Creep and Softening in Hot Work Tool Steel

A hot work tool fails at a temperature where the same steel would be safe in a cold work application, and the reason is that the property limiting it is no longer the yield strength. Above roughly a third of the melting point on the absolute scale, a metal deforms with time under a load it would otherwise carry, and the microstructure that gave the tool its hardness begins to change on its own. This page covers the temperature at which creep becomes the limiting property, the stages of the creep curve and the number used to estimate the life left, the changes that quietly remove hardness from a hot work die, and how a creep fracture differs from one caused by overload.

Where creep becomes the limiting property

Creep begins where the atoms in the lattice become mobile enough to rearrange under load, and the rule of thumb the source gives is that this happens at roughly one third to one half of the melting point measured on an absolute scale. The rule is a starting point rather than a design criterion, because what matters in practice is the temperature at which the strength of the metal becomes limited by creep rather than by its yield strength, and that transition has to be determined for each alloy separately. The two are not the same question, and for a steel they can differ by a wide margin.

For a hot work tool steel the consequence is direct. In the classification below, a chromium hot work steel is a low-alloy steel, and its service temperature in a die casting die or a hot forging die sits above the onset figure in that row. The tool is therefore operating in a regime where the load it can carry depends on how long it has been carrying it, which is why a hot work tool is specified by its hot hardness and its resistance to softening rather than by a room temperature tensile figure.

The temperature at which creep becomes an engineering consideration for each alloy group, expressed both in degrees and as a fraction of the melting point on an absolute scale. Source, Table 1 of the article Creep and Stress Rupture Failures.

The creep curve, and the number that predicts life

The classical creep curve has three stages. Primary creep is the period of adjustment right after loading, when the rate falls as dislocations tangle and harden the material. Secondary creep, also called steady state creep, is an equilibrium between work hardening and recovery, and its minimum rate is the number that is related empirically to the rupture life and used in engineering work on creep resistant materials. Tertiary creep is the accelerating region that ends in fracture. In many commercial creep resistant alloys the tertiary stage starts at a creep strain of 0.5 percent or less, so the design of a component for elevated temperature service rests on the time and the extension that precede it, and the duration of the tertiary stage itself is the margin that allows a failing part to be detected before it breaks.

Two cautions belong with those numbers. The strain measured in a test is not the whole strain in the part, because the total strain at temperature is the sum of the elastic strain, the thermal expansion strain and the creep strain, and in a failure analysis the first two cannot be left out. And creep data are usually gathered under a constant load, where the stress in the gauge section rises as the specimen extends, so a rupture life measured that way is shorter than the same material would give under constant stress. Neither point changes the shape of the curve, but both change the answer when the curve is read back onto a real component.

What changes in the steel at temperature

The microstructure of a part after a long period at temperature rarely resembles the structure it was delivered with, and for a hot work steel the change is the failure. Prolonged exposure brings precipitation of new phases, dissolution or growth of the phases that were wanted, grain growth, and, under the combined influence of stress and temperature, changes that a plain heat treatment cannot reproduce. A carbide that gave the die its hot hardness coarsens into a shape that no longer does. A phase that was finely dispersed agglomerates. A carbide that should have stayed inside the grain moves to the boundary, where a film of it acts as a crack path.

The source is careful about one point that matters when a suspect part is examined. The presence of a phase such as sigma does not by itself prove that the phase caused the failure, because the same phase can strengthen, weaken or do nothing depending on how much of it there is, where it sits and what shape it takes, and because other phases are usually present at the same time. A conclusion of that kind needs metallographic work that separates the phase that caused the failure from the ones that merely came along with it, which is why the microstructure near a fracture has to be compared with the microstructure away from it.

The changes the source lists in a steel held at temperature under load, what each one does to the strength of the part, and where it is found. Compiled from the metallurgical instabilities section of the article.

How a creep fracture looks

Stress rupture fractures are characterised by a multiplicity of creep voids next to the main fracture, and the voids are generally easy to find by optical examination of microsections. Whether the fracture looks ductile or brittle macroscopically depends on the alloy, the temperature and the strain rate, and the pattern is the opposite of what a shop expects. Lower creep rates, longer rupture times and higher temperatures promote intergranular fracture with little elongation or necking, while the transgranular fractures that carry pronounced elongation and necking tend to come from high applied stresses and high strain rates, failing by a void process similar to the dimple rupture of an ordinary tensile test. Fractographs often carry both paths at once, and in those cases the intergranular fissures usually came first, reduced the section, and raised the stress until the rest tore.

The temperature at which the path changes is called the equicohesive temperature. Below it the grain boundary region is stronger than the grains and the fracture runs through the grains, and above it the boundary is weaker and the fracture runs along it. That temperature is not a fixed property, since it moves with exposure time and stress, so the same steel can fail in either way depending on the duty. Two cavity shapes carry the same information on a smaller scale. Round cavities that form along the grain edges are the low stress, long time form, while wedge shaped cavities at grain boundary triple points are promoted by grain boundary sliding at high stress and lower temperature, and they leave a rough fracture surface with identifiable grain boundary precipitates.

For a component still in service the examination has to be done without cutting it up, and the field method the source describes is surface replication, which is the established way of assessing the condition of high temperature plant from its creep damage. Replication is what makes the damage levels in the next table usable, because it converts a cavity population on a polished surface into a number that can be tracked at successive inspections.

Estimating the life that is left

Two approaches are given. The first is the parametric one, and the parameter used most widely is the Larson-Miller parameter, written as the temperature in degrees Rankine multiplied by the quantity C plus the logarithm of the time in hours, with the result scaled by a thousandth. The constant C is about 20 for a low alloy steel. The parameter is used to determine the remaining rupture life at a given applied stress by comparing the service exposure with laboratory data, and it is the reason a creep test can be run for a few thousand hours and extrapolated over a service life measured in years.

The second approach is metallographic, and it is the one that gives a number for a part that is already in service. A cavity population read from a replica is matched to a damage level, and each damage level corresponds to a range of the life fraction already consumed. The remaining life factor is then calculated as the reciprocal of the consumed fraction minus one. Because the same cavity population is read at each inspection, the trend between inspections is more informative than any single reading, and the widest part of the range sits at the lowest damage levels, which is where the estimate is least certain.

Damage level against the fraction of life already consumed, with the remaining life factor calculated as 1/X minus 1. The damage levels are read from the cavity population seen on a replica, which is why a level 1 part can be left alone and a level 5 part has already failed. Source, Table 3 of the article Creep and Stress Rupture Failures.

Rupture ductility is the property that decides whether any of that warning time is real. Component designs are frequently based on a creep strain of one percent, and the source notes that in conventionally cast nickel-base heat resisting alloys a stress rupture ductility of about one percent is common at 760 °C, against values above five percent for the strongest wrought alloys. A low rupture ductility can therefore stop an alloy from being used to its full strength, because the part would break before it had visibly deformed. Where the amount of deformation is not critical, a high rupture ductility is worth choosing for a different reason, which is that the failure then arrives with warning rather than without it. Premature failures have come from exactly that lack of ductility in the tertiary stage.

What this means on a tool steel order

A hot work tool is bought against a service temperature, and the useful specification is the one that says what the steel retains at that temperature rather than what it reaches at room temperature. Four items are worth putting on the sheet. The service temperature and the time at it, because both enter the decision and a die that sees 600 °C for seconds behaves differently from one that sees 500 °C for hours. The tempering temperature, because a die tempered below its service temperature will continue to soften in use. The cleanliness of the steel and the way it was made, since the phase changes that remove hot strength are sensitive to composition control and to the amount of segregation in the bar. And the toughness at temperature, because the intergranular crack that ends the life of a hot work die is the one that grows along the boundaries where the phases have collected. Hot work grades and their cycles are on hot work tool steel heat treatment and H13 steel heat treatment, the repeated heating and cooling failure that usually arrives first is covered in thermal fatigue resistant tool steels, and the physical data a thermal model needs is on hot work steel thermal constants and tool steel thermal conductivity. The way red hardness is held at temperature in the high speed grades is on high speed steel heat treatment, the transformation diagram behind the tempering choice is in tempering and isothermal transformation, and the materials used for each die component are listed in die casting die component materials. Temper embrittlement, which arrives through the same grain boundaries, is collected in types of embrittlement in steels, and the stage by stage causes of a tool or die failure are set out in tool and die failure analysis.

Before you act on a creep assessment

A reference page, it is not an Aobo Steel specification, and the temperatures and limits above are the values published for the alloy groups and components named. Creep behaviour depends on the alloy, the stress, the temperature, the time and the environment together, and the onset figures are engineering rules of thumb rather than design limits for a particular tool. A damage level read from a replica is a reading of one small area, and a single inspection cannot distinguish damage that has grown slowly from damage that has just begun. Where a hot work die is being kept in service on the strength of a remaining life estimate, the estimate belongs with the inspection interval it came from.

Source: ASM Handbook, Volume 11, Failure Analysis and Prevention, ASM International, 2002.