ARTICLE pubs.acs.org/Langmuir
NMR Chromatography Using Microemulsion Systems Chava Pemberton,‡ Roy E. Hoffman,† Abraham Aserin,‡ and Nissim Garti*,‡ † ‡
Institute of Chemistry, Edmond J. Safra Campus, Givat Ram, The Hebrew University of Jerusalem, Jerusalem 91904, Israel The Ratner Family Chair of Chemistry, Casali Institute of Applied Chemistry, Institute of Chemistry, Edmond J. Safra Campus, Givat Ram, The Hebrew University of Jerusalem, Jerusalem 91904, Israel
bS Supporting Information ABSTRACT: NMR spectroscopy is an excellent tool for structural analysis of pure compounds. However, for mixtures, it performs poorly because of overlapping signals. Diffusion ordered NMR spectroscopy (DOSY) can be used to separate the spectra of compounds with widely differing molecular weights, but the separation is usually insufficient. NMR “chromatographic” methods have been developed to increase the diffusion separation but these usually introduced solids into the NMR sample that reduce resolution. Using nanostructured dispersed media, such as microemulsions, eliminates the need for suspensions of solids and brings NMR chromatography into the mainstream of NMR analytical techniques. DOSY was used in this study to resolve spectra of mixtures with no increase in line-width as compared to regular solutions. Components of a mixture are differentially dissolved into the separate phases of the microemulsions. Several examples of previously reported microemulsions and those specifically developed for this purpose were used here. These include a fully dilutable microemulsion, a fluorinated microemulsion, and a fully deuterated microemulsion. Log(diffusion) difference enhancements of up to 1.7 orders of magnitude were observed for compounds that have similar diffusion rates in conventional solvents. Examples of commercial pharmaceutical drugs were also analyzed via this new technique, and the spectra of up to six components were resolved from one sample.
’ INTRODUCTION NMR spectroscopy is usually the tool of choice for precise structural characterization of organic and biomolecules and is best suited to the analysis of pure compounds. However, NMR is limited and becomes difficult when presented with complex mixtures. Most organic syntheses and industrial processes initially yield a mixture of compounds that require separation prior to structural characterization by NMR. If analysis of intermediate products is desired, then they too must be separated prior to NMR analysis. These chemical separations might be difficult, take a lot of time and resources to achieve, and in many cases are impossible because of chemical sensitivity and/or structural similarity. Hence, efforts were made to use hyphenated techniques where the mixture is separated by chromatography then analyzed by NMR.1 To a certain extent, 2D and multidimensional NMR can resolve simple mixtures but a more suitable method is to separate the components according to their diffusion coefficients.2 This is achieved with pulsed magnetic gradients using self-diffusion (SD) NMR techniques, also known as diffusion ordered spectroscopy (DOSY).35 The chemical shift of the spectrum is plotted on one (usually the horizontal) axis, while the diffusion rate is plotted on a perpendicular (usually vertical) axis. Each component yields a separate regular 1D spectrum corresponding to its diffusion constant. This application of DOSY has been r 2011 American Chemical Society
dubbed NMR chromatography,6 although the term pseudochromatographic NMR is technically more correct and the term NMR chromatography is usually reserved for systems where the log(diffusion) difference has been enhanced as described below. The problem with the DOSY method is that in most cases there is insufficient separation on the diffusion axis to fully separate the components of the mixture. No matter which processing method711 is used, it is impossible to separate overlapping signals whose diffusions differ by less than 3050%.12,13 The separation can be enhanced by adding a solid chromatographic medium such as silica gel to the sample. However, solid silica broadens the signals to hundreds or thousands of Hertz (using conventional NMR techniques) due to inhomogeneous magnetic susceptibility, making it impossible to acquire a DOSY spectrum. In this study, we replaced the silica suspensions reported in our previous study14 with microemulsions. In this way, the highly inhomogeneous silica boundary that was so central to the previous NMR chromatography method15 is replaced by the more magnetically homogeneous droplet surface. Received: January 19, 2011 Revised: February 16, 2011 Published: March 15, 2011 4497
dx.doi.org/10.1021/la200232b | Langmuir 2011, 27, 4497–4504
Langmuir The advantages of microemulsions over silica for NMR chromatography are as follows: (1) A regular NMR spectrometer for liquid-state is all that is needed without an HR-MAS1620 probe or MAS spinning capability. (2) There is no need for highsusceptibility solvents14,15 that are usually brominated or iodated and, therefore, very photosensitive. (3) The sample is measured without spinning so the diffusion rate is not affected by spinning.21 (4) Microemulsions can be stable for years22 in sealed ampules so there is no need to prepare the mixture immediately prior to use. Silica suspensions tend to sink out of the solution over a period of minutes to hours.14 Therefore, silica suspensions must be prepared immediately prior to use. (5) The effect of residual inhomogeneity of magnetic susceptibility is too small to detect in microemulsions, while for silica suspensions the line widths cannot be reduced beyond 10 Hz. The magnetic susceptibilities of microemulsion components are very similar and the droplet size very small (550 nm), leading to a high degree of magnetic homogeneity over a range of temperatures. The line widths arising from microemulsions are comparable with regular NMR spectra, and line broadening is undetectably small, not exceeding 0.3 Hz.23 In this study, we used U-type, fully dilutable microemulsions of water-in-oil (W/O), bicontinuous, and oil-in-water (O/W) droplets22,24 giving the full range of functionalities that regular chromatography provides using regular and reversed phase silica. The effect of nanostructured liquids on diffusion rates was first noted in microemulsions and used to determine droplet size or provide evidence for bicontinuous phases.25,26 Many subsequent studies of microemulsion structure have used diffusion in this way.24,2734 The diffusion rate of dissolved substrates has also been observed to vary with the structure of microemulsions35 and micellar solutions.36 For the purposes of 1H NMR, it is most desirable to have a solvent that is free of protons. For conventional solvents, this is usually achieved by deuteration. Perchlorination (such as CCl4) or perfluorination (such as C6F6) are options that were often used in the early days of NMR but have fallen out of favor because the presence of deuterium in the solvent is required for fieldfrequency lock. Perdeuterating of a microemulsion is expensive, although in some cases it may be affordable as demonstrated later in this paper. A convenient alternative is to perfluorinate some of the microemulsion components while perdeuterating water and isopropyl alcohol (IPA). When first published,23 this was the first liquid microemulsion formula developed specifically for the purpose of NMR chromatography. One partially deuterated and fluorinated microemulsion (a mixture of perfluorohexane, tetraethylammonium perfluorooctanesulfonate in dilute D2O solution37) was previously reported for an NMR study but not for the purpose of NMR chromatography. In this study, we demonstrate the use of several microemulsions in two classes: perfluorinated deuterated3845 and fully deuterated. These are used to separate several model molecules of different polarity and demonstrate the applicability of NMR chromatography to the analysis of common pharmaceuticals.
’ EXPERIMENTAL SECTION All NMR experiments were performed with a Bruker AVII 500 spectrometer equipped with GREAT 1/10 gradients and a 5 mm BBI probe with a z-gradient coil with a maximum gradient strength of 0.536 T m1. Diffusion was measured using an asymmetric bipolar LED46,47 experiment with an asymmetry factor of 20%, ramping the strongest
ARTICLE
gradient from 2% to 95% of maximum strength in 32 steps. Gradient pulses of 14 ms and intergradient delays between 0.07 and 1 s were used in order to achieve a decay curve that decayed most of the way but not completely to zero in order to optimize the accuracy of the diffusion measurement. The spectrum was processed by a Fourier transform in the acquisition (t2) dimension and by a LevenbergMarquardt48,11 fit to decaying Gaussians, supplied with the Bruker TOPSPIN software, in the gradient ramp evolution (g) dimension. NMR spectra were recorded at 298 ( 0.5 K. Samples were prepared by weighing the components using an analytical balance. The viscosity of the solution is similar to that of water and sufficient to prevent convection effects from disturbing diffusion measurements. Water was TDW. Chemicals were obtained from the following sources: perfluoroheptyl-1-bromide (PFHepBr) and perfluorohexane (Alfa Aesar, Ward Hill, MA); ammonium hydroxide, n-butanol, ethanol, and potassium hydroxide (Frutarom, Haifa, Israel); lithium hydroxide, sodium dodecyl sulfate (SDS), and sodium hydroxide (J. T. Baker, Deventer, The Netherlands); isopropyl alcohol (IPA), n-pentanol, and propylene glycol (PG) (Merck, Darmstadt, Germany); 1,3-butanediol, cyclohexane, 4,4-dimethyl-4-silapentane-1-sulfonic acid sodium salt (DSS), hexamethyldisiloxane, R(þ)-limonene, perfluorododecanoic acid (PFDD), perfluorotetradecanoic acid, and Tween 60 (SigmaAldrich, St. Louis, MO); and butanol-d10, cyclohexane-d12, D2O, IPAd8, and SDS-d25 (Cambridge Isotope Laboratories Inc., Andover, MA). Pharmaceuticals were commercial samples: Dexamol (Dexon, Hadera, Israel), Advil (Pfizer, New York, NY), and Abitren (Teva Pharmaceuticals Ltd., Petach Tikva, Israel). The perfluorocarboxyl salts were prepared by first dissolving the acid in IPA and then adding a molar equivalent of the appropriate metal hydroxide dissolved in water. The solution was mixed and freeze-dried by lyophilization.
’ RESULTS AND DISCUSSION In order to demonstrate the advantages of using microemulsions as the chromatographic medium and to validate the techniques, we selected a few possible alternative approaches. A number of systems were studied: U-type microemulsions while observing the 19F diffusion spectrum; perfluoroemulsions including D2O and D2O/IPA-d8 dilution lines using perfluorohexane or PFHepBr and the lithium salt of perfluorododecanoate (Li[PFDD]); the nonfluorinated, deuterated D2O, SDS-d25, nbutanol-d10, cyclohexane-d12 system; and their application to the analysis of pharmaceutical drugs. U-Type Microemulsions and Perfluoro Compounds. The microemulsion used in the first set of experiments was a U-type microemulsion (as described below).49 U-type refers to microemulsions that can be diluted by the aqueous phase from the surfactant/oil-rich edge of the ternary (or pseudoternary) phase diagram to the aqueous-rich vertex. The system that we tested was previously developed by Garti et al.50 The system consists of five components: R(þ)-limonene and ethanol at a weight ratio of 1:1 constituting the oil phase, water and PG at weight ratio of 1:1 constituting the aqueous phase, and Tween 60 (ethoxylated sorbitan monostearate) as the surfactant. Measurements were made along the 7:3 wt ratio surfactant/oil dilution line in steps of 10 wt % from 0 to ca. 100 wt % of the aqueous phase. Upon dilution, the microemulsion transforms from W/O to bicontinuous to O/W. The exact composition of the Tween 60 varies from batch to batch and affects the precise dilutions at which the transitions occur. Different techniques also yield slight variations in the measured transition dilution. The results previously reported using a variety of methods for the 7:3 4498
dx.doi.org/10.1021/la200232b |Langmuir 2011, 27, 4497–4504
Langmuir
Figure 1. Diffusion constants against water content (in wt %) of three selected perfluorinated compounds along the 7:3 dilution line of the U-type microemulsion composed of R(þ)-limonene/ethanol (1:1), water/PG (1:1), and Tween 60.
dilution line of this U-type microemulsion51 are consistent with the bicontinuous region being between approximately 35 to 65 wt %, in line with the results of this work (Figure 1 and Supporting Information Table 1). Any component that is trapped in the inner phase or at the interface is expected to have slow diffusion coefficients. On the other hand, a component that is located in the continuous phase is free to move long distances and can diffuse relatively quickly. The components located at the interface of bicontinuous microemulsions are expected to diffuse the slowest and should give the greatest log(diffusion) difference in an NMR chromatography experiment. The log(diffusion) difference was determined for fluorinated compounds that were solubilized in the microemulsion using 19F SD-NMR. This overcame the problem of protons in the microemulsion swamping the signals of the substrates. Three model compounds were chosen, NaF which is an inorganic and very polar salt, perfluorohexane which is a nonpolar compound and very hydrophobic, and 2,4-dinitrofluorobenzene which has intermediate polarity. In the W/O region (up to ca. 30 wt % water), NaF diffuses slowly (on the order of 1011 m2 s1), indicating that it is located, as expected, inside the aqueous droplets. It is therefore assumed that the log(diffusion) difference enhancement will be large. However, the NMR signal is broad and relaxes quickly, making diffusion measurement difficult. A possible explanation for the rapid relaxation is that the fluoride ions are dehydrating the head groups of the surfactant and remain bound to them.52 At low water concentrations, the diffusion of NaF could not be measured because the signal relaxed completely during the diffusion time. At higher dilutions, the microemulsion converts into bicontinuous mesophases with increased diffusivity and finally inverts into O/W droplets for which the aqueous phase becomes the continuous phase, and since the NaF is dissolved in the water phase, the fluoride ions show a progressive increase in their diffusion rate. The lipophilic perfluorohexane diffuses rapidly in the W/O region indicating that it is located mostly in the continuous phase. In the bicontinuous region (ca. 30 to ca. 60 wt % water), perfluorohexane diffuses very slowly (on the order of 1012 m2 s1)
ARTICLE
at approximately the same rate as the surfactant,51 indicating clearly that it is trapped at the interface. The minimum diffusivity is reached at ca. 58 wt %, and thereafter it increases again as the system inverts to O/W. 2,4-Dinitrofluorobenzene has intermediate polarity and diffuses rather fast on the order of 1010 m2 s1, but without any significant change along all the dilution line, indicating that they have similar solubility in water and oil and are not attached to the interface. Knowing the character of each of these compounds, we could confirm that the diffusion coefficients accurately reflect the position of each compound in the microemulsion at any given dilution stage. The NaF always remains in the aqueous phase. The perfluorohexane prefers the oil phase but can partition between the oil phase and the interface depending on the nature of the interface. 2,4-Dinitrofluorobenzene has little preference between the oil and aqueous phase and presumably largely remains in the continuous phase along the dilution line, causing its diffusion rate to change little throughout. As an application of NMR chromatography, we have found that at different dilution stages (along the dilution line) the separation of compounds of various characteristics is feasible. Especially interesting is that the diffusion rate of perfluorohexane changes by 2 orders of magnitude in the bicontinuous phase compared to the W/O phase clearly, indicating that the best separation may take place in microemulsions with a bicontinuous structure. Ideally, the separation in diffusion rates (the log(diffusion) difference enhancement) should be as large as possible. Droplet sizes in O/W and W/O microemulsions dictate the log(diffusion) difference enhancement for selectively trapped entities. The larger the droplets, the slower the diffusion within them and at their interface,53 and the better the log(diffusion) difference enhancement. However, emulsions with large droplets are inherently unstable. Therefore, a compromise has to be made between log(diffusion) difference enhancement and the carrier’s (the emulsion’s) physical stability. Note that it is possible to use common microemulsions using 1 H SD-NMR; however, most substrate signals are swamped by the signals arising from the emulsion. Proton-Free Emulsions: Perfluorinated Emulsions. In order to make more general use of 1H SD-NMR, it was necessary to use a microemulsion that was substantially free of protons. As a result, fluorinated and/or deuterated surfactants were used to replace the Tween 60 system. Microemulsions derived from fluorinated compounds (as the oil phase), fluoro-amphiphiles (as surfactants/emulsifiers), isopropyl alcohol-d8 (as a cosurfactant), and D2O as the aqueous phase were constructed. One of the major requirements of the surfactant forming microemulsion is its full miscibility with water and with oil at all weight ratios. This requirement is achieved with nonfluorinated components by using Tween 60 as a surfactant and a mixture of solvent and alcohol as the oil phase. The ethanol served as cosurfactant/cosolvent as a prerequisite to facilitate the miscibility of the surfactant in water and in oil for U-type microemulsions.51 Fully fluorinated surfactants are not completely miscible with water and solvent (the oil phase), and the fluorinated solvents are also totally insoluble in water. The addition of various alcohols did not help to provide the required miscibility, and therefore, fluorinated oils such as hexafluorobenzene and cosurfactants such as methanol or ethanol did not produce microemulsions. On the other hand, the mixture of perfluorohexane and isopropyl alcohol (IPA) is a good 4499
dx.doi.org/10.1021/la200232b |Langmuir 2011, 27, 4497–4504
Langmuir
ARTICLE
Table 1. Effect of Counter Ion on Maximum Water Dilution Ranges at Which and the Samples Are Transparenta counter ion in PFDD
cloudy region
maximum difference in the
(wt % of water)
log(diffusion) domain
H Li
>70 >86
not measured 1.45
Na