"Mildly Flammable" Describes How an A2L Burns and Does Not Measure Ignition Odds
Class 2L requires a burning velocity of 10 centimeters per second or less, and the criteria set no ignition-energy limit. Minimum ignition energies in the sources reviewed run from 8 to 100 millijoules for R-32 and from 510 to 7,500 for R-1234yf.
- "Mildly Flammable" Describes How an A2L Burns and Does Not Measure Ignition Odds
- A2L Rules Vary City to City Because Codes Adopt Safety Standards Edition by Edition
- R-454B and R-32 Change Cylinders, Leak Detectors and Ignition Checks on the Job
- Swapping the Other Half of an R-410A Split Can Create a Prohibited New System
- R-32's Record Abroad Rests on Small Splits and Thin Incident Data
Chemours' U.S. safety data sheet for R-454B, revised March 5, 2025, carries hazard statement H220, "Extremely flammable gas." The Solstice technical data sheet for the same blend calls it "a mildly flammable refrigerant (ASHRAE class A2L)." The first label follows OSHA's hazard communication rule and the second follows ASHRAE 34. Chemours' 2022 European sheet for R-454C goes a third way, with H221, "Flammable gas."
ASHRAE's own heading is "Lower Flammability," in the 2018 addendum that made 2L a separate class. AHRI raised the wording in a July 2020 comment letter: "EPA may want to update references to 'mildly flammable refrigerants' to 'refrigerants with lower flammability' to align with ASHRAE definitions."
Standard 34 is sold, and its full text was not reviewed for this article, which relies on ASHRAE's free addenda. AHRI's website also refused automated downloads, so the final reports for AHRTI 9007-01, AHRTI 9008, AHRI 8017 and AHRI 8028 were not read in full. Their findings below come from a peer-reviewed paper from the 8017 grant, an AHRI presentation, an Oak Ridge report quoting 9007, short search excerpts, trade coverage and manufacturer literature, named where used.
Four Conditions, With Ignition Energy Not Among Them
Addendum g to Standard 34-2016, dated June 2018, sets out the classes this way. The letter A is a separate rating for lower toxicity.
| Class | Flame propagation at 140 F (60 C), 14.7 psia | LFL | Heat of combustion | Maximum burning velocity at 73.4 F (23 C), dry air |
|---|---|---|---|---|
| 1 | No | Not applicable | Not applicable | Not applicable |
| 2L, Lower Flammability | Yes | Above 0.10 kg/m³ (0.0062 lb/ft³) | Below 19,000 kJ/kg (8,169 Btu/lb) | 10 cm/s (3.9 in/s) or less |
| 2, Flammable | Yes | Above 0.10 kg/m³ | Below 19,000 kJ/kg | No limit |
| 3, Higher Flammability | Yes | LFL at or below 0.10 kg/m³, or heat of combustion at or above 19,000 kJ/kg | No limit | |
Burning velocity alone separates 2L from 2, and a 2015 KTH Royal Institute of Technology summary, citing Daikin, shows the effect: R-152a, an A2, clears the heat limit at 16 MJ/kg and misses the speed limit at 23 cm/s. The measurement is the maximum speed of a laminar flame relative to the unburned gas. Laboratories validate their method by reproducing 6.7 ± 0.7 cm/s for R-32 and 23.0 ± 2.3 cm/s for R-152a or by offering other evidence.
Flammability limits come from ASTM E681, using a spark in a 12-liter glass flask. Blends must meet the criteria at both the worst case of formulation and the worst case of fractionation. Addendum o to the 34-2024 edition, approved Aug. 31, 2026, spells out that LFL is judged at the composition with the lowest LFL and the 2L call at the composition with the highest burning velocity. EPA's 2021 SNAP rule describes the 34-2019 sequence slightly differently from the addendum: test at 73.4 F, then at 140 F only if no flame propagates at the lower temperature. Minimum ignition energy (MIE) appears nowhere in the criteria in Addendum g.
Five Refrigerants, and Numbers That Do Not Line Up
| Property | R-32 | R-454B | R-452B | R-454C | R-1234yf |
|---|---|---|---|---|---|
| Composition, weight percent (EPA) | HFC-32 | 68.9 R-32, 31.1 R-1234yf | 67 R-32, 7 R-125, 26 R-1234yf | 21.5 R-32, 78.5 R-1234yf | HFO-1234yf |
| LFL, percent by volume | 14.4 (ASHRAE, Chemours, AHRI) | 11.5 (ASHRAE; Chemours sheet citing it) 11.25 (Solstice) 11.3 (Chemours SDS) | 11.9 (ASHRAE, AHRI, Honeywell SDS) | 7.7 (ASHRAE, Chemours) | 6.2 (ASHRAE 2022, Chemours) 6.5 (worked example in ASHRAE Addendum o, 2026) |
| LFL, kg/m³ | 0.306 (ASHRAE) 0.307 (Chemours, AHRI) | 0.297 (ASHRAE, replacing 0.353 at worst-case fractionation) 0.278 (Climalife) 0.354 at worst-case fractionation (Chemours) | 0.3105 (ASHRAE) 0.310 (AHRI) 0.298 at worst-case fractionation (Chemours) | 0.292 (ASHRAE, replacing 0.290 at worst-case fractionation) 0.293 (Climalife) 0.289 at worst-case fractionation (Chemours) | 0.289 (ASHRAE, Chemours) |
| Burning velocity, cm/s | 6.7 (ASHRAE reference) 7.4 (Sadaghiani) | 5.2 (ASHRAE, Greenheck) | Under 4.0 (ASHRAE, AHRI) | Under 4.0 at worst-case fractionation (ASHRAE) | 1.5 (ASHRAE, Chemours) 2.7 (Kim and Sunderland table) |
| MIE, mJ | 8.0 ± 1.5 (Sadaghiani) 29 (Hihara) 30-100 (Chemours) 65 (Kim and Sunderland table) | 100-300 (Greenheck) | Not found | Not found | 510 ± 130 (Sadaghiani) 780 (Hihara) above 5,000 (Chemours) 7,500 (Kim and Sunderland table) |
| Heat of combustion, MJ/kg | 9.4 (Chemours) 9.5 (Greenheck) | 9.9 (Greenheck) | Not found | Not found | 10.7 (Chemours) |
| Autoignition temperature, C | 648 (Greenheck) 764 (Kim and Sunderland table) | 496 (Chemours SDS) | "No data available" (Honeywell SDS, 2018) | 444 (Chemours SDS) | 405 (Kim and Sunderland table) |
ASHRAE figures come from Addendum a to Standard 34-2022, approved Dec. 30, 2022, the latest table values read for this article. Later addenda and the 34-2024 text may differ. Greenheck and Chemours sell equipment or refrigerant. Check the current standard and the manufacturer's data before using any figure in a design.
"Variation of reported MIE values is very wide, which makes it difficult to use MIE as a flammability index," Kenji Takizawa of Japan's AIST told a 2015 workshop, putting the differences at up to three orders of magnitude. Even the lowest R-32 figure above is about 32 times Chemours' 0.25 mJ for propane. MIE also fails to predict everything: Takizawa reported that R-32 and R-1234yf did not ignite from a contactor spark at 220 volts and a 60-amp load, estimated at about 4 joules. University of Maryland researchers attribute the difficulty to long quenching distances, up to 25 millimeters. Takizawa's data tie quenching distance to burning velocity and estimated MIE to the cube of that distance, so the 10 cm/s limit bears on ignition indirectly even though it does not measure it.
Burning velocity is measured in dry air, while Takizawa's slide warned that "Humidity level in the current standard is not high enough for high humidity areas of the world." A Japanese research group's 2014 report found the flammable range of R-1234yf widens with humidity, and that R-410A, an A1, becomes flammable above 19 percent relative humidity at 60 C. In R-454C display-case tests, Stephen Spletzer of Chemours said in 2020, "high humidity was needed to produce a sustainable ignition."
Sadaghiani and colleagues found that 5 percent R-1234yf cut R-32's burning velocity by 11 percent and another 5 percent by more than 30 percent, so blends do not interpolate cleanly. A 2024 microgravity study with NIST coauthors, run at 3 bar and 333 K, attributes this inhibition partly to hydrogen fluoride (HF) formed in the flame trapping radicals.
A Research Program Built to Force Ignitions
The program's budget is reported as $5.2 million (Cooling Post, 2017), $5.8 million (Oak Ridge, 2019, and an AHRI deck, with different AHRI and California shares) and "$6 million plus" (AHRI's current page). Both itemized versions show DOE as the largest funder at $3 million.
UL ran AHRTI 9007-01, released in June 2017, across six room-scale scenarios: a packaged terminal air conditioner (PTAC) in a motel room, a rooftop unit serving a commercial kitchen, a walk-in cooler, a reach-in refrigerator, a split system in a utility closet with a servicing error, and a split with a failed hermetic terminal. AHRI's Karim Amrane told Cooling Post the tests were designed to create "relatively low-probability events." An AHRI deck from 2022, reporting the companion R-290 tests, is explicit about the limit: the testing was "more focused on understanding the ignition event severity, rather than the probability of the refrigerant ignition event."
In the PTAC comparison at the m1 charge level, 1.9 kg of R-452B released at 47.4 g/s produced a "small blue flame extinguished right away," while 110 grams of R-290 at 21 g/s propagated at floor level, with two deflagration vents bursting and one melting open. A related walk-in test, in a 12-by-14-by-8-foot box using R-455A and R-457A, reached peak temperatures of about 1,500-1,700 F, and leak location changed the outcome.
The 9007 preface, quoted by Oak Ridge, also delivered a finding that went beyond ignition. Earlier computer models "do not consider the possibility of a mechanism that can pool liquid refrigerant at the floor or droplet formation suspended in the air (refrigerant fog)," it says, and later R-410A work "confirmed refrigerant liquid could remain for hours depending on the leak methodology."
Hot Surfaces Gave Different Answers Under Different Conditions
Search excerpts of Exponent's AHRTI 9008 report say pure R-32, R-1234ze and R-452B did not ignite on a flat plate, with 5-gram charges and an 850 C target temperature. The report cautions that its results hold for its specific conditions and recommends testing other geometries. Chemours researchers later added a tube to confine the release and reported more differentiation between refrigerants at 800-850 C. Oil matters too: an AHRI FAQ, seen only in excerpt, says oil dominates hot-surface ignition, and a 2014 University of Maryland thesis recorded lowest hot-plate ignition temperatures of 764 C for R-32 alone and 649 C with 1 percent POE oil.
Carrier's Dick Lord described a full-size rig in ACHR News in 2020: a 1.5-ton ducted fan coil with an oversized 15 kW heater, limit switches jumped, airflow run well below allowable rates and leaks of 12 to 18 pounds per minute. "We confirmed that electric heater coils with surface temperatures up to 1,700°F will not ignite A2L refrigerants with airflows down to 115 cfm/ton for the typical blow-through configuration," he said. In a closet test with a standing-pilot water heater, flame extended at the pilot, and the cloud did not ignite because, Lord said, the heater's induced draft "pulled the combustion up the stack."
Still air tells a different story. Dennis Kim and Peter Sunderland's 2021 paper in Energies, funded by the AHRI 8017 grant with funders assisting in study design, tested 15 sources in a windowed 26-liter chamber. A nickel-chromium wire ignited well-mixed, quiescent 17 percent R-32 at 744 C, and those flames burned about 3 seconds. R-32 flames climbed to the ceiling and came back down to the floor, while R-452B and R-1234yf flames spread along the ceiling to the walls. Two wire tests lit mixtures slightly below the tabulated LFL, 13 percent R-32 and 10 percent R-452B. The authors note the LFL is defined with a weaker ignition source and a larger flame threshold.
Open flames, in the form of match heads and butane, also ignited premixtures. Candle and butane flames went out before any large deflagration when concentration rose slowly. Eleven sources produced no flame: a cigarette, a piezoelectric butane lighter, friction sparks, a plug and receptacle, a light switch, a hand mixer, a cordless drill, a toaster, a hair dryer, a hot plate and a space heater, with heating elements at 540 C or below. Results for hot surfaces in earlier work, the authors write, "are found to vary with the conditions and, often, to be stochastic." Chemours' summary of 8017 also counts a leak impinging on a candle as a viable source.
Tokyo University of Science found no propagation from a piezo gas lighter in accumulated R-32 and R-1234yf, but a kerosene cigarette lighter, lit in the test by an AC spark rather than its flint, ignited R-32 that had accumulated from a slow leak. The 2014 group report attributes pump-down explosions to diesel combustion of oil once air enters the tube, with the refrigerant then burning and A1 refrigerants behaving similarly.
Hydrogen Fluoride Appeared Without Any Refrigerant Fire
Tavares and colleagues give CH2F2 + O2 → 2 HF + CO2 for stoichiometric R-32 and a route to HF plus carbonyl fluoride for R-1234yf. By molecular weight, complete reaction would turn 1 pound of R-32 into about 0.77 pound of HF, a ceiling that real fires do not necessarily reach.
Tokyo University of Science released 800 grams of R-32, R-1234yf or R-410A at 10 or 60 g/min into a 7.8-square-meter room with a 2.4 kW radiant stove or 3.2 kW oil fan heater. Refrigerant stayed below 2 percent by volume and nothing ignited, yet the stove-run chart, read by eye, shows HF peaks of roughly 50 to 500 parts per million, with R-410A in the same range. NIOSH's immediately-dangerous level for HF is 30 ppm, and OSHA's limit is 3 ppm as an eight-hour average.
Lord's account conflicts on its face: "there is no HF generated even with temperatures 502°F above the autoignition temperature," he said of the electric-coil rig. The documents do not reconcile the two, though a combustion heater in a small room and an electric coil in a moving airstream are different conditions.
In fires, AHRI 8028, credited to George Hunter and run in cooperation with UL and the Firefighter Safety Research Institute, found hazardous HF from both classes, with HVACR News reporting "not a significant relative difference" between R-410A and the others in one scenario. A pooled-liquid basement scenario went the other way: R-32 ignited and spread across the room, and "levels of HF from the A2L refrigerant were significantly higher and faster developing." A 2023 California fire service article summarizes the study as A2Ls producing about 1-2 percent more HF, a figure that does not fit that scenario. Greenheck, citing 8028, reports the same 80 kW peak for R-32 and R-410A sprayed at a 120 kW burner.
What the Record Leaves Open
R-454B, which Greenheck names with R-32 as the top R-410A replacements, is absent from the test lists in the summaries read here: 9007 names R-32, R-452B, R-455A and R-457A, 9008 names R-32, R-1234ze and R-452B, and the 8017 paper names R-32, R-452B, R-1234yf and R-1234ze. Web searches for a published room-scale R-454B ignition test returned only manufacturer statements. The summaries read of AHRI 8028 and of Chemours' ten-refrigerant hot-surface work do not give full refrigerant lists, so R-454B cannot be ruled out of those two. R-452B is close in R-32 content, but the blend data above argue against assuming a match.
Oak Ridge concluded in 2018 that "there is no clear consensus on whether the flammable refrigerant charge limits are appropriate or effective for different scenarios," while also saying some assumptions may prove overly restrictive for A2Ls. Its leak tests on a 3-ton R-410A split, with 8.38 pounds of charge, averaged 3.9 to 14 pounds per minute over the first 80 percent of the release, compared with the roughly 2.1 pounds per minute implied by the four-minute release assumed in the 2016 version of IEC 60335-2-40. The high-rate cases simulated a full line break, while the low-rate cases aimed at "roughly" four minutes and still ran faster. Its model of 3.175 kg of R-452B lost in 18.36 seconds into a 167-square-meter home kept the monitored room concentration under 4 percent by volume, but "flammability risk was evident inside the ducts."
Within those limits, the record supports a narrow claim: compared with propane, A2Ls need several times the concentration, burn far more slowly and release roughly a fifth of the heat per kilogram. It does not support treating them as inert, since "an open flame can ignite any flammable refrigerant when a flammable concentration is reached," as Spletzer wrote in a Chemours paper. On any job, the unit's manual and the locally adopted code govern.
Why This Matters to Contractors
ASHRAE research project RP-1855 was scoped to cover HF and carbonyl fluoride exposure and how to clean up after an ignition event. AHRI's page still lists it as ongoing, while a search excerpt of an ASHRAE page says the final report came out in June 2023, and that report could not be read for this article. Until its findings circulate more widely, a service company called to a closet where an R-32 leak met a flame has more public data on how that fire behaved than on how to clean up after it.
Sources
- ASHRAE, Addendum g to ANSI/ASHRAE Standard 34-2016, June 2018. ashrae.org
- ASHRAE, Addendum a to ANSI/ASHRAE Standard 34-2022, Dec. 30, 2022. ashrae.org
- ASHRAE, Addendum o to ANSI/ASHRAE Standard 34-2024, Aug. 31, 2026. ashrae.org
- ASHRAE, Addendum ak to ANSI/ASHRAE Standard 34-2007, 2010. ashrae.org
- EPA, SNAP Rule 23 final rule, 86 FR 24444, May 6, 2021. govinfo.gov
- AHRI, comment letter, EPA Docket EPA-HQ-OAR-2019-0698, July 27, 2020. regulations.gov
- AHRI, "AHRI Flammable Refrigerants Research Initiative" (project list; final report PDFs refused automated download). ahrinet.org
- Xudong Wang, AHRI, "A3 Refrigerant R290 Leak and Ignition Testing in Whole Room Scale," June 13, 2022. heatpumpingtechnologies.org
- AHRI, "Low GWP Refrigerants FAQs," Safe Refrigerant Transition Task Force, seen in search excerpt only; automated download refused. ahrinet.org
- Cooling Post, "AHRI publishes first A2L report," June 13, 2017. coolingpost.com
- X. Wang, "A2L refrigerants leak and ignition testing for a walk-in cooler system," abstract, 25th IIR International Congress of Refrigeration, 2019. iifiir.org
- X. Wang, "AHRI research activities on flammable refrigerants," abstract, 2018. iifiir.org
- Xudong Wang and Pravin Gandhi, "A2L Refrigerants Leaks and Ignitions Testing under Whole Room Scale," abstract, Purdue, 2018. purdue.edu
- Patrick Coughlan, Nina Gray and Mary Koban, "Expanded Hot Surface Ignition Testing of A2L Refrigerants," abstract, Purdue, 2018. purdue.edu
- Dennis K. Kim and Peter B. Sunderland, "Viability of Various Sources to Ignite A2L Refrigerants," Energies 14, 121, 2021. umd.edu
- Adam Boussouf, University of Maryland, "Hot Surface Ignition of R-32 and R-410A Refrigerant Mixtures With Lubricating Oil," M.S. thesis, 2014. umd.edu
- Joanna R. Turpin, ACHR News, "The Results Are In: Experts Share A2L and A3 Refrigerant Test Data," November 2020. achrnews.com
- Mirhadi S. Sadaghiani and colleagues, Journal of Hazardous Materials 407, 124781, 2021, repository record. uwa.edu.au
- Raik Hesse and colleagues, "Exploring Nonlinear Flame Speed Inhibition Effects in Mixtures of R1234yf and R32," Proceedings of the Combustion Institute 40, 2024, abstract. nist.gov
- Justin K. Tavares and colleagues, "Effects of radiation heat loss on planar and spherical hydrofluorocarbon/air flames." nsf.gov
- Eiji Hihara, University of Tokyo, "Research Project on Risk Assessment of Mildly Flammable Refrigerants," ICR2015 workshop. jsrae.or.jp
- Kenji Takizawa, AIST, "Fundamental and Practical Flammability Properties of 2L Refrigerants," Aug. 20, 2015. jsrae.or.jp
- Tomohiko Imamura and Osami Sugawa, "Experimental Evaluation of Physical Hazard of A2L Refrigerant Assuming Actual Handling Situations," Aug. 20, 2015. jsrae.or.jp
- Japan Society of Refrigerating and Air Conditioning Engineers, 2014 progress report on mildly flammable refrigerants. jsrae.or.jp
- hvacrnews.com.au, "AHRI releases A2L report for firefighters," March 30, 2021. hvacrnews.com.au
- The California Fire Service, March-April 2023, page 19. csfa.net
- Greenheck, "The Safety of A2L Refrigerants in Direct Gas-Fired HVAC Equipment," MUA/101-24, 2024. greenheck.com
- Stephen Spletzer, Chemours, "What Are A2L Refrigerants and Why Do We Need Them?" chemours.com
- Chemours, Safety Data Sheet, Opteon XL41 (R-454B), version 6.0, March 5, 2025. hudsontech.com
- Chemours, Safety Data Sheet, Opteon XL20 (R-454C), version 13.0, March 5, 2025. hudsontech.com
- Chemours, European Safety Data Sheet, Opteon XL20 (R-454C), 2022. schiessl.cz
- Honeywell, Safety Data Sheet, R-452B, version 1.2, Aug. 22, 2018. hudsontech.com
- Solstice Advanced Materials, Solstice 454B technical data sheet. solstice.com
- Chemours, Opteon XL41 product information. opteon.com
- Chemours, Opteon XL20 product information. opteon.com
- Chemours, "Safety of Opteon XL Refrigerants." schiessl-kaelte.com
- Climalife, "A2L refrigerants guide." climalife.com
- KTH Royal Institute of Technology, "Understanding Refrigerant Flammability," 2015. kth.se
- Van Baxter and colleagues, Oak Ridge National Laboratory, ORNL/TM-2018/804, June 2018. ornl.gov
- Van Baxter and colleagues, Oak Ridge National Laboratory, ORNL/TM-2018/1058, March 2019. ornl.gov
- Ahmed Elatar and colleagues, Oak Ridge National Laboratory, ORNL/TM-2019/1331, Sept. 13, 2019. ornl.gov
- ASHRAE RP-1855 scope as listed in the HPT Annex 54 2019 annual report. heatpumpingtechnologies.org
- NIOSH, Pocket Guide to Chemical Hazards, hydrogen fluoride. cdc.gov
- OSHA, Occupational Chemical Database, hydrogen fluoride. osha.gov