Anal.Chem. 1996, 68, 1040-1046
`
`Microfabrication of a Planar Absorbance and
`Fluorescence Cell for Integrated Capillary
`Electrophoresis Devices
`
`Zhenhua Liang, Nghia Chiem, Gregor Ocvirk, Thompson Tang, Karl Fluri, and D. Jed Harrison*
`
`DepartmentofChemistry,UniversityofAlberta,Edmonton,Alberta,CanadaT6G2G2
`
`A micromachined absorbance and fluorescence detection
`cell for application to capillary electrophoresis within
`planar glass substrates (chips) is described. A microfab-
`ricated U-cell for absorbance provides a longitudinal path
`120-140 (cid:237)m long parallel to the flow direction and gives
`at least a 10-fold increase in absorbance compared to an
`absorbance path transverse to the flow direction. Absor-
`bance detection limits of 0.003 AU gave (cid:24)6 (cid:237)M detection
`limits for hydrolyzed fluorescein isothiocyanate dye. The
`same device can be used for longitudinal fluorescence
`excitation with a 20-fold improvement
`in signal-to-
`background levels due to reduced scattering, utilizing a
`form of sheath flow. Fluorescence detection limits of
`(cid:24)20 000 molecules and 3 nM were obtained for fluores-
`cein.
`
`Obtaining good limits of detection when using optical detection
`has been a challenge in capillary electrophoresis (CE), due to the
`short path lengths engendered by the small capillary diameter.1-3
`Fluorescence detection has proven to be very effective.4-6 Nev-
`ertheless, absorbance detection remains more generally accepted
`due to its wider applicability. Several methods, including the use
`of Z-cells,3,7,8 as in liquid chromatography detection,9 or widening
`of the capillary at the detection point, can improve detection limits
`by increasing the optical path length. With micromachined CE
`((cid:237)-CE) devices etched on planar glass plates, the path length
`problem is equally severe, so to date, fluorescence detection has
`been the principal optical detection method employed.10-14
`
`(1) Walbroehl, Y.; Jorgenson, J. W. J. Chromatogr. 1984, 315, 135-143.
`(2) Bruno, A. E.; Gassmann, E.; Pericles, N.; Anton, K. Anal. Chem. 1989, 61,
`876-883.
`(3) Bruin, G. J. M.; Stegeman, G.; Van Austen, A. C.; Xu, X.; Kraak, J. C.; Poppe,
`H. J. Chromatogr. 1991, 559, 163-181.
`(4) Green, J. S.; Jorgenson, J. W. J. Chromatogr. 1986, 352, 337-343.
`(5) Zare, R. N.; Gassmann, E. Eur. Pat. Appl. EP 21660, 1987.
`(6) Cheng, Y.-F.; Dovichi, N. J. Science 1988, 242, 562-564.
`(7) Chervet, J. P.; Van Soest, R. E. J.; Ursem, M. J. Chromatogr. 1991, 543,
`439-449.
`(8) Moring, S. E.; Reel, R. T.; Van Soest, R. E. J. Anal. Chem. 1993, 65, 3454-
`3459.
`(9) Stevenson, R. L. In Liquid Chromatography Detectors; Vickrey, T. M., Ed.;
`Chromatographic Science Series 23; Dekker: New York, 1983; Chapter 2,
`pp 23-86.
`(10) Harrison, D. J.; Manz, A.; Fan, Z.; Lu¨di, H.; Widmer, H. M. Anal. Chem.
`1992, 64, 1926-1932.
`(11) Harrison, D. J.; Fluri, K.; Seiler, K.; Fan, Z.; Effenhauser, C. S.; Manz, A.
`Science 1993, 261, 895-897.
`(12) Effenhauser, C. S.; Manz, A.; Widmer, H. M. Anal. Chem. 1993, 65, 2637-
`2642.
`(13) Raymond, D. E.; Manz, A.; Widmer, H. M. Anal. Chem. 1994, 66, 2858-
`2865.
`
`Micromachining allows complex capillary geometries to be
`formed so that optical components may be integrated into a planar
`In this way, micromachined Z-cells can be fabricated,
`device.10
`as was demonstrated by Verpoorte et al.15 in the realization of an
`absorbance Z-cell for liquid chromatography detection. That
`device used the crystal planes of silicon(111) to form mirror planes
`to reflect a beam from above the planar device into the capillary.
`CE devices are not readily fabricated in silicon due to the high
`voltages usually employed and the conductivity of Si.16 Unfortu-
`nately, mirror planes are not easily formed in amorphous glass
`since it etches isotropically,17 so an alternate strategy to launch
`and collect light is required.
`In this report, we describe the fabrication of a planar optical
`U-type cell in glass and its application for both fluorescence and
`absorbance detection. The cell provides a 10-fold improvement
`in absorbance detection limits by probing the capillary along the
`longitudinal rather than the transverse direction. It also enhances
`fluorescence detection limits, since the fluorescent signal may be
`detected well away from the point at which the light enters the
`capillary. This dramatically reduces the amount of scattered light
`collected, resulting in an improved signal-to-background ratio.
`
`CELL DESIGN
`Figure 1 shows the layout and dimensions of a (cid:237)-CE device
`with integrated optical components. Reservoirs A-C connect to
`the electrophoresis injector and separation capillaries, which form
`a U-shaped cell with a 100-140 (cid:237)m longitudinal path length.
`Reservoirs D and E connect to channels into which optical fibers
`were inserted to launch and collect light. These two reservoirs
`were designed for introducing index matching fluid to reduce
`scattering and reflection losses.
`The configuration used does not readily allow for lenses at
`the launch and collection points, although some shaping of the
`fiber tips to form lenses can be achieved by etching the optical
`fibers. To ensure that the light launched stays within the
`separation channel and is all collected, a launch fiber with small
`numerical aperture (NA) and small core diameter should be used.
`For a single-mode fiber, the beam waist, w, at the exit point is
`
`(14) Jacobsen, S. C.; Hergenro¨der, R.; Koutny, L. B.; Ramsey, J. M. Anal. Chem.
`1994, 66, 1114-1118.
`(15) Verpoorte, E.; Manz, A.; Lu¨di, H.; Bruno, A. E.; Maystre, F.; Krattiger, B.;
`Widmer, H. M.; van der Schoot, B. H.; de Rooij, N. F. Sens. Actuators 1992,
`B6, 66-70.
`(16) Harrison, D. J.; Glavina, P. G.; Manz, A. Sens. Actuators 1993, B10, 107-
`116.
`(17) Fan, Z.; Harrison, D. J. Anal. Chem. 1994, 66, 177-184.
`
`1040 AnalyticalChemistry,Vol.68,No.6,March15,1996
`
`0003-2700/96/0368-1040$12.00/0 © 1996 American Chemical Society
`
`Downloaded via UNIV OF VIRGINIA on November 6, 2018 at 18:26:01 (UTC).
`
`See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
`
`Agilent Exhibit 1274
`Page 1 of 7
`
`

`

`Figure 1. (a) Geometric layout of channels etched with glass plate.
`Letters mark reservoir access points. (b) Blowup of U-cell used for
`detection. Dimensions in millimeters.
`
`given by18,19
`
`+ 2.879
`
`V6 )
`
`w ) d(0.65 + 1.619
`V1.5
`V ) d(cid:240)
`NA
`(cid:236)
`
`(1)
`
`(2)
`
`where d is the core diameter and (cid:236) is the wavelength. For a 3.1
`(cid:237)m core, single-mode fiber with an NA of 0.1 and a wavelength
`of 488 nm, the waist is 3.93 (cid:237)m at the launch point. The angle of
`dispersion is given by
`
`ı ) arcsin(NA/n)
`
`(3)
`
`where n is the refractive index of the transmitting medium. The
`beam diameter, w¢, some distance x from the launch point, can
`be estimated as
`
`w¢ ) w + 2x tan ı
`
`(4)
`
`if traveling in a single medium. For a beam launched into water
`from a 3.1 (cid:237)m fiber with an NA of 0.1, the beam diameter will be
`(cid:24)22 (cid:237)m after passing through a 120 (cid:237)m long cell. Such a beam
`would be essentially contained within a 20 (cid:237)m deep channel, which
`would be a trapezoid 50 (cid:237)m wide at the top and 20 (cid:237)m wide at
`the bottom.
`
`(18) (a) Barnoski, M. Fundamentals of Optical Fiber Communications, 2nd ed.;
`Academic Press: New York, 1991. (b) Adams, M. J. An Introduction to
`Optical Waveguides; J. Wiley and Sons: Chichester, 1981; p 276. (c)
`Hunsperger, R. G. Integrated Optics: Theory and Technology, 2nd ed.;
`Springer-Verlag: Berlin, 1991; Chapter 6. (d) Sharma, A. B.; Halme, S. J.;
`Butusov, M. M. Optical Fiber Systems and Their Components; Springer-
`Verlag: Berlin, 1981; Chapters 1, 2.
`(19) Adams, M. J. An Introduction to Optical Waveguides; J. Wiley & Sons: New
`York, 1981; p 276.
`
`Figure 2. Geometry of channels and collection fibers around the
`U-cell of the flow channels. Flow channels are etched into the bottom
`plate only, and a top plate is thermally bonded.
`
`Since the launch fiber cannot be exactly at the detection cell
`edge due to fabrication constraints, there will be additional beam
`expansion. Assuming a 50 (cid:237)m separation between the aqueous
`sample channel and the fiber tip, and assuming that the value of
`n for this medium matches that of the glass chip, n ) 1.519, gives
`a beam width of 10.5 (cid:237)m at the cell-glass interface. Assuming
`that Snell’s law governs the refraction angle at the glass-water
`interface, and ignoring any lensing effect of the concave lens
`formed by the curved glass walls, the beam will expand to 28 (cid:237)m
`after traversing the 120 (cid:237)m cell and will be (cid:24)32 (cid:237)m in diameter
`at the collection fiber 50 (cid:237)m from the cell wall.
`Larger, multimode launch fibers could be used to increase light
`input. Assuming the same conditions as above, a beam launched
`from a 10 (cid:237)m fiber expands to 35 (cid:237)m across the cell and to 41
`(cid:237)m at the collection fiber. A 20 (cid:237)m launch fiber gives a beam 45
`and 51 (cid:237)m in diameter at the same two points. Consequently, a
`50 (cid:237)m collection fiber is adequate, but larger cross-section flow
`channels are needed for larger launch fibers if a 120 (cid:237)m path
`length is used. (We note that the angle of acceptance by the
`collection fiber is also given by eq 3, so with the same numerical
`aperture, a 50 (cid:237)m fiber will efficiently collect the light.)
`Alignment of the optical fiber core with the separation capillary
`can be achieved in several ways. Figure 2 illustrates the required
`geometry, in which the center of the channel for the optical fiber
`is offset from the plane of the glass-glass bond, so that the light
`path is aligned with the center of the separation channel. This
`was accomplished by etching the fiber channel after bonding the
`cover plate to the etched plate, as described later.
`The cell volume makes a contribution to band broadening.
`Using the standard formula,20 assuming a rectangular cell volume
`gives the height equivalent to a theoretical plate contributed by
`the detector cell, Hdet, as
`
`Hdet ) l2
`12did
`
`(5)
`
`where l is the cell path length and did is the injector-to-detector
`distance. A path length of 120 (cid:237)m will contribute 0.03 (cid:237)m of plate
`height for the device studied here and as little as 0.02 (cid:237)m for
`
`(20) Sternberg, J. C. Adv. Chromatogr. 1966, 2, 206-270.
`
`AnalyticalChemistry,Vol.68,No.6,March15,1996 1041
`
`Agilent Exhibit 1274
`Page 2 of 7
`
`

`

`typical devices reported recently,12 for which did (cid:25) 5 cm. Given
`the plate height range of 0.2-2 (cid:237)m that has been reported, a
`detector length of up to 250 (cid:237)m would introduce additional band
`broadening of 50-5%, respectively.
`
`EXPERIMENTAL SECTION
`Device Fabrication. A photomask glass used for photomasks
`was patterned photolithographically and etched with an HF/HNO3
`mixture, as described previously.17 Holes (1.5 mm diameter) were
`drilled in a cover plate with a diamond drill bit, and the two plates
`were then cleaned using a Model 2066 high-pressure cleaning
`station (MicroAutomation) under a class 100 clean hood. The
`plates were bonded by heating at 595 (cid:176)C for 6 h. Before bonding,
`the channels were etched either 20 (cid:237)m deep and 50 (cid:237)m wide with
`a transverse cell path length of 140 (cid:237)m, or 10 (cid:237)m deep and 30
`(cid:237)m wide with a cell path length of 120 (cid:237)m. Near reservoir C, the
`channel was expanded to 200 (cid:237)m wide.
`The optical fiber channels were enlarged after bonding the
`top plate by pumping 2-5% HF solution under 1.5 psi pressure
`for about 5-6 h. (Extreme care must be taken to avoid exposure to
`the HF solution.) This expanded the original 50 (cid:237)m (cid:2) 20 (cid:237)m
`trapezoidal channel to about 230 (cid:237)m wide by 150 (cid:237)m deep, giving
`an oval shape. The walls between the fiber and separation
`channels were reduced to (cid:24)10 (cid:237)m thickness.
`Optical fibers with nominally 125 (cid:237)m diameter glass cladding,
`250 (cid:237)m diameter jacket, and a core of 3.1 (F-SA), 7.9 (F-SS), or
`50 (F-MSD) (cid:237)m diameter were from Newport. Fibers were
`cleaved before use with a Newport F-BK2 fiber cleaver. The
`optical fibers were stripped of their polyimide coating and then
`etched in an HF/HNO3 mixture (10:1, using 49% HF and 70%
`HNO3). This reduced the outer cladding diameter slightly, to
`allow insertion into the fiber channel. Because the silica cladding
`etches faster than the core, the 3.5-4 min etch time created a
`rounded protuberance of core silica, giving a slightly convex lens
`at the fiber tip.21
`The fibers were inserted into the fiber channels in a class 100
`clean hood, with the assistance of a microscope and a fiber
`positioner (Newport FP-1 positioner mounted on Newport 423
`translation stages). Caution is required to avoid damaging the
`fiber during insertion. After insertion, index matching fluid, 1,4-
`dibromobutane (Aldrich, n ) 1.519), was introduced through
`reservoirs D and E. The entry points for the fiber were then
`sealed with epoxy resin (Araldite 5 min epoxy), and the reservoirs
`were sealed with parafilm.
`Reagents. The Na+ salt of fluorescein (Molecular Probes,
`Eugene, OR), fluorescein-5-isothiocyanate (FITC; Sigma, St. Louis,
`MO), and concentrated HF (Fisher, reagent grade) were used as
`received. The running buffer for all electrophoresis was 20 mM
`boric acid/100 mM tris(hydroxymethyl)aminomethane adjusted
`to pH 9.0. All fluorescein and FITC solutions were prepared in
`this same buffer to ensure identical ionic strengths and to avoid
`sample stacking effects. FITC solutions were allowed to hydrolyze
`overnight before use. Polymer waveguides were formed using
`Norland Adhesive, UV-curable types 71 and 88, or Epo-Tek 2 part
`optical adhesive (Epoxy Tech., Billerica, MA). Glass used for
`photomasks was obtained from AGFA Gevaert (Belgium) (73%
`SiO2, 13.5% Na2O, 1.8% Al2O3, 8.9% CaO, and 2.7% MgO by X-ray
`fluorescence).
`
`(21) Kayoun, P.; Puech, C.; Papuchon, M.; Arditty, H. J. Electron. Lett. 1981,
`17, 400-401.
`
`1042 AnalyticalChemistry,Vol.68,No.6,March15,1996
`
`Figure 3. Schematic of various detection geometries around the
`microstructure ((cid:237)-TAS). 1, Launch fiber; 2, fluorescence excitation
`lens; 3-5, fluorescence or transverse absorbance detection with lens,
`pinhole, and filter, respectively; 6 and 7, longitudinal absorbance
`collection fiber and filter, respectively; 8, transverse absorbance
`source using lens.
`
`Instrumentation. The computer-controlled power supply and
`relay arrangement has been described previously.17,22 Hamamatsu
`photomultiplier tubes (PMTs) were used to measure the fluores-
`cence and absorbance signals. Collection optics used were 25(cid:2)
`Leitz Fluotar (0.35 NA) or 7(cid:2) Rolyn (0.2 NA) objectives. The
`optical arrangements for detection are illustrated in Figure 3. The
`Uniphase Cyonics 488 nm laser was operated at 5.86 mW output
`power, measured with a Newport Model 835 power meter. The
`laser power was coupled into a 3.1 (cid:237)m fiber using a single-mode
`fiber coupler. The output power of the fiber was 0.143 (cid:237)W.
`A Gaertner L125B ellipsometer was used to determine the
`refractive index of the crown glass at 633 nm.
`Procedures. Devices were first filled with buffer, and then a
`sample dye was introduced at reservoir B. Sample was injected
`with a potential applied between reservoirs B and C (effective
`length, 6.5 cm). Separation was performed with a potential applied
`between reservoirs A and C (effective length, 5.8 cm). Here, the
`effective length of the 200 (cid:237)m wide segment near reservoir C is
`expressed as its equivalent 0.15 cm length of 30 (cid:237)m wide
`channel.23 The distance from injection to detection point, did, was
`3.7 cm.
`Several optical detection geometries were employed. Trans-
`verse absorbance involved illumination from below the chip
`(element 8 in Figure 3) with laser light delivered by mirrors and
`collection above the chip with a 25(cid:2) objective (5, Figure 3) and
`a 200 (cid:237)m pinhole at the image plane. A 488 nm notch filter was
`used for collection, and a 1% neutral density filter was used to
`
`(22) Seiler, K.; Harrison, D. J.; Manz, A. Anal. Chem. 1993, 65, 2637-2642.
`(23) Seiler, K.; Fan, Z. H.; Fluri, K.; Harrison, D. J. Anal. Chem. 1994, 66, 3485-
`3491.
`
`Agilent Exhibit 1274
`Page 3 of 7
`
`

`

`Figure 4. Scanning electron mirograph (SEM) of device cut and
`polished to display the launch and collection fiber channels and the
`U-cell optical path. Some SEM sample preparation damage is visible
`below the flow channel. Flow channel was 20 (cid:237)m deep.
`
`avoid photobleaching. Transverse fluorescence, our conventional
`detection method, was performed with excitation by laser light
`delivered by mirror and a lens at 45(cid:176)
`to the upper surface normal
`(2, Figure 3). Emitted light was collected with a 7(cid:2) objective, a
`200 (cid:237)m pinhole, a 508-533 nm bandpass filter (5, Figure 3), and
`a PMT. Longitudinal absorbance was performed by launching
`light into a 3.1 (cid:237)m single mode fiber (1, Figure 3) and collecting
`it with a 7.9 (cid:237)m optical fiber, unless otherwise indicated (6, Figure
`3). A 488 nm notch filter was also used. Longitudinal fluores-
`cence was performed using the 3.1 (cid:237)m launch fiber for excitation
`and a 7(cid:2)objective, a 200 (cid:237)m pinhole, and a 508-533 nm bandpass
`filter (5, Figure 3) for collection. The detector was positioned
`about midway along the 120 or 140 (cid:237)m optical cell path.
`
`RESULTS AND DISCUSSION
`Fabrication. The micromachined optical cell was fabricated
`in crown glass using conventional photolithography and chemical
`etching. The procedure described above, etching the fiber
`channel after bonding the cover plate, results in a fiber channel
`with a center that is offset from the bonding plane of the two
`plates. This aligns the fiber core with the center of the separation
`channel, as sketched in Figure 2. Figure 4 shows a side view of
`the assembled structure, in which the curvature of the walls is
`apparent. This curvature limits the approach of the fibers to the
`wall separating the fluid and fiber channels, as seen in the top
`view in Figure 5. The alignment of the fiber cores with each other
`is quite satisfactory, as seen in Figure 5. However, the fiber’s
`positioning relative to the fluid channel is slightly shifted. This
`problem could be resolved by redesigning the etch mask to shift
`the channel positions slightly.
`Alternatives to the fabrication procedures used here are
`possible. For example, the channels could be etched in top and
`bottom plates, although this requires careful alignment to (cid:24)1 (cid:237)m
`during bonding and the use of a two-mask process to define
`channels with two different depths. Another possibility is to form
`the optical waveguides in the glass by ion exchange doping rather
`than using conventional optical fibers, but this requires a waveguide
`that could survive the subsequent top-to-bottom plate bonding,
`which is usually done at elevated temperature. We did attempt
`
`Figure 5. Optical micrograph showing launch of 488 nm beam from
`lower fiber. (A) Beam path illuminated by fluorescein at 520 nm (503-
`533 filter) and (B) scattering at center wall and exit points seen with
`a 488 nm filter. Flow channel was 20 (cid:237)m deep, with a 140 (cid:237)m
`longitudinal optical path length.
`
`Table 1
`
`fluorescein
`concn ((cid:237)M)
`
`absorbance (AU)
`theora measd
`
`efficiencyb
`(%)
`
`absorbance path
`(length, (cid:237)m)
`longitudinal (140)
`fiber combination
`(launch/collection)c
`3.1/50
`3.1/7.9
`3.1/3.1
`
`transverse (20)
`
`62.5
`0.056
`0.0387
`62.5
`0.056
`0.0548
`62.5
`0.056
`0.0565
`12.5
`0.0112
`0.0112
`62.5
`0.008
`noised
`625
`0.080
`0.012
`64
`a Calculated from A ) (cid:15)bC, where b is path length and (cid:15) ) 6.4 (cid:2)
`104 M-1 cm-1. b Ratio of measured to theoretical absorbance. c Diam-
`eters of the launch and collection waveguides ((cid:237)m). d Too little signal
`to be distinguished from noise.
`
`69
`98
`101
`100
`
`to fill the channels with polymer-based waveguides, but the volume
`change upon curing resulted in numerous scattering defects in
`the waveguide.
`Optical Cell Performance. The efficiency of the optical
`absorbance cell was tested using several combinations of launch
`and collection fibers. The entire flow path was filled with a steady
`stream of fluorescein solution, and the absorbance was measured
`at 488 nm in a cell with a 140 (cid:237)m longitudinal path length. The
`data are presented in Table 1. Combining a 3.1 (cid:237)m launch fiber
`with either a 3.1 or 7.9 (cid:237)m collection fiber gave the theoretically
`predicted absorbances, within (2%. Larger collection fibers
`resulted in larger PMT signals due to increased light collection,
`
`AnalyticalChemistry,Vol.68,No.6,March15,1996 1043
`
`Agilent Exhibit 1274
`Page 4 of 7
`
`

`

`Figure 6. Simultaneous longitudinal absorbance and fluorescence
`detection of 50 (cid:237)M FITC, pH 9.0 at 2 kV applied (Vid ) 1.28 kV).
`
`but the cell’s absorbance efficiency, defined here as the ratio of
`observed to theoretical absorbance, was decreased. This is
`apparently due to stray light effects resulting from scattered 488
`nm light that reached the collection fiber without traveling entirely
`through the sample solution.
`In this sense, the theoretical
`calculation of suitable collection fiber diameter proved inadequate.
`A collection fiber of 7.9 (cid:237)m was used for the studies described
`below.
`A transverse absorbance path was also studied (elements 3,
`4, 5, and 8 in Figure 3), using a cell etched 20 (cid:237)m deep. The
`absorbance efficiency of this geometry could not be improved
`above 64%, indicating strong stray light effects. Consequently,
`the 140 (cid:237)m long longitudinal path gave a net 11-fold increase in
`absorbance, even though only a 7-fold increase would be expected
`on the basis of the path length difference.
`the beam
`that
`The calculations presented above predict
`diameter for the 140 (cid:237)m path in the cell of Figure 5 should be
`about 30 (cid:237)m where it exits the U-cell. However, due to the
`position of the cell off to one side, the beam is truncated by the
`contacting wall and exits with about a 22 (cid:237)m width. At the initial
`point of contact with the wall, the beam has traveled 50 (cid:237)m in
`water, at which point eq 4 predicts an 18 (cid:237)m diameter. Measure-
`ment of the beam width in Figure 5A at this point gives 18 ( 2
`(cid:237)m, in good agreement with theory.
`Figure 5A also illustrates that the beam is larger than the 7.9
`(cid:237)m collection fiber, leading to some losses. Further losses clearly
`occur from scattering at the flow channel-glass interface, as seen
`using a 488 nm bandpass filter, Figure 5B. However, the index
`matching fluid minimizes scattering at the other material inter-
`faces. It should be noted that losses in transmittance, which are
`independent of sample concentration, are not a serious problem
`unless the incident intensity is very low, whereas stray light effects
`will drastically reduce the cell performance.
`Figure 6 shows simultaneous determination of hydrolyzed
`FITC by longitudinal absorbance detection and longitudinal
`fluorescence excitation in a 120 (cid:237)m long cell. The separation
`efficiencies varied linearly with applied potential as expected
`(0.5-5 kV) but were the same for both longitudinal fluorescence
`and absorbance measurements at each potential (e.g., 19 000
`plates at 5000 V applied, Vid ) 3190 V, data not shown). Separation
`
`1044 AnalyticalChemistry,Vol.68,No.6,March15,1996
`
`Figure 7. Plot of normalized absorbance or concentration versus
`the calculated length of the injected sample plug for longitudinal (b)
`absorbance and (+) fluorescence measurements, with 50 (cid:237)M fluo-
`rescein, pH 9.0. Also plotted is the theoretical normalized concentra-
`tion obtained by numerically solving eq 7. The solid curves are for
`clarity only;
`the dashed curve shows the expected normalized
`absorbance in the absence of any disperson effects. Electrophoresis
`performed with Vsep ) 3 kV, tm ) 65.6 s.
`
`transverse
`efficiency was also measured with the previous,
`fluorescence detection configuration we employed, bringing the
`laser beam in from outside the chip at about 45(cid:176)
`to the collection
`optics, and observing a 30 (cid:237)m length of channel. This resulted
`in the same efficiency as obtained with longitudinal absorbance
`at a given potential, confirming that the larger cell length of the
`absorbance detector (120 vs 30 (cid:237)m) did not increase band
`broadening.
`Dispersion Effects. Dispersion due to diffusion plays a
`significant role in determining concentration at the detection point.
`This issue was examined in detail for the longitudinal cell. For
`the low linear velocities involved (j1 mm/s), the dispersion of
`an extended sample plug can be described by an error function
`expression given by Crank:24
`
`C(x) ) 1/2Co(erf
`
`h - x
`2xDdid/u
`
`+ erf
`
`h + x
`
`2xDdid/u)
`
`(6)
`
`where Co is the original concentration, 2h is the initial length of
`the plug, D is the diffusion coefficient, did is the injector-to-detector
`distance, u is the linear velocity in the capillary, and x is the
`longitudinal distance from the center of the absorbance cell.
`If
`the sample plug length is short enough relative to the optical cell
`path length, dispersion will result in a nonuniform concentration
`gradient within the cell. In this case, the Lambert-Beer law must
`be expressed as
`
`A ) (cid:15)s-a
`+aC(x) dx
`
`(7)
`
`where 2a is the width of the optical cell. Equations 6 and 7 are
`readily solved numerically using the Hi-Q software package
`(National Instruments, Austin, TX). The results of this dispersion
`
`(24) Crank, J. The Mathematics of Diffusion, 2nd ed.; Clarendon Press: Oxford,
`1975; pp 14-16.
`
`Agilent Exhibit 1274
`Page 5 of 7
`
`

`

`calculation are plotted as a function of plug length in Figure 7,
`for a detection cell length of 2a ) 120 (cid:237)m, D ) 3.3 (cid:2) 10-6 cm2/s
`(determined at pH 8.5 by polarography at 20 (cid:176)C), did ) 3.7 cm,
`and a linear flow velocity of 0.056 cm/s. (Velocity was calculated
`from the applied separation field (3 kV across 5.8 cm) and the
`measured mobility of 1.09 (cid:2) 10-4 cm2/V(cid:226)s.)
`Injected plug lengths were varied by controlling injection time
`and potential and were calculated from
`
`2h )
`
`tinjVinj
`6.5 cm (cid:237)
`
`(8)
`
`where (cid:237) is the mobility, determined from migration times
`measured as a function of applied field. We have previously shown
`that diffusion effects provide the main source of band broadening
`in the microchips, once sample plug size and detection volume
`are accounted for.10,17,22,23 Those results confirm that diffusion
`occurs in the channels with coefficients similar to values measured
`by other means. In the present experiments, the running buffer
`and sample solutions were prepared from the same stock buffer,
`so that sample stacking phenomena should be insignificant.
`Under these conditions, the solution to eqs 6 and 7 should
`describe the dispersion in peak height seen with varying sample
`plug size. Figure 7 shows how the peak absorbance and
`fluorescence varied with plug length, compared to the calculated
`effect of dispersion. The data exhibit lower signal than that
`calculated around the knee of the curve, but the theoretical trend
`matches the observed trend. The discrepancy is likely due to
`some inaccuracy in the diffusion coefficient due to differences in
`temperature and solvent conditions between the electrophoretic
`and polarographic measurements. For short plug lengths, diffu-
`sion at the injector has been shown to increase the amount
`injected over the calculated lengths.17,22,23
`Absorbance Detection Limits. The data in Figure 7 illustrate
`that varying the plug length allows control of the peak absorbance.
`We used this feature as a convenient means to vary absorbance
`of 50 (cid:237)M plugs of fluorescein and thus to obtain the absorbance
`detection limit for a cell with a 120 (cid:237)m path length. A plot of
`signal-to-noise (S/N) ratio versus absorbance, obtained by de-
`creasing the injected plug lengths of fluorescein, was linear, with
`a slope of 990 and an intercept of -0.6 (R2 ) 0.988). The S/N
`ratio reached 3 at about 0.003 AU.
`The absorbance detection limit obtained is rather high.
`However, it was limited by the poor coupling into the optical fiber
`that we achieved (>10 000:1 loss), and by the use of an Ar+ ion
`laser source. Use of a stabilized source and better coupling would
`improve detection limits with the same cell design.
`Concentration detection limits were measured by varying the
`concentrations of FITC and fluorescein for a fixed injection plug
`length. The absorbance calibration curve for the earliest eluting
`peak of hydrolyzed FITC was linear between 20 and 100 (cid:237)M
`formal concentration of FITC, measured in a 120 (cid:237)m path length
`cell. The S/N ratio increased linearly with concentration of FITC
`for plug lengths of 800 (cid:237)m, with S/N ) 9.1 ( 0.5 at 20 (cid:237)M FITC
`and a slope of 0.18. The concentration detection limit, in terms
`of formal injected concentration of FITC, was estimated to be
`about 6 (cid:237)M. Dispersion due to diffusion (a factor of (cid:24)0.8, vide
`supra) as well as the distribution of FITC between several
`hydrolysis products means the true concentration detection limit
`was somewhat lower.
`
`Figure 8. Fluorescence detection of 10 nM fluorescein, pH 9.0, with
`Vsep ) 3 kV and transverse fluorescence excitation using (a) 3.5 mm
`injection plug length (20 s at 1 kV), (b) 0.8 mm plug (5 s at 1 kV),
`and (c) 0.35 mm injection plug (1 s at 2 kV). This is contrasted with
`longitudinal fluorescence excitation using (d) 0.35 mm injection plug
`(1 s at 2 kV).
`
`The absorbance calibration curve was also determined by
`varying fluorescein concentrations in a 140 (cid:237)m longitudinal path
`length cell. Concentrations between 7 and 300 (cid:237)m were studied,
`giving a calibration slope of 750 M-1 and an intercept of -3 (cid:2)
`10-3 AU for a sample plug length of 800 (cid:237)m. The R2 factor was
`0.998. A plot of S/N versus concentration was also linear.
`Extrapolation to S/N ) 3 gave a concentration detection limit of
`5.6 (cid:237)M for fluorescein.
`Fluorescence Detection. Scattering of the excitation light
`creates a background that can raise the detection limits in
`fluorescence detection.6 Within the planar devices we have used,
`there is a fair amount of scattering from the glass and from the
`curved walls of the channels. The fiber launched design we
`employed here for absorbance can also be used for fluorescence
`excitation. In this case, the beam is launched longitudinally along
`the channel, and collection optics can be located at 90(cid:176)
`to the
`beam. Figure 5 shows that the light beam is smaller than the
`channel it is launched into, so the solution forms a sheath around
`the beam within the cell. This sheath geometry is similar to the
`surrounding water sheath in a sheath flow fluorescence cell,
`except the sample solution itself forms the “sheath”. The sheath
`flow cell provides good detection limits, since there is little light
`scattering from the solvent “cell walls”. For our cell design, the
`same sheath effect will occur so long as detection is performed
`away from the scattering at the entry and exit points of the cell.
`This can result in improved detection limits.
`Figure 8 shows the response to 10 nM fluorescein for
`transverse fluorescence excitation versus longitudinal excitation
`by the fiber. A 7(cid:2) objective with a 200 (cid:237)m pinhole was used for
`detection in both experiments, giving an observation zone about
`30 (cid:237)m long. For transverse excitation, an injection time of 20 s
`at 1 kV produced a measurable signal with S/N ) 7.2. An
`injection for 1 s at 2 kV or even 5 s at 1 kV produced no detectable
`signal for the transverse excitation. In contrast, a peak with S/N
`) 5.0 was seen when a 1 s injection was used at 2 kV and
`longitudinal excitation. As discussed above, the longer injection
`plug reduces the effect of dispersion on concentration, so this
`result indicates an improved detection limit. This improvement
`is even more remarkable considering that the laser power was
`unchanged, but the coupling efficiency to the fiber reduces the
`
`AnalyticalChemistry,Vol.68,No.6,March15,1996 1045
`
`Agilent Exhibit 1274
`Page 6 of 7
`
`

`

`photon flux more than 10 000-fold for the longitudinal excitation.
`Better coupling efficiency is possible, especially if a 5 or 8 (cid:237)m
`launch fiber is used. This is expected to result in further
`improvement in detection limit for longitudinal excitation. Use
`of a higher numeral aperture collection lens can also improve
`detection limits, though this will be true for both transverse and
`longitudinal excitation.
`Given the dispersion which occurs due to diffusion, the
`calculated