Anal. Chem. 1993, 65, 1481-1488
`
`1481
`
`Planar Glass Chips for Capillary Electrophoresis: Repetitive
`Sample Injection, Quantitation, and Separation Efficiency
`Kurt Seiler/ D. Jed Harrison/'f and A. Manz1
`Department of Chemistry, University of Alberta, Edmonton, Alberta, Canada T6G 2G2, and
`Central Analytical Research, Ciba-Geigy, Basel, Switzerland
`
`A sample injection system and a capillary elec-
`trophoresis channel integrated together on a pla-
`nar glass substrate are described, and the per-
`formance is evaluated. Voltages of at least 25 kV
`may be applied over an 11-cm-long capillary chan-
`nel with sufficient dissipation of the Joule heat
`generated (1.8 W/m). The glass substrate sustains
`at least 10® V/cm dc without dielectric breakdown.
`Using laser-induced fluorescence detection, mix-
`tures of fluorescein derivatives and fluorescein
`isothiocyanate-labeled amino acids were injected
`and separated. Up to 100 000 theoretical plates and
`obtained under optimized
`~20 plates/V were
`conditions, comparable to results with fused-silica
`capillaries. Quantitative analysis showed that
`peak areas were proportional to the amount
`in-
`jected for injected plug lengths greater than 200
`in
`µ  . The presence of two rectangular corners
`a 0.85-cm-long separation channel did not increase
`indicating a serpentine
`dispersion detectably,
`capillary channel 0.5 m in length could be fabri-
`cated in less than a 1-cm* 12 area.
`
`INTRODUCTION
`Capillary zone electrophoresis (CE) introduced by Mikkers
`et al.1 and Jorgenson and Lukács2·3 is an exciting separation
`technology that has attracted the attention of many analytical
`laboratories.4·5 It has shown the ability to resolve very complex
`mixtures of many different species within a short time and
`exhibits an efficiency that can exceed 106 theoretical plates.
`We have demonstrated that both an electrophoresis capillary
`and a sample injection system can be integrated together on
`a planar device, using microlithographic technology known
`as micromachining.6'7 Miniaturized systems for total chemical
`analysis (µ-TAS) of a complex sample have been previously
`proposed8"11 but only recently demonstrated experimentally.6·7
`In this report we present data for a planar electrophoresis
`* Author to whom correspondence should be addressed.
`* University of Alberta.
`! Ciba-Geigy.
`(1) Mikkers, F. E. P.; Everaerts, F. M.; Verheggen, Th. P. E. M. J.
`Chromatogr. 1979, 169, 11-20.
`(2) Jorgenson, J. W.; Lukács, K. D. Anal. Chem. 1981,53,1298-1302.
`(3) Jorgenson, J. W.; Lukács, K. D. J. Chromatogr. 1981,218,209-216.
`(4) Ewing, A. G.; Wallingford, R. A.; Olefirowicz, T. M. Anal. Chem.
`1989, 61, 292A-294A.
`(5) Capillary Electrophoresis: Theory and Practice; Grosman, P. D.,
`Colburn, J. C., Eds.; Academic Press: San Diego, CA, 1992.
`(6) Harrison, D. J.; Manz, A.; Fan, Z.; Lüdi, H.; Widmer,  . M. Anal.
`Chem. 1992, 64, 1926-1932.
`(7) Manz, A.; Harrison, D. J.; Verpoorte, E.; Fettinger, J. C.; Lüdi, H.;
`Widmer,  . M. J. Chromatogr. 1992, 593, 253-258.
`(8) Manz, A.; Fettinger, J. C.; Verpoorte, E.; Lüdi, H.; Widmer,  . M.;
`Harrison, D. J. Trends Anal. Chem. 1991, 10, 144-149.
`(9) Manz, A.; Graber, N.; Widmer,  . M. Sens. Actuators 1990, Bl,
`244-248.
`(10) Pace, S. J. U.S. Patent, 4.908.112, 1990.
`
`device obtained with a more highly optimized fluorescence
`detector design than we described previously.6 It should allow
`for better detection limits and a much lower contribution to
`band-broadening. Other instrumentation and experimental
`method improvements described here have provided precise
`control of the injection and separation processes and allowed
`the application of much higher applied potentials. Several
`features important to the performance of µ-TAS systems
`integrated electrophoresis device could be
`based on
`an
`examined due to these instrumental improvements.
`The ability to use higher potentials provides for more
`efficient separation of samples in electrophoresis,1"5 and it is
`important to demonstrate that high efficiencies can be
`obtained within the planar devices.
`In order to make a more
`compact design than the ~ 15 X 4 cm device we have studied,6
`it is important to explore the electrical breakdown charac-
`teristics of the glass substrate. Cross-talk between channels
`due to dielectric breakdown could limit how close together
`channels may be located on a device. Also, corners
`or curves
`will be required for any serpentine or coiled arrangement
`intended to increase capillary length within a small device
`area. Consequently, it is necessary to understand the effect
`of corners
`on band-broadening.
`We have shown that by use of electroosmotic pumping it
`is possible to control the direction of fluid flow in a manifold
`of intersecting capillaries without the use of valves.6 However,
`in a valveless device there will be both diffusion and convection
`phenomena. These will lead to leakage and mixing between
`solutions of different composition when they meet at channel
`intersections. Recently, Dose and Guiochon12 have derived
`a theoretical understanding of CE in this respect. They have
`pointed out that hydrodynamic flow and diffusion are two
`significant factors that need careful control to reduce quan-
`titative errors. Consequently, we have explored the effect of
`diffusion of sample at an intersection of capillaries in an
`In
`integrated manifold on the process of sample injection.
`addition, the problem of quantitative analysis of electroki-
`netically injected sample at a capillary channel intersection
`has been examined. Since diffusion may play a significant
`role in performance it
`is important to precisely control
`injection, separation, and delay times, and a computer-
`controlled system for this purpose has been developed. Such
`a system makes periodic cycling of injection and separation
`simple and so may facilitate automation of analysis in the
`future.7"9
`Our previous study examined a very simple separation of
`fluorescein derivatives and required 5 or 6 min. To show the
`ability to deal with more complex and realistic samples we
`present here the separation of several amino acids.13"15 Using
`(11) Gale, R. J.; Chowski, K. In Biosensor Technology, Fundamentals
`and Applications; Buck, R. P., Hatfield, W. E., Umana, M., Bowder, E.
`F., Eds.; Marcel Dekker: New York; 1990; pp 55-62.
`(12) Dose, E. V.; Guiochon, G. Anal. Chem. 1992, 64, 123-128.
`(13) Cheng, Y.-F.; Dovichi, N. J. Science 1988, 242, 562-563.
`(14) Liu, J.; Dolnik, V.; Hsieh, Y.-Z.; Novotny, M. Anal. Chem. 1992,
`64, 1328-1336.
`(15) Monnig, C. A.; Jorgenson, J. W. Anal. Chem. 1991, 63, 802-807.
`
`0003-2700/93/0365-1481 $04.00/0
`
`© 1993 American Chemical Society
`
`Downloaded via UNIV OF VIRGINIA on November 6, 2018 at 22:33:03 (UTC).
`
`See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
`
`Agilent Exhibit 1288
`Page 1 of 8
`
`

`

`1482 · ANALYTICAL CHEMISTRY, VOL. 65, NO. 10, MAY 15, 1993
`
`Figure 2. Schematic representation ot the Instrumental setup. For
`clarity only one power supply and one high-voltage (HV) relay are
`shown, although more were used. The computer controls the HV
`power supplies (I) and the HV relays, which are activated with a 24-V
`power supply via digitally controlled relays (II).
`
`One device was prepared with no Pt electrodes by etching the
`Pt away in 80 °C concentrated aqua regia before bonding the
`It was then cut with a glass saw to remove
`the contact
`plates.
`holes, and new holes were drilled by hand into the bonded device,
`using a low-speed drill and a diamond bit, as sketched in the
`inset of Figure 4. Following this the active channel lengths were
`reservoir 1 to beginning of narrow channel, 2 mm;
`as follows:
`from the end of the wide channel to the intersection, 9.3 mm;
`reservoir 2 to intersection, 90 mm; reservoir 3 to intersection, 100
`mm. This device was used only to study the application of high
`electric fields.
`Sample Introduction. The manifold of channels was initially
`flushed with the buffer by applying pressure to one of the
`It was very important to clean the
`reservoirs with a syringe.
`surface of the glass carefully with distilled water to remove
`conductivity paths from electrolyte solution on the surface. Fine
`Pt wires were inserted in the reservoirs to supply the electro-
`phoresis voltage. The sample was driven from reservoir 2 to the
`location of the detector by applying a voltage between reservoirs
`2 and 3, until a stable signal was obtained. The sample in the
`separation channel was flushed to waste by applying a voltage
`between reservoirs 1 and 3, before the cycling of the injections
`and separations was started.
`Instrumentation. A schematic representation of the instru-
`mental setup is shown in Figure 2. FUG Elektronik Model HCN
`2000 (0-2 kV) and Model HCN 12500 (0-12.5 kV) power supplies
`(Rosenheim, Germany) were used for sample injection and for
`separation, respectively. The current in the respective channels
`was monitored by measuring the voltage drop across a 10-k0
`resistor placed between one solvent reservoir and ground and
`recorded with a Hewlett-Packard dual-strip chart recorder Model
`7128A.
`Connections between the power supplies and ground to the
`devices were interrupted or switched between reservoirs using
`high-voltage relays (Model HVS10/S3, Kilovac, Santa Barbara,
`CA, and Model 24HV1A-100, Douglas Randall, Pawcatuck, CT).
`The relay armatures were activated with a 24-V (2.4-A) power
`supply (GHOF 2-24, GFC Power Ltd., Ontario, Canada). Contact
`of this supply to the high-voltage relay armatures was controlled
`by the computer using a set of digitally controlled relays (Control
`488/16, Iotech, Cleveland, OH).
`A Lab VIEW (National Instruments) program running on a
`Macintosh Ilci with a laboratory interface board (Model NB-
`MIO-16L, National Instruments) was used to control the output
`of the power supplies through two digital-to-analog converters.
`The Iotech relays were controlled via the digital 1-0 lines, and
`data acquisition utilized the analog-to-digital converters used in
`the double-ended input mode. A measuring cycle comprised
`generally the following operations:
`(1) activate (close) relay to
`sample reservoir; (2) apply injection voltage for the injection
`time; (3) deactivate (open) relay to sample reservoir and activate
`relay to buffer reservoir; (4) apply separation voltage. The rise
`and decay times of the power supplies when connected to the
`device were ~0.3 s (to 99%).
`Optical Setup and Signal Processing. The optical system
`was constructed on an optical breadboard (3 ft X 3 ft, Melles
`Griot). A 488-nm air-cooled Ar ion laser (Cyonics/Uniphase,
`Model 2011-20SL, Newport Research) operated at 4 mW served
`
`Figure 1. Top view of the layout of the channels Integrated on a planar
`glass substrate. The open circles show the numbering for the reservoirs
`contacting the channels, and the filled circles show the channel numbers.
`The channel leading to the edge of the device was plugged with epoxy.
`The location of several of the electrodes Integrated on the device Is
`shown (Pt). The detector was placed at (I) for most of the studies and
`at (II) for the study on the effect of corners. The overall dimensions
`are 14.8 cm X 3.9 cm X 1 cm. The Inset shows a magnification of
`the Intersection, where the sample was Injected.
`the higher voltages and improved detector now available, we
`were able to attain high separation efficiencies, and these
`compare favorably with conventional fused-silica capillaries.
`EXPERIMENTAL SECTION
`Reagents and Solutions. Three different buffer solutions
`were used for the experiments: a pH 8.0 buffer [50 mM boric
`acid, 50 mM tris (hydroxymethyl) aminomethane (tris)], apH 9.1
`buffer (10 mM sodium borate), and a pH 9.2 buffer (30 mM
`sodium carbonate) adjusted with HC1. The fluorescein sodium
`salt, the L-amino acids arginine (Arg), phenylalanine (Phe), and
`glutamine (Gin), and fluorescein 5-isothiocyanate (FITC) were
`obtained from Sigma (St. Louis). The fluorescein 5- (and 6-)
`sulfonic acid sodium salt (isomeric mixture) was obtained from
`Molecular Probes, Inc. (Eugene, OR). All were used as received.
`Labeling of Amino Acids. Labeling of the amino acids was
`performed as described in ref 15. To 1 mL of 6 mM FITC in
`acetone was added 3 mL of 3 mM solution of each amino acid
`in a pH 9.2 carbonate buffer (30 mM). This was allowed to react
`at room temperature in the dark overnight. It was further diluted
`with the mobile phase (pH 9.2,30 mM carbonate buffer) to obtain
`a solution of 10 µ  of each of the labeled amino acids. Before
`use, all solutions were passed through a 0.22-µ   filter (Millex-
`GV, Millipore, Bedford, MA) to remove particulates.
`Devices. The glass capillary electrophoresis-based TAS
`(CETAS) structures were fabricated under contract by Mettler
`AG (Switzerland) as discussed earlier.6-® A bottom glass plate
`had channels etched in it, while a top plate had Pt electrodes
`defined on it. These plates were melted together under controlled
`conditions, so that the channel shape was not distorted.6 The
`layout of the capillaries and some of the electrodes is shown in
`Figure 1. The two narrow channels, 2 and 3, are 30 µ   wide and
`10 µ   deep, while channel 1 is 1 mm wide and 10 µ   deep. Three
`holes were made through the top plate to contact the channels
`and serve as reservoirs. The channel
`lengths are as follows:
`reservoir 1 to beginning of narrow channel, 162 mm; from the
`end of the wide channel to the intersection, 9.3 mm; reservoir 2
`to intersection, 166.6 mm; reservoir 3 to intersection, 139 mm.
`Plastic pipet tips were shortened so that they fit into the holes
`of the top plate to form larger reservoirs for the electrolyte
`solutions. The Pt wires (0.3-mm diameter) inserted into these
`reservoirs were further isolated with a glass tube (wall thickness
`the reservoirs and electrodes.
`1.5 mm, length ~ 2 cm) placed over
`
`Agilent Exhibit 1288
`Page 2 of 8
`
`

`

`ANALYTICAL CHEMISTRY, VOL. 65, NO. 10, MAY 15, 1993 · 1483
`This could be done either from analyzing peak height and area
`from calculating the second statistical moment. There was
`or
`good agreement except when the signal-to-noise ratio was poor.
`RESULTS AND DISCUSSION
`The capillary electrophoresis TAS device (CETAS) layout
`is illustrated in Figure 1. The channels were etched into one
`glass plate, and a second was then bonded on top. This second
`piece had holes drilled into it to contact the channels. Pt
`electrodes were also fabricated on this plate, as shown in Figure
`1. The reservoir and channel numbering referred to in the
`text is also shown. As discussed previously,6 the fourth outlet
`was sealed with epoxy to prevent undesired hydrodynamic
`flow.
`Electrical Characteristics. When a voltage was applied
`between any pair of reservoirs in a device in which integrated
`Pt electrodes were present, the current was stable up to
`between 8 to 10 kV. The channel resistances evaluated from
`the current-voltage (I-V) response were in agreement with
`the channel geometry, indicating the current path was through
`the channels. As reported previously,6 9% of the potential
`was distributed between reservoir 1 and the channel inter-
`section point when reservoir 3 was at ground. The large
`channel (1) is separated by 1 mm from an integrated Pt
`electrode that contacts channel 2. With a potential applied
`between reservoirs 2 and 1 (1 at ground) this electrode will
`be floating at a potential very close to the applied voltage.
`Observation of a stable current at 10 kV which was consistent
`with the channel resistances indicates that glass can sustain
`at least 105 V/cm. The maximum potentials utilized in this
`work are about twice those achieved previously6 and were
`attained by insulating the Pt wires contacting the reservoirs
`with glass sleeves. At still higher applied potentials the
`in the channel became unstable, and this was
`current
`correlated with the evolution of gas in the channels at the Pt
`wires integrated in the device. While these wires were not
`connected to the power supplies they contact the channels
`at different locations and lead to the edge of the device, where
`the metal contact pads are located 1 mm from each other.
`Arcing between these contact pads, which are floating at the
`various channel potentials, is clearly the event which leads
`to current flow and water electrolysis at the Pt electrodes, as
`demonstrated below.
`To further evaluate the possible breakdown of the glass
`integrated Pt
`dielectric, a device was prepared without
`electrodes. The contact reservoirs were further separated by
`drilling new access holes in the bonded device, and the device
`was cut into two sections along a plane parallel to that running
`through the original reservoirs. The channels exposed at the
`edge by this cut were sealed with epoxy. The I-V curves for
`this device were stable and linear at potentials up to 25 kV
`(limited by the power supply), as shown in Figure 4. With
`10 mM borax buffer at pH 9.1 as electrolyte the channel
`resistances were 5.64 and 3.16 ± 0.01 Gfl between reservoirs
`2 and 3 and 1 and 3, respectively. The ratio of these values
`is in agreement with the channel geometry for this structure,
`indicating that the current flow is through the channels. A
`voltage of 25 kV corresponds to an electric field of 2.3 kV/cm
`when applied between reservoirs 1 and 3, which can be
`compared to the value of 0.3 kV/cm frequently used in CE
`in conventional fused-silica capillaries. The linearity of the
`I-V curve indicates that Joule heat is effectively dissipated
`at an energy density of at least 1.8 W/cm. This is larger than
`the value of 1 W/cm that has been suggested as the limit for
`CE in uncooled fused-silica capillaries.15 In addition, the
`potential through the glass between the channels contacting
`reservoirs 1 and 2 corresponds to 7.8 X104 V/cm, corroborating
`the good insulating ability of the bulk glass and the bonded
`joint between the top and bottom plates, as discussed above.
`
`Figures. Schematic of the optical detection system. Ar Ion laser light
`was focused with a lens (1) onto the separation channel, which was
`held In place with a plexiglass holder (2). Fluorescence emission (3)
`was collected with a microscope objective (4), focused onto an air slit
`(5), filtered (6), and then detected with a photomultiplier tube. Objective,
`air slit, and filter were mounted In a microscope body, which was fixed
`on x-y translation stages.
`
`It was focused onto the
`as the fluorescence excitation source.
`channel of the µ-TAS at ~45° to the surface using a lens with
`a focal distance of 30 cm, as shown in Figure 3. The surfaces of
`the CETAS device were roughened during the bonding process,
`so the beam diameter focused in the channels tended to vary
`between 100 and 200 µ  , depending on position. The laser
`position was shifted along the channel to study the effect of
`injector-to-detector distance. The paths of the reflected beams
`were arranged so that they did not strike the capillary channels
`elsewhere, in order to avoid photobleaching.
`The fluorescent radiation was detected with the optical axis
`of the assembly shown in Figure 3 perpendicular to the plane of
`the device. Light was collected with a microscope objective (10:
`1, NA 0.30, working distance 6 mm, Rolyn Optical) mounted on
`a microscope body (Melles Griot). A Hamamatsu Model R1477
`photomultiplier tube (PMT) powered by a Hamamatsu HC123-
`01 high-voltage supply was mounted in a sealed housing at the
`exit of the microscope. A 100-µ   air slit (Melles Griot) was placed
`at the field stop plane of the microscope in a specially made
`adaptor, to restrict the channel length viewed by the detector to
`~ 10 µ  . An Omega 518 DF25 interference filter (transmitting
`505-538 nm) was mounted over the entrance to the PMT housing
`to eliminate scattered laser light. The entire assembly was
`covered with a large box to reduce, but not eliminate, ambient
`light. Both the device mounting stage and the microscope
`assembly were mounted on Newport Research x-y translation
`stages.
`The PMT output was connected across a 1-  
`resistor. The
`potential developed was recorded with a Linear Instruments
`Model 500 strip chart recorder and was also filtered through two
`stages of active, six-pole Butterworth filters (Krohn-Hite Model
`3342, Avon, MS) both set at 50 Hz, before inputting the signal
`to the A-D converter. The response time of the detector was 40
`ms with this configuration. The sampling rate of the A-D
`converted varied between 10 and 20 counts per second. Once
`acquired digitally, the electropherogram peak parameters (center
`of gravity, peak area, variance) were obtained by statistical
`In this report, the time of migration of the
`moment analysis.
`center of gravity is used as the peak migration time. The number
`of plates was calculated after baseline correction for each peak.
`
`Agilent Exhibit 1288
`Page 3 of 8
`
`

`

`sample I (  = 8)
`fluorescein
`fluorescein 5- (and 6-)
`sulfonic acid
`sample II (  = 5)d
`Arg
`Phe
`Gin
`
`2.345 ± 0.007
`1.875 ± 0.007
`1.319 ± 0.006
`
`1193
`953
`670
`
`73.3 ± 2.4
`70.3 ± 2.4
`31.7 ±2.3
`" Parameters obtained for a series of separations at different
`separation voltages (from 2 to 10 kV). The injection parameters
`were constant (for sample 1,110 V/60 s; for sample II, 250 V/60 s).
`6 The mobilities (SD given) and the plug lengths were calculated
`with eqs 1 and 2, respectively. c The listed slopes (=k¡l¡C¡, SD given)
`were obtained from a proportionality analysis of area vs migration
`time (see eq 3). d Arg, arginine; Phe, phenylalanine; Gin, glutamine.
`
`1484# ANALYTICAL CHEMISTRY, VOL. 65, NO. 10, MAY 15, 1993
`
`Table I. Mobilities and Other Parameters*
`slope of peak
`mobilities6
`(X104)
`areas vs tmc
`(arb units/s)
`(cm2/V-s)
`
`caled plug
`length6
`(µ  )
`
`1.932 ± 0.006
`1.083 ± 0.004
`0.975 ± 0.004
`
`44.3 ± 1.6
`6.4 ± 0.1
`24.6 ± 0.5
`
`432
`231
`212
`
`POTENTIAL [kV]
`Figure 4. Current vs the applied potential difference between channels
`1 and 3 or 2 and 3 in a modified device similar to that shown in Figure
`1. The integrated Pt electrodes were not present, and new contact
`holes had been drilled (see inset and text).
`
`60 sec
`
`ELECTRIC FIELD DURING INJECTION [V/cm]
`
`ELECTRIC FIELD DURING SEPARATION [V/cm]
`_...........H _IÉ
`-HP
`
`émSmmm
`
`1
`
`......
`
`.1_1
`
`CYCLE C
`CYCLE B
`CYCLE A
`Figure 5. Three cycles (a-c), each including Injection and separation
`of a sample of 10 µ  fluorescein (I) and 10 µ  isomeric mixture of
`fluorescein 5- (and 6-) sulfonic acid (II and III) In a pH 8.0 tris-boric
`acid buffer. The length between Injection and detection was 6 cm. The
`figure also shows the potential program applied between reservoirs
`2 and 3 for sample injection and between 1 and 3 for separation.
`Repetitive Injections. Sample was introduced through
`reservoir 2 and small sample plugs were injected through the
`intersection point toward reservoir 3 by applying a low
`potential between reservoirs 2 and 3 for a short period, as
`discussed previously.6 A separation was then performed by
`applying a positive voltage between reservoirs 1 and 3 (3 at
`ground). The fluorescence detector could be located at
`different points between the intersection point and reservoir
`3. To control and automate this injection and separation
`procedure, an apparatus consisting of several high-voltage
`power supplies, high-voltage relays, and digitally controlled
`relays under the control of a microcomputer was assembled,
`as described in the Experimental Section. This apparatus
`intended to allow precise cycling and separation times
`was
`so that band-broadening effects due to diffusion may be
`minimized and controlled (vide infra).
`Three repetitive cycles, each consisting of sample injection
`(60 s), separation (280 s), and delay (10 s) steps are shown in
`Figure 5. The sample was a mixture of 10 µ 
`fluorescein and
`10 µ 
`fluorescein 5- (and 6-) sulfonate in 50 mM tris-boric
`
`acid buffer, pH 8.0. The first peak was
`identified as
`fluorescein, while the second two were due to the fluorescein
`sulfonate sample. These two peaks may represent the two
`the presence of an unidentified impurity. The
`isomers or
`figure shows qualitatively that
`the amount
`injected, as
`indicated by the peak heights and areas, decreased as the
`injection voltage was decreased. The reproducibility of peak
`in the slope of a plot of
`areas was estimated from the error
`vs migration time, as discussed below and presented in
`areas
`Table I. The standard deviation in peak area was about ±3 %.
`The peak migration times were reproducible to less than 1 %,
`and it was possible to make more than 40 repetitive injections.
`It can be seen in Figure 5 that the fluorescence intensity
`decreased during the delay and injection period. The laser
`intensity was sufficient to photobleach the dye essentially
`completely, so that when the flow rate was low or zero (i.e.,
`during the 10-s delay, or when the injection voltage was
`applied) the signal changed rapidly to background levels
`obtained in the absence of
`fluorescent dye. Once the
`separation voltage was reapplied, the higher velocity increased
`fluorescence to a higher background level. This level was
`controlled by the extent of leakage of dye from channel 2 into
`3.6
`Amino Acid Separation. To demonstrate the ability to
`deal with more complex samples, fluorescein isothiocyanate-
`labeled amino acids were separated. The electropherogram
`of three amino acids (Arg, Phe, Gin) is shown in Figure 6.
`This represents the first separation of amino acids in a planar
`device. The separation voltage of 10 kV applied between
`reservoirs 1 and 3 corresponds to 6.3 kV over
`the injection-
`detection distance of 9.6 cm or
`to an electric field of 655
`V/cm. The three amino acids are clearly separated in only
`identified by injecting each amino
`120 s. The peaks were
`acid separately. The glutamine peak exhibited a shoulder,
`which originated from the glutamine solution itself. The first
`peak eluted was identified as an impurity in the phenylalanine
`solution. The migration times of each component were
`linearly dependent on the injection-detector distance, d¡d,
`between 3.0 and 9.6 cm. This indicates the channel and
`electric field were
`spatially uniform in the longitudinal
`direction.
`The electrokinetic mobilities of the three principal peaks
`were evaluated from the slope of linear regression lines of the
`inverse of the migration time, tm vs the electric field, E, applied
`during separation. This relationship is given by
`l/tm = MiE/did
`(1)
`where µ,· is the electrokinetic mobility of component i. This
`mobility is the vector sum of the electroosmotic and elec-
`
`Agilent Exhibit 1288
`Page 4 of 8
`
`

`

`ANALYTICAL CHEMISTRY, VOL. 65, NO. 10, MAY 15, 1993 · 1485
`
`PLUG LENGTH [cm]
`Figure 7. Peak area as a function of the calculated plug length for
`the FITC-labeled amino acids Arg and Phe.
`Table II. Calculated Plug Length Dependence on Peak
`Areas®
`
`sample I
`fluorescein (  = 4)
`fluorescein 5-
`(and 6-)sulfonic
`acid {  = 3)
`sample II
`Arg (  = 4)
`Phe (  = 4)
`Gin (n - 4)
`
`intercept6
`(gm)
`
`slope6
`(gm/arb units)
`
`SEC (gm)
`
`35.7 ± 6.7
`64.9 ± 26.3
`5.7 ± 10.3
`
`0.089 ± 0.001
`0.40 ± 0.007
`0.037 ± 0.001
`
`141.2 ± 12.9
`99.1 ± 32.2
`39.7 ± 15.3
`
`0.978 ± 0.012
`0.694 ± 0.026
`0.822 ± 0.020
`
`±7.8
`±23.0
`±8.2
`
`±14.5
`±35.5
`±16.3
`
`0 Parameters of linear regression analysis of calculated plug length,
`Í¡nj (gm) vs peak area, A¡ (arb units) (/„,,, = slope X A, + intercept).
`The analysis includes plug lengths in a range from 200 to 1600 gm.
`The correlation coefficients of all linear regressions were larger than
`0.996.6 ±SD.c Standard error.
`
`with 9 V for 10 s, the largest with 1 kV for 60 s, whereas the
`time between the separations was kept constant at 70 s. Plots
`of the peak area vs the injected plug lengths calculated with
`eq 2 are shown in Figure 7 for Arg and Phe.
`For plug lengths greater than ~200 gm a linear correlation
`with peak area was observed, while the relationship deviates
`for smaller plugs. Data in the linear region were subject to
`linear regression of plug length on peak areas. The slope
`(1 /kitmC¡) and intercepts (gm) obtained for both fluorescein
`and the various derivatives are reported in Table II. At this
`state it is unclear whether the greater indeterminant error
`lies in the areas or the plug lengths, and we have arbitrarily
`chosen to treat l¡ as the dependent variable in the analysis.
`This offers the advantage that the uncertainty is expressed
`in units of plug length, which are easily interpreted in terms
`of the injection performance. The high linearity (R2 > 0.996)
`of the linear regressions shows that, by careful cycling of
`injection and separation, it is possible to inject a well-defined
`sample plug of length ¿200 gm into the separation channel.
`The standard error given in Table II corresponds to the error
`in the mean of the plug lengths studied (~900 gm) and
`indicates the average precision was about ±2% for a 900-gm
`plug. Plug lengths of ~ 200 gm are about the minimum length
`used with electrokinetic injections into fused-silica capillaries,
`whereas typical injection lengths are 5-10 times larger.17
`For all sample components the peak areas still decrease
`with smaller injected plug lengths, as calculated using eq 2
`and shown for Arg and Phe in Figure 7. Extrapolating from
`the linear range, the smallest values for the areas correspond
`to injection lengths in the range of 100 gm, corresponding to
`(17) Huang, X.; Coleman, W. F.; Zare, R. N. J. Chromatogr. 1989,480,
`95-110.
`
`0
`
`30
`
`90
`
`120
`
`60
`TIME (sec)
`Figure 8. Electropherogram of three 10 gM FITC-labeled amino acids
`(Arg, Phe, and Gin in a pH 9.2 carbonate buffer). Electrokinetic Injection
`was performed from channel 2 by applying a voltage of 250 V for 60
`s (~1-mm plug length). For the separation, a voltage of 10 kV was
`applied between reservoirs 1 and 3 (6.3 kV between injector and
`detector). The separation distance from Injection to detection was 9.6
`cm.
`
`trophoretic mobilities. The separation voltage was varied
`from 2 to 10 kV (131 to 655 V/cm) for a specific set of injection
`parameters, and the distance between the points of injection
`and detection, da, was 9.6 cm. Excellent
`linearity was
`obtained throughout for the plots of tm~l vs E(R2 > 0.9999),
`and migration times were reproducible within 1% for replicate
`injections at each potential. The data for both amino acid
`and fluorescein samples are given in Table I. The small
`standard deviations in g¡ and the linear dependence of tm-1
`on E confirm that Joule heat is sufficiently dissipated in the
`device up to at least 655 V/cm.
`Sample Injection and Detection. Electrokinetic injec-
`tion of sample from channel 2 results in different injected
`plug lengths for each ion, due to their differences in mobility.
`The plug length of an injected component, U, is given by16
`(2)
`h = ^inj®inj^i
`where Einj is the electric field applied during injection and t¡nj
`is the period it is applied for. From this expression it is
`possible to estimate the plug lengths, ignoring any diffusion,
`possible convective effects, or electric field distortions at the
`channel intersection.
`For a detector response that is proportional to analyte
`concentration the peak area, A;, will be proportional to the
`amount of sample injected and be described by
`A¿ = kteJCt
`(3)
`is the component’s concentration and k¡
`is a
`where C¡
`proportionality constant which depends on the detector
`design, the detector and device geometry, and the optical
`properties of the analyte. The same data used to evaluate
`integrated and subjected to proportionality
`the g, were
`analysis of peak area vs migration time. The slopes obtained
`(k¡l¡Ci) are given in Table I. The low standard deviation in
`the slopes indicates that the detector response is proportional
`to concentration in the range studied.
`The relationship between the calculated plug length and
`the experimentally measured peak area was evaluated by
`varying the injection parameters (injection voltage and
`injection time). The separation voltages were kept constant
`at 4 and 6 kV for the fluorescein derivatives and for the amino
`acid mixture, respectively. The smallest injection was made
`
`(16) Huang, X.; Gordon, M. J.; Zare, R. N. Anal. Chem. 1988,60,375-
`377.
`
`Agilent Exhibit 1288
`Page 5 of 8
`
`

`

`1486 · ANALYTICAL CHEMISTRY, VOL. 65, NO. 10, MAY 15, 1993
`
`Figure 8. Peak area for 1 µ  fluorescein, pH 8.0, vs square root of
`injection time, with no applied voltage during Injection. This time Is
`simply the period between two applications of the separation potential.
`
`an injected volume of 30 pL and thus to 300 amol of the
`component. The amount injected is larger than what would
`be estimated from the calculated injection plug length when
`it is below 200 µ  . This effect is likely due to the influence
`of diffusion during injection on the sample plug shape and
`concentration gradient at short times, as discussed below.
`The negative intercept observed in Figure 7 for extrapo-
`lation of the linear region of the data may arise from the fact
`the bac