Analytical Validation of Smartphone Spectroscopic Technic Used in an Educational Kinetic Study

Use of smartphone-based spectroscopy is showing a constant growth since last year. It presents the advantage of being widely available for everyone. Te most important thing is that it is still a low-cost method adapted to the education context. However, as all analytical methods, it should be validated to ensure the reliability of its results. In this study, we present the steps of the validation process with its statistical tests applied to the dosage of di-iode. Shapiro–Wilk test revealed that our method has a random character. Homogeneity of variance analyses using the Cochran test confrmed the precision of the method. Te Fisher test revealed the linearity of the model of correlation between I 2 concentration and the response. Te relation between response and concentration is A =1000C +0.002. From the parameters of the linear regression of the model, we deduced the limits of quantifcation and X Lq = 4 · 10 − 5 mol · L − 1 and X Ld =1 · 10 − 5 mol · L − 1 . Tanks to tightness of the sample, the method of I 2 dosage was successfully applied in iodine quantifcation to monitor acetone iodination during time in the context of kinetic studies with minimum system trouble. Being low cost, this method can facilitate access to physical methods in educational laboratories.


Introduction
Since the frst edition of the EURACHEM Guide in 1998, a number of important developments in analytical quality have taken place. A growing interest is being accorded to measurement and analytical methods noticeably in relation with the development of new methods [1]. Validation of analytical methods is one of the topics requiring a sharing of practices in order to defne of common guidelines for laboratories. Tis sharing ensures competence requirements for laboratories, profciency testing providers, and reference material producers. As indicated in Figure 1, the life cycle of an analytical method evolves through the following: (i) Te selection of the method is a crucial step. Its selection afects directly the results. (ii) Te optimization of the method, an important step, that ensures the suitability of the method and the operation conditions of the routine.
(iii) Te validation (internal or/and external) that ensures the verifcation of the results. (iv) Te routine use with a periodical control.
Smartphone-based spectroscopy is an emergent technic managed to quantify and describe physically human colour perception using a camera [2]. Nowadays, digitalizing images is becoming available for everyone [3]. Smartphone technology has now speared in every aspect of modern life. Since its frst commercialization in 1990s, until now, smartphone use has widely expanded. In Tunisia, a North African country, in 2016, 70% of the Tunisian population possesses smartphones connected to either mobile or Wi-Fi connections according to a report of the consumer lab Ericson. Te accessibility of these devices among high school students can encourage taking advantage of laboratory experiments and practical study. Tis type of spectroscopy is easy to handle and overcome technical problems related to material lack and damage [4][5][6][7]. It is important to be aware of the wide expansion of this type of spectroscopy which covers a lot of felds such as agriculture, biochemistry analyses, medical analyses, nanomaterial, and hazardous materials. Tis method can currently quantify copper [8], iron (III) [9], formaldehyde [10], water salinity [11], blood hematocrit [12], and acetazolamide [13].
Te acceleration of this method spread out due to its facility and low-cost appeal to the necessity of an easy procedure establishment for its analytical validation [14]. In this study, we tried to implement a simple method using smartphone for the quantifcation of di-iode in order to quantify it in a kinetic lab. We describe the steps of the procedure for the analytical validation. As it is a nonnormalized method, it has to be validated according to the EURACHEM Guide.
Te acceleration of this method spread out due to its facility and low-cost appeals to the necessity of an easy procedure establishment for its analytical validation [14]. In this study, we tried to implement a simple method using smartphone for the quantifcation of di-iode in order to quantify it in a kinetic lab. Physical methods are preferred to chemical ones in reaction monitoring for kinetic studies. Tey are fast and they do not disturb progress. Sampling during the time is more accurate. UV-visible spectroscopy is the widely used technique mainly for coloured solutions [15]. But, this spectroscopic method is sophisticated and expensive. Tis apparatus cannot be aforded anywhere. Even it exists, it requires maintenance and spare parts. Use of smartphone overcomes this problem since it is afordable for all the students. So our method ensures availability at lowcost for educational institutions. However, the operation of photographing and transferring image to laptop for treatment by image J is still an awful operation in the method. To improve its accessibility and inclusiveness, it will be interesting to develop a smartphone application that treats images directly on the smartphone and not on a laptop. At this stage, we provide the method of smartphone use of a solution quantifcation, and we describe the steps of the procedure for its analytical validation. As it is a nonnormalized method, it has to be validated according to the EURACHEM Guide [16]. Tis validation procedure has to be applied when developing smartphone new methods.

Materials and Methods
All reagents were handled while donning personal protection equipment (PPE), including a lab coat, gloves, and mainly eye protection under the hood. HCl is a strong acid, and skin contact should be avoided. Acetone and ethyl acetate are volatile organic solvents. Tey should be handled carefully. Iodine solutions must not be evacuated but stored to be correctly eliminated. Iodoacetone, a product of the reaction, is very powerful and harmful. All solutions containing this should be disposed of immediately after the experiment, and the apparatus was washed with plenty of water.
All used reagents are suitable for UV-visible spectroscopy. Acetone 99.5% and hydrogen chloride solution 1 mol·L −1 are delivered by Sigma-Aldrich. Iodine solution 0.5 mol·L −1 is provided by Merck.
In this study, samples of I 2 solutions were put in plates. Tis one was implanted in a carton-covered box with a small aperture. Camera was placed in front of the aperture to photograph the plate. Image acquisition was performed by a smartphone (Samsung Galaxy A31: android version 11, 48 megapixels back camera), used with fash. Ten, Beer-Lambert relation (1) allows the calculus of solutions concentration from the intensities of solutions color measured by RGB type 3 channels solution image treatment as follows: where I represents measurement intensity corresponding to I 2 solution and represents measurement intensity corresponding to I 2 blank. Te chosen regions of interest (ROI) were squares of 400 pixels centred on the circles of each plate wells where they were duplicated. Te distribution of RGB values of every pixel was contained in histograms by applying the macro shared in the supporting information.

Method Validation.
Reliable analytical data are a prerequisite for a correct interpretation of fndings in the evaluation of scientifc studies, as well as in daily routine work, analytical methods have to be validated [17]. According to the EURACHEM Guide recommendations, the required validation steps are random character, specifcity, accuracy, and linearity [18]. In the case of our application, the determination of the limits of detection and quantifcation is also needed since the application aims to monitor the process of degradation until the total completion of the reagent. It is also important to express the results with their uncertainty. Terefore, we accomplished the process of validation by uncertainty determination of our method response.
Te frst parameter to evaluate in the validation process is the normal character or the random character of a series of responses to ensure the absence of bias in the results. Shapiro and Wilk [18,19] proposed a statistic test verifying the hypothesis of normality for a random sample. Specifcity traduces qualitatively the extent to which substances interfere with the determination of a substance according to a given procedure [20]. It is also an essential parameter to be verifed in signal detection. It allows ensuring a negative response in the absence of measured species. Precision represents the closeness of agreement between independent test results using identical experimental procedure under stipulated conditions. It also proves the closeness between measured results and the true value of a standard sample [20]. Within a given range, the analytical responses may vary linearly with the concentration of the measured species. Linearity evaluation allows the determination of method sensitivity and method limits of detection and quantifcation.
Random character of the obtained measurements has to be checked and verifed, it allows the confrmation of their independence and their normal distribution. Terefore, we consider 3 solutions of iodine (2.5 × 10 −4 mol·L −1 ) prepared separately from the commercial one. Te dilutions and the measurements by our method were repeated for four days. Table 1 regroups the responses collected for 4 days.
To evaluate the normal character of the collected data, we use the Shapiro-Wilk test resumed in Table 1 [18]. It consists of calculating the median of responses (0.245). Te number of sequences R � 8 is found by determining the number of values lower or higher than median. By referring to the values of R α/2 and R 1 − (α/2) at the risk level α � 5% given by the table, we can see that the calculated value of R is ranged between R α/2 and R 1 − (α/2) . Tus, we conclude that the distribution of the measurements is normal [21,22].
We developed our method to study the kinetic of a reaction where I 2 is a reactant. Terefore, to ensure the absence of interference between the dosed species and the matrix, we evaluate the specifcity of the method. We dose I 2 in a mixture composed of acetone, HCl in aqueous medium, and ethyl acetate and I 2 in water. Te diferent solutions served to fll the wells of the same plate. Figure 2 illustrates the responses of the two series.
It shows that the two series have the same concentrations with a relative diference inferior to 5%. Specifcity test prove absence of interferences by the adjunction of the kinetic blocking mixture. Our method is consequently specifc and does not present a risk of interference with matrices [23].
To evaluate homogeneity of variance analyses, the Cochran test permits verifcation of the precision method [24]. Absorbance of iodine solution (2.5 × 10 −4 mol·L −1 ) is measured 3 times in the same plate and during 4 diferent days. Table 2 regroups all responses [16,18].
We can perceive that the calculated constant C Cal is less than the critical constant value at both risks of 5% and 1%. Terefore, using the Cochran test, we confrm that the variances are homogenous and there are no suspected measure [16,18,24]. Our method provides reliable responses with good precision.
From Table 3, we conclude that the model of correlation between the concentration of I 2 solutions and responses is linear. Te relation between the method response and the solution concentration in I 2 is given by the following equation: A � 1000C + 0.002, with correlation coefficient r 2 � 0.9999.

(2)
Te detection limit is the smallest concentration that can be distinguished from the blank with a risk of 0.13%. In this case, the statistical test of comparison of the response at the value 0 becomes signifcant. Te limit of quantifcation is determined with a risk of 0.05%. Teir values are, respectively, calculated by equations (3) and (4) [16,18,24]: International Journal of Analytical Chemistry Random distribution Conclusion Te distribution is random With Expression (3) gives X LD � 1·10 −5 mol·L −1 and expression (4) gives X LQ � 4·10 −5 mol·L −1 .
Te statistical precision of a response is expressed by calculating the confdence interval, which indicates the margin of error when generalizing an estimate obtained to a population of n samples. Te length of the interval centred on the mean value decreases as the sample size increases. We use the following formula to calculate the uncertainty U [27]: where U: uncertainty. z � value derived from the reduced centred normal distribution, equal to 1.96 if α � 0.05 (degree of trust); σ: standard deviation. n: the number of I 2 solutions with a concentration of 2.5 × 10 −4 mol·L −1 . Application of equation (7) to the results found in Table 3 indicates the U � 0.1 10 −4 mol·L −1 .
Tis method provides a numerical result on continuous scale from the measurement of a signal directly related to the amount of analyte. Table 4 recapitulates the steps of the validation of this method.

Monitoring of Acetone Iodination by Smartphone.
Di-iode is a yellow brownish species in aqueous solution [28]. Since it has marked colour, it can be easily adapted to smartphone spectroscopy quantifcation as described in the supporting information, and Figure 3 describes this procedure.
From the results mentioned above, we confrm the validation of the method used for the quantifcation of I 2 . We used our validated method to monitor I 2 concentration evolution during the reaction of acetone iodination. Tis method of quantifcation by smartphone, being easy to implement, was used to verify the mechanism and study the kinetic of acetone iodination reaction by di-iode. Te equation of the reaction [29] is as follows: Te model is not linear Conclusion Calibration model is linear

Precision Cochran test conform
Tere is no aberrant responses in the 12 measurements; the fraction of higher variance to the sum of variances is less than Cochran critical vale at the risk α � 5% and α � 5%.

Linearity
Fisher test conform Fisher test showed that the fraction of calculated residuals and experimental ones is inferior to the tabulated Fisher value for 5 levels repeated 3 times.

Calibration function
A � 1000 C + 0.002 Least square regression is involved to determinate the slop and the intercept of the calibration curve.
Correlation coefcient 0.9999 Tis value represents the fraction of the variation in one variable that may be explained by the other variable.
Limit of detection 1 10 − 5 mol·L −1 Statistical test of comparison of the response at the value 0 becomes signifcant. Te limit of detection is determined with a risk of 0.05%.
Limit of quantifcation 4 10 − 5 mol·L −1 Statistical test of comparison of the response at the value 0 becomes signifcant. Te limit of quantifcation is determined with a risk of 0.05%.
Because of the diference between the real value and the measured one, a degree of uncertainty will pertain to measurement. Uncertainty is the absolute range in which measured value can be accepted.

International Journal of Analytical Chemistry
Te acetone iodination mechanism in acid medium is complex. It was demonstrated that its law equation is as follows [29]: Terefore, the reaction rate is of the frst order toward [H + ] and [CH 3 COCH 3 ] and of order 0 toward I 2 [29][30][31].
At an ambient lab temperature of 298 K, we prepared 3 series of experiments carried out by mixing I 2 solution with acetone in acidifed aqueous medium according to the composition detailed in Table 5. Figure 4 describes the evolution of I 2 concentration in three diferent initial conditions as a function of time. All variations are linear with a correlation coefcient up to 0.97, which confrms the order pseudo-zero-order to I 2 . Te    Table 6. Partial orders (a) and (b) toward acetone and acid can be deduced using diferent initial concentrations of the three experiments. Te kinetic law is written for the experiments i � 1; 2; 3 are as follows: From the initial rate, we calculate the partial orders a � 1, b � 1, and the rate constant k � 0.0035 L 2 ·mol −2 ·s −1 . Tese results conform to those found before I 2 was quantifed with UV-visible spectroscopy [29].
From these results, we demonstrate that using smartphone spectroscopy is reliable for the determination of I 2 concentration.

Conclusions
Use of smart technologies such as USB cameras or smartphones constitutes an available method that can be used for education. Te quasi totality of students around the world processes such devices. Teir use can facilitate the study of reaction evolution with a reduced cost. Because of the large spread of this method, we proposed in this study a procedure of validation with smartphone for I 2 quantifcation: (i) We tested the random character of the method responses by the Shapiro-Wilk test. (ii) We proved the specifcity of the method. No diference was observed between the response of the method when I 2 is dissolved in water or in reaction mixture. (iii) We verifed that there is no suspected neither aberrant responses by the Cochran test. (iv) We applied the Fisher test and we found that the method is linear. Te equation of the calibration curve allowed us the determination of the limits of detection and quantifcation of the method. (v) We calculated the uncertainty of the method.
Te validated method of I 2 quantifcation is applied to the kinetic study of acetone iodation. Tis method can be more developed and used for other chemical reactions in laboratory. Our developed method serves the equity of studying kinetics with physical methods for the neediest institutions in sophisticated materials. Moreover, it ensures a considerable reduction of chemical quantities compared to the classical UV-visible spectrophotometer needing a minimum amount of solution to fulfl the cell. However, the operation of photographing and transferring image to laptop to be treated by image J is still an awful operation in the method. To improve its accessibility and inclusiveness, it will be interesting to develop a smartphone application that treats images directly on the smartphone and not on laptop.

A:
Absorbance I: Colour intensity of I 2 solution I 0 : Colour intensity of water C: Concentration of I 2 solution r 2 : Correlation coefcient A ij : Absorbance of a I 2 solution number i during the day j R: Number of sequences in the Shapiro-Wilk test S j : Standard deviation of data series j S max : Maximum standard deviation of data series C Cal : Cochran-calculated constant C Cri : Cochran critical constant a 1 : Slope of linear curve a 0 : Intercept of linear curve F Cal : Fisher test calculated value F Tab : Fisher test reference value X Lq : Limit of quantifcation X Ld : Limit of detection U: Uncertainty ri: Rate of the reaction a: Partial order toward C 3 H 6 O b: Partial order toward H + k: Constant rate.

Data Availability
Te authors confrm that the data supporting the fndings of this study are available within the article's Supplementary Materials. It contains a description of the details of the kinetic experience and the image J macro code and can be directly used.

Additional Points
Highlights. Validation means ft for purpose. Smartphone spectroscopy for quantifcation. Statistical test for analytical method validation.

Conflicts of Interest
Te authors declare that there are no conficts of interest regarding the publication of this paper.