Glossary

A single answer from Phases mixes two vocabularies. One is metallurgy you may already know: austenite, dilatation, Ms, nugget diameter. The other belongs to the product: sessions, runs, capabilities, parameter cards. When a phrase in a result looks like it should mean something and does not, it is almost certainly in one of the two lists below.

Terms are alphabetical within each group. Product terms link to the page that covers them in full.

Metallurgical terms#

Ae1 — On slow cooling, the last austenite transforms to pearlite at Ae1. What Phases plots under that name is the material card's pearlite critical temperature — a base value plus the card's per-element composition coefficients — not an equilibrium-diagram value, and not the separate equilibrium-fraction breakpoint the same card also carries. It does respond to a chemistry you apply: on a DC04 card, raising manganese from 0.2 to 2.0 wt% moves it from about 659 °C to 640 °C, and 3 wt% chromium raises it to about 707 °C. The carbon coefficient is zero on the cards tested, so changing carbon alone leaves it where it is. See known limits.

Ae3 — The temperature above which the steel is fully austenitic; on cooling, ferrite starts to precipitate below it. Like Ae1, what Phases plots is the card's ferrite kinetic critical temperature rather than an equilibrium value — but it is composition-corrected, so a chemistry you apply does move it. Raising carbon from 0.05 to 0.45 wt% on a DP600 card drops it from about 819 °C to 738 °C. It remains a card prediction, not a measured Ae3.

ASTM grain size — The standard number describing prior austenite grain size; a larger number means finer grains. It appears twice in Phases: as a kinetics factor authored on the material card, and as a required input for flow-stress prediction, which will not run without an explicit grain-size number. See the capability matrix.

Austenite — The face-centred-cubic high-temperature phase of steel. Every diffusional and martensitic transformation Phases models starts from austenite, which is why CCT sweeps impose a 100% austenite starting state.

Bainite — An intermediate transformation product formed between the pearlite and martensite ranges: fine ferrite plus carbides, harder than pearlite and tougher than untempered martensite. In the model bainite is allowed to consume all remaining austenite, which is one reason it competes strongly with pearlite at moderate cooling rates.

Bs (bainite start) — The temperature at which bainite begins to form on cooling. Phases derives it from the card's bainite critical-temperature base plus the card's composition coefficients, so it varies by grade and with an applied chemistry: across the base cards packaged with the solver it runs from about 547 °C (EN13261) to 650 °C (20MnB8), and raising carbon from 0.05 to 0.45 wt% on a DP600 card moves it from 596 °C to 572 °C. No published Bs correlation is used, and the bainite rate constants are shared across most non-boron grades, so a per-grade Bs does not make the kinetics grade-calibrated. See known limits.

CCT diagram — Continuous Cooling Transformation. Maps which microstructural phases form as steel cools continuously at a given rate. In Phases this is a sweep of several cooling rates plotted as transformation start and finish markers on a log-time versus temperature chart. See CCT and TTT.

Critical cooling rate — The slowest cooling rate that still suppresses diffusional transformation and gives the microstructure you want, usually full martensite. Phases does not solve for it directly; you bracket it by running a CCT sweep and reading where the ferrite, pearlite and bainite markers disappear. See find the critical cooling rate.

Dilatation — Length change during heating and cooling. Because phases differ in density, a transformation shows up as a departure from the straight thermal-expansion line. Phases computes it as a phase-weighted thermal term plus the volume change of the transformation itself.

Dilatometry — The experimental technique of measuring dilatation to detect transformation temperatures. The V-Gleeble CCT mode reproduces the dilatometry trace itself, which is a different output from the CCT diagram's phase-boundary map.

dt8/5 (Δt8/5) — The cooling time in seconds from 800 °C to 500 °C, the standard way welders characterise a weld thermal cycle. It is a direct input to the V-Gleeble HAZ mode: a smaller dt8/5 means a faster, harder cycle.

Faying surface — The sheet-to-sheet interface in a spot weld, where the nugget nucleates. It is one of the three contact interfaces the RSW solver models, along with the two electrode-to-sheet contacts.

Ferrite — The body-centred-cubic low-temperature phase, soft and formable. It dominates the room-temperature microstructure of the mild and HSLA grades in the catalog.

Flow stress — The true stress needed to keep a material deforming plastically at a given strain, strain rate and temperature. Phases has two separate flow-stress systems that are not interchangeable: per-phase flow-stress prediction from a material card, and the bulk flow-stress surface the V-Gleeble deformation modes require. See the capability matrix.

Gleeble — A physical thermomechanical simulator that subjects small specimens to precisely controlled heating, forces and cooling. "Virtual Gleeble" reproduces these tests numerically. See Virtual Gleeble.

HAZ (heat-affected zone) — The region beside a weld that was never molten but was hot enough to change microstructure. In the RSW post-processing Phases classifies each element by its peak temperature into nugget, upper HAZ, lower HAZ and base metal, and reports a diameter and phase composition for each zone; predicted hardness is added only when the model has cooled far enough by the end of the simulated time. See RSW validation.

HV (Vickers hardness) — The hardness scale Phases reports. The plots label it HV30, meaning Vickers hardness under a 30 kgf load. Every HV number Phases produces is a bulk rule-of-mixtures value computed from the phase fractions, the chemistry and the cooling rate — it is the hardness of the mixture, never the intrinsic hardness of one phase. See hardness validation.

Martensite — The hard, supersaturated phase formed by diffusionless shear when austenite is cooled below Ms. Phases models it with an athermal Koistinen–Marburger relation rather than a diffusional rate law, so the amount formed depends on how far below Ms you go, not on how long you hold there.

Ms (martensite start) — The temperature at which martensite begins to form on cooling. Phases derives it from the material card's martensite critical-temperature base plus the card's per-element composition coefficients, so it is per-grade and moves with a chemistry you apply. Across the base cards packaged with the solver the effective values span about 309 °C (EN13261) to 532 °C (DC06). The authored base itself is shared — 550 °C on the fifteen non-boron cards, 561 °C on the five boron cards — and the composition term does all the per-grade work on top of it. What Ms is not is an independently calibrated measurement: the coefficients are authored on the card and have not been validated grade by grade. See known limits.

Nugget — The solidified molten zone that forms the actual joint in resistance spot welding. Its diameter is the primary quality metric, usually judged against a minimum of five times the square root of the thinner sheet thickness in millimetres. See resistance spot welding.

Pearlite — The lamellar ferrite-plus-cementite product of slow cooling through the eutectoid. Phases tracks it as its own phase, though the model cannot tune pearlite kinetics independently of bainite — a documented limitation. See known limits.

Phase fractions — The proportions of austenite, ferrite, pearlite, bainite and martensite in the microstructure, which set the mechanical properties. Phases reports them as fractions between 0 and 1 or as percentages, and they are the primary output of every thermal simulation.

Press-hardening / boron steel — Grades such as 22MnB5, austenitised then quenched in the die to form martensite, giving very high strength. Common in automotive safety structures. Boron's grain-boundary effect is not modelled explicitly; the boron grades work because one of them was the original calibration basis.

Quench rate versus quench medium — A cooling rate in °C/s is a number, not a model of water, oil or air. Phases will not relabel a rate as a quench medium unless you supply that mapping yourself, because the solver has no bath, no agitation and no part geometry.

RSW — Resistance Spot Welding. Joining sheet metal by passing current through electrodes clamped on the stack; resistive heating melts a nugget at the faying surface. See resistance spot welding.

Satoh test — A specimen is heated and then cooled while axially restrained, so thermal contraction that cannot happen as strain appears as stress instead. It is used to characterise weld residual stress, and it is one of the V-Gleeble test modes.

TTT diagram — Time-Temperature-Transformation. As above, but for isothermal holds rather than continuous cooling. The transformation-diagram chat tool does not simulate TTT: the request is answered with an explicit unsupported status. Calculated TTT curves are available outside chat, in the Material Base transformation panel, derived from the card's own kinetics rather than from a simulated hold. See CCT and TTT.

Werkstoffnummer (material number) — The DIN/EN numeric designation for a steel, such as 1.2344 for X40CrMoV5-1 or 1.4404 for 316L. 398 catalog rows are labelled as a bare Werkstoffnummer, so typing one usually works — but it finds the row named that way, not a cross-reference from another naming system. 1.2344 and X40CrMoV51 are two separate rows with identical chemistry and different sources. See finding your steel.

Phases product terms#

Alloy group — The alloy-system label on a catalog material, such as Ni-Cr-Mo, Cr-Mo-V, C-Mn or Mn-B. The catalog uses 68 of them, and they cut across families: the largest are Ni-Cr-Mo (269 grades), Cr-Mo-V (244) and Cr-Mo (170). Ask for a group by name when you want the chemistry family rather than the application. See families and data sources.

Analysis recipe — A named, immutable calculation the agent may run over runs you own, instead of writing fresh code. The allowlist is short and deliberate: summary statistics over one run, an owned run comparison with endpoint deltas and residual statistics, and an aligned overlay of two compatible runs. A recipe refuses rather than guessing when two runs carry mismatched simulators, quantities, units or axes. See compare and analyse runs.

Artifact — Any file a run produced and stored: plots, animated GIFs, CSV data, the material card the run consumed, and the resolution manifest recording how that card was chosen. Artifacts live under the run's own storage prefix and are handed back as signed links, which expire — seven days for the plot links a simulation tool returns. See sessions, runs and artifacts.

Background task and replay — A long simulation keeps running on the server after you close the tab or lose your connection. When you come back, the client reconnects to the same task and replays the events it missed. That live buffer lives in the backend process and keeps a finished task for about ten minutes. A durable run is recorded separately and is not bound by that window: its finished result card is rebuilt in the conversation from the stored event log whenever you reopen the session. See long-running simulations.

Base template — The verified catalog material card whose structure a custom chemistry is rendered onto. Your composition supplies the numbers; the template supplies the curve structure, the kinetics parameters and the schema the solver can read. A template that cannot represent every element you specified is rejected and the next candidate is tried. See your own chemistry.

Capability — What a simulator needs from a material before it will run. There are four: base card, transformation kinetics, phase flow stress, and bulk flow surface. Heat treatment and CCT need the first two; V-Gleeble deformation modes also need a bulk flow surface; RSW needs the first two only when microstructure tracking is enabled. If a capability cannot be resolved, the run fails with a typed message naming it rather than quietly substituting something else. See the capability matrix.

Confirmation gate — The stop Phases puts in front of any custom or changed chemistry. The tool creates nothing, tells you which catalog grade the chemistry is closest to, and asks whether to continue or adjust. Confirming stores a stable material definition plus a compatible base template — it does not calculate kinetics, flow stress or any physics, and it is not a calibration claim. See your own chemistry.

Custom simulator — A simulator your workspace authored and published itself, rather than one of the four built in. It appears on the Custom tab of the simulator selector and runs through the run_simulator tool. Publishing one never claims scientific validation; it only says the definition is structurally valid and that an admin approved it. See custom simulators.

Distance badge (template match) — The good / moderate / high label shown beside a custom material. It measures how far your chemistry sits from the nearest catalog identity or template, and nothing else. It is not an accuracy score and not a calibration score; the simulator's own capability result is what tells you whether the physics is available. See your own chemistry.

Draft (simulator draft) — A custom simulator that has been written but not published. Drafts are versioned, can be edited, validated and test-run, and are visible only to the workspace that owns them. Nothing in a draft can run from the Custom tab until it is published. See custom simulators.

Durable run — A simulation executed as a recorded workflow rather than as a transient in-process call. The tool accepts the work and returns an operation identifier immediately; the solver continues on the server; the finished result is verified, stored and recoverable afterwards. Acceptance is not a result — the numbers only exist once the operation reaches a completed state. See long-running simulations.

Family — The application-oriented classification on a catalog material: structural steel, case-hardening, hot-work tool steel, martensitic stainless, cast iron, and so on. The catalog uses 43 of them. The largest are Structural steel (680 grades), General engineering steel (411) and General tool steel (210); several hold only one grade. A family is the useful unit when you want "something like this", not this exact grade. See families and data sources.

Fenced parameter block — The mechanism behind the parameter cards. The agent emits a fenced code block with a specific tag and single-line JSON inside; the chat replaces that block with an editable card instead of showing the code. The tags you will meet are phases-params, cct-params, rsw-params, rsw-config, vgleeble-params, material-params and flow-stress-params, plus flow-stress-results, which renders finished curves rather than an editable form. rsw-results is still emitted after a weld run, but the chat now removes it without rendering it; the figures you see come from the run's stored artifacts. See parameter blocks.

Guest session — Using Phases without an account. An anonymous identity is created for you automatically and gets a limited number of messages per fixed window that starts with your first message — ten per 24 hours unless the deployment is configured otherwise (GUEST_MESSAGE_LIMIT and GUEST_WINDOW_HOURS). Guests have no chat history sidebar, no materials workbench and no saved run browser. See guests and limits.

Material card — The LS-DYNA-format keyword file (.k) that carries everything the solver needs about a steel: chemistry, equilibrium curves, critical temperatures, kinetics parameters, thermal properties and the as-delivered phase fractions. Every run stores the exact card it consumed, as mat_Steel.k, alongside a material_resolution.json manifest recording how that card was resolved. See the material catalog.

Operation id — The identifier a durable simulation tool returns when it accepts work. Retain it to check whether the run has finished and fetch its verified result after a disconnect or host timeout. Over MCP, call ask_phases with this operation_id, the original product_session_id, and an empty message after the suggested polling delay. See Use Phases through MCP.

Operator — One typed, version-pinned step a custom simulator is allowed to use. The server-owned catalog is deliberately small — nine operators: resolve a material; run heat treatment, CCT, V-Gleeble or phase flow stress; expand a parameter sweep; take one position history out of a 1D model; draw a chart; write a report. RSW is not among them. A manifest may also declare its own equation and sandbox operators, which is where a workspace simulator computes something Phases does not already own; everything else a graph reaches is on that server list. See custom simulators.

Parameter card — The editable summary the agent shows before running a simulation: the parameters it understood, laid out as fields with a run button. You can correct any field before pressing run. For RSW the card is mandatory — the agent shows it first even when you have already stated every value. See how the conversation works.

Preserved-source kinetics — A named fallback, not a silent one. When a custom chemistry has no published kinetics model that applies to it, heat treatment, CCT and the V-Gleeble thermal modes will run using the verified template card's authored kinetics and report that choice as a warning. Those kinetics belong to the template, not to your chemistry, and must not be read as interpolated custom physics. See known limits.

Publication (publishing a simulator) — The step that turns a workspace simulator draft into something the Custom tab can run. It takes a server-issued review first and then an exact human confirmation on a later turn, and it is recorded as a specific pinned version. Publication is a governance act, not a scientific one: it never asserts that the model is validated. See custom simulators.

Run — One execution of one simulator, with its own record, its own storage prefix and its own artifacts. A CCT sweep across five cooling rates is a single run, not five. See sessions, runs and artifacts.

Run version — The sequential number identifying a run within its session: run_001, run_002, and so on, which is also the folder name in storage. Versions are reserved before execution and are never reused, so a deleted run's number does not come back. This is what you cite when you say "inspect run 3". See reuse and share runs.

Session — One conversation and everything attached to it: the message history, the active material, and the numbered runs. The active material persists across turns within a session, which is why you can load a grade once and then run several simulations against it. See sessions, runs and artifacts.

Session material — A material that exists only inside the current conversation. Loading a catalog grade, or confirming a custom chemistry, produces one. It is shown in simulation output as the closest grade label followed by (custom) when a chemistry was applied. It disappears with the session unless you save it. See your own chemistry.

Simulation study — A bounded set of related runs planned and executed together: a parameter sweep, a material comparison, a factorial design, a target search or a process-window search. The plan is written down as exact trials before anything runs, and only verified metrics take part in the ranking and the report. Studies are capped — at most 64 trials over at most three swept axes — so a study is a designed experiment, not an open-ended search. See run a simulation study.

Workspace material — A material saved into a workspace library, with a name you chose, available to everyone in that workspace and loadable in future sessions. Saving requires a workspace and a member or admin role, and Phases refuses to save a material whose chemistry is identical to its unmodified base grade. Editing a loaded workspace material in chat forks it into a session material rather than changing the stored row. See workspace materials and workspaces and roles.

Where these terms come from#

If a term you need is missing, the fastest way to get an authoritative answer is to ask in the chat itself — the agent will not invent a capability it does not have, and it names its tools and their limits plainly. The full list of what it can call is in the tool reference, and the boundaries it will not cross are collected in known limits.