UBC Undergraduate Research

Ultrafast microfluidic drop sorter Mulholland, Brendan 2011

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
Mulholland_B_et_al_APSC_459_2011.pdf [ 5.64MB ]
Mulholland_B_et_al_APSC_459_2011.pdf
Mulholland_B_et_al_APSC_459_2011.pdf
Metadata
JSON: 1.0074468.json
JSON-LD: 1.0074468+ld.json
RDF/XML (Pretty): 1.0074468.xml
RDF/JSON: 1.0074468+rdf.json
Turtle: 1.0074468+rdf-turtle.txt
N-Triples: 1.0074468+rdf-ntriples.txt
Citation
1.0074468.ris

Full Text

    Ultrafast Microfluidic Drop Sorter  Brendan Mulholland Daniel Da Costa Dale Eldridge  Project Sponsor: Dr. Carl Hansen  Applied Science 459 Engineering Physics The University of British Columbia April 4, 2011  Project Number 1107 ii  Executive Summary  This project sought to design, implement and test a high-throughput fluorescence measurement system capable of measuring the fluorescence of picolitre-volume aqueous droplets. Additionally, it is desired to sort these droplets into one of two populations based on the measured fluorescent intensity each individual droplet. While full implementation of this last objective was placed outside the scope of this project, research was conducted in order to direct future work. Ultimately, the system is intended to become a versatile tool that will be useful for a wide variety of applications in biology and biochemistry.  An optical system capable of measuring fluorescent emission from two distinct fluorescent dyes (specifically, fluorescein and Quasar 670, or equivalent) was designed.  Optical and mechanical components were specified, sourced and ordered. LabVIEW software was developed to preview and record data collected from the system.  MATLAB scripts were written to perform analysis of this data.  A microfluidic droplet sorting device described in a 2010 PNAS publication by Agresti et al. was replicated. Droplet-generating microfluidic devices were designed, fabricated and demonstrated to function as expected.  Measurement sensitivity (particularly to fluorescein), speed and robustness were identified as the critical indicators of system performance, and have been investigated in detail.  The sensitivity limit was determined to lie in the range 10 – 30 nM, the system was verified to be capable of measuring droplet fluorescence at droplets flow rates of at least 649 droplets per second and the system was demonstrated to make and record fluorescent measurement data continuously for almost 4 hours.  In order to physically sort droplets, it was determined that a high-voltage (~1kV), high-speed (1- 20kHz) amplification circuit was required. Research was conducted into the feasibility of constructing a custom amplification circuit.  It was determined that, while possible, this is a non-trivial design task which would require significant expertise and time commitment. Alternatives were investigated and it was determined that suitable commercial solutions are readily available, with the lowest end models starting at approximately $3000.  iii  Table of Contents Executive Summary ....................................................................................................................................... ii Table of Contents ......................................................................................................................................... iii List of Figures ................................................................................................................................................ v List of Tables .............................................................................................................................................. viii Glossary ........................................................................................................................................................ ix 1 Introduction ...................................................................................................................................... 11 2 Discussion ......................................................................................................................................... 13 2.1 Fluorescence ................................................................................................................................ 13 2.1.1 Dielectrophoresis ............................................................................................................... 14 2.1.2 Sorting ................................................................................................................................ 14 2.1.3 Circuit ................................................................................................................................. 15 2.2 Methods ...................................................................................................................................... 16 2.2.1 Droplets .............................................................................................................................. 16 2.2.2 Sorting ................................................................................................................................ 18 2.2.3 Optical System .................................................................................................................... 20 2.2.4 Data Acquisition ................................................................................................................. 23 2.2.5 Data Analysis ...................................................................................................................... 24 2.2.6 Safety .................................................................................................................................. 25 2.3 Testing Protocol ........................................................................................................................... 25 2.3.1 Droplet Stability .................................................................................................................. 25 2.3.2 Y-Junction Verification ....................................................................................................... 25 2.3.3 Droplet Steering ................................................................................................................. 26 2.3.4 Sensitivity ........................................................................................................................... 26 2.3.5 Speed .................................................................................................................................. 26 2.3.6 Robustness ......................................................................................................................... 27 2.4 Results ......................................................................................................................................... 27 2.4.1 Droplet Stability .................................................................................................................. 27 2.4.2 Y-Junction Verification ....................................................................................................... 27 2.4.3 Droplet Steering ................................................................................................................. 28 2.4.4 Sensitivity ........................................................................................................................... 28 iv  2.4.4.1 Series 1 ......................................................................................................................... 28 2.4.4.2 Series 2 ......................................................................................................................... 32 2.4.5 Speed ................................................................................................................................. 34 2.4.6 Robustness ........................................................................................................................ 35 2.4.7 Circuitry Issues .................................................................................................................. 37 2.4.8 Circuitry Alternatives ......................................................................................................... 38 3 Conclusions ....................................................................................................................................... 40 4 Project Deliverables .......................................................................................................................... 41 4.1 List of Deliverables ...................................................................................................................... 41 4.2 Financial Summary ...................................................................................................................... 41 5 Recommendations ............................................................................................................................ 45 5.1 Sorting ......................................................................................................................................... 45 5.2 Electrode Fabrication .................................................................................................................. 45 5.3 Optical Enclosure ......................................................................................................................... 45 5.4 Photon Counter Power Supply .................................................................................................... 46 5.5 Optical Alignment ........................................................................................................................ 46 5.6 Laser-Line Filters .......................................................................................................................... 46 5.7 Droplet Flow Rate vs. Sensitivity ................................................................................................. 46 5.8 Objective Lens ............................................................................................................................. 47 Appendix A – Original Project Description .................................................................................................. 48 Appendix B – Pictures of Experimental Setup ............................................................................................ 49 Appendix C – Dichroic Mirrors and Bandpass Filters .................................................................................. 60 Appendix D – Electrode Layout ................................................................................................................... 65 Appendix E – LabVIEW Interface ................................................................................................................. 67 Appendix F – MATLAB Data Analysis Scripts ............................................................................................... 80  Appendix H – Sensitivity Measurements, Series 2, Additional Figures ...................................................... 94 Appendix I – Robustness Experiment, Additional Figures ........................................................................ 100 References ................................................................................................................................................ 103 v  List of Figures Figure 1 - Absorbance and emission spectra for Quasar 670. .................................................................... 13 Figure 2 - Excitation and emission spectra for Fluorescein and Cy5 (similar to Quasar 670). .................... 14 Figure 3 - Microfluidic droplet sorting device ............................................................................................. 15 Figure 4 - Droplet generating microfluidic device ...................................................................................... 16 Figure 5 - Droplet generating junction geometry. ...................................................................................... 17 Figure 6 - Schematic representation of the microfluidic sorting device ..................................................... 18 Figure 7 - Measurement of the thickness of chrome plating used for electrode fabrication .................... 20 Figure 8 - Block diagram of optical measurement and droplet sorting systems. ....................................... 22 Figure 9 - Screen capture of front panel of LabVIEW interface .................................................................. 24 Figure 10 - Sorting junction showing all droplets flowing by default into the low impedance channel .... 28 Figure 11 - Raw data for sensitivity experiment using 50 μM Quasar, 50 nM Fluorescein ........................ 29 Figure 12 - Raw data for sensitivity experiment using 50 μM Quasar, 0 M Fluorescein ............................ 30 Figure 13 - Histogram for 50 μM Quasar, 50 nM Fluorescein .................................................................... 31 Figure 14 - Histogram for 50 μM Quasar, 0 M Fluorescein ........................................................................ 32 Figure 15 - Signal-to-noise ratios for Series 2 of sensitivity experiments ................................................... 33 Figure 16 - Sample of raw data collected during high droplet flow rate experiment ................................ 34 Figure 17 - Subset of raw data for robustness experiment ........................................................................ 36 Figure 18 - Histogram of data from robustness experiment ...................................................................... 37 Figure 19 - Front view of Optical Setup ...................................................................................................... 49 Figure 20 - Top view of Optical Setup ......................................................................................................... 50 Figure 21 - HeNe and Argon Ion Lasers ....................................................................................................... 51 Figure 22 - Front view of Optical Setup; Lasers on ..................................................................................... 51 Figure 23 - Top-down view of the focusing and fiber-optic optical setup .................................................. 52 Figure 24 - Side view of the focusing and fiber-optic optical setup ........................................................... 53 Figure 25 - Diagram Showing Mechanism of Optical Setup........................................................................ 54 Figure 26 - Photon Detector ....................................................................................................................... 55 Figure 27 - Photon Counter Power Supplies ............................................................................................... 56 Figure 28 - Power Supplies and Control For Lasers .................................................................................... 57 Figure 29 - Typical Microfluidic Device Mounting ...................................................................................... 58 Figure 30 - NI PCI 6601 Breakout Board ..................................................................................................... 59 vi  Figure 31 - Long Pass Dichroic..................................................................................................................... 60 Figure 32 - Dual-Edge Dichroic .................................................................................................................... 61 Figure 33 - Long Pass Dichroic; ................................................................................................................... 62 Figure 34 - Bandpass Filter; Transmits Wavelengths of 516 – 556 nm ....................................................... 63 Figure 35 - Bandpass Filter; Transmits Wavelengths of 665 – 705 nm ....................................................... 64 Figure 36 - Electrode Pattern for Droplet Sorting Device ........................................................................... 65 Figure 37 - Photo of fabricated chrome electrodes and microfluidic sorting devices ................................ 66 Figure 38 - Hierarchy diagram of the LabVIEW software. .......................................................................... 70 Figure 39 - Main LabVIEW VI block diagram screenshot showing code ..................................................... 72 Figure 40 - Main LabVIEW VI block diagram screenshot showing loop ...................................................... 73 Figure 41 - Main LabVIEW VI block diagram screenshot showing loop ...................................................... 73 Figure 42 - Main LabVIEW VI block diagram screenshot showing code which closes the output file. ....... 74 Figure 43 - Main LabVIEW VI block diagram screenshot showing code which opens the output file........ 74 Figure 44 - Main LabVIEW VI block diagram screenshot showing code ..................................................... 75 Figure 45 - Main LabVIEW VI block diagram screenshot showing code ..................................................... 75 Figure 46 - Main LabVIEW VI block diagram screenshot showing code ..................................................... 75 Figure 47 - Main LabVIEW VI block diagram screenshot showing code ..................................................... 76 Figure 48  Screenshot from the block diagram of the LabVIEW subVI “Data_to_string.vi”. ...................... 76 Figure 49 - Screenshot from the block diagram of the LabVIEW subVI “Update_Count_Array.vi". .......... 77 Figure 50 - Screenshot from the block diagram of the LabVIEW subVI “Get_Count_Array_Subset.vi" ..... 77 Figure 51 - Screenshot from the block diagram of the LabVIEW subVI “Get_Count_Array_Subset.vi" ..... 78 Figure 52 - Screenshot from the block diagram of the LabVIEW subVI “Get_Next_Sample.vi". ............... 78 Figure 53 - Screenshot from the block diagram of the LabVIEW subVI “ImageStream.vi". ....................... 79 Figure 54 - Sensitivity experiments, series 2: Raw data, [Fluorescein] = 100 nM. ..................................... 94 Figure 55 - Sensitivity experiments, series 2: Histogram, [Fluorescein] = 100 nM. .................................... 94 Figure 56 - Sensitivity experiments, series 2: Raw data, [Fluorescein] = 50 nM. ....................................... 95 Figure 57 - Sensitivity experiments, series 2: Histogram, [Fluorescein] = 50 nM. ...................................... 95 Figure 58 - Sensitivity experiments, series 2: Raw data, [Fluorescein] = 30 nM. ....................................... 96 Figure 59 - Sensitivity experiments, series 2: Histogram, [Fluorescein] = 30 nM. ...................................... 96 Figure 60 - Sensitivity experiments, series 2: Raw data, [Fluorescein] = 20 nM. ....................................... 97 Figure 61 - Sensitivity experiments, series 2: Histogram, [Fluorescein] = 20 nM. ...................................... 97 vii  Figure 62 - Sensitivity experiments, series 2: Raw data, [Fluorescein] = 10 nM. ....................................... 98 Figure 63 - Sensitivity experiments, series 2: Histogram, [Fluorescein] = 10 nM. ...................................... 98 Figure 64 - Sensitivity experiments, series 2: Raw data, [Fluorescein] = 0 M............................................. 99 Figure 65 - Sensitivity experiments, series 2: Histogram, [Fluorescein] = 0 M. .......................................... 99 Figure 66 - Average value of noise, in photons per millisecond, between each droplet ......................... 100 Figure 67 - Average value of signal, in photons per millisecond, counted in each droplet ...................... 101 Figure 68 - Ratio of photons per droplet, green signal, to photons per droplet, red signal..................... 102 viii  List of Tables Table 1 - Critical dimensions for droplet generating junction .................................................................... 17 Table 2 - Concentration series for sensitivity measurements .................................................................... 26 Table 3 - Signal-to-noise ratios for Series 2 sensitivity experiment ............................................................ 32 Table 4 - List of name, type and description of all numerical controls found on front panel of LabVIEW software. ..................................................................................................................................................... 68 Table 5 - List of name, type and description of all numerical indicators found on front panel of LabVIEW software. ..................................................................................................................................................... 69 Table 6 - List and description of all non-library subVIs found in the LabVIEW software. ........................... 70 ix  Glossary Collimated light Light whose rays are nearly parallel, and therefore will spread slowly as it propagates. Dichroic mirror A mirror that selectively reflects certain wavelengths of light and transmits others. Dielectrophoresis A phenomenon in which a force is exerted on a dielectric particle, which does not need to be charged, when it is subjected to a non- uniform electric field. Digital PCR A refinement of conventional PCR which allows precise quantification of nucleic acid concentration Dimethyl sulfoxide A colorless, liquid organosulfur compound with formula (CH3)2SO. DMSO is miscible in water. Used as the solvent for Quasar 670. FAM fluorescent dye A fluorescent dye which emits at 515nm when stimulated by 488nm light and is used to mark individual droplets. Fluorogenic substrate A non-fluorescent material that produces a fluorescent compound when acted upon by an enzyme. Fluorophore A fluorophore is a part of a molecule which absorbs light at a specific wavelength and subsequently emits light at a different wavelength. HFE-7500 A fluorocarbon oil used as a carrier fluid for aqueous droplets. Horseradish peroxidase An enzyme used in biochemistry to amplify weak signals and increase detectability of target molecules. x  Collimated light Light whose rays are nearly parallel, and therefore will spread slowly as it propagates. Neutral density filters A filter that reduces the intensity of all wavelengths of light equally. PEG Polyethylene glycol PFPE Oligomeric perfluorinated polyethers Polydimethylsiloxane (PDMS) A silicon-based polymer used as the body of microfluidic devices. Polymerase chain reaction (PCR) A scientific technique in molecular biology to amplify a single or a few copies of a piece of DNA across several orders of magnitude. Quasar 670 Carboxylic Acid An indocarbocyanine fluorescent dye which has an excitation maxima at 644nm and an emission maxima at 670nm. Surfactant A compound that lowers the surface tension of a liquid and is used to ensure that droplets do not coalesce. 11  1 Introduction  In 2010, Jeremy J. Agresti and the David A. Weitz Lab at Harvard University published a paper describing a novel method for screening large populations of biological samples using a microfluidic device. Droplets on the order of picolitre volumes were produced each containing a single yeast cell. The droplets were then sorted at rates of thousands per second to yield a population of the most efficient enzymes.  The droplets were sorted into one of two microfluidic channels according to the intensity of their fluorescence when excited by laser light. The fluorescence was produced by fluorophores present in each droplet. The highest fluorescent signals corresponded to the most efficient enzymes which were to be extracted.  Sorting was achieved by applying a high-voltage signal to electrodes embedded in the microfluidic device.  The signal created a time and spatially-varying electric field which in turn produces a force on the droplets in the device, directing them to the desired channel through the phenomenon of dielectrophoresis.  Commercial systems with similar objectives exist but do not compete in terms of speed and cost efficiency with the system developed by Agresti et al.  Fluorescence-activated cell sorting (FACS) allows high-throughput sorting of single cells at rates similar to those achieved by Agresti.  However, FACS systems require the cells to be suspended in a common solution and do not allow for isolation of the cells from each other in a droplet acting as a miniature bioreactor.  Advanced robotic systems are also available for large-volume testing which allow for isolation of single droplets but these cannot operate as fast or as economically as the method described in Agresti et al. (2010).  This project, sponsored by Dr. Carl Hansen from the Centre for High-Throughput Biology and UBC, sought to replicate the droplet screening and sorting methods developed at the Weitz lab at Harvard.  The development of an ultrafast microfluidic droplet sorter would allow for the biological testing such as directed evolution and digital PCR.  The objectives of the ultrafast microfluidic droplet sorter were firstly to design and implement a system for measuring the intensity of fluorescence produced by aqueous droplets suspended in oil. The fluorescence measurement system was to simultaneously acquire two fluorescence signals produced by 12  two different fluorescent dyes at droplet rates of approximately 1 – 2 kHz. Secondly, a system for sorting the droplets into two populations was to be constructed and verified. The sorting system was to consist of the microfluidic sorting junction chip itself and the electronic circuitry required to direct the droplets into one of the two channels on the chip. The ultimate goal of the integration of these two components into a fully functioning device however was deemed not feasible and was not included in the project objectives.  To quantifiably evaluate the performance of the fluorescence measurement system, the parameters measurement of sensitivity (particularly to Fluorescein), speed and robustness were identified as critical indicators of performance.  This report will include a discussion of the theory used in the course of the project, the testing procedures utilized, the results of these experiments, certain issues that were encountered, as well as conclusions and recommendations for the continuation of this project. The report is intended to convey all information learned and utilized in the design, fabrication and testing of the ultrafast microfluidic droplet sorter project to an audience including the project sponsor, the microfluidics lab personnel and the Engineering Physics project lab.   13  2 Discussion 2.1 Fluorescence  Fluorescence is produced by fluorescent molecules when they are excited by incident light at specific wavelengths. The emitted photons are slightly lower in energy than the exciting photons, and hence have longer wavelengths than the excitation wavelengths. This phenomenon is utilized in this project by measuring the fluorescence of droplets containing fluorescent dyes to characterize a population.  For this project, two lasers were used to excite the fluorescent dyes contained in the droplets and make fluorescent measurements. The two lasers are a 12mW red HeNe laser with wavelength of 632.8nm and a 20mW blue argon ion laser with a wavelength of 488nm. These lasers excite the fluorescent dyes Fluorescein and Quasar 670 (which is similar to Cy5), respectively. The peak emission wavelengths for these dyes are 521nm for Fluorescein, and 670nm for Quasar. Absorbance and emission spectra are shown in Figure 1 and Figure 2.  Figure 1 - Absorbance and emission spectra for Quasar 670. Note: from Bioresearch Technologies, Copyright 2011. 14    Figure 2 - Excitation and emission spectra for Fluorescein and Cy5 (similar to Quasar 670). Note: From Invitrogen Fluorescence SpectraViewer, Copyright 2010.  2.1.1 Dielectrophoresis Dielectrophoresis is a physical phenomenon where a force is exerted on a particle when subjected to a non-uniform electric field. Since the force does not rely on any charge on the particle, it is particularly suited to applications of microfluidics where producing a charge in a droplet is impractical or impossible. One complication of dielectrophoresis is that the force produced is highly dependent on the distances, field strength and geometries involved. For the case of a spherical particle of radius r in an oil with dielectric permittivity of εoil and an oscillation frequency in the kHz range, the time-averaged induced dipole force is    where k is a geometric factor determined by the location and shape of the electrodes (Ahn and Kerbage et al., 2006). 2.1.2 Sorting In order to sort the droplets into two populations a microfluidic device with a Y-junction was created (See Figure 3). Each path has associated impedance and the relative magnitude of the 15  impedance is determined by the length of a thin channel. This thin channel was designed to be twice as long along the keep channel than along the waste channel. The result is that one third of the oil and none of the droplets will go along this higher-impedance channel in the absence of an external force.  Figure 3 - Microfluidic droplet sorting device. Scale bar is 80um. (from Agresti et al. 2010) When a voltage is applied to the electrodes, a force is created on the droplets as they pass. This force will then pull the droplets into the higher-impedance channel. In this manner, the voltage across electrodes will directly dictate which of the two channels a droplet should enter. The pressure equalizing shunts to the right of the junction allow the junction to operate without interference from any pressure effects downstream (Agresti et al., 2010). In order to produce the required force to overcome the higher impedance of one channel, the required voltages are on the order of 1kV (Agresti et al., 2010) 2.1.3 Circuit As published in the Dielectrophoretic manipulation of drops for high-speed microfluidic sorting devices (Ahn and Kerbage et al., 2006), in addition to the high voltage requirements to produce a sorting force, the voltage should be pulsed at approximately 20kHz. This is to avoid charge build-up on the electrodes, which will cause DC screening and decreased effective voltage. Producing a logic on/off signal at the desired times and frequency is facilitated by the NI Breakout Board used to count photons. The output will be a pulsed 10V signal with a given duty cycle and frequency. In order to amplify this signal from 10V to 1.5kV, a high-voltage, high-speed amplifier is 16  required. This would turn on and off a voltage from an existing high voltage DC power supply to the electrodes. Of particular note is that this application has no load across the voltage. 2.2 Methods 2.2.1 Droplets  Droplets are formed using a standalone droplet generating device, as shown in Figure 4 below. This device forms droplets using a design that is commonly called flow-focusing geometry (shown in Figure 4 inset).  Figure 4 - Droplet generating microfluidic device  The dimensions of the junction were chosen to create droplets of approximately 30 to 60 μm in diameter with eight versions of varying junction width and height being fabricated. Since the ratio of oil flow to aqueous flow will affect droplet size, changing this ratio allows for some control in droplet diameter. In general, the droplet size decreases with a relative increase in oil flow rate to aqueous flow rate. Figure 5 below shows the critical dimensions in the droplet forming devices. Table 1 shows these dimensions for the constructed droplet generating devices. Two molds with heights 16 μm and 31 μm have been fabricated and are stored in the Hansen Lab cleanroom. Each of these molds will yield eight devices – two of each junction size. 17    Table 1 - Critical dimensions for droplet generating junction         The flow to produce droplets was driven using either two syringe pumps or a combination of one syringe pump and a low pressure regulator. When using both syringe pumps the syringe diameter must be correctly specified to obtain the desired flow rate. Experiments used 1.0 mL volume, 4.83 mm diameter syringes and typically used flow rates of 5.0 μL/minute for oil and 1.0 μL/minute for aqueous phase.These values were adjusted as needed to produce steady, uniform droplets.  The carrier fluid was HFE-7500 oil with a PFPE-PEG block copolymer surfactant, donated by RainDance technologies, added to prevent the coalescence of the droplets after formation. The D1 (μm) D2  (μm) Height ( μm) 8 27 16 12 40 16 20 67 16 30 100 16 8 27 31 12 40 31 20 67 31 30 100 31 Figure 5 - Droplet generating junction geometry. 18  surfactant was dissolved in the oil at 3% weight by weight. It has been shown by C. Holtze et al. that this type of PFPE-PEG block coplymer is suitable for stabilizing aqueous droplets in fluorocarbon oils while still maintaining biocompatibility (Holtze et al., 2008)  The aqueous solutions consist of the fluorescent dyes Fluorescein and Quasar 670 which were prepared in solution prior to injection into the droplet generating device.  The Quasar 670 was first dissolved in dimethyl sulfoxide (DMSO), due to low solubility in water, to create a stock concentration of 1 mM. This solution was then further diluted with water to the desired concentrations of 100 μM and 50 μM. Fluorescein was mixed into a stock solution at 10 mM in water. Series of mixtures with varying concentrations of Fluorescein from 1 μM down to 10 nM were mixed while the concentration of Quasar was kept constant at 100 μM throughout the series.  The mixed dyes were kept in the dark to avoid photobleaching effects while the stock solution of Quasar 670 was kept in a fridge. 2.2.2 Sorting  The droplet sorting chip was largely based on the design by the Weitz lab with some additional modifications to the basic structure (See Figure 6 below).  Figure 6 - Schematic representation of the microfluidic sorting device, adapted from the device presented by Agresti et al 19  The device is designed to sort droplets into two populations by using two electrodes to create a dielectrophoretic force directing the droplets into one of two channels. By design of the junction, all droplets will default into the upper channel in the absence of a dielectrophoretic force. The upper channel has half the fluidic impedance of the lower channel due to the length of the channel constrictions directly downstream of the junction. The smaller channels between both output channels are pressure equalizing shunts which isolate the junction from pressure  interference downstream (Agresti et al., 2010). The height of the output channels was increased in relation to the Weitz lab design. This increased the cross-sectional area of the channels, decreasing the droplet flow rate and allowing for observation of the droplets without the use of a high speed camera.  The microfluidic channels are fabricated in PDMS then plasma bonded onto a glass substrate slide. The electrodes were patterned in chrome (See Figure 7 below) on the glass substrate using a mask writer (Figure 37 in Appendix D). Since the chrome includes an anti-reflective layer which impedes conduction, the chrome contact pads were half etched to provide a conductive contact surface to attach signal leads. 20   Figure 7 - Measurement of the thickness of chrome plating used for electrode fabrication, showing that the thickness is on the order of 100 nm. This measurement was made using the surface profiling instrument found in the Hansen lab cleanroom (Alpha Step). As this instrument is normally used for measurements on the order of tens of micrometres, the accuracy of this measurement in unknown, but it does provide a useful order-of-magnitude estimate. Shown in the inset figure is an image from the instrument's camera prior to the measurement.  The white regions are the chrome plating and the darker region is the glass substrate.  The instrument uses a needle-like probe to scan from left to right, recording the change in height of the surface.  2.2.3 Optical System  The optical system for fluorescent measurement consists of the two lasers and the optics to combine and focus the beams onto the microfluidic chip. The optical system then collects the fluorescent light and transmits it to the photon counter.  A dichroic mirror first combines the lasers into a single beam. This beam is then directed by another mirror through a 50/50 beamsplitter (See Appendix C for detailed description of dichroic mirrors). The beamsplitter passes 50% of the beam through, while the remaining 50% is redirected towards a beam block. The beamsplitter is used later in the path to direct an image of the microfluidic 21  chip onto a CCD. This CCD is used to view real-time images of droplets flowing through the microfluidic device. Note that the CCD is only useful when used with a 10X magnification (or lower) objective, since higher magnifications restrict the field of view. With higher magnification objectives (20X and 40X) the beamsplitter was removed to decrease the power lost to the beam block. After the beamsplitter (assuming it is in place) the beam is redirected by a dual pass dichroic mirror. This mirror reflects the beam through the objective and onto a microfluidic device. The incident combined laser beam excites fluorescent dyes present in droplets as they flow through a microfluidic device. A portion of the stimulated emission will pass back through the objective used to focus the laser light.  Both the emitted fluorescence and any reflected laser light again encounters the dual pass dichroic mirror which transmits the two emission spectra while reflecting the laser light back towards the beamsplitter (if present). The two fluorescent wavelengths are split by a long-pass dichroic mirror, bandpass filtered then focused onto the fiber optic couplers. The fiber optic couplers carry each fluorescent signal to a channels on the photon counter. Figure 8 below shows the a block diagram of the fluorescent measurement and droplet sorting systems. 22   Figure 8 - Block diagram of optical measurement and droplet sorting systems.  Alignment of the lasers and optical components is a crucial in producing the strongest possible fluorescent signal. In order to properly align the components, the first step was to align the laser so that both beams were collinear. Since the blue argon ion laser cannot be adjusted, the red HeNe laser was adjusted for height and angle with both the adjustable laser mount and an adjustable mirror until both 23  laser spots were aligned at two different points along the beam path. Following combination of the lasers, the beam was directed toward the dual pass dichroic mirror. It is important that the beam strikes the dichroic at 45° ± 1.5° (See Appendix C). This dual pass dichroic reflects the beam through the objective to be focused on the droplets in the microfluidic device. Horizontal alignment and focusing of the laser spot on the microfluidic channel was accomplished with the manual X-Y stage and the linear Z stage using the image obtained from the beamsplitter and CCD. When the beamsplitter and CCD were not in place to view the device, alignment and focusing were accomplished by viewing the plot of the photon counts and aligning in order to obtain the maximum signal.  The final and perhaps most pivotal alignment and focusing step is that of the fluorescent spots to the fiber optic couplers. The coupler mounts have horizontal and vertical adjustability as well as focal adjustment. These adjustments are completed by viewing the photon counts and aligning to the maximum.  2.2.4 Data Acquisition  Data acquisition of the fluorescent signal is accomplished with the photon counter which is interfaced to a computer with a National Instruments data acquisition card (PCI 6601). The photon counter has four channels, two of which are used to count the fluorescent signals.  Whenever a photon is “counted” the photon counter outputs a voltage spike. Each voltage spike over a specific time period is counted by the NI DAQ. The NI DAQ uses interrupt based sampling which allows a maximum sampling rate of about 7kHz. Direct memory access (DMA) is another available acquisition mode with a maximum sampling rate of 44kHz but this mode only allows for sampling of one channel. The count is read from the DAQ by the PC running LabVIEW. A custom LabVIEW interface has been developed which allows for simultaneous viewing of the CCD image as well as incoming photon counts per millisecond (See Figure 9). The interface then allows for image or video capture and data recording. A capture card allows for interface of the CCD with the PC. Appendix E gives a detailed description of the LabVIEW interface and its operation. 24   Figure 9 - Screen capture of front panel of LabVIEW interface. On the left is the CCD image. On the right is the incoming raw data of photons per millisecond. 2.2.5 Data Analysis  MATLAB is used for all analysis of the collected fluorescence data. A script is used which includes all required steps to the data analysis. When the script is run, a smoothing function is first applied to the signal. This smoothing allows a threshold to be applied to the signal from the Quasar 670 without disturbances from data glitches that sometimes occur. A signal above this threshold indicates the presence of a droplet. Once the droplets have been detected, it is possible to determine the fluorescent magnitudes of the droplet.  The fluorescence magnitudes of each droplet can then be compared.  Although the use of a smooth signal before applying a threshold eliminates some noise, smoothing also has the effect of increasing the width of a detected droplet.  However, using a droplet width larger than the actual width creates a reduction in the calculation of the average photon rate per droplet, since some non-droplet photon counts are included in the average. The end result of this averaging is that the signal-to-noise rations are understated.  This has not been investigated in detail but, by inspection of several data sets, it is estimated that the effect may be as large as 25% or more  Following thresholding the photon counts which fall inside an identified droplet are integrated and normalized. The noise present in the absence of a droplet is also integrated and normalized to 25  provide a signal-to-noise ratio. Finally, a histogram is generated which shows the frequency versus fluorescent intensity of droplets in units of photons per millisecond. 2.2.6 Safety  Both the argon ion and HeNe lasers are designated class 3B. These lasers are hazardous to the eyes if directly exposed, but reflections from diffusing surfaces are not harmful. During testing and initial adjustment of the lasers, laser safety goggles were worn. When adjusting the red HeNe laser the blue goggles were worn and when adjusting the blue argon ion laser the orange goggles were worn.  In addition to laser goggles, the laser curtain was kept closed during operation of the lasers and Class 3B laser warning signs were posted. 2.3 Testing Protocol  To test the characteristics of the sorting and fluorescent measurement systems, several experiments were devised and performed. The tests run on the fluorescent measurement system included sensitivity, speed, robustness and stability over long run times. On the sorting system fewer tests were performed because the high voltage switching circuit circuit was not feasible to complete. The tests included verification of sorting junction action in the absence of an electric field and sorting tests with DC voltages. The following sections describe the protocols used to perform these tests. The Results section provides the results of these tests. 2.3.1 Droplet Stability The stability of droplets was tested to ensure that the surfactant would keep the droplets from coalescing over a period of time after formation and extraction. The droplets were formed with a droplet generating microfluidic device which output the droplets into a pipette tip. The droplets were then extracted with a pipette and transferred to a sample tube where they were stored overnight. The following day the droplets were observed under a microscope and reinjected into a sorting device to test that they could be transferred and reinjected. 2.3.2 Y-Junction Verification The droplet sorting device was first verified to direct all droplets into the waste channel in the absence of an electric field. This was done by flowing droplets into the device and imaging the output channels with the CCD. 26  2.3.3 Droplet Steering Further experiments were done to test if a DC electric signal applied to the sorting device’s electrodes would direct the stream of droplets into the keep channel. A 2.5 kV signal was applied to the on-chip electrodes while droplets flowed through the sorting junction. 2.3.4 Sensitivity  Sensitivity of the fluorescent measurement system was done by measuring a series of fluorescent dye mixtures with varying concentrations of Quasar 670 and Fluorescein (see Table 2). Experiments were done using two different objective lenses: 20X and 40X. Both of the lenses were Nikon Plan Fluor type. The experiment with the 20X objective had the beamsplitter in place while the experiments with the 40X objective did not. For all experiments the oil was driven by syringe pump while the aqueous dyes were driven either by another syringe pump or pressure. Rates and pressures are given in Table 2. All of these measurements were performed directly on the droplet generating device. The sampling rate for the first series was 2.5k kHz. The sampling rate for the second series was 5 kHz. Table 2 - Concentration series for sensitivity measurements Series Fluorescein (nM) Quasar 670 (μM) Oil injection rate (μL / min) Dye injection rate (μL / min) Dye injection pressure (psig) Junction width (μm) Junction height (μm) Objective Lens 1 50 50 5.0 0.5  30 31 20  0 50 5.0 0.5  30 31 20 2 100 100 1.0  1.8 20 16 40  50 100 1.0  1.8 20 16 40  30 100 1.0  1.8 20 16 40  20 100 1.0  1.8 20 16 40  10 100 1.0  1.8 20 16 40  0 100 1.0  1.8 20 16 40  2.3.5 Speed  To test the fluorescent measurement speed, a population of fluorescent droplets was formed using the syringe pumps and measured directly downstream of the droplet generating junction. The 27  junction was 31 μm high and 12 μm wide. The concentrations of Quasar 670 and Fluorescein were 50 μM and 10nM respectively.  The flow rates of both the aqueous phase and oil were increased until the aqueous phase became a streamline after which the flow rates were decreased until droplets reappeared. The resulting injection rates were 10 μL / min of oil and 0.4 μL / min of dye. After allowing the high speed droplets to stabilize, data was recorded at a sampling rate of 5 kHz for 32 seconds. The objective used was the 10X Nikon Plan Fluor and the beamsplitter was removed for this experiment. 2.3.6 Robustness  Testing the robustness of the data acquisition and stability of the measurements was done by measuring a large population of droplets over a long period of time. The concentration of Quasar 670 was 50 μM while the concentration of Fluorescein was 1 μM. Prior tests were done over a span of minutes while this test was carried out over almost four hours. The 10X Nikon Plan Fluor objective was used during this test with the beam splitter in place. Pressure of 1.5psi was used to drive the aqueous phase while a syringe pump rate of 1.0 μL/minute was used to drive the oil. The droplet producing junction used for this test was 16 μm high and 20 μm wide.  Samples were taken at 5000 kHz. 2.4 Results 2.4.1 Droplet Stability After storage overnight in a sample tube, a sample of the droplet emulsion was transferred with a pipette onto a microscope slide for inspection. The droplets were observed to be uniform in size which indicated that coalescence did not occur. 2.4.2 Y-Junction Verification All droplets were observed to flow into the low impedance channel as designed. This experiment was captured on video; Figure 10 shows a single frame from this video. 28   Figure 10 - Sorting junction showing all droplets flowing by default into the low impedance channel in the absence of an electric field. 2.4.3 Droplet Steering Initial testing of droplet steering using a time-invariant electric field alone was attempted early in the project (November 2010). These tests were intended to demonstrate that a continuous stream of droplets could be directed down either of the two channels by the application of a time-invariant electric field. The sorting device present in the methods section was used for this test and voltage of up to 2.5kV was applied to the electrodes. No effect on droplet direction was observed. 2.4.4 Sensitivity 2.4.4.1 Series 1 29   Figure 11  and Figure 12 below show subsets of the raw data collected during the first series of sensitivity experiments. The first run with 50nM Fluorescein acquired 61 seconds of data with an average of 33.8 drops per second. The signals-to-noise ratios of the first run were 143.0 and 8.4 respectively for Quasar 670 and Fluorescein. The control run with zero Fluorescein also acquired 61 seconds of data averaging 30.8 drops per second. The signals-to-noise ratios of the control run were 77.2 and 1.1 respectively for Quasar 670 and Fluorescein.  Figure 11 - Raw data for sensitivity experiment using 50 μM Quasar, 50 nM Fluorescein, sampling rate 2.5 kHz. Blue dotted line shows the threshold used for data analysis. 30   Figure 12 - Raw data for sensitivity experiment using 50 μM Quasar, 0 M Fluorescein, sampling rate 2.5 kHz. Blue dotted line shows the threshold used for data analysis.  Histograms for the first series of sensitivity runs are shown in Figure 13 and Figure 14. Figure 13 shows two distinct peaks on each channel corresponding to the noise and the signal. Figure 14 shows two peaks on the Quasar channel but only one peak corresponding to the noise on the Fluorescein channel.    31   Figure 13 - Histogram for 50 μM Quasar, 50 nM Fluorescein, sampling rate 2.5 kHz. 32   Figure 14 - Histogram for 50 μM Quasar, 0 M Fluorescein, sampling rate 2.5 kHz. 2.4.4.2 Series 2 Six sets of data were acquired for this series of experiments. Data analysis was completed on each set and the resulting figures of raw data and histograms are included in Appendix H. The signals-to- noise ratios, run times and average drops per second are summarized in Table 3 below. Table 3 - Signal-to-noise ratios for Series 2 sensitivity experiment Fluorescein (nM) Quasar 670 (μM) Fluorescein S/N Ratio Quasar 670 S/N Ratio Run time (s) Average droplets/s 100 100 4.1 9.1 614 12.6 50 100 3.5 2.7 714 15.2 30 100 6.0 9.4 888 16.0 20 100 4.0 10.6 815 24.7 10 100 1.86 6.0 679 10.8 0 100 1.29 7.2 986 5.6  33  The signals-to-noise ratios obtained in each experiment were compared and the resulting data set plotted in the Figure 15 below.  Figure 15 - Signal-to-noise ratios for Series 2 of sensitivity experiments and ratio of signal-to-noise ratios. The preliminary sensitivity measurements obtained in from Series 1 demonstrated that a clear distinction between 50 nM and 0 M of Fluorescein can be made using a 20X objective. The signal to noise ratio for 50 nM Fluorescein was 8.4 versus 1.1 for 0 M Fluorescein. Sensitivity measurements for Series 2 showed that the limit of detection is lower than 50 nM when using 40X objective lens. When performing these measurements, a strong dependence on alignment of the laser to the droplet channel and fiber-optic couplers to emission light was observed. Minor adjustments were required between runs; this requirement is likely cause of non-linearity in the signal-to-noise ratio versus concentration plots seen in Figure 15.  Nevertheless, the ratio of signal-to- 34  noise ratios was observed to be approximately linear for the range of 50 nM to 0 M Fluorescein, as would be ideally expected. Determination of a specific sensitivity limit depends on the requirements of the application; for the tested system sensitivity is in the range of 10 – 30 nM. 2.4.5 Speed A subset of the raw data acquired during this experiment is shown in Figure 16. The data was analyzed using the methods described in the Data Analysis section. Over the 32 seconds of data recorded, 20707 droplets were detected for an average flow rate of 649 droplets per second. This measurement rate was not limited by the data collection, but rather by the ability to form droplets at this speed. The signal-to-noise ratio from the Quasar 670 signal was 33.7 while the signal-to-noise ratio from the Fluorescein signal was 0.97. This indicates that Fluorescein was not detectable at 10 nM at these flow rates.  Figure 16 - Sample of raw data collected during high droplet flow rate experiment. Blue dotted line shows the threshold used in data analysis. 35   The above results show that droplets can be produced and measured at rates of at least 650 droplets per second. The upper limit of droplet measurement rate for this experiment was not limited by the data acquisition system but by the droplet generation device. At 650 droplets per second, a set of 1,000,000 droplets would be could be measured in approximately 26 minutes. If the population of droplets was formed separately prior to measurement, the flow rate could theoretically be increased passed the limit imposed by the use of syringe pumps by using pressure driven flow.   During this experiment, a 10nM Fluorescein concentration was not visible but this may be attributable to poor laser alignment and optical isolation. 2.4.6 Robustness A subset of the raw data acquired during the robustness test is shown below in Figure 17. The data was analyzed using the methods described in the Data Analysis section. Over the 3.83 hours of recorded data 185,476 droplets were detected for an average flow rate of 13.4 droplets per second. A total of 68,838,329 samples were collected at 5 kHz. A histogram of the data was created and is shown in Figure 18. A clear distinction is observed between the signal and the noise. The signal-to-noise ratio was calculated as 6.7 and 18.5 for the Quasar 670 and Fluorescein signals, respectively. Additional figures in Appendix I show the variability of the signal and noise levels with time. 36   Figure 17 - Subset of raw data for robustness experiment. Blue dotted line represents the threshold used to analyze data. 37   Figure 18 - Histogram of data from robustness experiment. Upper and lower peaks correspond to signal and noise, respectively.   This experiment demonstrated the ability of the system to run for an extended period of time and produce consistent data.  The only theoretical limit on the run time is dictated by the size of the computer’s hard disk. A set of 185,476 droplets was measured over 3.83 hours at an average rate of 13.4 droplets per second. At this rate, a population of 1,000,000 droplets would be measureable in approximately 21 hours. If the rate were increased to 100 droplets per second – well within the demonstrated droplet measurement rate – the same population would then take 2.8 hours. 2.4.7 Circuitry Issues Initially, it was thought that an amplification circuit could be trivially assembled using techniques similar to a lower voltage switching circuit. Unfortunately, investigation quickly revealed that the circuit was not as simple as initially expected. It was indicated that an ideal solution for such a circuit would use a half bridge driver to control the switching. Although half bridge drivers carry the added security of many forms of protections to ensure that external components are not broken, these protections are 38  also very complicated and guaranteeing that all required conditions are always met is a complicated task. Further, it was found that component selection for a half bridge circuit is very difficult and error- prone. Managing the charges and parasitic capacitance is a task that requires very careful component selection that depends on many output attributes that aren't necessarily known during the design phase. In particular, when doing high frequency switching, it is crucial to carefully manage parasitic capacitance in the various components (see Appendix G). Finally, any solid state based solution will require a minimum load. This load should be sufficient to cause a current value of 10% of the max current through the switching transistors (A. Dunford, personal communication, February 25, 2011). For the sourced transistors, this is on the order of 100mA, well above the 10mA maximum current supply from the power supply available (Stanford Research Systems, 2011). In addition to supply issues, a load required to produce a current of 100mA at 1.5kV will have to dissipate 150W of heat. 2.4.8 Circuitry Alternatives Due to the many issues surrounding a custom half bridge driver circuit, it was decided to abandon this portion of the project. Driving factors behind this decision were the difficulty level, time constraints, skepticism from experts in the field and a lack of high-voltage testing equipment. Alternative options are an avalanche transistor design and various commercial amplification solutions. The avalanche circuit design in Nanoseconds Switching for High Voltage Circuit Using Avalanche Transistors (Tapuri et al., 2009) switches voltages at up to 4.5kV with a falling time of 2.89ns. This is much more than this application could require, despite being a relatively simple design. Further, the design eliminates need for a minimum load and should remove issues with a minimum load. However, any custom circuit design will require proper testing equipment, namely a high voltage Oscilloscope. The total cost of this transistor based design is estimated at $300. However, there would also be associated design and assembly time requirements. Commercial solutions should solve all issues, as they are guaranteed to work, do not require any testing and include internal loads. One possibility is that the steering issue in not caused by a lack of voltage pulsing. This is certainly a possibility and, if so, there are some other parameters that could be easily changed to 39  improve the situation. In particular, the fabrication method of the probes means that the probes could be formed as a point closer to the center of the fluid channel, as in Dielectric Manipulation of Drops for High-Speed Microfluidic Sorting Devices (Ahn and Kerbage et al., 2006). Varying the shape and location of the probes could have a substantial affect on the effectiveness of the set up. It is also possible to use the external voltage control of the power supply to trigger some slower voltage pulses (16Hz update rate) (Stanford Research Systems, 2011). 40  3 Conclusions The objective of this project was to implement a system capable of measuring the fluorescence of picolitre-volume aqueous droplets and, based on these measurements, sort the droplets into one of two populations.  Measurement sensitivity (particularly to Fluorescein), speed and robustness were identified as the critical indicators of system performance. The system described by Agresti et al was identified as an ideal model system which met the goals of our project. The microfluidic droplet sorting device described by Agresti was replicated and demonstrated to operate correctly in the absence of an electric field. Droplet-generating microfluidic devices were designed, fabricated and demonstrated to function as expected. An optical system capable of measuring fluorescent emission from both fluorescein and Quasar 670 (or equivalent) was designed.  Optical and mechanical components were specified, sourced and ordered.  The optical system was implemented and the sensitivity, robustness and maximum measurement rate was assessed. The lower detection limit of the optical system to fluorescein fluorescent dye was determined to lie in the range of 10 – 30 nM.  Robustness of the system was assessed by allowing the system to run for almost four hours.  Over this period of time, 68,838,329 samples were recorded, representing 185,476 individual fluorescence measurements. Finally, the system was verified to be capable of measuring droplet fluorescence at droplets flow rates of 649 droplets per second and it is expected that the system will be able to function at even higher speeds. In order to physically sort droplets, it was determined that a high-voltage (~1kV), high-speed (1- 20kHz) amplification circuit was required. Research was conducted into the feasibility of constructing a custom amplification circuit.  It was determined that, while possible, this is a non-trivial design task which would require significant expertise and time commitment. Alternatives were investigated and it was determined that suitable commercial solutions are readily available, with the lowest end models starting at approximately $3000. 41  4 Project Deliverables 4.1 List of Deliverables Deliverable Status Details Handover Medium Optical Fluorescence Measurement Setup Working Various improvements to optical stability could be made. In Lab Amplification Circuit Incomplete Design recommendations for future project have been made. N/A LabVIEW software Working Currently only collects fluorescence data. Drop detection and monitoring is a future task. CD/DVD MATLAB scripts Complete Analyzes gathered data and performs data analysis. CD/DVD AutoCAD drawings of Microfluidic Devices Complete Includes all designs used during the course of the project. CD/DVD Silicon-photoresist molds and corresponding photo-lithographic masks Complete Includes all designs used during the course of the project. In Clean Room Some fully fabricated devices Complete Includes all designs used during the course of the project. In Lab Experimental data Complete Characterizes sensitivity and speed of the optical setup. CD/DVD Lab books N/A  Handover Proposal Complete  CD/DVD Recommendation Report Complete  CD/DVD and Handover Presentation Complete  Given in person  4.2 Financial Summary 42  Description Part Number Quantity Unit Price Total Price Supplier HeNe Laser, 632.8 nm, 12.0 mW, Random polarization HRR120 1 $1,620 $1,620 Thorlabs Laser Safety Glasses, Teal Lenses, 35% Visible Light Transmission (> OD3 @ 633nm) LG7 2 $203 $406 Thorlabs Kinematic V-Clamp Mount, One PM2 Clamping Arm Included   C1503 1 $221 $221 Thorlabs Ø1.5" Mounting Post, Length=10", 1/4"-20 Taps P10 2 $60 $119 Thorlabs Right Angle Kinematic Mirror Mount, 30 mm Cage System, SM1 KCB1 1 $104 $104 Thorlabs SM1 Threaded Kinematic Mount for Ø1" Optics KC1-T 1 $89 $89 Thorlabs Beam Block (Active Area: 0.7" X 1.4"), Includes TR3 Post LB1 1 $45 $45 Thorlabs Ø1/2" x 10" Stainless Steel Optical Post, 8-32 Stud, 1/4"-20 Tapped Hole TR10 4 $9 $36 Thorlabs 90° #8 Counterbored Ø1/2" Optical Construction Post TR3C 2 $14 $28 Thorlabs SM1 (Ø1.035"-40) Coupler, External Threads, 1" Long SM1T10  1 $20 $20 Thorlabs Alignment Plate for LMR1 Mount and SM1 Lens Tubes  LMR1AP 1 $20 $20 Thorlabs Adapter with External M27 x 0.75 Threads and Internal SM1 Threads SM1A18 1 $20 $20 Thorlabs Large Clamping Arm, 6-32 Threads PM2 1 $15 $15 Thorlabs Ø1" Lens Tube Cover SC1L24 1 $8 $8 Thorlabs Epoxy-Encased LED, 635 nm, 4 mW LED631E 2 $3 $6 Thorlabs Epoxy-Encased LED, 635 nm, 7.2 mW, Qty. of 5 LED630E 1 $6 $6 Thorlabs  STP4N150 MOSFET 497-5091-5- 5 $6 $32 Digikey 43  ND       536/40 nm BrightLine® single- band bandpass filter FF01- 536/40-25 1 $275 $275 Semrock 685/40 nm BrightLine® single- band bandpass filter FF01- 685/40-25 1 $325 $325 Semrock  BrightLine single-edge dichroic beamsplitter with edge at 605 nm FF605-Di02- 25x36 1 $225 $225 Semrock LaserMUX dichroic beamsplitter with edge at 503 nm LM01-503- 25 1 $195 $195 Semrock BrightLine dual-edge dichroic beamsplitter with edges at 500 & 646 nm FF500/646- Di01-25x36 1 $295 $295 Semrock Beamsplitter Mount BSM 2 $195 $390 Semrock  Bellofram High Flow Precision Air Regulator 1/4", 0-2 PSI 82150 1 $81 $81 Industrial Automation  Power Supply - Analog Triple Output DC 30V/3A 9294 1 $130 $130 Sparkfun Electronics  Coaxial Cable 7032K32 10 $1.33  $13.30 McMaster Carr Brass T 4429K251 2 $6.64  $13.28 McMaster Carr Standard-Wall Brass Threaded Pipe Nipple 1/4" Pipe Size X 2" L, 9/16" Thread Length 4568K133 3 $1.82  $5.46 McMaster Carr Standard-Wall Brass Threaded Pipe Nipple 1/4" Pipe Size X 3" L, 9/16" Thread Length 4568K135 3 $2.31  $6.93 McMaster Carr Pipe Sealant Tape, 0.25` 4591K11 2 $1.40  $2.80 McMaster Carr Pipe Sealant Tape, 0.5` 4591K12 2 $1.74  $3.48 McMaster Carr HI-Pressure Brass Single-Barbed Tube Fitting Adapter for 1/8" Tube ID X 1/4" NPT Male Pipe 50745K36 3 $9.10  $27.30 McMaster Carr 18-8 SS Male-Female Threaded Hex Standoff 3/8" Hex, 3/4" Length, 8-32 Screw Size 91075A164 4 $3.92  $15.68 McMaster Carr 44   Plug; Ber. Copper; SHV 517-1015 4 $24 $97 Allied Electronics connector, shv(high voltage) front mounting bulkhead receptacle, solder cup 517-1090 4 $11 $43 Allied Electronics  Quasar 670 Carboxylic Acid FC-1065-10 1 $50 $50 Biosearch Total: $4,989.23  45  5 Recommendations 5.1 Sorting Commercial solutions, although expensive, are guaranteed to work and would save significant amounts of time. For example, the Trek 677B is quoted at $3385 and includes an internal load (Trek Inc., 2011). However, this model just barely meets the amplification requirements. Specifically, it had a very slow rise time of 1ms and could supply 1.5kV max, which does not allow for testing at higher voltages. Trek also makes more appropriate choices, though these will, of course, cost more. The avalanche transistor design, as mentioned above, is a cheaper alternative. The component cost would be around $300. However, this will require testing and assembly time as well as the use of a high voltage oscilloscope. It is also recommended to test various changes to the probe design by creating a simple microfluidic chip that flows droplets into a large channel with a probe at the entrance. This manner, the force that a probe has on droplets can be observed in a quantifiable manner and probe designs can be tested and optimized. This design with a probe near the center of the stream should be investigated in detail. 5.2 Electrode Fabrication While the current method of fabricating electrodes (chrome plated glass substrates) is simple, reliable and inexpensive, droplet steering has not been demonstrated with electrodes made with this technique.  It is speculated that the geometry of these electrodes could lead to an unfavorable electric field distribution. Alternate fabrication methods, such as injecting a molten low-melting point alloy into microfluidic channel, can yield electrodes with 3-dimensional geometry. These electrodes may produce more favorable electric fields.  It is recommended that alternate fabrication methods be investigated. Computational modeling (i.e. COMSOL) of this problem may be highly useful. 5.3 Optical Enclosure Currently, the optical system is completely exposed to all ambient light in the room.  Enclosing the system in a light-proof enclosure is inexpensive and simple and would greatly reduce fluorescent measurement noise.  In addition, this should remove the inconvenient need to turn the lights out in the lab and improve the safety of the system. 46  5.4 Photon Counter Power Supply The power supplies used with the photon counter only marginally meet the requirements.  In particular, the warning indicator on the power supply used to supply 2 V often turns on. It is suspected that this may be leading to increased noise levels. It is recommended that the suitability of these power supplies be re-evaluated and replaced if necessary. 5.5 Optical Alignment Setting and maintaining accurate alignment of optical components has been a particularly challenging and time-consuming task throughout the course of this project.  Specifically, fluorescent measurement efficiency is extremely sensitive to the alignment of the fiber-optic couplers (positioning accuracy on the order of 100 µm or less is required).  Unfortunately, the post and post-holder mounting system used does not allow for easy adjustment, and is not ridged enough to maintain alignment if bumped.  It is recommended that the mounting system used with the fiber-optic couplers be re- evaluated. To aid in alignment of the fiber-optic couplers, it has been suggested that a light pen or diode laser be directed through the optical fibers in reverse and aligned with the primary lasers at the sample. 5.6 Laser-Line Filters Both lasers produce observable light components which overlap with the fluorescent spectrum of fluorescein.  This is easily demonstrated by passing either laser through the 535 nm bandpass filter used for fluorescein emission – faint green light will be transmitted through the filter.  For roughly $200 each, laser line filters are available which would greatly reduce this unwanted light and thus may reduce fluorescent measurement noise.  It is recommended that the potential reduction in measurement noise be evaluated and, if warranted, that laser-line filters be installed for one or both lasers. 5.7 Droplet Flow Rate vs. Sensitivity In the experimental results presented in this report, sensitivity was a major subject of investigation. In addition, droplet flow rate was increased and it was verified that measurements were possible at rates as high as at least 650 droplets per second.  However, sensitivity was never investigated as a function of droplet flow rate. It is expected that sensitivity will decrease with increasing droplet flow rate, though the actual behavior is unknown. It is recommended that this behavior be investigated. 47  5.8 Objective Lens  In this project, several objective lenses were used with good results (Nikon Plan Fluor 10X, 20X and 40X). However, no direct comparisons were made between these objectives.  The objective lens is a critical element of the optical system and it is therefore recommended that the choice of objective be evaluated. 48  Appendix A – Original Project Description As stated by the project sponsor: Objective 1: Students will build a microfluidic fluorescent droplet sorter. A microfluidic device will be provided to flow microdroplets, having diameters of ~30 microns, past a focused laser spot (Argon Ion laser, 488 nm) at a rate of ~1000 Hz.  Fluorescence from the droplets will be collected through a microscope objective and projected onto an avalanche diode photon counter.  Photon counting using a high-speed data acquisition card will be used to determine the distribution of fluorescent drops across hundreds of thousands of droplets.   An improvement to this will include the addition of a second red laser to allow for a relative measure of fluorescence.  Objective 2: The microfluidic device will be improved to include an electrode upstream of a junction leading to two collection channels.  Application of an electric field across the channel junction will be used to direct droplets to one of the collection channels, thus creating a droplet sorter.  Electronics will be designed and fabricated to allow for droplet sorting based on a defined fluorescent threshold 49  Appendix B – Pictures of Experimental Setup  Figure 19 - Front view of Optical Setup. Shows both lasers (far left); laser power supplies and control (right of lasers); pressure tubing and control (front center); photon counter (middle center); dichroic and directional mirrors (rear); optical setup, dichroics and opto-couplers (middle right); and control computer (back right). 50   Figure 20 - Top view of Optical Setup. Shows Argon Ion laser (bottom left); laser power supplies and control (right of laser); pressure tubing and control (bottom center); photon detector (middle center); NI PCI 6601 breakout board (left of keyboard); dichroic and directional mirrors (rear); optical setup, dichroics and opto-couplers (bottom right); and control computer (top right). 51    Figure 21 - HeNe and Argon Ion Lasers. Shows HeNe laser (top) and Argon laser (bottom). Note the exhaust tubing for the higher powered argon ion laser; this goes to a fan that removes excess heat.   Figure 22 - Front view of Optical Setup; Lasers on. The far left shows the mirror directing the red light from the HeNe laser along the back wall. Immediately to the right is a dichroic mirror reflecting the blue light from the argon ion laser (and passing the red laser) along the same path as the red laser. On the right, the combined beam is directed towards the objective and microfluidic device. 52     Figure 23 - Top-down view of the focusing and fiber-optic optical setup. In the foreground is the objective in use (interchangeable); the picture shows a 10X Nikon objective. Below the objective, not visible, is a series of dichroic mirrors that direct the fluorescent light to the left or top of the picture into the mounted assemblies (opto-couplers). The assembly on the left (wrapped in purple) receives the green fluorescence signal and directs this light into a fiber-optic cable. The assembly on the top (wrapped in green) receives the red fluorescence and directs this light into a fiber-optic cable. Above this assembly is a beam splitter, which reflects parts of the image to a CCD for live video of the microfluidic device. At the top left of the picture, part of the photon detector is visible; at the top right, part of the CCD. 53   Figure 24 - Side view of the focusing and fiber-optic optical setup. The assembly in the foreground (wrapped in purple) receives the green fluorescence signal and directs this light into a fiber-optic cable. The assembly at the top (wrapped in green) receives the green fluorescence signal and directs this light into a fiber-optic cable. Above this assembly is a beam splitter, which reflects parts of the image to a CCD for live video of the microfluidic device. The CCD is visible at the top middle of the picture. The reverse-L-shaped parts on the right are composed of 4 main components. On the top (not visible) is an interchangeable objective. Below the objective, the component with a downwards-facing white arrow is a long-pass dichroic mirror mounted at 45° that transmits the fluorescent signal. Both laser beams come in from the beam splitter, on the left, and reflect off the dichroic mirror upwards towards the microfluidic device. The stimulated fluorescence then passes back downwards and through the aforementioned dichroic to a mirror at the bend in the L-shaped part. This mirror is also mounted at 45°, which reflects all light towards the left. Next, the fluorescence hits another dichroic mirror that separates the two types of fluorescence. The green fluorescence is directed towards the assembly in the bottom left of the picture (wrapped in purple), which uses a band pass filter to eliminate noise. The red fluorescence passes through the dichroic towards the assembly on the left of the picture (wrapped in green), which also uses a band pass filter to eliminate noise. 54   Figure 25 - Diagram Showing Mechanism of Optical Setup. This diagram is a 3D illustration of the L- shaped components below the XY translation stage. Note that the laser beams from the HeNe and Argon Ion lasers come in from the left in the illustration, whereas the physical setup uses an entry port for the laser beams facing the same direction as the tube at the bottom of the assembly. Laser light comes in and reflects off a dichroic upwards towards a microfluidic device. This light stimulates fluorescent emission (labeled as Quasar 670 and Fluorescein), some of which comes downwards from the sample and passes through the dichroic that reflects the laser light. The fluorescence is then reflected to the horizontal and split into two separate beams using another dichroic.  55   Figure 26 - Photon Detector. This receives the light passed through the fiber-optic coupler and outputs a voltage spike each time a photon is received. Note that this is a 4 channel photon detector but only two channels are in use; the two capped knobs on the right are the other two unused channels. 56   Figure 27 - Photon Counter Power Supplies. Two power supplies were required to meet the stringent requirements of the photon detector, which needs 3 voltage supplies. Of particular note is that the lower supply (+2V) is on its own ground circuit, as requested by the photon detector. The top power supply is used for 5V and 30V. Both power supplies must be turned on for correct photon detector operation. 57    Figure 28 - Power Supplies and Control For Lasers. The top box controls the HeNe (red; 12mW) laser; this only has an on/off key switch. The Argon Ion (blue; 20mW) laser is controlled by the lower box. Since the Argon Ion laser is more powerful, it has a few more controls than its on/off key. In particular, it also has an on/off switch (bottom left of box), a high/low power switch (bottom right of box; down is low power) and a current or intensity control, top of box. Only one of intensity or current may be controlled at a time; this is selected by the switch between the two knobs. 58    Figure 29 - Typical Microfluidic Device Mounting. The microfluidic device is taped across the aperture in a custom slide mounted onto the XY translation stage. 59    Figure 30 - NI PCI 6601 Breakout Board. This receives the voltage spikes from the photon detector and counts them. The count is then read periodically from the computer in order to determine how many photons have arrived since the last sample.   60  Appendix C – Dichroic Mirrors and Bandpass Filters  The optical system contains three dichroic mirrors and two bandpass filters. Each was specifically chosen to meet the requirements of the optical system and all were purchased from Semrock Inc.  Figure 31 through Figure 35 show plots of transmission (%) vs. wavelength (nm) for each dichroic mirror and bandpass filter, as provided by Semrock on their website. Beam Combiner  The long-pass dichroic mirror, Semrock part number LM01-503-25, is used to combine the two laser beams into a single beam.  It is a long-pass dichroic beam splitter with an edge wavelength of 503 nm as shown in Figure 31. Semrock specifies transmission of >95% for wavelengths in the range 514.5 to 647.1 nm and reflection of >98% for wavelengths in the range 473.0 to 491.0 nm. This is a circular component, with a 25 mm diameter and a thickness of 2.0 mm.  Figure 31 - Long Pass Dichroic; Edge Wavelength of 503 nm (Beam Combiner)  61  Emission-Excitation Separator  The dual-edge dichroic beam splitter, Semrock part number FF500/646-Di01-25x36, is used to separate the excitation light (laser light) from the emission light (fluorescent light emitted from sample). It has edge wavelengths at 500 nm and 646 nm, as shown in Figure 32. Semrock specifies reflection of >95% for wavelengths in the ranges 486 to 490 nm and 632 to 634 nm. Semrock also specifies transmission of >90% for wavelengths in the ranges 505 to 613 nm and 653 to 750 nm. It is very important to note that these specifications only apply for collimated light with an incidence angle of 45 ± 1.5 degrees. This is a rectangular component, with dimensions of 25.2 mm x 35.6 mm x 1.05 mm.  Figure 32 - Dual-Edge Dichroic; Blocks Wavelengths of 486 – 490 nm and 632 – 634 nm  62  Fluorescent Separator  The long-pass dichroic beam splitter, Semrock part number FF605-Di02-25x36, is used to separate the fluorescent emission of fluorescein from that of Cy5 (or equivalent). It has an edge wavelength of 605 nm, as shown in Figure 33. Semrock specifies transmission of >93% for wavelengths in the range 612 to 950nm and reflection of >98% for wavelengths in the range 350 to 596 nm.  This is a rectangular component, with dimensions of 25.2 mm x 35.6 mm x 1.05 mm.  Figure 33 - Long Pass Dichroic; Edge Wavelength of 605 nm  63  Fluorescein Emission Filter  The bandpass filter used to isolate fluorescein emission, Semrock part number FF01-536/40-25, has a center wavelength of 536 nm. A plot of transmission vs. wavelength is shown in Figure 34. Semrock specifies transmission of >93% for wavelengths in the range 516 to 556 nm.  In addition, Semrock specifies blocking of optical density (OD) > 6.5 at 488 nm (argon ion laser) and OD > 5 at 633 nm (HeNe laser).  This is a circular component with a radius of 25 mm.  Figure 34 - Bandpass Filter; Transmits Wavelengths of 516 – 556 nm  64  Quasar 670 Emission Filter  The bandpass filter used to isolate Quasar 670 emission, Semrock part number FF01-685/40-25, has a center wavelength of 685 nm.  A plot of transmission vs. wavelength is shown in Figure 35. Semrock specifies transmission of >90% for wavelengths in the range 665 to 705 nm.  In addition, Semrock specifies blocking of OD > 6 at 488 nm (argon ion laser) and 633 nm (HeNe laser).  This is a circular component with a radius of 25 mm.  Figure 35 - Bandpass Filter; Transmits Wavelengths of 665 – 705 nm   65  Appendix D – Electrode Layout  Figure 36 - Electrode Pattern for Droplet Sorting Device. The orange regions shown are chrome that will not be etched off of the glass slide. The overall dimensions are 127mm x 127mm. Note the large squares on both sides of the chip; these are electrical contact pads for connecting an external voltage source.  66   Figure 37 - Photo of fabricated chrome electrodes and microfluidic sorting devices. The overall dimensions of the glass substrate are 127mm x 127mm.    67  Appendix E – LabVIEW Interface A program written using National Instruments’ (NI) LabVIEW is used to collect data from the photon counter (via the NI PCI 6601) and to collect images from the camera (via an NI PCI frame grabber). At time of writing, the latest version of this program had the filename “Photon Counter MAIN 31-03-2011”. The main purpose of the software is to display a preview of the data being collected from both the photon counter, view the camera output, and save data from the photon counter and/or camera. Program Requirements The computer running the software must have LabVIEW 2010 and NI Imaq (for the camera) installed. Drivers for the NI PCI 6601 (counter) and frame grabber must also be installed. Output Data Formats Camera Single images will be saved as uncompressed tiff images. Videos will be saved as uncompressed AVI videos. These AVI videos will be very large files (~500MB/min) and post-processing compression is usually required to reduce the files to a manageable size. Photon Counter Data from the photon counter is saved to text files. Each file has three tad-delimited columns of data. From left to right, the columns are: counter 0 photon count (green signal), counter 1 photon count (red signal) and time in milliseconds. Photon count data is given as the number of photons counted since the most recent sample. For example, if the sampling rate is 2000Hz, then photon counts would be in units of photons per half millisecond. When analyzing data recorded with sampling rates near or above 1000Hz, it is important to note that the recorded timestamps are given in milliseconds, and therefore consecutive timestamps may often have equal value.   68  Setting the Sampling Rate The NI-DAQmx task ‘GatePulse-Ctr3’ is used to generate a square wave which is internally routed to the gate of each counter on the NI PCI-6601. The count is then sampled on each rising edge of this signal. In order to change the sampling rate, this task must be edited through NI ‘Measurement & Automation Explorer’. Once the program is running, browse to ‘My System -> Data Neighborhood -> GatePulse-Ctr3’. Once there, the high and low time for the signal can be specified. Generally, a 50% duty cycle is used. Note that National Instruments specifies a maximum sampling rate of 7 kHz when using interrupt based sampling in continuous operation. If the sampling rate is set above this value, the software will be unable to keep up and an error message will be displayed.  The alternative to interrupt-based sampling, direct-memory-access (DMA), allows sampling rates as high as 44 kHz. Unfortunately, the PCI 6601 only supports one DMA channel, not the required two. See the NI 660x User Manual (in particular, table 2-1) for more information. Front Panel – Photon Counter Press the button ‘Start Counting Photons’ button to begin collecting and displaying data from the photon counter. Press this button again to stop the preview. Press the button ‘Begin Saving Photon Count Data to Disk’ to begin streaming data to a text file. Press the button again to stop streaming data and close the file. When this button is pressed, the user will be prompted to enter a path and filename for the output file. This prompt will default to the path specified by the control ‘Output File Start Path’, located on the secondary front panel tab ‘Settings and Stuff’. Table 4 lists each numerical control and gives its name, data type and a description. Table 5 lists each numerical indicator and gives its name, data type and a description. Table 4 - List of name, type and description of all numerical controls found on front panel of LabVIEW software. Name Type Description Disp. Data Points Long Integer (I32) Specifies the number of photon count samples which are displayed at any given time. 69  Plot Update Period (ms) Long Integer (I32) Specifies the time, in milliseconds, between successive updates to the photon count plots, as well as the average counts, Elapsed Time and Actual Sampling Rate. Increasing this value may improve performance. Num Samples to Avg Long Integer (I32) Specifies the number of samples with are averaged in order to compute the average photon counts (‘Cntr 1, Mean’ and ‘Cntr 2, Mean’). This value must be less than or equal to ‘Count Array Size’. Sampling Rate (kHz) Double Specifies the expected sampling rate. This value is used to convert the photon count rate into photons per ms. Note that this only affects displayed values – values saved to disk will have units of photons per time T, where T is the sampling period. Count Array Size Long Integer (I32) Specifies the size of the circular array which stores the photon count data for preview. Decreasing this value may improve performance, but limits the number of samples which can be displayed at any given time. Changes made to this control while the program is running will not come into effect until the program is restarted. yMax Double Specifies the upper limit on the y-axis of the photon count display. yMin Double Specifies the lower limit on the y-axis of the photon count display.  Table 5 - List of name, type and description of all numerical indicators found on front panel of LabVIEW software. Name Type Description Cntr 0 (Grn) Double Instantaneous photon count on counter 0 (green), in units of photons per ms. Cntr 0, Mean Double Average photon count (counter 0, green) over the most recent x samples, where x is specified by the control “Num Samples to Avg”. Units are photons per ms. Cntr 1 (Red) Double Instantaneous photon count on counter 1 (red), in units of photons per ms. Cntr 1, Mean Double Average photon count (counter 1, red) over the most recent x samples, where x is specified by the control “Num Samples to Avg”. Units are photons per ms. Actual Sampling Rate (Hz) Double The actual sampling rate, calculated as (Total Samples Collected)/(Elapsed Time). Total Samples Collected Long Integer (I32) The total number of samples collected since “Start Counting Photons” was last pressed. Elapsed Time String The elapsed time since “Start Counting Photons” was last activated.  70  Front Panel Controls – Camera Press the button ‘Start Camera’ to begin previewing the image from the camera. Once the preview is running, the controls ‘Capture AVI’ and ‘Capture Single Image’ can be used to save an uncompressed AVI video or uncompressed tiff image to the disk, respectively. In both cases, the user will be prompted to enter a path and filename for the output file. As for the photon count output file, this prompt will default to the path specified by the control ‘Output File Start Path’, located on the secondary front panel tab ‘Settings and Stuff’. To end AVI capture, press the ‘Capture AVI’ button a second time. The control ‘Rotation Angle’ will rotate both the displayed image and any saved images or videos counter-clockwise by the specified angle (in degrees). List of SubVIs and VI Hierarchy Figure 38 shows the hierarchy of the LabVIEW software (only non-library sub-Vis are shown). Table 6 lists all non-library subVIs and provides a description of each.  Figure 38 - Hierarchy diagram of the LabVIEW software.  Table 6 - List and description of all non-library subVIs found in the LabVIEW software. VI Filename (Icon Text) Description Get_Count_Array_Subset (GET COUNT Returns a subset of stored photon counts of specified length. Used for calculation of the mean photon count rate and for the 71  ARRAY SUBSET) graphical displays. Get_Elapsed_Time (GET ELAPSED TIME) Given a start and end time in ms, this VI calculates the time difference and returns this value in numerical format and string format. Data_to_string (DATA TO STRING) Converts the photon count rates and timestamp to a string which will be written to the output text file. Update_Count_Array (UPDATE COUNT ARRAY) Writes the photon count rates and timestamp to the global array “count array”. This is a custom-implemented circular array. Get_Next_Sample (GET NEXT SAMPLE) Gets the next available sample from both counters. Calc_Frame_Rate (FRAME RATE Hz) Used to calculate the camera’s frame rate. Averages the rate over the previous 20 frames. ImageStream (IMAGE STREAM) Creates a new, empty IMAQ image. Images acquired from the camera are written onto this image.  Block Diagram Screenshots Screenshots have been taken of critical block diagram elements of both the main VI and key non-library subVIs (Calc_Frame_Rate and Get_Elapsed_Time are omitted). Screenshots from Main VI 72   Figure 39 - Main LabVIEW VI block diagram screenshot showing code which aquires data from the photon counter and writes this data to file. 73   Figure 40 - Main LabVIEW VI block diagram screenshot showing loop which acquires images from camera (vieo capture enabled).  Figure 41 - Main LabVIEW VI block diagram screenshot showing loop which acquires images from camera (video capture disabled). 74   Figure 42 - Main LabVIEW VI block diagram screenshot showing code which closes the output file.  Figure 43 - Main LabVIEW VI block diagram screenshot showing code which opens the output file.  75  Figure 44 - Main LabVIEW VI block diagram screenshot showing code which sets the scale of the photon count preview plots.  Figure 45 - Main LabVIEW VI block diagram screenshot showing code which calculates the actual sampling rate and elapsed time.  Figure 46 - Main LabVIEW VI block diagram screenshot showing code which updates the photon count preview plots. 76   Figure 47 - Main LabVIEW VI block diagram screenshot showing code which calculates the average photon count. Screenshots from SubVIs  Figure 48  Screenshot from the block diagram of the LabVIEW subVI “Data_to_string.vi”.  77   Figure 49 - Screenshot from the block diagram of the LabVIEW subVI “Update_Count_Array.vi".  Figure 50 - Screenshot from the block diagram of the LabVIEW subVI “Get_Count_Array_Subset.vi", for the case where we have filled the circular array.  78   Figure 51 - Screenshot from the block diagram of the LabVIEW subVI “Get_Count_Array_Subset.vi", for the case where the circular array has not been completely filled.   Figure 52 - Screenshot from the block diagram of the LabVIEW subVI “Get_Next_Sample.vi". 79   Figure 53 - Screenshot from the block diagram of the LabVIEW subVI “ImageStream.vi".  80  Appendix F – MATLAB Data Analysis Scripts Custom MATLAB scripts have been written to analyze the large data sets generated by the fluorescence measurement system.  The main goal of this analysis is to allow visualization of these data sets and to calculate quantitative metrics from which the performance of the system can be judged.  The analysis is divided into steps, each of which has its own m-file.  The script “main_script.m” is the highest-level script and is used to execute all the others. The flow chart shown in figure x shows each of the basic steps taken in the analysis and provides a brief explanation.  This basic procedure was followed in generating all fluorescent measurement data presented in this report.   setup_data.m • Creates vectors of raw data for use in later calculations • Calculates the sampling period and frequency and the the length of input data vectors • Creates a vector of evenly spaced timepoints smooth_data.m • Creates a copy of the  original photon count data vectors and applies the MATLAB function smooth.m to them • Smooth.m applies a moving average filter to the data • Smoothing reduces thresholding errors caused by random noise threshold_data.m • The smoothed red channel photon count data is thresholded in order separate signal from noise • The threshold is chosen emperically plot_raw_data.m • Plots the raw data for a specified starting index and number of indices (generally, the data files will be too large to plot in full) • Also plots the smoothed raw data with the threshold value shown 81   integrate_data.m • Integrates the photon counts accross each region of signal and noise separetely, and stores the results in two vectors.  This is done for both the red channel data and the green channel data, using the results from thresholding the red channel data. • Normalizes these integrals by dividing y the time interval which was integrted over, effectively calculating the average photons per ms for each signal or noise interval. • Generates several plots which are useful in assessing the effectivesness of the algorithm. hist.m • Create combines the normalized signal and noise integral vectors for both the red channel data and the green chanel data and generates two histograms, one fore each channel. calc_info.m • Calculates useful information from the results of the preceeding analysis.  This includes the number of samples analyzed, number of droplets analyzed, total experiement time and the droplet flow rate. Also calculated is the mean, mode and variance of the fluorescent intensity for the noice and signal on both channels. • The signal to noise ratio is calculated by taking the ration of the average photons per ms for the signal to the average phtons per ms of the noise. savetodisk.m • This script automates the process of saving figures to the disk. • All figures whose handles which were previously stored in a vector are saved, as both MATLAB figure files (.fig) and image files (.tif). 82  main_script.m % Start diary % This records everything displayed to the command window to a file diary('Command Window Print.txt'); diary on;  % Store a vector of handles to each plot we create % This allows savetodisk.m to automatically save each figure to the disk plt_hndl_vec = [];  % Plot Formatting NL = sprintf('\n');     % Newline character ax_fnt_sz = 14;         % Font size, plot axis labels ttl_fnt_sz = 16;        % Font size, plot titles  setup_data;             % Setup data, calculate sampling rate, ect.  smooth_data;            % Apply smoothing function to data  threshold_data;         % Apply threshold to data  plot_raw_data;          % Create plot of raw data and smoothed data  integrate_data;         % Integrate data over each droplet  hist;                   % Create histogram  calc_info;              % Calculate some relevant statistics  diary off;              % Stop diary  savetodisk;             % Save all figure to disk, as .fig and .tif  setup_data.m % Setup Data to analyze  % data_file is a Lx3 matrix % The columns of data_file are the green photon count, red photon count, % and timestamps (ms) y = data_file(1:end,:);  y_grn = y(:,1);               % Photon count, green channel y_red = y(:,2);               % Photon count, red channel t_actual = y(:,3);            % Recorded timestamps, in ms  L = length(y);                % Number of samples  % Calculate sampling frequency (Hz) Fs = 1000*L / (t_actual(end)-t_actual(1)); 83   T = 1/Fs;                     % Sampling period (s) t = (0:L-1)*T;                % Time vector (s) total_time = (L-1)*T;         % Total time (s)  clearvars y;                  % Delete y, to ensure it isn't used later  smooth_data.m is_smoothing_enabled = true;  % Set to false to bypass smoothing  smooth_size = 50;    % Sets the number of samples which are averaged  span = smooth_size/L;  if(is_smoothing_enabled)     % Smooth data     y_grn_smth = smooth(y_grn,span,'lowess');     y_red_smth = smooth(y_red,span,'lowess'); else     % Bypass smoothing     y_grn_smth = y_grn;     y_red_smth = y_red; end  threshold_data.m % Specify threshold thresh = 200;  % Apply Threshold to data dig_sig_rd = y_red_shift_smth > thresh;  y_thresh_red = dig_sig_rd.*y_red(:); y_thresh_grn = dig_sig_rd.*y_grn(:);  plot_raw_data.m num_points = 2000;      % Number of samples to plot start_index = 1;        % First sample to plot  x_i = start_index:(start_index+num_points); t_i = 1:num_points+1;  % y-axis plotting limits yMin_r = 0; yMax_r = 5000; yMin_g = 0; yMax_g = 1500;  84  num_subplots = 2;  % Create Figure figure; set(gcf,'name','Raw Data'); plt_hndl_vec=[plt_hndl_vec,gcf]; % Set plot properties that are common to each subplot for n = 1:num_subplots     subplot(num_subplots,1,n); hold on; grid on;     xlim([t(t_i(1))  t(t_i(end))]);     xlabel('Time (s)','fontsize',ax_fnt_sz);     ylabel(['Fluorescent Intensity',NL,'(Photons per ms)'],...         'fontsize',ax_fnt_sz); end  subplot(num_subplots,1,1) title('Channel 1 - Quasar 670 (100\mu{}M)','fontsize',ttl_fnt_sz); plot(t(t_i),y_red(x_i)*(Fs/1000),'r'); plot([t(t_i(1)) t(t_i(end))],[thresh thresh]*(Fs/1000),'--b'); ylim([yMin_r yMax_r]);  subplot(num_subplots,1,2); title('Channel 2 - Fluorescein (1\mu{}M)','fontsize',ttl_fnt_sz); plot(t(t_i),y_grn(x_i)*(Fs/1000),'g'); ylim([yMin_g yMax_g]);  % Create Figure figure; set(gcf,'name','Raw Data, Smoothed'); plt_hndl_vec=[plt_hndl_vec,gcf]; % Set plot properties that are common to each subplot for n = 1:num_subplots     subplot(num_subplots,1,n); hold on; grid on;     xlim([t(t_i(1))  t(t_i(end))]);     xlabel('Time (s)','fontsize',ax_fnt_sz);     ylabel(['Fluorescent Intensity',NL,'(Photons per ms)'],...         'fontsize',ax_fnt_sz); end  subplot(num_subplots,1,1) title('Channel 1 - Quasar 670 (100\mu{}M) (Smoothed)','fontsize',ttl_fnt_sz); plot(t(t_i),y_red_smth(x_i)*(Fs/1000),'r'); plot([t(t_i(1)) t(t_i(end))],[thresh thresh]*(Fs/1000),'--b'); ylim([yMin_r yMax_r]);  subplot(num_subplots,1,2); title('Channel 2 - Fluorescein (1\mu{}M) (Smoothed)','fontsize',ttl_fnt_sz); plot(t(t_i),y_grn_smth(x_i)*(Fs/1000),'g'); ylim([yMin_g yMax_g]);  integrate_data.m % Integrate thresholded data int_vec_rd = [];                % Integral of red signal int_vec_grn = [];               % Integral of green signal 85  int_vec_inv_rd = [];            % Integral where no signal, red int_vec_inv_grn = [];           % Integral where no signal, green int_vec_count = [];             % Samples per droplet int_vec_inv_count = [];         % Samples between droplets  droplet_start_times = [];       % Index of first sample in each integral droplet_start_times_inv = [];   % Index of last sample +1 in each integral  int_red = 0; int_grn = 0; int_count = 0;  currently_integrating = false; for n = 1:(L-grn_lag)     if( (dig_sig_rd(n) ~= 0) && (~currently_integrating) )         currently_integrating = true;         if(int_count~=0)             int_vec_inv_count = [int_vec_inv_count,int_count];             int_vec_inv_rd = [int_vec_inv_rd,int_red];             int_vec_inv_grn = [int_vec_inv_grn,int_grn];             droplet_start_times_inv = [droplet_start_times_inv,n];         end         int_red = 0;         int_grn = 0;         int_count = 0;     elseif((dig_sig_rd(n) == 0) && (currently_integrating))         currently_integrating = false;         if(int_count~=0)             int_vec_count = [int_vec_count, int_count];             int_vec_rd = [int_vec_rd, int_red];             int_vec_grn = [int_vec_grn, int_grn];             droplet_start_times = [droplet_start_times,n];         end         int_red = 0;         int_grn = 0;         int_count = 0;      end     int_red = int_red + y_red(n);     int_grn = int_grn + y_grn(n);     int_count = int_count+1; end  % Calculate Normalized signal and noise int_norm_grn = int_vec_grn./int_vec_count*(Fs/1000); int_norm_red = int_vec_rd./int_vec_count*(Fs/1000);  int_norm_grn_noise = int_vec_inv_grn./int_vec_inv_count*(Fs/1000); int_norm_red_noise = int_vec_inv_rd./int_vec_inv_count*(Fs/1000);  % *** Generate some figures which are useful for evaluating analysis % performance  % Plot: Signal, Averaged over each Droplet 86  figure; plt_hndl_vec=[plt_hndl_vec,gcf]; set(gcf,'name','Signal, Averaged over each Droplet'); plot(t(droplet_start_times)/60,int_norm_red,'r'); hold on; grid on; plot(t(droplet_start_times)/60,int_norm_grn,'g'); xlabel('Time (min)','fontsize',ax_fnt_sz); ylabel(['Fluorescent Intensity',NL,'(Photons/ms)'],'fontsize',ax_fnt_sz); title(get(gcf,'name'),'fontsize',ttl_fnt_sz); legend('Quasar 670 (100\mu{}M)','Fluorescein (1\mu{}M)');  % Plot: Noise, Averaged between each Droplet figure; plt_hndl_vec=[plt_hndl_vec,gcf]; set(gcf,'name','Noise, Averaged between each Droplet'); plot(t(droplet_start_times_inv)/60,int_norm_red_noise,'r'); hold on; grid on; plot(t(droplet_start_times_inv)/60,int_norm_grn_noise,'g'); xlabel('Time (min)','fontsize',ax_fnt_sz); ylabel(['Fluorescent Intensity',NL,'(Photons/ms)'],'fontsize',ax_fnt_sz); title(get(gcf,'name'),'fontsize',ttl_fnt_sz); legend('Quasar 670 (100\mu{}M)','Fluorescein (1\mu{}M)');  % Plot: Droplet Detection Times figure; plt_hndl_vec=[plt_hndl_vec,gcf]; set(gcf,'name','Droplet Detection Times'); plot(droplet_start_times,'.-r'); hold on; grid on; plot(droplet_start_times_inv,'.-g'); xlabel('Droplet Index','fontsize',ax_fnt_sz); ylabel('Sample Index','fontsize',ax_fnt_sz); title(get(gcf,'name'),'fontsize',ttl_fnt_sz);  % Normalized Photons per Droplet plot figure; plt_hndl_vec=[plt_hndl_vec,gcf]; set(gcf,'name','Photons per Droplet - Green Normalized to Red'); plot(t(droplet_start_times)/60,int_norm_grn./int_norm_red,'-g'); hold on; grid on; xlabel('Time (min)','fontsize',ax_fnt_sz); ylabel('Normalized Fluorescent Intensity','fontsize',ax_fnt_sz); title(get(gcf,'name'),'fontsize',ttl_fnt_sz);  hist.m num_bins = 150;  % Number bins in each histogram  % **** Create Histograms [N_rd_all,X_rd_all] = hist([int_norm_red_noise, int_norm_red],num_bins); [N_grn_all,X_grn_all] = hist([int_norm_grn_noise, int_norm_grn],num_bins);  % Create Figure figure; set(gcf,'name','Histogram'); plt_hndl_vec=[plt_hndl_vec,gcf];  % Plot Red Histogram subplot(2,1,1); hold on; grid on; ylabel('Frequency','fontsize',ax_fnt_sz); 87  title('Channel 1 - Quasar 670 (100\mu{}M) - Histogram','fontsize',ttl_fnt_sz); bar(X_rd_all,N_rd_all,'r'); xlabel(['Fluorescent Intensity',NL,'(Photons/ms)'],'fontsize',ax_fnt_sz);  % Plot Green Histogram subplot(2,1,2); hold on; grid on; ylabel('Frequency','fontsize',ax_fnt_sz); title('Channel 2 - Fluorescein (1\mu{}M) - Histogram','fontsize',ttl_fnt_sz); bar(X_grn_all,N_grn_all,'g'); xlabel(['Fluorescent Intensity',NL,'(Photons/ms)'],'fontsize',ax_fnt_sz);  calc_info.m % *** Calculate information num_droplets = length(int_vec_rd); num_droplets_grn = length(int_vec_grn); meas_freq = num_droplets/total_time; duty_cycle = 100*mean(dig_sig_rd); % In Percent!  % *** Display results disp(['Threshold Used: ',num2str(thresh)]); if(is_smoothing_enabled)     disp(['Smooth_size: ',num2str(smooth_size)]); else     disp('No smoothing.'); end  disp([NL,'Sampling Rate (actual): ',num2str(Fs),' Hz']); disp(['Total # Samples Analyzed: ',num2str(L)]); disp(['Total # Droplets Analyzed: ',num2str(num_droplets)]); disp(['Total Time: ',num2str(total_time),' s']); disp(['Droplets per second: ',num2str(meas_freq)]); disp([NL,'Duty Cycle: ',num2str(duty_cycle),' %']); disp(['Single-Measurement Duration: ',...     num2str(10*duty_cycle/meas_freq),' ms']);  disp([NL,'Photons per Droplet, Red:']); disp(['Mean: ',num2str(mean(int_vec_rd))]); disp(['Median: ',num2str(median(int_vec_rd))]); disp(['Standard Deviation: ',num2str(std(int_vec_rd))]);  disp([NL,'Photons per Droplet, Green:']); disp(['Mean: ',num2str(mean(int_vec_grn))]); disp(['Median: ',num2str(median(int_vec_grn))]); disp(['Standard Deviation: ',num2str(std(int_vec_grn))]);  disp([NL,'Photons Between Droplets (noise), Red:']); disp(['Mean: ',num2str(mean(int_vec_inv_rd))]); disp(['Median: ',num2str(median(int_vec_inv_rd))]); disp(['Standard Deviation: ',num2str(std(int_vec_inv_rd))]);  disp([NL,'Photons Between Droplets (noise), Green:']); 88  disp(['Mean: ',num2str(mean(int_vec_inv_grn))]); disp(['Median: ',num2str(median(int_vec_inv_grn))]); disp(['Standard Deviation: ',num2str(std(int_vec_inv_grn))]);  disp([NL,'Samples per Droplet: ']); disp(['Mean: ',num2str(mean(int_vec_count))]); disp(['Median: ',num2str(median(int_vec_count))]); disp(['Standard Deviation: ',num2str(std(int_vec_count))]);  disp([NL,'Samples Between Droplets:']); disp(['Mean: ',num2str(mean(int_vec_inv_count))]); disp(['Median: ',num2str(median(int_vec_inv_count))]); disp(['Standard Deviation: ',num2str(std(int_vec_inv_count))]);  disp([NL,'Signal-to-noise, red: ',...     num2str( (mean(int_vec_rd)/mean(int_vec_count)) .*...     (mean(int_vec_inv_count)/mean(int_vec_inv_rd)) )]); disp(['Signal-to-noise, green: ',...     num2str( (mean(int_vec_grn)/mean(int_vec_count)) .*...     (mean(int_vec_inv_count)/mean(int_vec_inv_grn)) )]);  % Display Date disp([NL,'Current Date: ',date]);  savetodisk.m % Saves all figures with handles stored in the vector plt_hndl_vec to disk % Figures are saved as MATLAB .fig files and as .tif image files % After saving, figures are closed  for n = 1:length(plt_hndl_vec)     plt_hndl = plt_hndl_vec(n);     % Get next figure handle     ttl = get(plt_hndl,'name');     % Get name of figure, for filename      % Maximize figure     set(plt_hndl,'units','normalized','outerposition',[0 0 1 1]);      saveas(plt_hndl,ttl, 'fig');    % Save figure as MATLAB figure (.fig)     saveas(plt_hndl,[ttl,'.tif']);  % Save figure as image file (.tif)     close(plt_hndl);                % Close figure end   94  Appendix H – Sensitivity Measurements, Series 2, Additional Figures  Figure 54 - Sensitivity experiments, series 2: Raw data, [Fluorescein] = 100 nM.   Figure 55 - Sensitivity experiments, series 2: Histogram, [Fluorescein] = 100 nM. 95   Figure 56 - Sensitivity experiments, series 2: Raw data, [Fluorescein] = 50 nM.   Figure 57 - Sensitivity experiments, series 2: Histogram, [Fluorescein] = 50 nM. 96   Figure 58 - Sensitivity experiments, series 2: Raw data, [Fluorescein] = 30 nM.   Figure 59 - Sensitivity experiments, series 2: Histogram, [Fluorescein] = 30 nM. 97   Figure 60 - Sensitivity experiments, series 2: Raw data, [Fluorescein] = 20 nM.   Figure 61 - Sensitivity experiments, series 2: Histogram, [Fluorescein] = 20 nM. 98   Figure 62 - Sensitivity experiments, series 2: Raw data, [Fluorescein] = 10 nM.   Figure 63 - Sensitivity experiments, series 2: Histogram, [Fluorescein] = 10 nM. 99   Figure 64 - Sensitivity experiments, series 2: Raw data, [Fluorescein] = 0 M.   Figure 65 - Sensitivity experiments, series 2: Histogram, [Fluorescein] = 0 M. 100  Appendix I – Robustness Experiment, Additional Figures   Figure 66 - Average value of noise, in photons per millisecond, between each droplet.  This shows the variability in noise over time. 101   Figure 67 - Average value of signal, in photons per millisecond, counted in each droplet.  This shows the variability in signal over time.  102   Figure 68 - Ratio of photons per droplet, green signal, to photons per droplet, red signal.   103  References Agresti, J. J. et al. (2010). Ultrahigh-throughput screening in drop-based microfluidics for  directed evolution. Proceedings of the National Academy of Sciences, 107, 4004-4009.  Ahn, K. and Kerbage, C. et al. (2006). Dielectrophoretic manipulation of drops for high-speed  microfluidic sorting devices. Applied Physics Letters, 88, 024104.  Biosearch Technologies. (2011). Retrieved April 3, 2011, from  http://www.biosearchtech.com/store/product.aspx?catid=224,140,117,206&pid=714#  Fluorescence SpectraViewer. (2010). Accessed November 20, 2010, from Invitrogen:  http://www.invitrogen.com/site/us/en/home/support/Research-Tools/Fluorescence-  SpectraViewer.html  Holtze, C. et al. (2008). Biocompatible surfactants for water-in-fluorocarbon emulsions. Lab on a Chip,  8,1632 – 1639. Stanford Research Systems. (2011). High Voltage Power Supply, PS300 Series: PS325. Retrieved  April 4, 2011, from Stanford Research Systems Web site:  http://www.thinksrs.com/products/PS300.htm  Tamuri, A. et al. (2009). Nanoseconds Switching for High Voltage Circuit Using Avalanche  Transistors. Applied Physics Research, Vol. 1, No. 2, Nov 2009.  Trek Inc. (2011). High Voltage Amplifiers: 677B - Datasheet. Retrieved April 4, 2011, from Trek  Inc.  Web site: http://www.trekinc.com/products/HV_amp.asp 

Cite

Citation Scheme:

    

Usage Statistics

Country Views Downloads
China 18 10
United States 3 0
Japan 2 0
City Views Downloads
Beijing 18 0
Daly City 2 0
Tokyo 2 0
Ashburn 1 0

{[{ mDataHeader[type] }]} {[{ month[type] }]} {[{ tData[type] }]}

Share

Share to:

Comment

Related Items