Have you attended an online call with Public Lab? We'd love your feedback!

Public Lab Research note

Separating Oil from Water by Freezing the Water

by Matej | November 13, 2015 17:28 13 Nov 17:28 | #12405 | #12405

Matej was awarded the Basic Barnstar by stevie for their work in this research note.

What I want to do

I attempted to extract various types of oils previously mixed with (tap) water by freezing the water. This idea is based on a fact that oil has usually lower freezing point than water so I should be able to collect the oil from the top of the ice. This technique has been already used to separate natural oils from water. Please keep in mind that all of the amounts of substances are not exact because I did not used calibrated pipettes. Also parts of some more viscose products remained attached to the walls of the eye dropper or a plastic flask.

My attempt and results

I mixed the following oils with water in ratio 1:20 (1.5ml oil to 30ml water): 5W-30 Crude ConEd Dielectric Oil Fluid 20W-50 80W-90



After that I froze the the mixtures in my freezer for one day. The Product Oils (PO) separated from the frozen Product Water (PO) on the surface in different volumes for each type of PO, probably mainly due to different viscosities of the oils (but also some other properties). Following image is showing all the recovered amounts in comparison to 1ml of clean water sample. a) 5W-30 recovered b) Crude recovered c) ConEd recovered d) 20W-50 recovered e) 80W-90 recovered w) water 1ml for comparison



In the first two attempts, a) and b) samples, I was using the eyedropper to collect the oil. After that I realized that the more viscose POs are attaching to the eyedropper walls so I stoppered using it. I poured the remaining products directly directly from the flask into the cuvettes.

IMG_6490.JPG I noticed that the 5W-30 (only) dramatically changed it's appearance and consistence after freezing. It's color changed from transparent honey to non-transparent milk.

Samples of Pure Initial Oil Spectra Compared and Recovered PO Spectra

I took triplicate spectra of each oil before mixing with water and after recovering the PO. Here below are some examples. You can find all spectra at https://spectralworkbench.org/profile/matej

ConEd DIelectric Fluid IN (pure)

ConEd DIelectric Fluid PO (recovered)

Crude IN

Crude PO

5W-30 IN

5W-30 PO

Questions and next steps

Next steps are as referred to in my initial research note here: http://publiclab.org/notes/Matej/11-04-2015/oil-sheen-testing

This was 1:20 ratio testing. Seems like the amount of recovered PO widely varies from awesome (approx.) 75% to not so awesome but still workable (approx.) 20% or less. In case of low amounts you might have to use the Q-tip & mineral oil method previously mentioned here and here However I did not tested the PO with mixed with mineral oil yet. I am also wondering what could be such Oil to Water ratio in an actual oil spill? This method of collecting PO is certainly working and is very accessible. However, it might be changing chemicals and spectral properties of some oils. This is still TBD, In this particular case I found major differences just in the 5W-30. Product water spectra will be also be the next step. Note: Perhaps use of "thinner" cuvettes will be more helpful in case of using less recovered PO.

Why I'm interested

I am interested to test figure out a way how to prepare best specs for the OTK and to test the ability of different different DIY techniques to separate oil from water. This is one of the methods I am trying.

I did this Help out by offering feedback!

People who did this (0)

None yet. Be the first to post one!


Nice Work

Reply to this comment...

Sorry - for those using HTTPS/SSL, the embeds won't be working at the moment, as we just upgraded PublicLab.org to use extra security, but haven't yet done the same for Spectral Workbench. We should have this resolved shortly.

This is very exciting work, Matej - I wonder if you could estimate the dilution of a field-collected sample by trying to use the minimal # of drops of an oil to reproduce a sheen. But that sounds like it'd make a toxic mess. Perhaps doing a surface collection from the Gowanus, or an oily puddle on a rainy day could be a real-world analogue of your tests here.

Keep up the great work!

Reply to this comment...

The embeds are working again. Thanks for your patience!

Reply to this comment...

Hi matej - are the differences in the 5w30 just by visual inspection or in the spectrum as well?

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

Awesome technique! the look of the 5w-30 reminds me of thawed out frozen diesel fuel. I think the milkiness in diesel from freezing is from wax crystal formation-- If the 5w-30 is heated a bit (just a bit!) I would imagine the waxes could be re-suspended. There are fuel-line and fuel tank heaters in some diesel vehicles to do this. I suspect that some of the waxy components might be missing from the re-suspended mixture because they are still stuck in the frozen container, but I could be wrong.

in terms of comparison, it looks like there might have been some clipping on the 5w-30 and ConEd spectra:


Reply to this comment...

HI Guys, Thanks for your interesting comments!

Jeff: I will certainly try to do that. In general I am trying to find out what could be a realistic ratio of oil:water. To start we figured out 1:20, I am using right now, but I would like to start working with a smaller one. Any suggestions what that ratio could be? BTW Randy Austin, chief of Spill Response Program NYC DEC, mentioned that an oil spill in general creates a sheen only if the water is quite saturated. That might mean that the ratio might not be as low as a few drops. I think I am going to do a small quick experiment on that - just adding drops to water and looking for the amount it will create a sheen.

Matthew: Very interesting! I did not know about the wax. That might be it. Will try to heat it up and take couple of spectra. Clipping of 05W30 spectra: Yes. This one is burned out (clipped) because I wanted to show the "blue tail" between approx. 255 - 350nm and approx. 430 - 530nm. Usually I take 3 spectra of each specimen. The ones I took from the recovered 05W30 are not showing much for some reason: http://spectralworkbench.org/spectrums/61164 http://spectralworkbench.org/spectrums/61163 http://spectralworkbench.org/spectrums/61162 I think I will take the spectra again.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

@Matej Lots of interesting things here! Anecdotally from the spill response team I work with, it only takes a few drops of oil to create a sheen, so if this is not the case, I'd like to correct that misconception!

I was also wondering about the other separation methods you're trying. In your 20:1 mixes, if you let them settle on their own, do they form two distinct layers? Is it enough in the field to skim liquid from the surface, and let gravity separate the oil samples? I'd also be interested in the result of reheating the frozen samples, even the ones that didn't visibly change as much as the recovered 05W30.

For oil spills that happen in the field, does the oil mostly remain on the surface (where at least it could be skimmed and sampled) or does it form suspensions, or sink to the bottom... or do certain types of "oil" or fluid dissolve in the water. Maybe a better question is, what gets spilled in different types of situations (bakken crude spill, gulf oil spill, industrial contaminants...) and where would you expect to find those contaminants -- on the surface, dissolved or suspended, or on the bottom? And if you are doing a general investigation, where should you sample from?

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


Great work, Matej- Do you have any thoughts on applying your research to collecting samples from frozen soil? How will frozen earth affect collecting samples? My kit isn't here, yet, so perhaps collecting samples in advance would work. The ground is just starting to freeze, although, with El Nino, it's been abnormally warm & wet.

There is light snow cover in the Upper Midwest today-


Maybe I can collect samples now, before the kit comes? If the ground is frozen hard, I could try collecting soil samples by pouring warm water on it. If that would mess up the results, please let me know. Many of the soils that I wish to test are mucky & slurpy, not like the beach sand demonstrated at the barnraising. Lots of colloidal suspension, I expect. The organic content of the soils may be high, too.

Thanks for your original ideas and experimental techniques.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

HI Mary, I am not 100% familiar with methods of collecting oil from soil. I think @eustatic might know better. I saw he had some diluted soil samples with him at the Barnraising. Anyways I would approach it in two separate ways: 1. I think collecting samples from frozen soil would be pretty much same process as collecting samples from not frozen soil. I would say just defrost the sample in a room temperature. Do not add more water to it. I would even try to dry the sample in the room temp, to get rid of as much water as possible. After that I will use the mineral oil which should attract the oil from the targeted sample. I would try to purify the sample to get rid of the soil remains and than try to measure the spectra. 2. The other way around might be actually (as you are saying) defrost the soil and mixing it with more water. I would wait until the mixture settles down and than I would try to freeze it and collect oil from the surface with a cotton swab. ... another way might be adding a centrifugation process to help the filtering / settling of the soil particles.

Reply to this comment...

@pdhixenbaugh: There are different oil concentrations which can be detected by naked eye. We had a very interesting debate about this at this last Barnraising. Here is a link to a note Jeff wrote during the session: https://publiclab.org/notes/warren/11-21-2015/oil-sample-concentration-session-at-barnraising Also the concentration might depend on may different factors. Oil is hydrophobic and floats on the top, but that is not always the case. in case of an oil spill (eg. BP) it took really long time for the oil to reach the surface, sometimes many miles away from the source. That means that oil could be virtually in any layer of the water level. Also I probably depends on what kind of oil, etc. We chose the ratio 1:20 as a testing model in more less "ideal" environment. That said it was 1.5ml of oil in 30ml of water. I tried to recover as much oil as possible. Recently even in repeated cycles, but I was not able to recover all. In the best case I was able to recover 1ml max. or less. So that means that the rest of the oil stayed in the water.

Yes, my 1:20 mixes after some time form two layers.

Is it enough in the field to skim liquid from the surface, and let gravity separate the oil samples? - yes that is what I would try to do but it is good to freeze the mixture because you will get more actual oil. And you can very easily collect it with the cotton swab from the ice, avoiding collecting the water by accident. Also right now I am testing the PIG Oil Only absorbent mats. That might be interesting for you. Will post that in couple of days.

check out my new research note on re-heating and re-freezing the samples here: https://publiclab.org/notes/Matej/12-15-2015/separating-oil-from-water-by-freezing-the-water-part-2 - you can find here spectra of the reheated samples. Also look at my Spectral Workbench profile. I have there much more measurements than published here in the research note: https://spectralworkbench.org/profile/matej

what gets spilled in different types of situations (bakken crude spill, gulf oil spill, industrial contaminants...) and where would you expect to find those contaminants -- on the surface, dissolved or suspended, or on the bottom? And if you are doing a general investigation, where should you sample from? - See Jeff's research note from the barnraising posted on the top of this comment. also look at the HAZCAT guide from Randy Austin, Chief, Spill Response Program, Region 2 (NY City), Division of Environmental Remediation, New York State Department of Environmental Conservation. Ideally I would take larger samples from different layers of water, but also in different places. There might be quite a difference between sampling moving body of water (river) and a static body of water where most of the oil will most likely float on the surface. Generally the oil is always trying to float to the top. Good approach would be trying to locate the suspected source and sample there. Also sometimes happens that a "pocket" of oil is trapped under ground and that can leak from the bottom of the river bed. There is really a lot of variables that come into play here. I remember that @eustatic https://publiclab.org/profile/eustatic mentioned lot of interesting things related to your question. Perhaps try to contact him. This is certainly worth writing a separate research note. I will look into that.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

Oh, just noticed -- on your dielectric fluid scan, you need to select a new cross section -- the top row of pixels is blank. Great post!

Reply to this comment...

And do you have a link to the HAZCAT guide from Randy? Thanks!

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

Login to comment.