Identification of Silicones
In positive-ion DART-MS, PDMS products deliver signals over a comparatively wide m/z range. Generally, three major polydimethylsiloxane ion series are observed . Characteristically, each of the ion series of polydimethylsiloxanes is spaced at Δm/z = 74.018791, which reflects the mass of the OSi(CH3)2 monomeric building blocks. The first ionic series is attributed to cyclic fragments minus one methyl group, i.e., [(OSi(CH3)2)n–CH3]+, presumably formed by loss of CH4 or larger moieties from protonated molecules; ions therefore occur at nominal m/z = (n × 74)–15. The second ion series corresponds to cyclic silicone fragments of the type [(OSi(CH3)2)n + H]+, delivering ions at nominal m/z = (n × 74) + 1. The third series is obtained by ammonium ion adduct formation rather than protonation in analogy to the above, i.e., incorporates [(OSi(CH3)2)n + NH4]+ ions that yield peaks at m/z = (n × 74) + 18, i.e., at nominal vales of m/z 536, 610, 684, 758, and so on. In positive-ion DART-MS, this third series dominates by far . Here, the signals usually started from m/z 610, i.e., at [C16H52NO8Si8]+, but often only from slightly higher m/z values, and extended into the m/z 1800–2000 range. This upper limit corresponds to PDMS ions containing about 25 Si atoms, e.g., [C50H154NO25Si25]+, m/z 1868.50361.
The identification of the signals observed as ions from PDMS is explained along the partial DART spectrum of a Pavoni braided ring baking mold in the m/z 500–1000 range (Figure 1a). The ion series observed is in accordance with the [(OSi(CH3)2)n + NH4]+ ions described above and it exhibits mass differences within 0.0025 u of the calculated Δm/z = 74.01879 for OSi(CH3)2 units. This spectrum was obtained based on an external mass calibration. The PDMS ionic formulas calculated for the signals at m/z 536.16442, 610.18301, 684.20162, 758.22019, 832.23886, 906.25741, and 980.27616 are also provided (Figure 1b). Correct formulas are identified by their equal number of Si and O atoms and are highlighted in this list. They correspond to ammonium adduct ions; hence, there is one nitrogen contained. The ionic formulas are found within 2 ppm mass error. Furthermore, the isotopic patterns of all signals reveal the presence of numerous Si atoms. As a proof, the isotopic pattern of the signal at m/z 832.23886 is compared with the theoretical pattern as calculated for the assigned ionic formula [C22H70Si11O11N]+ (Figure 1c).
Silicone septa are frequently used in different laboratory applications. They should neither easily be extracted by solvents nor should they release such material upon thermal stress. In the context of the present study, two types of silicone septa were included to serve as a reference of silicone objects. They should provide DART spectra of authentic silicone rubber compared with silicone oil and grease, and allow recognizing characteristic features of those spectra in case of doubt about the origin of signals. The identification of silicones is also supported by equal appearance of spectra obtained from silicone septa or silicone oil. Thus, the origin of the spectra is clearly related to the release of PDMS from the objects under study (i.e., from the baking mold discussed in the above paragraph). All other DART spectra obtained in this study exhibit close similarity (Figures S5–S17 of the Supplementary Material).
Standardizing the DART Gas Temperature
For reliable comparison of all articles, a standard temperature for the helium gas had to be defined. This value was chosen after scanning the DART gas temperature. Here, the influence of the temperature is demonstrated along the DART spectra of a Kaiser rectangular baking mold as obtained upon exposure to DART gas at 150 to 350°C (Figure 2). The variation of the DART spectra with rising temperature is quite obvious. At 150°C, only low-mass silicones are observed and the intensity of the most abundant signal corresponds to 1.1 × 106 counts. As the temperature is increased in increments of 50°C, the distribution expands to higher mass, shows more signals, and the intensity of the signals rises up to 3.8 × 106 counts. At 350°C, the series of silicone ions expands from m/z 610 to beyond m/z 1400. Based on these observations, a standard temperature of 300°C was defined and applied to analyze all silicone rubber articles.
As a consequence, the role of eventually occurring thermal decomposition needs to be considered. Baking molds are supposed to withstand heat for an elongated period, typically 45–60 min. This is confirmed by the imprinted temperature limits for use, which are explicitly specified to be 260°C in case of the Farberware dough scraper and even 280°C for the Pavoni braided ring baking mold (Figure S3 of the Supplementary Material). The manufacturer’s statement is obviously intended to provide confidence for the potential user that this product is safe for use in direct contact with food even when exposed to high temperature for about 1 h. Admittedly, exposure of a pacifier to the DART gas at 300°C does not exactly reflect the conditions of normal use; it may, however, simulate the effect of hours of use under physiological conditions. Baby articles under study are all intended for long-time use in babies’ mouths either to supply milk or other drinks or to calm down as is the case with pacifiers and teething rings. They are thus operated under aqueous extracting conditions and, therefore, potentially prone to release low molecular weight oligomers directly into the child’s intestinal tract. Wrist watch bands and other wrist bands are frequently made of silicone rubber. They are exposed to moist skin and sweat, in particular because of increased transpiration in the contact zone with the polymer. One no-name wrist watch band has been included here for comparison.
None of the items did show visible marks, discoloration, or the like after analysis (for a demonstration with the Playtex ortho PRO pacifier cf. Figures S3 and S4 of the Supplementary Material). Overall, the conditions chosen for analysis seem not to affect the surface integrity of the objects. Thus, the release of medium molecular weight silicones is not the consequence of major decomposition due to overly harsh conditions. Rather, it represents an intrinsic property of those silicone rubbers to release medium molecular weight silicones as “natural” components of these materials analogous to what has been observed in case of stabilizers and plasticizers in other polymeric materials [28, 29].
Release of Additional Compounds
Most of the object analyzed released almost only PDMS. The spectra of some sample showed additional ionic series, typically in the low-mass range of the spectra. For example, the DART spectrum of the NUK Happy Days pacifier reveals the release of some low-mass polyethylene glycols that are detected as [PEG + NH4]+ ions (Figure 3). This spectrum has been recalibrated internally on silicone signals in the range of m/z 684 to 1202 prior to formula assignment to the additional compounds.
As a second object, delivering ion series in addition to silicones is presented by the Kaiser rectangular baking mold (Figure 2). Its DART spectra show tightly spaced signals in the m/z 300–600 range. Based on previous actual uses of this mold, they were first assumed to correspond to residual components of fats (e.g., diacylglycerols). However, after internal calibration on the silicones, formulas of the type [CnH2nN5]+, [CnH2nN5O]+, and [CnH2nN5O2]+were retrieved. This series also did not appear in the spectra of the otherwise very similar Pavoni baking molds. The identification of these additional compounds has not been pursued further. Other occasionally occurring additives occurring in the DART spectra were plasticizers, for example.
Quantification of Silicone Release
It is not straightforward to determine the amount of PDMS released from a silicone rubber surface upon exposure to the DART gas at 300°C. Nonetheless, a realistic estimate can be obtained by comparison of the signal intensity generated by DART ionization of known amounts of silicone oil under comparable conditions. Therefore, solutions of silicone oil in dichloromethane were prepared at concentrations ranging from 0.01 to 10 mg μL–1 and spots of several microliters were applied to a glass slide. After evaporation of the solvent, the residual silicone oil spot was positioned analogous to the silicone rubber items and subjected to DART analysis. Glass slides were employed as sample support because (1) glass has low thermal conductivity and should heat up similar to silicone, (2) it is chemically inert, and (3) it offers a plane surface similar to the spatial arrangement achieved with the objects under study.
To determine the spectral background, the DART spectrum of a clean glass slide was measured as a blank (Figure S18 of the Supplementary Material). The low-intensity signals at m/z 536, 610, 684, and 758 in this spectrum are due to silicone background; the ions causing them are also present in the silicone oil and other silicones. These four background peaks were observed ever since the installation of the source and are not due to residual sample from excessive analysis of silicone oil and rubber samples. In fact, their occurrence induced previous work to establish a mass calibration based on silicone oil.
It turned out that in the order of 1 μg of PDMS was required on the glass slide to obtain signals of reasonable signal-to-noise ratio (s/n). Positive-ion DART mass spectra of selected runs of silicone oil for creation of a calibration curve corresponding to 1, 4, 6, 20, and 100 μg of silicone oil applied to the glass slide are shown in Figure S19 of the Supplementary Material. Silicone peaks having relative intensities above 2% and of s/n >3 were retrieved from the spectrum list, exported to Microsoft Excel software, and their individual intensities (in counts) were summed up to yield a total signal count. The increase of the total signal count was not linearly correlated to the amount of sample; instead it was best represented by a logarithmic fit (Figure 4).
This non-linear behavior may be interpreted in terms of saturation as the upper end of the curve corresponded to the DART analysis of a visible silicone oil spot of 3–4 mm in diameter. The signal obtained thereof represented the maximum of what could be released from such a surface. The objects analyzed had a larger spatial expansion than such a spot and, thus, sample ions may have been generated and collected from an even larger surface, which in turn caused a higher amount of sample ions per run than could be obtained from a silicone oil spot. The sample spot size effect was also observed during the analysis of the SGE yellow silicone septum that was only 5 mm in diameter and was, thus, found to release only 20 μg of PDMS per run (Table 2). With such a tiny object, slight variations in position and angle caused more pronounced variation of the signal than they did during the analysis of somewhat larger items.
Some of the objects delivered total signal intensities well within the range of the calibration curve, whereas most were even beyond the upper level. The amount of silicone released ranged from 1 μg to more than 100 μg. The lowest value was measured for the NUK Happy Kids latex pacifier, which is astonishing in that the latex product was expected to release no silicone at all. Nonetheless, the signals must be attributed to this pacifier as they were clearly visible in the m/z 1100–1900 range, reproducible, and definitely different from the silicone background signals in the m/z 536–758 range.
The majority of the silicone rubber products, irrespective of the category of use, were found to release several tens to more than hundred micrograms per analysis (Table 2 and for a detailed compilation of data cf. Supplementary Table S1). Of course, the present quantification is limited in that the system to build the calibration curve cannot exactly mimic the properties of the silicone rubber surfaces. Some values also show noteworthy standard deviations of one-third or even one-half of the average value, whereas others exhibit variations of just a few percent.
Overall, these experiments deliver a solid estimate of the amount of silicone released and reveal that any silicone rubber item does so to a substantial degree. From a consumer’s point of view, it is particularly frustrating that the baking molds and the scraper that are all designated to elongated use at high temperature belong in the group of most efficient PDMS releasers.
Decomposition of the sample material might occur when a spot is exposed to the hot DART gas repeatedly or over an elongated period. Then, a continually increasing fraction low molecular weight oligomers would be formed and be more efficiently ionized, thereby causing an overestimation of PDMS release. Therefore, a control experiment was run on the dough scraper that compared two series of measurements. First, each of seven runs was taken from a fresh spot along one rim of the dough scraper with the spots set 1 cm apart. Second, the same spot was subjected to seven repeat runs corresponding to 112 s of total exposure with intervals of about 15 s between runs, which corresponded to the time required to start the subsequent run on the instrument. Quantification of the PDMS release from those runs yielded an average of 1.9 ± 0.4 × 108 counts for the series with shifting positions and 2.0 ± 0.4 × 108 counts for stationary repeat runs on the same spot. As the average value alone might obscure a trend in any of the series, the individual values were depicted in a graph (Figure 5). Obviously, there is no tendency towards an increase of PDMS release due to sample decomposition when a spot is investigated repeatedly.
Another concern is related to the exact position and orientation of the items subjected to DART analysis, which could be critical for the resulting spectra, in particular as the objects generally have curved surfaces. The reproducibility has, therefore, been tested by repeatedly measuring the same item. For each run, the item has been manually held in the ionization zone of the DART interface and fully removed to be repositioned for the next run. The variations were comparatively small, showing no noteworthy effect on the resulting spectra, either in ion distribution or in abundance (Figures S5–S17 of the Supplementary Material). It turned out that minor variations in angle and distances to ionization source and entrance capillary are well tolerated. A measure for shot to shot variation may be deduced from the variation of the quantitation results in last section of the discussion.
Some difference is to be expected between the DART analysis of a silicone rubber object and its ability to release PDMS under the conditions of its normal use. To provide some insight, small slices of baking molds of two manufacturers were placed in rapeseed oil for 1 h at 180°C to simulate the baking process where the mold is in contact with fatty dough. The oil was then placed on a glass slide and analyzed by DART-MS. A neat rapeseed oil sample was treated the same way. Signals attributable to the rapeseed oil mainly occurred in the range m/z 600–700 and were identified as ammonium adducts of diacylglycerides. In contrast, after exposure to silicone rubber baking molds, the oil was found to contain notable amounts of PDMS and the corresponding signals exhibited intensities several times higher than those of the oil (Figure 6). Neglected matrix effects of the rapeseed oil the quantification procedure were also applied to these spectra to provide an estimate of the amount. The signal intensities from PDMS corresponded to 2 μg of PDMS in 2 μL of oil in case of the Pavoni mold and 2.5 μg in case of the Kaiser mold. In terms of concentration, this indicates roughly 1 μg mg–1 PDMS in the fat content of a cake.
The aqueous extraction of a pacifier and a nipple (1.0 g in 10 mL water at 35°C for 15 h) did not result in detection of PDMS from the water. At least the release of PDMS into water is very restricted; for milk, however, the situation may change.