Limited-Sample NMR Using Solenoidal Microcoils, Perfluorocarbon

Anal. Chem. 1998, 70, 5326-5331

Limited-Sample NMR Using Solenoidal Microcoils, Perfluorocarbon Plugs, and Capillary Spinning Babak Behnia† and Andrew G. Webb*

Department of Electrical and Computer Engineering and Beckman Institute for Advanced Science and Technology, University of Illinois at UrbanasChampaign, 405 North Mathews Avenue, Urbana, Illinois 61801

This study demonstrates three improvements to masslimited NMR using solenoidal microcoils as detectors: (1) sample confinement using liquid perfluorocarbon plugs to increase the observe factor, (2) design and incorporation of a capillary spinner to improve spectral line widths, and (3) facile sample changing via the use of a capillary insert. The probe is constructed to spin a fused silica capillary of 530 µm i.d., 700 µm o.d. inside a solenoidal coil wound around a 0.8 mm i.d., 1 mm o.d. glass capillary. The smaller capillary contains the sample, and capillaries with different samples can be exchanged easily. In high-resolution limited sample microcoil NMR studies published thus far, the length of the sample plug has been 7-10 times the length of the solenoid to avoid line broadening from volume magnetic susceptibility (χv) mismatches at both ends of the sample. This arrangement is not efficient since it places most of the sample volume outside of the coil observe volume. It is shown here that the observe factor cannot exceed 23% if the sample plug is bracketed by air, without substantial line broadening occurring. However, if the sample is bracketed by two liquid perfluorocarbon plugs, the observe factor can be increased to 70% while maintaining high spectral resolution. This is equivalent to improving the limit of detection by a factor of 3, or reducing the total data acquisition time for a given signal-to-noise by a factor of 9. It is also shown that, for the 440-nL sample plug used in this study (bracketed by the perflurocarbon plugs), sample spinning can improve the spectral resolution from 1.5 (nonspinning) to 0.6 Hz (spinning). This corresponds to a further improvement in the limit of detection of 2.5, or just over a factor of 6 decrease in data acquisition time. Nuclear magnetic resonance (NMR) spectroscopy is a powerful method for determining molecular structure. However, its relatively low sensitivity, compared to other analytical techniques, limits detection of small amounts of material. To compensate for this, a number of researchers have developed solenoidal microcoils, miniature-sized radio frequency (rf) coils, as NMR detectors.1-9 * Address correspondence to this author. E-mail: [email protected]. Phone: (217) 333-7840. Fax: (217) 244-0105. † E-mail: [email protected]. Phone: (217) 333-0188. Fax: (217) 244-0105. (1) Wu, N.; Peck, T. L.; Webb, A. G.; Magin, R. L.; Sweedler, J. V. Anal. Chem. 1994, 66, 3849-3857.

5326 Analytical Chemistry, Vol. 70, No. 24, December 15, 1998

Two recent review articles summarize work in this area in both static and dynamic (NMR-coupled separations) applications.5,7 These investigations have utilized the fact that, given a limited amount of sample, the signal received from the sample is maximized when the sample is concentrated in the minimum volume of solvent possible, and the smallest rf coil is constructed to enclose this sample volume. However, as a result of the proximity of the sample and the coil, magnetic field distortions within the sample are considerably greater than those for conventionally sized coils. It is important to minimize such field distortions since they result in line broadening, and hence a loss of information, and a lower signal-to-noise ratio (SNR). One source of field distortions is the copper wire typically used in wrapping the microcoil. It has been shown that, by immersing the microcoil in Fluorinert (FC-43, a mixture of C12 perfluorobutylamines), which is a fluid with a volume magnetic susceptibility (χv) similar to that of copper, line broadening can be reduced significantly.3 This arrangement approximates an infinite medium with uniform χv surrounding the cylindrical sample. It is known that this geometry establishes a uniform field inside the sample, even if the χv of the surrounding medium is different than that of the sample.10-13 A second source of magnetic field distortions is the discontinuity in χv at both ends of the sample, which is typically shaped as a cylindrical plug. To minimize field distortions in the observe volume (Vobs) of the coil, the plug length should be significantly longer than the coil length.7,9 This arrangement places most of the sample outside of the Vobs, thus lowering the sample sensitivity (Ssmp), defined here as the SNR per unit volume of the total (2) Peck, T. L.; Magin, R. L.; Lauterbur, P. C. J. Magn. Reson. Ser. B 1995, 108, 114-124. (3) Olson, D. L.; Peck, T. L.; Webb, A. G.; Magin, R. L.; Sweedler, J. V. Science 1995, 270, 1967-1970. (4) Webb, A. G.; Grant, S. C. J. Magn. Reson. Ser. B 1996, 113, 83-87. (5) Webb, A. G. Prog. Nucl. Magn. Reson. Spectrosc. 1997, 31, 1-42. (6) Olson, D. L.; Lacey, M. E.; Sweedler, J. V. Anal. Chem. 1998, 70, 645650. (7) Olson, D. L.; Lacey, M. E.; Sweedler, J. V. Analytical Chemistry, 1998, 70, 257A-264A. (8) Subramanian, R.; Webb, A. G. Anal. Chem. 1998, 70, 2454-2458. (9) Subramanian, R.; Lam, M. M.; Webb, A. G. J. Magn. Reson, 1998, 133, 227-231. (10) Callaghan, P. T. J. Magn. Reson. 1990, 87, 304-318. (11) Bhagwandien R.; Van Ee, R.; Beersma R.; Bakker, C. J. G.; Moerland M. A.; Lagendijk J. J. W. Magn. Reson. Imaging 1992, 10, 299-313. (12) Griffiths, D. J. Introduction to Electrodynamics; Prentice Hall Inc.: Englewood Cliffs, NJ, 1989; pp 245-269. (13) Ludeke, K. M.; Roschmann, P.; Tischler, R. Magn. Reson. Imaging 1985, 3, 329-343. 10.1021/ac9808371 CCC: $15.00

© 1998 American Chemical Society Published on Web 11/06/1998

sample. This also corresponds to a decrease in the previously defined observe factor (Fobs), the volume of sample inside the coil divided by the total sample volume. This study aimed to maximize the value of Fobs for static NMR investigations while maintaining high spectral resolution. FC-43 liquid was placed not only outside the coil as previously, but also on both sides of the sample plug within the sample capillary. This approach is similar in principle to using solid susceptibilitymatched plugs14,15 at larger scale for conventional probes. As a result, the line widths obtained with Fobs of 70% were measured to be the same as those obtained with an “infinite length sample”. If the sample is bracketed by air, this value of Fobs drops to approximately 25%. A second widely used technique to improve spectral resolution using standard NMR probes is sample spinning, where an airdriven turbine spins the vertically oriented sample tube about its long axis. The NMR signal behaves as if the nuclei experience a single field that is the average of the fields along the trajectory; i.e., the effects of any field gradients that are perpendicular to the axis of rotation are reduced. For maximum efficiency, the rate of rotation should be greater than the line width obtained from the sample when stationary.16 To reduce the susceptibility induced line broadening in microcoils, a newly designed and simple-to-build sample spinner has been fabricated. The spinner rotates a 530-µm-i.d./700-µm-o.d. capillary, containing a total of 440 nL of sample, inside a 0.8-mm-i.d./1-mm-o.d. capillary supporting the microcoil. It is shown that, in this size regime, sample spinning reduces spectral line width by a factor of 2. EXPERIMENTAL SECTION Chemicals. D2O (D, 99.9%) was obtained from Cambridge Isotope Laboratories (Andover, MA) and sucrose from EM Science (Gibbstown, NJ). All chemicals were used without further purification. FC-43 was obtained from 3M Corp. (St. Paul, MN). Microcoil Fabrication. The solenoid coil was fabricated by wrapping 75-µm-diameter copper wire (Cu 99.99%, California Fine Wire, Grover Beach, CA) around a 0.8-mm-i.d./1-mm-o.d. glass capillary (Drummond Scientific, Broomall, PA) using previously described fabrication techniques.3,4 The 14-turn solenoid coil used in these studies was 1.4 mm long, the extra length being due to the finite thickness of the wire insulation. Vobs was 310 nL for the 530-µm-i.d./700-µm-o.d. sample capillary. Electrical Characterization of the Microcoils. Radio frequency measurements of the microcoil were performed using an impedance analyzer (HP 4396B, Hewlett-Packard, Palo Alto, CA) with the microcoil immersed in FC-43 and 8-mm wire leads extending from the solenoid. An impedance test adaptor (HP 43961A) and a test fixture (HP 16191A) were used to carry out the measurements. The inductance, Q at 250 MHz, and the selfresonance frequency were 160 nH, 63, and 587 MHz, respectively. After construction of the capillary spinning probe, the impedance of the solenoid was matched to 50 Ω at 250 MHz with three variable nonmagnetic capacitors (0.6-4.5 pF Gigatrim, Johanson (14) Wilmad Glass NMR products catalog. http://www.wilmad.com/html/nf/ DotyPlugs.html (accessed June 1998). (15) Doty, F. D.; Entzminger, Y.; Yang, A. Concepts Magn. Reson. 1998, 10, 133-156. (16) Freeman, R. A Handbook of Nuclear Magnetic Resonance: John Wiley & Sons: New York, 1988; pp 194-197.

Figure 1. Schematic of the spinner probe. The 5-mm-long bearings are supported to the coil capillary via a larger capillary that houses both the bearing and the coil capillary (top portion of figure). Epoxy adhesive is placed in the space between the larger capillary and the smaller capillaries for mechanical support. The polyethylene bottle (not shown) is glued to the edges of the circular hole in the PCB. A coaxial cable is soldered onto the appropriate copper pads.

Mfg. Co., Boonton, NJ) using a network analyzer (HP 8751A, Hewlett-Packard) and an S-parameter test set (HP87511A). Design and Construction of Capillary Spinning Probe. A detailed schematic of the probe assembly is shown in Figure 1. The glass capillary was cut to a length of 1.8 cm, with the solenoid centered along the long axis. Two 5-mm-long sections of 700µm-i.d./850-µm-o.d. polyimide-coated fused silica capillary (Polymicro Technologies Inc., Phoenix, AZ) served as sleeve bearings. They were centered and glued to both ends of the coil capillary using ultraviolet (UV) curable adhesive (Optical Adhesive No. 63, Norland Products Inc., New Brunswick, NJ), as shown in the top portion of Figure 1. A printed circuit board (PCB, Kepro Circuit Systems, Fenton, MO) was patterned with a 1.4-cm-diameter hole drilled in the center. The capillary was positioned on the PCB surface such that the coil windings were above the center of the hole. Using quick setting epoxy (Loctite, Rocky Hill, CT), both ends of the capillary extending beyond the hole were glued to the PCB surface. The wire leads from the microcoil were soldered onto the appropriate copper pads on the PCB. Three variable nonmagnetic capacitors were also soldered on the appropriate copper pads for impedance matching. A 1.4-cm-diameter polyethylene bottle (not shown in Figure 1) was cut in half and each half glued on either side of the hole on the PCB. This arrangement created a cavity that housed the microcoil and the susceptibility matching fluid (FC-43). A 4-cm-long piece of 530-µm-i.d./ 700-µm-o.d. polyimide-coated fused silica capillary (Polymicro Technologies Inc.) served as the sample capillary. A lightweight Analytical Chemistry, Vol. 70, No. 24, December 15, 1998

5327

Figure 2. Photograph of two views of the spinner probe assembly, including the polyethylene bottle containing the fluorinert liquid.

turbine was constructed and glued to one end of the sample capillary. After the sample capillary was loaded with sample, it was sealed at both ends using UV curable adhesive. A flat glass surface (capillary positioning block in Figure 1) was attached to the PCB in order to support one end of the sample capillary with minimal friction as it spins. The assembled probe was mounted on an aluminum cylinder that fitted into the magnet bore. An air jet from a 3-mm-diameter aperture on a glass tube placed 5 mm away struck the turbine and caused both the turbine and the sample capillary to spin. Air was supplied to the glass tube via 6-mm-i.d. rubber tubing. An optical tachometer described previously17 was constructed to measure the rate of rotation under such conditions. Using this arrangement, the spinner operates at a rate of 100-150 Hz. A photograph of the spinner probe assembly is shown in Figure 2. Sample Preparation and Handling. Four different sample configurations were used throughout this study, as depicted in Figure 3. In configuration A, the sample plug was bracketed by air on both sides. The length (l) of the sample plug was varied from 2 to 6 mm. In configuration B (“infinite sample”), the sample plug was 15 mm in length and bracketed by a short plug of FC43 on both sides. The FC-43 served only to mechanically contain the sample during spinning and also prevented the sample from being contaminated by the UV curable adhesive. In configuration C, the sample plug was 2 mm in length and bracketed on both sides by solid Teflon beading (500 µm o.d., Zeus, Orangeburg, SC). In configuration D, the sample plug was also 2 mm in length but was bracketed on both sides by FC-43 liquid. For all configurations, the sample, adhesive, and FC-43 were loaded into the capillary tube by separately drawing each from their respective vials under a microscope. The adhesive plug farthest away from the turbine was drawn last and cured inside the capillary via exposure by UV light. Since UV light does not transmit through the polyimide coating of the sample capillary, it was guided through an optical fiber and directed toward the capillary aperture. The other end of the capillary was sealed by injecting UV curable (17) Behnia, B. M.Sc. Thesis, University of Illinois at UrbanasChampaign, 1998.

5328 Analytical Chemistry, Vol. 70, No. 24, December 15, 1998

Figure 3. Sample configurations used in this study: A (airbracketed), B (infinite sample), C (Teflon-bracketed), and D (FC-43bracketed). For configuration A, the length, l, of the sample plug is varied from 2 to 6 mm.

adhesive using a smaller 250-µm-i.d./350-µm-o.d. capillary and an attached syringe. NMR Spectroscopy. All NMR experiments were carried out at 250 MHz (5.875 T) using an 89-mm wide-bore Oxford Instruments magnet (model 2280), a Tecmag (Houston, TX) Libra console running MacNMR 5.5, a 50-W power amplifier (American Microwave Technology, model M3137, 200-500 MHz), and a broadband preamplifier (AU-1054, 1-500 MHz, 1.09 dB noise figure, 32 ( 0.3 dB gain, Miteq, Hauppauge, NY). After pulse

Figure 4. (a-d). Spectra of 5% H2O/95% D2O with varying sample lengths. All were acquired with a spectral width of 2 kHz, 4096 complex data points, and 1 scan; all spectra are displayed on the same scale in this figure. The vertical scale is kept constant for Figures 4-6.

Figure 5. (a-c). Spectra of 5% H2O/95% D2O using the sample configurations shown in Figure 2. All were obtained with a spectral width of 2 kHz, 4096 complex data points, and 1 scan; all are displayed on the same scale.

calibration, each sample was shimmed using an automatic shimming procedure (simplex algorithm) that starts from zero shim values and manipulates all 12 shims. For the H2O/D2O samples used, to determine that radiation damping was not occurring, the tip angle was varied from 90° to 5°. The line width and resonant frequency remained unchanged, indicating no radiation damping. All spectra were acquired with a 90° pulse length which was 5.5 µs with the 10 dB attenuator at the input to the 50-W amplifier. SNR and line width measurements were performed using the NUTS software package (Acorn NMR Inc., Fremont, CA). Spectrometer Noise Characterization. In addition to coil sensitivity and figures of merit discussed by Olson et al.,7 both the SNR and limits of detection (LODs) are dependent upon the spectrometer noise (including noise from the preamplifier), which will vary for different systems. The noise level must be taken into account for comparing results from different spectrometers. This measurement was performed using an rf signal generator (Fluke 6060A, 100 kHz-1050 MHz, John Fluke Mfg. Co. Inc., Everett, WA). A -80-dBm (equivalent to 64 µV peak-to-peak), 250-MHz sinusoidal signal was fed directly into the preamplifier (50 Ω input impedance). The intermediate frequency (IF) and audio gains of the spectrometer were set at values low enough to ensure that saturation did not occur. A “pseudoscan” was acquired with a spectral width of 2 kHz and the signal level of the sinusoid measured. Since this level was known to be 64 µV, this measurement provided a calibration for the receiving system. A 50-Ω load was then placed at the input of the preamplifier, a second pseudoscan collected, and the standard deviation of the noise signal (time domain) measured. The signal was converted to units of volts using the conversion ratio previously established, with a value in this case of 0.174 µV. Using this procedure, LODs from different hardware systems can be directly compared, since noise levels may vary widely. For example, if the same NMR experiments performed in this study were performed on a spectrometer with an equivalent input noise of 0.087 µV, the SNR would be twice the SNR obtained here. RESULTS AND DISCUSSION Four sets of experiments were performed. The first illustrates limitations of the value of Fobs if the sample is bracketed by air.

The second experiment demonstrates improvements in Fobs by using a bracketing plug that is better susceptibility matched than air to the sample plug. The third shows further improvements in resolution from sample spinning. The last set of experiments shows data from a limited sample of a small organic molecule (sucrose) using the combination of sample spinning and an FC43 susceptibility plug. The first experiment was designed to determine the length of a sample plug required for undistorted spectra to be obtained if the plug is bracketed by air. One “infinite sample” capillary (Figure 3, configuration B) and three different air-bracketed samples (Figure 3, configuration A) were prepared with sample lengths of 2, 4, and 6 mm. The sample plug was a solution of 5% H2O/95% D2O. The NMR spectra corresponding to each sample configuration are shown in Figure 4. The line width for the airbracketed sample plugs decreased from 6.4 to 2.4 and finally to 1.5 Hz as the plug length increased from 2 to 4 and to 6 mm, respectively. It is clear that the line width for an infinite-sample plug (1.4 Hz) was almost fully recovered when the air-bracketed sample plug was elongated to 6 mm, corresponding to a value of Fobs of only 23%. It should be noted that, with the nonspinning configuration used in these experiments, some magnetic susceptibility-induced distortions were still present, even in the spectrum from the infinite sample. The second set of experiments showed that the value of Fobs can be increased by bracketing the sample plug with FC-43 liquid plugs. Experiments using solid Teflon plugs were also performed. Three different sample capillaries were prepared, corresponding to configurations B (infinite sample), C (Teflon-bracketed), and D (FC-43 bracketed) in Figure 3. Figure 5 shows spectra of 5% H2O/95% D2O using the three sample configurations. From this set of experiments, it is clear that the line widths improve from air-bracketed to Teflon-bracketed to FC-bracketed samples. The reported values for χv of air, Teflon, FC-43, D2O, and fused silica are +0.38 × 10-6, -10.5 × 10-6, -8.2 × 10-6, -8.8 × 10-6, and -11.9 × 10-6, respectively (all values are converted to SI units).3,15,18 These results are consistent with the reported values because bracketing plugs with χv closest in value to that of the (18) Keyser, P. T.; Jefferts S. R. Rev. Sci. Instrum. 1989, 60, 2711-2714.

Analytical Chemistry, Vol. 70, No. 24, December 15, 1998

5329

Figure 6. (a-c). Spectra of 5% H2O/95% D2O using sample configurations B, C, and D from Figure 2, showing the effects of spinning. All spectra were acquired with a spectral width of 2 kHz, 4096 complex data points, and 1 scan, and are all displayed at the same scale. (d) Same spectrum as (c) expanded to show the spinning sidebands at 130 Hz offset from the main peak. The intensity is approximately 2.5% that of the main peak. Other smaller peaks in (d), such as the small blips at +48 and -54 Hz, are due to coherent noise from the spectrometer.

Figure 7. Spectra of 220 nmol of sucrose. The total volume used is 440 nL. Both data sets were acquired with a spectral width of 800 Hz, 1024 complex points, 32 scans, and 3.3 s delay between scans. Both spectra are shown using the same scale for direct comparison. (a) Spectrum for the air-bracketed sample which is not spinning. (b) Spectrum for the FC-43-bracketed sample which is spinning.

sample should result in the lowest line width. The line width in the spectrum from the FC-43-bracketed sample capillary with a 70% Fobs is comparable to that of the infinite sample plug. An increase from 23% to 70% corresponds to a 9.3 factor decrease in total acquisition time for a given SNR, or to an improvement in the LOD by a factor of 3. In terms of maintaining sample purity, it should be noted that the solubility of most organic and inorganic compounds in perfluorocarbons is extremely low, although in certain cases this particular susceptibility matching fluid would not be appropriate. This sample preparation technique is easiest for high-viscosity, high-boiling-point solvents such as D2O. With careful preparations, we have also been able to handle more volatile solvents such as ethanol. The third set of experiments illustrated the effect of sample spinning on the line width. The same samples as in the previous set of experiments were used, and the resulting spectra shown in Figure 6. The line widths for the infinite sample, Teflon-bracketed, and FC-43-bracketed sample plugs are 0.6, 1.2, and 0.6 Hz, respectively. It was not possible to spin the air-bracketed samples because the sample plug slid inside the capillary. However, it is clear that spinning improves the line width considerably. Also, it should be noted that, for the FC-43-bracketed sample, the line 5330 Analytical Chemistry, Vol. 70, No. 24, December 15, 1998

width is almost identical to that from the infinite sample. Therefore, in addition to the improvement in Fobs, the improvement in line width from 1.5 Hz nonspinning to 0.6 Hz spinning can reduce the data acquisition time by a further factor of approximately 6. Figure 6d shows the level of spinning sidebands associated with this setup: the sample was spun at 130 Hz, and the sideband level was measured to be 2.5%. As commonly done with other spinners, it is feasible to vary the sample spinning rate to blur out the spinning sidebands. The final set of experiments illustrates the overall advantage of using FC-43-bracketing and sample spinning. Two 440-nL (length ) 2 mm) samples of 500 mM (220 nmol total) sucrose were prepared in 4% CH3OH/96% D2O: one as an FC-43-bracketed sample and the other as an air-bracketed sample. The spectrum of the FC-43-bracketed sample was acquired with spinning; the spectrum of the air-bracketed sample was acquired without spinning. Both spectra are displayed in Figure 7 at the same scale for direct comparison. While the anomeric splitting is absent in the air-bracketed spectrum, the spectrum of the FC-43-bracketed sample displays a splitting of 80%. The SNR of the largest peak in the sucrose spectra from the spinning probe is 140. Given that 220 nmol of sample was used, the LOD for sucrose defined as the amount of sample needed to yield an SNR of 3 can be

calculated to be 4.7 nmol for the spectral parameters used here. The high concentration of sucrose used, and the addition of a small amount of methanol, were necessary due to the absence of a lock channel in the Tecmag system. Since the protons of the residual water in the D2O solution exchange-broaden with the OH groups of sucrose, the natural T2 of the H2O peak is short, and shimming to subhertz line widths is not possible. However the CH3 protons of methanol do not exchange, giving rise to a longer T2 relaxation time. In addition, without a lock channel, the resonance frequency drifts slightly over long acquisition times. This drift is particularly detrimental when frequency shifts are in the same order as the spectral line widths, causing significant broadening when the spectra are averaged. To keep the total acquisition time to 2.5 min and thus minimize this broadening effect, a high concentration of sucrose (500 mM) was chosen. For spectrometers with a lock channel, such a high sample concentration and addition of methanol are not necessary.

In conclusion, the addition of liquid perfluorocarbon susceptibility-matched plugs and the design of a sample spinner should greatly improve the quality of NMR spectra from mass-limited samples, as well as allowing quick sample changing, which is a prerequisite for many industrial applications. ACKNOWLEDGMENT This work was supported by a grant from the National Institutes of Health (PHS 1 R01 GM53030-01). We acknowledge the assistance and technical support of Dr. Raju Subramanian, Dr. Tim Peck, Dr. Dean Olson, Mike Lam, Sam Grant, and Mike Lacey.

Received for review July 29, 1998. Accepted September 29, 1998. AC9808371

Analytical Chemistry, Vol. 70, No. 24, December 15, 1998

5331