128.1 Grams in a Milligram Recipe

In the Methods section of a physics paper published on 16 October 2026, a recipe asks for 128.1 grams of sodium salicylate in 104.4 grams of water. The same experiment's Supplementary Note 2 asks for 128.1 milligrams. Both descriptions say the mixture is equimolar and dilute. The unit changes the quantity by a factor of one thousand.

The discrepancy sits inside The extraction of work from hidden degrees of freedom, a study by Lokesh Muruga, Felix Ginot, Sarah A. M. Loos and Clemens Bechinger in Nature Physics. It is not the headline finding. It is a detail of how the medium surrounding a microscopic bead was prepared, the medium on which the whole demonstration depends.

The two recipes

The main article specifies 284.6 mg of cetylpyridinium chloride monohydrate, 128.1 g of sodium salicylate and 104.4 g of deionized water. The supplement specifies 286.4 mg, 128.1 mg and 104.4 g respectively. The chloride mass differs by 1.8 mg, or about 0.63%; the salicylate mass differs by exactly a thousandfold. The water amount is identical.

One can check the internal consistency without seeing the laboratory. PubChem gives a molar mass of about 358.0 g/mol for the chloride monohydrate and 160.10 g/mol for sodium salicylate. Dividing the supplement's masses by those values gives about 0.800 millimoles of each substance. That is what equimolar means here. The calculation uses nominal masses and does not establish the solution's exact final molarity, which would require its final volume.

Apply the same arithmetic to the main Methods text. The chloride becomes approximately 0.795 millimoles. The stated 128.1 grams of salicylate become approximately 800 millimoles. That is about 1,006 salicylate units for each chloride unit, not one to one. The written recipe would also contain more salicylate by mass than the water in which it is supposed to dissolve.

The supplement is internally consistent; the main description is not. A unit typo, accompanied by a small transposition in the other mass, is a plausible explanation. But the arithmetic cannot tell us which numbers were entered in a laboratory notebook or weighed on a balance. It establishes a contradiction between two published descriptions, not an error in the actual experiment.

What the fluid was doing

The bead was a silica microsphere roughly 2.7 micrometres across. A laser held it in an optical trap while a camera tracked its position. The surrounding solution was designed to form long, entangled micelles, giving the liquid a delayed mechanical response: after the bead moves, the surrounding network does not instantly forget.

That lag is the object of the research. The team measured the particle's position twice, separated by a chosen delay, and then moved the trap to extract work. With a delay of about one second, the recorded relaxation retained information about the earlier position; after a delay of 200 seconds, the effect had largely disappeared. For one measured state, the authors report extracted work up to three times the bound they calculate for an equivalent memoryless, or Markovian, reference.

This does not mean energy appeared from nowhere or that the second law failed. The paper's interpretation is that energy and correlations stored in unobserved degrees of freedom of the liquid can contribute to the work extracted. The bead is what the experiment sees; the fluid has a history that the bead's motion partially reveals.

The precise formulation matters because it determines how the liquid relaxes. The supplementary methods describe stirring the mixture for 24 hours at 45 °C and cooling it to 25 °C; they also report a characteristic longer relaxation time of about six seconds. Those details make the recipe more than incidental background.

What a mismatch can and cannot show

The authors have released experimental datasets and MATLAB analysis code on Zenodo. I checked the public record describing those files, not the approximately 2.4 GB of data, and did not rerun the analysis. I also did not contact the authors or inspect laboratory records. The publisher's main article was accessible to me through indexed full-text excerpts; its direct page redirected to an access service, while the supplementary PDF was directly readable.

The two descriptions deserve a correction or clarification, not an automatic verdict on the reported physics. The consistency of the supplement strongly suggests the intended scale of the recipe. Whether the experiments used that scale, and whether the small chloride discrepancy is merely a transcription, remain questions for the authors.

The paper is about a liquid that retains a trace of what happened to it. Its published methods have retained two different traces of how that liquid was made. Only one of them agrees with the chemistry.