How the world’s longest running experiment breaks physics

Most everyday materials such as glass, plastic, even mayonnaise have a combination of solid and liquid properties.


Skanda Vivek

3 years ago | 8 min read

One drop falls every 10 years — questioning the very foundations of statistical physics. Is it solid or liquid? The answer it turns out, is not so simple and apart from being a fundamental unsolved problem in physics, cascades into our everyday lives.

In 1927, Professor Thomas Parnell started an experiment to demonstrate to students that some substances have both solid and liquid properties. Parnell poured a heated sample of pitch into a sealed funnel and allowed it to settle for three years. Turns out, three years was far too less to play out the entire experiment.

To this day, droplets fall roughly once every 10 years. But if you were to smash it with a hammer, it would shatter like a solid.

Pitch after being hit by a hammer
Pitch after being hit by a hammer

While pitch might sound like a special material, it is in fact not. Pitch is derived from petroleum or coal tar, and is used to make road asphalt and in commercial roofing as a waterproofing agent. In fact —

Most everyday materials such as glass, plastic, even mayonnaise have a combination of solid and liquid properties.

So why is this problem still unsolved? Enter the glass transition.

Liquid — Solid phase transition versus the glass transition

Abrupt change in volume accompanying the melting phase transition | Skanda Vivek
Abrupt change in volume accompanying the melting phase transition | Skanda Vivek

Statistical physics predicts an abrupt discontinuity of a measurable physical quantity accompanying the phase transition from solid to liquid or vice versa. Take volume for example: when a typical solid is heated, it’s volume increases with increasing temperature. At the melting point, there is an abrupt jump in volume accompanying the solid — liquid phase transition.

Along with this jump in volume, there is a latent heat (or enthalpy of fusion) associated with the melting of the solid. This heat is the energy required to overcome forces of attraction between molecules in the solid, at a constant temperature so that the material makes the leap abruptly from solid to liquid. So how is that different from the liquid — glass transition?

Debenedetti and F. H. Stillinger. “Supercooled liquids and the glass transition”. Nature, Vol 410, 8 March 2001.
Debenedetti and F. H. Stillinger. “Supercooled liquids and the glass transition”. Nature, Vol 410, 8 March 2001.

A liquid approaching the glass transition bypasses the melting point and instead continues to be a liquid at low temperatures. This liquid is now called supercooled. Supercooled liquids tend to have large viscosities that get even larger as temperature decreases.

A good example of a supercooled liquid is pitch from the pitch drop experiment I discussed earlier. In an earlier post I discuss what viscosity is. Essentially viscosity is the ease by which a liquid can be spread. More the viscosity, the harder it is to spread. For example cold butter is more viscous and thus harder to spread on toast than warm butter.

Through measuring the time pitch drops have taken to fall, scientists place the viscosity of pitch as 2.3x10¹¹ times that of water —

Pitch is 230 billion times more viscous than water!

Once temperatures are even lower however, this leads to the conundrum of the glass transition:

If a liquid is absurdly hard to spread — on geological timescales, is it a solid or an extremely slow liquid?

Humanity has known glass blowing since the 1st century BC. Once glass is blown at a high temperature and cooled, it eventually becomes hard like a solid.

What you might not have known is that when glass is blown and then cooled, it might take millions of years to reach a final equilibrium configuration (An interesting study looked at the glass transition of 20 Million year old Amber fossils). Compared to those timescales, the pitch drop experiment is still a baby!

Here’s the conundrum:

In crystals, molecules form an ordered structure such that viscosity is infinite: solids cannot be spread (you can’t spread butter on toast unless you deform or fracture it). Is there a similar glass transition temperature below which viscosity becomes infinite? But how do you experimentally verify this infinite — considering a liquid near the glass transition might take millions of years to flow?

Is there hidden order in disorder?

At the glass transition, things slow down logarithmically. After cooling freshly blown glass, it might take a minute for the first round of solidification where the viscosity increases by say a factor of 10. To increase viscosity a factor of 20 it takes 10 minutes, a factor of 60 increase in viscosity takes a million minutes.

Soon you realize that things get really long, really quick. An old running joke among researchers is the first data point takes a minute, second 10 minutes, third 100 minutes, but by the 5th or 6th data point you exceed a PhD student’s time at school.

Waiting for infinity to occur is not the only way to determine if there is a glass transition or not. The transition from liquid — crystal is accompanied by a distinct change of structure: disordered liquid to ordered crystal.

Lucy Reading-Ikkanda/Quanta Magazine
Lucy Reading-Ikkanda/Quanta Magazine

From a quick glance, liquid and glass have the same structure. Whereas crystals have an ordering of molecules.

The great debate is whether or not disordered glass has some sort of hidden order, not entirely obvious from a quick glance.

There are multiple competing theories on whether or not there is a hidden order associated with the glass transition. Some popular theories predict an abrupt ideal glass transition temperature, below which all motion ceases. At this ideal glass transition temperature, molecules are perfectly locked together such that each molecules motion is correlated with all the other molecules.

Many times this is thought to be the same as if spheres were randomly packed together as close as possible — there’s no obvious order but no sphere is moving anywhere. But how do you measure this sort of order? Whether or not there is a structural signature of random order is still an unresolved question.

Random Close Packing of Different Sized Spheres | Vasili Baranaua and Ulrich Tallarek, published by The Royal Society of Chemistry

Hidden order in our everyday world?

A tantalizing question is whether or not our everyday world has some sort of hidden order. From stock markets to housing crashes and epidemics: this touches on knowing whether these emergent phenomena have hidden underlying order and can thereby be predicted.

Recent research found that indeed, there is a very specialized order associated with disorder, called hyperuniformity. In 2014, it was found that a Chicken’s eyes have different colored receptors oriented in a seemingly random structure, but actually had hidden order.

Lucy Reading-Ikkanda for Quanta Magazine
Lucy Reading-Ikkanda for Quanta Magazine

While it may seem like there is a similar order associated with other disordered phenomena like traffic jams and large crowds, keep in mind that hyperunifomity was discovered only in certain systems. In our quest to find hidden order in everything we see, we must not forget why we are trying to do so in the first place. This brings me to the final discussion:

What we should do when the laws of physics fail to predict many everyday phenomena from the liquid — solid transition to traffic jams and stock markets?


When do the laws of physics fail? What are the practical consequences we have to live with?

We mustn’t forget that physical laws are as important as the phenomena they predict. Also, physical laws and the laws of nature in general might be set in stone. But even stone is subject to change. Statistical physics was envisioned to understand materials and changes from one form to another.

It is a law of averages. Individual atoms and molecules are random and unpredictable. However, centuries of work have found that statistical physics is an excellent approximation of group behavior of molecules that make a solids, liquids and gases. But that’s what physical laws are — they are approximations.

Physical laws are an approximation of reality, not the other way around.

When the right conditions are not met, the laws of physics fail. It’s like looking at a lawn. In the picture below, the lawn can be approximated as a green lawn, and the small bald patches don’t change that large-scale approximation.

But if instead of the bald patch, there are small highly infectious patches of weeds, that might change the entire landscape and you don’t have a beautiful lawn anymore.

Approximately a green lawn | Courtesy of the author

There are multiple reasons why established laws in statistical physics fail to describe many phenomena in the everyday world. These include:

  1. Lack of averaging out: In traditional crystals, atom motions average out so that they are neatly arranged in lattices, whereas in glasses atom motions don’t average out so nicely. This can be extended to larger societal aggregations as well. Sometimes one annoying driver or accident can lead to huge traffic jams or a presidential election change the trajectory of a country.
  2. Finite size effects: The world is not infinite sized, whereas materials can be approximated as containing infinite number of molecules.
  3. Timescales: The solid to liquid transition can be approximated as discontinuous or taking zero — time. However, the transition from liquid to glass might take millions of years (we don’t know for sure). Or the timescale for societal events is much longer relative to our lived experience. Thus things become harder to predict during this short time. Statistical physics is not just a law of average space, but also average time.

But it is not all bad. We can have a good approximation of reality and predict as well as shape events in the future — much better than we are currently doing. Just not everything. For example, in hindsight many scientists correctly predicted COVID would be a huge pandemic without social distancing.

Much harder to predict would be exactly when a pandemic occurs. Physics has a huge benefit, in that —

Physical approximations are rooted in causation, rather than correlation

While vast amounts of data has led to burgeoning of data science and machine learning methods to predict the future such as stock markets and housing prices, many of these methods horribly fail with a slight perturbation, because they are correlative rather than causative.

For example, simple distortions to stop signs can fool complex neural networks that have not been trained on these distorted images.

Subtle perturbations cause a neural network to misclassify stop signs | Evtimov et al
Subtle perturbations cause a neural network to misclassify stop signs | Evtimov et al

In conclusion — physical laws fail when their underlying assumptions are not met. In some cases, another law is found to replace the existing law. But in many other cases, there might not be a magic law that describes everything.

The theme of this blog is a tantalizing reminder of this struggle. Is it possible to resolve whether pitch is a solid or a liquid : something that has the structure of a liquid but is a solid on most practical timescales? — and is there a hidden order underneath the disorder? If so, would that mean that underneath the unpredictability of our everyday world lies a beautiful order?

Or, maybe the answer is that somethings are more predictable than others? And rather than thinking in terms of predictable or unpredictable, we need to have a spectrum of predictability?

Either scenario shouldn’t discourage understanding why existing laws fail, thereby helping us better understand our complex world. The advantage of a balanced approach to find out the right approximations behind complex emergent phenomena lies in building and shaping sustainable and resilient futures.


Created by

Skanda Vivek

Senior Data Scientist in NLP. Creator of







Related Articles