Bake eight hours, not twelve: the claim rests on 0.1 points

This site has a page on what hydrogen embrittlement is and on why its test is not an acceptance test. Neither answers how long to bake, because the standards this site reads do not answer it either. So this page does something the site has barely done before: it reads a research paper, in full, and checks its arithmetic.

What was measured

The paper is Influence of baking time on hydrogen embrittlement in alkaline nickel-zinc electroplated steel fasteners, in the Indian Journal of Chemical Technology, volume 32, July 2025, pages 478–484. It was submitted in July 2024 and accepted in March 2025.

Declare the interest first: four of the six authors are at Norm Coating, a fastener coating company in İzmir, and two are at İzmir Katip Çelebi University. A coating plant publishing on how long its own process needs to run is not disqualifying, and the paper is open access so the working is visible. It is a thing to know while reading a conclusion whose practical effect is a shorter oven cycle.

The paper does not report how many bolts were pulled per condition. That is worth saying out loud rather than assuming a number, because the differences the conclusion turns on are small.

One detail lands on another page here. The SEM work found the coating on the thread region distributed unevenly because of the edge effect, thicker on the crest and thinner in the root, while the head came out uniform. That is the same geometry problem as the coating eating the thread tolerance, observed from the other side.

The table

Four conditions, four rows. HDR is hydrogen ductility recovery, the percentage gain in elongation over the unbaked reference.

The first thing in that table is not the conclusion. It is that strength barely moves. The paper says so directly: the maximum difference between the strength values did not exceed 1 %. Yield and tensile sit within a percent of each other across an unbaked bolt and a bolt that spent twelve hours at 200 °C. Everything that changes, changes in the elongation column.

That is the practical warning buried in a paper about ovens. A tensile test on an embrittled bolt reports a bolt that meets its class. This site says that already, on the page about why every check passes; here it is with numbers attached, from a bolt that was never baked at all.

Recomputing the third column

HDR is defined in the paper as the strain of the baked sample against the strain of the reference. Running that on the elongations printed in the same table:

Two rows reproduce to the decimal. The 8 h row does not. To print 11,4 the elongation would have to be about 15,49, and the table prints 15,4. The likeliest explanation by far is a rounded or truncated elongation column, and this page is not suggesting anything else.

It matters because of where that row sits. The conclusion is that eight hours is enough and twelve is unnecessary, and the evidence for it is 11,4 against 11,5 — one tenth of a percentage point. Recomputed from the printed elongations the same comparison is 10,79 against 11,51, a gap of seven tenths. Still small. Seven times larger than the one the conclusion rests on.

The body text and the table also disagree with each other: the prose gives the 4 h and 8 h recoveries as 7,1 % and 11,3 %, the table as 7,19 and 11,4.

What the paper does establish

Reading it sceptically is not the same as dismissing it, and three findings survive the arithmetic without depending on the 8 h row at all.

Four hours was not enough. Whichever figure you take, the 4 h recovery is well under the other two, and the paper states that four hours did not provide sufficient conditions for a total recovery of ductility. Four hours is also the minimum the paper cites from the general literature, which puts the commonly quoted floor at the bottom of the useful range rather than in the middle of it.

Baking did not change the microstructure. Grain counts and the ferrite to pearlite ratio came out close across all four conditions, so what the oven changed was the hydrogen, not the steel. For a part whose class was set by heat treatment, that is the reassuring half of the result.

An acoustic method tracked the same trend. Electromagnetic acoustic spectroscopy showed dislocation mobility highest in the unbaked samples and falling as bake time rose, agreeing with the elongation results. It is non-destructive, which is the interesting part: the established test is not. The standard hydrogen embrittlement test says of itself that it is not an acceptance test, and part of the reason is that you destroy what you test.

The standard the paper leans on is the one ISO just cut loose

For its baking guidance the paper cites ISO 19598, summarising it as the instruction that the recommended baking time rises with the strength of the material.

ISO 4042:2022 removed every reference to ISO 19598. Its introduction gives the reason: the latest editions of those general electroplating standards are not adequate to cover the requirements for electroplated fasteners, especially with regard to hydrogen embrittlement and baking. That is the precise subject of this paper.

ISO 19598 remains a current standard and citing it is not an error. But a 2025 study of fastener baking resting on the document ISO had just declared inadequate for fastener baking is worth seeing side by side. What else went when those references went is on another page.

There is one more overlap. The footnote under that ISO 4042 table reads: for alkaline zinc-nickel electroplatings, and nickel content from 12 % to 16 %, product testing shall be considered as part of in-process control and is not mandatory for each manufacturing lot. The paper’s coating is alkaline zinc-nickel and its EDX put the nickel at about 14 %. These bolts sit inside the one window where the standard relaxes per-lot testing.

What to take to a plater

• Ask for the temperature and the duration in writing. The standard says they shall be specified. A certificate that says baked and stops has not specified anything.
• Do not carry eight hours across to a different job. One coating, one class, one diameter, one paper. The paper itself notes the optimum depends on the coating, its permeability and thickness, and the substrate.
• Do not read strength as evidence. An unbaked bolt in this study met 12.9 on yield and tensile. The column that moved was elongation.
• Ask when the part was baked, not only whether. The interval between plating and baking is its own variable and it is covered separately.

This page covers step 5, the finish. The whole order is substrate, thread, head, drive, finish, documentation, and why doing it out of order is rework rather than a tweak is in specifying a screw.

Common questions

How long does ISO 4042 say to bake?

It does not give a number. In the table of measures against internal hydrogen embrittlement, the column for fasteners with a core hardness above 370 HV requires baking, product testing for each manufacturing lot, and that the baking temperature and duration shall be specified. Who specifies them is left to the parties.

Is eight hours at 200 °C enough to remove hydrogen?

That is the conclusion of one 2025 study on 12.9 grade M8 zinc-nickel plated bolts, which compared 4, 8 and 12 hours and reported hydrogen ductility recoveries of 7,19, 11,4 and 11,5 per cent. The gap between eight and twelve hours in those figures is a tenth of a percentage point. Recomputing the recoveries from the elongations printed in the same table gives 7,19, 10,79 and 11,51, so the eight hour row does not reproduce, most likely because the elongation column is rounded. One paper, one coating, one class and one diameter is not a rule for your parts.

Can a tensile test tell me whether a bolt is embrittled?

Not on its own. In that study the unbaked bolts and the twelve hour baked bolts differed by less than one per cent in yield and tensile strength, and the paper says the maximum difference between the strength values did not exceed one per cent. The difference showed up in elongation, in a slow strain rate test run at 1,0 × 10⁻⁶ per second.

Does baking damage the heat treatment of a 12.9 bolt?

The study found no meaningful change. Grain counts and the ferrite to pearlite area ratio were close across the unbaked reference and all three baking times, and it reports that the heat treatment had little to no effect on the microstructure. Yield and tensile strength moved by under one per cent across the same set.

Why does the paper cite ISO 19598 when ISO 4042 dropped it?

ISO 19598 is still a current standard, so citing it is not an error. It is worth noticing, though, that ISO 4042:2022 removed all references to ISO 2081 and ISO 19598, and its introduction says those general electroplating standards are not adequate for electroplated fasteners especially with regard to hydrogen embrittlement and baking, which is the subject of the paper.

References

• Can E., Atagur M., Uzgur N., Tilkioglu D., Uslucan E., Ertugrul O. — Influence of baking time on hydrogen embrittlement in alkaline nickel-zinc electroplated steel fasteners. Indian Journal of Chemical Technology 32 (July 2025) 478–484
• ISO 4042:2018 — the table of measures against internal hydrogen embrittlement and its footnote on alkaline zinc-nickel, read from the free preview
• ISO 4042:2022 — foreword and introduction, for the removal of the ISO 2081 and ISO 19598 references

The paper was read in full, as the published version, from the open-access platform of the publisher; it is not quoted from its abstract. The recomputation of the hydrogen ductility recovery column is this page’s own arithmetic on the elongation figures printed in the paper, not a correction issued by anyone. Two of the three rows reproduce exactly and one does not, and the most likely reason is a rounded elongation column; no other explanation is asserted. The paper does not state how many bolts were tested per condition, so no sample size is given here. Four of its six authors work at a fastener coating company, which is stated in the text rather than left for the reader to find. The ISO 4042 material comes from the free previews of the 2018 and 2022 editions, which carry the foreword, introduction, scope and, in the 2018 edition, the internal hydrogen embrittlement table on page 7; clause B.4, which that table points to, is outside the preview and is not quoted. This page gives no baking time or temperature of its own, for any coating or class. The single study described here is one coating, one property class and one diameter, and nothing in it transfers to other parts without testing them.