Mode-Enhanced Surface Plasmon Resonance in Few-Mode Fibers via Dual-Groove Architecture

Abstract

We propose a dual-groove few-mode fiber surface plasmon resonance sensor that exploits the LP₁₁ mode for enhanced plasmonic sensing. The device incorporates two physically separated grooves with distinct metallic coatings, enabling dual-channel operation via wavelength-division multiplexing. Finite element method simulations show that the optimized design achieves a maximum sensitivity of 14,800 nm/RIU within the RI range of 1.33–1.40. The introduction of a TiO₂–Au bilayer enhances mode coupling and ensures complete spectral separation, thereby improving stability and reducing environmental interference. Biosensing simulations at 37 °C further confirm the practicality of the proposed architecture. Channel 1, filled with ethanol as a temperature-sensitive medium, provides temperature monitoring, while Channel 2 successfully distinguishes between normal and tumor cells, reaching a sensitivity of up to 9428.57 nm/RIU for Jurkat cells. Overall, the TiO₂-enhanced dual-channel FMF-SPR sensor combines ultra-high sensitivity, spectral independence, and biosensing capability, demonstrating strong potential for next-generation fiber-optic sensing and biomedical applications.

1. Introduction

The rapid development of artificial intelligence and biomedical engineering has substantially increased the demand for high-performance sensing systems [1,2]. Surface Plasmon Resonance (SPR) has emerged as a leading platform for biological detection and medical diagnostics due to its high sensitivity and specificity [3,4,5]. Originating from the resonant interaction between surface free electrons and incident electromagnetic waves, SPR is highly responsive to refractive index (RI) variations in the surrounding medium, making it particularly suited for biochemical sensing [6].

Since Jorgenson et al. first reported fiber-optic SPR sensors in 1993 [7], optical fibers have been widely studied due to their flexibility, compact size, and ease of integration. However, conventional Single-Mode Fibers (SMFs) support only the fundamental mode, which limits their coupling efficiency with SPR [8]. Few-Mode Fibers (FMFs), by contrast, can excite higher-order modes such as LP₁₁. This property enhances the localized electric field and significantly improves sensing performance. Yet FMF-based SPR sensors remain sensitive to temperature drift and environmental perturbations, which compromise stability [9,10,11,12,13].

Several strategies have been proposed to address these issues. Traditional approaches, such as differential detection [14], dual-channel configurations [15], and multilayer metal coatings [16], can improve performance but also increase structural complexity and risk modal interference [17,18,19]. More recent efforts integrate SPR into photonic crystal fibers and novel material platforms. Examples include D-shaped photonic crystal fiber methane sensors [20], tunable metamaterial absorbers based on Dirac semimetals [21], photonic crystal fiber SPR sensors for dual-parameter magnetic field and temperature detection [22], and quad-band terahertz sensors [23]. Despite their innovation, these designs still struggle to simultaneously achieve high sensitivity, robust stability, and structural simplicity.

This gap motivates the present study, which introduces a new approach to overcome these limitations. In this work, we make three key contributions. First, we propose a novel dual-slot FMF-SPR sensor that leverages the LP₁₁ mode and a dual-channel wavelength-division multiplexing (WDM) architecture. Second, the two spatially separated sensing slots, each coated with a distinct metallic layer, allow for independent signal acquisition and cross-validation, thereby enhancing stability without adding structural complexity. Third, systematic finite element method (FEM) analysis demonstrates a maximum sensitivity of 14,800 nm/RIU over an RI range of 1.33–1.40. Biosensing simulations at physiological temperature (37 °C) further confirm reliable differentiation between multiple cancer cell types. Together, these results highlight the practicality of the dual-channel WDM architecture and its strong potential for biomedical sensing and clinical diagnostics.

2. Materials and Methods

2.1. Optical Fiber Structure Design

A symmetric dual-groove FMF-SPR sensor is proposed, as shown in Figure 1. The sensor is based on a standard FMF with a core radius of 7 μm and a core RI of 1.4485. The cladding has a radius of 62.5 μm and an RI of 1.444. Two semi-open grooves, each 10.5 μm in width, are symmetrically etched along the y-axis. This configuration partitions the cladding into two physically isolated sensing regions, enabling the simultaneous detection of different analytes, preventing cross-contamination, and enhancing resistance to external disturbances.

A gold (Au) film with a thickness of tAu is deposited at the base of each groove as the plasmonic excitation layer. In the lower groove, an additional titanium dioxide (TiO₂) layer with a thickness of tTiO₂ is coated above the Au film, forming a heterogeneous bilayer. This bilayer introduces distinct optical responses in the two sensing channels, thereby facilitating signal separation and cross-validation. Moreover, it enhances the coupling between the LP₁₁ mode and the surface plasmon polariton (SPP) mode, resulting in improved RI sensitivity.

2.2. Theoretical and Numerical Framework

2.2.1. Sensing Principle

SPR occurs when the phase-matching condition between a guided fiber mode and the SPP mode at the metal–dielectric interface is satisfied [24]. In the proposed sensor, the LP₁₁ mode couples with the SPP mode at specific wavelengths, producing localized resonance. Variations in the RI of the sensing medium alter the propagation constant of the SPP mode, thereby shifting the phase-matching condition and causing a corresponding resonance wavelength shift [25].

The resonant energy loss from the LP₁₁ mode appears as a sharp dip in the transmission spectrum. The propagation loss is calculated as follows:

$$\alpha =\left|{\displaystyle \frac{2\pi }{\lambda }}\times 8.686\times \mathrm{Im}\left(n_{\mathrm{e}\mathrm{f}\mathrm{f}}\right)\right|$$

(1)

Amplitude sensitivity $S_A$, defined as the rate of loss variation with RI $n_a$, is expressed as follows:

$$S_A=\left|{\displaystyle \frac{1}{{\alpha }_{\left(\lambda ,n_a\right)}}}{\displaystyle \frac{\delta {\alpha }_{\left(\lambda ,n_a\right)}}{\delta n_a}}\right|$$

(2)

Wavelength sensitivity $S_{\lambda }$, defined as the resonance wavelength shift per unit change in RI, is given by the following:

$$S_{\lambda }={\displaystyle \frac{\Delta {\lambda }_{\mathrm{peak}}}{\Delta n_a}}$$

(3)

2.2.2. Numerical Modeling and Parameter Settings

The proposed sensor was numerically modeled using the FEM implemented in COMSOL Multiphysics 6.2. A two-dimensional cross-sectional model with Perfectly Matched Layers (PMLs) at the outer boundaries was employed to suppress artificial reflections.

The dielectric function of Au was described using the Drude–Lorentz model:

$${\epsilon }_{\mathrm{gold}}={\epsilon }_{\mathrm{\infty }}-{\displaystyle \frac{{\omega }_p^2}{\omega \left(\omega +i{\omega }_{\tau }\right)}}$$

(4)

with constants: ${\epsilon }_{\mathrm{\infty }}$ = 9.75, ${\omega }_p^2$ = 1.36 × 10¹⁶ rad²/s² and ${\omega }_{\tau }$ = 1.45 × 10¹⁴ rad/s.

The RI of TiO₂ was defined by an empirical dispersion relation:

$$n_{Ti{O}_2}=\sqrt{5.913+{\displaystyle \frac{2.441\times {10}^7}{{\lambda }^2-0.803\times {10}^7}}}$$

(5)

Simulations were carried out over the 600–1500 nm wavelength range. The RI of the sensing medium was varied from 1.33 to 1.40 RIU. The LP₁₁ mode was used for excitation, and the transmission loss spectra together with resonance wavelength shifts were analyzed to evaluate sensitivity and optimize structural parameters.

2.3. Fabrication Feasibility and Mode Excitation

The fabrication feasibility of the proposed dual-groove FMF-SPR sensor can be ensured by leveraging well-established micro- and nano-fabrication techniques. As illustrated in Figure 2, the dual-groove structure can be inscribed on the sidewall of the FMF using femtosecond laser micromachining, followed by electric-arc polishing to obtain smooth groove surfaces. Au film can then be deposited inside the grooves via magnetron sputtering. In accordance with the proposed design, a heterogeneous bilayer is realized by subsequently depositing a TiO₂ layer onto one of the Au-coated grooves through the same sputtering technique. Such processes are consistent with previously reported fabrication approaches for bilayer fiber-based SPR sensors [18], indicating that the proposed structure can be practically implemented.

A further consideration lies in the selective excitation and stable maintenance of the LP₁₁ mode in the FMF, which is critical to the sensor’s operation. Because mode coupling and bending generally favor the fundamental LP₀₁ mode, additional mode-control strategies are required. Practical methods include precise launch alignment, polarization adjustment, mode filtering, and optimization of the bending radius to suppress LP₀₁ propagation. Recent experimental advances also demonstrate the feasibility of high-purity LP₁₁ operation in FMFs. Liu et al. reported stable LP₁₁ oscillation with >90% modal purity in an all-fiber laser using mode-selective FMF Bragg gratings [26]. Heng et al. further demonstrated a switchable all-fiber Brillouin laser in which the LP₁₁ mode was maintained with 97% purity and long-term stability [27]. These results provide strong evidence that the LP₁₁ mode required in the proposed sensor can be reliably excited and sustained using existing techniques.

3. Numerical Simulation and Design Discussion

This section presents the FEM-based modeling of the proposed dual-groove FMF-SPR sensor. The LP₁₁ mode is first evaluated and compared with the LP₀₁ mode to identify the optimal guiding configuration. Subsequently, the performance of multiple higher-order modes is analyzed, followed by optimization of structural parameters—such as Au film thickness and residual cladding thickness—to improve sensitivity. The dual-channel responses under varying RIs are then examined, and a TiO₂ layer is incorporated to enhance spectral separation. Finally, biosensing simulations at physiological temperature (37 °C) are conducted to validate the sensor’s capability in distinguishing cancer cell types.

3.1. Mode Analysis and Selection in the FMF-SPR Sensor Structure

3.1.1. Guiding Behavior and Resonance Characteristics

To determine an effective sensing mode, simulations were performed at an analyte RI of 1.35. The coupling behavior between the LP₁₁ mode and the SPP mode was examined and compared with that of the fundamental LP₀₁ mode.

As shown in Figure 3a, the effective RI of the LP₁₁ and SPP modes intersect as a function of wavelength, indicating phase matching and enabling strong mode coupling. This leads to a pronounced resonance dip in the loss spectrum, resulting from efficient energy transfer to the metal surface. Figure 3b compares the loss spectra of the LP₀₁ and LP₁₁ modes under identical conditions. The LP₁₁ mode exhibits a sharper and deeper resonance peak, reflecting stronger coupling with the SPP mode. In addition, it has a longer evanescent field at the metal interface, enhancing its sensitivity to RI variations.

These results show that the LP₁₁ mode provides stronger resonance, higher analyte–field overlap, and enhanced sensing performance, making it the optimal choice for subsequent optimization.

3.1.2. Performance Evaluation of Multiple Modes

To further assess sensing performance across different guided modes, four representative modes—LP₀₁, LP₁₁, LP₂₁, and LP₀₂—were systematically compared. The analysis considered transmission loss spectra, resonance wavelength shifts, and evanescent field penetration depths, as these parameters collectively determine the sensing range and detection accuracy.

As shown in Figure 4a–d, all four modes exhibit a progressive redshift in resonance wavelength as the analyte RI increases from 1.33 to 1.37. This trend confirms that phase matching with the SPP mode is effectively maintained across the tested RI range. Notably, the higher-order modes—LP₁₁, LP₂₁, and LP₀₂—produce sharper and deeper resonance peaks than the fundamental LP₀₁ mode, indicating stronger coupling efficiency and greater RI sensitivity.

Table 1. Quadratic fitting of resonance wavelength versus RI for different modes.

Mode	Fitting Equation	R2
LP01	y = 37,086 − 55,937x + 21,428x2	0.9999
LP11	y = 34,458 − 52,060x + 20,000x2	0.9999
LP21	y = 38,192 − 57,715x + 22,142x2	0.9999
LP02	y = 46,052 − 69,327x + 26,428x2	0.9995

summarizes the quadratic fitting of resonance wavelength versus RI for the four modes. The fitted curves, shown in Figure 5, exhibit smooth monotonic behavior with excellent agreement to the simulation results. All R² values exceed 0.999, demonstrating the accuracy and reliability of the fitting model for calibration and prediction.

Furthermore, the evanescent field penetration depth—defined as the distance at which the electric field intensity decays to 1/e of its surface maximum—was calculated for each mode at RI = 1.35. The results, summarized in

Table 2. Resonance wavelength and penetration depth for each mode at RI = 1.35.

Mode	Resonance Wavelength (nm)	Penetration Depth (nm)
LP01	625	192.45
LP11	627	194.88
LP21	632	194.74
LP02	626	193.04

, range from 192 nm to 195 nm, indicating similar decay behavior across all modes.

Taken together, these results demonstrate that the LP₁₁ mode provides the best balance between resonance intensity, sensitivity, and penetration depth, making it the most suitable candidate for further optimization in the proposed sensor design.

3.2. Optimization of Structural Parameters in the FMF-SPR Sensor Configuration

3.2.1. Optimization of Residual Thickness After Polishing

To evaluate the effect of residual cladding thickness (Rc) on the FMF-SPR structure, simulations were performed with tAu fixed at 40 nm. Both sensing channels were filled with analytes of RI = 1.33 and RI = 1.35.

As shown in Figure 6a, decreasing Rc reduces the distance between the fiber core and the metal interface, thereby enhancing coupling between the LP₁₁ mode and the SPP mode. This produces a redshift in the resonance wavelength and an increase in the loss peak. In contrast, increasing Rc weakens modal interaction, leading to a blueshift and a lower resonance peak. Figure 6b further shows that amplitude sensitivity decreases progressively with increasing Rc. The highest sensitivity is observed at Rc = 0 μm; however, complete removal of the cladding is impractical due to fabrication limitations and the risk of fiber core damage. Accordingly, Rc = 0.1 μm is selected as the optimal value, balancing sensing performance with structural robustness.

Beyond spectral response, Rc also affects the evanescent field distribution.

Table 3. Resonance wavelength and penetration depth at different Rc values (RI = 1.35).

Rc (μm)	Resonance Wavelength (nm)	Penetration Depth (nm)
0.0	630	95.0
0.1	627	194.8
0.2	626	295.1
0.3	625	397.8

summarizes the resonance wavelength and penetration depth at RI = 1.35. The penetration depth increases markedly with Rc, ranging from 95.0 nm at Rc = 0 μm to 397.8 nm at Rc = 0.3 μm, accompanied by a slight blueshift in resonance wavelength.

In summary, Rc governs a trade-off between coupling efficiency and sensing depth: smaller Rc enhances modal interaction and sensitivity, whereas larger Rc provides deeper evanescent field penetration. Considering both performance and fabrication feasibility, Rc = 0.1 μm is identified as the optimal residual thickness.

3.2.2. Optimization of Excitation Layer Thickness

With Rc fixed at 0.1 μm, the effect of tAu on sensor performance was investigated. Simulations were conducted for tAu ranging from 30 nm to 60 nm in 10 nm increments. The two sensing grooves were filled with analytes of RI = 1.33 and RI = 1.35 to simulate differential sensing conditions.

As shown in Figure 7a, increasing tAu causes a gradual redshift in the resonance wavelength. At tAu = 30 nm, the resonance peak is sharpest, and the transmission loss is highest, indicating strong coupling between the LP₁₁ mode and the SPP mode. At tAu = 40 nm, the resonance remains well-defined with slightly reduced loss, suggesting good stability. However, at tAu = 50–60 nm, the resonance dips become significantly shallower, reflecting reduced modal coupling and degraded sensing precision.

In addition to spectral response, the evanescent field penetration depth was cal-culated to assess field–analyte interaction.

Table 4. Resonance wavelength and penetration depth for different tAu at Rc = 0.1 μm, RI = 1.35.

tAu (nm)	Resonance Wavelength (nm)	Penetration Depth (nm)
30	615	186.4
40	621	213.7
50	628	242.1
60	635	245.5

presents the resonance wavelength and penetration depth for different tAu values at Rc = 0.1 μm and RI = 1.35. The penetration depth increases monotonically with tAu, but the growth rate saturates beyond 50 nm.

These findings indicate a trade-off between evanescent field penetration and resonance sharpness. While thicker Au layers extend the field and improve amplitude sensitivity, they also reduce coupling efficiency and spectral clarity. Considering both sensing stability and field interaction, tAu = 40 nm is identified as the optimal excitation layer thickness for reliable operation.

3.3. Sensing Performance of the Optimized FMF-SPR Sensor Structure

3.3.1. Dual-Channel Consistency Analysis

To evaluate the dual-channel consistency of the optimized FMF-SPR sensor, both channels were symmetrically filled with dielectric media of RI ranging from 1.33 to 1.40. The corresponding loss spectra are shown in Figure 8a. As the RI increases, the resonance wavelength exhibits a continuous redshift. For each RI, only a single resonance dip is observed, indicating that the SPR responses of both channels remain identical under symmetric conditions.

To further evaluate the sensing performance, the wavelength sensitivity was calculated from the resonance wavelength shifts, as presented in Figure 8b. The sensitivity increases monotonically with RI, reaching a maximum of 5500 nm/RIU at n = n₁ = n₂ = 1.40. At this RI, the resonance spectra of both channels fully overlap, forming a single, symmetric, and sharply defined resonance peak. These results demonstrate the FMF-SPR sensor’s excellent dual-channel consistency, characterized by high resonance stability and strong signal enhancement.

As shown in

Table 5. Resonance wavelength and penetration depth under symmetric dual-channel configuration.

RI	Resonance Wavelength (nm)	Penetration Depth (nm)
1.33	595	145.2
1.35	627	196.6
1.37	673	230.6
1.39	752	297.4

, both the resonance wavelength and the penetration depth increase steadily with rising RI. The extended evanescent field penetration facilitates stronger analyte interaction, thereby improving sensing performance and broadening the effective detection range.

3.3.2. Dual-Channel Asymmetric RI Configuration

To investigate the anti-interference capability of the FMF-SPR sensor under more complex conditions, the spectral response was analyzed under an asymmetric dual-channel configuration. In this case, the RI of Channel 1 (CH1) was fixed at n₁ = 1.33 as a reference, while the RI of Channel 2 (CH2), denoted as n₂, was varied from 1.33 to 1.40 to simulate analyte-induced changes.

As illustrated in Figure 9, when the RIs of the two channels differ, the overall resonance intensity decreases significantly compared with the symmetric case. This reduction arises from the loss of phase-matching synchronization between the guided modes and the SPP modes on both sides. The phase mismatch suppresses coherent superposition, leading to mode splitting in the coupling spectrum.

When n₂ ≥ 1.36, two distinct resonance peaks emerge. One peak remains near 595 nm, corresponding to SPR excitation in the reference channel (CH1) with constant RI. The other peak exhibits a progressive redshift as n₂ increases, reflecting the enhanced effective RI of the SPP mode in CH2. This behavior indicates that the phase-matching point between the guided mode and the plasmonic mode shifts toward longer wavelengths as the analyte RI rises.

Compared with the symmetric configuration, the asymmetric RI distribution results in clear spectral separation between the reference and sensing signals. This separation enables precise identification of localized RI variations, while enhancing robustness against environmental fluctuations. These results confirm the feasibility of the dual-channel design for independent multi-point detection and improved noise immunity in practical sensing applications.

3.4. Optimization and Performance Enhancement of FMF-SPR Sensors

3.4.1. Performance Analysis of the TiO₂ Dielectric Layer

To further enhance the channel separation capability of the dual-channel FMF-SPR sensor, a high-index TiO₂ dielectric layer was introduced on one side of the Au surface, forming an asymmetric multilayer structure. In this configuration, the side coated solely with Au acts as the reference channel (CH1), while the side with the Au + TiO₂ composite layer serves as the sensing channel (CH2). Owing to the high RI of TiO₂, the resonance wavelength in CH2 undergoes a pronounced redshift, thereby expanding the spectral gap between the two channels. This structural modification enhances anti-interference performance and enables independent dual-channel detection.

The effect of tTiO₂ was analyzed to balance resonance wavelength shift and resonance loss intensity. Figure 10a–c presents the simulated loss spectra and sensitivity results for tTiO₂ = 10 nm under different RI conditions.

As shown in Figure 10a, when n= 1.33, the resonance wavelength separation between CH1 and CH2 reaches approximately 125 nm, indicating effective spectral decoupling. This increased separation improves detection stability and anti-interference capability. Moreover, CH2 exhibits a stronger resonance loss peak, confirming that the TiO₂ layer enhances the coupling efficiency between the guided mode and the Au interface, thereby reinforcing SPR excitation. In Figure 10b, CH2 consistently demonstrates higher wavelength sensitivity across the RI range, with a marked increase when n > 1.38. While the maximum sensitivity of CH1 reaches 5500 nm/RIU, CH2 achieves up to 7500 nm/RIU, highlighting the structural advantage of TiO₂-enhanced sensing.

However, Figure 10c shows that when CH2 is filled with n₁ = 1.33 and CH1 with n₂ = 1.38, the resonance peaks of the two channels converge, leading to partial spectral overlap. This result indicates that the 10 nm TiO₂ configuration does not fully guarantee channel independence under all RI conditions.

Therefore, increasing tTiO₂ is necessary to induce a larger redshift in CH2, thereby enlarging the spectral gap and ensuring complete channel separation. Such optimization enables fully independent dual-channel detection with complementary verification across the entire RI range.

3.4.2. Optimization of TiO₂ Layer Thickness

Although the introduction of the TiO₂ dielectric layer enhances spectral separation and sensing performance, partial overlap between the resonance peaks of CH1 and CH2 may still occur under certain RI combinations. To ensure fully independent channel responses across the sensing range, it is necessary that the loss spectra of the two channels remain non-overlapping under all RI pairings.

For this purpose, the RI of CH1 was fixed at n₁ = 1.40 and that of CH2 at n₂ = 1.33, while tTiO₂ was systematically varied. Figure 11a shows that increasing tTiO₂ causes a continuous redshift of the resonance wavelength in CH2. At smaller thicknesses, spectral overlap is observed, whereas at 30 nm the resonance peaks of the two channels are completely separated across the investigated RI range.

To gain further insight into the effect of tTiO₂, the confinement loss of the core mode was simulated for 0, 10, and 30 nm over an RI range of 1.33–1.40. As shown in Figure 11b, wavelength sensitivity increases with tTiO₂. At 30 nm, the maximum sensitivity reaches 14,800 nm/RIU, compared with lower values at smaller thicknesses or without coating.

In addition to sensitivity, resonance sharpness was evaluated. The figure of merit (FOM) was introduced and defined as follows:

$$\mathrm{F}\mathrm{O}\mathrm{M}={\displaystyle \frac{S_{\lambda }}{\mathrm{F}WHM}}$$

(6)

where $S_{\lambda }$ is the wavelength sensitivity (nm/RIU), and $\mathrm{F}\mathrm{W}\mathrm{H}\mathrm{M}$ is the full width at half maximum of the resonance dip.

As shown in Figure 12a,b, the resonance dips broaden as RI increases, particularly at tTiO₂ = 30 nm. Figure 12c indicates that the FOM remains above 200 across most of the sensing range, with a maximum value of approximately 350 within the RI range of 1.34–1.35.

Based on the combined evaluation of wavelength sensitivity, spectral separation, and FOM, a TiO₂ thickness of 30 nm provides the best overall performance. At this thickness, dual-channel independence is maintained, and spectral separation can be resolved even when RI differences are as small as 0.01.

Table 6. Comparison of the Proposed Au–TiO₂ FMF-SPR Biosensor with Previously Reported Multilayer SPR Sensors.

[Ref.]	RI Range (RIU)	Coating Structure	Guided Mode	Max. Sensitivity (nm/RIU)
[28]	1.26–1.42	TiO2-Au	LP01	7500
[29]	1.33–1.41	MXene-Au-TiO2	LP01	9400
[30]	1.33–1.40	Ag/Au/Graphene/MoS2	LP01	11,775
This work	1.33–1.40	Au-TiO2	LP11	14,800

summarizes a comparison between the proposed Au–TiO₂ FMF-SPR biosensor and previously reported multilayer SPR sensors [28,29,30]. The proposed design achieves a maximum sensitivity of 14,800 nm/RIU within the RI range of 1.33–1.40, which is higher than the values reported for the other structures.

3.5. Analysis of the Optimized Biosensor for Tumor Cell Detection

The dual-channel FMF-SPR sensor has potential applications in multiple biosensing scenarios, with cancer cell detection as a representative case. To evaluate the performance of the optimized design, three representative mammalian cell models were investigated: skin (Basal), cervical (HeLa), and blood (Jurkat), each compared between its normal and tumor states.

For temperature sensing, CH1 was filled with ethanol, whose RI varies with temperature as follows [31]:

$$n\left(T\right)=1.3605-3.94\times {10}^{-4}\cdot T$$

(7)

The simulation range was set from −10 to 40 °C. As shown in Figure 13a, the resonance wavelength exhibits a blue shift and reduced loss with increasing temperature. The temperature sensitivity is defined as follows [31]:

$$S_T={\displaystyle \frac{\Delta {\lambda }_{Peak}}{\Delta T}}$$

(8)

The maximum sensitivity obtained is −1.2 nm/°C, and the linear regression yields R² = 0.99734 (Figure 13b), confirming stable and predictable thermal response in CH1.

For cell detection, the system temperature was fixed at 37 °C [32], which corresponds to the physiological culture condition for mammalian cells and minimizes temperature-induced refractive-index drift, ensuring reliable results. CH2 was filled with Basal, HeLa, and Jurkat samples in both normal and tumor states [33,34,35]. Their refractive indices are listed in

Table 7. RI parameters of normal and tumor cells.

Tissue Type	Cell Type	Normal Cell RI	Tumor Cell RI
Skin [33]	Basal	1.360	1.380
Cervix [34]	HeLa	1.368	1.392
Blood [35]	Jurkat	1.376	1.390

.

As shown in Figure 14a–c, for each cell type, the tumor sample exhibits a clear red shift compared to its normal counterpart, while the CH1 reference peak remains unchanged. This demonstrates that the sensor can reliably distinguish the normal and tumor states of the same cell type.

As summarized in

Table 8. Wavelength sensitivity comparison with prior studies.

[Ref.]	Tumor-Related Cells	Wavelength Sensitivity (nm/RIU)
[33]	Basal cell	2500
	HeLa	3750
	Jurkat	4285
[35]	Basal cell	3150
	HeLa	4333
	Jurkat	4642
[36]	Basal cell	3800
[37]	Basal cell	4950
This work	Basal cell	6250
	HeLa	8875
	Jurkat	9428

, the wavelength sensitivities obtained in this work are compared with those reported in previous studies [33,34,35,36,37]. The results demonstrate that the optimized dual-channel FMF-SPR sensor achieves consistently higher sensitivities across all three types of cells, confirming the advantage of the proposed structure.

In summary, the dual-channel FMF-SPR sensor enables simultaneous temperature monitoring and tumor-cell identification on a single platform, with evident improvements in sensitivity over existing designs. Beyond this demonstration, the architecture is naturally extensible: by integrating serial micro-grooves (multiple sensing regions) along a single fiber, it enables multiplexed detection of different biomarkers and simultaneous readout of multiple parameters (e.g., temperature, and potentially pH) within one measurement. Such serial integration increases practical throughput and aligns well with clinical workflows that require multi-sample and multi-target analysis within constrained assay time, thereby offering a promising pathway for early cancer screening and clinical diagnostics.

4. Conclusions

In this study, a dual-groove FMF-SPR sensor based on the LP₁₁ mode is proposed, which shows improved resonance characteristics compared with the LP₀₁ mode. By implementing a dual-channel WDM structure with different coatings (Au for the reference channel and Au/TiO₂ for the sensing channel) on physically separated grooves, the device achieves clearly separated resonance valleys within the same spectrum, ensuring stable dual-channel operation. Simulations indicate a maximum sensitivity of 14,800 nm/RIU over 1.33–1.40, with the FOM exceeding 200 in most of the range.

Further simulations at 37 °C demonstrate that the sensing channel can distinguish Basal, HeLa, and Jurkat cells between normal and tumor states, with a maximum sensitivity of 9428.57 nm/RIU for Jurkat. These results confirm that the TiO₂-enhanced, LP₁₁-assisted dual-channel FMF-SPR sensor provides high sensitivity, spectral independence, and stability, showing strong potential for biosensing and biomedical detection.