Self-Assembled Pd Nanocomposites into a Monolayer for Enhanced Sensing Performance

Abstract

To date, the advanced synthetic approaches for palladium nanoparticle-based catalysts involve multistep, toxic and high-cost fabrication routes with low catalytic and sensing performances. In this work, we introduce a new one-shot approach to produce highly sensitive Pd nanocomposites using a large-area polymer self-assembly strategy. This synthesis method allowed us to control the Pd nanoparticle shape and to tailor their plasmonic band positions in a wide light spectral range from ~350 to ~800 nm. We thus determined the critical synthesis conditions that give rise to a ringlike morphology in a reproducible manner. There is no need for a reducing agent and preliminary functionalization of the surface supporting the nanoparticles upon synthesis. To the best of our knowledge, few works have demonstrated the good performance of PdNPs in sensing. Here, we have demonstrated a robust SERS response for 4-mercaptopyridine with an enhancement factor of 4.2 × 10⁵. We were able to exceed this high value, which matches the current maximum found in the literature, by decreasing the gap distances between Pd nanorings in relation to the high density of hotspots and the exacerbation of the coupling effect between PdNPs. These tailored products provide new insights for the use of Pd nanomaterials in photocatalysis applications, according to the well-established catalytic performance of Pd materials obtained in this work.

1. Introduction

Recently, there has been large interest in the fabrication of ordered 2D arrays of nanoparticles due to the coherent interparticle interactions through localized surface-plasmon resonance within them [1,2]. When the interparticle distance approaches the wavelength of the interacting light, far-field radiative coupling heavily alters the plasmon resonance energy and width. In linear nanoparticle chains with the interparticle distance comparable to the single particle resonance wavelength, a narrow plasmon mode polarized perpendicular to the nanoparticle chain emerges due to far-field interference. As the interparticle distance is decreased to quantities equal to or less than the nanoparticle diameter, the near-field interactions become significant, and in linear chains, it splits the plasmon resonance into two propagating plasmon modes polarized parallel and perpendicular to the principal chain axis [3,4,5,6,7,8]. The production of arrays with homogeneous particle size and shape, as well as regular interparticle distance, using different techniques has provided profound insights into the far- and near-field interactions of surface plasmons. The plasmon resonance redshift due to near-field interactions has already been utilized in colorimetric technology for detecting molecular recognition events with DNA-functionalized gold nanoparticles [9]. Distance-dependent plasmon coupling can be used in plasmon rulers for the measurement of the size of biopolymers in solution [10]. In recent years, the importance of the linear arrangement of nanoparticles in the ring form and its impact on plasmon coupling has been investigated because of their potential applications as optical, magnetic and electronic resonators and sensors [11,12,13].

Ring-like nanostructure (RLN) properties can be simply varied by their diameter and wall thickness, which make them the subject of many theoretical and experimental studies [14,15]. For example, plasmonic nanorings have been exploited for ultrasensitive bio- and chemical sensing due to the simple tuning of their plasmonic resonance band [16,17]. In addition, nanoscale ring resonators made from semiconductors can be used in the fabrication of nanolasers with fine tunability of the emission wavelength [18,19], and the specific response of magnetic nanorings make them the main candidates for applications in dense information storage [20,21,22].

Heretofore, bottom-up chemical synthesis has provided a tool for the preparation of a wide range of nanostructures, from particles to two-dimensional (2D) structures [23]. Several methods for the production of ordered structures have been introduced, including thiol self-organization, assembly on modified substrates, application of premade templates, etc. [24]. Most of these chemical approaches were not able to precisely control the deposition of metals or metal particles as nanometer-sized ring structures, although fascinating properties are expected for such structures [25,26].

Recently, the template-based approach is common since it is high throughput, simple and low-cost. More specifically, a self-assembly strategy with amphiphilic di-block copolymers has been presented for the synthesis of RLNs. The mechanism of this strategy is based on block copolymer vesicle formation due to inconsistency between polymer diblocks. In this method, a micro-phase separation is employed and self-organized nanoscale polymer particles are created. Spin coating of loaded vesicles with a metallic precursor on a stiff substrate followed by chemical reduction produces monodisperse nanoparticles (MNPs) in the porous polymer film [27,28]. Nevertheless, this strategy includes several steps for MNP preparation and cannot be used for a wide variety of MNPs. Intricacies and limitations of copolymer-templated techniques for the synthesis of specific NPs and the importance of the ringlike nanostructure in sensing applications was a motivation for introducing a general approach for the incorporation of a wide range of metallic salts into a ringlike organized nanostructure using polymer self-assembly.

In this work, we have synthesized monolayer of Pd nanoparticles (PdNPs) with a simple, fast and label-free self-assembly shape-controlled technique in order to develop 2D materials for photocatalysis. Here, we were particularly interested in Pd since it is the plasmonic material with the best catalytic activity. Subsequently, we evaluated our samples as sensors as a first step to highlight their good plasmonic features. In the future, we will focus on their performance in plasmonic photocatalysis reactions. The novelty and innovation of this work relies on the one-step facile synthesis of anisotropic PdNPs with a low interparticle distance in a tunable manner (size and shape) that leads to a good SERS enhancement. In fact, in similar papers, syntheses of Pd are generally complex, multi-step processes and especially are not successful in producing performant SERS substrates.

PdNPs are spontaneously formed by spin coating the Pd precursor–PMMA dispersion on a N-doped silicon wafer due to the rapid evaporation of the volatile solvents from PMMA. Specifically, upon spin coating, the self-assembly of PMMA into micelles containing Pd²⁺ occurs, and then the evaporation of the volatile solvents (PMMA and precursor solvents) leads to nanoporous PMMA film formation, and the PdNPs are localized around the inner walls of the holes.

2. Materials and Methods

2.1. Synthesis Process

This synthesis method is very similar to the one that was formerly used by our group for fabrication of gold nanocubes. This synthesis method is based on self-assembly of PMMA into micelles and vapor-induced phase separation due to evaporation of volatile solvents that are used in the process. N-doped silicon substrate is used in order to reduce the Pd precursor. Silicon plays an important role in providing electrons to Pd²⁺ for PdNPs production. This spontaneous reduction of precursor eliminates the need to use any external reducing agent, so the samples are prepared by a one-step procedure. This stems from the high solubility of metallic salts in alcohol, which is a non-solvent of PMMA. Consequently, multiple sizes of micelles containing palladium ions, ethanol and acetone are distributed on the substrate surface. Evaporation of volatile solvents causes micelle explosion and the formation of PMMA nanoholes containing PdNPs with different average diameters. Here, the PdNO₃/PMMA dispersion was prepared by mixing two immiscible solutions, i.e., poly methyl methacrylate)/methyl isobutyl ketone (PMMA/MIBK) (PMMA mass = 365,000 g/mol, C = 30 g/L) and palladium (II) nitrate dihydrate (>99%)/ethanol. Then, a monolayer of Pd solution was spin coated on a clean silicon wafer in order to form monodisperse PdNPs on the surface. Different concentrations of Pd solutions (20, 40, 60, 80, 100 and 120 mM) were spread on the silicon substrates at different spin coating speeds: 3000, 5000 and 10,000 rpm. All the samples were carefully prepared using 30 s spin coating time and 3000 rpm.s⁻¹ acceleration. In order to obtain a homogeneous layer, the acceleration value must be lower than the spin coating speed. Pd nanoparticle characterization was carried out according to the following explanations.

2.2. Surface Characterization

PMMA thickness measurements were performed using Scan-Asyst Atomic Force Microscope (AFM-ICON), Palaiseau, France, operating in contact and noncontact modes. The probes have a half-cone angle at 200 nm from apex <10, resonant frequency = 330 kHz, spring constant = 42 N/m and tip radius of curvature <5 nm. A scratch was made in our samples in order to expose the Si wafer and achieve an accurate measurement of the depth of the holes. This was accomplished by means of an AFM technique to predict the overall thickness of the PMMA. Line profile was used to measure the film thickness.

SEM images were obtained by a Hitachi SU8030, Tokyo, Japan, operating in secondary electron imaging mode with 5–15 kV accelerating voltage, magnification = 30–250 K, 8 mm working distance and current density of 10 μA. The samples were coated with a 5 nm Pt/Pd metal layer. For EDS analysis, we studied margins of samples that contained Pd agglomerates. For nanoparticle size/shape distribution, 120 SEM images (1100 NPs) were investigated for each sample. The samples were sputtered by carbon since it is an ideal conductive coating for EDS analysis and its single low-energy X-ray peak does not overlap with any other elements.

Optical absorption measurements were performed in air at room temperature using a home-built extinction spectroscope in reflection mode (Olympus optical microscope). An incident beam of a halogen lamp was weakly focused at normal incidence on the substrates during all measurements. A N-doped Si substrate containing a PMMA layer was considered as the reference sample for the extinction test in reflection mode. The reflected light was collected through an optical microscope and focused at the end of an optical fiber. The latter was connected to a spectrometer (Ocean Optique QE65000, Duiven, The Netherlands). The microscope was equipped with a camera and allowed us to align the detection area with the focal spot. For the visualization of the extinction spectra, a 10× objective was used in order to detect light from a large surface area.

2.3. SERS Experiment

Raman and SERS measurements were performed using a Horiba Labram HREvo instrument, Lille, France, operated with a laser (532 nm) and a cooled CCD camera. Both excitation and collection were through a long-distance objective lens (50×/0.5 NA). All the SERS spectra measurements were performed at 1.25 mW laser power during 5 s of acquisition time. The focused light had a 1.52 μm beam diameter. For SERS measurements, a 10 μL drop of 4-mercaptopyridine (4-MPY) was deposited on substrates (on 7 points of the sample) and kept until dried. For Raman measurements, a 10 μL drop of 4-MPY was deposited on an As-doped silicon surface or on PMMA film and the spectra was measured with a 50× objective.

2.4. Simulation Conditions

Plasmonic nanostructure simulation was performed using 3D FDTD simulations (Lumerical solutions, Vancouver, BC, Canada). The commercial software package was installed on a station based on an Intel CPU with 512 GB RAM capacity. The simulated system corresponds to a single PdNPs or a dimer of PdNPs embedded in a PMMA thin layer (50–120 nm range), which is deposited on Si substrate (infinite thickness). We investigated spherical/cubic shapes, and we also studied the size influence by varying the size of PdNPs in the 30–180 nm range. The plasmonic structure was surrounded by air (n = 1) on top. To obtain accurate results, 1 nm mesh size was used in the calculation. The simulation covered a zone of −1000–+1000 nm. The computational time for a single structure was ~4 h. Optical excitation was performed from the top (“+Z” direction) using a broadband plane wave source. The wavelength range of the simulation was set between 200 nm and 800 nm, which is the UV–near infrared light range. The source type was a total-field/scattered-field (TSFT), which is often used for studying the scattering characteristics of nanoparticles. Perfectly matched layer (PMI) boundary was used to calculate the absorption cross-section. In order to record broadband near-field enhancement results, a box monitor was placed close to the structure. The broadband local-field enhancement spectrum was obtained from an integral volume average of |E/E0|² [29,30,31].

Principle of Extinction Properties

The incident source is used to detach two distinct regions from the computational region. One contains the total field (the sum of the incident field and the scattered field) while the second region contains only the scattered field. The scattering cross-section, б_scat, can be represented by the following formula:

б_sca(ω) = P_sca(ω)/I_int(ω)

(1)

where P_scat(ω) and I_int(ω) are the total scattered power and the intensity of the incident source. The total scattered power is obtained by adding the power of the scattered area power monitors. In addition, the absorption cross-section can be calculated using the following equation:

б_abc(ω)=P_abc(ω)/I_int(ω)

(2)

where P_abs(ω) is the total power absorbed by the Pd nanoparticles, which can be obtained by calculating the sum of the power flowing into the monitors located in the full field area. According to the scattering field and total field obtained by numerical simulation, the extinction cross-section is obtained by using the following formula:

б_ext(ω)=б_abs(ω) + б_sca(ω)

(3)

2.5. XRD Analysis

X-ray diffraction measurements (XRD) were performed using a Rigaku SmartLab X-ray diffractometer (Les Ulis, France) with CuKα radiation (λ = 1.54059 Å) in Bragg–Brentano geometry, with a step-size of 0.01° in the range of 10 to 85°.

3. Results and Discussion

3.1. Adjusting Pd Precursor Concentration

Increasing the concentration of the Pd precursor from 20 to 100 mM (for ethanol/MIBK solvents) at medium spin coating speed (5000 rpm) showed an increase in the number of PdNPs, as shown in Figure 1. When the concentration of the Pd precursor was lower than 60 mM, randomly shaped nanoparticles were obtained. For 80 mM, a ring arrangement of cubic/spherical NPS was observed. A high surface coverage by spherical NPs was observed for the 100 mM precursor concentration. The repulsive interaction between the coupled PMMA non-solvent/metallic salt and the PMMA chains resulted in a micro-phase separation. This micro-phase separation showed itself through the emergence of vesicles containing the coupled non-solvent/Pd²⁺, which were then immobilized on the substrate surface after the evaporation of the PMMA solvent and acetone; then, the explosion of the vesicles evaporated the non-solvent so that the rings containing the salts were formed. The strong concentration of the salt around the holes and the formation of the ring configuration of the NPs indicates that the metallic salt will be located at the PMMA/PMMA non-solvent interface of the vesicle. The latter contains ester groups, and since the substrate is a N-doped semiconducting surface, the salt will be reduced spontaneously and will form distributed nanoparticles in rings. All the details on the ring configuration formation of nanoparticles can be found in our previous work [2].

The NP density was relatively constant with increasing Pd precursor concentration from 20 to 40 mM, but it highly increased from 60 mM. A concentration of 100 mM was considered to be a critical concentration of Pd precursor since it yielded a fully covered surface of PdNPs. The increase in Pd²⁺ concentration beyond this value led to the formation of PMMA aggregates inside the solution. The increase in PdNPs due to precursor concentration is attributed to an increase in seed numbers. In fact, a higher precursor concentration provided more nucleation sites and seeds during the reduction process. As shown in Figure 2, this increase resulted in a redshift of the plasmon resonance band from 400 to 505 nm, which can be attributed to the coupling effect due to a decrease in gap distance between nanoparticles. For increases in concentration from 20 mM to 60 mM, an increase in the NP size can be observed, which results in an LSPR redshift, as larger a NP causes increased radiative damping, which then lowers the resonance frequency. On the other hand, since the extinction intensity of samples prepared at 80 mM or 100 mM is higher than those prepared at low precursor concentrations (20 and 40 mM), it might be attributed to the higher NP density for 60–100 mM precursor concentrations, as mentioned earlier. We have also studied the coupling effect for dimers made of PdNPs using FDTD simulations in order to understand the dependence of Pd spherical/cubic (common structure for 80 mM and 100 mM samples) NP plasmonic properties on this effect (Figure 3).

We simulated a dimer made of Pd nanocubes and nanospheres with different gap distances (d). It is well known that the interparticle gap distance plays a key role in the local field enhancement factor of nanoparticle dimers illuminated by a laser. For the excitation polarization perpendicular to the interparticle axis of the two nanoparticles, the interaction between the metal nanoparticle dimers and the incident light is very weak. For this reason, FDTD calculations were carried out to quantitatively investigate the influence of near-field coupling on the extinction spectra of Pd nanosphere dimers, made of particles with a diameter D = 50/100 nm, with the parallel-polarized excitation field (Figure 3a,c). The dimer made of cubic NPs showed a larger resonance wavelength compared to the spherical dimer for the same gap distance and the same NP size, as shown in Figure 3b,d. Cubic dimers are more sensitive to NP size compared to spherical dimers, as a higher redshift of the LSPR is obtained for cubic dimers (from 449 to 800 nm compared to redshift from 350 to 550 nm for the spherical dimer) with the same increase in NP size and the same gap distance (5 nm).

The increase in PMMA thickness (from 50 to 118 nm) as a function of Pd concentration, shown in Figure 4, might be related to an increase in the repulsive interaction between the PMMA solution and the PdNPs. Therefore, PMMA micelles formed on the silicon surface can assemble into larger ones and this leads to a slower evaporation rate of the volatile solvents, so a higher film thickness is obtained. The increase in PMMA thickness with increasing precursor concentration can also be a parameter determining plasmon peaks, as shown in Figure 2. The increase in PMMA thickness leads to a redshift of the LSPR peak due to an increase of the effective refractive index around the PdNPs.

3.2. Solvent Evaporation Rate (ER) Mediated PdNPs Synthesis at High Precursor Concentrations

As described in Section 2.1, PdNPs are formed upon evaporation of the solvents from the PMMA micelles that appear on the substrate surface immediately after spin coating. The ER of the solvents is proportional to the spin coating speed that controls the PMMA thickness and the size of the PMMA nanoholes. The role of this crucial parameter in the synthesis process is illustrated in Figure 5, Figure 6, Figure 7 and Figure 8 for the 80 mM precursor concentration. RLNs were observed for the samples prepared at the 5000 rpm speed. The mechanism of the formation of the hole-structure-containing NPs has been explained in our other work [32], and it has been mentioned that NPs tend to be formed in the inner part of the holes and around the PMMA/hole interface. This tendency is attributed to the precursor assembly at the interface of non-solvent vesicles and PMMA chains after phase separation. However, the ring arrangement of NPs can be observed clearly only for samples obtained from 5000 rpm that contain a homogeneous and medium-size hole. For high ERs, the holes are so small and each hole contains only one or two nanoparticles, so the ring arrangement of the NPs cannot be seen despite NP formation around the inner walls of the holes. At lower speeds (3000 rpm and lower), RLNs cannot be seen clearly because of low the NP density. The hole size distribution is more homogeneous for the 5000 rpm spin coating speed compared to 3000 and 10,000 rpm. At this speed, smaller holes were obtained compared to samples prepared with other spin coating speeds. Due to the faster evaporation of the solvents at high coating speeds, the micelles do not have enough time to coalesce. As a result, small RLNs were obtained above 5000 rpm.

According to these findings, the 80 mM precursor concentration and the 5000 rpm spin coating speed seem to be the reaction conditions which are suitable to prepare ringlike structure with better monodispersity in terms of hole and PdNPs dimensions. Specifically, anisotropic/spherical PdNPs self-assemble into a ring around the inner PMMA hole outline. Thus, the chemical functions that are present in this outline constitute the growth site for PdNPs. The number of these nanoparticles can be then controlled by Pd concentration and the ER, based on the observations in Figure 1 and Figure 5. A narrow size distribution for the PMMA holes at 5000/10,000 rpm was obtained, as shown in Figure 6 (~850 and ~340 for 5000 rpm and 10,000 rpm, respectively), which might be related to an equilibrium between the hole coalescence and the hole formation rate. A lower spin coating speed promotes PMMA hole coalescence in addition to bigger and polydisperse holes. According to Figure 5b, the samples prepared at a moderate speed (5000 rpm) mainly show random shapes of PdNPs, and the 10,000 rpm samples contain mainly spherical NPs. Another major difference between different spin coating speeds is the particle size distribution, as shown in Figure 6. One can also clearly see that the PdNPs size is more homogeneous for the 5000/10,000 rpm samples compared to the 3000 rpm samples. This level of morphology homogeneity leads to a homogeneous LSPR wavelength/SERS enhancement that is highly interesting for SERS applications.

In Figure 7a, the LSPR bands for samples prepared with different speeds have been compared. The maximum wavelength, λmax, of the second plasmonic band is redshifted for the samples prepared at 5000 rpm (~465 nm) compared to those prepared at 3000 rpm (~440 nm) and 10,000 rpm (~455 nm). This result can be attributed to the larger PdNPs that exhibit an anisotropic shape, as seen in Figure 5. For these samples, the nanoparticles were localized around the holes and at close distances from each other on some sites, so it could intensify the coupling effect that is suitable for SERS. For a better understanding of this property, we will discuss it later in Section 3.3. In Figure 7b, a decrease in the PMMA thickness (from 110 to 76 nm) can be observed with the increase in spin coating speed, which is attributed to the repulsive interaction between the hydrophilic Pd nanoparticles and the hydrophobic PMMA, and this interaction is exacerbated at higher spin coating speeds because of the higher density of NPs. It must be noted that a decrease in the PMMA thickness at higher spin coating speeds leads to a blue shift of the LSPR peak while an LSPR redshift can be obtained due to higher a NP number at a higher speed. Therefore, there is competition between the NP density and the PMMA thickness for influence on the LSPR position. For this reason, the importance of the PMMA thickness compared to the NP density was investigated, as shown in Figure 8, based on an FDTD simulation. It must be mentioned that the influence of the PMMA thickness on the LSPR wavelength is not considerable, according to the results of our previous work on gold NPs in the same composite structure [33].

In Figure 8, the redshift of the LSPR peak is observed with an increase in the PMMA thickness. However, this redshift (497 to 536 nm) in the maximum range of PMMA thickness in our method (50 to 120 nm) is not as strong as the redshift due to particle size and density, as shown in Figure 3 (from 322 to 800 nm). Therefore, the PMMA thickness influence on the LSPR position is less important than the shape, size and gap distance in ranges that can be obtained for each of them in our synthesis method. The lower impact of the PMMA thickness might also be attributed to the indirect contact of the PdNPs and the PMMA layer in our composite.

As can be observed in Figure 9, for a high precursor concentration (100 mM), a high density of NPs can be obtained for all spin coating speeds from 3000 to 10,000 rpm. The difference between the samples prepared from different spin coating speeds is related to the NP monodispersity. Despite the narrower NP size distribution at higher spin coating speeds, an inhomogeneity in gap distances is observed under these conditions. One can see that dimer or trimer structures are formed at this concentration. For the samples prepared at 3000 rpm, a dimer made of different nanoparticle sizes can be obtained, which leads to inhomogeneous plasmonic properties. Therefore, a reproducible SERS enhancement cannot be obtained for this type of structure. As seen in Figure 10, the influence of size on the near-field coupling of the Pd nanosphere dimers under the parallel polarization excitation mode was evaluated.

As shown in Figure 10, all the dimers were characterized by two plasmonic bands. The first band was located at 234, 314 and 355 nm for the 100, 150 and 180 nm nanosphere diameters, respectively. The second band was located at ~497, ~700 and >700 nm. This result indicates that both of the LSPR peaks redshift strongly as the NP size increases. Therefore, for a high number of NPs, the importance of homogeneous NP size must be taken into consideration in order to obtain a reproducible LSPR peak and a uniform SERS enhancement.

Energy dispersive X-ray (EDX) was performed for the samples prepared with suitable conditions for the preparation of monodisperse Pd nanorings, i.e., with 80 mM Pd concentration and 5000 rpm. The obtained spectrum is displayed as an inset in Figure 11 and

Table 1. EDS analysis on element content for sample shown in Figure 5b.

Element Line	Net Counts	Weight %	Weight Error	Atom%	Atom % Error	Formula
C K	423	5.7	0.23	20.6	1.8	C
N K	35	1.9	1.2	6	7.2	N
O K	151	6.5	0.7	17.7	4	O
Si K	1693	18.1	0.4	28	1.2	Si
Pd L	2294	67.7	2.2	27.6	1.8	Pd
Total		100		100
C K	423	5.7	0.2	20.6	1.8	C

. Herein, Pd and Si elements are predominantly detected. This EDX analysis confirms the existence of PdNPs on the Si layer. In fact, the very low intensity of oxygen compared to Pd and Si demonstrates that Pd is not present as an oxide in the sample. Another important point in this regard is the stability of the EDS results over time. The element percentages were quite similar at different time intervals from one month to seven months after sample preparation. Therefore, the stability of Pd over time can be concluded.

3.3. SERS Performance

We can conclude from this comprehensive synthetic study that the Pd precursor concentration plays a crucial role in shape control and thus a critical concentration (80 mM) is suitable to obtain uniform nanorings. On the other hand, the role of the ER is to control the number of PdNPs in the PMMA holes, and it allows good tuning of gap distances between them. Since both parameters gave rise to promising SERS platforms, we investigated their impact on sensing the features of PdNPs. Figure 12 shows better SERS enhancement at high ERs, in relation to the presence of the higher number of coupled nanoparticles and the higher intensity of the plasmonic resonance. As a result, closely separated nanoparticles above 5000 rpm led to the formation of a higher number of hotspots that caused a better SERS enhancement.

We must note that Pd has a low SERS efficiency compared to gold and silver due to its low Raman cross-section, so SERS enhancement might be obtained by probes that have good affinity to it. 4-MPY has a good affinity to PdNPs, which allows us to follow the SERS results under different conditions of synthesis [1]. We thus studied the role of morphology on SERS properties by the analysis of samples prepared with different Pd concentrations. From this, we obtained SERS enhancement only for the samples prepared at high precursor concentrations, including 60, 80 and 100 mM. For 60 and 100 mM, we were not able to collect a homogeneous SERS signal from different points of the substrate for 3000 and 10,000 rpm. It is expected that the RLNs of closely separated nanoparticles can exacerbate the coupling effect. The samples prepared with 80 mM showed the lowest SERS enhancement compared to the 60 and 100 mM samples (for the 5000 rpm spin coating speed). This might be attributed to the long distance between the nanoparticles on most zones of the 80 mM samples that cannot satisfy a coupling effect. It is worthwhile to note that the SERS measurement is a collective response from a large excited surface. Therefore, the higher the number of nanoparticles on the substrate, the higher the SERS signal. For this reason, better sensing efficiency was observed when the substrate was covered by the maximum number of hotspots, i.e., at the highest precursor concentration (100 mM) condition, as shown in Figure 13.

Among the samples prepared with different spin coating speeds, the SERS enhancement homogeneity is relatively identical for the 5000/10,000 rpm samples, but it is lower for the 3000 rpm samples. This is attributed to the lower LSPR wavelength homogeneity for the 3000 rpm samples compared with other spin coating speeds, as shown in

Table 2. SERS enhancement/LSPR wavelength homogeneity of samples prepared with different spin coatings for 80 mM precursor concentration.

Spin Coating Speed (rpm)	LSPR Band (nm)	ISERS/INORMAL
3000	440	20
5000	465	25
10,000	455	33

.

As shown in Figure 13, SERS measurements were performed for nanorings obtained at 80 mM (10⁻³ and 10⁻⁵) and a high density of spherical PdNPs at 100 mM with lower concentrations of 4-MPY (10⁻⁵ and 10⁻⁶ M). The measurements of the Raman and SERS signals were performed for 5 s. It can be noticed that for 80 mM at a 4-MPY concentration lower than 10⁻⁵ M, we found no significant 4-MPY signature. However, for 100 mM, the SERS signal could be detected for up to 10⁻⁶ M of 4-MPY. These findings are related to the high surface coverage by PdNPs when we prepared samples with a higher amount of precursor, as previously discussed.

The SERS/Raman peak positions are listed in

Table 3. Assignments and Raman shifts (cm⁻¹) for ordinary Raman and SERS spectra of 4-MPY.

Raman	SERS	Plane	Assignment
429 645 688 721 781 1003 1042 1053 1117 1215 1225 1252 1287 1467 1488 1500 1595 1620 Str = stretching; bend = bending; def = deformation; wag = wagging; Sciss = scissoring; breath = breathing.	425 677 701 716 773 812 996 1009 1032 1051 1088 1214 1275 1450 1487 1575 1605	Out Out Out In Out In Out Out Out Out Out Out In In In In In Out In In In In In In In In In	Wag C2H8, C5H10 Wag C1H7, C4H9 Wag N-H Str R Def R, wag N-H Str R, str C-S Def C-H Wag C1H7, C2H8 Wag C4H9, C5H10 Def N-H+ Wag C-H Wag C-H Ring breath Trigonal ring breath Def R, bend C-H, N-H Trigonal ring breath, str C-S Sciss C-H Bend C-H Sciss C-H Bend C1H7, C5H10, N-H Def R, bend C-H Bend C-H in the same direction Trigonal ring breath, be C-H Def R, bend C-H, N-H, str C-N Wag C-H Str C@C with deprotonated nitrogen Str C@C with protonated nitrogen

with the appropriate assignments which refer to the published literature [34,35].

As shown in Table 3, for 4-MPY, there are two vibrational modes: in-plane modes and out-of-plane modes. In-plane modes are strong, whereas out-of-plane modes are very weak. The 1009, 1088, 1214, 1275 and 1575 cm⁻¹ shifts correspond to in-plane modes but 996 cm⁻¹ is related to the out-of-plane mode. The coordinating sites involved in the adsorption of 4-MPY on metal surfaces are the nitrogen atom, the sulfur atom and the pyridine ring. Here, the interaction via the p-electron of the pyridine ring cannot be considered because there is no spectroscopic evidence of the out-of-plane aromatic ring vibrations [35]. As can be observed in Figure 13 and Table 3, most of the enhanced peaks are assigned to the in-plane vibrations.

A significant downshift (from 1595 to 1575 cm⁻¹ due to the interaction of 4-MPY with Pd metal), a broadening of the ring C=C stretching band (1575 and 1605 cm⁻¹ of SERS) and the existence of a weak peak around 996 cm⁻¹ that is assigned to the (N-H⁺) out-of-plane deformation mode justify a tilted orientation of 4-MPY with respect to the PdNPs. The ratio of the peak intensity of 1575 cm⁻¹ to 1605 cm⁻¹ is about three and it proves that 4-MPY molecules on Pd surfaces are mainly in the N-deprotonation form, while the ratio of deprotonated and protonated 4-MPY for the reference sample (4-MPY in solution) is almost equal. 4-MPY on Pd surfaces might be more inclined to a perpendicular orientation rather than to a flat one because of the strong enhancement of in-plane modes [36,37,38,39,40]. The enhancement factor (EF) of the Si substrate containing PdNPs (obtained from the 100 mM precursor concentration) was estimated by the following equation:

EF = (I_SERS/N_SERS)/(I_Normal/N_Normal)

(4)

where the I_SERS and I_Normal are the intensities of the same bond for the SERS and normal Raman spectra, respectively.

For the peak at 1575 cm⁻¹, the ratio of I_SERS/I_Normal is 78. The SERS and normal Raman peaks are estimated based on the 10⁻⁶ and 10⁻³ M 4-MPY concentrations, respectively.

N_SERS and N_Normal are the numbers of adsorbed molecules on the Si substrates containing PdNPs and the Si reference substrates (without any NPs) within the laser spot area. In this experiment, 10 μL of an aqueous 4-MPY solution was dropped onto the substrates. N_Normal and N_SERS were estimated using the following equation:

N_bulk = Ahρ/M

(5)

where A is the laser spot area (1.81 μm²), h is the laser penetration depth (23.2 μm for 514 nm) and ρ and M are the density (1.161 g/cm³) and molecular weight (111.17 g/mol) of 4-MPY, respectively. Thus, N_Normal = 4.3 × 10⁻¹³ molecules.

N_SERS was calculated as below:

N_SERS = CVA/S

(6)

where C is the 4-MPY concentration (0.001 mM), V is the sample volume (10 μL), A is the laser spot area (1.81 μm²) and S is the sample area (2.24 mm²). Thus, N_SERS = 8.08 × 10⁻¹⁷ mol.

Based on the calculation, EF = (78) × (4.3 × 10⁻¹³)/(8.08 × 10⁻¹⁷) = 4.2 × 10⁵.

For a better understanding of the SERS features, the crystalline properties of the prepared palladium nanoparticles for the sample prepared at 80 mM (Figure 9c) were investigated by X-ray diffraction (XRD). As shown in Figure 14, there are five distinct reflections in the diffractogram at 2θ = 40.48°, 48.72°, 67.8° and 82.3°. These characteristic reflections can be indexed to the face centered cubic (fcc) structure of Pd, according to the JCPDS card 05-0681 [41]. As shown in Figure 14, two peaks related to silicon and palladium were superimposed at 2θ = 48.72°. According to the XRD results, we can confirm the formation of Pd nanostructures and the direct link between their plasmonic characteristics and the obtained SERS enhancement.

The ability to control the synthesis products and their sensing properties is seen in the SERS results, and it seems promising to extend their use to other applications including photocatalysis, lab-on-chip, etc. Yin et al. [42] emphasized the high catalytic effect of MXene-Pd nanocomposites on nitro compounds. This system is quite similar to our self-assembled Pd nanocomposites, which open new perspectives on the use of PdNPs composite platforms in various photocatalysis reactions.

4. Conclusions

In summary, a novel self-assembly route using the physicochemical interactions between the polymer, the metallic ions, the solvents and the substrate surface has been reported for the synthesis of nanoring-like Pd nanostructures. The parameters that play a crucial role in the formation of this specific morphology are spin coating speed and Pd precursor concentration. We demonstrated that a high concentration of Pd precursor and a moderate spin coating speed are required to obtain nanorings with relatively uniform diameters on the whole substrate. The samples prepared with Pd²⁺ concentrations higher than 80 mM, 5000 rpm and MIBK/ethanol solvents demonstrated good performance for the detection of the 4-MPY molecule using SERS. This is due to the high coverage of the substrate surface by the PdNPs and hotspots. FDTD simulations were performed in order to study the influence of different parameters, including NP size, shape, gaps between NPs and the thickness of the PMMA, on the plasmonic features. As a result, decreasing the gap distances between PdNPs has a major contribution to increasing the extinction efficiency and hotspot density by surface unit, in relation to the higher SERS enhancement factor.