Introduction

Why an UltraViolet Spectrometer?

High-Performance-Liquid-Chromatography (HPLC) is expensive. It's like. Really. Really. Expensive. The need for a reliable and cheap (at least in comparison) PET degradation assay has led to many innovations, including titrimetry, fluorimetry, chromatography, and spectrometry. However, they are all limited by their discrete measurement, often limited to once every 15-30 minutes. By continuously monitoring the degradation progress, we are also able to much more effectively visualize and analyse the different properties of our variants, such as MHET inhibition. Our hardware aims to bridge the gap between expensive systems and the need for affordable, continuous data collection by developing a low-cost UV spectrometer.

What is an UV Spectrometer

The hardware is placed on a hot plate set to the degradation temperature. Since the quartz cuvettes are immersed in a water bath, the system requires time to reach thermal equilibrium. A quartz collimator lens is used to parallelize the light emitted from the UV LED as it passes through the quartz cuvette. Because of their benzene ring, many of the byproducts from PET depolymerization absorb UV light in the 240–260 nanometer range. Our sensor specifically measures the light intensity at 254 nm. By monitoring changes in light intensity at this wavelength, we can estimate the concentrations of these byproducts.

Design Iterations

Throughout development, we went through multiple iterations to refine our UV spectrometer design. Below are entries for each iteration:

Design: Our first iteration attempted to minimize the cost and increase accessibility by using an inexpensive 28byj 48 stepper motor to power the spectrometer. The design uses 2 3d printed spur gears. One for the 28byj stepper motor and one for the sensor and UV-LED. The sensor spins on the outside around the cuvette and the LED spins in the middle.

The cuvettes sit inside as shown, with water on both sides between each cuvette and the next. We included the water between the cuvettes because we planned on putting the whole hardware in the air convection oven. While it is generally decent at temperature control, it may fluctuate too much for optimal PET degradation. The addition of water and its high specific heat capacity meant that it would act to stabilize the temperature of the hardware.
Build: We were able to quickly 3d print a prototype to test outside of the lab.
Test: The prototype was made as a proof of concept for the rotating mechanism for the spectrometer. After spinning up the prototype for a few minutes, it became evident that the 28byj 48 stepper motor was not nearly powerful enough to accurately spin the spectrometer wheel. The stepper began to heat up and skip steps.
Learn: From this prototype, we were able to gain a better understanding of what electronics we would need to be able to accurately run the spectrometer for the time required to degrade PET.

Design: After the first prototype failed because of the weak stepper motor we had used, we decided to swap it out for a more accurate servo motor. But because most positional servo motors could only rotate 90-180 degrees, which wouldn’t work unless we geared the servo up, we had to use continuous servos instead.
Build: For the prototype, we once again printed it out in PLA.
Test: We didn’t get to comprehensively test this version of the hardware because we realized the mistake we made while we were wiring the system.

Learn: One big mistake that we had made was that we had not realized that continuous servo motors could not accurately move to specific angles. This was a huge problem because the spectrometer would have to move with incredibly narrow margins of error for the UV readings to be accurate. After some further research, we realized the use of servo motors would have to be abandoned completely. This was because workarounds, such as implementing encoders, would either require an absolute encoder or a rotary encoder. An absolute encoder added unnecessary costs, and a rotary encoder could slowly accumulate errors as the degradation went on. This probably the most avoidable mistake that we made in the whole hardwaring iteration process. We were feeling pretty stupid after this one.

Design: For the third iteration of the hardware, we decided to switch to a much more powerful motor. After the first two attempts, we decided that overkill was, in fact, the perfect amount of kill. While it was still an affordable stepper motor, it provided some much-needed power and accuracy. In this version, we also switched from using gears to belts, which we thought would be more accurate, as it provides nearly zero backlash.

For the design of this iteration, we also moved the cuvettes up a little to allow for the water to underneat and around the cuvettes, equalizing the temperature across the hardware more efficiently and further stabilizing the degradation parameters.

Build: The electronics for the beefy stepper motor required a bit more electronics and the addition of a 24-volt power supply. We decided to stick with the PLA because it was still early in our prototyping.
Test: This version worked wonderfully. The big stepper motor was definitely a much-needed addition. After a few hours of testing, the motor had maintained a very accurate positioning without the need for a closed-loop control system.
Learn: There were two main issues with our design. The first was that we found the belts a little too finicky and hard to work with. We felt like the belt and pulleys weren’t necessary, and the design constraints it put on us would mean that the mechanism would either have to be overcomplicated or too tall and clunky. The second issue was that we felt the water between the cuvettes wasn’t really needed and was adding to the complexity of the spectrometer.

Design: One problem that we had with the last iteration was that it had the big stepper motor standing up. There are 2 main issues with this. The tall stepper motor took up a lot of space. This also meant that the spinning part of the hardware had to be extended up. This made it wobbly and created lots of errors and backlash. Furthermore, we found that with the use of the pulleys, the flex of the 3d prints made it not as effective as we had hoped. With this iteration, we tried to both minimize the height of the hardware as well as reduce the backlash for more accuracy. And so instead of using spur gears or pulleys, we landed on using beveled gears. This meant that the stepper motor would be able to lie down, drastically reducing the space it would take up. We also decided to use helical teeth to smooth out the gears and increase accuracy.
Build: We 3d printed everything out of PLA. Because this was the first version where the stepper motor was lying down, it took a few tries to figure out how to mount it properly.
Test: The new system with the helical bevel gears had very little backlash while still taking very little space. We tested the hardware again by running it for a few hours and seeing if the sensor and LED still aligned through the cuvette after a few hours. This time, not only did the motor stay accurate but so did the hardware.
Learn: One thing that was a bit concerning about the hardware was that the top bevel gear attached to the sensor and LED was kind of wobbly. It didn't really build up any errors while testing. However, it was still rather concerning. This version of the hardware still hadn’t actually used the sensors and LED because we believed it would be a lot easier and better if we tested the sensor and LED on a proof of concept jig (version 5).

Design: This iteration is not so much an iteration but more a test, a proof of concept. The goal of this hardware is to be a simplified version of the 8 slot hardware. We planned for it to just hold the LED and sensor in place with a cuvette in the middle. This was just to 1 test the effectiveness of our sensor and LED and their accuracy. The point is just to reduce the number of variables by making a proof of concept which doesn’t move at all. One problem we had while designing this was that it was very hard to hit the right balance of ease of assembly and accuracy.
Build: This was somehow the easiest and the hardest iteration to assemble. On one hand, theres not much to assemble, its just one piece of PLA. However, on the other hand, having to maneuver your fingers to install the LED and sensor took way too long.
Test: While testing, we found that our screwdriver couldn’t quite fit to screw in a screw. Because of the extreme time constraints we were under, we used a soldering iron to melt a hole in the plastic. The testing for this hardware was just chaos. We started by testing the electronics with a breadboard off the hardware. Everything was working great. We were wearing our UV goggles, and the sensor was getting relatively consistent results. The problem came when we tried to connect the legs of the LED to Dupont cables. It is unclear if it was the fault of the fupont cables or the LED, but they could not maintain a stable connection. While this didn’t sound that terrible, debugging this hardware was a nightmare. The cables weren’t our only problem, the breaboards were also acting up. The resistors weren’t connecting to the breadboard. Overall, lots of small issues with the hardware dragged and delayed our testing. Finally, when we decided to solder the Dupont cables directly onto the legs of the LED, they just snapped. After hours and hours of bending, unbending, and rebending, the little legs just gave way and snapped.
Learn: What we learned from this iteration, while abundant, was more life lessons than anything. If it sounds too good to be true, then it probably is. The <$1 breadboard and wires were not it. Patience. Patience. Patience. If only, if only. If only we had worked more slowly, then the legs might not have broken off.

Materials

Components and Functions

Our design of UV-Vis Spectrometer includes multiple components, including:

Component	Price	Function
Arduino Uno	$25	The arduino is the brain of the operation. It handles all the controls and data logging.
UV LED (254 nm)	$58	The LED is used as the UV source. We chose 254 nm because its around the peak UV absorbance of TPA, MHET, and BHET
Photodiode Sensor	$2	We chose this specific photodiode because its sensitivity also peaks around 254nm.
Quartz Cuvettes	$43.70500*8	This is likely the most expensive and crucial part of the hardware. Quartz is necessary because glass and many other materials also absorb UV light.
Stepper Motor and Driver	$49	Not something we should have skimp on in the first place. While high torque nema stepper motors may seem like overkill (and it is), in the case of this hardware, which will be run for at least 48 hours overkill is underkill. Because stepper motors can't be directly controlled by arduinos, the motor controller acts as an interface between them.
Power Supply	$72.39	A power supply is needed to power the motor becuase it draws more voltage then the arduino can provide.
Wires	$1	Well... Wires

Results

While we initially planned to run multiple rounds of degradation assays with our hardware, time constraints and unexpected complexities with the electronics prevented us from obtaining consistent and repeatable data. Our biggest obstacle was integrating the UV-LED with the rest of the spectrometer due to faulty wires and breadboards. This significantly set us back since we had spent hours debugging the electronics to no end.

We had planned to perform several experiments as a proof of concept. However, just as we were setting up the hardware, we accidentally bent the legs of our only UV-LED. Normally, this wouldn’t be a major issue, but after a week of frequent adjustments and testing, the LED’s legs had already been weakened. This final bend caused one of its legs to snap off, abruptly halting our experiments until we can order a replacement after the wiki freeze.

Despite significant setbacks, especially near the end, we will continue to iterate and improve our hardware. We plan to perform 4 rounds of experiments and assays:

Assay Validation: Testing the effectiveness and accuracy of our sensor using varying concentrations of TPA in HEPES buffer.
Initial PET Degradation: Running continuous PET degradation data with the wild-type TfCut using the smaller proof-of-concept hardware.
Assay Validation: Testing on the same standards but using the 8-slot hardware.
Full-Scale Assay: Performing the assay validation round on the final 8-slot hardware to collect continuous data simultaneously on the wild-type TfCut, six variants, and a blank control.

Our goal is to be able to consistently obtain accurate and continuous data on wild-type TfCut, our 6 variants, and a blank at the same time. In spite of numerous setbacks, we will continue to design, build, test, and improve our hardware so that we will be able to provide a more thorough analysis of the characteristics of our mutants.

Hardware