Prism-based surface plasmon resonance sensors are widely used for detecting small changes in the refractive index of liquids, gases, biological layers, and functional coatings. Their apparent structural simplicity can be misleading. A conventional Kretschmann sensor may contain only a prism, a thin metallic film, and an analyte, yet its predicted resonance angle, minimum reflectance, linewidth, field enhancement, and sensitivity can change substantially with the material dataset, polarization definition, film thickness, boundary conditions, mesh density, and angular sampling strategy.
A reliable simulation should therefore reproduce the physical coupling mechanism rather than merely generate a reflectance dip. The numerical model must correctly represent the phase matching between the incident optical field and the surface plasmon polariton, include the complex and dispersive optical properties of the metal, resolve the electromagnetic field across the thin metallic layer, and distinguish genuine plasmonic absorption from numerical loss or boundary reflection.
If you are interested in collaborating on surface plasmon resonance-based research, we are open to discussions. We have our own. software algorithm and experimental facility in-house.
This article discusses the most important considerations for developing, validating, and optimizing prism-based SPR sensor simulations using the transfer matrix method, finite-element method, finite-difference time-domain method, or comparable full-wave electromagnetic techniques.
Begin with the Correct Physical Interpretation
Phase Matching in the Kretschmann Configuration
In the Kretschmann configuration, a thin metal film is deposited directly on a high-index prism. Light enters through the prism and reaches the prism-metal interface at an angle greater than the relevant critical angle. Under these conditions, the optical field penetrates through the metal as an evanescent field and can excite a surface plasmon polariton at the metal-analyte interface.
For a planar interface between a metal with relative permittivity $\varepsilon_m$ and a dielectric analyte with relative permittivity $\varepsilon_d$, the complex propagation constant of the surface plasmon polariton is approximately
$$
k_{\mathrm{SPP}} = k_0
\sqrt{
\frac{\varepsilon_m\varepsilon_d}
{\varepsilon_m+\varepsilon_d}
},
$$
where
$$
k_0=\frac{2\pi}{\lambda_0}
$$
is the free-space wavenumber.
The tangential component of the incident wavevector inside the prism is
$$
k_x=k_0n_p\sin\theta,
$$
where $n_p$ is the prism refractive index and $\theta$ is the internal angle of incidence measured relative to the surface normal.
The approximate resonance condition is therefore
$$
k_0n_p\sin\theta_{\mathrm{SPR}}
\approx
\operatorname{Re}(k_{\mathrm{SPP}}).
$$
This relationship is useful for estimating the expected resonance angle before running a numerical sweep. However, it should not be treated as an exact prediction for a finite metal film. The actual resonance position is influenced by the metal thickness, prism loading, optical loss, adhesion layers, intermediate coatings, and the finite coupling strength between the prism field and the surface plasmon mode.
A useful model should reproduce both the reflectance minimum and the expected field localization at the metal-analyte interface. A reflectance dip without strong interface-localized electromagnetic energy may originate from thin-film interference, lossy absorption, or a numerical artifact rather than true surface plasmon excitation.
Use Transverse Magnetic Polarization
Surface plasmon polaritons at a conventional metal-dielectric interface can be excited only by transverse magnetic polarization, commonly called TM or p polarization. In this polarization, the electric field has a component normal to the interface, allowing it to drive the collective oscillation of free electrons in the metal.
Transverse electric polarization, also called TE or s polarization, does not excite the conventional SPP mode in an isotropic planar structure. A TE simulation can still be useful as a control case. The TE reflectance should not contain the characteristic SPR dip seen in the TM response. If similar dips appear in both polarizations, the model may be showing an interference resonance, an incorrect polarization definition, or an improperly normalized port response.
Polarization conventions differ between software packages. In a two-dimensional model, one formulation may solve for the out-of-plane magnetic field, while another solves for the out-of-plane electric field. The user should verify the actual field components rather than relying only on labels such as TM, TE, p, or s.
Establish a Validated Planar Baseline
Do Not Begin with an Overcomplicated Multilayer Sensor
A common modeling mistake is to start immediately with graphene, transition-metal dichalcogenides, dielectric enhancement layers, affinity coatings, nanoparticles, or bimetallic films. These additional layers introduce several coupled variables and can conceal fundamental errors in the base model.
The first simulation should contain only the prism, the metal film, and a homogeneous analyte. For a gold-based sensor, a film thickness around several tens of nanometers is generally an appropriate starting region, although the optimum value must be determined for the selected wavelength, prism, analyte, and optical-constant dataset. The objective of the baseline model is not to maximize sensitivity. It is to establish a physically consistent reference response.
The baseline simulation should produce a smooth TM reflectance curve, a clear resonance minimum above the critical-angle region, strong electric-field localization near the metal-analyte boundary, and an acceptable energy balance. Only after these characteristics have been confirmed should additional layers be introduced.
Use the Transfer Matrix Method as a Benchmark
For an optically infinite planar multilayer structure, the transfer matrix method is usually faster and more transparent than a full-wave finite-element or FDTD model. It provides an excellent reference for checking the resonance angle, minimum reflectance, and approximate linewidth.
A practical validation strategy is to build the prism-metal-analyte structure using the transfer matrix method and independently reproduce it using the full-wave model. The two approaches should show close agreement when identical optical constants, layer thicknesses, wavelength, polarization, and angle definitions are used.
Small differences may arise from interpolation, solver tolerances, port normalization, or finite-domain effects. Large differences usually indicate an inconsistency in material definitions, angle implementation, polarization, geometry orientation, or boundary treatment.
Full-wave simulation becomes especially valuable when the structure contains finite-width beams, surface roughness, patterned layers, gratings, nanoparticles, microfluidic boundaries, nonuniform sensing films, or coupled thermal and electromagnetic physics. For a perfectly planar sensor, however, agreement with the transfer matrix method should be demonstrated before trusting more complicated results.
Define Material Properties Carefully
Use Complex and Dispersive Metal Permittivity
The optical response of the metal is the most influential material input in an SPR simulation. Gold, silver, aluminum, and copper exhibit strongly wavelength-dependent complex permittivity. Assigning a constant real refractive index to the metal removes the physical absorption mechanism and produces an invalid plasmonic response.
The complex refractive index is commonly written as
$$
\tilde{n}=n+i\kappa,
$$
where $n$ is the refractive index and $\kappa$ is the extinction coefficient. The corresponding relative permittivity is
The sign of the imaginary component depends on the time-harmonic convention used by the software. It is therefore safer to use the software's built-in optical material model or verify its convention before entering complex permittivity manually.
Frequently used experimental data for gold and other noble metals include the Johnson and Christy optical constants. Analytical Lorentz-Drude and Brendel-Bormann representations are available from the Rakić optical-property models. Convenient tabulated versions of several datasets are also available through the refractiveindex.info database.
The selected dataset should always be reported because different datasets can predict noticeably different resonance depths and linewidths. Experimental thin films may also differ from bulk reference data because of grain size, surface roughness, deposition conditions, void fraction, contamination, and adhesion layers.
Do Not Mix Incompatible Optical Datasets
An SPR simulation may combine prism data from one temperature, gold data from another temperature, and analyte data measured at an unrelated wavelength. This can produce apparently precise but physically inconsistent predictions.
For angular interrogation at a fixed wavelength, every material should be evaluated at that wavelength. For wavelength interrogation, each material should be represented by an appropriate dispersion function or interpolation table over the entire wavelength range.
The interpolation method should be checked near dataset boundaries. Extrapolating metal optical constants outside the measured spectral range can produce nonphysical permittivity values. A spectral sweep should remain within the valid range of all material models unless a properly validated analytical dispersion model is used.
Include Prism Dispersion
The prism determines the maximum tangential wavevector that can be supplied to the surface plasmon mode. Consequently, the prism refractive index directly influences the accessible resonance-angle range.
BK7 glass is frequently used in visible-wavelength SPR systems, while higher-index glasses or crystals may be selected when additional momentum is required. The refractive index should ideally be represented using a Sellmeier equation or manufacturer-provided dispersion data rather than a single approximate value.
A typical Sellmeier relation has the form
$$
n^2(\lambda)=1+
\sum_j
\frac{B_j\lambda^2}
{\lambda^2-C_j},
$$
where $B_j$ and $C_j$ are material-specific coefficients.
The wavelength unit used in the Sellmeier equation must be verified. Many published coefficient sets require wavelength in micrometers. Entering wavelength in meters or nanometers without conversion can produce completely incorrect prism indices.
Represent the Analyte Realistically
The analyte should not always be modeled as a perfectly lossless material with a constant refractive index. Water, biological fluids, solvents, polymers, and gas mixtures may exhibit wavelength and temperature dependence. Absorption may also become relevant in infrared spectral ranges.
If the purpose of the sensor is bulk refractive-index sensing, the analyte can initially be represented as a homogeneous semi-infinite dielectric. For biosensing, however, the actual measurand may be a nanometer-scale binding layer rather than a uniform change throughout the analyte domain. Bulk sensitivity and surface sensitivity are related but distinct quantities.
A model intended for practical comparison should distinguish between a change in the entire analyte refractive index and the formation of a finite-thickness biological layer. Treating a molecular binding event as a uniform bulk refractive-index increase can significantly overestimate the response.
Include Adhesion Layers When Comparing with Experiments
Gold films are often deposited on chromium or titanium adhesion layers. Although these layers may be only a few nanometers thick, they are lossy and are located directly within the high-field region. They can broaden the resonance, increase minimum reflectance, modify the resonance angle, and reduce field enhancement.
A theoretical structure without an adhesion layer may therefore show a considerably deeper and narrower resonance than an experimentally fabricated sample. Adhesion layers can be omitted during idealized optimization, but they should be included when the purpose is experimental prediction or device design.
For metal films approaching only a few nanometers in thickness, bulk optical constants may no longer be reliable. Film continuity, surface scattering, island formation, and nonlocal effects may become important. Such ultrathin structures require experimentally informed material models rather than straightforward bulk-property assignment.
Choose the Appropriate Model Dimensionality
Use a Two-Dimensional Unit Cell for Uniform Planar Sensors
A conventional prism-metal-analyte stack is translationally invariant along the interface. A two-dimensional model is usually sufficient when the incident field is a plane wave and the layers are uniform.
The finite width of the modeled region should not be interpreted as the physical sensor width when periodic or Floquet conditions are used. The computational domain represents a repeating section of an effectively infinite planar structure.
A three-dimensional model is normally unnecessary unless the sensor contains finite lateral features, asymmetric particles, localized defects, nonuniform microfluidic structures, finite illumination spots with lateral variation, or polarization coupling that cannot be represented in two dimensions.
Reducing a valid three-dimensional problem to two dimensions can remove important physics. Conversely, using three dimensions for a strictly planar problem increases computational cost without improving the physical prediction.
Use a Finite Beam Only When It Serves the Research Question
Most analytical SPR calculations assume plane-wave illumination. Experimental systems use beams with finite angular divergence and finite spatial width. A finite Gaussian beam can broaden the observed resonance and influence the apparent minimum reflectance, especially when the intrinsic SPR dip is extremely narrow.
Finite-beam modeling should be introduced only after the plane-wave response has been validated. Otherwise, beam divergence, lateral propagation, boundary truncation, and detector integration can obscure basic model errors.
For many sensor-design studies, plane-wave angular interrogation is appropriate. A Gaussian beam becomes relevant when investigating imaging SPR, lateral Goos-Hänchen shifts, beam deformation, finite detector apertures, angular-spectrum averaging, or direct comparison with a specific optical arrangement.
Implement the Incident Angle Correctly
Define the Angle Inside the Prism
The resonance angle reported for a prism-based SPR sensor is usually the internal incidence angle within the prism, measured from the normal to the metal surface. Confusion arises when the external laboratory angle, prism-face angle, source angle, and internal interface angle are treated as identical.
The wavevector components inside the prism should be parameterized as
$$
k_x=k_0n_p\sin\theta,
$$
$$
k_y=k_0n_p\cos\theta.
$$
The signs of the components depend on the direction of propagation and the coordinate system. A sign error may reverse the propagation direction or cause the port to launch an outgoing rather than incoming mode.
When modeling the complete prism geometry, Snell's law must be used to relate the external angle to the internal angle. When modeling only the layered interface, it is usually clearer to define the internal angle directly.
Check the Critical-Angle Region
For a simple prism-dielectric interface, the approximate critical angle is
$$
\theta_c=
\sin^{-1}
\left(
\frac{n_d}{n_p}
\right),
$$
provided that $n_p>n_d$.
The SPR minimum normally occurs in the attenuated-total-reflection region, although the exact behavior depends on the complete multilayer structure. A resonance located far below the expected critical-angle region should be examined carefully because it may correspond to ordinary thin-film interference rather than plasmon coupling.
The critical angle is only a diagnostic reference. The prism is not directly in contact with the analyte in the complete Kretschmann stack, and the metal film modifies the optical response. Nevertheless, calculating $\theta_c$ helps determine an efficient initial sweep range.
Select Appropriate Excitation and Boundary Conditions
Use Ports or Floquet Excitation Consistently
For an infinite layered model, periodic or Floquet boundaries along the lateral direction are often the most efficient choice. The phase shift across the periodic pair must correspond to the tangential wavevector $k_x$ of the obliquely incident wave.
The input boundary should launch the TM-polarized incident mode. The output boundary should absorb transmitted or evanescent components without introducing artificial reflections. In port-based formulations, the reflected power can often be obtained from the port scattering parameter, but only when the port modes and power normalization are correctly defined.
The official COMSOL example for Otto and Kretschmann excitation provides a useful reference implementation using periodic ports, Floquet periodicity, TM excitation, and an angular parameter sweep. Regardless of the software used, the same underlying requirements apply: correct tangential phase, correct polarization, correct propagation direction, and reflection-free truncation.
Do Not Use Perfect Electric Conductor Boundaries for Open Optical Domains
Perfect electric conductor boundaries are generally inappropriate for truncating prism or analyte regions. They reflect electromagnetic waves and can create artificial standing-wave patterns that distort the SPR dip.
Open boundaries should be represented using ports, scattering boundaries, absorbing boundaries, or perfectly matched layers. The correct choice depends on whether the outgoing field is a propagating plane wave, an evanescent component, or a mixture of modes.
Perfectly matched layers should be sufficiently separated from the metal-analyte interface so that they do not interact with the physically relevant evanescent field. They should also be meshed appropriately. A poorly designed PML can introduce more reflection than a simpler absorbing boundary.
Make the Analyte Domain Optically Semi-Infinite
The sensing region should extend far enough from the metal surface that the SPP field decays substantially before reaching the outer boundary. If the analyte domain is too thin, the outer boundary may interact with the evanescent field and shift the resonance.
Instead of selecting an arbitrary domain thickness, inspect the field decay normal to the interface. The analyte thickness should span several decay lengths unless an absorbing treatment accurately represents a semi-infinite region.
The SPP penetration depth in the dielectric can be estimated from the imaginary part of the normal wavevector. If the analyte field varies as $\exp(-\alpha_d y)$, the amplitude decay length is approximately
$$
\delta_d=\frac{1}{\operatorname{Re}(\alpha_d)}.
$$
The exact form depends on the coordinate convention and complex wavevector definition. The practical objective is simply to ensure that the computational boundary does not truncate a significant portion of the sensing field.
Develop a Mesh for the Metal and Interface Fields
Resolve the Metal Thickness Explicitly
The metal layer may be only a few tens of nanometers thick, while the free-space wavelength may be hundreds of nanometers. A default wavelength-based mesh can therefore place too few elements across the metal.
A useful starting point is to place several well-shaped elements across the metal thickness, often approximately six to ten for a first quadratic-element FEM model. This is not a universal requirement. The final mesh must be justified through convergence testing.
Mapped elements are especially effective for planar multilayers because they maintain controlled element thickness through the metal and coating layers. Highly skewed triangular elements at thin interfaces should be avoided because they may degrade the field-gradient representation.
Consider the Wavelength and Skin Depth in the Metal
The mesh size should not be based only on the free-space wavelength. The complex refractive index of the metal produces a shorter effective wavelength and a finite electromagnetic skin depth.
The field variation in the metal is governed by a complex wavevector. A practical mesh should resolve both the phase variation and exponential attenuation. Strong field gradients also occur at the metal-dielectric boundary, where the normal electric-field component changes according to the electromagnetic boundary conditions.
Local refinement should therefore be concentrated within the metal, the adhesion layer, any nanometer-scale sensing coating, and the immediate neighboring regions. Uniformly refining the entire prism is computationally inefficient.
Perform a Mesh-Convergence Study on Sensor Metrics
A visually smooth field plot does not demonstrate mesh convergence. Convergence should be evaluated using quantities that determine the sensor performance, including the resonance angle, minimum reflectance, FWHM, absorptance, and interface field enhancement.
The mesh should be refined until these quantities change by less than a predefined tolerance. The acceptable tolerance depends on the intended claim. A model used to compare large qualitative trends may tolerate a larger numerical variation than a model claiming millidegree resonance shifts.
Mesh convergence and angular-step convergence should be tested separately. Refining the mesh cannot correct an inadequately sampled resonance curve.
Design the Angular or Spectral Sweep Intelligently
Use a Coarse-to-Fine Angular Sweep
A single uniform sweep over a very broad angular range is inefficient and may miss a narrow resonance. Begin with a coarse sweep to identify the approximate resonance region. Then perform a refined sweep around the dip.
The required angular increment depends on the resonance linewidth. A broad lossy resonance may be adequately resolved with steps of several hundredths of a degree. A narrow optimized structure may require steps of a few thousandths of a degree or smaller.
The resonance angle should not automatically be assigned to the lowest sampled point. A local polynomial fit, spline fit, or one-dimensional minimization around the dip can estimate the minimum more accurately than the sweep spacing. The fitting interval should remain local so that the baseline curvature does not bias the result.
Refine Each Analyte Case Independently
When the analyte refractive index changes, the resonance angle shifts. A fixed narrow sweep centered on the original resonance may truncate the dip for later analyte values.
Each analyte case should either use a sufficiently broad sweep or employ an adaptive procedure that relocates and refines the resonance region. The same fitting and linewidth-extraction method should be applied consistently to all cases.
Artificially large sensitivity values can appear when one resonance is fully captured but another is sampled only along one side of the dip. Every reflectance curve should therefore be inspected before calculating performance metrics.
Treat Wavelength Interrogation as a Fully Dispersive Problem
In wavelength interrogation, the incidence angle is held fixed while the wavelength is varied. The resonance wavelength satisfies a wavelength-dependent phase-matching condition.
All materials must be dispersive in this simulation. Keeping the metal or prism permittivity constant during a wavelength sweep can produce a mathematically smooth but physically meaningless spectrum.
The spectral resolution should be selected according to the resonance linewidth, and the resonance wavelength should be refined using interpolation or local minimization. Wavelength-dependent source normalization and port modes should also be checked, especially over broad spectral ranges.
Calculate Reflectance and Energy Balance Correctly
Verify the Reflectance Definition
In a power-normalized port formulation, reflectance may be calculated as
$$
R=|S_{11}|^2.
$$
This expression is valid only when $S_{11}$ represents the reflected amplitude of the correctly normalized incident mode. If several diffraction orders or modes are present, their reflected powers must be summed.
An alternative approach is to integrate the normal component of the time-averaged Poynting vector across the input boundary and normalize it by the incident power. This method is useful for confirming port calculations and for models containing nonstandard excitations.
Reflectance should be dimensionless and should normally remain between zero and one for a passive single-input system. Values outside this range usually indicate incorrect normalization, sign conventions, integration orientation, or mode decomposition.
Check the Electromagnetic Energy Balance
For a passive model,
$$
R+T+A=1,
$$
where $R$ is reflectance, $T$ is transmittance, and $A$ is absorptance.
The absorptance may be calculated from
$$
A=1-R-T
$$
or by integrating the volumetric electromagnetic loss in the metal and other absorbing materials. Agreement between these methods provides a valuable consistency check.
Near the SPR condition, reflectance decreases because incident power is coupled into the plasmonic field and dissipated primarily through metal absorption. If the reflectance minimum is not accompanied by increased absorption or interface field enhancement, the dip should not automatically be interpreted as SPR.
Small deviations from unity may arise from numerical discretization. Large deviations indicate mesh insufficiency, boundary reflection, incorrect port power, omitted outgoing modes, or material-sign errors.
Extract Meaningful Sensor Performance Metrics
Resonance Angle and Angular Sensitivity
The angular resonance position is commonly defined as the angle of minimum TM reflectance,
Angular sensitivity is then calculated as
$$S_{\theta}\frac{\Delta\theta_{\mathrm{SPR}}}
{\Delta n_a}$$
where $\Delta n_a$ is the change in analyte refractive index. The sensitivity is usually expressed in degrees per refractive-index unit.
For nonlinear sensor responses, a single sensitivity calculated between widely separated analyte indices may conceal local variation. A more rigorous approach is to calculate the local slope,
$$S_{\theta}(n_a)\frac{d\theta_{\mathrm{SPR}}}
{dn_a}$$
A regression model can be used when several analyte values are available. The reported range, fitting method, and coefficient of determination should be included.
Resonance Width and Figure of Merit
Sensitivity alone does not describe how accurately a resonance shift can be detected. A broad dip may show a large shift but poor angular discrimination.
The conventional angular figure of merit is
$$\mathrm{FOM}\frac{S_{\theta}}
{\mathrm{FWHM}}$$
where FWHM is the full width at half maximum of the resonance feature. For a reflectance dip, the phrase FWHM can be ambiguous. It is more accurately treated as a full width at half depth.
If $R_{\min}$ is the minimum reflectance and $R_{\mathrm{base}}$ is the local baseline reflectance, the half-depth level can be defined as
$$R_{1/2}R_{\min}+\frac{R_{\mathrm{base}}-R_{\min}}{2}.
$$
The linewidth is the angular separation between the two points where the reflectance curve crosses $R_{1/2}$. Using $R=0.5$ as the half-depth level is incorrect unless the local baseline is exactly one and the minimum is exactly zero.
Some studies define detection accuracy as
$$\mathrm{DA}\frac{1}{\mathrm{FWHM}}$$
These metrics are convenient for comparing simulated structures, but they do not by themselves provide the practical limit of detection. Experimental detection limits also depend on source stability, detector noise, angular resolution, fitting uncertainty, temperature drift, surface chemistry, and sample-handling variability.
Minimum Reflectance
A low minimum reflectance indicates efficient coupling between the incident field and the plasmonic mode. However, the lowest possible reflectance is not always the best sensor-design objective.
A configuration with an extremely deep but broad dip may provide poorer refractive-index resolution than a slightly shallower but much narrower dip. Conversely, an exceptionally narrow resonance may be difficult to measure if the optical system has substantial beam divergence or limited angular resolution.
Minimum reflectance, sensitivity, linewidth, and field overlap should therefore be optimized together. Reporting only the highest sensitivity usually provides an incomplete view of sensor performance.
Field Enhancement and Sensing Depth
The electric-field distribution should be evaluated at the resonance angle and compared with an off-resonance angle. True SPR excitation should produce strong field localization at the outer metal-analyte interface and exponential decay into both media.
A useful enhancement measure is
$$\eta_E\frac{|E|^2_{\mathrm{interface}}}{|E_0|^2}$$
where $E_0$ is the incident-field amplitude under a clearly defined normalization.
The location used for evaluating the field must be specified. The global maximum may occur at a sharp corner, a mesh irregularity, or within an intermediate layer rather than within the analyte. For sensing analysis, field intensity and modal overlap within the analyte are generally more meaningful than the absolute global maximum.
The sensing depth is related to the field penetration into the analyte. Bulk refractive-index sensitivity depends on the complete evanescent field, whereas sensitivity to a thin molecular layer depends strongly on the field intensity within the first few nanometers from the metal surface.
Optimize the Metal Thickness Systematically
The metal thickness controls the coupling strength between the prism evanescent field and the SPP at the metal-analyte interface. If the film is too thick, the evanescent field is strongly attenuated before reaching the sensing interface. The reflectance dip becomes weak or may disappear.
If the film is too thin, the mode may become strongly radiative, the dip may broaden, and transmission through the metal may increase. Very thin films may also be experimentally discontinuous, even if the numerical model treats them as perfectly uniform.
The optimum thickness depends on the metal, wavelength, prism index, analyte index, adhesion layer, and any additional coatings. The thickness should therefore be swept over a physically fabricable range while recording resonance angle, minimum reflectance, linewidth, figure of merit, and interface field enhancement.
The optimum value for one performance metric may not be optimum for another. A thickness that minimizes reflectance may not maximize the figure of merit. The final selection should be tied to the intended measurement system and fabrication capability.
Distinguish Bulk Sensing from Surface-Layer Sensing
Many simulated SPR sensors are evaluated by changing the entire analyte refractive index. This describes bulk refractive-index sensing. It is appropriate for concentration monitoring, liquid composition measurements, or environmental sensing when the refractive-index change extends throughout the evanescent field.
Biosensing often involves adsorption of a molecular layer at the functionalized metal surface. Such a layer can be modeled as a finite-thickness dielectric region with refractive index $n_b$ and thickness $t_b$. The resonance shift can then be evaluated as a function of $t_b$, $n_b$, surface mass density, or binding state.
The surface response is usually nonlinear with layer thickness because the plasmonic field decays away from the metal. A simplified dependence may be represented as
$$
\Delta\theta_{\mathrm{SPR}}
\propto
1-\exp
\left(
-\frac{2t_b}{\delta_d}
\right),
$$
where $\delta_d$ is an effective intensity-decay length in the analyte.
A model claiming biosensing performance should therefore not rely exclusively on bulk refractive-index sensitivity. Both bulk and surface-layer responses may be reported, but they should be identified clearly.
If you're working on related challenges in this area and would find guidance helpful, feel free to reach out: CONTACT US.
Validate Every Claimed Improvement
Compare Against the Conventional Reference Structure
When proposing a modified SPR sensor, the new structure should be compared against a conventional prism-metal-analyte sensor using identical simulation conditions. The prism, wavelength, analyte range, metal dataset, angular resolution, fitting method, and performance definitions should remain consistent.
Changing both the sensor structure and the evaluation method makes it difficult to identify the true source of improvement. For example, a modified structure should not be compared using a finer angular grid than the reference structure.
The baseline and modified reflectance curves should be shown together whenever possible. This allows the reader to evaluate whether the reported sensitivity improvement is accompanied by broader linewidth, higher minimum reflectance, secondary resonances, or a reduced usable dynamic range.
Verify the Phase-Matching Trend
Increasing the analyte refractive index generally increases the SPP propagation constant. A larger tangential incident wavevector is then needed, producing a shift in the resonance angle or resonance wavelength, depending on the interrogation method.
If the simulated resonance shifts in an unexpected direction, examine whether the tracked dip is the same physical mode across all parameter values. Multilayer structures can contain several resonances, avoided crossings, guided modes, or hybrid plasmonic modes. Automatically selecting the global reflectance minimum may cause the analysis to jump from one branch to another.
Mode tracking should consider resonance continuity, field distribution, absorption location, and interface localization rather than reflectance magnitude alone.
Conduct Numerical and Physical Robustness Tests
A credible simulation should be stable against reasonable changes in mesh size, domain dimensions, boundary placement, solver tolerance, and angular resolution. It should also be tested against realistic fabrication variations.
Metal thickness may vary by one or more nanometers across a deposited sample. Adhesion-layer thickness, surface roughness, coating thickness, prism index, analyte temperature, and optical constants also carry uncertainty.
A tolerance analysis can reveal whether an apparently optimized structure is robust or depends on unrealistic parameter precision. If a one-nanometer thickness variation destroys the claimed improvement, the design may have limited experimental value despite excellent nominal performance.
Account for Temperature and Experimental Conditions
Temperature changes the analyte refractive index, prism refractive index, metal permittivity, film dimensions, and potentially the binding chemistry. For aqueous sensing, the thermo-optic response of the liquid can produce a measurable resonance shift even when the target concentration remains unchanged.
A temperature-dependent model may use
$$
n(T)=n(T_0)+
\frac{dn}{dT}(T-T_0),
$$
provided that the thermo-optic coefficient remains valid over the selected range.
The thermal expansion of a nanometer-scale metal film is often a secondary effect compared with the thermo-optic response, but it may become relevant in high-temperature plasmonic systems. Temperature-dependent metal losses can also broaden the resonance.
For experimental comparison, the simulation should use the actual operating wavelength, temperature, prism material, deposited-film thickness, adhesion layer, analyte composition, and optical geometry. Agreement should not be forced by adjusting several uncertain parameters simultaneously without independent measurements.
Avoid Common Modeling Errors
Using Lossless Metal Properties
A lossless negative permittivity may support an idealized surface mode, but it cannot reproduce the absorption-driven reflectance dip of a realistic SPR sensor. Complex metal permittivity is essential.
Selecting the Wrong Polarization
TE excitation should not produce conventional planar SPR. Always inspect the electric and magnetic field components to confirm TM excitation.
Measuring the Angle from the Surface
Optical incidence angles are normally measured from the surface normal. Using the complementary angle shifts the entire sweep and can lead to incorrect comparison with published results.
Using Too Few Elements Across the Metal
An apparently converged prism mesh does not ensure that the nanometer-scale metal film is resolved. The thin layers require dedicated local or mapped refinement.
Ignoring Material Dispersion
Constant refractive indices may be acceptable for a narrow angular sweep at one wavelength. They are generally unsuitable for a broad wavelength sweep.
Applying Reflective Boundaries to Open Regions
Perfect electric conductor or magnetic conductor boundaries can produce artificial standing waves. Use ports, periodic conditions, scattering boundaries, or PMLs as appropriate.
Treating the Deepest Dip as the SPR Mode
Thin-film interference, guided modes, and hybrid resonances may also create reflectance minima. Verify the field localization and absorption mechanism.
Reporting Sensitivity Without Linewidth
A high angular shift does not guarantee high resolution. Sensitivity should be considered with FWHM, figure of merit, minimum reflectance, and practical measurement constraints.
Using an Arbitrary Half-Maximum Definition
For a reflectance dip, the linewidth should normally be calculated at half depth relative to the local baseline, not automatically at $R=0.5$.
Ignoring Thin Adhesion and Functional Layers
Nanometer-scale layers can significantly alter the response because they occupy the highest-field region of the sensor.
Optimizing Unrealistic Thicknesses
A mathematically optimal film may be too thin to fabricate as a continuous layer or too sensitive to manufacturing variations. Fabrication constraints should be incorporated during optimization.
A Reliable Simulation Workflow
A disciplined workflow begins with a three-layer prism-metal-analyte structure at a fixed wavelength. The metal is assigned a verified complex optical dataset, the prism uses appropriate dispersion data, and the analyte is initially represented as a lossless homogeneous dielectric.
An analytical estimate of the critical angle and phase-matching angle is then calculated. A transfer matrix model is used to obtain an initial reflectance curve and identify the likely resonance region.
The full-wave model is constructed using TM plane-wave excitation, periodic or Floquet lateral boundaries, and suitable open boundaries along the propagation direction. The metal and adjacent interfaces receive dedicated mesh refinement.
A coarse angular sweep identifies the reflectance dip. A refined local sweep, followed by interpolation or numerical minimization, determines the resonance angle. The result is checked against the transfer matrix prediction.
The electric-field distribution, absorptance, and energy balance are then examined. A genuine SPR condition should show localized field enhancement at the metal-analyte interface and increased dissipative loss in the metal.
Mesh density, boundary distance, domain thickness, and angular resolution are varied independently to demonstrate numerical convergence. The analyte refractive index is then changed over the required sensing range, and the resonance is tracked consistently for each case.
Only after the baseline model is validated should additional layers or patterned features be introduced. Every modification should be compared with the same reference structure using identical material data, numerical resolution, and metric definitions.
Finally, realistic uncertainties in film thickness, adhesion layers, optical constants, temperature, and fabrication tolerance should be evaluated. This last stage frequently determines whether a theoretically improved sensor is experimentally practical.
Conclusion
Accurate simulation of a prism-based SPR sensor depends less on constructing a visually elaborate geometry and more on preserving the correct electromagnetic coupling mechanism. The model must use TM polarization, complex dispersive metal properties, a correctly defined internal incidence angle, sufficient resolution through the metal film, and boundaries that represent an optically open structure.
The resonance should be confirmed through reflectance, absorption, energy balance, and interface-localized field enhancement. Transfer matrix calculations provide an essential benchmark for planar models, while full-wave techniques are most valuable when finite dimensions, patterned surfaces, nonuniform coatings, roughness, or coupled physical effects are introduced.
Sensor optimization should not focus on sensitivity alone. Resonance linewidth, minimum reflectance, figure of merit, sensing depth, fabrication tolerance, material uncertainty, and experimental angular resolution must be considered together. A model that remains physically interpretable and numerically stable under these tests provides a credible basis for sensor design, publication, and experimental development.
Our research is publishing a reputed journal. We are also open for collaboration. You can read our publication here
Check out YouTube channel, published research
you can contact us (bkacademy.in@gmail.com)
Interested to Learn Engineering modelling Check our Courses 🙂
--
All trademarks and brand names mentioned are the property of their respective owners.The views expressed are personal views only.
