Ultra-trace elemental determination of Si by means of graphite furnace-atomic absorption spectrometry

Abstract

The significance of Si determination has increased with the rise of production of silanes, siloxanes and silicones. The determination of Si plays an important role to determine the presence of such compounds in solution and provide quantitative data. For most methods, lower limits of detection of Si so far have been in the mg L⁻¹ to μg L⁻¹ range. However, to comply with the maximum levels for Si absorption through food and dietary supplements, trace determination of Si in the very low μg L⁻¹ to high ng L⁻¹ range is required. In this study, a method for determining ultra-trace amounts of Si in food simulants, ethanol, and acetic acid, was developed using GF-AAS. The method was validated according to the guidelines set by the FDA, and the results showed Si quantification limits of 0.2 μg L⁻¹ and 0.4 μg L⁻¹ for ethanol and acetic acid, respectively.

---

Introduction

At Evonik, a Si-containing coating agent was manufactured to be used to coat glass food containers. In an article published by the European Food Safety Authority (EFSA) it is stated¹ that for a migration of a substance, in this case a silane, <50 μg kg⁻¹ of food, only a limited toxicological data set is sufficient. This requirement was the driver to develop a method that is able to detect silane by means of Si in contents <10 μg kg⁻¹. Since Si is the second most abundant element on earth,² its determination in ultra-trace amounts is a challenge for elemental analysis. Along with sample pre-concentration, meticulous cleaning and expensive hardware are required.^3,4

The method of choice for ultra-trace elemental analysis is usually ICP-OES or quadrupole-based ICP-MS. For Si, however, ICP-based techniques are hampered by isobaric interferences and high backgrounds. Thus, detection limits are usually in the mg L⁻¹ to μg L⁻¹ range.^5–7

We perform ultra-trace elemental determination of Si by means of graphite furnace-atomic absorption spectrometry (GF-AAS) in the food simulant acetic acid¹ and in ethanol⁸ exposed to silanes. We showcase a methodology used in an extraction study that is more sensitive than comparable ICP-MS approaches while requiring less setup and labour.

Experimental

Chemicals

The following chemicals were used in this study. Unless noted otherwise, all chemicals are of analytical grade. Ethanol (EMSURE, Merck, Darmstadt, Germany), Si stock solution (1000 mg L⁻¹ Si, from SiO₂ in 0.5 M NaOH, Merck, Darmstadt, Germany), hydrolysed amino silane (calculated mass concentration of Si was β_Si = 26 277 mg L⁻¹, Evonik Operations GmbH, Rheinfelden, Germany), nitric acid (67–69%, Ultrapure, VWR, Radnor, USA).

Data acquisition, instrumentation and parameters

Unless noted otherwise, instruments and equipment were purchased from Thermo Fisher Scientific. AAS analyses were performed on the iCE 3500 with integrated graphite furnace (GFS35Z) and autosampler (for iCE 3400 and iCE 3500 Zeeman B/C). A Si hollow cathode lamp was used as radiation source and an omega platform extended lifetime cuvette (lifetime: approximately 750 measurements) was used as the atom cell. Ar was used as the inert gas. The furnace programs were dependent on the sample matrix and are given in Table 1.

Table 1 Furnace temperature programs for the analysis of samples in 3% acetic acid (A) and 95% ethanol (B)

T A/°C	T B/°C	t/s	Ramp/°C s^−1
125	100	20	15
1150	1000	20	60
2650	2700	3	0
2800	2800	3	0

---

Drying, ashing and atomisation temperatures were optimised for both acetic acid and ethanol (see ESI, Tables 1 and 2†). To optimise the drying temperature, the liquid level in the graphite tube was observed. The ashing and atomisation temperatures were optimised by variation of one parameter while keeping the other constant. The highest net signal for either parameter gave the respective ashing and atomisation temperatures.

Analyses were performed at λ_Si = 251.6 nm with a spectral resolution of 0.5 nm and a lamp current of 90% at 15 mA. For background correction the Zeeman effect was used. A working volume of 20 μL was selected. In accordance with the manufacturer's guidebook (see ESI†), no modifiers were used. Rather, by adopting matrix-matched standards the influences of volatile Si-species were accommodated. The samples were concentrated five-fold by adopting multiple sample injections into the graphite furnace and analysed in triplicate. The evaluation of the measurement was carried out using peak height rather than peak area. While the signal-to-noise ratio between the calibration standard and the background was comparable for both height and area, the noise-corrected net signal was significantly larger for the peak height. With respect to ultra-trace elemental analysis, considering both signal-to-noise ratio and noise-corrected net signal, peak height was preferred. All dilutions were prepared in conditioned (cleaned using ultrapure dilute nitric acid followed by matrix-matching with acetic acid and ethanol, respectively) centrifuge tubes (Falcon, Fisher Scientific). The solutions were then poured into conditioned (same as above) polypropylene sample cups (Thermo Fisher Scientific).

Cleaning of vessels and preparations for the measurement

The sample vessels were filled with dilute nitric acid, left for at least 30 min, then rinsed twice with ultrapure water and allowed to dry. This cleaning step was repeated with a cleaning solution adapted to the sample matrix (either 3% acetic acid or 95% ethanol). For analysis, sample cups were rinsed twice with the sample solution. Prior to the analysis, a new sample capillary was installed. The furnace chamber and the windows of the graphite furnace were cleaned using ethanol. Initially, three consecutive determinations of blanks were performed, to confirm that the background extinction of Si was <0.010 (less than the first calibration point).

Filtration of the sample

The samples were filtered through a syringe filter. The syringe (Injekt Luer Solo, B|Braun, Elsenfely, Germany) and filter (Chromafil Xtra PTFE-45/25, 0.45 μm, Macherey-Nagel, Düren, Germany) were rinsed five times with 2 mL of cleaning solution (either 3% acetic acid or 95% ethanol) and once with 2 mL sample solution. Three solutions were prepared for each sample, each using its own syringe and filter.

Calibration

Calibration solutions in a concentration range of 2 μg L⁻¹ to 25 μg L⁻¹ were diluted from a 50 μg L⁻¹ Si stock solution by the autosampler. The 50 μg L⁻¹ stock solution was prepared from hydrolysed amino silane with a mass concentration of 26 277 mg L⁻¹ Si, calculated stoichiometrically from its composition. The same formulation was used to coat the bottles in the extraction study. A quality control standard with a Si concentration of 1 μg L⁻¹ was prepared from the 50 μg L⁻¹ stock solution. All solutions were produced using solvents that correspond to the sample matrix (either 3% acetic acid or 95% ethanol). When using ethanol, solutions were prepared gravimetrically, otherwise they were prepared volumetrically.

Validation of the method

A seven-point calibration between 2 μg L⁻¹ to 25 μg L⁻¹ was used for quantification. The linearity of the method was gauged by the calibration's coefficient of correlation. The sample was measured in a threefold determination. A spike experiment was carried out at three different concentration levels (1 μg L⁻¹, 2 μg L⁻¹, 3 μg L⁻¹), each in a threefold determination. The mean values of the recovery rates for each concentration level were calculated to determine the accuracy of the method. The precision of the method was acquired by the relative standard deviation of the nine individual recovery rates of the spike experiment. The limits of determination and quantitation (LOD and LOQ, respectively) were calculated from the standard deviation σ of blank measurements (n = 10) and the slope of the calibration curve m. To correct for the concentration that resulted from a five-fold injection of the sample into the graphite tube, the nominal LOD and LOQ, respectively, were calculated by taking the number of injections N into account. The calculation of the respective limits is performed in accordance with the formula given in DIN 32645:2008-11 ⁹ and dividing the result by the number of injections, N = 5. After 14 d, the analysis of the spiked solutions at a spike level of 2 μg L⁻¹ was repeated. The mean recovery rate was calculated and used to check the stability of solution. This procedure was carried out for each of the sample matrices (3% acetic acid and 95% ethanol).

Results

Filtration of samples

The cut surfaces of the glass samples introduced trace amounts of glass dust into the food simulants. Preliminary experiments (data not shown) demonstrated deviating responses, which were caused by Si in the glass dust. To ensure accurate and reproducible results, it was necessary to separate the glass dust from the food simulants by means of syringe filtration. All results shown here were acquired from filtered samples.

Usage of matrix-matched standards

Comparability of the calibration solutions and the samples was achieved in two ways. First, instead of commonly used inorganic ICP-standards (usually 1000 mg L⁻¹ in 0.5 M NaOH), we opted for organic silane standards with known Si contents. These silanes were additionally used to coat the sample bottles, furthering comparability between standard and sample. Second, we prepared the calibrants in either 3% acetic acid or 95% ethanol, depending on the samples, which ensured comparable behaviour within the autosampler and GF-AAS.

Validation parameters and results

The validation of the Si determination in the food simulants was performed in accordance with acceptance criteria specified in the Guidance for Industry: Preparation of Premarket Submissions for Food Substances.¹⁰ Hence, validation criteria 1–5 (linearity, accuracy, precision, LOD, LOQ and stability of solution) are considered. The parameters, corresponding acceptance criteria (AC) and corresponding results are listed in Table 2. All acceptance criteria are met.

Table 2 Validation parameters, acceptance criteria according to Guidance for Industry: Preparation of Premarket Submissions for Food Substances and results for ethanol and acetic acid

No.	Parameter	Acceptance criterion	Ethanol	Acetic acid	AC met?
1	Linearity	R ≥ 0.995	0.9986	0.9971	Yes
2	Accuracy				Yes
2a	1 μg L^−1 spike	Recovery rate 60–110%	82%	97%	Yes
2b	2 μg L^−1 spike	Recovery rate 60–110%	106%	106%	Yes
2c	3 μg L^−1 spike	Recovery rate 60–110%	89%	107%	Yes
3	Precision	RSD of recovery rates max. 20%	18%	7%	Yes
4a	Nominal LOD	None	0.07 μg L^−1	0.1 μg L^−1	N/A
4b	Nominal LOQ	≤1 μg L^−1 Si	0.2 μg L^−1	0.4 μg L^−1	Yes
5	Stability	Min. 60% recovery rate at 2 μg L^−1	89%	74%	Yes

---

Application of this methodology to routine analysis

From a recent food contact migration study, the Si contents in acetic acid (HOAc-X.Y) and ethanol (EtOH-X.Y) were determined by means of the methodology reported here. For acetic acid, Si was determined in the high ng L⁻¹ range, while low μg L⁻¹ contents were determined in ethanol. The results are summarised in Table 3. Evidently, trace amounts of Si can be determined and quantified in food contact simulants of real applications by the method reported here.

Table 3 Resulting Si contents in food contact solvents acetic acid and ethanol. Two samples were received each and measured in duplicate. Mean values and corresponding RSDs are given

Sample	Si/μg L^−1	Mean/μg L^−1	RSD/%
HOAc-1.1	0.85	0.82	5
HOAc-1.2	0.79
HOAc-2.1	0.72	0.72	0.5
HOAc-2.2	0.71
EtOH-1.1	3.6	3.3	11
EtOH-1.2	3.4
EtOH-2.1	3.5	3.5	3.4
EtOH-2.2	3.7

---

Conclusions

We introduce a new method for Si determination in food simulants that is able to determine the elemental Si concentration down to 0.2 μg L⁻¹ and 0.4 μg L⁻¹ for acetic acid and ethanol, respectively. The method was successfully validated by meeting all acceptance criteria regarding linearity, accuracy, LOD, LOQ, precision and stability as given by the FDA's Guidance for Industry: Analytical Procedures and Methods Validation for Drugs and Biologics.¹⁰ Cleaning steps (rinsing of sample vessels with acid, then with sample matrix) before the measurement to prevent contamination from the vessels and equipment and automated dilution by the autosampler minimised risks of contamination. Multiple injections of the sample into the graphite tube concentrated the samples, resulting in a reduction in a low nominal LOD and LOQ, respectively. Matrix-matched calibration by use of a hydrolysed silane of known composition as the calibration substance provided comparability of the calibration solution and the sample solution. We reach lower limits of detection than common ICP-MS-based approaches while avoiding the use of expensive special equipment, such as high-resolution ICP-MS, while minimizing time-consuming processes and facilitating ultra trace determination of Si in aqueous solutions.

Data availability

The data supporting this article have been included as part of the ESI.†

Author contributions

Conceptualisation: AS and CV. Methodology: AS, PA, CK and CV. Software: AS. Validation: AS, MVBK and CV. Formal Analysis: AS. Investigation: AS. Resources: AS, PA, CK and CV. Data curation: AS and MVBK. Writing – original draft: AS and MVBK. Writing – review & editing: MVBK, PA, CK and CV. Visualisation: AS and MVBK. Supervision: CV. Project administration: CV. Funding acquisition: CV.

Conflicts of interest

At the time of writing, all authors are employed at Evonik Operations GmbH. The authors state that there are no conflicts of interest to declare.

Notes and references

1. EU-Guideline, EFSA CEF Panel (EFSA Panel on Food Contact Materials, Enzymes, Flavourings and Processing Aids), 2008. “Note for Guidance for the preparation of an application for the safety assessment of a substance to be used in plastic Food Contact Materials”, EFSA J., 2008, 6(7), 41 , 21r..
2. A. Yaroshevsky, Geochem. Int., 2006, 44, 48–55.
3. W. Boschetti, L. M. G. Dalagnol, M. Dullius, A. V Zmozinski, E. M. Becker, M. G. R. Vale and J. B. de Andrade, Microchem. J., 2016, 124, 380–385.
4. Y. Takaku, K. Masuda, T. Takahashi and T. Shimamura, J. Anal. At. Spectrom., 1994, 9, 1385.
5. S. Hauptkorn, J. Pavel and H. Seltner, Fresenius. J. Anal. Chem., 2001, 370, 246–250.
6. M. F. Gazulla, M. Rodrigo, M. Orduña, M. J. Ventura and C. Andreu, Talanta, 2017, 164, 563–569.
7. Q. Wang and Z. Yang, Silicon, 2018, 10, 1–10.
8. E. Bradley, R. Brandsch, J. Bustos, D. Dainelli, B. Faust, R. Franz, E. J. Hoekstra, O. Kappenstein, R. Rijk, C. Simoneau and M. Vints, Training Workshop ‘Safety of Food Contact Materials: Technical Guidelines for Testing Migration under Regulation (EU) No. 10/2011’, 2014.
9. Deutsches Institut für Normung, DIN 32645:2008-11: Chemical Analysis – Decision Limit, Detection Limit and Determination Limit under Repeatability Conditions – Terms, Methods, Evaluation, 2008.
10. U.S. Food & Drug Administation, Guidance for Industry: Preparation of Premarket Submissions for Food Contact Substances (Chemistry Recommendations), 2018.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4ja00360h

---