encrypted chemical signals

Based on the inlet spacing different stable regions are obtained resembling a multi-stable behavior, which is a phenomena responsible for

memory storage in neurons

Transforming the data into radial coordinates shows

fractal behavior

Simulated data from the loop device that resembles a

kiwi fruit

**Design of droplet based microfluidic devices for Lab-On-Chip applications**

Jeevan Maddala

**Applications**

**Background**

**Challenges**

Point of care diagnosis

Biomimicry

Drug Discovery

How to check if a particular drug is allergic to you?

Droplet microfluidic technology

Fast to detect

Detection with minimum amount of blood

Can be made mobile

Challenges:

Designing these devices

Control of droplets

Chemistry

Reaction mechanism

Advantages using microfluidic technology

Screening 1 Million drops/Sec

Easy scale up

Precise results at lab scale and large scale

Less raw material cost

Challenges:

How to rationally design the structure to perform combinatorial chemistry?

Devices mimicking biological systems to understand and exploit the designs made by nature

Water

T-Junction

Whitesides., Nature, 2006

Manuprakash et al., Science 2007

Song et al., Angew. Chem. Int. Ed, 2006

Piotr Garstecki lab group

Microfluidics is the science of manipulating small (10 to 10 liters) amounts of fluids in micro channels

Advantages of droplet based microfluidic devices

Precisely control concentration of molecules in space and time

Throughput ( 1 Million/sec)

Oil

Understanding the behavior of droplets

Controlling droplet position in space and time

Designing the networks to achieve desired objective

Proposed GA approach

Objective 1: Design Ladder network to synchronize droplets coming at fixed delay with minimum bypasses

Results

Initial population is doubled by several mutations and one point crossovers

N to 2N

Selection operator is used to select N structures from 2N for next generation

2N to N

Selection contains the following metrics

Fitness function

Diversity

.

Genetic operators – Selection

Changes the structural properties drastically

Helps in locating all the optimal solutions

Implemented in two tiers:

Tier1: Changes the overall bypass number

Tier2: Searches among the existing structures by changing 1 bypass

Genetic operators - Mutation

Ladder network has the following components:

Total bypasses (TB) – Binary values

Value (V) – Real values

Choice of top segment (SC ) – Binary values

Choice of bottom segment (SC ) – Binary values

Coding Decoding

As number of bypasses(n) increases, number of equations to solve is in order of (n )

Simple analytical model is not possible due to droplet motion

Structure functionality is nonlinear and discontinuous

Modifying single bypass could lead to a drastic change in functionality

Therefore finding optimal designs using conventional gradient based approaches is hard

Possible solutions to find optimal structures is through: Evolutionary algorithm or MINLP

Challenges in designing multi-bypass ladder networks

Objective 2: Design ladder network to synchronize droplets with input spacing

Results

GA is used to optimize the design based on functionality

Structural component converted to numerical strings

Coding

Genetic operators used to change structure

Mutation

One point cross over

Selection

Functionality is given in the fitness function

Fabrication limitations are posed as constrains in GA

Two off springs are produced by two existing structures

Off springs have properties of both the parents

Searches locally around the existing structures

Genetic operators –One point cross over

Synchronization of drops

Ladder networks

Functionality?

To synchronize droplets

Application:

Reducing input fluctuations for downstream processing

Optimal design?

Design & Functionality

Summary

Proposed solution

Problem:

Design the ladder structures based on objective fucntion ?

(Optimization)

3

Converting structure to numbers

One - one mapping

The coding scheme should span all the domain

Problem statement

Evolutionary optimization: Genetic Algorithm

**Design of ladder networks**

Dynamics of drops at bifurcating networks

Observations

**Droplet behavior**

We observe micro scale bifurcating networks in Nature, for e.g. Blood vessels, Leaf venation etc.

We observe that these networks are not only useful for optimal transport of fluids but also have the following advantages:

Encrypted Chemical Signaling

Memory Storage

Pattern formation

Binary droplet decision leads to Chaos

Results

Sigmoid function

Droplet encoder

Proposed an Artificial Intelligence (AI) based approach to design ladder networks

Uncovered ladder designs that are:

Robust to inlet fluctuations

Produce nonlinear transformation of spacing

Encode and decode droplet pairs

-9

-15

**Future Work**

**Thank you!**

Acknowledgments:

Dr. Raghunathan Rengaswamy

Dr. Siva Vanapalli

William S Wang

Loop Dynamics

R = R + n R

Used to predict the droplet dynamics

Postulates:

Conservation equation at the node

Flow is conserved at the node

Constitutive relationship

Pressure drop is equal to product of flow and resistance

Velocity of the droplet is linear function of flow rate

V = Q

Each droplet increases the resistance by the same amount

Resistance in the channel is summation of channel resistance, resistance due to droplets and valve resistance

Network model

M Schindler and A Ajdari, Physcal Review Letters,2008

(Palacios A M et al, PNAS, 2011)

(Hosaya et al., Exp Brain Res, 1991)

(Roth-Nebelsick et al., Annals of Botany, 2001)

T

c

d

c

Develop large scale software that could design a generic microfluidic network based on objective function

Implement MINLP frame to design these systems

Cytotoxicity studies

For e.g. loop device

For e.g. Ladder Networks

No diversity leads to local minima

10% Elite

80% Diverse

10% Random

t

b

Manipulate bypasses

Manipulate genes