Journal of’ Chromatography, 593 (1992) 253-258 Elsevier Science Publishers B.V., Amsterdam CHROMSYMP. 2444 Planar chips technology for miniaturization and integration of separation techniques into monitoring systems Capillary electrophoresis on a chip Andreas Manz*, D. Jed Harrison*, Elisabeth M. J. Verpoorte, James C. Fettinger, Aran Paulus, Hans Li.idi** and H. Michael Widmer Central Analytical Research, Ciba-Geigy Ltd., CH-4002 Basle (Switzerland) ABSTRACT Miniaturization of already existing techniques in on-line analytical chemistry is an alternative to compound-selective chemical sensors. Theory on separation science predicts higher efficiency, faster analysis time and lower reagent consumption for microsystems. Micromachining, a well known photolithographic technique for structures in the micrometer range, is introduced. A first capillary electrophoresis experiment using a chip-like structure is presented. INTRODUCTION The continuous monitoring of a chemical param- eter, usually the concentration of a chemical spe- cies, is gaining increasing attention in biotechnol- ogy, process control, and the environmental and medical sciences. Chemical sensors exhibit only a minimal number of applications for measurements of combustion gases, certain ions and enzyme sub- strates. The state-of-the-art strategy is called “total chemical analysis system” (TAS), which periodical- ly transforms chemical information into electronic information. In such a system, sampling, sample transport, necessary chemical reactions, chromato- graphic or electrophoretic separations and detec- * Permanent address: University of Alberta, Edmonton, Can- ada. ** Present address: Ciba-Corning Diagnostic Corp., Medfield, MA, USA. tion are performed automatically. Some examples of TAS, such as a gas chromatographic monitor [l] and an on-line glucose analyser [2], have been re- ported. Recently, we proposed a general concept for a miniaturized TAS [3-6]. As far as separation techniques are concerned, miniaturization has been heavily discussed for many years. Improved separation performance at shorter retention times is predicted by theory. Mini- aturization has been experimentally realized using small-diameter particles or open capillaries. Devia- tions from theoretical predictions have usually been caused by inhomogeneity in column packings or capillary diameters, inappropriate injections or large detection volumes. At least two publications have been presented in the literature on the use of photolithographically fabricated microstructures for gas [7] and liquid chromatography [8]. Recently, we proposed a 15-nl detector cell for absorption measurements (optical pathlength 1 mm [4-61). This 0021-9673/92/$05.00 0 1992 Elsevier Science Publishers B.V. All rights reserved
`
`Agilent Exhibit 1286
`Page 1 of 6
`
`

`

`254
`
`the parameter of inter- est as a function of the variables to be miniaturized (space and time), shows the major trends of a pa- rameter during its down-scale. In the case of capil- lary separation systems, the two approaches are equally interesting. Table I depicts the results of an analysis carried out according to approach 1. Choice of the desired A. MANZ et al. number of theoretical plates at a given retention time, as well as the heating power per length (in the case of CE), allows comparison of the resulting cap- illary dimensions and operation conditions for CE, liquid chromatographic (LC) and supercritical fluid chromatographic (SFC) separation experiments. The microchannels must be a few micrometers in diameter (2.8-24 pm), a few centimeters in length (6.5-20 cm) and need small-volume detectors (3.3- 94 pl). Although these values cannot replace experi- mental results, they give an indication of values for- bidden by theory. Approach 1 is very meaningful if the values of the given parameters are clear and if the optimum performance is well defined, as is the case with the Golay equation for capillary LC and SFC. In the case of CE, the optimum performance is basically determined by the maximum voltage ap- plied to the system. The higher the voltage, the bet- ter the separation performance and, at the same time, the faster the analysis. The limitation is usu- ally given by the heat produced in the capillary. Three parameters may be relevant: the power per TABLE I CALCULATED PARAMETER SETS FOR A GIVEN SEPARATION PERFORMANCE OBTAINED WITH CE, LC AND SFC Assumed constants are: diffusion coefficients of the sample in the mobile phase, 1.6
`paper presents a technique for the manufacture of entire microchannel systems with very high preci- sion. Such systems allow injections in the pl or nl range, dilutions, pre- or post-column reactions and sophisticated small-volume detections to be com- bined with, for example, capillary electrophoresis (CE). THEORY AND MINIATURIZATION Two approaches provide information on the be- haviour of a simple flow system when it is miniatur- ized: (1) a set of numerical values can be calculated, using standard formulae to give the order of magni- tude for a specific paramater; and (2) consideration of the proportionalities,
`
`lo-'
`
`i.e.
`
`m2/s (CE, LC) and IO-” m2/s (SFC); viscosities of the mobile phase, 10m3 Ns/mZ (CE, LC) and 5 IO-’ Ns/m2 (SFC); electrical conductivity of the mobile phase, 0.3 S/m (CE); electrical permittivity x zeta potential 5.6 x 10-i’ N/V (CE) Parameter Symbol (unit) CE (micellar) Capillary LC Capillary SFC Number of theoretical plates N 100 000 100000 100 000 Analysis time t(k’ = 5) (min) 1 1 1 Heating power PIL (W/m) 1.1 _ _ Capillary inner diater d (pm) 24 2.8 6.9 Capillary length L (cm) 6.5 8.1 20 Pressure drop AP (atm) _ 26 1.4” Voltage
`(kV) 5.8 _ - Signal bandwidth flX (mm) 0.21 0.56 1.4 Signal bandwidth 4, (ms) 42 70 70 Signal bandwidth oc. (Pl) 94 3.3 52 Ratio length/diameter of an eluting peak gz,ld ca. 10 cu. 200 ca. 200 Detection-volume requirements c,/2 (Pl) <47 < 1.6 ~26 Optical pathlength parallel to flow
`105 < 280 < 700 Optical pathlength perpendicular to flow d (pm) <24 <2.8 < 6.9 Response time requirements ~$2 (ms) <21 <35 <35 ’ The pressure needed to maintain the mobile phase in the supercritical state may exceed this value, e.g. for carbon dioxide the inlet and outlet pressure could be 75.4 and 74 bar, respectively.
`
`o,i2 b-4
`
`<
`
`AU
`
`Agilent Exhibit 1286
`Page 2 of 6
`
`

`

`h =
`
`d
`
`2/v l/d Constant d N = L/(d. h) L Lld Ll@ ECCU Constant Iid 11s U= E.L Lld L/8 I cc v &/L s d Constant cc IJ. I/(d’ . L) Constant 118 118 = lJ. I/L Constant
`ATccI= d2 Constant unit volume, the power per unit length, and the tem- perature difference generated in a steady-state ther- mal diffusion system. An example of a proportion- ality analysis according to approach 2 is shown in Table II. The miniaturization of a CE system is characterized by the inner diameter
`
`CE ON A CHIP 255 TABLE II EXAMPLE OF A PROPORTIONALITY ANALYSIS FOR CE The given miniaturization factors are d and L. Three arbitrarily chosen time dependencies are shown here. The remaining parameters are then calculated using the basic definition of d, L and time Parameter Diameter of capillary Length of capillary Time Linear flow-rate P&let number Reduced plate height Number of theoretical plates Electric field Applied voltage Electric current Power per volume Power per length Temperature difference Symbol L system d. L system dz L system d d d d L L L t L d.L 2-L u = L/t Constant l/d 118 vocu.d d Constant l/d
`of the capil- lary, by its length L and by the time t. Time cannot be set into a well defined relation with d or L, which means that we have one degree of freedom. Out of the numerous possible dependencies, a set of three have been chosen: time proportional to L (length), to d . L (area) and to d2 . L (volume), with power per unit volume, power per unit length and temper- ature difference as a constant, respectively. All of the remaining parameters of interest are then strict- ly based on d and L. The L system is characterized by a time scale forced into proportionality with L. For example, a ten-fold shorter capillary implies a ten-fold shorter retention time if the electric field strength (and the linear flow-rate) are kept constant. The number of theoretical plates must decrease. The influence of d is restricted to the current flow and the power gener- ated in the capillary. In the d . L system, the time scale depends on d multiplied by L. This implies that all the reduced variables used in capillary LC, e.g. P&let number or reduced plate height [9], are kept constant. In the same way the power per unit length, which is often used as a measure of thermal effects [lo], remains constant. In this case, an improvement of both sep- aration performance and analysis time can be achieved if d is miniaturized more drastically than L. For example, a ten-fold decrease in A and a five- fold decrease in L would double the number of the- oretical plates in lj5Oth of the retention time. Start- ing from a known CE experiment (d = 70 pm, L = 1 m, n = lo6 and t = 30 min), we would obtain 2 . lo6 theoretical plates within 36 s using a capillary of 20 cm x 7 pm I.D. The increase in plate number can be understood as arising from the increased electric field and from the decreased migration times, since longitudinal diffusion plays a major role. This system has been experimentally proven to be true by Monnig and Jorgenson [lo]. The d2 . L system is based on a constant temper- ature difference in the capillary when steady tem- perature diffusion is taken into account
`
`W
`
`It is far more optimistic than the d . L system. Experimen- tally, it has not yet been possible to prove this sys- tem to be valid at all.
`
`[l 11.
`
`Agilent Exhibit 1286
`Page 3 of 6
`
`

`

`256 A. MAN2 et ul. MICROMACHINING Originated by the microelectronics industry, the photolithographic patterning of layer structures on the surface of silicon wafers has become a well known and high-tech standard procedure. In addi- tion to its semiconductor qualities, monocrystalline silicon is abundant and inexpensive, can be pro- duced and processed controllably to extremely high standards of purity and perfection, has excellent mechanical and chemical properties (yield strength better than steel, Young’s modulus identical to steel, Knoop hardness comparable to quartz, chem- ical inertness comparable to glass) and is highly amenable to miniaturization (down into the mi- crometer range). The surface treatment to obtain mechanical structures is called micromachining [ 121 POUSHED Si WAFER THERMAL OXIDE DEPOSITION Si02 , PHOTORESIST PHOTORESIST DEPOSITION PHOTO- LlTHOGRAPHY DEVELOPING ETCHING THE OXIDE REMOVING THE PHOTORESIST ETCHING THE Si MICROSTRUCTURE LIGHT \ MASK STRUCTURED Si WAFER 1 I Fig. 1. Process steps of a standard one-mask micromachining procedure to etch a channel structure into silicon. and includes fabrication steps such as film deposi- tion, photolithography, etching and bonding. A simple process for obtaining a channel in silicon is shown in Fig. 1. It is obvious that the two-dimen- sional shape of the channel layout is given by the photomask, but the particular pattern does not af- fect the complexity of the process at all. As soon as a variation in depth (third dimension) or material (e.g. a metal layer) is needed, additional processes must be added to the sequence. There are mainly four different groups of processes: (1) Film deposition includes spin coating, ther- mal oxidation, physical vapour deposition (PVD) and chemical vapour deposition (CVD), low-pres- sure CVD, plasma-enhanced CVD, sputtering, etc. A large variety of metals, inorganic oxides, poly- mers and other materials can be deposited using these techniques. (2) Photolithography, a technique used to trans- fer a layout pattern from a mask onto a photosensi- tive film, can be done using visible light for struc- tures larger than 1 pm. For special applications such as submicron patterning, UV, X-ray or elec- tron beam lithography is used. (3) Etching is performed either as a wet chemical process or as a plasma process. Isotropic as well as anisotripic processes are known. (4) Bonding means the assembly of pieces of sil- icon onto silicon, glass or other substrates. The sub- ject of micromachining has been dealt with in great detail in many sources in the literature. For more information, see for instance ref. 13, which gives a good review of this huge field. Silicon-, quartz- and glass-based physical and chemical sensors and actuators are currently a focus of interest [14]. Compared with conventional ma- chining, photolithographic processes allow cheap mass fabrication of complicated microstructures. Hundreds to thousands of structures may be fab- ricated in the same batch. The precision and repro- ducibility of the structure elements are excellent. Silicon allows monolithic integration of electronics, sensors and actuators, but micromachining has to be done under clean-room conditions and needs high-tech instrumentation. However, in the last few years, the number of companies offering custom- made silicon, quartz and glass microstructures has significantly increased.
`
`Agilent Exhibit 1286
`Page 4 of 6
`
`

`

`MOBILE PHASE
`
`DETECTOR CELL
`
`257
`
`W.iSTE
`
`SAMPiE
`
`0
`
`fluorescein. Background electrolyte: 50 mM bo- rate, 50 mM Tris, pH 8.5,3000 V on 13 cm. Detection at 6.5 cm, fluorescence, excitation 490 nm, collection 520 nm, injection through side channel, 500 V for 30 s. other 10 ,um by 1 mm, meet at one point. The in- tersection has a volume of 9 pl, which means that no extra dead volume exists. The cross-section of the channels is not exactly rectangular, but rounded at the corners (compare with refs. 17-19). Detection was done using a laser fluorescence set-up similar to the one described previously [20,21] located some- where downstream from the electrophoresis capil- lary (O-135 mm after the point of injection). The buffer reservoirs containing the platinum electrodes were pipette tips mounted directly into the drilled holes at the ends of the channels. To set the experiment up, the background elec- trolyte and the fluorescent sample mixture were both driven past the injection and detection points by externally applied voltages, allowing positioning of the detector. The carrier electrolyte was then dri- ven through the electrophoresis channel to flush it out. A 30-s pulse of 500 V applied to the sample channel provided the injection. To run the electro- pherogram, 3000 V were applied to the carrier elec- trolyte and the separation capillary (200 V/cm). The resulting separation of two fluorescent dyes is shown in Fig. 3. For calcein, a performance of 18 000 theoretical plates has been obtained. The height equivalent to a theoretical plate is 3.6 pm, which is comparable to a non-optimized standard CE experiment. The standard deviation of the peak in terms of time, length and volume was 1.4 s, 0.49 mm and 145 pl, respectively. CONCLUSIONS Consideration of hydrodynamics and diffusion processes indicates faster and more efficient chro- matographic separations, faster electrophoretic sep- arations and shorter transport times for miniatur- ized TAS. The consumption of carrier, reagent or electrophoresis buffer is dramatically reduced. Mi- cromachining, especially photolithographic processes, offers access to novel analytical micro- structures such as branched-channel systems having no dead volume. The access to electrophoretic separations within a planar glass structure shown here is a first step to- wards an integrated microflow system using CE to- gether with injection, sample pretreatment and post-column reactions, etc. We anticipate that high- er voltages can be applied to the structure to speed
`\ ELECTROPHORE.SIS II’llECTlON CAPILLARY Fig. 2. Glass microstructure for injection and CE. Size 15 x 4 cm. Electrophoresis channel 30 x 10 pm. The external laser fluo- rescence detector was positioned 6.5 cm from the point of in- jection. CAPILLARY ELECTROPHORESIS ON A CHIP We have made six different structures in silicon (covered by Pyrex glass) [ 151 and one in amorphous glass [ 161. The silicon structures, even with state-of- the-art insulating films (SiOZ and Si3N4), exhibited poor voltage breakdown characteristics. In the best case, 950 V could be applied on a single device for a few minutes. Better results have been obtained with amorphous glass or quartz. The photolithographically fabricated glass device shown in Fig. 2 was used to perform a first CE ex- periment “on a chip”. The overall size of the struc- ture is 150 X 40 X 10 mm. It consists of two glass plates, one of them containing the etched channels and the other serving as a cover. The three channels, two of them being 10 pm deep and 30 pm wide, the I
`
`1 2 3 4 5 8 7 3 min Fig. 3. CE separation of two fluorescent dyes. Sample: 20
`
`pM
`
`pM
`
`Agilent Exhibit 1286
`Page 5 of 6
`
`CE ON A CHIP
`calcein, 20
`

`

`258 A. MANZ et al. up the separation and to increase its performance. Silicon structures would show a problem at high applied voltages. The reasonable range might be up to 200 V. This, of course, implies a poor efficiency (small number of theoretical plates), but relatively short retention times. 9 A. Manz and W. Simon, Anal. Chem., 59 (1987) 74. 10 C. A. Monnig and J. W. Jorgenson, Anal. Chem., 63 (1991) 802. 11 12 13 REFERENCES 14 I H. M. Widmer, J. F. Erard and C. Grass. Int. J.Environ. Anal. Gem., 18 (1984) 1. 2 M. Garn, P. Cevey, M. Gisin and C. Thommen, Biotechnol. Bioeng., 34 (1989) 423. R. B. Bird, W. E. Stewart and E. N. Lightfoot. Transport Phenomena, Wiley, New York, 1960, pp. 267--211. K. E. Petersen, Proc. IEEE, 70 (1982) 420. W. H. Ko and J. T. Suminto, in W. Gopel, J. Hesse and J. N. Zemel (Editors), Sensors, a Comprehensive Surve_v, Vol. 1, VCH Weinheim, 1989, pp. 107-168. Proceedings of‘ Transducers ‘84; Sens. Actuators, A21-A23 (1990) and Sens. Actuators, B 1 (1990); Proceedings y/ Trans- ducers ‘91, Digest of‘ Technical Papers, IEEE, Piscataway, NJ, 1991. 15 3 A. Manz, N. Graber and H. M. Widmer, Sens. Actuators, Bl (1990) 244. 16 D. J. Harrison, A. Manz and P. G. Glavina, in Transducers ‘9 I; Digest q/ Terhnicul Papers, IEEE, Piscataway, NJ, I99 1, p. 792. A. Manz, D. J. Harrison. J. C. Fettinger, E. Verpoorte, H. Liidi and H. M. Widmer, in Transducers ‘91; Digest
`Tech- nical Papers, IEEE, Piscataway NJ, 1991, p. 939. T. Tsuda, J. V. Sweedler and R. N. Zare. Anal. Chem.. 62 (1990) 2149. M. Jansson, A. Emmer and J. Roeraade, J. High Resolut. Chromatogr. Chromatogr. Commun., 12 (1989) 797. J. F. Brown and J. 0. N. Hinckley. J. Chromatogr., 109 (1975) 225. J. W. Jorgenson and K. D. Lukacs, Sciencr i Washington. D.C.), 222 (1983) 266. E. Gassmann, J. E. Kuo and R. N. Zare, Science (Washing- ton, D.C./. 230 (1985) 813. E. Verpoorte, A. Manz, H. Liidi and H. M. Widmer, in Transducers ‘91. Digest of Technical Papers, IEEE, Piscata- way, NJ, 1991, p. 796. A. Manz, J. C. Fettinger, E. Verpoorte, H. Ltidi, H. M. Widmer and D. J. Harrison, Trends Anal. Chem., 10 (1991) 144. E. Verpoorte, A. Manz, H. Liidi, A. E. Bruno, F. Maystre, B. Krattiger, H. M. Widmer, B. H. van der Schoot and N. F. de Rooij, Sens. Actuators, submitted for publication. S. C. Terry, J. H. Jerman and J. B. Angell, Z_/X_E Trans. Electron. Devices, ED-26 (1979) 1880. A. Manz, Y. Miyahara, J. Miura, Y. Watanabe, H. Miyagi and K. Sato, Sens. Actuutors, Bl (1990) 249. 17 18 19 20 21
`
`of
`
`Agilent Exhibit 1286
`Page 6 of 6
`
`