Introduction: Our whole system was triggered by light[1], so E. coli needed to be illuminated when cultured. To make quantitative analysis, we needed to strictly control light intensity and wave length. Our hardware series kept upgrading as our demands increased.

Material: Considering the small volume and high efficiency, we chose 3W LED as our light source after careful calculations[2]. We managed to buy 11 kinds of LEDs with different wave lengths (390-400nm, 420-430nm, 430-440nm, 450-455nm, 460-470nm, 495-500nm, 515-530nm, 590-595nm, 600-605nm, 620-630nm, 655-660nm).

Design: To satisfy multiple needs, we designed two kinds of hardware. One for solid medium, the other for fluid medium.

1. Hardware 1.0 for solid medium

Process: Using thermal conductive adhesive, we fixed the LED beads with substrate to our radiator, which looked like a sunflower. In the following picture, you can see our light source component as a whole, “sunflower”!

Figure 1. Our light source component- “Sunflower”

We managed to dig a hole in a container, which our sunflower perfectly fitted in. We used wires to connect the sunflower with a power source, and the plates inside the container could be illuminated by our lovely light! You would see some of them in the following pictures.

Figure 2. Inside the container.

We made four containers illuminated by LEDs. Then we put four containers into a huge incubator. We called these containers our “Hardware1.0”

Figure 3. Four containers in the incubator.

Generally, this system worked pretty well, and we incubated several colored mediums with sufficient illumination all night.

2. Hardware 1.5 for fluid medium

The space was limited for our sunflowers in the 96-well plate shaker. We used aluminum foil instead of sunflowers. And it worked well.

The DC current source whose highest voltage was 12V could provide energy for four LEDs at the same time.

Then we fixed them inside the lid of the shaker. Four plates are separated from each other by aluminum foil to avoid light leakage.

Figure 4. Internal structure of our hardware for liquid.

Figure 5. External look of our hardware for liquid.

We tested our hardware and found some problems.

1. Light intensity was not adjustable.

2. The distribution of light intensity at the bottom was uneven, especially in the 96-well plate shaker where light intensity was high.

We used different colors in areas which had different fluorescence values or optical densities. The deeper the color was, the value that it stood for was higher. The following pictures visually display the fluorescence or optical density distribution in a certain area in a 96-well plate.

Figure 6. The distribution of optical density value of E.coli in the 96 hole plates irradiating by blue light(460-470nm). & Figure 7. Correspondence between color and value.(From small to large)

Figure 8. The distribution of fluorescence intensity of E.coli in the 96 hole plates irradiating by green light(515-530nm).

Figure 9. The distribution of fluorescence intensity of E.coli in the 96 hole plates without illumination.

The light intensity could reach 35000 lux at the center of the 96-well plates, while it was 15000 lux in marginal area.


1. As for light intensity, on one hand, we used PWM to change current waveform so that the valid value can be manipulated; on the other hand, we altered distance to change light intensity subtly.

[1] PWM

Figure 10. Pulse Width Modulation (PWM)

Figure 11. Concise schematic diagram of PWM

[2] Distance

Experiment results: The illuminance was in exponential attenuation with the increasing of the vertical distance.

y = 25151x-1.496

R^2 = 0.9804

Figure 12. Line chart of vertical distance and illuminance

[3] In order to make the distribution of light intensity more even, we built a model to simulate bottom light distribution.

Assumptions: The luminous flux of LED is the same at all angles above the surface. Only specular reflection occurs on the side face and no reflection on other surfaces. Only reflections for the first time is considered.

You could see it in detail in our modeling part.

The experiment proved that a cylinder was better to solve the intensity problem, as our modeling suggests.

3. Hardware 1.7 for fluid medium

To make the distribution of light intensity more even, we soldered some patch LEDs on a circuit broad. The patch LED was small in size, around a few millimeters. We adjusted the position of the LEDs so that each well of a 96-well plate could be illuminated by one LED evenly.

The next picture is semi-finished product. We connected these LEDs in series and used some resistances to balance the current.

Figure 13. Circuit board in process of being soldered.

We soldered additional LEDs to fit the 96-well plate. Here are three primary colors of light.

Figure 14. The successful result of the three primary colors.

4. Hardware 2.0 for solid medium

To get mixed light, we put three LED beads in one carton. We used the three primary colors of light and they produced much more beautiful light! When the red and green beads were turned on, they turned to bright yellow. Red and blue turned to purple. Green and blue turned to cyan-blue, a pretty color.

Figure 15. Bright yellow mixed by red and green.

Figure 16. Purple mixed by red and blue.

Figure 17. Cyan-blue mixed by blue and green.

5. Hardware 3.0 for solid medium

It was time for our beautiful rose. Obviously LED beads could not satisfy our demands, so we purchased a projector. By simply using a reflection mirror[3], we cast an image of a rose on a piece of filter paper.

Figure 18. Our projector and reflection mirror in an incubator.

Figure 19. The rose picture cast by our projector.


6. Final version of our hardware

To make our hardware more portable, we made a more advanced and fancy-looking one. With good human-machine design interface and high level of integration, our final hardware was very flexible and easy to operate. Its appearance was designed as a carton train. We named it Forest Train and the name meant that its destination was our rose forest. Anyone “takes” it (that means he/she put a plate in Forest Train) would see beautiful rose produced by E. coli.

We ordered transparent polycarbonate boards (PC boards) from a factory online. PC boards had high biological affinity and we used them to build the shell. To block the outside light, we paint the inner layer black. We used acrylic pigment to blacken and decorate our boards, which is waterproof and colorful.

Figure 20. Transparent PC board.

Figure 21. Blackened PC boards.

Inside Forest Train there was a projector, and we used a professional reflection mirror for projectors to illuminate a plate.

As for placing plates, we used iron stands, tested tube holders and a tripod to establish a stable structure. It was a delicate structure and it functioned pretty successful, as the plates could be placed properly without blocking the light.

Figure 22&23 The supporting device for plates.

Besides, we also used a temperature controller and two fans in our Forest Train so that it could function as an incubator. Fans were used to refresh the air in the device and accelerate the process of heat dissipation. An ultraviolet lamp was fixed on the ceiling to sterilize the device after using.

Anything unsealed on the hardware was hid in our appearance design harmoniously (such as the fan in the wheel and the “window”). The “window” was a door to access our plates and the fans were hidden in the first wheels of the train. With the well-designed hardware (2.0 version), we made some remarkable progress in lab.

Figure 24 “Window” to access plates.

Figure 25 Fans hidden in the wheels.

Figure 26 Internal structure of our Forest Train.

Figure 27 & 28 Appearance of our Forest Train

Future work

We wish in the future we can automate the process in our Forest Train and improve its efficiency. Then you can see a longer Forest Train with a customized projector, and every window can open and pick or take out a plate automatically. The customized projector can illuminate every plate. In that case we can produce beautiful products combined art and science continuously and efficiently, maybe just like a flower shop.


[1] Fernandez-Rodriguez J, Moser F, Song M, et al. Engineering RGB color vision into Escherichia coli[J]. Nature Chemical Biology, 2017, 13(7):706-708.

[2] Gerhardt K P, Olson E J, Castillo-Hair S M, et al. An open-hardware platform for optogenetics and photobiology[J]. Sci Rep, 2016, 6:35363.