An Improved Methodology for Determination of Fluorine in Biological Samples Using High-Resolution Molecular Absorption Spectrometry via Gallium Fluorine Formation in a Graphite Furnace

Nowadays growing attention is paid to the control of fluorine content in samples of biological origin as it is present in the form of various biologically active organic compounds. Due to the chemically-rich matrix of biological tissues, the determination of fluorine becomes a very difficult task. Furthermore, a required complex sample preparation procedure makes the determination of the low contents of F by ion chromatography UV-Vis or ion-selective electrodes not possible. High-resolution continuum source graphite furnace molecular absorption spectrometry (HR-CS GF MAS) seems to be the best option for this purpose due to its high robustness to matrix interferences, especially in the presence of carefully selected modifiers. In this work the possibility of quantitative F determination in water and animal tissues was examined by measuring the molecular absorption of gallium monofluoride (GaF) at 211.248 nm with the use of a commercially available HR-CS GF MAS system. Experimental conditions for the sensitive and precise determination of fluorine were optimized, including the time/temperature program as well as addition of gallium and modifier mixture in combined mode. Under these conditions the fluoride present in the sample was stabilized up to 600 °C, and the optimum vaporization temperature for GaF was 1540 °C. Palladium and zirconium deposited onto the graphite surface served as solid modifiers; sodium acetate and ruthenium modifiers were added directly to the sample. The limit of detection and the characteristic mass of the method were 0.43 μg/L and 8.7 pg, respectively. The proposed procedure was validated by the use of certified reference materials (CRMs) of lake water and animal tissue; the acceptable recovery was obtained, proving that it can be applied for samples with a similar matrix.


Introduction
The interest in the determination of fluorine increased over decades, which is due to its significant role in environment as well as in biological system. Although not a first one, though interesting work was published in 1941 by the U.S. National Bureau of Standards [1] on the quantitative determination of fluorine in organic compounds. Since that time a large number of papers has been published on the analytical procedure used for the determination of fluorine. The application of commonly used optical spectroscopy (inductively coupled plasma optical emission spectrometry-ICP-OES; atomic absorption spectroscopy-AAS) is limited due to the resonance lines of the fluorine being below gallium monofluoride (GaF) followed by monitoring of its molecular absorption [2]. The instrument is equipped with a 300 W xenon short-arc lamp (XBO 301, GLE, Dortmund, Germany) operating in a hot-spot mode as a continuous radiation source for the wavelength range from 185 nm-900 nm, high resolution double monochromator consisting of a prism for pre-dispersion, an echelle monochromator and a charge-coupled device (CCD) array detector. All measurements were performed using pyrolytically coated graphite tubes with integrated PIN-platform (Analytik Jena); the molecular absorption of gallium monofluoride (GaF) was measured at 211.248 nm, within the spectral range 211.1215 nm and 211.3732 nm. The optimized graphite furnace temperature/time program used for fluorine total content determination is shown in Table 1. More details are provided in Section 3.1. Table 1. Graphite furnace temperature/time program for Pd/Zr modifier and Ga deposition on graphite furnace (steps 1-5); addition of sodium acetate and ruthenium modifier (step 6). Sample thermal pretreatment (steps 7 to 10) and GaF molecules evaporation (step 11). Step

Sample Preparation
A microwave system UltraWAVE (Milestone, Sorisole, Italy) was used for the digestion of biological CRMs for the determination of total content of fluorine. The digestion of around 300 mg of samples (bovine liver or cod tissues) in concentrated nitric acid was performed under the following program: 20 min up to 220 • C and 15 min at 220 • C. In order to protect the graphite surface, the cooled digests were slightly diluted with water to a final concentration of HNO 3 of around 20%, still exhibiting low pH below 2.

Method Development
In the case of HR-CS MAS the method development covers the optimization of temperatures of main processes, e.g., the process of thermal deposition of modifiers, then thermal pretreatment of the sample as such, and finally the formation of di-molecules containing fluorine. The conditions for pyrolysis of the sample and for the process of GaF formation as well as its vaporization are crucial, as they affect the sensitivity of the determination of fluorine. The initial experimental conditions were adopted from [27], but as they did not provide the expected sensitivity and precision of results, the careful optimization of all parameters were carried out. Then the selection of the wavelength range for the monitoring of molecular absorption signal for GaF as well as its closed surrounding leaded to the range between 211.1215 nm and 211.3732 nm, which is very close to values published in recent paper (e.g., 211.1230 nm-211.3718 nm [35]). The selected range cover also the atomic line of Ga (211.196 nm) [33], which ensure the monitoring of the presence of gallium atoms in gaseous phase.

Pyrolysis and Vaporization Conditions
The pyrolysis and vaporization curve for an aqueous solution of NaF (20 µL of the 100 µg/L solution) are given in Figure 1. Interestingly, changing the pyrolysis temperature, integrated absorbance increased for about 15% within the range from 300 • C to 580 • C, then a narrow range between 580 • C and 640 • C with a plateau was observed. After that a typical dropdown of the integrated absorbance was noticed and a complete analyte loss at 1300 • C.

Sample Preparation
A microwave system UltraWAVE (Milestone, Sorisole, Italy) was used for the digestion of biological CRMs for the determination of total content of fluorine. The digestion of around 300 mg of samples (bovine liver or cod tissues) in concentrated nitric acid was performed under the following program: 20 min up to 220 °C and 15 min at 220 °C. In order to protect the graphite surface, the cooled digests were slightly diluted with water to a final concentration of HNO3 of around 20%, still exhibiting low pH below 2.

Method Development
In the case of HR-CS MAS the method development covers the optimization of temperatures of main processes, e.g., the process of thermal deposition of modifiers, then thermal pretreatment of the sample as such, and finally the formation of di-molecules containing fluorine. The conditions for pyrolysis of the sample and for the process of GaF formation as well as its vaporization are crucial, as they affect the sensitivity of the determination of fluorine. The initial experimental conditions were adopted from [27], but as they did not provide the expected sensitivity and precision of results, the careful optimization of all parameters were carried out. Then the selection of the wavelength range for the monitoring of molecular absorption signal for GaF as well as its closed surrounding leaded to the range between 211.1215 nm and 211.3732 nm, which is very close to values published in recent paper (e.g., 211.1230 nm-211.3718 nm [35]). The selected range cover also the atomic line of Ga (211.196 nm) [33], which ensure the monitoring of the presence of gallium atoms in gaseous phase.

Pyrolysis and Vaporization Conditions
The pyrolysis and vaporization curve for an aqueous solution of NaF (20 μL of the 100 μg/L solution) are given in Figure 1. Interestingly, changing the pyrolysis temperature, integrated absorbance increased for about 15% within the range from 300 °C to 580 °C, then a narrow range between 580 °C and 640 °C with a plateau was observed. After that a typical dropdown of the integrated absorbance was noticed and a complete analyte loss at 1300 °C. A pyrolysis temperature of 600 °C was chosen as an optimum for all further experiments. Comparing to previous reports, where 500 °C [14,30,31,36] and 550 °C [2,[27][28][29] were applied, selected conditions assure more efficient removal of volatile matrix components. Indeed, even higher pyrolysis temperature was reported (700 °C [33]) with the mixture of zirconium and calcium as chemical modifiers. Despite reaching a higher pyrolysis A pyrolysis temperature of 600 • C was chosen as an optimum for all further experiments. Comparing to previous reports, where 500 • C [14,30,31,36] and 550 • C [2,[27][28][29] were applied, selected conditions assure more efficient removal of volatile matrix components. Indeed, even higher pyrolysis temperature was reported (700 • C [33]) with the mixture of zirconium and calcium as chemical modifiers. Despite reaching a higher pyrolysis temperature, it is worth to notice possible interferences of CaF on the GaF, thus this option was withdrawn from further consideration.
Changing the vaporization temperature, the appearance of gallium monofluoride in gaseous phase can be registered above 1000 • C. Then the integrated absorbance increases rapidly between 1000 • C-1400 • C reaching the plateau within 1400 • C-1600 • C. The further increase of vaporization temperature leads to a consistent absorbance decrease that refers to an analyte loss caused most probably by thermal decomposition of GaF, as it occurs in the case of CaF [21]. More detailed measurements were carried out within the plateau range to establish conditions assuring best sensitivity accompanied with the best precision of measurements, expressed as lowest relative standard deviation. The optimum temperature for GaF vaporization turned out to be 1540 • C and as it is consistent with those previously reported [2,27,31,36], it was used in all further investigations.

Effect of Pd/Zr Modifier
Indirect measurement of F in a form of GaF molecule requires the application of palladium as a modifier to enhance the efficiency of the transformation into GaF, to avoid the low-temperature fluorine losses, mainly in the form of volatile HF, and to obtain a sharp molecular absorption signal [2]. Additional use of Zr was also proposed, which usually results in a considerable increase in sensitivity [14]; moreover is known to support suppression of phosphorus monoxide formation (PO) [37]. An attempt was made to use a mixture of Pd/Zr as a modifier, thermally pretreated before each sample injection; mainly to find out the optimum modifier masses in order to achieve the best performance in respect to the sensitivity.
For this purpose different amounts of the mixture of Pd/Zr were deposited onto the graphite furnace in order to find optimal amount and still to compare with values reported by Gleisner et al. [2] (m 0 , 5 µg of Pd and 0.1 µg of Zr). It turned out that the application of 1.5 times higher masses of modifiers than previously reported values results in an increase in sensitivity of about 10% ( Figure 2). However, further addition of higher masses of Pd/Zr modifier led to a gradual and considerable decrease in signal intensity. The addition of 7.5 µg of Pd and 0.15 µg of Zr was considered as the best option and was used in all further investigations.
temperature, it is worth to notice possible interferences of CaF on the GaF, thus this option was withdrawn from further consideration.
Changing the vaporization temperature, the appearance of gallium monofluoride in gaseous phase can be registered above 1000 °C. Then the integrated absorbance increases rapidly between 1000 °C-1400 °C reaching the plateau within 1400 °C-1600 °C. The further increase of vaporization temperature leads to a consistent absorbance decrease that refers to an analyte loss caused most probably by thermal decomposition of GaF, as it occurs in the case of CaF [21]. More detailed measurements were carried out within the plateau range to establish conditions assuring best sensitivity accompanied with the best precision of measurements, expressed as lowest relative standard deviation. The optimum temperature for GaF vaporization turned out to be 1540 °C and as it is consistent with those previously reported [2,27,31,36], it was used in all further investigations.

Effect of Pd/Zr Modifier
Indirect measurement of F in a form of GaF molecule requires the application of palladium as a modifier to enhance the efficiency of the transformation into GaF, to avoid the low-temperature fluorine losses, mainly in the form of volatile HF, and to obtain a sharp molecular absorption signal [2]. Additional use of Zr was also proposed, which usually results in a considerable increase in sensitivity [14]; moreover is known to support suppression of phosphorus monoxide formation (PO) [37]. An attempt was made to use a mixture of Pd/Zr as a modifier, thermally pretreated before each sample injection; mainly to find out the optimum modifier masses in order to achieve the best performance in respect to the sensitivity.
For this purpose different amounts of the mixture of Pd/Zr were deposited onto the graphite furnace in order to find optimal amount and still to compare with values reported by Gleisner et al. [2] (m0, 5 μg of Pd and 0.1 μg of Zr). It turned out that the application of 1.5 times higher masses of modifiers than previously reported values results in an increase in sensitivity of about 10% ( Figure 2). However, further addition of higher masses of Pd/Zr modifier led to a gradual and considerable decrease in signal intensity. The addition of 7.5 μg of Pd and 0.15 μg of Zr was considered as the best option and was used in all further investigations.

Effect of Ru(III) Modifier
It is also know that some transition metals, such as ruthenium, can significantly enhance the formation of GaF by reducing the interfering formation of GaO, which usually results in further sensitivity increase [2]. In our approach, instead of Ru(III) nitrosyl nitrate [2], the solution of RuCl 3 in HCl was used and the mass of modifier was optimized versus the intensity of GaF signal. The results shown in Figure 3 indicate that the optimum mass of ruthenium introduced in a form of RuCl 3 in 20% HCl is around 1 µg. Thus with the applied time/temperature program for graphite furnace, the amount of 3 µg of Ru recommended in literature [2,34] is not necessary for the best sensitivity. hance the formation of GaF by reducing the interfering formation of GaO, which usually results in further sensitivity increase [2]. In our approach, instead of Ru(III) nitrosyl nitrate [2], the solution of RuCl3 in HCl was used and the mass of modifier was optimized versus the intensity of GaF signal. The results shown in Figure 3 indicate that the optimum mass of ruthenium introduced in a form of RuCl3 in 20% HCl is around 1 μg. Thus with the applied time/temperature program for graphite furnace, the amount of 3 μg of Ru recommended in literature [2,34] is not necessary for the best sensitivity.

Optimization of the Amount of Gallium
The last step of the optimization process was the search for the highest yield of GaF molecules formation by adding increasing amounts of gallium(III) nitrite monohydrate as a molecule-forming agent. For this purpose, we used the combined mode of the application of gallium as molecule forming agent solution, meaning that part of gallium was deposited on the surface of graphite along with the Pd/Zr modifier (step 1-5 in Table 1), and part of gallium was added directly to the sample, then introduced together at step 7 (Table  1).
In order to optimize the amount of gallium introduced in both modes, firstly different masses, up to 600 μg of Ga, were deposited onto the graphite surface, when 20 μg of Ga were still added with the sample as in our previous work [34]. The results show that the integrated absorbance grows as the amount of Ga grows (Figure 4), indicating that the formation of GaF is correlated with the mass of Ga deposited on the graphite. A value of 500 μg Ga(III) (25 μL of the 20 g/L solution) was selected for further investigation due to the following technical limitations: (1) it is not recommended to inject more than 50 μL of all solutions into the graphite tube with integrated PIN platform; (2) the solutions of Ga(NO3)3 become turbid at the Ga(III) concentration higher than 25 g/L.
The alternative option is the addition of gallium(III) nitrite monohydrate solution into the furnace together with the sample solution. In this work we use the combined mode of the use of gallium, as to support the GaF formation on two levels, deposited on the graphite surface and being present in the solutions. Two amounts of Ga (60 μg and 800 μg) were selected for the evaluation of the performance of gallium as GaF molecular formation agent.

Optimization of the Amount of Gallium
The last step of the optimization process was the search for the highest yield of GaF molecules formation by adding increasing amounts of gallium(III) nitrite monohydrate as a molecule-forming agent. For this purpose, we used the combined mode of the application of gallium as molecule forming agent solution, meaning that part of gallium was deposited on the surface of graphite along with the Pd/Zr modifier (step 1-5 in Table 1), and part of gallium was added directly to the sample, then introduced together at step 7 (Table 1).
In order to optimize the amount of gallium introduced in both modes, firstly different masses, up to 600 µg of Ga, were deposited onto the graphite surface, when 20 µg of Ga were still added with the sample as in our previous work [34]. The results show that the integrated absorbance grows as the amount of Ga grows (Figure 4), indicating that the formation of GaF is correlated with the mass of Ga deposited on the graphite. A value of 500 µg Ga(III) (25 µL of the 20 g/L solution) was selected for further investigation due to the following technical limitations: (1) it is not recommended to inject more than 50 µL of all solutions into the graphite tube with integrated PIN platform; (2) the solutions of Ga(NO 3 ) 3 become turbid at the Ga(III) concentration higher than 25 g/L.  According to Figure 5, samples with a low amount of Ga used for pretreatment have usually 20%-30% lower light absorbance than those from the second set with a high mass of Ga even at high contents of Ga taken together with the sample directly before the measurement. In order to obtain much more robust results and higher signal intensity it is recommended to introduce higher masses of Ga during thermal pretreatment than together The alternative option is the addition of gallium(III) nitrite monohydrate solution into the furnace together with the sample solution. In this work we use the combined mode of the use of gallium, as to support the GaF formation on two levels, deposited on the graphite surface and being present in the solutions. Two amounts of Ga (60 µg and 800 µg) were selected for the evaluation of the performance of gallium as GaF molecular formation agent.
According to Figure 5, samples with a low amount of Ga used for pretreatment have usually 20-30% lower light absorbance than those from the second set with a high mass of Ga even at high contents of Ga taken together with the sample directly before the measurement. In order to obtain much more robust results and higher signal intensity it is recommended to introduce higher masses of Ga during thermal pretreatment than together with the sample. According to Figure 5, samples with a low amount of Ga used for pretreatment have usually 20%-30% lower light absorbance than those from the second set with a high mass of Ga even at high contents of Ga taken together with the sample directly before the measurement. In order to obtain much more robust results and higher signal intensity it is recommended to introduce higher masses of Ga during thermal pretreatment than together with the sample.  Figure 6 shows an example of the time-resolved molecular absorbance spectrum in the wavelength range of 211.1215 nm-211.3732 nm obtained during the analysis of F in the form of GaF under different conditions. A sharp absorption band from gallium monofluoride with a spectral width of around 8 pm is situated in the middle of the analyzed spectral range. A signal that appears afterwards and at lower wavelength comes from Ga atoms [38] due to their presence in the entire system [33]. Some other signals from Ga atoms may be expected in the target spectral range, but they are significantly smaller, than the one that can be seen on Figure 6, thus can be neglected. The application of the proposed modifiers set suppresses this signal, so it does not interfere with the band of GaF. As it can be seen at Figure 6, no other signals of interferences can be found in the studied spectral range. In particular, the signal from phosphorus monoxide could originating  Figure 6 shows an example of the time-resolved molecular absorbance spectrum in the wavelength range of 211.1215 nm-211.3732 nm obtained during the analysis of F in the form of GaF under different conditions. A sharp absorption band from gallium monofluoride with a spectral width of around 8 pm is situated in the middle of the analyzed spectral range. A signal that appears afterwards and at lower wavelength comes from Ga atoms [38] due to their presence in the entire system [33]. Some other signals from Ga atoms may be expected in the target spectral range, but they are significantly smaller, than the one that can be seen on Figure 6, thus can be neglected. The application of the proposed modifiers set suppresses this signal, so it does not interfere with the band of GaF. As it can be seen at Figure 6, no other signals of interferences can be found in the studied spectral range. In particular, the signal from phosphorus monoxide could originating from the biological matrix, but it is not observed ( Figure 6D). The signal at 211.248 nm is proportional to the amounts of gallium. It is essentially important that plots on Figure 6C,D for water and bovine liver digest, respectively, are very similar, which means that the possible matrix effects are limited. from the biological matrix, but it is not observed ( Figure 6D). The signal at 211.248 nm is proportional to the amounts of gallium. It is essentially important that plots on Figure 6C and 6D for water and bovine liver digest, respectively, are very similar, which means that the possible matrix effects are limited.

Method Validation
Validation covers the entire analytical procedure, including biological sample preparation (acid digestion) as well as instrumental parameters (time/temperature program, reagents amount, wavelength) that provided the highest signal response from GaF. The selected reagent concentrations and volumes are listed in Table 2, together with their mode of application. Sodium acetate was used as to create a buffer and prevent losses of fluorine in the form of HF during the drying and pyrolysis stages. The improved procedure, which includes the process of acid digestion, allows for the determination of fluorine in samples

Method Validation
Validation covers the entire analytical procedure, including biological sample preparation (acid digestion) as well as instrumental parameters (time/temperature program, reagents amount, wavelength) that provided the highest signal response from GaF. The selected reagent concentrations and volumes are listed in Table 2, together with their mode of application. Sodium acetate was used as to create a buffer and prevent losses of fluorine in the form of HF during the drying and pyrolysis stages. The improved procedure, which includes the process of acid digestion, allows for the determination of fluorine in samples at low pH values of their solutions (up to C(H + ) = 7 mol/L), which is unique and advantageous in the context of the analysis of biological samples. Notably, fluorine determination independent of its solution pH is inaccessible by the widely used ISE method.
The equation for quantification of F taken from the calibration curve was y = 0.102x + 0.1302, where x-the concentration of fluorine; y-integrated absorbance. The coefficient of determination (R 2 ) and the characteristic mass (m 0 ) of the reported method were 0.9992 and 8.7 pg of F, respectively. The calculated LOD of the method was constituted on the basis of three times the standard deviation of 10 repetitive blank measurements (using all the reagents listed in Table 2 and replacing the sample with deionized water) divided by the slope of the calibration curve. Thus, the LOD for 20 µL of sample volume was 0.43 µg/L of F and 8.5 pg of absolute F, respectively. This is comparable with other reports that applied HR-CS GF MAS with GaF formation (4 pg-33 pg of absolute F [2,14,27,28,30,33]). It is worth mentioning that such a low LOD cannot be reached using other molecule-forming agents (Ca, Sr or Al) [15,17,19]. The linear range of calibration was found for F concentrations to be 2.0 µg/L-2000 µg/L (40 pg-40 ng of F).
In order to check the robustness of the proposed procedure to potential matrix effects, first a digest of biological CRM (NIST 1577c or MODAS-5) was added directly into the furnace. The concentration of F was below the limit of detection, so its content in CRMs could be neglected. Then the same set of samples was spiked with a known amount of F standard before acid digestion and they were analyzed according to the proposed procedure. In addition, the samples with high fluorine content were measured in different dilutions and selected results were compared with the fluorine reference values for available water CRMs. An acceptable recovery factor (93-106%) confirmed the negligible influence of matrix components on the determination of fluorine in biological samples (Table 3). This proves that the improved procedure allows for the determination of fluorine in samples at low pH values of their solutions after calibration against aqueous standard solutions.

Conclusions
An improved analytical procedure has been developed and validated for the determination of trace amount of fluorine in biological samples using the molecular absorption of the GaF molecule. Modifiers and molecule-forming agents were applied in combined mode, both via thermal deposition in graphite furnace and addition to the sample. Thus, the calibration against aqueous standard solutions was possible. By using the complex matrix mixture, it was possible to enhance the pyrolysis temperature, which provides more efficient removal of matrix components. The commonly used ruthenium(III) nitrosyl nitrate was substituted by RuCl 3 in 20% HCl. The sensitivity and the limit of detection obtained are comparable with other reports that applied HR-CS GF MAS, including the one developed in our laboratory. The validated analytical protocol allows to determine low amount of fluorine directly after acid digestion of biological samples, which was impossible using other methods.

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