238 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS levels of NDE1A. Recent studies indicate that Bronopol can act as a nitrosating agent for diethanolamine (8). Third, NDE1A was separated by LC and detected by a thermal energy analyzer which had previously only been coupled to a gas chromatograph to determine volatile nitrosamines. Numerous methods for the trace analysis of nitrosamines have been developed and published (9). However, the majority of nitrosamines studied were volatile and presented little separation problems from the matrix. The cosmetic chemists problem is more difficult since the non-volatile trace contaminant (NDEIA) in the cosmetic product is often in a multi-complex emulsion system. The complexity of a finished cosmetic formulation and possible trace levels of NDE1A, originating from the raw materials, indicate that the NDE1A analysis must start with the raw materials used to formulate these products. Since the cosmetic raw materials di- and triethanolamine can readily form NDEIA by nitrosation, the latter more slowly than the former, these raw materials were the first to be studied for nitrosamine contamination. NDEIA and other polar nitrosamines are very amenable to separation by reverse phase chromatography using either water or water/alcohol mobile phases (10). The water/al- cohol mobile phase provides good solubility for the raw material matrices eliminating the need to perform multiple isolation steps prior to analysis. The thermal energy analyzer's inability to handle mobile phases or samples containing water are alleviated by quantitating the nitrosamines by a UV detector. Rahn (10) has reported that NDEIA in the ethanolamines can be analyzed by ttPLC while Rosenberg and co-workers have reported on the analysis of NDE1A in Lauramide DEA by HPLC (11). The major difference in the two methods was the design of the UV detector employed for quantitation. The difference between a fixed wavelength and a variable wavelength detector will be detailed later in the discussion section. The data presented in this report is the result of a joint study between laboratories on the applicability of HPLC for the routine analysis of NDE1A. Both normal and reverse phase high pressure liquid chromatography coupled with UV detection for the isolation, identification, and quantitation of NDEIA in other cosinetic raw •naterials and a prototype product have been explored. The precision and accuracy of the method and the extension of the method into the part per billion (ppb) range is reported. It is our intention to also give the reader some insight as to how UV detection compares with that of a thermal energy analyzer for the quantitative analysis of polar nitrosamines such as NDE1A. EXPERIMENTAL A. APPARATUS AND REAGENTS The liquid chromatographic separations were performed on a Waters Associates Model 204, equipped with two M6000A Solvent Delivery Systems, a Model U6K Universal Injector, a Model 440 Fixed Wavelength Detector operated at 254 nm, a Model 660 Solvent Programmer, and a M730 Data Module or a Brinkman Model 2541 recorder. The Variable Wavelength Detector used was a Perkin-Elmer LC-75. Where indicated, a Waters Associates Radial Compression Separation System (RCSS) was

NITROSAMINE ANALYSIS 239 substituted for the standard stainless steel HPLC columns. Confirmation of structure by Mass Spectroscopy was performed on a Finnigan 4000 GC-MS data system. Standard NDE1A was obtained from Columbia Organic Chemical Company (Colum- bia, SC). Methanol used for analysis was HPLC-grade and the chloroform used was J. T. Baker's Photrex reagent-grade solvent. Distilled water was purchased from the Electrified Water Company or prepared from a Milli-Q system. The HPLC column used for the reverse phase separations was/.tBondapak C•8, while the normal phase separations were on a /.tPorasil column. Both columns were manufactured by Waters Associates. For some of the analysis, a Whatman Partisil PXS 10/25 ODS was also acceptable. B. METHOD OF ANALYSIS FOR ETHANOLAMINES HPLC parameters.' Column/.tBondapak C•8 (octadecylsilane), Waters Associates, 3.9 mm I.D. x 30.0 cm, stainless steel or Whatman Partisil PXS 10/25 ODS. Solvent A--distilled H20. Solvent B--methanol. Flow rate--l.O ml/min. Pressure (-6perating range)--O-6000 psi. Detector sensitivity--O-O.005 aufs. Recorder a) Range--10 my. b) Chart speed--10 ram/min. Pre column--Waters Associates, 3.9 mm I.D. x 3.0 cm, C•8 porasil stainless steel (optional). Solve n t progra m: a) Initial conditions--Solvent A (100%). b) Hold time--7 min. c) Run time (Solvent A to Solvent B)--3 min. d) Final Conditions--Solvent B (80%). e) Hold for 20-25 min. f) Reverse time (Solvent B to Solvent A)--10 min. g) Final conditions--Solvent A (100%) (allow 30 min. to equilibrate between runs). h) Programming is not needed for NDEIA standard runs, use only Solvent A. Procedure Preparation of N- nitrosodiethanolamine (NDElA ) standards: 1. Weigh 0.1000 g (0.08-0.12 g) of NDE1A into 100-ml volumetric flask. 2. Add 20 ml of Solvent A, swirl until NDE1A is dissolved, then dilute to volume with Solvent A. This gives a stock solution of (800-1200 ppm) NDE1A. 3. Pipet 1 ml of stock solution into 100-ml volumetric flask. Dilute to volume with Solvent A. This gives a working stock solution of 10 ppm (8-12 ppm) NDE1A.