Two transitions, one bicritical point
A glass is a liquid that forgot to crystallize. Cool a liquid fast enough, and it never finds its crystal. Instead, it thickens, slows, and eventually stops flowing — trapped in a disordered state that looks like a solid but has the structure of a liquid. For a century, the nature of this transition has been one of the deepest open problems in condensed matter physicsThe branch of physics that studies how large numbers of particles behave collectively — solids, liquids, magnets, superconductors. Where most phase transition research happens..
The reason it stayed open is that everyone was looking for one transition. There are two.
In 2025, Pettini and colleagues published a study that resolved the glass transition into two distinct events, each with its own thermodynamic signature and its own topological fingerprint in the energy landscapeAn abstract surface where each point represents a possible arrangement of all the molecules, and the height represents the system’s energy. Valleys are stable configurations; ridges are barriers between them..1
The first is a dynamic transition. At a temperature physicists call TMCT (the mode-coupling temperatureThe temperature at which the mathematical theory of mode-coupling — which tracks how different molecular vibration patterns interact — predicts that flow should stop. Named after the theory, not the physical mechanism.), the liquid’s molecular rearrangements slow catastrophically. Molecules can still vibrate in their cages, but they can no longer escape them. The system loses ergodicityThe ability of a system to eventually visit all its possible states. An ergodic system forgets its initial conditions; a non-ergodic system gets trapped. Like a marble in a rugged landscape that can no longer roll out of the valley it happens to be in. — its ability to explore all of its possible configurations. This transition is weakly first-orderA first-order transition is an abrupt jump between states (like water boiling). “Weakly” means the jump is tiny — barely detectable but still discontinuous, not a smooth glide.: there is a small but measurable discontinuity, like a tiny jump.
The second is a thermodynamic transition. At a lower temperature called TK (the Kauzmann temperatureThe temperature at which a supercooled liquid would have the same entropy as a crystal — the “entropy crisis.” Named after Walter Kauzmann, who identified this paradox in 1948.), the number of distinct configurations available to the glass drops toward zero. The configurational entropyThe number of distinct ways molecules could arrange themselves while still forming a glass. It’s a measure of structural possibilities. When it reaches zero, the system has run out of options. — a measure of how many ways the molecules could rearrange — vanishes. This transition is second-orderA smooth, continuous transition — no abrupt jump. Properties change gradually, and warning signs (like critical slowing down) grow strongest. Think: a system that degrades gracefully, giving you time to detect and respond.: continuous, with a diverging correlation lengthHow far apart two molecules can be and still influence each other. Near a second-order transition, this length grows without bound — distant parts of the system start behaving in lockstep..
Each transition has its own order parameterA quantity that distinguishes between phases. It’s zero on one side of the transition and nonzero on the other. For a magnet, it’s magnetization. For glass, each transition has its own: one measures memory, the other measures structural options.. Each produces its own peak in the specific heatHow much energy you need to raise the temperature by one degree. Near a phase transition, the system absorbs extra energy as it reorganizes, producing a peak. Two peaks = two transitions.. Each leaves its own topological mark on the geometry of the energy landscape. They are not two views of one event. They are two events.
Here is where the story turns. Glass-forming liquids come in two broad varieties. Strong glass formers — like silica (SiO2) and boron trioxide (B2O3) — have their two transitions close together. Fragile glass formers — like ortho-terphenyl (OTP) and toluene — have them far apart. Physicists have measured this distance for decades. It is called fragilityA measure of how dramatically a liquid’s viscosity changes near its glass transition temperature. Strong glass formers change gradually (Arrhenius); fragile ones change dramatically (super-Arrhenius). It’s one of the most useful empirical numbers in glass science., and it is one of the most useful empirical quantities in glass science.2
But what if fragility is not just a useful number? What if the distance between the two transitions is the distance from a bicritical pointA point in parameter space where two independent phase transitions converge and merge. Their order parameters couple, and the system enters a new universality class — fundamentally different physics from either transition alone. — a special location in parameter spaceThe abstract space of all possible combinations of a system’s control variables (temperature, pressure, density, etc.). Each point represents a different set of conditions. Phase diagrams are maps of parameter space. where the two transitions merge?
At a bicritical point, two independent phase transitionsQualitative changes in a system’s behavior — not just “more of the same” but a fundamentally different operating mode. Water boiling, a magnet demagnetizing, a network congesting. converge. Their order parameters couple. The universality classA group of physically different systems that share the same mathematical behavior near their tipping points. A magnet and a fluid can be in the same class — same equations, completely different stuff. Multicritical transitions belong to different classes than single-critical ones. changes. New physics emerges from the coupling that neither transition produces alone.
The strong-glass limit — where TMCT equals TK — is exactly this point. Strong glass formers are near the bicritical point. Fragile glass formers are far from it. Fragility is distance from bicriticality.◆ our hypothesisReframing fragility as distance from a bicritical point is an original proposal of this research program, not established glass-physics consensus. The two transitions (Pettini et al. 2025) are published; the bicritical interpretation is ours and remains to be tested.
This reframing turns an empirical classification (strong vs. fragile) into a structural prediction: systems near the bicritical point should show coupled critical behavior with a new universality class — and systems far from it should show two independent transitions with their individual universality classes. The classification is no longer a label. It is a measurement of distance from a codimension-two pointA special point that requires tuning two independent parameters simultaneously to reach. You need to adjust both temperature and (say) density to land exactly on it. The “two” refers to two constraints, not two dimensions..
Drag the slider to change how fragile the glass former is. Watch the two specific heat peaks — and see what happens when they meet.
Below TK, a third transition appears. Called the Gardner transitionA transition within the glass phase where each metastable state shatters into a fractal hierarchy of sub-states. Named after Elizabeth Gardner (1957–1988). It connects glass physics to the physics of tightly packed particles., it splits each glass state into a fractal hierarchy of sub-states — marginally stableBalanced on the edge of instability. The smallest push can trigger a rearrangement. Not unstable (not falling apart), not stable (not firmly locked) — right at the boundary. configurations separated by vanishing barriers. The Gardner transition produces the scaling lawsMathematical relationships describing how properties change near a tipping point. They follow power laws — doubling the distance from the critical point changes the observable by a fixed ratio, not a fixed amount. that govern jammingThe physics of disordered particles packed as tightly as possible — like sand grains in a jar. Jammed systems are rigid but structurally disordered, sharing deep mathematical connections with glasses.: the physics of tightly packed disordered particles.3
This creates a chain that connects the glass transition to an entirely different physical system. The chain runs: glass bicriticality → Gardner transition → jamming scalings → protein core packing. We follow this chain in Tutorial 04.
The glass transition is the simplest system in this series: two transitions, one coupling parameter, one codimension-two point. It introduces the core idea — that the coupling between transitions matters, not just the transitions themselves — in the most stripped-down setting possible.
Dynamic transition (TMCT): The order parameter is the Edwards-Anderson overlapA measure of how “frozen” a system is — how similar its configuration at time t is to its configuration much later. Named after the spin glass pioneers Sam Edwards and Philip Anderson. qEA, defined as the long-time limit of the density autocorrelation functionHow similar the molecular arrangement is now compared to some time in the past. High autocorrelation means the structure persists; low means the molecules have reshuffled.. Below TMCT, qEA > 0: the system remembers its configuration. Above, qEA = 0: full ergodicity. In mean-field theoryAn approximation that replaces all the detailed interactions between particles with a single average effect. Exact in infinite dimensions; approximate but often qualitatively correct in real systems. (infinite dimensions), this transition is sharp. In finite dimensions, it becomes a crossover, but the underlying singularity shapes the dynamics.
Thermodynamic transition (TK): The order parameter is the configurational entropy Sc, measuring the logarithm of the number of distinct amorphous packings. In the Random First-Order Transition (RFOT)The leading theoretical framework for the glass transition, combining ideas from mean-field spin glass theory with real-space nucleation. Predicts both TMCT and TK as distinct physical events. framework,4 Sc → 0 at TK: the system runs out of configurations. This is a genuine second-order transition with a diverging point-to-set correlation lengthHow large a region you must freeze before the interior becomes trapped in a single configuration. Near TK, this length diverges — the system develops long-range amorphous order..
Pettini et al. detected both transitions through independent topological signatures in the energy landscape geometry: inflection in the mean curvature of energy level sets at TMCT, and a peak in scalar curvature variance at TK. The transitions leave distinct geometric fingerprints, confirming they are independent events with different topological character.
The TMCT / TK ratio correlates with the Angell fragility indexA single number that captures how a glass former’s viscosity depends on temperature. Invented by C. A. Angell in 1995. Small m = strong (gradual change). Large m = fragile (dramatic change). m.2 For strong glass formers (SiO2): TMCT / TK ≈ 1.1–1.2, m ≈ 20. For fragile glass formers (OTP): TMCT / TK ≈ 1.5–2.0, m ≈ 80–200.
In the bicritical framework, the ratio TMCT / TK − 1 measures distance from the codimension-two point. A Ginzburg-Landau analysisA general mathematical framework for describing phase transitions in terms of a free energy function and order parameters. Named after Vitaly Ginzburg and Lev Landau. The standard starting point for analyzing how transitions interact. with two coupled order parameters (qEA and Sc) predicts O(2)/XY universalityA specific universality class with continuous rotational symmetry in two components. The same mathematical structure governs superfluid helium, thin-film superconductors, and certain magnetic systems. at the bicritical point — the same symmetry class as the Kosterlitz-Thouless transitionA topological phase transition where vortex-antivortex pairs unbind. Discovered by Kosterlitz and Thouless (Nobel Prize, 2016). It’s driven by topological defects rather than symmetry breaking.. If confirmed, the bicritical point is where topological defect physics enters the glass problem.
Preliminary correlation: r = 0.56, p = 0.09 (N = 10 glass formers).◆ our resultAn original, not-yet-peer-reviewed computation from this research program — distinct from established consensus. At N=10, p=0.09, this is a hint, not a finding. Suggestive but not significant. Needs careful TMCT / TK compilation for 20+ systems with well-characterized fragility values.
The Gardner transition occurs within the glass phase, below TK. Predicted by mean-field theory for all glass formers, it marks the onset of marginal stability — each metastable glass basin shatters into a fractal hierarchy of sub-basins. The signature: full replica symmetry breakingA mathematical phenomenon where copies (“replicas”) of the same system, analyzed together, reveal a hidden hierarchical structure of states within states. Developed by Giorgio Parisi (Nobel Prize, 2021). of the Parisi type.5
In 2024, Charbonneau and colleagues demonstrated that the Gardner transition produces the same scaling exponentsNumbers that describe exactly how properties change as a system approaches a tipping point — how fast correlations grow, how order emerges. Different types of transitions have different exponents. They’re the mathematical fingerprint of the transition. that govern jamming — the physics of maximally dense random packings.3 The connection is: glass → Gardner → jamming scalings → protein packing fraction.
If glass bicriticality is real, then strong glass formers are where the Gardner transition should be most clearly resolved (closest to bicritical = strongest coupling). This is the opposite of current research focus, which studies Gardner in fragile hard-sphere-like systems.
Claim: The two transitions identified by Pettini et al. interact at a codimension-two point in parameter space, making the glass transition a bicritical system. The universality class at the bicritical point is O(2)/XY.
Status: Theoretical framework complete. Five-point bicriticality framework applied. Ginzburg-Landau free energy with two coupled order parameters constructed. Predicts crossover exponent φT ≠ 1 at the bicritical point.
Test: Simulate a binary Lennard-Jones mixture at multiple densities. If the two specific heat peaks merge at some critical density ρc with a universality class change, bicriticality is confirmed. If they remain independent at all densities, bicriticality is ruled out.
Claim: Fragility m scales as (TMCT / TK − 1)γ for some exponent γ. If true, fragility is not an independent material property — it is a derived quantity measuring distance from bicriticality.
Status: Preliminary. r = 0.56, p = 0.09 (N = 10 glass formers). Direction is correct (higher ratio = higher fragility) but sample is too small for significance.
Next step: Compile TMCT / TK ratios for 20+ glass formers with well-characterized Angell fragility values. If the correlation strengthens, it supports the bicritical interpretation. If it vanishes, fragility and the TMCT / TK ratio are measuring different things.
Connected through the Gardner/jamming chain. The Gardner transition below TK produces the scaling laws that govern jamming — which in turn sets the universal packing fraction (φc ≈ 0.55) found in protein cores. The glass bicritical point sits at the top of a causal chain that ends in the geometry of living molecules.
Two order parameters (qEA and Sc) on the same node set — molecular configurations in phase space. A relational gradient operator extracts differences for each order parameter independently along declared edges.
If the transitions are decoupled, antisymmetric circulation across components should produce zero cross-component contribution — no coupling signal. As the system approaches bicriticality (the strong-glass limit where TMCT / TK → 1), the physical coupling between transitions grows.
The question: Does the cross-component circulation signal grow monotonically as TMCT / TK → 1? If so, it measures what fragility measures — distance from the codimension-two point — but detects it through relational field structure rather than through viscosity curves.
Pettini, M., Gori, M., Clementi, C., & Pettini, G. (2025). The glass transition: A topological perspective. Entropy, 27(3), 258.
Berthier, L. & Biroli, G. (2011). Theoretical perspective on the glass transition and amorphous materials. Rev. Mod. Phys., 83, 587–645.
Angell, C. A. (1995). Formation of glasses from liquids and biopolymers. Science, 267, 1924–1935.
Charbonneau, P. et al. (2024). Gardner transition and jamming scalings. arXiv:2410.23797.
Parisi, G. & Zamponi, F. (2010). Mean-field theory of hard sphere glasses and jamming. Rev. Mod. Phys., 82, 789–845.
Kirkpatrick, T. R. & Thirumalai, D. (2015). Colloquium: Random first order transition theory concepts in biology and physics. Rev. Mod. Phys., 87, 183.