Reducing occupational risks and greening a traditional sample
treatment of industrial fluorine containing-materials by application of
design of experiments
JUAN GONGORA1, NEREA AYARZA2, CHRISTIAN ZOBARAN1, JOSE MIGUEL SAENZ2,
OSCAR PEREZ2, ROSA MARIA ALONSO1
1Department of analytical chemistry, Faculty of Science and Technology, University of Basque
Country (UPV/EHU), 48080 Leioa, Bizcay, SPAIN
2Derivados Del Flúor DDF SA, 39706 Ontón, Cantabria, SPAIN
Abstract: - Despite the new technological advances, some traditional methodologies concerning wet chemistry
must be used as reference methods when dealing with complex matrices or when no certified reference materials
are available. In this context, Willard-Winter distillation is nowadays still employed as a reference technique for
fluorine extraction in day-to-day analysis. However, this procedure requires strong acid mixtures, increasing
waste treatment procedures/costs and the potential risks associated with their use.
The present work reports the application of design of experiments (DoE) to improve the analytical methodology
of reference for fluorine extraction through Willard-Winter distillation by substituting perchloric acid. Variables
affecting the sample treatment of fluorine-containing compounds, anhydrite, fluorspar, cryolite and aluminium
fluoride were studied to ensure complete dissolution and total extraction of fluorine.
Volume of sulfuric acid, sample amount, volume of distilled solution including volume of melt and amount of
NaOH for fluorspar and the extended fluoroaluminate compounds were the variables studied. Predicted
experimental conditions were performed and validated in the target compound, obtaining fluorine concentrations
comparable to those obtained by the reference methodology.
By this modified approach, not only harmful effect of manipulation of perchloric acid is reduced but also costs
of the analytical procedure do so. Besides, a greener performance is achieved by avoiding chlorinated species,
reducing waste dangerousness and its treatment.
Key-Words: - fluorine distillation, anhydrite, fluorspar, cryolite, aluminium fluoride, fluorine analysis, design of
experiments
Received: February 5, 2023. Revised: October 25, 2023. Accepted: November 27, 2023. Published: December 31, 2023.
1 Introduction
Innovation and development of new fluorinated
chemicals and its applications situates fluorine
industry in a privileged and little-known position
though its great importance in many areas of daily
life. The increase of hydrofluoric acid in the market
(valued at US$1,839.1 million in 2022) gives an idea
of the widespread demand existing for its use in
different fields (e.g. petroleum refining, glass
treatment, metallurgic industry, production of
electronics, pharmaceuticals and agrochemicals).
Both economic aspect and tithing applicability make
the fluoride chemical industry of special relevance in
end-users’ industry and technological research
centers [1,2].
Likewise, there is an increased concern to
improve analytical procedures as one of the biggest
challenges from the safety and environmental point
of view while simultaneously creating economic,
environmental value for all employees and society. In
this sense, hazardous substances are the main target
to be replaced by others less toxic and less dangerous
to reduce risks and complicated treatment of hazard
residues [3,4].
Fluorinated compounds are synthetized by
hydrofluoric acid, which is mainly produced from the
mineral fluorspar, also called fluorite. The reaction of
fluorspar with sulfuric acid in excess produces HF
and anhydrite, which is a by-product used for further
applications, mainly in the cement industry [5]. The
demand for hydrofluoric acid is focused on the
electronic industry for the manufacture of silicon-
based semiconductor devices, on nuclear stations to
achieve 235U growing process, on
fluororganic/fluoropolymers compounds and on the
aluminium manufacture where aluminium fluoride
MOLECULAR SCIENCES AND APPLICATIONS
DOI: 10.37394/232023.2023.3.8
Juan Gongora, Nerea Ayarza,
Christian Zobaran, Jose Miguel Saenz,
Oscar Perez, Rosa Maria Alonso
E-ISSN: 2732-9992
89
Volume 3, 2023
and cryolite compounds play an important role in the
so-called: Hall-Héroult electrolysis process [1,6-11].
Different methodologies have been proposed to
achieve the extraction of fluorine from the sample
and its further quantification. Distillation,
pyrohydrolysis and alkaline fusion were the most
used for sample treatments, whereas potentiometric,
volumetric or spectrophotometric measurements
were the most usual methods for fluorine
determination [12-20] Furthermore, instrumental
breakthroughs lead to the implementation of faster,
simpler alternatives such as X-ray fluorescence
spectroscopy, with good reproducibility results for
the direct analysis of fluorine from solid samples
without any pretreatment of the sample [21].
However, these methods have not replaced the
official methodologies where Willard-Winter
distillation is still the selected procedure for the
extraction of fluorine in fluorspar, aluminium
fluoride and cryolite [22-24].
In HF factory laboratories, quality control
analysis are performed routinely to ensure the
adequate quality of raw materials and the best
performance of the reaction in intermediate products
to fulfill customers’ requirements in final products.
Considering the current policies committed to ensure
safety and health/environment protection, this study
deals with the substitution of perchloric acid for
sulfuric acid to extract and analyze fluorine from the
target compounds, anhydrite, fluorspar, cryolite and
aluminium fluoride.
For this aim, a chemometric study using design of
experiments (DoE) was conducted to find the best
experimental conditions for fluorine recovery, when
the mixture of perchloric/phosphoric acid is replaced
by sulfuric acid as the most suitable. A first screening
study was carried out to select the main variables
before the achievement of the most suitable
experimental conditions in the optimization phase
applying central composite design. Once the new
experimental conditions were fulfilled using
sulphuric acid, the new modified procedure was
validated and compared to those from reference ones
to check the viability for its implementation in
routine analysis.
2 Experimental
2.1 Reagents and solutions
Internal reference compounds (anhydrite, fluorspar,
cryolite and aluminium fluoride) were provided by
the company DDF, S.A. (Ontón, Spain) with a
particle size < 350 µm. Sodium hydroxide for the
alkaline fusion and phosphoric (85%), perchloric
(70%) and sulfuric acids (95%) as fluorine extracting
agents were used in the distillation process supplied
by Merck (pro analysi quality, Darmstadt, Germany).
Silica (chromatography quality) with a diameter pore
between 0.04 and 0.063 mm was used as silicon
source for fluorine distillation and provided by
Merck (Darmstadt, Germany).
Distilled solutions were neutralized at pH 6.8 – 7.0
before potentiometric determination by using a
solution of sodium hydroxide 5.0 M and
bromothymol blue solution (0.04 % as indicator
supplied both by Panreac Química S.A. (Barcelona,
Spain).
Sodium citrate (0.1 M) and sodium chloride (1.5 M)
solutions, both supplied by Merck (Darmstadt,
Germany), were used as TISAB buffer (Total Ionic
Strength Adjuster Buffer). This buffer solution was
used to set pH values’ solutions before
potentiometric determination of fluoride, to maintain
the ionic strength at a constant and high value and to
form complexes of interfering anions. Calibration
curves were performed by serial dilution from the
stock solution of sodium fluoride (1000 µg/mL,
quality Titrisol) provided by Merck (Darmstadt,
Germany).
Distillation performance, buffer and standards
solutions were prepared with purified water from a
Milli-Q Element A10 water system (Millipore,
Milford, MA, USA).
2.1 Reference methodology
The methodologies described below are routinely
used in the fluorine industry DDF, SA and are
considered as the reference methods in this study. A
sample pretreatment by alkali fusion is performed not
only in aluminium derivatives, which are insoluble in
acidic solutions but also, in fluorspar to ensure
fluorine extraction.
Sample pretreatment consisted of a first melting step
by an alkali fusion before fluorine distillation. This
step was carried out in a nickel crucible where
samples were deposited between a sand of 5 g NaOH.
The crucible is covered and introduced in the oven at
500 ºC for 15 minutes until total alkaline fusion.
Then, the mixture was dissolved in high-purity water
until room temperature, transferred into a 250 mL
calibrated flask and finally diluted with high-purity
water.
MOLECULAR SCIENCES AND APPLICATIONS
DOI: 10.37394/232023.2023.3.8
Juan Gongora, Nerea Ayarza,
Christian Zobaran, Jose Miguel Saenz,
Oscar Perez, Rosa Maria Alonso
E-ISSN: 2732-9992
90
Volume 3, 2023
The distillation process was performed by
introducing the sample amount (in the case of
anhydrite) or 50 mL of the alkaline solution
(fluorspar and aluminium derivatives), 60 mL of a
mixture of perchloric acid and phosphoric acid
(50%), 0.5 g of silica to promote fluoride distillation
as a combination HF:SiF4 according to Kleboth’s and
Dahle’s works[25,26], and 200 mL distilled water.
The distillate was collected into a vessel up to 800
mL that contained previously 100 mL of distilled
water, Table 1.
Table 1. Experimental values of reference
methodology for the sample pretreatment of
anhydrite, fluorspar, cryolite and aluminium fluoride.
(*NA: Not applicable).
Variables
Anhydrite
Fluorspar, cryolite and
aluminium fluoride
Vac (mL)
60
60
Vdes (mL)
800
800
m (g)
2.0
0.5.-0.6 (fluorspar)
0.4-0.5 (cryolite)
0.3-0.4 (AlF3)
Vmelt
(mL)
NA*
50
mNaOH (g)
NA*
5+5
Potentiometric calibration curves were built at the
potential defined for the fluoride content according to
the Nernst equation. Calibration curves were
prepared from 1 mg/L to 10 mg/L for anhydrite, 10
to 50 mg/L for fluorspar and cryolite and, from 30 to
70 mg/L for aluminum fluoride. The stock solutions
were prepared in 100 mL calibrated flasks that
contained 50 mL of 0.1 M TISAB solution. The
measurement of samples was performed by mixing
10 mL of distilled solution and 10 mL of TISAB.
After fluorine measurement by ISE-F, fluoride
recovery (%) was calculated as follows:
F (g/100g sample) = (f*CF*Vdes/m) * 100 Eq.1
f: factor (anhydrite 0.002; fluorspar and
fluoroaluminate species: 0.5/Vmelt)
CF: fluoride concentration obtained by interpolation
from the calibration curve
Vdes: volume of distillate
m: sample amount
2.1 Chemometric optimization
2.4.1. Variables and response selection
The start point of the chemometric optimization was
stated on defining the response and the most
significant variables that may affect the distillation
process. In this work, the fluorine content expressed
as fluoride (%) was considered the response in all the
target compounds except for fluorspar, which was
expressed as CaF2 (%). The volume of sulfuric acid
(Vac mL), sample amount (m, mg) and volume of
distilled solution (Vdes mL) were the variables
studied, including the volume of melt sample (Vmelt
mL) and amount of NaOH (mNaOH, g) for fluorspar
and the extended fluoroaluminate compounds which
are insoluble in acidic solutions as it was above
mentioned.
The volume of sulphuric acid was the main variable
selected regarding the goal of this research. The
volume of distillate was studied since it is related to
the analysis time and the volume of waste generated.
Sample amount and volume of melt were chosen
because they are key variables in the sample
treatment. More concretely, the volume of melt was
chosen for those compounds requiring previous
solubilization via alkali fusion whose neutralization
effect may affect the free acid in the solution to
extract the fluorine from the matrix. Other parameters
such as flow, temperature distillation and distillates’
pH values were monitored during the distillation
process.
2.4.2. Screening phase
A screening step was performed to study the effects
of the variables by a full factorial design (2k+ 2) for
anhydrite and a fractional factorial design (2k-1 + 2)
for the rest of the target compounds. The screening
phase was applied with two replicas in the central
point and using high, medium and low levels of
variables.
Fluorine recovery values were fitted to a
mathematical model using a multiple regression
algorithm, based on ordinary least squares
regressions. These regression equations (one per
analyte) were statistically evaluated by analysis of
ANOVA at 5% significance level, to estimate and
determine effects and interactions.
This analysis compares the variance of the responses
with the residual variance, which summarizes
experimental error. These ratios have a statistical
distribution, which is used for significance testing.
Effects were declared significant (+/−) or non-
significant (NS) regarding p-values. Variables with
MOLECULAR SCIENCES AND APPLICATIONS
DOI: 10.37394/232023.2023.3.8
Juan Gongora, Nerea Ayarza,
Christian Zobaran, Jose Miguel Saenz,
Oscar Perez, Rosa Maria Alonso
E-ISSN: 2732-9992
91
Volume 3, 2023
p-values lower than 0.05 (significance level of 5%)
were considered “statistically significant”. The grade
of significance increased (++/−−) when p-value <
0.01 [27,28].
Model suitability was checked regarding the obtained
R2 (percentage of variance explained) for each
response model and studying residuals distribution.
In all cases, R2 values showed a good fit and
residuals´ distributions did not diverge significantly
from the normal distribution.
Five blanks were evaluated along each factorial
design as a quality control to check the suitability of
the whole procedure, ensuring that no fluorine was
recovered after performing experiments at different
conditions.
2.4.2.1. Anhydrite
Anhydrite was the simplest sample to study since it
does not require an alkaline fusion step. A full
factorial design was carried out with three variables:
volume of sulfuric acid (Vac mL), sample amount (m,
mg) and volume of distilled solution (Vdes mL). A
total of eight experiments (23) including two replicas
of central point were randomly performed, Table 2.
Table 2. Variables and intervals of variables at low
(-1), medium (0) and high (+1) levels studied in the
screening design corresponding to anhydrite,
fluorspar and the target fluoroaluminate compounds.
(*NA: Not applicable).
Some limitations were found when experiments were
performed using less than 40 mL of acid and
collecting 800 mL. In these conditions, the
distillation temperature overcame 140 °C and the pH
value of the distillate solution was lower compared to
the reference’s one. These observations may be
explained by the presence of other acid than
hexafluorosilicic acid in the distillate, as it was
reported by Dahle et col. [29]. When the temperature
exceeds 140 °C in the distillate, sulfuric acid
decomposition to SO3 is produced, which explains
the low pH values of those distillates. Moreover,
technical problems arose when employing 10 mL of
sulphuric acid by resistance overheating to the point
of breaking the covered glass.
2.4.2.2. Fluorspar and fluoroaluminate compounds
A fractional factorial design was selected in this case
considering a compromise between the laborious
analysis for fluoride determination and the additional
pretreatment step via alkali fusion. The alkali fusion
is crucial to achieve dissolution but at the same time,
reduces the free acid in the distillator for fluorine
extraction. Accordingly, the volume of melt solution
and mass of NaOH were the selected variables to take
into account in the screening phase but studied
separately. The volume of melt was included in the
fluorspar design and the mass of NaOH in the cryolite
design.
A total of twelve experiments (23-1 + 2*3+2) with two
replicas in the central point were performed. The
volume of sulfuric acid (Vac mL), sample amount (m,
mg) and volume of melt solution (Vmelt mL) were the
variables studied in the fluorspar screening phase.
And the volume of sulfuric acid (Vac mL), sample
amount (m, mg) and volume of NaOH (VNaOH mL) in
the cryolite screening phase, Table 2.
Technical problems were observed in the case of
fluorspar, when the second replica of the central point
was performed. Distillation ran irregularly causing
resistance overheating till the point of breakage of the
covered glass.
Based on the company’s experience with
fluoroaluminate compounds analysis, it was assumed
that the overall information obtained from the
screening phase of fluorspar and cryolite may be
extrapolated to aluminium fluoride. For this reason, a
screening phase of aluminium fluoride was not
conducted before CCD for the optimization phase.
2.4.3. Central composite design, CCD
The CCD was performed to evaluate the response as
defined previously but excluding variables of lacked
significance among those studied in the screening
phase. The CCD allows to model surface responses
with a number of experiments equal to (2k+2k+n),
with k the number of variables and n the number of
extra points at the central point of the design. A CCD
consisting of a cube samples (2k) with star points
(2×n) placed at ±α from the central point of the
experimental domain. The axial size (α) was 1.68
which establishes the rotatability condition.
The five-level CCD parameter variations and
consequent responses conduct fitting of a quadratic
model to the data. For an experimental design with
MOLECULAR SCIENCES AND APPLICATIONS
DOI: 10.37394/232023.2023.3.8
Juan Gongora, Nerea Ayarza,
Christian Zobaran, Jose Miguel Saenz,
Oscar Perez, Rosa Maria Alonso
E-ISSN: 2732-9992
92
Volume 3, 2023
three variables, the model including linear, quadratic
and cross terms can be expressed as:
Y =ß0 + ßAXA + ßBXB + ßCXC + ßABXAXB + ßACXAXC
+ ßBCXBXC + ßAAXA2 + ßBBXB2 + ßCCXC2 Eq. 2
Where Y is the response to be modelled, β is the
regression coefficients and XA, XB and XC represent
sample amount (A), volume of acid (B) and volume
of melt (C), respectively.
Upon the basis of the obtained responses, a multiple
linear regression model (MLR) for each response was
defined by the program based on ordinary least
squares regression, and evaluated by ANOVA to
estimate and determine effects and interactions. To
select the optimal conditions, response surface plots
were built based on the adjustment parameters
obtained after carrying out ANOVA analysis.
Model suitability was checked regarding the
percentage of variance explained for each response
and verifying the normality distribution of residuals.
Once the model’s suitability was checked, response
surfaces were plotted in three-dimensional space and
optimal values were found according to each
response surface [30].
Intervals of variables studied in the optimization
phase with a central composite design (CCD) for
each variable are shown in Table 3. In all cases, the
explained variance (R2) values were adequate, greater
than 92.0% and distributions of residuals were
significantly random.
Table 3. Variables and intervals of variables in
the optimization phase with a central composite
design (CCD) at (±1), star (±α) and center (0)
levels. (NA: Not applicable).
2.4.3.1. Anhydrite
A simple CCD employing twelve experiments (22 +
2 * 2 + 4) was applied in anhydrite samples to achieve
the optimal values for fluorine recovery.
The volume of acid and sample amount were the
significant variables studied. The lowest level of
volume of acid (20 mL) was reconsidered and set at
40 mL to avoid problems related to overheating of the
distillation system, as was previously observed in the
screening phase.
2.4.3.1. Fluorspar and fluoroaluminate
compounds
A fractional factorial design (Resolution III) was
applied using twelve experiments (23-1+2*3 +2)
where one variable was combined with the others in
a balanced way.
Fluorspar. A fractional factorial of CCD was built
considering the variable with the highest p-value
obtained in the screening phase as a variable
MOLECULAR SCIENCES AND APPLICATIONS
DOI: 10.37394/232023.2023.3.8
Juan Gongora, Nerea Ayarza,
Christian Zobaran, Jose Miguel Saenz,
Oscar Perez, Rosa Maria Alonso
E-ISSN: 2732-9992
93
Volume 3, 2023
confounded with two-variable interaction. The
volume of acid (A) was chosen as a combination of
the others: sample amount (C) and melt solution (D).
Fluoroaluminate compounds. For these materials, the
volume of melt (D) was included in the CCD and
defined as a combination of the other main variables:
volume of sulfuric acid (A) and sample amount (C).
2.5 Validation of the models’ prediction
The optimal conditions obtained were performed
experimentally by applying the predicted conditions.
Three samples of each material were analyzed in
triplicate to validate the prediction of the models.
Fluorine recovery values were calculated as: F (%) ±
t*s/√N at a 95 % confidence level. Values in
fluorspar were expressed as CaF2 (%).
For evaluating the reliability of results, accuracy
and precision were calculated and expressed as
relative error RE (%) and relative standard
deviation, RSD (%).
Calibration curves were obtained at a potential value
defined by: E (v)=(84.0 ± 0.1) – (-58.4 ± 0.2) log [F]
for anhydrite; E (v)=(87.2 ± 0.2) – (58.1 ± 0.4) log
[F] for fluorspar, E (v)=(84.3 ± 0.2) + (-59.0 ± 0.2)
log [F] for cryolite and E (v)=(94.0 ± 0.7) + (-60.0 ±
0.4) log [F] for aluminium fluoride samples with a R2
higher than 0.9998 in all cases.
3 Results
3.1 Screening phase
3.1.1. Anhydrite
Results obtained in the screening phase are
summarized in Table 4. The volume of distillate
(Vdes) and the interaction (volume of acid and volume
of distillate) had a significant effect (p-value < 0.05).
Even if the volume of acid did not show to be
significant itself, its interaction did. Then, it was
included in the optimization phase since it played an
important role during the distillation process as it was
observed throughout the experimental performance.
Table 4. a) Significance (p values) of variables and
b) their interactions studied in the optimization
procedure with a screening design for anhydrite (full
factorial design), fluorspar and cryolite (fractionated
factorial design). The significant values (p < 0.05) are
in bold, and the effect in parenthesis. Model
suitability expressed as R2. (NS: No significative. NA
Not applicable.).
Regarding the opposite signs of the volume of
distillate and anhydrite mass, it makes sense to
extract a higher fluorine amount by employing a
higher volume of distillate, and less sample amount
promotes yielding more efficient fluorine extraction
during distillation.
Even if the volume of distilled effect was not
negligible, it was finally decided to fix it at 800 mL
as in the reference method for subsequent
experiments and ensure maximum fluorine
extraction, indeed.
3.1.2. Fluorspar and Fluoroaluminate compounds
In fluorspar samples, the three variables significantly
influenced the response and were inversely
proportional to the response regarding the negative
value of the coefficients. This means that the yield of
the extraction is more effective when less amount of
mass is used. The significance of variables and their
interactions are gathered in Table 4.
These findings were coherent by the fact that less
amount of sample led to a total dissolution of the
sample, and then less volume of melt solution is
MOLECULAR SCIENCES AND APPLICATIONS
DOI: 10.37394/232023.2023.3.8
Juan Gongora, Nerea Ayarza,
Christian Zobaran, Jose Miguel Saenz,
Oscar Perez, Rosa Maria Alonso
E-ISSN: 2732-9992
94
Volume 3, 2023
required to accomplish the extraction of fluorine
during the distillation. This fact leads to a higher
availability of free sulfuric acid in the distillator to
extract fluorine from the matrix, increasing fluorine
recovery in the distillate.
Considering cryolite samples, similar results were
obtained as in the latter case regarding the volume of
acid and sample amount effects however, NaOH
amount was not a significant variable (p-value >
0.05). Thus, it was considered to study the effect of
volume of melt in the sample pretreatment of
fluoroaluminate compounds for the optimization
phase of these compounds,
3.2 Optimization phase
The significance of variables and their interactions p-
values are compiled in Table 5 for each of the
fluorinated samples studied.
Table 5. a) Significance (p values) of variables and
b) their interactions obtained in the optimization
procedure with a CCD for anhydrite (full factorial
design), fluorspar and fluoroaluminate compounds
(fractionated factorial design). The significant values
(p < 0.05) are in bold, and the effect in parenthesis.
Model suitability expressed as R2. (NS: No
significative. NA: Not applicable.).
3.2.1. Anhydrite
According to the results, sample amount was the only
significant variable as well as its interaction.
Likewise, the prediction of the model was only
governed by the quadratic term, being a two-
dimensional response plot enough to evaluate the
affecting sense of the model, Fig. 1a.
The model predicted a maximum response with 0.13
g sample and 60 mL of sulfuric acid to perform the
distillation process.
3.2.2. Fluorspar and fluoroaluminate compounds
3.2.2.1. Fluorspar
Different response surfaces were obtained and
plotted in Fig. 1 b,c,d..
Figure 1. Response surfaces obtained after MLR
regression in the sample treatment optimization
design (CCD) corresponding to anhydrite a) and
fluorspar b,c,d) samples. Predicted conditions and
responses’ value (ypred) were anhydrite a) m 0.130 g,
Vac 42 mL, Ypred 3.44 % F, b) Vmelt 30 mL, Vac 40 mL,
ypred 96.7 %CaF2, c) m 0.4750 g, Vmelt 40 mL, ypred
93.5 %CaF2 and d) m 0.8g, Vac 80 mL ypred 109.0
%CaF2 and m 0.2g, Vac 40 mL, ypred 116.4 % CaF2.
Several maximum points of the response appear in
the prediction curve. Among them, the optimal
conditions predicted in Fig. 1c) were discarded since
the predicted response (93.5%) was lower than the
reference value (97.1%). The rest of the experimental
conditions according to the prediction models were
assayed without any satisfactory results. Lower
values of fluorine recovery than those predicted by
the model and by the reference methodology were
obtained.
Despite the prediction models failing, conditions of
the optimal point indicated in Fig. 1d were selected
MOLECULAR SCIENCES AND APPLICATIONS
DOI: 10.37394/232023.2023.3.8
Juan Gongora, Nerea Ayarza,
Christian Zobaran, Jose Miguel Saenz,
Oscar Perez, Rosa Maria Alonso
E-ISSN: 2732-9992
95
Volume 3, 2023
(m 0.2g, Vac 40 mL), but increasing the volume of
acid at 60 mL (instead of 40 mL). This consideration
was established taking into account the following
reasons: i) an increase of the volume of acid may
extract a higher amount of fluoride from the matrix
as it was observed in the screening phase and ii) it
was experimentally verified that when the Vac:Vmelt
ratio increased, the free acid is higher which
enhances fluoride extraction. Reconsidering these
aspects, it was obtained that 20 mL melt solution
containing 0.2 g of sample followed by distillation
employing 60 mL of acid were the optimal values of
the variables that yield comparable results to the
values of the reference methodology.
3.2.2.2. Fluoroaluminate compounds
Cryolite. Based on the results, the amount of cryolite
and volume of sulfuric acid were the most significant
variables followed by their interactions (Table 5).
Response surfaces were plotted to find the optimum
predicted by the model, Fig. 2a,b,c.
Figure 2. Response surface obtained after MLR
regression in the optimization design (CCD)
corresponding to cryolite samples. Predicted
conditions and responses’ values (ypred) were a) m
0.25g, Vac 40 mL, ypred 52.78 % F; b) Vac 40 mL, Vmelt
20 mL, ypred 55.63% F and c) m 0.25 g, Vmelt 20 mL,
ypred 55.63 % F.
According to the response plots, 0.25 g cryolite, 20
mL volume of melt and 40 mL of acid were the
predicted variables to achieve maximum fluorine
recovery.
Aluminium fluoride. Attending to the significant
variables test, the mass of aluminium fluoride was the
only significant variable that explained the model
prediction for the fluorine extraction, Table 5.
Likewise, and according to the graphical depictions
of the response surfaces plotted in Fig. 3a,b,c, 0.3 g
aluminium fluoride and 20 mL volume of melt to
distillate by using 40 mL of acid were the optimal
conditions predicted by the model.
Figure 3. Response surfaces obtained after MLR
regression in the optimization design (CCD)
corresponding to aluminium fluoride samples.
Predicted conditions and responses’ values (ypred)
were a) m 0.25 g, Vmelt 20 mL ypred 61.98% F b) m
0.25g, Vac 40 mL, ypred 62.36%; and c) Vac 40 mL,
Vmelt 20 mL, ypred 62.56 %F.
3.3. Validation of the models’ prediction
The optimal conditions predicted for each model
were applied for the analysis of three samples of the
targeted materials produced in industrial processes.
The obtained results compared to those predicted by
the optimization performed and the reference values
obtained by the DDF reference methodology are
summarized in Table 6.
Table 6. Predicted values obtained from the curve
response of anhydrite, fluorspar, cryolite and
aluminium fluoride samples; compared to the
experimental values and those obtained by the
reference methodology. Values were calculated as: x
(%) ± t*s/√N at 95 % confidence level). Fluoride
content was expressed as F (%) except for fluorspar
(CaF2 (%)).
MOLECULAR SCIENCES AND APPLICATIONS
DOI: 10.37394/232023.2023.3.8
Juan Gongora, Nerea Ayarza,
Christian Zobaran, Jose Miguel Saenz,
Oscar Perez, Rosa Maria Alonso
E-ISSN: 2732-9992
96
Volume 3, 2023
Based on the results, extraction of fluorine by
distillation with sulfuric acid suggests satisfactory
recovery values of fluorine under the optimized
experimental conditions performed. Results agree
within the confidence interval compared to the DDF
reference values. Except for anhydrite materials,
accuracy and precision obtained were very good and
less than 1.0 % and 0.8 %, respectively. Higher
values were obtained for anhydrite samples (RE 8 %
and RSD 31 %) whose fluorine content is less than in
the other target materials. This implies higher values
of RE and RSD for lower percentages of fluorine,
being the precision the most affected by the
application of the methodology at low fluorine
content.
4 Discussion
Optimized methodology
In general, fewer sample amounts were required in
the optimized conditions than in the reference
methods, in agreement with the negative signal effect
of this variable obtained along the DoE performed.
Fluorspar and fluoroaluminates ratio values (volume
of acid/volume of melt) were triple and double
respectively compared to the ratio in the reference
methods (1.5). This observation could be related to
the strength of the acids to extract fluorine from the
matrix, indicating that a greater volume of sulfuric
acid is required to balance this aspect compared to the
strength of the perchloric/phosphoric acids mixture.
Likewise, fluorspar samples demanded more volume
of melt to distillate compared to the fluoroaluminates,
which could be explained due to their physical-
chemical properties. More severe conditions were
required to overcome the forces that hold the lattice
between Ca-F. In fact, fluorspar is the compound
with higher lattice energy regarding to the melting
point series among the target compounds CaF2
>AlF3>cryolite [31]. Moreover, the high purity level
of fluorspar samples (>90%) compared to the
fluoroaluminate ones (50-65%) has to be additionally
considered.
Comparative with ISO norms
The optimized methodologies of fluoroaluminate
derivatives require less volume of acid than ISO
standard norms [23,24]: 40 mL sulfuric acid in the
optimized procedure compared to 50 mL sulfuric
acid in ISO standards. In the case of fluorspar,
distillation is achieved using more volume of acid (60
mL sulfuric acid) than in the ISO standard norm [22],
in which 35 mL of perchloric acid is employed.
However, costs, safety and environmental benefits
are balanced in this latter case.
Another aspect to consider concerning ISO norms is
the method to determine fluorine compared to the
method used in this study by F-ISE potentiometry.
Considering the determination of fluorine in ISO
norms, titration with LaNO3 (fluorspar) or with
Th(NO3)4 (fluoroaluminate derivatives) is
performed, which is not in agreement with the
practice of green analysis by reducing or avoiding
harmful reactants, even more in the case of handling
thorium due to its radioactive characteristics.
Costs, environmental and safety benefits
The optimized experimental conditions reduce not
only the volume of acid but also the amount of
sodium hydroxide, which represents a saving in the
consumption of reagents and a reduction in the
volume of waste compared to the reference method.
In fact, a saving of up to more than 20% per litter is
achieved using sulfuric acid instead of perchloric
acid according to the market price. On the other hand,
the optimized procedure avoids the use of perchloric
acid which is potentially dangerous due to the risk of
explosion that involves its manipulation and
treatment of chlorinated wastes, greening the analysis
for routine purposes [32,33].
MOLECULAR SCIENCES AND APPLICATIONS
DOI: 10.37394/232023.2023.3.8
Juan Gongora, Nerea Ayarza,
Christian Zobaran, Jose Miguel Saenz,
Oscar Perez, Rosa Maria Alonso
E-ISSN: 2732-9992
97
Volume 3, 2023
Applicability
This optimized approach may be applied not only for
the quality control of raw materials and intermediate
products in the HF industry but also, for the analysis
of by-products from fluorapatite (Ca5(PO4)3F) used
in fertilizers manufacturers or by-products from the
electrolysis of cryolite in the smelting process.
Furthermore, the importance of quality control is
crucial in these fluorine-containing materials, as it
allows timely adjustments of the parameters of the
HF manufacturing process and, therefore, adequate
compliance with product specifications.
Environmental labs require as well analytical
approaches for the assessment of fluorine distribution
in wastes from landfills and in fluorine-contaminated
soils near storage sites from industry [19].
In addition, these methodologies allow ensuring the
implementation of new instrumental techniques, and
even as an alternative methodology in situations
where instrumental failures may occur.
5 Conclusion
The chemometric approach for the optimization of
the fluorine separation procedure of fluorinated target
compounds has been demonstrated in terms of
experimental design. Results from factorial designs
showed that the amount of sample and volume of acid
were the most influential parameters on fluorine
extraction.
Fluorine values obtained by ISE-potentiometry
agreed with those obtained by the reference method.
Thus, the modifications of the optimized conditions
for the sample pretreatment allowed the substitution
of the perchloric/phosphoric acids mixture by
employing sulfuric acid. The implementation of the
new optimized conditions ensures economic and
occupational exposure benefits in agreement with the
principles of the laboratories’ quality control
systems, including a reduction of wastes in
conformity with the actual consensus for more
greening procedures of analysis.
These alternative approaches may be applied to those
fluorine-containing materials that require fluorine
extraction and analysis for: quality purposes, fluorine
environmental assessment in soils and industrial
wastes, and HF manufacturing process control which
depends on fluorine content analysis to monitoring
reactors’ parameters.
More investments are needed in research for the
development and acquisition of new
instrumentation for the analysis of fluorine to obtain
fast and reliable results in fluorine-containing
materials, whose chemical complexity makes them a
great challenge for the development of new analytical
methods.
Acknowledgement:
The authors thank the University of the Basque
Country (UPV/EHU) for financial support (Project
GIU19/068), and the Company DDF, SA for its
collaboration and contribution to this project.
References:
[1] https://www.eurofluor.org/ (Last accessed 13th
September 2023).
[2] https://www.futuremarketinsights.com/reports/
hydrofluoric-acid-market (Last accessed 13th
September 2023).
[3] ISO guide 17025:2017. Exigences générales
concernant la compétence des laboratoires
d'étalonnages et d'essais. International
Organization for Standardization, 2017,
Switzerland.
[4] https://echa.europa.eu/fr/regulations/reach/under
standing-reach (Last accessed 13th September
2023).
[5] F. Rais, A. Jelidi, A. Kamoun, M Chaabouni, M
Sergent, R Phan-Tan-Luu. Use of an ellipsoidal
subregion of interest in the space of mixture
components to the optimization of a
fluoroanhydrite-based Self-Leveling floor
composition. Chemometric and Intelligent
Laboratory Systems. Vol.74, 2004, pp. 253-261.
[6] F. Ullmann, Encyclopedia of Industrial
Chemistry, A11, Willey-VCH, 2002, pp. 315-
316.
[7] E.P. Gillis, K.J. Eastman, M.D. Hill, D.J.
Donnelly, NA Meanwell. Applications of
fluorine in medicinal chemistry, Journal of
Medical Chemistry. Vol.58, No.21, 2015, pp.
8315-8359.
[8] Q-D. Shen, Preparation, Structure and properties
of fluorine-containing polymers, in: Zhi-Min
Dang (Ed.) Plastics Design Library, Dielectric
Polymer Materials for High-Density Energy
Storage, William Andrew Publishing, 2018, pp
59-102. ISBN 9780128132159
[9] Mineral Profile; Fluorspar. British Geological
Survey, www.mineralUK.com (Last accessed
13th September 2023).
[10] D. Othmer, D. Herman, Encyclopedia of
Chemical Technology, Willey & Sons, 1980, pp.
540- 541.
[11] J. Metson, Production of alumina, in: Roger
LumLey (Ed.) Metals and Surface Engineering,
MOLECULAR SCIENCES AND APPLICATIONS
DOI: 10.37394/232023.2023.3.8
Juan Gongora, Nerea Ayarza,
Christian Zobaran, Jose Miguel Saenz,
Oscar Perez, Rosa Maria Alonso
E-ISSN: 2732-9992
98
Volume 3, 2023
Fundamentals of Aluminium Metallurgy,
Woodhead Publishing, 2011, pp. 23-48.
[12] V.L. Dressler, D. Pozebon, E.L.M. Flores,
J.N.G. Paniz, E.M.M. Flores. Potentiometric
determination of fluoride in geological and
biological samples following pyrohydrolytic
decomposition, Analytica Chimica Acta,
Vol.466, No.1, 2002, pp. 117-123.
[13] E. Berns, P. Van der Zwaan, The pyrohydrolytic
determination of fluoride, Analytica Chimica
Acta, Vol.59, No.2, 1972, pp. 293-297.
[14] V. Conrad, W. Brownlee, Hydropyrolytic-ion
chromatographic determination of fluoride in
cool and geological materials, Analytical
Chemistry, Vol.60, No.4, 1988, pp. 365-369.
[15] V. Haff, C. Butler, J.D. Bisso. Tritimetric
determination of fluorine, particularly in
aluminium fluoride, Analytical Chemistry,
Vol.30, 1958, pp. 984-989.
[16] K. Anazawa, T. Tomiyasu, H. Sakamoto.
Simultaneous determination of fluorine and
chloride in rocks by ion chromatography in
combination with alkali fusion and cation-
exchange pretreatment. Analytical Science,
Vol.17, No.1, 2001, pp. 217-219.
[17] S.Y. Oh & Y.D. Seo, Distribution characteristics
and analytical protocols for remediation of
fluorine‑contaminated soils at a
phosphate‑gypsum waste landfill,
Environmental Earth Sciences, Vol.82, No.26,
2023, pp. 82:26.
[18] J.V. Gimeno Adelantado, V. Peris Martinez, A.
Checa Moreno, F. Bosch Reig,
Spectrophotometric determination of fluoride in
fluoride.bearing minerals after decomposition by
fusion with sodium hydroxide, Talanta, Vol.32,
No.3, 1985, pp.224-226.
[19] T.A. Palmer, Direct determination of fluorine in
aluminium reucton materials by using an ion-
selective electrode, Talante, Vol.19, 1972, pp.
1141-1145.
[20] M.S. Corbillón, M.P. Carril, J.M. Madariaga, I.
Uriarte, Fast determination of total fluoride by
direct potentiometry in samples of aluminium
fluoride an cryolite, Analyst, Vol.120, 1995,
2227-2231.
[21] O. E. Bezrukova, S.D. Kirik, S.G. Ruzhnikov,
I.S. Yakimov, P.S Dubinin. An accurate
determination of cryolite ratio in industrial
aluminum baths by wavelength-dispersive X-ray
fluorescence spectrometry, Spectroscopic Acta
Part B, Vol.152, 2019, pp. 52-58.
[22] ISO Guide 5439:1974. Acid grade fluorspar
Determination of available fluorine content –
Potentiometric method after distillation,
International Organization for Standardization,
1974, Switzerland.
[23] ISO Guide 2362:1972. Aluminium fluoride for
industrial use -- Determination of fluorine
content -- Modified Willard-Winter method,
1972, Switzerland.
[24] ISO Guide 1693:1976. Cryolite, natural and
artificial -- Determination of fluorine content --
Modified Willard-Winter method, 1976,
Switzerland.
[25] K. Kleboth, Zur Theorie der
Fluorokieselsäuredestillation, Monatshefte für
Chemie, 1969, Vol.100, pp.1494-1498.
[26] Dahle, D., Wichemann, H. J., “Further Studies
of Fluorine Distillation”, Association of Official
Agricultural Chemists (1937), 20 (2), 297-303.
[27] J.N. Miller, J.C. Miller, Statistics and
chemometrics for analytical chemistry, 6th
edition, Prentice Hall/Pearson, New York, 2010.
[28] K.H. Esbensen, Multivariate data analysis—In
practice, sixth edition, ©CAMO Software AS,
Oslo (Norway), 2019.
[29] D. Dahle, H.J. Wichmann, A quantitative study
of fluorine distillation, Journal of the Association
of Official Agricultural Chemists, Vol.19, No.2,
1936, pp. 312-320.
[30] P.W. Araujo, R.G. Brereton, Experimental
design II. Optimization, Trends in Analytical
Chemistry, Vol.15, 1996, pp. 63–70.
[31] C. Malherbe, B. Gilbert, Direct determination of
the NaF/AlF3 molar ratio by Raman
spectroscopy in NaF NaF/AlF3 molar ratio by
Raman spectroscopy in NaF−AlF3−CaF2 melts at
1000 °C, Analytical Chemistry, Vol.85, 2013,
pp, 8669−8675.
[32] ASQ Chemical and Process Industries Division
Chemical Interest Committee, Quality Assurance
for the Chemical and Process Industries: A
Manual of Good Practices. ASQ Quality Press
(2nd Edition), USA, 1999.
[33] S. Garrigues, FA. Esteve-Turrillas, M. de la
Guardia, Greening the wastes, Current Opinion
in Green Sustainable Chemistry, Vol.19, 2019,
pp, 24-29.
Contribution of Individual Authors to the
Creation of a Scientific Article (Ghostwriting
Policy)
Juan Góngora: formal analysis, investigation,
methodology, writing – original draft & editing.
Nerea Ayarza: methodology, review & editing.
Christian Zobaran: formal analysis,
investigation, methodology.
MOLECULAR SCIENCES AND APPLICATIONS
DOI: 10.37394/232023.2023.3.8
Juan Gongora, Nerea Ayarza,
Christian Zobaran, Jose Miguel Saenz,
Oscar Perez, Rosa Maria Alonso
E-ISSN: 2732-9992
99
Volume 3, 2023
José Miguel Saenz: conceptualization,
methodology, supervision.
Oscar Pérez: conceptualization, methodology,
supervision.
Rosa Maria Alonso: methodology, supervision,
writing-review.
Conflict of Interest
The authors have no conflicts of interest to declare
that are relevant to the content of this article.¨
Creative Commons Attribution License 4.0
(Attribution 4.0 International, CC BY 4.0)
This article is published under the terms of the
Creative Commons Attribution License 4.0
https://creativecommons.org/licenses/by/4.0/deed.en
_US
MOLECULAR SCIENCES AND APPLICATIONS
DOI: 10.37394/232023.2023.3.8
Juan Gongora, Nerea Ayarza,
Christian Zobaran, Jose Miguel Saenz,
Oscar Perez, Rosa Maria Alonso
E-ISSN: 2732-9992
100
Volume 3, 2023
Sources of Funding for Research Presented in a
Scientific Article or Scientific Article Itself
No funding was received for conducting this study.
