Thermal hysteresis (TH) and ice recrystallisation inhibition (IRI) are two different things an antifreeze molecule can do, and they do not come as a package. Thermal hysteresis lowers the temperature at which ice grows below the melting point, opening a gap where existing ice is frozen but static. Ice recrystallisation inhibition stops small ice crystals from merging into large damaging ones over time. A molecule can be strong at one and weak at the other — and for snow and cold-chain applications, IRI is the property that matters.
The distinction is not academic hair-splitting. It determines which molecules are worth designing, how you measure them, and what you can honestly claim. This piece separates the two mechanisms, shows why the field increasingly treats them as decoupled, and explains why DeepSnow's synthetic-antifreeze work is specified on IRI rather than TH.
Key takeaways
- Thermal hysteresis (TH) = the gap between melting and freezing points an ice-binding molecule creates; it holds ice static within a temperature window.
- Ice recrystallisation inhibition (IRI) = suppression of the coarsening by which small ice crystals grow into large ones; measured as mean-grain-size reduction.
- The two are decoupled: strong IRI can occur with little or no TH (Meister group, 2023, PNAS).
- For snowmaking and cold-chain, IRI is the load-bearing property — it governs crystal structure and durability, not a frozen-but-static temperature window.
- Synthetic antifreeze glycoproteins reach 89–97% mean-grain-size reduction depending on chain length (Deleray et al. 2024).
What is thermal hysteresis?
Thermal hysteresis is the difference between the temperature at which ice melts and the (lower) temperature at which it starts to grow, created when an antifreeze molecule binds to the ice surface. By adsorbing onto specific ice faces, the molecule locally pins the ice front; growth cannot resume until the temperature drops far enough to overcome that pinning. The size of that gap — often a fraction of a degree to a couple of degrees — is the thermal hysteresis.
TH is what the classic Antarctic-fish antifreeze proteins are famous for. Notothenioid fish survive in −1.9 °C seawater because their antifreeze glycoproteins produce enough TH to keep the small ice crystals in their body fluids from growing. Within the hysteresis gap, ice is present but frozen in place. The measurement is direct: you observe a single ice crystal in solution and record how far you can cool it before it grows. Crucially, TH is about arresting growth within a temperature window, not about what happens to crystal structure over time.
What is ice recrystallisation inhibition?
IRI is the suppression of recrystallisation — the process by which, in an already-frozen sample, large ice crystals grow at the expense of small ones. Even below freezing, ice is not static: smaller crystals have higher surface energy and dissolve, feeding larger crystals in a coarsening process. That coarsening is what damages frozen cells, degrades frozen food texture, and turns fine made-snow into glassy grains. An IRI-active molecule slows or halts it.
The physics behind recrystallisation is Ostwald ripening, the same thermodynamic drive we describe for snow in Ostwald ripening and why snow crystals coarsen. IRI is measured differently from TH: a "splat" assay freezes a thin film, holds it at a sub-zero temperature, and measures how much the mean grain size grows over time versus a control. Less growth means more inhibition. The result is reported as mean-grain-size (MGS) reduction, and it is the metric that matters when the goal is a fine, stable crystal structure rather than a static temperature window. Our visual primer covers the assay in what is ice recrystallisation inhibition.
Why are TH and IRI different superpowers?
Because they arise from different modes of ice binding, and a molecule can have one without the other. This is the key modern finding: work from the Meister group (2023, PNAS) shows that reversible, dynamic ice binding can deliver strong IRI with little or no thermal hysteresis, whereas the strong, near-irreversible binding that produces large TH is a distinct regime. Potent IRI does not require — and often does not come with — large TH.
| Property | Thermal hysteresis (TH) | Ice recrystallisation inhibition (IRI) | |---|---|---| | What it does | Creates a melt–freeze temperature gap; holds ice static | Stops small crystals coarsening into large ones | | Binding mode | Strong, near-irreversible ice adsorption | Reversible, dynamic ice-surface interaction | | Measured by | Single-crystal growth vs cooling | Splat assay; mean-grain-size reduction over time | | Governs | A temperature window | Crystal structure and durability over time | | Matters most for | Sub-freezing organism survival | Snow quality, frozen food, cryopreservation |
The practical upshot is that you can engineer a molecule for potent IRI without chasing TH — which is fortunate, because TH-heavy molecules bring their own problems, including ice-shaping effects that can be damaging.
Which property matters for snow and cold-chain?
IRI, by a wide margin. In snowmaking, the objective is a fine, dense, durable crystal structure that grooms well and resists coarsening — exactly what IRI protects — not a sub-zero temperature window in which ice sits frozen. In frozen food and cryopreservation, the damage that IRI prevents (large crystals rupturing cell membranes and degrading texture) is the whole problem; a thermal hysteresis gap does little to address it. This is why the reviews of engineered ice-control compounds treat IRI as the workhorse property for these applications.
A necessary honesty note: IRI is a primary, measurable mechanism, but it is not the entire cryoprotection story. Antifreeze (glyco)proteins also interact with cell membranes, and full cryoprotection in living systems involves more than recrystallisation control. We do not claim IRI alone explains cryopreservation — only that, for the snow and cold-chain jobs we target, it is the decisive, directly measurable property.
How does DeepSnow's synthetic-antifreeze work use this?
By designing candidates specified on IRI, using cheap, synthesisable backbones rather than harvested natural proteins. The DS-100 line is built on alanine/glutamate polypeptides made by NCA polymerisation — chemistry documented in Deleray et al. (2024, Chem. Mater.), which reports 89% MGS reduction for a 28-mer, 94% for a 57-mer, and 97% for a 170-mer, and in McPartlon et al. (2025, Adv. Mater.) on ultra-economical antifreeze polypeptides. Alanine's role as the smallest known IRI-active unit is covered in why alanine.
Specifying on IRI has two consequences. First, the target metric is MGS reduction from a splat assay, a clean and reproducible number, rather than a TH value that would push toward a different and less useful class of molecule. Second, an AI-assisted design loop can rank candidate sequences on predicted IRI and let the wet lab synthesise and assay the best — the tooling ranks and proposes, the bench confirms; it does not model native glycosylation end-to-end. The pipeline is described in DS-100: how AI-designed synthetic antifreeze glycoproteins work. Note that the foundational patent in this space, WO2024258965A2, is assigned to the University of Utah Research Foundation (inventors Kramer, Deleray); DS-100 variants are designed to sit outside those claims.
Getting the vocabulary right
The reason to insist on the distinction is that TH and IRI get used interchangeably in loose writing, and the confusion leads to bad design and overclaiming. A molecule sold on its "antifreeze" credentials because it produces thermal hysteresis may do nothing useful for snow durability or frozen-food texture. The right question for a cold-chain or snow application is always: what is the mean-grain-size reduction? That is the IRI number, and it is the one worth designing toward.
Talk to us
DeepSnow designs ice-control chemistry specified on the property that actually governs snow quality and cold-chain stability — ice recrystallisation inhibition, measured as mean-grain-size reduction. To follow the DS-series R&D or discuss the science, join the waitlist or get in touch.
DS-100 and the DS-series are in R&D and are pre-commercial; MGS figures cited are from published synthetic-AFGP literature, not DeepSnow product data. AI-design tooling ranks and proposes candidates that the wet lab synthesises and assays.