Optimization of compound-specific chlorine stable isotope analysis of chloroform using the Taguchi design of experiments

Rationale: Chloroform, a probable human carcinogen, is commonly detected in various concentration levels in many surface water and groundwater sources. Compound-specific chlorine stable isotope analysis (Cl-CSIA) is significant in investigating the fate of chlorinated contaminants in the environment. Analytical conditions should, however, be thoroughly examined for any isotopic fractionation. In this study, we simultaneously optimize three analytical parameters for a robust online Cl-CSIA of chloroform using theTaguchi design of experiments. Methods: For Cl-CSIA, a purge-and-trap autosampler coupled to a gas chromatograph in tandem with a quadrupole mass spectrometer, with electron ionization in selected ion monitoring (SIM) mode, was used. Using the Taguchi


| INTRODUCTION
Chloroform (CHCl 3 ), a toxic chlorinated hydrocarbon, is detected at various concentrations in tap water, surface water reservoirs (natural and manmade), and some aquifers. In addition to the natural formation of chloroform, 1,2 chlorination of water supplies may result in the formation of trihalomethanes (THMs), as the chlorine reacts with organic matter, 3,4 with chloroform being one of the main products. 5 Moreover, its wide manufacture and use as an industrial solvent, and its commercial use contribute to its common occurrence and detection. In groundwater systems, chloroform could result from various sources including improper handling and disposal, spills and leakage, and recharge of chlorinated water such as treated wastewater or water from public water supplies. 5 The Environmental Protection Agency in the United States of America (US EPA) has classed chloroform as a probable human carcinogen (Group B2). 6 Compound-specific isotope analysis (CSIA) has been found to be very useful to investigate contamination sources, degradation pathways and fate of organic contaminants in a variety of environmental systems. [7][8][9][10][11][12][13][14][15][16] For instance, Sturchio et al 17  Online stable chlorine isotope measurements, with the use of molecular and/or fragment ions of compounds formed in the ion source, have formerly been performed using quadrupole mass spectrometry (qMS), isotope ratio mass spectrometry (IRMS), and other online instrumental techniques, on several compounds. [21][22][23][24] The qMS method was first developed by Sakaguchi-Söder et al. 22 Round robin tests of gas chromatography (GC)/IRMS and GC/qMS for Cl-CSIA on TCE by Bernstein et al, 25 and on chloroform and carbon tetrachloride by Heckel et al, 26 have shown that the methods agree very well in terms of precision and accuracy. The use of the GC/qMS technique is therefore preferable, as it is comparatively simple, widely applicable for a range of chlorinated compounds, and does not require upstream conversion.
In both the IRMS and the qMS techniques, the environmental concentrations of target compounds are often too low for CSIA measurements. As a result, online sample enrichment techniques, such as solid-phase microextraction (SPME) and purge and trap (P&T), for C-CSIA using GC/IRMS have successfully been demonstrated. 9,[27][28][29] The P&T method was found to be more suitable, as it resulted in less isotopic fractionation. 28 Various optimization methods for carbon isotope analyses (C-CSIA) are documented. 9,30,31 For instance, the effect of different purging times on carbon isotope ratios was demonstrated with the use of a continuous-flow P&T system 30 and a P&T system with a liquid autosampler. 28 With respect to sample enrichment for online Cl-CSIA using qMS, headspace injection techniques are commonly used. 18,19,23,26 However, it was shown that the isotope ratios and the precision may vary, due to various analytical conditions. 21,26,32 As a result, different optimization approaches for Cl-CSIA using a variety of instrumental configurations, peripherals, and compounds have been published. 21,23,[32][33][34][35][36][37] For example, Heckel et al 26  Analyzing the effect of each analytical parameter separately may not result in the highest overall precision when combined. Therefore, methods are needed for the simultaneous optimization of various parameters to determine their best combination. In the engineering sciences, the Taguchi design of experiments method is widely applied to optimize the performance of processes or the quality of a product with reduced costs. [38][39][40] With the orthogonal array of the Taguchi method, multiple parameters and multiple levels are run independently at the same time and evaluated subsequently, thereby reducing the number of experiments needed to aim at a certain target. 38,41 In this study, we applied the Taguchi design of experiments method to systematically optimize and significantly improve the determination of compound-specific stable chlorine isotope ratios, with chloroform as a model compound, using a simple online P&T-GC/qMS technique. We specifically looked at the three parameterspurging time, transfer time, and dwell timeto select their best combination for a stable and precise isotope analysis.
A P&T system (PTA 3000; IMT, Vohenstrauß, Germany) was used to extract the chloroform from an aqueous phase. The P&T system was connected to a 6890N gas chromatograph coupled to a 5973N quadrupole mass spectrometer (both from Agilent, Santa Clara, CA, USA), with a transfer line heated at 200 C. Using this P&T-GC/qMS system, the Cl-CSIA analyses were performed online.
Purging of chloroform in the aqueous phase was carried out at 40 C with a purging flow of 20 mL/min (helium). The trap was filled with Tenax as the adsorbent (M16T 08; IMT, Moosbach, Germany) and operated at −35 C. When purging was completed, the trap was rapidly heated to 190 C. The desorbed chloroform was transferred splitless to the GC/qMS system for Cl-CSIA. Between measurements, the trap was conditioned at 190 C for 23 min to avoid memory effect caused by possible residuals on the trap. For the conditioning, nitrogen gas was used.
The gas chromatograph was equipped with a DB 624 capillary column (60 m × 0.25 mm, 1.4 μm film thickness; Agilent J&W, Santa Clara, CA, USA). The temperature program started at 40 C which was held for 2 min. The temperature was then raised to 90 C at 8 C/min, held for 4 min, raised to 125 C at 6 C/min, raised to 220 C at 15 C/min, and finally held for 5 min. The ion source was operated at 70 eV using electron ionization (EI) and the mass spectrometer was set in the selected ion monitoring (SIM) mode. The masses to be analyzed for the Cl-CSIA of chloroform were m/z 83 and 85.

| Mathematical equation and peak area correction
For chloroform, the targeted masses of the CHCl 2 + fragment ion (m/z 85 and 83; 12 C 1 H 35 Cl 37 Cl + and 12 C 1 H 35 Cl 2 + , respectively) were chosen. These are the two highest peaks in terms of their relative abundance, enabling low detection limits, and they were also used by others. 26,18 The chlorine isotope ratio R Cl was determined using the following equation: where 85 I and 83 I are the peak signal intensities of the fragment ions at m/z 85 and 83, respectively, reflecting the ratio of the heavy to light isotopes.
Isotope analyses typically show a mass, i.e. peak area (PA), dependency that has to be corrected. 42,43 We used an in-house standard of chloroform at three different concentrations (59.6, 104.5, and 149 μg/L) that were run in each sequence. In this concentration range, there was an evident linear relationship between the chlorine isotope ratios of the standards and the peak area of the m/z 83 fragment ion ( Figure S1, supporting information). We used the slope of this relationship for peak area correction and the final calculation of the isotopic composition as in Sakaguchi-Söder. 42

| Taguchi experimental design and evaluation method
The analytical conditions of the P&T-GC/qMS system, namely, purging time, transfer time and dwell time, were optimized using an L 9 orthogonal array of the Taguchi experimental design. 44 For each sequence, a total of nine experiments, with four replicates, were designed under nine systematic but non-identical combinations of three parameters (P1-3) and three levels per parameter i.e., P1: Purging time (5, 10, 15 min), P2: Transfer time (80, 120, 160 s) and P3: Dwell time (20, 60, 100 ms). The selected parameters and levels are summarized in Table 1, and Table 2 shows the layout of L 9 array for this study.  (2): where i is the experiment number 1 to 9: S/N i is the signal-to-noise ratio, μ i is the mean isotope ratio (R), and σ i is the standard deviation of the isotope ratio of the i th experiment. The S/N values were further evaluated to determine the mean of S/N per parameter per level. For each parameter, the level with the largest mean of S/N is considered to be the most favorable one. Furthermore, the parameter with the T A B L E 1 Parameters and levels selected for optimizing the Cl-CSIA method for chloroform with Taguchi-method L 9 (3 levels × 3 parameters) using P&T-GC/qMS largest range in S/N for the three levels is considered to be the most sensitive parameter, with highest rank, in this L 9 test.

| Results of Taguchi experiments
The isotope ratios (R) determined for the nine experiments, L 9 , with four repetitions, are presented in Table 3. For each of the nine experiments, a S/N was determined using Equation (2). The arithmetic means of the S/N, for each level in each parameter, were calculated based on the L 9 layout from Table 3 and are summarized in Table 4. As described in section 2, the range was calculated for the evaluation of the degree of influence that the selected parameters have on the precision of determining the isotope ratios.
Having a S/N of 1.60, the smallest range among the three parameters, it is evident that the transfer time has the lowest influence on the precision of the method. The isotope ratios of chloroform showed very similar results for transfer times of 80 and 120 s (see also Table S1, supporting information).
Furthermore, the influence of the selected levels of the dwell time on the precision of the method was limited, with a S/N range of 2.23 ( Table 4). The S/N for a dwell time of 100 ms was the largest among the levels within the parameter, indicating the optimal condition. The smallest relative standard deviation (RSD) in the isotope ratio was achieved under the same dwell time (Table S2,  selected 50 ms as the optimal dwell time for that compound. As indicated in Table 4, the range of S/N for purging time was  Rcorr* Raw R-values were corrected for a given peak area based on slopes in Figure S1 (supporting information).
T A B L E 4 Summary of means of S/N per level per parameter. Each mean was determined from the S/N (Table 3)

| Further optimization of purging time
In our Taguchi approach we focused on the highest precision of the isotope results by optimizing transfer time, dwell time, and purging time. However, we observed in our experiments that, although a purging time of 5 min had the highest precision, the extraction efficiency was lowest compared with purging times of 10 and 15 min, which have similar efficiencies ( Figure S2, supporting information).
Low extraction efficiencies could influence isotope ratios. 28,30,45 We therefore conducted additional experiments with purging times of 5, 7, 8, 9 and 10 min, each in triplicate (CHCl 3 = 100 μg/L), wherein samples were randomly sequenced to refine the method. In these measurements, a transfer time of 120 s and dwell time of 100 ms were fixed. The results are presented in Figure 1.
Analyses of variance (ANOVA) on the peak areas, isotope ratios and the corresponding purging times were carried out. The results (Table S4,

| CONCLUSIONS
We have applied the Taguchi  time and 100 ms of dwell time were found to be the optimum conditions for Cl-CSIA of chloroform, and a precision of 0.28 ‰ was achieved. This shows that a direct and accurate determination of the stable isotope ratios of chloroform from aqueous samples, using GC/qMS, without upstream offline chemical conversion or other online conversion techniques is possible.
The Taguchi experimental design proved to be very useful for the optimization of analytical conditions for stable isotope analyses.
In principle, it can be applied to various compounds and analytical methods by selecting appropriate parameters and conditions. However, additional fine-tuning may be necessary for specific purposes.