The Master Plan
This document is intended to serve as an up to date record of the planned system design, changes and reasoning. On the most basic level the device consists of 4 systems: fluidics, optics, electronics and signal processing. The design and functionality of each system impacts the others, thus they cannot be designed and implemented in isolation. For this reason it is important that we keep track of the bigger picture goal and how each system is designed towards achieving that goal. As we go through the design and prototyping of each system and subsystem, it is natural that constant changes will be made. It is however important that these changes and the reasoning behind them is documented for when writing my thesis.
The Overarching Goal
The aim of this project is to develop a low cost microfluidics based flow cytometer that can be used in POC environments. Specifically it will focus on analysing yeast cells with the aim of aiding in drinking water safety analysis. While a device for use in the laboratory would be acceptable, the ultimate goal is to develop a fully autonomous device that could be deployed directly in a river for long term analysis. This goal leads to certain requirements.
The Main Requirements:
- Low cost (Goal: sub R25k)
- Ability to detect and count yeast cells (10 μ\mum)
- Compact Desktop Form-factor
- Implement standard 488nm laser for compatibility with common tags
- Add at least one fluorescent channel
- Ease-of-use
Additional Requirements:
- Portability/Battery Power
- IOT/Off-Grid functionality
- Automated Sampling, Tagging and Analysis
With these requirements in mind we can design the overall system to meet them.
The Fluidics
The main purpose of the fluidics is to control the flow of cells through the detection area. The cells need to pass through the detection area one at a time (without overlap) and be centred to ensure reliable and reproducible results. Faster flowing sheath fluid is used to centre the sample stream within the channel through hydrodynamic focussing. This ensures that the position of cells can be precisely controlled while still allowing for a channel diameter much larger than the cells (easier to manufacture and helps avoid clogging). By controlling both sheath and sample fluid flow rates, the sample can be centred and it is possible to ensure that cells are spread out enough to not overlap within the detection window.
Traditional flow cytometry uses flow cells made from quartz. This results in an extremely optically clear flow cell and allows for the use of an inner sample tube and outer sheath tube for 3D hydrodynamic focussing. These flow cells are, however, extremely expensive. With modern manufacturing techniques such as photolithography and SLA 3D Prinitng improving in quality and becoming more accessible, microfluidics presents a viable alternative.
Microfluidics centers around a planar fluidics chip that achieves the same goal as traditional flow cells.
The requirements:
- Center cells in the flow channel both horizontally and vertically
- Control flow rate to ensure single cell detection
- User replaceable
- Can be reliably manufactured with reproducible performance
- Can be manufactured with equipment available at SANDLab
- Low cost
These requirements present some design challenges. The main challenge is with regards to the hydrodynamic focussing, since the chip is manufactured in a planar fashion.
Focussing Approach:
-
Microfluidic drifting is currently the planned approach
Possible problems could include layout and space constraints.
-
An alternative approach is using contraction-expansion arrays
Chip Manufacture:
While SLS printing has sufficient resolution for the fluidics, it is not sufficient for integrating lenses into the design, thus SLS will be used for prototyping of the fluidics but lithography for the final chip.
There are two possible approaches to lithography, direct lithography and soft lithography. Soft lithography involves manufacturing a negative mold and then casting the chip itself using PDMS. This does have some advantages, but the major disadvantage is inconcistency and time. PDMS requires significant time in a vaccuum chamber to remove air bubbles that could ruin a chip. Additionally PDMS shrinks while drying and this shrinkage is dependent on environmental conditions, thus the performance and viability of the produced chip can vary run by run.
The proposed approach is thus directly manufacturing the chip on a glass slide with lithography. Due to the planar nature of lithography the chip will need to have an open channel design and then be sealed afterwards. Plasma bonding is often proposed as a solution, however this is time consuming and does not reliable yield high bonding strength. Based on [@keciliAdhesiveBondingStrategies2020] the proposed approach involves spincoating a thin layer of PDMS or resin onto a glass slide, pressing the chip into this and then pressing it against the sealing layer before curing. For SLS prototype chips, resin will be used as the bonding agent and a clean glass slide will be used as sealing layer. The final SU-8 chip will need to use PDMS as the sealing layer (as inlets and outlets will need to pass thorugh this layer) and thus use PDMS as the bonding agent.
Possible issues and possible solutions:
- Drifting not working as simulated: Contraction-expansion arrays
- Difficulty spincoating whole microscope slide: Double width slide or smaller design
- Resin/PDMS filling channels during sealing: Switching to using APTES activated bonding or plasma bonding
The Optics
What to look for in filters:
- OD of filters (good = 5-6)
- Extintion ratio
- Transmission %
- Bandwidth
- Sharpness of edge
Two filters, either BP and BP or LP and then BP
Fibre Choice
The near ultra-violet range of frequencies involved in this project drastically reduce the available fibre options, since the telecoms industry are the major market for fibre optics and related equipment and they operate in the 100nm+ range.
Excitation fibre
The excitation fibre needs to be single mode to ensure a gaussian distribution of light and most light thus hitting the centre of the channel. There are limited options for 488nm single mode fibres with options from ThorLabs and Coherent. Thorlabs was chosen as they are a reliable supplier already on the university's systems. ThorLabs offers the following options:
These fibers enable single mode transmission from 400-680 nm and feature an acrylate jacket. The S405-XP and SM400 fibers both consist of an undoped, pure silica core, and the SM400 fiber is surrounded by a depressed, fluorine-doped cladding. Since these two fibers do not contain germania ( GeO_2\mathrm{GeO}\_2 ), which causes electronic defects and color centers associated with the Ge-O bond, the primary cause of photodarkening is greatly reduced. The transmission-limiting effects caused by other nonlinearities (e.g., stimulated scattering) or even thermal damage are also reduced compared to those of a conventional silica fiber doped with germanium.
Fiber Specifications
| Item # | Operating Wavelength | Mode Field Diameter b { }^{\text {b }} | Cladding Diameter | Coating Diameter | Cut-Off Wavelength | Price |
|---|---|---|---|---|---|---|
| S405-XP | 400−680 nm400-680 \mathrm{~nm} | 3.3±0.5μ m3.3 \pm 0.5 \mu \mathrm{~m} @ 405 nm 4.6±0.5μ m4.6 \pm 0.5 \mu \mathrm{~m} @ 630 nm | 125.0±1.0μ m125.0 \pm 1.0 \mu \mathrm{~m} | 245.0±15.0μ m245.0 \pm 15.0 \mu \mathrm{~m} | 380±20 nm380 \pm 20 \mathrm{~nm} | $18.04/Meter |
| SM400 | 405−532 nm405-532 \mathrm{~nm} | 2.5−3.4μ m@480 nm2.5-3.4 \mu \mathrm{~m} @ 480 \mathrm{~nm} | 125±1μ m125 \pm 1 \mu \mathrm{~m} | 245±15μ m245 \pm 15 \mu \mathrm{~m} | 305−400 nm305-400 \mathrm{~nm} | $22.13/Meter |
| SM450 a { }^{\text {a }} | 488−633 nm488-633 \mathrm{~nm} | 2.8−4.1μ m@488 nm2.8-4.1 \mu \mathrm{~m} @ 488 \mathrm{~nm} | 125±1.0μ m125 \pm 1.0 \mu \mathrm{~m} | 245±15μ m245 \pm 15 \mu \mathrm{~m} | 350−470 nm350-470 \mathrm{~nm} | $9.22/Meter |
| 460HP | 450−600 nm450-600 \mathrm{~nm} | 3.5±0.5μ m3.5 \pm 0.5 \mu \mathrm{~m} @ 515 nm | 125±1μ m125 \pm 1 \mu \mathrm{~m} | 245±15μ m245 \pm 15 \mu \mathrm{~m} | 430±20 nm430 \pm 20 \mathrm{~nm} | $13.23/Meter |
| Item # | Short-Term Bend Radius | Long-Term Bend Radius | Attenuation (Max) | Proof Test Level | NA | Core Index | Cladding Index |
|---|---|---|---|---|---|---|---|
| S405-XP | ≥6 mm\geq 6 \mathrm{~mm} | ≥13 mm\geq 13 \mathrm{~mm} | ≤30.0 dB/km\leq 30.0 \mathrm{~dB} / \mathrm{km} @ 630 nm ≤30.0 dB/km\leq 30.0 \mathrm{~dB} / \mathrm{km} @ 488 nm | ≥200kpsi(1.4GN/m2)\geq 200 \mathrm{kpsi}\left(1.4 \mathrm{GN} / \mathrm{m}^2\right) | 0.12 | Call d{ }^{\mathrm{d}} | Call d { }^{\text {d }} |
| SM400 | ≥10 mm\geq 10 \mathrm{~mm} | ≥30 mm\geq 30 \mathrm{~mm} | ≤50 dB/km\leq 50 \mathrm{~dB} / \mathrm{km} @ 430 nm ≤30 dB/km\leq 30 \mathrm{~dB} / \mathrm{km} @ 532 nm | 11 % ( 100 kpsi ) | 0.12-0.14 | 405 nm:1.46958e 405 \mathrm{~nm}: 1.46958{ }^{\text {e }}467 nm:1.46435e467 \mathrm{~nm}: 1.46435{ }^{\mathrm{e}}532 nm:1.46071e532 \mathrm{~nm}: 1.46071{ }^{\mathrm{e}} | 405 nm:1.46382e 405 \mathrm{~nm}: 1.46382{ }^{\text {e }}467 nm:1.45857e 467 \mathrm{~nm}: 1.45857{ }^{\text {e }}532 nm:1.45491e532 \mathrm{~nm}: 1.45491{ }^{\mathrm{e}} |
| SM450 | ≥5 mm\geq 5 \mathrm{~mm} | ≥25 mm\geq 25 \mathrm{~mm} | ≤50 dB/km\leq 50 \mathrm{~dB} / \mathrm{km} @ 488 nmc488 \mathrm{~nm}^{\mathrm{c}} | 11 %(100 \mathrm{kpsi}) | 0.10-0.14 | 488 nm:1.46645f488 \mathrm{~nm}: 1.46645^{\mathrm{f}} | 488 nm:1.46302f488 \mathrm{~nm}: 1.46302^f |
| 514 nm:1.46501f514 \mathrm{~nm}: 1.46501{ }^f | 514 nm:1.46159f514 \mathrm{~nm}: 1.46159{ }^{\mathrm{f}} | ||||||
| 460 HP | ≥6 mm\geq 6 \mathrm{~mm} | ≥13 mm\geq 13 \mathrm{~mm} | ≤30 dB/km\leq 30 \mathrm{~dB} / \mathrm{km} @ 515 nm | ≥200kpsi(1.4GN/m2)\geq 200 \mathrm{kpsi}\left(1.4 \mathrm{GN} / \mathrm{m}^2\right) | 0.13 | Call d{ }^{\mathrm{d}} | Call d { }^{\text {d }} |
a. The wavelength range is the spectral region between the cut-off wavelength and the bend edge and represents the region where the fiber transmits the TEM M_00\mathrm{M}\_{00} mode with low attenuation. For this fiber, the bend edge wavelength is typically 200 nm longer than the cut-off wavelength. b. MFD is a nominal, calculated value, estimated at the operating wavelength(s) using a typical value of NA & cut-off wavelength. Please see the MFD Definition tab for details. c. Stated attenuation is a worst-case value, quoted for the shortest design wavelength. d. Please contact us to learn more about the refractive index of this fiber, as we are not permitted to publish this information on our website. e. The indices provided are for an NA of 0.13 . f. The indices provided are for an NA of 0.10 .
As seen from the above excerpt the SG405-XP and SM400 are pure silica fibres, ensuring that they can withstand long term deployments and high intensity near UV light without degrading over time. This however adds significant cost. Our use case, where the fibre will be attached to the chip (since we are using a laser that is already fibre coupled) and be replaced with the chip, combined with the relatively low power of our excitation laser, thus negates the need for these more expensive fibres. The 460HP is a high performance fibre with increased strength with the aim of increasing yield when used in high volume production of RGB components. This is not relevant to our use case and thus with the aim of making the consumable fluidics chip as affordable as possible (while maintaining performance), the SM450 is the perfect choice.
Collection Fibre
Fibre connectors
Fibre coupled lasers are commonly available with two connectors, FC and SMA905. SMA905 is an older connector and uses a metal ferrule, with much looser tolerances than the ceramic ferrules used in FC connectors. Additionally, SMA connectors do not typically make direct contact between fibres (relying on a small air gap), due to them not being spring-loaded. This results in significantly higher transmission losses and worse alignment. FC is thus the clear choice for the laser fibre connector.
FC connectors can either have a flat edge (PC or UPC) or a 8\textdegree\textdegree angled connector (APC). The angled connector reduces back reflections significantly and avoids optical resonance causing fluctuations in laser power. While this would be preferred, it is not feasible to make and polish APC connectors in a lab setting due to the precision required and the fragility of the fibre cut at an angle. Additionally, our laser is relatively low power and could possibly be used in a burst mode which would negate the effect of these back reflections. For the collection optics, these back reflections are also entirely irrelevant.
FC connectors have the additional advantage of being quick to connect and disconnect reliably, thus improving the field setup ease.
SU-8 and its optical properties
Experimental data for transmittance of a 30 μm thick SU-8 photoresist after hard-baking.
Transmission characteristics through different thicknesses of SU-8 in the UV region.
The above figures shows high transmittance for the wavelengths of interest (488nm and 510nm).
Note from the article regarding the ripples at small widths:
Figure 2 also shows interference signals from the samples in the visible light region in the form of a modulation of the transmission curves. The ripples in the signal for the 1.1 μm\mu m sample between 400 and 800 nm, for example, has a lower frequency than the signal from the 7.0 μm\mu m sample. Their frequencies could be, in principle, used to calculate the SU-8 film thickness when the complex index of refraction of soft-baked SU-8 for the used wavelengths is known. These interference signals arise when the measurement setup switches to the tungsten iodide lamp at 360 nm, as mentioned in Sec. II B. Since this wavelength range falls in the VIS spectrum and is beyond the scope of this work, the interference signals are ignored and are only presented for the sake of completeness. The inset in Fig. 2 shows that the difference in transmittance for particular wavelength regions in the UV range can be as high as an order of magnitude for different resist thicknesses.
An excerpt from "Realization and characterization of SU-8 micro cylindrical lenses for in-plane micro optical systems" regarding SU-8 transmittance:
To measure the in-plane transmittance of SU-8 structures, a 488 nm laser light emitted from a single-mode fiber was arranged to travel the same distance through an SU-8 structure and an air gap, respectively. Both outputs of the laser were collected by multi-mode fibers and transferred to a power meter. Transmittance was than determined from the ratio of the two power values. The experimental results revealed that the transmittance through the SU-8 with a structural width of 65 μm\mu m, 125 μm\mu m, and 250 μm\mu m was 93.1%, 92.3%, and 95.6%, respectively. The discrepancy in these values may arise from fluctuation of the power meter and the light source. The transmittance of SU-8 was therefore high, since at least 92% of the incoming light passed through the SU-8 structure.
From another article tested 250 μm\mu m films of SU-8 and a less fluorescent 1002F photoresists at 485nm excitation:
Fluorescence of blocks of SU-8 and 1002F resin and photoresist. The blocks were excited at 485 nm with a ±2 nm band-pass, and the fluorescence was measured from 490 to 650 nm with a ±2 nm band-pass. The upper solid and dashed lines are SU-8 photoresist and SU-8 resin, respectively. The dotted and dash-dotted line are 1002F photoresist and 1002F resin, respectively. The lower solid line is PDMS. For technical reasons, the photoresist blocks were not treated with UV exposure and postexposure baking.
Spectral properties of SU-8 and 1002F photoresists after UV exposure and postexposure bake. (A) Fluorescence of SU-8 (solid squares) and 1002F (open circles) of lines (100 μm wide) of varying heights. The fluorescence was collected with a filter set designed for fluorescein (excitation filter 450−490 nm, dichroic 500 nm long pass, emission 520 nm long pass). (B and C) Same as (A) but the fluorescence was collected using a filter set for tetramethyl rhodamine (excitation filter 528−553 nm dichroic 565 nm long pass, emission 590−650 nm) (B) or Cy5 (excitation filter 590−650 nm, dichroic 660 nm long pass, emission 665−740 nm) (C). The error bars represent the standard deviation of five measurements. (D) The absorbance of glass (dotted line), 1002F photoresist (dashed line), and SU-8 photoresist (solid line) is shown at varying wavelengths.
Filters
Lenses
The Electronics
Excitation
- Laser
- Power Supply
- Switch/Control
Laser Choice
For the laser, a 488nm laser was chosen as it would allow for the use of the wide variety of flourochromes developed over the years for flow cytometry. Longer wavelength lasers in the 600nm range are much cheaper and more common, but would be a significant shift from the industry standard 488nm that most users would be trained on and for which most flourochromes are available.
This however massively limited the number of options, with most options being high-power lasers aimed at manufacturing and laser engraving (expensive and likely to damage cells) or from scientific suppliers such as ThorLabs (extremely expensive). Ideally a fibre-coupled laser would be used which allows for easy coupling of light into the fluidic chip, however this adds additional costs and reduces options even further. Ultimately this 25mw fibre-coupled 488nm laser from LaserTree stood out as ideal. It is relatively affordable at $236 and meets all other requirements.
Using a raw laser diode instead of a module cuts down significantly on cost and while it does increase the complexity of the electronic system that needs to be designed, it also offers greater flexibility.
Measurement
- Avalanche Photodiodes
- APD bias supply
- Amplifier
Photodiode Choice
The choice of photodiode has a significant impact on the sensitivity of the system and is one of the most important component decisions. Regular photodiodes simply lack the necessary sensitivity to detect the fluorescence from a single cell. Avalanche photodiodes (APDs) offer significant improvements in sensitivity, but they require high voltage DC driving circuits and are significantly more expensive. There are only a couple of APD manufacturers (Laser Components, Hamamatsu and Excelitas) producing diodes in for lower wavelengths (rather than the 1000+ nm required for telecoms) and it is surprisingly difficult to directly purchase these diodes. APDs are manufactured either on a silicon (Si) or indium gallium arsenide (InGaAs) substrate. InGaAs diodes have a peak sensitivity in the 1000+ nm range and thus are not suitable for this project. Even Si diodes are mostly sensitive in the 400 - 1000 nm range, thus specialised UV-enhanced diodes are needed which cover the 200 - 1000nm range. While these diodes have a peak sensitivity around 600nm, they do still have much higher sensitivity at 488nm than normal photodiodes. Additionally, the higher sensitivity at wavelengths closer to 600nm could prove useful as it means increased sensitivity to our fluorescent wavelengths which will be much dimmer than the 488nm laser light.
Excelitas does not produce small area UV-enhanced diodes and thus are not an option. Laser Components produces a single UV-enhanced diode, the SUR500x. While this is a good option, it is not available for purchase except via specialised suppliers and on a request for quote basis, implying extremely high prices for small batches.
ThorLabs sells APD modules that use Hamamatsu photodiodes, but these are exorbitantly expensive. The Hamamatsu S12053-05 is however directly available from RS Components at a somewhat reasonable price and was thus chosen as it meets the requirements.
APD Bias Supply
Matsusada TR-0.2P: Temp Compensated, 200V XP Power A02P-12T: Only a DC-DC boost converter, up to 200V, available at RS Components Laser Components dBC-220-3S: Temp Compensated, 220V, Current Limit Custom Designs:
- This one by Jim Williams
- This application note from Analog Devices (also written by Jim Williams) is more detailed
Control Systems
- Micro Pumps
The Signal Processing
Progress tracking:
- Plan optics approach
- Identify Optical Components
- Design FLuidics
- Identify Electronic Components
- Test Fluidics
- Order Optical Components