Paul Yager Research Group
Bioengineering Department, Box 352255, University of Washington, Seattle, WA 98195, USA

Excerpts from Final Report on

A Microfluidic Sample Preconditioning System for CBW Agent Detection and Quantification
(Separations Project*)

Funded by the DARPA MTO Office MicroFlumes Program
May 1997 to July 2000

for information on Pumps and Controls, see www site of Prof. Forster in the UW Department of Mechanical Engineering


Accomplishments: Separations Section

Personnel 

Faculty and Staff

  • Prof. Paul Yager, P.I.
  • Prof. James P. Brody, Co-Investigator
  • Stephanie Frisbee, Secretary
  • Mary Smith, Secretary
  • Shirley Verzosa, Secretary
  • Gary Holman, Lab Technician
  • Darrel Bell, Lab Technician

Graduate Students

  • Catherine Cabrera (Ph.D. candidate)
  • Andrew Kamholz (Ph.D. candidate)
  • Anson Hatch (Ph.D. candidate)
  • Kenneth Hawkins (Ph.D. candidate)
  • Matthew Munson (Ph.D. candidate)
  • Eric Schilling (M.S.E. candidate)

Postdoctoral Fellows

  • Mark Holl, Ph.D.
  • Roya Sabeti, Ph.D.
  • Dong Qin, Ph.D.
  • Katerina Macounova, Ph.D.
  • Shelli Dennis, Ph.D.

Undergraduates

  • Mark St. John
  • Todd Bui

The project had the following goals:

This project has led to the following general activities:

These, in turn, led to the following 6 specific accomplishments:

  1. Demonstration of continuous microfluidic preconditioning devices based on transverse sedimentation, electrophoresis and isoelectric focusing of synthetic particles, proteins, bacteria and their spores.
  2. Demonstration of continuous microfluidic separation and concentration of these species.
  3. Development of polymeric laminate rapid microfluidic prototyping and flexible interconnect system.
  4. Demonstration of integrated microfluidic pumping and sedimentation separation system, and integrated sedimentation and IEF system.
  5. Demonstration of continuous quantification of proteins and small molecules using fluorescence-based assays, such as DIA, in the T-sensor.
  6. Development of models incorporating both hydrodynamic, electrophoretic and chemical information to facilitate design of future systems and understanding of processes occurring in devices.

These are described in more detail below.


CBW AGENT SAMPLE PRECONDITIONING SYSTEM DESIGN

As shown in the highly idealized schematic drawing below, we designed a system that would be capable of plugging directly into an air sampler, incorporating pumps P and flow control modules interconnecting three sequential microfluidic components: a sedimentation device A to remove the large and rapidly sedimenting interferent particles such as sand and dust, an electrokinetic element B to purify the target analyte, and an electrophoretic concentrator C to increase the concentration of the analyte.

In the original design device, B was to be a zone electrophoresis (ZE) device. Such a device would have required precise positioning of entrances and outlets to allow accurate operation. As it developed, it was possible to use isoelectric focusing (IEF) as an alternative to ZE, thereby allowing devices B and C to be consolidated into a single device.


MODELING DEVELOPMENT

Transverse Diffusive Transport in Microchannels--Modeling the "Butterfly Effect"

Initial modeling of transport across channels in our laboratory, particularly before we gained access to Microcosm's FlumeCAD package of software, was exclusively with 1D models of transport in channels. In these simulations all effects of dispersion of flow velocity were ignored, and the variation in flow rates along the optical path were ignored. Because the Poiseuille flow profile in ducts under positive displacement pumping is complex, the motion of particles from one region in the microfluidic device to another is also complex. To predict separation and monitoring of chemicals and larger particles in microchannels, the exact shape of the interdiffusion zone must be known. For this reason more detailed modeling was undertaken. This modeling initially included just diffusion of species perpendicular to the flow direction.

It immediately became obvious that distribution in solute residence time in the channels leads to differential lateral diffusion near the walls and along the device midline. For example, if two fluids enter the device shown at left from the two input channels, their contents will interdiffuse, but that interdiffusion will appear to progress further in laminae immediately adjacent to the device upper and lower surfaces, than along the device midline. The green zone is a crude representation of the region in which interdiffusion between the two streams has occurred. We call this phenomenon the "butterfly effect".

A 2D computational model for simulation of analyte diffusion in the parallel flows of analytes as in H-filters and T-sensors was developed in our laboratory. This model was coded into MATLAB and run on a PC. It allowed for consideration of the true flow velocities in microchannels.

The parabolic velocity profile across the w-dimension has considerable impact on the spatial distribution of analytes at steady state.

o The conditions for the simulation on the left: aspect ratio d/w ª 2.4, L = 4000 µm, flow rate ª 2 µl/s. The top frame shows diffusion of analyte (gray) as flow proceeds downstream (from lower left to upper right). The original (pre-diffusion) fluid interface is shown by a dotted line.

o The bottom frames are a set of contour plots of analyte concentration at the distances downstream noted in the upper-left-hand corners.

o Quantitative analysis of the diffusion profile was submitted to Biophysical Journal in July 2000 (Kamholz et al.).

Note that there are other sources of complication, including the fact that at the contact point of the two fluids their velocity is 0. Fluid accelerates along the device midline until fully developed flow occurs at a predictable distance downstream. These factors are particularly important considering that we quantify the interdiffusion zone by imaging. This line of modeling expanded to include not only diffusion but also chemical reaction and physical complexation of multiple species in the channels; this allowed the development of quantitative models of such processes as immunoassays in microchannels that agree very well with experimental data.

Modeling Isoelectric Focusing in Microchannels

Additional modeling efforts in the group were directed toward prediction of electrophoresis and isoelectric focusing in microchannels. The first step in the modeling was to accurately depict the development of pH gradients in channels under the influence of electrochemical activity caused by passing a current perpendicular to the flow (vide infra). This modeling involved similar features to the diffusion-model above, except that it added creation of species at the channel wall, and electrophoresis of all charged species in the presence of an imposed electric field. This led to some startlingly good agreement with experimental data.

Example of early 1D modeling of the positional dependence of pH and the Na+ ion concentration in a microchannel under the influence of current across that channel. No buffer was used in this simulation

A later time in the evolution of the pH and Na+ ion profiles in the same microchannel. Note that very sharp pH jump that persists at long times in the channel. This was confirmed experimentally as well.

Initial modeling was also 1D, which was ultimately replaced by a 2D model that took into account the actual distribution of flow velocities in real channels.

SEDIMENTATION SEPARATION

The first separation device that studied was that originally proposed--a simple microfluidic device that would remove the rapidly sedimenting and large particles from the feed stream by sedimentation. Many experiments were carried out in small systems and monitored by optical microscopy. We then designed polymeric laminate devices that were capable of separating rapidly sedimenting particles from a stream of slowly sedimenting particles at flow rates (100 µL/sec) that would be compatible with existing commercially available air samples.

The initial device had an internal volume of about 100 µL, and had walls of PMMA. It was demonstrated at the PI meeting in Pittsburgh, where it worked for 4 hours as driven by one of the NMPV pumps (see picture later in the report).

Concept for sedimentation separation devices. The one at left was the concept for the original design. The one at center incorporated at least one gas permeable PDMS membrane for extraction of gases from the input solution.

As the operation of the first sedimentation separation device progressed at the Pittsburgh PI meeting, it became filled with bubbles because of saturation of the fluid in the channel with air. These bubbles became pinned within the PMMA device and were impossible to dislodge. This resulted in "channeling" of the flow through the maze of bubbles, as well as a reduction in the effective fluid internal volume of the device, accelerating the flow through the device. This set of problems was solved in the design of the next generation of sedimentation separators, at the cost of a great increase in the complexity of the device. This 13-layer laminate separated particles by sedimentation rate at 100 µl/min and removed any trapped gas to prevent disturbance of the flow by adherent bubbles.

Design of the gas-removing continuous sedimentation separation device. The internal volume of the flow channel was ~100µL. Individual layers were made from laser-cut Mylar, alternating with layers of Mylar covered with adhesive on both sides. The top and bottom frames were made from thick layers of rigid PMMA. A key feature was the presence of two PDMS membranes for extraction of gases trapped in the fluid input to the system. The whole device was designed around a modular interconnect system that allowed rapid connection to convention PEEK HPLC tubing.


ELECTROKINETIC SEPARATIONS

"Mesoscopic" Microfluidic Test System

These electrokinetic flow cells were mounted in custom holders on our 2 Zeiss inverted optical microscopes. These holders allow optical monitoring of the cells while connected to Kloehn computer-controlled syringe pumps. Ultimately as many as 6 pumps could be connected to a single device using conventional PEEK HPLC tubing.

Mounted on the inverted microscope is the test device, which can be viewed in epifluorescence or transilluminated. The replaceable "sandwich" of glass slide, cover slips and gold electrodes is clamped within the aluminum parts shown.

Viewed from above, note the red wires connected to the electrodes. PEEK tubing (Upchurch) is connected via 0-dead-volume connectors of our own design. Four of six possible inlet ports are connected to this device.

Zone Electrophoresis (ZE)

By adding electrodes to the walls of a T-sensor, we found that we could use electrical forces to move biological particles (from proteins to parasites) from side to side in channels. The simplest form of electrokinetic separation, zone electrophoresis (ZE) is based on differences in electrophoretic mobility of the particles to move analytes from one side of a channel to the other a different rates. This generally requires injection of the mixture in the channel along a line.

The initial problem was to create devices that generated a uniform electric field across the bulk of the fluid in the microchannels. After attempts involving copper electrodes and copper electrodes onto which a thin layer of gold had been electroplated, we settled on a method of forming the channels with gold evaporated on Mylar. The Mylar pieces were then bent into the appropriate shape for the channel walls. These devices were our first long-term usable cells.

An intermediate version of the electrokinetic laminate device. Gold sputtered on Mylar was sandwiched between layers of Mylar and held together with pre-applied adhesive layers. This device is approximately the size of a standard microscope slide.

A later version of the electrokinetic laminate device in which the gold electrodes were replaced by palladium electrodes sandwiched between layers of Mylar, held together with pre-applied adhesive layers. Inter-electrode gap was either 2.5 or 1.27 mm.

Isoelectric Focusing (IEF)

Biological particles have surfaces with positive and negative titratable groups. Above a critical voltage, electrolysis of water at the surfaces of the electrodes creates a pH gradient across the channel perpendicular to the flow direction--alkaline at the cathode, acid at the anode. In isoelectric focusing (IEF), a particle migrates electrophoretically until it reaches fluid at its isoelectric pH (where it has no net charge). Isoelectric focusing is superior to "simple" electrophoresis because particles come to rest at a particular position in the channel, simplifying separation.

In a microfluidic channel between two electrodes, into which a buffer solution is pumped, we predicted that the initial uniform pH of the solution will be converted to a stable pH gradient as the rate of creation of acid and base would be just matched by their rates of recombination along the channel midline. The proteins, bacteria, or other (relatively large) charged particles would respond to both the field and the pH and equilibrate after the establishment of the pH gradient. They could then be sorted into separate outlet streams for analysis or storage. It was also possible that they could be identified optically by their position between the two electrodes.

Proper design of such a system would require modeling of the multiple processes occurring in the system.

This was potentially a very attractive method for manipulation of biological and inorganic particles in microchannels. However, there were many questions as to whether this would work. The first questions were:

Experiment Matches Modeled pH Gradients

We rapidly found that high concentrations of conventional pH indicator dyes could be used to monitor the pH in channels no thicker than a few hundred µm. Initial attempts to use sophisticated ratiometric fluorescent indicators like SNARF met with limited success. We ultimately developed a fully quantitative method for mapping the channel pH using optical absorptive dyes such as those described below. We found that when the concentration of the buffer was kept at or below about 1 mM, it was possible to generate pH gradients with wide ranges.

Demonstration of the evolution of a pH gradient in a microchannel using a pH indicator dye. This is a mosaic of a set of images taken at different times. Dye2- + H+ = HDye-. The pKa = 6.3 for bromocresol purple. V = 2.0V, I = ~6mA. Starting pH 6.3, electrode area = 0.41mm2

Location of acid and alkaline 'fronts': experimental vs. predicted

Once the first gold-on-Mylar cells were developed, which eliminated electrochemical degradation of the anodes found with gold-on-copper electrodes, it was found that experimental pH profiles matched the results from Cabrera's simple one-dimensional models within experimental error. As shown above, the columns are predicted values for percent of dye that is protonated (<10%=alkaline, >90%=acid); the lines are experimental values based on image intensity profiles taken from the image above. While agreement was best at about 10 minutes into an experiment, persistent problems occurred in static fluid with slow acidification of the channels at very long times. We continue to attribute that phenomenon to some form of oxidation the materials within the channels.

Formation of bubbles in the microchannels would have been a severe problem. We found that as long as the potential between the electrodes was kept below about 2.3V, bubbles did not form. Conversion from Au to Pd electrodes increased the current almost 10-fold, with no apparent negative effects. This does not mean that O2 and H2 were not present. In fact, we found evidence of high partial pressures of O2 in the channels, particularly in static fluid.

pH Gradient Formed in Flowing Saline

We soon found that pH gradients were formed not only in static fluid, but in flowing buffer as well. Shown below (as a mosaic of images from the channel entry at upper left to the channel exit a lower right) are results from a channel formed by 40 mm gold electroplated electrodes with a thickness of 0.127 mm. The space between the electrodes was 2.54 mm. A 0.2 mM solution of Phenol Red in 5 mM NaCl was used. The flow rate of the solution was 0.083 ml/s (99.3mm/s). A potential of 1.2 V was applied; current density ~ 1 mA/mm2.

Note that the pH gradient stabilized in the second window, and remained rock steady until the flow was distorted by the fact that all the fluid was exiting through a single port. This was extremely encouraging.

µIEF of a Protein

Having established that stable pH gradients could be established, we proceeded to attempt to focus three types of analytes--synthetic microparticles, bacteria (both vegetative and in spore form) and proteins. All of these can be considered CBW agent simulants. Shown below is a comparison of positions of the protein bovine serum albumin (BSA) during IEF for different initial pH values of the buffer used. Note that the pI of BSA is 4.6; focusing works best when the starting pH is near to the pI of the protein. The reasons for this have not been fully resolved, but are the subject of ongoing work in the laboratory at this time.

Flowing Transverse IEF Separation of Proteins

If each protein is focused to its own optimal location between electrodes, it should be possible to separate proteins. Initial efforts to do this were encouraging, but far from completely successful, as shown below.

Three images (above) near the outflow of a flowing stream. Note that while some separation is evident, there is strong overlap of the two labeled proteins, which differ by 1 pH unit in their reported isoelectric points.

Mean retention time of 4 min, total flow rate 4.8ml/min. 1mM MES, initial pH 5.41, 2.3V, 7-8 mA, electrode gap 1.25 mm, electrodes 400 mm high.

The separation between proteins was often found to be best shortly after turning on the voltage, and then to deteriorate over time (and space).

To our surprise, excellent separation was often found if the potential on the channel was flipped in polarity after establishment of a steady state separation. This was found to be true whether the fluid was static or flowing.

For example (at left), better separation of BSA-Alexa 594 (orange) from neutravidin-Bodipy-Fl (green, pI=~7) is seen soon after polarity switched. Conditions similar to those above.

We now believe that this improvement on switching of the polarity (and the related worsening in separation over time if the voltage is not switched) is related to parasitic electroösmotic pumping in the microdevices. There are several possible ways to exploit this phenomenon. However, we have found a convenient way to suppress the electroösmotic pumping and are now examining whether the "normal" operation of the system can be improved thereby. We are also using a combination of experiment and modeling to determine the ultimate resolution of the technique.

Bacteria Focus

Bacterial cells were also found to focus well using IEF, as well as to concentrate when the buffer concentration is so high that ZE occurs.

Shown above is a flowing bacterial stream segregating and focuses into two streams upon application of an electric field. A stream of E. coli (~1e6 cfu/ml) stained with Syto 13, a DNA intercalating fluorescent dye, and suspended in 1mM histidine, enters the bottom inlet; buffer enters from the top. The bottom outlet is blocked. The electrode separation is 2.54 mm; the electrode thickness was 0.4 mm, and channel length was 40 mm. The applied voltage was 2.3V, current 8 mA. The flow rate was 0.08 ml/s for each inlet. Total in-channel retention time ~4 min. It is not know what is in the upper stream, although we consider it possible that that was "naked" DNA that had leaked from lysed bacteria.

Bacteria Are Focused Well, but...

We have now repeatedly demonstrated concentration of bacteria using IEF into narrow streams in a ~5 minutes in low fields. Because of what we believe to be the aforementioned problems with parasitic electroösmotic pumping, we have had inconsistent results in separating and concentrating bacteria to date. We have seen formation of two streams from two different bacterial species, but it is not clear that the presence of two streams was not an artifact. This is under investigation at this time.


FABRICATION TECHNIQUES

Silicon Microfabrication

There were three approaches to device fabrication in this project. In the first, conventional Si microfabrication was used, combined with conventional anodic bonding to form channels. This has persisted because of the rigidity of the materials, and remains our preferred method for making channels that have a high aspect ratio. Such channels have been used extensively in our studies of the diffusion rates in microchannels.

By selective wet or dry etching of Si (in the WTC Microfabrication Laboratory), channels are formed along one surface and through it for I/O ports. Pyrex is anodically bonded to one surface to make a window for optical monitoring using absorbance or fluorescence. Optical imaging allows measurement of chemical concentrations vs. position across the channel.

Si microfabrication was one of our primary fabrication methods. Shown at left is a cartoon of an H-filter in operation. It includes the shape of devices now in use, in which the inlet channels curve. Note that there is still a dead spot in the flow velocity at the point(s) where the inlet and outlet channels meet. Si is shown as transparent to show through-holes.

A second method was the use of PDMS channels using either Si or SU-8 masters. This proved to be less convenient than the third method, described below.

WTC/Micronics Laser Prototyping System

Most of our microfluidic work in the last 2 years of the program was based on microdevices fabricated of Mylar and laminated. This process was originally developed jointly by UW and Micronics, Inc. Micronics established a facility for producing these laminates at their facility in Redmond, and a duplicate systems was subsequently established at the Washington Technology Center at no cost to this grant.

The Universal Laser Systems CO2 laser cutting system shown at left allows conversion from CAD file to assembled polymeric laminate in hours. This is a powerful new tool for rapid prototyping. Shown at right is a schematic of a portion of the high-throughput electrokinetic separation cell to indicate how alternating layers of Mylar (clear) and adhesive (blue) can be assembled to form very precise (and low cost) microdevices.


SCALE-UP AND SYSTEM INTEGRATION

High Throughput Electrokinetic Device

By rotating the flow direction of the H-filters, much greater contact area between the adjacent streams is possible, and scaling up throughput is just a matter of increasing the device width. It can be fabricated in Si or both moldable and laminated materials. A prototype high throughput system-compatible electrokinetic device based on the low-volume H-filters tested as shown above was fabricated in Mylar and palladium. This module was and tested for its ability to carry out a simple operation that would be the basis of future use in a high throughput CBW agent sample preconditioning device.

The design of an 11-layer Mylar laminate device high-throughput (100µL/min) flow cell with one inlet port and two outlet ports. It contains thin Pd foils (100 µm) on the top and bottom of the channel for electrodes. Two Mylar ridge supports that run the length of the channel that both support the channel walls and the gold spring probes.

Photograph of the device as tested. The knob-shaped item in the foreground is one of two gold spring-loaded hemispherical probes used to make electrical contact with the Pd foils. The electrode channel dimensions: 47 mm x 23 mm x 400 µm, and the overall device dimensions are 5 cm x 9.5 cm.

The operation of the device was tested using beads as model target antigens. The initial test of this device was to see if the imposition of the electric field could direct the beads to one of the outlets or the other, as shown below.

Test of high throughput electrokinetic device. Negatively charged 3 µm polystyrene beads (fluorescently labeled) were suspended in the same weakly buffered solution that we have used for IEF experiments (1 mM MES, 1 mM Na2SO4). The bead suspension was pumped into the device by its single inlet at a flow rate of 100 µL/min. Negatively charged particles focused as predicted in response to the applied electric field and the pH gradient created in the device. Particles could be steered to either outlet.

The longer residence time near the electrodes (walls) seems to lead to a lag in passage of particles driven to those surfaces. This will be studied in future studies.

System Integration

The ultimate goal of the project was to produce an integrated system that incorporated pumps, flow control, sedimentation, and the electrokinetic separation and concentration devices. We accomplished most but not all of these goals. The first demonstration integrated system included a battery operated NMPV pump pulling fluid from a stirred tank through the first sedimentation separation device. The stirred tank contained a mixed suspension of dense silica microspheres (to simulate an interferent in the output of an air sampler) and effectively neutrally buoyant blue latex microspheres (to represent a spore-like target analytes). The task for the system was to trap all the silica particles in the sedimentation separator, and to allow the latex microspheres to pass unimpeded. This system operated continuously for over 4 hours at one of the PI meetings, interrupted only by battery changes. As noted earlier, the separator suffered from accumulation of bubbles. The revision of the system containing the gas-removal system is shown below.

The second integrated system (at left) demonstrated the continuous removal of gases from the fluid in the sedimentation separation device. Vacuum for gas removal was generated by a simple syringe and stored in a commercial "vacutainer". All components were interconnected using PEEK tubing. The system was used to separate blue latex beads (sg=1) from silica (sg=1.8). Complete separation observed at an optimum flow rate of 200 mL/min.

The third and final integrated system of the project is shown above. This system included the same sedimentation separation and gas removal system as previously tested, followed by the electrokinetic separation module. The sedimentation separation module is held at 45°, and the electrokinetic device is held vertical to eliminate any effects of sedimentation on the part of the polystyrene latex microspheres. Two pumps (attached by green PEEK tubing and beyond the right edge of this photograph) were used to suck out the analyte-enriched and analyte-depleted suspensions.


 FUTURE SEPARATION SYSTEM CHALLENGES


 SELECTED RECENT SEPARATIONS PROJECT PUBLICATIONS

 

A.E. Kamholz and P. Yager, "Analysis of molecular diffusion in pressure-driven microfluidic channels", Biophysical Journal, submitted

M.R. Holl, K. Macounová, C.R. Cabrera, and P. Yager, "Microfluidic device and methods for continuous-flow transverse electrokinetic separations", Electrophoresis, submitted

C.R. Cabrera, B.A. Finlayson, and P. Yager, "Formation of natural pH gradients in a microfluidic device under flow conditions: Model and experimental validation", Analytical Chemistry, in revision

A. Hatch, A.E. Kamholz, G. Holman, P. Yager, and K.F. Böhringer, "A ferrofluidic magnetic micropump", Journal of MEMS, in revision

A. Hatch, A.E. Kamholz, B.H. Weigl, and P. Yager, "Microfluidic diffusion-based analysis: Immunoassay in a T-Sensor", Nature Biotechnology, in revision

K. Macounová, C.R. Cabrera, M.R. Holl, and P. Yager, "Development of natural pH gradients in microfluidic channels for use in isoelectric focusing", Analytical Chemistry, 2000; 72(16); 3745-3751.

P. Yager, C. Cabrera, A. Hatch, K. Hawkins, M. Holl, A. Kamholz, K. Macounová, and B.H. Weigl, "Analytical devices based on transverse transport in microchannels, .in Micro Total Analysis Systems 2000, van den Berg, Olthuis and Bergveld, eds., Kluwer Academic Publishers, Dordrecht, 2000, pp. 15-18.

M.R. Holl, K. Macounová, and P. Yager, "A microfluidic sedimentation particulate capture device with internal degassing membranes", in Micro Total Analysis Systems 2000, van den Berg, Olthuis and Bergveld, eds., Kluwer Academic Publishers, Dordrecht, 2000, pp. 319-322.

A.E. Kamholz, B.H. Weigl, B.A. Finlayson and P. Yager, "Quantitative analysis of molecular interaction in a microfluidic channel: the T-sensor", in Analytical Chemistry, 1999, vol. 23, pp. 5340-5347

B. H. Weigl, P. Yager, "Microfluidic diffusion-based separation and detection," Science, 1999, Vol. 283(5400), pp. 346-347.

P. Yager, D. Bell, J.P. Brody, D. Qin, C. Cabrera, A. Kamholz, B.H. Weigl, "Applying microfluidic chemical analytical systems to imperfect samples," in Micro Total Analysis Systems '98, D. J. Harrison and A. van den Berg, eds., 1998, Kluwer Academic Publishers, Dordrecht, pp. 207-212.

B.H. Weigl, P. Yager, "Silicon-microfabricated diffusion-based optical chemical sensor," in Sensors and Actuators B (Chemical) 1997, Vol. B39, pp. 452-457.

J.P. Brody, P. Yager, "Diffusion-based extraction in a microfabricated device," in Sensors and Actuators A (Physical), 1997, Vol. A58(1), pp. 13-18.


 Return to Yager group MicroFlumes Project Home Page

Return to Yager's Home Page

 

revised 8/30/03