Front-face ﬂ uorescence spectroscopy of tryptophan and ﬂ uorescein using laser induced ﬂ uorescence and excitation emission matrix ﬂ uorescence

Herein we study the responses of two standard molecules, tryptophan and ﬂ uorescein, over a large concentration range in boric acid via two front-face ﬂ uorescence spectroscopies: excitation emission matrix ﬂ uorescence and laser-induced ﬂ uorescence. The responses in relation to the tryptophan content were measured and were found to depend on the measurement technique. For ﬂ uorescein, no response was found for the excitation emission matrix, and an inverse relationship with the solid content was observed for the measured laser-induced ﬂ uorescence. Aiming to enhance the very weak ﬂ uorescence response of ﬂ uorescein, a novel protocol using a gel interface was established. These methods represent a future premise to investigate molecules with low ﬂ uorescence e ﬃ ciency on surfaces using a logarithmic response. components for the EEM ﬂ measurement of ﬂ uorescein (FLU) in the presence of the gel layer.


Introduction
Front-face uorescence spectroscopy (FFFS) is a powerful method to detect and determine uorescent molecules in solid matrices without contact and sample preparation. It allows rapid and efficient characterization and quantication in pharmaceutical or alimentary domains. In environmental or agricultural applications, this technique could be a powerful tool for pollutant or nutrient quantication. However, FFFS was developed essentially in the agronomic domain using the uorescence of the chlorophyll family to determine plant health, 1 for lake sediment surveys 2 and for red wine monitoring. 3 Very few applications were based on other samples such as soils, 4 sewage or compost. 5 The application of FFFS in the environmental domain is under development because the samples present complex mixtures of several mineral matrices as well as complex and simple organic molecules. Hence, before using FFFS as a routine technique for heterogeneous surfaces in environmental systems, it is necessary to study several key parameters including diffusive effects, color or absorbance, the chemical and physical states of surfaces, and theoretical models. These parameters cannot be treated simultaneously, and simpler systems must be studied rst to progressively increase the complexity. This work addresses the relationship between the FFFS responses and the molecule content in a solid. Compacted boric acid or simple powder mixture matrices were studied. Tryptophan, a well-known and wellstudied molecule, and uorescein, a molecule used as a uorescence standard for solutions, were chosen to investigate two FFFS techniques: laser induced uorescence (LIF) and emission excitation matrix uorescence (EEM). Tryptophan is a uorescent amino acid and is present in all proteins and consequently, in every biological material. Its uorescence properties depend on its chemical environment and its structural location. 6 Whatever its state, solid or in solution, the more efficient excitation wavelength domain is 250 to 280 nm, and the emission wavelength domain ranges from 310 to 380 nm, 6-8 depending on the chemical conditions. Pure tryptophan is a white yellowish solid. Fluorescein uorescence in solution occurs in the excitation wavelength domain of 440 to 500 nm and its emission occurs from 510 to 516 nm without an emission wavelength shi except under extreme conditions. 9,10 Solid uorescein is a dark red powder. No previous report was found on the solid-state uorescence of uorescein. In this work, tryptophan and uorescein were studied in the solid state over a wide range of content, from 0.0002% to 7.618% and from 0.002% to 77.92% for uorescein and tryptophan, respectively. Fluorescence emission was measured by FFFS with LIF and EEM to establish the relationship between the molecule content and uorescence response in a solid boric acid matrix.

Sample manufacturing
Boric acid powder (BA, Fisher Chemical) was used to prepare solid pellets with different tryptophan (TRP) and uorescein (FLU) content. An initial solid mixture was created by nely grinding a boric acid mass with a TRP or a FLU mass, giving the highest solid content of C0-TRP and C0-FLU, respectively (Table  1). Aer this, known masses of C0-TRP and C0-FLU were added to another mass of BA to prepare the next set of samples C1-TRP and C1-FLU, respectively, and so on, to obtain a nearlogarithmic decrease of TRP and FLU content in the sample set ( Table 1). The zero content pellet was obtained using only the BA solid. For each solid mixture, one part of the powder was pressed at 8 tons per cm 2 to obtain two solid pellets. Eventually, measurements were performed on the pellets and powders for 13 and 14 different content samples ranging from 2 Â 10 À3 to 78% for TRP and 2 Â 10 À4 to 7.6% for FLU, respectively. The content in percentage of mass for TRP and FLU is denoted as % TRP and % FLU, respectively. All solids were conserved in a dark, dry, enclosed plastic bag to avoid contamination and degradation. All amounts and references are given in Table 1.

EEM acquisition
EEM measurements were taken on a Perkin Elmer LS50B spectrouorimeter using a solid holder module while measuring the pellet under quartz glass. The excitation range was from 200 to 400 nm varying every 10 nm with the excitation slit xed at 5 nm. The corresponding emission spectra were acquired from 200 to 600 nm with a scan speed of 250 nm min À1 and a slit of 5 nm. No emission lter was used. The response integration time and the photomultiplicator tension were both set as automatic. The 10 by 10 nm extractions of the excitation emission matrices were obtained via the 3D export tool from the Perkin Elmer program FLWinLab.

PARAFAC analysis
To determine the component's source contribution to the EEMs, the CP/PARAFAC algorithm was used. First all EEMs were cleaned from the scattering signals: the Rayleigh by cutting the scattering band (20 nm) and the Raman from rst and second diffraction orders by applying the Zeep procedure. 11 Decomposition was then performed from two to six components on the 28 corrected EEMs, and the higher number of components giving a CORCONDIA test over 60% was selected as the optimal component number. 12,13 No outliers were present in the data and 3 components were found. Because the uorescence gives a response depending on the apparatus, the component contribution was noted by relative contribution to the uorescence (RCF). The RCF of a component should be related to the molecule content and is comparable from one sample to another. 12 The CP/PARAFAC model components were indexed as CX TRP-FLU , where X is the component number and TRP-FLU indicates the name of the database.

LIF measurement
The laser-induced uorescence was measured with a portable LIF system (lab-made equipment developed by EMBRAPA Instrumentation, Sao Carlos, Brazil) equipped with a 405 nm laser source. 14 An optical ber was coupled with a solid sample holder. Each face of each pellet was measured under the following conditions: emission wavelength range from 450 to 800 nm, integration time 1 s, boxcar 4, and average 5. For each measurement, the integration of the uorescence emission spectra was denoted as A X LIF expressed in arbitrary units of uorescence, a.u. by nm, where X is the studied molecule TRP or FLU.

Gel interface measurements
A gel interface was tested to enhance the uorescence response of the solid state. A hydroxy-gel was synthesized on the same day of the measurement. Tetraethylorthosilicate (TEOS) was added to HCl until the pH ¼ 3. Hydrolysis was performed under gentle stirring for 36 h. Aer this, ethanol was added to the solution, and the gel was extracted by centrifugation, cleaned with ethanol and conserved in a glass container at 4 C in the dark. Two grams of the lump gel was placed on the quartz glass of the sample holder. By applying gentle pressure with the counter piece, the gel sliced and became transparent. Aer removal of the counter piece, the gel remained as a layer, sticking to the quartz windows, with slight cracks. Ten milligrams of powder was then deposited on the gel surface, and the system was pressed again until the gel cracks disappeared. The gel layer width was approximately 1 to 2 mm following pressure application. During the stabilization time, the emission uorescence (l ex /l em ¼ 490/510 nm) was measured every minute to monitor the response evolution. This protocol was performed for the 14 pellets containing uorescein and the boric acid pellet.

Direct solid EEM decomposition
The CP/PARAFAC decomposition of the fourteen EEM measurements of the TRP and FLU pellets gives three independent components (CORCONDIA 81.71%) presented in Fig. 1. Component C1 TRP-FLU (Fig. 1a) is dened by a peak maximum at (l ex /l em ) ¼ 250/325 nm with an ellipsoidal shape. Component C2 TRP-FLU (Fig. 1b) is slightly shied compared to C1 TRP-FLU with a peak maximum at (l ex /l em ) ¼ 280/380 nm and a sharper ellipsoidal shape. These two locations are near the uorescence of tryptophan in solution. [6][7][8]15,16 One can note that the C1 TRP-FLU excitation range is narrower than that of C2 TRP-FLU . The third component, C3 TRP-FLU (Fig. 1c), shows a horizontal ellipsoidal peak centered on (l ex /l em ) ¼ 210/400 nm with a large emission range from 320 to 500 nm. C3 TRP-FLU seems to be related neither to uorescein nor tryptophan but to a non-specic solid surface response. 17 The relationship between components C1 TRP-FLU and C2 TRP-FLU with the TRP content in percent ( Fig. 1d and e) shows a rapid increase of the response from 0 to 1% and a saturation from 1% to 80% for the tryptophan content. The same behavior was observed for C2 TRP-FLU with a higher RCF for the powder than for the pellet. Moreover, this difference persists at the low range of tryptophan (0 to 1%), and it is obvious that a linear relationship exists only for the powder state with a good correlation coefficient for both C1 TRP-FLU (r 2 ¼ 0.93, slope ¼ 2355 RCF% of TRP À1 ) and C2 TRP-FLU (r 2 ¼ 0.94, slope ¼ 610 RCF% of TRP À1 ) as shown in the insets of Fig. 1(d) and (e), respectively. For the pellets, the RCF directly plateaus just below the saturation value ( Fig. 1d and e). The only difference between the powder and the pellets is the solid surface texture, and reection or diffusive effects could explain this difference. Concerning C3 TRP-FLU , which was attributed to the solid state, it presents no variation (r 2 ¼ 0.38, slope ¼ 103 RCF% of TRP À1 ) (Fig. 1f), except for the low % of TRP that could be attributed to an incomplete separation of the TRP uorescence by CP/PARAFAC; i.e., TRP uorescence emission occurs up to 380 nm and this could inuence the decomposition, leading at low content to a linear relationship with the % of TRP.
Direct solid LIF spectra LIF spectra measurements are presented in Fig. 2a-d; each individual spectrum is the average of the four measured spectra from the two pellets (two faces for two pellets of the same TRP content) and two measured spectra for the powder (duplicate). There is little difference in the spectrum shapes of the powder and pellet states for the signal of TRP (Fig. 2). The correlation between the pellet and powder responses is very strong; r 2 ¼ 0.94 and r 2 ¼ 0.95 for the TRP area and maximum intensity, respectively, whereas it is less correlated for FLU, where r 2 ¼ 0.83 and r 2 ¼ 0.84 for the area and maximum intensity, respectively. This means that for TRP, the solid state does not inuence the response, whereas for FLU, it is more signicant. Regardless of this, the uorescein response presents a maximum intensity, I FLU LIF,Max , at 520 nm and several slight shoulders at 480 nm, 575 nm, and 620 nm. This emission location, 520 nm, is slightly red shied compared with the FLU emission in solution, which is at 510 nm. Concerning tryptophan, the maximum intensity, I TRP LIF,Max , is centered at 490 nm, and only one shoulder appears at 520 nm. This emission peak  position for TRP was not expected compared to the emission maximum for TRP in solution, which is between 310 nm and 380 nm. The variation of the area under the curve A FLU LIF and A TRP LIF versus percent of chemical, and the corresponding log(A FLU LIF ) and log(A TRP LIF ) versus the logarithm of the chemical content, are described in Fig. 2 for both powder and pellet states. For FLU, there is little difference between the pellet and powder responses. Hence, for pellets or powders, A FLU LIF decreases exponentially with increasing FLU content; this is unexpected behavior for a specic uorescence response. This could be because the uorescein solid state color induces a strong inner lter effect. However, because uorescence is isotropic, some part of the uorescence emitted by molecules located at the surface should be observed because it irradiates directly out of the solid. Another hypothesis could be drawn by understanding this phenomenon as a strong quenching effect by the neighbor of the uorescein molecules, either the boric acid or the uorescein itself. This hypothesis could be supported by the linear correlation (A FLU LIF ¼ 50 010 Â % of FLU + 408) observed at a very low content of FLU (<0.1%), which is strong (r 2 ¼ 0.95). The tryptophan LIF response, A TRP LIF , shows a logarithmic trend curve with a lower response for the powder than for the pellet state in Fig. 2. This saturation phenomenon could be explained by an inner lter effect like in solution 18 but must be theoretically demonstrated for solids. However, for the low % of TRP, the relationship is linear until log(% of TRP) ¼ À2. Below this % of TRP, A TRP LIF does not vary when the blank response is reached. By using a linear regression it is possible for the tryptophan to determine a detection limit, in the solid boric acid matrix, of 0.022% tryptophan in mass.

EEM with gel interface results
The emission spectra of uorescein extracted using a gel are shown in Fig. 3(a), for three different stabilization periods (0, 10 and 35 min) aer gel contact and for the uorescein powder C8 (0.2022% uorescein in mass). During the diffusion of the uorescein in the gel, the intensity increases and there is no shi in the peak position. Fig. 3b presents the maximum uorescence intensity (510 nm) normalized to the initial intensity (t ¼ 0 s), a few seconds aer the deposition of the powder in the gel, as a function of time, except for the C3-FLU content (550 nm), owing to the saturation of the detector at the maximum wavelength. For the C3-FLU content, the intensity increases slightly and decreases regularly with time (Fig. 3b, black squares). This is due to the inner lter effect of uorescein, which becomes highly concentrated in the gel. For lower concentrations (from C8-FLU to CD-FLU), the stabilization follows a logarithmic rule, and the stabilization time is reached more rapidly when the content is low. For the CD-FLU content (Fig. 3b, light-gray triangles), the intensity is very low, and the signal is too noisy to detect a stabilization trend. Considering this stabilization time, the FFFS-EEM uorescence measurement was performed aer 1 h of stabilization to ensure that there was no evolution with time that inuenced the results. The system was then stabilized for 1 h, and the FFFS-EEM measurement was taken with the same parameters described previously.
The CP/PARAFAC decomposition of all the gel-extracted uorescein samples gives three components (CORCONDIA 88.51%), which are presented in Fig. 4. Component C1 FLU-GEL shows a large domain of excitation with a maximum at 400 nm and a straight emission band centered at 510 nm. It is slightly blue shied compared to the emission range of the uorescein in solution. The second component, C2 FLU-GEL , presents an excitation maximum at 450 nm and the same emission as C1 FLU-GEL . The range of excitation for C1 FLU-GEL and C2 FLU-GEL of the uorescein in the gel is not consistent with those found in the literature, 480-500 nm. This blue shi could be explained by conformational inuence due to the gel network. However, the redshi excitation position of C2 FLU-GEL compared to that of C1 FLU-GEL could be attributed to the inner lter effect for the uorescein with high concentration. 18 Indeed, it was not possible to apply an inner lter effect correction to the FFFS measurements because the surface absorbance measurement is difficult and because no theoretical model allows this correction. When the concentration of the uorescein increases in the gel, the auto-absorption process shis the excitation peak to higher wavelengths such as those that occur in solution. This redshi of emission wavelength, normally observed when the uorescein is in solution, does not occur in this experiment. The C3 FLU-GEL component uorescence maximum is located at the (l ex /l em ) ¼ 200/400 nm position. It seems that this component is similar to C3 TRP-FLU and specic to the FFFS-EEM technique and not related to the samples in this work.
The RCF of C1 FLU-GEL , C2 FLU-GEL and C3 FLU-GEL aer 1 h of stabilization is presented in Fig. 5. The RCF of C1 FLU-GEL is linear for the low content of uorescein of <0.02% in mass (Fig. 5, inset, r 2 ¼ 0.99) but dramatically decreases aer this  content limit (Fig. 5a). The C2 FLU-GEL component contribution increases with increasing uorescein content. However, there is a difference of slope between the lower content and the higher content ( Fig. 5b) owing to the very low contribution of C2 FLU-GEL from 0 to 0.005% of uorescein in mass. In fact, assuming that C1 FLU-GEL and C2 FLU-GEL are related to the two responses of uorescein in the presence of the inner lter effect in the gel, it can be considered as the sum of contributions between C1 FLU-GEL and C2 FLU-GEL . Hence, the relationship clearly presents two lines of correlation, one under 0.022% of uorescein (r 2 ¼ 0.96, slope ¼ 14 825 RCF% of FLU À1 ) and another one over this limit, with a lower correlation coefficient and slope (r 2 ¼ 0.87, slope ¼ 630 RCF% of FLU À1 ) (Fig. 5c). The C3 FLU-GEL component is unrelated to the uorescein content and presents a non-linear increase for the rst point, aer which it decreases regularly to reach an average contribution of 500 RCF (Fig. 5d).

Conclusions
The solid state uorescence responses of tryptophan and uorescein in boric acid depend on the spectroscopic technique and content. For tryptophan, the FFFS-EEM gives a response that ts well with the uorescence spectral location of tryptophan in solution and it conrms the possibility of measuring this compound directly in solid samples. The CP/PARAFAC decomposition in a data-set of EEM with a mass content between 0 and 80% of uorescein or tryptophan gives three components. Two are highly related to the tryptophan uorescence and give a linear response for low content (0 to 1% of TRP) with a saturation of the contribution over 1%. The relative contribution to the uorescence response of the components at a low content of TRP clearly depends on the solid state: the powder gives a linear response, whereas the pellets saturated the detector, probably due to the diffusive or reective effects. The FFFS-LIF response of tryptophan shows a linear trend with different slopes for the powder and pellet forms. However, the LIF uorescence occurs at an unusual wavelength (l ex /l em ) ¼ 405/495 nm for tryptophan and further investigations have to be performed to understand which physical processes are involved. For uorescein, no response was obtained for the FFFS-EEM technique, whereas using FFFS-FIL, a decreasing spectral response was obtained in the emission wavelength range for uorescein in solution. This behavior could be due to a very strong inner lter effect attributed to the uorescein color (red-brown), and this hypothesis should be tested.
By interposing a silicon gel between the quartz glass and solid powder, it was possible to obtain a uorescein response using the FFFS-EEM technique. The CP/PARAFAC components show a uorescence peak at the same emission position as uorescein in solution, but with a 90 nm blue shi for the excitation wavelength. A piecewise linear relationship was observed with a steep slope from 0 to 0.02% of uorescein in mass and a less steep slope over this limit. No saturation of the signal was observed with the gel interface during this experiment. For all FFFS-EEM measurements, a component located at (l ex /l em ) ¼ 200/400 nm was found aer CP/PARAFAC separation, which can be attributed to the solid-state response or to the apparatus response. For uorescein, it was possible to measure some uorescence by FFFS-EEM using a silicon gel layer, but further investigations should be performed to obtain quantitative results. However, even without preparing a solid surface measurement, this gel method could be developed to easily and rapidly measure heterogeneous surfaces like soils or coated materials that normally exhibit weak or no uorescence.

Conflicts of interest
There are no conicts to declare.