Article pubs.acs.org/ac
Contactless NMR Spectroscopy on a Chip Herbert Ryan,† Suk-Heung Song,† Anja Zaß,‡,⊥ Jan Korvink,‡,§ and Marcel Utz*,†,∥,⊥ †
Department of Mechanical and Aeropspace Engineering, University of Virginia, Charlottesville, Virginia 22904, United States Department of Microsystems Engineering, University of Freiburg, Germany § Freiburg Institute for Advanced Studies, Germany ∥ Department of Chemistry, University of Virginia, Charlottesville, Virginia 22904, United States ⊥ Center For Microsystems For The Life Sciences, University of Virginia, Charlottesville, Virginia 22904, United States ‡
S Supporting Information *
ABSTRACT: Inductively coupled planar resonators offer convenient integration of high-resolution NMR spectroscopy with microfluidic lab-on-a-chip devices. Planar spiral resonators are fabricated lithographically either by gold electroplating or by etching Cu laminated with polyimide. Their performance is characterized by NMR imaging as well as spectroscopy. A single-scan limit of detection LODt = 0.95 nmol s1/2 was obtained from sample volumes around 1 μL. The sensitivity of this approach is similar to that obtained by microstripline and microslot probes.
M
As has been shown both theoretically18−22 and experimentally,23−25 the specific sensitivity (per mole of sample) of NMR measurements increases dramatically as the receiver coil is scaled down. This has been exploited extensively to devise “hyphenated” techniques2,3,26−29 such as high-performance liquid chromatography (HPLC)−NMR, in which the sample is injected after chromatographic separation into a capillary that runs through a miniaturized solenoid coil or forms the dielectric in a microstripline. Microfluidic lab-on-a-chip devices, by contrast, have a planar geometry. Obtaining NMR signals with optimum sensitivity from such a device requires receiver coils with a similar form factor. Planar spiral coils can be made by lithographic techniques and have been successfully integrated with simple microfluidic systems.7−10 High sensitivity NMR signals can also be obtained from microfluidic systems by remote detection.4,5,30−33 In this case, the signal is encoded by a large transmitter coil that fits around the entire microfluidic device. Detection, however, is done using a microsolenoid wrapped around a capillary downstream from the microfluidic device. In all these cases, the integration of microfluidics with NMR spectroscopy requires either an electrical contact or a fluidic connection to the lab-on-a-chip device. This is inconvenient and severely limits high-throughput applications. Here, we present contactless microfluidic chips based on planar microcoils that can be coupled inductively to the NMR console by simply inserting them into a probe assembly equipped with a probe coil that is large enough to
icrofluidic lab-on-a-chip devices represent a rapidly emerging technology based on the paradigm of integrating complex biological or biochemical assays on a compact, expendable platform.1 Substantial progress has been demonstrated in recent years, in particular in the development of on-chip methods for the culture, manipulation, sorting, and characterization of cells. Integration of such devices with highresolution spectroscopy offers obvious advantages. NMR spectroscopy is one of the few tools available that allows one to quantify the metabolome of biological fluids noninvasively, without the use of fluorescent labels and without destroying the sample. Because of the limited sample volumes, microfluidic NMR spectroscopy benefits from the use of miniaturized receiver coils. Several designs, such as solenoids wound around capillaries,2−6 planar microcoils,7−10 microstriplines,11−14 and microslot probes15 have been described. However, these approaches require a fixed fluidic infrastructure to accommodate the sample. This runs counter to the paradigm of expendable LOC devices, and is difficult to combine with highthroughput clinical screening applications. By contrast, our approach relies on planar rf resonators that are integrated into the microfluidic platform. These are essentially 2D metal structures designed to resonate at the Larmor frequency, which are coupled inductively to the NMR receiver and transmitter. This mode of operation allows for easy insertion and removal of the microfluidic device from the spectrometer. It is important to note that inductive coupling does not incur a significant cost in sensitivity.16,17 In fact, we found the sensitivity of spiral-coil resonators to exceed results that have been reported for conventionally coupled spiral coils of similar geometry.7,8 © 2012 American Chemical Society
Received: January 19, 2012 Accepted: March 12, 2012 Published: March 12, 2012 3696
dx.doi.org/10.1021/ac300204z | Anal. Chem. 2012, 84, 3696−3702
Analytical Chemistry
Article
removed by brief etching in aqua regia, and a layer of 10 μm of SU-8 was spun onto the structure for protection of the resonator. Process B relied on prefabricated polyimide laminate sheets of 50 μm thickness with a one-sided coating of 18 μm Cu (Upisel-N, UBE Industries, Ltd.). The as-received laminate was cleaned with acetone, and a layer of AZ 1500 series positive photoresist was spin coated onto the Cu side. The coil pattern was exposed using a high-resolution printed transparency mask, which was obtained from the AutoCAD drawings, on a Karl Süss MJB3 mask aligner. After development of the photoresist, the unwanted Cu was removed by etching in H2SO4/H2O2/ H2O (1:2.3:20). Finally, a thin layer of MEMS wax was coated on to protect the Cu coil from oxidation. Micrographs of coils fabricated by processes A and B are shown in Figure 3.
accommodate the entire chip. The principle is illustrated in Figure 1. Inductive coupling is widely used in NMR
Figure 1. Integration of microfluidic chip with microfabricated resonator: (A) top view, (B) cross-section view, and (C) assembly. (1) Cu/Au self-resonant planar microcoil, (2) fluidic network, (3) protection dielectric (MEMS wax or SU-8), (4) coil substrate (polyimide/glass), (5) cover layers (PMMA/glass).
imaging.34,35 Recently, Sakellariou et al. have shown that substantial gains in sensitivity are possible in magic angle spinning experiments on mass-limited samples by including inductively coupled resonators into the MAS rotor.36 As discussed in the remainder of this paper, planar spiral microcoils with the required self-resonance frequencies in the range of several hundred MHz and quality factors above 50 are easily obtained by simple lithographic techniques. Microfluidic devices can be integrated with NMR spectroscopy by simply attaching such resonators to their surface. Thus equipped, the microfluidic device is then just as easy to introduce and remove from the NMR probe as a conventional 5 mm sample tube. This offers considerable advantages, particularly in view of future clinical applications based on NMR metabolomics.
Figure 3. Planar microfabricated resonators. (Left) Self-resonant coil using the stray capacitance between turns for tuning and (right) Cu coils fabricated on Upilex preplated polyimide foil.
Fluidic Channels. Glass and poly(methylmethacrylate) microfluidic chips were used in conjunction with the microresonators. The glass chips were fabricated from borofloat glass by first sputtering Cr and spin coating a layer of AZ 1600 photoresist. The fluidic pattern, drawn in AutoCAD, was then exposed using a direct laser writer and developed. After Cr etching, the fluidic channels were etched in hydrofluoric acid to a depth of 150 μm. Inlet and outlet holes were drilled into the substrate using diamond drill bits. The chips were then covered by a second glass layer, which was thermally fusion bonded. PMMA chips were fabricated from commercial PMMA sheets (McMaster-Carr, Santa Fe Springs, CA) of 200 μm thickness. The fluidic pattern was cut into the PMMA using a VersaLASER system 3.50 (Universal Laser Systems, Scottsdale, AZ). The fluidic pattern was drawn in CorelDraw. Each chip consisted of three PMMA layers: a bottom layer, a middle layer with the fluidic pattern, and a cover layer with inlet and outlet access holes. The layers were bonded together using two bonding layers made from double-sided adhesive sheets (Adhesives Research, Inc., Glen Rock, PA) of 157 μm thickness, into which the fluidic pattern was cut. This resulted in a well depth of 514 μm.
■
EXPERIMENTAL SECTION Microcoils. Planar NMR resonators were fabricated by two different routes. Process A departed from a borofloat glass substrate, onto which a Cr adhesion layer of 5 nm and gold seed layer of 50 nm thickness were sputtered after thorough cleaning with piranha solution and acetone. The sample was then spin-coated with 16 μm of AZ 4500 series positive photoresist. The coil pattern was drawn in AutoCAD and exposed onto the specimen by a Heidelberg DWL66 direct laser writer at a resolution of 4 μm. After development of the photoresist and brief O2 -plasma cleaning, the coil structure was electroplated in Au up to a thickness of 7 μm. The photoresist was then stripped with dimethyl sulfoxide. The seed layer was
Figure 2. Gradient echo images of a planar resonator immersed in doped water at increasing excitation power. 3697
dx.doi.org/10.1021/ac300204z | Anal. Chem. 2012, 84, 3696−3702
Analytical Chemistry
Article
the width of the peak at −3 dB from the maximum as Q = Δω−3 dB/ω0. Materials. D2O (99.99%) and α-D-glucose (puriss.) were obtained from Aldrich. Glucose solutions were stored for at least 24 h at room temperature after dissolution to ensure complete isomerization before the NMR measurements.
NMR Experiments. All NMR experiments were conducted with a Varian VNMR spectrometer, operating at a static field strength of 14.1 T. Images have been acquired with a Doty 12 mm microimaging probe using a sample of 150 mM NaCl and 10 mM CuSO4 in H2O. Gradient echo images were obtained using a Gaussian excitation pulse profile with 2 ms duration. Slice thickness was 0.5 mm; a gradient echo time of 2 ms and a relaxation delay of 500 ms were used. The latter ensured that the image contrast was not affected by T1 relaxation. A dedicated NMR probe head was built for inductive coupling of microfluidic chips. The commonplace pneumatic sample transport mechanism was modified to accommodate rectangular chips instead of conventional 5 mm NMR sample tubes. The design of this probe is shown in Figure 4. The probe head
■
RESULTS AND DISCUSSION The resonance frequencies and Q factors obtained from a series of coils are shown in Figure 5. The coils have been fabricated
Figure 4. (A) Schematic of the probe assembly and (B) 3D CAD drawing of the sample orientation/support structure. (1) Probe coil, made from self-adhesive Cu tape mounted on (2) support glass slide, (3) matching capacitor (2−10 pF), (4) matching adjustment rod, (5) ground plane, (6) semirigid coaxial cable, and (7) probe sheath.
structure shown in the figure was manufactured directly from the CAD drawings using a uPrint SE 3D fused deposition modeling system (Stratays, Inc., Eden Prairie, MN) from acrylonitrile-butadiene-styrene (ABS) copolymer. The primary coil structure was cut from self-adhesive copper tape (3M) and glued to two microscope slides that snapped into the sample holder on either side of the pillars shown in Figure 4. All spectra and images were processed using our own routines written in the Mathematica language. Radiofrequency Characterization. The electrical response of the resonators was tested using two circular loops on a printed circuit board arranged such that their mutual inductance largely cancels. Bringing a tuned resonator into proximity of the overlap region of the loops leads to interference with this cancellation. The resonance was observed by measuring the transmission loss (S12) through the double loop device with a HP 8735D vector network analyzer. While the position of the resonance peak directly indicated the resonance frequency, the quality factor was determined from
Figure 5. (a) Resonance frequency and (b) quality factor of planar coil resonators as a function of the number of turns. Results for coils of 5.75 and 6.00 mm outer diameter are shown.
from Cu-coated polyimide film, as discussed in the Experimental Section. The width of the conductor path was 100 μm, with a gap between turns of 50 μm in all cases. The number of full turns was varied between 6 and 8, and coils with outer diameters of 6 and 5.75 mm were fabricated. As the data in Figure 5 demonstrates, the resonance frequency drops with an increasing number of turns and coil diameter, as expected. By contrast, the Q factors seem largely constant with a median value of 53 over the range of geometries explored here, even though there is substantial experimental scatter. In order to test the behavior of planar coil resonators in a magnetic resonance context, a coil was placed vertically in a 12 mm sample tube filled with doped water. Figure 2 shows gradient echo images obtained from this system as a function of 3698
dx.doi.org/10.1021/ac300204z | Anal. Chem. 2012, 84, 3696−3702
Analytical Chemistry
Article
complete review has recently been given by Jacquinot and Sakellariou;17 and a few specific points relevant to the present work are summarized in the Supporting Information. The experiments described here were carried out under conditions of strong coupling, with the probe coil tuned to a frequency much higher than the Larmor frequency. Under these conditions, the sensitivity is governed by the geometry and quality factor of the microcoil and largely independent of the probe coil. A quantitative assessment of probe efficiency and sensitivity was obtained by combining a microfluidic glass chip with a Cu/ polyimide coil. A sample chamber with the geometry shown in Figure 7A was made from borofloat glass, as described in the
excitation pulse intensity. The local enhancement of the radio frequency field due to the resonance of the coil is clearly visible in the images. The contrast stems from two sources: on the one hand, the current in the resonator increases the local rf field strength, leading to a larger pulse excitation angle. In addition, the resonant coupling of the microcoil to the probe causes the resonance signal from the vicinity of the microcoil to be recorded with higher sensitivity. The signal intensity from the area close to the resonator and the background is plotted in Figure 6. The enhancement due to
Figure 6. Dependence of signal strength on excitation power.
the resonator is manifest in two different ways: the maximum signal close to the resonator is a factor of 3.5 higher than in the background, and it is reached at a power level of 200 mW, while the background signal peaks at 2000 mW. Since the B1 field strength is proportional to the square root of the applied power, this corresponds to a factor of 3.16. This close agreement of the enhancement factors constitutes a nice confirmation of the correspondence principle that relates the sensitivity of the NMR experiment to the probe efficiency (magnetic field strength generated per square root of power). In these experiments, the resonator coil was coupled inductively to the tuned Litz coil built into the Doty microimaging probe. If both the resonator and the probe are tuned to the same frequency, the coupling between them leads to a splitting of the resonance. The resonance frequency has to be chosen such that one of the two resonance lines coincides with the Larmor frequency. With proper matching of the probe, the available power is then split evenly between the probe coil and the resonator coil. In the case of the experiments shown above, the coupling splits the resonance by about 40 MHz. Unfortunately, the tuning range of the microimaging probe does not reach up to the required 620 MHz. The experiment was therefore carried out with a resonator at lower selfresonance, and the tuning of the probe was kept close to 600 MHz. This approach leads to inefficient coupling, with the majority of the available power dissipated in the probe coil instead of the resonator. The magnification factor of 3.16 reported above is therefore not indicative of the maximum achievable probe efficiency. Nonetheless, the experiment allows one to validate the correspondence principle and provides useful information on the shape of the rf field generated by the resonator. The theory of inductively coupled NMR detection systems has been discussed by a number of authors; a particularly
Figure 7. (A) Single scan 1H spectrum of 5% v/v ethanol in D2O obtained in a sample chamber of the dimensions shown and 150 μm depth. (B) Nutation spectrum from the same setup, at an rf pulse power of 6.3 W.
Experimental Section, consisting of a long linear channel of 2 mm width, 150 μm depth, and 12 mm length. This chamber was combined with a Cu/polyimide resonator of 6.75 turns and 6 mm outer diameter. The active sample volume is approximately 2 × 4 × 0.15 mm3 = 1.2 μL. The chip was then inserted into the probe assembly shown in Figure 4. A sample of 5% ethanol in D2O was used to obtain the spectrum shown in Figure 7A. A single transient of 10 000 data points over a spectral window of 5 kHz was recorded following a 90° pulse. The sensitivity was determined from the signal/noise ratio (SNR) in the time domain (defined as the maximum signal height in the free induction decay divided by twice the root-mean-square noise amplitude, which was SNR t = 955 ± 10
(The proton content of the D2O used in this experiment was determined by an independent experiment at 0.5%, and has been properly taken into account.) The time-domain limit of detection, defined as the number of spins that have to resonate in a 1 Hz bandwidth to give a signal as strong as the noise, is therefore11 3699
dx.doi.org/10.1021/ac300204z | Anal. Chem. 2012, 84, 3696−3702
Analytical Chemistry
LODt =
Article
The inherent line width (fwhm) was determined by fitting the triplet line with a Lorentzian line shape of 0.0075 ppm (4.5 Hz) width. It has become customary to assess the sensitivity of micro-NMR systems through the anomeric proton of sucrose. In the present case, we use the anomeric proton (position 1) of glucose for this purpose. As is well-known, glucose isomerizes into α-D-glucopyranose and β-D-glucopyranose (top left in Figure 8) in aqueous solution over several hours at room temperature. The sample was therefore prepared at least 24 h before the NMR measurements were made. The anomeric protons for the α and β isomers appear at 4.6 and 5.2 ppm, respectively.37 The signal from the β anomeric proton was used for assessment of the sensitivity. The single-scan SNR was 106. Assuming a concentration of β-D-glucopyranose of 250 mM, this corresponds to a specific SNR of 320 μmol−1. By comparison, Massin et al. have reported a value of 88 μmol−1 for a conventionally coupled planar microcoil of very similar dimensions. Their experiment was carried out at a proton resonance frequency of 300 MHz, but the quoted value was scaled to 600 MHz and is therefore directly comparable to our result. The mass sensitivity can be converted into a limit of detection, corresponding to the amount of sample that needs to be present to yield a SNR of 3 for an acquisition time of 1 s. The normalized limit of detection, thus defined, is given by15
Ns SNR t Δf
where Ns represents the number of protons contributing to the signal, and Δf is the width of the resonance line. The sum of the widths of all peaks in Figure 7A is Δf = 45 Hz. The sample volume exposed to the rf field of the resonator is about 1.2 μL. We obtain a single-scan limit of detection of LODt = 0.95 ± 0.1 nmol s1/2. By comparison, Bart et al.11 have reported a value of 0.47 nmol s1/2 for a microstripline probe, obtained from a sample of pure isopropanol, also at 600 MHz. Their value, however, was obtained using four scans rather than one and should therefore be doubled for comparison with ours. Also, their sample volume was 600 nL, compared with 1.2 μL in our case. The mass sensitivity increases with the inverse cube root of the coil volume.19 Scaling our LODt to 600 nL (a factor of 1.25) and 4 scans (a factor of 2), we obtain a value of 0.38 nmol s1/2, which is very similar to the one reported by Bart et al. Figure 7B shows a nutation spectrum recorded at an excitation power of 6.3 W. The 90° pulse length is 18 μs, which corresponds to a probe efficiency of 0.184 mT/√W or (for protons) 7.83 kHz/√W. It should be noted that this is the efficiency of the entire transmission pathway, including losses in the cables, preamplifier setup, filters, and the probe. The A450/ A90 ratio is 56%, and A810/A90 is 35%. The radio frequency homogeneity is not very good when compared to a standard 5 mm liquid-state NMR probe. However, it compares favorably with what other groups using planar microcoils have reported. Figure 8B shows a single scan spectrum of a sample of 500 mM α-D-glucose in D2O . The water signal was suppressed by 1 s presaturation at 0.25 mW at a frequency corresponding to δ = 4.76 ppm. A total of 8000 complex data points were acquired after a 90° pulse over a spectral width of 10 kHz and processed with Gaussian line broadening of 1.5 Hz.
nLODm = 3
Ns t Exp SNR
where tExp denotes the time for the experiment. From the SNR reported above, we obtain nLODm = 22.2 nmol s1/2, assuming an experiment time of 5.5 s (which includes the relaxation delay). Bart et al. have reported an almost identical value of nLODm = 22.3 nmol s1/2 for the sucrose anomeric proton in a sample volume of 0.6 μL in a microstripline probe. By contrast, lower limits of detection have been obtained from solenoid microcoil probes.2,3 For example, Lacey et al. have reported a value of 2.0 nmol s1/2 for a solenoid microcoil probe with a coil diameter of 850 mm and a sample volume of 0.62 mL.38 It is likely that this discrepancy is due to the better spectral resolution that has been achieved with solenoid coils as well as higher filling factors. As demonstrated in Figure 9, it is possible to acquire homonuclear two-dimensional spectra using the same setup. This is of particular importance in the context of the analysis of complex mixtures. Figure 9 shows a correlation spectrum (COSY) obtained from 1.2 μL of 500 mM glucose in D2O. The sample chamber in the chip design shown in Figure 7 extends substantially beyond the active area of the resonator on either side. This avoids horizontal boundaries associated with differences in magnetic susceptibility in the active sample volume but has the disadvantage that only a fraction of the fluid contained in the chip is actually subjected to the measurement. Using a round sample chamber concentrates a larger fraction of the available sample in the active region but was found to lead to impractically large susceptibility broadening (cf. Supporting Information).
■
CONCLUSIONS Microfluidic devices integrated with self-resonant planar coils can be coupled inductively to the spectrometer, allowing highresolution NMR spectroscopy with sensitivities similar to those obtained from other micro-NMR approaches. The inductive
Figure 8. 1H spectrum of glucose obtained from 1.2 μL of sample in an inductively coupled microfluidic chip. Spectral resolution is about 4.5 Hz. 3700
dx.doi.org/10.1021/ac300204z | Anal. Chem. 2012, 84, 3696−3702
Analytical Chemistry
Article
5R01EB011591; as well as the Excellence Initiative of the German Federal and State Governments. We gratefully acknowledge Dr. Kailiang Wang’s contribution to the early stages of this work.
■
(1) Manz, A.; Graber, N.; Widmer, H. Sens. Actuators, A 1990, 1, 244−248. (2) Wu, N.; Webb, A.; Peck, T.; Sweedler, J. Anal. Chem. 1995, 67, 3101−3107. (3) Webb, A. Magn. Reson. Chem. 2005, 43, 688−696. (4) Hilty, C.; McDonnell, E.; Granwehr, J.; Pierce, K.; Han, S.; Pines, A. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 14960−14963. (5) Bajaj, V. S.; Paulsen, J.; Harel, E.; Pines, A. Science 2010, 330, 1078−1081. (6) Mohmmadzadeh, M.; Baxan, N.; Badilita, V.; Kratt, K.; Weber, H.; Korvink, J. G.; Wallrabe, U.; Hennig, J.; von Elverfeldt, D. J. Magn. Reson. 2011, 208, 20−26. (7) Massin, C.; Boero, G.; Vincent, F.; Abenhaim, J.; Besse, P.; Popovic, R. Sens. Actuators, A 2002, 97, 280−288. (8) Massin, C.; Vincent, F.; Homsy, A.; Ehrmann, K.; Boero, G.; Besse, P.; Daridon, A.; Verpoorte, E.; de Rooij, N.; Popovic, R. J. Magn. Reson. 2003, 164, 242−255. (9) Wensink, H.; Benito-Lopez, F.; Hermes, D.; Verboom, W.; Gardeniers, H.; Reinhoudt, D.; Berg, A. Lab Chip 2005, 5, 280−284. (10) Ehrmann, K.; Saillen, N.; Vincent, F.; Stettler, M.; Jordan, M.; Wurm, F.; Besse, P.; Popovic, R. Lab Chip 2007, 7, 373−380. (11) Bart, J.; Janssen, J. W. G.; van Bentum, P. J. M.; Kentgens, A. P. M; Gardeniers, J. G. E. J. Magn. Reson. 2009, 201, 175−185. (12) Bart, J.; Kolkman, A. J.; Oosthoek-de Vries, A. J.; Koch, K.; Nieuwland, P. J.; Janssen, H. J. W. G.; van Bentum, P. J. M.; Ampt, K. A. M.; Rutjes, F. P. J. T.; Wijmenga, S. S.; Gardeniers, H. J. G. E.; Kentgens, A. P. M J. Am. Chem. Soc. 2009, 131, 5014. (13) van Bentum, P.; Janssen, J.; Kentgens, A.; Bart, J.; Gardeniers, J. J. Magn. Reson. 2007, 189, 104−113. (14) Gruschke, O.; Baxan, N.; Clad, L.; Kratt, K.; von Elverfeldt, D.; Peter, A.; Hennig, J.; Badilita, V.; Wallrabe, U.; Korvink, J. G. Lab Chip 2012, 12, 495−502. (15) Krojanski, H.; Lambert, J.; Gerikalan, Y.; Suter, D.; Hergenröder, R. Anal. Chem. 2008, 80, 8668. (16) Utz, M.; Monazami, R. J. Magn. Reson. 2009, 198, 132−136. (17) Jacquinot, J.-F.; Sakellariou, D. Concept. Magn. Reson., A 2011, 38A, 33−51. (18) Hoult, D.; Richards, R. Proc. R. Soc. London, A 1975, 344, 311− 340. (19) Hoult, D.; Richards, R. J. Magn. Reson. 1976, 24, 71−85. (20) Sidles, J.; Rugar, D. Phys. Rev. Lett. 1993, 70, 3506−3509. (21) Hoult, D. I. In Encyclopedia of NMR; Grant, D., Harris, R., Eds.; J. Wiley & Sons: Chichester, U.K., 1996; pp 4256−4266. (22) Hoult, D. J. Magn. Reson. Imaging 2000, 12, 46−67. (23) Wu, N.; Peck, T.; Webb, A.; Magin, R.; Sweedler, J. J. Am. Chem. Soc. 1994, 116, 7929−7930. (24) Olson, D.; Peck, T.; Webb, A.; Magin, R.; Sweedler, J. Science 1995, 270, 1967−1970. (25) Fratila, R. M.; Velders, A. H. Annu. Rev. Anal. Chem. 2011, 4, 227−249. (26) Subramanian, R.; Kelley, W.; Floyd, P.; Tan, Z.; Webb, A.; Sweedler, J. Anal. Chem. 1996, 68, 4431−4435. (27) Wolters, A.; Jayawickrama, D.; Webb, A.; Sweedler, J. Chem. Rev. 1999, 99, 3133−3152. (28) Schewitz, J.; Pusecker, K.; Gfrörer, P.; Götz, U.; Tseng, L.; Albert, K.; Bayer, E. Chromatographia 1999, 50, 333−337. (29) Trumbull, J.; Glasgow, I.; Beebe, D.; Magin, R. IEEE Trans. Biomed. Eng. 2000, 47, 3−7. (30) Telkki, V.-V.; Zhivonitko, V. V.; Ahola, S.; Kovtunov, K. V.; Jokisaari, J.; Koptyug, I. V. Angew. Chem., Int. Ed. 2010, 49, 8363− 8366.
Figure 9. COSY spectrum of 1.2 μL of 500 mM glucose in D2O with water suppression by presaturation. A total of 256 t1 increments were acquired over a total experiment time of 4 h. The data was extended to 512 t1 points by linear prediction with 64 coefficients.
coupling mode allows for easy exchange of the sample. While the demonstration experiments reported here have used very simple fluidic geometries, this opens the possibility to equip more complex lab-on-a-chip devices with planar resonators. The manufacturing cost, especially for the laminated foil devices, is sufficiently low to point the way toward dispose-andrecycle NMR modalities of use. The spectral resolution of 7.5 ppb (4.5 Hz at 600 MHz proton frequency) is limited by the susceptibility mismatch between the fluidic chip materials and the sample. Currently, this requires avoidance of edges perpendicular to the magnetic field near the sample volume; using a round sample chamber with a diameter similar to the microcoil leads to a severely distorted line shape (cf. Supporting Information). This is a severe restriction in the design of NMR LOC devices. Efforts toward eliminating this constraint are currently underway in our group and will be reported at a later occasion.
■
ASSOCIATED CONTENT
* Supporting Information S
Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.
■
REFERENCES
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Present Address ⊥
MicroPelt GmbH, Freiburg, Germany.
Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS This work has been supported by the U.S. National Science Foundation under Grant Number CHE-0809795 and by the National Institute of Health under Award Number 3701
dx.doi.org/10.1021/ac300204z | Anal. Chem. 2012, 84, 3696−3702
Analytical Chemistry
Article
(31) Telkki, V.-V.; Zhivonitko, V. V. J. Magn. Reson. 2011, 210, 238− 245. (32) McDonnell, E.; Han, S.; Hilty, C.; Pierce, K.; Pines, A. Anal. Chem. 2005, 77, 8109−8114. (33) Harel, E.; Hilty, C.; Koen, K.; McDonnell, E. E.; Pines, A. Phys. Rev. Lett. 2007, 98, 017601. (34) Hoult, D. I.; Tomanek, B. Concept. Magn. Reson. 2002, 15, 262− 285. (35) Bilgen, M. BioMed. Eng. OnLine 2006, 5, 3. (36) Sakellariou, D.; Le Goff, G.; Jacquinot, J. Nature 2007, 447, 694−697. (37) Gurst, J. J. Chem. Educ. 1991, 68, 1003. (38) Lacey, M. E.; Subramanian, R.; Olson, D. L.; Webb, A. G.; Sweedler, J. V. Chem. Rev. 1999, 99, 3133−3152.
3702
dx.doi.org/10.1021/ac300204z | Anal. Chem. 2012, 84, 3696−3702
