**Use of a Mixed Cationic-Reverse Phase Column for Analyzing Small Highly Polar Metabolic Markers in Biological Fluids for Multiclass LC-HRMS Method**

#### **Marco Roverso \*, Iole Maria Di Gangi, Gabriella Favaro, Paolo Pastore and Sara Bogialli**

Department of Chemical Sciences, University of Padua, Via Marzolo, 1, 35131 Padova, Italy; iole.digangi@gmail.com (I.M.D.G.); gabriella.favaro@unipd.it (G.F.); paolo.pastore@unipd.it (P.P.); sara.bogialli@unipd.it (S.B.)

**\*** Correspondence: marco.roverso@unipd.it; Tel.: +39-049-8275181

Received: 17 September 2020; Accepted: 12 October 2020; Published: 14 October 2020

**Abstract:** The determination of small highly polar metabolites at low concentrations is challenging when reverse-phase (RP) chromatography is used for multiclass analysis. A mixed cationic-RP column coupled to high-resolution tandem MS (HR-MS/MS) was tested for highly polar compounds in biological fluids, i.e., trimethylamine N-oxide (TMAO) and the isobaric molecules beta-methylamino-L-alanine (BMAA) and 2,4-diaminobutyric acid (DAB). The efficient retention and separation of the above compounds were obtained with common and MS-friendly RP conditions, reaching high selectivity and sensitivity. The method was firstly assessed in plasma and urine, showing good linearity in the range 50–1000 μg/L and 500–10,000 μg/L for TMAO and both BMAA and DAB, respectively. Excellent precision (RDS < 3%) and good accuracies (71–85%) were observed except for BMAA in plasma, whose experimental conditions should be specifically optimized. Preliminary tests performed on compounds with biological relevance and a wider range of polarities proved the effectiveness of this chromatographic solution, allowing the simultaneous analysis of a larger panel of metabolites, from very small and polar compounds, like trimethylamine, to quite lipophilic molecules, such as corticosterone. The proposed LC-HRMS protocol is an excellent alternative to hydrophilic interaction liquid chromatography and ion-pairing RP chromatography, thus providing another friendly analytical tool for metabolomics.

**Keywords:** polar amino acids; mixed cationic-RP column; LC-HRMS; BMAA; TMAO

#### **1. Introduction**

The separation of small strongly polar compounds is challenging for reverse-phase (RP) chromatography. The weak interaction between the hydrophobic stationary phase and analytes usually leads to short retention times, matrix interferences and poor selectivity [1]. Derivatization steps and ion-pairing chromatography are possible solutions. However, derivatization procedures are usually time-consuming, and both reagents and secondary reactions may interfere with the analysis. Ion pairing is another useful trick but may result in low sensitivity with LC-MS. Hydrophilic interaction liquid chromatography (HILIC) is nowadays a quite widespread approach for the analysis of highly polar molecules such as amino acids, saccharides, nucleic acids, and phosphate-containing molecules [2]. In this case, very common drawbacks are peak distortion, large consumption of organic solvents, long equilibration times, and lack of solubility of some polar compounds in organic solvents. Mixed-mode liquid chromatography is an interesting alternative to the RP and HILIC stationary phases. These phases, firstly described in 1984 by Bischoff and McLaughlin [3], can be very effective for the simultaneous analysis of polar and non-polar compounds, due to the combined effects of different interaction mechanisms [4].

Beta-*N*-methylamino-l-alanine (BMAA) is a polar amino acid with a supposed neurotoxic activity. BMAA was described as a metabolite of several cyanobacterial strains, which can be bioaccumulated and biomagnified across the food chain [5]. Chronic exposure to BMAA has been associated with degenerative neurological conditions such as amyotrophic sclerosis, Parkinson's disease and dementia [6]. The risk assessment of BMAA is highly debated in the scientific literature since results are difficult to compare, because of the possible false-positives obtained by using different analytical techniques [7]. The main issue is the presence of several isomers of BMAA, such as 2,4-diaminobutyruc acid (DAB), so that the chromatographic separation and the correct quantification in biological and food samples remain challenging [8,9]. Most of the existing analytical procedures are based on the derivatization of the analytes with 6-aminoquinolonyl-*N*-hydroxysuccinimdyl prior to RP chromatography, or HILIC and ion-pairing RP separation without derivatization [5,8–11]. Solid-phase extraction performed using strong cation exchange sorbents is also reported for improving the sensitivity of the method and for samples clean up [12]. Recent works take advantage of tandem MS to improve the selectivity of the method and avoid false positives [13].

Trimethylamine *N*-oxide (TMAO) is a highly polar metabolite whose production is controlled by gut microbiota and liver enzymes. The production of TMAO is strictly related to the dietary consumption of l-carnitine and lectin rich food. Recently, the alteration of TMAO concentration in plasma was positively associated with an increased risk for cardiovascular diseases, heart failure, obesity, impaired glucose tolerance, diabetes, and colorectal cancer [14,15]. The quantitative analysis of TMAO is usually performed with RP or HILIC chromatography coupled to electrospray (ESI) mass spectrometry [16–18]. BMAA and TMAO, such as other polar metabolites, are potentially useful biomarkers to be monitored in various biological fluids, but the experimental conditions conventionally used for RP mode can inhibit the simultaneous multiclass determination of very different compounds, which is a goal of the metabolomic approach.

In this preliminary research work, a new separation approach, based on the use of a mixed cation-RP stationary phase, is evaluated to obtain the best retention and selectivity in the direct analysis of highly polar compounds such as TMAO, BMAA, and DAB. The analytical method is based on LC coupled to high-resolution tandem mass spectrometry (LC-HR-MS/MS) and applied to various biological matrices, such as plasma and urine, for clinical applications. Furthermore, the present study outlines future applications of this protocol to a wider panel of polar and non-polar analytes in the framework of metabolomic investigations.

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

#### *2.1. Chemicals*

Analytical grade BMAA, DAB, TMAO, deuterated trimethylamine *N*-oxide (D9, 98%) (TMAO-d9), trimethylamine, kynurenic acid, dopamine, homocysteine, carbidopa, picolinic acid, L-DOPA, 3-hydroxykynurenine, corticosterone and formic acid (FA) were purchased from Sigma-Aldrich Italy (Milan, Italy). LC-MS solvent grade acetonitrile was purchased from Carlo Erba Reagents (Milano, Italy). Ultrapure-grade water was produced by a Pure-Lab Option Q apparatus (Elga Lab Water, High Wycombe, UK). Standard stock solutions of the analytes under study were prepared in water/acetonitrile 50:50 at 1000 mg/L and stored at −20 ◦C until use. A mixed working standard solution was obtained by suitable dilutions in mobile phases. TMAO-d9 was used as an internal standard (IS).

#### *2.2. Instrumentation*

The LC-MS/MS system was an Ultimate 3000 UHPLC coupled to a Q-Exactive hybrid quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA).

A Luna HILIC (100 mm × 2.0 mm i.d. × 3 μm, Phenomenex, Bologna, Italy) and an Acclaim Mixed-Mode WCX-1 (150 mm × 2.1 mm i.d. × 3 μm, Thermo Fisher Scientific, Waltham, MA, USA) were used as columns thermostated at 15 ◦C. The elution was performed in a gradient mode (0–6 min 0% B, 6–15 min 0% B to 100% B, 15–20 min 100% B, equilibration time 10 min) at a flow rate of 0.25 mL/min, using water (eluent A for Acclaim, B for HILIC) and acetonitrile (eluent B for Acclaim, A for HILIC) both 10 mM FA. This flow rate was a compromise between the chromatographic selectivity and the ESI-MS responses. The injection volume was 5 μL. The MS conditions were the following: electrospray (ESI) ionization in positive mode, resolution 35,000 in MS and 17,500 in MS/MS (at *m*/*z* 200), AGC target 3 <sup>×</sup> 10<sup>6</sup> and 2 <sup>×</sup> 105 in MS and tandem MS, respectively; max injection time 200 ms, scan range 50–750 Da, isolation window 4.0 *m*/*z*, normalized collision energy 35 in HCD mode. The capillary voltage was 3.5 kV, the capillary temperature was 320 ◦C, auxiliary gas, and sheath gas was nitrogen at 40 and 20 a.u., respectively, while sweep gas was not used. Calibration was performed with Pierce™ ESI Positive Ion Calibration Solution (Thermo Fisher Scientific, Waltham, MA, USA). The MS data were analyzed with the Xcalibur 4.0™ software (Thermo Fisher Scientific, Waltham, MA, USA).

#### *2.3. Sample Preparation and Calibrations*

A five-point external calibration curve was prepared from stock solutions of TMAO, BMAA and DAB. The concentrations of BMAA and DAB in the five solutions were 50, 100, 250, 750 and 1000 μg/L, while TMAO was present at 5, 10, 25, 75 and 100 μg/L. IS was added in all the calibration solutions to a final concentration of 250 μg/L. The peak area ratio (A/AIS) related to the selected fragment ions acquired in tandem MS mode for each compound was plotted against concentrations for calibration purposes. Linearity was assessed using the least-squares regression.

Mouse plasma and urine samples were collected as previously reported [19]. For the matrix-matched calibration, plasma samples were pooled from ten different animals and split into 100 μL aliquots. Five aliquots were spiked with both BMAA and DAB at concentrations of 500, 1000, 2500, 7500 and 10,000 μg/L and TMAO at concentrations of 50, 100, 250, 750 and 1000 μg/L, respectively; one aliquot was treated as blank. The basal concentration of TMAO, which is a metabolite physiologically present in plasma and urine, was preliminarily estimated by the standard addition method. Plasma aliquots were extracted with 400 μL of ice-cold acetonitrile added with 10 mM FA and 625 μg/L of IS, vortexed and centrifuged at 14,000× *g* and 4 ◦C for 10 min. 100 μL of the supernatant was further diluted 1:1 with water 10 mM FA and injected for the analysis. The same protocol was used for urine samples. The selected analytes were analyzed in their free form, i.e., not bound to proteins, which were removed during the sample preparation.

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

The chromatographic separation of BMAA, DAB and TMAO is hindered by poor retention in RP mode, and HILIC conditions have been used to overcome these problems. Figure 1A shows the extracted ion chromatogram (EIC, mass accuracy 10 ppm) of [M+H]<sup>+</sup> precursor ions at *m*/*z* 119.0815, 76.0757 and 85.1322 for isobaric BMAA and DAB, TMAO and TMAO-d9, respectively, obtained from the analysis of a working standard solution at 2500 μg/L in HILIC mode. Although the isobaric BMAA and DAB are greatly retained by this stationary phase (rt = 7.03 and 7.17 min, respectively), their separation is not effective. Conversely, TMAO and TMAO-d9 are not retained and are detectable at the column dead time. Although several attempts were made to increase the efficiency of the BMAA and DAB separation and the TMAO retention by varying both additive type and concentration (in particular, by changing FA concentration from 10 mM to 25 mM and by substituting FA with ammonium acetate, from 1 mM to 10 mM in both water and acetonitrile), no significant improvements were obtained. The effective separation of TMAO, and/or BMAA and DAB using HILIC chromatography are anyway reported in the literature, but the proposed methods are based on stationary phases that are quite different from the one herein described, e.g., ZIC-HILIC [8], ethylene-bridged hybrid (BEH) particles [12] or amide-based stationary phase [17].

**Figure 1.** (**A**) Overlapped EIC chromatograms (accuracy 10 ppm) for BMAA ([M+H]+, *m*/*z* 119.0815, black, rt = 7.03 min), DAB ([M+H]+, *m*/*z* 119.0815, black, rt = 7.17 min), TMAO ([M+H]+, *m*/*z* 76.0757.0815, red, rt = 1.27 min) and TMAO-d9 (IS) ([M+H]+, *m*/*z* 85.1322, green, rt = 1.27 min) obtained with Luna HILIC. (**B**) Overlapped EIC chromatograms (accuracy 10 ppm) for BMAA ([M+H]+, *m*/*z* 119.0815, black, rt = 5.00 min), DAB ([M+H]+, *m*/*z* 119.0815, black, rt = 5.30 min), TMAO ([M+H]+, *m*/*z* 76.0757.0815, red, rt = 7.32 min) and TMAO-d9 (IS) ([M+H]+, *m*/*z* 85.1322, green, rt = 7.29 min) obtained with and Acclaim Mixed-Mode WCX-1. The concentration is 2500 μg/L for each analyte.

Better results, in terms of peaks separation and retention, were obtained with a mixed cationic-RP column in reverse-phase conditions, as reported in Figure 1B. In this case, the isobars BMAA (rt = 5.00 min) and DAB (rt = 5.30 min) were retained and satisfactorily separated. Even for TMAO and TMAO-d9 (rt = 7.29 min) retention was considerably improved.

Greater sensitivity and selectivity can be achieved by acquiring data in high-resolution MS/MS mode, in particular for BMAA and DAB. The comparison between the MS/MS spectra of BMAA and DAB (Figure 2, panel A), shows the presence of two significantly different fragment ions for DAB and BMAA: signal at *m*/*z* 88.0399 was related to the loss of substituted amino moiety of BMAA, and signal a *m*/*z* 101.0714 due to the loss of the terminal NH3 of DAB. Once selected the precursor ions, the EIC (mass accuracy: 10 ppm) related to the fragment ions at *m*/*z* 58.0880, *m*/*z* 68.1302, *m*/*z* 101.0714 and *m*/*z* 88.0399 for TMAO, TMAO-d9, DAB and BMAA, respectively, were acquired in parallel reaction monitoring mode (Figure 2, panel B), thus improving the MS sensitivity and the selectivity obtained by the chromatographic separation. The obtained separation is anyway necessary when lower selective detection systems, such as LC coupled to single-stage or low-resolution MS are used. In the last cases, the shared fragment ions can produce inaccuracies or false-positive results.

**Figure 2.** *Cont.*

**Figure 2.** (**A**) Comparison between the MS/MS spectra of BMAA and DAB, both obtained in HCD mode with a normalized collision energy (NCE) of 35. (**B**) EIC MS/MS chromatograms (accuracy: 10 ppm) obtained with Acclaim Mixed-Mode WCX-1: BMAA: EIC for *m*/*z* 88.0394 (parent ion: *m*/*z* 119.0815, NCE:35); DAB: EIC for *m*/*z* 101.0709 (parent ion: *m*/*z* 119.0815, NCE:35); TMAO: EIC for *m*/*z* 58.0660 (parent ion: *m*/*z* 85.1322, NCE:35); TMAO-d9 (IS): EIC for *m*/*z* 68.1302 (parent ion: *m*/*z* 85.1322, NCE:35). The concentration is 2500 μg/L for each analyte.

The method linearity, precision and accuracy, the last intended as the combined contribution of matrix effects and recoveries, were evaluated in plasma and urine to perform a preliminary evaluation of the performances. Matrix-matched calibration curves were performed in triplicate and were obtained by spiking blank samples at the same nominal concentrations of the external calibration curves. A very good linearity (R<sup>2</sup> > 0.99) was observed for all the selected analytes in water, plasma and urine matrices. Precision, evaluated from the standard deviations of the regression slopes, was excellent and showed relative standard deviations (RSD%) < 3% for all the matrices taken into consideration. Matrix effects and recoveries were evaluated from the percent slope ratio of the matrix-matched and external standard calibration curves. Values within the range 70–110% are generally acceptable, as strong matrix effects and poor recoveries can be excluded. Considering the excellent precision of the method, it is possible to assume good recoveries and a limited matrix effect for BMAA and DAB in urine, as the values obtained for the percent slope ratio were 71% and 85%, respectively. Results obtained for plasma are acceptable for DAB (73%) and quite good for TMAO (81%). Such method performances were already proved to be reliable for the quantification of TMAO in mouse plasma (RSD of the IS < 15% for N = 77 samples) [19]. Nevertheless, the combined matrix effect, in terms of ESI signals variation and recovery, was not acceptable in plasma for BMAA, as values are lower than 50%. In this last case, probably due to the interfering compounds co-extracted from this complex matrix, it will be necessary to increase the dilution of the sample before analysis or modify the extraction procedure or the chromatographic conditions. The choice of a suitable IS could anyway improve its accuracy. LODs, assessed from the lowest point of the matrix-matched calibration curve, and corresponding to an S/N value of 3, were estimated to be 10 μg/L for BMAA and DAB and 2 μg/L for TMAO.

Preliminary quantitative data in plasma and urine samples were obtained only for TMAO, which was present as an average (*n* = 5) basal concentration of 121 ± 8 μg/L in plasma and 193 ± 9 μg/L in urine, respectively. BMAA and DAB were always lower than LODs in both plasma and urine samples, but the collection of specimens suspected to be positively correlated to BMAA has to be specifically planned, and it is beyond the focus of this work.

Further tests are in progress in our laboratory using the chromatographic method herein reported in order to expand the panel of analytes potentially quantifiable using this mixed cationic-RP column. The novel set of analytes included: other isomers of BMAA, such as beta-amino-*N*-methylalanine and *N*-(2aminoethyl) glycine; levodopa, carbidopa and dopamine, which are compounds involved in the Parkinson's disease, and may be useful for the assessment of the possible adverse effects of BMAA; trimethylamine, short-chain fatty acids, e.g., butyric, isobutyric, valeric, isovaleric, hexanoic and acetic acid are key compounds, together with TMAO, linked to the gut microbiota, whose alteration was recently associated with the development of type 2 diabetes and obesity [20]; metabolites such as picolinic and nicotinic acid, tryptophan, kynurenic acid, 3-hydroxykinurenine were selected as representative of other specific metabolic pathways, e.g., the tryptophan metabolism. Corticosterone, an important intermediate for the synthesis of glucocorticoid hormones in humans, was selected in order to evaluate the chromatographic interactions of the mixed-RP stationary phase with lipophilic substances. Preliminary results regarding the chromatographic separation of some of the selected analytes obtained by the Acclaim Mixed-Mode WCX-1 column are reported in Figure 3. As shown, this column was suitable for retaining compounds with a wide range of polarities, from trimethylamine to corticosterone.

**Figure 3.** EIC chromatograms obtained with an Acclaim Mixed-Mode WCX-1 in ESI(+)-HRMS. (**A**) Trimethylamine, [M+H]+, EIC at *m*/*z* 60.0808. (**B**) TMAO, [M+H]+, EIC at *m*/*z* 76.0757. (**C**) TMAO-d9 (IS), [M+H]+, EIC at *m*/*z* 85.1322. (**D**) BMAA (left peak) and DAB (right peak), [M+H]+, EIC at *m*/*z* 119.0815. (**E**) Kynurenic acid, [M+H]+, EIC at *m*/*z* 190.0499. (**F**) Dopamine, [M+H]+, EIC at *m*/*z* 154.0863. (**G**) Homocysteine, [M+H]+, EIC at *m*/*z* 136.0427. (**H**) Carbidopa, [M+H]+, EIC at *m*/*z* 227.1026. (**I**) Picolinic acid, [M+H]+, EIC at *m*/*z* 124.0393. (**L**) L-DOPA, [M+H]+, EIC at *m*/*z* 198.0761. (**M**) 3-hydroxykynurenine, [M+H]+, EIC at *m*/*z* 225.0870. (**N**) Corticosterone, [M+H]+, EIC at *m*/*z* 347.2217. Mass accuracy: 10 ppm. The concentration is 2500 μg/L for each analyte.

#### **4. Conclusions**

In this work, the effective chromatographic retention of selected highly polar metabolites was carried out by using a mixed cationic-RP column, simultaneously obtaining an efficient separation of the isobaric BMAA and DAB without derivatization and ion pairing. The selectivity of the method was increased by HR tandem MS, avoiding the contribution of the partial co-eluted peaks. A preliminary evaluation of the method performances showed good linearity, acceptable recoveries and matrix effects for all the analytes in urine, and DAB and TMAO in plasma. The full validation of the method, including the assessment of the LOD, LOQ, repeatability and reproducibility, is in progress. Further evaluation of the column retention and selectivity started on a larger panel of analytes with different chemical properties and related to the metabolism of tryptophan, the short-chain fatty acids, and other isomers of BMAA and molecules related to the Parkinson disease. The versatility of this alternative chromatographic method is of particular interest in the field of metabolomics, where it is essential to analyze simultaneously various classes of molecules with very different chemical properties, in terms of polarity and molecular weight.

**Author Contributions:** Conceptualization, M.R., S.B. and I.M.D.G.; methodology, M.R., S.B. and I.M.D.G.; validation, M.R. and I.M.D.G.; formal analysis, M.R. and I.M.D.G.; investigation, M.R., S.B. and I.M.D.G.; resources, S.B. and P.P.; data curation, M.R., G.F. and I.M.D.G.; writing—original draft preparation, M.R., S.B. and G.F.; writing—review and editing, M.R., S.B. and P.P.; visualization, S.B. and P.P.; supervision, S.B.; funding acquisition, S.B. and P.P. All authors have read and agreed to the published version of the manuscript.

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

**Acknowledgments:** The authors are grateful to the research group of Gian Paolo Fadini (University of Padua) for collaboration with the TMAO project.

**Conflicts of Interest:** The authors declare they have no conflicts of interest.

#### **References**


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

© 2020 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/).

*Article*

### **Determination of Non-Steroidal Anti-Inflammatory Drugs in Animal Urine Samples by Ultrasound Vortex-Assisted Dispersive Liquid–Liquid Microextraction and Gas Chromatography Coupled to Ion Trap-Mass Spectrometry**

**Pasquale Avino, Ivan Notardonato, Sergio Passarella and Mario Vincenzo Russo \***

Department of Agricultural, Environmental and Food Sciences, University of Molise, via De Sanctis, I-86100 Campobasso, Italy; avino@unimol.it (P.A.); ivan.notardonato@unimol.it (I.N.); sergio.passarella@studenti.unimol.it (S.P.)

**\*** Correspondence: mvrusso@unimol.it; Tel.: +39-0874-404-717

Received: 26 May 2020; Accepted: 29 June 2020; Published: 6 August 2020

**Featured Application: The paper would like to show an easy, rapid, and a**ff**ordable protocol to be used for determining four non-steroidal anti-inflammatory drugs (NSAIDs) (i.e., acetylsalicylic acid, ibuprofen, naproxen, and ketoprofen) in urine samples at trace levels. The method could be routinely used in several situations, from medicine and veterinary to doping issues.**

**Abstract:** A low solvent consumption method for the determination of non-steroidal anti-inflammatory drugs (NSAIDs) in animal urine samples is studied. The NSAIDs were extracted with CH2Cl2 by the ultrasound vortex assisted dispersive liquid–liquid microextraction (USVA-DLLME) method from urine samples, previously treated with β-glucuronidase/acrylsulfatase. After centrifugation, the bottom phase of the chlorinated solvent was separated from the liquid matrix, dried with Na2SO4, and derivatized with N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA) + trimethylchlorosilane (TMCS) (99 + 1). After cooling at room temperature, the solution was concentrated under nitrogen flow, and 1 μL of solution was analyzed in gas chromatography/ion trap-mass spectrometry (GC-IT-MS). The enrichment factor was about 300–450 times and recoveries ranged from 94.1 to 101.2% with a relative standard deviation (RSD) of ≤4.1%. The USVA-DLLME process efficiency was not influenced by the characteristics of the real urine matrix; therefore, the analytical method characteristics were evaluated in the range 1–100 ng mL−<sup>1</sup> (R2 <sup>≥</sup> 0.9950). The limits of detection (LODs) and limits of quantification (LOQs) were between 0.1 and 0.2 ng mL−<sup>1</sup> with RSD <sup>≤</sup>4.5% and between 4.1 and 4.7 ng mL−<sup>1</sup> with RSD <sup>≤</sup>3.5%, respectively, whereas inter- and intra-day precision was 3.8% and 4.5%, respectively. The proposed analytical method is reproducible, sensitive, and simple.

**Keywords:** non-steroidal anti-inflammatory drug (NSAID); urine; doping analysis; dispersive liquid–liquid microextraction (DLLME); gas chromatography mass spectrometry (GC-MS)

#### **1. Introduction**

Anti-inflammatory drugs, used for reducing inflammation, are of two types, i.e., cortisone-based and non-steroidal anti-inflammatory drugs (NSAIDs). The latter are, in all likelihood, the best known and most used category of anti-inflammatory drugs in therapy [1]. NSAIDs are a wide class of drugs showing anti-inflammatory, analgesic, and antipyretic action and include some of the best-known molecules used to fight pain [2]: ibuprofen, nimesulide, ketoprofen, naproxen, and diclofenac. They are able to stop the inflammation process by their mechanism of action, i.e., interfering with the synthesis of prostanoids; molecules that play a fundamental role in these processes [3]. To do this, the NSAIDs block one or more passages of the metabolism of arachidonic acid, which is the precursor of prostaglandins [4]. Further, NSAIDs can also be used as pain relievers and antipyretics [5,6].

NSAIDs are associated with a small increase in the risk of a heart attack, stroke, or heart failure [7]. However, even in this case, the real danger depends on the type of molecule taken, the duration of the treatment, and the doses taken. Short-term use can instead trigger less serious but sometimes serious adverse effects, such as ulcers, gastric bleeding, and kidney damage [8–10]. In addition, NSAIDs can trigger allergic reactions and interfere with the activity of antihypertensive drugs [11].

Furthermore, NSAIDs are commonly used in animal medicine in different inflammatory situations (e.g., for curing musculoskeletal problems in equines) [12–14]. On the other hand, these drugs are improperly used for masking inflammation and pain of an animal, especially before horse racing. NSAIDs are substances prohibited in horse competitions and are considered one of the main doping agents [15–18]. For instance, salicylic acid, a NSAID used for the treatment of pain and fever, has an allowed threshold of 750 μg mL−<sup>1</sup> in urine, or 6.5 μg mL−<sup>1</sup> in plasma, for equines [19].

NSAIDs are considered safe drugs, but acute overdose or chronic abuse can give serious toxic effects [20,21]. They are weak in acid (pKa 3–5) and some of them show short half-lives (e.g., ibuprofen 2–3 h [22]), whereas others show long half-lives (e.g., phenylbutazone residual can also be detected after 24 h [23]). A screening procedure is necessary for detecting such drugs in urine samples. Different analytical methods are present in literature, mainly based on liquid–liquid extraction (LLE) or solid-phase extraction (SPE), followed by chromatographic methods (i.e., HPLC with fluorescence detector HPLC-FLD, HPLC-diode array detection (DAD), gas chromatography mass spectrometry (GC-MS), GC-MS/MS, UHPLC-MS/MS, capillary electrophoresis CE-DAD, and CE-MS) [20,24–33]. Further, a derivatization step is necessary before the GC-MS analysis [30,31,34].

Recently, Rezaee et al. introduced the dispersive liquid–liquid microextraction (DLLME) [35]. The extraction is based on the addition of both an immiscible solvent with higher density to the aqueous sample and a dispersant solvent for increasing the contact between the two immiscible solvents. For many years, researchers have deepened this method by applying it to different matrices [36–38], especially for avoiding (at least, for reducing) the use of highly toxic chloro-solvents [39]. In this way, several protocols based on ultrasound vortex assisted DLLME (USVA-DLLME) for determining toxic compounds in foodstuffs have been investigated and set up [40–44].

The aim of this study was to develop a simple method for the simultaneous screening and confirmation of four NSAIDs, i.e., acetylsalicylic acid (ASA), ibuprofen (IBP), naproxen (NAP), and ketoprofen (KPF), in animal urine samples. The entire procedure, not previously reported in literature, starts with the extraction procedure, i.e., the USVA-DLLME method, followed by the NSAID derivatization step with N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA)-trimethylchlorosilane (TMCS) to form the relative trimethylsilyl (TMS) derivates: gas chromatography coupled with an ion trap-mass spectrometry detector (GC-IT-MS) has allowed us to detect the NSAID residues in real samples.

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

#### *2.1. Materials*

Ethanol, C2Cl2, CHCl3, C2H4Cl2, C2H2Cl4, and acetone were of pesticide grade (Carlo Erba, Milan, Italy), whereas NaCl, acetic acid, NaOH, HCl, and Anhydrous Na2SO4 were of analytical grade (Carlo Erba). Standards of acetylsalicylic acid, ibuprofen, naproxen, and ketoprofen were purchased as powder from Sigma–Aldrich (Milan, Italy), whereas anthracene, used as the internal standard (IS), was provided by LabService Analytical (Anzola Emilia, Bologna, Italy). Beta-glucuronidase/arylsulfatase and BSTFA-TMCS (99 + 1) solutions were given by Sigma–Aldrich.

The solutions (1 mg mL−1) of each analyte, i.e., acetylsalicylic acid, ibuprofen, naproxen, and ketoprofen (Table 1), were prepared in acetone. These solutions were further diluted for preparing

final working standard solutions for spiking both the blank solutions (simulated urine samples) and real samples.

**Table 1.** The non-steroidal anti-inflammatory drugs (NSAIDs) investigated in this paper, with their corresponding abbreviations, Chemical Abstracts Service (CAS) number, chemical structure, molecular weight (MW), target, and qualifier ions (selected ion monitoring (SIM), abundance 100%).


<sup>a</sup> Abbreviations: acetylsalicylic acid (ASA), ibuprofen (IBP), naproxen (NAP), ketoprofen (KPF); <sup>b</sup> target and qualifier ions of the trimethylsilyl derivates.

The anthracene solution (1 mg mL<sup>−</sup>1) was prepared in ethanol, and by further dilution the working solution was obtained. NaOH 1 M, HCl 1 M, and CH3COOH 1 M was used to adjust the pH of the blank, and real samples were prepared with ultrapure water (resistivity 18.2 MΩ cm<sup>−</sup>1) and obtained by means of a Milli-Q purification system (Millipore, Bradford, MA, USA).

#### *2.2. Sample Preparation*

#### 2.2.1. Preparation of Simulated Urine Samples

For simulating a urine sample, an aqueous solution containing the most present components was prepared as follows: urea 14 g L−1, creatinine 0.4 g L−1, uric acid 0.05 g L−1, glucose 0.06 g L−1, mono potassium phosphate 0.2 g L<sup>−</sup>1, and sodium chloride 13 g L−1.

#### 2.2.2. Preparation of Animal Urine Samples

Animal urine samples were provided by small farm owners near Campobasso (Molise, Italy). Each sample was filtered through a 0.45 μm pore size cellulose acetate filter and buffered at pH 5 with few drops of acetic acid, with the addition of a few μL of NaOH 1 M. Before performing the extraction and derivatization procedures, the animal urine samples were subjected to enzymatic hydrolysis. With total of 9 mL of sample and 100 μL of β-glucuronidase/arylsulfatase [45], the IS (5 μL of anthracene, 60 ng μL<sup>−</sup>1) were incubated overnight at 37 ◦C.

#### 2.2.3. USVA-DLLME and Derivatization Procedure

The extraction procedure was performed as follows: the mixture of dispersive (1 mL of acetone) and extraction (250 μL of CH2Cl2) solvent was injected above the sample level of the solution previously kept at room temperature at pH 3 with a few μL of HCl [46]. The solution was subjected to vortex for 1 min and ultrasounds for 2 min: This occurrence was repeated three times, followed by centrifugation for 10 min at 4000 rpm at room temperature. The organic phase was withdrawn with a micro-syringe and placed in a vial with the addition of a few grains of anhydrous sodium sulphate. A total of 50 μL of BSTFA + TMCS (99 + 1, v + v) were added [47] and the vial was closed and heated up to 50 ◦C for 30 min. Afterwards, the vial was cooled at room temperature and the organic phase was concentrated to a final volume of 20–50 μL under a slight nitrogen flow and 1 μL were analyzed in GC-IT-MS.

#### *2.3. GC-IT-MS Apparatus*

Analysis and data acquisition were performed using a gas chromatograph Finnigan Trace GC Ultra, equipped with an ion trap mass-spectrometry detector Polaris Q (Thermo Fisher Scientific, Waltham, MA, USA), a programmed temperature vaporizer (PTV) injector, and a PC with a chromatography station Xcalibur 1.2.4 (Thermo Fisher Scientific).

A fused-silica capillary column with a chemically bonded phase (SE-54, 5% phenyl-95% dimethylpolysiloxane) was prepared in our laboratory [48–50] with the following characteristics: 30 m × 250 μm i.d.; N (theoretical plate number) 132,000 for *n*-dodecane at 90 ◦C; K', capacity factor, 7.0; df, (film thickness) 0.246 μm; uopt (optimum linear velocity of carrier gas, hydrogen) 39.5 cm s<sup>−</sup>1; utilization of theoretical efficiency (UTE) 95%. A 1 μL sample was injected into the PTV injector in the splitless mode. A total of 10 s after, the injection the vaporizer was heated from 110 ◦C to 290 ◦C at 800 ◦C min−1; the splitter valve was opened after 120 s (split ratio 1:50). The transfer line and ion source were held at 270 ◦C and 250 ◦C, respectively. Helium (IP 5.5) was used as a carrier gas at a flow rate of 10 mL min<sup>−</sup>1. The oven temperature program was as follows: 100 ◦C for 60 s, 10 ◦C min−<sup>1</sup> up to 290 ◦C, and held for 120 s. The IT/MS was operated in the electron ionization mode (70 eV), and the analytes were qualitatively identified in the full-scan mode (*m*/*z* 100–500) and quantified in the selected ion monitoring (SIM) mode (Table 1). The quantitative analysis was performed by calibration graphs of ratio Area(NSAID)/Area(IS) plotted versus each NSAID concentration (ng mL−1). All the samples were determined in triplicate.

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

For USVA-DLLME extraction of the four investigated NSAIDs from animal urine samples, several parameters that control the optimal extraction performance were investigated and optimized using the one variable at a time method. It should be highlighted that the entire analytical methodology has been studied by means of simulated urine samples, prepared according to what reported in Section 2.2.1 and after applied to real urine samples. Simultaneously, the use of β-glucuronidase was welcome because it increased the IBP detection [33].

#### *3.1. Parameter Optimization*

The parameter optimization was addressed to find out the best analytical conditions for achieving high recoveries and accurate and precise determinations of the NSAIDs in animal urine samples. In this way, extraction solvent and volume, sample pH, and NaCl effect were deeply investigated.

First, the study dedicated its attention on the choice of organic extraction solvent. This issue plays a key role in the extraction efficiency. Chlorinated solvents are generally used because they show characteristics (higher density than water, low solubility in water) appropriate to obtaining high extraction efficiency and worthy gas chromatographic performance. Following these considerations, our attention was focused to five solvents: dichloromethane (CH2Cl2; d = 1.3255 g mL−1), chloroform (CHCl3; d = 1.4788 g mL−1), carbon tetrachloride (CCl4; d = 1.5940 g mL−1), 1,2-dichloroethane (C2H4Cl2; d = 1.2454 g mL−1), and 1,1,2,2-tetrachloroethane (C2H2Cl4; 1.5953 g mL<sup>−</sup>1). Table 2 reports the results of the performance of a 300 μL volume of each solvent on simulated urine samples spiked with 20 ng mL−<sup>1</sup> of each NSAID: Dichloromethane shows the best recoveries, ranging between 94.6% and 98.5% for IBP, NAP, and KPF, respectively, and 82.5% for ASA with a relative standard deviation (RSD, %) below 3.0. The recoveries are calculated as the accuracy (IS added before the extraction) [51].

The extraction recovery, defined as the percentage of the total analyte (n0), that was extracted to the sediment phase (nsed) has been determined according to the formula reported in a previous paper [36]. Over the extraction solvent choice, another quite important parameter is its volume, used to achieve the highest recoveries. The strength of the DLLME regards an extraction solvent volume as low as possible for obtaining good performance. Leong and Huang [39] highlighted that an extraction solvent volume leads to a change in the sediment phase volume and therefore in the enrichment factors (EFs). For these reasons, the effect of different dichloromethane volumes (200, 250, 300 μL) were investigated (Table 3): a volume of 250 μL is sufficient to obtain good recoveries for all the NSAIDs, i.e., 94.2% for ASA, 100.1 for IBP, 99.8 for NAP, and 101.2 for KPF with RSDs ≤3.1.

**Table 2.** Effect of different extraction solvents on the NSAID recovery accuracy (%). The conditions were as follows: 9 mL of simulated urine samples spiked with NSAIDs (20 ng mL−<sup>1</sup> of each), 1 mL of acetone, 300 μL of extraction solvent, and 5 μL of anthracene (I.S.; 60 ng μL<sup>−</sup>1). In brackets are reported the relative standard deviations (RSDs, %); each analysis was in triplicate.


**Table 3.** Effect of different volumes of CH2Cl2 on the NSAID recoveries (%). The conditions were as follows: 9 mL of simulated urine samples spiked with NSAIDs (20 ng mL−<sup>1</sup> of each), 1 mL of acetone, different volumes of CH2Cl2 as extraction solvent, and 5 μL of IS (60 ng μL<sup>−</sup>1). In brackets are reported the RSDs (%); each analysis was in triplicate.


Another parameter influencing the extraction is the pH of the solution. In fact, it should be remembered that NSAIDs are weak acids. Particularly, ASA shows a pKa of 3.5 [52], IBP of 5.3 [53], NAP of 4.14, and KPF of 4.45 [54]. Solutions of simulated urine samples at different pH were tested for studying the best acidic conditions. Table 4 evidences that the best recoveries and RSDs are obtained at pH 3: in fact, they range between 93.5 and 100.1% and between 3.4 and 4%, respectively.

**Table 4.** The effect of pH on the NSAID recoveries (%). The conditions were as follows: 9 mL of simulated urine samples spiked with NSAIDs (20 ng mL−<sup>1</sup> of each), 1 mL of acetone, 250 μL of CH2Cl2, and 5 μL of IS (60 ng μL<sup>−</sup>1). In brackets are reported the RSDs (%); each analysis was in triplicate.


Finally, the effect of different NaCl quantities on the NSAID recoveries was evaluated. Table 5 shows that the salt decreased the NSAID solubility (salting out) below and above 13 g L−<sup>1</sup> concentration. Further, the decision to perform the whole study at NaCl concentration of 13 g L−<sup>1</sup> was essentially due to two considerations: (1) this concentration was the average of those reported in the real urine samples, which was between 10 and 16 g L−<sup>1</sup> of NaCl [55]; (2) the percentage NSAID recoveries obtained and reported in Table 5 were very similar to each other for NaCl concentrations between 10 and 15 g L<sup>−</sup>1.

Finally, it should be highlighted that two other interesting parameters, such as vortex time and ultrasonication time, were extensively studied in previous papers by this group [41–44].

**Table 5.** Effect of different NaCl amounts on the NSAID recoveries (%). The conditions were as follows: 9 mL of simulated urine samples spiked with NSAIDs (20 ng mL−<sup>1</sup> of each), pH 3, 1 mL of acetone, 250 μL of CH2Cl2, and 5 μL of IS (60 ng μL<sup>−</sup>1). In brackets are reported the RSDs (%); each analysis was in triplicate.


#### *3.2. GC-IT-MS Method Validation*

Using optimized parameters, all the analytical data were investigated. Table 6 shows the correlation coefficients (R2) in the range 1–100 μg L<sup>−</sup>1, along with the limits of detection (LODs) and limits of quantification (LOQs), repeatability (as intra-day precision) and reproducibility (as inter-day precision), and EFs of each NSAID considered. LODs and LOQs were determined according to Knoll's definition [56,57], i.e., an analyte concentration that produces a chromatographic peak equal to three times (LOD) and seven times (LOQ) the standard deviation of the baseline noise. All the compounds show a good linearity in the investigated range (≥0.995) and LODs and LOQs between 0.1–0.2 <sup>μ</sup>g L−<sup>1</sup> and 4.1–4.7 <sup>μ</sup>g L<sup>−</sup>1, respectively, with high intra- and inter-day precision (≤3.8 and <sup>≤</sup>4.5, respectively). The EFs, defined as the ratio between the analyte concentration in the sediment phase (Csed) and the initial analyte concentration (C0) in the sample (EF = Csed/C0) [35], were also studied, ranging between 350–450.

**Table 6.** Correlation coefficients (R2) calculated in the range 1–100μg L<sup>−</sup>1, limit of detection (LOD;μg L<sup>−</sup>1) and limit of quantification (LOQ; μg L<sup>−</sup>1) and inter- and intra-day precision (expressed as RSD, %) of each NSAID determined by GC-IT-MS.


Finally, for a complete analytical methodology evaluation, the recoveries have been studied in the investigated matrices, i.e., animal urine, at two different spiked NSAID concentrations (20 ng mL−<sup>1</sup> and 50 ng mL−1). Table 7 shows these data: recoveries in animal urine samples between 93.8 and 102 with RSDs ≤3.2.


**Table 7.** Average NSAID recoveries (%) obtained at different spiking concentrations on real urine samples. In brackets are reported the RSDs (%); each analysis was in triplicate.

<sup>a</sup> Goat urine sample.

Finally, Table 8 shows a comparison among different methods present in literature [58–63] for analyzing NSAIDs. The extraction methods were different: three papers were based on hollow-fiber liquid microextraction [59,60,62], whereas two papers were on rotating disk sorptive [63] and liquid–liquid extraction [61]. According to the analytical techniques, three studies used HPLC with ultraviolet (UV) [58,61,63] and one the diode array detection (DAD) [59], one used the ultra-performance liquid chromatography coupled with tandem mass spectrometry (UPLC-MS/MS) [62], and one used GC equipped with a flame ionization detector (FID) [60]. Looking at the comparison among the different studies with parameters developed in this study, the main advantages regard LODs and LOQs, recoveries, and RSDs, whereas EFs are good except those reported by Payan et al. [59]. On the other hand, the whole procedure can be routinely applied and does not require particular technology, such as the use of rotating disks or hollow fiber.



chromatography-diode-array detection; 3 hollow-fiber liquid membrane-protected solid-phase microextraction-gas chromatography flame ionization detector; 4 solid phase extraction combined with supramolecular solvent-high performance liquid chromatography-UV; 5 hollow-fiber liquid-phase microextraction-ultra performance liquid chromatography-tandem mass spectrometry; 6 rotating disk sorptive extraction HPLC-UV; 7 diclofenac (DIC), fenoprofen (FPC), mefenamic acid (MFA); 8 not reported.

#### *3.3. Application to Real Animal Urine Samples*

Using the entire analytical USVA-DLLME-GC-IT-MS protocol previously developed (briefly resuming: 9 mL of simulated urine sample solution at pH 3 containing 5 μL of I.S., 60 ng mL<sup>−</sup>1, addition of 1 mL of acetone and of 250 μL dichloromethane as extraction solvent, three times of 1 min vortex and 2 min ultrasounds, centrifugation for 10 min at 4000 rpm, 1 μL injection into GC-IT-MS), some animal urine samples have been analyzed, particularly three animal urine samples, i.e., two from goats and one from a sheep. All the subjects were healthy. No residues (i.e., levels below the LODs) were found in all the samples. The analysis allows us to investigate the presence of such compounds at trace levels in these matrices, but it does not furnish evidence as to whether there was a previous assumption of such molecules. As an example are shown in Figure 1, the gas chromatograms in SIM mode of a simulated sample of urine (a) and one of goat urine sample (b) both additions with 30 ng mL−1 of each NSAID. The peaks are well-solved and the determinations are precise and accurate.

**Figure 1.** Gas chromatography/ion trap-mass spectrometry (GC-IT-MS) chromatograms in SIM mode of (**a**) simulated urine and (**b**) goat urine samples, both spiked with 30 ng mL−<sup>1</sup> of each NSAID. For experimental conditions, see text. Peak list: 1. acetylsalicylic acid; 2. ibuprofen; internal standard (IS); 3. naproxen; 4. ketoprofen.

#### **4. Conclusions**

This paper highlights an affordable method for analyzing NSAIDs in animal urine samples. The method used for the animal urine samples in this study can also be applied to human urine samples, as a lead on to the discussion about athletes. In fact, athletes often make excessive use of anti-inflammatories in order to compete, even in less than optimal physical conditions. Many athletes take NSAIDs to compete or even simply train, even in the presence of pain, joint inflammation, trauma etc. Incorrect use of these drugs can lead to serious damage to health. Further, with regards to "premedication" in the sports field, it should be highlighted that the NSAIDs are not among the substances prohibited by the anti-doping measures and are therefore only drugs at risk of easy inappropriate abuse. Equine doping can also be defined as "the use of any exogenous agent (pharmacological, endocrinological, hematological, etc.) or clinical manipulation, which, in the absence of suitable and necessary therapeutic indications, is aimed to improve performance, outside the adjustments induced by training. In this view, this paper shows a simple, rapid, and sensitive method for determining four NSAIDs in animal urine samples. The very low LODs and LOQs and the high precision reached by means of a modified DLLME method coupled with GC-IT-MS allow us to apply the entire procedure to routine screening and monitoring of such compounds in doping cases or other similar situations.

**Author Contributions:** Conceptualization, M.V.R.; methodology, M.V.R. and P.A.; validation, I.N. and S.P.; formal analysis, S.P.; investigation, I.N.; resources, M.V.R.; data curation, M.V.R. and P.A.; writing—original draft preparation, P.A.; writing—review and editing, M.V.R. and P.A.; supervision, M.V.R.; project administration, M.V.R. All authors have read and agreed to the published version of the manuscript.

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

**Acknowledgments:** The authors would like to thank Alessandro Ubaldi for his helpful suggestions in the data interpretation.

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

#### **References**


© 2020 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/).

### *Article* **Fast and Low-Cost Synthesis of MoS2 Nanostructures on Paper Substrates for Near-Infrared Photodetectors**

**Neusmar J. A. Cordeiro 1,\*, Cristina Gaspar 2, Maria J. de Oliveira 2, Daniela Nunes 2, Pedro Barquinha 2, Luís Pereira 2, Elvira Fortunato 2, Rodrigo Martins 2, Edson Laureto <sup>1</sup> and Sidney A. Lourenço 3,\***


**Featured Application: This work presents an analysis of time and temperature influence of microwave-assisted hydrothermal synthesis on the direct growth of 2D-MoS2 nanostructures on cellulose paper substrates, and the production of MoS2-based low-cost photosensors with high responsivity and detectivity values.**

**Abstract:** Recent advances in the production and development of two-dimensional transition metal dichalcogenides (2D TMDs) allow applications of these materials, with a structure similar to that of graphene, in a series of devices as promising technologies for optoelectronic applications. In this work, molybdenum disulfide (MoS2) nanostructures were grown directly on paper substrates through a microwave-assisted hydrothermal synthesis. The synthesized samples were subjected to morphological, structural, and optical analysis, using techniques such as scanning electron microscopy (SEM), X-ray diffraction (XRD), and Raman. The variation of synthesis parameters, as temperature and synthesis time, allowed the manipulation of these nanostructures during the growth process, with alteration of the metallic (1T) and semiconductor (2H) phases. By using this synthesis method, two-dimensional MoS2 nanostructures were directly grown on paper substrates. The MoS2 nanostructures were used as the active layer, to produce low-cost near-infrared photodetectors. The set of results indicates that the interdigital MoS2 photodetector with the best characteristics (responsivity of 290 mA/W, detectivity of 1.8 <sup>×</sup> <sup>10</sup><sup>9</sup> Jones and external quantum efficiency of 37%) was obtained using photoactive MoS2 nanosheets synthesized at 200 ◦C for 120 min.

**Keywords:** MoS2; microwave-assisted hydrothermal synthesis; low-cost photosensors

#### **1. Introduction**

Since its discovery in 2004, graphene has become one of the nanomaterials of great interest in the construction of devices, due to its high electronic conductivity, mechanical flexibility, and low production cost [1]. Despite the good results obtained with graphene [2,3], the absence of energy bandgap restricts its application in some devices, such as photodetectors, mostly due to low intrinsic responsivity. This led to the development of a series of other two-dimensional materials with different characteristics, such as the hexagonal boron nitride [4], silicene [5], borophene [6], black phosphourous [7], and two-dimensional transition metal dichalcogenides (2D TMDs) [8]. The latter have been thoroughly explored in recent years for several applications [8].

Among the 2D TMDs materials, composed of a transition metal (M) and a chalcogen (X) with generalized form MX2 in which M = Mo, W, Nb, Ta, Hf, Pt, and so on, and

**Citation:** Cordeiro, N.J.A.; Gaspar, C.; Oliveira, M.J.d.; Nunes, D.; Barquinha, P.; Pereira, L.; Fortunato, E.; Martins, R.; Laureto, E.; Lourenço, S.A. Fast and Low-Cost Synthesis of MoS2 Nanostructures on Paper Substrates for Near-Infrared Photodetectors. *Appl. Sci.* **2021**, *11*, 1234. https://doi.org/10.3390/ app11031234

Academic Editor: Samuel B. Adeloju Received: 30 December 2020 Accepted: 25 January 2021 Published: 29 January 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/).

X = S, Se, Te, the MoS2 nanostructures has generated great interest in the scientific community [8–10]. Featuring a metastable metallic phase (1T), which can be stabilized or converted to a stable semiconductor phase (2H) by the appropriate heat treatment [11] or use of microwave radiation [12], MoS2 has unique characteristics, such as high carrier mobility, strong electron-hole confinement and variable bandgap (1.20 to 1.89 eV) [13,14]. The bandgap of the MoS2 increases with the decrease of the crystal thickness, to below 100 nm, due to the quantum confinement effect [15], and reaches 1.89 eV for a single monolayer [13]. Thus, it can cover an extent NIR (near-infrared) electromagnetic spectrum (6560 to 10,332 nm) by changing the monolayer number. Additionally, the material changes its electronic structure from an indirect bandgap, in its bulk form, to a direct bandgap in the MoS2 monolayer [16], enabling optical applications. The relationship between its optical and electrical characteristics with the number of stacked layers and the control of the 1T and 2H phases turns the MoS2 into a material of great interest for applications in chemical sensors [17], hydrogen evolution reaction [18], electronic devices [19], sodium-ion battery [20], and photodetectors [21,22].

The most commonly used synthesis methods of the two-dimensional MoS2 are exfoliation (liquid [23,24] and mechanical [24]), chemical vapor deposition (CVD) [25], and hydrothermal [26]. Due to a better relationship between uniformity, the quantity of produced material and production cost, the hydrothermal method has been the most applied in the synthesis of MoS2 nanostructures, usually presenting synthesis times between 20 and 24 h [27,28]. The microwave-assisted hydrothermal method has proved to be a good alternative to the conventional hydrothermal method, offering shorter synthesis times due to "molecular heating", consequently lower energy consumption, enhanced reaction selectivity, and homogeneous volumetric heating [29–31]. This method also allows the direct growth of the material on a substrate, being an alternative to conventional deposition or costly transfer methods [28,32]. Synthesis parameters, such as time and temperature, as well as the chemical processes involved in the production of the material, allow the direct growth of the MoS2 nanostructures on flexible substrates [28], like cellulose-based substrates [33–35]. These advantages will have impact in the development of paper-based electronics that emerge as a future alternative to traditional electronics, seeking low-cost electronic systems and components which can be environmentally friendly [36–38]. To meet the requirements imposed on this new generation of devices, different methods of producing materials and printing technologies are employed for the production of electronic devices [36,38–40], like solar cells [41], thin film transistors [42], light emitting devices [43], electrochromic devices [44], and photosensors [28], among others.

In this work the authors adopted the microwave-assisted hydrothermal method, with different synthesis parameters, for direct growth of MoS2 nanostructures on tracing paper substrates. Structural and morphological characterizations of the MoS2 nanosheets were performed, where it was possible to observe the dispersion of the 2H and 1T phases of the nanostructures grown on paper, making these samples potentially interesting for applications in optoelectronic devices. Thus, near infrared photodetectors of MoS2 nanostructures, grown on cellulose substrates, were built with different synthesis parameters. These interdigital MoS2 photodetectors (with detection at 980 nm) were obtained with high responsivity (290 mA/W) and detectivity of 1.8 × <sup>10</sup><sup>9</sup> Jones. These devices presented higher responsivity values than that reported in the literature, as compared with MoS2 photodetectors produced by the conventional hydrothermal method.

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

Molybdenum disulfide (MoS2) nanosheets were grown on cellulose paper substrates by a microwave-assisted hydrothermal method optimizing the hydrothermal temperature and reaction time. Sodium molybdate dihydrate (Na2MoO4.2H2O) and thiourea (CS(NH2)2) were used as precursors of molybdenum and sulfur, respectively. Cellulose paper was used as a substrate, which was previously washed by sonication in deionized water and by acetone and isopropyl alcohol, both for 15 min.

For direct growth on the cellulose fibers, the substrates were previously immersed in a seed solution composed of sodium molybdate and thiourea (1:4) for 60 min. After that time, the substrates were dried on a hotplate at 80 ◦C for 60 min. Substrates with deposited seeds were placed in a hydrothermal reactor with nutrition solution also composed of sodium molybdate and thiourea (1:4). Then, they were taken to the microwave oven for synthesis that was carried out with a power of 100 W, maximum pressure of 280 psi, and with different values of temperature and time of synthesis. Three different synthesis temperatures were used: 190 ◦C, 200 ◦C, and 220 ◦C. These were chosen to apply the shortest possible time for the growth of the nanostructures, corresponding to each temperature used, until the maximum time of 120 min. Thus, the different temperatures and synthesis times used were: 190 ◦C for 30, 45 and 120 min, 200 ◦C for 15, 30, 45, 60 and 120 min, 220 ◦C for 05, 15, 30, 45, 60, and 120 min. After the growth of the MoS2 on the cellulose fibers, the paper with the grown nanostructures were washed with ethanol in ultrasound for 15 min to remove the nanostructures not fixed on the sample's surface. Finally, the samples were dried in a nitrogen flow.

Interdigital contacts were deposited by the screen-printing technique using commercial conductive ink CRSN2442 Suntronic, composed of silver nanoparticles dispersed on the solvent (solvent-based), acquired from Coats Screen Inks GmbH. The interdigital electrodes have a total dimension of 10.0 × 6.0 mm, 500 μm electrode wide and a 500 μm active channel between the electrodes.

Structural and optical properties of the MoS2 nanostructures, grown on cellulose paper substrate, were studied by X-ray diffraction (XRD), scanning electron microscopy (SEM), and micro-Raman spectroscopy. The crystal phase was obtained by X-ray diffraction (X'Pert PRO MPD), using Cu-k<sup>α</sup> radiation (<sup>λ</sup> = 1.540598 ´ Å). The XRD patterns were collected in symmetric configuration in the 2θ range of 5−70◦, using a step angle of 0.05◦, step time of 2 s, operating at 40 kV tension and 30 mA current. The morphology was studied by scanning electron microscopy (Carl Zeiss AURIGA CrossBeam workstation). Raman measurements were performed in a Renishaw inVia Reflex micro-Raman spectrometer equipped with an air-cooled CCD detector and a HeNe laser operating at 50 mW of 532 nm laser excitation. The spectral resolution of the spectroscopic system is 0.3 cm−1. The laser beam was focused with a 50 × Leica objective lens (N Plan EPI) with a numerical aperture of 0.75. An integration time of 3 scans of 1.5 s each was used for all singlescan measurements to reduce the random background noise induced by the detector, without significantly increasing the acquisition time. The intensity of the incident laser was 2.5 mW. Triplicates were taken for all spectra. The spectrograph was calibrated between different Raman sessions using the Raman line at 521 cm−<sup>1</sup> of an internal Si wafer for reducing possible fluctuations of the Raman system. All spectra were recorded at room temperature. Raw data were collected digitally with the Wire 5.0 software for processing. The characterization of the photodetectors was performed with the use of an AUTOLAB potentiostat (PGSTAT204) for ixV electrical measurements and chronoamperometry and a 980 nm diode laser for irradiation of the sensors, under ambient conditions. The power of the laser light was recorded using a handheld digital power meter console (PM100D) from THORLABS. Neutral density filters were used to vary the excitation power incident on the devices.

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

The microwave-assisted hydrothermal synthesis method was used for the direct growth of MoS2 nanostructures on cellulose paper substrates due to the ease of the technique's application, as well as its ability to change synthesis parameters to optimize the produced structures [29–31]. The growth was carried out using a two-step process, where the synthesis is performed on a substrate with a thin layer of deposited seeds.

For construction of NIR photodetectors, a layer of interdigital silver electrodes was deposited on the layer of the MoS2 nanostructures using the screen-printing technique. Further details on the growth of nanostructures and the construction of the devices can be

found in the Experimental Section. Figure 1 shows a diagram of the MoS2 nanostructures synthesis and the photodetector production.

**Figure 1.** Scheme of the synthesis process of the molybdenum disulfide (MoS2) nanostructures in paper substrates by the microwave-assisted hydrothermal method, and the production of infrared radiation photodetectors.

Figure 2 shows SEM images of the samples synthesized with a synthesis temperature of 220 ◦C at different times.

**Figure 2.** Scanning electron microscopy (SEM) images of the MoS2 nanostructures grown directly on cellulose paper substrates with a synthesis temperature of 220 ◦C and time of: (**a**) 05 min, (**b**) 15 min, (**c**) 30 min, (**d**) 45 min, (**e**) 60 min, and (**f**) 120 min.

Microscopy images show that the beginning of the growth process is characterized by the formation of vertically aligned MoS2 nanosheets on the paper surface (see also Figure S1 in the Supplementary File), while the paper fibers are not fully covered. For longer synthesis times, the surface of the paper becomes completely covered by vertically aligned MoS2 nanosheets, regardless the temperature. The effect of temperature is noticed on the time required to observe the first nanosheets on the paper surface. Such a result is expected since, for lower temperatures, the solution is exposed to lower energies during the synthesis process, requiring a longer time for nanomaterial formation. At 190 ◦C, it is possible to observe that, even for the longest synthesis time (120 min), the profile of the paper fibers is still perceived, indicating that it was not possible to form a thick and uniform nanostructure layer. On the other hand, at 200 ◦C the fibers are uniformly coated with a thick layer of MoS2 nanosheets. It is important to note that the samples synthesized at this temperature had a more uniform surface profile than the samples produced by the conventional hydrothermal method reported in the literature [28,33,35], which generally have agglomerations of MoS2 nanosheets in microspherical profile. A surface totally covered with nanostructures is also observed in the samples synthesized at 220 ◦C. However, at this temperature, it is possible to observe that, for longer synthesis times, the sample surface has regions with a spherical profile very similar to those synthesized by the conventional hydrothermal method [28,32]. This indicates that there was the production of a large number of nanostructures during the synthesis process, where they tend to agglomerate in spherical shape, microflowers, formed by nanosheets. The growth kinetics of MoS2 nanostructures on cellulose paper and other flexible substrates was recently discussed by Sahatiya et al. [28] through classical nucleation and growth theory. These spherical structures are formed when there is excessive nanosheets growth, which generally occurs for longer synthesis times or for MoS2 synthesis without the presence of the substrate [26]. In the synthesis with direct growth on substrate, the excess of grown nanostructures ends up detaching from the paper surface and agglomerating into spherical particles to decrease the surface energy. Note the early formation of these small flowers for samples synthesized at 200 ◦C (Supplementary Figure S1).

Figure 3 shows XRD diffractograms at different temperatures and synthesis times of the MoS2 nanostructures, in addition to the XRD diffractogram of cellulose paper and the JCPDS n◦ 37-1492 card pattern of 2H-MoS2 for comparison purposes [28,45]. Dashed vertical lines are used to facilitate the identification of phases in the diffractograms obtained at different temperatures and synthesis times. It is possible to identify regions, close do 15◦ and 22◦, with the characteristic peaks of the cellulose diffractograms, the 2H-MoS2 phases with XRD peaks at ~32.8◦, ~49.8◦, and 57.8◦ relative to (100), (105), and (110) planes, respectively. For 2θ > 30◦ the XRD peaks fit well with JCPDS card n◦ 37-1492, where the presence of (100) and (110) peaks, ate ~32.8◦ and 57.3◦ respectively, show that the samples present similar atomic arrangement along the basal planes with the bulk MoS2.

According to the standard 2H-MoS2 diffractogram pattern, represented by the JCPDS card n◦ 37-1492 shown in Figure 3, there is a diffraction peak at 14.4◦ regarding the plane (002) of the layered material. However, two other peaks appear in the diffractogram, at 9.3◦ and 18.6◦, identified in Figure 3 by # symbols. MoS2 is a representative 2D layered material and the weak van der Waals interaction between MoS2 layers favors the intercalation of molecules and ions in this space [32,46]. The interlayer distance of the (002) plane, referring to the 14.4◦ peak, of intercalation-free MoS2 is 0.62 nm, while the corresponding interlayer distances of 9.4◦ and 18.6◦ peaks, present in the diffractograms of the synthesizes samples, were 0.93 nm and 0.49 nm, respectively, calculated using the Bragg equation. Several groups attribute the interlayer distances of 0.93 nm and 0.49 nm to the distance between two adjacent MoS2 expand layers—resulting from the intercalation of different materials as CTAB [47], mesoporous carbon (MoS2/m-C) [48], NH4 <sup>+</sup> [49], and NH3 molecules [46]—and that distance between intercalated molecules and the adjacent MoS2 layers, respectively [32,47,48,50]. Other authors have attributed the 18.6◦ peak to a secondorder diffraction [49] It is important to note that, while some authors have attributed the

shifted 9.4◦ and 18.6◦ peaks to the presence of 1T-MoS2 phase [49,51], Lei et al. [52] have alerted about the difficult to use the expanded interlayer spacing of MoS2 as direct evidence to confirm the existence of the 1T-MoS2 phase.

**Figure 3.** X-ray diffraction patterns of MoS2 nanosheets growth on cellulose paper substrate at different synthesis temperatures and times and the JCPDS 37-1492 card pattern of 2H-MoS2. The peaks identified by circles are related to the cellulose paper. The # symbol is used to identify the peaks at 9.3◦ and 18.6◦.

Liu et al. have shown that the use of the Na2MoO4.2H2O and CS(NH2)2 with high thiourea concentration induce the formation of ammonium ion intercaled into MoS2 layer [49]. In the synthesis procedure presented in this work, Na2MoO4.2H2O and thiourea were used as precursors of molybdenum and sulfur, respectively, in the proportion of (1:4). Thus, we believe that the MoS2 nanosheets grown on cellulose paper substrates by a microwave-assisted hydrothermal method present NH4 <sup>+</sup> intercaled between layers of MoS2. It should also be noted that there is an decrease in the intensity of the cellulose diffraction peaks when subjected to heat treatments (probably associated with crystallinity of the cellulose) [53]; even so, the diffraction peaks of MoS2 stand out in the diffractogram for high synthesis times, due to the greater amount of nanomaterial grown on the paper surface, for (100), (105) and (110) planes, or the possible increase of the intercalation with the synthesis times, for the 9.4◦ and 18.6◦ peaks.

It can be noted that there is a continuous increase in peak intensity at 9.2◦ up to 30 min, in samples synthesized at 220 ◦C. Then the intensity of the entire diffractogram, including those peaks related to cellulose paper, starts to decrease for higher times. The high temperature leads to degrading the paper structure and can affecting the growth process for longer synthesis times, leading to less production of nanostructures on the substrate surface. This result is in line with what is reported in the literature, which leads to cellulose degradation processes starting above 200 ◦C [53]. In summary, the growth process of the MoS2 on paper substrate seems to be better at 200 ◦C.

For a complementary analysis of the presence of 1T and 2H-MoS2 phases, Raman spectroscopy measurements were performed, looking for the presence of characteristic peaks of these phases on the samples. After confirming the presence of both phases in the synthesized samples by Raman spectroscopy, an analysis of the spatial distribution of 1T and 2H-MoS2 was performed by the micro-Raman technique. By mapping an area of 225 μm [2] and using steps of 1 μm, the regions of the spectra between 146 to 148 cm−<sup>1</sup> (referring to the main vibrational mode of the 1T phase), and between 378 to 385 cm−<sup>1</sup> (referring to the E1 2g vibrational mode of the 2H phase) [12,46], were highlighted. The Raman spectrum and the micro-Raman mapping of the samples synthesized at 220 ◦C, for times of 05 and 120 min, are shown in Figure 4.

**Figure 4.** Raman spectra and micro-Raman mapping of the MoS2 nanosheets synthetized at 220 ◦C with synthesis times of (**a**,**c**) for 05 min and (**b**,**d**) for 120 min. The blue and green mappings represent 1T- and 2H-MoS2 phases in the samples.

As can be seen in Figure 4a,b, it was possible to identify characteristic Raman peaks of the 1T (vibrational modes J1, J2 and J3) and 2H (vibrational modes E1 2g and A1g) phases in the sample spectrum [46], thus confirming the presence of both phases in the synthesized nanostructure. Figure 4c,d shows the micro-Raman mapping, confirming the presence of both phases, the 1T (blue) and the 2H (green), in both samples and in the same space region. These phases are not occupying well-defined regions, but they are dispersed throughout the mapping area. This behavior is observed for all samples (regardless synthesis time and temperature), as shown in Figure 4 and Figure S2 of the supplementary information. The dispersion of the metallic and semiconductor phases of MoS2 in the samples may be associated with the continuous production of the metallic phase during synthesis. Over time, the 1T phase is converted to a 2H phase, due to the low stability of the metallic phase and presence of microwave radiation during the synthesis process [12]. Thus, the 1T and 2H phases are present in the samples, as can be seen in the diffractogram of the sample synthesized at 200 ◦C for 120 min. The greater dispersion of phases on the surface of the substrate can be good for certain applications, since it increases the metal/semiconductor interface (1T/2H) which leads to greater carrier mobility due to the presence of the 1T phase.

As shown in Figure 1, low-cost interdigital photodetectors were built on cellulose paper substrate with a MoS2 active layer synthesized by the microwave-assisted hydrothermal method at different temperatures and times. Figure 5a–c show the current x voltage curves under laser illumination (980 nm) of the photodetectors with a MoS2 photosensitive

layer synthesized at 190, 200, and 220 ◦C, respectively. The curves are linear and symmetric for the small bias voltage, indicating a ohmic-like contact as observed for the single-layer MoS2 phototransistor synthesized using the CVD technique [54], or by photodetector built with a multilayer MoS2 synthesized by the conventional hydrotermal method, using in this case Ag NPs (nanoparticles) as contact [28]. The small bandgap energy of the multilayer MoS2 may generate a very small Schottky barrier between Ag NPs electrodes and MoS2. After illumination, carriers are injected into the small conduction band of the multilayer MoS2 generating photocurrent. The photocurrent increase can be associated with a bias voltage through to the reduction of the carrier transit time (τtransit = l2/μV), where l is the length of the channel, μ is the carrier mobility, and V the bias potential [22].

**Figure 5.** Characterization of infrared photosensors produced with MoS2 samples grown directly on paper substrates at different times, being: (**a**) IxV of samples synthesized at 190 ◦C, (**b**) IxV of samples synthesized at 200 ◦C, (**c**) IxV of samples synthesized at 220 ◦C, (**d**) Ixt of samples synthesized at 190 ◦C, (**e**) Ixt of samples synthesized at 200 ◦C, and (**f**) Ixt of samples synthesized at 220 ◦C. The Ixt measurements were performed with a bias voltage of 4 V, lighting cycles lasting 40 s and power of 20 mW.

The curves in Figure 5b,c, for longer synthesis times, present a strong increase in the current x bias voltage (with a non-linear behavior) for bias voltages higher than ~2V. This non-linearity in the IxV curve is less intense in the samples synthesized at lower temperature, 190 ◦C—Figure 5a, and in the samples with shorter synthesis times for the synthesis temperatures of 200 and 220 ◦C, possibly due to the smaller amount of MoS2 material. Figure 5d–f show that there is an increase in both the dark current and the photocurrent, a result that goes according to the discussion held for Figure 5a–c. The increase in the dark current with synthesis time can be associated with the presence of the metallic phase and trap states.

To evaluate the infrared photosensors, figures of merit Responsivity (R) and Specific Detectivity (D\* ) were used, which can be calculated using the equations *R = (Ion-I0ff)/P* and *D\* = R/*√*((2eI0ff)⁄A)*. Here, Ion is the generated photocurrent in the device under illumination, *Ioff* is the dark current, P is the incident light power on the effective area of the device, *e* is the elementary charge and *A* is the effective surface area. Figure 6 shows that R and D\* values increase with increasing of the synthesis time for three synthesis temperatures and for a bias voltage of 4V. The device built with a sample synthesized at 200 ◦C for 120 min has the highest value for both D\* and R. There is also a big increase of these values for synthesis time longer than 60 min (see yellow curve). The decrease in the R value obtained in the sample synthesized at 220 ◦C for 120 min, when compared to the sample synthesized at 200 ◦C for 120 min, may be attributed to the degradation of the cellulose paper.

**Figure 6.** The Responsivity (**a**) and Detectivity (**b**) according to the synthesis time, for samples synthesized with temperatures of 190 ◦C (red triangles), 200 ◦C (yellow circles) and 220 ◦C (blue squares). Responsivity was measured with a laser power of 20 mW. Graph's inset shows the external quantum efficiency (EQE) values as a function of time and synthesis temperatures of the MoS2 photoactive layer.

Therefore, the experimental data show that the temperature of 200 ◦C and time of 120 min were the best conditions for the production of near infrared photodetectors based on MoS2 active layer grown directly on cellulose paper substrates by the microwaveassisted hydrothermal method. This R value is above the value reported in the literature for photodetectors with active MoS2 layer produced by the hydrothermal method with the same structure. Nahid Chaudhary et al. [55] for example, obtained the maximum R value of 23.8 μA/W for excitation at 635 nm, and Parikshit Sahatiya et al. [28] obtained the R value of 60 μA/W for excitation at 554 nm.

It is important to note that temperatures above 200 ◦C have generally been used in the literature for synthesis of MoS2, or heat treatments have been made after the synthesis to increase the fraction of phase 2H over phase 1T. At 200 ◦C (as in this work), we still have a large proportion of the 1T phase when compared with samples synthesized by the conventional method followed by heat treatment [13,28]. Therefore, we believe that the increase in the 1T/2H interface, caused by the simultaneous and continuous formation of the metallic phase and its conversion to the semiconductor phase, and the consequent increase in the mobility of carriers in the sample, is of great importance to obtain high values of responsivity in MoS2 photodetectors. The interface between the paper substrate and the MoS2 layer may be another important characteristic for devices, once it can be a source of trap states [54] and temperature and time synthesis had a greater effect on the photoconductivity. Thus, the quick increase of the photocurrent to bias voltage (>2 V), as observed in Figure 5, can be associated with the detrapping charge from trap states. It is known that photogain in photoconductors increases as carrier lifetime and decreases as transit time *G = τlife/τtransit* [22].

Figure 6 shows the external quantum efficiency (EQE) of the devices. EQE was obtained by using *EQE(%) = Rhc/(λe)x100*. Here, R is the responsivity, *h* is the Planck's constant, *c* is the speed of light in vacuum, *e* is the elementary charge, and λ is the wavelength of the excitation light. As can be seen, the EQE increases as the time synthesis increases, similar to R and D\*. The device built from MoS2, synthesized at 200 ◦C for 120 min, presents an EQE value of 26%.

The MoS2 photodetector has a relatively high R value, compared to other similar devices based on MoS2 synthesized by the hydrothermal method. The rise and the decay response time has high values (τrise = 3.7 s e τdecay = 4.7 s) which may be associated with the high concentration of the metallic phase in MoS2 and trap states in the MoS2 and/or paper/MoS2 interface. In comparison, the HfO2-gated single-layer MoS2 phototransistors, with high responsivity due to the photogate effect, present a slow response of 0.6–9.0 s. In this case, the slow response time and the high responsivity was induced by the trap states in the MoS2 or MoS2/SiO2 interface [54] Jin et al. [56] deposited Ag NPs on monolayer MoS2 phototransistor and obtained a very high R value and slow response time of 18.7 s, showing other strategy to induce photoconductive effect. The improvement was attributed to the localized surface plasmon resonance of Ag NPs with the increase of light absorption and carrier injection from Ag NPs to MoS2 under illumination. Thus, we believe that, in our device, the surface plasmon resonance in the Ag NPs electrode is light excited and generates hot electrons via nonradiative decay plasmonic resonance, since the Ag NPs show a plasmonic resonance signal in the near-infrared region, as shown in Figure S3 in the supplementary file. The energy of the hot electrons may be greater than the small Schottky barrier between Ag NPs and MoS2, and it can be injected into the MoS2 multilayer generating a photocurrent [57–59]. Additionally, the effect of the MoS2 phases (1T and 2H) on the electrical performances of hybrid PDs was compared by Wang et al. [60] and results indicated that the metallic 1T phase exhibited a high *R* value. However, these devices showed low on/off ratios (<2) and slow response times (0.75 s) due to their metallic conducting nature similar to that of the graphene. On the other hand, the semiconducting 2H phase demonstrated lower *R* value and fast response time (<25 ms). Thus, we believe that our PDs built with MoS2 synthesized at 200 ◦C for 120 min presented higher metallic phases than other devices built with materials synthesized for other temperatures and times, and that the hot electrons from Ag NPs could be contributing to the photocurrent improvement. Therefore, the contact between Ag NPs and the MoS2 multilayer must convert from a small Schottky junction to an Ohmic-like contact when the device is illuminated, as observed in Au nanorods/Perovskite photoconductor [56,61]. The device works as a light illuminated photoconductor, explaining the slow response time and high R compared to a similar device based on a MoS2 synthesized by the conventional hydrothermal method [28,55].

Figure 7 shows the dependence of the photocurrent (*Ion-Ioff*) as a function of the light power on the photosensor synthesized at 200 ◦C for 120 min.

**Figure 7.** (**a**) Current dependence due to the decrease in the power of the light source (laser 980 nm, ON, OFF) on the sensor. The inset shows the relationship between the generated photocurrent and the irradiation power of the light source on the sensor for the synthesized sample at 200 ◦C for 120 min. The bias voltage used here was 4 V. (**b**) Responsitivity, Specific Detectivity, and EQE (inset) under different irradiation powers of the device regarding the synthesized sample with a temperature of 200 ◦C and a time of 120 min.

Figure 7a shows a decrease in photocurrent behavior with decreasing light power intensity. This is an expected result, since with the decrease in the incident light power, there is a decrease in the number of photons reaching the sensor and, consequently, there is a decrease in the number of carriers promoted from the valence band to the conducting band of the MoS2 nanostructures. With the decrease in the number of carriers in the conduction band, there is a decrease in the generated photocurrent (*Ion-Ioff*). An analysis of the relationship between the generated photocurrent and the incident light power on the photodetector is showed in the inset of Figure 7a, where a fitting with power law is used (*I* = *APθ*). Here *A* is a proportionality constant and θ is the exponent that determines the relationship between the photogenerated current and the light source power. From the adjustment, a linear relationship is observed (*θ* = 1.1 ± 0.1). Similar behavior of the photocurrent with light intensity was observed for the single-layer MoS2 phototransitor [54]. Figure 7b shows the relationship between responsivity/detectivity and the light source power used. Note that the highest responsivity and detectivity values (290 mA/W and 1.8 × <sup>10</sup><sup>9</sup> Jones) occur for the power of 2 mW. In the range of 6 to 2 mW, the responsivity and detectivity values increase rapidly with decreasing power (these results indicate that for lower radiation intensity, in the μW range, the photodetector can present very high responsivity values—these results are not showed here). This behavior has been associated with the increase in the recombination probability as carriers photoexcitation increases [59]. EQE measurements (inset of Figure 7b) obtained at various excitation powers show that the EQE gradually increases as the power decreases, similar to the results for the graphene– MoS2–graphene devices (responsivity of ≈ 0.2 A W−<sup>1</sup> and EQE = 55%) [62].

#### **4. Conclusions**

MoS2 nanostructures were grown directly on cellulose paper substrates using the microwave-assisted hydrothermal synthesis method with different temperatures and synthesis times. Variations were observed in the amount of MoS2 grown on top of cellulose fibers as a function of the synthesis parameters. The analysis also showed the distribution of the metallic and semiconductor phases of the MoS2 nanostructures throughout the sample, generating metal/semiconductor interfaces. Near-infrared sensors were produced from the synthesized samples, where the screen-printing technique was used to deposit interdigital Ag NPs contacts in the samples. The results of the electrical characterizations showed an improvement in the performance of the sensor with the increase in nanostructures synthesis time, for the three temperatures of synthesis used. Such results can be explained by the increase in the amount of MoS2, according to the model created for the

sensors operation. Based on the data analysis, the photodetector produced from the sample synthesized at 200 ◦C and 120 min presented a high responsivity value (290 mA/W), a specific detectivity value (1.8 × <sup>10</sup><sup>9</sup> Jones) and an external quantum efficiency of 37%, with response times τrise = 3.7 s and τdecay = 4.7 s. The sensors produced from the samples synthesized at a temperature of 220 ◦C also showed high values of responsivity, however, the high temperature causes substrate degradation, making the device fragile and the application difficult. The improvement in the responsivity value of the photodetector can be associated with the photoconductive effect due to three probably factors: (i) the high concentration of the metallic 1T phase in MoS2 layers; (ii) trap states in the MoS2 and/or paper/MoS2 interface; and (iii) hot electron injection from surface plasmon resonance in the Ag NPs electrodes to MoS2 nanomaterial.

An analysis of the photocurrent dependence as a function of the incident radiation power on the sensor showed a linear dependence of the device's photocurrent in relation to the power of the light source. The responsivity showed a great increase with the reduction of the irradiation power, at the value of 290 mA/W for the power of 2 mW. This result must be associated with the increase in the probability of recombination with the increase in the photoexcited carriers. This work provides a functional example, as well as a promising strategy, to improve the performance of 1T/2H MoS2-based photodetectors. Further studies on the surface state passivation for device optimization, for example, will be conducted to improve photocurrent in MoS2 photodetector synthesized by the microwave-assisted hydrothermal synthesis method.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2076-341 7/11/3/1234/s1, Figure S1: SEM images of the MoS2 nanostructures grown directly on cellulose paper substrates with different temperature and time of synthesis, Figure S2: Micro-Raman mapping and detachment of phases 1T (blue) and 2H (green) in samples of MoS2 nanostructures grown directly on paper substrates with different temperature and time of synthesis, Figure S3: Absorbance spectrum of silver nanoparticles ink in an isopropyl alcohol solution.

**Author Contributions:** Conceptualization, N.J.A.C., C.G., P.B., L.P., E.F., R.M., E.L. and S.A.L.; Data curation, N.J.A.C., C.G., M.J.d.O., D.N., P.B., L.P., E.L. and S.A.L.; Formal analysis, N.J.A.C., C.G., M.J.d.O., D.N., P.B., L.P., E.L. and S.A.L.; Investigation, N.J.A.C., C.G., D.N., R.M. and S.A.L.; Methodology, N.J.A.C., C.G., M.J.d.O., D.N. and S.A.L.; Project administration, P.B., L.P., E.F., R.M., E.L. and S.A.L.; Resources, D.N., P.B., L.P., E.F., R.M. and S.A.L.; Software, N.J.A.C., C.G., M.J.d.O. and D.N.; Supervision, C.G., D.N., P.B., L.P., E.F., R.M., E.L. and S.A.L.; Validation, N.J.A.C.; Visualization, N.J.A.C. and S.A.L.; Writing—original draft, N.J.A.C., M.J.d.O., D.N., L.P., R.M., E.L. and S.A.L.; Writing—review & editing, N.J.A.C., L.P., R.M., E.L. and S.A.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior—Brasil (CAPES)—Finance Code 001.

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

**Informed Consent Statement:** Not applicable.

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

**Acknowledgments:** The authors gratefully acknowledge the financial support from the following Brazilian agencies and programs CNPq, Capes, and Fundação Araucária. This research was also funded by FEDER funds through the COMPETE 2020 Programme and National Funds through FCT—Portuguese Foundation for Science and Technology under project number POCI-01-0145- FEDER-007688, Reference UID/CTM/50025. The authors would like to acknowledge the European Commission under project NewFun (ERC-StG-2014 GA 640598) and BET-EU (H2020-TWINN-2015, GA 692373). The authors would like to thank the Multiuser Laboratory of Universidade Tecnológica Federal do Paraná—Londrina Campus—for the performed analyses.

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

#### **References**


### *Article* **Weekly and Longitudinal Element Variability in Hair Samples of Subjects Non-Occupationally Exposed**

**Pasquale Avino 1,2, Monica Lammardo 3, Andrea Petrucci 4,\* and Alberto Rosada <sup>4</sup>**


**Featured Application: Knowledge of weekly and longitudinal variability of metals in the hair of non-professionally exposed people is of particular importance for understanding the levels/effects of chemicals on workers.**

**Abstract:** Hair is an ideal tissue for tracing the human health conditions. It can be cut easily and painlessly, and the relative clinical results can give an indication of mineral status and toxic metal accumulation following long-term or even acute exposure. Different authors have found outdoor pollution phenomena, such as the levels, significantly alter metal and metalloid hair contents. This paper investigates the element concentration variability in hair samples collected from a not-exposed teenager, neither environmentally nor professionally. The sampling was carried out for one week, and the samples were collected from different locations on the scalp. A nuclear analytical methodology, i.e., the Instrumental Neutron Activation Analysis, is used for determining about 30 elements. Some differences have been found among the samplings as well as between the proximal and distal sections. A deep comparison with other similar studies worldwide present in the literature has been performed for evidencing the relationships and the differences due to different ethnical origins, lifestyles, diets, and climates among the different young populations.

**Keywords:** hairs; variability; week; longitudinal; element; metals; INAA; occupational exposure; unexposed subject

#### **1. Introduction**

During these last few decades, the human biomonitoring through biological fluids (blood, urine) or tissues (hair, nails) has been largely used for the assessment of health effects following an occupational or environmental exposure [1–9], particularly for the absorbed element content. Basically, the researchers have focused their attention on identifying baseline element values in different population samples, living in different areas characterized neither by air/water/soil contamination nor by exposure to chemicals. Different countries conducted large-scale surveys to assess the exposure profile of different populations and better understand serious environmental public health problems [10–14] and to establish the baseline ranges of trace elements in their populations. Other examples are the analysis of trace elements in the U.S population by the U.S. National Health and Nutrition Examination Survey (NHANES) [15], in Canada [16] and in European countries [17–21]. This topic is considered relevant both to obtain finger-print data related to a certain area [22] and, in forensic studies, to the provenance of subjects in a specific site [23–26]. Hairs are also important because they can be used to understand the element variation in case of prolonged intoxication [27–29]. Metals are integrated into hairs during their growth: in this sense the element composition of hairs is the signature of the living

**Citation:** Avino, P.; Lammardo, M.; Petrucci, A.; Rosada, A. Weekly and Longitudinal Element Variability in Hair Samples of Subjects Non-Occupationally Exposed. *Appl. Sci.* **2021**, *11*, 1236. https://doi.org/ 10.3390/app11031236

Academic Editor: Samuel B. Adeloju Received: 27 December 2020 Accepted: 25 January 2021 Published: 29 January 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/).

area and the lifestyle of each person [30,31]. Although there is in literature a great deal of papers regarding the determination of the elemental content in hairs [32–35], to the authors' knowledge, no paper is focused on the element variability during a week. On the other hand, a few studies concern the longitudinal element distribution along human hairs: Yukawa et al. [36] reported the variation of the trace element concentration in long human hairs, showing the profiles according to the distance from the scalp; Kempson and coauthors [37,38] discussed how the exogenous contamination does not influence the levels of some elements (for instance, Zn); on the contrary, Kempson and Skinner [39] report that some other elements, such as Al, are subjected to being accumulated during pollution events and such as Ca, which was demonstrated to be sensitive to endogenous and exogenous contributions [40]; Park et al. discussed the longitudinal association between toenail zinc levels and the incidence of diabetes among American young adults [41]. Finally, Maurice et al. [42] reported a forensic case of poisoning by thallium: the authors developed an analytical method in order to determine the Tl profile all along the hair. The present authors agree with the consideration reported in a recent paper focused on the variation of longitudinal concentration of trace elements in elephant and giraffe hairs [43]. In here, it is highlighted that an important role for this concentration is played by the animals' behavior traits, which suggests that these traits have to be considered also in the study of human hair. Such studies are necessary for having knowledge of endogenous and exogenous roles of hair elements.

Starting from these considerations, this paper would like to investigate the element concentration variability in hair samples collected from a not-exposed teenager, neither environmentally nor professionally. The sampling was carried out for one week, and the samples were collected from different locations on the scalp. Each hair sample was divided into two parts in order to study the section closest (proximal) to the scalp and the most distant (distal) from it. The analyses were performed by Instrumental Neutron Activation Analysis (INAA), and about 30 elements were determined. The use of such analytical technique allowed minimizing the pre-treatment of the samples and hence the relative positive/negative artifacts and performing a multi-elemental determination simultaneously [44–46].

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

#### *2.1. Sample Collection and Storage*

Hair samples were taken over the course of a week from a female young person (10 years old) chosen among the primary school students. The samples were taken in the right, central, and left nape area on alternate days, with strands of about 10 cm (Figure 1a). All samples were taken by cutting the hair in the chosen area as close as possible to the root (Figure 1b). The cut was made by means of stainless-steel scissors, with zero release of elements [12], in order to avoid any possible contamination caused by the friction between the blades of the scissors and the surface of the hair. The hair is light brown in color, and it is not brittle. In the period preceding the sampling, the subject did not wash her hair, did not use any type of treatment, and did not bathe in the sea. The subject is a non-smoker and was not subjected to second-hand smoke; she lives in a medium-small suburban center (9000 inhabitants) of Central Italy, where there are no industrial settlements, whereas farms that employ biological agriculture are far from the site, and the vehicle traffic is limited except on Saturday morning for the weekly food market.

Immediately after collection, the wisp of hair was placed in high-purity Kartell nuclear grade containers. In laboratory, the containers with the wisp of hair were immediately placed in a silica gel dryer and kept in the dark in an environment with a temperature between 20 and 27 ◦C, until the time of pre-treatment and subsequent weighing.

**Figure 1.** Scheme of the sampling area (**a**) and representation of the proximal and distal hair zones analyzed (**b**).

#### *2.2. Hair Pre-Treatment*

The sample pre-treatment was performed by means of a standard procedure, already adopted in previous samplings [12], following a protocol suggested by the International Atomic Energy Agency (IAEA) [47]: washing the samples in 25 mL of ultrapure acetone for 10 min and subsequently rinsing them with ultrapure deionized H2O (resistivity 18 MΩ × cm), repeated three times.

After drying in an oven at 40 ◦C for 15 min, all the samples were placed back in the desiccator, brought back to room temperature, and weighed in containers of nuclear-grade polythene for neutron irradiation. The weighing was carried out by means of the AE 160 analytical balance (Mettler-Toledo GmbH, Greifensee, Switzerland) having a sensitivity of ±0.1 mg.

#### *2.3. Neutron Irradiation and Gamma Measurements*

The neutron irradiation of the samples was carried out in the rotating rack of the TRIGA Mark II research nuclear reactor in the ENEA Casaccia Research Centre. The irradiation time was 24 h, and the neutron flux was <sup>Φ</sup> = 2.34 × <sup>10</sup><sup>12</sup> <sup>n</sup> × cm−<sup>2</sup> × <sup>s</sup>−<sup>1</sup> (corresponding to fluence F = 1.997 × 1017 <sup>n</sup> × cm−2). These parameters were decided in order to detect those elements that, once activated, produce radionuclides whose half-life was between 3 h and several years [48,49].

After the irradiation in the rotating rack, two sets of measurements were carried out: the first one was after 20 h of decay and lasted between 40 min and 1.5 h; the second one was after 15 days of decay and lasted between 20 and 72 h. The position of the samples in the two sets of measurements was vertical at a distance of 4 cm from the detector for the first set and vertical but in contact with the detector for the second set (Figure 2).

The measurements were carried out by an HPGe coaxial Canberra detector with a resolution (FWHM) of 1.88 keV at 1332.5 keV, a relative efficiency of 42.1, and a Peakto-Compton ratio of 65.8:1. The system has a Canberra multichannel analyzer with 8192 channels.

The energy and efficiency calibrations for the different counting geometry were carried out respectively with a source of 137Cs and 60Co and with a source of 152Eu, whose activities were certified by the Centre Communitaire de Référence (CEA).

Primary standards and secondary reference materials (SRMs) were used: as primary standards, single-element solutions at a concentration of 1000 μg mL−1; as SRM material with similar composition to the investigated matrix, the BCR CRM 397 (human hair), just used in a previous study [50], as well as the NIST1515 (Apple Leaves) for an overall checking.


**Figure 2.** Instrumental Neutron Activation Analysis (INAA) master scheme (tir irradiation time; td decay time; t <sup>1</sup> 2 half-life time). In brackets for each product, nuclides are reported as well as some nuclear analytical chemistry parameters, i.e., peak energy (keV), half-life time (m minute, h hour, d day, y year), and limit of detection (LOD).

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

#### *3.1. Quality Control and Quality Assurance (QC/QA)*

The environmental studies regarding the correlation between pollutants and biological tissues, such as heavy elements in human hair, need very sensitive and accurate analytical techniques in order to determine contaminants at trace and ultra-trace levels [51]. Among the different possibilities (e.g., spectroscopy, electrochemistry, etc.), the nuclear methods are still the main available techniques to address such stringent requirements. Their high accuracy and precision and the very low limits of detection (LODs) allow investigating a matrix deeply [52]. Furthermore, the possibility of avoiding chemical–physical treatments or performing radiochemical separations is of fundamental importance for achieving these results. Even if INAA is considered a primary analytical technique—i.e., it is possible to analyze the sample just knowing all the nuclear parameters (such as all the nuclear crosssections of each radionuclide and the nuclear reactor data)—the easier way to analyze a sample is by comparing its activity with that of a standard irradiated in the same conditions. The comparison between the data analysis obtained by INAA and the certified values is the first step for assessing the quality assurance and the quality control (QA/QC). For this aim, standard reference materials (SRMs) such as BCR CRM 397 (Human Hair) and NIST 1515 (Apple Leaves) were chosen according to the matrix similarity and biological origin. Table 1 reports the differences (Δ expressed as %) between our data and the certified values of both the SRMs for each element.

It should be noted that the Human Hair SRM shows a Δ below 6.5%: Se and Zn show high precision and accuracy as well as the indicative or informative element values (i.e., As, Co, Cr, Cu, Fe, Hg, and Mn). The comparison profile for the Apple Leaves is slightly different: among the certified elements, good Δ (below 16%) are achieved for Ba, Ca, Fe, K, Mn, Na, Rb, Sr, and Zn, as also reported in previous paper [53], whereas Hg and Ni show quite relevant differences such as −48.4% and −21.9%, respectively. For Ni, the reason is due to the poor INAA sensitivity. Mercury shows two different results: for Human Hair SRM, the difference between measurements is 5.7%, whereas for Apple Leaves, the SRM is −48.4%. According to our evaluation, the cause has to do with the different Hg levels in

the two SRMs. In the first SRM, Hg is at 12.3 μg g<sup>−</sup>1, whereas the authors find 13.0 μg g−<sup>1</sup> with a coefficient of variation (CV%, defined as the ratio between standard deviation and mean value × 100) of 6.1, a good value according to the references [49,54,55]. In the second SRM, the certified Hg value is 0.0432 μg g−<sup>1</sup> (or 43.2 ng g−1), whereas the authors find 0.0223 μg g−<sup>1</sup> with a CV% of 45.7, which is an unsatisfactory value according to the ref [54]. The second SRM shows an Hg content almost 300 times lower than the first SRM: this occurrence justifies the high Δ between ours and the certified data in the second SRM.

**Table 1.** Analytical standard comparison (mean <sup>±</sup> s.d.; <sup>μ</sup>g g−1) between our data and certified values: BCR CRM 397 (human hair) and NIST-SRM 1515 (Apple Leaves).


<sup>1</sup> Product nuclide; <sup>a</sup> indicative values expressed as μg g−1; <sup>b</sup> informative values expressed as μg g−1; n.d.: not detected; Δ: difference (%) between our mean values calculated and certified one as (*ourvalue*−*certi fiedvalue*) *certi fiedvalue* × 100. The standard deviation is calculated among five replicas.

#### *3.2. Element Content in Human Hair Samples*

Before performing the analysis, the authors worried about the effects of cleaning and cutting hair before sampling. The problem of hair cleaning is largely discussed as well as the effect of washing using different procedures, as reported in the literature [56–62]. The IAEA recommends a cleaning procedure for hair [47]. Frequent head washing does not affect (the significance limit greater than 0.05) the content of some elements (Br, Co, Cu, Mn, Se, Zn) or of some pollutants (Ni, Hg); on the other hand, As does get washed out but with no such great amount (lower by 1.7 times) [39,47,63]. The hairs were cut by metallic scissors: before that, the oxides were carefully removed. Any effect of the friction between blades and hair shaft were limited in order to prevent possible sample contamination.

Table 2 shows the level of 24 elements measured in the samples investigated along with the minimum and maximum values and the CV%. First, two important elements, i.e., As and Hg, considered hazardous elements for the human health and of exogenous origin, are below the corresponding LODs: this is a preliminary confirmation that the subject is not exposed to sources of As and Hg. Other considerations could be drawn about essential elements, i.e., Cr, Cu, Fe, Mn, Se, and Zn: their variability, expressed as CV%, is below 40% except for Cr and Cu, which have 66.2% and 68.8%, respectively.


**Table 2.** Element levels and single daily concentrations (μg g<sup>−</sup>1) determined in all the investigated independent samples (five independent measurements).

<sup>1</sup> s.d. standard deviation.

Br, Ca, K, and Na, the last three considered labile elements (because they are strongly influenced by washing [64]) show a low CV% below 27%, confirming that the hair was not washed. Silver and gold are present at low concentrations, 740 and 22 ng g<sup>−</sup>1, respectively: their presence could suggest a previous use of shampoos containing nanoparticles of these two elements into the composition for antimicrobial activity [65,66]. Other elements such as La, Rb, Sc, Th, and W can be considered of environmental origin: their CVs% are high, above 60%, especially Th, up to 200%. Finally, Ba, Br, and Sb show low CVs%: their presence and levels could be due to anthropogenic pollution and particularly to airborne particulate matter, as just evidenced by authors in previous papers [46,49].

Table 3 shows the concentration trend (the number of samples per nuclide is too low only three scalp locations and two longitudinal positions—to obtain any actual reasonable trend that could be considered real and not obtained by chance) in the areas of the nape where the sampling was carried out (left, center, and right) as well as the longitudinal variation of the concentrations along the hair.

If the trends of Br, Ca, K, and Na from Tables 2 and 3 are taken into account, similar concentrations are noted for Br both during the days and in correspondence with the sampling zones and in the longitudinal variation: mean value 26.5 ± 3.46 <sup>μ</sup>g g−1; 1st day 25.0 ± 3.0 <sup>μ</sup>g g<sup>−</sup>1; 2nd day 24.1 ± 0.6 <sup>μ</sup>g g<sup>−</sup>1; 3rd day 30.4 ± 2.3 <sup>μ</sup>g g−<sup>1</sup> (Table 2); left zone 27.1 ± 2.1 <sup>μ</sup>g g<sup>−</sup>1; central 23.7 ± 0.45 <sup>μ</sup>g g<sup>−</sup>1; right 28.8 ± 1.7 <sup>μ</sup>g g<sup>−</sup>1; hair proximal section 26.5 ± 2.6 <sup>μ</sup>g g<sup>−</sup>1; distal 26.5 ± 4.9 <sup>μ</sup>g g<sup>−</sup>1. The same stability in the concentration trend is evident for Na, Ca, and K. These four elements, i.e., Br, Ca, K, and Na, are ubiquitous in the environment; besides, Ca, K, and Na are among the fundamental components of the tissues and biological fluids in the human body, and K and Na are also present in the body secretions (exudate, etc.). In the samples, Br, K, and Na show an increasing concentration trend between the first and third day (Table 2). Similarly, the same trend is shown, but in a much more marked way, by elements of environmental origin such as La, + 47.5%

increase in concentration on the third day compared to the first; Rb, + 216%; Th, + 98%; and W + 158%. The authors would like to underline that a fundamental requirement of this study was to have unwashed hairs, in order to understand the natural element levels in human hair. Confirming this statement (i.e., unwashed hair), a similar trend is found for all the elements mentioned above, with the exception of bromine which remains constant, whereas greater increases are found for the elements of environmental origin.

**Table 3.** Element concentrations (μg g−1) in the nape different areas and along the hair (proximal and distal zones), (five independent measurements).


<sup>1</sup> s.d. standard deviation.

A similar trend is shown by the daily trend (the number of samples per nuclide is too low—only 3 days—to obtain any actual reasonable trend that could be considered real and not obtained by chance) of some elements considered essential, such as Cr, Fe, and Mn; for the last two, the increase is significant: + 74.5% and + 50.3% respectively. Cr and Fe increase their level in the distal area with respect to the proximal one, and a similar trend is also shown by Se and Zn, which, instead, have a daily trend with a maximum on the second day.

However, there are elements that maintain a constant concentration both with regard to the daily trend (Table 2) and in the nape sampling areas and in the distal and proximal ones (Table 3), as highlighted by the coefficient of variation (CV%) calculated taking into account all the concentrations obtained. These elements are the ones present either in the body's tissues or in the biochemical systems, such as Ca, K, Na, and Br (CV% 15.2%, 16.6%, 7.76%, and 8.64%, respectively), and those considered essential such as Mn, Se, and Zn (15.7%, 7.73%, and 11.4%, respectively) and those considered pollutants of anthropogenic origin: Ba, Sb, and Co (10.7%, 18.7%, and 19.2%, respectively). For the last ones, it is possible to hypothesize a hair contamination deriving from their levels in the environment, whereas the level stability of the essential elements is due to regular biochemical systems/reactions. This occurrence allows defining basal elements concentration levels in the hair analysis both in

good health conditions and in "anomalies" caused by occupational and/or environmental exposure.

On the other hand, the concentration of elements of environmental origin (i.e., La, Rb, Sc, Th, and W) is less stable: their CVs vary between 38.4%, tungsten, and 126.9%, thorium. The hypothesis that can be advanced is that their presence in the environment, and therefore their levels in the hair, is connected to the variation of local climatic conditions: wind direction, wind speed, temperature, humidity, etc.

A trend similar to that of the elements of environmental origin is also shown by two chemicals considered essential, i.e., Cr and Cu, and, albeit to a lesser extent, by Fe (CV% 45%, 53.2%, and 23.5%, respectively). For these three elements, the hypothesis of a double origin could be put forward; i.e., there is an overlapping between the element basal levels (i.e., the concentration levels naturally present in the human body) and the concentrations deriving from environmental pollution. This hypothesis can be confirmed because these three elements (i.e., Cr, Cu, and Fe) significantly increase their concentrations in the distal area of the hair (i.e., the section more in contact with the environment) compared to the proximal area (Table 3), even if their average concentration level agrees with hair data reported for Rome and Italy (Table 4). Finally, it is interesting to note that some elements of environmental origin but coming from anthropogenic pollution and considered particularly toxic such as As, Hg, and Ni were not revealed by INAA, i.e., they show levels below the LOD (Figure 2). Therefore, it can be assumed that As, Hg, and Ni are < 1 ng g−1, < 5 ng g<sup>−</sup>1, and < 80 ng g−1, respectively. Very low levels of Eu and Tb, considered Rare Earth Elements (REEs) of environmental origin, were measured just in one hair sample at concentrations of 1 and 20 ng g−<sup>1</sup> respectively (LOD for Eu 0.03 and for Tb 0.3 ng g<sup>−</sup>1).

**Table 4.** Concentration (μg g<sup>−</sup>1) comparison between these data and those found in other Italian and international studies.


*3.3. Comparison with Studies on Adult and Teenager Population*

Table 4 shows a comparison between these data and similar levels found in a previous study performed in Rome [50].

First, the presence of elements such as Ag, Au, and Co can be due to cosmetics and/or personal hygiene products. In particular, Ag shows concentrations comparable with the data of the Rome group (0.740 ± 0.271 <sup>μ</sup>g g−<sup>1</sup> versus 0.40 ± 1.58 <sup>μ</sup>g g−1) [50] and the Italian population (0.83 ± 1.96 <sup>μ</sup>g g<sup>−</sup>1) [67] as well as Au (0.022 ± 0.014 <sup>μ</sup>g g−<sup>1</sup> vs. 0.036 ± 0.038 <sup>μ</sup>g g−<sup>1</sup> for the Italian population) [50], whereas Co shows decidedly lower levels than the Italian population (0.041 ± 0.013 <sup>μ</sup>g g−<sup>1</sup> vs. 0.145 ± 0.133 <sup>μ</sup>g g<sup>−</sup>1), but they fall within the international data (0.002–15 μg g<sup>−</sup>1) [68–76].

Levels of Ca, K, and Na are in line with the data of the Rome and Italy groups. In particular, Ca and K show similar levels to the light brown hair in the Rome group [50] (Ca: 2010 ± <sup>512</sup> <sup>μ</sup>g g−<sup>1</sup> vs. 1776 ± <sup>1776</sup> <sup>μ</sup>g g−1; K: 249 ± <sup>66</sup> <sup>μ</sup>g g−<sup>1</sup> vs. 258 ± <sup>280</sup> <sup>μ</sup>g g−1) with a much lower standard deviation. Similar considerations can be also put forward for K, which are compatible with the relevant standard deviation of both Rome and Italian data. Br is instead at concentration levels equal to more than double the concentration data of all the Rome and Italian samples, even though it is slightly above the variability of the Italian data (26.5 ± 3.5 <sup>μ</sup>g g−<sup>1</sup> vs. 12.9 ± 10.1 <sup>μ</sup>g g−1) but it falls within the international data range (0.15–490 μg g<sup>−</sup>1).

Among the essential elements, Cr, Cu, Fe, Se, and Zn show good agreement with the data of Rome and Italy groups, whereas Co presents lower levels than those of Rome and Italy groups and Mn higher than the ones of the Rome samples.

As for the elements of environmental origin, the levels of Rb, Sc, and Th are in full agreement with the data reported in Table 4, whereas W is absent in all the other comparison samples, and La is at significantly higher levels. For La, as well as for Mn and Br, all the concentration data are higher than the comparison values [50]: 0.957 ± 0.592 <sup>μ</sup>g g−<sup>1</sup> vs. 0.038 ± 0.031 <sup>μ</sup>g g−<sup>1</sup> in the Italian reference group, i.e., 25 times higher; Mn 3.75 ± 1.24 <sup>μ</sup>g g−<sup>1</sup> vs. 0.42 ± 0.32 <sup>μ</sup>g g−<sup>1</sup> in the entire series of the Rome samples, 9 times higher; Br 26.5 ± 3.5 <sup>μ</sup>g g−<sup>1</sup> versus 12.9 ± 10.12 <sup>μ</sup>g g−<sup>1</sup> in the Italian group, 2 times higher. This anomaly can be explained by environmental contamination, since these three elements are present in the urban atmosphere, especially in airborne and dust depositions [77,78]. In previous papers [79,80], the authors report data on the elements in PM10 and PM2.5 airborne sampled in downtown Rome along with the trends during the last three decades: both La and Mn and Br show significant levels (particularly, La 170 ± 101 pg m−<sup>3</sup> in PM10; Mn 60 ± 44 ng m<sup>−</sup>3, Br 17.1 ± 13.9 ng m−<sup>3</sup> in PM2.5).

Table 5 shows a large comparison among our data and data collected from teenagers (boys and girls) worldwide [81–91]. The table also reports some reference data from Korean [92], Canadian [93], and American [94] teenagers hair values. First of all, it could be seen that our data are broader and more accurate than those determined in other studies, considering the high performance of the analytical method used, i.e., INAA. This occurrence allows drawing only few significant considerations on the content of some metals: Ca (from 2 to 10 times) and Mn (up to 10 times; ours and some other data are similar, an exception is present in the data collected from girl teenagers in Rome, i.e., 69.2) are higher than those reported in the other studies [81–91] and in the reference values contained in [92–94], whereas Cr and As are always less than those reported in the literature. Finally, Ag, Cu, Fe, Se, and Zn show levels close to those reported worldwide.

For a better correlation and significance of this comparison, the authors have reported the correlations in Table 6 and the relative plots in Figure 3 between our data and those determined in girl teenagers hair worldwide [81–91]. As it can be seen, there is high correlation (above 0.9) with the data found, especially in West Europe (different Italian locations, Spain), and low correlation (ranging between 0.1 and 0.6) with those determined in Artic area (Arkhangelsk) and Korea, for this latter also for boys. This occurrence can be interpreted considering the different ethnical origins, lifestyles, diets, and climates (including the presence of different and massive anthropogenic and/or natural sources) among the different young populations.




*Appl. Sci.* **2021** , *11*, 1236

**Table 5.** *Cont.*


**Table 6.** Linear regression and coefficient of determination (R2) between our data and each dataset for different girl teenagers hair determined worldwide [81–91].

**Figure 3.** Correlation plots among our data and similar data determined in girl teenagers' hair worldwide [81–91].

#### **4. Conclusions**

The data determined in this study are a first tentative study of the element profile both in different hair sections and in different days of a week, also in relationship to the anthropogenic and/or natural sources present in an area, specifically in a suburban area in Central Italy. The element content in the hairs of a girl is compared with studies reporting the relative levels in hairs collected from girl teenagers worldwide: the diverse correlations found with West Europe and East Europe/Central Asia values highlight differences depending on habits, lifestyles, and nutritional diets.

**Author Contributions:** Conceptualization, A.R. and P.A.; methodology, M.L. and A.R.; software, A.P. and M.L.; validation, A.P. and A.R.; formal analysis, M.L. and A.R.; investigation, P.A.; resources, A.P. and M.L.; data curation, P.A.; writing—original draft preparation, A.R. and P.A.; writing—review and editing, A.P., A.R. and P.A.; visualization, A.R.; supervision, A.P. and P.A. All authors have read and agreed to the published version of the manuscript.

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

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board of ENEA (approval June 2006).

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

**Data Availability Statement:** Data is available upon request by contacting the corresponding author.

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

#### **References**


#### *Article*
