Greetings, all. It’s been a hot minute since I’ve posted on the forum. For newer members who do not know me, my name is Nick. I am an extractor in New England with a background in ethanol, CO2, and hydrocarbon. Recently I moved on from running individual labs and started my own consulting business. Most of my work is focused on the northeast, mainly in Massachusetts, but I also work with facilities across the country.
Recently I contracted with a facility in Maine that was suddenly experiencing what many are now calling the “Medusa Stone” issue with instant crashing and rapid solvent purging while trying to make diamonds. It would start immediately after a pour with visible microcrystalline forming in the extract. It would then quickly progress to rapid crashing and purging until, within 2 hours, the solution would crash into a puck. They also noted that many of their extracts did not have the full smell and flavor profile that they were used to seeing from the strains.
This facility is known for its diamonds, so this issue was greatly impacting their business. By the time I became involved they had been having the issue for 3 weeks. It was affecting every extraction they did on two different extractors. Initially, as many have done when encountering this problem, they believed it was an issue with the solvent. While I don’t like ruling anything out too early in an investigation, contaminated solvent was low on my list of probable causes. Firstly, since their initial encounter with the issue they had received several different batches of solvent from several different suppliers, all with the same result. Second, they, like many in the northeast, use Cannagas Supply for their solvent provider. Cannagas being a Massachusetts company they have a substantial footprint in this region, including all the facilities I work with. Since the issue was not being seen by everyone all at once it seemed unlikely that it was a solvent-based issue. My leading theory going in was that it was a contaminant introduced to the system at the point of extraction, likely by something applied to the biomass during the flower cycle. I have experienced compounds coming through into extract due to foliar applications during the flower stage on many occasions, so this seemed the most likely cause. However, according to the facility they used no foliars at all during flower cycle. They are a large toll processor, so they handle a lot of different material and do not always know how it was treated. However, the problem initially occurred when they were running their own material. It also occurred on a machine that was running live and another that runs exclusively cured material. The live lined up as foliars that are applied to material taken for live extract do not have time to degrade like they would on flower that is dried down as it is frozen at the time of harvest, stopping the degradation process. The fact that it was also occurring with cured material was a significant wrinkle in my assumption.
Once I arrived at the facility I inspected the equipment. They had done a teardown just over a week earlier, so the systems were fairly clean. I did note that there was a substantial amount of water build up in both the system that runs live as well as the system that runs exclusively cured. This was unusual given how recently the systems were cleaned out and the fact that they use mol sieves that are regularly regenerated. It was also very curious that so much moisture had built up in a system that only runs cured material.
As a baseline test I took samples from each working tank by draining the feed line into a clean mason jar. The machine used for live ran straight n-butane and the machine for cured ran a tri blend 50/25/25 n-butane/Iso-butane/propane. I bled some solvent off to flush the line prior to running the solvent into the jars.
The samples from the working tanks were all highly reactive. Rapid purging was apparent and there was a small amount of white debris visible in the bottom of the jars. I also took samples from fresh stock tanks in the same manner to compare. The solvent from the stock tanks appeared entirely stable.
Initial working tank and stock tank samples
At this point we decided the best approach was to attempt to start clean, transfer the solvent form the tanks back into stock tanks, and do full tear downs of the systems. To be thorough I instructed the staff to clean everything in isopropyl followed by a wipe down with RO, then a clean in an Alconox solution, followed by another rinse in RO, followed by air drying overnight.
The next day while rebuilding the system I inquired as to the last time the MVP recovery pump had been cleaned. It turned out that it had not been opened since before the issue started 3 weeks earlier. Upon inspecting the diaphragms a substantial amount of white powder was observed.
While doing the clean up other mechanical issues were detected with the MVP causing us to remove it from the system and perform all of our tests by running passively. We also decided to pack the columns rather than using reusable bags in an effort to limit cross contamination variables.
Once the system was back together we took a sample from the stock tank and observed it to be stable. We then filled the working tank with approximately 30 pounds of n-butane. We then took a sample from the working tank and observed it to be stable.
Stable stock tank and working tank samples
We then ran some of the same material they had remaining from the run they were doing several weeks earlier when the issue first occurred. Despite starting with a clean system and verifying the solvent we started with was stable the extract began reacting immediately.
Reactive pour from clean system and stable solvent
We then did another full breakdown following the same procedure as above.
On the next day we again took a sample from the stock tank and the filled working tank. This time we also poured off and collected the mystery oil fraction from the distillation as well as a sample where we flushed the solvent from the clean working tank through the clean system and collected it from the recovery vessel. All were observed to be stable
Stable samples from stock tank, mystery oil, working tank, and system flush
On this day we ran some material that was bought in from another grow. The first pour had no issues at an hour after the pour. The second pour began crashing within 20 minutes.
Stable first pour, highly reactive second pour
We then took a sample from the working tank and found that it was now reactive.
Reactive working tank sample after running material on clean system
It was at this point that I began to really question my assumption that it was a contaminant in the material and not in the solvent. Despite starting each day with apparently stable solvent after running material from two different sources we were seeing not only reactivity in the extract but also seeing that reactivity making its way to the working tank. I began to wonder if it was a compound that was in the solvent to start with that we were concentrating with each recovery, causing the contaminant to build in the working tank and make the solvent more and more reactive. We ended that day as we did every other day—with a full tear down.
The next day I came up with a plan to attempt to test if we were concentrating a contaminant that came to us in the solvent. To do this we started as we had each day with a sample from the stock tank, the working tank, the mystery oil, and the flush through the clean system. All stable. We then went through repeated rounds of filling the empty material column and then recovering all but a single mason jar. We then collected that as a sample and also sampled the working tank with each round. With each round of redistillation the sample from recovery was utterly unreactive. The first couple of rounds of samples from the working tank did have small amounts of reactivity. However, by the third of redistillation the reactivity from the working tank had all but disappeared. And what reactivity had existed dissipated relatively quickly and became stable. This is the opposite result of what we would expect if we were concentrating a contaminant in the solvent.
At this point we ran one column of the same material as we had run the day before. The extract was instantly reactive and the working tank sample after the run was highly reactive. There was also unexplained white particulate in the sample from the working tank.
Reactive extract and working tank sample
This, again, pointed me back at the material. It was also discovered at the end of this day that each time the system had been cleaned it was not completely isolated from cross contamination. The same molecular sieve socks had been used each time. We once again did a full tear down and fully removed the sieves.
That evening on the way home I was speaking to the owners of the company trying to identify the source of the contaminant. They mentioned to me that several months earlier they had installed machines in every room of the grow that are designed to take ambient moisture in the air and catalytically convert it to ionized hydrogen peroxide. They asl said that when this started was the first time they had extracted material that had been grown with these machines active.
Early on when these issues started happening in the industry as a whole I had suspected hydrogen peroxide as a cause. I had found an article on Pubmed about using hydrogen peroxide with LPG to make more efficient natural gas engines. The addition of hydrogen peroxide to the LPG (a 70/30 mixture of propane and butane) dramatically increased the volatility of the LPG, allowing it to ignite and burn at a lower temperature, leading to more efficient combustion. My working theory was that the increased volatility caused the rapid purging, which lead to rapid crashing. The decomposition of hydrogen peroxide is also an exothermic reaction, which would only increase the purge rate, leading to more crashing. Hydrogen peroxide is also a mild acid, which, again, will increase the rate of crystallization. Lastly, crystallization itself is an exothermic process, further increasing the purging, further pushing the reaction forward until it essentially runs away to completion. Lastly, the products of hydrogen peroxide decomposition are an oxygen radical and (dramatic music) good old H2O (recall the excessive water found in the working tanks when I arrived). It also turned out that the bought in material we had run the previous couple of days was from a grow that uses the same equipment in their grow.
This called for a test! I asked the owners to have someone at the grow place clean mason jars over the business end of one of these emitters. Since the amount of H2O2 they created was relative to the amount of moisture in the air I asked them to do it in a freshly populated dry room. I asked for samples to be exposed for 30 minutes, 1 hour, 2 hours, 6 hours, and overnight. The jars were to be lidded immediately after exposure, wrapped in cloth to keep out light, and placed in a freezer to simulate what would happen if the hydrogen peroxide were on fresh frozen flower.
The next day we rebuilt the system, this time with no sieves, took our samples (all stable) and began running toll material that we knew had no exposure to any hydrogen peroxide.
While we were running the material the jars that were exposed to the hydrogen peroxide emitters arrived. We cracked open a fresh stock tank of n-butane, took a sample, and observed it to be stable. I then added stable gas to the 30 minute exposure jar.
Instant reaction. Rapid, rapid purging. I made a mark to show the solvent level and moved on to the 1 hour jar.
Again, instant reaction. I began periodically checking the purge rate qualitatively. With longer exposures I was seeing significantly faster purging.
The 2 hour, 6 hour, and overnight jars all reacted with increasing volatility.
I then consolidated all of the H2O2 jars and placed them next to a jar of stable solvent from the stock tank. I marked the level of solvent in each and checked the level every 5 minutes. For the first 30 minutes the H2O2 jars were purging at roughly twice the rate as the stable jar. This began to level off over time. At about an hour out the H2O2 jars were purging roughly 1/3 faster. It was interesting to note that even after the apparent reactivity of the H2O2 jars had essentially stopped the purge rate was still observed to be faster than the stock tank solvent.
At this point the first pour from that day’s run was ready. Recall that this was the first time we were running with a truly clean, cross contaminant free system that was known to not be exposed to H2O2. It came out appearing clean. No microcrystalline or rapid purging. 30 minutes later it was still stable.
Clean jar 30 minutes from pour
Jar two from that run came off about 30 minutes later—it also appeared to be clean.
So, it seemed like we had found our culprit. All the tests seemed to line up. However, the conclusion is not without wrinkles. While hydrogen peroxide would make sense for the fresh frozen material it doesn’t explain the cured material causing the same issue nor does it explain the highly persistent nature of the cross contamination. H2O2 has a half life of just a few hours. There is some evidence that ionized hydrogen peroxide might stick around for a few days, but that still doesn’t explain why cured material with hold on to it for so long or why the introduction of anything that had been exposed to a previous impacted extraction would instantly carry the contamination over. Other operators that I knew of who had with problem were able to resolve it with a single teardown. My client kept having it come back over and over again with even the smallest cross contamination vector. Again, I was stumped.
So, again, here we go to Pubmed. I started researching what could extend the half life of hydrogen peroxide. It turns out there are a number of ways to make it last significantly longer. Two of which involve mixing the hydrogen peroxide with either sodium citrate or sodium phytate. While I do not have direct evidence that either of those salts are specifically produced by cannabis they are produced in abundance in most plant species, particularly sodium phytate. If you mix even a 1% by volume amount of sodium citrate with H2O2 its half life becomes several weeks. If you mix a small amount of sodium phytate with hydrogen peroxide the half life jumps from a few hours to a few months! Most hydrogen peroxide exposure to the plants comes in the form of spays that periodically expose the plants for brief periods of time. This is enough to potentially cause some mixing and stabilization and, for fresh material, if done close to harvest, could have an immediate but short term impact on the extraction operation. In the case of my client the equipment they had installed was exposing the plants to the hydrogen peroxide continuously for the entire lifecycle of the plant. If even a small percentage of the hydrogen peroxide were being stabilized and that kept happening continuously for months it could easily have a cumulative effect, building up ever greater amounts of stabilized hydrogen peroxide. That stabilized hydrogen peroxide would not only transfer between extractions of sieves and reusable socks but would also survive the dry down and be present on the cured material as well. There is also the factor of the white particulate that we saw not only in the beginning in the MVP but also in the working tank on the clean system. My assumption is that this was microcrystalline THCA that had been sucked into the recovery and made it’s way both to the MVP and the working tank. It is conceivable that some amount of hydrogen peroxide could have become trapped in the crystal structure when it rapidly formed. That could allow for the THCA itself to be a vector for cross contamination as well. I know there are a few assumptions here but they are educated. Ultimately, it adds up pretty well.
I want to be clear that while I believe the data here supports the conclusion that hydrogen peroxide contamination was the cause of the reaction in question I am not declaring that every organization that is experiencing the “Medusa Stone” issue is caused in this way. I will say that hydrogen peroxide applications as well as equipment such as that used in this grow are becoming more and more prevalent in the cannabis growing world. I am posting my thoughts, methods, and conclusions in the hope that they can be tested and repeated by others. I am an extractor but I will always be a scientist first. Do your own tests, form your own conclusions, and come back here to share them. That’s how we move the ball forward as a community.
Stay medicated, my friends!
Nick
