163 | Nigel Goldenfeld on Phase Transitions, Criticality, and Biology
What this covers
Nigel Goldenfeld leads a conversation about how the mathematics of phase transitions—the physics governing abrupt shifts in systems like water freezing or magnets becoming magnetized—offers tools for understanding biological complexity and the origin of life. The core insight is that scale interference, the way microscopic detail refuses to disappear near critical points, and universality, the surprising fact that wildly different systems obey identical mathematics at transition points, reveal hidden structure in phenomena ranging from turbulence to evolution. Goldenfeld demonstrates how renormalization group methods that resolved infinities in quantum electrodynamics also explain why the genetic code is optimized against translation error, and he connects the physics of critical phenomena to questions about how life began and what defines living systems.
The conversation spans territory that moves from solid state physics into biology and cosmology. Goldenfeld argues that scale-free behavior observed in nature—evolution, sand piles, ecological collapse—often masks a hidden tuning parameter, so observing a pattern tells you little about the process that generated it. He discusses how Carl Woese's discovery of archaea altered understanding of the tree of life, proposes that turbulence onset obeys directed percolation physics, and develops a model in which life's function is to accelerate planets toward equilibrium by harnessing chemical-potential gradients. He also contests the notion that sand piles and other systems exhibit self-organized criticality without external tuning, showing that apparent scale invariance can arise from omitted details in the models themselves. The argument weaves between settled physics and speculative biology, grounded throughout in how transferring mathematical frameworks across domains can reframe intractable questions.
Goldenfeld argues that the physics of phase transitions and critical phenomena—especially the lesson that small scales cannot be neglected near criticality (scale interference) and that very different systems share universal behavior—provides a powerful, transferable toolkit for understanding complex systems from turbulence to evolution and the origin of life.
- Near a phase transition, the naive assumption that microscopic scales wash out is wrong, and this 'scale interference' is what the renormalization group resolves.
- Universality means wildly different systems (magnets, percolating coffee, turbulent pipes, predator-prey ecosystems) obey identical mathematics at their critical points.
- Apparent scale-free behavior in nature (evolution, sand piles) often hides a tuning parameter, so observing a pattern does not reveal the generating process.
Phase transitions occur when local interactions collectively generate global order without external coercion or fine-tuning.
- A phase transition is a cooperative phenomenon in which spins interacting only with their nearest neighbors can, below a critical temperature, collectively all align in the same direction, so that local interactions propagate to produce global order without any global coercion.
“So it is a conspiracy. Phase transition is a conspiracy. It's a co-operative phenomenon. You're not interacting with everybody in the world, but just by interacting with your neighbors, everybody ends up doing the same thing.”
- A phase transition is a trade-off between energy (minimized by spins aligning with neighbors) and entropy (disorder induced by thermal fluctuations); higher temperature favors the entropic effect and lower temperature favors the energetic effect, so dialing temperature drives the system between order and disorder.
“And so the phase transition is a trade-off between energy and entropy. And the higher the temperature, the more the entropic effect, the lower the temperature, the more the energetic effect.”
- Unlike the Ising model where an external actor must carefully tune the temperature, in systems like turbulence and evolution there is internal dynamics that naturally drives the system toward the critical point—e.g. the evolutionary timescale being far larger than the ecological timescale acts as the hidden tuning parameter—so criticality emerges intrinsically rather than being imposed.
“in examples like this, I think probably both turbulence and in evolution, there's some internal dynamics that naturally takes you to that critical point. Is that a fair thing to say?”
- Critical phenomena are the cluster of effects—total breakdown of usual physical assumptions, scale interference, scale-invariant fluctuations—that occur in a small temperature interval around the critical temperature where order first appears, but only in continuous (gradual) transitions, not in abrupt transitions like ice melting.
“all the phenomena, the epiphenomena, if you will, that are associated with this weird scale interference, all of those epiphenomena happen in a small temperature interval around that temperature, and the things that you see there are like a total breakdown of the usual physical assumptions.”
- Some phase transitions, like liquid-to-ice, occur suddenly (all the water turns to ice at once), while others, like magnetization or superconductivity, develop order gradually as temperature lowers; the gradual case is where calculating the amount of order proved to be an extraordinarily hard problem.
“Something like a liquid-gas transition is something that occurs just suddenly... In other kinds of phase transitions, like a magnet, like a super-conductor, things like this, but many other things, that transition occurs gradually”
- In the Ising model, spins represent electronic dipole moments of atoms on a crystal and interact only with their nearest neighbors, with no global coercion.
“the spins are meant to represent say electronic dipole moments of atoms on a crystal, and they interact with their nearest neighbors, but otherwise there is no global coercion for them to do anything.”
Scale interference between microscopic and macroscopic scales is the fundamental mechanism underlying all critical phenomena.
- The reason it took until roughly the early 20th century (with Einstein's work on Brownian motion) to establish that matter is made of atoms is that nearly all mechanical and thermal properties of the world can be explained treating matter as a continuum, because microscopic structure washes away at large scales—except at phase transitions where this breaks down.
“you can explain pretty much everything that you see in the world around you, mechanical properties, thermal properties and so on, without having to assume the existence of atoms. Everything happens in a way that looks just like some continuum”
- When physicists calculated the order parameter near a continuous phase transition, the answer disagreed completely with experiment, and crucially nothing in the model's description could account for the discrepancy even in principle—it was not a matter of tweaking a parameter—which signaled that a fundamental assumption was wrong.
“actually, there is nothing you can think of in your description that is responsible for this discrepancy. It's not like, well, you just sort of change this parameter just a little bit... It's like you can't even understand in principle how this discrepancy can occur.”
- In phase transition calculations the small expansion parameter is related to dimension (e.g. expanding around four dimensions, as in Wilson's calculation in 3.99 dimensions), and the same infinities appear, but here they are understood: they are the mathematical signature of the false assumption that scale interference can be ignored, so putting the small scale back into the problem resolves the infinities.
“the calculation breaks down because it's fundamentally making a wrong assumption. It's making the wrong assumption that the scale... You don't have scale interference... That is wrong, and the failure of that common-sense assumption shows up in the mathematics as infinities in the theory.”
- Near a phase transition, the common-sense assumption that a very small length scale (the atomic lattice spacing) can be neglected relative to the large correlation length is wrong; the small scale cannot be lost sight of, and this 'scale interference' between microscopic and macroscopic scales is precisely what produces critical behavior and was the assumption that wrecked earlier calculations.
“Just because something is small doesn't mean it's negligible, and in the phase transition problem, this fact that you cannot lose sight of the small scale, if you do, you will not... You cannot possibly explain the experiment.”
- In critical systems there is a wide intermediate range of length scales between the lattice spacing and the correlation length where nothing tells you which scale you are at; shown a picture of spins near the critical point without a scale bar, you could not identify the scale, which is scale invariance and is mathematically identical to a fractal.
“if I didn't put the scale bar on the picture, you wouldn't know what scale you're at, and so we call that scale invariance and it leads to very particular mathematical consequences.”
Directed percolation and phase transition mathematics apply to turbulence, ecology, and evolution as non-equilibrium critical phenomena.
- Water filtering through coffee grounds illustrates percolation: there is a critical packing threshold—neither too loose nor too dense—at which water just manages to get through, and this percolation point is a second-order phase transition that mathematically exhibits the same critical phenomena as a magnet becoming magnetic at its critical temperature, but it is non-equilibrium and 'directed' because gravity prevents flow upward.
“that percolation point is a second order phase transition, and it behaves in the way that I've told you. It turns out that it really mathematically exhibits the same phenomena as, say, a magnet becoming magnetic at the critical temperature.”
- Directed percolation processes are anisotropically scale invariant—the scaling in the driven (vertical/time) direction differs from the scaling in the transverse direction because the system is out of equilibrium and the flow can go sideways but not back up—and this is mathematically the same as Reggeon field theory used to model high energy hadron scattering.
“it's mathematically identical to what happens in high energy hadron scattering... Reggeon field theory was an attempt to understand high energy scattering. And it turns out mathematically to be the same as this process of directed percolation.”
- In ecology, the balance between how voraciously predators eat prey and how fast prey reproduce produces a coexistence point—where neither population goes to zero—that behaves as a directed percolation phase transition, with the absorbing state (extinction, from which the system cannot recover) being the key feature, and at the critical birth rate the behavior is scale-free.
“depending on how voraciously the predators eat the prey... you can get a co-existence... And there's a point where they just co-exist. And at that point it's like a phase transition as well. And it also is a directed percolation transition.”
- The onset of turbulence as flow speed increases in a pipe is not merely analogous to a phase transition but is an exact prediction of a non-equilibrium phase transition in the universality class of directed percolation, and this prediction has been verified experimentally in at least one system and is being confirmed in actual pipes.
“it's a non-equilibrium phase transition 'cause the fluid is definitely not in equilibrium... It's what's called a non-equilibrium phase transition in the universality class of directed percolation.”
- Turbulence appears robust rather than delicately tuned, but the puzzle is fake: the analog of tuning to the critical temperature is making the fluid as turbulent as possible, which corresponds to driving the viscosity toward zero; at intermediate viscosity you see turbulence over only a limited range of scales, and as viscosity decreases the scale-invariant range grows, just as the fractal range grows approaching the critical temperature in a magnet.
“the analog of tuning the temperature to the critical temperature is the analog of making the fluid as turbulent as it possibly can be... as you make the viscosity going to zero of whatever fluid you're talking about, that tunes you towards a point at which you would see complete scale invariance in a turbulent fluid.”
Phylogenetic trees exhibit fractal, scale-invariant branching patterns from interplay between evolutionary and ecological timescales.
- The fractal scale-invariance of phylogenetic trees can be an outcome of the interplay between ecology and evolution—when organisms (like beavers damming rivers) change their environment, that activity can alter the evolutionary history of the entire ecosystem—and the hidden tuning parameter is in time: it requires the evolutionary timescale to be vastly larger than the ecological timescale (their ratio tending to infinity), analogous to scale interference in space for magnets.
“this could be an outcome of the interplay between ecology and evolution, and the idea being that when organisms change their environment by their activity.”
- Phylogenetic trees built from genome sequencing—where branch lengths reflect numbers of DNA mutations—turn out to be mathematically fractal: when you characterize the branching asymmetry (how many clades emanate from a species relative to total connected branches) the pattern follows scale-invariant behavior, a result discovered about a decade ago.
“those trees turn out to be mathematically fractal. If you draw them out they look fractal, and if you study them topologically... the characterisation of that asymmetry of the trees turned out to follow a scale invariant behavior”
- Before about 3.8 billion years ago, evolution was not tree-like: organisms could collectively swap genes with unrelated organisms (horizontal gene transfer), and this network effect—rather than mere vertical copying with variation—produces hugely accelerated evolution rates that drive the system to a unique and optimal genetic code; only afterward did evolution become tree-like (vertical), and the process self-terminates because once a good ribosome exists, foreign genes are unlikely to improve it.
“before 3.8 billion years ago, actually, organisms could swap genes, collectively, not just transmit them to their offspring... that process is a very, very fast way for an evolving system to discover solutions to problems... because it's now utilizing a network effect rather than just making a copy of itself with small variations.”
- Although Earth is about 4.6 billion years old and was inhospitable early on, the last universal common ancestor existed by about 3.8 billion years ago, and within less than a billion years life evolved essentially the complexity of the modern cell; after that, despite producing dinosaurs and humans, there were no qualitatively very significant changes in the architecture of the cell or the information it transmits—a paradox of rapid early evolution followed by relative stasis.
“within less than a billion years, you essentially evolved the complexity, more or less the complexity of the modern cell. And then after that evolution didn't really do much... from a point of view of the architecture of the cell... there wasn't really very, very significant changes from the qualitative point of view.”
Universality means vastly different systems exhibit identical mathematical properties at transitions, determined by symmetry and dimension alone.
- Solving the phase transition problem did not just affect material science and solid state physics; it changed how physicists think about high energy particle physics, cosmology, and quantum mechanics, because it revealed that fundamental assumptions built into prior thinking were wrong.
“it completely changed the way that scientists and physicists think about modelling nature. It revolutionized not just the study of phase transitions... it changed the way we think about energy particle physics, cosmology, high energy physics, quantum mechanics, everything changed”
- One of the astounding insights from phase transition physics is universality: systems that are totally disconnected and apparently different—a magnet versus a liquid-gas transition, or water percolating through coffee versus a fluid becoming turbulent—have exactly the same mathematical properties at their transition point, so to predict critical phenomena a physicist only needs the symmetry group and spatial dimension of the system.
“systems which are totally disconnected from each other, they seem apparently different, nevertheless have exactly the same mathematical properties.”
- Universality is a double-edged sword: it gives huge explanatory power because you can predict a phenomenon's critical behavior without knowing microscopic details (only symmetry and dimension), but precisely because so many different microscopic processes give the same pattern, you cannot work backwards from an observed pattern to the underlying process—seeing a pattern does not mean you understand the process that generated it.
“universality is a double-edged sword. On the one hand, it means you can explain a phenomenon without having to know very much about it... but then if you want to ask, well, what's really going on at the microscopic level, why is it like that, you don't really... You don't really know.”
Life is self-organizing matter with evolvability as its defining feature, helping planets dissipate free energy toward equilibrium.
- Goldenfeld's view is that life is a generic, inevitable consequence of the laws of physics—matter can self-organize out of equilibrium hierarchically into structures that are evolvable—and that it is evolvability (the capacity for open-ended growth of complexity), not reproduction or metabolism, that defines living systems.
“my prejudice is that life is a generic consequence of the laws of physics, and that matter can self-organize out of equilibrium hierarchically into structures that are then evolvable. And it's the evolvability that in my view defines living systems.”
- Goldenfeld proposes that the purpose of life is to help planets approach equilibrium: life takes the chemical-potential (free-energy) differences present in a far-from-equilibrium young planet, short-circuits the slow chemical reactions that would dissipate them, and uses that free energy to power living systems—competing with ordinary chemistry but gaining an advantage by exchanging information through genes and evolutionary adaptation.
“what life is, is a process by which a planet approaches equilibrium... the purpose of life is to help planets approach equilibrium, because what life does is it takes chemical potential differences in the environment.”
The genetic code evolved to near-optimal error-minimization through horizontal gene transfer and dynamic optimization, not frozen accident.
- Carl Woese was the first to empirically answer the question of the evolutionary history of all life using molecular sequencing (the forerunner of genomics), discovering that life divides into three domains, including a third domain of round blobby organisms that look like bacteria but have a completely different evolutionary history (now called archaea).
“my colleague and friend Carl Woese was really the first person come up with a way to answer this question empirically... that was the discovery of the... What's called the third domain of life.”
- Francis Crick argued in 1968 that the genetic code is a 'frozen accident' that cannot evolve—because changing the code mid-history would garble all proteins and kill the organism—but Goldenfeld and colleagues showed this was wrong: the code did evolve dynamically, reaching a unique, optimal, and fast-evolving state.
“Francis Crick realized this in 1968, and so basically the idea is the genetic code is a frozen accident, it can't evolve... Okay, so we know that that was wrong. And so how did it evolve?”
- The genetic code (the map from 64 nucleotide triplets/codons to 20 amino acids) is structured so that the most frequent translation errors yield an amino acid chemically similar to the intended one, so the protein still has a chance of folding correctly; this near-optimal error-minimizing structure—what an intelligent designer would build—was instead produced by evolution, as discovered by Haig and Hurst in the 1990s.
“the genetic code has this structure, it is pretty much, not quite, but pretty much optimal in the sense of having a structure so it will minimize the most frequent errors that you're going to get in this gene expression process.”
- X-ray structural evidence shows the ribosome evolved as successive onion-like layers, with the core peptidyl transferase center performing translation very crappily; this crappy, malleable translation at the core is precisely what made the ribosome able to evolve, because a highly malleable process can be improved over time before accuracy increases in later layers.
“in the middle you've got a thing called the peptidal transferase center, which does translation, but it does it really, really crappily... And this crappy translational procedure that's at the core of the ribosome is essential for its ability to evolve.”
Observed patterns alone cannot distinguish underlying mechanism from statistical artifact; neutral and detailed models yield identical distributions.
- The famous claim that sand piles self-organize to criticality without tuning was misleading: Sid Nagel's experiments showed real sand piles are not scale-invariant because the models had neglected grain inertia, and ignoring inertia is equivalent to setting it to zero, which is equivalent to tuning the system to a critical point—so apparent self-organized criticality can be an artifact of modeling simplifications.
“in making the models, people had neglected the inertia of the grains... by making the mathematical models, you actually had inadvertently tuned it to a critical point by saying, well, I don't need to worry about inertia, so I'll just set that at equal to zero. Ignoring something is the same as setting it equal to zero.”
- For the well-documented Barro Colorado Island ecosystem, a 'neutral model' that treats species as non-interacting and functionally equivalent (analogous to the ideal gas) produces a species abundance distribution virtually identical to one derived from detailed biological understanding of how organisms compete for light and nutrients, meaning the pattern alone cannot distinguish whether community structure reflects pure statistics or real biological process—the true answer being a mixture of both.
“now you look at what you predict from that, and you look at what you predict from actually understanding the biology of the organisms, and you get two distributions and they're virtually exactly the same.”
- Complex systems such as the brain exhibit the same epiphenomena associated with critical phenomena; specifically, neuroscience has established that there are neural avalanches in the brain that behave essentially like critical phenomena.
“in neuroscience we know now that there are neural avalanches in the brain which behave pretty much like critical phenomena and so on and so forth”
- Goldenfeld's theory of scientific impact is that impact equals the ratio of what you can do (numerator, bounded above by luck, funding, and intelligence) to what everyone else does (denominator); since the numerator is bounded, the strategy is to minimize the denominator by working on problems no one else is working on, exploiting 'intellectual arbitrages' where knowledge from one field can be transferred to another.
“my theory of impact in science is that your impact is the ratio of what you can do, the numerator, to what everybody else does, in the denominator... What I can do is make the denominator as small as possible. And so, I try to work on problems that other people aren't working on.”
- Despite being so fundamentally important, condensed matter physics has no popular trade book describing what the field is, and most people do not even know the field exists.
“nobody has written the trade book that describes what condensed matter physics is. And nobody even knows the field exists, yet it's so fundamentally important.”
Renormalization in quantum electrodynamics and phase transitions both handle infinities by resolving scale interference mathematically.
- Feynman, Schwinger, and Tomonaga handled the infinities arising in quantum electrodynamics calculations by redefining unobservable quantities like the electron mass so that the infinities were absorbed, yielding finite results; this 'renormalizable' procedure worked but was treated with suspicion—even by Feynman who called it mathematical legerdemain—because it lacked a clear physical understanding of why it worked.
“Feynman himself didn't believe this was a good strategy, he said many times it's sort of, it's some legerdemain, it's a mathematical trick, and why the heck this works I have no idea”