A plant team often holds a calorimetry report and no clear way to turn its temperatures into a decision. The data are all there: the process temperature, the highest temperature after a cooling failure, a decomposition limit and the vessel's technical limit. Stoessel criticality classes put those four temperatures in order and sort an exothermic process into five classes, from low risk at Class 1 to the highest at Class 5. A wrong class leaves either an unprotected hazard or money spent on protection the process does not need. This guide explains how the five classes are defined, what each one tells a plant owner to do, how a process moves between them and where the method stops. It draws on Francis Stoessel's published work and the process safety literature, within our chemical reaction engineering practice.
What Stoessel Criticality Classes Are and the Scenario Behind Them
Stoessel criticality classes answer one narrow question: if cooling fails at the worst moment, how far does the batch get before something stops it? The answer depends on four temperatures. The five classes, also called the Stoessel criticality index and drawn as the Stoessel criticality diagram, are five characteristic ways of ranking them. Francis Stoessel set the method out in his textbook Thermal Safety of Chemical Processes and applied it to protection planning in a 2009 paper.
Every criticality class rests on a cooling-failure scenario, which treats the loss of cooling as the worst case and the reactor as adiabatic from that moment. The scenario unfolds in a fixed sequence, set out in an IChemE Technical Process Safety workshop that follows Stoessel:
- Cooling is lost while the batch runs at its process temperature (Tp).
- Unreacted material reacts and lifts the batch to the maximum temperature of the synthesis reaction (MTSR).
- A secondary reaction may start, such as a decomposition, if MTSR is high enough, and it drives the temperature to a final value.
- The technical limit may intervene. The maximum temperature for technical reasons (MTT) is a boiling point or a pressure limit of the vessel.
The scenario describes what the chemistry and the equipment allow once control is lost. It does not say how often control is lost, and that separation is deliberate. Likelihood enters later, through the probability reading described below.
The four temperatures come from different measurements and carry different uncertainties. The next section shows where each one originates.
Where the Four Inputs Come From: The Data Behind a Class
A criticality class is only as good as the four temperatures behind it, and each comes from a different part of the work. Two describe the process and its chemistry, one describes the equipment and one describes the thermal stability of the reaction mass. A mistake in any single input can move a process by a whole class. The table shows what each input is, what defines it and where the data originate.
| Input | Defined by | Data source | Common weakness |
|---|---|---|---|
| Process temperature, Tp | The mode of operation | The recipe. For non-isothermal processes, the instant at which a cooling failure has the worst consequences | Using the nominal set point when the worst instant is elsewhere in the batch |
| MTSR | Accumulation of unreacted material and Tp | Reaction calorimetry: heat of reaction, heat capacity and accumulation under the real feed | Accumulation measured on a lab feed profile the plant does not reproduce |
| MTT | The equipment | The boiling point in an open system. In a closed system, the temperature at which pressure reaches the relief set pressure | Gas-releasing reactions, where pressure can reach the limit before the liquid boils |
| TD24 | Thermal stability of the reaction mass | DSC screening and adiabatic calorimetry. TD24 is where the time to maximum rate under adiabatic conditions (TMRad) is 24 hours | Autocatalytic or induction-time behaviour that a global kinetic model misses |
Reaction calorimetry supplies the data behind MTSR, and our guide to what reaction calorimetry measures and why scale-up needs it covers that work. Thermal stability comes from DSC screening and adiabatic calorimetry. Adiabatic tests correct for the heat absorbed by the test cell through the phi-factor, the heat capacity of sample plus cell divided by that of the sample alone. Our guide to adiabatic temperature rise and MTSR explains how accumulation sets MTSR.
With the four inputs in hand, ranking them is mechanical. The ranking gives the five classes in the next section.
The Five Stoessel Criticality Classes, One by One
Each class is a different ordering of the four temperatures, and each ordering changes what stands between a cooling failure and a decomposition runaway. In Classes 1 and 2 the synthesis reaction cannot reach either the technical limit or the decomposition. In Classes 3 and 4 it reaches the technical limit, and in Class 5 the decomposition is triggered with no technical barrier before it. The diagram and table set the orderings side by side, reading upward from the lowest temperature.
| Class | Order, low to high | After loss of cooling | Guidance in the literature |
|---|---|---|---|
| 1 | Tp < MTSR < MTT < TD24 | Neither the technical limit nor the decomposition is reached. | Low thermal risk. No special measures, but avoid holding the batch under heat accumulation for long. Evaporative cooling or emergency relief could act as a barrier. |
| 2 | Tp < MTSR < TD24 < MTT | As Class 1, but TD24 sits below MTT, so a long hold under accumulation could trigger decomposition and reach MTT. | Low thermal risk if heat accumulation is avoided. The boiling point could not act as a barrier. |
| 3 | Tp < MTT < MTSR < TD24 | The technical limit is reached. The decomposition is not triggered. | Risk-reducing measures required. Safety depends on the heat release rate of the synthesis reaction at MTT. Evaporative cooling, controlled depressurisation, backup cooling, dumping, quenching or a relief system may be used. |
| 4 | Tp < MTT < TD24 < MTSR | The technical limit is reached and the decomposition could be triggered. | Risk-reducing measures required. Safety depends on the heat release rate of both reactions at MTT. If control measures fail, the secondary reaction is triggered. |
| 5 | Tp < TD24 < MTSR < MTT | The decomposition is triggered. The technical limit is reached only during the secondary runaway. | No safety barrier between the two reactions, and the technical limit may not serve as one. Alternative process design, such as a change from batch to semi-batch, is worth considering. |
The guidance column follows the IChemE workshop summary of Stoessel's method. Two pairs of classes reward a second look. Classes 1 and 2 differ only in whether MTT or TD24 sits higher, and that swap decides whether boiling could ever protect against a decomposition. Classes 3 and 4 differ only in whether MTSR lies above TD24, which decides whether a failure of the technical barrier lets the decomposition start.
A class also points to a protection strategy, which is the subject of the next section.
From Criticality Class to Protection Strategy: What Each Result Triggers
A class matters because it points to a protection strategy before any equipment is chosen. Stoessel's 2009 paper takes exactly this approach, planning protection measures against runaway reactions from the class. The principle of inherently safer design, promoted by AIChE's Center for Chemical Process Safety (CCPS), puts changes to the process ahead of added protection. The class tells a team where on that ladder it must start.
- Classes 1 and 2 — the guidance is that no special measure is needed beyond avoiding long holds under heat accumulation. The work becomes confirming that feed control and hold times keep the batch away from accumulation.
- Classes 3 and 4 — risk-reducing measures are required, and the technical limit can act as a barrier only if the behaviour at MTT is determined quantitatively.
- Class 5 — no barrier exists between the two reactions, so the first question is whether the process itself can change.
The class is only half of the reading. The other half is how severe and how fast a runaway would be, and Stoessel combines the two in a three-by-three matrix. Severity follows the adiabatic temperature rise (ΔTad), and probability follows the time to maximum rate under adiabatic conditions (TMRad). The table gives the bands as reproduced in the IChemE workshop and in a Purdue University review of the method.
| Severity | Low probability: TMRad of 24 h or more | Medium: 8 to 24 h | High: under 8 h |
|---|---|---|---|
| Low: ΔTad below 50 K, no pressure increase | No measures required | No measures required | No measures required |
| Medium: 50 to 200 K | Technical measures | Technical measures | Redesign the process |
| High: above 200 K | Technical measures | Redesign the process | Redesign the process |
Neither tool replaces the other. The class shows whether a technical barrier sits between the synthesis reaction and a decomposition, while the matrix shows how severe and how fast the runaway would be. A Class 3 process with a large adiabatic temperature rise is a different problem from a Class 3 process with a small one. Both readings depend on how the process is run, so a team can change them, and the next section covers how.
How a Process Moves Between Criticality Classes: The Design Levers
The most useful property of the classification is that it is not fixed. Each input depends on how the process is run, so a deliberate change can move the criticality class and an unrelated change can move it by accident. The table lists the levers that act on each temperature, with the check each one needs. Every row is a question for the assessment team, not an instruction.
| Lever | Input it moves | Direction | What to check |
|---|---|---|---|
| Controlled feed instead of charging at once | MTSR, through the accumulated fraction of reactant | Lowers MTSR | Calorimetry must show that the feed is consumed as it is added. If the reaction is kinetically limited, heat release continues after the feed stops. |
| More solvent, lower concentration | MTSR, because the heat released per kilogram of reaction mass falls | Lowers ΔTad and MTSR | The effect on MTT, TD24 and vessel capacity. |
| Lower process temperature | MTSR, which is Tp plus the accumulated rise | Lowers MTSR directly | A slower reaction can accumulate more reactant, which pushes MTSR back up. Confirm accumulation at the new temperature. |
| Higher design or relief set pressure | MTT | Raises MTT | An MTT above TD24 changes the ordering and can remove the barrier that boiling provided. |
| Different solvent | MTT in an open system, and the heat capacity | Moves MTT and MTSR | Re-measure TD24 in the new medium. |
| Change of impurity profile, water content, catalyst or conversion state | TD24 | Can raise or lower it | TD24 belongs to the composition tested. Re-measure it rather than carry it over. |
Controlled feed is a standard way to limit accumulation, and our guide to batch, semi-batch and continuous reactor selection explains how the reactor mode is chosen. Knowing which lever to pull is one question. Knowing which method answers which safety question is the next.
Stoessel Criticality Classes vs Risk Matrix, HAZOP and LOPA
These methods are often named together and sometimes confused, yet each answers a different question at a different point in a safety study. The criticality classes rank a thermal scenario from measured temperatures. The matrix ranks consequence and speed, while HAZOP and LOPA work on deviations and protection layers. A common sequence uses the first two to feed the second two.
| Method | Question it answers | Main inputs | Output |
|---|---|---|---|
| Stoessel criticality classes | After a cooling failure, what stands between the batch and a decomposition? | Tp, MTSR, MTT, TD24 | A class from 1 to 5 and a protection strategy |
| Severity and probability matrix | How severe and how fast would the runaway be? | ΔTad and TMRad | A zone: no measures, technical measures or redesign |
| HAZOP (IEC 61882) | Which deviations from design intent can occur, and what follows? | Process description, P&IDs and guide words | Recorded causes, consequences, safeguards and actions |
| LOPA (a CCPS method) | Do the independent protection layers reduce the scenario frequency enough? | An initiating event, a consequence and layer data | Scenario risk compared with the company's criteria |
Read as a chain, the methods fit together. Calorimetry produces the inputs, the class and matrix rank the scenario, HAZOP finds the other ways the process can deviate, and LOPA tests whether the chosen protection is enough. Each step feeds the next, so a weak input at the start carries through to the protection decision.
Those inputs often change between the laboratory and the plant, which is where most surprises begin.
Troubleshooting: When the Class Changes Between Lab and Plant
Many unpleasant scale-up surprises are class changes that nobody was watching for. A criticality class assigned from a lab study describes the lab's feed profile, concentration and vessel limits. At plant scale the cooling, the addition time and the relief pressure all differ, and one input can cross another. The table connects the symptoms a plant team sees to the part of the class picture to examine first.
| Symptom | What it usually points to | Where to look first |
|---|---|---|
| Unexpected exotherm on scale-up, or temperature still climbing after the feed stops | Accumulation higher than assumed, so MTSR is higher than the class used | Accumulation measured under the plant's actual feed time and temperature |
| Cooling capacity insufficient for the reaction at plant scale | The first question of the scenario fails: the cooling system cannot control the process temperature | Heat release rate against jacket capacity at plant scale, before any classification |
| Reaction mass boils during a cooling failure, with the relief valve lifting on the batch reactor | The technical limit was reached, which happens in Classes 3, 4 and 5 | The MTT used in the study against the relief set pressure and boiling point of the real vessel |
| TD24 lower than expected after a recipe change | The composition changed: impurity profile, concentration or conversion state | Whether TD24 was measured on this composition or carried over |
| A second, faster exotherm appears above the process temperature | MTSR above TD24, which is Class 4 or 5 | The quality of the TD24 data and how close the two values are |
| Product held hot for hours during transfer or hold-up | A long hold under heat accumulation, the caveat attached to Classes 1 and 2 | Hold time compared with the time to maximum rate at the hold temperature |
Each symptom is a prompt to rerun the classification with plant-scale inputs, which is a management-of-change decision as much as a technical one. The control side of the same gap, where loops and jackets fail to hold temperature, is covered in our guide to cascade temperature control of jacketed reactors. Rerunning a classification does not remove its limits, which the next section sets out.
What Stoessel Criticality Classes Cannot Tell You
A criticality class is a temperature ranking for one scenario, so its limits matter as much as its strengths. It does not see pressure directly, it does not cover every way a batch can go wrong, and it rests on inputs that are themselves estimates. Knowing where it stops keeps a Class 1 result from being read as a clean bill of health.
- Pressure — the inputs are temperatures, and pressure enters only through MTT. For gas-releasing reactions the Purdue University review notes that MTT can sit below the boiling point, because the pressure limit may be reached first. HEL's 2025 presentation to the Purdue Process Safety and Assurance Center states that pressure is usually the real hazard.
- One scenario — the method addresses the cooling-failure scenario. Overcharging, loss of agitation or contamination fall outside it, and a HAZOP is the tool that finds them.
- Single-step reactions — the review notes that the method is aimed at single-step reactions and can miss interactions between steps. Jiang and co-authors proposed new methods for two-step syntheses in 2018.
- Changing composition — if all the solvent boils off, HEL notes, the more concentrated mixture may exceed MTT and change the class.
- Estimated inputs — empirical TD24 estimates generally assume Arrhenius kinetics and a global mechanism, while competing, sequential or autocatalytic reactions can behave differently, as HEL's presentation points out.
- Undrawn orderings — the diagram draws five of the six possible orderings of MTSR, MTT and TD24 above the process temperature. The sixth, with TD24 lowest and MTT below MTSR, needs an assessor's judgement. Stoessel's physical-operations version groups both orderings of the two upper limits in Class 5.
These limits are why the class opens a study rather than ending one. They also show how the same logic travels beyond the reactor.
Criticality Classes for Drying, Distillation and Storage
The same logic now extends beyond reactors. In September 2025 Francis Stoessel, Annik Nanchen and Anne-Florence Tran-Van presented criticality classes for physical unit operations, in which no reaction should take place in the process temperature range. It covers drying, distillation, extraction, granulation, extrusion, filtration, centrifugation, storage and transport. The question is unchanged: can the equipment push the contents past the point where a secondary reaction starts?
| Element | Reactions | Physical unit operations |
|---|---|---|
| Failure scenario | Cooling failure | Heating failure |
| Energy source | Heat of the synthesis reaction, giving MTSR | The heating system, giving THmax, the maximum temperature it can deliver |
| Stability limit | TD24 | TSTAB, which may be TD24 or another limit |
| Equipment limit | MTT | MTT, often related to pressure |
The stability limit can come from several tests. It may be TD24 for short processes, or the temperature at which the time to maximum rate is four times the process time. It may also be taken from storage tests such as Dewar, basket or Grewer oven tests.
Their paddle-dryer example shows how much the heating medium matters. With a steam supply the dryer lands in Class 5. Replacing steam with hot water, so that the heating system can no longer reach the stability limit, brings it to Class 2. The class falls because the design removes the scenario, not because protection was added. The reading of the criticality classes carries over from reactions.
Whether the scenario is a reactor or a dryer, the assessment is specialist work. The next section describes what it involves.
What a Competent Classification Involves, and What You Should Receive
Assigning a criticality class looks like a sorting exercise, but each of the four inputs rests on specialist measurements and judgement. A competent assessment measures heat release and accumulation on the real recipe, characterises the thermal stability of the real reaction mass and checks equipment limits against the actual vessel. Process safety engineers carry it out, and they can tell when two values that look different are effectively the same. In Great Britain the HSE guide to designing and operating safe chemical reaction processes covers the same ground.
A plant owner commissioning the work should expect to receive:
- A scenario definition — the cooling failure assumed, the point in the batch and the accumulation.
- The four inputs — each with its data source, test conditions and uncertainty.
- The class and its sensitivity — including what would change it.
- The severity and probability reading — with the matrix zone it falls in.
- The consequences for protection — the barrier concept the class supports and the follow-up studies required, such as behaviour at MTT.
- Reassessment triggers — the changes that require the classification to be redone.
The classification described here is site- and recipe-specific. It requires thermal data measured on the actual reaction mass at conditions representative of the intended process. It must be performed and interpreted by qualified process safety engineers against the regulatory regime that applies to the site. In the United States that is typically OSHA's Process Safety Management standard (29 CFR 1910.119) and the EPA's Risk Management Program, and equivalent regimes apply elsewhere. This article explains what the classes mean. It is not a substitute for that assessment, and none of the orderings, bands or guidance quoted in it should be applied to a real process without it.
Knowing what to expect from the work makes the next decisions easier to frame.
Which Questions to Ask First: A Decision Framework
A criticality class is the start of a safety conversation, so the useful question is what to do with it. The answer depends on where the process stands: on paper, in the laboratory, heading for a pilot plant or already running. The sequence below turns a classification into decisions, from the cheapest to the most expensive.
- Before scale-up — confirm that heat release, accumulation and a stability limit have been measured for this recipe. Without them the class is a guess.
- Margins — list which of the four inputs sit close to each other. Those pairs carry the uncertainty and deserve repeat measurement.
- Class 1 or 2 — record what keeps the batch away from heat accumulation, such as feed control and hold times, in the operating documents.
- Class 3 or 4 — commission the quantitative study of behaviour at the technical limit, because relief and emergency design depend on it.
- Class 5 — ask first whether the process can change: feed mode, concentration, temperature or heating medium. Inherent options come before added protection.
- After any change — reassess when the recipe, scale, solvent, vessel or relief pressure changes.
Plants without in-house capability usually commission the first classification as a thermal hazard assessment from a process safety consultant. The result then feeds a wider process safety review and, where needed, HAZOP facilitation. Whoever does the work needs the same data package: heat release, accumulation and thermal stability measured on the real recipe. That holds for an in-house team, a process safety consultant or external process hazard analysis consultants alike.
Our scale-up support engagements start with this work for any process heading toward a larger vessel. Our process and EHS consulting carries the follow-up as contract process engineering, including for manufacturers in India and the Gulf.
Frequently Asked Questions
What are Stoessel criticality classes?
Stoessel criticality classes are a five-level ranking of the thermal risk of an exothermic process, set out by Francis Stoessel. The ranking compares four temperatures in a cooling-failure scenario. They are the process temperature, the maximum temperature of the synthesis reaction (MTSR) and the maximum temperature for technical reasons (MTT). The fourth is TD24, the temperature at which the adiabatic time to maximum rate is 24 hours. Class 1 is the lowest risk and Class 5 the highest.
The classes guide the protection strategy a process needs. They are read with a severity and probability assessment, and they are one output of a thermal hazard study, not a replacement for it.
How do you determine the criticality class of a reaction?
A competent assessment measures the heat release and the unreacted material that can accumulate on the real recipe, normally by reaction calorimetry. Thermal stability of the real reaction mass is measured by DSC and adiabatic calorimetry. The equipment limit is the boiling point in an open system, or the temperature at which pressure reaches the relief set pressure in a closed one.
Ranking the four values gives the class. That step takes minutes, but the data behind it do not, and they must come from tests on the actual system rather than from a datasheet.
What does criticality class 5 mean?
In Class 5 a loss of control of the synthesis reaction triggers the decomposition, because MTSR is above TD24. The technical limit is reached only during the secondary runaway. In the IChemE workshop summary of Stoessel's method, no safety barrier exists between the two reactions.
That summary says the heat release of the decomposition at the technical limit may be too high for boiling or relief to serve as a barrier. It adds that alternative process design, such as a change from batch to semi-batch, is worth considering. For physical operations, Stoessel's 2025 presentation says the process or its conditions must be revised. A Class 5 result is a prompt to question the design before adding protection.
Is criticality class 1 completely safe?
No. Class 1 describes a low thermal risk in the cooling-failure scenario, because the synthesis reaction reaches neither the technical limit nor the decomposition. Even so, the guidance is to avoid holding the reaction mass for a long time under heat accumulation conditions.
Class 1 also says nothing about deviations other than a cooling failure, and it treats pressure only through MTT. A classification is therefore read alongside a HAZOP and pressure data, not instead of them.
Why does the classification use TD24 instead of an onset temperature?
An onset temperature is the point where a particular instrument first detects heat release, so it moves with the sensitivity of the apparatus. TD24 is defined by time. It is the temperature at which the time to maximum rate under adiabatic conditions is 24 hours. HEL's 2025 presentation notes that onset temperatures were once used with a safety margin chosen from the instrument sensitivity.
Tying the limit to time links it to the question that matters after a cooling failure: whether time remains for emergency measures. For physical unit operations, Stoessel's 2025 presentation allows other stability limits, such as the temperature at which the time to maximum rate is four times the process time.
When should a criticality class be reassessed?
Reassess whenever an input could have changed. Examples include a new recipe or concentration, a different feed profile or solvent, a new scale or vessel and a changed relief set pressure. A new raw-material source that alters the impurity profile counts too.
Each of these can move MTSR, MTT or TD24, and the class is not fixed. In facilities covered by OSHA's Process Safety Management standard, process changes fall under management of change, so the reassessment belongs in that procedure.
Can criticality classes be used for drying, distillation or storage?
Yes, with an adapted scenario. Stoessel, Nanchen and Tran-Van presented a version for physical unit operations in September 2025, covering drying, distillation, extraction, granulation, extrusion, filtration, storage and transport. The reaction energy is replaced by the maximum temperature of the heating system (THmax). A heating failure replaces the cooling failure, and TD24 becomes a thermal stability limit (TSTAB).
The reading of the five classes carries over, and in Classes 3, 4 and 5 the behaviour at MTT must be determined quantitatively. Their paddle-dryer example shows how much the heating medium matters, because the dryer moves from Class 5 to Class 2 when steam is replaced by hot water.
Do criticality classes include pressure?
Not directly. The four inputs are temperatures, and pressure enters only through MTT. In an open system MTT is the boiling point, and in a closed one it is the temperature at which pressure reaches the relief set pressure. For reactions that release gas this is a weak link. A Purdue University review notes that MTT can then sit below the boiling point, because the pressure limit may be reached first.
HEL's 2025 presentation puts it plainly: the parameters are temperatures, but pressure is usually the real hazard. A classification for a gas-releasing reaction therefore needs pressure data from calorimetry alongside the temperatures.
Holding a Calorimetry Report and Unsure Which Class You Are In?
Global Formulation provides thermal hazard assessment support, process safety review and HAZOP facilitation for chemical manufacturers, starting from the calorimetry and stability data for your own recipe.
Talk to Our Process Safety Team