Spectral Editing of Organic Mixtures into Pure Components Using

Anal. Chem. 2008, 80, 186-194

Spectral Editing of Organic Mixtures into Pure Components Using NMR Spectroscopy and Ultraviscous Solvents Andre´ J. Simpson,* Gwen Woods, and Omid Mehrzad

Department of Chemistry, University of Toronto, 1265 Military Trail, Scarborough, ON M1C 1A4, Canada

A general technique is described that permits the extraction of a complete 1H NMR spectrum for components in organosoluble mixtures. The approach should find a wide range of applications considering that pure component spectra can be generated without the need for physical separation. This technique is especially significant for synthetic organic chemistry and the pharmaceutical industry due to the potential to isolate a product spectrum even in the presence of overlapping starting materials, byproducts, or degradation products. A viscous oil-based solvent system that can be temperature-manipulated from essentially a solid at one extreme to a freely flowing liquid at the other is employed. The system contains no protons and is miscible with common organic solvents. Through careful control of the temperature and thus solvent viscosity, the behavior of small molecules moves from the positive to the extreme of the negative NOE regime. Under such conditions, all protons in a molecule correlate with all other protons as propagation by spin diffusion becomes highly efficient, behavior normally only observed with rigid macromolecules in conventional solvents. Therefore, as long as one proton (or carbon signal in hybrid experiments) is resolved for a component in a mixture, the entire proton spectrum for that molecule can be cleanly extracted from a 2D NOESY spectrum (or from selective 1D NOE-based analogues). Preliminary results are highly encouraging, indicating that the approach may be feasible for a wide range of molecules and mixtures; however, in practice the exact types of structures, combinations of structures, and range of concentrations that can be cleanly extracted will become evident as the technique becomes better established. Whether a synthetic organic chemist faced with a mixture of starting materials and products or a pharmaceutical researcher confronted with the complexities of drug decomposition, unraveling mixtures is a reality of modern chemistry. Quantification and identification of components is often the common goal. Obtaining NMR spectra, whether essential for identification or required by protocol, frequently requires physical separation that can pose a variety of drawbacks. For example, chromatography can be impossible if products are highly reactive, undesirable if key * To whom correspondence should be addressed. Tel: 1-416-287-7547. Fax: 1-416-287-7279. E-mail: [email protected].

186 Analytical Chemistry, Vol. 80, No. 1, January 1, 2008

associations are disrupted by separation, or simply too timeconsuming or costly. Through careful interpretation of conventional NMR approaches such as HMBC and TOCSY, it is possible to differentiate the signals from components in relatively simple mixtures.1 Furthermore diffusion ordered spectroscopy (DOSY) is a powerful tool for the separation of components based on diffusivities.2 In DOSY, however, similar diffusion coefficients or significant spectral overlap can hinder the clean separation of all components in a mixture. Here we demonstrate an alternative approach that uses NMR spectroscopy to subedit a mixture into its individual components without the need for physical separation and should be highly complimentary to existing methods.1,2 The technique uses an ultraviscous NMR solvent to greatly reduce the molecular tumbling of small molecules such that they are more akin to tumbling rates of large proteins. The nuclear Overhauser effect (NOE) becomes large and of negative sign.3 Furthermore, dipolar interactions, which are averaged by isotropic tumbling in nonviscous solvents, become considerable. This effect provides the framework for the efficient propagation of spin diffusion resulting in spin population disturbance throughout the entire molecule. Thus, if the viscosity of the NMR solvent can be “controlled”, conditions are established that allow all protons in a molecule to correlate with all other protons in the same molecule. Under these conditions, simple “NOE”-based experiments can be employed to cleanly extract the spectrum of a component as long as one proton or carbon signal is resolved. Alteration of solvent composition to manipulate the NOE signal from small molecules has been shown previously.4-10 Most common applications aim to change the solution conditions such that the NOE, which for some medium-sized molecules can be close to zero, can be observed and used for structural conformation.11 This study takes the concept further and utilizes the combination of an inert oil, reduced temperatures, and a long (1) Lin, M. G.; Shapiro, M. J. Anal. Chem. 1997, 69, 4731-4733. (2) Morris, K. F.; Stilbs, P.; Johnson, C. S. Anal. Chem. 1994, 66, 211-215. (3) Bothnerby, A. A.; Johner, P. E. Biophys. J. 1978, 21, A2. (4) Luck, L. A.; Landis, C. R. Organometallics 1992, 11, 1003-1005. (5) Adamy, S. T.; Kerrick, S. T.; Jonas, J. Z Phys. Chem. 1994, 184, 185-203. (6) Fesik, S. W.; Olejniczak, E. T. Magn. Reson. Chem. 1987, 25, 1046-1048. (7) Williamson, M. P.; Williams, D. H. J. Chem. Soc., Chem. Commun. 1981, 165-166. (8) Bothnerby, A. A.; Johner, P. E. Biophys. J. 1978, 24, 779-790. (9) Williamson, M. P.; Neuhaus, D. J. Magn. Reson. 1987, 72, 369-375. (10) Kartha, G.; Bhandary, K. K.; Kopple, K. D.; Go, A.; Zhu, P. P. J. Am. Chem. Soc. 1984, 106, 3844-3850. (11) Neuhaus, D.; Williamson, M. P. The nuclear Overhauser effect in structural and conformational analysis, 2nd ed.; Wiley-VCH: New York, 2000. 10.1021/ac702119d CCC: $40.75

© 2008 American Chemical Society Published on Web 12/04/2007

experimental mixing time to force small molecules to the extreme of the negative NOE regime. Under these conditions, it will be demonstrated that complete spectral profiles can be extracted from a single resolved resonance using selective 1D or 2D NMR spectroscopy. The protocols described should find a wide range of applications as the pure spectrum of a component can be extracted from a mixture without physical separation. One key application may be in synthetic organic chemistry as it should be possible to isolate a pure spectrum for a product even in the presence of overlapping starting materials, byproducts, or both. Furthermore, this report demonstrates that correlations over distances of >20 bonds can be observed even in extremely flexible, long chains. The transfer of spin diffusion through flexible chains is not normally efficient due to the local dynamics in the chain; thus, long chains represent the most challenging structural entities for study by the techniques outlined here. By altering the mixing time or the viscosity (determined by the operating temperature of the oil system described here) “ultra-long-range” spatial interactions can be recorded. Such information will prove complementary to that information gained from existing approaches and should provide information as to the structure, conformation, rigidity, and dynamics of molecules that may be difficult or impossible to gather using conventional NMR techniques. MATERIALS AND METHODS The 1000N Halocarbon oil was kindly donated by the Halocarbon Corp. (www.halocarbon.com). A 80:20 ratio indicates that 80% oil was mixed with 20% solvent (chloroform-d, unless stated otherwise). NMR solvents were of 99.9% purity and purchased from Cambridge Isotope Laboratories Inc. (Andover, MA). Samples were prepared on a weight basis using 1.98 g/mL as the density of the oil at room temperature. The 10-mg aliquots of each component were used in all studies, with the exception of the “hybrid” HSQC-NOESY experiments in which 50 mg was used. Solution-State NMR. Experimental Details. NMR data were acquired on a Bruker Avance 500-MHz spectrometer using a 1HBB-13C TBI probe fitted with an actively shielded Z gradient. Temperature control was performed using a Bruker variabletemperature (BVT-2000) unit in combination with a Bruker cooling unit (BCU-05) to provide chilled air. The 1-D solution-state 1H NMR experiments were performed with 32 scans, a recycle delay of 3 s, and 16 348 time domain points. Spectra were apodized through multiplication with an exponential decay corresponding to 1 Hz line broadening in the transformed spectrum and a zero filling factor of 2. Two-dimensional nuclear Overhauser effect spectroscopy (NOESY) spectra were acquired in phase-sensitive mode, using time proportional phase incrimination. Eight scans and 2048 data points were collected for each of the 256 increments in the F1 dimension. Studies of NOE with changing solvent composition and temperature were carried out between neighboring protons, and a 250-ms mixing time was employed. For all other experiments, a mixing time of 1 s was employed to promote spin diffusion. Both dimensions were processed using sine-squared functions with a π/2 phase shift and a zero-filling factor of 2. The 1D selective ‘steady-state” NOESY experiments were carried out using a band optimized Gaussian pulse tailored to selectively excite the region of interest. Selective NOESY was performed with preirradiation of 2 s, relaxation delay of 3 s, and 16 dummy scans. Experiments were recorded “on” and “off”

resonance, alternating per scan for a total of 32 transients. For off-resonance, irradiation was carried out at 100 ppm (no sample signals) and on-resonance (peak of interest selected) difference spectra were created by appropriate phase cycling of the receiver. Spectra were apodized by multiplication with an exponential decay corresponding to 1-Hz line broadening in the transformed spectrum and a zero filling factor of 2. Heteronuclear single quantum coherence-nuclear Overhauser effect spectroscopy (HSQCNOESY) spectra were collected in phase-sensitive mode using echo/antiecho gradient selection. A total of 32 scans were collected for each of the 256 increments in the F1 dimension. 2048K data points were collected in F2, using a 1J 1H-13C (145 Hz), a 600-ms mixing time, and a relaxation delay of 2 s. Both dimensions were processed using sine-squared functions with a π/2 phase shift and a zero-filling factor of 2. RESULTS AND DISCUSSION The NOESY spectrum for a mixture of brucine, strychnine, and cholesteryl acetate dissolved in chloroform is shown in Figure 1A. As expected for small molecules, all NOESY cross-peaks are positive (indicated by a sign antiphase to that of the spectrum diagonal) and correlations are only observable between protons in proximity to each other. This result is expected as the molecules in the nonviscous solvent fall in the positive NOE regime where spin diffusion is extremely inefficient.11 In contrast, Figure 1B shows the same mixture dissolved in an 80:20 mixture of oil/ chloroform. It is immediately apparent that all NOE cross-peaks have become negative (the same phase as the spectral diagonal) and more intense. The spectrum is characterized by numerous additional cross-peaks running both vertically and horizontally. These additional correlations result from long-range interactions propagated by spin diffusion such that all protons in a molecule correlate with all other protons in the molecule. The intensity of the NOE cross-peaks and efficiency of spin diffusion in the viscous solvent arise from very slow molecular tumbling and thus long correlation times, transitioning this system into the extreme of negative NOE regime. (See later in this article, which demonstrates the theoretical maximum of -100% NOE is easily achieved for small molecules in this viscous oil-based system.) Therefore, if a row or column is selected at a resolved proton, a 1D spectrum is extracted that will be a pure subspectrum (albeit not quantitative) for that specific molecule. It is essential to note that throughout this paper slices from the 2D spectra are chosen for protons or carbons positioned at the extremity of a given molecule. The reasoning for these positions is to verify that, with the longest distances across a molecule, all correlations are observable. If a central position is selected in a molecule, correlations to all other protons are easier to observe as these correlations are over shorter distances. The importance of observing extremity correlations becomes evident when examining a mixture of novel products or components where only one proton can be resolved. This resolved position may be at the extreme of the molecule, giving no choice as to the row/ column to be extracted. Thus, it is necessary to demonstrate that the observed correlations encompass the whole and not just a section of the molecule. In the case of the cholesteryl acetate, neither the terminal acetate (position 31) or the methyls (positions 24 and 30) at either end of the molecule are resolved in the mixture; thus, position 16 has been chosen as the site that is Analytical Chemistry, Vol. 80, No. 1, January 1, 2008

187

Figure 1. Nuclear Overhauser spectra of brucine (B), strychnine (S), and cholesteryl acetate (C). Both spectra are obtained at 288 K with a mixing time of 1 s. (A) is dissolved in chloroform-d. (B) is dissolved in 80:20 oil/chloroform-d. Numbers correlate to the protons for which spectral slices were extracted. Structures and slices are shown in Figure 2.

188 Analytical Chemistry, Vol. 80, No. 1, January 1, 2008

closest to the molecular extremity while still being completely resolved. Later in this paper, correlations over even greater distances are demonstrated. Figure 2A illustrates the considerable overlap in the 1D spectrum of the mixture. Figure 2C corresponds to a row extracted from the NOESY spectrum for proton 16 in cholesteryl acetate. The numbered molecule is shown at the top right of Figure 2, and the corresponding horizontal slice of the proton at position 16 (“C-16”) is shown on the F1 dimension of the NOESY spectrum in Figure 1B. It is clear by comparison of the control 1H NMR spectrum of cholesteryl acetate (Figure 2B) that all of the peaks from cholesteryl acetate have been cleanly extracted from the mixture. However, the correlation from the C-16 proton to the terminal CH3 protons at the opposite end of the molecule (positions 24 and 30) are relatively weak (correlations indicated by red arrows in Figure 2C and D). By reducing the acquisition temperature to 278 K, these long-range correlations become much stronger (Figure 2D), albeit accompanied by a reduction in line shape. These long-range correlations are observed over a distance of ∼18-19 Å or a distance of ∼15 bonds. This distance is quite staggering considering that, in conventional NOESY experiments performed in nonviscous solvents, distances of >5 Å are rarely observed.12 Pure spectra of both brucine (Figure 2E and F) and strychnine (Figure 2G and H) can also be extracted from the mixture. In the case of brucine, proton 21 is chosen, and for strychnine, proton 22; these positions are indicated in Figure 1 as “B21” and “S22”. The methoxyl groups in brucine overlap with other components in the spectrum and so cannot be selected. In both cases, a complete spectrum can be extracted that contains all peaks expected for the molecule. This clearly demonstrates that conditions can be obtained such that all protons in each molecule interact with all other protons in the molecule, which in turn permits spectral editing. The approach of spectral editing described here is based upon extraction of rows from 2D NOESY spectra; thus, there will be a reduction in the observed resolution, when compared to a conventional 1D spectrum, from both the influence of the oil and lower inherent resolution of the 2D data sets. In this paper, only rows have been extracted given the greater digital resolution in the F2 dimension (only 256 points were collected in F1). Figure 3 shows an expansion of the aromatic region for the mixture described above. Position 24 in brucine has been selected as an example. The full width at half-height (fwhh) is increased from ∼2.5 Hz in the 1D spectrum to 3.16 Hz in the projection from the NOESY (Figure 3A and B, respectively) as a result of the low digital resolution of the NOESY spectrum. In an 80:20 oil/ chloroform mixture at 288 K, the fwhh increases to 6.42 and 6.54 Hz in the 1D and NOESY projections (Figure 3C and D, respectively). Due to the broader line shape with the oil present, the lower resolution in the NOESY projection has little effect on the spectral quality. The presence of the oil however reduces the spectral resolution by ∼2-3 times. This is the tradeoff for creating conditions where spin diffusion is prolific and spectral editing of whole molecules is possible. Readers should consider that the size and rigidity of structures studied, as well as temperature and oil concentration, will ultimately influence the broadening ob(12) Van, Q. N.; Smith, E. M.; Shaka, A. J. J. Magn. Reson. 1999, 141, 191-194.

served in oil and should remember the above serves as a mere example only. Finding the correct temperature for spectral editing can be potentially challenging. Figure 2 (top left) depicts the percent NOE enhancement observed for the three different molecules in 70:30 and 80:20 oil/chloroform with varying temperature. The green box highlights the 288 K temperature at which the spectral editing, described above, was performed. The percent NOE enhancements were measured in the intact mixture and between the protons colored red on the individual structures. These protons were chosen as they are easily resolved at all temperatures allowing integration of cross-peaks. The NOE correlation with temperature varies for both individual positions within a molecule and between different molecules. The question thus arises, “Is the technique applicable to the vast majority of (or potentially all) organosoluble molecules?” Extremely Challenging Molecules. Small molecules with fast molecular tumbling may not be efficiently suppressed by a viscous solvent. Molecules with long, flexible chains are not efficient at propagating spin diffusion due to local motion within the chain and could also prove problematic. To investigate these issues further, a second mixture containing phenanthracene and hexadecanophenone was used. Phenanthracene, a small and spherical molecule, undergoes fast molecular tumbling, while hexadecaphonone contains a long and flexible chain (see Figure 4 for chemical structures). As molecules become larger and increasingly rigid, the propagation of spin diffusion becomes more efficient. Thus, hexadecanophenone, which is >25 Å “end to end”, of low molecular weight, and extremely flexible represents one of the most challenging molecules for study by this approach. Essentially, if it is possible to subedit a spectrum of hexadecanophenone from a mixture, it should be possible to subedit the vast majority of (or possibly all) organosoluble molecules by this approach. The behavior of both phenanthracene and hexadecanophenone in the various oil compositions with temperature is shown in Figure 4. With phenanthracene, a complete spectrum can be extracted from both 70:30 and 80:20 oil mixtures at low temperature (data not shown). Hexadecanophenone, however, proves more challenging and less than -10% NOE enhancements are seen in both 70:30 and 80:20 compositions for protons 9 and 10 (see structure). For this molecule, a solvent system of 90:10 is required. In 90:10 with chloroform-d as the solvent, it is necessary to use an additional external locksa thin capillary insert down the center of the NMR tube. Details of this and other considerations are outlined in a “practical guide” which can be downloaded as Supporting Information. A shifting of ∼0.57 ppm is seen for all resonances in the 90:10 oil as depicted by the inset in Figure 4D. Interestingly, the tetramethylsilane in the oil is severely broadened, indicating that even this small, completely spherical molecule has a tumbling rate considerably restricted in the 90:10 system. Figure 4A shows the total mixture while (B) and (C) demonstrate the phenanthracene spectrum and the extraction of the 1D profile of phenanthracene from the NOESY spectrum. Panels D and E in Figure 4 show that all resonances from hexadecanophenone can be cleanly extracted in the 90:10 mix even when the CH3 group is selected as the position for extraction. This result should not be underestimated as it definitively Analytical Chemistry, Vol. 80, No. 1, January 1, 2008

189

Figure 2. Top left: plots of the relative % NOE observed with varying oil/chloroform ratios and temperatures for a mixture of brucine, strychnine, and cholesteryl acetate. Top right: chemical structures; the red positions indicate the protons between which the NOEs were measured. These protons were chosen as they are easily resolved at all temperatures. The green positions indicate the positions at which slices were taken from the NOESY spectra (also see Figure 1B); these positions were chosen as they present the positions that are closest to the molecular extremity while still being fully resolved in the 1H NMR spectrum. (A)-(H) are 1H NMR for the mixtures and components in 80:20 oil/chloroform. Unless otherwise stated, all spectra were acquired at 288 K (see green box on plot, top left). (A) mixture, (B) cholesteryl acetate, (C) cholesteryl acetate extracted from the NOESY spectrum, (D) same as (C) but at 10 K lower (278 K), (E) brucine, (F) brucine extracted from the NOESY spectrum, (G) strychnine, and (H) strychnine extracted from the NOESY spectrum.

demonstrates that long-range interactions over distances of >25 Å can be mediated by spin diffusion even in the most challenging 190 Analytical Chemistry, Vol. 80, No. 1, January 1, 2008

molecules (i.e., long, unrestricted, flexible chains). If spin diffusion can mediate a correlation over such a distance in a flexible

Figure 3. 1H NMR spectra for brucine, strychnine, and cholesteryl acetate highlighting digital resolution (DR) and fwhh. (A) Conventional 1H spectrum at 298 K in 100% chloroform-d. (B) Positive projection from the NOESY spectrum collected under the same conditions. (C) Conventional 1H spectrum at 288 K in 80:20 oil/chloroform-d. (D) Positive projection from the NOESY spectrum collected under the same conditions. Arrows highlight position 24 in brucine (see Figure 2 for structure).

molecule, the approach should be applicable to the vast majority of (or potentially all) organosoluble molecules. One question does, however, arise: “In a mixture of unknown composition what ratio of oil/solvent should be used?” Unfortunately, no definitive answer to this question exists. The simplest approach is to dissolve the mixture in an 80:20 mixture and cool the sample. If at low temperature there are still signals of interest that have not significantly broadened, a mixture of 90:10 will be required. In the authors’ limited experience (with ∼40 molecules), 80:20 is suitable for the majority of molecules and 90:10 is only required for challenging cases such as hexadecanophenone. Even dimethylaminobenzaldehyde, a single ring aromatic structure, can be extracted from a NOESY in 80:20 (data not shown). It is significant to add that, with hexadecanophenone in 80:20, correlations from “head to tail” are still possible when the steadystate NMR approach is employed. In the steady-state NOE method, a signal is preirradiated for a set period of time and the NOE enhancement is measured by difference. This method is more sensitive than transient NOE methods11 but does rely on subtraction to create the final spectrum. Considerable artifacts from incomplete subtraction can be problematic. Figure 4F illustrates the result of the steady-state experiment, where the CH3 has been selectively irradiated. The artifacts from incomplete

subtraction are unfortunately considerable, but the result does confirm that ultra-long-range “head to tail” correlations in hexadecanophenone are detectable even in an 80:20 mixture. Any NOEbased detection scheme including the 1D steady state, The 1D selective pulsed field gradient spin echo (PFGSE) and 2D NOESY techniques can be employed for the subediting of mixtures. While the steady-state method has largely been replaced by the PFGSE approach, the steady-state method does allow the user to “force” spin diffusion via a long preirradiation period and thus may still prove useful for spectral editing when used in combination with modern gradient-assisted transient approaches. Hybrid 1H-13C Techniques. Thus far, we have demonstrated that as long as one proton signal is resolved from a molecule, the complete spectrum can be extracted from a mixture utilizing spin diffusion. However, in situations of spectral overlap, a compound of interest may not have a 1H signal that is completely resolved. In such cases, utilizing the larger chemical shift dispersion of 13C may prove useful. By employing simple techniques such as 2D HSQC-NOESY, a complete proton spectrum can be obtained from a molecule with only a single carbon resonance resolved. Figure 5 demonstrates this concept. Figure 5A shows a 2D HQSC-NOESY correlation for cholesteryl acetate in 80:20 oil/chloroform at 288 K. Under these conditions, all protons correlate with all other protons in the molecule, and thus in the 2D HSQC-NOESY, all protons also correlate will all other protonated carbons in the sample. A horizontal slice through any carbon chemical shift will therefore produce a complete 1H spectrum of cholesteryl acetate (blue arrow, Figure 5A). Figure 2 shows the proton spectrum of cholesteryl acetate (B) and compares it to the proton spectrum extracted from the carbon at position 16 (see Figure 2C for structure). These results demonstrate that, as long as a carbon resonance is resolved, a complete 1H proton spectrum for the molecule can be recovered. Interestingly, a vertical slice can also be taken from the 2D HSQC-NOESY (red arrow, Figure 5A). In this case, a spectrum is created that contains all the protonated carbons in the molecule. Figure 5D compares the protonated carbon spectrum projected from an HSQC for cholesteryl acetate to the carbon spectrum obtained by extracting the vertical plane for proton 16 (Figure 5E). The ability to extract all the protonated carbon chemical shifts from an individual component in a mixture may prove a very useful tool in the assignment of 13C resonances for molecules within mixtures. CONCLUSIONS We have demonstrated that, through the use of a viscous solvent system, careful temperature control, and a long mixing time, subediting the pure spectrum of a component from a mixture is possible. With sufficient material such that a hybrid 2D HSQCNOESY can be obtained, this isolation can be carried out if either a single carbon or a proton signal is resolved. With lesser material, a resolved proton resonance is required. In general, a solvent system of 80:20 oil/solvent should work for the majority of molecules. In cases of long flexible chains (alkanes, fatty acids, etc.) or extremely small molecules, a 90:10 oil/solvent may be required. The mixture of interest should first be dissolved in 80: 20 with a reduced temperature of ∼258-263 K. If sharp peaks remain in the 1D NMR spectrum, a 90:10 mixture may be required for these components. A ratio of 70:30 (or less) is only recommended if the mixture has limited solubility or the line shape at Analytical Chemistry, Vol. 80, No. 1, January 1, 2008

191

Figure 4. Top left: plots of the relative % NOE observed with varying oil:chloroform ratios and temperatures for a mixture of phenanthracene and hexadecanophenone. Top right: chemical structures; the red and green positions have the same meaning as in Figure 2. Unless otherwise stated, all spectra were acquired at 278 K (see green box on plot, top left) and in 90:10 oil/chloroform. (A) Mixture, (B) phenanthracene, (C) phenanthracene extracted from the NOESY spectrum, (D) hexadecanophenone, (E) hexadecanophenone extracted from the NOESY, and (F) hexadecanophenone in 80:20 oil/chloroform at 268 K acquired via the steady-state NOE method. 1, chloroform in external lock capillary; 2, chloroform in oil; 3, TMS in internal lock capillary; 4, artifacts from incomplete subtraction.

298 K in 80:20 is too broad as may be the case for large or very rigid molecules. If the mixture contains a range of molecules with varying sizes and chain length, the mixture may need to be examined in a range of oil/solvent combinations and at different 192

Analytical Chemistry, Vol. 80, No. 1, January 1, 2008

temperatures. Most modern spectrometers can acquire experiments at different temperatures under complete automation. While, under the correct experimental conditions, it should be possible to extract a component from a single slice of a NOESY

Figure 5. (A) 2D HQSC-NOESY spectra of cholesteryl acetate, (B) 1H NMR spectra of cholesteryl acetate, (C) 1H slice extracted from the 2D HSQC-NOESY by slicing at carbon 16 (indicated by the blue arrow; see Figure 2 for structure), (D) protonated carbons in cholesteryl acetate (from HSQC projection), and (E) protonated carbons extracted from the 2D HSQC-NOESY by slicing at proton 16 (indicated by the red arrow). All spectra were acquired for cholesteryl acetate dissolved in 80:20 oil/chloroform at 288 K.

spectrum, inspection of a number of rows is advisable. Regardless of overlap in subsequent slices with other compounds, these slices

should contain all the peak information from the molecule of interest. By comparing a number of rows, it should be possible Analytical Chemistry, Vol. 80, No. 1, January 1, 2008

193

to unambiguously confirm all the resonances from the component of interest. However, if a mixture contains large and rigid components, these may become extremely broad at high oil concentrations and complicate analysis. In such cases, relaxationbased filters may be useful to remove “background” signals. It is important to point out that, in mixtures containing a wide range of concentrations, signals from trace components may become masked by those from more concentrated components after the addition of oil. In practice, the loss of information from “trace” components will depend on many factors including the exact broadening of the components in the oil (partially determined by the rigidity of the individual structures), the dynamic range of the concentrations, and spectral overlap, as-well as resolution of the 2D data set. As such, the extent to which types of structures, combinations of structures, and range of concentrations can be cleanly extracted will only become evident as the technique becomes better established. Ultimately the biggest drawback of the approach described here is mixture solubility. As the mixture has to be dissolved in an 80:20 oil mixture (assuming a 1-mL volume, 5-mm standard tube) only 200 (80:20) or 100 µL (90:10) is available as solvent. The oil is miscible with most common organic solvents including, ethanol, chloroform, acetone, and benzene. The oil is, however, completely immiscible with water and DMSO. It is important to note that mixtures containing benzene freeze between 278 and 268 K, ruling out lower temperatures in this solvent. The main hardware requirement for this type of study is the ability to run NMR spectra at temperatures down to 253 K. While many modern NMR spectrometers are now delivered with such

194

Analytical Chemistry, Vol. 80, No. 1, January 1, 2008

an accessory and most large chemistry departments should have this capability, low-temperature control may not be available in all NMR facilities. Air cooling units for accurate temperature control are available from the major spectrometer manufacturers and third party vendors. The work here demonstrates that ultralong correlations, mediated by spin diffusion, can be observed in viscous oils. These correlations in turn allow extraction of pure component spectra from mixtures. The observation in flexible chains of a correlation over 25 Å is highly encouraging and suggests that with carefully chosen experimental conditions, the system should work for the vast majority of compounds soluble in organic solvents. ACKNOWLEDGMENT We thank the Natural Science and Engineering Research Council (NSERC), the Canadian Foundation for Climate and Atmospheric Sciences (CFCAS) (GR-520), the International Polar Year (IPY), and the Ontario government for an Early Researcher Award (A.J.S) for funding this research. We thank the Halocarbon Corporation for the kind donation of the oil used in this study. SUPPORTING INFORMATION AVAILABLE Additional information, in the form of a “Practical Guide” outlining how to best perform the described studies. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review October 15, 2007. Accepted October, 18, 2007. AC702119D