Why Wine Climbs the Glass Before It Falls in Tears
The Climb Is Not the Tear
The visible part goes down. That is what makes the invisible part worth asking about.
The streaks on the inside of a wine glass—the “tears” or “legs”—return liquid to the drink. But a descending tear does not explain how liquid keeps arriving above the drink in the first place. A thin film can climb the glass against gravity, supplying the region from which tears fall.
I went into the question with a tidy explanation: alcohol evaporates, surface tension pulls liquid upward, gravity brings it back. The explanation is useful. It also stops twice before the interesting part is finished.
Evaporation does more than change the alcohol concentration. And explaining why a film climbs is not the same as explaining how it becomes tears.
A surface with a direction
Surface tension is often introduced as something a liquid simply has. For this problem, the important property is that it can be different at neighboring places.
Ethanol lowers the surface tension of a water–ethanol mixture. As ethanol preferentially evaporates from a thin climbing film, the liquid farther uphill can become more depleted in ethanol than the liquid feeding it. The resulting surface-tension gradient exerts a tangential stress toward the higher-surface-tension region. That stress can draw liquid uphill while gravity favors drainage.
This is Marangoni flow. The name is less useful than the direction: a difference along the surface drives motion along the surface. A uniformly high surface tension is not, by itself, an uphill pump.
The first correction to my short explanation came from a 2015 experiment by David Venerus and David Nieto Simavilla. They asked what evaporation was doing to the film’s temperature, not just its composition.[1]
Evaporation requires energy. As liquid loses that energy, it cools. For the mixtures and conditions in their analysis, cooling also raises surface tension. A film that becomes both cooler and poorer in ethanol uphill therefore has two contributions pushing its surface-tension gradient in the same direction.
The researchers used infrared thermography on wine and cognac films. In the wine, they reported a temperature slope of roughly minus 100 kelvin per meter: about a one-degree drop over a centimeter, if that slope extends over that distance. That is a conversion of their reported gradient, not a measurement I made.
One degree sounds small until the relevant object is a film only tens of micrometers thick. The question is not whether the entire glass becomes impressively cold. It is whether the temperature changes enough along the film to contribute a meaningful surface stress.
Their measurements and model indicate that it does. But there is an important seam in that statement: temperature was measured; the division of the driving stress into temperature and composition contributions was inferred using a model. The thermal and composition effects were not separately switched off in an otherwise identical glass.
The model simplified the transport of heat and matter—for example, omitting water evaporation from its main derivation—and it did not explain every feature of the temperature profiles. It supports taking evaporative cooling seriously, not assigning a fixed thermal percentage to every glass of wine.
Evaporation is doing two jobs. It removes a component and moves energy. The short explanation had kept only the first.
Why a swirl is more than a flourish
Now suppose we have explained the uphill supply. We still owe an explanation of the pattern: where does the thicker region form, and why does it shed downward fingers instead of remaining an even film?
A separate research team—Yonatan Dukler, Hangjie Ji, Claudia Falcon and Andrea Bertozzi—made a particularly revealing distinction in a 2020 paper: a film spontaneously advancing onto relatively dry glass is not the same initial problem as a film moving through a coating left by swirling.[2]
Their preswirled experiment used 18-percent-alcohol-by-volume port in a conical, stemless martini glass. They covered the glass to suppress evaporation, gave it a brief slow swirl, and waited with the cover on until the obvious swirling motion had settled. Then they removed the cover.
They reported a front emerging from the meniscus—the curved region where the drink meets the wall—and traveling up the glass. That front subsequently destabilized into tears descending toward the drink.
This is the point where “up” and “down” can become misleading. The rising front is a traveling change in film thickness. It need not be the path of one identifiable lump of liquid. A moving pattern and the liquid moving through that pattern are different things.
In their model, the film behind this particular rising front is thinner than the film ahead. The researchers propose that the front is a reverse undercompressive shock: a specialized kind of traveling thickness transition that can shed disturbances downward even while the front itself travels upward.
“Shock” here does not mean a sonic boom or a violent collision. The mathematical interest is in how a relatively sharp thickness transition travels, and whether small disturbances collect at it or pass through it. The glass is hosting a slow, thin-film wave, not an explosion.
The proposal connects their observed direction of motion with thin-film theory and numerical solutions. It is not a complete, directly measured map of every tear forming. Their model assumes a constant surface-tension gradient, an assumption they say eventually fails as the film climbs. Their study leaves a full two-dimensional account of the later breakup—and a quantitative prediction of tear spacing—as further work.
The swirl nevertheless changes the question. It does not merely give the liquid a shove that later runs out. It leaves a coating: a different starting thickness profile through which the subsequent evaporation-driven flow develops. Their experiments reproduced the qualitative sequence, even though manual swirling did not give them precise control over that coating’s thickness. Their one-dimensional simulations supported the proposed front mechanism, not the full breakup into tears.
More than one way to make a ridge weep
There is a tempting way to write this story: the newer shock explanation replaces the older explanation. The papers do not justify such a clean victory.
The 2015 study also considered tears falling from a ridge near the film’s upper contact line. Approximating that ridge as a liquid cylinder gave a plausible breakup spacing. But the geometry mattered: one approximation predicted about 18 millimeters, another about 8, against an observed spacing around 15 millimeters. That is evidence for a useful instability picture, not an exact fingerprint that uniquely identifies the mechanism.[1]
The 2020 paper, meanwhile, explicitly allows that downward tears can also come from liquid accumulated in a stationary ridge higher on the glass. Its moving-front proposal addresses a particularly reproducible preswirled setting. It does not establish that every tear in every glass has the same history.[2]
I find the distinction more satisfying than a single magic phrase. There are at least three things to explain: the stress that drives flow, the film shape that develops, and the instability that turns that shape into descending fingers. They are connected. They are not interchangeable answers.
The next question I would want an experiment to answer is how independently controlling the glass temperature and the initial coating changes the front that appears. The studies give reasons to ask it; they do not mean I have already tested it.
For now, the visible streak has become a little less simple. It is not merely liquid losing a fight with gravity. It is the downward part of a process in which evaporation can first help liquid climb—and in which the glass’s recent history helps decide what happens next.
Sources and limits
[1] David C. Venerus and David Nieto Simavilla, “Tears of wine: new insights on an old phenomenon”, Scientific Reports 5, 16162 (2015). Main text, methods and captions read; relevant equations checked in the retained PDF. The infrared experiment and the model-based stress allocation are distinguished above.
[2] Yonatan Dukler, Hangjie Ji, Claudia Falcon and Andrea L. Bertozzi, “Theory for undercompressive shocks in tears of wine”, Physical Review Fluids 5, 034002 (2020). Read via the complete author manuscript, arXiv:1909.09898v2, including its appendix. This independent team’s port experiment is not independent replication of the earlier temperature measurements; some historical data and theoretical foundations are shared.
This is a reading-based account. I did not perform a wine experiment or tasting, inspect figure images, view the cited videos, or reanalyze raw measurements. Simple gradient and model-arithmetic checks are not new physical evidence.