Optical Determination of Low-Level Water Concentrations in Organic

A new method for the sensitive optical detection of low-level water ...... An optical fibre sensor for remotely detecting water traces in organic solv...
1 downloads 0 Views 79KB Size
Download PDF
Anal. Chem. 2001, 73, 5339-5345

Optical Determination of Low-Level Water Concentrations in Organic Solvents Using Fluorescent Acridinyl Dyes and Dye-Immobilized Polymer Membranes Daniel Citterio,†,‡ Katsuya Minamihashi,† Yuka Kuniyoshi,† Hideaki Hisamoto,†,§ Shin-ichi Sasaki,† and Koji Suzuki*,†,‡

Department of Applied Chemistry, Keio University, 3-14-1 Hiyoshi, Kohoku-ku, Yokohama 223-8522, Japan, and Kanagawa Academy of Science and Technology (KAST), 3-2-1 Sakado, Takatsu-ku, Kawasaki 213-0012, Japan

The fluorescent acridinyl indicators 4-(9-acridinyl)-N-(5hexenyl)-N-methylaniline (KD-F0011), 6-(9-acridinyl)1,2,2,3-tetramethyl-2,3-dihydro-1H-perimidine (KDF0021), and 6-(9-acridinyl)-2-(3-butenyl)-1,2,3-trimethyl2,3-dihydro-1H-perimidine (KD-F0022) were designed, synthesized, and applied for highly sensitive optical determination of low-level water in organic solvents. All these dyes were found useful as fluorescence indicators for the detection of water below 1% (v/v) in different solvent media with a low detection limit of 0.002% (v/v) or 20 mg/L (22 ppm by weight) for KD-F0021 in THF solution. Sensing membranes made from poly(ethylene glycol) dimethacrylate by photocopolymerization with the indicator KD-F0011 were also prepared. Using the membrane sensor, the lowest detection limit of 0.001% (v/v) or 14 mg/L (20 ppm) water was achieved in diethyl ether samples. This system enables the continuous monitoring of the water content in a flow-through arrangement, where single-wavelength excitation (404 nm) and single-wavelength detection (532 nm) can be used for the fluorescence determination, allowing a simple measurement setup. In a continuous-flow experiment using THF samples, fully reversible and fast signal changes with t95% ) 1-2 min for water concentrations up to 0.50% (v/v) were observed. A detection limit of 0.004% (v/v) or 40 mg/L (45 ppm) water in THF was achieved. These characteristics make this type of sensor a useful tool for the online continuous monitoring of water present as an impurity in organic media, which is difficult to achieve using a Karl Fischer instrument. The detection and quantification of small amounts of water present as impurities in organic solvents are important in several fields of chemistry and industry. To fulfill government regulations * Corresponding author: (e-mail) [email protected]; (fax) +81-45-5645095. † Keio University. ‡ Kanagawa Academy of Science and Technology. § Present address: Department of Applied Chemistry, Graduate School of Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan. 10.1021/ac010535q CCC: $20.00 Published on Web 09/22/2001

© 2001 American Chemical Society

and to provide products of high quality, samples such as gasoline are checked routinely for their content of water. The presence of water is also monitored in industrial production lines. Producers of solvents and chemical reagents have to guarantee low levels of water in their products to satisfy the needs of their customers. Currently, the determination of the moisture content in sample substances is mostly done electrochemically according to the wellestablished Karl Fischer method.1 Other classical procedures, such as drying to constant weight, distillation, or physical methods, such as measuring density or electrical conductivity, often yield unsatisfactory results or are too time-consuming for large-scale applications. The Karl Fischer method has several useful characteristics: When up-to-date instrumentation based on coulometric rather than volumetric titration is used, it is a quite rapid method. It features high sensitivity, and modern coulometric analyzers can detect amounts of water in organic media down to a level of 5 ppm. Furthermore, with the exception of strongly oxidizing or reducing agents, or substances that directly react with the ingredients of the Karl Fischer reagent, this method is applicable to a large variety of samples. On the other hand, the method has some disadvantages: First, this method allows batch analysis only. Although most of the required procedures may be automated, a real-time continuous on-line monitoring is not possible. Second, specialized equipment is required, and being a wet chemical method, consumed reagents have to be used. For many applications in industry, a flow-through sensor for on-line monitoring of low-level water in organic media would be most interesting. For these purposes, easy-to-use portable instruments with no need of excessive sample pretreatment are required. Although there exist many different types of sensors for the quantification of humidity in air,2 only a few sensing systems suitable for the measurement of water in organic solvents have been presented in the literature. A method suitable for on-line analysis based on direct potentiometry was first presented by Kakabadse and co-workers in 1988.3 Systems relying on optical (1) Fischer, K. Angew. Chem. 1935, 48, 394-396. (2) Bu ¨ hlmann, P.; Pretsch, E.; Bakker, E. Chem. Rev. 1998, 98, 1593-1687 and references therein. (3) Kakabadse, G. J.; Al-Aziz, M. S.; Hamilton, I. C.; Olatoye, E. O.; Perry, R.; Tipping, A. E.; Vaudrey, S.; Al-Yawer, N. F. N. Analyst 1988, 113, 13651368.

Analytical Chemistry, Vol. 73, No. 21, November 1, 2001 5339

detection include the following examples: Kessler et al. developed a method based on a membrane-immobilized fluorescent solvent polarity probe to determine water in the 1-20% level range.4 Bai and Seitz described a fiber-optic sensor based on polymer swelling.5 Chang et al. presented a fluorescence lifetime-related solid sensor for water.6 The sensing principle is based on the fluorescence lifetime and intensity changes of a metal-ligand compound containing osmium(II) bound to a resin and then incorporated into a thin sol-gel layer. The detection limit is dependent on the solvent, ranging from 0.02% (v/v) in ethyl acetate to 0.68% in methanol. Angel and co-workers reported on the use of the luminescent ruthenium complex Ru(phen)2dppz2+ embedded into a Nafion layer as a water-sensitive transducer for imaging sensors.7,8 Although very sensitive to the presence of water, the interference of oxygen (O2) due to fluorescence quenching might be a serious drawback of this system. Recently, Liu and co-workers published a paper describing the application of solvatochromogenic flavone dyes for the detection of water in acetone.9 However, the application of the flavone dyes was limited to experiments in solution using acetone as the solvent. An optical sensor based on the quenching of fluorescence by hydrogen bond donating compounds such as alcohols, has been used by Orellana and coworkers to quantify lower alcohols in gasoline products.10 This work reports on the development of an optical chemical sensing scheme based on fluorescence for quantifying water present at ppm levels in organic solvents. The fluorescence-type optical transduction of the chemical signal is advantageous for the development of an optical sensor due to the high sensitivity of fluorescence emission detection and due to its high potential for miniaturization.11 To allow continuous on-line monitoring, the signal change of the sensor system has to be reversible, and increasing as well as decreasing concentration changes have to be detectable. Additionally, a reagent-free operation of the system is desirable. Our approach is based on the use of fluorescent dyes, which undergo a pronounced intramolecular charge transfer (ICT) upon excitation with visible light. The partial charge separation in the excited state results in strong hydrogen bond interactions between the indicator dye, acting as the hydrogen bond acceptor, and water, acting as the hydrogen bond donor. The indicators 4-(9acridinyl)-N-(5-hexenyl)-N-methylaniline (KD-F0011), 6-(9-acridinyl)-1,2,2,3-tetramethyl-2,3-dihydro-1H-perimidine (KD-F0021), and 6-(9-acridinyl)-2-(3-butenyl)-1,2,3-trimethyl-2,3-dihydro-1H-perimidine (KD-F0022) (Figure 1) were designed to show high sensitivity to the presence of water and, in the case of KD-F0011 (4) Kessler, M. A.; Gailer, J. G.; Wolfbeis, O. S. Sens. Actuators, B 1991, 3, 267-272. (5) Bai, M.; Seitz, W. R. Talanta 1994, 41, 993-999. (6) Chang, Q.; Murtaza, Z.; Lakowicz, J. R.; Rao, G. Anal. Chim. Acta 1997, 350, 97-104. (7) Carter, J. C.; Egan, W. J.; Nair, R. B.; Murphy, C. J.; Morgan, S. L.; Angel, S. M. Book of Abstracts, PITTCON ’99 (The Pittsburgh Conference) 1999, 1143. (8) Glenn, S. J.; Cullum, B. M.; Carter, J. C.; Angel, S. M. Book of Abstracts, PITTCON ’99 (The Pittsburgh Conference) 1999, 659. (9) Liu, W.; Wang, Y.; Jin, W.; Shen, G.; Yu, R. Anal. Chim. Acta 1999, 383, 299-307. (10) Orellana, G.; Gomez-Carneros, A. M.; de Dios, C.; Garcia-Martinez, A. A.; Moreno-Bondi, M. C. Anal. Chem. 1995, 67, 2231-2238. (11) Prasanna de Silva, A.; Nimal Gunaratne, H. Q.; Gunnlaugsson, T.; Huxley, A. J. M.; McCoy, C. P.; Rademacher, J. T.; Rice, T. E. Chem. Rev. 1997, 97, 1515-1566.

5340 Analytical Chemistry, Vol. 73, No. 21, November 1, 2001

Figure 1. Chemical structures of the dye molecules investigated in this study: (a) KD-F0011; (b) KD-F0021 (R1dR2dR3 ) CH3) and KD-F0022 (R1dR2 ) CH3; R3 ) CH2CH2CHdCH2).

and KD-F0022, to be covalently immobilized to a hydrophilic membrane support that is insoluble to most organic solvents and water. As a result, these fluorescence dyes and the sensing membranes are useful for the determination of low-level water with detection limits of 20-70 ppm, which are to our knowledge the lowest values of fluorescence-based sensors to date. EXPERIMENTAL SECTION Materials. Reagents used for the synthesis of the indicators were purchased from Tokyo Kasei (Tokyo, Japan), Wako Pure Chemical Industries (Osaka, Japan), Junsei Chemical (Tokyo, Japan), and Aldrich (Milwaukee, WI). The detailed synthetic procedures, the method of purification, and 1H NMR data for KDF0011, KD-F0021, and KD-F0022 are provided separately as Supporting Information. The membrane material poly(ethylene glycol) dimethacrylate (PEGDMA) was obtained from Aldrich. The photoinitiator benzoin methyl ether was bought from Tokyo Kasei. All reagents for synthesis and membrane preparation were used as received. THF was distilled over sodium/benzophenone under a dry nitrogen atmosphere immediately prior to use. Diethyl ether was distilled over LiAlH4 whereas other solvents were distilled over CaCl2 under a nitrogen atmosphere and stored over molecular sieves (4 Å) for at least 12 h. Instruments. The absorbance spectra of dissolved dye solutions and of sensor layers were recorded on a Hitachi U-2001 double-beam spectrophotometer (Hitachi Co. Ltd., Tokyo, Japan). Fluorescence excitation and emission spectra were measured at 25 ( 1 °C on a Hitachi F-4500 spectrophotometer using either a standard fluorescence quartz cell (d ) 1 cm) in a temperaturecontrolled cell holder or a small-volume flow-through cell. Samples were kept in a temperature-controlled water bath before measurement. The determination of water in THF samples for reference purposes was done with a CA-06 moisture meter (Mitsubishi Chemical, Tokyo, Japan) based on coulometric Karl Fischer titration. Preparation of Sensing Layers. The covalent immobilization of the dyes to the polymer membrane was done in a way similar to that described by Hisamoto and co-workers.12 The dye KDF0011 or KD-F0022, respectively (1 wt %), benzoin methyl ether (3 wt %), and the membrane material (96 wt %) were mixed by ultrasonification. Due to solubility problems, a small amount of CHCl3 was added to the mixture in the case of KD-F0022 and then evaporated under reduced pressure after complete dissolution. Several drops of the resulting membrane cocktail were (12) Hisamoto, H.; Manabe, Y.; Yanai, H.; Tohma, H.; Yamada, T.; Suzuki, K. Anal. Chem. 1998, 70, 1255-1261.

Table 1. Spectral Properties of KD-F0011 and KD-F0022 in Various Solvents KD-F0011

KD-F0022

solvent

solvent polarity ET(30) (kcal/mol)

CT absorption λmax (nm)

CT fluorescence λmax (nm)

diethyl ether THF ethyl acetate dichloromethane acetone acetonitrile

34.6 37.4 38.1 41.1 42.2 45.6

385 404 396 405 403 403

497 532 539 544 572 587

a

CT fluorescence λmax (nm)

Imaxa

1.00 0.36 0.27 0.40 0.03 0.01

421 432 426 433 433 431

556 592 577 538 541 547

1.00 0.17 0.12 0.18 0.07 0.07

In relation to the most intensive emission measured in diethyl ether (I ) 1.00).

dropped on a glass support and covered with a quartz plate, using a microscope cover glass (thickness 1 mm) on each side to keep the distance between the cover and the base support. Subsequent photopolymerization was performed in a nitrogen atmosphere by irradiation with UV light (30 W) during 3 h. Finally, the membranes were removed from the glass support by soaking in methanol with simultaneous ultrasonification. They were further washed with methanol to remove any unbound dye. All membranes were stored in methanol in a dark place. The average thickness of the membranes after soaking in methanol was found to be 170 µm. Preparation of Sample Solutions. The sample solutions of organic solvents containing different amounts of water were freshly prepared immediately prior to the measurement. All solvent transfers were performed under a dried nitrogen or argon atmosphere. For THF, the actual water content was verified using a Karl Fischer instrument. Before recording the fluorescence spectra, all samples were degassed by ultrasonification during 15 min. RESULTS AND DISCUSSION Spectral Properties of KD-F0011, KD-F0021, and KDF0022 in Solution. The physical-chemical background of the absorption and fluorescence spectroscopic properties of 9-acridinyl, anthryl, and other aryl derivatives of aromatic amines have been studied extensively by several authors.13-16 The donor (D)acceptor (A) compounds KD-F0011, KD-F0021, and KD-F0022 containing an aromatic amine as the electron donor and acridine as the acceptor show a low-energy charge-transfer (CT) absorption band that undergoes a small red shift with increasing solvent polarity. The ICT leads to a large dipole moment in the excited state. According to the Franck-Condon principle, the CT fluorescence emission spectra are expected to show far stronger influences of the environment than the CT absorption spectra, leading to larger Stokes shifts with increasing solvent polarity. This fact is responsible for the strong positive solvatofluorchromic shift observed in the case of KD-F0011 investigated in a series of organic solvents with increasing polarity (Table 1) due to the ICTinduced dipolar character of the dye in the excited state. At the same time, the absorption spectra show only a very moderate positive solvatochromism. (13) (14) (15) (16)

Imaxa

CT absorption λmax (nm)

Herbich, J.; Kapturkiewicz, A. Chem. Phys. 1991, 158, 143-153. Kapturkiewicz, A. Chem. Phys. 1992, 166, 259-273. Herbich, J.; Kapturkiewicz, A. Chem. Phys. 1993, 170, 221-233. Herbich, J.; Kapturkiewicz, A. J. Am. Chem. Soc. 1998, 120, 1014-1029.

The dihydroperimidine subunit is known to be a very effective π-electron-donating group17 with higher donor strength than an anilinic subunit. This fact was made use of in the design of the fluorescent dyes KD-F0021 and KD-F0022 for highly sensitive water sensing. Longer wavelength absorption and fluorescence emission as well as a stronger intramolecular charge transfer can be expected relative to KD-F0011. The position of the CT absorption band was shifted by ∼30 nm in comparison to KDF0011. Large Stokes shifts are found as well, leading to fluorescence emission bands between 540 and 600 nm depending on the solvent. However, no pronounced positive solvatochromism can be observed for the fluorescence emission spectra (Table 1). At the same time, the fluorescence quantum yields are lower compared to KD-F0011 with the anilinic donor subunit (data not shown). It is assumed that the enlarged steric hindrance to coplanarity between the electron-donating perimidine subunit and the electron-accepting acridine subunit for KD-F0021 and KDF0022 is the main reason for different spectral solvent effects in comparison with KD-F0011. Spectral Changes Induced by the Presence of Water. Because of the enhanced electron density at the nitrogen center in the acridine subunit in the ICT excited state, this atom becomes a stronger acceptor for hydrogen bonds. Hydrogen bonding increases the possibility for nonradiative relaxation of the excited dye, leading to a decrease in fluorescence quantum yield in the presence of hydrogen bond donating compounds, as for example water molecules. In Figure 2a, the absorption spectra of KD-F0011 in THF solution with a water content between 0 and 10% (per volume) are presented. The CT absorption band can be observed as the longest wavelength band in the spectra. Similar to KDF0011, the absorption spectra of KD-F0021 and KD-F0022 do not undergo significant changes on changing from pure THF to THF containing 10% (v/v) water. A small positive solvatochromic shift can be observed for a water content above 5% (v/v). The fluorescence emission spectra of KD-F0011 and KD-F0021 in THF solution are shown in Figure 2b and c, respectively. The spectral effects observed for KD-F0022 are very similar to the ones of KDF0021. The concentration of the indicators was selected at a sufficiently low level (10-5 M range) in order to prevent selfquenching effects. In the case of KD-F0011, in contrast to the absorption spectra, the position of the emission band undergoes a pronounced red shift with increasing concentration of water in THF. The position of the emission band shifts from 532 nm in (17) Griffiths, J. Modern Colorants: Synthesis and Structure; Blackie Academic & Professional: London, 1995; Chapter 2.

Analytical Chemistry, Vol. 73, No. 21, November 1, 2001

5341

Figure 2. (a) UV/visible absorption spectra of KD-F0011 (c ) 4.0 × 10-5 M) in THF solution in the presence of increasing amounts of water (0-10%), (b) fluorescence emission spectra of KD-F0011 (exc 404 nm, c ) 4.0 × 10-5 M), and (c) KD-F0021 (exc 429 nm, c ) 6.0 × 10-5 M) in THF solution in the presence of increasing amounts of water.

Figure 3. Relative fluorescence emission intensity of the acridinyl dyes as a function of the water content in THF: (b) KD-F0011 (exc 404 nm/em 534 nm); (9) KD-F0021 (exc 429 nm/em 598 nm); (2) KD-F0022 (exc 432 nm/em 592 nm). The inset shows the magnified low-level water range with highest sensitivity between 0 and 1% (v/ v) water.

pure THF to 568 nm in THF containing 10% (v/v) water. At very high water contents, the determination of the emission maximum becomes uncertain due to the very low fluorescence intensity. This behavior can be attributed to a specific water-fluorophore interaction18 and also partially to the increase in polarity of the solvent mixture. At the same time, the fluorescence intensity is strongly quenched by the presence of water, acting as a hydrogen bond donor. The dihydroperimidine dyes KD-F0021 and KD-F0022 show an even more efficient quenching of the fluorescence emission by water which is expected according to the design concept of the indicaotrs with their stronger electron donor groups resulting in a more pronounced charge separation in the ICT excited state. The position of the emission band however, is not significantly influenced. The dependence of the relative fluorescence intensity on the concentration of water in THF is summarized in Figure 3. As can be seen from the graph, the sensitivity toward the presence of water is highest below 1.00% (v/v), leading to a reduction in relative fluorescence intensity of nearly 65% for KD-F0011 and 90% for KD-F0021 and KD-F0022. Due to their close structural similarity, no significant differences between the latter two dyes (18) Lakowicz, J. R. Principles of Fluorescence Spectroscopy, 2nd ed.; Kluwer Academic/Plenum Publishers: New York, 1999; Chapter 6.

5342 Analytical Chemistry, Vol. 73, No. 21, November 1, 2001

are observed. At water contents higher than 5.00% (v/v), the fluorescence emission is quenched almost completely for all fluorophores. The quenching of the fluorescence of the investigated fluorophores due to the presence of water in THF or various other organic solvents can be described by the Stern-Volmer equation at low-level water concentration. The Stern-Volmer quenching constant K indicates the slope of the linear calibration function and is therefore a measure for the sensitivity of the method (Table 2). For water contents higher than 1% (v/v), this simple linear relationship is not suitable anymore to describe the experimentally observed data. The highest sensitivity was observed for the dihydroperimidine indicator KD-F0021, and consequently, this fluorophore was applied for the determination of water in different kinds of organic solvents widely used in organic chemical synthesis (THF, ethyl acetate, CH2Cl2). In the case of dichloromethane, the linear response range was limited to ∼0.15% (v/ v) due to the low solubility of water in this solvent, resulting in the saturation of the sample. The detection limits for the fluorometric water determination method with KD-F0021 given in Table 2 were estimated from the linear Stern-Volmer plot on the basis of 3 times the standard deviation as measured for the solution of the indicator in the pure solvent. The differences can be explained by the increase in polarity in the solvents, resulting in a relatively lower fluorescence quenching in more polar solvents. In the case of THF, with the lowest detection limit of the examples studied, the value is only ∼4 times higher compared to a modern coulometric Karl Fischer analyzer with a detection limit of 5 ppm, corresponding to 0.0004% (v/v) or 4 mg/L in THF. Membrane-Immobilized KD-F0011 and KD-F0022. To develop a reversibly working chemical sensor for the determination of water in organic solvents, allowing real-time, on-line monitoring of the water concentration, the fluorophores KD-F0011 and KD-F0022, both possessing an olefinic double bond, were covalently immobilized to a polymeric membrane matrix. PEGDMA with an average of nine ethylene glycol units between the two terminal methacrylate groups was found to be a suitable material for the covalent binding of the olefinic side chain of the indicator by photocopolymerization. This polymer shows hydrophilic character to allow the uptake of water and is practically insoluble to most organic solvents. It allowed the fabrication of visually clear, transparent membranes in combination with the fluorophores. After the first thorough washings with methanol as described in the Experimental Section, no further leaching of dye

Table 2. Calibration Parameters and Detection Limits for Water Determination in Various Solvents indicator KD-F0011 KD-F0021 KD-F0022 KD-F0011 (immobilized to PEGDMA)

solvent

K (M-1)a

THF THF ethyl acetate dichloromethane THF diethyl ether THF ethyl acetate acetonitrile

3.19 ( 0.06 11.86 ( 0.15 11.33 ( 0.38 6.64 ( 0.07 7.05 ( 0.27 2.95 ( 0.04 1.43 ( 0.04 1.41 ( 0.01 0.77 ( 0.04

R2b 0.995 0.998 0.986 0.999 0.988 0.998 0.991 0.999 0.976

detection limit ndc 0.002% (v/v); 20 mg/L; 22 ppm 0.003% (v/v); 30 mg/L; 34 ppm 0.004% (v/v); 40 mg/L; 30 ppm ndc 0.001% (v/v); 14 mg/L; 20 ppmd 0.003% (v/v); 29 mg/L; 33 ppmd 0.003% (v/v); 30 mg/L; 34 ppmd 0.005% (v/v); 54 mg/L; 70 ppmd

a Stern-Volmer quenching constant: F /F ) 1 + K[H O], with F and F representing the fluorescence intensity in the absence and in the 0 2 0 presence of water, respectively. b A minimum of five data points was used for each linear calibration curve. c nd, not determined. d Determined in a batch system.

from the polymer support could be observed. The flexibility of the membranes was sufficient to allow the handling of the sensing films with the flow-through cell. However, it was strongly dependent on the applied solvent, being highest in dichloromethane and considerably lower in THF. It is reported in the literature19 that higher flexibility and water uptake could be achieved by copolymerization of PEGDMA with 2-hydroxyethyl methacrylate. But the presence of free hydroxyl groups in the polymer film would strongly decrease the sensitivity due to inherent hydrogen bond interaction with the fluorophore. Compared to the spectral properties of the dye in solution, the position of the CT fluorescence emission band is almost independent of the water concentration in the case of membraneimmobilized KD-F0011. For THF samples, the induced spectral shift between 0 and 10% (v/v) water content is reduced to 6 nm compared to 36 nm for the dissolved dye. Furthermore, the emission wavelength is not influenced by the polarity of the sample solvent. No solvatochromic spectral shifts are observed, indicating that the microenvironment around the fluorophore in the membrane network is mostly determined by the polymer matrix and not by the sample. The Stern-Volmer linearized response curves for four organic solvents of different polarity were investigated, and the relevant parameters are summarized in Table 2. In all cases, the identical excitation (404 nm) and emission (532 nm) wavelength was used for the measurement. For the example of THF, the sensitivity of the dye-containing polymer film (K ) 1.4 M-1) toward water was reduced by a factor of 2 in contrast to the experiment in solution (K ) 3.2 M-1). Since the signal-to-noise ratio is strongly improved compared to the measurement in free solution, however, the detection limit for water achieved with the KD-F0011-sensing membrane is comparable to the one estimated for the more sensitive dye KD-F0021 in solution. It was calculated to be between 0.001% (v/v) for diethyl ether (14 mg/L or 20 ppm) and 0.005% (v/v) for acetonitrile (54 mg/L or 70 ppm). The reproducibility for water determinations with KD-F0011 immobilized to PEGDMA membranes was investigated. The slopes of the calibration functions for water in THF for three replicate calibrations with the same membrane in a batch system were checked for equality by a statistical test.20 No significant differences were observed, indicating a good reproducibility of (19) Berkowicz, B. D.; Peppas, N. A. J. Appl. Polym. Sci. 1995, 56, 715-720. (20) GraphPad Prism 3 for Macintosh, GraphPad Software Inc., San Diego CA. The method used is described in: Zar, J. H. Biostatistical Analysis, 2nd ed.; Prentice Hall: Englewood Cliffs, NJ. 1984; Chapter 18.

Figure 4. Response function, response time, and reversibility of the PEGDMA membrane sensor containing covalently immobilized KD-F0011, upon step changes of the water concentration in THF (measured in the continuous-flow mode at 25 ( 1 °C; λexc ) 404 nm; λem ) 532 nm).

the calibration function. For three water determinations with three different membranes, the relative standard deviation for the normalized emission reading at 532 nm (F0/F) was 1.4, 1.4, and 2.9% for 0.01, 0.15, and 1.0% (v/v) of water in THF (n ) 3 × 3). PEGDMA membranes incorporating KD-F0011 were mounted into a flow-through cell, and the fluorescence emission was monitored under continuous-flow conditions using THF with different water contents as the sample solution. The response profile is shown in Figure 4. The observed forward response time is 1 min (for 95% of the full signal change) between water-free THF and a 0.01% (v/v) water-containing sample and 2 min for a sample containing 0.50% (v/v) water. The intensity decrease is fully reversible, showing no loss of signal. The detection limit for water in THF measured in continuous flow was calculated based on 3 times the noise level of a freshly prepared sensing film in pure THF. It was estimated to be 0.004% (v/v), corresponding to 40 mg/L (45 ppm). Photodecomposition of the indicator is not observed under continuous irradiation of a membrane film in THF at 404-nm wavelength during 17 h in the stopped-flow mode. Sensor membranes kept in methanol in a dark place remained fully functional even after 3 months of storage. On the basis of the results obtained from the experiments in solution, use of KD-F0022 covalently immobilized to a PEGDMA membrane was expected to result in a sensing film of higher Analytical Chemistry, Vol. 73, No. 21, November 1, 2001

5343

Table 4. Determination of Water in THF by Coulometric Karl Fischer Moisture Analysis and by PEGDMA-Immobilized KD-F0011 H2O found (% (v/v))a

a

Figure 5. Response of membrane-immobilized KD-F0011 (1 wt %) to water (9), methanol (b), and ethanol (2) in THF (exc 404 nm). Table 3. Effect of the Presence of Ethanol and Methanol on the Quantification of Low-Level Water Concentrations in THF alcohola

addition (% (v/v))

H2O added (% (v/v))

foundb (% (v/v))

recovery (%)

EtOH

0.1

MeOH

0.1

0.050 0.100 0.050 0.100

0.052 0.099 0.072 0.142

103.8 98.7 144.2 141.8

a

Abbreviations: EtOH, ethanol; MeOH, methanol. b n ) 3.

sensitivity than KD-F0011. However, the low fluorescence quantum yield of the dihydroperimidine indicator did not allow the successful application of this dye in a membrane environment. Despite the higher concentration of the dye in the membrane in comparison to the solution, the relatively short optical path length in the thin film is insufficient to generate an emission output signal to be detected sensitively. Since the working principle of the sensor is based on hydrogen bond interaction between the fluorophore and water, it is obvious that other hydrogen bond donating compounds such as alcohols or amines will interfere in the determination of the water concentration. For this purpose, the effect of methanol and ethanol on the fluorescence emission signal was investigated. In Figure 5, the response of membrane-immobilized PEGDMA to the presence of methanol and ethanol in THF is shown. The hydrogen bond acidity of aliphatic alcohols decreases with increasing length of the alkyl chain.21 This characteristic is reflected in the degree of fluorescence quenching of KD-F0011 induced by methanol and ethanol. Therefore, an error in the determination of water in an organic solvent originating from the presence of various lower alcohols has to be considered when working with real samples. Additionally, the quantification of water using the sensing membrane in protic solvents such as ethanol or methanol is possible but is analytically not very useful since the sensitivity is decreased. Table 3 shows the apparent changes in water concentration caused by the presence of ethanol or methanol on the quantification of low water levels in THF with PEGDMA-immobilized KD-F0011 (21) Joerg, S.; Drago, R. S.; Adams, J. J. Chem. Soc., Perkin Trans. 2, 1997, 2431-2438.

5344

Analytical Chemistry, Vol. 73, No. 21, November 1, 2001

sample

Karl Fischer

KD-F0011

1 2 3 4 5 6 7

0.113 ( 0.0070 0.166 ( 0.0043 0.218 ( 0.0095 0.416 ( 0.0119 0.564 ( 0.0038 0.812 ( 0.0014 1.055 ( 0.0103

0.077 ( 0.0179 0.158 ( 0.0037 0.236 ( 0.0116 0.430 ( 0.0063 0.589 ( 0.0048 0.806 ( 0.0020 1.027 ( 0.0086

Mean value ( standard deviation (n ) 3).

in a batch system. Ethanol with the lower hydrogen bond acidity does not interfere when present in equal or 2-fold higher amounts relative to water. The stronger hydrogen bond donor methanol, however, significantly interferes with the determination of water at low concentration levels. Finally, since the Karl Fischer method is a widely recognized standard method for determination of water in solvents, a direct comparison between the newly proposed fluorometric method (batch analysis) and the coulometric Karl Fischer method was performed. As a preliminary application to real samples, THF samples with different contents of water, obtained from freshly distilled THF by standing in ambient air or by adding a small amount of water, were simultaneously measured by both procedures. The results of the two methods were compared by a linear regression analysis. The results of the reference method are represented by x and the results of the optode membrane by y, respectively. The interrelationship between the fluorometric method and the Karl Fischer method is approximated by the linear function

yfluorometric ) (0.993 ( 0.029)xKarl Fischer + (0.000 ( 0.017) (1) with r2 ) 0.996 for seven paired results. The hypothesis of linearity with a slope of unity and an intercept of zero is not contradicted. Therefore, the results from the two methods can be regarded as identical. Table 4 lists the analytical data for the determination of water in THF using the two different methods. Although some differences between the results are found, there is a good agreement between the results obtained with the newly proposed fluorometric method and the well-established standard Karl Fischer method. CONCLUSIONS A new method for the sensitive optical detection of low-level water concentrations in various organic solvents has been established. The sensing principle is based on novel fluorescent donoracceptor acridinyl fluorophores KD-F0011 and KD-F0021, which show increased radiationless relaxation from an intramolecular charge-transfer excited state upon hydrogen bond interaction with water. Two possible application patterns have been investigated: (1) the direct measurement in solution with the highly sensitive KD-F0021 and (2) a thin polymeric film incorporating covalently

immobilized KD-F0011 as an optical sensing film. The large Stokes shift for both methods allows the use of simple fluorescence spectrometric instrumentation for the reagent-free measurement of water in organic solvents. The advantages of a membrane sensor over the established Karl Fischer method include the following points: (a) real-time, continuous monitoring is possible, (b) faster analysis times are achieved, (c) the amount of sample required for the analysis is strongly reduced (order of 1 mL for the Karl Fischer technique compared to order of 100 µL for the membrane sensor), and (d) the weight determination of the individual sample is not required. The detection limits for the fluorometric method (low-ppm levels) performed with a laboratory-scale fluorescence spectrometer were found to be very close to the ones for the wellestablished Karl Fischer method. A preliminary application to real samples indicated a good agreement with the latter one. When more sophisticated instruments with higher sensitivity in emission detection are used, further improvements in the performance of this new analytical technique can be expected. Considering the simplicity of the fluorometric assay and its ability to be used for continuous monitoring, the sensor is useful to replace the Karl

Fischer method for specific applications, at least in the upper concentration range of the latter. ACKNOWLEDGMENT This work was supported by the Ministry of Education, Science and Culture of Japan and the Kanagawa Academy of Science and Technology (KAST), which is gratefully acknowledged. D.C. gratefully acknowledges a postdoctoral fellowship by the Japan Society for Promotion of Science (JSPS) and a research fellowship granted by the Science and Technology Agency (STA) of Japan. We also thank Dr. Shichi, Mr. Kanno, and Ms. Hashizume of Nissan ARC Ltd. for the moisture determinations using the Karl Fischer method. SUPPORTING INFORMATION AVAILABLE Detailed synthetic procedures, purification methods, and 1H NMR data for KD-F0011, KD-F0021, and KD-F0022. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review May 9, 2001. Accepted July 30, 2001. AC010535Q

Analytical Chemistry, Vol. 73, No. 21, November 1, 2001

5345