# Flows in micro fluidic networks: from theory to simulation

Danny vam Noort Ecology and Evolutionary Biology Dept., Guyot Hall, Princeton University. Princeton. 08544 NJ, USA dnnnvriiinrinceton.2dii John S. McCaskill Fraunhofer Society, BioMIP, lnstitutszentrum Birlinghoven, SchloP Birlinghoven, 53754 Sankt Augustin. Germany inccnskill[~biomin.~~.de

Abstract- When complex flow structures are designed, such as in DNA computing [l],it is essential to be able to predict the flow pattern of the solutions in the fluidic network. A model based on the resistance of the channels and flow velocities of the inlets can eliminated re-iterative design steps. We have constructed a symbolic model (using Mathematicam)to determine the desired flow pattern based on the equations of Ohm and Kirchoff. The values from this simulation were used in a flow simulation program.

were made with a microfluidic simulation software package from Coventor'" (Coventor Inc.. NC. USA).' First we present the necessary equations to calculate the resistance of a rectangular shaped tube after which the reactor modules and their working principle will he introduced.

2 Theory
When flows are laminar. the non-linear term v.R in the Navier-Stokes equation becomes zero (e.g. Happel and Brenner 161). Exact solutions of the remaining linear equations are then possible, especially if the flow is steady as is the case considered here. Indeed. the microfluidic architecture was deliberately chosen to avoid dynamic flow switching. For flow in the x-direction. the Navier-Stokes law implies:
V Z U = -1 dp

1 Introduction
Complex flow structures. like the DNA-computer we are developing [l.Z]. are difficult to construct without the use of calculation and simulation tools. Simple open networks. i.e.. with one input and one output can be constructed ad hoc. However. when there are multiple inputs and outputs. it is a necessity to know the exact ratio between the flows and volume. The fluidic network should he balanced by determining the channel resistance and the input velocities required to realize the direction and volume of flows. The resistance and flow properties are calculated in a similar fashion to that used in electric circuit theory, i.e. the same laws apply. The resistance can he calculated using Ohm's law while the first law of Kirchhof states that the sum of currents is zero in nodes of the circuit. The analogy between fluidic and electric circuits is that the electric resistance (depending on the wire dimensions. the dielectric properties of the material) corresponds to the hydrodynamic flow resistance (depending on the channel dimensions, the material properties of the channel) while the current (the flow rate of electrons in the wire) corresponds to the flow velocity (the flow rate of the liquid in the channels). The voltage drop over a wire corresponds to the pressure drop over a channel. The use of equivalent resistance circuits in microfluidic design was pioneered in the diploma thesis of Arne Bochman [3] (Univ. Jena. 1996). Other recent research use theoretical models to perform calculation, e.g., flow rates of a membrane pump or the impedance for microfluidic systems, were performed by Bardell and Forster [4,5]. In our model the resistances of the micro channels were calculated algebraically using the computer propram Mathematicam (Wolfram Research Inc, IL. USA). assuming an ideal situation. Visualisation of flows

(1)

P dx

with U is the flowrate a n d p is the viscosity. The pressure gradient 4 is constant for a steady state
dx

and is written as

*=-P dx I

(2)

where ~p > 0 is the pressure difference between the ends of the channel of length 1. The end effects are neglected. So combining (1) and (2) gives v f u = -I AP (3)
P '

The volumetric flow rate Q is: i AP Q=
R I

(4)

where R is the specific resistance. The total resistance over the channel is
R, = RI

(5)

where I is the length of the channel. Substituting this in eq. (4) gives -dp=R,Q (6) (which corresponds to Ohm's law Al'=RI ). In the case of a rectangular channel with sides a and h. the volumetric flow rate is [SI:

0-7803-7804-0 1031117 00 D ZOO3 IEEE

2537

To optimise the transfer of only the appropriate sequences. lb). After each selection step. which depends on the channel resistance in the system. a washing step has to be performed so as to rinse off the non-specifically bound DNA sequences. In the standard approach.o:l B C By substituting eq 7 in eq 6.. Because of the change in pH at the denaturation step.o=*lub . la). each DNA sequence encodes binary sequence ( S ) .1 (Q 2 + b 21(7) n5n=1(2n-ly to determine the resistances and the desired inputs and output velocities. which stops the beads from disappearing down the channels but creates additional hydrodynamic resistance. the resistance of the channels can be calculated. on which single stranded selectorDNAs (reverse complementary to a sub-sequence) are immobilised [I].. Wa out N Ti" dH 4 Microreactor Structure The selection procedure described above can be implemented in a microreactor. To transfer actively the selected DNA sequences to the appropriate output. A [n"/un~(~~:+b'/anh(~. a subsequent neutralisation step is necessary after each selection stage before flowing the selected DNA into the next module. . To prevent the beads from flowing to other STMs. Denaturation is performed by using an alkali solution (NaOH). 1) at a certain position in a sequence (S. a bead-barrier has been added to the design. The DNA strands in the template solution hybridise to the selector-strands and are thus transferred to another channel in the microflow reactor where they are denatured and passed on to the next selection module (Fig.3 .2. Different DNA sub-sequences are selectors used to represent single bit values (0. To this end we have developed a selection transfer module (STM) which is able to make positive selections from a population of specific DNA sequences. Here the wash is performed with the neutralisation solution. i=1.). The principle of the selection module is shown in Fig.pi 24 __ 4 8 . they are conveyed from one flow to another by moving paramagnetic beads.. : 3 Selection procedure In DNA Computing [7]. Therefore a microreactor model has to be made 2538 . DNA sequences are used as information carriers and as tools for the calculations. Each selection module is coded with short single stranded selection-DNA strand (ssDNA) from a finite set of predefined sequences.I 1 I / j 1 i j . A selection is made by picking out that word which has the desired bit necessary for further processing. adjusted in concentration to the common melting temperature of the hybridised DNA strands. the sub-population is passed on to the next selection step. 1. This is only possible if the flows have the correct flow pattern to do an adequate mixing (Fig.

The selection transfer module to be designed was shown in Fig. since we are only interested in the ratio between the resistors and velocities) except Rsh=3 and the velocity v2=4vI. flow simulations were made with CoventorTM. There are 3 inputs (template.esismrs. colculnrions /o l . and v3. they should be at least 3 times wider with the same depth main channels. Table I shows all the resistors and flow velocity values. To ensure that the resistance of the supply and waste channels don't corrupt the overall resistance. Other constraints are that the pressure p7 is zero (there is the assumption that there is no residual pressure on the output relative atmospheric).=4 and v. u'e have designed a DNA computer to solve a particular NP-complete family of problems: the maximum clique graph problem.irc~rerlru irzprir 3 iri order.io ollou. To solve the problem we are working on. Ourpur 7 is con. neutralisation. A sche. a/ [ 2 ] . de-hybridisation) and 2 outlets (waste and template). I . which means that template out is connected to template in (node 4 is connected to node 3). 3 2 1 Since we are working under different buffer conditions (denaturation solution should be neutralised to allow hybridisation conditions again in the next microreactor).=2vl. Using the values of these parameters gives us the correct fluidic network for the desired flow behaviour. Fiyre 2 Figure 2 shows the module represented by resistors used for the calculation of the channel resistances and flow velocities.n seyrrence qf sac11 w ImClOlJ. the values of resistance were taken as ratios and not as absolute values. ratios between the different buffers should be fixed and have been chosen l : l . then R. ~ l o c i r i ~ i B 2539 .5 Microreactor Model The characteristics of the microfluidic system are dependent on the particular class of problems being addressed. The calculation is being made with R34 as a variable and all the other values fixed. In reality the problem is limited to a finite number of these reactors in sequence. Supply channels are needed to deliver the solutions to the STMs. Details for this procedure can be found in McCaskill [ I ] and van Noort er. I . should be equal. 6 Flow Simulation To visualise the calculated resistances and velocities used to design the microflow system.for. Tablr I A The cliannel resistances nnd flow i . we need at least 15 of these reactors in sequence.nnric rqwrsriirarion of rltc firicroflm ~yncror cmsisriq qf 7 r. The horizontal part is the reaction chamber in which the selected DNA strands are transferred by magnetic beads from left to right where they are released to the template output. As mentioned before. The network has been made iterative just for the calculations. "" ~ I. This means that the velocity v. In this case.. If all the other resistors are set to 1 (these values being arbitrary.