Epiphyte is capturing CO2!

There are two qualitative indications that a sorbent-based DAC system like Epiphyte is successfully capturing carbon dioxide:

  1. When the sorbent is heated, the concentration of CO2 in the closed plenum chamber rises (desorption phase).
  2. When air is blown through the sorbent, the CO2 concentration on the output (leeward) side is less than that on the input (windward) side (adsorption phase).

Recent tests have shown definitively that Epiphyte is meeting the first criterion, and there are good signs of the second. This picture shows the location of the CO2 sensors (Sensirion SCD30); the airflow is from left to right; the sensors also measure relative humidity (RH) and air temperature:

Desorption Test

I started with the desorpton test, also called the reset bake, as the goal was not only to measure how much CO2 was captured previously, but also to get the sorbent to release all the captured CO2 and water in anticipation of the adsorption phase to follow.

The diagram below shows schematically the configuration of Epiphyte in the desorption phase, and the locations of the sensors. Note that the CO2 sensor on the output side is not located in the same corresponding position as on the input; this was due to difficulty that I encountered in drilling through the wall of the damper assembly.

I closed the dampers after making sure the plenum was well-aired out, and left the fan off. Then I set the heating wires to 140C, then to 160C, then to 180C. The temperature profile over time is shown here:

The red and pink traces show the measured temperature of the heating wires, which contact the front and back sides of the sorbent; the green trace is the temperature of the sorbent in the middle; and the blue trace is the temperature of the aluminum panel that holds the sorbent in place. We can see, as reported before, that the sorbent heats up much more slowly than the wires; and the frame parasitically draws a good deal of the heat away, an inefficiency that will need to be improved in successive builds.

At about 51 minutes into the test, I opened the dampers and blew air through the Epiphyte with the fan for about 2 minutes, before resuming the test with the closed chamber. This was to see whether the sorbent had released all the captured CO2 and H2O. The graph below shows the CO2 concentration (in ppm) and the relative humidity (in %) measured by the CO2 sensors on both sides of the sorbent panel:

The CO2 concentration before starting the test (around 500ppm) can be regarded as the baseline, and any rise above this can be assumed to be due to CO2 released from the sorbent; the same applies to the RH. So the fact that the concentration rises significantly is our proof that Epiphyte has captured some CO2 and has passed the “Hello World” test!

Note that much more CO2 and water vapor are released on the input side; this is likely due to the simple fact that the side first receiving the air will adsorb more of these molecules. Also, due to the fact that the CO2 reaches a peak and then drops off (due to leakage from the ductwork), and does not rise again after the temperature is increased or the chamber is exhausted, is a strong sign that we have managed to release all the captured CO2 from this sorbent panel, and the sorbent has been fully “reset” and ready to start adsorbing.

From this data, and the dimensions of the Epiphyte chamber, it is possible to estimate the total amount of CO2 and H2O that were previously adsorbed and then released (in a followup post I will calculate this).

Adsorption Test

The diagram shows the configuration for the adsorption test: Dampers open, heater off, fan blowing air through the sorbent panel:

The graph below shows measured levels of CO2 and relative humidity using the same sensors as before, but this time with air flowing past:

The first feature that jumps out is that the CO2 measurements are now very irregular. My guess is that this is due to turbulence from the moving air and the resulting changes in local air pressure. The RH measurement is much better-behaved. However, despite the ugliness of the CO2 traces, we may be able to gain some useful insight. First, let’s plot the difference (delta) between the measured CO2 concentration and RH on the input vs. the output:

[Note the position of the y-axis]. This graph for CO2 still has a lot of noise, but we can get the impression that it is positive more than it is negative, indicating some adsorption of the CO2; the RH trace makes it clear that water is also being adsorbed as expected.

Finally we can plot the cumulative delta, or the integral of these traces:

This can be interpreted as a measurement of how much CO2 and H2O have been adsorbed, although to get a useful answer, we need to know the rate of airflow and the dimensions of the ductwork; measuring these will be the next order of business in the lab.

From these results, we can see clearly that adsorption is taking place. However, after about an hour, the integrated CO2 line begins to decrease while captured H2O continues to rise. This may mean that the continued adsorption of water is crowding out the CO2 and forcing the sorbent to release some of it even without heat. Further work is needed to determine the optimal operating conditions for efficient capture of CO2.


In one of my recent presentations, I reported a problem where the CO2 sensor was putting out glitchy or missing data. I discovered that the supply voltage that was reaching the SCD30 was below the specified operating voltage, due to loss in the wire connecting it with the processor, and to the presence of an unnecessary voltage regulator on the Adafruit breakout board. I ran a separate, larger supply wire and bypassed the regulator; I also made sure that the wires to the sensors were not close to the high-current wires feeding the heating circuit. After that, the problem went away.