Real-Time Monitoring of Fragrance Release from Cotton Towels by

An innovative headspace sampling and injection system for gas chromatography was designed using a longitudinally modulating cryogenic system mounted ...
1 downloads 3 Views 886KB Size
Anal. Chem. 2010, 82, 729–737

Real-Time Monitoring of Fragrance Release from Cotton Towels by Low Thermal Mass Gas Chromatography Using a Longitudinally Modulating Cryogenic System for Headspace Sampling and Injection Olivier P. Haefliger,* Nicolas Jeckelmann, Lahoussine Ouali, and Ge´raldine Leo´n Firmenich SA, Corporate R&D Division, P.O. Box 239, CH-1211 Geneva 8, Switzerland An innovative headspace sampling and injection system for gas chromatography was designed using a longitudinally modulating cryogenic system mounted around the sampling loop of a two-position loop injector. The setup was hyphenated to a fast low thermal mass gas chromatograph, allowing transient concentrations of semivolatile analytes to be monitored in real time with a time resolution of 4.5 min. The performance of the instrument, and in particular its cryotrapping efficiency, was characterized using a mixture of long-chain alkanes, methyl esters, ethyl esters, and alcohols of different volatilities. The device was found to be ideally suited to the analysis of semivolatile compounds with boiling points ranging between 190 and 320 °C, which are typical for a majority of perfumery raw materials. The new instrument was successfully used to monitor the release of eight odorant compounds from cotton towels to which fabric softener had been applied that alternatively contained the fragrance in free form or in microencapsulated form. The analytical results, unprecedented in their level of precision and time resolution for such an application, evidenced the major impact of microencapsulation technology on the kinetics of fragrance release during the drying of the towels and on the triggering of additional fragrance release by applying mechanical stress to the fabric to rupture the microcapsule walls. Microencapsulation in polymeric shells has emerged as an excellent approach to improve the performance of fragrance compositions in home care and body care consumer products.1-3 Perfumery raw materials which would otherwise not be stable enough in some surfactant-based matrixes, for example, due to the basic pH of most detergents, can be protected. Also, fragrance deposition on targets such as fabric during laundry washing processes can be enhanced. This is especially useful in the case * To whom correspondence should be addressed. Phone: (+41) 22 780 3239. Fax: (+41) 22 780 3334. E-mail: [email protected]. (1) Park, S. J.; Arshady, R. Microcapsules for Fragrances and Cosmetics; MML Series; Arshady, R., Boh, B., Eds.; Citus Books: London, 2003; vol 6. (2) Thies, C. A Short History of Microencapsulation Technology; MML Series; Arshady, R., Ed.; Citus Books: London, 1999; vol 1. (3) Quellet, C.; Schudel, M.; Ringgenberg, R. Chimia 2001, 55, 421–428. 10.1021/ac902460d  2010 American Chemical Society Published on Web 12/21/2009

of ingredients with a high water solubility that tend to be washed off. Finally, the release of volatile odorant compounds can be controlled more accurately by tailoring the permeation properties of the membrane rather than by just relying on the evaporation kinetics of the individual molecules, which is what happens when the fragrance is applied in nonencapsulated form. Overall, fragrance microencapsulation can therefore deliver compositions with hedonic attributes that are preferred by the consumers, which exhibit an enhanced odor intensity, and which are longer-lasting. An essential aspect of the characterization of microencapsulated perfume formulations is the assessment of their odorant release after application. One part of this task is typically achieved using a sensory panel, with evaluators rating attributes such as pleasantness and odor intensity.4 The other part, which consists in accurately determining the evolution over time of the gas-phase concentrations of individual compounds, falls within the general category of dynamic sampling headspace analysis. Such work is typically conducted by gas chromatography (GC), and a wide variety of sampling approaches are known.5-7 Considering that typically at most several tens of milligrams of an individual fragrance compound are placed in an entire washing machine containing at least a few kilograms of laundry to wash, it is obvious that only low analyte levels may reasonably be expected in the headspace samples to analyze. Accordingly, a preconcentration step during sampling is a prerequisite, and approaches based on full loop injections cannot be expected to afford a sufficient sensitivity.8-10 The most commonly used preconcentration approach consists in adsorbing the analytes on a sorbent trap, from which they are injected later into the GC column by heating the trap.11-13 Another well-documented (4) Melgaard, M.; Civille, B.; Carr, T. Sensory Evaluation Techniques, 3rd ed.; CRC Press: Boca Raton, FL, 1999. (5) Nunez, A. J.; Gonzalez, L. F.; Janak, J. J. Chromatogr. 1984, 300, 127–162. (6) Kolb, B. J. Chromatogr., A 1999, 842, 163–205. (7) Wang, Y.; McCaffrey, J.; Norwood, D. L. J. Liq. Chromatogr. Relat. Technol. 2008, 31, 1823–1851. (8) Negelein, D. L.; Bonnet, E.; White, R. L. J. Chromatogr. Sci. 1999, 37, 263–269. (9) White, R. L. Anal. Chem. 2008, 80, 9812–9816. (10) White, R. L. Chromatographia 2009, 69, 129–132. (11) Mitra, S.; Xu, Y. H.; Chen, W.; Lai, A. J. Chromatogr., A 1996, 727, 111– 118. (12) Feng, C.; Mitra, S. J. Chromatogr., A 1998, 805, 169–176.

Analytical Chemistry, Vol. 82, No. 2, January 15, 2010

729

technique is cryogenic trapping.14-19 In this latter case, the volatiles are trapped in a piece of bare or coated tubing cooled by a cryogen such as carbon dioxide or liquid nitrogen from which they are released once the cooling process is interrupted by one way or another. Most reported applications deal with analytes typically classified as volatile organic compounds (VOCs), and the case of more semivolatile compounds such as perfumery ingredients has been addressed less frequently. A further requirement for the study of the release kinetics of fragrances from targets is that measurements have to be repeated at frequent time intervals, for a large number of times, and if possible in a fully automated manner. Furthermore, since perfumery molecules exhibit a wide range of physical and chemical properties, the use of a temperature gradient rather than isothermal conditions is mandatory. To maintain a short enough duty cycle, fast instruments should preferably be used.8-10,14,15,17,20-22 In this context, a new instrumental design is described and characterized.23 Its principle consists in aspirating the headspace to analyze with a vacuum pump. The analytes are then trapped in a piece of metal capillary mounted on a two-position valve held at a low temperature using a commercially available longitudinally modulating cryogenic system (LMCS) cooled by liquid carbon dioxide and intended to be used for two-dimensional GC applications.24-27 When trapping is complete, the analytes are remobilized by actuation of the valve followed by the motion of the LMCS. They are then refocused at the head of a GC capillary column and subsequently eluted with a temperature gradient. Total duty cycles not exceeding minutes are possible since a low thermal mass gas chromatograph amenable to heating rates up to 30 °C/s and requiring about 90 s to cool down is used.28 The ideal suitability of the instrument to the real-time monitoring of low levels of semivolatile compounds will be illustrated by comparing the release of a mixture of perfumery molecules deposited on cotton towels after application of fabric softener when the fragrance composition was incorporated to the base into the classical free form versus when it was formulated as a dispersion of microcapsules. (13) Tienpont, B.; David, F.; Witdouck, W.; Vermeersch, D.; Stoeri, H.; Sandra, P. Lab Chip 2008, 8, 1819–1828. (14) Mouradian, R. F.; Levine, S. P.; Sacks, R. D. J. Chromatogr. Sci. 1990, 28, 643–648. (15) Klemp, M.; Peters, A.; Sacks, R. Environ. Sci. Technol. 1994, 28, 369A– 376A. (16) Rankin, C. L.; Sacks, R. D. J. Chromatogr. Sci. 1994, 32, 7–13. (17) Sacks, R. D.; Nowak, M. L.; Smith, H. L. ISA TECH/EXPO Technology Update Conference Proceedings 51, 1996; 97-101. (18) Volatile organic compounds by vacuum distillation. US EPA Method 5032, 1996. (19) Borgerding, A. J.; Wilkerson, J. Anal. Chem. 1996, 68, 701–707. (20) Sacks, R.; Smith, H.; Nowak, M. Anal. Chem. 1998, 70, 29A–37A. (21) Hope, J. L.; Johnson, K. J.; Cavelti, M. A.; Prazen, B. J.; Grate, J. W.; Synovec, R. E. Anal. Chim. Acta 2003, 490, 223–230. (22) Akard, M.; Sacks, R. D. J. Chromatogr. Sci. 1994, 32, 499–505. (23) Haefliger, O. P. World Patent Application WO 2007/141718, June 1, 2007. (24) Marriott, P. J.; Kinghorn, R. M. Anal. Chem. 1997, 69, 2582–2588. (25) Kinghorn, R. M.; Marriott, P. J.; Dawes, P. A. J. Microcolumn Sep. 1998, 10, 611–616. (26) Marriott, P. J.; Ong, R. C. Y.; Kinghorn, R. M.; Morrison, P. D. J. Chromatogr., A 2000, 892, 15–28. (27) Kinghorn, R. M.; Marriott, P. J.; Dawes, P. A. HRC, J. High Resolut. Chromatogr. 2000, 23, 245–252. (28) Sloan, K. M.; Mustacich, R. V.; Eckenrode, B. A. Field Anal. Chem. Technol. 2001, 5, 288–301.

730

Analytical Chemistry, Vol. 82, No. 2, January 15, 2010

Figure 1. Experimental setup for automatic headspace analysis by low thermal mass gas chromatography using a longitudinally modulating cryogenic system (LMCS) for analyte sampling and injection.

EXPERIMENTAL SECTION Headspace Sampling Gas Chromatograph. The instrumental setup is illustrated in Figure 1. The innovative headspace sampling and injection system was made up of two key elements, a two-position 10-port valve (Valco, Houston, TX) and a LMCS (Everest model, Chromatography Concepts, Doncaster, Australia) which were integrated in an easy to customize gas chromatograph (8610C with thermostatted heated valve oven, SRI Instruments, Torrance, CA) used only for its oven capability and to control all timed events. It was hyphenated to a standard benchtop chromatograph (6890N, Agilent, Palo Alto, CA) equipped with a low thermal mass gas chromatographic system (MACH, Gerstel, Mu¨lheim an der Ruhr, Germany), which was also used for data acquisition from its flame ionization detector (FID). Data processing was performed using ChemStation B.03.01-SR1 software (Agilent). Headspace sampling, through a 1 m long 1/16 in. o.d. and 0.040 in. i.d. piece of fused-silica-lined stainless steel tubing placed in a heated hose held at 150 °C that was a standard part of the first gas chromatograph, was achieved using a membrane pump (KNF Neuberger, Balterswil, Switzerland). The flow rate, typically around 10 mL/min, was continuously recorded using a flowmeter (Flow Tracker 1000, Agilent, Palo Alto, CA) connected to a personal computer and installed after the loop so that the headspace samples would not have to flow through it. In the trapping position of the two-position 10-port valve, labeled as A in Figure 1, the headspace samples passed through a 0.25 mm i.d. deactivated stainless steel sampling loop (UADTM5, Frontier Laboratories Ltd., Fukushima, Japan) about 1 m in length labeled as loop 1. During this step, the analytes present in the headspace were trapped in a 2 cm long segment of loop 1 placed inside the cryogenic trap of the LMCS and held at -50 °C using carbon dioxide. Desorption and injection of the trapped analytes into the GC column followed, at the end of the trapping stage after a time interval which was varied between 15 s and 30 min. This step started with a switch of the two-position 10-port valve that was followed by actuation of the cryogenic modulator,

first in the direction opposite to the carrier gas flow and then back to its original position 0.2 min later. During that time interval, the trapping segment of loop 1 was exposed to the heat, typically 200 °C, of the oven in which it was placed. The setup was kept in that configuration until the beginning of the next sampling event, with the exception of the cooling of the cryogenic modulator that was initiated 1 min earlier to allow the target temperature to be reached. A two-position six-port valve would have delivered the same performance as the two-position 10-port valve in the described setup, which had two ports permanently connected by a jumper and a second loop labeled as loop 2 that was not effectively used. A two-position six-port valve was indeed part of the initial instrumental design.23 However, the setup with the two-position 10-port valve had the advantage of minimizing the sudden pressure changes upon switching of the valve. Furthermore, it offers interesting perspectives for future improvements of the system, for example by using the second loop more efficiently in combination with the same or with another LMCS. After desorption, the desorbed analytes were directly injected into the second gas chromatograph through a piece of 180 µm fused-silica capillary (Agilent) maintained at 250 °C using a 1 m long heated hose (H320C-010-02-250C; Hillesheim, Wagha¨usel, Germany). Gas chromatographic separation took place on a 10 m × 0.18 mm capillary GC column wound in a 5 in. low thermal mass cage (DB-1 ms, Agilent for the work with the standards, and RTX1, Restek, Bellefonte, PA for the work with the towels). The temperature was held at 35 °C for the first 1 min after initiation of the desorption and was then ramped to 260 °C at 200 °C/min, a final temperature which was maintained for 20 s. Helium was used as the carrier gas, at a constant pressure of 2.8 bar. The observed dead time was around 0.3 min. The signal from the FID detector was acquired at 50 Hz. Chemicals. Unless stated otherwise, all 42 compounds used for the characterization of the system had purities of 99% or more. Tridecane (A13), pentadecane (A15), hexadecane (A16), heptadecane (A17), and octadecane (A18), butanol (H4), pentanol (H5), methyl heptanoate (M7), ethyl hexanoate (E6), and ethyl dodecanoate (E12) were from Acros (Geel, Belgium), decane (A10), undecane (A11), dodecane (A12), heptanol (H7, 98%), octanol (H8), nonanol (H9, 98%), decanol (H10), undecanol (H11), methyl undecanoate (M11), ethyl pentanoate (E5), ethyl heptanoate (E7), ethyl decanoate (E10), and ethyl undecanoate (E11, 97%) were from Aldrich (Buchs, Switzerland), ethyl butanoate (E4), ethyl octanoate (E8), ethyl nonanoate (E9, 97%), and ethyl tridecanoate (E13) were from Alfa Aesar (Karlsruhe, Germany), nonane (A9, 95%), tetradecane (A14), nonadecane (A19), hexadecanol (H16), dodecanol (H12), methyl pentanoate (M5), methyl hexanoate (M6), methyl octanoate (M8), methyl nonanoate (M9), methyl decanoate (M10), methyl dodecanoate (M12, 97%), methyl tridecanoate (M13), methyl tetradecanoate (M14), methyl pentadecanoate (M15), ethyl tridecanoate (E13), and ethyl pentadecanoate (E15, 96%) were from Fluka (Buchs, Switzerland). The model fragrance used during the study of release kinetics was formulated by mixing equal weights of nine perfumery raw materials, which were used in their industrial form. Their structures and selected properties are displayed in Table 1. (±)trans-(E)-1-(2,6,6-Trimethyl-3-cyclohexen-1-yl)-2-buten-1-one (δ-

damascone), tetrahydro-4-methyl-2-pyran (Doremox), and (±)-2propylheptanenitrile (Jasmonitrile) were manufactured by Firmenich (Geneva, Switzerland). (±)-2-tert-Butyl-1-cyclohexyl acetate (Verdox) was manufactured by IFF (New York, NY). Allyl heptanoate, benzyl acetate, benzyl alcohol, 3-(4-tert-butylphenyl)-2-methylpropanal (Lilial), and 2-phenylethanol (phenethylol) were received from the manufacturing plant at Firmenich. Generation of Headspace Samples with Known Concentrations. Headspace samples with known concentrations to be used for the characterization of the instrument or its calibration were generated using an olfactometer described elsewhere.29 Briefly, solutions containing typically 1 g/L of the analytes of interest in 1-butanol were infused at 1 µL/min by a syringe pump (11plus, Harvard Apparatus, Holliston, MA) into an injection chamber held at 150 °C by a heating tape (H-So 4045-1.0, Hillesheim, Wagha¨usel, Germany) and continuously flushed at 1 L/min with nitrogen. The resulting vapor was then mixed with 9 L/min of air in a diluter held at 95 °C. The inlet of the headspace sampling gas chromatograph was placed in the middle of the 35 mm wide outlet of the diluter, after a right angle neck. Preparation of Fragrance-Loaded Microcapsules. Core-shell fragrance-loaded microcapsules were prepared by interfacialpolycondensationofanetherifiedmelamine-formaldehyde resin using a copolymer of acrylic acid and acrylamide as the colloidal stabilizer.30-32 Particle size analysis by optical microscopy and laser diffraction measurements (MasterSizer, Malvern Instruments Ltd., Malvern, U.K.) revealed a particle size distribution centered around 21 µm. Preparation of Fragranced Cotton Towels. Cotton towels fragranced by application of fabric softener containing either free perfume or encapsulated perfume were prepared by a standard European-style laundry cycle in batches of 60 29 cm × 29 cm items weighing about 45 g each. New towels (Lavana AG, Wetzikon, Switzerland) which had previously been conditioned by five successive laundry cycles at 95 °C in the presence of unfragranced detergent were used. A computerized washing machine (Miele Novotronic W970, ID157, Gu¨tersloh, Germany) ensured perfect control and reproducibilty of the washing cycles. A 30 min long main wash at 40 °C with 16 L of water in the presence of 100 g of unfragranced laundry detergent was followed by three 4 min long rinses of 12 L each. The fragranced fabric softeners, which had been prepared by adding free or encapsulated fragrance to a standard unfragranced base of the esterquat type 30 min earlier to allow equilibration, were added at the very beginning of the final rinse in the form of a dispersion in 1 L of water to facilitate a homogeneous spreading on all cotton towels. The final spinning cycle was as follows: spin, distribute 30 s at 90 rpm, 20 s at 350 rpm, 20 s at 450 rpm, 20 s at 600 rpm, 30 s at 800 rpm, 300 s at 1000 rpm. Investigation of Perfume Release from Cotton Towels. Shortly after the end of the washing cycle, 1.5 cm × 28 cm strips were cut from a towel. These dimensions were chosen so that weight of the swatch would represent precisely 1:20 of the weight (29) Vuilleumier, C.; van de Waal, M.; Fontannaz, H.; Cayeux, I.; Rebetez, P. A. Perfum. Flavor. 2008, 33, 54–61. (30) Sinclair, P. (The Wiggins Teape Group, Ltd.). U.K. Patent Application GB 2 073 132, March 26, 1981. (31) Dietrich, K.; Herma, H.; Nastke, R.; Bonatz, E.; Teige, W. Acta Polym. 1989, 40, 243–251. (32) Kumar, A.; Katiyar, V. Macromolecules 1990, 23, 3729–3736.

Analytical Chemistry, Vol. 82, No. 2, January 15, 2010

731

Table 1. Chemical Structures and Selected Properties of the Nine Investigated Perfumery Raw Materials

of the entire towel. The strips were then placed as a loose coil in a 2 cm × 25 cm jacketed glass tube held at 25 °C with a circulating water bath. Air circulation, at 1 L/min, was then initiated through the glass tube. Immediately, its outlet was placed next to the inlet of the headspace sampling gas chromatograph and the monitoring of the evolution of the headspace concentrations was initiated. The determination of gas-phase concentrations from peak areas was performed using response factors measured during the analysis of a headspace sample containing about 100 ng/L of each of the nine perfumery raw materials under investigation. To facilitate the assessment of mass balances, the gas-phase concentrations were then converted to nanograms per minute released by an entire towel exposed to a hypothetical flow of 20 L/min using the figures mentioned above. RESULTS AND DISCUSSION Characterization of the Experimental Setup. Figure 2 displays a typical chromatogram recorded with the new experimental setup when a headspace sample containing about 100 ng/L 732

Analytical Chemistry, Vol. 82, No. 2, January 15, 2010

each of 11 long-chain alkanes, 11 long-chain methyl esters, 12 longchain ethyl esters, and 8 long-chain alcohols, generated by the olfactometer, was sampled for 1 min at 10 mL/min. It was especially satisfactory considering that the use of a low thermal mass gas chromatograph made a total method time of less than

Figure 2. Typical chromatogram recorded from a 10 mL headspace sample containing about 100 ng/L each of 11 long-chain alkanes, 11 long-chain methyl esters, 12 long-chain ethyl esters, and 8 long-chain alcohols. Abbreviations are defined in the Experimental Section.

Figure 3. Comparison of the normalized response factors recorded from a 10 mL headspace sample (1 min of sampling at 10 mL/min) containing about 100 ng/L each of 11 long-chain alkanes, 11 longchain methyl esters, 12 long-chain ethyl esters, and 8 long-chain alcohols. Error bars represent the standard deviation of five measurements.

2.5 min possible, including the first 1 min under isothermal conditions at 35 °C. Some overlap between alkanes and methyl esters can be seen. Those peaks were integrated manually and split at the valley. Considering the time required to cool the low thermal mass gas chromatograph, a new data acquisition could be started at least every 4.0 min plus the cryotrapping time. Cryotrapping Efficiency. The cryotrapping efficiency of the LMCS used for analyte preconcentration was assessed by comparing the response factors recorded for analogues of different volatilities within one family of substances and between different families. This task was facilitated by the use of a FID for detection which could detect all investigated classes of compounds. The first step of data processing consisted in determining for each analyte the corresponding peak areas, expressed in picoamp seconds. Response factors were then calculated by dividing these values with the amounts of analytes sampled that were derived from the knowledge of the headspace concentrations, the sampling rate, and the sampling time. The response factors measured for the different analytes were then normalized using correction factors taken from a database that uses methyl octanoate as the reference to allow them to be directly compared.33 Figure 3 displays these normalized response factors as a function of the number of carbon atoms in the main chain for alkane, methyl ester, ethyl ester, and alcohol analogues. A range, roughly determined by the boiling point (BP) of the analytes, can be observed over which all analogues exhibit a constant and maximal response factor. For alkanes this range spans from dodecane (BP 215-217 °C) to nonadecane (BP 330 °C), for ethyl esters from ethyl heptanoate (BP 188-189 °C) to ethyl tetradecanoate (BP 320 °C), and for methyl esters from methyl octanoate (BP 193-194 °C) to methyl pentadecanoate (BP 339 °C). More volatile analytes, with a lower BP, exhibited lower normalized response factors. This was most likely the consequence of too small a breakthrough volume at the applied flow rate and cryotrapping temperature. Heavier analytes also showed lower normalized response factors. One initial hypothesis for this (33) Cicchetti, E.; Merle, P.; Chaintreau, A. Flavour Fragance J. 2008, 23, 450– 459.

Figure 4. Comparison of the normalized response factors recorded from a headspace sample containing about 100 ng/L of 42 analytes when the sampling time, at 10 mL/min, was varied between 15 s and 30 min. Selected data illustrate one analyte which is optimal for the experimental setup (ethyl decanoate), one analyte which has too high a volatility to be quantitatively cryotrapped (ethyl hexanoate), and one analyte which has too low a volatility and is subject to loss caused by adsorption (ethyl pentadecanoate). Error bars represent the standard deviation of five measurements.

phenomenon was incomplete desorption. However, this was ruled out by the observation that the normalized response factors for these compounds did not increase when the temperature of the oven in which the cryotrap is mounted was raised to 250 °C, nor did it decrease when that temperature was lowered to 150 °C. Therefore, the effect was attributed to adsorption effects in the zone around the valve. Adsorption phenomena could also explain why the alcohols exhibited a slightly different behavior compared to the other three families. Efficient cryotrapping of pentanol with a boiling point of only 138 °C was possible, and the normalized response factors stayed constant at least up to dodecanol (BP 260-262 °C). Overall, considering the physical-chemical parameters of perfumery raw materials, the working range of the instrument is ideally suited to the investigation of this class of compounds, with the exception of high boiling point musks and very volatile top notes. The performance of the instrument was further characterized by investigating the influence of the cryotrapping time on the normalized response factors, as reported in Figure 4. For analytes in the optimal working range of the instrument, represented in Figure 4 by ethyl decanoate, normalized response factors remained perfectly constant when the sampling times were varied between 15 s and 30 min, corresponding to sampled volumes ranging between 2.5 and 300 mL at a sampling rate of 10 mL/ min. This is proof of the excellent linearity of the instrument and of its suitability to quantitative analyses. In the case of highvolatility compounds, such as ethyl hexanoate, the normalized response factors decreased with increasing sampling time, indicating that not only are they not quantitatively cryotrapped but also partially released from the cryotrap while sampling is still proceeding. This drawback could possibly be improved by replacing the bare capillary used in the cryotrap by one that would be coated with a retentive stationary phase. The impact of such an approach on the remobilization of the least volatile analytes would, however, have to be assessed. Low-volatility compounds, such as ethyl pentadecanoate, exhibited the exact opposite trend. Analytical Chemistry, Vol. 82, No. 2, January 15, 2010

733

Figure 5. Release over a 178 min time period of eight perfumery ingredients from a wet cotton towel exposed to a 20 L/min air stream right after fabric softener application. The fragrance was formulated in free form. The crosses represent individual chromatograms, separated by 4.5 min intervals.

In this case, the normalized response factors increased as the sampling time became longer until values comparable to the ones observed for higher volatility compounds were reached. At that point the amount of analytes lost by adsorption had become negligible when compared to their total amount in large sampled volumes. Release of Odorant Fragrance Molecules from Cotton Towels. Figure 5 displays the results of the monitoring of the release rate, expressed in nanograms per minute, of eight fragrance raw materials from a freshly washed and still wet cotton towel exposed to a 20 L/min air stream. The perfume had been deposited on the fabric through a fabric softener formulated with free fragrance raw materials. The sampling time was 30 s, which was the longest possible duration that did not result in the trap being clogged up with ice during the initial stages of the drying process. Each cross on the plot represents data extracted from individual chromatograms acquired at 4.5 min time intervals. This 734

Analytical Chemistry, Vol. 82, No. 2, January 15, 2010

way of displaying the data was selected to illustrate the performance of the instrument, but for better clarity the following plots will only contain lines without crosses. The release rate plots are overlaid with the integrated amounts of released fragrance in dotted lines. These latter quantities are expressed as the percentage of the hypothetical fragrance load on the towel assuming an ideal but generally unrealistic case in which 100% of the fragrance originally present in the fabric softener would have been deposited on the cotton towels. Figure 5 does not display any data for benzyl alcohol even though this compound was present in the fragrance mixture because it was never detected. This finding can readily be explained by the relatively low log P value and high water solubility of this molecule which makes it likely to be washed off in the washing machine rather than being deposited on cotton. Figure 5 shows that for all analytes except Lilial, the least volatile one, the release rates dropped below the detection limits corresponding to about 20 ng/min before the end of the 3 h long

Figure 6. Release over a 356 min time period of eight perfumery ingredients from a wet cotton towel exposed to a 20 L/min air stream right after fabric softener application. The fragrance was formulated in microencapsulated form.

monitoring. This was the final consequence of a steady decrease of the release rates, starting from initial values in the range of a few tens of micrograms per minute. The fact that for Lilial, in contrast with all other analytes, the highest concentration was observed in the second rather than the first data point was attributed to adsorption phenomena on the walls of the glass tube in which the cotton swatch had been placed. The dynamics of the system were, however, excellent, with signals dropping as soon as the investigated cotton swatches were removed from the glass tube. Almost 100% of the δ-damascone expected to be present on the towel assuming an ideal deposition yield of 100% could be accounted for, which is a further proof of the good suitability of the instrument for quantitative analyses. The lower recovery of phenethylol, benzyl acetate, allyl heptanoate, and Jasmonitrile can easily be understood by the partial wash-off of these more watersoluble compounds during the laundry procedure. The true meaning of the recovery of about 60% of Verdox and 70% of Doremox and Lilial is more difficult to assess as the parameters

that govern perfume deposition and release are known to be extremely complex.34 Figure 6 displays the results of the exact same experiment, but this time when the fragrance was formulated in the fabric softener in microencapsulated form rather than in free form. For the three analytes with a low log P value and a relatively high water solubility, benzyl alcohol, which again was never detected, phenethylol, and benzyl acetate, the release profiles were roughly similar to the ones observed during the study with free, nonencapsulated fragrance. These types of compounds are always partially present in the aqueous phase of the microcapsules suspension and are also prone to partitioning into the rinse liquor during the laundry cycle. They had therefore been deposited on fabric mainly in free form and only marginally as part of the content of the fragrance-loaded microcapsules. (34) Escher, S. D.; Oliveros, E. J. Am. Oil Chem. Soc. 1994, 71, 31–40.

Analytical Chemistry, Vol. 82, No. 2, January 15, 2010

735

Figure 7. Release of eight perfumery ingredients from a dry cotton towel on which fragrance-loaded microcapsules had been deposited but which initially was not releasing any detectable amount of fragrance at the end of the experiment depicted in Figure 6. The towel was rubbed at 0, 178, and 356 min to initiate fragrance release by rupture of the microcapsules and then exposed to a 20 L/min air stream.

Figure 6 also shows that for five other fragrance compounds, allyl heptanoate, Jasmonitrile, Verdox, δ-damascone, and Doremox, the release kinetics were dramatically different when they were formulated as a microcapsule suspension compared to when the fragrance was not encapsulated. While initial rates were up to 1 order of magnitude lower, fragrance release lasted for a much longer time. In particular, when the monitoring was ended after 6 h, two of the five compounds, δ-damascone and Doremox, were still released from the towels at levels around 50 ng/min. The evolution of the rate of fragrance release was also remarkable, evidencing three successive regimes. During the first one, a rapid decrease of the fragrance release was observed. This regime was attributed to the evaporation of the fraction of the fragrance which had partitioned into the aqueous phase during the laundry cycle and had therefore been deposited on the towels in free form. A second regime started after about 40 min and featured a burst of fragrance release peaking after around 90 min. This regime was 736

Analytical Chemistry, Vol. 82, No. 2, January 15, 2010

explained by the diffusion of the fragrance compounds through the membrane of the microcapsules followed by their evaporation. For all five compounds, the diffusion rates increased as the water content decreased upon drying of the towels. However, they were also influenced by the specific interactions between the membrane and the individual fragrance compounds, resulting in peaks with different amplitudes. The third regime started after the maximal fragrance release had been reached. It was characterized by the microcapsule walls becoming less and less permeable to the fragrance compounds, resulting in continuously decreasing release rates. This was probably a consequence of the structure of the microcapsule membranes becoming more dense under conditions in which the water level was highly reduced due to extended drying. The release curve observed in Figure 6 for Lilial evidenced a very slow and low release of this compound from the microcapsules. This is most likely due to the combination of a high log P

value, the lowest volatility of all investigated compounds, and the presence in its chemical structure of an aldehyde functional group thatcaninteractespeciallystronglywiththemelamine-formaldehyde structures of the microcapsule membranes. Total fragrance recoveries were generally lower during the monitoring of fragrance release from cotton towels with applied microencapsulated perfume (Figure 6) compared to free perfume (Figure 5). At first glance, this finding could be considered surprising since microcapsules would be expected to be deposited in high yield on the cotton towels. The first impression is, however, wrong. Indeed, Figure 6 only displays the fragrance that was released during the drying of the towels and which had either been deposited in free form or that had diffused through the microcapsule walls before they had become hermetic. At that point, additional perfume release could be triggered by mechanically breaking the walls of the microcapsules. This was achieved by rubbing the cotton swatch against itself, thus mimicking the mechanical stress applied to clothes as they are worn. The results of this experiment, repeated three times at t ) 0 min, 178 min, and 356 min, are displayed in Figure 7. Each rubbing experiment resulted in the release of more fragrance. δ-Damascone, among others, was an especially interesting case as the amount of fragrance released in 3 h as a consequence of the first rubbing event was roughly equivalent to the total amount that had been released during the initial 6 h drying phase. A key feature is that the release profiles induced by rubbing followed the same steady decline pattern previously observed in the case of free fragrance, without a burst. This is clear proof for the rupture of microcapsule walls during the rubbing, thus depositing an additional dose of free fragrance on the cotton towels. CONCLUSIONS AND PERSPECTIVES The results displayed in Figures 5-7, obtained with an innovative instrumental setup, are unprecedented in their simultaneous level of precision and time resolution when investigating the release of a mixture of fragrance compounds from a real application. The methodology is a powerful tool to understand the impact of microencapsulation on fragrance release. It also

allows the performance of these systems to be further improved by providing a detailed understanding of the influence of the processing parameters. Since the described analytical system could have many other applications, not only to investigate fragrance release but also in environmental chemistry or for process monitoring, further improvements are anticipated in the future. The first one will consist in replacing the FID, which was convenient at this stage because of its universal response and excellent linearity, by a mass spectrometer. This should result in an improvement of the sensitivity by orders of magnitude. Second, the possibility to efficiently trap analytes more volatile than the ones investigated by replacing the bare capillary placed in the cryotrap by a capillary with a coating will be assessed. Third, efforts will be made to reduce the size of the sampling and trapping setup to decrease the total surface exposed to the headspace samples, with the intention of minimizing the impact of adsorption effects on the analysis of those perfumery molecules with the lowest volatility such as musks. This task could be performed in parallel with an improvement of the inertization of the system. Fourth, the replacement of the LMCS by other cryotrapping systems that could enable a more active remobilization of the analytes will be investigated. Finally, the system will be parallelized by fully utilizing the features of the two-position 10-port valve as explained in the Experimental Section. ACKNOWLEDGMENT The authors are indebted to Gre´gory Seiler and Denis Perroux for the great mechanical and electrical engineering support provided during the customization of the gas chromatographic system. They also acknowledge their appreciation for Dr. Christine Vuilleumier and Pierre-Andre´ Rebetez who provided assistance with the olfactometer. Finally, they thank Dr. Fre´de´ric Begnaud for helpful discussions. Received for review December 2, 2009.

October

29,

2009.

Accepted

AC902460D

Analytical Chemistry, Vol. 82, No. 2, January 15, 2010

737