*Article* **Hg2**<sup>+</sup> **Optical Fiber Sensor Based on LSPR Generated by Gold Nanoparticles Embedded in LBL Nano-Assembled Coatings**

**María Elena Martínez-Hernández 1,\*, Javier Goicoechea 1,2 and Francisco J. Arregui 1,2**


Received: 15 October 2019; Accepted: 7 November 2019; Published: 10 November 2019

**Abstract:** Mercury is an important contaminant since it is accumulated in the body of living beings, and very small concentrations are very dangerous in the long term. This paper reports the fabrication of a highly sensitive fiber optic sensor using the layer-by-layer nano-assembly technique with gold nanoparticles (AuNPs). The gold nanoparticles were obtained via a water-based synthesis route that use poly acrylic acid (PAA) as stabilizing agent, in the presence of a borane dimethylamine complex (DMAB) as reducing agent, giving PAA-capped AuNPs. The sensing mechanism is based on the alteration of the Localized Surface Plasmon Resonances (LSPR) generated by AuNPs thanks to the strong chemical affinity of metallic mercury towards gold, which lead to amalgam alloys.

**Keywords:** fiber optic sensor; gold nanoparticles; localized surface plasmon resonance; mercury; ppb

#### **1. Introduction**

The presence of mercury in the environment is a real concern nowadays. It is well known that mercury not only causes damage to the environment, but also to human health [1]. It is a highly toxic element known to cause DNA damage, lipid peroxidation, and protein oxidation and deactivation and is also associated with cardiovascular diseases [2]. This has become a priority matter in the European Union (EU) and the United States Environmental Protection Agency (US-EPA) [3], which seek to take actions against diverse harmful agents that attack the environment like solvents, hydrocarbons, pesticides, and heavy ions (mercury among them). Their objective is to improve the determination of human exposure through integrated monitoring of the environment and food [4,5]. In 2006, the International Conference on Chemicals Management adopted the "Dubai Declaration on International Chemicals Management", the "Overarching Policy Strategy", and endorsed the "Global Plan of Action", in which priority attention is given to mercury [6].

Those international institutions have regulated that any water source and aquifer as well as some specific food products should be monitored and controlled in order to guarantee that they have admissible levels of a series of dangerous contaminants [7], but this task cannot be done nowadays because these tests would require unaffordable costs and complexity. That is why there are many research works that are focused on finding simpler, better, and more accurate ways to detect mercury, where the biggest challenge is to obtain quick, cost-effective and accessible results.

To solve this problem, new methods and perspectives with the use of different sensor devices have been reported. Classical approaches include the monitoring of electrochemical reactions, using techniques like galvano-static techniques, impedance measurement, electrochemiluminescence, and others [1–3]. However, most suffer some reproducibility and stability problems [4].

Among the different sensing materials, gold nanoparticles are one of the most interesting materials for optical sensing applications [5], because of their stability, compatibility with the aqueous medium, easy surface functionalization along with miniaturization [6], and their optical properties. When gold nanoparticles interact with light, there is a resonant light-matter energy coupling known as Localized Surface Plasmon Resonances (LSPR) which can be used as a sensing signal [7–10]. The LSPR occurs thanks to the energy transfer from incident light to certain collective oscillation modes of the free electrons within the nanoparticles that creates an intense optical absorption band. The location of this resonant peak in the visible or infrared region depends on multiple factors such as shape, size, aggregation state, distribution or interaction of the nanoparticles [9]. The consequences of exciting the LSPR are the selective absorption of certain excitation wavelengths and the generation of locally enhanced or amplified electromagnetic fields (EM) on the surface of the nanoparticles and their resonant condition is very sensitive to refractive index variations of the close surrounding medium and the surface chemistry of the nanoparticles [11]. Some studies base their sensing mechanism on the variation of the optical absorbance intensity of the LSPR bands [12–14] of the gold nanoparticles simply due to their surface interaction with mercury ions. Sensors of this type can have low detection limits [15,16]. Furthermore some of the reported works require the use of additional measuring techniques such as ellipsometry [17], SPR reflectometry [18] or involving biological reactions, such as aptamer-based recognition [15], allowing highly sensitivities (LODs around), but increasing the sensors complexity and their cost. Those approaches suppose a limitation for the practical use of such sensors, and the development of more robust, simple and effective sensitive coatings is still a challenge nowadays.

Fiber optic sensors can be a simpler and powerful alternative to these nanoparticles dispersions analysis because have small size, electromagnetic immunity, electrically passivity, and biocompatibility [19]. One of the most common approaches to create optical fiber sensors is the immobilization of the sensitive material onto the surface of the optical fiber. In this manner, the guided light is altered by the interaction with the sensitive material whose optical properties can be affected by the presence of the target to be measured. Therefore, the photonic signal traveling through the fiber will be also modified, which constitutes one of the most common transduction principles of optical fiber sensors for chemical measurements [20]. So far, gold nanoparticles are the most popular solution for the development of highly-sensitive mercury fiber optic sensors, thanks to their stability, small size, low cost, and outstanding optical properties. In most of the approaches, additional molecules or biomolecules are needed to cause this LSPR variation, such as the tendency of mercury to form complexes with certain proteins [13] or the use IgG–anti IgG as bioreceptor–analyte pair [14]. Other sensors study the change of LSPR resonance wavelength. For example, it has been reported the plasmon-coupling effect in gold nanoparticles core-satellites nanostructures linked by thymine(T)-rich DNA hybridization [21]. It is known that the shape and distribution of gold nanoparticles can generate changes in the LSPR, causing wavelength shifts [11]. The process of Au–Hg alloy is able to modify the shape of gold nanoparticles causing changes remarkable blue shifts. Such changes occurred because of the chemical modification of the nanoparticles near their surface (Hg–Au amalgam formation) modify their effective size and shape [22] altering the LSPR resonant condition.

In this work, it is proposed the embedding of gold nanoparticles in a polymeric matrix that allow the interaction with mercury ions (Hg2+). The sensing mechanism is based on the strong chemical affinity of metallic mercury (Hg0) towards gold [18] to form stable amalgam-like alloys [17], and consequently altering the LSPR resonance of the gold nanoparticles, therefore, providing a wavelength-based sensing signal. It has been already reported that the reaction of metallic mercury on the surface of the gold nanoparticles can cause the change of their shape, affecting to the LSPR resonance conditions [22].

The layer-by-layer nano-assembly technique is used here for such embedding of the metallic nanoparticles in the matrix that can facilitate the gold-mercury interaction. This sensing mechanism is simpler than the previous approaches reported in the literature and does not involve the utilization of auxiliary biomolecules with the gold nanoparticles.

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

#### *2.1. Materials*

The polymer poly (allylamine hydrochloride) (PAH) (Mw~15.000) was used as polycation during the LbL process. For the synthesis of AuNPs it was used poly (acrylic acid) (PAA) 35 wt% solution in water, Borane dimethylamine complex (DMAB) and Gold (III) chloride trihydrate. The pH of the solutions were adjusted using HCl and NaOH. The mercury samples were prepared with Mercury (II) chloride (HgCl2) in buffer phosphate. For the buffer solutions it was used sodium phosphate dibasic (Na2HPO4) and sodium phosphate monobasic (NaH2PO4). Piranha solution was also used, which is the combination of sulfuric acid (H2SO4) with hydrogen peroxide (H2O2), 3:1 ratio. All materials were supplied by Sigma Aldrich and aqueous solutions were prepared using ultrapure water with a resistivity of 18.2 MΩ·cm.

#### *2.2. Synthesis Method of the PAA-Capped AuNPs*

There are other works that describe different synthesis routes for metallic nanoparticles of various morphologies [23–26]. In this case, to AuNPs synthesis it was used a chemical reduction route carried out in water-based solutions in which the PAA act as a stabilizer [23]. Gold nanoparticles have been prepared by adding 20 mL of HAuCl4·3H2O (5 mM) to 120 mL of PAA (10 mM). This solution was stirred for 2 h. Afterwards 1 mL of fresh DMAB (0.1 M) solution was added under vigorous stirring, and the reaction was left overnight. All operations were performed at room conditions. UV-VIS absorption spectra of the synthesized nanoparticles dispersions were characterized using a Jasco V-630 spectrophotometer. The UV-VIS absorption spectra of the PAA-AuNP dispersions showed a LSPR absorption band centered at 540 nm. Transmission electron microscopy (TEM) has been used to determine the morphology of the AuNPs, resulting in spherical shape particles, with a diameter ranging from 10 to 20 nm [23].

#### *2.3. Optical Detection Setup*

Optical fiber sensors were made from multimode optical fibers 200 μm-core diameter with polymeric cladding, 0.39 NA (THORLABS FT 200EMT). The sensor structure was based on the mechanical removal of the acrylate cladding of a segment of approximately 2 cm of the optical fiber. This removal was performed with the help of a few drops of dry acetone and a blade, exposing the bare optical fiber core, in its entire cylindrical section. Subsequently, this optical fiber segment was immersed for 5 min in piranha solution to eliminate the acetone that could remain. The ends of the optical fiber were terminated using temporary SMA connectors (THORLABS BFT1). The sensor was excited from one of the connectors with a halogen white source and the other end collect the optical response with a CCD spectrometer (HR4000-UV Ocean Optics).

#### *2.4. Layer-By-Layer Nano-Assembly*

Using layer-by-layer nano-assembly (LbL) it is possible the deposition of oppositely charged polyelectrolyte ultra-thin layers by dipping the substrates into a sequence of solutions. A solution of PAH (10 mM) was used as polycationic solution, and PAA-capped AuNPs (PAA-AuNPs) dispersion was used as polyanion. The optical fiber substrates were immersed into each charged solution for 5 min. After every polyelectrolyte adsorption step it is necessary to rinse the assembly in ultrapure water with same pH of the polyelectrolytes [24,25]. Each polycation/polyanion layer combination is called bilayer. In this work, a total of six bilayers of (PAH/PAA+AuNPs) are deposited onto the cladding removed optical fiber segment (Figure 1). All solutions were adjusted to pH 7. Before starting the deposition of layer by layer, the entire fiber segment where the cladding was removed was immersed in KOH (1M) for half an hour to achieve substrate surface electrostatic charge

**Figure 1.** Layer-by-layer nano-assembly built-up of the sensitive coatings. Construction of fiber optic sensor with (PAH/PAA+AuNPs)6 over a cladding-removed 200 μm-core optical fiber.

#### *2.5. Mercury Samples*

Phosphate buffer (PB) was prepared dissolving 2.198 g of Na2HPO4 in 400 mL of ultrapure water. This solution was stirred for 15 min, then 0.62 g of NaH2PO4 was added and 100 mL of ultrapure water, and stirred for 15 min, obtaining a pH = 7.6. The different concentrations of Mercury (II) chloride (HgCl2) were dissolved in the PB. Different concentration mercury samples must be in metallic form to interact with the AuNPs, consequently, before exposing the optical fiber sensor to the mercury ions, it is necessary to reduce Hg2<sup>+</sup> to Hg<sup>0</sup> using DMAB (12 mL of a freshly prepared DMAB stock solution (0.1 M) as reducing agent. The reaction was stirred at room temperature conditions for 2 h (kept away from direct sunlight). The Hg concentrations analyzed were 1, 2, 4, 8, 10, and 20 ppb. All samples were kept under stirring until the moment of measurement.

In order to vary only the mercury concentration and keep the rest of parameters constant, the DMAB amount is corrected for every sample just adding a certain amount of blank PB stock solution. For every measurement, the optical fiber sensor was immersed into the Buffer PB + DMAB solution prior to the exposure to the mercury ion stock solutions, in order to get a stable baseline for the latter mercury detection.

#### *2.6. Sensors Regeneration*

It is known that nitric acid forms highly instable complexes with Hg2<sup>+</sup> and favors the separation of mercury from gold nanoparticles [26]. The regenerating solution was prepared starting from a stock PB (pH 7.6) and HNO3 was added dropwise until the pH was lowered to 4.6 and the dissolution was keep at a constant temperature of 55 ◦C.

#### *2.7. Data Processing*

During the immersion in the mercury solutions all spectra were recorded continuously and the LSPR maxima were estimated using a Matlab® algorithm. This provides live information about the time response of the sensors. The results obtained will be estimated by their dynamic response as a way to obtain parameters for rapid estimation before the responses obtained from the sensor.

#### *2.8. Cross-Sensitivity to Other Metals*

There are other metals whose presence in the organism is necessary because they are involved in biological functions, however, when they exceed a certain threshold they can be considered toxic, among them we can find zinc and nickel [27]. Consequently it is very important to characterize the mercury sensor cross-sensitivity against other metal ions such as Fe2<sup>+</sup>, Ni2<sup>+</sup>, Pb2<sup>+</sup>, Cd2<sup>+</sup>, and Zn2+. All the solutions were prepared under the same conditions as the mercury samples. All ionic species for the cross-sensitivity test has been set to the maximum concentration used with the mercury (20 ppb). This concentration of the other ionic species are significantly higher than the limits required by the regulation [28] like for Fe (6.2 ppb), Cd (3.4 ppb) and Zn (1.8 ppb).

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

#### *3.1. E*ff*ects of Hg0 on AuNPs in Dispersion*

Prior to the construction of the fiber optic sensor, a preliminary study was made in order to characterize the effects of mercury on the optical properties of the gold nanoparticles in dispersion. It is known the Hg<sup>0</sup> can be bonded onto the surface of Au-based nanomaterials to form a solid amalgam-like alloy [29,30].

Samples in dispersion were analyzed with different Hg concentration and keeping constant the volume and concentration of AuNPs solutions. US-VIS spectra of the dispersions showed a dramatic change of LSPR resonance wavelength clearly seen with the naked eye as a color change (Figure 2).

**Figure 2.** Change of Localized Surface Plasmon Resonances (LSPR) wavelength of PAA-AuNP dispersion with different Hg concentration. Cuvette 1: as prepared only with AuNPs. Cuvette 2: PAA-AuNP + DMAB. Cuvette 3: PAA-AuNP + DMAB and mercury (2.7 ppb). Cuvette 4: PAA-AuNP + DMAB and mercury (27 ppb).

In Figure 2 it is shown the UV-VIS spectrum of cuvette 1 that contain only a PAA-AuNPs dispersion as prepared that shows an violet-reddish color. The spectrum shows the typical LSPR attenuation band centered in 540 nm which is compatible with the synthesis routes available in the literature. Since it is necessary to reduce mercury ions to their metallic form (Hg0) to enable the amalgam interaction, cuvette 2 is equal to cuvette 1, except that 200 μL of 0.1 M DMAB were added. Here it is observed a slight blue shift of the LSPR resonance wavelength of 8 nm, that remained stable in time. This LSPR variation is probably due to the modification of the polymeric PAA stabilization cap thanks to the interaction with the DMAB. For lower mercury concentrations such as 2.7 ppb (cuvette 3), the displacement also occurs, but to a lesser extent, in 18 nm respect to the cuvette 2 that is the optical reference with no mercury. When 300 μL of Hg (10−<sup>3</sup> M) were added to the dispersion keeping the same concentration and volume of PAA-AuNPs and DMAB solution, it was obtained 27 ppb of mercury concentration and the LSPR resonance experimented a stronger blue shift, almost disappearing, yielding a clear yellowish color.

In the synthetic process of AuNPs, the reduction of gold ions (*Au*3+) to gold nanoparticles (*Au*0) is possible thanks to the use of a protective agent (PAA), which contributes to control the shape and size of the resultant nanoparticles, preventing their agglomeration or precipitation in the colloidal solution and the DMAB that acts as a reducing agent [9]. The small displacement of LSPR resonance wavelength that occurred in the case of cuvette 1 as a result of interaction between AuNP-PAA with the additional DMAB present in the sample solutions (with no mercury in cuvette 1). This LSPR

wavelength shift could be induced by the refractive index variation in the optical fiber immersion media. However, cuvettes 2, 3, and 4 have the same concentrations of AuNP-PAA and DMAB, they only differ in a very small mercury concentration that induces a more severe LSPR resonance displacement thanks to the chemical modification of the AuNPs.

#### *3.2. Obtaining the AuNPs LSPR onto the Fiber Optics*

The absorbance of the (PAH/PAA+AuNPs)n coating onto the optical fiber was registered during the LbL process. With every bilayer increment the absorbance spectrum show an increasing of the intensity around the 540 nm wavelength suggesting a homogeneous growth of the LbL coating. After an optimization study of the LbL process it was found that using six bilayers it is possible to obtain a well-defined LSPR absorption band (Figure 3) consequently this number of bilayers was kept constant for all the sensors in this work.

**Figure 3.** Absorption spectra of LSPR resonance wavelength for every layer of (PAH/PAA+AuNPs) deposited on 200 μm-core optical fiber.

The absorbance of the LbL (PAH/PAA+AuNPs)n films (being n the number of bilayers) is shown in Figure 3. The absorbance spectra confirm the existence of an absorption band centered at 540 nm, which corresponds to the LSPR of the AuNPs. This demonstrates that the absorption band of the coatings matches with that of the AuNPs dispersion initially synthesized by chemical reduction seen in Figure 2 (cuvette 1).

#### *3.3. Detection of Mercury Ions with Fiber Optic Sensor*

In Figure 4, it is shown an initial immersion of the sensor in the Buffer PB + DMAB solution, it was registered a small displacement of LSPR resonance wavelength (approximately 5 nm) and after a few minutes it remained stable. In this work, all sensors were kept in this solution for one hour in order to have a stable baseline for the later mercury detection stage. Nevertheless, shorter immersion times could be also acceptable. After the sensor it was immersed in a 20 ppb mercury sample and there was a variation of absorption with respect to the condition of the baseline. In addition, for the mercury concentration of 20 ppb there is a change in the LSPR resonance wavelength of 15 nm with respect to the Buffer PB + DMAB solution.

**Figure 4.** Displacement in wavelength of the LSPR for 20 ppb of mercury concentration.

Different sensors were fabricated with the same materials and methods mentioned in Section 2, and each one was used to detect a particular mercury concentration (shown in Figure 5). Although the wavelength shift of the LSPR band was easily visible in a few seconds for the highest mercury concentration, the (PAH/PAA+AuNPs)6 sensors showed a settling time of nearly 3000 s (from 10% to 90%). Consequently, all the sensors were immersed in the mercury solution for 50 min. All sensors' LSPR bands experimented a blue-shift when exposed to mercury. The absolute wavelength shift increases with the mercury concentration; for 20 ppb of mercury solution, the LSPR maximum wavelength change is 16 nm, and for 1 ppb is 1.11 nm.

**Figure 5.** Dynamic response of the sensors to different Hg concentration, ranging from 20 ppb to 1 ppb.

This wavelength-based response can be seen in Figure 6, where it is shown the maximum variation in wavelength |Δλmax| for each mercury concentration and the linear fitting.

**Figure 6.** Maximum variation in wavelength (Δλmax) for each mercury concentration (1, 2, 4, 8, 10, and 20 ppb) after 3000 s.

From the continuous data acquisition of the baseline during the immersion in Buffer PB + DMAB (0 ppb), it is possible to calculate the standard deviation (σ). The limit of detection (LOD) of the sensor can be estimated as 3σ, that is 0.147 nm, which is equivalent to 0.7 ppb, which is below the 2 ppb detection limit stated by the US-EPA and 1 ppb for the EU [29,30].

The results presented so far are accurate enough to provide reliable measurements of aqueous samples without any further chemical or biological agent, and they could be performed in the field. Nevertheless, the sensors still need relatively long time measurements. In order to overcome this, a measurement technique is proposed to obtain faster measurements. In this sense, the values of the slopes of each dynamic curve (Figure 5) could be used as a fast estimation parameter. In Figure 7, the slope of the sensor's response approximated by the linear fitting of the first 500 s (roughly 8 min) is plotted for every Hg concentration.

**Figure 7.** Fast estimation of the optical fiber sensor response using the slope of the linear fitting of the samples of the first 500 s. The Hg concentration has been varied from 1 to 20 ppb.

As can be observed, the value of the slope increases with the increase in mercury concentration, getting lineal response (Figure 8a). In Figure 8b it is shown the high correlation between the absolute wavelength shift and the slope of each curve, meaning that it can be reliably used as a fast response estimator. These results allow estimating the behavior of the sensor in different mercury concentrations in a faster way, after 500 s.

**Figure 8.** (**a**) Slopes of the linear fitting of the first 500 s vs. mercury concentration. (**b**) Correlation between the mercury measurement using the absolute wavelength shift of the LSPR, and the fast slope estimation.

#### *3.4. Sensor Regeneration*

Another critical aspect is the reusability of a single optical fiber sensor for multiple measurements. In fact, it is possible to regenerate the sensor in a HNO3 solution, the sensor was deposited in the solution mentioned in Section 2.6 for 1 h. During the immersion in the regenerating solution it was observed a red shift of the LSPR resonance wavelength, which is a similar to the first reaction curve. After regeneration the sensor was submerged again in a second Hg dissolution of 20 ppb. As can be seen in Figure 9, the comparison of two measurements of the same optical fiber sensor against two different samples of mercury (20 ppb) is represented. The first measurement corresponds to the freshly fabricated sensor that was deposited in a first mercury sample (20 ppb). After reacting to mercury was deposited in the regenerative solution that allowed the sensor to recover the initial conditions, so it was deposited in a new sample of mercury (20 ppb), thus obtaining the second measurement, yielding a very similar wavelength shift as in the first measurement.

**Figure 9.** Regeneration of the optical fiber sensor. Repeatability of the sensor under 20 ppb of mercury after its regeneration in a dilute nitric acid solution. The final wavelength-shift is very similar in both cases, and the response slope is even more stable.

#### *3.5. Cross Sensitivity*

Finally, the selectivity of optical fiber sensor against different heavy metal ions (Fe2<sup>+</sup>, Ni2<sup>+</sup>, Pb2<sup>+</sup>, Cd2+, and Zn2+) is also studied the same sensor has been exposed to the same concentration (20 ppb) of the different metal ions, and all solutions were prepared using the same protocol as in the previous

mercury tests (PB + DMAB @pH 7.6). The results of the final wavelength shift after the immersion in the different ion solutions are showed in Figure 10.

**Figure 10.** Selectivity analysis against the most common heavy ions. The measurements were carried out using the same sampling preparation process, and with the same concentration (20 ppb).

As it can be seen in Figure 10, the proposed optical fiber sensor showed a significantly higher response to the presence of mercury compared to the other metal ions, nearly 400% higher than the second more reactive cross-contaminant (in this case iron) enjoying a high selectivity towards mercury.

The metal ions analyzed have detection limits allowed in water higher than of mercury [28], for example Ni (15 ppb), Fe (6.2 ppb) among others. Therefore, in normal samples, our sensor would be more selective to mercury than to other metal ions. A further study would require the evaluation of the present devices in real aquifer water samples, but this is out of the scope of the present contribution. This work presents a competitive approach for a mercury optical fiber sensor, with a simple and direct measurement of mercury in water.

#### **4. Conclusions**

In this work, a simple and highly sensitive mercury optical fiber sensor has been proposed. Its sensing mechanism is based on the wavelength shift of the AuNPs LSPR, thanks to the strong chemical affinity of mercury towards gold. The gold nanoparticles were obtained by synthesis method of the PAA-capped AuNPs, using PAA as a stabilizing agent and DMAB as reducing agent. The LBL nano-assembly technique was used for the incorporation of gold nanoparticles onto optical fiber in a stable sensitive thin-film, (PAH/PAA+AuNPs)6. The absolute wavelength-shift was a reliable and robust signal with a relatively long response time of around one hour. In order to obtain faster measurements, the slope of the wavelength variation proved to give reliable results in only 8 min. It is possible to reuse the sensor, something that reduces costs and manufacturing time. In addition, this sensor showed low cross sensitivity towards other metal ions. It was achieved a limit of detection of 0.7 ppb, which is lower than the standard limits recommended by the European Union (1 ppb) and US-EPA (2 ppb). The sensor proposed in this work could be competitive alternative for mercury detection, a problem of global concern.

**Author Contributions:** Conceptualization and Methodology, M.E.M.-H., J.G. and F.J.A.; Investigation and Validation, M.E.M.-H.; Writing-Original Draft Preparation, M.E.M.-H. and J.G.; Writing-Review & Editing, F.J.A.; Supervision, J.G. and F.J.A.; Project Administration and Funding Acquisition, F.J.A.

**Funding:** This work was supported in part by the Spanish Agencia Estatal de Investigación (AEI) and the European Regional Development Fund (ERDF) under the Project TEC2016-79367-C2-2-R, and Public University of Navarra pre-doctoral research grants.

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

#### **References**


© 2019 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 (http://creativecommons.org/licenses/by/4.0/).

### *Communication* **Analysis of Trace Metals in Human Hair by Laser-Induced Breakdown Spectroscopy with a Compact Microchip Laser**

**Makoto Nakagawa <sup>1</sup> and Yuji Matsuura 1,2,\***


**Abstract:** A laser-induced breakdown spectroscopy (LIBS) system using a microchip laser for plasma generation is proposed for in-situ analysis of trace minerals in human hair. The LIBS system is more compact and less expensive than conventional LIBS systems, which use flashlamp-excited Q-switched Nd:YAG lasers. Focusing optics were optimized using a Galilean beam expander to compensate for the low emitted pulse energy of the microchip laser. Additionally, hundreds of generated LIBS spectra were accumulated to improve the signal-to-noise ratio of the measurement system, and argon gas was injected at the irradiation point to enhance plasma intensity. LIBS spectra of human hair in the UV to near IR regions were investigated. Relative mass concentrations of Ca, Mg, and Zn were analyzed in hairs obtained from five subjects using the intensity of C as a reference. The results coincide well with those measured via inductively coupled argon plasma mass spectrometry. The lowest detectable concentrations of the measured LIBS spectra were 9.0 ppm for Mg, 27 ppm for Zn, and 710 ppm for Ca. From these results, we find that the proposed LIBS system based on a microchip laser is feasible for the analysis of trace minerals in human hair.

**Keywords:** laser-induced plasma spectroscopy; microchip laser; hair analysis

#### **1. Introduction**

Laser-induced breakdown spectroscopy (LIBS) is a technique that measures emission spectra from luminous plasma generated by irradiation with nano-, pico-, and femto-second laser pulses and is useful for multi-elemental analysis of various target materials [1,2]. A microchip laser [3–6] that emits a pulse energy of hundreds of microjoules has become popular because it makes the LIBS system more compact and lower in cost than conventional systems, which use flashlamp-excited Q-switched Nd:YAG lasers [7,8]. Portable LIBS systems with microchip lasers and compact fiber-coupled spectrometers have been developed [9]. These compact systems have been used for quantitative analysis of steel composition [10,11] and aluminum alloys [12,13].

LIBS techniques are useful for qualitative and quantitative analysis of biological samples and have been applied in the diagnosis of some diseases, such as cancer [14]. In biomedical applications, one of the advantages of the LIBS technique is that pretreatment of samples is not required, unlike in other elemental analysis methods such as inductively coupled argon plasma-atomic emission spectroscopy (ICP-AES) or mass spectrometry (ICP-MS) [15]. For healthcare applications, such as nutritional status monitoring, analysis of easily harvested biological specimens, such as nails and hair, is useful. ICP-AES and ICP-MS have already been applied to the analysis of a variety of biological samples, including nails and hair [16]. However, as mentioned above, ICP-AES and ICP-MS need relatively complicated pretreatment processes and, therefore, real-time analysis is difficult. Additionally, the large-scale and high-cost equipment that is necessary for those analysis techniques is not cost-effective for most healthcare applications. Therefore, many groups have proposed LIBS techniques for the analysis of biological samples. LIBS spectra

**Citation:** Nakagawa, M.; Matsuura, Y. Analysis of Trace Metals in Human Hair by Laser-Induced Breakdown Spectroscopy with a Compact Microchip Laser. *Sensors* **2021**, *21*, 3752. https://doi.org/10.3390/ s21113752

Academic Editor: Gabriela Kuncová

Received: 13 April 2021 Accepted: 25 May 2021 Published: 28 May 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/).

of fingernails have been measured for the diagnosis of diseases, such as diabetes. As nails are relatively hard tissues with high mechanical and chemical strength, stable LIBS measurements are possible [17–20].

Hair is another target tissue for LIBS measurements because the concentrations of trace elements in hair are generally higher than that in other biological tissues. Haruna detected Ca in human hair using LIBS and qualitatively analyzed Ca variation with age and sex [21]. Corsi et al. quantitatively analyzed mineral content (Mg, K, Ca, Na, and Al) in human hair and compared their results with those obtained through a commercial analytical laboratory [22]. Emara measured trace elements in horsehair and compared the results with those obtained via atomic absorption spectroscopy [23]. More recently, Zhang combined LIBS with ultrasound-assisted alkali dissolution to analyze Zn and Cu more accurately in human hair [24]. In these LIBS applications, conventional Q-switched Nd: YAG lasers are used. This is because conventional Q-switched Nd: YAG lasers can obtain relatively high plasma intensity by irradiation with pulse energy of more than 10 mJ, which is necessary to detect trace elements in hair. The use of these lasers makes the LIBS system large in scale and not easy to handle. In addition, irradiation with high energy pulses easily induces severe damage to hair and sometimes the hair is torn off, which can cause changes in the detected signal while obtaining the LIBS spectra.

In this paper, we propose LIBS analysis of human hair while using a microchip laser for plasma generation. To obtain plasma intensity that is sufficiently high for trace element analysis, we optimally designed the focusing optics. We then accumulate hundreds of LIBS spectra generated by the microchip laser emitting at high repetition rate pulses while Argon gas is injected at the irradiation point to enhance plasma intensity. Table 1 presents the typical concentrations of trace elements contained in the black hair of Japanese adults [25]. Among these elements, we analyzed the relative mass concentrations of Ca, Mg, and Zn of hairs obtained from five subjects while using the intensity of C as a reference. The concentration results coincide well with those measured via ICP-MS.



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

In our experiment, a passively Q-switched microchip laser (L11038–11, Hamamatsu Photonics, Hamamatsu, Japan) emitting optical pulses with an energy of 2 mJ and pulse duration of 1 ns at a wavelength of 1064 nm was used as the light source. The measurement setup is shown in Figure 1. The experiment was conducted under atmospheric pressure. The plasma generated from the sample was delivered to the spectrometer by a step-index silica-glass fiber with a core diameter of 600 μm and NA of 0.22. The distal end of the fiber was located at a distance of 2 mm from the irradiation point. The spectra were measured using a fiber-coupled spectrometer (HR2000+, Ocean Insight, Orlando, FL, USA) which was synchronized with the laser pulses using an external pulse generator. The integration time of the spectrometer was set to 1 ms, which was the shortest possible setting, and was not gated. Therefore, the plasma emission generated by a nanosecond optical pulse was completely detected. Three types of spectrometers were used to obtain LIBS spectra from the ultraviolet (UV) to visible (Vis) regions. With the combined use of three spectrometers, the wavelength ranges of 200–343 nm, 355–474 nm, and 480–597 nm were covered with a wavelength resolution of 0.14 nm.

**Figure 1.** Measurement setup.

Dozens of human hairs were bundled and fixed on a metal substrate to create samples. The substrate had a hole with a diameter of 12 mm at the center to avoid generating the plasma from the underlying substrate. The hairs were washed twice with pure water after degreasing with acetone. When irradiating with optical pulses, the sample was moved slowly along the longitudinal direction at a speed of approximately 2 cm/min to avoid severely damaging the hair sample. Our protocol was approved by the ethical committee on the Use of Humans as Experimental Subjects of Tohoku University (No. 20A-29), and informed consent was obtained from the examinees.

We first focused the laser beam, using a spherical BK7 lens with a focal length of 100 mm, to obtain the LIBS spectrum shown in Figure 2. For each measurement, 300 consecutive spectra were accumulated. Because of the limited data acquisition speed of the spectrometer, the repetition rate of the laser was set to 10 Hz, and thus, it took 30 s for a single measurement. To identify elements from the LIBS spectra, we referred to the NIST Database [26] and OSCAR Database [27]. Although we confirmed emission peaks of Ca at Ca I: 422.7 and 445.5 nm, Ca II: 393.3 and 396.8 nm, and of the C–N bond at 388.29 nm, we could not detect peaks of another material because of the low signal-to-noise ratio (SNR) of the LIBS system.

**Figure 2.** Laser-induced breakdown spectroscopy (LIBS) spectrum of hair measured by a focusing system with a single lens.

To improve the SNR, we designed focusing optics to increase the optical power density at the focusing spot. To obtain a more focused beam spot, we built a Galilean

beam expander with 5× magnification. The focused beam size was reduced to 0.12 mm from 0.37 mm by introducing the expander. The power density at the focal spot was approximately 4.5 × 109 W/cm2. Figure <sup>3</sup> shows a LIBS spectrum of hair measured with and without the beam expander. We found that the peak intensity was enhanced by approximately 10 times.

**Figure 3.** LIBS spectra of hair measured with and without an expander.

To further increase the sensitivity, we injected argon gas onto the focusing spot. It was reported that the optical emission intensity of plasma induced by laser irradiation is enhanced in the argon atmosphere [28]. One of the reasons for this is that the energy decay of free electrons in plasma is suppressed in an argon atmosphere. In addition, high plasma temperature is maintained in argon because of its lower thermal conductivity. Figure 4 shows LIBS spectra of human hair measured with and without argon injection. Argon gas was injected from a nozzle set at a distance of 2 mm from the focused spot with a flow rate of 1 L/min. It was found that the peak intensity was enhanced approximately 1.8 times using argon injection; small peaks that can be attributed to C–N bond were observed at around 358 nm.

**Figure 4.** LIBS spectra of hair measured with and without argon gas injection.

To find the optimum number of spectrum accumulation in a LIBS measurement, we changed the accumulation number from 1 to 900 and measured the LIBS spectra of human hair. Figure 5 shows the correlation between the SNR of the obtained spectra and the accumulation number. The SNR was defined as (Peak area of Ca II at 393.3 nm)/[(standard deviation of background signal) × (full width of the Ca II peak at half maximum)]. We found that the SNR was almost saturated at an accumulation number of 300, and therefore, we set the accumulation number to 300 in consideration of the measurement time.

**Figure 5.** Correlation between signal-to-noise ratio (SNR) of obtained spectra and accumulation number.

#### **3. Results and Discussions**

Figure 6 shows a LIBS spectrum of human hair measured from 480 to 597 nm. In this wavelength region, emission peaks for the C–C bond at 512.8 and 516.4 nm, Ba I at 553.5 nm, and Na I at 589.6 nm are observed. Figure 7 shows a spectrum measured at 200–340 nm. In the measurement results for the UV region, we found that there was less background noise from calcium emission; thus, we changed the focal length of the lens from 100 mm to 50 mm to further increase the energy density at the focused spot. The focal spot size was reduced to 0.09 mm from 0.12 mm. In this region, we observed peaks of C I (247.8 nm), Mg II (279.5 nm), and Ca II (315.8 nm). Figure 8 shows an enlarged spectrum at around 330 nm. Although the peak intensities are relatively low, we found that the peaks coincide with that of Zn I at 328.26, 330.26, and 334.51 nm. Since it was found that the peak at 330.26 nm was affected by the peak of Na I at 330.3 nm, we hereafter use the peak at 328.26 nm for analysis of Zn I, which is one of the trace minerals essential for human life.

**Figure 6.** LIBS spectrum of hair measured in the visible wavelength region.

**Figure 7.** LIBS spectrum of hair measured in the ultraviolet wavelength region.

**Figure 8.** An enlarged LIBS spectrum of hair measured at around 330 nm.

We performed LIBS measurements to observe individual differences in the relative concentrations of the trace minerals. We collected hair samples from five volunteers, aged 23–25, and compared the results of LIBS analysis to those obtained via ICP-MS analysis. LIBS measurements were used to observe individual differences in the relative concentrations of these trace minerals. As the concentration of trace elements in hair depends on the position along the hair length, we analyzed the hair around 2.5 cm from the root for all the measurements. We utilized the commercial service of the Kyorin Preventive Medicine Institute [29] for ICP-MS analysis of hair samples that were taken from the same subjects at the same time. For ICP-MS analysis, the same part of hair as described above was used for comparisons.

As the intensity of observed peaks varies widely because of the small diameter of the hair samples, the relative concentrations obtained via LIBS analysis were calculated by setting the peak intensity of C I as the reference because C is a primary composition component of hair; as such, the individual difference should be relatively small. Table 2 shows the observed peak intensities of Mg II (279.5 nm), Zn I (328.26), and Ca II (315.8 nm) measured for samples taken from the five subjects. The coefficients of variation (CV) were calculated from the results of three measurements for each subject, and we confirmed that the variations reduced considerably upon using the peak intensity of C as a reference. We

also tried to normalize the peak intensities by the total emission intensity of measured LIBS spectra. However, we did not obtain better results because of the relatively large background noises in the measurement. The calculated total emission intensity largely changed with baseline correction processing.

**Table 2.** Effect of normalization based on the peak intensity of C I at 247.8 nm. In the table, "X" and "X/C" show the coefficients of variations (CV) before and after normalization, respectively.


Figure 9 shows the relative mass concentrations of Mg, Zn, and Ca measured for the five subjects compared with the absolute concentrations analyzed via ICP-MS. In the LIBS results in Figure 9, the dots are the average values of three measurements, and the error bars show the minimum and the maximum measured values. The measurement variability was sufficiently small to see individual differences and the trends between the subjects coincided with the results of ICP-MS analysis.

We confirmed good linearity between the results of the LIBS and ICP-MS methods, and the determination coefficient R<sup>2</sup> was 0.921 for Mg, 0.670 for Zn, and 0.952 for Ca. We observed relatively small correlation for Zn; this may be because of the small peak intensity compared to the ones of Mg and Ca. The lowest detectable concentrations, defined by SNR = 3, were 9.0 ppm for Mg, 27 ppm for Zn, and 710 ppm for Ca. Since these values are lower than typical concentrations of these trace minerals, we confirmed the feasibility of the proposed LIBS using a microchip laser for analysis of relative mass concentrations in human hair. In the above analysis, we did not consider the variability of plasma temperature, electron density, and upper energy levels of the observed transitions [30]. For more accurate analysis of trace elements in human hair, corrections based on these factors may be necessary.

**Figure 9.** *Cont*.

**Figure 9.** Relative mass concentrations of (**a**) Mg, (**b**) Zn, and (**c**) Ca measured for the five subjects and compared with the absolute concentrations analyzed via ICP-MS.

#### **4. Conclusions**

As a compact and low-cost LIBS system for in-situ analysis of trace minerals in human hair, we proposed a system using a microchip laser for plasma generation. Since the pulse energy emitted from a microchip laser is lower than that of conventional flashlamp-excited Q-switched Nd:YAG lasers, we optimally designed the focusing optics utilizing a Galilean beam expander to obtain plasma intensity sufficiently high for analysis of trace elements in hair. Additionally, we accumulated hundreds of LIBS spectra to improve the SNR of the measurement system and injected argon gas on the irradiation point to enhance the plasma intensity. After investigating LIBS spectra of human hair in the UV to near IR regions, we focused on the spectra in the UV region because of the location of emission peaks of Mg and Zn, which are trace minerals essential for human life.

We analyzed relative mass concentrations of Ca, Mg, and Zn in hairs obtained from five subjects while using the intensity of the C peak as a reference. The results coincided well with those measured via ICP-MS. We estimated the lowest detectable concentrations from the SNR of the measured LIBS spectra: 9.0 ppm for Mg, 27 ppm for Zn, and 710 ppm for Ca. From these results, we have concluded that the proposed LIBS system based on a microchip laser is feasible for the analysis of trace minerals in human hair.

Owing to the low-cost and compact proposed system, we expect that it will be a useful biomedical sensor for health-care applications based on non-invasive and real time analysis of hair.

**Author Contributions:** M.N. performed the experiments and analyzed the data; Y.M. conceived and designed the experiments. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by JSPS KAKENHI Grant Numbers JP 20H00231 and JP 20K12615.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the he Ethical Committee on the Use of Humans as Experimental Subjects of Tohoku University (20A-29, 9 October 2020).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy restrictions.

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

#### **References**

