*Review* **An Overview of Artificial Olfaction Systems with a Focus on Surface Plasmon Resonance for the Analysis of Volatile Organic Compounds**

**Marielle El Kazzy 1, Jonathan S. Weerakkody 1, Charlotte Hurot 1, Raphaël Mathey 1, Arnaud Buhot 1, Natale Scaramozzino <sup>2</sup> and Yanxia Hou 1,\***


**Abstract:** The last three decades have witnessed an increasing demand for novel analytical tools for the analysis of gases including odorants and volatile organic compounds (VOCs) in various domains. Traditional techniques such as gas chromatography coupled with mass spectrometry, although very efficient, present several drawbacks. Such a context has incited the research and industrial communities to work on the development of alternative technologies such as artificial olfaction systems, including gas sensors, olfactory biosensors and electronic noses (eNs). A wide variety of these systems have been designed using chemiresistive, electrochemical, acoustic or optical transducers. Among optical transduction systems, surface plasmon resonance (SPR) has been extensively studied thanks to its attractive features (high sensitivity, label free, real-time measurements). In this paper, we present an overview of the advances in the development of artificial olfaction systems with a focus on their development based on propagating SPR with different coupling configurations, including prism coupler, wave guide, and grating.

**Keywords:** surface plasmon resonance; olfactory sensors; electronic noses; volatile organic compounds; odorants

#### **1. Introduction**

Over the last few decades, the detection of gases including odorant molecules and volatile organic compounds (VOCs) has attracted great interest and has become increasingly in demand in various field. VOCs constitute a large class of low-molecular-weight (<300 Da) carbon-containing compounds. They can exhibit odorous properties and are characterized by a high vapor pressure (≥0.01 kPa at 20 ◦C) and a high-to-moderate hydrophobicity [1]. These small volatile molecules have a wide range of sources, both natural (plants, animals, bacteria etc.) and anthropogenic (fossil fuels, automobile exhaust gas etc.). The majority of VOCs have inimical effects on human health such as headaches and nose, eye and throat irritation [2]. Consequently, monitoring the nature and concentration of these compounds in indoor or outdoor environments can be very important, and sometimes, vital. Additionally, they can be considered as chemical messengers. In fact, their analysis has been shown to reveal a considerable amount of information. For instance, studies in medical diagnostics have identified gases associated with different diseases such as rheumatoid arthritis, cancer, and schizophrenia [3]. Furthermore, a recent study showed the possibility of detecting viral infections such as COVID 19 through exhaled breath analysis [4]. VOC and odor analysis can also have applications in the food, beverage and fragrance industries for quality assessments. Finally, gas sensing can be very useful for

**Citation:** El Kazzy, M.; Weerakkody, J.S.; Hurot, C.; Mathey, R.; Buhot, A.; Scaramozzino, N.; Hou, Y. An Overview of Artificial Olfaction Systems with a Focus on Surface Plasmon Resonance for the Analysis of Volatile Organic Compounds. *Biosensors* **2021**, *11*, 244. https:// doi.org/10.3390/bios11080244

Received: 17 June 2021 Accepted: 14 July 2021 Published: 23 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

security applications (detection of drugs, explosives etc.), environmental monitoring or other usages under development such as augmented/virtual reality [5]. Nowadays, the gold standard for VOC detection involves the use of trained human or canine noses or gas chromatography coupled with mass spectrometry (GC-MS). Indeed, to control the quality of raw materials or final food and perfume products, industries often have recourse to human sensory panels. Trained dogs are commonly employed for security controls or even for the detection of diseases such as prostate and breast cancers [6,7]. Although very sensitive and efficient for field studies, the use of the biological nose presents several drawbacks. For instance, human panels may yield biased subjective results and are prone to fatigue. Dogs require expensive training and their application fields are limited and sometimes risky. The second method, namely, GC-MS, is a highly sensitive and accurate analytical technique that allows separating, identifying and quantifying different VOCs in a mixture. However, analyses require skilled operators and are time consuming and expensive [8]. Therefore, there is a need for an affordable, reliable, portable and sensitive device that allows for a rapid analysis of gases including VOCs. Such a context has prompted many researchers to work on the development of alternative technology such as artificial olfaction systems that overcome the various drawbacks mentioned above.

Herein, artificial olfaction systems include gas sensors, olfactory biosensors and an electronic nose (eN). A gas sensor or olfactory biosensor is a single-sensor device which is able to detect gases and that consists of a receptor coupled with a transducer and a data processing system. Olfactory biosensors use biomaterials as receptors. On the other hand, as stated by Julian W. Gardner and Philip N. Bartlett in 1994 [9], an eN is "an instrument, which comprises an array of electronic chemical sensors with partial specificity and an appropriate pattern-recognition system, capable of recognizing simple or complex odours". By its very nature, the eN is, in fact, a biomimetic device that replicates the odor discrimination principle of the mammalian olfactory system. Thanks to considerable research efforts on natural olfaction, and especially, the Nobel prize winning work of Linda B. Buck and Richard Axel (1991) [10], we know that, in order to distinguish among a myriad of odors, the biological nose uses cross-reactive olfactory receptors (ORs) (about 400 different types in the human nose). This particular feature of ORs (i.e., cross reactivity or partial specificity) allows each receptor to interact with different odorant molecules with differential affinities. Therefore, in the same manner as barcodes, odors are encoded by a combination of olfactory receptors, which consequently allows the nose to have this large detection spectrum. Moreover, to transduce an olfactory stimulus, the biological odor sensor uses an extensively studied "molecular switch": the G protein. Indeed, Buck and Axel showed that ORs belong to the large family of G protein coupled receptors (GPCRs). They are located in the plasma membrane of the cilia, i.e., the dendritic extrusions of the olfactory neurons projected into the mucus covering the olfactory epithelium. When a VOC binds to an OR, the G protein transduction cascade is initiated and the binding event is converted into an electrical signal processed by the olfactory bulb and deciphered by the olfactory cortex. Figure 1 shows the analogy between biological and electronic noses.

**Figure 1.** Analogy between the biological and the electronic nose (eN). Figure adapted from [11].

The history of artificial odor detection starts in 1920. In their work on spray electricity and waterfall electricity, Zwaardemaker and Hogewind [12] found that the addition of odorant molecules (e.g., phenol, thymol, citrol) to water markedly raised the spray electricity which could therefore be used to detect these molecules. Subsequently, in 1950, Tanyolac and Eaton [13] attempted to detect air contaminants by measuring variations in the surface tension of a liquid drop. They showed that when contaminated air was in contact with a drop of distilled water, mineral oil or water-stabilized mercury, a considerable change in the surface tension of the drop could be observed. Based on their results, they suggested that an instrument able to classify and measure air contamination at low concentrations could be developed. The first prototype of an electronic device capable of detecting odorants was introduced by Hartman in 1954 [14]. The system was based on polarized microelectrodes as sensing elements. Following this, in 1961, using a thermistor as a transduction device, Moncrieff [15] investigated various coating materials (e.g., polyvinyl chloride, cellulose acetate, milk casein) which interacted differently with odorants. He claimed that using an array of sensors with different coatings could broaden the detection spectrum and, thus, allow for the discrimination of a large number of odors. In 1962, Seiyama et al. [16] developed a gas sensor using semiconductive thin films. The gas detection principle of their system was based on changes in electrical conductivity. A similar study was published in 1965 by Buck et al. [17]. In the same year, Dravnieks and Trotter [18] developed a vapor detector based on the thermal modulation of contact potential. Shaver [19] described a method to enhance the sensitivity of a tungsten oxide gas detector by the addition of a catalytic material such as platinum in 1967. The following year, Taguchi fabricated the first metal oxide semiconductor (MOS) gas sensors for home and industrial usage employing tin oxide as sensitive coating material, which he subsequently patented in 1971 [20]. His company, Figaro Engineering Inc., became the main manufacturer of MOS gas sensors. In 1979, Wohltjen and Dessy [21] introduced the first surface acoustic based gas sensor. However, it was not until 1982, with Persaud and Dodd [22], and then in 1985, with Ikegami and Kaneyasu [23], that the first electronic nose systems based on an array of intelligent chemical sensors emerged. In order to understand the discrimination mechanism of the sense of smell, Persaud and Dodd designed a model of the nose using three Figaro sensors with a differential response spectrum. As a result, their device was able to distinguish among a wide variety of odors, and highlighted the importance of nonspecific interactions in the odor discrimination mechanism. As shown in Figure 2, over the following decades, an exponential number of studies were carried out in order to develop gas sensors and electronic noses. Different sensor systems employing chemiresistive, electrochemical, piezoelectric, and optical transducers [24] have been deployed and assembled in an array to construct eN systems.

**Figure 2.** Number of publications on gas sensors and electronic noses since 1982 and the percentage of studies carried out on each type of transduction technique. The data was obtained from Scopus (Keywords used for the histogram: "gas sensor" or "electronic nose" or (gas or vapor or "volatile organic compounds") "sensor array" or multisensor. For the pie chart: "gas sensor" or "electronic nose" or (gas or vapor or "volatile organic compounds") "sensor array" or multisensor and (semiconduct\* or chemores\* or chemires\* or "conducting polymer") or (optical or "surface plasmon resonance" or colorimetric or fluorescen\*) or (acoustic or piezoelectric or gravimetric) or (electrochemical).

To date, most eN systems have used chemical layers (metal oxide semiconductor, polymers, etc.) as sensing elements. However, these systems suffer from limited diversity of sensor coatings and poor selectivity. To improve the odor sensing performance, the latest trend consists of using natural biological elements such as ORs and odorant binding proteins (OBPs) or their analogues, such as peptides as sensitive materials [25,26]. Indeed, the sensitivity and selectivity of such receptors have been naturally improved and optimized by millions of years of evolution, making them ideal candidates for odor detection. However, integrating them into an electronic device and maintaining their bioactivity in nonoptimal conditions is very challenging. Promisingly, great improvements have been made in this novel field of olfactory biosensors and electronic noses [25–30].

A large number of reviews have presented the operating principles of the various sensor systems that have been developed so far for VOC and gas detection [8,24,31–36]. In addition, several reviews have focused on the development of gas sensors and eNs based on the main techniques, namely, chemiresistive [11,37–41], gravimetric [42,43], amperometric [44], optical fibers [45], colorimetric and fluorometric [46]. However, to the best of our knowledge, no review has emphasized the development of gas sensors, olfactory biosensors and eNs based on another popular technique, namely, surface plasmon resonance (SPR). Indeed, SPR offers many advantages compared to other techniques, including label free measurement with quantitative and qualitative data, real-time monitoring with information on the affinity and the kinetics of the studied interaction, compatibility with multiplex and high-throughput analyses, reusable sensor chips, and repeatable measurements. Accordingly, in this review, we aim to first give a brief overview of artificial olfaction systems based on various sensor systems, and then a focus on the advances made using SPR.

After this introduction, the second section will review the most common sensing systems currently employed for VOC and gas detection. The third part will be dedicated to advances in SPR-based gas sensors, olfactory biosensors and eNs. It includes a brief description of the theoretical principles of the SPR technique followed by an overview of research works using SPR with different coupling configurations.

#### **2. Gas Sensors and Electronic Noses Based on Various Sensing Systems**

As stated, many ingenious systems with different types of sensing materials and transduction techniques have been devised and studied. In the following section, we present a brief overview of the most commonly used sensing platforms for VOC and gas detection. For each system, we will underline the transduction principle, strengths, weaknesses, and present some illustrative examples from the literature.

#### *2.1. Chemiresistive Sensors*

This category mainly includes three types of gas sensors, i.e., using MOS, conducting organic polymers (CP) and carbon-based materials [47]. These sensors have a common operating principle whereby the binding of VOCs induces a variation in the electroconductivity. They also have a similar structure that essentially consists of an active layer deposited on a substrate with two electrodes to measure changes in resistance upon exposure to target molecules [39,40,48]. In the following part, popular MOS sensors and CP-based sensors will be discussed more in detail. Gas sensors using carbon material (graphene, carbon-nanotubes, etc.) are not discussed here. More information can be found in recent reviews [49,50].

#### 2.1.1. MOS Sensors

MOS-based sensors are the most commonly used systems for gas and VOC detection among all the sensing technologies [39]. They were first manufactured and marketed by Taguchi in 1968 for gas leak detection [31,35]. These sensors are typically made of a ceramic substrate coated with either n-type (mainly SnO2, TiO2, ZnO) or p-type (e.g., NiO) metal-oxide semiconducting film between two electrodes. The ceramic substrate usually contains a heating element that allows the device to reach its operating temperature, generally ranging between 200 and 500 ◦C [32]. The transduction mechanism of these sensors is based on variations in their conductivity or resistance upon gas molecule binding, which was well addressed in a recently published review [51]. Various factors, such as the bulk resistance, surface effect, grain boundary and contact between the grain interface and the electrode, can affect the electrical properties of gas sensing materials in MOS-based sensors. The detection spectrum and sensitivity of the sensors can be tuned by doping the semiconductor film with noble catalytic metal (e.g., Pt, Pd) [52] or by modifying the working temperature. The grain size, the thickness, and the microstructure and morphology of the coating film can also affect the binding affinity of the device [32,40,53].

These sensors are attractive candidates for eN as they offer high sensitivity with fast response and recovery times. They are also robust and easy to use. Moreover, advances in micro- and nano- fabrication technologies have enabled low-cost production of miniaturized sensor arrays [41,54]. The major drawbacks of these sensors are the lack of selectivity, their susceptibility to humidity and the high operating temperature which leads to high power consumption and reduced lifespan [39,55]. Nevertheless, great efforts have been made to overcome these drawbacks. Low-power microheaters have been designed and new porous structures have been explored [39,54,55]. Moreover, room temperature operating MOS sensors have been developed following different strategies, and involve the use of metal oxide nanostructures such as nanowires, nanotubes and nanobelts [56,57]. MOS sensors and particularly SnO2-based systems have been extensively studied, miniaturized and combined into arrays for the detection of a large panel of VOCs. Hundreds of outstanding works on experimental and commercial eN systems can be cited. However, this not being the subject of the present review, more details about these systems can be found in the cited reviews [11,34,39,41,54,55,58].

#### 2.1.2. Conducting Organic Polymers Sensors

CP based sensors have received considerable attention since the early 1980s [59]. They are probably the most widely used systems for VOC detection after MOS sensors, and were used in the earlier generations of electronic nose systems [35,36]. CP based sensors are generally composed of a substrate (e.g., glass microscope slide, silicon wafer), on which a film of conducting polymer is deposited between two parallel or interdigitated electrodes [31]. Intrinsic conducting polymers (ICPs) such as polypyrrole, polyaniline, polythiophene and their derivatives have been typically employed for sensor applications [37]. They are usually deposited by electro-polymerization [35]. As for MOS sensors, the transduction principle of these devices relies on variations in the conductivity of the sensors in the presence of VOCs. Several studies have investigated the interaction between the ICPs and the target molecules and suggested different mechanisms [37,38,60]. Reversible modulation of conductance is detected by measuring variations in the current flowing through the polymer when a voltage is applied across the electrodes [31]. The sensing performance of the CPs can be adjusted by modifying the polymer molecular structures, changing the dopants and incorporating a second component into conducting polymers [37]. The addition of a second component gives rise to an original new category of sensing elements called hybrid or composite conducting polymers (CCPs). Further information and examples of CCP-based sensors can be found in the following reviews [37,61].

Unlike MOS sensors, CP-based systems can operate at room temperature, and thus, consume less power. They also exhibit good sensitivity and have short response times [37]. In addition, they are easy to fabricate and resistant to poisoning [8,24]. However, these devices suffer from a lack of selectivity and baseline drift. Moreover, their sensitivity can be affected by humidity and temperature and they can be overloaded by some VOCs resulting in a short lifetime [24,35,54]. Hundreds of papers about CP-based sensors and eNs can be found in the literature [37,38,54]. CP-based gas sensor arrays have been developed for many applications. For instance, Yu et al. have designed a portable array of polypyrrole sensors for the analysis of diabetic patient's breath [62]. Li et al. detected aromatic organic compounds using nanofibers of conducting polyaniline [63]. CP have also been used as sensitive coatings and combined with different sensing platforms such as quartz crystal microbalance [64] and field effect transistors [65].

#### *2.2. Electrochemical Sensors*

This family of sensors includes three main categories classified according to their measurement approaches: amperometric, potentiometric and conductimetric/impedimetric sensors [44]. These electroanalytical techniques generally involve monitoring the modulation of an electrical property (current, potential, conductivity or impedance) associated with the interaction of odorant molecules with the working electrode [24]. The working electrode is usually made of gold or platinum and covered with sensing materials, for example, in certain cases, a porous membrane that acts as a transport barrier [35].

These sensors have the advantages of being robust and can function at room temperature [24]. They are also low cost, have low power consumption and can be miniaturized [66], which are all suitable characteristics for eN systems. Additionally, the reactivity of these gas sensors can be customized by adding metal layers, polymers or biological sensing materials to the working electrode surface [34]. However, due to their sensing methodology, some of these sensors have a narrow detection spectrum with a high sensitivity only to a limited number of electrochemically active gases [36]. Several groups have explored the potential of different categories of electrochemical sensors for the detection of VOCs and odorant molecules. For instance, Buttner et al. [67] have demonstrated the usability of an amperometric sensor for in situ detection of explosives in soil. Barou et al. [68] presented a proof of concept for the detection of odorant molecules using square wave voltammetry. Liu et al. [69] designed an olfactory biosensor based on electrochemical impedance spectroscopy (EIS). Also using EIS technique, Hou et al. [70] were able to detect odorant molecules by monitoring the electrical properties of a Langmuir-Blodgett film

containing OBPs. In another study [71], employing the same electroanalytical method, the team reported a novel odorant detection strategy using a rat olfactory receptor. As a part of the European project SPOT-NOSED, Akimov et al. [72] worked on the development of nanobiosensors that consist of a single olfactory receptor anchored between nanoelectrodes that detects odorant binding using EIS.

#### *2.3. Field Effect Transistor (FET)*

There are several types of FET gas sensors, including thin-film transistor, catalytic metal gate FET, suspended gate FET, capacitively coupled FET and horizontal floating-gate FET. The transduction principle of these devices is mainly based on the modulation of the threshold voltage or the drain source current. Each type of FET sensor has a specific structure, sensing mechanism and characteristics with different advantages and drawbacks. Hong et al. [73] recently published a paper that explains and reviews the operating principle, features and performance of each type of FET sensor.

Many research groups have studied and explored this type of sensor for VOC detection applied to different areas and using various types of sensing materials. For example, Haick's group has extensively worked on the development of silicon nanowire field effect transistors (SiNW FET). The SiNW FET surfaces were modified with different types of organic molecules in order to detect different kinds of VOCs and specially disease biomarkers [74–76]. Park's team developed a highly sensitive FET based bioelectronic noses using single walled carbon nanotubes or polypyrole nanotubes conjugated with human ORs [65,77]. Johnson's group designed and studied VOC sensor arrays using DNAdecorated carbon nanotubes FETs [78–80] and graphene FETs [81]. Kotlowski et al. [82] described an olfactory biosensor employing reduced graphene oxide FET functionalized with OBPs. Liao et al. [83] demonstrated that organic thin-film-transistors are suitable for electronic nose development.

#### *2.4. Gravimetric or Piezoelectric Sensors*

Two types of piezoelectric sensors are mainly used for VOC and gas detection: surface acoustic wave (SAW) sensors [8,35] and bulk acoustic wave (BAW) also called quartz crystal microbalance (QCM). A SAW sensor, in delay line configuration, basically consists of two inter-digitated transducers (IDTs) placed on top of a piezoelectric substrate such as quartz or Lithium niobate. To detect target molecules, a sensitive membrane (e.g., conducting polymers, lipids, biomolecules, etc.) is deposited between the IDTs [8]. A QCM sensor comprises a quartz disc coated with two gold electrodes connected to either side of the disc and a layer of sensitive material [35]. Despite their structural differences, both sensors have similar transduction principles. They detect odorant molecules by measuring variations in the resonant frequency caused by a change in mass after VOCs adsorption [8,31,32]. When an alternating voltage is applied across the piezoelectric element, it oscillates at a specific frequency driven by its mechanical properties [31]. This produces 2-dimentional acoustic waves (Rayleigh waves) that propagate along the surface at a frequency between 100 and 400 MHz in SAW sensors. Whereas, in QCM devices, 3-dimentional waves that travel through the bulk at a frequency of 10 to 30 MHz are generated [31].

QCM and SAW sensors have short response time and they are able to work at room temperature. Moreover, the detection spectrum of these devices can be tailored by modifying their sensitive membrane (the sensing materials) [8]. However, they suffer from complex circuitry and limited multiplexing capacity for large sensor array system. Additionally, the coating technologies are poorly controlled resulting in sensors having poor batch-to-batch reproducibility [31]. To tackle this issue, Chevalier et al. [84] showed that diamond nanoparticles can promote homogenous and reproducible coating of SAW sensors. A large number of studies have focused on the development of SAW and QCM based gas sensors and eNs using various sensitive materials. Rapp et al. [85] presented an improved array of eight SAW sensors for the detection of organic gas and an in-built multiplexing technique that allows an easy optimization of signal to noise ratio. They expanded the

choice of coatings for the SAW sensors and improved the sensor to sensor reproducibility for a certain coating material. Matatagui et al. [86] recently designed a portable low-cost eN based on SAW sensors and using ferrite nanoparticles as sensing materials for the detection of BTX (benzene, toluene and xylene), which are hazardous gases. Panigrahi et al. [87] worked on the detection of a VOC associated with *Salmonella* contamination in meat using a QCM system coated with synthetic polypeptides. Compagnone et al. reported a QCM sensor array using peptide modified gold nanoparticles for the detection of food aromas [88]. In another study [89], they have investigated the use of metallo porphyrins coated QCM platform for quality control of chocolate. Likewise, Di Natale et al. [90] designed an array of eight QCM sensors coated with metallo porphyrins for the detection of lung cancer. Park's group [91] and Wang's group [92,93] have developed QCM and SAW olfactory biosensors by employing ORs as sensing materials. Furthermore, several studies have explored the performance of QCM based sensors coupled to molecularly imprinted polymers (MIPs) for the detection of VOCs [30].

Other types of gravimetric sensor systems based on film bulk acoustic resonator [94–96], cantilevers [97–101], capacitive micro-machined ultrasonic transducer [102,103] have also been explored and optimized for the detection of VOCs. The following reviews [34,42,43,104] provide more details about these sensors and bring together different research articles that focus on the development of this technique.

#### *2.5. Optical Sensors*

This category of sensors detects odorants by measuring variations in the optical properties (e.g., refractive index, fluorescence, absorbance) of the sensing material by monitoring light properties modulation (e.g., wavelength, intensity, phase). They involve the use of a large assortment of techniques including different categories of optical fibers and a diversity of light sources and light-sensitive photodetectors [24]. Depending on the operating principle (i.e., the optical property that is monitored), it is possible to distinguish among several types of optical sensors, each having advantages and drawbacks. It is important to mention that optical spectroscopy (near infrared, infrared, Raman, etc.) is also very promising for gas sensing. Herein, it is not in the scope of this paper and thus will not be considered. More information can be found in a recently published review [105].

The simplest optical sensors effective for electronic nose development are colorimetric sensors. These sensors are based on the measurement of UV−vis absorbance or reflectance and involve the use of chemoresponsive dyes (chromophore) such as metalloporphyrins that will change color upon exposure to VOCs [106]. They have the advantages of being low cost, easy to manufacture and allow real-time multiplexed monitoring of VOCs. However, their main drawback is that they do not offer quantitative measurements [107]. Suslick's group pioneered this technique. They have extensively developed this type of sensors with a large number of published articles where they showed efficient detection of VOCs with very low detection limit for different applications [46]. Hou's group also developed a colorimetric sensor array for the detection of aldehydes and lung cancer biomarkers [108] and for the discrimination of Chinese liquors [109].

Fluorometric or fluorescent sensors are more sensitive than colorimetric sensors and involve the use of fluorophores. They can be categorized into different types based on the fluorescence parameter that is measured (e.g., fluorescence intensity, anisotropy, lifetime, emission and excitation spectra, fluorescence decay, and quantum yield) [24,46,110]. Walt and co-workers pioneered multiplexed fluorescent sensors combined with optical fibers [111]. Indeed, fiber optic platforms are widely used for optical sensor development thanks to their attractive features, including remote and multiplexed sensing capability, biocompatibility, miniaturized structure, light weight, flexibility and immunity to electromagnetic interference [112]. Another main advantage of these systems lies in the temporal response obtained with the kinetic information compared to the equilibrium response obtained with most other sensing technologies. In the field of VOC detection, Walt et al. [111,113] developed an array of optical fibers with a solvatochromic dye (Nile

red) encompassed in different polymer matrices with diverse polarity, flexibility, hydrophobicity, pore size and swelling tendency in order to obtain sensors that interact differently with VOCs. The sensitive polymer/dye combinations were deposited at the distal end of the fiber. Changes in the fluorescence intensity at a given wavelength upon the exposure to VOCs were recorded over time thanks to a CCD camera. In another study [114,115], they developed an array of fluorometric fiber optic-based sensors (FOS) where the fluorescent dyes were incorporated into different classes of microbeads. The beads were then immobilized in microwells at the tip of the imaging fiber. Kang's group also developed fiber optic-based fluorometric sensors for VOCs detection [107,116,117]. In particular, the team presented an array of five FOSs using four different types of solvatochromic dyes and two different polymers to form sensitive membranes. The sensing materials were deposited on side-polished optical fibers and pulse width modulations were measured as a response to VOCs [116]. More details and examples about colorimetric and fluorometric sensors can be found in the following reviews [35,46,114,118].

Another important family of optical sensors is based on surface plasmon resonance and involves the excitation of surface plasmons that are extremely sensitive to variations in the refractive index of the sensing materials. In 1982, Nylander et al. [119] investigated the possibility of employing SPR as a transduction technique for gas detection. Using an organic film as a sensing material, their system demonstrated a sensitivity to halothane in the parts per million (ppm) range. Since then, this optical sensing technique has gained substantial popularity. Owing to its prominent attractive features, namely, high sensitivity, label free detection and real time measurements, SPR constitutes a very powerful tool for sensor development comparing to other optical techniques. It has proven to be very useful for monitoring and studying interactions and affinities especially between biological elements (e.g., antibody-antigen, ligand-receptors). Consequently, SPR has been extensively employed for a large number of applications including diseases diagnosis, drug discovery and other bioanalysis [120,121]. Additionally, SPR sensors have been used for the detection of chemical species such as VOCs. Indeed, many research groups have developed efficient gas sensors, olfactory biosensors, and electronic nose systems using SPR as sensing technique. This will be the focus of the following section of the review. To illustrate the progress in this domain, examples of studies with different SPR coupling configurations will be presented and discussed.

#### **3. Propagating SPR-Based Gas Sensors and Electronic Noses**

The SPR phenomenon was first observed by Wood [122] in 1902. In his study, he pointed out inexplicable peculiarities in the spectrum of light diffracted by a diffraction grating. To understand this phenomenon, in 1941, Fano [123] re-examined Wood's observations and showed that the anomalous diffraction pattern was caused by the excitation of "polarized quasi-stationary waves" present at the surface of the metallic gratings. In 1952, Pines and Bohm [124] suggested that the energy losses of fast electron passing through foils were caused by the excitation of plasma oscillations or "plasmons" i.e., oscillations of the electronic density in the conducting media. Hereafter, this energy loss and its association with surface plasma oscillations were studied by Ferrell and Stern [125,126], Ritchie [127], Powell [128] and many others. In 1968, Otto [129] presented a method for the excitation of nonradiative surface plasma waves and showed that it resulted in a strong attenuation of the reflected light intensity. Moreover, in the same year, Kretschmann and Raether [130] described another configuration that enabled the excitation of the nonradiative surface plasmons (SPs).

A plasmon corresponds to the collective oscillation of the free electrons in a noble metal [131]. Surface plasmons are collective oscillations of electrons that take place at the interface between two media having dielectric constants of opposite signs typically a metal (e.g., gold, silver) and a dielectric (e.g., air, water) [132]. The SPs are not arbitrary events, they occur upon the excitation or coupling to an electromagnetic photon wave (i.e., light). In fact, when a photon beam interacts with the free electrons of a metal, these electrons will

respond by coherently oscillating in resonance with the light wave. This phenomenon is known as surface plasmon resonance and corresponds to the excitation of the SPs.

SPs can be classified into two categories: propagating or localized.

Propagating SPs, also known as surface plasmon polaritons (SPPs) or surface plasma waves (SPWs), are typically produced at the surface of thin metallic layers. SPPs can be considered as electromagnetic waves that propagate along the planar surface of a metal interfacing a dielectric (Figure 3a). The excitation of the SPs in such structures requires the use of coupling elements (e.g., prism, waveguide, gratings) that allow to achieve resonance or matching conditions leading to SPR.

**Figure 3.** Schematic representation of (**a**) propagating surface plasmon (SP) and (**b**) localized surface plasmon.

On the other hand, localized surface plasmon resonance (LSPR) occurs when light interacts with metallic nanostructures (e.g., gold nanoparticles) that are smaller than the incident wavelength [133,134]. The electric field of the light causes the localized free-electrons in the nanostructure to oscillate with a specific frequency. When the electron cloud is displaced relative to the nuclei a restoring force, generated by the Coulomb attraction between the electrons and the nuclei causes the electron cloud to oscillate relative to the nuclear framework [135] (Figure 3b). This has three main consequences: an enhancement in the local electromagnetic field near the particle's surface and a strong light scattering as well as a sharp spectral absorption with a maximum at the plasmon resonant frequency [136]. For gold nanoparticles (size ranging from few to hundreds of nanometers), a strong absorption pic is observed in the visible light leading to their red color in solution [136]. Unlike the SPR phenomenon, which takes place at the surface of a metallic film, LSPR does not require coupling elements and does not propagate hence its localized character. However, it is likewise sensitive to changes in the local dielectric environment. In particular, the extinction peak (namely the resonance wavelength) is highly affected by the refractive index of the surrounding. Thus, for sensing applications, molecular interactions occurring at the surface of the nanoparticles are typically detected by monitoring shifts in the LSPR wavelength [137]. LSPR sensing platforms consist of either metallic nanoparticles (e.g., nanospheres, nanorods, nanostars), suspended in solution or deposited on a solid support, or micro- and nano- fabricated metallic structures arrays on a solid support (e.g., nanopillar array) [138]. The LSPR peak wavelength can be tuned corresponding to the desired application by modifying the size, shape and material of the nanostructures, which represents an advantage for sensor development [137]. Thanks to the improvement in nanofabrication, various LSPR-based nanosensors have been developed for diverse applications including the detection of various biomolecules such as DNA, disease biomarkers, hormones, amino acids etc. [137,139–141], and different chemical compounds such as inorganic gases [142,143] and VOCs [2,144–146].

In the present review, we will exclusively focus on propagating SPR-based sensor developed for the detection of VOCs. In the literature, a considerable number of reviews and articles that describe and explain the theory behind this phenomenon (propagating SPR) as well as its application for sensor devices can be found [121,147–154].

In the following sections, we will first present a brief theoretical overview of propagating SPR. Then, we will make a comprehensive review of the progress made in the development of gas sensors and electronic noses that employ this technique. In particular, the different systems will be classified based on their coupling configuration.

#### *3.1. The Theory of Propagating SPR*

Let us consider a semi-infinite metal with a frequency dependent complex permittivity or dielectric function *ε<sup>m</sup>* and a semi-infinite dielectric with a permittivity *εd*, separated by a planar interface. The solution of Maxwell's equations under appropriate boundary conditions suggests that s-polarized surface oscillations cannot be supported by this type of interface. Consequently, SPWs are transverse-magnetic (TM) or p-polarized waves, i.e., their magnetic field vector is parallel to the interface and perpendicular to the propagation direction [121,154]. Moreover, the existence of surface plasmon requires that the real part of *ε<sup>m</sup>* is negative and its absolute value is greater than *εd*. At optical wavelength (visible and near infrared), this condition is satisfied for various metals including gold which is commonly used for sensor applications [153]. From the analysis of Maxwell's equations, it is also possible to derive the frequency dependent wavevector also called the dispersion relation or propagation constant of the SPW on a smooth surface that is given by [151,155]:

$$k\_{SPP} = \frac{\omega}{c} \sqrt{\frac{\varepsilon\_d \varepsilon\_m}{\varepsilon\_d + \varepsilon\_m}} \tag{1}$$

where *ω*/*c* is the free space wave vector of an optical wave.

The propagation of SPWs along the interface undergoes strong attenuation due to high Ohmic losses in the metal which, consequently, limits the propagation length [149]. This damping is associated with the imaginary part of the wavevector that depends on the metal's permittivity at the oscillation frequency of the SPW [121,149]. The propagation length along the interface is a few microns or even a few tens of microns depending on the metal and the excitation wavelength used [147]. This length can be expressed as [149]:

$$L\_{SPP} = \frac{1}{|2\operatorname{Im}\{k\_{SPP}\}|}\tag{2}$$

Confined at the vicinity of the interface, the electromagnetic field associated with the wave decays evanescently into the metal and the dielectric. However, as shown in Figure 4a the distribution of this field is asymmetric and mostly concentrated in the dielectric [148]. This disparity in the penetration depth is due to the fact that the dielectric constant of the metal is greater than that of the dielectric. The field decay from the surface in the adjacent medium is determined by the dispersion relations of the SPW in the direction perpendicular to the interface (i.e., in the dielectric *kzd* and in the metal *kzm*) [155]:

$$k\_z = \begin{cases} \frac{\omega}{c} \sqrt{\frac{\frac{\omega^2}{\varepsilon\_d + \varepsilon\_m}}{\varepsilon\_d + \varepsilon\_m}} & \text{in } dielectric; \\\ \frac{\omega}{c} \sqrt{\frac{\frac{\omega^2}{\varepsilon\_d + \varepsilon\_m}}{\varepsilon\_d + \varepsilon\_m}} & \text{in } metal. \end{cases} \tag{3}$$

**Figure 4.** (**a**) Distribution of the electromagnetic field of the surface plasmon polaritons (SPPs) along the *z*-axis (perpendicular to the surface), the intensity of this field is maximum at the surface and decays exponentially away from it. With *Ld* and *Lm* the penetration depth in the dielectric and the metal, respectively. (**b**) Dispersion curve of: free photons propagating in a dielectric (blue solid line), x-component of free photons propagating in a dielectric (red dashed line) and SPPs (black solid line).

The decay length also called penetration depth or skin depth of the SPW in the adjacent medium corresponds to the distance from the interface at which the intensity of the field falls to 1/e of its maximum value [148,155]. This value can be expressed as [155]:

$$L\_{zi} = \frac{1}{|Im\{k\_{zi}\}|} \text{ with } i = \text{metal or dielectric} \tag{4}$$

To give an order of magnitude, the penetration depth is a few hundred nm (~200 nm) in the dielectric and a few tens of nm (~25 nm) in the metal [147].

The excitation of surface plasmons or the generation of SPWs at the planar interface requires special configurations. Indeed, for the same frequency, the propagation constant (the wavevector) of the surface plasmon at the metal-dielectric interface (black solid line) is higher than the wavevector of photons in the dielectric (blue solid line) (Figure 4b). This mismatch has two consequences. First, the SPPs cannot radiate in light, and are bound to the surface. Second, they cannot be directly coupled or excited by a conventional light illuminating the metal/dielectric interface. Attenuated total reflection (ATR) or diffraction endows the excitation wave with additional momentum to overcome the mismatch and excite SPPs. In practice, this can be achieved using different coupling systems (couplers) such as prim, waveguide and grating couplers [121,147–149]. The excitation of the SPPs manifests itself by a resonant transfer or absorption of the incident light energy resulting in SPR.

As mentioned earlier, SPR is extensively used as transduction technique for optical sensor development and enables the detection of analytes by monitoring changes in the refractive index (*nd*) or permittivity (*ε<sup>d</sup> with ε<sup>d</sup>* = *n*<sup>2</sup> *<sup>d</sup>*) of the dielectric where the sensing material is deposited. Indeed, since the electromagnetic field of the SPWs is mostly concentrated in this medium, the propagation constant of the wave is strongly affected by its optical properties, namely, its refractive index. The characteristics of the exciting light (i.e., its intensity and phase), are altered upon the interaction with the SPPs and, thus, variations in these parameters can be correlated with changes in the propagation constant of the SPWs and thus the refractive index of the dielectric. In other words, binding-induced modulation in the refractive index at the sensor surface and, consequently, the propagation constant of the SPWs can be detected by measuring changes in the output light properties. Finally, it is worth noting that, since the penetration depth of the field in the dielectric is few hundreds of nm (~200 nm), the SPR can only detect binding events taking place below this limit.

In the following parts, we present the different coupling strategies and review the various studies that employed SPR for the detection of odorant molecules.

#### *3.2. Prism Coupler-Based Sensors*

The excitation of SPPs via ATR and prism coupler was first demonstrated by Otto then by Kretschmann and Raether. The Kretschmann configuration is the most commonly used method. This configuration consists of a thin metal film usually gold (about 50 nm thick) deposited on the surface of a prism on top of which the sensitive material is deposited. As shown in Figure 5a, to provoke the coupling, the prism is illuminated with a p-polarized light wave (since the SPW are p-polarized) at an incident angle greater than the critical angle. When the light reaches the prism-metal interface, it is totally internally reflected and an evanescent photon wave is generated at the interface [147]. The high refractive index or permittivity *ε <sup>p</sup>* of the glass prism allows to enhance the momentum or wavevector of the evanescent wave that can thus excite the SPPs [156]. Resonance occurs when the in-plane component of the incident light (photon) wave vector *kph*,*<sup>x</sup>* (red solid line), which corresponds to the propagation constant of the evanescent wave, matches that of the SPWs. Consequently, a transfer of energy from the incident light to SPWs occurs and is manifested by a sharp dip in the intensity of the reflected light. To satisfy the matching conditions, the angle of incidence or the wavelength of the exiting light can be adjusted since the propagation constant of the evanescent wave is dependent on these parameters. The terms resonance angle and resonance wavelength correspond to values of incident angle and wavelength at which almost 100% efficient coupling and energy transfer are achieved [156]. The resonance condition can be expressed as [121]:

$$k\_{\rm pl,x} = \frac{\omega}{c} \sqrt{\varepsilon\_p} \sin \theta = \text{Re}\{k\_{spp}\} \tag{5}$$

**Figure 5.** (**a**) Excitation of SPPs by prism coupling in Kretschmann configuration. (**b**) Dispersion curve of: x-component of free photons propagating in a dielectric (red dashed line), x-component of free photons propagating in a prism (red solid line), SPPs (black solid line). (**c**) Intensity interrogation principle.

The same resonant conditions apply for the Otto configuration. The only difference is that, in this configuration the metal film is separated by a small gap from the surface of the prism [129].

In practice, for sensing applications, the sensitive materials are deposited on top of the metal layer, which allows to customize the sensitivity and selectivity of the sensor. Four main measurement methodologies are employed to detect the kinetic interaction of target molecules with the sensitive materials: intensity interrogation, spectral or wavelength interrogation, angular interrogation and finally phase interrogation [157].

In the first method (namely intensity interrogation), variations in the intensity of the reflected light are monitored over time at a fixed wavelength (i.e., using a monochromatic light source) and fixed incident angle (known as the working angle) (Figure 5). The working angle is usually chosen close to the resonance angle (*θres*) where small variations in *θres* caused by modulation of the surface refractive index will result in large shifts in the intensity of the reflected light. On the other hand, for spectral/wavelength interrogation, a broadband or polychromatic light source is used to excite the SPPs at a fixed incident angle and modulations of the resonance wavelength are monitored. Conversely, in the case of angular interrogation, variations in the resonance angle are measured at a fixed wavelength. Finally, for phase interrogation, shifts in the relative phase difference between p- and s-polarization components are monitored at a specific wavelength and angle. This last interrogation technique offers the highest sensitivity but suffers from a narrow dynamic detection range [158]. The different interrogation methods and especially intensity interrogation allow for simultaneous monitoring of binding events occurring on multiple sensors which is particularly beneficial for electronic nose systems. This multiplexing technique is called surface plasmon resonance imaging (SPR imaging) [159].

Many gas sensors and eN systems can be found in the literature based on this configuration and using a large diversity of sensitive materials including biological elements (e.g., olfactory receptors, odorant binding proteins and peptides) and chemical elements (e.g., polymers and calixarenes). The different systems can be classified into two categories depending on the detection medium, i.e., in the liquid or in gas phase.

#### 3.2.1. Detection of VOCs in Liquid Phase

Prism coupler-based SPR has been widely employed to develop biosensors/biochips for the analysis of large biological molecules. However, it is often considered unsuitable or limited for the analysis of low weight molecules such as VOCs (molecular mass < 300 Da) in the liquid phase. To overcome this limitation, it is essential to couple the optical transduction systems with appropriate sensitive materials in order to generate detectable signals upon their interaction with VOCs. Different biological sensing materials (ORs, OBPs, etc.) have been used for such applications. Very often, signal amplification strategies are needed to obtain reliable SPR signals, which will be highlighted in this review.

Selected and improved by natural evolution, olfactory receptors are very attractive candidates. Since their identification and isolation by Buck and Axel, these proteins have been extensively studied [10]. Great research efforts has been made to deorphanize these receptors [160] and improve their large-scale production that was found to be challenging in some early works [161,162]. The use of OR as sensing materials for the development of olfactory biosensor and eNs presents many assets including high sensitivity and selectivity. In addition, they can be genetically modified to facilitate their purification and immobilization. However, being transmembrane proteins, the presence of a lipid bilayer environment is crucial to maintain their three-dimensional structure and retain their activity when immobilized on sensor chips. This task has been a major drawback and challenge for the development of OR-based olfactory biosensor. Nevertheless, several ingenious strategies have been employed to provide the lipidic environment such as the use of plasma membrane fractions, nanovesicles and nanodiscs [26]. Consequently, OR-based sensors were proven to be effective for the detection of VOCs using different transduction systems including QCM [92], FET [163], electrochemical [71].

SPR platforms have also been associated with ORs. Pajot-Augy's group [164] demonstrated the possibility of using mammalian OR as sensing elements for highly sensitive olfactory biosensors. In their study, they first co-expressed rat ORI7 and human OR17-40 and their associated Gαolf subunit in yeast cells. To maintain their structure, the ORs were encompassed in membrane fractions that formed nanosomes with a diameter of approximately 50 nm. The nanosomes were then immobilized on a Biacore sensor chip L1, which consisted of a gold-coated glass support functionalized by a covalently linked carboxylated dextran polymer hydrogel grafted with long alkyl chains (Figure 6). Nanosomes were effectively bound by those alkyl anchors. A BIAcore 3000 was used to perform measurements. This type of setup allows the measurement of resonance angle shifts and consists of a near-infrared LED light source for SPR excitation and a linear array of light sensitive diodes to monitor the reflected light. As reported, no SPR signal was observed when VOCs were injected alone due to poor signal/noise ratio. To solve the problem, an indirect ingenious amplification strategy was designed. It consists of taking advantage of the presence of Gαolf anchored to the nanosomes to monitor receptor activation by an odorant ligand, through the desorption of Gαolf subunit from the lipidic bilayer. In such a way, when a target odorant binds to the OR, the subunit is activated and then desorbs from the lipidic membrane, resulting in a much stronger SPR signal, as illustrated in Figure 6. To trigger this mechanism, VOCs were injected in the presence of guanosine-5 -triphosphate (GTP). The study demonstrated that ORs retained their functionality in membrane fractions even after immobilization and the obtained olfactory biosensor exhibited high sensitivity and selectivity. The sensor chip kept the same activity level for up to eight injection cycles.

**Figure 6.** (**a**) BIAcore sensor chip L1 functionalized with nanosomes. No surface plasmon resonance (SPR) response was observed when nanosomes were stimulated either with odorant alone (**b**), or guanosine-5 -triphosphate (GTP) alone (**c**), as compared to the control stimulated with water. The SPR signal was only observed when odorant and GTP were injected simultaneously (**d**). The signal relative to the release of the Gα subunit can be further enhanced four-fold by replacing GTP by GTPγS (**e**) [164].

In a complementary study [165], using this SPR sensing strategy, they investigated the molecular mechanisms underlying odorant detection, in particular, the role of OBPs in the dynamic interactions between OR and odorant ligands. They showed that OBPs play an active role in preserving the conformation and activity of OR especially at high odorant concentration. This finding revealed another role of OBPs in olfaction, in addition to their role in transporting odorants through the olfactory mucus. Importantly, their study showed that SPR-based olfactory biosensors can be used not only for the analysis of odorant molecules, but also for the investigation of basic biologically relevant questions in olfaction.

Furthermore, in collaboration with Jaffrezic-Renault's team, they demonstrated the importance of the surface chemistry on the performance of the system [166]. Human OR17-40 modified with a cmyc tag on the N-terminus and its Gαolf subunit were coexpressed in yeast cells (*S. cerevisiae*). The receptors carried by nanosomes were attached to the sensor chip through specific antibody-directed immobilization using Anti-cmyc monoclonal antibodies. Two strategies involving different biofilm architectures were explored: one with controlled antibody orientation and the other with random orientation, as illustrated in Figure 7. A Kretschmann-type SPR spectrometer NanoSPR-6 with two optical channels and a diode light source (650 nm wavelength) was used to perform the study. The response of the system was measured in terms of resonance angle modulation. The setup included a double channel Teflon flow cell that allowed signal acquisition in both custom and differential modes (delta between working and reference channels). They showed that the density of nanosomes and the multilayer bulk thickness are crucial factors for the performance of the olfactory biosensor. The biofilms prepared following the first surface chemistry strategy had higher thickness and nanosome density. However, the corresponding olfactory biosensor exhibited a lower sensitivity for the target odorant molecules compared to the OS based on the second surface chemistry. Indeed, the second strategy provided biofilms with lower thickness and higher porosity that allowed a better accessibility of Gαolf to GTPγS, and thus, increased sensitivity.

**Figure 7.** Schemes of the two-surface chemistry employed for the immobilization of olfactory receptors (ORs) in the nanosomes, which were specifically captured via anti-cmyc antibody attached to the gold-coated substrate in an orientated (**a**) or random way (**b**).

Another strategy to exploit the potential of OR for sensing applications is to use so-called artificial olfactory cells, which are genetically modified cells that express olfactory receptors. Park's team developed a sensitive and selective SPR-based olfactory biosensor using whole cells expressing olfactory receptors ORI7 as sensitive materials [167]. The cells were attached to a gold-coated glass slide using poly-D-lysine. The slide was put into optical contact with a prism using a refractive index matching fluid. A p-polarized laser light with a wavelength of 670 nm was used as the probe beam. Thanks to a photodiode detector, variations in the reflected light intensity were monitored as a response to analytes.

In this system, the SPR signal was not directly ascribed to the conformational change of the OR or to the desorption of the Gα subunit. In fact, the olfactory receptors expressed on the surface of the cell were not in the detectable range of the SPR (approximately 200 nm above the gold surface), since the size of the cell was several micrometers. However, the G-protein transduction cascade induced by odorant binding generated changes in the intracellular components, mainly with an increase in Ca2+ ions. Such changes generated a variation in the local refractive index consequently leading to an SPR signal. In a previous study [168], the group had already demonstrated the feasibility and effectiveness of such a system (i.e., an SPR-based sensor with artificial olfactory cells expressing OR) for the detection of odorants (Figure 8).

**Figure 8.** Principle of cell-based measurement of odorant molecules using SPR. An olfactory cell expressing OR was adhered to the gold surface of the sensor chip, and activated by odorant molecule diacetyl. The specific binding of diacetyl to the OR triggered the G protein transduction cascade inside the cell and thus an SPR signal [168].

Although the cell-based olfactory biosensor is very interesting, it is limited by the short lifetime of the sensitive materials. In addition, the system is easily influenced by environmental conditions. Therefore, in another work [169], Park's team explored a different strategy to provide a natural lipidic environment to maintain the stability and biological function of ORs and that is to use liposomes (Figure 9). They controlled the size of the liposome to 40–50 nm, making them fall within the detectable range of the SPR. The liposomes were then immobilized on the poly-D-lysine-coated SPR sensor chip. Their study demonstrated that the reconstituted ORs carried by liposomes were effective sensitive materials for odorant detection.

**Figure 9.** Schematic diagram of reconstitution of OR and SPR analysis. The partially purified OR was reconstituted using lipid/detergent mixed micelle and immobilized on the gold surface of SPR to detect the odorant binding [169].

In a similar work, Sanmartí-Espinal et al. [170] prepared nanovesicles from yeast membranes, with a size of about ~100 nm in diameter, to carry ORs as sensitive materials. Their SPR-based olfactory biosensor had good selectivity. Based on the SPR signal, they even tried to quantify the number of odorants that interacted with a given olfactory receptor.

In addition to ORs, odorant binding proteins also have great potential as sensitive materials in the field of olfactory biosensors. OBPs are small proteins (~20 kDa) highly concentrated in the nasal mucus of vertebrates [171] and in the sensory lymph of insects [172]. Vertebrate OBPs belong to the lipocalin family, characterized by β-barrel structure with eight antiparallel β-sheets that enclose a hydrophobic binding cavity for odorants also known as calyx [173]. Thanks to their binding pocket, OBPs can reversibly bind odorant with micromolar dissociation constant and a broad affinity spectrum (i.e., can interact with different chemical classes) [173]. These proteins are thought to act as shuttles that facilitate the transport and diffusion of hydrophobic odorants across the aqueous mucus to reach the olfactory receptors [174].

Unlike ORs, OBPs are soluble proteins, and thus, do not require a lipidic environment. This also facilitates their large-scale production and purification. They exhibit good stability to high temperature and pH variations, as well as low susceptibility to proteolytic degradation [27]. Moreover, they have a broad specificity and can be genetically modified to tailor their binding properties or facilitate their immobilization. Despite their high stability, maintaining the activity of these proteins over time after their immobilization on the sensor chip and/or after exposure to VOCs is challenging especially in a dry working environment. Nevertheless, many studies have largely investigated the suitability of these sensitive materials for the development of olfactory biosensors and eNs. Indeed, OBPs have been coupled to different transduction platforms (e.g., SAW [175], FET [176]) and their performance were evaluated in both liquid and gas phase [26].

Recently, our team successfully demonstrated the feasibility of a SPR-based OS with OBPs as sensing elements [177]. For that study, three rat OBP3 derivatives with customized binding properties were designed and produced, including OBP3-w, OBP3-a and OBP3-c. The first protein corresponded to the wild type form while the two others were genetically modified mutants. Thanks to site-directed mutagenesis, the binding affinities of the OBPs were customized by varying certain amino acid residues of their binding site. OBP3-a was tuned to have good affinity for aldehydes by introduction of a lysyl residue, while OBP3-c was modified with bulky amino acids to block the binding pocket. Consequently, it could no longer interact with VOCs and was used as negative control. The recombinant proteins were all expressed in *E. coli*. They were immobilized by self-assembly on gold-coated prism by means of a cysteine group that was introduced to their N-terminus, located on the opposite side of the binding cavity. This functionalization strategy allowed easy and orientation-controlled protein immobilization with the OBP at the vicinity of the gold surface. The SPR measurements were performed using a commercial SPR imaging apparatus (SRRiPlex from Horiba). The microarray was illuminated with p-polarized light at 663 nm wavelength. The intensity modulation of the reflected light at a fixed working angle of all the sensors was monitored simultaneously thanks to a CCD camera upon addition of VOCs (Figure 10).

The obtained SPR-based olfactory biosensor had a very low detection limit (DL), e.g., 200 pM for the odorant β-ionone. This result is among the lowest DL reported in the literature. Moreover, the SPR system was able to detect odorants with a molecular weight of 100 g/mol (hexanal) which is lower than DL in mass commonly admitted for commercial SPR imaging, namely, 200 g/mol. Indeed, the intensity of the SPR signal obtained could not be explained solely by the increase in mass after the binding of VOCs on the chip. It is very likely that the binding of VOCs to OBPs induced a conformational change, which led to a variation of the local refractive index with amplified SPR imaging signals. This was possible thanks to our functionalization strategy that enabled the immobilization of the OBPs at the vicinity of the gold surface. Moreover, at low VOC concentration, the olfactory biosensor exhibited an extremely high selectivity with great potential for trace VOC detection.

Biomaterials unrelated to the olfactory system were also used as sensitive materials. Dung et al. developed an efficient SPR-based olfactory biosensor for the detection of toluene using the toluene binding domain (TBD) [178]. TBD belongs to the TodS protein present in the bacterium *Pseudomonas putida*. In this study, a direct immobilization strategy was also employed by introducing three cysteine residues to the N-terminus of the TBD protein. This allows, on the one hand, to control the protein orientation to ensure good accessibility of the binding pocket, on the other hand, to detect SPR signal induced by the conformational change of TBD upon toluene binding. Shifts in reflected light intensity were monitored by a photodiode receptor as a response to analytes. The TBD-based olfactory biosensor showed not only good sensitivity for the target VOC, with DL at 15.62 μM, but also high specificity, with no response for other aromatic hydrocarbons, such as p-xylene and benzene.

The Table 1 summarizes the conditions for VOCs detection of SPR-based olfactory biosensors and their performances in liquid phase.

#### 3.2.2. Detection of VOCs in Gas Phase

The first studies showing the feasibility of prism coupler-based SPR for gas detection date back to the early 1980s [119,179]. However, very few examples were reported in the literature before 2000 [180–184]. These systems were limited in terms of sensitivity and selectivity based on only one or few sensitive chemical layers. Since 2000, an increasing number of articles can be found in the literature using both biological and organic sensitive materials [185–204]. It has been demonstrated that SPR is very effective for sensing VOCs in the gas phase. In fact, when using air as the analysis medium, the detection noise remains relatively low thanks to the low optical index of this medium. Consequently, the binding of the small VOCs can generate reliable SPR signal with very high signal/noise ratio.

For the development of SPR-based olfactory biosensors and eNs for VOC detection in the gas phase, the use of biomolecules such as ORs and OBPs as sensitive materials is limited by their stability under such conditions. Their peptide analogues are particularly interesting alternatives. Indeed, peptides, and in particular, short ones, do not require specific conditions (i.e., humidity, temperature, phospholipidic matrix) to maintain their activity. Moreover, they are much easier to produce and immobilize onto a sensing platform.


**Table 1.** SPR-based olfactory biosensors in the Kretschmann configuration for the detection of VOCs in liquid phase.

Recently, our group developed an innovative optoelectronic nose using biomimetic peptides based on SPR imaging for the detection of VOCs in the gas phase [185]. For this purpose, a homemade SPR imaging system based on the Kretschmann configuration was constructed, shown in Figure 11. A polarized LED light beam with a 632 nm wavelength was used to excite SPs and a 16-bit CDD camera was used to simultaneously monitor the reflectivity of all the sensors on the chip in real-time. Variation in the reflectivity at a fixed working angle was measured over time upon the exposure of the sensor microarray to VOCs, providing a temporal response.

**Figure 11.** Schematic presentation of the home-made SPR imaging setup.

Such an SPR imaging system is very promising for the development of eN. First, a chip consisting of a large sensor array can be easily prepared and used. The number of sensors is only limited by the resolution of the microarray printing of the sensitive materials. Second, thanks to the imaging mode, the interactions between VOCs and all sensors can be simultaneously monitored using the same instrument. Finally, SPR imaging can provide

temporal responses with additional kinetic information compared to a simple equilibrium response obtained with most of the existing eNs.

The peptides were all terminated by a cysteine for their direct immobilization on the gold surface of prism. Thanks to their diverse physicochemical properties and crossreactivity for VOCs, the obtained eN was found to be very effective in sensing VOCs of different families. In particular, it exhibited extremely high selectivity, capable of discriminating between VOCs differing by a single carbon atom. Additionally, it showed good repeatability and stability under repeated use and prolonged storage.

In order to improve the performance of our eN, in another study [186], we investigated the influence of the wavelength of the LED on the sensitivity of the system by combining numerical simulations with experimental validation. The results showed that the angular sensitivity increased with the wavelength but the angular linearity range decreased due to the narrowing of the plasmon resonance curve at high wavelength. Therefore, a compromise must be made to choose the optimal wavelength depending on the study purposes. Under optimal conditions, the detection limits of our eN reach low parts per billion (ppb) range for VOCs such as 1-butanol.

Furthermore, we investigated the optical contributions to the sensitivity of the SPR imaging [187]. For this, an original characterization method, which was independent of the carrier gas, was established for the SPR prism sensitivity based on pressure jumps [205]. In this work, the impact of different adhesive layer (Cr, Ti) as well as surface topography on the system sensitivity was evaluated. It was found that even though slightly higher sensitivities were theoretically achieved using Ti/Au prism, Cr/Au prisms were more suitable for eN applications since they showed lower sensitivity variabilities, noise, and signal drift due to better adhesive properties. Furthermore, the sensitivity loss due to Au grain-related SPP damping was fully characterized and numerically validated to be free from additional fitting parameters. The adsorption of water vapor was later characterized for such Au surfaces to understand humidity related effects on the eN system. Finally, our study showed that prism sensitivity decreased with increasing temperature [206].

In order to diversify the sensitive materials for eN development, in collaboration with Compagnone's team [191], we tested six novel penta-peptides and nine hairpin DNA selected by virtual screening. Thanks to the complementarity of their binding properties, the obtained eN was able to discriminate not only between VOCs of different chemical families, but also VOCs from the same family with only 1-carbon difference such as 1-butanol and 1-pentanol.

Considering the outstanding potential of our eN system and its great ability to detect and discriminate VOCs, a miniaturized version, called NeOse Pro, was further developed by the company Aryballe. Using the same biomimetic peptide-based chip, Maho et al. [188] demonstrated that NeOse Pro was even able to discriminate between two chiral forms ((*R*) and (*S*)) of Carvone and Limonene (Figure 12). Such performance is exceptional for eN system.

**Figure 12.** (**a**) Portable NeOse Pro and the experimental set-up for VOC sampling, (**b**) working principal and (**c**) raw image of the prism surface with each spot corresponding to a sensor [188].

NeOse Pro is a very promising tool for field analysis, although, as with most eNs, its use for the headspace analysis of highly humid samples remains a challenge, since its performance may be deteriorated by the presence of a high background signal generated by water vapor from aqueous samples. Slimani et al. [189] have tackled this issue by using a miniaturized silicon preconcentrator packed with hydrophobic adsorbent coupled to the NeOse Pro (Figure 13). As a result, the eN showed not only a great improvement in the detection limit (lowered by 125-fold) for target VOCs, but also an enhancement in the discrimination ability demonstrated by the analysis of eight different flavored waters.

**Figure 13.** NeOse Pro and μ-preconcentration system coupling. (**a**) Experimental setup, with the sample vial. (**b**) Schematic view of the NeOse Pro/micro preconcentrator (μPC) system and (**c**) View of the preconcentration chip on the metalized side [189].

In a recent study, Fournel et al. [190] compared the performance of the NeOse Pro with human olfaction. They found that the responses of the eN were not a mere reflection of the chemical space of odorants, but rather, that semantic dimensions were also prominent, similar to natural olfaction.

Besides biomolecules, chemical sensitive materials such as cavitands (calixarenes, cyclodextrins) were also used for the detection of gaseous VOCs with prism coupler-based SPR. They are very interesting for trapping VOCs thanks to their molecular structures with cavities, whose sizes, shape and physicochemical properties can be tuned using a wide variety of functional groups.

Daly et al. [192] ingeniously designed new cavitands containing a carboxylic acid group at the upper rim of the cavity for the detection of organophosphorus vapors, and in particular, the sarin nerve gas stimulant dimethylmethylphosphonate (DMMP). The formation of a hydrogen bond between the COOH moieties and P = O group of DMMP was expected. Two different cavitands with four alkyl feet (five carbons long) were produced and studied. Both molecules had similar cavities but with the carboxylic acid group pointing either out of or into the cavity. Their sensitivity to DMMP was compared with that of fluoropolyol, a commonly used polymeric sensing layer for DMMP detection. Cavitands and fluoropolyol layers were deposited on gold-coated glass slide by spin coating and Langmuir-Blodgett technique for comparison. Both techniques allow for the preparation of uniform and homogeneous thin films with a controlled thickness. To perform measurements, a variable wavelength SPR setup in the Kretschmann configuration was used. The interaction between the DMMP and the sensing layers was monitored by measuring the shift in the SPR wavelength at a fixed incident angle upon the exposure to analytes. The results showed that both cavitand layers exhibited almost the same sensitivity and were able to detect ppb levels of DMMP with a rapid and reversible response. The orientation of the COOH group had no effect on DMMP binding, but had strong impact on water uptake. The cavitand-based gas sensor outperformed the fluoropolyol-based one in terms of DMMP sensitivity and with less interference from water vapor and alcohol. Therefore, such a gas sensor is promising for sensitive and specific detection for nerve gas

agents. Moreover, the use of cavitands as sensitive materials for SPR based detection of aromatic vapors was also reported by Feresenbet et al. [193].

In a recent study, ¸Sen et al. [194] worked on the development of gas sensor for the detection of VOCs and in particular acetone using synthesized tetranitro-oxacalix[4]arenes. To perform the study, three nitro-substituted heterocalix[4]arenes were synthesized. Thin films of the three sensing materials were deposited on a substrate by spin coating. Their sensing properties for acetone, chloroform, toluene, ethanol and benzene vapors were evaluated by SPR. A BIOSUPLAR 6 Model spectrometer was used to perform SPR measurements. A p-polarized light with a wavelength of 632.8 nm was used to excite the SP. The intensity of the reflected light at a fixed working angle was recorded by a photodetector as a function of time upon the injection of VOCs. The sensing performance of the three films were investigated at room temperature and the VOCs were carried by dry air to avoid the effect of water vapor. As a result, two of the three thin films showed high sensitivity and selectivity to acetone with a detection limit of 3.8 ppm. The system also exhibited a fast and reproducible response with short recovery times (few seconds).

Other chemical sensitive materials such as polymers were also explored and combined with prism coupler-based SPR for the development of gas sensor. Capan et al. [195] investigated the performance of poly(methylmethacrylate) (PMMA) film as a sensitive material for the detection of BTEX (benzene, toluene, ethylbenzene and m-xylene). PMMA films with different thicknesses were deposited onto gold coated glass substrates by spin coating. The different films were obtained by varying the concentration of the polymer solution and the spin speed. The SPR measurements were performed using a Kretschmann type optical setup and a p-polarized monochromatic light at a wavelength of 633 nm was used to excite the SPs. Optical contact between the substrates and a semicylindrical prism was achieved using an index-matching liquid. Two interrogation methods were used to monitor the response of the system upon VOC injection: modulation in the reflection intensity over time at a fixed working angle and shifts in the resonance angle. As a result, among all the BTEX gases, benzene produced the highest SPR response when exposed to PMMA films. Moreover, the response to the other VOCs was very low which indicated that the gas sensor had high selectivity to benzene. Additionally, the team studied other sensitive materials such as calix-4-resorcinarene films [196] and poly[3-(6-methoxyhexyl)thiophene] derivatives films [197] for sensing BTEX and other VOCs using SPR.

Nanto et al. [198] also used synthetic polymer thin films as sensing membranes for the detection of harmful gases such as ammonia and amines with an SPR-based sensor in the Kretschmann configuration. An LED emitting at a wavelength of 660 nm was used as light source and the reflected light was measured by a CCD camera. The response of the system was measured in terms of modulation of the resonance angle as a function of time upon VOC injection. The sensitivity of two types of polymers was investigated: acrylic acid and styrene. A thin film (several tens of nm) of each polymer was deposited on the gold-coated surface of a prism using plasma chemical vapor deposition (CVD). The response of both membranes was tested against eleven harmful gases: ammonia, acetaldehyde, propionaldehyde, xylene, toluene, trimethylamine, triethylamine, dimethyamine, hormaldehyde, acetic acid and butyl acetate. As a result, the gas sensor with the acrylic acid membrane responded only to the basic gases (i.e., ammonia and amines) with high sensitivity and selectivity. In contrast, the OS with the styrene membrane exhibited a 200 times lower sensitivity. The system with the acrylic acid membrane also exhibited a linear response for ammonia in the range of 50–300 ppm and with an estimated detection limit of several ppm. Finally, the study showed that the thickness of the sensing membrane can be optimized to improve the sensitivity. In another study, using a similar system, Nanto's team [199] successfully demonstrated the feasibility of multiplexing with a two-channel odor sensor able to simultaneously detect ammonia and acetic acid with high selectivity. The sensor was based on the same SPR setup but with two sensing membrane, namely, acrylic acid and N,N-dimethylacetamide thin films deposited on one chip by CVD. Two channels of the CCD camera were used to monitor the response to VOCs.

To improve the sensitivity of polymer-based gas sensors, one strategy is to introduce nanoparticles (NPs) such as gold NPs. According to the literature the incorporation of Au NPs in SPR sensors could enhance the sensitivity of the device [207]. Indeed, with a rational design, coupling between the localized surface plasmons of the Au NPs and the propagating surface plasmons of the Au substrate may take place, which can result in a larger plasmon angle shift and changes in reflectivity.

Sih et al. [200] developed an SPR-based gas sensor for the detection of alcohol vapors. In the study, the performance of polythiophene (PT) films as a sensing material was compared with that of Au NPs thin films capped with conjugated oligothiophenes. SPR measurements were performed using a Kretschmann configuration setup and a p-polarized light at a wavelength of 632.8 nm was used to excite the SPs. To prepare the chips, the Au NP/oligothiophene (NPOT) film (~60 nm thickness) was electrodeposited on a gold-coated glass slide and the PT film (~7 nm thickness) was deposited by electropolymerization. The response of the sensors was monitored by measuring the shift of the resonance angle. The performance of the two sensitive materials was tested upon exposure to vapors of five solvents: hexanes, toluene, ethanol, methanol, and water. As a result, the PT layer responded to ethanol, methanol and toluene whereas the NPOT film responded exclusively to alcohols. Therefore, there was an improvement in selectivity in incorporating Au NPs. However, in this study no significant improvement in sensor sensitivity was observed.

Another advantage of nanostructures for gas sensor application is the high surface to volume ratio. Alwahib et al. [201] tested the efficiency of a SPR-based OS with a reduced graphene oxide/maghemite (rGO/γ-Fe2O3) nanocomposite film as sensing layer for hydrocarbon vapor detection. They used a kretschmann-based SPR setup with a helium-neon (He-Ne) laser at 633 nm emission wavelength. A chopper and a polarizer were used to generate the p-polarized excitation beam and a photodetector to monitor the reflected light (Figure 14). Trilayer and bilayer sensing membranes were prepared and compared. The former consisted of a nanocomposite layer (3 nm thick) sandwiched between two gold layers (bottom layer: 37 nm thick, top layer: 2.7 nm thick). For the latter, the rGO/γ-Fe2O3 film (3 nm thick) was deposited on top of a gold layer (49 nm thick). The sensing membranes were placed on microscope glass slides and then brought into contact with a high index prism using an index-matching liquid. The response of the system, upon the exposure to acetone, ethanol, methanol and propanol, was monitored in terms of modulation of the resonance angle. The SPR signal resulted from the adsorption of hydrocarbon vapors that diffused through the pores of the sensing layer inducing a change of the refractive index. As a result, the trilayer-based gas sensor showed higher sensitivity to acetone compared to the other hydrocarbons. Furthermore, it was more stable and had shorter response time comparing to the bilayer-based gas sensor. The authors concluded that this improvement was due to the presence of the third gold layer, which promotes better interactions.

Apart from the improvements of the SPR sensitivity by the optimization of the optical parameters and the use of NPs as described earlier, other approaches have been proposed in the literature based on active plasmonics to add active functionalities to SPR-based devices. For example, Manera et al. [202] reported a study where magnetic field was used to control the SPR. They compared the sensing performance of a magneto-optical SPR (MO-SPR) sensor with that of a traditional SPR sensor for the detection of alcohol. A home-made setup with the Kretschmann configuration was employed to perform the measurements and a p-polarized light with a wavelength of 632.8 nm was used to excite the surface plasmons. To prepare the MO-SPR sensor, a multilayer of Cr/Au/Co/Au was deposited on a glass substrate. Then, a nanoporous TiO2 thin film, used as sensitive material was deposited on top of the multilayer by glancing-angle deposition (GLAD). For comparison, a substrate for classical SPR was also prepared by depositing the TiO2 layer on top of a gold-coated glass substrate. Three VOCs were analyzed, including ethanol, methanol and iso-propanol. The MO-SPR based gas sensor exhibited a significant improvement in sensitivity. Furthermore,

its sensitivity was also much higher than that of their previous SPR-based gas sensor using TiO2 thin films [203] and nanometric polyimide films [204].

**Figure 14.** (**a**) SPR setup for detection of hydrocarbon vapor using trilayer Au-rGO/γ-Fe2O3-Au sensor. (**b**) SPR signals of the acetone vapor detection using the reduced graphene oxide/maghemite (rGO/γ-Fe2O3) sensing layer. (**c**) its resonance angle shift for increasing concentrations of different hydrocarbon [201].

The Table 2 summarizes the conditions for VOCs detection of SPR-based artificial olfaction systems and their performances in gas phase.

#### *3.3. Wave Guide Coupler-Based Sensors*

The fundamental coupling principle using a waveguide is similar to that of the prism, whereby the excitation of surface plasmons is achieved by an evanescent wave generated by ATR. To clarify the terms, an optical fiber is a special type of waveguide, and one that is widely used. Indeed, fiber optics are less expensive than waveguides and have good flexibility, remote sensing capability and other important features presented earlier. The VOC sensors systems that will be presented in the following section will exclusively involve the use of fiber optic SPR (FO-SPR) sensing platforms.

Fiber optic-based SPR sensors can be elaborated based on either transmission or reflection configuration. A typical fiber optic consists of high refractive index material (the core) sandwiched with a lower refractive index layer (the cladding) which allows light guidance through a succession of total internal reflections (TIRs). In the case of an FO-SPR in transmission configuration, a small region of the optical fiber cladding is removed and replaced by a metal layer where the SPR phenomenon will take place (Figure 15). In the reflection configuration, a thick metal layer deposited at the end of the fiber allows for the SPWs generation and plays the role of a mirror. In both cases, the sensitive materials are deposited on top of the metal layers. As with prism coupling, resonance occurs when the propagation constant of the evanescent wave generated by ATR of the guided mode matches the propagation constant of the SPWs [156]. In SPR-based optical fiber sensors, most of the interrogation methods are based on the detection of loss in the transmitted/reflected light at the resonance. Spectral or wavelength interrogation of the transmitted or the back-reflected light is the most commonly used measurement method. However, fiber-optic sensors based on intensity or phase interrogation have also been reported [208]. Theoretically, the sensitivity of waveguide-based SPR sensors is approximately the same as that of the corresponding ATR configurations [147].


**Table 2.** SPR-based artificial olfaction systems in the Kretschmann configuration for the detection of VOCs in gas phase.

**Figure 15.** Typical fiber optic SPR (FO-SPR) sensor in transmission configuration.

The first fiber optic based SPR sensor with a conventional geometry (as the one presented in Figure 15) and using spectral interrogation as measurement methodology was proposed by Jorgenson et al. [209] in 1993 for a chemical sensing application. Since then, a large number of studies have experimentally and/or theoretically explored diverse geometry-modified single mode or multimode fibers including side and tip implemented FOS, fiber gratings (e.g., long period fiber gratings and tilted fiber Bragg gratings) and specialty fibers [208] (Figure 16). Different plasmonic coatings (e.g., gold, silver) have also been explored. Moreover, configurations involving the excitation of SPPs on continuous thin metallic layers (i.e., propagating SPPs) as well as those involving LSPR phenomena in metallic nanoparticles at visible and near-infrared wavelengths have been reported and reviewed [210].

**Figure 16.** Schematic representation of the different plasmonic fiber-optic sensors **I**: (**a**) Unclad/etched/tapered fiber SPR probe; (**b**) Hetero-core structure; (**c**) Side-polished/D-shaped SPR probe; (**d**) U-shaped SPR probe. **II**: (**a**) Flat fiber tip SPR probe with end mirror; (**b**) Angle polished flat fiber tip SPR sensor; (**c**) Tapered tip SPR probe; (**d**) LSPR fiber tip probe. **III**: (**a**) Wagon-wheel fiber SPR sensor with triangular hole geometry; (**b**) Microstructured optical fiber SPR sensor with crescent-shaped holes; (**c**) Photonic crystal fiber SPR sensor with circular holes; (**d**) Microcapillary fiber SPR sensor geometry. **IV**: (**a**) Etched Fiber Bragg Grating SPR sensor; (**b**) Long Period Fiber Grating SPR sensor; (**c**) Tilted Fiber Bragg Grating SPR sensor [208].

The effectiveness of these sensors has been extensively investigated for physical (e.g., temperature, humidity), chemical (e.g., pH, gas, VOCs) and biological (e.g., DNA, proteins) sensing applications [156,208,210–214]. In the following section, we will focus on the FO-SPR sensors with different configurations developed for the detection of VOCs. Just like prism-based SPR sensor systems, most of the studies on FO-SPR sensors for the detection of VOCs have been reported after the year 2000 [2,215–225]. Only a few papers were published in the 1990s [226,227].

With the aim of achieving simple, low-cost and selective detection of aldehydes (known as cytotoxic and carcinogenic compounds) present in the environmental water, Cennamo et al. [215] developed an SPR sensor using plastic optical fiber (POF). To perform the study, butanal was used as the target VOC and porcine OBP (pOBP) as the sensing material. A plastic optical fiber consisting of a PMMA core of 980 μm and a fluorinated polymer cladding of 20 μm was used to elaborate the sensing platform. For that, the cladding of the POF along half the circumference and about 10 mm in length was removed. The exposed core was then coated with a photoresist buffer (1.5 μm thick) on top of which a 60 nm gold layer was deposited (Figure 17). For signal amplification purposes, a competitive assay was designed. For this, instead of OBP, butanal moieties were immobilized on the gold surface of the POF. Then, to test the detection performance, the sensor was exposed to OBPs pre-incubated with/without butanal. Binding events were detected by monitoring variations in the resonance wavelength. A halogen lamp with a wavelength emission range from 360 nm to 1700 nm was used as light source and the transmitted light spectrum was measured using a spectrum analyzer with a detection range of 200 nm to 850 nm. In a first step, the sensor was subjected to OBP (not pre-incubated with butanal), an increasing response was observed for increasing concentration of OBP. This result confirmed that pOBPs bind to the butanal moieties fixed on the chip, which is a prerequisite for the competitive assay. Next, the sensor was exposed to different concentrations of butanal pre-incubated with a fixed concentration of OBP. The results showed that the lower the concentration of butanal (in the pre-incubation solution), the higher the optical signal obtained. Indeed, the lower concentration of butanal resulted in more free OBPs available to bind to butanal moieties on the sensor surface. The obtained olfactory biosensor was able to detect butanal in aqueous solution for concentrations ranging from 20 μM to 1000 μM.

**Figure 17.** The olfactory biosensor using SPR based plastic optical fiber [215].

In a previous study [216] the team combined the SPR based POF platform with MIP as sensing material to achieve selective sensing of explosives such as 2,3,6-trinitrotoluene (TNT) in aqueous medium. The system exhibited a detection limit of 5.1 × <sup>10</sup>−<sup>5</sup> M and a sensitivity of 2.7 × <sup>10</sup><sup>4</sup> nm/M. The authors concluded that despite its limited sensitivity, the sensor was suitable for the detection of TNT with good selectivity. Additionally, the system was easy to prepare and suitable for rapid measurements that did not require any particular skill.

Vandezande et al. [217], designed a FO-SPR sensor for the detection of alcohol vapors. The sensor consisted of an optical fiber with a diameter of 400 μm, from which the inner technology enhanced clad silica (TECS) cladding and the outer protective cladding had been removed from the end. The exposed glass core was then coated with a 39 nm thick gold layer. Metal organic frameworks (MOFs) and more specifically zeolitic imidazolate framework (ZIF) were used as sensing materials and deposited on top of the gold layer Figure 18. MOFs consist of metal ions or a metal oxide cluster interlinked by polydentate linkers into a crystalline 3D framework. These porous materials have large surface area and tunable pore size, which are attractive features for gas and VOC sensing applications [228]. ZIFs were selected among other MOFs because of their high chemical stability and their small pore sizes. In this study, the sensing ability of two ZIF materials: ZIF-8 and ZIF-93, was explored for the detection of different alcohol vapors including methanol, ethanol, isopropanol, and n-butanol. The response of the system was expressed in terms of changes in the refractive index converted from the SPR response. This made it possible not only to monitor the mass and density changes during layer formation of the ZIFs, but also to investigate sorption behavior of VOCs on these layers. The obtained FO-SPR sensors were able to detect VOCs with ppm concentrations and with a detection limit of 2.5 ppm for methanol. However, a significant drift was observed after extended analysis periods. In this study, the authors claimed that the difference in recognition behavior of the hydrophobic ZIF-8 and more hydrophilic ZIF-93 could be exploited to generate qualitative information regarding the vapor composition.

**Figure 18.** Schematic representation of a metal organic framework FO-SPR probe, not drawn to scale [217].

Gupta's group published several studies on the detection of VOCs and other odorant molecules using FO-SPR sensors [218–220]. In one of these studies [220], they explored the sensing ability of graphene-carbon nanotubes/poly(methyl methacrylate) (GCNT/PMMA) hybrid composites for the detection of methane gas. Their sensitivity and selectivity were compared to that of three other sensing materials including reduced graphene oxide (rGO), carbon nanotubes (CNT), reduced graphene oxide-carbon nanotubes (GCNT). To fabricate different probes, 24 cm long plastic clad silica optical fibers (core diameter 600 μm, numerical aperture 0.4) were used. About 1 cm length of the cladding was removed from the middle portion of the fibers and the uncladded core was coated with a silver layer via thermal evaporation technique. Finally, the sensing materials were deposited on top of the silver by dip coating. To test the performance of the fabricated system, the probe was installed in a gas chamber and a polychromatic light from a tungsten halogen lamp was launched at the input end of the fiber. The spectrum of the transmitted light was recorded with a spectrometer at the other end. The FO-SPR sensors were exposed to different concentrations of methane (ranging from 10 to 100 ppm) and their performance was analyzed in terms of resonance wavelength shift. To evaluate their selectivity, the sensors were exposed to different gases: methane, ammonia, hydrogen sulfide, chlorine, carbon dioxide, hydrogen, and nitrogen. The FO-SPR sensor based on (GCNT/PMMA) hybrid composites showed the best sensitivity and selectivity to methane gas comparing to the three others using rGO, CNT, and GCNT as sensing materials. The authors attributed

this performance to the high aspect ratio and the large defect level in the nanocomposite material, which could provide more active sites for VOC adsorption.

Photonic crystal fiber (PCF) is a class of optical fiber characterized by a flexible structure design, which presents a unique light controlling capability with light confinement characteristics not achievable using conventional optical fiber. Combined with SPR, PCF can form a very attractive platform for optical sensing. Accordingly, Lui et al. [221] proposed a novel PCF-SPR sensor to detect mixture of methane and hydrogen. As presented in Figure 19, the PCF-SPR sensor consisted of four ultra-large side-holes symmetrically introduced into the cladding layer. These holes allowed to improve the sensitivity to VOCs since the refractive index variation due to concentration change is usually very low. In practice, the two rows of smaller air-holes along the angle of 45◦ and 135◦ surrounding the fiber enabled the introduction of the ultra-large side-holes much closer to the fiber core, which, consequently, led to higher sensitivity. The inner surfaces of the left and top ultra-large air-holes were coated with a gold layer on top of which a film of sensing material was deposited. A film of Pd-WO3 deposited via the sol-gel scheme was used for the detection of hydrogen. The methane-sensitive film consisted of a kind of ultraviolet curable fluoro-siloxane nanofilm with the inclusion of cryptophane A. It was deposited on the gold layer via a capillary dip-coating technique. The sensing performance and response of the system were characterized by analyzing the confinement loss spectra. As a result, the study showed that using polarization filtering, the concentration of methane and hydrogen in a gas mixture could be accurately measured without interfering with each other. The authors suggested that this approach could be broadened to achieve qualitative identification of multiple gases.

**Figure 19.** The schematic and cross section of photonic crystal fiber SPR sensor. (**a**,**b**) structural parameters and (**c**) experimental scheme [221].

Arasu et al. [222] reported a single mode fiber Bragg grating (FBG)-based FO-SPR sensor coated with graphene oxide (GO) layer for ethanol sensing in an aqueous medium. To fabricate the sensor, a standard single mode FBG with a 9 μm core diameter and 125 μm cladding diameter was used. The polymer coating directly over the Bragg grating was removed. Then, a 45 nm thick gold layer was deposited over the grating area without removing the cladding. Finally, a nanostructured GO layer was put on top of the gold surface by drop-casting technique. A tungsten halogen white light source was employed to generate the input signal and the output light was analyzed by a spectrometer. Wavelength

interrogation was used to monitor the response of the system upon the addition of different concentrations of ethanol in water.

In order to make sure that the FBG was effective for SPR sensing without the removal of the cladding layer, the team compared the beam profile of a gold coated FBG to that of a standard gold coated single mode fiber (SMF). The results confirmed that, in contrast to the standard SMF, the FBG was able to scatter the light from the fiber core into the cladding, producing TIR at the cladding-air interface and, thus, an evanescent wave that could be exploited for SPR. They also compared the intensity spectrum of a bare FBG, a gold coated FBG and a gold coated FBG with the GO layer, as well as the sensing performance of the last two for ethanol. It was clear that the GO layer enhanced both the sensitivity and accuracy of the FO-SPR sensor thanks to its excellent electrochemical and physical properties.

Wei et al. [223] proposed a long period fiber grating (LPFG) SPR sensor combined with a monolayer of graphene as sensing material. To fabricate the sensor, a single mode fiber with a core diameter of 10 μm, a cladding diameter of 125 μm, and a numerical aperture of 0.22 was used. The long period grating was first inscribed on the fiber core by a CO2 laser and then the SiO2 surface of the fiber was coated with an Ag film (50 nm thick) on top of which a monolayer of graphene was deposited by CVD. A schematic representation of the sensor structure is given in Figure 20a,b. To test the performance of the sensor chip, an experimental setup (Figure 20c) comprising a gas flow control system, a wide spectral range light source and a spectrometer was used. The LPFG SPR sensor was exposed to different concentrations of methane carried by a nitrogen gas flow. Wavelength interrogation was employed to monitor changes in the refractive index and thus detect variations in the concentration of VOCs in contact with the sensor. The obtained graphene coated LPFG SPR sensor exhibited a dose dependent linear response to methane and improved sensitivity compared to an uncoated LPFG sensor and an Ag-coated LPFG SPR sensor. The sensor also demonstrated good response repeatability and a baseline recovery (with a recovery time of 65 s). Finally, using finite element simulation, the team showed that the graphene layer enhanced the intensity of the electric field surrounding the sensing layer, which could explain the sensitivity enhancement observed in the presence of this layer.

**Figure 20.** (**a**) Schematic representation of the graphene-based long period fiber grating SPR sensor. (**b**) Longitudinal section of the sensor. (**c**) The experimental setup used [223].

The Table 3 summarizes the conditions for gas or VOCs detection of fiber optic SPRbased artificial olfaction systels and their performances in liquid and gas phase.


**Table 3.** Fiber optic SPR-based artificial olfaction systems for the detection of VOCs in liquid and gas phase.

#### *3.4. Grating Coupler-Based SPR Sensors*

Based on light diffraction effects, the grating coupler is another approach to excite surface plasmons. This method was first observed and described by Wood [122] in 1902. Basically, when a light wave reaches a periodically distorted metal-dielectric interface, it is diffracted into a series of beams that propagate away from the surface at different angles [147] (Figure 21). Coupling occurs when the momentum component along the interface of a scattered order is equal to the propagation constant of the SPPs. The coupling condition can by expressed as [121]:

$$\frac{2\pi}{\lambda}\sqrt{\varepsilon\_d}\sin\theta + m\frac{2\pi}{\Lambda} = \pm \text{Re}\left\{k\_{spp}\right\} \tag{6}$$

where *λ* is the wavelength of the incident p-polarized light, *θ* the incidence angle, *m* the diffraction order and **Λ** the diffraction grating period. To perform measurements using this type of SPR sensors, angular, spectral, phase or intensity interrogation can be employed.

**Figure 21.** Excitation of SPs by grating coupler.

This category of sensors is much less popular and poorly developed compared to those presented above, because they are generally less sensitive than smooth metal-film coupling based sensors (i.e., prism and optical fiber). Several theoretical and experimental studies have been carried out in attempts to improve the performance of these sensors [132,229,230]. For instance, Nazem et al. [229] recently demonstrated (theoretically and experimentally) the feasibility of a sensitive SPR sensor based on Ag-MgF2 grating. Similarly, Dai et al. [230] experimentally demonstrated a high sensitivity of an SPR sensor with silver rectangular grating coupling. A higher sensitivity than that of a prism-coupled SPR sensor was obtained in the negative order diffraction excitation mode. Borile et al. [231] reported a grating-coupled SPR senor integrated into a microfluidic chamber for label-free monitoring of cell adhesion and cell-surface interaction. Cai et al. [232] worked on the improvement of the sensitivity of grating-based SPR sensors by designing sharp dips of the higher diffraction orders and developing double-dips method. Finally, in the field of VOC sensing, Sambles's group [233,234] presented a prototype gas sensor employing SPR on gratings in the beginning of 1990s. Since then, this field has not been developed much further.

#### **4. Conclusions and Outlook**

The reliable analysis of VOCs is of great interest in various fields. To complement traditional analytical methods (GC-MS) and biological noses, great progress has been made in the development of artificial odor detection systems such as gas sensors, olfactory biosensors, and eNs based on diverse sensing technologies. As demonstrated in this paper, propagating SPR with different coupling configurations (prism coupler, wave guide, and grating) is very efficient for such applications. In particular, prism coupler-based gas sensors have been widely studied for sensing VOCs, either in the liquid or gas phase. For VOC analysis in the liquid phase, as highlighted in this review, signal amplification strategies are necessary by selecting appropriate sensitive materials and immobilization techniques to generate reliable SPR signals. In contrast, in the gas phase, the binding of small VOCs on the sensing materials can generate reliable SPR signals with good signal/noise ratios, since the detection noise remains relatively low under such conditions. Moreover, based on SPR imaging mode, a novel generation of eNs with large-scale multiplexed arrays has been developed. Combined with peptides as sensing materials, such eNs offer exceptional performance in terms of sensitivity and selectivity, with the ability to discriminate among chiral forms of VOCs. Regarding wave guide coupler-based gas sensors, most systems use optical fiber in different configurations. They are very interesting thanks to their remote and multiplexed sensing capability, as well as their miniaturized structures. Finally, grating coupler-based gas sensors are much less popular because their sensitivity is still limited.

Although the different systems that we have presented are efficient and sensitive for the detection of VOCs, the current trend is toward the development of more miniaturized sensors. Accordingly, nano plasmonic sensors based on localized SPR are attracting

more and more attention, and are being developed for different nanoscale applications including the detection of VOCs. Moreover, the sensing performance of these systems can be optimized by simply varying the size and shape of the nanostructures, which is very advantageous for sensor development. The improvement in nanofabrication processes has made it possible to explore diverse nanostructured geometries to achieve optimal LSPR nanosensors [137].

To further improve the performance (sensitivity, selectivity, and stability) of SPR-based gas sensors, olfactory biosensors, and eNs, it is essential to design novel sensing materials that are able to mimic the binding properties of biomolecules such as ORs and OBPs, but with higher stability. One trend is to use peptides as alternatives. Indeed, peptides are much more robust than proteins, cheaper to synthesize, and could potentially be integrated into industrial devices. On top of that, their selectivity towards target VOCs can be easily tuned through rational designs based on molecular modeling, virtual screening, and phage display. Finally, eNs will benefit greatly from the accelerating growth of artificial intelligence that will allow for more efficient data processing. There is no doubt that novel SPR-based gas sensors and eNs will play a more important role in the field of VOC detection and will find applications in various new domains.

**Author Contributions:** Conceptualization, M.E.K. and Y.H.; writing—original draft preparation, M.E.K. and Y.H.; revision and validation by M.E.K., Y.H., J.S.W., C.H., R.M., A.B. and N.S.; funding acquisition, Y.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** French National Research Agency (ANR) and AID/DGA supported this work (ANR-18- CE42-0012).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** The authors thank greatly the following organizations for funding the PhD scholarship of M.E.K (ANR and Labex LANEF du Programme d'Investissements d'Avenir: ANR-10- LABX-51-01, of J.S.W. (Fondation Nanosciences) and of C.H. (CEA). The authors also thank ANR and AID/DGA for the financial support (ANR-18-CE42-0012). SyMMES laboratory is part of Labex LANEF, Labex ARCANE and CBH-EUR-GS (ANR-17-EURE-0003).

**Conflicts of Interest:** Yanxia Hou is scientific counselor and co-founder of the company Aryballe, which manufactures and commercializes the portable device NeOse Pro.

#### **Abbreviations**



#### **References**


## *Review* **Plasmonic Metal Nanoparticles Hybridized with 2D Nanomaterials for SERS Detection: A Review**

**Caterina Serafinelli 1,2,3,4,\*, Alessandro Fantoni 1,3, Elisabete C. B. A. Alegria 1,2 and Manuela Vieira 1,3,4**


**Abstract:** In SERS analysis, the specificity of molecular fingerprints is combined with potential singlemolecule sensitivity so that is an attractive tool to detect molecules in trace amounts. Although several substrates have been widely used from early on, there are still some problems such as the difficulties to bind some molecules to the substrate. With the development of nanotechnology, an increasing interest has been focused on plasmonic metal nanoparticles hybridized with (2D) nanomaterials due to their unique properties. More frequently, the excellent properties of the hybrids compounds have been used to improve the drawbacks of the SERS platforms in order to create a system with outstanding properties. In this review, the physics and working principles of SERS will be provided along with the properties of differently shaped metal nanoparticles. After that, an overview on how the hybrid compounds can be engineered to obtain the SERS platform with unique properties will be given.

**Keywords:** SERS analysis; plasmonic metal nanoparticles; hotspots; hybrid materials

#### **1. Introduction**

The SERS technology fits very well in the scenario of the rapidly emerging new technologies, realizing the ultimate goal of analytical chemistry: the detection, analysis, and manipulation of single molecules, namely the single-molecule detection. For example, a growing interest is focused on groundbreaking "single-molecule electrical approaches" methods that translate chemical or physical processes into detectable electrical signals at the single-event level on the platform of single-molecule electronic devices [1]. Among the different strategies developed in recent years, graphene−molecule−graphene singlemolecule junctions (GMG-SMJs) are particularly attractive, showing great potential for the routine applications [2]. Other approaches to reach single-molecule sensitivity are focused on mechanical strategies such as Atomic Force Microscopy (AFM). Owing to the atomically well-defined tip apex and its mechanical flexibility, AFM has revealed an intriguing potential to directly characterize the molecular structure in real space with an outstanding single-molecule sensitivity, and on these bases, a new generation of AFM, the single-molecule AFM (sm-AFM), has been developed for the characterization and manipulation of single molecules [3]. Other approaches are based on optical strategies: for example, new fluorescence microscopy techniques surpassing the diffraction limit of the traditional optical microscopes are receiving a great interest. The development of superresolved fluorescence microscopy to achieve "super-resolution" led to wide applications in many scientific fields and was recognized by the Nobel Prize in Chemistry in 2014. Other approaches are based on LSPR of plasmonic nanoparticles [4]. A small overview has been depicted on the techniques aiming to reach the single-molecule sensitivity, but

**Citation:** Serafinelli, C.; Fantoni, A.; Alegria, E.C.B.A.; Vieira, M. Plasmonic Metal Nanoparticles Hybridized with 2D Nanomaterials for SERS Detection: A Review. *Biosensors* **2022**, *12*, 225. https:// doi.org/10.3390/bios12040225

Received: 2 March 2022 Accepted: 6 April 2022 Published: 9 April 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

there is still a lot of work to do for their massive use as commercial biosensing platforms. Due to its great potential to be implemented in commercial use, in recent years a growing interest has been concentrated in SERS analysis, so that on the basis of extensive research being produced to go deep inside the mechanisms and in its applications, numerous reviews have been published with the aim to help the researchers in the next steps of their studies [5,6]. However, despite the large number of works published covering a wide range of topics, only a few works have been focused on the use of hybrid structures composed of plasmonic metal nanoparticles and 2D nanomaterials in SERS analysis [7]. Stimulated by the enormous progress made on the knowledge and applications of the hybrid compounds, in this review we highlight their use in SERS analysis in order to fill the gap in this direction. The following work is organized as follows: in Section 1, an introduction illustrating the scenario of SERS analysis has been presented, whereas in Section 2, physics and working principles of the SERS technique are provided. In Section 3, the work on the early use of hybrid structures is addressed, focusing attention on spherical plasmonic metal noble nanoparticles and graphene or its derivates in SERS analysis. In Section 4, anisotropy in plasmonic metal nanoparticles is introduced and it is explained how the performances in SERS analysis are enhanced. The use of anisotropic metal nanoparticles in hybrid compounds to improve the performance of SERS analysis is depicted in Section 5. After discussing the effect of particle anisotropy in hybrid compounds, in the following sections, the different approaches to improve the performance of the SERS technique will be highlighted. In Section 6, the engineering of the bidimensional material is discussed, while considering how the hybrid compounds are evolving towards three-dimensional structures with the aim of improving the SERS properties even more in Section 7. The use of bidimensional material as a nano spacer is presented in Section 8, taking into account the use of plasmonic metal nanoparticles veiled with bidimensional material in Section 9. Considering different directions in which the research of SERS is focused, a general overview on the future of SERS will be provided in Section 10, along with the conclusion in Section 11.

#### **2. Physics and Working Principles**

Raman spectroscopy is a valuable technique to study chemical and intramolecular bonds by producing and examining inelastic scattering generated from molecules, thus providing a vibrational fingerprint, unique for each molecule, becoming a powerful tool to identify chemical species, supplying both qualitative and quantitative molecular information from any sample. Unfortunately, this is a very weak process, but it has been found that when the molecules are located near a rough metal surface or metal nanoparticles (NP), a Raman scattering boost occurs, greatly enhancing the signal intensity [8], thus opening the way to a new emerging research field: the surface-enhanced Raman scattering (SERS) spectroscopy, that has proved to be a powerful technique for non-invasive, rapid and reliable sensing of chemicals and biomolecules [9–11]. The Raman enhancement in SERS spectroscopy occurs when an incident electromagnetic field interacts with surface plasmon resonance at a metal surface (SERS effect) [12]. The unique optical properties of metal nanostructures are related to the presence of strong localized plasmon resonances (LSPR), excitation of coherent, collective oscillations of delocalized electrons in the conduction band by an external electromagnetic (EM) field as the driving force [13]. When metal nanostructures interact with a light beam, part of the incident photons are absorbed, and part are scattered in different directions: both absorption and scattering are greatly enhanced when the LSPR is excited [14]. The LSPR generated by metal nanostructures produces extremely intense and highly confined electromagnetic fields within the gaps between metallic nanostructures termed "hotspots" [15] that have been claimed to provide extraordinary enhancements of up to 10<sup>15</sup> orders of magnitude to the surface-enhanced Raman scattering (SERS) signal [16]. Generally accepted mechanisms of SERS enhancement are attributed to two effects: the first one is an electromagnetic mechanism (EM) related to the striking increase in the local electromagnetic field near the metal nanoparticle surface, whereas the second one is a chemical mechanism (CM) involving a charge transfer [17]. The EM is a process originated from the electromagnetic interaction between the metal nanoparticles and the molecules, implicating two mechanisms: the first is the result of the interaction of the metal nanoparticles with the incident beam, whereas the second one is a re-radiation phenomenon [18]. When a metal nanoparticle interacts with an incident field *E*<sup>0</sup> at a wavelength *λ*<sup>0</sup> surface plasmon resonance (SPR), oscillating dipoles are induced and the resulting polarization generates large fields around the particle. The enhanced local field *Eloc* around the metal nanoparticles generated by the interaction with incident light is proportional to the incident field *E*<sup>0</sup> and to a factor called local enhancement factor *Mloc* (*λ*0) and can be quantified as:

$$E\_{\rm loc} = E\_0 \, M\_{\rm loc} \, \text{(}\lambda\_0\text{)}\tag{1}$$

The enhanced local field *Eloc* excites the molecule inducing a dipole so that the molecule scatters the Raman signal in all directions radiating an enhanced scattered field *Escat* at a Raman wavelength *λ<sup>r</sup>* shifted from *λ*<sup>0</sup> with an intensity proportional to the molecule polarizability *α* and the enhanced local field *Eloc*.

$$E\_{\text{scat}} = \alpha \ E\_{\text{loc}} = \alpha \ E\_0 \ M\_{\text{loc}} \ (\lambda\_0) \tag{2}$$

In these circumstances, the Raman signal is already enhanced with respect to the case in which molecules do not undergo the presence of the nanoparticle at its vicinity. On the second step, the scattered field *Escat* interacts with the metal nanoparticle and as a result, it is enhanced by a re-radiation process that is expressed by:

$$E\_{\rm rad} = E\_{\rm scat} \,\, M\_{\rm rad} \,\, (\lambda\_r) = \text{a} \,\, M\_{\rm rad} \,\, (\lambda\_r) \,\, M\_{\rm loc} \,\, (\lambda\_0) \,\, E\_0 \tag{3}$$

The chemical enhancement is a process originated from a change in the molecule polarizability (along with the Raman cross-section of its vibrational modes), resulting from the physicochemical interaction between the substrate and the molecule. When the molecule interacts with the substrate by means of physisorption or chemisorption, its geometrical and electronic structures are modified, so that the Raman cross-section of its vibrational modes is different compared to that of the free molecule. The chemical enhancement arises from two different mechanisms. The first one is the *non-resonant chemical effect*: when the molecular orbitals have energies far from the Fermi level of the metal, new electronic states are not formed but the electronic and geometrical structures of the molecule are transformed, resulting in the modification of the Raman shifts and the intensity of the vibrational modes. The second mechanism is a *resonant charge transfer chemical effect*: the interaction between the metal and the molecule leads to a metal–molecule charge transfer state (CT) and in the case in which the laser source is in resonance with the CT state, the Raman modes are strongly enhanced. In addition, the chemical effect can also originate from a temporary electron transfer between the molecule and the metal, a "transient" charge transfer [19]. Although amplified by both EM and CM, SERS efficacy greatly benefits from EM enhancements, with a contribution of 108 or more, whereas the contribution of the chemical effect would not exceed a factor 100, thus suggesting that the EM enhancement is the dominant contribution to the SERS sensitivity [20]. Figure 1 presents a schematic showing the SERS signal arising from a molecule localized in the hotspot formed in the junction between two gold nanoparticles.

A key parameter in quantifying the overall signal increasing is the SERS enhancement factor (*EF*) that is experimentally evaluated by means of SERS intensity measurements for the adsorbed molecule on the metal surface, relative to the normal Raman intensity of the same, "free" molecule in the solution. One of the most used equations to calculate *EF* is expressed by:

$$EF = \frac{I\_{SERS} / N\_{Surf}}{I\_{RS} / N\_{Vol}} \tag{4}$$

where *NVol = cRSV* is the average number of molecules in the scattering volume (V) for the Raman (non-SERS) measurement, and *NSurf* is the average number of adsorbed molecules in the scattering volume for the SERS experiments while *ISERS* and *IRS* are the intensities of the same band for the SERS and bulk spectra [21].

**Figure 1.** Schematics showing the SERS signal arising from a molecule localized inside a hotspot created in the space between two plasmonic metal nanoparticles.

#### **3. Hybrid Nanocomposites**

Plasmonic nanoparticles, in condition of resonance, are able to generate a strong electromagnetic field [22] and in the past they have been exploited as excellent SERS substrates, mainly due to their huge enhancement induced by the EM effect. Despite the huge potential of SERS analysis, the low affinity of non-thiolated molecules to metals, such as gold or silver, is still a challenge because the affinity determines the retention of analytes. To bypass this complication, several approaches have been developed. For example, coreshell colloidal material comprising gold nanoparticles coated with a thermally responsive poly-(N-isopropylacrylamide) (pNIPAM) (Au@pNIPAM) has been developed as a SERS substrate [23]: the plasmonic metal core provides the enhancing properties, whereas the pNIPAM shell is exploited to trap and get the molecules close to the metal core to obtain the SERS signal. Other applications include other coatings such as calixarenes [24]. In this context, bidimensional (2D) nanomaterials such as graphene and its derivatives, 2D metallic oxides, hexagonal boron nitride (h-BN)) etc., have attracted great interest and have been investigated as SERS substrates for their intriguing properties such as an enhanced photogeneration rate, the plasmon-induced "hot electrons" and improved conductivity [25]. Between 2D nanomaterials, transition metal dichalcogenides (TMDCs) is a large family of materials of MX2 type, where M is a transition metal element from group IV, V or VI, for example, Mo or W and X is a chalcogen (S, Se or Te). They possess a layered structure, and each layer includes three atomic planes with an arrangement of X-M-X type, where a hexagonally packed plane of transition metal atoms M is enclosed within two atomic planes of chalcogen X resulting in a thickness of 6–7 Å for each single layer [26]. Inside the TMDCs family, MoS2 and Tungsten disulfide (WS2) received great interest because of their properties and have been exploited to produce advanced SERS substrates. Graphene and its derivative such as graphene oxide (GO) or reduced graphene oxide (r-GO) have been the most extensively investigated and have emerged as a material for SERS substrates, due to their intriguing properties. Firstly, its unique structure is favorable for interactions with analytes via π-π stacking and hydrophobic interactions thus facilitating the adsorption of non-thiolated molecules, resulting in an enhancement of the SERS signal. Secondly, it has been shown that it contributes to the SERS signal enhancement with a magnitude depending on the degree of GO chemical reduction [27]. The signal SERS enhancement is dominated by chemical mechanism [28] rather than the electromagnetic mechanism, although the EF is not dramatic as in the case of metal substrates. Moreover, graphene offers the additional advantage of fluorescence quenching, improving the SERS efficacy. Larger cross-sections for the fluorescent signal, compared to Raman signal, are observed, so that Raman characteristics are often interfered with, or even submerged, by the intense fluorescence background, thus lowering the quality of SERS analysis. A fluorescence quenching effect of fluorescent dyes rhodamine 6G (R6G) and protoporphyrin IX (PPP) adsorbed on graphene was first reported by Zhang [29]. The approximate evaluated quenching factor has been found on the order of 10<sup>3</sup> and the quenching effect has been assigned to a resonance energy transfer process according to the results obtained by Swathi [30]. The chemical enhancement on three types of different 2D nanomaterials, graphene, hexagonal boron nitride (h-BN) and molybdenum disulfide (MoS2) (each having different electronic properties) has been investigated by Ling et al. [31] by means of the copper phthalocyanine (CuPc) molecule as a probe. It has been observed that different vibrational modes showed different enhancement factors depending strongly on the substrates. These inconsistencies have been related to three different enhancement mechanisms determined by the distinct electronic properties and chemical bonds exhibited by the three substrates. Graphene is zero-gap semiconductor and has a nonpolar C-C bond [32], so that the Raman enhancement is assigned to the strong charge transfer with the CuPc molecule and not to the weak dipoledipole interactions due to the nonpolar nature of graphene. Differently, h-BN is highly polar (due to the strong B-N bond) and insulating with a large band gap (5.9 eV) [33] so that the dipole interactions are dominant while the charge transfer interactions are negligible, and the signal amplification results from dipole-dipole interactions between h-BN and CuPc molecules. In addition, MoS2 is semiconducting [34] and less polar compared to h-BN so that, although both dipole-dipole interactions and charge transfer occur, they are weaker thus resulting in a weaker signal enhancement. In view of the huge progress in designing and developing plasmonic metal nanoparticles as SERS substrates originating from their EM enhancement, and in the use of graphene and bidimensional nanomaterials as effective Raman enhancement substrates due to their CM, further advances in reaching greater performances in SERS analysis have been obtained by the combination of plasmonic metal nanoparticles with 2D nanomaterials. The obtained hybrid composites displayed advanced enhancement properties, arising from the synergistic effect of the EM (originating from the high local electric field at hotspots formed in metallic NPs) and CM (deriving from the charge transfer between 2D nanomaterials and probe molecule). In the early works, hybrid composites have been produced by growing spherical metal nanoparticles on graphene nanosheets. A hybrid system (GO/PDDA/AgNPs) has been developed, self-assembling Ag spherical nanoparticles on the surface of graphene oxide by means of poly (diallyldimethyl ammonium chloride) (PDDA) and successively tested as SERS platform sensing for folic acid detection, showing strong SERS activity. The SERS spectra of p-aminothiophenol (p-ATP) collected using Ag colloid and GO/PDDA/AgNPs showed a greater enhancement signal for spectrum obtained with Ag colloid due to the fact that GO did not assist the adsorption of p-ATP on AgNPs of GO/PDDA/AgNPs. In the case of folic acid detection, the SERS signals on GO/PDDA/AgNPs were much stronger than that on Ag nanoparticles [35]. Stimulated by the striking properties arising from the synergistic effect of EM and CM in the hybrid nanocomposites, Chen [36] developed a SERS platform based on p-aminothiophenol (PATP)-functionalized silver nanoparticles supported on graphene nanosheets (Ag/GNs) for the sensitive and selective detection of 2,4,6-trinitrotoluene (TNT). As a first step, graphene nanosheets were decorated with silver nanoparticles by reducing silver nitrate with sodium citrate and then the Ag/GNs composites have been functionalized with p-aminothiophenol (PATP) obtaining the PATP-Ag/GNs composites. In presence of TNT, π-π conjugated structures between TNT and PATP are created, promoting the effective charge transfer from the electron-rich PATP to the electron-poor TNT that leads to the enhanced Raman signals. With a LOD of 5.0 × <sup>10</sup>−<sup>16</sup> M, the PATP-Ag/GNs hybrid displayed a great sensitivity towards TNT detection. A SERS substrate has been prepared according to the one-step strategy in the work of Wei [37]. Hybrid structures have been produced by simultaneous reduction of GO and HAuCl4 with sodium citrate and ammonia, and then an RGO/AuNP film has been deposited on a silicon

wafer and on a poly (ethylene terephthalate) (PET) substrate to create the SERS platform. To explore the potential of the created RGO/AuNP composite, 4-aminothiophenol (4-ATP) has been exploited as probe molecule in SERS analysis, finding a value of 5.6 × 105 for the enhancement factor (EF), such that the RGO/AuNP hybrid displayed remarkable performances in 4-ATP detection. On the basis of the results obtained, the composite has been exploited as a SERS substrate to detect 2-thiouracil (2-TU) with a low concentration to 1 μM, thus confirming the great ability as a biodetection platform. However, due to their attractive properties, not only graphene has been exploited to create hybrid structures with plasmonic metal nanoparticles, but several types of 2D nanomaterials have also been used to produce composites to be utilized as SERS substrates. A SERS active substrate has been constructed by Chao [38] by growing Au nanoparticles on Molybdenum disulfide (MoS2) nanosheets. The gold precursor has been directly reduced on the MoS2 surface in the presence of carboxymethyl cellulose (CMC) as a stabilizer in an aqueous solution to create the AuNPs@MoS2 nanocomposite. Several AuNPs@MoS2 nanocomposites have been prepared with different amounts of Au nanoparticles and the one with the best performance has been selected by means of the standard probe rhodamine 6G (R6G), finding a value of 8.2 × <sup>10</sup>−<sup>5</sup> for the enhancement factor (EF). The amplification of the SERS signal has been assigned to the hot spots generated by the little aggregation and closeness of Au nanoparticles. An approach to exploit the unique properties of a nanohybrid formed by gold nanoparticles (AuNPs) deposited onto exfoliated nanosheets of tungsten disulfide (WS2) was presented in the work of Sabherwal [39]: an active SERS platform based on an Au NPs/WS2 nanohybrid has been developed for the label-free detection of Myoglobin, a cardiac biomarker. The AuNPs/WS2 nanohybrid has been prepared by the in situ reduction of gold salt precursor, and then the surface was functionalized with specific aptamers to impart high selectivity towards Myoglobin. The prepared nanohybrid has been tested for SERS detection. The obtained results showed the synergistic use of the unique properties of chemical and electromagnetic enhancement of both WS2 and AuNPs for a many fold increase in the SERS signal intensity. The AuNPs/WS2 nanohybrid system allowed the Myoglobin detection with a LOD of 10−<sup>2</sup> pg mL<sup>−</sup>1, considerably lower with respect to that measured in other works. In Table 1, the results obtained for SERS analysis using hybrids of plasmonic metal nanoparticles and graphene or other 2D nanomaterials are listed.

**Table 1.** Performances of the SERS platform based on hybrids of plasmonic metal nanoparticles and graphene or other 2D nanomaterials.


#### **4. Effect of Nanoparticle Shape**

Plasmon resonances in spherical nanoparticles can be tuned in a limited range of wavelengths by changing the particle diameter, while introducing anisotropy in the particle shape provides a fine control of plasmon resonance thus broadening the range of wavelengths from the visible through the mid IR, by varying the aspect ratio (AR) of the NPs [40]. On these presuppositions, the early attempts to further improve the SERS efficacy of the hybrid composites exploited the engineering of particle morphology tailoring the particle anisotropy along with the plasmonic properties and intrinsic electromagnetic "hotspots".

#### *4.1. Au Nanorods*

Rod-shaped nanoparticles are nanostructures where one dimension is longer than the other two, so that the term nanorods indicates elongated nanoparticles. They exhibit two plasmon resonances bands: the first one is longitudinal mode parallel to the long rod axis, whereas the second one is a transverse mode perpendicular to the long axis of the rod [41]. Different from the transverse band and characterized by a low intensity and independence from the aspect ratio (AR), the longitudinal mode is much more intense and strongly independent from AR, so that when controlling the AR of AuNRs, it is possible to tune the plasmon resonance across the UV-visible region of the spectrum [42]. From theoretical discrete dipole approximation (DDA) and experimental electron energy loss spectroscopy (EELS), near-field maps illustrated an intense EM field in the proximity of the AuNRs originating from plasmon modes [43,44]. From the calculations, results show a high electromagnetic (EM) field enhancement at the Au nanorods' tips (hotspots) under longitudinal excitation, whereas the enhancement is moderate on the NR lateral sides for transverse excitation. In addition, the AR also controls the SERS efficacy of the AuNRs: exploiting crystal violet (CV) as a molecule probe, it has been shown that the EF increases when the AR values become greater, so that controlling the AuNRs' aspect ratio enables a fine tuning of plasmon resonance and the EF for an optimized SERS analysis [45]. For the sake of having a more complete view on the properties of elongated particles, it is worth mentioning that, with a tight and precise control over the synthesis conditions, other elongated structures have been obtained, such as gold nano bipyramids. Depending on their sharp nanotips, gold nano bipyramids (AuNBPs) revealed a stronger local field enhancement compared to AuNRs [46]. Due to the sharp tips, in AuNBs the radiation can be centralized into a strong local electric field and enhancing at the same time the local density of photonic states [47], resulting in the beneficial application for spectroscopy, photocatalysis, detection and biomedicine [48].

#### *4.2. Au Nanotriangles*

The term nanoplates indicates nanoparticles in which one dimension is much smaller than the other two, and the specific case in which the base is triangular, the nanoparticles are referred as nanotriangles (NTs). Nanotriangles are characterized by intriguing properties originating from the morphology: the EM fields can be confined by their intrinsic sharp corners and edges, thus showing a strong enhancement. EES mapping indicates a strong EM field enhancement near the tips corresponding to a dipolar mode, whereas for increasing energies other modes start to appear presenting the highest enhancement at the edges and at the center of the NT [49]. It is possible to select the wavelengths of plasmon resonance by changing the aspect ratio (edge/thickness) of the AuNTs so that, as in the case of AuNRs, the LSPR of AuNTs are characterized by a fine tunability. Furthermore, depending on the synthetic procedure, AuNTs display a different grade of truncation that enable the dipolar LSPR tuning, leading to a blue-shift in the spectrum when increasing the snip size of the missing corner [50]. As in the case of AuNRs, the ability to generate great EM field enhancement and the plasmon resonance tunability by means of morphology renders AuNTs an intriguing substrate for SERS analysis. As demonstrated by Tan [51], there is a strong relationship between the SERS enhancement and the LSPR of the corresponding substrate: when laser excitation wavelengths match the LSPR, the SERS signal is found

about two orders stronger compared to the case in which the laser excitation wavelength is far away from the LSPR band.

#### *4.3. Au Nanostars*

Gold nanostars are particles with a star-like morphology comprising a spherical central core from which radial, acute tips branch out, and over the last few years have raised particular attention thanks to their features such as the ease of synthesis for large scale production, the high surface-to-volume ratio useful for improving drug loading efficiency, and above all, thanks to their unusual optical and plasmonic properties [52,53] which pave the way to a great potential for nanomedicine applications. Despite the irregularity of the particle star-like morphology, the extinction spectra show well-defined localized surface plasmon features. The UV-Visible spectra of Au nanostars display an intense band centered at ca. 650–900 nm and a weaker band localized at ca. 500–600 nm [54]. Assigning the LSPR of Au nanostar bands by means of theoretical models for the resolution of the Maxwell's equations, such as Finite-Difference Time-Domain (FDTD) [52], Discrete Dipole Approximation (DDA) [55] or Boundary Element Method (BEM) [54], produced the same results: the band at lower energy is assigned to dipolar resonances localized at the individual tip, whereas the band localized at higher energy is assigned to dipolar resonances localized at the central core. Interestingly, FDTD calculations have demonstrated, despite the general observation that the UV-Visible spectra are dominated by the LSPR band associated with tip oscillations, that the plasmon modes of Au nanostars arise from the hybridization of resonances associated with the core and the tips generating bonding and antibonding nanostar plasmon. Thus, a contribution from the core plasmon to the tip plasmon is proposed. The core plasmons have larger frequencies than the tip plasmons, so that the conduction electron of the core structure can adiabatically follow the lower frequency tip plasmon oscillations. This results in an "antenna effect" which is responsible for an increase in the extinction cross-section (a factor of 4-fold with respect to the individual tip plasmons) as well as in the electric field enhancement. In Au nanostars, the central sphere acts as an electron reservoir. Regarding the influence of the specific morphological details, it has been found that the aperture angle and the roundness of the tip are of major importance in affecting the energy of the LSPR tip mode and thus its position in the UV-Visible spectra, since small changes lead to a significant shift in the main band. BEM calculations have been confirmed when investigating the spatial distribution of the plasmon modes of gold nanostars by electron energy-loss spectroscopy (EELS) mapping performed on a single particle in a scanning transmission electron microscope (STEM) [56], which also showed a high localization near the tips. Similar to Au nanorods, the plasmon resonance in AuNSs can be tuned through modifications in the aspect ratio and/or in the tip sharpness. When the aspect ratio increases, it is possible to observe a red-shift in LSPR, and larger Au nanostars present an increased number and longer tips along with a red-shift in plasmon resonances at an increase in AuNS size. Despite the great potential, understanding how the morphology can impact the efficacy of AuNSs as a field enhancer is a controversial task. In an early work, Vo-Dinh [57] investigated the properties of AuNSs as SERS substrates and compared with respect to the size. Changing the ratio between seeds and HAuCl4, PVP-functionalized AuNs with sizes in the range from 45 nm to 116 nm have been synthesized. Even though the morphology of AuNSs with distinct sizes is different, exploiting p-mercaptobenzoic acid (p-MBA) as a probe molecule, EF around 5 × <sup>10</sup><sup>3</sup> has been achieved, thus revealing no great differences among the different sizes. Nevertheless, from the results obtained by Ganesh [58] it has been found that when the size and length of the spikes of Au NS have been changed, the intensity of the SERS signal is significantly altered. Exploiting AuNSs with two different morphologies, short spike (SSNS) and long spike (LSNS) Au nanostars, from the SERS experiment demonstrated that SSNS exhibit a higher SERS enhancement compared to LSNS, and it has been justified with a size effect. According to the work of Hong and Li [59], the optimized size of Au nanoparticles for obtaining maximum enhancement in Raman signal is around 50 nm and the core size

of SSNS falls within this regime exhibiting higher intense Raman signals. Regardless of the controversy in correlating the AuNSs morphology to SERS activity, their own optical properties such as the strong near field enhancement and plasmon resonance tunability enable Au nanostars to be employed in SERS analysis with great potential.

In Figure 2, the electromagnetic field around a gold nanoparticle with a different morphology shown.

**Figure 2.** (**a**) Electromagnetic (EM) field (bright spot) around a gold nanostars' (AuNS) spike and (**b**) EM field around a gold nanorod (AuNR) tip.

With the aim to evaluate how the nanoparticles' morphology can affect the SERS activity, Kundu, in his work [60], used rhodamine 6 G (R6G) as probe molecule exploiting nanoparticles with different shapes (Au nanospheres, Au nanorods, Au nanowires and Au nanoprisms) as SERS substrates. Calculating the enhancement factor (EF) according to Equation (1), it has been observed that Au nanoprisms showed much higher value of EF compared to other shapes as follows: nanoprisms > nanowires > nanorods > nanospheres. The enhanced SERS activity for nanoprisms has been related with the presence of a greater number of edges: the largest electric field is localized near the sharpest surface or at the sharp ends of the NPs so that Raman enhancement reaches its maximum value at the sharpest surface and, in addition, edges can interact strongly with R6G compared to the smooth surface of the nanospheres. The synergistic effect of EM field enhancement and strong interactions with probe molecules results in the highest EF values. Similar results have been obtained by Wu [61]. The effect of particle shape has been investigated choosing three types of Au nanoparticles (nanospheres, nanorods and nanostars) as SERS substrates and using malachite green isothiocyanate (MGITC) molecules as probes. From the spectra collected under excitation from the 532 nm and 785 nm lasers, it was shown that Au nanostars displayed the highest SERS enhancement factor (EF) while the nanospheres possessed the lowest SERS activity under excitation with 532 and 785 nm lasers. The experimental results have been combined with theoretical calculations in order to provide a deep insight into the relationship between the particle morphology and its SERS activity. The field distribution has been investigated by means of finite-difference time-domain (FDTD) simulation, revealing considerable differences in the distribution of the EM field around the diverse nanoparticle types induced by their localized surface plasmon resonance (LSPR) under both 532 and 785 nm incident lasers. From FDTD simulations, it can be seen that the maximum electric field intensity is concentrated around the sharp tip in the anisotropic structure generating 'hot spots' thus leading to the dominant contribution to the SERS intensity. According to the particle morphology and anisotropy, the Au nanoparticles follow the order nanostars > nanorods > nanospheres in terms of enhancement factors. The size and morphology of plasmonic metal nanostructures have been discussed as key factors in determining the formation of the hotspots' network in a SERS substrate, but other factors, such as their spatial arrangement and nanogap distances need to be controlled in order to obtain optimal plasmonic properties and maximal Raman signal

amplification. When plasmonic metal nanostructures are connected, a direct exchange of free electrons is possible producing different types of plasmons, such as the longitudinal antenna plasmon mode (LAP), bonding dipole plasmon mode (BDP) and charge transfer plasmon mode (CTP) [62]. The type of plasmon created (BDP, CTP or LAP) will affect the performances of the SERS substrate. The particle surface morphology is a further key factor to be considered to control the performances of the substrate sensing. In addition, the shape of plasmonic metal nanostructures has been engineered to obtain an enhanced EM field that is beneficial to the SERS analysis, but it is worth mentioning other approaches aiming to obtain promising results in this direction and that are based on the combination of LSPRs with propagating surface plasmons (PSPs). Often, surface plasmons (SPs) are classified in two categories: PSPs, that are running surface waves, and localized surface plasmons (LSPs), collective surface charge oscillations in the form of the standing waves confined in a metal nanoparticle [63]. The incident light cannot easily excite the PSPs and in order to achieve SPRs' excitation, a specific configuration is required and needs to be assembled [64]. As opposed, the LSPRs are excited by directly shining the light on an assembly of plasmonic metal nanoparticles. Very recently, it has been found that the enhancement factors in SERS analysis are greatly enhanced when the excitation of LSPs is produced by PSPs generated as surface EM waves, and excited from a special coupling medium (prism, waveguide, fiber, or grating) [65], thus paving the way to the development of new devices.

#### **5. Effect of Plasmonic Metal Particle Morphology in Hybrid Compounds on SERS**

Taking into account the effect of particle size on the SERS efficacy, it has been easy to combine anisotropic nanoparticles with bidimensional material to further improve the SERS efficacy of the hybrid composites thus exploiting the engineering of particle morphology by tailoring the particle anisotropy along with the plasmonic properties and intrinsic electromagnetic "hotspots". In Figure 3, a schematic of a hybrid compound formed by a graphene layer and an Au nanorods with SERS signal is represented.

**Figure 3.** Au nanorods under the illumination of a laser beam and the resulting SERS signal.

In the work of Liu [66], for the first time, the fabrication of a hybrid composed of AuNRs and GO for SERS (GO–AuNR) analysis has been reported. The GO–AuNR composite material has been developed exploiting an electrostatic self-assembly strategy and then tested as a SERS substrate. To test the SERS efficacy of the composite system, the Raman signals of four dye molecules (crystal violet (CV), neutral red (NR), trypan blue (TB) and ponceau S (PS)) on different substrates (SiO2/Si, AuNRs, GO and GO–AuNR) have been compared. From the results obtained, it has been possible to observe that strong SERS activity is observed when CV and NR are deposited on the GO–AuNRs substrate whereas no SERS effect has been detected for TB and PS dye molecules. In the case of cationic dye (CV and NR), the molecules are electrostatically attracted by the negatively charged GO–AuNRs substrate, whereas the anionic dye TB and PS due to their negative charge

are prevented to interact with the substrate, thus weaking the Raman signal. Although exploiting hybrid composites containing AuNRs has revealed improved performances in SERS analysis, the substrate is still not properly adequate for detecting real samples, that can be negatively charged. Further improvement has been reported in successive works. For example, Jiang et al. [67] fabricated a composite containing aggregated Ag nanorods and GO showing good stability and used as a SERS platform to detect Rh6G molecules and trace I− ions in the solution exploiting the SERS quenching due to the formation of the Rh6G-I complex thus finding a LOD of 0.2 nmol/L for Rh6G and a LOD of 0.004 μmol/L. A different strategy exploiting core–shell Au@Ag nanorods has been developed by Gao et al. [68] to obtain an advanced SERS substrate. Using Au@Ag core–shell nanorods hybridized with reduced graphene oxide (GO-Au@AgNRs) Rh6G molecules have been detected with an enhancement factor (EF) up to (5.0 ± 0.2) × 108, 4-fold higher compared to that obtained with rGO-AuNRs and pesticide thiram with a limit of detection (LOD) of 5.12 × <sup>10</sup>−<sup>3</sup> <sup>μ</sup>M has been revealed. An early hybrid system containing Au nanotriangles has been developed by Jiang [69]. Reduced graphene oxide/silver nanotriangle (rGO/AgNT) composite sol was prepared by the reduction of silver ions with sodium borohydride in the presence of H2O2 and sodium citrate and exploited for SERS detection of Dopamine (DA). The detection is based on the competitive adsorption occurred between DA and the molecular probe acridine red (AR) onto the reduced graphene oxide (GO) nanosheets. Depending on the formation of multiple hydrogen bonding and π-π stacking, DA molecules display a much stronger affinity towards the GO nanosheets compared to AR molecules, so that, when added to the system, DA molecules competed with AR for similar adsorption sites on the rGO surface thus leading to the desorption of AR molecules from the rGO surface. The desorbed AR molecules can be successively adsorbed on the surface of Au nanotriangles thus enhancing the SERS signal: when increasing the DA concentration, the amount of AR molecules adsorbed onto AuNTs rises with SERS intensity responding linearly with DA concentration. Stimulated by intense local electric field enhancements of silver nanoplatelets (Ag-NPs) caused by the anisotropic morphology and sharp corners compared with other morphologies, such as nanospheres, Meng [70] developed hybrid systems containing AgNPs and graphene nanosheets (Ag-NP@GH) to be exploited as SERS substrates taking advantage of their unique properties. The SERS enhancement capability of the Ag-NP@GH composite has been tested calculating the enhancement factor (EF) using rhodamine 6G (R6G) as a probe molecule and a value of 4.7 × 108 for EF has been found, thus confirming the good performance of Ag-NP@GH in SERS analysis. After testing, the composite has been exploited for organic pesticide detection, including thiram and methyl parathion (MP), and their mixtures finding a LOD of 40 nM for thiram and 600 nM for MP. The improved sensitivity of the Ag-NPs@GH results from the combination of the EM effect of the AgNPs and the CM effect of the graphene. Hotspots are generated by AgNPs staying close to each other, whereas graphene nanosheets contribute to SERS efficacy with a chemical enhancement (CM) due to the strong adsorption capability by means of π-π interactions. Aiming to exploit the unique plasmonic properties of gold nanostars (AuNSs), in the work of Krishnan [71] hybrid systems containing graphene oxide (GO) and AuNSs have been produced and used as SERS-active substrates. A simple and eco-friendly synthetic route based on a deep eutectic solvent (DES) has been developed and evaluated as a SERS substrate using crystal violet (CV) as a probe molecule and a value of 1.7 × 105 for the enhancement factor EF and a limit of detection (LOD) of 10−<sup>11</sup> M have been found. The improved performances of the composite as a SERS substrate have been explained with the large number of nanogaps between two contiguous AuNSs generating the SERS hotspots and to its morphology, permitting to the CV analyte molecules to diffuse inside the structure. One of the early hybrid structures comprising branched Au nanoparticles was developed by Ray in his work [72] in a four-step process, binding the Au-branched nanoparticles, (termed Au nanopopcorns) by means of Cysteamine molecules on graphene oxide (GO) nanosheets. To evaluate the SERS enhancement capability of the composites, SERS spectra were collected using Rh6G dyes as probe molecules and GO, Au nanopocorns and hybrid nanostructures

as substrates. Calculating the enhancement factor (EF), it has been found that the higher value of EF is for the composites, followed by Au nanopopcorns and GO. The signal enhancement in the case of GO has been explained with the chemical mechanism, whereas the effect of nanoparticles has been further investigated. When exploiting hybrid composites containing differently shaped (spherical, cage and popcorns) Au nanoparticles, the highest EF for the popcorn shape has been found, depending on the presence of sharp tips. As well as in combination with GO, the differently shaped Au nanoparticles showed the highest EF for nanopopcorns. The hybrid structures containing Au nanopopcorns revealed the best performances in the HIV-1 gag-gene DNA and Staphylococcus aureus (MRSA). Animated by the intriguing properties of hybrid structures, Li [73] developed a SERS substrate for bilirubin detection integrating composites containing AuNS-decorated graphene oxide (GO) nanosheets on common filter paper. The AuNSs/GO hybrids have been assembled by means of electrostatic interactions created by a deposited layer of Poly (diallyldimethyl ammonium chloride) (PDDA) on GO nanosheets and then exploited for bilirubin detection. The resulting SERS substrate combines the EM effect originated by hotspots generated from AuNSs and the ability of GO nanosheets to adsorb bilirubin molecules by means of strong electrostatic and π-π interactions as shown by kinetics measurements. In addition, the SERS performances are improved by the superquenching of fluorescence by both GO nanosheets and GNSs. The SERS substrate showed an LOD as low as 0.436 μM for free bilirubin in blood serum, thus holding considerable properties for clinical translation in accurate diagnosis of jaundice and its related diseases. In Table 2, the performance of hybrid compounds containing metal nanoparticles with different morphologies is reported.

**Table 2.** Performances of the SERS platform based on different shaped nanoparticles.



**Table 2.** *Cont.*

The effect of nanoparticle morphology has been reviewed in this section, but it is noteworthy to mention that in other works, structures formed by plasmonic metal nanohole arrays hybridized with graphene have been exploited with promising results in SERS analysis [74,75].

#### **6. Engineered 2D Nanomaterials**

Tailoring the anisotropy in nanoparticles morphology amplifies the SERS signal by means of EM field intensification near the nanoparticles; nevertheless, the SERS enhancement also relies on the number of the hotspots created by the structure used as a substrate, so that alternative approaches optimize the hybrid composite structure in such a way as to create the greatest possible number of hotspots by engineering the bidimensional material according to different strategies. Some strategies, for example, exploited the intrinsic properties of the bidimensional material to amplify the SERS efficacy. Recently, in the work of Su [76], a SERS substrate composed of 1T-MoS2 nanosheets decorated with silver nanocubes (1T-MoS2/AgNCs) and assembled on filter paper has been developed and used for thiram (TRM) and thiabendazole (TBZ) residues in apple fruit detection. The two different phases of MoS2, the 1T (metallic, trigonal) and 2H (semiconducting, hexagonal), have been used to create the hybrids with Ag nanocubes (1T-MoS2 /AgNCs and 2H-MoS2/AgNCs) and tested in SERS analysis with the model molecules rhodamine 6G (R6G) in order to evaluate the effect of the MoS2 phase on the SERS performances. The best performances obtained for the composites containing the metallic 1T phase have been explained with a superior electron transfer from Ag to 1T-MoS2 compared to the 2H-MoS2 phase, depending on the absence of band gap, the lower binding energies of 1T-MoS2 compared to 2H-MoS2 and the

more abundant density of state (DOS) near the Fermi level. In the work of Koratkar [77], a SERS substrate composed of monolayer MoS2 decorated with AuNPs has been engineered to obtain higher SERS analysis performances. By means of low-power focused laser cutting, artificial edges have been sculpted in monolayer MoS2, on which AuNPs, when deposited by drop-casting, tend to predominantly accumulate. The huge density of AuNPs along these artificial edges concentrates the plasmonic effects in this region, so that hotspots are generated exclusively along these artificial edges. Calculations of first-principles density functional theory (DFT) suggested that AuNPs are strongly coupled to the artificial edges through dangling bonds that are widespread along the unpassivated edges cut by the laser. Moreover, according to DFT calculations, as a result of AuNP binding, there is an enriched availability of conduction channels around the Fermi level so that artificial edges decorated with AuNPs displayed a higher electrical conductivity. The dense assemblage of AuNPs and the increased electrical conductivity generate along the artificial edges regions of mobile charge oscillating in phase with the laser light that drastically enhanced the magnetic field and the SERS response. Using Raman mapping it has been possible to localize the hotspots along the MoS2 edges cut by the laser. Inspired by the intrinsic properties of boron nitride (BN) nanosheets, Li et al. [78] have developed a SERS substrate composed of faceted Au nanoparticles synthesized over BN nanosheets by a simple sputtering and annealing method. The stronger resistance to oxidation renders more advantageous the use of BN nanosheets as reusable SERS substrates as they support the heating at high temperatures in air necessary to remove the analyte molecules for reusing. Furthermore, different from graphene which introduces intrinsic Raman band of high intensity in SERS spectra [79], BN nanosheets only display a Raman G band [80] of a low intensity that are barely enhanced by AuNPs, so that interferences are not created and only Raman signals from analytes are shown. The performances in SERS analysis have been tested using rhodamine 6G (R6G) as the probe molecule and silicon oxide (SiO2/Si), atomically thin BN and bulk hBN substrates decorated with AuNPs obtaining the greater enhancement signal for structures comprising BN nanosheets. Even though AuNPs are able to adsorb a certain quantity of R6G molecules, as a result of π-π interactions the BN surface can adsorb a drastically higher number of molecules that are localized in the hotspots between AuNPs, thus enhancing the SERS signal. As expected on the basis of the stronger resistance to oxidation of BN, the experimental results confirmed the ability of hybrid composites to be reused as SERS substrates, being able to sustain multiple thermal regeneration cycles. The performances in SERS analysis for hybrid compounds containing engineered 2D nanomaterials are reported in Table 3.

**Table 3.** Performances of the SERS platform based on hybrid compounds containing engineered 2D nanomaterials.


#### **7. Three-Dimensional Structures**

Despite the huge steps forward, the development of the SERS platforms is still mainly limited to plasmonic metal nanoparticles hybridized with bidimensional nanomaterial, but three-dimensional (3D) nanostructures start to be used to create hybrid materials. In his work [81], Yin prepared three-dimensional MoS2 nanohybrids according to the microwave irradiation hydrothermal synthesis strategy (3D MoS2-NS@Au-NPs) and the system created has been compared with bidimensional MoS2/Au nanoparticle hybrids and tested as a SERS platform for melamine in milk detection. From comparison, it can be seen that the SERS activity of 3D MoS2-NS@Au-NPs structures is improved by almost 56.4-fold in EF compared to 2D MoS2-NST@Au-NPs hybrid structures. The amplification of EF has been ascribed to the larger surface area for adsorbing probe molecules and the higher number of hot spots generated to benefit the SERS performances supplied by the three-dimensional structure. Once optimized, the effectiveness of the generated hot spots by tuning the size and the density of the Au nanoparticles of the composite, the optimized structures have been tested for melamine quantitative detection in milk, finding a LOD of 1 ppb, a value lower than the maximum level of melamine as 2.5 ppm in food prescribed by the U.S. Food and Drug Administration. In the same year, Wang et al. [82] fabricated hierarchical MoS2 microspheres (MoS2-MS) decorated with "cauliflower-like" AuNP arrays (CF-AuNPs), by means of a new synthetic route, which have been successively investigated and tested for SERS analysis. According to the obtained results, it is possible to tailor the average size of CF-AuNPs@MoS2-MS nanocomposites by tailoring the molar ratio between MoS2-MS and HAuCl4 and, once optimized to achieve the best performances in SERS analysis, they have been tested for molecular detection. In addition to R6G and methylene blue (MB) molecule sensing, that showed an LOD, respectively, of 10−<sup>14</sup> M and 1015 M, the composites have been tested for the detection of various metabolites in human early morning urine with promising results. Furthermore, the CF-AuNPs@MoS2-MS nanocomposites have been inserted in cellulose acetate membrane (CAM) to fabricate flexible wafer-scale flexible SERS substrates. From the results obtained by exploiting three-dimensional nanostructures hybridized with metal plasmonic nanoparticles as substrates in SERS analysis, it has been shown that the main advantage of these composites resides in their huge surface area. In fact, due to their large surface area the composites can adsorb a higher number of probe molecules compared to composites containing bidimensional materials and, in addition, are able to generate hotspots distributed in the space that are more efficient in enhancing the SERS signal by means of electrochemical mechanism. The effect of surface area in SERS analysis efficacy has been explored by Singh [83]. By means of facile hydrothermal method, a series of MoS2 nanoflowers with a surface area ranging from 5 m2/g to 20 m2/g has been synthesized and tested in SERS analysis using the R6G as a probe molecule. It has been found a linear dependency of SERS signals originating from different substrates with the surface area, thus correlating the surface area of the composites with the intensity of the Raman signal. In Table 4, the performances of the hybrid compounds with a threedimensional structure are reported.


**Table 4.** Performances of the SERS platform based on hybrid compounds with a three-dimensional structure.

#### **8. Nanospacers**

Other approaches introduced bidimensional nanomaterial as a nanospacer between layers of plasmonic structures in order to create dense three-dimensional hotspots that support the striking SERS enhancement. In the early work of Zhu [84], the interactions between light and the sub-nanometer gap were systematically investigated. With a simple fabrication technique, a structure composed of graphene sandwiched between two layers of vertically stacked Au NPs has been developed and investigated as a SERS substrate. Performing numerical simulations based on the Finite element method (FEM) to investigate the effect of graphene in the sandwich structure, it has been found that the electric field is strongly amplified in the gap defined by the graphene film between two vertically stacked layers of AuNPs, leading to electric field enhancement of up to 88 times, much higher compared to that of 14 times in the horizontal gaps between Au nanoparticles without graphene. In addition, by changing the number of graphene films it is possible to control the nanogap between the vertical AuNPs layers, and, as consequence, the SERS enhancement. To investigate the effect of the gap induced by a different number of graphene layers, composites of two vertical layers of AuNPs containing a various number of graphene film have been created and used as a substrate in SERS analysis, obtaining the best performance for structures containing graphene monolayer, thus deducing that the coupling between the layers of plasmonic structures decays exponentially with their distance. Using the composite containing the graphene monolayer (4 nm Au/1LG/4 nm Au) as a SERS substrate, an LOD of 10−<sup>9</sup> M for rhodamine B (RhB) has been found, hence showing a higher sensitivity compared to 4 nm Au/4 nm Au films, that showed an LOD of 10−<sup>7</sup> for RhB molecules when used as substrates. The same results have been obtained for R6G molecules, with composites containing graphene resulting in a higher sensitivity. The use of 4 nm Au/1LG/4 nm Au in practical applications has been tested on by means of Sudan III and methylene blue detection, finding an LOD of 0,1 nM, thus showing a potential use in areas of food safety, medical diagnostics, biological imaging and environmental pollutant detection. Similar results showing the extraordinary performances in SERS enhancement obtained by exploiting graphene as a nanospacer have been obtained in the work of Man [85]. In this work, Au nanoparticles, (AuNPs), silver nanoparticles (AgNPs) and graphene have been combined in order to form a sandwiched structure, AgNPs/graphene@AuNPs, to be exploited in order to achieve unique performances in SERS

analysis. By means of rhodamine R6G and crystal violet (CV) molecules, the composite has been experimentally tested in SERS analysis generating a huge amplification in Raman signal intensity. The excellent signal SERS enhancement obtained has been explained as the combination of chemical mechanism (CM) and electromagnetic mechanism (EM). The graphene film induced the CM, thus enhancing the SERS activity, but can also act as a nanospacer able to control the hot spots' size by changing the number of graphene layers. In fact, the electromagnetic enhancement was the result of three-dimensional hotspots generated by lateral nanogaps (AuNPs-AuNPs, AgNPs-AgNPs) and vertical nanogaps (AgNP-AuNPs), tunable by graphene layers. To investigate the effect of the graphene film, different composites containing a different number of graphene layers have been explored using R6G molecules as probes, finding that the graphene bilayer offers the best performances. A single layer of graphene induced the plasmon tunneling phenomenon due to the short distance (<0.5 nm) between Ag and Au nanoparticles which reduces the plasmonic coupling effect weakening the Raman signal, whereas for higher numbers of graphene layers, the nanogap also increases so that the electromagnetic enhancement is reduced because the enhanced local electric field will exponentially decay with distance. This composite system AgNPs/graphene@AuNPs has been tested for Malachite green (MG) detection in sea water finding an LOD of 10−<sup>11</sup> M, thus demonstrating a potential ability in practical applications. Motivated by the few studies exploiting WS2 bidimensional nanomaterial in SERS analysis and on the presumption that, due to its structure, WS2 could promote both chemical enhancement (CM) by means of charge transfer between substrate and probe molecules, and electromagnetic enhancement (EM) by means of the strong coupling between WS2 and metallic nanostructures through surface plasmon excitation, thus enhancing the SERS signal, Jiang in his work [86] exploited bidimensional WS2 nanomaterial as a nanospacer in hybrid nanostructures. A remarkable SERS platform based on AuNPs/WS2@AuNPs nanohybrids has been designed and developed in a multi-step process. Firstly, annealing an Au film deposited onto a SiO2 substrate, a layer of Au NPs has been created. Successively, by means of a thermal decomposition process, a bilayer WS2 film has been grown onto the AuNPs surface, and finally, a second layer of AuNPs was deposited onto the WS2 film by means of a further annealing, thus obtaining the AuNPs/WS2@AuNPs composites. Introducing the bilayer WS2 film as a nanospacer between the two layers of plasmonic structures, a highly enhanced local electromagnetic field has been generated. Dense 3D hotspots occurring through this hybrid plasmonic nanostructures are responsible for the greatly enhanced SERS performances. Using rhodamine R6G as a probe molecule to test the performance in SERS analysis, the AuNPs/WS2@AuNPs nanohybrids showed an excellent sensitivity with the minimum detectable concentration of 10−<sup>11</sup> M. In addition, the AuNPs/WS2@AuNPs nanohybrids showed extremely satisfying performances in detecting other probe molecules such as crystal violet (CV) molecules. The results obtained in the presented works illustrate the role of bidimensional nanomaterial used as a nanospacer between the layer of plasmonic metal nanostructures in the enhancement of the SERS signal. When used as a nanospacer, the bidimensional nanomaterial is able to create three-dimensional hotspots generated by the combination of lateral nanogaps (gaps inside plasmonic layer) and vertical nanogaps (gaps between plasmonic nanolayers). The great benefit of using a nanospacer is the possibility to finely tune the vertical gap by changing the number of bidimensional nanomaterial layers and taking into account that an exponential decay controls the coupling between two plasmonic layers, so that increasing the nanogap with a higher number of 2D nanomaterial layers reduces the SERS enhancement, but also for too small nanogaps, the Raman signal is weakened by the plasmon tunneling phenomenon. The performances of hybrid structures containing nanospacers in SERS detection are listed in Table 5.


**Table 5.** The performances of the SERS platform based on hybrid compounds engineered with nanospacers are listed.

#### **9. Bidimensional Nanomaterials Used to Veil AuNP Arrays**

In the majority of the hybrid composites that have been developed, the bidimensional nanomaterial acts as a support for the plasmonic metal structures; however, this configuration only provides a limited number of contact points between 2D nanomaterials and metal nanostructures: to attain a large enhancement of the SERS signal, an efficient contact between the metal framework and the bidimensional nanomaterial is necessary [87]. On this basis, different strategies wrapped the 2D nanomaterial around plasmonic metal nanoparticles in such a way that the number of contact points is increased, thus optimizing the Raman enhancement. A SERS substrate has been created veiling an array of silver nanocubes (AgNCs) with a graphene oxide (GO) film by means of a simple GO deposition process based on GO self-assembly on the metal surface [88]. According to the finite element method (FEM) calculations, the maximum intensity is 70% reduced when a 7 nm-thick GO layer is added on the metal structure surface and a more spread E-field distribution along the cluster edges and at the interface between particles is generated after supporting a 7 nm-thick GO layer on AgNCs, in contrast with well-localized hot spots observed for bare cubes. In conformity with FEM calculations, a reduction in SERS efficacy is expected, but the experimental results pointed out a superior SERS activity including more resolved peaks with higher signal intensity and larger reproducibility. The excellent SERS activity has been explained with a chemical enhancement deriving from a combination of π-π interactions and charge transfer from the oxygen-rich functional groups of GO to the probe molecules and in addition, with the GO ability to catch different compounds that therefore are accumulated on its surface thus intensifying the SERS signal. Comparing the performances, it has been found that wrapping plasmonic nanostructures enables a greater enhancement of the SERS signal than supporting on the bidimensional material. Intrigued by the great potential of Au nanoparticles hybridized with graphene nanosheets as SERS substrates, Cerruti [89] produced Au nanostars wrapped by graphene oxide (GO) nanosheets which further improved the SERS platform. Previously synthesized AuNSs have been functionalized with the positively charged Cysteamine, that create electrostatic interactions with negatively charged GO, so that GO-wrapped AuNSs (Au NSt@nGO) have been produced and tested in SERS analysis. As a result of AuNSs being wrapped with GO, the Raman signal of nGO by 5.3-fold compared to samples in which nGO is in contact with the nanostars but does not wrap them, whereas there is a higher enhancement for wrapped AuNSs, thus confirming the efficiency of wrapping to improve the SERS signal. SERS signals of typical Raman reporter such as rhodamine B (RhB), crystal violet (CV) and R6G sandwiched between AuNSs and GO nanosheets were higher compared to the signal obtained when the molecules are adsorbed on the nanostar surface. Together with an increase in SERS efficiency, wrapping AuNSs with GO results in a greater physiological stability, depending on a prevented RhB desorption in physiological conditions. A more

detailed interpretation of SERS efficacy generated by plasmonic nanoparticle wrapping with a bidimensional material is given in the work of Chen [90]. Inspired by the strong surface adsorption of airborne hydrocarbon and aromatic molecules of thin boron nitride (BN) nanosheets, a SERS platform has been created placing an atomically thin BN nanosheet over an Au nanoparticle array produced via physical processes. Using R6G as probe molecules and BN nanosheets with different thicknesses, it has been found that Raman signals were most prominent for the lower thickness of BN, but reduced when the layer thickness increased and the stronger Raman signals were attributed to hotspots. Thinner BN nanosheets, due to their greater flexibility are able to better conform to the underlying AuNPs so that the analyte molecules were closer to the plasmonic hotspots. For increasing thickness, the BN nanosheets were much less deformed so that analyte molecules were more distant from the plasmon-induced EM field which decays exponentially with the distance. In addition, more R6G molecules were attracted depending on the strong adsorption ability of BN nanosheets towards aromatic molecules by means of π-π interactions. Conformational changes explain the stronger adsorption capability, with BN nanosheet polarity not contributing to such effect. Wrapping plasmonic metal nanostructures with graphene nanosheets has revealed great potential, thus expanding applications of hybrid composites from SERS detection to SERS bioimaging. Graphene oxide (GO)-wrapped Au nanorods (GO@GNRs) have been created by Wu [91]. Assessing the cytotoxicity showed a greatly enhanced biocompatibility of GO@GNRs, provided by the encapsulation enabling a reduced contact with the surrounding environment, thus decreasing the amount of residual CTAB that induces cytotoxicity. The SERS activity in the near infrared (NIR) has been investigated using six dye molecules as probes showing extremely intense SERS signals and highly enhanced activities of NIR SERS multiple effects, such as Au nanorods LSPR, the charge transfer between graphene nanosheets and probe molecules and the enrichment of dye molecules on the GO sheets. The enhanced NIR SERS activity and the improved biocompatibility enable a successful application of GO@GNRs as a robust nanoplatform for ultrafast NIR SERS bioimaging. On these bases, exploiting bidimensional materials to veil arrays of plasmonic nanostructures produces a greater SERS activity compared to that obtained when 2D material is used as a support. Due to its flexibility, the bidimensional nanomaterial is able to bring analyte molecules into the proximity of the hotspots originated from plasmonic structures thus enhancing the SERS efficacy. Moreover, using 2D nanomaterials as veiling medium offers the additional advantage of protecting the arrays of plasmonic nanostructures, thus increasing the stability of the SERS substrate and broadening the composite range of applications. Wrapping plasmonic nanostructures imparts a biocompatibility to the system, thus that can also be exploited in the biomedical field.

#### **10. Future Perspectives**

In order to give a better understanding of its enormous potentiality, the future perspectives of the SERS technique will be provided considering a range of directions in which the research in this field is focused. Since its development, SERS has affirmed a great potential as a powerful technique to detect simple and more complex molecules, in contact or adjacent to a plasmonic substrate that generally consists of a metal surface, and recently, also of hybrid materials. Such enormous potential, combined with advances in the development of associated instrumentation has produced an outbreak of research, moving forward many different applications ranging from materials and environmental science through biology and medicine. Some clinical implementations of SERS that could find an application in the near future are: 1) the detection of tumor margins during surgery [92], as it has been demonstrated that tags were sufficient to improve tracking of tumor margins employing a portable Raman microscope instead of a benchtop instrument; 2) in optical fiber-guided imaging procedures such as endoscopy, colonoscopy, or others used to detect and to visualize superficial diseased tissues within the body [93]; and 3) in liquid biopsy, a term widely comprising the identification of disease biomarkers in blood or other bodily fluids [94]. Liquid biopsy holds great potential to simplify disease detection and

monitoring and make it less painful for the patient. In particular, the monitoring of the disease progresses and the response to therapy would ideally be allowed on a daily basis, thus avoiding the dependence on imaging approaches that could not detect changes in the size of the tumor. The SERS analysis has also been revealed as a valid tool to detect inorganic and highly toxic organic pollutants as well as to monitor bacterial contaminant with a detection threshold in the parts for a billion range. In food analytics, detection limits in quality control and nutrient quantification down to the nanomolar range have been reached. The difficulties in analyzing the surface residues, an issue of significance in a variety of areas including health and safety, homeland security, forensics, etc., paved the way to the development of flexible SERS substrates, still a young research area within the progress of the SERS technique. The greater part of recent flexible substrates designed for point of care analysis can be classified according to two categories: sticky "SERS tapes" and adsorptive "SERS swabs". Typically, SERS tapes are adhesive and flexible plastic films, bearing plasmonic particles on the surface, that can be pressed and peeled from the sample surface to extract the molecules for in situ analysis [95], whereas SERS swabs can be exploited to collect chemical compounds by dabbing the surface of the sample [96]. A main obstacle in the present applications of SERS is the difficulty in analyzing complex real samples, containing a large variety of chemical species and micro/macro-contaminants in addition to the target analyte molecules. Often, these impurities are present in much higher concentrations than the analyte, thus they can interfere with the analysis resulting in significantly reduced sensitivity and reproducibility. The method developed to address this issue still requires sophisticated equipment for the pre-treatment of the samples so that the analyses are restricted to a laboratory setting and must be performed by trained professionals. In this scenario, the difficulties in analyzing SERS spectra from complex samples for disease diagnosis and food analytics pave the way to the implementation of Artificial Intelligence [97]: the acquired SERS spectra of such samples could be spectrally unmixed so that the concentration profiles for better quantification could be estimated. The algorithm offers the additional advantages to characterize the disease while evaluating the adulterants and toxins in food processing as well. Thus, while SERS has experienced enormous progress, a broader range of use is limited by high running costs, as well as their inefficacy in point-of-care analysis. In view of the tremendous progress in the implementation of the SERS technique as an analytical tool, a great challenge is a rapid evolution in commercial products such as compact setups, tailored sensing platforms or efficient imaging methods that are able to compete or complete goods in current use in a wide range of technologies. In the next steps, SERS must be developed such that its processes of application will be simplified and its use of routine will be enabled by non-specialists. In light of the commercialization, adequacy for automated mass production and stability during storage must be examined.

#### **11. Conclusions**

Between the plethora of approaches focused on improving the performances and making SERS analysis a routine technique to enhance the quality of life of people, in recent years the development of sensing platforms based on structures composed of plasmonic metal nanoparticles hybridized with bidimensional nanoparticles have attracted great interest due to their outstanding properties. In this review, the different approaches used to enhance the outcomes of hybrid compounds have been discussed. Initially, the simplest structures based on spherical metal nanoparticles hybridized with graphene and its derivates (the most common bidimensional nanomaterial) have been considered, and then it has been considered how the structures have been evolved by changing the particle shape or engineering the bidimensional material to obtain the highest values in the enhancement factor (EF). Due to a greater number of analyte molecules adsorbed and the number of hotspots formed, three-dimensional (3D) structures of 2D nanomaterials and plasmonic nanoparticles showed the best results, so that, most likely, the future steps will be focused on the development of such compounds. The future perspectives of the technique have

been discussed from a broader point of view considering the various directions in which the technique is progressing and not only from a point of view of the hybrid compounds. In this scenario, multifunctional substrates combining several of the characteristics considered will be developed in order to render SERS a daily technique for better living conditions.

**Author Contributions:** Conceptualization, C.S., A.F. and E.C.B.A.A.; methodology, C.S.; writing original draft preparation, C.S., A.F. and E.C.B.A.A.; writing—review and editing, C.S., A.F. and E.C.B.A.A.; supervision, A.F., E.C.B.A.A. and M.V.; funding acquisition, A.F., E.C.B.A.A. and M.V. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by EU funds through the FEDER European Regional Development Fund project LISBOA-01-0145-FEDER-031311, by Portuguese national funds provided by FCT—Fundação para a Ciência e Tecnologia, through grant SFRH/BD/09347/2021 and PTDC/QUI-QIN/29778/2017 project and Instituto Politécnico de Lisboa (IPL/2021/WASTE4CAT\_ISEL and IPL/2021/MuMiAS-2D\_ISEL).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **A Novel Enzyme-Based SPR Strategy for Detection of the Antimicrobial Agent Chlorophene**

**Gabriela Elizabeth Quintanilla-Villanueva 1,2, Donato Luna-Moreno 3,\*, Edgar Allan Blanco-Gámez 1,2, José Manuel Rodríguez-Delgado 4, Juan Francisco Villarreal-Chiu 1,2 and Melissa Marlene Rodríguez-Delgado 1,2,\***


**Abstract:** Chlorophene is an important antimicrobial agent present in disinfectant products which has been related to health and environmental effects, and its detection has been limited to chromatographic techniques. Thus, there is a lack of research that attempts to develop new analytical tools, such as biosensors, that address the detection of this emerging pollutant. Therefore, a new biosensor for the direct detection of chlorophene in real water is presented, based on surface plasmon resonance (SPR) and using a laccase enzyme as a recognition element. The biosensor chip was obtained by covalent immobilization of the laccase on a gold-coated surface through carbodiimide esters. The analytical parameters accomplished resulted in a limit of detection and quantification of 0.33 mg/L and 1.10 mg/L, respectively, fulfilling the concentrations that have already been detected in environmental samples. During the natural river's measurements, no significant matrix effects were observed, obtaining a recovery percentage of 109.21% ± 7.08, which suggested that the method was suitable for the fast and straightforward analysis of this contaminant. Finally, the SPR measurements were validated with an HPLC method, which demonstrated no significant difference in terms of precision and accuracy, leading to the conclusion that the biosensor reflects its potential as an alternative analytical tool for the monitoring of chlorophene in aquatic environments.

**Keywords:** SPR biosensor; enzyme; laccase; chlorophene; emerging pollutant; water sample

#### **1. Introduction**

Emerging pollutants are persistent chemicals in the environment, classified as pharmaceutical compounds or their metabolites (human and veterinary drugs). These include personal care products (e.g., disinfectants, fragrances, insect repellents, cosmetics and sunscreens) and endocrine disrupting compounds (e.g., bisphenol A, triclosan and pesticides) [1]. In particular, halogenated phenolic compounds comprise the vast majority of the active ingredients employed in the manufacture of personal care products [2]. In this sense, chlorophene (4-chloro-2-(phenylmethyl)phenol) is an antimicrobial agent widely applied in disinfectants for cleaning activities and for farming, industrial and household environments [3,4], as well as preservatives in cosmetics and wood [5]. According to the Environmental Protection Agency (EPA), chlorophene has been included in the list of priority toxic pollutants [6,7]. It has been related to mutagenic effects in mammals [5],

**Citation:** Quintanilla-Villanueva, G.E.; Luna-Moreno, D.; Blanco-Gámez, E.A.; Rodríguez-Delgado, J.M.; Villarreal-Chiu, J.F.; Rodríguez-Delgado, M.M. A Novel Enzyme-Based SPR Strategy for Detection of the Antimicrobial Agent Chlorophene. *Biosensors* **2021**, *11*, 43. https://doi.org/10.3390/bios11020043

Received: 14 January 2021 Accepted: 6 February 2021 Published: 9 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

fertility alterations and kidney damage through prolonged exposure [4]. The occurrence of chlorophene (CP) has been reported in water [7] and soil [8]. For example, concentrations up to 0.13 mg/L of CP were detected in a backwater stream in Kerala (India) [7]. Meanwhile, 50 mg/L was quantified in activated sludge sewage, and 10 μg/L was quantified in treatment plant effluent [9].

These micropollutants enter the environment through anthropogenic pathways [7]. A trace amount results in ecological risks, such as biomagnification along the food chain due to accumulation in organisms by hydrophobic properties [3]. Such is the case of CP's occurrence in male bream bile from the Dommel river (7 μg/mL) [10]. The presence of this emerging pollutant has been commonly detected by high performance liquid chromatography mass spectrometry (HPLC-MS) [4] and gas chromatography mass spectrometry (GC-MS) [3], powerful analytical methods for detecting and quantifying trace amounts of compounds. For instance, Rayaroth et al. (2015) [7] identified the presence of chlorophene in a backwater stream using the liquid chromatography quadrupole time of flight MS (LC-QTOF-MS), with a C18 column set at 35 ◦C and a gradient elution of acetonitrile:formic acid in water (0.1%). Meanwhile, Chen et al. (2018) [11] established the quantification of CP using an HPLC instrument with a UV absorbance detector, an SB-C18 column and a mixture of formic acid as a mobile phase and methanol. On the other hand, the use of the GC-MS technique has also been reported, employing a 5% phenyl methyl siloxane capillary column, splitless injection at 250 ◦C, an oven temperature from 70 to 280 ◦C (10 ◦C/min) and helium as the carrier gas [3]. It is worth highlighting that in prior chromatography−mass spectrometry analysis, sample pretreatment needed to be performed, commonly a purified process by solid-phase extraction (SPE) using cartridges [11] or solvent extraction followed by a concentration step [3]. Consequently, the time-consuming sample preparation and lab environments' restrictions remain significant drawbacks that limit chromatographic techniques. Thus, there is an increasing interest in developing new analytical tools that provide fast, sensitive, and in situ measurements, such as biosensor systems. In this sense, the surface plasmon resonance (SPR) technique has had significant relevance in the environmental field. For example, it was employed in the detection of endocrine disruptors (estrogen [12] and bisphenol A [13]), organophosphate pesticides like chlorpyrifos [14] and industrial pollutants such as polychlorinated biphenyls [15]. Nevertheless, no attempt to use biosensors to detect CP has been explored.

On the other hand, diverse treatment processes have been applied to remove CP, such as MnO2 oxidation, persulfate treatment and ozonation [11]. However, the use of laccase enzymes in removing chlorophene and dichlorophene is worth noticing [16]. Laccases are phenoloxidases produced in extracellular form by a diverse variety of organisms, from higher plants to fungi [17,18] and bacteria [19]. These enzymes catalyze the oxidation of organic compounds by the concomitant reduction of oxygen [20]. In particular, the removal of CP by laccase catalysis was demonstrated by Shi et al. (2016), suggesting a direct polymerization as the principal mechanism for elimination [16].

Therefore, this work establishes the immobilization of laccase enzymes for their use as a receptor in the detection of chlorophene using an SPR technique. The use of enzymes as recognition elements is very uncommon in SPR techniques [21,22], considering that most applications rely on antigen–antibody interactions [23], aptamer recognition [24,25] and nucleic acid hybridization [26]. In particular, studies of laccase as a bioreceptor in SPR are very scarce [27]. The proposed enzyme-based SPR biosensor's analytical parameters, such as the limit of detection, sensitivity and working range, were studied. Finally, fortified real water samples were analyzed by the SPR biosensor, and the results obtained were compared in terms of accuracy and precision against a well-known HPLC method.

#### **2. Materials and Methods**

#### *2.1. Reagents*

All the chemical compounds employed in the enzyme immobilization and the biosensing process, such as 16-mercaptohexadecanoic acid (MHDA), 11-mercaptoundecanol

(MUD), ethanolamine hydrochloride, N-hydroxysuccinimide (NHS) and 1-ethyl-3-(3 dimethylamino-propyl) carbodiimide hydrochloride (EDC)) were purchased from Sigma-Aldrich (St. Louis, MO, USA). The laccase enzymes (*Rhus vernicifera*) and salts employed in buffers were purchased from Sigma-Aldrich (St. Louis, MO, USA).

HPLC analysis was performed by employing a Zorbax ODS C18, 25 cm × 4.6 mm, 5 μm column, which was purchased from SUPELCO Analytical (St. Louis, MO, USA) with an HPLC system model YL9100 (Younglin Instrument Co., Ltd., Gyeonggi-do, Korea). The HPLC-grade acetonitrile and water were purchased from Merck (Darmstadt, Germany).

The standards of chlorophene (analytical grade, Sigma-Aldrich, Mexico City, Mexico) were prepared as stock solutions (10 mg/mL) in ethanol:water (90:10, %*v*/*v*) 96 and completed at 10 mL with ultrapure water. From these stock solutions, working solutions were prepared by serial dilution in a water:phosphate-buffered saline solution with a pH of 7.3 (90:10, %*v*/*v*) in a concentration range from 0 to 10 mg/L.

Real water samples were obtained from a river and filtered with Whatman grade 40 filter as the only pretreatment. Then, one level of fortification was prepared by spiking the river samples with chlorophene at a concentration of 3 mg/L, followed by its analysis by HPLC and the SPR technique.

#### *2.2. Sample Collection and Characterization*

Water samples were collected from a river located in León city, Guanajuato-México (21◦09 54.0 N, 101◦43 30.6 W). Sample collection was performed following the Mexican standards established in NOM-230-SSA1-2002 [28]. Briefly, at the sampling site, water samples were collected in polyethylene bottles (pre-rinsed with distillate water). The temperature and pH were measured in the area with a multiparameter probe (WTW Multi 350i). Analyses of the sulfate, total alkalinity and acidity were performed according to the Mexican standards NMX-AA-036-SCFI-2001 [29], as well as the hardness following the methods of NMX-AA-072-SCFI-2001 [30] and the chlorides according to NMX-AA-073- SCFI-2001 [31]. The total organic carbon (COT) and heavy metals were measured by a Shimadzu analyzer by catalytic oxidation of combustion (TOC-L, Shimadzu, Kyoto, Japan) and atomic absorption (Thermo Jarrell Ash Scan1, Franklin, MA, USA), respectively, under NMX-AA-051-SCFI-2016 [32].

#### *2.3. Cr/Au Thin Film Deposition*

Homogeneous thin films of Cr/Au were deposited on thin glass substrates by electron gun evaporation, following the method described by Luna-Moreno [33]. Briefly, the chromium layer was evaporated up to a 3 nm thickness. Then, a gold film of 50 nm was deposited by thermal evaporation at a rate of 5 Å/s and 8 × <sup>10</sup>−<sup>6</sup> mbar. The thickness of the thin films was evaluated by employing a quartz crystal microbalance thickness monitor (Leybold Inficon XTC/2 Depositions controllers).

#### *2.4. SPR Instrumentation*

The SPR setup was a homemade platform described previously by Sánchez-Alvarez et al. (2018) [34], based on a Kretschmann configuration and comprising two stacked rotation plates, configured for synchronized movement according to a θ-2θ system by a stepper motor. The measuring cell in the SPR system consisted of a sandwich configuration integrated by a Teflon cell, a gold thin film chip and a hemicylindrical BK7 glass prism. The substrate's glass surface was optically coupled to the prism using an oil matching index (n = 1.51). Meanwhile, the chip's gold-coated surface was facing against the flow of the Teflon cell, which had an inlet and outlet that allowed the solutions to come in contact with the gold through its inner channel (Figure 1). Our design allowed for adjusting to customized measurement chips (different size and thickness), depending on the desired application compared to commercial cells.

**Figure 1.** (**a**) Surface plasmon resonance (SPR) setup and (**b**) scheme of the measuring cell.

The chemical solutions continuously flowed through the measuring cell via a syringe pump (Legato 100) at a rate of 30 μL min−1. Furthermore, a photodetector (Hamamatsu, model S1226-8Bk) was used to capture the reflected light of a He-Ne laser (Uniphase mod. 1101P) that passed through the prism.

#### *2.5. Enzymatic Activity*

The laccase enzymatic activity was measured through the spectrophotometric UV-Vis assay established by Zhang et al. (2018) [35]. For the assay, 200 μL of the enzyme was added to a reaction mixture (2 mL) containing 10 mM of 2,2 -azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) (ABTS) in a 0.1 M sodium acetate buffer with a pH of 4.5. The reaction occurred at room temperature, and absorbance changes were recorded at 420 nm in a UV-Vis spectrophotometer (Cary 50, Varian Inc., Palo Alto, CA, USA).

Enzyme activity was expressed as a function of the amount of enzyme necessary to produce 1 μM of product per minute (U) and was calculated by the following equation:

$$Activity = \frac{\left[\left(\frac{\Delta A bs}{\min}\right) \times V\_t\right]}{\varepsilon \times 10^4 \times 1 \times V m}$$

where Δ*Abs* is the change in absorbance, *Vt* is the total volume of the cell, ε is the molar extinction coefficient of ABTS (36,000 M−1cm−1) at 420 nm and *Vm* is the volume of the laccase sample [35].

#### *2.6. Enzyme Bioreceptor: Chip Functionalization and Laccase Immobilization*

Before the immobilization process, a functionalization treatment on the gold substrate (50 nm chips) was performed. Briefly, the gold chips were cleaned by consecutive immersion in acetone and ethanol (30 s in each solvent) and then dried with air. Then, the clean chips were immersed for 12 h at room temperature in a solution of alkanethiols MHDA:MUD (250 μM in ethanol) [36]. The sulfur group from the alkanethiols bound to the gold. Meanwhile, the free carboxylic group on the other end of the chain served as the binding site for the further immobilization of the enzyme.

Once the chip's surface was functionalized with the alkanethiols, the carboxylic groups were activated using the EDC/NHS crosslinkers [37]. A solution of EDC/NHS (EDC 0.2 M/NHS 0.05 M) in an MES buffer (100 mM, 500 mM NaCl, pH 5.0) was flowed on the gold surface, allowing the formation of carbodiimide esters. Then, a solution of 200 U mg-1 of laccase was injected. By creating an amide bond between the amino acids of the enzyme and the activated carboxylic terminal group, the laccase's attachment occurred, ending the immobilization process. Finally, the remaining active esters were deactivated with a solution of ethanolamine (1 M, pH 8.5), preventing unspecific bindings. Once the immobilization process concluded, a washing step was performed, flowing over the sensor surface a phosphate buffer solution (PBS) to remove non-bonded molecules.

#### *2.7. SPR Measurements: Chlorophene Detection*

Once the chip was mounted on the SPR setup, the working angle was established at the slope's midpoint, formed in the SPR curve (approaching the critical angle). At this point, greater sensitivity to changes in light intensity, caused by the interaction of the receptor with the analyte, was achieved.

The analysis of chlorophene (CP) was performed by a direct enzyme–substrate assay, where the immobilized laccase enzymes catalyzed the oxidation reaction of CP in the sample. The obtained signals (enzyme–substrate binding) were directly proportional to the concentration of the analyte in the samples, since a shift in the conformation of the enzyme occurred as a result of CP binding in the active site of the laccase, observed as a change in the refractive index measured by the photodetector [38]. The PBS buffer was set as a running solution during the measurement process, and the samples containing CP (0–10 mg/mL) were flowed at 30 μL/min over the sensor surface. The sensor surface was then washed with a PBS buffer injection to remove weakly bound CP from the biofunctionalized chip. Finally, a regeneration solution (NaOH 10 mM) was injected for 20 s to release the bioreceptor and prepare it for a new measurement cycle. All measurements were performed in triplicate. The obtained average SPR signals were plotted as a CP concentration function in the sample. The calibration curve generated was employed to establish the analytical parameters of the biosensor. The sensitivity of the method corresponded to the slope of the curve. The detection limit was evaluated as three times the standard deviation of the baseline, while the limit of quantitation was 10 times the standard deviation. The recovery and reproducibility of the analytical procedure were established using spiked real samples. The determination of the recovery and precision of the SPR was also performed on natural samples, evaluating possible matrix effects.

#### *2.8. HPLC Measurements*

The HPLC analysis (YL9300, Thermofisher-USA) was performed according to the method previously described [4]. Briefly, a solution of acetonitrile:water (85:15) was employed as a mobile phase at a flow rate of 1 mL/min, using a Zorbax ODS C18, 25 cm × 4.6 mm (5 μm particle size) column and a UV detector at 290 nm [7]. The calibration standards of the CP stock were prepared in ethanol from a stock solution of 100 mg/L (working range from 0–10 mg/L). The recovery and reproducibility of the analytical procedure were established using spiked real samples.

#### **3. Results and Discussion**

#### *3.1. Chip Functionalization and Laccase Immobilization*

The immobilization process of the enzymes was initiated with the functionalization of a gold-coated chip. After 12 h of incubation, the binding alkanethiols were activated through the EDC/NHS cross-linkers, forming an amide bond that attached to the laccase enzymes. Figure 2a shows the critical angle displacement of the SPR curve of the alkanethiol-coated chip (blue line) in comparison with the SPR curve after laccase immobilization occurred (red line), showing a shift of 3.2 degrees (from 69.2◦ to 72.4◦) in the resonance angle.

**Figure 2.** (**a**) Reflectance spectra obtained by the angular sweep of a sensor chip with alkanethiols (blue line) and after immobilization of the laccase (red line). (**b**) SPR sensorgram of laccase immobilization.

The shift was attributed to the increase of the mass density due to the bound enzymes on the surface. In this sense, several studies have quantified the immobilization yield on a surface through a conversion factor of 1 ng/mm2 of biomolecules or protein. The conversion factor was related to a change of 0.1◦ in the SPR angle (1000 refractive units) [39–41]. Therefore, the angle displacement obtained in this study would represent a density of 32 ng/mm<sup>2</sup> of enzyme onto the gold-coated chip.

The immobilization process was also monitored in real time at a fixed angle of 66.8◦ (highest sensitivity from the SPR curve slope), obtained by the angular sweep of the immobilized SPR chip (Figure 2a). The sensorgram obtained from the immobilization process in-flow is observed in Figure 2b. The increase of the SPR baseline signal was notable after the subsequent addition of the EDC/NHS crosslinkers, the laccase enzyme and the ethanolamine. However, after the washing step, the signal decreased, suggesting the removal of those molecules weakly bonded on the surface. Once the washing process concluded, it was noticeable that the SPR signal was higher than the initial baseline (prior immobilization), inferring the successful linkage of the molecules that remained on the surface. These results agree with the ones obtained by the angular swept measurements (Figure 2a).

#### *3.2. SPR Measurements: Chlorophene Detection*

A direct enzyme–substrate assay was performed to detect chlorophene ranging from 0–10 mg/mL, using an SPR gold-coated chip immobilized with laccase enzymes. In this biosensor, the immobilized laccase catalyzed the oxidation reaction of CP in the sample, observed as a change in the refractive index (SPR signal) due to a shift in the conformation of the enzyme as a result of the concentration of analyte binding to its active site [38]. The active site of the laccase enzymes comprised four copper atoms (type I, type II and two type III copper atoms) [42]. The enzyme's catalytic mechanism involved the substrate oxidation in the type I copper site, followed by an internal electron transfer from the reduced type I atoms to the type II and type III trinuclear cluster, where the reduction of dioxygen to water occurred [42].

According to Enguita et al. (2004) [43], apolar groups in chemical structures are attached to a hydrophobic binding site in laccase, which is located in proximity to the type I Cu site of the enzyme through the His497 residue (one of the type I copper ligands). In this sense, the phenyl group of the chlorophene molecule might present a close approach toward the aromatic ring of the His497 residue in laccase, favoring the electron transfer from CP (oxidation process) to the type I copper and subsequent internal transfer to the trinuclear cluster [43]. Jabbari et al. (2017) [27] reported this electron transfer mechanism during the study to detect the bromocriptine drug by the SPR technique using laccase from Bacillus sp. HR03 [27]. The SPR signal value at the plateau (saturation in the binding event) obtained from the sensorgrams was plotted as a function of the CP concentrations to generate a calibration curve (Figure 3).

**Figure 3.** SPR sensorgrams for chlorophene (CP) detection at different concentrations and calibration curves in PBS (*n* = 3).

The analytical parameters obtained in this study are summarized in Table 1. The results obtained meet the concentration detected in a backwater stream in Kerala (India) [7] and in activated sludge sewage [9] by HPLC-MS and GC-MS. Additionally, the immobilized enzyme withstood 35 regeneration cycles by using 0.1 M NaOH before any significant loss of recognition capacity was observed. Currently, there is a lack of biosensors that address the detection of CP, since the analytical tools have been mainly limited to chromatographic techniques. Thus, this research represents the first approach to the fabrication of a robust SPR platform for the routine monitoring of chlorophene in water.

**Table 1.** Analytical parameters of SPR-based biosensors for chlorophene detection.


#### *3.3. Evaluation of SPR Performance with River Water: Study of Matrix Effects*

Certain components in the real sample matrix could lead to false positives or unspecific bindings that interfere with bioreceptor recognition [44]. Thus, this is a critical issue to determine the performance of a method during the analysis of real samples. Possible matrix effects due to river water composition need to be evaluated. Thus, a natural water sample was injected as a control. Apparently, no significant SPR signal was shown in the preliminary sensitivity assay result with the river water, suggesting that no significant enzyme–substrate reaction occurred. However, it is essential to perform a selectivity assay in the presence of related compounds (interferences) to confirm the method's feasibility for the selective detection of chlorophene. Then, spiked river samples with 3 mg/mL of CP (triplicate) were measured, obtaining nearly identical SPR responses among them under these conditions (see Figure 4).

**Figure 4.** Evaluation of nonspecific signals due to matrix effects from the river water.

The characterization of the river samples can be observed in Table 2, where it is worth noticing that the concentration of organic matter did not represent a considerable effect on the recovery percentage analysis, showing a recovery percentage of 109.21% under the method conditions (see Table 3). No significant differences were found between the theoretical concentration and the one obtained experimentally (*p* = 0.05, *n* = 3).

**Table 2.** Components and significant ion concentrations (mg/L) in the river water samples.



On the other hand, the laccase enzymes were stable under the concentration of dissolved chlorides and the alkaline pH, since those are conditions that tend to affect the enzymatic activity of these enzymes [45,46].

#### *3.4. Comparison of SPR Protocol with the HPLC Method*

Samples spiked in a range from 0 to 30 mg/L were analyzed by HPLC. The linear regression analysis provided a correlation of 0.9995, a limit of detection (LOD) of 0.07 mg/L and a limit of quantification (LOQ) of 0.22 mg/L, with an operating range of 0–30 mg/L.

Then, the river samples spiked with chlorophene at concentrations of 3 mg/L were analyzed. A Student's *t*-test with 95% confidence was evaluated, and no significant discrepancies were found when comparing the SPR and HPLC methods, indicating excellent agreement between the techniques (Table 3).

#### **4. Conclusions**

This work established the use of a homemade SPR biosensor based on using laccase enzymes as a bioreceptor for the real-time detection of the hazardous antimicrobial chlorophene in real waters. To the best of our knowledge, this study is the first attempt to develop a biosensor to detect chlorophene. The analytical parameters accomplished by the SPR biosensor fulfilled the concentrations that have already been detected in natural water samples. The biosensor method resulted in a limit of detection and quantification of 0.33 mg/L and 1.10 mg/L, respectively. Although no apparent matrix effects were detected in the analytical response of the SPR measurements of the river samples, it is essential to perform a selectivity assay to confirm the method's feasibility in the selective detection of chlorophene. Furthermore, the method's reliability could be improved by analyzing certified samples to complement fortified natural water results. Finally, the comparison of SPR measurements with an HPLC conventional method demonstrated no significant difference in precision and accuracy. These results show a considerable advantage, due to its lack of a pretreatment process, which is required by traditional techniques, suggesting a suitable and straightforward analysis of this contaminant in natural water.

**Author Contributions:** Conceptualization, M.M.R.-D. and D.L.-M.; methodology, G.E.Q.-V., D.L.-M., J.M.R.-D. and E.A.B.-G.; validation, G.E.Q.-V. and M.M.R.-D.; formal analysis, G.E.Q.-V., J.F.V.-C. and M.M.R.-D.; investigation, G.E.Q.-V., D.L.-M., J.M.R.-D. and E.A.B.-G.; writing—original draft preparation, G.E.Q.-V. and M.M.R.-D.; writing—review and editing, D.L.-M., J.F.V.-C. and M.M.R.-D. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by UANL's Programa de Apoyo a la Investigación Científica y Tecnológica (PAICYT), grant number CE866-19.

**Acknowledgments:** The authors would like to thank the Consejo Nacional de Ciencia y Tecnología (Conacyt) for Gabriela Quintanilla-Villanueva scholarship #740156.

**Conflicts of Interest:** The authors declare no conflict of interest, personal, financial, or otherwise, with the manuscript's material.

#### **References**


## *Article* **Optimization of High-Density Fe-Au Nano-Arrays for Surface-Enhanced Raman Spectroscopy of Biological Samples**

**Giovanni Marinaro 1,\*, Maria Laura Coluccio <sup>2</sup> and Francesco Gentile <sup>2</sup>**


**Abstract:** The method of realizing nanostructures using porous alumina templates has attracted interest due to the precise geometry and cheap cost of nanofabrication. In this work, nanoporous alumina membranes were utilized to realize a forest of nanowires, providing a bottom-up nanofabrication method suitable for surface-enhanced Raman spectroscopy (SERS). Gold and iron were electroplated through the straight channels of the membrane. The resulting nanowires are, indeed, made of an active element for plasmonic resonance and SERS as the hexagonal distribution of the nanowires and the extreme high density of the nanowires allows to excite the plasmon and detect the Raman signal. The method to reduce the distance between pores and, consequently, the distance of the nanowires after electrodeposition is optimized here. Indeed, it has been predicted that the light intensity enhancement factor is up to 10<sup>12</sup> when the gap is small than 10 nm. Measurements of Raman signal of thiol groups drying on the gold nanowires show that the performance of the device is improved. As the thiol group can be linked to proteins, the device has the potential of a biosensor for the detection of a few biomolecules. To assess the performance of the device and demonstrate its ability to analyze biological solutions, we used it as SERS substrates to examine solutions of IgG in low abundance ranges. The results of the test indicate that the sensor can convincingly detect biomolecules in physiologically relevant ranges.

**Keywords:** plasmonic nanowires; molecular sensing; surface-enhanced Raman spectroscopy; porous alumina

#### **1. Introduction**

Over the past years, surface-enhanced Raman spectroscopy (SERS) has become a powerful tool allowing non-destructive, highly sensitive studies of molecules, chemicals or biological samples [1–3]. Further improvements of this technique could spur considerable progress in areas such as single-molecule sensing, early cancer detection and in situ analyte detection in microfluidics.

SERS takes advantage of highly packed sensitive nanostructures elements positioned in areas of few squared micrometers. Similar devices, fabricated thanks to recent advances in nanotechnology, enable the ultrasensitive, label-free detection of analytes. This detection can be enhanced, in turn, through integration with microfluidics that allows tight control over the volumes, flows and velocities of the biological liquids under examination. Microfluidics-assisted SERS has found applications in several fields, including biomedical engineering, proteomics, life science, and cellomics [4,5]. Notably, the combination of nanoscale devices and the manipulation of nano-liquids has enabled, among other things, the separation and identification of complex mixtures in very low abundance ranges [6,7].

While the technique achieves high sensitivity and ultra-low detection limit, nevertheless, the characteristics of sensitivity, precision, and selectivity have to be improved—this

**Citation:** Marinaro, G.; Coluccio, M.L.; Gentile, F. Optimization of High-Density Fe-Au Nano-Arrays for Surface-Enhanced Raman Spectroscopy of Biological Samples. *Biosensors* **2021**, *11*, 181. https:// doi.org/10.3390/bios11060181

Received: 3 May 2021 Accepted: 2 June 2021 Published: 5 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

would make SERS devices suitable for the detection of biomarkers in complex solutions and biological fluids.

SERS devices amplify the Raman signal. Raman spectroscopy is a method that provides structural, chemical and conformational information about biomolecules such as proteins and DNA. In Raman spectroscopy, visible light and infrared radiation interact with a molecule, producing, as a result, a spectrum that describes the energy associated to the vibrational states of that molecule. Raman spectroscopy has a number of features, such as requiring minimal sample preparation and being label-free, non-destructive and non-invasive. The principal limitation of this spectroscopy is the extremely low Raman scattering cross-section (typically about 10−<sup>24</sup> to 10−<sup>27</sup> photons per events per molecule [8]) which is insufficient to characterize many biological systems, especially in low concentration ranges. Therefore, it is of primary importance to provide means to increase the signal intensity [9]. Among others, an efficient method for enhancing the Raman signal is to use plasmonic surfaces. Similar surfaces, typically made of a metal nanomaterial, manipulate and enhance the local electromagnetic (EM) field of several orders of magnitude. The EM field enhancement is the result of the collective, resonant oscillation of the electrons on the nano-metal's surface [10]. The EM field increment has, as a consequence, the *surface enhancement* of the *Raman spectroscopy* signal, from which the acronym SERS is derived. The SERS effect provides access to otherwise unattainable information of biological systems, drugs, diluted analytes and biomolecules that are not detectable with conventional techniques of analysis [3].

The theoretical upper bound for SERS enhancement is 1012 [11]. Remarkably, since the SERS efficiency shows a very high sensitivity on the *geometry* of the substrate, in recent years, a variety of techniques have been developed to fabricate nanoscale structures with maximum resolution, maximum precision, and minimal tolerances [12–17]. Moreover, the design and fabrication methods for efficient SERS substrates should allow reliability and reproducibility over sufficient large areas to provide, at the same time, enhancement of the EM field, stability over time, and the ability to resist mechanical and environmental vibrations and noise. This leaves a lot of room for improvements in the design and fabrication of the final devices. Due to the topological requirements, the fabrication of SERS substrates involves nanotechnology techniques. Some representative examples are given in the following: Kattumenu et al. exploited nanorod-decorated nanowires to observe the Raman enhancement of thiolic molecules [18]. A super-hydrophobic surface made of micropillars was used to concentrate and detect a few molecules dissolved in a droplet [2]. Optical properties of a hexagonal array of metal nanopillars for plasmonic applications were investigated by Zhang et al. [19]. Menvod et al. functionalized graphene nanosheets with cationic poly (diallyldimethylammonium) (PDDA) and citrate-capped gold nanoparticles (AuNPs) for SERS bio-detection application [20]. Zhang et al. fabricated large-scale Au nanodisk arrays on Si substrate via x-ray interference for the detection of Rhodamine 6G as low as 10−<sup>5</sup> mM [21]. Gentile et Al. dispersed silver nanoparticles into the pores of a superhydrophobic surfaces to guarantee superior SERS capabilities [1].

The present approach regards the use of plasmonic devices with large surface area and high-density hotspots whose increased detection efficacy is due to the strong plasmonic coupling of the nanowires. The sensitive device area is in the range of centimeters. While the single plasmonic elements can be made from 30 to 300 nm, their coupling distance can be adjusted between 3 nm and 20 nm. All these properties in the same device enable sensitive analysis of biological solutions, statistical significance, reliability, and repeatability. Moreover, the proposed technology is cheap and can be used in future translational biological medicine studies.

In this paper, the method has been used to detect Benzenedithiol molecules which were chemisorbed on the gold nanowire surface and, in another case, immunoglobuline IgG. The performance and the results of the biosensors will be presented in the next sections.

#### **2. Materials and Methods**

#### *2.1. Nanowires Fabrication*

In order to efficiently and reproducibly fabricate NW-based substrates, the electrochemical deposition of NWs into nanoporous alumina templates was utilized (Figure 1). For a detailed description of the porous alumina fabrication, the protocol in a previous paper [8] was considered. The process involves two steps of anodization of a one-inch aluminum disk which create a thin layer of aluminum oxide with a highly ordered nanopore distribution. The first anodization step yields an alumina layer with poorly organized pores at the top, but high regularity at the bottom.

**Figure 1.** Sketch of Fe-Au electrodeposition through the pores of porous alumina: iron (**A**), addition of gold (**B**), and 3D illustration (**C**).

The pore distribution and size homogeneity are shown in Figure 2. The pores of alumina templates have a diameter of about 60 nm and are distributed in a hexagonal lattice with a constant distance of 40 nm. The electrolyte for the iron segment growth was realized with 6 g of iron sulfate heptahydrate (FeSO4·7H2O, Sigma-Aldrich), 1 g of boric acid (H3BO3, Sigma-Aldrich) and 1 g of ascorbic acid (C6H8O6, Sigma-Aldrich). Boric acid was added to improve the purity of iron while ascorbic acid was added to adjust the pH to 3.

**Figure 2.** SEM micrographs. Porous alumina membranes with a distance between pores of 60 nm (**A**), 85 nm (**B**) and 95 nm (**C**). The white scale bar is 200 nm.

The electrolyte for the gold segment growth was prepared with 0.1 potassium dicyanoaurate (KAu(CN)2, Sigma-Aldrich, St. Louis, MO, USA) and 4 g of boric acid (H3BO3, Sigma-Aldrich). Boric acid was added in order to adjust the pH and work in safe acidic conditions. In order to realize three different sizes of the pores, and consequently, three different aspect ratios of nanowires, the membranes were dipped in 5% H3PO4 in water (*w*/*w*) for different lengths of time. Porous alumina was then etched away in an acidic

chrome solution (CrO3/H3PO4 in water) at 40 ◦C overnight (Figure S1) in order to obtain free-standing nanowires.

Removal of the alumina reveals highly ordered features on the surface of the Al substrate, which facilitates the growth of ordered pores upon the second anodization step. The pores obtained with the second anodization are straight channels arranged in a hexagonal pattern. The diameter of pores and the distance between them can be tuned by controlling the parameters of anodization, while their length depends on the anodization duration. One of the main challenges in using porous membranes for SERS is to reduce the gap between the active elements of the nanostructure because, based on numerical predictions, the light intensity shows the best enhancement factor (up to 1012) when the gap is smaller than 10 nm.

A way to address the problem to a solution is to gently dip porous alumina in a diluted phosphoric acid solution in order to widen the pores and, consequently, reduce the inter-distance to a few 10s of nanometers. Simulations have already predicted that nanostructures obtained by this method are suitable for SERS devices [8].

After removing the backlayer substrate of alumina with an etching solution, an electric contact was made using a PECVD by depositing a thin layer of 20 nm of ITO (Indium Tin Oxide) on the surface of the porous alumina membrane. The electrolyte for the iron segment growth was realized with 6 g of iron sulfate heptahydrate (FeSO4·7H2O, Sigma-Aldrich), 1 g of boric acid (H3BO3, Sigma-Aldrich) and 1 g of ascorbic acid (C6H8O6, Sigma-Aldrich). Boric acid was added to improve the purity of the iron while ascorbic acid was added to adjust the pH to 3. The electrolyte for the gold segment growth was prepared with 0.1 potassium dicyanoaurate (KAu(CN)2, Sigma-Aldrich) and 4 g of boric acid (H3BO3, Sigma-Aldrich). Boric acid was added in order to adjust the pH and work in safe acidic conditions.

#### *2.2. SEM Analysis*

SEM analysis of the plasmonic nanowires was conducted with a Zeiss GeminiSEM 500 at Dresden Center for Nanoanalysis (DCN), TU Dresden. The samples were already conductive, so there was no need to sputter gold. The samples were fixed on a stub with a long pin and then mounted on a carousel 9 × 9 mm sample holder. In order to fix the samples, a small amount of silver paint was applied between the edge of the aluminum disc and the stub. A further copper lever was screwed in order to secure the sample on the stub. Several images of metal nanowires were acquired in high vacuum mode at 5 kV, a magnification of 300,000 and a working distance of about 3 mm with an InLens Detector (ZEISS) for secondary electrons. In order to reduce the noise, a frame integration (N = 14) was performed. With this setup, every frame was scanned and averaged 14 times.

#### *2.3. Light Transmission Measurements*

A Nikon Eclipse Ni with an integrated Thor spectrometer (a ray diagram is shown in Figure 3) was used to measure the wavelength band of white light in transmission.

To do so, a variation in the fabrication step was realized: an ITO nanofilm was sputtered to the back of the alumina membrane as a conductive layer for the electrodeposition instead of sputtering gold. ITO is a material that is conductive and transparent at the same time. This guaranteed the electrodeposition of nanowires through the pore channels and the ability to perform optical light measurements.

**Figure 3.** Ray diagram of light transmission setup. The onset shows the plasmonic device illuminated from the backside.

#### *2.4. Raman Spectroscopy*

A Renishaw InVia Raman spectrometer with a 1200 line/mm grating for the SERS measurements was used for the measurement of the Raman signals. The samples were excited by 830 nm laser line in backscattering configuration through 100× objective (NA = 0.9) using the respective edge filters to stop the laser lines. The scattering was collected in the range of 200 to 2000 cm−1. The spectra were analyzed with WiRE 3. Benzenedithiol 1,4 molecules were deposited on the substrates by immersion in solution and subsequent rinsing in MilliQ water. A droplet of about 10−<sup>1</sup> mM of benzedithiol 1,4 in ethanol was gently deposited on the surface of the device with a pipette and allowed to dry for about one hour in order to stabilize the disulphuric links of thiols with the gold nanowires. SERS spectrum of Benzenedithiol 1,4 was excited with a laser power of 2 mW. Further investigations were carried out with an HR 800 Raman spectrometer with a micro-Raman spectral acquisition images. The samples were excited with a 795 nm laser line through a 100× objective (NA = 0.9). The spectra were exported as text files and analyzed with an in-house script written and run in Matlab (R2017b MathWorks).

#### *2.5. SERS Analysis of IgG*

A drop of immunoglobulin (IgG) at 0.44 mg/mL was positioned on the substrate and left to dry. The Raman spectra of IgG adsorbed on the substrate were collected by an InVia Raman spectrometer with a 1200 line/mm grating, equipped with a 100× optical microscope objective. Samples were excited by an 830 nm laser line, setting the laser power at 1.6 mW. The Raman signals were recorded on maps of different sizes in a spectral range of 800 to 1800 cm−<sup>1</sup> and an integration time of 2 s for each point. The map spectra were analyzed with WiRE 3 and elaborated with Wolfram Mathematica (The Wolfram Centre, Oxford, United Kingdom), analyzing each map on the basis of characteristic peak intensities (1250, 1330 and 1450 cm−1). Individual Raman spectra were baseline corrected using a polynomial passing through at least eight points uniformly distributed in the spectral range. Then, spectra were rescaled in the intensity direction using min–max normalization, whereby the minimum value of a spectrum intensity is transformed into a 0, the maximum

value is transformed into a 1, and every other value is transformed into a decimal between 0 and 1 [22].

#### **3. Results—SERS Device**

*3.1. Characterization of Nanowires*

The SERS device, realized according to the description in the previous session, was characterized both with SEM during different steps of fabrication and with a spectrometer connected to an optical microscope, as described in the materials and methods session. Figure 2A shows the alumina membrane after the two steps of anodization. Figure 2B shows the template after 100 min of the widening process, which allows 15 nm of pore distance; Figure 2C shows the template after 120 min of the widening process, with a homogeneous pore distance of 5 nm. An ITO nanofilm was sputtered to the back of the alumina membrane as a conductive layer for the electrodeposition. Porous templates were then used to grow metal nanowires through the pores after electrodeposition. The electrodeposition Fe-Au nanowires form a composite material together with the alumina template near the bottom (Figure 4A), and after the removal of the alumina, free-standing Au-Fe nanowires were finally obtained(Figure 4B,C).

**Figure 4.** SEM micrographs. (**A**). Porous alumina with electroplated nanowires at the bottom near the metal contact. The thickness of the membrane is roughly 20 μm, as indicated from the green line. (**B**). Nanowires after alumina etch( scale bar 5 μm). (**C**). Magnified region marked as red square in Figure 3B; the scale bar is 500 nm.

> Since three different typologies of porous alumina templates based on pore diameter were realized, an increasing plating time was applied for small, medium and large pores in order to fill the different hollow volumes of the nanoporous material and ensure about the same height of nanowires. Figure 5A shows the top of the gold nanorods obtained with a porous alumina template without the widening process. Figure 5B,C shows Au-Fe nanowires with a distance, respectively, of 15 and 5 nm.

**Figure 5.** SEM micrographs. Fe-Au nanowires with distance between wires 40 nm (**A**), scale bar = 200 nm, 15 nm (**B**), scale bar = 90 nm and 5 nm (**C**), scale bar = 180 nm.

Transmission spectra of the nanowires was acquired and then convoluted by using a Savitzky–Golay filter. The peaks at 747 nm, 820 nm and 810 nm were transmitted with higher intensity by the nanowires with smaller distance (Figure S2), suggesting that the plasmonic resonance of the gold nanowires is higher when they are excited with an infrared laser line. For this reason, Raman spectroscopy was conducted with a laser line of 833 nm.

#### *3.2. Detection of Benzenedithiol and IgG Solution Dried on the Biosensor*

Free-standing Fe-Au nanowires were fabricated using the electrochemical method after two consecutive steps of electrodeposition. The SERS spectrum of Benzenedithiol 1,4 is shown in Figure 6. The performance of the device with 5 nm gap nanowires, as shown in Figure 6, is much better than the other cases when the distance is larger. The estimation of the signal intensity is of the order of 104 with respect to the thicker nanowire device. The signal of Benzenedithiol for the small gap spacing was measured at different points by mapping a region of the biosensor. The signal in Figure 6 (blue curve) is extracted from one point of the map. We repeated the spectra acquisition several times at that point and did not observe relevant changes. We mapped the other surfaces (with higher gaps between nanowires) and detected a poor signal.

**Figure 6.** Raman spectra of Benzenedithiol 1,4 chemiosorbed on Au nanowires with large distance (green line), nanowires with middle distance (red), and nanowires with small distance between nanowires (blue line).

As predicted by simulations in a previous work, the SERS signal of Benzenedithiol 1,4 shows a higher intensity due to the fact that the dipoles of gold nanorods on the head of nanowires, realized in this work, have, in the infrared region, a better electric-field enhancement, which was estimated to have a factor of 104. In Figure S4 of the Supporting Information, we report the SERS signal coming from Benzenedithiol (BDT) measured by the nanowire sensor device with three different configurations (large, middle, small gap) compared to the Raman spectrum of BDT acquired over a flat non-SERS substrate (flat silicon surface). In the image, all spectra are individually normalized to the maximum peak in the spectral range. Remarkably, the SERS and Raman signal are significantly different, proving the selective adsorption and enhanced Raman activity of BDT over the SERS nanowires device, compared to the spontaneous Raman scattering of the molecule in standard conditions.

To further assess the performance of the device, we used the biosensor to examine a solution of antibodies in low abundance ranges. Upon casting a sample drop containing IgG with an initial concentration of 0.44 mg/mL on the sensor surface, we waited until the complete evaporation of the solvent and analyzed the residual using the Raman

setup described in the methods of the paper. The Raman maps reported in Figure 7 were collected over a region of the sensor device partially coated with the sample drop. The maps in Figure 7a report the normalized Raman spectrum intensity measured at 1250 cm<sup>−</sup>1, 1330 cm−<sup>1</sup> and 1450 cm<sup>−</sup>1, respectively. Notably, the signal distribution in the maps matches with very high accuracy with the originating layout of the sample drop on the device (Figure 7b). The signal is high within the contact area of the sample with the sensor surface, then it sharply falls to nearly zero, moving away from the biological sample towards the free sensor surface. The very high correspondence between the expected spatial distribution of the biological sample with the Raman maps indicates that the device and the entire method is effective in performing the analysis of biological solutions. The spatial frequencies of 1250 cm−1, 1330 cm−<sup>1</sup> and 1450 cm−<sup>1</sup> that we have used as a reference in the analysis are central lines where the IgG vibrational activity is preferentially expressed. They represent the fingerprint of IgG.

**Figure 7.** SERS maps of IgG positioned on the sensor device, acquired at 1250, 1330 and 1450 cm<sup>−</sup>1, respectively (**a**). Optical image of the sample drop after evaporation on the device, and the region of the sample surface interrogated through SERS analysis (**b**). Collection of nine spectra randomly sampled from the SERS maps, where the characteristic peaks of IgG have been highlighted (**c**).

IgG molecules are characterized by a significant percentage (47%) of β-sheet conformation and only 7% of α-helix [23]. The β-sheet secondary structure is identified by the amide I broad band at 1673 cm−<sup>1</sup> and by the amide III region with a slight band at around 1243 cm<sup>−</sup>1. In the amide III region, the band at 1336 cm−<sup>1</sup> is also observable, evidencing the α-helix secondary structure portion [24]. The CH2 deformation (ρCH2) band is observed at around 1450 cm<sup>−</sup>1, associated with the protein structures. Other minor peaks are related to amino-acidic residues (e.g., phenylalanine at 1004 cm−1) and to the backbone skeletal vC–C vibration bands around 1030 to 1170 cm−<sup>1</sup> [25].

In Figure 7c, we report a number of Raman spectra extracted from the full-field Raman maps described above. Each of those spectra, randomly sampled from the maps, exhibit characteristic peaks at no less than one of the following frequencies: 1250 cm<sup>−</sup>1, 1330 cm−<sup>1</sup> and 1450 cm−1. Remarkably, above roughly 1500 cm−1, the Raman spectra in the grid seem to convey no or little information about the biological sample, perhaps because environmental or instrumental noise obscure the sample emission. Below the 1500 cm−<sup>1</sup> limit, the frequency content of the signal is consistent with the biological sample being IgGs.

The Raman spectra that we have reported in Figure 7c are representative examples of a larger set of data, all having the property of showing a very high Raman signal at 1250 cm<sup>−</sup>1, 1330 cm−1, 1450 cm−1, or a combination of these three frequencies. This is illustrated from the Raman maps reported in Figure 7a, where the signal calculated in correspondence with those reference values shows minimal irregularity. To demonstrate repeatability more convincingly, we report in Figure S3 of the Supporting Information section the complete set of Raman spectra acquired over the active area of the sensor device.

#### **4. Discussions**

The SERS signal coming from molecules in close proximity to a plasmonic nanomaterial shows a very high sensitivity to the distance of the molecules to the surface and the molecule orientation, which are parameters that are not or are minimally influenced by the operator during the measurement. As a result, for its nature, the SERS analysis of a compound and corresponding Raman spectra show poor reproducibility and repeatability. This limits the use of the technique as a quantitative method of analysis of biological systems. Nevertheless, while not perfectly identical, the Raman spectra that we have shown in Figure 7c all show a pattern similarity. Centered either at 1250 cm<sup>−</sup>1, 1330 cm−<sup>1</sup> or 1450 cm<sup>−</sup>1, the spectra have a peak that is characteristic of IgG, as reflected by the maps reported in part of the same figure, where the Raman intensity calculated for those peaks shows very high uniformity over the sample sensing area.

To understand the clinical implications of the results, it is useful to put the IgG concentration of 0.44 mg/mL that we have used in this work in context. The typical values of IgG normally found in adults fall in the 7 to 15 mg/mL interval, called the reference range [26], which are between 15 and 30 times higher than our study's sample concentration. Thus, even without testing the device with ultra-low concentrated solutions, the results of the work indicate that this biosensor is suitable to detect IgG fluctuations downward or upward relative to the reference range. Upward oscillations (high levels of IgG) may be indicative of pathological states including chronic infection, such as HIV, multiple myeloma, chronic hepatitis, and multiple sclerosis. On the other hand, downward oscillations (low levels of IgG) occur, such as, for example, in macroglobulinemia; in some types of leukemia; and in nephrotic syndrome, a type of kidney damage. Further to this end, in a more sophisticated evolution of the device that will be developed over time, the gold sensor surface will be functionalized with antibodies [27] or aptamers [28] for the selective capture of biomarkers. This sensing device will achieve the recognition of antigens in complex mixtures in very low abundance ranges, combining the characteristics of low cost, high resolution, high sensitivity, and selectivity.

#### **5. Conclusions**

In the present work, a method of SERS device nanofabrication using alumina template was optimized for molecular sensing. To do so, alumina templates were fabricated according to two anodization steps. The SERS device, realized according to the description in the previous sessions, was characterized geometrically and optically with SEM during different steps of fabrication and with a spectrometer. Porous templates were used to grow metal nanowires through the pores after electrodeposition. A series of SERS measurements at 830 nm using a forest of nanowires, organized in a hexagonal lattice, with a gap of only 5 nm to detect the Raman signal of chemisorbed benzenedithiol 1,4 shows a large increase in the local intensity with respect to the forest of nanowires with larger distance. As reported in a previous paper [8], the enhancement factor of the Raman signal for this device is of the order of 104. As the thiol group can be linked to many biomolecules of interest such as proteins, the device has the potential to be used as a biosensor for the detection of a few biomolecules in different fields such as microfluidics, proteomics and optoelectronics. Further analysis of solutions of biomedical interest—i.e., IgG—in low abundance ranges confirm the ability of the device to detect biomolecules with potential applications in the treatment and diagnosis of diseases.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/bios11060181/s1, Figure S1: Etching process of porous alumina. The samples are upside down and floating on chrome solution which is kept at 40 ◦C. Figure S2: Etching process of porous alumina. The samples are upside down and floating on chrome solution which is kept at 40 ◦C. Figure S3: Complete set of Raman spectra acquired over the active area of the sensor device. Figure S4: SERS signal coming from Benzenedithiol (BDT) measured by the nanowires sensor device with three different configurations (big, middle, small gap) compared to the Raman spectrum of BDT acquired over a flat non-SERS substrate (flat Silicon surface). In the image, all spectra are individually normalized to the maximum peak in the spectral range.

**Author Contributions:** G.M. fabricated the SERS devices, acquired the SEM images, prepared the solution of BDT, performed the measurements of BDT sensing using Raman spectroscopy and wrote the original draft, M.L.C. performed the measurements of IgG sensing using Raman spectroscopy and contributed to the analysis of the results. F.G. analyzed and discussed the results. G.M. and F.G. wrote, reviewed and edited the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data that support the findings of this study are available from the corresponding authors upon reasonable request.

**Acknowledgments:** G.M. would like to thank the Dresden Center for Nanoanalysis of the Technische University Dresden for support provided for the SEM analysis.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Plasmonic Interferometers as TREM2 Sensors for Alzheimer's Disease**

**Dingdong Li 1, Rachel Odessey 1, Dongfang Li <sup>1</sup> and Domenico Pacifici 1,2,\***


**\*** Correspondence: Domenico\_Pacifici@brown.edu

**Abstract:** We report an effective surface immobilization protocol for capture of Triggering Receptor Expressed on Myeloid Cells 2 (TREM2), a receptor whose elevated concentration in cerebrospinal fluid has recently been associated with Alzheimer's disease (AD). We employ the proposed surface functionalization scheme to design, fabricate, and assess a biochemical sensing platform based on plasmonic interferometry that is able to detect physiological concentrations of TREM2 in solution. These findings open up opportunities for label-free biosensing of TREM2 in its soluble form in various bodily fluids as an early indicator of the onset of clinical dementia in AD. We also show that plasmonic interferometry can be a powerful tool to monitor and optimize surface immobilization schemes, which could be applied to develop other relevant antibody tests.

**Keywords:** TREM2 sensors; Alzheimer's disease; plasmonic interferometry; optical biosensor; surface functionalization

#### **1. Introduction**

Alzheimer's disease (AD) is a chronic neurodegenerative disorder that affects more than five million Americans and approximately 50 million people worldwide [1]. AD causes loss of memory followed by loss of ability to think and communicate and, finally, loss of life [2]. As it progresses, AD has devastating effects on the ability of subjects to carry out the events of their day-to-day lives and can create significant mental and emotional distress for loved ones whose identities and relationships to the patient are forgotten. AD is one of the costliest disorders to society, costing over a quarter of a trillion dollars in 2017 alone in the United States [3]. Despite the many incentives and the correspondingly tremendous efforts of biopharmaceutical researchers, no disease-modifying therapy is yet available for AD and the drug candidates put forward to treat or prevent the onset of AD symptoms continue to fail in clinical trials [4].

AD is traditionally diagnosed by monitoring subjects' behavioral changes because it is challenging to diagnose more definitively without an invasive examination of the brain. This inexact method, which has a misdiagnosis rate of up to 45%, contributes to the untenably high attrition rate of drug candidates by creating an incomplete understanding of disease etiology and, as such, a lack of robust and valid biomarkers on the causal path of the disease [5]. Such biomarkers are essential to effective patient care and, specifically, the efficient development of drug treatments because they enable early and more accurate (i) diagnosis and stratification of patients during trial enrollment, (ii) measurement of target engagement and modulation, and (iii) testing of the therapeutic hypothesis in clinical trials that are already extremely long and costly [6].

#### *1.1. TREM2 as Biomarker for Early-Onset Detection of AD*

Most cases of AD have a complex, highly polygenic architecture [7]. A number of causal genes for AD have been identified in recent years, many of which play important

**Citation:** Li, D.; Odessey, R.; Li, D.; Pacifici, D. Plasmonic Interferometers as TREM2 Sensors for Alzheimer's Disease. *Biosensors* **2021**, *11*, 217. https://doi.org/10.3390/bios11070217

Received: 1 June 2021 Accepted: 26 June 2021 Published: 1 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

roles in myeloid cells such as microglia, immune cells in the brain [8–10]. One such gene is Triggering Receptor Expressed on Myeloid Cells 2 (TREM2) [11–13], which senses brain tissue damage due to aging or neurodegeneration by triggering a microglial response aimed at scavenging and clearing brain tissue debris [14–17]; genetic variants which impair this function also increase the risk of AD more than threefold [18,19]. Recent studies have shown that TREM2 was abnormally elevated 5 years before the expected onset of symptoms in AD patients [13,16,17]. These findings suggest that TREM2 is in the causal path to disease and among the strongest genetic risk factors for AD [20]. Moreover, TREM2 could be used as an effective biomarker for early stage detection of AD [12] and other neurodegenarative diseases [21].

TREM2 may be found in its soluble form in cerebrospinal fluid (CSF) and other bodily fluids, such as saliva [22]. CSF concentrations of soluble TREM2 have been shown to be higher in AD cases than in controls, they correlate with markers of neurodegeneration, and may be used to quantify glial activation in AD [23]. Recently, soluble CSF TREM2 has also been proposed as a surrogate immune biomarker of neuronal injury in Parkinson's disease [24]. Because of its extremely low concentration in CSF (∼1–5 ng/mL [13,16,17]), TREM2 is conventionally detected and quantified by enzyme-linked immunosorbent assay (ELISA) [25], a well-established and highly sensitive plate-based assay which uses a multilayered format with a labeled secondary antibody. ELISA, while widely in use and highly sensitive, involves a multi-step incubation protocol that usually calls for 2 to 8 h of preparation time for each step. This, together with its stringent washing and blocking protocols and fluorescent labeling steps, makes ELISA a time-consuming assay that is hardly integrable into point-of-care, portable, or multiplexed biosensing platforms.

#### *1.2. Plasmonic Interferometry for Sensing Applications*

Several optics- and nanostructure-based alternatives to conventional biosensing methods have been developed, including sensing based on magnetic nanoparticles [26], carbon nanotubes and other nanostructures [27], quantum dots [28], and surface plasmon polaritons (SPPs) in gold, especially gold nanoparticles [29]. SPPs are collective oscillations of electrons that may occur when light interacts with the interface between a dielectric and a metal. One promising use of SPPs for biosensing is in plasmonic interferometry, which has been demonstrated to host biosensors with extremely high sensitivity and selectivity [30–44]. In this method, SPPs are generated and propagate within a micrometer-scale optical interferometer such as the one in Figure 1a,b: light is incident on a groove-slit-groove (GSG) geometry patterned onto a metal film; optical scattering at the subwavelength-width grooves couples a fraction of the incident light field into SPPs that travel across the surface of the metal, toward the slit. SPPs accrue a propagative phase that depends on the physical distance traveled (that is, the interferometer arm length) and on the SPP refractive index, which is given by [45]:

$$
\hat{\eta}\_{\rm SPP}(\lambda) = \eta\_{\rm SPP}(\lambda) + i\kappa\_{\rm SPP}(\lambda) = \sqrt{\frac{\epsilon\_{\rm m}(\lambda)\epsilon\_{\rm d}(\lambda)}{\epsilon\_{\rm m}(\lambda) + \epsilon\_{\rm d}(\lambda)}},\tag{1}
$$

where *<sup>n</sup>*SPP and *<sup>κ</sup>*SPP are the real and imaginary parts of *<sup>n</sup>*SPP, respectively, and <sup>m</sup> and <sup>d</sup> are the complex dielectric functions of the corresponding metal and dielectric material. A small change in <sup>d</sup> (which may be due, for example, to the presence of molecules near the surface) can produce a significant change in the optical path length traveled by an SPP across the surface of the metal [30]. When the counter propagating SPPs arrive at the subwavelength-width slit, they interfere with each other and with the optical beam incident at the slit location before coupling back into free space through the slit. This interference process modulates the light intensity that is transmitted through the slit and detected in the far-field [30,44,46,47]. Since SPPs are highly confined near the metal-dielectric interface, their propagative phase is highly sensitive to the refractive index at the surface. Therefore, the presence of an analyte adsorbed to the surface can be detected even in submonolayer concentrations by observing changes in the transmission spectra determined by plasmonic interferometry [30,39,42,43,46]. The sensitivity of a device based on plasmonic interferometry can be enhanced by simply increasing the interferometer arm length, and the signal to noise ratio can be improved by simply changing the geometrical parameters (such as slit/groove width, depth, and length) and the incident wavelength [30,42,44,46].

**Figure 1. Design for TREM2 sensor chip based on plasmonic interferometry.** (**a**) Cross-section schematic of the groove-slit-groove (GSG) architecture, which shows a slit flanked by two grooves, from which SPPs are excited by light diffraction and propagate towards the slit aperture, where they interfere and are then transmitted back into free space for far-field detection; diagram includes an example of an antigen complex, further described in Figure 2. The bottom slab represents quartz, the middle titanium, and the top layer gold. (**b**) Scanning electron micrograph (SEM) of a GSG interferometer with *p*<sup>1</sup> = 7.65 μm, *p*<sup>2</sup> = 8.15 μm. (**c**) Schematic of plasmonic interferometer sensor chip layout. The chip contains four nominally identical sensing spots enabling multiplex sensing applications. The yellow area indicates quartz covered by gold and the blank area is an uncoated quartz window used for optical alignment. (**d**) Schematic of a representative active sensing area. Each sensing area contains two columns of single slits and two columns of nominally identical asymmetric GSG interferometers with separation distance of 300 μm. The slit/grooves in each interferometer are ∼20 μm long and, within each column, the distance between two adjacent interferometers is ∼40 μm.

**Figure 2. Surface immobilization protocol for capture of TREM2 in solution.** Chip surface was treated with (**i**) an RCA1 cleaning procedure followed by (**ii**) (3-Aminopropyl)triethoxysilane (APTES) to form an amino-terminated surface. Sulfo-NHS-biotin (sulfo-N-Hydroxysulfosuccinimide biotin) covalently attaches to the amino groups of the surface (**iii**) and subsequently captures streptavidins (**iv**). Finally, the streptavidin functionalized chip is bound by the biotinylated TREM2 antibody (**v**) for sensing of the TREM2 molecule (**vi**). The green dot in (**v**) represents the sulfo-NHS ester of biotin that acts as the biotinylation reagent and allows to form a stable bond between the antibody and the streptavidin already bound to the sensor surface, as reported in (**iv**).

Compared to more conventional surface plasmon resonance (SPR) sensing platforms, sensors based on plasmonic interferometry (PI) retain high sensitivity and low detection limits whilst providing several advantages, including: (1) broadband (as opposed to single wavelength) operation that allows for spectroscopic capabilities and detection of refractive index changes at multiple wavelengths of interest, simultaneously [30,46]; (2) wide-angle excitation of SPPs (as opposed to the specific angle required to excite the surface plasmon resonance on the metal surface) that enables less stringent alignment requirements [30,43,46] and the use of incoherent light sources, which can also be integrated directly on the sensor surface [42]; (3) smaller sampling volumes and sensing areas, which lead to higher levels of device integration and multiplexing, since the same sensor chip can contain millions of individually addressable devices (over an area of just 1 cm2) that can be used to detect multiple analytes and perform screening on multiple patients at the same time [30,39,42].

Here, we employ plasmonic interferometry to develop and assess a surface functionalization protocol designed to detect TREM2 in solution. Specifically, we (i) monitor each step of the proposed functionalization protocol using the intensity change in the transmitted spectra of plasmonic interferometers and (ii) employ this surface functionalization scheme to detect biological levels of TREM2 in solution.

#### **2. Biosensing Chip: Design, Fabrication and Implementation**

#### *2.1. Fabrication of Biosensing Chip Based on Plasmonic Interferometry*

The proposed biosensing chip comprises four arrays of nominally identical GSG plasmonic interferometers that were designed and fabricated with asymmetrical arm lengths (*p*<sup>1</sup> = 7.65 μm, *p*<sup>2</sup> = 8.15 μm) to optimize the device sensitivity to refractive index change caused by TREM2 adsorbed to the sensor surface [30,39,42,44,46]. First, a 4 nm-thick titanium adhesion layer was deposited by electron-beam evaporation onto a previously cleaned 1 mm-thick fused quartz slide followed by a ∼200 nm-thick gold layer. The thickness of the titanium layer was determined experimentally in order to cause strong

surface adhesion of the gold layer, which would otherwise tend to delaminate and form blisters if directly deposited on glass. Four sensing spots were milled onto the metal film with a focused ion beam (FIB) using a gallium ion source. Each sensing spot contains two columns of seven nominally identical GSG interferometers and two columns of single slits for the purpose of statistical analysis and normalization, as shown in Figure 1c,d. The distance between two parallel columns and two adjacent interferometers in the same column are 300 μm and 40 μm, respectively. The area of each sensing spot is about 0.2 mm2. Figure 1b shows a scanning electron micrograph of a representative GSG interferometer with left arm length 7.65 μm and right arm length 8.15 μm, within 2% fabrication error. Each groove is approximately 20 μm long, 200 nm wide, and ∼50 nm deep; each slit is 20 μm long, 180 nm wide, and ∼200 nm deep. These parameters were determined by analyzing SEM/FIB cross-sections (not reported). The actual values were chosen to optimize signalto-noise light transmission ratio, SPP excitation efficiency, amplitude of SPP interference, and overall device sensitivity to refractive index change. Figure 1a shows a schematic illustration of a cross-section of the plasmonic interferometer, with two grooves flanking a slit in order to facilitate incoupling of incident light into SPPs (by optical scattering from each groove) and outcoupling of SPPs back into free space (through the slit).

#### *2.2. Surface Functionalization of Optical Biochip*

TREM2 antibodies were immobilized on the gold surface by using the protocol shown in Figure 2. The gold chip was first cleaned in an RCA1 solution, a mixture of 20 mL 29% ammonium hydroxide, 20 mL 30% hydrogen peroxide, and 100 mL deionized water (DI water), at 75 ◦C for 10 min. RCA1 cleaning removes organic residues from the gold surface and forms hydroxyl groups that facilitate silanol groups binding to the surface, as shown in Figure 2i. Next, the cleaned chip was soaked in 8 mL freshly prepared 2% (3-aminopropyl)triethoxysilane (APTES) solution for 1 h at room temperature to form an amino terminated surface, as shown in Figure 2ii. The 2% APTES solution was obtained by serial dilution of 99% APTES (Sigma-Aldrich, Burlington, MA, United States) in DI water. The APTES treatment time and concentration were chosen based on a standard surface plasmon resonance (SPR) functionalization protocol for a gold chip [48,49]. Then, the chip was rinsed thoroughly with DI water to remove loosely adsorbed APTES molecules that hadn't formed any covalent bonds with the hydroxyl terminated surface.

Next, biotinylation of the surface was achieved by soaking the chip in a 0.5 mg/mL sulfo-NHS-biotin solution (sulfo-N-Hydroxysuccinimide biotin ester sodium salt, Thermo Fisher Scientific, Waltham, MA, United States) for a duration of 2.5 h, followed by rinsing with phosphate buffer solution (PBS) and DI water. The sulfo-NHS-biotin solution was prepared in 0.01 M PBS (pH = 7.4, Sigma-Aldrich) immediately before using. On an amino-terminated gold surface, sulfo-NHS-biotin will covalently attach to the amino groups via ester linkage, as shown in Figure 2iii. Afterwards, the chip was treated with streptavidin by 1 h immersion in a 5 mg/mL fresh streptavidin (Iyophilized powder, Sigma-Aldrich) PBS (0.01 M, pH = 7.4) containing 0.05% Tween 20, which was used to minimize the non-specific binding of streptavidin to the biotinlyated gold surface. The resulting streptavidin-coated chip was subsequently washed three times by washing buffer (0.01 M PBS with 0.05% Tween 20) to remove loosely adsorbed streptavidin molecules and then followed by PBS (0.01 M, pH = 7.4) to wash off the Tween 20 solution, leaving behind a streptavidin-functionalized metal surface, as shown in Figure 2iv.

In order to saturate the sulfo-NHS-biotin binding sites on the APTES-functionalized chip and to form a monolayer of streptavidins and antibodies on the surface, the chip was optically characterized after every hour of chemical treatment until no significant spectral peak shifts were observed in the optical transmission spectra through the GSG interferometers; the concentration ratio of sulfo-NHS-biotin and streptavidin was optimized to achieve a high biotinylated antibody covering rate. Subsequently, the chip was incubated with 0.1 mg/mL biotinylated TREM2 antibody PBS for 2.5 h, as shown in Figure 2v. Finally, human TREM2 biotinylated antibody was purchased from R&D (Cat. #BAF1828, R&D) with 50 μg bovine serum albumin (BSA) per 1 μg as a carrier protein; the yielded solution contains 0.5% BSA in PBS. In sensing experiments, when TREM2 binds to the antibody the resulting chip surface is schematically shown in Figure 2vi.

#### *2.3. Optical Characterization of Surface Functionalization Steps with Plasmonic Interferometry*

After each functionalization step, the sensor chip was rinsed with DI water to remove unbound molecules/proteins and thoroughly dried under a stream of nitrogen gas to remove any residual water droplets/molecules that might affect the final wavelength shift measurement. This washing and drying protocol was enforced throughout the experiments to make sure that actual protein-protein binding events were measured, which allowed us to better validate the proposed surface functionalization and sensing methods. After surface rinsing and drying, the transmitted intensity of each plasmonic interferometer on the sensor chip was measured. To perform the optical characterization, a plasmonic interferometer sensor chip containing functionalized plasmonic interferometers was placed on the controllable moving stage of a Nikon Eclipse Ti Series inverted microscope. The collimated white light beam from a xenon arc lamp coupled with an optical lens system was focused onto the plasmonic interferometer sensor chip, transmitted through the slit, and collected by an objective lens (0.6 NA, 40×). The collected light was further dispersed by a monochromator and detected by a CCD camera. Transmitted spectra of the single slit accompanying each plasmonic interferometer were taken as well for spectral normalization.

Each solid line in Figure 3 corresponds to the mean value of seven normalized transmitted light intensities after each functionalization step,

$$I\_{\text{n,mean}} = \frac{1}{7} \sum\_{k=1}^{7} \frac{I\_{\text{GSG},k}(\lambda)}{I\_{\text{SS},k}(\lambda)},\tag{2}$$

where *I*GSG,*<sup>k</sup>* is the background-corrected light intensity transmitted through the *k*th GSG interferometer and *I*SS,*<sup>k</sup>* is the background-corrected light intensity transmitted through the *k*th single slit, which serves as reference to estimate the light intensity baseline. The light gray shading represents the standard deviation that results from the proposed normalization and averaging procedure.

As shown in Figure 3, several maxima (minima) are observed as a result of constructive (destructive) interference between the counter-propagating SPPs and the original beam at the slit location. To fix the ideas and make it easier to follow the evolution of the various functionalization steps, we choose a specific peak as a reference, although the shift caused by the addition of reagents that are adsorbed to the surface can be tracked across the whole spectrum, which is an additional advantage of plasmonic interferometry compared to more conventional SPR that typically operates at single wavelength. After RCA1 cleaning of the sensor gold surface, the measured peak wavelength of the reference peak is around 588.1 nm (Figure 3i). After each functionalization step, the refractive index near the surface is modified by another layer of biomolecules, resulting in a spectral peak shift towards longer wavelengths. For the APTES and sulfo-NHS-biotin treatment steps (Figure 3ii,iii, respectively) the relative spectral peak shifts compared to their previous steps, are both ∼1.3 nm, while the relative spectral peak shift caused by the addition of streptavidin is ∼2.1 nm (Figure 3iv); finally, surface capture of the biotinylated TREM2 antibody produces an additional shift of ∼1.3 nm. A total spectral peak shift of ∼6 nm was observed after the entire functionalization procedure. The data shown in Figure 3 suggest that plasmonic interferometers can be effectively used to monitor the evolution of complex functionalization steps that involve adsorption of monolayer- and submonolayer-thick molecules to the sensor surface.

**Figure 3. Tracking functionalization steps through plasmonic interference spectra.** Measured results of transmitted intensity spectra after (**i**) RCA1, (**ii**) APTES, (**iii**) sulfo-NHS-biotin, (**iv**) streptavidin, and (**v**) biotinylated TREM2 antibody treatment. Solid lines represent the mean value of normalized intensity spectra averaged over seven nominally identical GSG interferometers after each functionalization step, as illustrated by the lower left insets. Light gray areas represent standard deviation. The vertical dashed line indicates the position of a representative transmission peak (588.1 nm) that results from constructive SPP interference after RCA1 cleaning. The black arrows mark the wavelength shift (Δ*λ*) in this reference peak as the result of new constructive interference conditions after each functionalization step.

#### **3. TREM2 Biosensing Experiment: Results and Analysis**

#### *3.1. Uniformity Study of Surface Functionalization Steps with Plasmonic Interferometry*

The sensing chip was functionalized and stored in a refrigerator at 4 ◦C prior to testing. Immediately before TREM2 detection, the functionalized chip was blocked by 0.5∼1% BSA PBS in order to saturate nonspecific binding sites and prevent false positive results [50]. Figure 4 shows the results of experiments carried out on each of the four sensing spots of the chip, tracking the spectral shift due to each functionalization step and after every hour of BSA blocking. Due to saturation of the blocking agent on the surface over time, the spectral shift due to BSA stabilized, enabling optimization of the BSA blocking time. We estimate that BSA almost saturated the surface of the chip after 4 h of blocking; therefore, the chip was blocked for 7 h at room temperature to ensure complete saturation. After the

given blocking time, the chip was rinsed three times with a washing buffer (PBS with 0.05% Tween 20) followed by PBS and DI water, then dried under purified nitrogen gas flow.

**Figure 4. Uniformity study of surface functionalization steps across four sensing spots.** Circles represent the mean value of wavelength shift (Δ*λ*) measured from 7 GSG plasmonic interferometers after each functionalization step, labelled in the horizontal axis. Error bars represent the standard deviation. Lines and symbols with different colors indicate data measured from different sensing spots.

#### *3.2. Sensing TREM2 Antigen-Antibody Binding Interaction with Plasmonic Interferometry*

For sensing experiments, various concentrations of TREM2 in buffer solution were obtained by diluting a stock solution of TREM2 (0.27 mg/mL recombinant Human-TREM2 Fc Chimera PBS (Cat. #BAF1828, R&D)) in 0.5% BSA PBS (0.01 M, pH = 7.4). BSA serves as a protein stabilizer to maintain the integrity of TREM2 at low concentration.

TREM2 binding experiments were performed by directly dispensing 40 μL TREM2 BSA PBS on the chip and then drying it to verify that surface capture and adsorption had effectively taken place. More specifically, after a given binding time interval, the chip was rinsed three times with a washing buffer solution (PBS with 0.05% Tween 20) followed by PBS and then DI water, dried under nitrogen gas flow. Then, transmitted spectra through the slit of each GSG plasmonic interferometer were measured to assess the presence of TREM2 at the sensor surface. This process was repeated to obtain TREM2 kinetic binding curves as a function of time and for various initial TREM2 concentrations in buffer solution.

Figure 5 shows the observed peak shift Δ*λ* as a function of time in a binding experiment performed with a 2.7 ng/ml TREM2 0.5% BSA PBS. Blue circles and error bars represent the mean value and standard deviation, respectively, of Δ*λ* measured from seven GSG plasmonic interferometers. The inset of Figure 5 displays the average of normalized intensity spectra across all seven GSG plasmonic interferometers at different time steps, showing a spectral shift towards longer wavelengths as the reaction progresses. The time-domain sensing curve (or "sensorgram") reported in Figure 5 shows that by tracking the wavelength shift as a function of time we can indeed monitor the capture of TREM2 antigens from antibody binding sites and subsequent formation of a sub-monolayer of TREM2 on the sensor surface.

**Figure 5. Sensing temporal evolution of TREM2 surface binding kinetics with plasmonic interferometry.** Blue circles represent the mean peak shift (Δ*λ*) averaged over seven nominally identical GSG plasmonic interferometers as the result of temporal evolution of antigen-antibody binding reaction for a 2.7 ng/ml TREM2 0.5% BSA PBS. Error bars represent the standard deviation. Bottom right inset illustrates the normalized transmitted spectra (averaged over seven identical GSG interferometers) measured at each time step. Color changing from dark red to yellow represents increasing reaction time from 0 to 60 min.

To understand the kinetic interaction of the binding between TREM2 and its antibody, we first consider an equilibrium model for the chemical reaction [39,51–56]. Ideally, the antibody-antigen interaction is a reversible reaction:

$$
antidodyt + antigen \rightleftharpoons complex. \tag{3}$$

The time-dependent rate equation that governs this reaction can be expressed by:

$$\frac{d[complex]}{dt} = k\_f[Ab][Ag] - k\_b[complex] \tag{4}$$

where [*complex*] is the molar concentration of antibody-antigen complex, [*Ab*] is the molar concentration of unoccupied antibodies, [*Ag*] is the molar concentration of antigen in the solution, *k <sup>f</sup>* is the forward reaction constant and *kb* is the backward reaction constant. When the reaction reaches equilibrium, we have:

$$k\_f[Ab]\_{eq}[Ag]\_{eq} = k\_b[complex]\_{eq} \tag{5}$$

$$K\_d \equiv \frac{k\_b}{k\_f} = \frac{[Ab]\_{eq}[Ag]\_{eq}}{[complex]\_{eq}} \tag{6}$$

where *Kd* is the dissociation constant.

In addition, if we assume that the total number of antibodies immobilized at the sensor surface is fixed and the concentration of antigens is constant due to the large volume (or continuous dispensing) of the sample solution onto the surface, the following expression holds true at all times:

$$|Ab|\_0 = |Ab| + |
complex|\,\tag{7}$$

where [*Ab*]<sup>0</sup> is the initial concentration of the antibodies immobilized on the sensor surface. Applying Equations (6) and (7) to Equation (4) and taking a time integral, we can obtain:

$$\left[\text{complex}\right] = \frac{[Ag][Ab]\_0}{[Ag] + K\_d} [1 - e^{-(k\_f[Ag] + k\_b)t}].\tag{8}$$

Equation (8) implies that: (a) the concentration of the antibody-antigen complex at the surface increases exponentially as a function of time; (b) the concentration of the antibodyantigen complex at equilibrium is always lower than the initial concentration of antibodies immobilized at the sensor surface; and (c) by increasing the antigen concentration in solution, the binding kinetics should occur with a faster rate and the equilibrium complex concentration should also increase. Figure 5 reports an exponential fit of the experimental data, where we assumed that the relative peak shift Δ*λ* was directly proportional to the number of TREM2 (the antigen) proteins captured by the immobilized antibodies and binding at the sensor surface over time. The data were fit by <sup>Δ</sup>*<sup>λ</sup>* <sup>=</sup> *<sup>a</sup>*(<sup>1</sup> <sup>−</sup> *<sup>e</sup>*−*t*/*τ*), where *a* = 7.12 nm and *τ* = 13.44 min. Several studies have shown that protein adsorption is a very sophisticated process, strongly influenced by experimental parameters, such as surface wettability, pH, protein structure, and other factors [57,58]. Moreover, proteins could be adsorbed to the surface in a multilayer fashion, especially on a hydrophobic surface, which would produce a higher (and therefore more readily detectable) Δ*λ* than that caused by the formation of a single monolayer.

To further validate the model, we performed sensing experiments to detect TREM2 in solution with three different physiological concentrations (1.35 ng/ml, 2.7 ng/ml, and 8.1 ng/ml) by using three of the four sensing spots of the chip, separately, and investigated the concentration dependence of the binding kinetics. Note that the chip was first regenerated to bare gold by 15 min of RCA1 cleaning, then functionalized and blocked again with BSA PBS before TREM2 detection. After regeneration, we confirmed that the normalized transmission spectra went back to the initial spectra, well within the confidence intervals reported in Figure 3. Figure 6 shows the sensing results averaged over the three functionalized sensing spots after interaction with 0.5% BSA PBS spiked with different TREM2 concentrations. The Δ*λ* associated with the binding reaction corresponding to 1.35 ng/ml and 2.7 ng/ml TREM2 concentration showed a single-exponential time dependence <sup>Δ</sup>*<sup>λ</sup>* <sup>=</sup> *<sup>a</sup>*(<sup>1</sup> <sup>−</sup> *<sup>e</sup>*−*t*/*τ*), where *<sup>a</sup>* <sup>=</sup> 2.8 nm, *<sup>τ</sup>* <sup>=</sup> 41.08 min for 1.35 ng/ml and *<sup>a</sup>* <sup>=</sup> 6.89 nm, *τ* = 15.04 min for 2.7 ng/ml, which is consistent with the parameters identified for the same concentration in Figure 5. When the concentration of TREM2 increased from 1.35 ng/ml to 2.7 ng/ml, the binding time constant *τ* decreased, which confirms that the reaction occurred faster in the presence of a higher concentration of antigens, as predicted by the model. Additionally, the parameter *a*, which represents the total wavelength shift at equilibrium, increased with the concentration of antigens. These experimental findings are consistent with the simple model described by Equation (8).

Interestingly, the binding reaction data corresponding to the higher concentration (8.1 ng/ml TREM2) displayed a kinetic curve that has a second slow rising stage, which could be fit with a double-exponential function: <sup>Δ</sup>*<sup>λ</sup>* <sup>=</sup> *<sup>a</sup>*1(<sup>1</sup> <sup>−</sup> *<sup>e</sup>*−*t*/*τ*<sup>1</sup> ) + *<sup>a</sup>*2(<sup>1</sup> <sup>−</sup> *<sup>e</sup>*−*t*/*τ*<sup>2</sup> ), where *a*<sup>1</sup> = 5.83 nm, *τ*<sup>1</sup> = 8.1 min, *a*<sup>2</sup> = 6.08 nm, and *τ*<sup>2</sup> = 476.19 min. This could indicate the presence of multiple adsorption mechanisms with distinct dissociation constants *Kd* that were active at higher analyte concentrations. These different values of *Kd* were manifested in the double-exponential curve by unique binding time constants *τ*<sup>1</sup> and *τ*<sup>2</sup> and associated wavelength shifts *a*<sup>1</sup> and *a*2. Previous studies reported similar two-stage adsorption kinetics, which may be due to a number of possible mechanisms. For instance, the initial stage may be caused by formation of a single monolayer at the surface, while the second slow rising stage could correspond to multilayer condensation of proteins on the surface [57,59]. The longer time constant may also arise from more subtle adsorption mechanisms the various proteins may be subjected to: for example, for BSA, conformational changes may be the primary reason for multilayer conformation, while for the TREM2 antibody (immunoglobulin G), multilayer conformation behavior could be significantly

influenced by long-range electrostatic interaction [60]. In addition, higher-order interaction effects during the adsorption process could cause the combined kinetics to differ significantly from a summation of two single component adsorption kinetics [61,62], especially at higher analyte concentrations.

**Figure 6. Binding times for different TREM2 concentrations.** Temporal evolution of peak wavelength shifts measured from normalized transmission spectra for different TREM2 concentrations. Error bars represent standard deviation from 7 GSG interferometers. Dashed lines represent exponential fit using the kinetic model provided in the text.

Although a more complex model may be needed to fully describe all of the possible microscopic phenomena underlying the actual reaction kinetics, to first approximation the data presented in Figure 6 can be fit using the simple exponential model which allows us to generate reliable calibration curves that accurately describe the proposed sensing mechanism. For instance, by choosing a given incubation time of *t* = 10 min, the wavelength shifts observed at three different concentrations are statistically different, enabling us to infer the concentration based on the observed Δ*λ*. However, we believe that a fit to the full time-resolved data set is an overall more statistically significant analysis of the acquired data and can lead to a better discrimination between different analyte concentrations, as shown in Figure 6. The full set of data presented so far validates the possibility of using plasmonic interferometry coupled with surface functionalization as a viable sensing scheme for TREM2 detection in solution.

#### **4. Conclusions**

We demonstrate a biophotonic sensing platform to monitor the chemical reactions associated with surface functionalization of a metal in time series through wavelength shifts in the transmission spectra of plasmonic interferometers. This work helps us devise and assess a functionalization protocol for detecting TREM2, a biomarker associated with the development of AD and other neurodegenerative diseases, in solution. We devise a chemical equilibrium model which we fit with the reported data. The fitting parameters from the data reinforce the theoretical model and are consistent between experiments, suggesting that the experimental process is repeatable. The chemical equilibrium model allows us to generate calibration curves for the reported data which are able to differentiate between different concentrations of TREM2. The results reported here open up the possibility to employ the proposed sensing platform based on plasmonic interferometry for the detection of physiological concentrations of TREM2 in CSF or other bodily fluids. In the future, plasmonic interferometry may be a promising method to test functionalization protocols for deployment on other scalable platforms, such as colloidal nanoparticles, which are widely

used in biological testing, to help validate the importance of TREM2 to the development pathway of AD. Additionally, plasmonic interferometry may be adapted to other surface functionalization protocols for different antibody-antigen pairs to enable development of a wide array of testing schemes for a variety of clinically relevant biomarkers.

**Author Contributions:** Conceptualization, D.P., D.L. (Dingdong Li) and D.L. (Dongfang Li); data curation, D.L. (Dingdong Li), D.L. (Dongfang Li) and D.P.; formal analysis, D.L. (Dingdong Li), D.L. (Dongfang Li) and D.P.; investigation, D.L. (Dingdong Li); methodology, D.L. (Dingdong Li), D.L. (Dongfang Li) and D.P.; software, D.L. (Dongfang Li) and D.P.; validation, D.L. (Dingdong Li), D.L. (Dongfang Li) and D.P.; project administration, D.P.; resources, D.P.; visualization, D.L. (Dingdong Li), D.L. (Dongfang Li), R.O., and D.P.; writing—original draft preparation, D.L. (Dingdong Li); writing—review and editing, D.L. (Dongfang Li), D.L. (Dongfang Li), R.O. and D.P.; supervision, D.P.; project administration, D.P.; funding acquisition, D.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the National Science Foundation (NSF), CBET Div. Of Chem., Bioeng., Env., and Transp. Sys., ENG Directorate For Engineering, Grant. No. 1842605, "EAGER: Development of Surface Chemistry and Plasmonic Interferometers for Early-Onset Detection of Alzheimer Disease;" Program BIOSENS-Biosensing, Program Ref. Codes No. 7916, 9150.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author.

**Acknowledgments:** The authors would like to thank Jing Feng and Tinayi Shen for useful contributions and engaging discussions. D.P. would like to thank Alessandro, Paola, and Emma for their understanding and support throughout this research project.

**Conflicts of Interest:** The authors declare that there are no conflicts of interest related to this article.

#### **Abbreviations**

The following abbreviations are used in this manuscript:


#### **References**

