Polymer Nanochannels Fabricated by Thermomechanical

UV-ablation nanochannels in micro/nanofluidics devices for biochemical analysis. Chen Wang , Jun Ouyang , Hong-Li Gao , Heng-Wu Chen , Jing-Juan Xu ...
0 downloads 0 Views 474KB Size
Anal. Chem. 2005, 77, 2252-2258

Polymer Nanochannels Fabricated by Thermomechanical Deformation for Single-Molecule Analysis Ponniah Sivanesan,† Kenji Okamoto,‡ Douglas English,‡ Cheng S. Lee,‡ and Don L. DeVoe*,§

Calibrant Biosystems, 7507 Standish Place, Rockville, Maryland 20855, and Department of Mechanical Engineering, and Department of Chemistry and Biochemistry, University of Maryland, College Park, Maryland 20742

A simple method for fabricating nanoscale channels based on thermomechanical deformation of rigid polymer substrates is demonstrated. Polycarbonate preforms containing microchannels with cross-sectional dimensions on the order of tens of micrometers are controllably deformed to produce submicrometer dimensions. The reduced channel dimensions are achieved by heating the preform while applying a uniaxial tensile force to reduce channel cross sections through the Poisson effect. Nanochannels with circular or elliptical cross sections are defined by varying the channel position and preform geometry prior to deformation. Arrays of parallel nanochannels with critical dimensions down to 400 nm are described. Using the fabrication method, a nanochannel network is fabricated for the detection of single protein molecules via confocal fluorescence microscopy. The chip includes a detection channel with cross-sectional dimensions approaching the confocal volume dimensions of the detection optics and a larger adjacent reference channel used to optimize focusing. Detection of fluorescently labeled bovine serum albumin at 15 and 150 nM concentrations is presented, demonstrating the ability to perform singlemolecule fluorescence measurements in polycarbonate chips using visible wavelengths for excitation and detection. A critical issue in single-molecule analysis is the ability to spatially constrain molecules of interest within a nanoscale confinement zone. Spatial localization serves to reduce Brownian motion of low molecular weight analytes and to provide a fixed location for interrogation of the molecule of interest.1 A variety of approaches to this goal have been reported, including the use of suspended microdroplets2 and trapping molecules within nanoscale porous gels.3 More recently, confining molecules within sub* Corresponsding author. Phone: 301-405-8125. Fax: 301-403-9477. E-mail: [email protected]. † Calibrant Biosystems. ‡ Department of Mechanical Engineering, University of Maryland. § Department of Chemistry and Biochemistry, University of Maryland. (1) Lyon, W.; Nie, S. Anal. Chem. 1997, 69, 3400. (2) Barnes, M. D.; Ng, K. C.; Whittenm, W. B.; Ramsey, J. M. Anal. Chem. 1993, 65, 2360. (3) Dickson, R. M.; Norris, D. J.; Tzeng, Y.-L.; Moerner, W. E. Science 1996, 274, 966.

2252 Analytical Chemistry, Vol. 77, No. 7, April 1, 2005

micrometer silica capillaries has received attention.1,4-6 For example, in an early demonstration by Nie and Lyon, submicrometer detection channels fabricated from silica capillaries were used to extend the observation time of single molecules.1 The nanoscale capillaries were fabricated by using a CO2 laser to locally heat a silica capillary while pulling along the capillary length, thus necking down the detection channel, with final inner diameters of ∼500 nm. However, capillary nanochannels cannot readily support the fabrication of multiple nanofluidic elements or the integration of nanochannels with larger microfluidic networks capable of providing system-level functionality such as analyte separation, reagent and analyte delivery, and integrated sensing. Thus, to avoid the limitations of capillary nanofluidics, lab-on-a-chip technology using nanochannels fabricated on planar substrates has been widely investigated as an alternative analytical platform. In one such approach, molecular confinement has been achieved within microscale channels by focusing a stream of fluid containing the analyte. In a method described by Mathies and Haab,7 hydrodynamic8 or electrostatic focusing9 was used to concentrate an analyte stream into the center of a larger microchannel fabricated in a planar glass substrate to measure fluorescent bursts from single DNA molecules. However, because increased focusing leads to higher velocity of the analyte molecules, the total residence time within the detection window is reduced using this method, leading to a reduced signal-to-noise ratio as discussed by Mathies and Peck.10 In addition, focusing is only performed in the lateral dimension, so that a mismatch between confocal optical detection volume and channel height remains a potential limitation for typical microchannel dimensions. Several approaches have focused on the use of traditional topdown methods to create nanoscale channels with submicrometer cross-sectional dimensions. For example, nanochannel fabrication based on the removal of ultrathin lithographically patterned (4) Lee, Y. H.; Maus, R. G.; Smith, B. W.; Winefordner, J. D. Anal. Chem. 1994, 66, 4142. (5) Guenard, R. D.; King, L. A.; Smith, B. W.; Winefordner, J. D. Anal. Chem. 1997, 69, 2426. (6) Zander, C.; Drexhage, K. H.; Han, K.-T.; Wolfrum, J.; Sauer, M. Chem. Phys. Lett. 1998, 286, 457. (7) Haab, B.; Mathies, R. Anal. Chem. 1999, 71, 5137. (8) Nguyen, D. C.; Keller, R. A.; Jett, J. H.; Martin, J. C. Anal. Chem. 1987, 59, 2158. (9) Seiler, K.; Harrison, J. D.; Manz, A. Anal. Chem. 1993, 65, 1481. (10) Mathies, R. A.; Peck, K. Anal. Chem. 1990, 62, 1786-1791. 10.1021/ac048923q CCC: $30.25

© 2005 American Chemical Society Published on Web 02/10/2005

sacrificial films encapsulated by polymer11 or silicon-based12-18 materials has been widely demonstrated. Alternately, nanochannel fabrication via bulk micromachining has been reported using photolithography on silicon19,20 or e-beam lithography on SiO221 followed by a shallow substrate etch and capping layer bond. Other approaches take advantage of nonconformal deposition processes to form enclosed nanochannels using materials such as parylene22 or SiO2.23 Other methods include hot embossing of thermoplastics with nanoscale channel features,24 nanoimprinting into thin polymer films,25 using carbon nanotubes as sacrificial templates for forming SiO2 nanochannels,26 forming sacrificial polycarbonate nanofibers by electrofilament deposition followed by SiO2 deposition to create the nanochannel walls,27 employing nanoporous membranes to couple microfluidic channels,28,29 and manipulation of surfaceimmobilized lipid vesicles to form nanotubes ranging from 100 to 300 nm in diameter connecting discrete vesicles within the network.30 In the present work, a new concept for nanochannel fabrication based on thermomechanical deformation of a polymer preform containing channels with relatively large (tens of micrometers) critical dimensions is described. By deforming the preform above its glass transition temperature using a uniaxial load oriented parallel to the channel length, the cross-sectional dimensions may be controllably reduced by several orders of magnitude through the Poisson effect. Unlike previously established methods, this nanofabrication technique offers several unique advantages. Because the initial substrate contains relatively large features, standard lithography with low-resolution optical masks may be used for preform fabrication. Furthermore, since only a portion (11) Eijkel, J. C. T.; Bomer, J.; Tas, N. R.; van den Berg, A. Lab Chip 2004, 3, 161. (12) Tas, N. R.; Mela, P.; Kramer, T.; Berenschot, J. W.; van den Berg, A. Nano Lett. 2003, 3, 1537. (13) Foquet, M.; Korlach, J.; Zipfel, W. R.; Webb, W. W.; Craighead, H. G. Anal. Chem. 2004, 76, 1618. (14) Nam, W. J.; Bae, S.; Kalkan, A. K.; Fonash, S. J. J. Vac. Sci. Technol., A 2001, 19, 1229. (15) Stern, M. B.; Geis, M. W.; Curtin, J. E. J. Vac. Sci. Technol., B 1997, 15, 2887. (16) Harnett, C. K.; Coates, G. W.; Craighead, H. G., J. Vac. Sci. Technol., B 2001, 19, 2842. (17) Foquet, M.; Korlach, J.; Zipfel, W.; Webb, W. W.; Craighead, H. G. Anal. Chem. 2002, 74, 1415. (18) Tas, N. R.; Berenschot, J. W.; Mela, P.; Jansen, H. V.; Elwenspoek, M.; van den Berg, A. Nano Lett. 2002, 2, 1031. (19) Han, J.; Craighead, H. G. Science 2000, 288, 1026. (20) Haneveld, J.; Jansen, H.; Berenschot, E.; Tas, N.; Elwenspoek, M. J. Micromech. Microeng. 2003, 13, S62. (21) Hibara, A.; Saito, T.; Kim, H.-B.; Tokeshi, M.; Ooi, T.; Nakao, M.; Kitamori, T. Anal. Chem. 2002, 74, 6170. (22) Ilic, B.; Czaplewski, D.; Zalalutidinov, M.; Schmidt, B.; Craighead, H. G. J. Vac. Sci. Technol., B 2002, 20, 2459. (23) Cao, H.; Yu, Z.; Wang, J.; Tegenfeldt, J. O.; Austin, R.; Chen, E.; Wu, W.; Chou, S. Appl. Phys. Lett. 2002, 81, 174. (24) Studer, V.; Pepin, A.; Chen, Y. Appl. Phys. Lett. 2002, 80, 3614. (25) Guo, L. J.; Cheng, X.; Chou, C.-F. Nano Lett. 2004, 4, 69. (26) Melechko, A. V.; McKnight, T. E.; Guillorn, M. A.; Merkulov, V. I.; Ilic, B.; Doktycz, M. J.; Lowndes, D. H.; Simpson, M. L. Appl. Phys. Lett. 2003, 82, 976. (27) Czaplewski, D. A.; Kameoka, J.; Mathers, R.; Coates, G. W.; Craighead, H. G. App. Phys. Lett. 2003, 83, 4836. (28) Kuo, T. C.; Kim, H. K.; Cannon, D. M., Jr; Shannon, M. A.; Sweedler, J. V.; Bohn, P. W. Angew. Chem., Int. Ed. 2004, 43, 1862-1864. (29) Zhang, Y.; Timperman, A. T. Analyst 2003, 128, 537-542. (30) Karlsson, A.; Karlsson, M.; Karlsson, R.; Sott, K.; Lundqvist, A.; Tokarz, M.; Orwar, O. Anal. Chem. 2003, 75, 2529.

of the initial substrate is deformed, the final device can contain both microscale and nanoscale channels in the same chip. This allows the fabrication of microscale features for system-level functionality such as fluid delivery, with a minimal dead-volume interface to the nanochannels, and with smooth and continuous transitions between microscale and nanoscale channels. The deformation process results in nanochannels with elliptical or circular cross sections, which can improve the uniformity of flow and electric field distributions while also reducing flow resistance compared with channel cross sections containing sharp corners.27 Nanochannel geometry may be controlled in a robust manner, and channels with widely varying cross sections may be prepared on the same chip by choosing different initial channel widths and locations relative to the preform edge. The process requires minimal fabrication equipment and is exceptionally high-throughput and cost-effective. While the process is demonstrated here using polycarbonate as the substrate material, it is applicable for a wide range of rigid plastics, potentially enabling the fabrication of nanochannels that can benefit from a wealth of polymer chemistry options. EXPERIMENTAL SECTION Fabrication. The thermomechanical deformation process is derived from the well-known optical fiber draw method, in which the bottom of a cylindrical preform of fused silica is heated above its glass transition temperature, while a uniform force is applied to pull an initial preform diameter, typically ∼2 cm, to a final diameter of ∼100 µm in a distance of a few centimeters. Thus, as the heated region lengthens, preform cross-sectional dimensions are proportionally reduced through the Poisson effect. Flow in the neck-down region is well-layered and highly controllable because of the high viscosity of the silica. The fiber draw process has been thoroughly characterized due to the demands of the telecommunications industry, and a number of detailed models of the process have been developed.31-33 The same process is also used for fabricating silica capillaries with inner diameters as small as 100 nm.34 In this work, the fiber drawing process has been extended from cylindrical fused silica to rectangular polycarbonate (PC) preforms, where the preforms contain one or more microchannels. Polycarbonate was chosen as the substrate material for several reasons. With a glass transition temperature (Tg) of ∼150 °C, thermal requirements for pulling PC substrates are substantially lower than fused silica with a Tg of 1200 °C. Compared to other thermoplastics such as poly(methyl methacrylate) or cyclic olefin copolymer, which typically exhibit favorably lower autofluorescence than PC, we have found experimentally that PC is particularly easy to deform with excellent yield. Microchannels were fabricated by hot embossing into a PC substrate from features patterned in a silicon wafer template by bulk anisotropic etching. Sealing of the microchannels was achieved by thermally bonding the microchannel substrate to a second PC layer into which fluidic access holes were mechanically milled. Before pulling, the (31) Gupta, G.; Arruda, E. M.; Liu, X. Rheol. Acta 1996, 35, 584. (32) Gupta, G. J. Nonlinear Mech. 1997, 33, 151. (33) Pouzirjev, V. A.; Rashidov, J. R; Ashurbekov, R. K. Proc. Sci. Technol. Conf. Inf. Meas. Syst. 1987, 61. (34) Beloglasov, V. I.; Soukhoveev, S. P.; Suetin, N. V. Micro Syst. Tech. 2000, 1, 6.

Analytical Chemistry, Vol. 77, No. 7, April 1, 2005

2253

Figure 1. Schematic of the nanochannel draw tower apparatus.

preforms were mechanically reshaped to define the desired width of the heated region, with preform dimensions chosen based on the required nanochannel dimension, the original microchannel dimensions, and the physical location of microchannels in the preform. To shape the final nanochannels, the PC preforms were placed in a custom draw tower shown schematically in Figure 1. The tower employs two nickel-chromium wires as resistive radiant heater elements, positioned on either side of the substrate for uniform temperature control. Two single-axis linear motors are used to define the pull distance, with the top and bottom ends of the preform pulled simultaneously to keep the center of the preform positioned between the heaters. Springs are positioned between the motor stages and preform clamps. The springs serve to maintain a linearly decreasing pulling force over the pull distance, ensuring that as the preform cross section reduces during the pulling process, the applied force does not exceed the maximum tensile strength of the preform. Preforms are oriented such that all channels in the neck-down region are parallel to the draw force. As the heated region of the preform reaches the glass transition temperature of PC, the microchannels lengthen while the cross-sectional dimensions are reduced. Power to the heaters is switched off as soon as the preform reaches its maximum pull length defined by the selected motor positions. The process is performed vertically to prevent gravity-induced deformation of the nanochannel region. Single-Molecule Measurements. To test the viability of single-molecule detection in thermomechanically fabricated nanochannels, the detection of single bovine serum albumin (BSA) molecules was demonstrated in a nanofluidic network consisting of adjacent detection and reference channels. Both channels were 2254

Analytical Chemistry, Vol. 77, No. 7, April 1, 2005

prefilled with 2-propanol (IPA). The reference channel was next filled with a high concentration solution of 40-nm-diameter fluorescent labeled polystyrene beads (FluoSpheres, 505-nm excitation and 515-nm emission, Molecular Probes) suspended in deionized water with 1% v/v Triton X-100 surfactant as a suppressing agent for dye adsorption on the channel walls. The fluorescent beads in the reference channel act as a bright marker allowing the detection channel position to be readily visualized. After preparing the reference channel, the detection channel was then filled with a solution containing BSA prelabeled with Alexa Fluor dye (497-nm excitation and 520-nm emission, Molecular Probes). Measurements using two BSA concentrations were performed using different buffer solutions. In the first case, the buffer was 90% v/v 2-propanol in 2 mM Tris-HCl and the protein concentration was 15 nM. In the second case, the protein concentration was 150 nM and the buffer was 20 mM Tris-HCl supplemented with 2% v/v Triton X-100 surfactant. All filling was performed through nanoports (Upchurch Scientific) using high-pressure syringe pumps (PHD2000, Harvard Apparatus). A constant flow rate of 0.05 µL/min was employed for prefilling all solutions as well as for transient measurements. Measurements were performed on an inverted microscope (Carl Zeiss Axiovert 200) modified for sample-scanning laser confocal imaging. An argon ion laser (Melles Griot) at 488 nm was used for imaging and single-molecule detection. For singlemolecule experiments, the excitation power was 18-20 mW and the laser beam was focused on the channel through a dichroic mirror using a LD Plan-Neofluar 63x/0.75 Korr objective lens (Carl-Zeiss). The Gaussian width of the laser spot is estimated to be ∼600 nm based on images from immobilized fluorescent beads. Fluorescence light was collected via the same objective lens. The collected light is sent through the dichroic mirror to an avalanche photodiode (APD, Perkin-Elmer, SPCM-AQR). A Raman notch filter (Kaiser Optical) and a long-pass filter (XF3080, Omega Optical) were used to reduce noise due to scattered excitation light and autofluorescence from the polycarbonate substrate. Single-photon counts from the APD were buffered and sent to a computer for analysis. RESULTS AND DISCUSSION Nanochannel Characterization. A typical nanofluidic chip fabricated using the thermomechanical deformation method is shown in Figure 2. This particular chip contains two 700-nmdiameter circular channels drawn down from initial preform channels with trapezoidal cross sections of 30-µm depth and 20-µm width at half-depth. The microchannels were fabricated in 3.2-mm-thick PC with a similar bonding plate, resulting in a preform with a total thickness of 6.4 mm. Prior to thermomechanical deformation, the preform was shaped using a milling machine to reduce the heated region width to 10 mm. Figure 3 shows SEM images of one of the nanochannels after cleaving the chip. Cleaving was performed by first dipping the chip in liquid nitrogen to ensure a clean break line and minimal distortion of the channel cross section. As seen in Figure 3, the thermally pulled nanochannels exhibit circular or elliptical cross sections rather than the original trapezoidal shape resulting from the bulk etched silicon templates used to form the initial microchannels due to the nonuniform stress field imposed during the pulling

Figure 2. Typical fabricated nanofluidic chip.

Figure 3. (a) Far-field and (b) high magnification electron micrographs showing a single nanochannel with a circular cross section 700 nm in diameter.

process. Final channel dimensions are repeatable, with less than 5% deviation in cross-sectional dimensions observed during multiple pulling tests on preforms with identical initial geometry. The effects of several fabrication parameters on final nanochannel dimensions were evaluated experimentally. The relationship between pull length and overall preform dimensions was first explored. Figure 4 shows the measured percentage change in preform width and thickness as a function of pulled distance when using preforms with different initial cross-sectional dimensions, namely (a) 14.5 mm × 5.84 mm, and (b) 17.5 mm × 1.54. In both cases, the preform width changes quickly at the onset of pulling, but as the pull distance reaches approximately twice the preform thickness, the change in width becomes significantly less rapid. In contrast, the thickness changes approximately linearly with the pull length within the tested range of 0-3 cm. The effect of initial pulling force, defined by the product of the spring constant chosen for the extension springs connecting the preform to the draw tower and the set pull length, was also

considered. Although higher forces were found to result in higher strain rates during the pulling process, no effect on the final preform or nanochannel dimensions was observed over a wide range of initial preform geometries and spring constants. Because the preforms do not possess cylindrical symmetry like traditional fiber preforms, the stress and strain fields during the pulling process are not uniform across the preform. Hence, the location of the original microchannel within the preform also affects the final dimensions of the resulting nanochannels. Limited modeling of temperature distribution35 and mechanical deformation36 in the drawing of glass sheets from rectangular preforms has been reported, but detailed analysis of the process remains a complex and unsolved thermomechanics problem. While a detailed model of the deformation process is beyond the scope of this investigation, several qualitative observations are worth noting here. First, channel width tends to experience the greatest (35) Booth, F.; Bourne, D. E.; Harrison, M. C. Glass Technol. 1972, 13, 22. (36) Milutinovic-Nikolic, A.; Jancic, R.; Aleksic, R. Glass Technol. 1998, 39, 166.

Analytical Chemistry, Vol. 77, No. 7, April 1, 2005

2255

Figure 4. Percentage change in width and thickness vs pulled distance for preforms with initial cross-sectional dimensions of (a) 14.5 mm × 5.84 mm and (b) 17.5 mm × 1.54 mm.

Figure 6. (a) Optical image of 30 parallel nanochannels and (b) electron micrograph showing two adjacent channels.

Figure 5. Percent reduction in channel width as a function of channel position within a 6.4-mm-thick preform following a pull distance of 28 mm (10.6 mm wide before pull and 1.92 mm wide after pull).

reduction during the pulling process for channels located near the edges of a preform, while channels at the preform center exhibit the smallest reduction. This effect is displayed in Figure 5 for a 10.6-mm-wide and 6.4-mm-thick preform following a 28-mm-long pull distance. As shown in the figure, channels located at the center of the preform experience a 75% reduction in width, while channels 100 µm from the preform edge exhibit a 95% reduction, following a nearly parabolic trend. Note that the measured channel positions are unevenly spaced and nonsymmetric about the center line since only a limited number of 2256 Analytical Chemistry, Vol. 77, No. 7, April 1, 2005

channels were measured from the final chip. The difference in width reduction is most noticeable for preforms with a large ratio of width/thickness, while preforms that have been shaped to a square heated region exhibit minimal variation in final channel dimensions. In contrast, the maximum reduction in channel height occurs for channels located near the preform center. In general, channels at the preform edge maintain their aspect ratio during the pulling process, so that microchannels with the same initial width and depth result in circular nanochannels, while channels at the preform center exhibit elliptical cross sections with major axes oriented parallel to the preform width. As the preform thickness is increased, channels at the center of the preform become increasingly circular. Figure 6a shows an optical image of 30 parallel nanochannels fabricated in a 9-mm-wide and 6.4-mm-thick preform with a 33-mm pull distance. Final channel dimensions range from approximately 400 nm high and 1.8 µm wide for channels near the preform center to 800 nm high and 800 nm wide at the preform edge. The SEM image of two adjacent channels in Figure 6b reveals this trend, with the channel closer to the center (left) wider and shallower than the channel closer to the edge (right). For a given preform geometry, the desired

Figure 7. Confocal fluorescence image showing wide reference channel filled with fluorescent labeled polystyrene beads (left) and narrow detection channel filled with Tris buffer (right).

nanochannel cross section may be achieved by selecting an appropriate initial microchannel geometry and a specific location relative to the shaped preform edge. Detection of Single Protein Molecules. A key goal of this study is to demonstrate the viability of single-molecule detection of proteins in thermomechanically formed nanofluidic channels produced from inexpensive polycarbonate substrates. Previous work by Soper and co-workers37 demonstrated the efficient detection of single molecules in polycarbonate microchannels

using near-infrared excitation and detection to reduce the high autofluorescence at visible wavelengths. In contrast, the present work considers single-molecule detection in polycarbonate substrates using conventional fluorophores excited at visible wavelengths. One noteworthy aspect of the detection platform that enables good detection performance is a substrate thickness that is significantly reduced along with channel dimensions during the pulling process, so that less polycarbonate lies within the optical detection path, resulting in a reduction in autofluorescence from the substrate and an improved signal-to-background ratio. Another aspect is the use of a long working distance objective with a correction collar, combined with a reference channel used to both locate and optimize focusing within the detection channel. Imaging of the reference channel positioned immediately adjacent to the detection channel allowed optimization of the objective’s correction collar setting to achieve the best image quality as judged by image sharpness and signal-to-background ratio. This approach provides accurate correction for the effective thickness and index of refraction of the polycarbonate substrate and ensures optimal focusing for the detection channel since both channels appear in the same vertical image plane of the microscope. Note that optimal focusing is very difficult without the large reference channel, since the smaller detection channel is virtually invisible during imaging. A fluorescence image of the detection channel (700 nm tall and 1.8 µm wide) used for single BSA molecule measurements is shown in Figure 7. While the channel height is well matched to the optical detection volume, the width of the detection channel is ∼3× larger than the optical spot diameter. While this slightly larger channel prevents complete detection of all molecules passing the detection point, it was used in this initial study to

Figure 8. Fluorescent emission measurements for Alexa Fluor-conjugated BSA in (a) IPA and (b) Tris buffer solutions.

Analytical Chemistry, Vol. 77, No. 7, April 1, 2005

2257

simplify alignment of the detection optics and reduce the pressure needed for filling the channel with buffer solution, while demonstrating the feasibility of single-molecule measurements in thermomechanically deformed polycarbonate chips. In addition to the detection channel, a parallel reference channel (700 nm tall and 15 µm wide) located ∼30 µm from the detection channel was filled with a high concentration of fluorescent beads. After optimizing the optics using the bright reference channel, single-molecule detection was achieved by positioning the focused laser spot on the center of the detection channel and monitoring detected photons as a function of time. Typical transient measurements acquired in this way are shown in Figure 8 for BSA in IPA and in Tris buffer. Transient photon bursts appearing above the constant background signal are absent in pure solvents and compare well with the expected burst intensities for a diffusing single molecule.37 The acquisition time for each point in Figure 8 is 0.1 ms. In both cases, the average burst width was measured to be 0.22 ( 0.13 ms, providing further evidence that the measured bursts result from single molecules traversing the detection region. The variation in peak height in Figure 8 is expected, due to molecules passing through the detection region along longitudinal paths with different radial positions, resulting in different excitation energy and photon capture efficiency. In all experiments, the burst intensities were significantly greater for BSA in Tris buffer relative to BSA in IPA, though the pump speed and laser intensities were kept constant. To explore the source of this difference, we measured the relative quantum yield for Alexa Fluor-labeled BSA in the two solvent environments. While the shape and spectral maximum of the resulting absorption spectra were found to be nearly identical in both IPA and Tris buffer, the fluorescence efficiency (quantum yield) was 3.4 times higher in Tris buffer than in IPA, thus accounting for the difference in peak heights. Additional burst height variation may arise from minor variation in laser power, focusing, or surface interactions between BSA and the nanochannel walls for the two solutions, but these effects are minimal. However, it should be noted that the single-molecule observation rates shown in Figure 8 are as much as 1 order of magnitude lower than expected for the flow rates and BSA concentrations used. There are several likely contributions to this observation. First, the flow rates were imposed using a syringe pump, but the actual flow rates may be significantly lower due to the high flow resistance presented by (37) Wabuyele, M. B.; Ford, S. M.; Stryjewski, W.; Barrow, J.; Soper, A. Electrophoresis 2001, 22, 3939-3948.

2258

Analytical Chemistry, Vol. 77, No. 7, April 1, 2005

the nanochannels. It is also likely that actual BSA concentrations were lower than the reported concentrations due to protein absorption during sample preparation, injection, and manipulation in the nanofluidic system. CONCLUSIONS The presented thermomechanical deformation process for nanochannel fabrication allows the use of low-cost microscale patterning methods combined with a simple thermal pulling step to achieve channels with submicrometer geometry. While nanochannels with critical dimensions as small as 400 nm have been successfully fabricated using preforms with initial channels with minimum features on the order of 20 µm, significantly smaller channels are feasible by reducing the preform channel dimensions. In addition to simplicity and repeatability, a key advantage of the fabrication method lies in the ability to seamlessly integrate microscale fluidic systems, which are self-aligned to the nanochannels, with no discontinuities between the elements. While the thermomechanical deformation process was demonstrated in a single-pull system, the concept may be extended for deforming a microfluidic substrate in multiple locations along one or more axes, enabling the fabrication of individual nanochannel arrays coupled together by microscale fluidic elements. Unlike direct patterning methods previously used for nanochannel fabrication, the thermomechanical deformation process is somewhat limited in the range of channel geometries which can be achieved. For example, all nanochannels must be straight, and multiple channels formed in a single pulling step must be parallel and of equal length. Furthermore, the nanochannel length is dependent on the cross-sectional geometry of the preform, although the preform may be mechanically shaped prior to pulling to achieve a wide range of final channel lengths. Despite these limitations, the process provides unique features which should prove useful for a range of lab-on-a-chip applications where singlemolecule detection coupled with system-level functionality via microfluidics is desired. ACKNOWLEDGMENT The authors gratefully acknowledge the NIH National Institute of Biomedical Imaging and Bioengineering for providing funding for this research under Grant 1 R43 EB00453. Received for review July 23, 2004. Accepted January 5, 2005. AC048923Q