Effects of thickness and roughness on plasmonic characteristics of gold thin films deposited on polished optical fiber

The thickness and roughness of metal layers substantially affect the performance of surface plasmon resonance (SPR)-based sensors. The deposition methods, control parameters, and substrate characteristics influence the layer thickness and roughness. This study investigates the SPR characteristics of a polished optical fiber surface coated with gold (Au) metal of different thicknesses. The Au layer is deposited via the thermal evaporation method, and its thickness is varied by controlling the deposition time (3–6 min). A proportionality relationship between thickness and deposition time is observed. Island-shaped structures in gold (Au) morphology are formed due to low adhesion to the substrate. The shape of this island creates gaps in the layer, causing scattering. In addition, the roughness on the gold surface triggers the Localized Surface Plasmon Resonance (LSPR) phenomenon. As a result, the measured dielectric characteristics differ from the reference. The SPR curve calculation simulation was carried out based on reference optical parameters and measurement results by an ellipsometer, which were then compared with experiments. The obtained results show that the substrate roughness, morphology, and thickness of the Au layer play an essential role in determining the characteristics of the SPR curve in a fiber optic plasmonic sensor. As a result, in basic experiments, the sample with an Au thickness of 27.37 nm (deposition time = 3 min) shows better characteristics (half-maximum full width, minimum transmittance, and resonance wavelength) compared with the sample with an Au thickness of 53.97 nm (deposition time = 4 min), Although 53.97 nm is the optimal thickness from the simulation using reference optical parameters (smooth substrate surface and smooth gold layer).


Introduction
Thin nanostructured metal layers have attracted broad interest for their potential applications in plasmonic technology [1,2].Surface plasmon resonance (SPR)-based sensors are among numerous plasmonic technologies actively developed in recent decades [3].Surface plasmons (SPs) are coherent oscillations of free electrons at the metal-dielectric interface.SPR describes the strong absorption phenomenon of transverse magnetic light caused by resonance, which occurs when the wave vector of the incident light matches that of the SPs [4,5].SPR-based sensors are reported to be highly sensitive and enable real-time detection of heavy metals [6] and biomolecules [7].These sensors are usually designed in the Kretschmann configuration, in which a thin metal layer on a glass prism generates an evanescent field under light illumination [8].Gold (Au) is the preferred thin metal layer in sensor fabrication owing to its high chemical stability and excellent detection performance [9].In some studies, prisms have been replaced with fiber optics to improve the versatility of SPR-based sensors.Such sensors are called fiber optic plasmonic sensors (FOPS).The evanescent fields in FOPS are generated via electromagnetic wave guidance in the fiber optic core.They then interact with SPs in the metal layer deposited on the optical fiber [10].
To optimize the sensor performance, many researchers have modified the morphological properties of the metal layer.Such investigations are extensively documented in both theoretical and experimental studies.In one approach, the optical fiber structure of the plasmonic sensor is modified by polishing [11].After polishing, the inhomogeneous metal layer morphology caused by the cylindrical structure of optical fibers becomes a flat optical fiber surface.In fact, different polishing procedures can generate diverse characteristics and exceptionally diverse surface roughness [12].Numerous studies have reported the thickness of Au deposition in the fabrication of SPR sensors for polished optical fibers is 50 nm, which is determined based on the optimum thickness of a prism-based configuration.However, the impact of surface roughness on polished optical fibers is frequently overlooked [3,10].
The surface roughness of polished optical fibers still requires serious attention.According to several theoretical and experimental articles on prism-based configurations, the characteristic of the SPR curve is affected not only by the thickness [13] and homogeneity [14] of the metal layer but also by the roughness of the layer [15].The methods and control parameters implemented during the deposition of the thin metal layer also play an important role [16].Furthermore, when the deposition parameters are adjusted to form layers of identical thickness, distinct morphological characteristics caused by the deposition method and control parameters will affect the SPR.
The frequency shift of the SPR is highly correlated with morphological modifications of the layer, implying changes in the matter-light interaction behavior (absorption, reflection, and transmission).This phenomenon can be understood through the complex dielectric constant of the material, which is composed of real and imaginary components representing light deflection and absorption, respectively [17].The association between the complex dielectric constant of plasmonic materials and layer morphology, particularly thickness, has been reported elsewhere.It was found that both the real and imaginary components of the complex dielectric constant of Au decrease with increasing thickness of the Au layer.As the layer thickness increased from 20 to 50 nm, the average crystal size and sensor performance increased proportionally [18,19].
In this study, SPR-based sensors were fabricated by coating Au as the plasmonic material on a polished optical fiber substrate through thermal evaporation deposition.In this simple deposition technique, the film thickness can be precisely controlled by setting the chamber pressure, filament current, and exposure time [20].
Here, the deposition time (metal exposure time) of the Au layer was altered to achieve different thicknesses at the same deposition rate.The deposition time was varied by adjusting the vacuum pressure and filament current.As shown in theoretical studies, SPR is generated within the 20-100 nm range of metal thicknesses [13].Consequently, the deposition times were controlled to give thicknesses within this range.This research aims to analyze the effect of substrate surface roughness and the morphological properties of the thickness resulting from thermal evaporation deposition with different deposition times on the SPR curve characteristics on polished fiber optic substrates.

Method
Multimode optical fibers were produced by Molex USA (No: 538-106803-1004).The fibers' core/jacket diameter ratio and numerical aperture were 600/1040 and 0.37, respectively.The fibers were polished using the wheel polishing technique illustrated in figure 1(a).The tip of each fiber was placed in a homemade holder, and the polishing width and depth were controlled with an adjustable displacement platform.Real-time monitoring of the polishing process is facilitated using a microscope (HiView).Using 5000-grit abrasive paper rolled onto a polishing wheel, the fibers were polished to 300 μm and 1.5 cm depth and width, respectively (figure 1(b)).Before deposition, the polished optical fiber section was washed with acetone (purity: 99.5%) by Smartlab, followed by distilled water; this washing cycle was repeated three times.
The Au metal (purity: 99.9%) by SDGM Inc. was deposited using the thermal evaporation (Emitech OM-E6500MHVE).The deposition time was varied to 3, 4, 5, and 6 min.Besides that, the deposition rate was kept constant and adjusted to achieve a certain layer thickness between 20 to 100 nm by adjusting the vacuum pressure (2 × 10 -4 mbar) and the filament current (8 A ).
The morphology of the gold layer on the polished fiber surface was examined using a scanning electron microscope (FEI: Inspect-S50).The metal layer thickness (t) was determined using the Beer-Lambert law as follows: Where I and I 0 signify the incoming and transmitted light intensities, respectively.Transmission measurements were conducted using a digital optical power and energy meter (PM100D Thorlabs) on Au-coated glass slides under probe conditions.The laser was a He-Ne laser with a wavelength of λ = 633 nm.At this wavelength, the absorption coefficient of Au is α = 681390 cm -1 [21,22].
The SPR-based sensor was characterized as depicted in figure 2. The terminal of each sensor probe was connected to a light source (OSL2 Fiber Illuminator) and a spectrophotometer (Ocean Optics USB 4000) through a fiber adaptor.The SPR curve (T SPR ) characteristics were investigated in the 400-800 nm wavelength range.It was calculated from the optical fiber transmittance data before deposition (R λ ) and after deposition (Sλ) obtained from the sensor system using the equation:

( )
The roughness surface was observed through both the glass slide and polished optical fiber using a topography measurement system (Polytech TopMap) and was calculated as the roughness average (Ra), representing the average deviation of the peaks and valleys over the measured length of the surface.The sample measurement length for Ra is 1 millimeter.

Results
The deposition time was varied as 3, 4, 5, and 6 min to obtain Au films with various thicknesses.The prepared samples were labeled S1, S2, S3, and S4, respectively (see table 1).Thickness increases with deposition time.This increase in thickness also indicates an increase in the amount of gold (Au) detected via SEM-EDX mapping (figure 3). Figure 3 shows the morphological characteristics of the gold layer.The layers form an island-like structure, resulting in small gaps between the Au (gold) consistently observed throughout the layers in all samples.As deposition time increases there is a striking trend showing the enlargement of this gap as deposition time increases.Figures 4 and 5 show the surface profiles of the glass slide and Au-coated polished optical fiber, respectively.Figure 1(c) shows that the polishing method has successfully achieved a flat surface with the Ra value in table 2. The measured roughness values differed between the surface of the glass slide and the polished optical fiber coated with Au due to the initial differences in surface substrate roughness.
The ellipsometer data were adjusted using the thickness and roughness determined by the Beer-Lambert law and TMS as adjustment parameters, showing high agreement, as seen in figure 6.The agreement indicates that the Beer-Lambert law method effectively measures the thickness of the gold layer.
As the layer thickness increased from S1 to S3, the real and imaginary components of the dielectric constant decreased and converged towards the reference value, as reported in several previous studies [18,19].However, these trends reversed in S4 (figure 7(b)).Examining the data in table 2, one observes that S3 and S4 possess very similar thicknesses but distinctly different roughness levels.Based on the optical parameters reffer by Johnson and Christy, and the results obtained from ellipsometry characterization, the transmittance curve was calculated using the equation below [23]: T(λ) represents the transmittance normalized for wavelength, λ is the wavelength of the light source, and n eff is the effective index of the surface plasmon mode.L is the length of the sensor region.Calculations are carried out based on the dimensional construction shown in figure 1(b).SPR curve characterization is carried out by observing parameters such as full-width at half maximum (FWHM), Resonance Wavelength, and Resonance Transmission intensity.The FWHM value is determined through curve fitting using a Gaussian Amplitude approach (table 3).
Figure 8(a) depicts the Surface Plasmon Resonance (SPR) curve obtained under optimal conditions, depicting the gold layer and the flat and smooth surface of the optical fiber.Meanwhile, figure 6(b) illustrates the SPR curve produced by a layer of gold, with the measured optical parameters, coated on a smooth and flat optical fiber surface.Meanwhile, figure 6(c) depicts the SPR curve from the experimental results.The simulation results generally show good agreement, with resonance wavelengths ranging from 469.7-508.6 and transmission resonances between 41.9%-49.4%.The optimal thickness for Au resonance has been determined as 50 nm [3].At this thickness, resonance is expected to reduce the transmittance to zero.However, in this study, the resonance does not reach zero because the simulation and experimental configurations use one side of the polishing area.The condition causes light (transverse magnetism) to interact with the layer only partially, and light moving perpendicular to the gold layer will be transmitted.

Discussion
Although polishing methods yield a flat substrate surface, they introduce surface roughness.The roughness degree of the polished surface directly depends on the grit size of the abrasive paper [24].Within this context, the Au layer on the polished fibers is rougher than that deposited on glass substrates because the fiber substrate is  inherently rougher than glass substrates.Even under the same deposition conditions, a layer on a rough surface will possess a coarser texture than a layer deposited on a smooth surface [12].
The layer structure that forms islands, as observed via SEM (figure 3), occurs because gold (Au) has low surface energy, resulting in low adhesion to the substrate [25].Therefore, gold tends to interact and stick to each other, forming clusters and islands.In other words, gold exhibits high cohesion.Cohesion higher than adhesion to the substrate ultimately forms gaps in the morphology of the gold layer.The increase in the amount of gold (Au) detected through SEM-EDX mapping explains that the gap becomes larger as the deposition time increases due to interactions between many gold atoms.
Meanwhile, the real parts of the dielectric constants already demonstrated metallic behavior in the tested samples as per the reference.Absorption by the LSPR phenomenon causes the larger imaginary part (responsible for light absorption).The layer gaps formed in the morphology cause light scattering.
Localized Surface Plasmon Resonance (LSPR) is a phenomenon in which the polarization of a material absorbs light because its size is smaller than the wavelength of the light involved [10].Thus, thin layers tend to produce high imaginary epsilon values, as shown in this study.In S4, the increase in the imaginary part arises from the small peak area of a sample with increasing smoothness, triggering LSPR.The trend was also reported in another study, where the study noted an increase in the imaginary component of the dielectric constant as the thickness decreased in the range of 14 to 49 nm [19].
Under ideal conditions (figure 8(a)), where the surface of the gold layer and substrate is set to be smooth, a decreasing trend in the FWHM value is observed, which is proportional to the thickness of the gold.The decreasing trend occurs because the thicker the layer, the fewer evanescent waves penetrate the gold surface due to the narrowed penetration depth.Meanwhile, there is a change in the resonance phase, which can be seen from the shift in the resonance wavelength.Changes in the resonant phase by the influence of layer thickness are described by the Airy equation [25].
Figure 8(a), In the comparison between S1 to S2 and S2 to S4, a significant phase change can be seen at S1 to S2, which then tends to be constant at S2 to S4.The change is caused by the thickness of S1, which is below optimal (50 nm).This behavior was also reported by Benaziez et al in simulations on silver metal [13].
The overall resonance wavelength in figure 8(b) is shifted in a smaller direction compared to figure 8(a).Meanwhile, in the results of figure 8(c) (experiment), there is a shift in a greater direction compared to figure 8(a).The resonance phase shifting towards larger wavelengths means that the resonance phase occurs when the penetration depth of the evanescent wave is deeper, which means the sensor is more sensitive and vice versa [26].
Uneven surfaces cause phase differences due to different angles of incidence of the normal line.In general, the contribution of roughness (Δk rough ) to the corresponding evanescent wave phase (k ew ) and surface plasmons (k sp ) is formulated by the equation [27]: In the development of SPR sensors, achieving a high resonance wavelength with a narrow full-width at half maximum (FWHM) and maintaining minimum transmittance is the ideal condition to be achieved [3,10].But as seen in figure 8(b) compared to figure 8(a), it can be seen that the shift in the resonance wavelength becomes smaller in proportion to the decreasing FWHM value.Likewise, in figure 8(c), the shift to a larger resonance wavelength makes the FWHM wider.The FWHM value is inversely proportional to sensor accuracy, the wider the FWHM, the sensor accuracy decreases [28].The width of the FWHM at high resonance wavelengths is caused by the reduced propagation length of surface plasmons at large wavelengths, while the penetration depth of evanescent waves is deeper [29].In figure 8(a), the sample with a thickness of 53.97 nm (S2) shows good performance in terms of Full Width at Half Maximum (FWHM), minimum transmittance, and resonance wavelength.This may occur because the thickness of S2 is close to the optimal thickness (50 nm) determined by the prism configuration of the Surface Plasmon Resonance (SPR) sensor [3].However, the experimental results show that S1 outperforms S2.Moreover, in figure 8(b), the SPR signal on S1 is not visible or recognized.The SPR propagation length is directly proportional to the ratio (|em1|/em2) [30], where em1 and em2 are the real and imaginary parts of epsilon for metal, respectively.This explains that S3 (figure 8 ), the ratio (|em1|/em2) is high, while the penetration depth is narrow, and the FWHM width is high, so the SPR curve is difficult to recognize.

Conclusion
Our study revealed distinct differences in the SPR curves of Au deposited at different thicknesses (27.43-69.67nm, achieved by varying the deposition time) on a roughened polished fiber surface.This study confirms the need to carefully consider the thickness, morphology, and roughness aspects of the Au layer as well as the surface roughness of the polished optical fiber to optimize the performance of the polished FOPS.However, it should be noted that this research has not extensively explained the specific contribution of the formed roughness pattern to the phase shift (Δk rough ), thus requiring further research.

Figure 1 .
Figure 1.a) Set up wheel polishing technique b) schematic of size and dimensions of the polished fiber area c) photograph polished optical fiber before gold film deposition.

Figure 2 .
Figure 2. Characterization system for the SPR sensor.

Figure 7 .
Figure 7. Dielectric constants of the thin Au layers on Samples S1-S4 and the reference sample: a) real and b) imaginary components.

Figure 8 .
Figure 8. SPR curves of Samples S1-S4 in air (n air = 1) a) simulation based on reference, b) simulation based on ellipsometry fitting, c) Experiment.

Table 2 .
Surface roughness of polished optical fiber and glass slide coated with au layer.
Surface roughness glass slide samples (nm)Surface roughness of polished optical fiber samples (μm)

Table 3 .
Characteristics of the SPR curves.