Somewhere in your material specification there is a line that reads something like 27 J at −20°C, average of three. It is easy to treat as a box to tick. It is actually the single requirement standing between a structure that bends when something goes wrong and one that breaks.
Here is what the Charpy test measures, why the numbers in your spec are the numbers they are, and what to do when a set fails.
What the test actually does
A standard specimen — 10 mm × 10 mm × 55 mm, with a precisely machined V-notch — is held at both ends and struck by a swinging pendulum. The pendulum has a known energy at release. Whatever it does not spend breaking the specimen carries it up the other side, and the height it reaches gives the energy absorbed. That absorbed energy, in joules, is the result.
The notch matters as much as the material. It concentrates stress at a defined point and forces the fracture to start there, which is what makes results from two different laboratories comparable. A notch cut to the wrong radius, or a fraction off centre, produces a number that means nothing — which is why specimen preparation is a controlled process rather than a machining job.
So the test is not measuring strength. A tensile test measures strength. Charpy measures toughness: how much energy the material soaks up before it fractures, and — just as importantly — whether it tears or snaps.
Why temperature is in the specification
This is the part that makes Charpy worth doing at all.
Ferritic steels do not have one toughness value. They have a curve. Above a certain temperature they absorb a lot of energy and fail in a ductile way — the specimen bends, tears, and the fracture face is dull and fibrous. Below it, the same steel absorbs very little and fails in a brittle way — almost no deformation, a bright crystalline fracture face, and the crack runs.
The transition between those two behaviours happens over a surprisingly narrow band of temperature, and it is called the ductile-to-brittle transition. Austenitic stainless steels and aluminium alloys do not really show it; carbon and low-alloy steels emphatically do.
That is why your spec names a temperature. Testing at 20°C tells you what the steel does in a workshop. Testing at −20°C tells you what it does on a January night in Cumbria, or on the outside of a cryogenic vessel, or at the seabed. The requirement exists because somebody worked out the lowest temperature the component will ever see and demanded evidence that the steel is still on the ductile side of its curve when it gets there.
Get this wrong and the consequence is not a slightly weaker structure. It is a structure that behaves entirely differently — one where a small flaw that would have been harmless propagates all the way across.
Reading a result properly
A Charpy certificate that reports only absorbed energy is telling you less than half of what the test produced.
Absorbed energy is the headline number, reported for each of three specimens plus the average. Most specifications set both a minimum average and a lower minimum for any individual specimen, because one bad result among three matters.
Lateral expansion is how much the specimen spread sideways at the notch as it deformed. It is a direct measure of ductility, and some codes — particularly in pressure equipment — specify it instead of, or alongside, energy.
Percentage shear is the proportion of the fracture face that failed in a ductile manner. It is assessed visually against reference standards. A result of 27 J with 80% shear and a result of 27 J with 10% shear are the same number describing two different materials: the first is comfortably ductile, the second is sitting on the edge of its transition and only just scraped the value.
If you are only reading the joules, you are missing the signal that tells you how much margin you have.
Where the specimen comes from
Two things about sampling change the answer, and both get overlooked.
Orientation. Rolled plate is not the same material in every direction. Specimens taken transverse to the rolling direction routinely give lower toughness than those taken longitudinally, because the elongated inclusions from rolling line up to help a crack across the plate. Your specification will state the orientation, and it is not arbitrary — it should match the direction a crack would actually run in service.
Location. For weld procedure qualification, specimens are taken from defined positions: weld centreline, fusion line, and set distances into the heat-affected zone. Those positions exist because the HAZ is where toughness is most likely to have been damaged by the welding thermal cycle. Testing the weld metal alone and calling the procedure qualified misses the region most likely to fail.
When a set fails
It happens, and the response is a procedure rather than a panic.
Most standards allow a retest under defined conditions — typically when the average passes but one specimen falls below the individual minimum, or when the average falls slightly short. The retest usually means three more specimens from the same material, and all of them have to meet the requirement. What the standards do not allow is repeated retesting until a set passes; that is fishing, and it will not survive an audit.
Before retesting, it is worth asking what actually failed:
- Was the material right? A grade substitution, or a plate that is not what the certificate says, shows up here first. Positive material identification is quicker than a retest.
- Was the heat treatment right? Normalising and post-weld heat treatment both change toughness substantially. A missed or mistimed cycle is a common cause.
- Was the welding heat input controlled? Excessive heat input coarsens the grain structure in the heat-affected zone, and coarse grains are less tough. This is one of the main reasons WPSs cap heat input.
- Was the specimen prepared correctly? Notch geometry, orientation and location are all candidates before you conclude the steel is at fault.
A failure is information. Treating it purely as an obstacle to a retest wastes it.
Why three specimens
Charpy results scatter. Two specimens cut from adjacent positions in the same plate, tested at the same temperature, can differ by a wide margin — and near the transition temperature they can differ enormously, because a small difference in local microstructure decides which side of the transition that specimen falls on.
That scatter is a property of the material, not a laboratory error, and it is why every specification asks for three and judges the average. A single specimen is not evidence of anything.
It is also why the individual minimum exists alongside the average. Three results of 40, 40 and 5 J average 28 and would pass a 27 J average requirement, while describing a material that failed catastrophically once in three attempts. The individual minimum is there to catch exactly that pattern, and if you are specifying impact requirements yourself it is the clause worth getting right.
Where scatter is unusually wide, the honest response is to look at why rather than to average it away. Wide scatter at the specified temperature usually means you are sitting on the transition, and sitting on the transition means the margin you thought you had does not exist.
What our laboratory does
Charpy impact testing here runs to 300 J, with sub-zero conditioning for low-temperature requirements. Specimens are machined in-house, so notch geometry is controlled by the same quality system that reports the result rather than being someone else's variable.
We report absorbed energy per specimen and averaged, lateral expansion, and percentage shear, with the specimen orientation and location stated — because a certificate that omits those is not traceable to anything.
Where a set fails, we will tell you what the fracture faces suggest before you commission a retest. Metallographic examination of a failed specimen frequently identifies the cause in a day, and that is usually cheaper than three more specimens and a second failure.
Impact testing sits alongside tensile, hardness, bend and fracture testing in the same laboratory, which matters for procedure qualification where all of them feed one WPQR. One provider, one set of records, one certificate.
The short version
Charpy tells you whether your steel tears or snaps at the coldest temperature it will ever see. The temperature in your specification is the whole point of the test, not a detail. Read shear percentage and lateral expansion alongside the joules, because they tell you how much margin you have rather than just whether you passed. And when a set fails, find out why before you retest — the answer is usually in the material certificate, the heat treatment record or the notch, and only rarely in the steel itself.
If you have coupons that need testing, or a specification you are not sure how to sample against, send us the drawing and the material grade and we will tell you what the test programme should be.
TECHNICAL REVIEW — DELETE FROM THE LINE ABOVE, DOWN
Scheduled for 2026-08-18. Not to go live until a Level 3 or the RPA has read it.
Check: HIGHEST-EVIDENCE PIECE: 716 impressions and ZERO clicks at best position 7.6 over 90 days. Lab to confirm the 300 J capability, sub-zero range and that we machine specimens in-house. Check the retest wording against the standards we actually work to.
Image to shoot: Pendulum mid-swing on the impact rig, plus a fracture-surface close-up showing the ductile/brittle contrast with a scale in frame.
Internal links already in the text: /charpy-impact-testing/, /destructive-materials-testing/, /tensile-testing/, /flawed/ — confirm they read naturally, do not add more.
