The FA9000

Photocatalytic conversion of terephthalic acid to 2-hydroxyterepthalic acid.

To test the photocatalytic performance of a semiconductor photocatalyst, multiple methods can be used. A popular choice is a fluorescence probe, typically performed using terephthalic acid for the assay. This involves illuminating the catalyst in a basic solution of terephthalic acid, generating hydroxyl radicals that bind to form 2-hydroxyterepthalic acid. The significance of this reaction is the conversion of a compound that does not fluoresce, to a compound that does. This enables the ability to track the increase in concentration of a catalytically produced species over time.

Manually operated in-situ fluorescence probe.

To take the fluorescence probe to the next level, I had to make a fixture to operate the experiment in-situ. Typically, the reaction takes place separately and aliquots are taken at certain intervals. However, this introduces a lot of errors including other variables to consider, such as wait times. To accommodate this, I redesigned the cuvette holder for the Cary Eclipse Fluorescence Spectrometer from Agilent such that it can accommodate a LED light fixture.

However, this system had serious drawbacks. The LED controller utilizes signal wires that modulate the power to the LED based on the resistance between them. In this instance the dial is applying that resistance and is acting as a rheostat. The screen monitors the voltage, current and subsequently the power in watts going to the LED.

The issue arises from an increased resistance due to heat when the LED turns on. Overtime the LED will need more power to maintain a constant wattage output. This system is not accommodating for that. Furthermore, it is completely manual. We typically illuminate the film in five minutes intervals followed by a respective scan for fluorescence. This involves manually plugging and unplugging the LED power supply, relying on a timer to do so, and then quickly taking the scan. The LED light also goes straight into the photomultiplier tube (PMT) which causes an overvoltage error in the detector. To mitigate this a blocker must be inserted and removed manually.

The tipping point personally, followed the completion of The 8 Shot reaction vessel. To get data prior to a ACS poster presentation, I sat in front of the fluorometer for over eight hours, manually running the equipment. This gave me data to present, however, like the experiments involving aliquots taken, I am certain human error is a contributing factor. This leads me to believe an experiment functioning this way is bad science.

Remodeled and 3D-printed door for Cary Eclipse Fluorometer.

I became intent on not having to sit in front of the machine for hours on end to run the experiments, while ensuring the data from the experiment is as accurate as possible. So I began developing a method to remove as much reliance on human intervention as I could, in essence, an automation.

Fluorometer door version 2, integration of automation electronics.

To start, I began redesigning the door to the fluorometer. I knew there were going to be a mess of wires that needed to be routed into this dark chamber, and I did not want to have to cut them into the original piece. After some trial and error to replicate this odd shaped piece, I got something that fit well and I was satisfied with.

I realized after finishing my model of the door that I can fit all the electronic components within its shell. This would be ideal to consolidate the pieces for the experiment and not have to route wires in weird ways through the door. I began figuring out what electronics I need to perform this task, and modeling how they would fit into the door assembly.

Automation electronics prototype.

I decided to utilize the Arduino Uno R4 Wifi as a microcontroller platform. I chose this component based on the capacity to broadcast wirelessly, and control computers by simulating mouse and keyboard commands. Furthermore, there is a host of cheap electronic components that are directly compatible with the board.

Next, I performed research on what I would need to achieve my goals. I wanted to be able to monitor the power coming out of the LED, as well as control it digitally. This involved purchasing a power meter breakout board, to monitor the voltage and current coming out of the LED power supply.

To control the power, I needed some way to apply a resistance to the signal wires (pink/purple). I found a device called the DigiPot 6 Click which is a type of digital potentiometer. When used in rheostat mode it has 255 wiper settings where it can sweep from 70kohm to 100kohm. When used with the LED power supply, this gives me control over the power level within 0.03W. This is important, as mentioned before, because the LED will drop in power as it heats up. I can program the Arduino to monitor the power and compensate for drops below 0.03W.

An important variable during the reaction is the temperature of the solution. To monitor this factor, I decided to integrate thermocouples into the system such that I can monitor the temperature of the cuvette and the LED.

And lastly, I integrated a OLED display on the front of the system to indicate the parameters set for the experiment. Specifically, it will indicate the name of the sample being run, the illumination time, the number of cycles, the current temperatures being monitored, and the ETA of the experiment.

Webserver for controlling the automation parameters and running the experiment.

The automation functions through a webserver broadcasted by Arduino. It allows the user to connect to the Arduino through the WIFI network it broadcasts, and then through searching the webserver’s IP address in a browser. Here one can set all the parameters and control the automation wirelessly, while monitoring live graphs of temperature and power.

To run the scans on fluorometer’s computer, a python script was developed. The python script is activated by the Arduino through simulated keyboard commands, after which the python script completely controls the computers end of the job. The python program communicates back and forth between the Arduino to let each other know when illumination periods are complete and when scan periods are complete. This includes running a blank scan at the beginning of the experiment so the computer can inform the Arduino how long it takes. It uses the length of the scan, and the parameters set, to display the ETA on the OLED screen such that the user can come leave and come back to change the sample.

Model of the final automation electronic assembly, fit with thermocouple connectors, and GX16 output for LED assembly power and logic.

I confirmed that the system was working as intended and moved to cleaning up the model and the design. This involved creating connections that can easily be installed and removed to set up the experiment and pack up. A C14 power connection is installed in front of the device to receive 120V AC and power all the internals. A USB C connector enables wired connection to the scan computer for sending and receiving data. Lastly, the rear panel which outputs to the inside of the fluorometer features thermocouple connectors and a GX16 connector which handles all power out to the LED fixture.

Production model of the FluorAutomator 9000.

Previously, I used a filament 3D-printed prototype to test the fitment of the electronics. Once I was sure that the design was functional, I moved towards printing the case in UV resin using the Anycubic Photon Mono M7. This enabled me to get very high-resolution printing and a pitch black texture to prevent light scattering.

The FluorAutomator 9000 running an experiment in-situ.

Automated experiment run for 30 minutes total illumination time in 5 minute intervals, with 20 scans between illumination periods, once per minute.

To test the capabilities of the instrument and probe the effects of waiting after illumination, I ran a unique experiment. This involved the typical illumination time set to 5 minutes, for a total of 6 cycles. However, I included the option to wait for a minute after the illumination period and take a scan, repeated for a total of 20 scans. This resulted in a total of around 200 scans taken over a roughly 5 hour period. This experiment was a total success, as it ran without an errors for the entirety of the experiment.

Graph depicting the plotted power, and temperatures of the LED and Cuvette during an experiment.

Due to the sensors in place, and the communication between the Arduino and the computer. I can record all the sensor metadata as a function of time in CSV format. This allows me to plot a graph showing how the power maintains a constant level, as set, and a depiction of how the solution temperature changes throughout the experiment. This enables me to dive deeper into the accuracy of the science, without neglecting important variables that affect the outcome of the data.