resonance SPP modes-evanescent wave coupling. The greater the difference is, the narrower the FWHM will be. Meanwhile, some researchers claim that the observed SPR curve band is the result of the convolution of the resonance dips excited by separate optical modes [41], and the Ohmic loss of metal and the radiative loss of SPPs on the parallel axis [in Fig. 1(c)] lead to the broadening of the resonance dip [11].

The sensitivity (Sλn) of the sensor is defined as the resonance wavelength shift (∂λres) caused by the unit average RI change (∂ns) surrounding the sensing region, as expressed in Eq. (1) [42]. Its magnitude mainly depends on the overlap integral of the SPP electric field intensity on the vertical axis, as shown in Fig. 1(c), in the analyte region [11]. The DA is defined as the reciprocal of the FWHM value in the resonance dip, as expressed in Eq. (2) [42]. Some research works also employ the depth-width ratio in the SPR curve to describe the DA [43]. Both definitions have similar properties. The narrower the FWHM is, the easier it is to find the resonance wavelength and the higher the DA is. The signal to noise ratio (SNR) is inversely proportional to the FWHM, as expressed in Eq. (3) [44]. The figure of merit (FOM) and the quality factor (QF) are comprehensively quantitative indicators for characterizing the performance of the sensor, as expressed in Eqs. (4) and (5) [45], respectively.

The limit of detection (LOD) of the sensor represents the minimum concentration or the quantity of the target components that can be detected from the test sample within a given degree of reliability. Equations (6)–(8) [46–48] display the different ways to evaluate LOD. Slightly different from LOD, the limit of quantification (LOQ) represents the lowest detectable analyte concentration near blank considering the standard deviation of calibration curve, as expressed in Eq. (9) [46]. In general, the kinetic adsorption curve, acquired from biomass detection by using the sensor, is obtained by fitting the Langmuir adsorption isotherm equation [49] or logistic regression equation [50]. The sensitivity (Sλn) utilized to calculate the LOD and the LOQ can be obtained via the first-order derivative or the linear fitting of the low-concentration data (near zero) in the adsorption curve [46,47,51]. The aforementioned performance indicators of SPR sensors are listed in Table 1.Table 1.

Parameter Indices to Evaluate the Performance of SPR Sensors

The use of different commercial optical fibers as the substrate to construct SPR sensors will obtain different sensing characteristics. PCF-based SPR sensors possess high sensitivity due to the high confinement loss of SPPs excited by a Gaussian-like leaky mode of the PCF element [15]. HCF-based SPR sensors can realize all-fiber distributed sensing, since the entire inner wall of the fiber can be coated with chemicals to form the sensing area [34]. Fiber grating-based SPR sensors can obtain a smaller FWHM and a higher SNR [36].

SPR is the electromagnetic resonance stimulated by the evanescent field leakage when the total reflection occurs at the substrate interface. Therefore, research works are carried out to increase the evanescent field leakage for enhancing the sensitivity by customizing the fiber structure through micro/nano processing (e.g.,  fused tapering [52], side polishing [53], bending into U-shape [54]). In 2014, Cennamo et al. [55] fabricated a tapered plastic fiber SPR sensor, of which the specific sensing structure is shown in Fig. 2(a). When the light enters the tapered gradient region, it will be diffused into the cladding due to the reduction of the core diameter. At this point, fundamental modes in the core become high-order modes, and the light in the cladding is more likely to excite the evanescent wave on the metal layer [14]. The resonance spectra obtained in such a sensor to detect nicotine are shown in Figs. 2(b) and 2(c). The sensitivity and the LOD of the sensor achieved 1.3×104nm/M and 1.86×10−4M (1 M = 1 mol/L), respectively.

When the cladding and even the core of the fiber on one side are polished away, the evanescent field energy will leak from the polished area. This allows the side-polished fiber (SPF)-based SPR sensor to have a higher sensitivity [53]. In 2009, Tsai et al. [58] proposed a TM mode field coupling-based multi-step SPF-SPR sensor. The polished area consisted of three Au layer coated areas, leading to a two fold sensitivity enhancement, compared to the sensor with a single Au layer coated area.

The signal spectral response of the fiber SPR sensor has a wide wavelength range; and thus, multi-parameter sensing can be realized by capturing different central wavelengths in different bands [59,60]. The implementations are mainly categorized into the following three branches. First, multiple detection lanes are constructed based on different fiber substrates [61–65]. In 2021, Xiao et al. [61] developed a fiber SPR sensor with three detection lanes through coating Au-graphene composite layer on the surface of the side-polished PCF and filling two air holes with magnetic fluid and polydimethylsiloxane (PDMS), respectively, as shown in Fig. 3(a). Since the resonance conditions in the measurement of the three lanes were satisfied in different bands, the loss spectrum of the sensor contained three peaks, which were used for the measurement of magnetic field, RI, and temperature, respectively, as shown in Fig. 3(b). Multi-parameter detection can also be realized via constructing multiple detection lanes based on different mechanisms, e.g.,  localized SPR (LSPR) and interferometric sensing [68], as well as, birefringence phenomenon [69].

Second, double detection lanes are constructed in the same sensing area [66,70]. In 2019, Alonso-Murias et al. [66] achieved three-parameter measurement by coating PDMS onto part of the Au layer covered area in the SPR fiber tip sensor, as shown in Fig. 3(c). The segment coated with PDMS was used for the measurement of temperature, and the bare part was used for the measurement of RI and liquid level. The response signal in the liquid level measurement was reflected based on the intensity modulation, and the other two were expressed based on the wavelength modulation, as shown in Fig. 3(d).

Thirdly, multi-parameter measurement is carried out based on the sensitive characteristics of dielectric materials [67,71]. In 2020, Wang et al. [67] developed a fiber LRSPR sensor based on the dielectric buffer layer with temperature-sensitive photoabsorption characteristic, as shown in Fig. 3(e), where the dual-parameter measurement of RI and temperature was realized. The buffer layer had an absorption effect on light near 350 nm, and the absorption intensity decreased with the increase of the temperature. Therefore, the transmission spectrum of the sensor contained a characteristic absorption peak and a long-range resonance dip, as shown in Fig. 3(f).

The optical fiber sensors used for multi-parameter measurement can meet the needs of different complex scenarios, which is of great significance to application.

Multi-parameter measurement is generally used to solve the cross-influence among parameters. In 2019, Zhang et al. [56] developed an SPF-SPR sensor, and the sensing structure is shown in Fig. 2(d). A segment of LPFG was embedded in the SPF to realize the self-compensation for temperature. The produced sensor was applied to detect the concentration and the temperature of the glucose, and the obtained transmission spectra are shown in Figs. 2(e) and 2(f). The relationship between the wavelength shift (Δλres) in the SPR resonance dip and the change in the concentration (ΔC) and the temperature (ΔT) of the glucose is given as follows: Δλres=KCa·ΔC+KTa·ΔT.(10)

The relationship between the wavelength shift in the optical grating resonance peak (ΔλLPFG) and the change in the concentration (ΔC) and the temperature (ΔT) of the glucose is given as follows: ΔλLPFG=KCb·ΔC+KTb·ΔT,(11)where KCa and KTa are the sensitivity coefficients of the resonance dip for detecting the concentration and the temperature of the glucose, respectively. KCb=0 and KTb are the sensitivity coefficients of the resonance peak for detecting the concentration and the temperature of the glucose, respectively. Combining Eqs. (10) and (11), the following relation matrix can be obtained: [ΔCΔT]=[KCaKTa0KTb]−1[ΔλresΔλLPFG].(12)

According to Eq. (12), the resonance wavelength shift caused by the change in the concentration and the temperature of the glucose can be calculated separately. Meanwhile, the SPR sensing unit and the LPFG temperature compensation unit were set close to each other, and thus the sensor could realize the in situ temperature self-compensation in the detection of the glucose concentration.

Similarly, Weng et al. [72] implemented a double-SPF-SPR sensor, which contained two polished areas in 2016. The temperature and the RI sensitivities in the X-pol loss spectrum were 0.1371 nm/°C and 561.4286 nm/RIU, and the temperature and the RI sensitivities in the Y-pol loss spectrum were 0.04736 nm/°C and 457.1429 nm/RIU, respectively. The sensor theoretically eliminated the measurement errors caused by the fluctuation of the ambient temperature in the measurement of RI.

When the fiber SPR sensor is employed to detect the biomass based on the antigen-antibody specific binding, the resonance wavelength shift caused by the interaction between the antigen and the non-specific binding site on the sensor surface will affect the detection reliability. In 2021, Wang et al. [57] realized a U-shaped fiber SPR sensor, as shown in Fig. 2(g). When the light passed through the bending part, part of the light would no longer meet the condition of the total reflection and would leak into the cladding due to the bending of the fiber core. Meanwhile, the cladding was also bent. The light that met the total reflection condition in the cladding can stimulate the SPR, and the light that did not meet the condition would leak to the ambient environment and would strengthen the evanescent field [14]. The sensing region is divided into the antibody modified detection lane constructed by the MMF and the unmodified reference lane constructed by the SMF. The sensor was applied for the detection of RI and human immunoglobulin G (IgG) solution doped with bovine serum albumin (BSA). Two resonance dips appeared in the transmission spectrum, as shown in Fig. 2(h), corresponding to the two lanes. The influence of the non-specific adsorption of BSA on the detection branch can be calculated according to the wavelength shift of the reference lane, as expressed in Eq. (13): Δλd=Δλr×Sd/Sr,(13)where Δλd and Δλr are wavelength shifts of the detection lane and the reference lane, respectively, and Sd and Sr are RI sensitivities of the detection lane and the reference lane, respectively. Therefore, the introduction of the reference lane can theoretically eliminate the measurement errors caused by the non-specific binding of BSA in the IgG detection in the detection lane. Furthermore, similar to the terminal reflective sensing structure [73,74], the U-shaped fiber SPR sensor allows an easy operation since it can be directly inserted into the analyte solution.

Similarly, Wang et al. [75] developed a cascaded dual-lane optical fiber SPR biosensor, consisting of an Au layer coated detection lane and an Ag layer coated reference lane, in 2020. The resonance dip positions corresponding to the two lanes were different, and the sensor can theoretically eliminate the measurement errors caused by the non-specific binding and the temperature cross-sensitivity. When the sensor, of which the transmission spectrum contains multiple resonance dips, is exposed to the external mechanical interference, the entire resonance spectrum will exhibit a shift, but the relative wavelength difference between the resonance dips remains unchanged. In this case, the sensor provides an in situ sensing self-reference function, and this leads to a higher detection reliability [13,76].

LRSPR refers to an electromagnetic resonance mode stimulated by adding a lossless dielectric buffer layer (e.g.,  Cytop [9], Teflon [77], MgF2 [78], nano-SiO2 film [79]), with an RI similar to that of the intended buffer layer but lower than that of the substrate, between the substrate and the metal layer. The layer configuration is shown in Fig. 4(a). Compared with the traditional SPR sensor, in which there only exists a single row of SPPs, the SPPs in the LRSPR sensor exist on both sides of the metal layer. The constructive interference between the two rows of SPPs forms a symmetrical mode, i.e., long range SPPs (LRSPPs) [43]. Its energy is mainly focused in the dielectric on both sides of the metal layer, which shows a property of low damping [82]. The penetration depth of the LRSPP’s electromagnetic energy in the dielectric layer and the propagation distance at the interface between the dielectric layer and the metal layer are both an order of magnitude higher than that of traditional SPPs [6]. Therefore, the LRSPR sensor exhibits more advantages in detecting biological macromolecules. In 2019, Zhang et al. [78] proposed an LRSPR sensor for the BSA detection by successively evaporating the MgF2 layer and the Ag layer on the side-polished large-core fiber surface. The specific sensing structure and experimental results are shown in Figs. 4(b) and 4(c), respectively. The FWHM of the sensor was as low as 47.8 nm, and the sensitivity reached 0.765 nm/(mg/mL). Although the LRSPR sensor has a narrower FWHM, the sensitivity of the sensor significantly relies on the symmetrical configuration [43]. This would restrict its detection range, to a certain extent, in the field of biochemical detection.

The CPWR sensor incorporates a thick (typically 500 nm [11]) dielectric layer with high-complex permittivity (e.g.,  ITO [83], ZnO [80], TiO2 [80]) as the waveguide layer over the traditional SPR sensor surface, as shown in Fig. 4(d). When the plane wave is incident on the waveguide layer in the TM or the transverse electric (TE) mode, the light will be both reflected and refracted at the waveguide layer interface. The resulted multiple rays will interfere due to the existence of the optical path difference, to form the waveguide resonance mode and to generate the waveguide mode resonance [43]. In 2018, Liu et al. [80] established a simulation model of optical fiber CPWR sensor, and the specific sensing structure is shown in Fig. 4(e). The output spectrum corresponding to the incident light with different polarization states and the mode field distributions of the CPWR mode as well as the SPR mode were simulated, as shown in Fig. 4(f). Simulation results indicated that the FWHM of the interference valley according to the sensing model, to detect the RI, was as low as 26.37 nm, and the sensitivity achieved 1360.61 nm/RIU. Due to the large distance between the analyte layer and the SPPs on the metal surface, the CPWR sensor shows a lower sensitivity than the SPR sensor, but it has a narrower FWHM leading to higher SNR and DA [11].

Similar to CPWR, NGWSPR can be stimulated by coating a thin (typically 10 nm [11]) dielectric layer with high real part of the dielectric constant (e.g.,  TiO2 [84,85], ZrO2 [11,85], PbS [86], SiO2 [85]) on the metal layer surface. Since the thickness of the dielectric layer is insufficient to support a guided mode, this leads to the spread of SPPs along the dielectric layer, where part of the evanescent field energy is coupled to the covered materials (to be measured) [11]. In 2020, Prakashan et al. [85] coated the core-shell nanoparticles doped SiO2-TiO2-ZrO2 ternary sol on the fiber SPR sensor, which avoided the oxidation of the Ag layer and achieved the detection of mercury with a low LOD of 10 nM.

WCSPR refers to an electromagnetic resonance mode generated by coating an additional metal layer on the surface of the waveguide layer, based on CPWR. The layer configuration is shown in Fig. 4(g). When the light ray is incident at a small angle, the SPPs propagating along the interface between the lower metal layer and the waveguide layer will be excited by the TM polarized light. With the increase of the incident angle, the SPPs on the interface between the upper metal layer and the waveguide layer will also be excited [43]. Due to the coupling between the SPR and the waveguide resonance modes generated by the interference between multiple rays, narrow resonance dips and interference valleys that do not shift with the change of the external RI appear in the transmission spectrum [81]. In 2019, Ma et al. [81] developed a WCSPR sensor by sputtering Au-ZnO-Au layers successively on fiber core surface, and the specific sensing structure is shown in Fig. 4(h). Simulation results in Fig. 4(i) indicated that the sensor with optimal structural parameters realized the RI measurement with a sensitivity range of 2198–4732 nm/RIU, which was 7%–131% higher than that of the traditional SPR sensor. Compared with traditional SPR, the WCSPR sensor has slightly higher sensitivity and narrower FWHM, which means a higher DA. WCSPR can be regarded as a compromise between the traditional SPR and the CPWR [43].

Due to unique sensing characteristics of ATR-based SPR modes, corresponding sensors are suitable for different applications, e.g.,  stimulating stronger surface enhanced Raman scattering (SERS) signal for pH sensing [87] and obtaining biological information regarding the cellular micromotion [88] and the DNA hybridization [89]. Different types of ATR-based SPR modes are summarized in Table 2 for a comparison.Table 2.

Comparison of Different ATR-Based SPR Modes

Zero-dimensional nanomaterials are confined in the nanoscale range (1–100 nm) in all three directions, and their electrons cannot move freely, e.g.,  nanospheres [90] and nanoparticles [91]. In 2020, Wang et al. [75] implemented an SPP coupled optical fiber biosensor, as shown in Fig. 5(a), for the human IgG detection. When the antigen modified on the Au nanoparticles (AuNPs) was combined with the antibody modified on the sensor surface, the near-field electron coupling, between the LSPR arising from the AuNPs and the SPR arising from the Au layer, would greatly amplify the SPR response signal. Experimental results in Fig. 5(b) indicated that the LOD of the sensor achieved as low as 0.015 μg/mL. In 2016, Shi et al. [92] presented a rapid method for the fabrication of the fiber SPR sensor through the polydopamine (PDA)-accelerated electroless plating (ELP), as shown in Fig. 5(c). Spherical Au seeds were at first adsorbed onto the PDA functionalized fiber surface by amino and imino intermediates generated during the polymerization, and then an Au layer was deposited. The developed sensor showed the advantages of fast fabrication and good stability, and realized the high-sensitivity detection for human IgG, as shown in Fig. 5(d).

Due to the assistance of the coupling between the inner and the outer walls, the hollow gold nanoparticles (HGNPs) have a stronger plasmonic field compared to solid GNPs [93]. In 2021, Cheng et al. [51] implemented an HGNPs modified fiber LRSPR sensor as shown in Fig. 5(e). Based on the Schiff base and Michael addition reaction between dopamine and amino or thiol-terminated biomolecules [94], the sensor utilized PDA as a coupling agent to immobilize antibodies for detecting the human IgG. The achieved LOD was 0.20 μg/mL while the FWHM of the SPR curve was wider due to the increased radiation damping caused by the light scattering of the HGNPs, as shown in Fig. 5(f). In addition, the utilization of liposome particles [95], magnetic nanoparticles (with super paramagnetic property) [91], polymer particles [96], quantum dots [97], and the further combination of them and the sandwich immunoassay, to increase the molecular weight on the sensor surface [93], can greatly improve the performance of SPR sensors.

Based on the above discussion, typical application examples of zero-dimensional nanomaterials applied in improving the SPR sensing performance are listed in Table 3.Table 3.

Application Examples of Zero-Dimensional Nanomaterials in SPR Sensing

One-dimensional nanomaterials are confined in the nanoscale range in two directions and are extended to the macroscale range in the last direction. Their electrons can only move freely in one direction, e.g.,  nanotubes [38], nanorods [104], nanowires [105], and nanofibers [106]. Carbon nanotubes (CNTs) have been widely applied in the fields of biochemical sensing, nano-optics, and photothermal conversion owing to their extraordinary chemical, electrical, and optical characteristics. In 2021, Pathak et al. [38] embedded palladium nanoparticles (PdNPs) in polypyrrole (ppy) coated MWCNTs and established a fiber optic SPR sensing system, as shown in Fig. 6(a), for the hydrazine detection. CNTs have a high RI, which makes it easier for the evanescent field energy to be coupled into the outer space of the metal layer. This thereby enhanced the light-matter interaction and promoted the perception capability of the sensor to the ambient environment changes [107]. The electronegativity of the CNT’s surface promoted the transfer of electrons from CNTs to the Au layer surface and thereby strengthened the electric field intensity on the surface of the sensing region [38]. The superposition of the electric field stimulated by adjacent CNTs further strengthened the confined electric field on the sensor surface. Moreover, the CNT’s modified sensor surface can absorb more molecules than the bare Au layer due to the large specific surface area and the excellent biocompatibility of CNTs [108]. Therefore, experimental results in Fig. 6(b) indicated that the LOD was as low as 20 nM.

The modification of Au nanorods on the surface of SPR sensor can effectively enhance the sensitivity of the sensor because of the transverse and the longitudinal plasmon electric field coupling between nanorods and the Au layer [91]. In 2020, Xia et al. [104] developed a multilayer Au nanorod modified PCF-SPR sensor, as shown in Fig. 6(c). The sensitivity of the sensor in detecting the RI, for the range of low-concentration biological solutions, can achieve up to 25,642.65 nm/RIU. Experimental results in the detection of human IgG solutions with different concentrations are shown in Fig. 6(d), which indicated that the LOD was as low as 4.6 ng/mL. If the Au nanorods form an array structure and the distance between nanorods is smaller than the wavelength, the formed metamaterials layer supports a guided mode with the field distribution decided by the plasma-mediated interaction between nanorods [109]. Based on the overlap between the detection field, bioactive substances in metamaterial layer, and the confined plasma-mediated energy, the sensitivity of the sensor will be greatly enhanced [109]. Similarly, Kant et al. [106] employed an electrospinning setup for the synthesis of xanthine oxidase (XO) entrapped Ta2O5 nanofibers and modified them on the surface of the fiber SPR sensor, as shown in Fig. 6(e) for the xanthine detection. Experimental results in Fig. 6(f) indicated that the sensor had a high sensitivity and the LOD was as low as 12.70 nM.

Based on the above discussion, typical application examples of one-dimensional nanomaterials applied in improving the SPR sensing performance are listed in Table 4.Table 4.

Application Examples of One-Dimensional Nanomaterials in SPR Sensing

Two-dimensional nanomaterials are confined in the nanoscale range in one direction and are expanded to the macroscale range in other two directions. Their electrons can only move freely in such two directions. Two-dimensional nanomaterials popularly used for improving the performance of SPR sensors are listed in Table 5. Graphene possesses lower energy losses (e.g.,  Ohmic loss and radiative loss), high electron transport mobility, and high surface-to-volume ratio, and it can accommodate more organic and biological molecules because of the π stacking interaction between the carbon-based ring structure of molecules and hexagonal cells of graphene [122]. The position of each graphene atom can be involved in the interaction between biomolecules, which means that graphene may eventually realize the single-molecule detection. GO has a hexagonal carbon structure similar to graphene and also contains the oxygen functional group such as hydroxyl, alkoxy, and carboxyl. This results in its higher solubility and easier surface functionalization for various types of biological receptors [122]. Phosphorene has also been applied in photonic sensor research works due to its high carrier mobility, strong carrier confinement, tunable direct bandgap, and anisotropic electron and photon characteristics [123,124]. The high carrier mobility and light absorption of TMDCs enable the effective use of light energy in the electron transmission. This leads to a strong coupling on the interface, thereby will enhance the surface energy density of the sensing layer and will amplify the response of the sensor to the RI disturbance in the sensitive layer [125].Table 5.

Popular Two-Dimensional Materials for SPR Sensor Performance Improvement

Different two-dimensional layered materials can be vertically stacked together and integrated to form van der Waals heterostructures and superlattices. This can manipulate optical and electronic characteristics of their interlayer interactions, to provide technical support for the development of next-generation devices [126]. In 2019, Rahman et al. [127] theoretically designed a phosphorene-graphene/TMDC heterostructure-based fiber SPR biosensor, as shown in Fig. 7(a). BP improved the sensing performance by amplifying the interlayer coupling of van der Waals forces [130]. The stacking of the layered structure allows the thickness to be precisely controlled and the avoidance of the metal oxidation. Meanwhile, the sensitivity of the sensor can be controlled by the number of layers in the two-dimensional nonomaterials. The proposed biosensor also exhibited vast application potential in the DNA detection, as shown in Fig. 7(b).

MXene can realize the attractive biosensing due to its chemical stability, fully functionalized surface termination, high metal conductivity, tunable direct bandgap, small work function, large adsorption energy, and hydrophilic biocompatible surface [123,126,131]. In 2020, Chen et al. [128] developed a Ti3C2Tx MXene improved fiber SPR sensor [in Fig. 7(c)] for the RI detection. Experimental results in Fig. 7(d) demonstrated that, for the first time, characteristics of Ti3C2Tx MXene could enhance the sensitivity of fiber biosensors. The modification of some perovskite materials with high RI and low dielectric loss on the SPR sensor surface to stimulate NGWSPR can also strengthen the intensity of the evanescent electromagnetic field on the sensing layer surface [129,132,133]. In 2020, Xia et al. [129] successively coated Au layer and ferroelectric crystal-BaTiO3 on the polymer-tipped optical fiber (PTOF), as shown in Fig. 7(e), and experimental results in Fig. 7(f) indicated the improvement of the performance of the SPR sensor.

Based on the above discussion, typical application examples of two-dimensional nanomaterials applied in improving the SPR sensing performance are listed in Table 6.Table 6.

Application Examples of Two-Dimensional Nanomaterials in SPR Sensing

Three-dimensional nanomaterials represent composites composed of one or more basic structural units from zero-dimensional, one-dimensional, and two-dimensional materials, e.g.,  three-dimensional superlattice structures [146,147] and nanostructured metal architectures [148,149].

In 2020, Li et al. [150] implemented a U-shaped MMF LSPR biosensor based on the three-dimensional hyperbolic super-material structure of multilayer graphene coated Au nanoparticles. The fabrication procedure of the biosensor is shown in Fig. 8(a). In Fig. 8(b), the multilayered structure formed by the repeated synthetization of the nanoparticle layer and the graphene enhanced the vertical charge density oscillation and transformed the plasma oscillation mode from a planar structure to a three-dimensional structure. This led to a strong LSPR effect. It was found in Fig. 8(c) that this biosensor displayed excellent performance in the DNA hybridization kinetics measurement. In 2018, Daems et al. [151] designed three types of three-dimensional DNA origamis as bioreceptor carriers to functionalize the fiber SPR sensor for the detection of thrombin, as shown in Fig. 8(d). The produced biosensor showed the better performance in the absence of backfilling and a wider linear range. Experimental results in Fig. 8(e) indicated that the LOD was as low as 6.1 nM.

Based on the quantum size effect, the surface effect, and the dielectric confinement effect, nanomaterials with new functions are expected if there are further breakthroughs in the structure, the composition, the arrangement, and the size of three-dimensional nanomaterials.

Compared with the transmission-type optical fiber SPR sensor [152] shown in Fig. 9(a), the end-face reflective optical fiber SPR sensor can be inserted into a small piece of sample or cramped space to directly acquire the signal in the dip-and-read way. Combined with mature fiber communication technology, this can achieve the high-efficiency, low-damage, and invasive measurement. At present, there are mainly three types of end-face reflective fiber SPR sensing probes [14], as shown in Fig. 9(a). On the cylindrical surface at the end of fiber, the cladding is removed, and a metal layer is coated to stimulate the SPR. Meanwhile, the end-face of fiber is also coated with the metal to serve as a mirror. Moreover, when the light is reflected from the end-face and enters the fiber again, the optical path is doubled, which enhances the plasmon resonance effect. In 2019, Kong et al. [153] fabricated a SiO2-microtip probe on the SMF end-face based on arc discharge technology. The probe surface was coated with an Au layer and was connected to the laboratorial testbed in Fig. 9(b) to realize the biochemical detection at a single-cell level. A bio-probe was produced in Fig. 9(c) by immobilizing the antibody on the Au layer surface via a nano-polyelectrolyte film of chitosan/polysodium styrene sulfonate (Cs/PSS), and it was utilized for the human IgG detection. Experimental results in Figs. 9(d) and 9(e) indicated that the sensitivity and the LOD of the bio-probe reached 0.19154 nm/(ng/mL) and 0.1 ng/mL, respectively.

A highly ordered plasmonic nanoarrays, produced by self-assembly technology [154], focused ion/electron beam lithography [155], and two-photon laser direct writing [37], can stimulate various coupling optical effects such as Fano resonance [156], guided mode resonance [157], surface lattice resonance (SLR) [158], and SERS [154]. The combination of above effects and the SPR sensing can further strengthen the light-matter interaction, can reduce the radiative loss, and can thus improve the sensor performance. The periodic nanoarray also provides tunable structure parameters, wider spectral range, and higher degree of freedom. The SPR sensor based on the nanoarray has higher sensitivity and FOM, which expands the application scenarios of SPR. Furthermore, it is compatible with existing imaging equipment, microfluidic chips, and photonic circuits so that the fabricated devices can be highly integrated and miniaturized. Based on the above advantages [157], nanoarray structures have attracted more and more attention.

In 2016, He et al. [159] implemented an SPR device on the SMF end-face, as shown in Fig. 10(a). The end-face of the fiber was first coated with a 50 nm Au layer, and then the Au layer was engraved with a square nanoslot array as illustrated in the shadow area on the Au layer. The first-order spatial Fourier component of the nanoslot array was grating coupled with the fiber-guided wave. The second-order spatial Fourier component restricted the SPP transmission and can thereby form a plasmonic crystal cavity. Such a plasmonic crystal cavity on the SMF end-face was utilized to detect the RI, and experimental results in Fig. 10(b) manifested an FOM of 68RIU−1. Based on the coupling between the SPR within a dense metal layer and multiple resonance modes stimulated locally within anisotropically shaped nanostructures [160], the plasmonic nanoarray can also be utilized to enhance the sensitivity of the traditional reflective fiber SPR sensor. In 2017, Antohe et al. [160] deposited nanotriangular arrays on the Au layer coated reflective fiber end-face, as shown in Fig. 10(c), and achieved an increment of 25% in the sensitivity according to Fig. 10(d).

In 2018, Wang et al. [161] lithographed nanotrimer arrays at the fiber end-face, as shown in Fig. 10(e), which was, for the first time, the excitation of the SLR phenomenon on the optical fiber sensor. SLR originates from the coupling of the diffracted light propagating along the array plane and the localized surface plasmons [157]. It was found that the FWHM of the SLR-based sensor, shown in Fig. 10(f), was 89.23% lower than that of the LSPR-based sensor, shown in Fig. 10(g). The FOM was increased by approximately 10 times [158].

Lab on tip based probe devices provide high-precision dynamic tuning of plasmonic antenna and high-resolution detection through the strong interaction between plasmonic hotspots and samples. The test objects include the molecular adsorption or deformation [162], molecular vibration spectrum [163], quantum tunneling [164,165], etc.

As shown in Figs. 11(a) and 11(b), the coupling between the tapered fiber (or dielectric waveguide) and metal nanowires (or SPP waveguide) can greatly improve the light wave coupling and sub-wavelength focusing efficiency of the near-field optical microscope probe. This realized the scanning Raman spectroscopy imaging with a spatial resolution of 1 nm [166,169]. As shown in Fig. 11(c), it was observed by Danckwerts et al. that a sharp increase in the four-wave-mixing produced signal would occur when the Au nanoparticle-fiber probe approached another Au nanoparticle [167]. Due to the quantum tunneling effect, a sudden change of the increase trend occurred when the gap became very small. The Au nanoparticle-fiber probe was also employed to approach a single molecule. As shown in Fig. 11(d), the molecular fluorescence experienced the process of from enhancement to quenching when the probe approached a single molecule [168]. This indicated that the dynamically tuned plasmonic antenna was more conducive to the characterization of the light-matter interaction in the plasmonic hotspot.

Although the fiber SPR sensors can realize the detection of physical and biochemical parameters, there is still some space to improve fiber SPR sensors in terms of accuracy, cross-sensitivities, selectivity, costs, and reliability. In the future research, it will be promising to involve microelectromechanical systems [170,171] for improving the stability of the device and to employ liquid metals [172] for developing flexible optical devices. The integrated and miniaturized design of fiber SPR sensors can utilize the two-photon three-dimensional micro/nano film printing to achieve the accurate, real-time, and on-site detection [171,173]. Terahertz biosensors based on metamaterials can also be vigorously developed to achieve the non-invasive detection for biomarkers or biochemical indicators in living organisms [156,173]. According to the sensing characteristics of fiber SPR sensors modified by different nanomaterials, nanomaterials can be reversely designed at the atomic level to modify sensors to obtain desired sensing characteristics, e.g.,  to locate the resonance valley in the optical band. In order to improve the performance of fiber SPR sensors from different aspects, various approaches can be jointly applied in the implementation of sensors. For example, the modification of nanomaterials on the surface of the LRSPR sensor with improved metal layer quality can enhance the sensitivity of the sensor while ensuring the DA [174,175].

This review work summarized current major approaches for improving the performance of fiber SPR sensing and the applications of the manufactured sensors mainly in the field of biochemical detection. The sensing principle and the performance indicators including determinants of SPR sensors were demonstrated to clarify the research direction of the performance improvement. The application examples and the internal mechanisms in the SPR sensing performance improvement were reviewed from three aspects, corresponding to the basic layer configuration of fiber SPR sensors, namely substrate, intrinsic layer, and surface nanomaterial modification. Relevant contents were summarized and compared in detail. In general, the performance improvement of fiber SPR sensors mainly lies in heightening the overlap integral of the SPP electric field intensity on the sensor surface to enhance the sensitivity, reducing the energy losses of SPP to narrow down the FWHM of SPR curve and to improve the DA and the SNR, and introducing the sensing self-compensation to enhance the detection reliability. Furthermore, the development direction of fiber SPR sensors was discussed based on the state of the art in plasmonic detection technologies.