# Capillary electrode array/proposal import 2007

The Capillary Electrode Array: A CMOS DNA Separation System

\section{Introduction}

This proposal is motivated by a key problem in molecular biology and biochemistry: starting from a heterogeneous ensemble of biopolymers in an aqueous solution with a homogeneous concentration profile, create an aqueous solution with a highly heterogeneous concentration profile of localized concentrations of certain biopolymer species. %% (Figure \ref{cea_conc_profile}). Although this procedure is important for all of the biological macromolecules in an organism, DNA molecules are the exclusive focus of this report.

%% \begin{figure}[h] %% \begin{center}

%% \includegraphics{cea_homo_conc.eps}

%% \includegraphics{cea_hetero_conc.eps}

%% \end{center} %% \caption{Concentration profiles in separation process: (\textbf{top}) %% Before separation, three species in solution have spatially homogeneous concentration %% profiles. (\textbf{bottom}) After separation, the three species are %% separated and a spatially heterogeneous concentration profile %% exists. } %% \label{cea_conc_profile} %% \end{figure}

Currently, the most widely used technique for DNA analysis is electrophoresis (see, for example,\cite{viovy00}) in which an electric field is applied to a gel or cross-linked polymer solution. Negatively charged DNA molecules migrate through the solution, with the molecules of the separation medium slowing the DNA molecules through mechanical interactions. Longer DNA molecules experience more obstructions than smaller DNA molecules, and move more slowly through the medium.

In a seminal 1991 paper\cite{prost91}, Ajdari and Prost proposed the use of an inhomogeneous electric field, transverse to the direction of flow, for free-flow (i.e. without a gel-based medium) separation of molecules. Their theoretical analysis indicated separation efficiency significantly greater than that of conventional electrophoresis systems.

Here, we propose the use of CMOS microfabrication and maskless postprocessing to create a molecular separation system that operates on the principle described in \cite{prost91}. This device contains a microchannel, an array of electrode pairs surrounding the microchannel, and a microheater element to actuate, by bubble creation, bulk flow through the microchannel. While the bubble actuator creates the hydrodynamic force to drive DNA molecules through the microchannel, the electrode pairs create electric fields perpendicular to the direction of the microchannel flow. In the following, we describe the basic architecture of the proposed system, and outline areas requiring further development for realization of a practical system for preparative DNA separation.

\section{Fabrication}

In the following, our focus will be directed towards the 1.5 micron ABN process offered by AMI Semiconductor. This process provides two levels of polysilicon for MOS transistor gates and capacitors formed between the polysilicon layers, and two layers of aluminum metallization for electrical interconnection between circuit elements. The minimum width of a polysilicon line in this process is 1.5 $\mu$m, thus the description above. This process is significantly behind the cutting edge of CMOS processes, having been introduced in 1987; currently (2004) MOSIS offers a process with a .13$\mu$m polysilicon line width.

A process flow for fabrication of the proposed device is shown in Figure \ref{cea_process_flow}.

\begin{figure}[h] $\xymatrix{ *+[F]{ \txt{CMOS fabrication} } \ar[d] \\ *+[F]{ \txt{XeF}_2 \txt{ etching} } \ar[d] \\ *+[F]{ \txt{Wafer scribing to expose microchannel}} \ar[d] \\ *+[F]{ \txt{Fluidic connection to microchannel inlet,\\ electrical connections to IC pads}}  }$ \caption{Process sequence for capillary electrode array fabrication.} \label{cea_process_flow} \end{figure}

A cross-section view of the microchannel before and after processing is shown in Figure \ref{cea_cross_section}. Prior to etching with XeF$_2$, a polysilicon-2 minimum-width line has polysilicon-1 and metal-1 electrodes below and above, respectively.

\begin{figure}[h] \begin{center}

\includegraphics[scale=.8]{cea_cross_section.eps}

\end{center} \caption{Microchannel cross section: (\textbf{A}) A

 polysilicon-2 line has electrodes below (polysilicon-1) and above
(metal-1).  (\textbf{B}) After etching with XeF$_2$, the
polysilicon-2 line is removed to form a microchannel.}


\label{cea_cross_section} \end{figure}

Figure \ref{cea_layout_annotated} shows 10 electrode pairs. Each poly-1 electrode is 1.6$\mu$m wide, and each metal-1 electrode is 2.4$\mu$m wide. The spacing between the metal-1 electrodes is 1.6$\mu$m, for a center-to-center electrode spacing of 4$\mu$m.

\begin{figure}[h] \begin{center}

\includegraphics[scale=.8]{cea_layout_annotated.epsi}

\end{center} \caption{Plan view of 10 electrode pairs.} \label{cea_layout_annotated} \end{figure}

The fabrication of the microchannel opening is shown in Figure \ref{cea_opening_cross_section}.

\begin{figure}[h] \begin{center}

\includegraphics[scale=.8]{cea_opening_cross_section.eps}

\end{center} \caption{Microchannel opening cross section (not to scale):

 (\textbf{A}) Polysilicon-2 is exposed to the atmosphere, with a
polysilicon-1 electrode beneath. (\textbf{B}) After etching with
XeF$_2$, the polysilicon-2 line is etched to form the opening to the
microchannel.  }


\label{cea_opening_cross_section} \end{figure}

%% CEA as practical realization of \cite{prost91} dielectric traps

%% bulk flow instead of electrophoretic potential difference \cite{dovichi1999,dovichi2002,yamamoto2000}

A crucial difference between the system analyzed in \cite{prost91} and the device proposed here is the use of bubble-actuated bulk flow instead of an electrophoretic potential to drive molecules through the microchannel. This mechanism is conceptually very similar to that of a bubble-driven inkjet print head, and has been demonstrated in a CMOS-fabricated microsystem \cite{westberg97}. DNA microarray formation using a standard bubble-jet print head has been demonstrated \cite{yamamoto2000}, and it is expected that the mechanism in the proposed system will not damage DNA molecules. One concern is the effect of heating on separation efficiency. An increase in temperature will have the effect of broadening the concentration profile, countering the effect of the separation process.

%% It is expected that this bulk flow will enable the proposed device to %% be used in preparative separation of DNA molecules, by ejecting %% droplets containing purified DNA solutions.

%% \cite{kim96}

\bibliography{cea_proposal-2006-10}

\end{document}

target audience is people familiar with MEMS design and fabrication

mathematical analysis of separation performance not necessary for

  initial proposal


what are the performance requirements for circuitry

 process limitations
straight microchannel -> interelectrode impedance not highly variable
fractal microchannel -> highly variable interelectrode impedance


first present straight microchannel with electrode array

 perhaps not even necessary to go into fractal geometry in first
proposal iteration
fractal microchannel geometry sure to confuse people more :)


specific steps toward realization

 how much circuit design is necessary to justify fabrication?
preliminary device to evaluate microchannel etch clearance?
impedance sensor for evaluating etch clearance / molecular presence?
sensitivity of CMOS impedance sensor - single DNA molecule
detection not possible
difficulty of designing!



technologies required

 CMOS
micron-scale - open layer for simple design
submicron - CMU MEMS a possibility for exposing polysilicon for
etching
reducing complexity / risk of first realizations is important -
minimize process steps
internal fab data would be nice for understanding design... not
necessary
XeF2
bubble actuator
simulation



motivation

 separation in free solution
dispensing unmodified DNA molecules
ideal: printing DNA microarray from DNA solution
preparative / analytical separation



methods

 CMOS microfabrication / micromachining
dielectrophoresis
ratchets
bubble-jet printing mechanism


issues

 addressing complexity of electrokinetic phenomena in CEA
experimental investigation of separation phenomena - too complex for
simulation-based estimates?
relationship between DNA length and channel dimensions / separation
efficiency
clearance of microchannel - long etch times, SiN etched


DNAjet printer: DNA molecules printed on standard surfaces

first step: provide essential insights to enable collaboration with others

 necessary to collaborate with people who can help with system integration
delivering DNAjet printer requires significant effort


essential insights:

 CMOS microfabrication / postprocessing
importance of CMOS economics
importance of CMOS circuitry for complex field generation
transverse inhomogeneous field for separation
DNA separation dynamics



microsystems for CE known to MEMS practitioners dielectrophoresis, transverse-field separation not widely known or used

leverage CMOS microfabrication and novel separation mechanisms to

create DNA processing system with unprecedented(?) capabilities

 dream: integrated upstream system isolates DNA from crude biofluids
for CEA preparation - DNA microarray or integrated high speed sequencing downstream


three dimensional CMOS integration for microfluidics as well as

microelectronics


rotating microfabricated system for molecular separation

 drop-in replacement for DVD, 3D patterned on nanoscale for molecular
information system, molecular analysis system


electromagnetic phenomena in rotating microfabricated systems

potentially relevant to new RF systems...

microantenna array from rotating microscale elements?
DVD player as antenna system? :)


'People always overestimate how much will change in the next three years, and they underestimate how much will change over the next 10 years.'

CEA an incremental step

visual representation difficult in preliminary report

 requires solid modeling
solid model of representative electrode pair / microchannel


representation of electric field in microchannel also difficult

 fractal geometry of microchannel means single electrode pair can create e-field at several points on 'microchannel coordinate'


bubble-jet mechanism

 too complex to satisfactorily address in preliminary effort - ANSYS, other simulation?
key for preparative separation
CMOS-compatible


fractal microchannel geometry

 maximizes length
complex e-field within length of microchannel
manufacturing: XeF2 not perfectly selective
stresses in thin-film layers


charge transport in microchannel

 diffusion
vertical electrophoresis
dependent on switching frequency



molecular transport mechanisms

 vertical electrophoresis
(laminar) bulk flow from bubblejet
diffusion
dielectrophoresis
molecular interaction with microchannel inner surface


\section{Introduction}

This report is focused on an operation with great importance in molecular biology and analytical biochemistry: starting from a heterogeneous ensemble of biopolymers in an aqueous solution with a homogeneous concentration profile, create an aqueous solution with a highly heterogeneous concentration profile of localized concentrations of certain biopolymer species (figure \ref{concentration_profile}). Although this procedure is important for all of the biological macromolecules in an organism, DNA molecules are the exclusive focus of this report.

Current systems for DNA separation require many copies of identical DNA molecules for analysis. However, it is not necessary to have multiple copies of a DNA molecule to characterize the base sequence of the DNA molecule. The polymerase chain reaction provides a straightforward means of creating an essentially unlimited number of copies of a single DNA molecule, making it possible to use current sequencing techniques and PCR to analyze a single isolated DNA molecule. The motivation for analyzing individual or small numbers of DNA molecules is twofold. Single copies of locally unique DNA molecules can contain sequence structure that determines the morphology or fate of a cell or organism (as in the case of an antibody producing B- or T-cell in which DNA rearrangement produces antibody diversity, or a cancerous cell proliferating out of control due to a mutated chromosomal DNA molecule). Secondly, in some cases, such as forensic applications, the amount of DNA present in the orginal sample may be minimal, justifying a process for dealing with small numbers of DNA molecules.

One approach to creating a device that can manipulate single DNA molecules is to use microfabrication \cite{madou97}, which generally refers to the creation of structures with dimensions on the order of a micron or less. Microfabrication is an extremely important technology for our society, and is most notably used to create integrated circuits (ICs) for computer and communications systems.

Generally, microscale devices with mechanical (including fluidic) structures and electrical inputs or outputs are called microelectromechanical systems (MEMS). Many MEMS devices are fabricated using custom processes that are tailored to the production of a specific device. Any microfabrication process typically has high initial costs for startup of the manufacturing process \cite{sze96}. In a custom process with low manufacturing volume, the costs are amortized over fewer devices, increasing the per-unit costs. Additionally, microfabrication processes are typically perfected with extensive use, and a low-volume custom process does not benefit from the gradual improvements seen in high-volume microfabrication processes. The use of standard processes that are adapted to the production of a novel device can improve the manufacturability and economic attractiveness of such a device. Additionally, the use for MEMS of processes that enable the creation of transistor-based circuitry makes it possible to integrate control, sensing, and actuation functions on a single chip.

The primary focus of the IC industry's effort is the fabrication of digital and analog complementary metal-oxide-semiconductor (CMOS) integrated circuits, and CMOS IC technology has surpassed all others in miniaturization and cost of fabrication. In addition to providing packaged chips based on designs created in-house, many CMOS IC manufacturers provide fabrication services for designs that are created by their customers. This service dramatically reduces the resources necessary to create new devices using CMOS IC technology. The competitive market for IC devices and IC fabrication services makes it necessary for IC producers to rapidly improve their technology or risk business failure. This economic imperative for continuous improvement of CMOS IC technology has created a rapid increase over time in the number of transistors on a chip, and a rapid decrease in the size and cost of individual transistors in an IC.

CMOS fabrication technology is optimized for the creation of transistors and metal interconnect between transistors. There is considerable flexiblity in the patterning of the thin films on the surface of a silicon wafer that compose the integrated circuit itself, although there is generally no flexibility at all regarding the thickness of the layers. Carefully designed thin film patterns on a CMOS chip can be coupled with processing performed on the chip after fabrication of circuitry to provide the chip with new functionality. Selective removal by etching of certain thin film layers (so-called \emph{sacrificial} layers) in the integrated circuit can create suspended structures and holes in the stack of materials deposited in the fabrication of an integrated circuit. With careful design, these structures can be used to create sensors (to link mechanical inputs to electrical outputs) and actuators (to link electrical inputs with mechanical outputs).

For the purpose of DNA manipulation and analysis, many devices within the CMOS process can be used. Electromagnetic fields created by voltages placed across electrodes can selectively effect the movement of negatively-charged DNA molecules. The impedance between electrodes, which is affected by the DNA molecules and other ions present, can be sensed using on-chip circuitry as a means of analyzing the composition of the solution between the electrodes \cite{ling95}. Resistive heating within on-chip resistors can be used to melt double-stranded DNA, and can vaporize fluid to effect bulk flow of a solution through a microchannel (in a process similar to bubble-jet printers). The temperature dependence of some electrical properties in CMOS devices, such as the resistivity of polysilicon and certain transistor properties, can be used to sense the temperature on the chip. Photodiodes integrated on-chip can be used to sense photons emitted by fluorophores that interact with DNA molecules, and thereby sense the presence of DNA molecules.

In the following, some prior work using novel schemes to electrically separate DNA molecules, and novel microfabricated devices to separate DNA molecules is described. We then describe a new microchannel structure with fractal geometry, fabricated in CMOS, possessing paired electrodes about the body of the microchannel. We sketch a possible application of the device for DNA separation and microarray printing, and propose future theoretical and experimental analyses to verify the utility of the device.

In the following, we present a brief description of

All complementary metal-oxide-semiconductor integrated circuit fabrication processes enable the creation of MOS transistors with positive charge carriers (p-channel, or pMOS) and negative charge carriers (n-channel, or nMOS). The primary advantage of CMOS is the fact that logic elements draw significant current only during transitions from one state to another, and very little current between transitions. This has lead to the overwhelming, and increasing, dominance of CMOS for digital integrated circuit design \cite{rabaey96}. The substantial cost savings that are realized by integrating several functions of an electronic system on a single chip, along with the continuing rapid progress of CMOS technology \cite{sze00}, has motivated the increasing popularity of CMOS for analog integrated circuit design \cite{gray93,razavi01} as well.

A basic circuit in CMOS digital systems is the inverter. Figure \ref{cmos_inverter_schematic} shows a schematic view of a n-channel MOS (abbreviated nMOS) transistor and a p-channel MOS (abbreviated pMOS) transistor connected to form a CMOS inverter.

\begin{figure}[h] \begin{center}

\includegraphics{cmos_inverter_schematic.eps}

\end{center} \caption{A schematic view of the CMOS inverter.} \label{cmos_inverter_schematic} \end{figure}

Figure \ref{cmos_inverter_plan_view} shows a plan view of the layers composing the CMOS inverter. This is the representation of the components of the integrated circuit that is used for layout of the IC. The size of the features on the individual layers, and the placement of different layers relative to one another, is dictated by the \emph{design rules} for the process, which are dependent on the dimensions and tolerances achievable in the process. The design rules are a key part of the well-defined interface between design and fabrication of CMOS integrated circuits \cite{mead80,weste92,pucknell92}. The MOSIS Scalable CMOS (SCMOS) \cite{MOSIS:SCMOS} design rules are intended to be technology and process independent (between CMOS processes), with a scaling factor $\lambda$ used to parameterize the design rules. This makes it possible to use different fabrication processes to realize a particular design, simply by changing the scaling factor to reflect the dimensions achievable in the process used. The layers necessary for specification of an integrated circuit design are generally standard between different processes, due to the basic structure of the MOSFET and the interconnections present in an integrated circuit.

While the two-dimensional shape of the layers in the process can be controlled by the designer, the parameters of the process, such as the sequence of the process steps and the thickness of the layers, are set by the manufacturer, and are not adjustable by the designer. This is particularly relevant for micromachined transducers fabricated in a CMOS process, since the layer thicknesses are rarely optimized for the creation of sensors and actuators. Additionally, the technical details of a process, such as the layer thicknesses, are typically treated by CMOS manufacturers as proprietary secrets that are shared with outside parties either under a restrictive non-disclosure agreement or not at all.

In the following, our focus will be directed towards the 1.5 micron ABN process offered by AMI Semiconductor. This process provides two levels of polysilicon for MOS transistor gates and capacitors formed between the polysilicon layers, and two layers of aluminum metallization for electrical interconnection between circuit elements. The minimum width of a polysilicon line in this process is 1.5 $\mu$m, thus the description above. This process is significantly behind the cutting edge of CMOS processes, having been introduced in 1987; currently (2001) MOSIS offers a process with a .18$\mu$m polysilicon line width.

Figure \ref{cmos_inverter_fabrication_cross_section} shows a cross-section view of the thin film layers comprising the CMOS inverter, realized in the ABN process. In this process, as in virtually all modern CMOS processes, the transistor is formed using a polysilicon gate, with the gate serving as a block for the dopants that form the drain and source regions of the transistor.

The irregular topography of the thin-film stack in figure \ref{cmos_inverter_fabrication_cross_section} limits the resolution achievable during photolithography steps, relative to a flat surface. Modern submicron CMOS processes use a dual damascene process to create planarized surfaces for metallization, as shown in figure \ref{dual_damascene}. In the figure, tungsten is used to fill the via that provides interlayer metal connection, as is common in modern CMOS processes. The dual damascene process avoids the use of potentially troublesome metal etch processes, and instead shapes metal interconnections using oxide trenches and removes excess metal in a chemically-assisted mechanical polishing process.

The metallization schemes for submicron CMOS pose difficulties for MEMS fabrication. The selective etching of certain layers in a CMOS process to create microsystems is dependent on the exposure of the layers to the atmosphere and thereby the wet or dry etchant used. The use of planarization for metal layers makes all but the top layer of metal inaccessible in this processing method. This is in contrast to the AMI ABN process (and generally other processes with feature size larger than 1 micron) which has a non-planar surface and layers that can be exposed to the atmosphere, with an appropriate design. Additionally, the use of tungsten, with its unique etch chemistry, complicates the processing necessary to realize MEMS using CMOS, which has focused on well-known Al, silicon, and silcon oxide etchants.

The central feature of the current design is the microchannel to be formed by sacrificial etching. The microchannel is formed by etching a material within the CMOS thin-film stack, with the material in the shape of a line, coated with silicon oxide along its length, and exposed to the etchant at its ends. The etchant used to create the microchannel must be extremely selective, because of the extreme difference between the width and the length of the microchannel.

CMOS-compatible microchannels have been formed by etching aluminum with an aluminum-specific wet etchant \cite{baltes97_1}. This technique has been used to create a resonant flow sensor \cite{baltes99_1} and a CMOS-compatible droplet ejection system \cite{westberg97}. Aluminum etching is not suited for this application, for two reasons. Due to its low resistivity, aluminum should be used, where possible, for the microelectrodes that create electromagnetic fields within the volume of the microchannel, and for signal transmission within the circuitry on the chip. In addition, wet aluminum etchants, despite their selectivity, etch silicon dioxide at a measurable rate. The microchannel in the current design is an order of magnitude longer than other CMOS-compatible microchannels. This requires a longer etch time to clear the microchannel, resulting in more oxide etching and potential damage to the circuitry that is protected by the oxide passivation.

Polysilicon is another material that can be sacrificially etched in a CMOS process to create microchannels, using xenon difluoride \cite{kim96}. XeF$_2$, which is a white odorless solid at standard temperature and pressure, sublimates at XeF$_2$ vapor rapidly etches mono- and poly-crystalline silicon (at up to 10 $\mu$ per minute \cite{williams96}), but does not measurably etch silicon oxide, aluminum, or photoresist. XeF$_2$ is known to etch titanium and tungsten, two components of many modern metallization processes; however, since these metals are not present in the 1.5 micron AMI ABN process, this issue will not be addressed here.

\section{DNA separation}

In order to effect the transport of electrically charged molecules in controlled way, electric fields can be imposed on a solution, in a process called \emph{electrophoresis} \cite{viovy00} (phoresis, Greek for force). Although this effect occurs with any charged molecule, by far the most widespread application of electrophoresis in analytical chemistry is the separation of DNA.

The most-widely used DNA electrophoresis approach uses a static electric field to force DNA molecules to migrate through separation medium consisting of a cross-linked polymer solution or gel. The constituents of the separation medium act as obstacles, impeding the motion of the DNA molecules (one analogy is the gravitational transport of a rope through a close-packed array of posts extending horizontally from a vertical surface). In this process, longer DNA molecules experience more hindrance to their movement from the gel, and have a lower mobility in the gel than smaller DNA molecules. The mobility $\mu$ of a molecule undergoing electrophoresis is $\bf{V} = \mu \bf{E}$, where $\bf{V}$ is the time-averaged velocity of the molecule, and $\bf{E}$ is the electric field in the solution. A medium is necessary for the electrophoretic separation of DNA in a constant field, because the mobility of DNA molecules in solution is independent of size.

The use of a polymer solution for separation of DNA molecules is less than ideal. The complex structure of the medium makes nearly impossible the development of a quantitative model for predicting molecular dynamics. Although computer simulations have been successful in describing the dynamics of gel electrophoresis of DNA molecules, the results provide only a qualitative understanding of the process. The practical impossibility of a precise model of the gel structure prevents even computer simulations from being used in a predictive manner.

On the other hand, the dynamics of polyelectrolytes in free solution, while still quite complicated, can be much more accurately modeled using hydrodynamic concepts, treating the aqueous solution as a continuum governed by the Navier-Stokes equation.

Several proposals have been made for electrophoretic separation of DNA molecules in solution using asymmetric time-varying electric fields. Called Brownian ratchets or stochastic ratchets\cite{slater97a,slater97b}, these schemes use the Brownian motion of DNA molecules over potential barriers to effect size-dependent transport.

It is well-known that an electric dipole experiences a net force in a nonuniform electric field. This force on a molecule or particle (such as a cell) can be used to effect movement of the object. This process is called dielectrophoresis \cite{pohl78,jones95}.

In a seminal 1991 paper \cite{prost91}, Ajdari and Prost proposed the use of an inhomogeneous electric field, transverse to the direction of flow, for free-flow (i.e. without a gel-based medium) separation of molecules.

It has been shown experimentally that DNA molecules in aqueous solution can be manipulated with microfabricated electrodes \cite{washizu90,washizu95,engh98}. The .

Electrophoresis is the most widely used method for separating DNA molecules by size. In this process, an electric field is applied to a gel or cross-linked polymer solution. Negatively charged DNA molecules

 preparative separation of DNA


what is the current state-of-the-art?

 electrophoresis through viscous medium
microfabricated structures for electrophoresis


what is the key conceptual or technological contribution of this work?

 implementation of novel physical process for separation
use of multiple control elements available in CMOS for separation


 proposal contains, at least, description of device and manufacturing process



what is the plan for success?

 fabrication, experimental process to determine functionality


instead of wafer scribing to expose dielectric tube, use opening on top of tube for XeF2 exposure and fluid discharge

 invert chip for more effective droplet ejection
microscale fluids less affected by gravitational forces
back-side fluid input


somehow use afm nanooxidation technique of \cite{boisen2003} in conjunction with cea to efficiently fabricate dna alumina nanoarrays

how to integrate afm tip with CMOS CEA