During the past decade, so-called microarray-basedbiochips and lab-on-a-chip microfluidic devices havefound increasing use in biotechnology for drug discov-ery, DNA analysis, high-throughput screening of pa-tients, and combinatorial syntheses.1–10 Potentialbenefits of introducing miniaturized analytical sys-tems include improved accuracy, multiplexing, lowersample consumption, disposability, and volume reduc-tion. In situations when chemical reactions are to beperformed and the amount of material is limited orcostly, volume reduction is advantageous because highconcentrations can easily be maintained. Multiplex-ing �parallelism� allows for improved throughput and
When this research was performed H. Blom �[email protected]�was with the Department of Microelectronics and InformationTechnology, Royal Institute of Technology, Electrum 220, SE-16440 Kista, Sweden. M. Johansson and S. Hård were with theDepartment of Microelectronics, Chalmers University of Technol-ogy, SE-412 96 Gothenburg, Sweden. M. Johansson is now withOptronic Consulting, Svangatan 2B, SE-416 68 Gothenburg, Swe-den. M. Gosch and R. Rigler were with the Department of Med-ical Biophysics, Karolinska Institutet, SE-171 77 Stockholm,Sweden. T. Sigmundsson and J. Holm were with ACREO ABElectrum 236, SE-164 40 Kista, Sweden. T. Sigmundsson is nowwith Åmic Production, AB, Dag Hammarskjoldsvagen. 52B, SE-751 83 Uppsala, Sweden. J. Holm is now with IBSEN Microstruc-tures A�S, Gammelgaardsvej 65, DK-3520 Farum, Denmark.
thereby for more-cost-effective analyses, as large num-bers of experiments can be run in parallel. Today’sbiochip systems feature multiplexing numbers thatreach from a few �as low as 4 � 4� to tens of thou-sands11 or even hundred of thousands,12 whereas lab-on-a-chip multiplexing numbers can reach into thehundreds.8–10 Laser-induced fluorescence is oftenused as the methodology for reading out data frommany of these miniaturized systems. The laser in-duced fluorescence technique has ultrahigh detectabil-ity and can permit analysis even at the single-molecule�monochromophore� level. Single-molecule measure-ments are at present accomplished by scanning of aconfocal laser focus across the miniaturized systems.To allow for a higher detection speed in microarray-based biochips we have improved the laser-induceddetection methodology and achieved multiplexing�parallelism� without sacrificing ultrahigh sensitivityand high spatial resolution.13 The method permitsanalyses at the single-molecule level14 and also allowsreal-time dynamical analyses to be made.15 The prin-ciple of the method is based on laser induced fluores-cence, but it is combined with correlation analyses andwith multifocal generation by a diffractive optical fan-out element. This combination allows us quickly andeasily to make dynamic measurements in parallel on amicroarray biochip. In an attempt to contribute tothe lab-on-a-chip research field of microfluidics, wepresent a fluidic characterization method that can beused to measure flow properties in parallel at severaldifferent places inside small microchannels. Themethod can easily be applied to multispatial detection
in many separate channels, too. This characteriza-tion method is also based on correlation analysis andmultifocal generation by a diffractive optical fan-outelement. Several authors have already explored theuse of multifocal arrangements with fluorescence cor-relation analysis for microfluidic measurements.16–19
The measurements were previously always accom-plished by use of only two laser foci, which were gen-erated by numerous mirrors, beam splitters, andprisms. In addition to a somewhat large optical ta-bletop area, these approaches also demand delicatefine adjustment and control of all optical parts thatwere used to generate the two foci. These drawbackshave probably deterred further attempts to developmore foci. To allow for a larger number of foci, thusreaching a higher degree of multiplexing, one can takeadvantage of multifocal generation by diffractive opti-cal elements �DOEs�. The so-called DOE manipu-lates light through modulation of the phase of animpinging laser beam, and the result is a diffractiveredistribution of the laser intensity.20 If the desiredintensity of a diffraction pattern in an illuminated areais much larger than a diffraction-limited spot, the DOEis referred to as a beam shaper. An example of beamshaping is the redistribution of the Gaussian intensityprofile from a normal laser into a rectangular intensityprofile.21 Fan-out is the term that one uses when theDOE creates an illuminated area with a number ofdiffraction-limited spots.22 The major advantage ofthe DOE is that it is small and compact and thereforeobviates the necessity of using a large optical tabletoparea and optical fine adjustment to generate multiplefoci from a single laser beam. Through proper designand fabrication the so-called fan-out DOE can give alarge number of foci arranged in many patterns. Fab-ricated fan-outs have previously exhibited some prob-lems in achieving uniform intensity among the foci,23
but a newly developed design algorithm that over-comes first-order fabrication errors can give uniformintensity in all foci.24 The algorithm has been appliedto the fabrication of the 4 � 1 fan-out DOE used in thisstudy. In what follows, we describe the theory of flowstudies by fluctuation analysis through autocorrela-tion and cross correlation, the design of the DOE andthe microstructure, our multifocal diffractive opticalfluorescence correlation spectroscopy setup, the probeand its loading, and finally all experimental results ofthe microfluidic studies made inside the microstruc-tures.
2. Theory: Fluorescence Correlation Analysis
A concise method of fluctuation analysis that canexpress the degree to which two dynamic propertiesare correlated over a period of time is given by thecorrelation functions. Fluorescence correlationanalysis is based on observing the intensity fluctua-tions of individual fluorescent molecules moving inand out of a small open volume element, defined bythe excitation light’s intensity distribution and opti-cally monitored by a detector.25–29 The normalizedcross-correlation function between two fluorescent
detection signals that results from two spatially dif-ferent laser foci can be defined as18
gc��� � 1 ��� P1�r� P2�r��G�r, r�, ��drdr�
C 2 � P1�r�dr � P2�r��dr�
. (1)
P�r� is the combined distribution of the collection ef-ficiency function and the excitation intensity from aposition r � �x, y, z� as seen by the detector. Thedistributions are approximated by three-dimensionalGaussian distributions,29 � �x, y�,
P1 � P01 exp�2� 2
w02��exp�2
z2
z02� ,
P2 � P02 exp�2��� � R�2
w02 ��exp�2
z�2
z02� ,
where w0 and z0 are the radius and the height of thevolume element seen by the detector and R � �Rx, Ry�is the vector from the center of one volume element tothe other. Because the volume elements are all inthe same plane, R � ez � 0, where ez is the unit vectorin the z direction. G�r, r�, �� � ��C�r, t��C�r�, t ���is the concentration cross-correlation function anddescribes the concentration fluctuations of the fluo-rescent molecules, �C�r, t� � C�r, t� C, about themean concentration, C� . Concentration fluctuationsthat are due to diffusion and flow of individual mol-ecules are determined by the hydrodynamic differen-tial equation d�C�r, t��dt � D�2�C�r, t� V�r� ���C�r, t�, where D is the diffusion coefficient of thefluorescent molecules and V�r� is the vectorial flowvelocity. Solving the hydrodynamic differentialequation28 and then performing the integrations ofnormalized cross correlation result in the followinganalytical expression18:
gc��� � 1 �1N ��1 �
�
�D��1 �
w02
z02
�
�D�1�2�1
� exp�1
w02
�V� � R�2
1 � ����D�
�1
z02
Vz2�2
1 � �w02��z0
2�D�� . (2)
In Eq. �2�, 1�N is the amplitude of the correlationcurve and gives the mean number of molecules in theexcitation volume, and �D is the characteristic diffu-sion time of the molecules. In small microstructuresthe flow can be considered laminar, and thez-dependant flow in the cross-correlation expressioncan be neglected. The second term in the exponen-tial can therefore be dropped. If the shift betweenthe different volume elements R goes to zero, thisexpression reduces to the familiar analytical autocor-relation function for diffusion and an applied flow.28
This autocorrelation function is suitable for evaluat-ing and characterizing flow speed down to �0.5 mm�s.16,18,30 If no flow is applied and the shift betweenthe different volume elements goes to zero, then theexpression reduces to the analytical autocorrelationexpression for pure diffusion.25 The cross-correlation flow time, �F � �R���V�, can be defined bythe ratio of the distance between the foci to the two-dimensional flow velocity. However, in our case theflow can be written as V � Vy�x, z�, with the laserbeam propagating along the z axis and the micro-structure and the four foci from the 4 � 1 fan-outDOE oriented in the y direction. These simplifica-tions then give the following expression:
gc��� � 1 �1N ��1 �
�
�D��1 �
w02
z02
�
�D�1�2�1
� exp�Vy
2
w02
�� � �F�2
1 � ����D�� . (3)
This cross-correlation function is suitable for evalu-ating and characterizing flow speed down to �0.1mm�s.18 To the two cases discussed above, we mayadd that the part that describes the pure diffusion�nonexponential term� can be neglected if the flowvelocity is high �above 1 mm�s�.
3. Experimental Designs
A. Diffractive Optical Element
The 4 � 1 fan-out DOE �see Fig. 1� was designed witha newly developed iterative Fourier-transform algo-rithm24 and fabricated by direct-write electron-beamlithography �JEOL, JBX-5DII� with an accelerationvoltage of 50 kV and an electron-beam current of 3nA. No compensation for the proximity effect wasmade. Resist �PMGI, SF15� was spin coated onto aquartz substrate to a thickness of 2 �m. Before theexposure, a 20-nm-thick Ni�Cr layer was evaporatedon top of the resist to prevent charge-up of the resistlayer during the exposure. Sixteen electron-doselevels were used. After the exposure the Ni�Crlayer was removed and the resist developed stepwiseto a maximum depth of 1.0 �m, corresponding to aphase modulation of 2� for the wavelength of 532 nmthat was used. Measurements showed that the in-tensity uniformity, i.e., the difference between thespot with maximum light power and the spot withminimum light power, was �3% and that 78% of theincident laser light was found in the four desiredfan-out foci.
B. Microstructures
All microstructures �see Fig. 1� used in the paralleldetection experiments were fabricated by Acreo ABby standard micromachining technologies, which in-clude thin-film deposition, photolithography, dryetching, and wafer bonding. In the first step, differ-ent square-sized microchannels �5, 10, 20, and 50 �m�and round 1-mm diameter connection holes were pat-terned onto oxidized 300-�m-thick silicon wafers
�Topsil, Frederikssund, Denmark� by photoresist andoptical lithography. The patterned oxide maskswere subsequently dry etched by reactive ion etching.Deep reactive ion etching was thereafter used to etchmicrochannels and connection holes with verticalsidewalls into the silicon wafers; 170-�m thick Pyrex7740 glass wafers �Corning, Danville, Va.� were an-odically bonded to the silicon wafers as a lid. Thewafers were finally sawed into 22 mm � 9 mm mi-crostructures.
C. Apparatus
The experimental setup for parallel excitation anddetection is displayed in Fig. 2. A single-line �532-nm, 50-mW� output of a diode-pumped solid-stategreen laser �Kimmon DPSS Laser, HK-5526� was ex-panded six times by a beam expander �OWIS, f � 25mm and f � 150 mm� to fully illuminate the diffrac-tive optical element and the backaperture of the mi-croscope objective �radius, 4 mm�. The output powerfrom the laser was adjusted to 1.5 mW by transfer ofa neutral-density filter �optical density, 0.5–4� intothe laser beam. With the help of two plano–convexlenses � f � 150 mm� and the diffractive optical ele-
Fig. 1. Scanning-electron microscopy pictures showing a 10 �m �10 �m square microstructure with a superimposed fluorescencepattern from the 4 � 1 diffractive optical fan-out element. Thefoci in the left-hand pattern numbered 1–4 from top to bottom inthe text, and the foci in the right-hand pattern are numbered 1–4from left to right.
Fig. 2. Schematic illustration of the setup used for parallel flowmeasurements in microstructures with a multifocal 4 � 1 diffrac-tive optical designed fan-out element. OD, optical density.
ment, the expanded beam was focused and recolli-mated before it was sent into a microscope �Olympus,IX-70�. The function that the fan-out DOE performsin this collimator setup is to multiplex the incomingGaussian converging spherical beam into four iden-tical and laterally displaced Gaussian convergingbeams. In the focal plane of the collimator unit, thelateral size of the foci that were generated was cal-culated to be 51 �m, determined by the focal length ofthe first plano–convex lens, the wavelength, and theincoming beam radius. The multifocal recollimatedexcitation light was finally reflected by a dichroicmirror �Chroma, 565LP� into a 60� water immersionobjective �Olympus, UplanApo, NA 1.2�, corrected tothe thickness of a 170-�m coverslip. Over the objec-tive, either a cover glass or a microstructure could beattached to a self-manufactured holder built on top ofthe x–y movable microscope table. The fluorescenceemission from the parallel detection element was col-lected by the same objective and passed through abandpass filter �Chroma, HQ585�40� that discrimi-nates against Rayleigh and Raman scattered light.The fluorescence emission was further magnified�1.5�� with a lens in the Olympus IX70 microscopeand then focused onto four 62.5-�m multimode fibers�Ribbon� situated in a commercial fiber bundle con-tact �MPO� with 250-�m spacing. A translator builtin house that held the multimode fiber ferules per-mitted spatial fine adjustment of the fiber bundle.The output of the fibers was connected to four ava-lanche photodiodes �EG&G Model SPCM-100�, andthe photoinduced transistor–transistor logic pulsesfrom the detectors were passed to two PC-based cor-relator cards �ALV-5000E� that displayed the fluores-cence photon bursts and calculated and displayed theautocorrelation and the cross-correlation functioncurves. The autocorrelation curves were analyzedby a computer program written in house and based ona nonlinear least-squares minimization algorithmthat can specify, among other things, the number ofmolecules and the diffusion times of the molecules.The cross-correlation curves were analyzed with aself-written fitting program that used the built-inminimization Levenberg–Marquardt functions of theOrigin 6.0 data analysis software program.
D. Probe and Loading
The fluorescent probe used in the measurementswere tetramethylrhodamine attached nucleotides,TMR-4-dUTP �Amersham Pharmacia Biotech�.The probes were highly diluted �1010 M� in high-performance liquid chromatography–graded water be-fore the fluorescence measurements were made.During the experiments the microstructure was fixedin the self-manufactured holder connected with poly-ethylene tubing connections ��0.58 mm, L � 100 cm�.Loading and hydrodynamic flow generation in themicrostructures were then obtained by use of a nee-dle ��0.25 mm� attached in the tubing and by eleva-tion of a probe-filled syringe ��12 mm�.
4. Measurements and Discussion
A. Calibrations
In a droplet experiment performed with a 0.3-nMTMR-4-dUTP solution on top of a cover glass, thetime required for free diffusion through the four fan-out laser foci was measured. The free diffusion timewas deduced as �diff � 0.075 ms. The lateral radiusof the 4 � 1 fan-out volume element was finally cal-culated to be 0.4 �m, with an error of �5% that is dueto uncertainties in the reference diffusion constantand the measured diffusion time.29 This calibrateddiffusion time and the volume element radius werefixed in the following autocorrelation flow experi-ments. With prior knowledge of the laser foci ra-dius, we could calculate the flow velocity simply bydividing the radius by the flow time.
B. Autocorrelation Analyses
Figure 3 shows parallel experimental autocorrelationcurves from fluorescent TMR-4-dUTP molecules inthe four fan-out foci inside a 20 �m � 20 �m squaremicrochannel. The fan-out was here oriented per-pendicularly to the sidewalls of the microchannel�pattern at the left in Fig. 1�, with focus 1 situated onthe sidewall at approximately middle height in thechannel. A continuous hydrodynamic flow was ob-tained from a syringe reservoir that contained ahighly diluted TMR-4-dUTP solution set 80 cm abovethe microstructure. By applying the analytical au-tocorrelation function described in Section 2, with anassumed zero distance for the shift between the dif-ferent volume elements,28 we evaluated the experi-mental curves. From the fit, the average numbers ofmolecules within the volume elements were deducedto be N4 � 10.2, N3 � 10.0, N2 � 8.3, and N1 � 11.1,as were the flow times, �Flow4 � 0.037 ms, �Flow3 �0.042 ms, �Flow2 � 0.055 ms, and �Flow1 � 0.19 ms, allwithin errors of �5%. One property of the curvesthat we can observe directly �without analytical fit-ting� by looking at Fig. 3 is that the flow time wasshortest in focus 4. This focus was situated in thecentral part of the microchannel. The flow time
Fig. 3. Parallel autocorrelation curves of a TMR-4-dUTP solutioninside a 20 �m � 20 �m square microchannel. Curves from leftto right: focus 4 �solid�, focus 3 �stars�, focus 2 �crosses�, and focus1 �solid�.
then increased toward the sidewall �foci 3, 2, and 1�,as it should for a laminar hydrodynamic flow gradient�Hagen-Poiseuille�.30 Another property, in additionto its much longer flow time, is the slightly differentcurve shape of focus 1 from the other foci. The rea-son for this phenomenon is most likely the interactionbetween a fluorescent molecule and the wall. Thewall acts as protection against photodestruction, andthe molecule can live longer than normally.31
Therefore the autocorrelation curve’s shape ischanged �toward a flatter profile� for shorter time, asthis time range describes the photophysical reactionsof molecules. A third property that can be observedfrom Fig. 3 is the amplitudes of the curves. As theflow is fastest in the middle, more molecules can passfocus 4 �for a fixed time� than the other foci, and theamplitude of focus 4, 1�N, is therefore lower than forfoci 3 and 2. The amplitude should also be increasedat the wall, as fewer molecules per time pass focus 1,but such is not the case. The wall may to someextent distort the shape32 of this focus and make itbigger such that more molecules are detected at everyinstance of time. The lower amplitude may also bethe result of an increased number of molecules thatslowed down or got stuck on the wall, which may alsoexplain the small fraction of longer diffusion timesseen from 0.4 to 1 ms in Fig. 3. A combination of allthe effects probably explains focus 1’s lower ampli-tude.
After our initial microfluidic application of the 4 �1 fan-out DOE, we measured, in parallel, flow profilesat various flows inside the microstructure. We ap-plied these flows by moving the reservoir with aTMR-4-dUTP solution in steps of 5 cm from 40 to 85cm. The 4 � 1 fan-out foci were placed in the middleof the 20-�m microchannel. Figure 4 illustratesthese measurements, which were made after the flowtimes had been transformed into flow velocities, byuse of the autocorrelation curves of each focus atevery reservoir height. The measurements showhyperbolic profiles caused by the flow gradient di-rected toward the center of the microchannel. Byscanning the channel with the diffraction-limited la-ser foci one could also, in a simple and precise man-
ner, perform high-spatial-resolution flow profiling ofthe microstructure. The scanning method can de-liver a full two-dimensional flow profile withmicrometer-sized resolution in the microchannelquickly and easily.30 Using the 4 � 1 fan-out DOEwould even further improve the speed of the method.The parallel measurement procedure shown in Figs.3 and 4, could also be used, for example, to measureflow profiles and flow gradients at the interface of twomixing fluids inside a microfluidic device.33
C. Cross-Correlation Analyses
Turning the microstructure, thereby causing all fourfoci to be parallel to the channel walls �at the right inFig. 1�, made cross-correlation flow analyses possible.The syringe reservoir was moved in steps of 5 cmfrom 90 to 10 cm, as the fluorescent signals from foci1 and 2 were cross correlated. These measurementsare shown in Fig. 5. The flow times were also mea-sured, simultaneously, between foci 3 and 4, and thatmeasurement showed the same results. All mea-sured data of the flow times could be fitted to thecross-correlation equation deduced in Section 2.However, as the flow is relatively high �greater than1 mm�s�, and the curves show distinct maxima, thedata can be evaluated without a fitting procedure ifflow times are the only desired information. Theflow times can be determined directly as the timeduring which the maximum of each curve is situated.Such a reading results in flow times of 0.18 ms for the90-cm reservoir height and 2.8 ms for the 10-cmheight. To convert to flow speed, one simply dividesthe different flow times by the foci spacing �R � 2.78�m�. From Fig. 5 it is also possible to deduce di-rectly the pure diffusion between the centers of thefoci. Connecting the peaks of all cross-correlationcurves leads to the autocorrelation function of diffu-sion between these centers. As we look at Fig. 5,pure diffusion becomes increasingly more importantwith decreased flow speeds, as larger numbers of mol-ecules leaving focus 1 may miss focus 2. The ampli-tudes of the curves in Fig. 5 are therefore decaying.Their shapes are also influenced by the diffusion as
Fig. 4. Parallel flow profiling inside the 20-�m microchannel atseveral reservoir heights in 5-cm steps from 85 cm �top� to 40 cm�bottom�.
Fig. 5. Cross-correlation curves between foci 1 and 2 inside the20-�m microchannel, with the reservoir height moved in steps of 5cm, from 90 cm �left� to 10 cm �right�.
increased random movement broadens the curves.By using cross-correlation it is also possible to mea-sure the direction of the flow. The microstructure orthe fan-out DOE, together with the fiber detector, canbe turned in small steps to map out the angle of anyflow component.18
In our final application of the 4 � 1 fan-out weswitched the output of the detectors and cross-correlated all four foci. That made it possible tostudy transportation effects of diffusion as a functionof flow and distance. In Fig. 6 we show such a mea-surement: Foci 1 and 2, 1 and 3, and finally 1 and 4are cross correlated at a reservoir height of 90 cm.As expected, the flow time increases from 0.18 ms �cf.Fig. 5� to 0.37 ms to 0.56 ms, because the moleculesnow have to move two and three times longer, respec-tively, than in the previous measurement. The am-plitudes tell us how many of the molecules thatstarted out from focus 1 and went into focus 2 madeit first to focus 3 and then also finally to focus 4. Thedata analyzed here show that 89% of all moleculesmade it to focus 3 �the same number also made it fromfocus 2 to focus 4� and that �80% went all the way tofocus 4. With this parallel detection method singlefluorescent molecules could be followed on their waythrough a miniaturized analytical system. Themethod with its high spatial resolution might also beused to follow biomolecules inside a cell. Moving theDOE �cf. Fig. 2� makes possible even smaller or biggerdistances among the four foci in the present setup.However, our fiber holder does not permit this extradegree of freedom today. Parallel excitation and de-tection of as many microchannels or ports as thereare on lab-on-a-chip microfluidic devices seems to bea natural goal. A large increase in the number ofspots generated by the diffractive optical element issimple with present-day fabrication techniques.34 Itis even possible to integrate multifocal generationinto the surfaces35 of microstructures and therebyhave a chance to miniaturize future analysis systemseven further. A parallel detector arrangement, with10 � 1 or even 100 � 1 multimode fibers connected tosingle-photon avalanche photodiodes, is possible by
use of commercially available fiber bundles. Withsuch an arrangement the flow profile inside a micro-structure could be measured in one single shot withhigh spatial resolution. The cost of the detectorswill be the limiting factor here, as the cost of amultichannel correlator does not increase much com-pared with that of a two-channel version. Discus-sions with manufacturers of single-photon avalanchephotodiodes and also with representatives of aca-demia with a view to improving detector arrays havebeen initiated. Combining such detector arrayswith present-day diffraction fan-out elements �sever-al hundreds of foci� may in the future lead to evenfurther multiplexed analysis of microfluidics in lab-on-the-chip.
5. Conclusion
We have developed a multifocal diffractive opticalfluorescence correlation spectroscopy system for par-allel flow analyses of tetramethylrhodamine labeledbiomolecules in microstructures. Multifocal excita-tion was made possible through the use of a 4 � 1fan-out diffractive optical element that produces uni-form intensity in all four foci. Fiber-coupled ava-lanche photodiodes and two PC-based correlationcards were used for parallel detection of the autocor-relation or cross correlation analyses or both of thefluorescence emission. Parallel flow analyses insidea 20 �m � 20 �m square microchannel, with the fourfoci perpendicular to the channel walls, made it pos-sible to monitor different flows and flow profiles in allfour foci simultaneously. We were able to makecross-correlation flow analyses by turning the micro-structure, thereby having all four foci parallel to thechannel walls. Transport effects of the diffusion as afunction of flow and distance could then be studiedtoo. Future transport effects, flow monitoring, flowprofiling, and prolonged fluorescence molecule detec-tion possibilities, in artificial microstructures or incells, can benefit from this parallel fluorescence de-tection principle.
This study was supported by grants from the NyaKomponenter Och Funktionella Material for Mor-gondagens Naringsliv program of the SwedishNational Board of Industrial and Technical Develop-ment, the Swedish Foundation for Strategic Researchto Andreas Plucktun and Rudolf Rigler, the HumanScience Research Frontier Program, and the SwedishScience Council. We acknowledge expert workshopassistance from Lennart Wallerman and stimulatingdiscussions with and proofreading of the manuscriptby Gunnar Bjork.
References1. M. Schena, D. Shalon, R. W. Davis, and P. O. Brown, “Quan-
titative monitoring of gene expression patterns with a comple-mentary DNA microarray,” Science 270, 467–470 �1995�.
2. D. Whitecomb, C. R. Newton, and S. Little, “Advantages inapproaches to DNA-based diagnostics,” Curr. Opin. Biotech-nol. 9, 602–608 �1998�.
3. N. S. Gray, L. Wodicka, A.-M. W. H. Thunnissen, T. C. Nor-man, S. Kwon, F. H. Espinoza, D. O. Morgan, G. Barnes, S.
Fig. 6. Cross-correlated flows at a reservoir height of 90 cm.Left to right: foci 1–2, foci 1–3, foci 1–4.
LeClerc, L. Meijer, S.-H. Kim, D. J. Lockhart, and P. G.Schultz, “Exploiting chemical libraries, structure, and genom-ics in the search for kinase inhibitors,” Science 281, 533–538�1998�.
4. M. J. Marton, J. L. DeRisi, H. A. Bennett, V. R. Iyer, M. R.Meyer, C. J. Roberts, R. Stoughton, J. Burchard, D. Slade, H.Dai, D. E. Bassett, L. H. Hartwell, P. O. Brown, and S. H.Friend, “Drug target validation and identification of secondarydrug target effects using DNA microarrays,” Nature Med. 4,1293–1301 �1998�.
5. L. C. Usich-Wilson, X.-D. Xiang, and P. G. Schultz, “Lessonsfrom the immune system: from catalysis to materials sci-ence,” Acc. Chem. Res. 29, 164–170 �1996�.
6. D. J. Lockhart and E. A. Winzeler, “Genomics, gene expressionand DNA arrays,” Nature 405, 827–836 �2000�.
7. D. R. Walt, “Bead based fiber-optical arrays,” Science 287,451–452 �2000�.
8. A. van den Berg and P. Bergveld, eds., Micro Total AnalysisSystems �Kluwer Academic, Dordrecht, The Netherlands,1994�.
9. D. J. Harrisson and A. van den Berg, eds., Micro Total AnalysisSystems ’98 �Kluwer Academic, Dordrecht, The Netherlands,1998�.
10. A. van den Berg, W. Olthuis, and P. Bergveld, eds., Micro TotalAnalysis Systems 2000 �Kluwer Academic, Dordrecht, TheNetherlands, 2000�.
11. D. G. Wang, J.-B. Fang, C.-J. Siao, A. Berno, P. Young, R.Sopolsky, G. Ghandour, N. Perkins, E. Winchester, J. Spencer,L. Kruglyak, L. Stein, L. Hsie, T. Topaloglou, E. Hubbel, E.Robinson, M. Mittmann, M. S. Morris, N. Shen, D. Kilburn, J.Rioux, C. Nusbaum, S. Rozen, T. J. Hudson, R. Lipshuts, M.Chee, and E. S. Lander, “Large-scale identification, mapping,and genotyping of single-nucleotide polymorphisms in the hu-man genome,” Science 280, 1077–1082 �1998�.
12. R. J. Lipshutz, S. P. Fodor, T. R. Gingeras, and D. J. Lockhart,“High density synthetic oligonucleotide arrays,” Nature Genet.21, 20–24 �1999�.
13. H. Blom, M. Johansson, A.-S. Hedman, L. Lundberg, A. Han-ning, S. Hård, and R. Rigler, “Parallel fluorescence detection ofsingle biomolecules in microarrays using diffractive opticaldesigned 2 � 2 fan-out element,” Appl. Opt. 41, 3336–3342�2002�.
14. R. A. Keller, W. P. Ambrose, P. M. Goodwin, J. H. Jett, J. C.Martin, and W. Ming, “Single-molecule fluorescence analysisin solution,” Appl. Spectrosc. 50, 12A–32A �1996�.
15. R. Rigler and E. S. Elson, eds., Fluorescence CorrelationSpectroscopy—Theory and Applications �Springer-Verlag, Ber-lin, 2001�.
16. M. Brinkmeier and R. Rigler, “Flow analysis by means of flu-orescence correlation spectroscopy,” Exp. Tech. Phys. 41, 205–210 �1995�.
17. K. Q. Xia, Y. B. Xin, and P. Tong, “Dual-beam incoherentcross-correlation spectroscopy,” J. Opt. Soc. Am. A 12, 1571–1578 �1995�.
18. M. Brinkmeier, K. Dorre, J. Stephan, and M. Eigen, “Two-beam cross-correlation: a method to characterize transportphenomena in micrometer-sized structures,” Anal. Chem. 71,609–616 �1999�.
19. D. J. LeCaptain and A. Van Orden, “Two-beam fluorescence
cross-correlation spectroscopy in an electrophoretic mobilityshift assay,” Anal. Chem. 74, 1171–1176 �2002�.
20. M. R. Taghizadeh, P. Blair, B. Layet, I. M. Barton, A. J. Wad-die, and N. Ross, “Design and fabrication of diffractive opticalelements,” Microelectron. Eng. 34, 219–242 �1997�.
21. S. C. Holswade and F. M. Dickey, “Gaussian laser beam shap-ing: test and evaluation” in Current Developments in OpticalDesign and Engineering VI, R. E. Fischer and W. J. Smith,eds., Proc. SPIE 2863, 237–245 �1996�.
22. J. Bengtsson, “Design of fan-out kinoforms in the entire scalardiffraction regime with an optimal-rotation-angle method,”Appl. Opt. 36, 8435–8444 �1997�.
23. J. A. Cox, B. S. Fritz, and T. R. Werner, “Process error limi-tations on binary optics performance,” in Computer and Opti-cally Generated Holographic Optics, 4th in a Series, I. Cindrichand S. H. Lee, eds., Proc. SPIE 1555, 80–88 �1991�.
24. J. Bengtsson and M. Johansson, “Fan-out diffractive opticalelements designed for increased fabrication tolerances to lin-ear depth errors,” Appl. Opt. 41, 281–289 �2002�.
25. D. Magde and E. L. Elson, “Fluorescence correlation spectros-copy. I. Conceptual basis and theory,” Biopolymers 13,1–27 �1974�.
26. M. Magde, E. L. Elson, and W. W. Webb, “Fluorescence corre-lation spectroscopy. II. Experimental realizations,”Biopolymers 13, 29–61 �1974�.
27. M. Ehrenberg and R. Rigler, “Rotational Brownian motion andfluorescence intensity fluctuations,” Chem. Phys. 4, 390–401�1974�.
28. M. Magde, W. W. Webb, and E. L. Elson, “Fluorescence corre-lation spectroscopy. III. Uniform translation and laminarflow,” Biopolymers 17, 361–376 �1978�.
29. R. Rigler, U. Metz, J. Widengren, and P. Kask, “Fluorescencecorrelation spectroscopy with high count rate and low back-ground: analysis of translational diffusion,” Eur. Biophys. J.22, 169–175 �1993�.
30. M. Gosch, H. Blom, J. Holm, T. Heino, and R. Rigler, “Hydro-dynamic flow profiling in microstructures by single moleculefluorescence correlation spectroscopy,” Anal. Chem. 72, 3260–3265 �2000�.
31. S. A. Soper, H. L. Nutter, R. A. Keller, L. M. Davis, and E. B.Shera, “The photophysical constants of several fluorescentdyes pertaining to ultra sensitive fluorescence spectroscopy,”Photochem. Photobiol. 57, 972–977 �1993�.
32. S. A. Zugel and F. E. Lytle, “Two-photon excitation in squareand cylindrical capillaries,” Appl. Spectrosc. 54, 1203–1207�2000�.
33. B. H. Weigl, R. L. Bardell, N. Kesler, and C. J. Morris, “Lab-on-a-chip sample preparation using laminar fluid diffusioninterfaces—computional fluidic dynamics model results andfluidic verification experiments,” Fresenius J. Anal. Chem.371, 97–105 �2001�.
34. D. S. Mehta, C. Y. Lee, and A. Chiou, “Multipoint parallelexcitation and CCD-based imaging system for high-throughput fluorescence detection of biochip micro-arrays,”Opt. Commun. 190, 59–68 �2001�.
35. H. Martinsson, J. Bengtsson, M. Ghisoni, and A. Larsson,“Monolithic integration of VCSEL and diffractive optical ele-ment for advanced beam shaping,” IEEE Photon. Technol.Lett. 11, 503–505 �1999�.
