Physical Metallurgy of Steel

Class Notes and lecture material for MSE 661 (formerly 651.01)
Notes revised by: Robert H. Wagoner, Smith Chair
Notes originally compiled by: Glyn Meyrick, Professor Emeritus
Web installation by: Wei Gan, Graduate Research Associate
Last revision date: 9/27/02
SECTION 1 - INTRODUCTION
Foreword
This document is intended to augment formal lectures on the general topic of the physical
metallurgy of steels, presented within the MSE Department at the Ohio State University.
This document is not intended for publication and is restricted for use in MSE 661
(formerly MSE 651.01).
These notes are based on a variety of texts and published articles and also on the personal
experience of Professors Wagoner and Meyrick. Specific references to sources are made
within the document. However, the material is often in the form of knowledge that has
been accumulated by the work of many people and is "well-known" by experts in the
field. A detailed acknowledgment of the work of each contributor to the field is not
attempted.
Source Material
The following sources are referenced frequently in these notes, and an annotated
bibligraphy is available on the course web site under “Closed Reserve.” As indicated,
these books are on closed reserve at the Science and Engineering Library and may be
consulted in situ by any student in MSE 661. The following books will sometimes be
referred to in brief form, for example “Bain and Paxton,” or “Leslie’s book.”
1. Edgar C. Bain and Harold W. Paxton: Alloying Elements in Steel, 2
nd
edition,
American Society for Metals, 1966. (OSU: TN756 B16 1966, Lib. Congress: 63-
29304)
2. George Krauss: Steels: Heat Treatment and Processing Principles, ASM
International, 1990. (OSU: TN751 .K73 1990, ISBN: 087170370X)
3. R. W. K. Honeycombe and H. K. D. H. Bhadeshia, Steels: Microstructure and
Properties, 2
nd
edition, Edward Arnold, 1995. (OSU: TN693.I7 H65 1995, ISBN/1
st
edition: 0713127937).
R. H.Wagoner MSE 661 – Steel Section 1, Page 2
4. Carbon and Alloy Steels (an ASM Specialty Handbook), ed. J. R. Davis, 1996, ASM
International. (OSU: TA479.C37 C37 1996, ISBN: 0-87170-557-5).
5. Stainless Steels (an ASM Specialty Handbook), ed. J. R. Davis, 1994, 1996, ASM
International (no OSU number, ISBN 0-87170-503-6).
6. George Roberts, George Krauss, Richard Kennedy: Tool Steels, 5
th
edition, ASM
International, 1998. (OSU: TS320 .R6 1998, ISBN: 0-87170-599-0).
7. Atlas of Time-Temperature Diagrams for Irons and Steels, ed. George F. Vander
Voort, ASM International, 1991. (OSU: TN693 I7 A88 1991, ISBN: 0-87170-415-
3)
8. William D. Callister, Jr.: Materials Science and Engineering: an Introduction, 5
th
edition, Wiley, 2000. (OSU: TA403 .C23 2000, ISBN: 0471320137)
9. William C. Leslie: The Physical Metallurgy of Steels, McGraw-Hill, 1981. (OSU:
TN693.I7 L45, ISBN: 0070377804)
10. Fundamentals of Dual-Phase Steels, ed. R. A. Kot and B. L. Bramfitt, The
Metallurgical Society of AIME, 1981. (OSU: TN 693 I7 F86, ISBN: 0-89520-383-
9).
11. Case-Hardened Steels: Microstructural and Residual Stress Effects, ed. Daniel E.
Diesburg, The Metallurgical Society of AIME, 1984. (OSU: TN 751 C37 1984,
ISBN: 0-89520-471-1)
12. The Making, Shaping and Treating of Steel, 9
th
edition, ed. Harold E. McGannon,
United States Steel, 1971. (OSU: TN730 .U55 1971, Libr. Cong: 76-30310)..
13. Hardenability Concepts with Applications to Steel, eds. D. V. Doane and J. S.
Kirkaldy, The Metallurgical Society of AIME, 1978. (OSU: TN693 I7 H37, ISBN:
0-89520-141-0)
14. Karl-Erik Thelning: Steel and Its Heat Treatment, 2
nd
edition, Butterworths, 1984
(OSU: TN693 I7 T47 1984, ISBN: 0-408-01424-5)
15. Jonathan Beddoes and J. Gordon Parr: Introduction to Stainless Steels, 3
rd
edition,
ASM International, 1999. (OSU: TA479.S7 B4 1999, ISBN: 0-87170-673-3)
R. H.Wagoner MSE 661 – Steel Section 1, Page 3
Definition of Steel
Steel is a family of materials that is derived from ores that are rich in iron, abundant in
the Earth’s crust and which are easily reduced by hot carbon to yield iron. Steels are very
versatile; they can be formed into desired shapes by plastic deformation produced by
processes such as rolling and forging; they can be treated to give them a wide range of
mechanical properties which enable them to be used for an enormous number of
applications. Indeed, steel is ubiquitous in applications that directly affect the quality of
our lives. Steel and cement constitute about 90% of the structural materials that are
manufactured (Westwood, Met and Mat Trans, Vol. 27 A, June 1996, 1413).
What, then, is steel?
A precise and concise definition of steel is not an easy thing to present because of the
very large variety of alloys that bear the name. All of them, however, contain iron. We
might reasonably begin by describing a steel as an alloy which contains iron as the major
component. This is only a beginning because there are alloys in which iron is the major
constituent, that are not called steels; for example, cast irons and some superalloys. The
major difference between a cast iron and a steel is that their carbon contents lie in two
different ranges. These ranges are determined by the maximum amount of carbon that can
be dissolved into solid iron. This is approximately 2% by weight (in FCC iron at 1146
°C). Steels are alloys that contain less than 2% carbon. Cast irons contain more than 2 %
carbon. Many steels contain specified minimum amounts of carbon. This does not mean
that all steels must contain substantial quantities of carbon. In some steels the carbon
content is deliberately made very small and, also, the amount actually in solution is
reduced further by the addition of alloying elements that have a strong tendency to
combine with the carbon to form carbides.
Steels can be divided into two main groups; plain carbon steels and alloy steels. The latter
can then be subdivided into many groups according to chemistry ( e.g. standard low alloy
steels), applications (e.g. tool steels ) or particular properties (e.g. stainless steels) etc. Let
us begin with plain carbon steels; this group is the simplest to understand and it
comprises steels that are used in the greatest tonnage.
Plain carbon steels
A plain carbon steel is essentially an alloy of iron and carbon which also contains
manganese and a variety of residual elements. These residual elements were either
present within the raw materials used in the production process e.g. iron ore and scrap
steel additions, or they were added in the production process for a specific purpose, e.g.
deoxidization by means of silicon or aluminum. Hence they are called residual elements
to distinguish them from alloying elements that are deliberately added according to
specified minimum amounts. The American Iron and Steel Institute (AISI) has defined a
plain carbon steel to be an alloy of iron and carbon which contains specified amounts of
Mn below a maximum amount of 1.65 wt. %, less than 0.6 wt. % Si, less than 0.6 wt. %
R. H.Wagoner MSE 661 – Steel Section 1, Page 4
Cu and which does not have any specified minimum content of any other deliberately
added alloying element. It is usual for maximum amounts (e.g. 0.05 wt. %) of S and P to
be specified. ( see The Making, Shaping and Treating of Steel, USS 1970)
Steelmaking, Briefly
The manufacture of steel is a subject that is well beyond the ambit of these notes. But we
should be aware that various manufacturing practices can affect the oxygen, nitrogen and
sulfur content and hence the cleanliness of the product. Here is a very brief accounting of
the principles of iron and steel making.
Iron Making - For simplicity, let’s consider the critical temperatures for pure iron by
referring to the iron-carbon phase diagram, Figure 1-1. Up to 910 deg. C, iron has the
alpha, or BCC crystal structure. From 910 to 1390 deg. C, the gamma or FCC crystal
structure is stable. From 1390 to 1537 deg. C the delta, or high-temperature BCC phase
is stable, and above 1537 deg. C (about 2800 deg F), the metal is a liquid.
Figure 1-1: The iron-carbon phase diagram. Equilibrium lines are solid
and metastable lines are shown as dashed.
Graphite
+
Liquid
Liquid
γγγ γ +++ + Graphite (Stable Phase)
γγγ γ +++ + Cementite (Metastable Phase)
ααα α +++ + Graphite
ααα α +++ + (Cementite)
ααα α +++ + γγγ γ
γγγ γ
γγγ γ +++ + Liquid
δδδ δ
δδδ δ +++ + Liquid
1400
1200
1000
800
600
0.52 %
0.16 %
0.11
0.020%
0.021 %
0.76 %
0.68 %
2.01 %
2.04 % 1146 C
4.27 %
0 1 2 3 4 5
Carbon, wt. %
1499 C
1152 C
910 C
738 C
727 C
1537 C
1390 C
T
e
m
p
e
r
a
t
u
r
e
,


C
R. H.Wagoner MSE 661 – Steel Section 1, Page 5
Iron is made from various kinds of iron ores, but any one can be selected to illustrate the
principles. Let us say that the principal ore is of the form of FeSO. Very high
temperatures and very reducing atmospheres are are needed to break this down because
the sulfide/oxide it is much more stable than the native metal under normal conditions.
Furthermore, in order to get reasonable transport rates inside the metal, it is clear that it
must be a liquid.
It is very difficult to achieve the melting temperature of iron. Gasoline fires won’t do it.
Propane, butane and so on won’t do it. Instead, a very exothermic reaction must be used.
In a blast furnace, the iron ore is reduced by “burning” coke, a pure and highly energetic
form of carbon:
FeSO (solid) + C (solid) -> Fe(liquid) + CO (gas) + S (in solution) (1-1)
At the end of the iron making, the products are iron liquid saturated (or nearly so) with
carbon, sulfur, and nitrogen. (The nitrogen is carried in with the air injected into the blast
furnace, although we haven’t shown that here explicitly.) There are also whatever
“tramp” elements may have come along with the ore and possibly oxygen, depending on
how completely the iron ore was reduced.
Steel Making – Once the liquid iron is obtained with its undesirable “passengers,” steel
making involves removing these impurities, mainly carbon and sulfur. The trick here is
almost the opposite of the iron making. Oxygen is bubbled through the liquid iron and it
combines with carbon dissolved in the melt to produce carbon monoxide:
O (in solution) + C (in solution) ->CO (1-2)
A similar reaction may be written for nitrogen, which is removed at the same time. Also,
sulfur in solution is attracted to the O / CO bubbles rising through the melt and is carried
out to the top of the melt, where it is trapped by a molten salt slag on the surface. After
just the right amount of time of this process (before significant iron is again oxidized
back to its ore form!), the basic oxygen process has the products:
Fe (liquid) + O (in solution) + C (v. small) + S (v. small) + slag (silicates) (1-3)
The slag is usually removed by either pouring it off, or tapping the steel from the bottom,
although inevitably some of it remains entrained in the melt because of the violent
bubbling action as the oxygen combines with carbon.
The final step in producing steel product is the casting and cooling of the molten product
shown in Equation 1-3. The problem here is that there is an abundance of oxygen
dissolved in the molten iron and the solubility limit of oxygen in solid iron at lower
temperatures is much lower. Therefore, the oxygen (and nitrogen, which we’ve been
ignoring) will form pores (i.e. porosity) in the final solid steel, which is very undesirable.
For normal, old-fashioned batch casting (i.e. pouring into a mold and letting it cool, then
R. H.Wagoner MSE 661 – Steel Section 1, Page 6
pouring another batch into another mold, etc.), the quality of the steel was often referred
to by the method of controlling porosity from oxygen and nitrogen. Here are the
traditional choices:
• Killed Steel – For fairly high-carbon steels (greater than 0.3%) Si and Al are added
in small amounts to tie up the remaining oxygen as Al2O3 or SiO2. Similar
reactions occur for nitrogen. This is called “killing” because adding aluminum or
nitrogen to a violently bubbling steel cauldron (ladle) effectively stops the bubbling
by tying up the available oxygen (thus “killing” the ladle bubbling).
• Semi-killed Steel – Less Al or Si is added to remove just part of the oxygen and
nitrogen, leaving some in solution.
• Rimmed steel – The ingot is allowed to cool only slowly so that oxygen bubbles
have time to migrate to the surface of the ingot before solidification takes place.
The surface (or “rim”) of the ingot is cut off to get rid of the highly porous areas.
This works only partly and leaves a fair amount of porosity throughout.
• Capped steel – The ingot is kept warm by insulating the top so that freezing
proceeds from the bottom, thus promoting the oxygen bubbles to migrate to the top
of the ingot. The top or “cap” of the ingot is then cut off. This is the cheapest
method (because only the end of the ingot must be thrown away) but removes the
least oxygen.
Discussion of Steelmaking Consequences - The term “cleanliness” usually refers to the
amounts of various phases such as oxides, sulfides and silicates that can be present in
steel. The smaller the amount of these phases, the cleaner is the steel. These phases are in
the form of particles called inclusions and they can have significant effects on the
properties of steel and are normally undesirable. Thus, variations in the manufacturing
process can have significant effects on the properties of the steel, so some brief
comments are in order.
There have been major advances in the production of steel during the last 20 years and
continuous casting, in which great attention is being paid to the cleanliness of the steel,
has become the dominant production method. According to R. D. Pehlke in
Metallurgical Treatises, published by the Metallurgical Society of the AIME and
reprinted in 1983, most steel produced by continuous casting is fully killed. Vacuum
deoxidization is also being used to eliminate the oxygen (and sulfur), and the steel is
protected by argon atmospheres in covered tundishes thereby yielding a cleaner steel with
respect to its inclusion content. This is beneficial with respect to the mechanical
properties and uniformity of the final product. Continuous casting also produces a
product that is much closer to the final shape that is required for the application of the
steel.
More and more steel sheet is now being produced in what are called mini-mills. These
mills charge electric arc furnaces with scrap steel and iron, melt the charge, control the
R. H.Wagoner MSE 661 – Steel Section 1, Page 7
composition, and, via a tundish continuously cast it into slabs several inches thick and a
few feet wide. This slab is immediately fed through a long furnace in which it is heat
treated and from which it emerges at the temperature desired to begin hot-rolling and
passes directly into a series of hot rolls. Problems can arise in controlling the residual
element contribution to the composition of the product because of variations in the
quality of the scrap steel charged into the electric furnaces. Copper is particularly
undesirable because it is not easily removed from liquid steel, and as its concentration
increases, it can produce cracks due to hot-shortness. This phenomenon arises because
when the hot steel is exposed to air the iron is oxidized to form scale, much of which
spalls off but the copper is not oxidized and accumulates beneath the scale at the surface
of the steel. The copper penetrates the steel along grain boundaries and causes grain
boundary cracking to occur during rolling.
Classification of Plain Carbon Steels
The main numbering systems for plain-carbon steels are either the S.A.E. / A.I.S.I.
system or the unified numbering system (UNS). The first two systems are very similar
and use a numerical code of 10xx and 10xxx. The more modern, unified numbering
system uses G 10xxx etc. In these codes the 10 indicates that the steel is a plain carbon
steel while the xxx's indicate a range for the carbon content . Typically compositions run
from 1005 to 1095 but 10125 would have 1.25 wt.% C.
Note the three steels 1044, 1045 and 1046 in the A.I.S.I. system. Here the first two digits
inform us that the steels are all plain carbon steels. The xx' s are now 44, 45 and 46.
Because these digits are used to indicate the carbon content within a range of a few
points, it would appear that the steels are indistinguishable. With respect to the carbon
content this is true. Each has 0.43 to 0.5 wt.% C as the specified range. However, the
ranges for Mn differ, being 0.3 to 0.6, 0.6 to 0.9 and 0.7 to 1.00 wt.%. In fact, it is
possible for one steel to satisfy the specifications for more than one standard steel. The
point to note is that when the carbon ranges are the same, the higher Mn limit
corresponds to the larger xx. Generally speaking, the numerical code gives you a fair
idea of what the steel is. Then it is sensible to have access to tabulated compositions in
Metal Handbooks, rather than to attempt to memorize many details. Table 1-1 is an
example of such classification.
For the sake of completeness we should note that carbon steels with more than 1.00 %
Mn are identified by replacing the first 0 with 5 i.e. 15xx. There is a group of steels that
possess more than 1 wt. % Mn and also have a minimum range of Si specified as between
0.15 and 0.35 wt. %. This is 13xx and its members, strictly speaking, are alloy steels.
Also, whereas it is usual to keep the sulfur and phosphorus contents low to avoid
embrittlement problems that they can cause, there are occasions when ductility is not a
major concern and machinability is more important. In these cases, S and P are
deliberately added because they endow the steel with good machinability properties .
Hence we have the 11xx and 12xx series for which the first is resulfurized and the second
is enhanced in both S and P, e.g. 1146 has 0.08 to 0.13 S and 1215 has 0.26 to 0.35 S and
.04 to 0.09 P. Finally, some steels also have Pb (also for improving machinability)
R. H.Wagoner MSE 661 – Steel Section 1, Page 8
additions whereupon an L is included e.g. 12L14. Typically, the lead content lies
between 0.15% and 0.35%.
Table 1-1 - Chemical Composition Limits for Standard Plain-Carbon Steels.
Chemical Composition Limits (Ladle Analysis), Per Cent ( 3 )
SAE Corresponding
Number C Mn P, max. S, max. AISI Number
1005 0.06 max. 0.35 max. 0. 04 0. 05 -
1006 0.08 max. 0.25-0.40 0. 04 0. 05 1006
1008 0.10 max. 0.30-0.50 0. 04 0. 05 1008
1010 0.08-0.13 0.30-0.60 0. 04 0. 05 1010
1011 0.08-0.13 0.60-0.90 0. 04 0. 05 -
1012 0.10-0.15 0.30-0.60 0. 04 0. 05 1012
1013 0.11-0.16 0.50-0.80 0. 04 0. 05 -
1015 0.13-0.18 0.30-0.60 0. 04 0. 05 1015
1016 0.13-0.18 0.60-0.90 0. 04 0. 05 1016
1017 0.15-0.20 0.30-0.60 0. 04 0. 05 1017
1018 0.15-0.20 0.60-0.90 0. 04 0. 05 1018
1019 0.15-0.20 0.70-1.00 0. 04 0. 05 1019
1020 0.18-0.23 0.30-0.60 0. 04 0. 05 1020
1021 0.18-0.23 0.60-0.90 0. 04 0. 05 1021
1022 0.18-0.23 0.70-1.00 0. 04 0. 05 1022
1023 0.20-0.25 0.30-0.60 0. 04 0. 05 1023
1025 0.22-0.28 0.30-0.60 0. 04 0. 05 1025
1026 0.22-0.28 0.60-0.90 0. 04 0. 05 1026
1029 0.25-0.31 0.60-0.90 0. 04 0. 05 -
1030 0.28-0.34 0.60-0.90 0. 04 0. 05 1030
1035 0.32-0.38 0.60-0.90 0. 04 0. 05 1035
1037 0.32-0.38 0.70-1.00 0. 04 0. 05 1037
1038 0.35-0.42 0.60-0.90 0. 04 0. 05 1038
1039 0.37-0.44 0.70-1.00 0. 04 0. 05 1039
1040 0.37-0.44 0.60-0.90 0. 04 0. 05 1040
1042 0.40-0.47 0.60-0.90 0. 04 0. 05 1042
1043 0.40-0.47 0.70-1.00 0. 04 0. 05 1043
1044 0.43-0.50 0.30-0.60 0. 04 0. 05 -
1045 0.43-0.50 0.60-0.90 0. 04 0. 05 1045
1046 0.43-0.50 0.70-0.90 0. 04 0. 05 1046
1049 0.46-0.53 0.60-0.90 0. 04 0. 05 1049
1050 0.48-0.55 0.60-0.90 0. 04 0. 05 1050
1053 0.48-0.55 0.70-1.00 0. 04 0. 05 -
1055 0.50-0.60 0.60-0.90 0. 04 0. 05 1055
1060 0.55-0.65 0.60-0.90 0. 04 0. 05 1060
1064 0.60-0.70 0.50-0.80 0. 04 0. 05 1064
1065 0.60-0.70 0.60-0.90 0. 04 0. 05 1065
1069 0.65-0.75 0.40-0.70 0. 04 0. 05 -
1070 0.65-0.75 0.60-0.90 0. 04 0. 05 1070
1074 0.70-0.80 0.50-0.80 0. 04 0. 05 1074
1075 0.70-0.80 0.40-0.70 0. 04 0. 05 -
1078 0.72-0.85 0.30-0.60 0. 04 0. 05 1078
1080 0.75-0.88 0.60-0.90 0. 04 0. 05 1080
1084 0.80-0.93 0.60-0.90 0. 04 0. 05 1084
1085 0.80-0.93 0.70-1.00 0. 04 0. 05 -
1086 0.80-0.93 0.30-0.50 0. 04 0. 05 1086
1090 0.85-0.98 0.60-0.90 0. 04 0. 05 1090
1095 0.90-1.03 0.30-0.50 0. 04 0. 05 1095
NOTE: LEAD. When lead is required as an added element to a standard steel, a range of 0.15 t o
R. H.Wagoner MSE 661 – Steel Section 1, Page 9
Classification of General and Alloy Steels
It follows from the definition of a plain carbon steel, that if one or more additional
elements are deliberately added to produce specified minimum contents then the product
is an alloy steel. In general, this is so, but we should also note that small additions of rare
earth elements such as cerium can be made to a plain carbon steel for inclusion control.
One of the important roles of manganese is to combine with sulfur to prevent the
formation of iron sulfides which can embrittle the steel. It has been discovered that
additions of rare earth elements can produce high-melting point, complex sulfides and
oxysulfides that are smaller and less harmful to the mechanical properties of steels than
manganese sulfides can sometimes be.) The steel would not then be referred to as an
alloy steel.
The definition given is a very broad one and it indicates that a clear, concise, nice little
sub-division scheme to describe all steels is not easily produced. As we shall see, there is
a group of low-alloy steels for which the compositions are specified in this country
according to schemes originally recommended by the AISI and the Society of
Automotive Engineers(SAE), but alloy steels are also classified on the basis of some
important property ( e.g. stainless steels) or on the basis of use (e.g. tool steels) or even as
a consequence of a particular heat-treatment (e.g. maraging steel).
To add to the confusion, there are steels that have particular names coined by the
manufacturer and also steels that are characterized by an American Society for Testing
Materials (ASTM) specification, e.g. A306 which specifies the mode of manufacture,
permissible amounts of S and P but leaves the manufacturer to choose the carbon levels
necessary to achieve specified tensile properties. The tensile properties determine the
grade of steel, e.g. A306 grade 50 specifies a tensile strength of 50 to 60 KSI. The
picture appears to be quite confusing. Our objective is to eliminate the confusion by
understanding how the properties of steel are related to their microstructures, how
particular microstructures are produced and what effects are produced by the alloying
elements.
There is, of course, a wide range of properties. Mostly we will concentrate on
mechanical properties. The ASM. handbooks and the book "Making, Shaping and
Refining of steel" put out by the old U. S.. Steel Co. and in a revised tenth edition in
1985 by W. T.Lankford, Jr. et al. published by the Association of Iron and Steel
Engineers, present many details of properties and production processing. Our purpose is
to develop a fundamental understanding. In order to do this, I propose to begin with pure
iron, proceed to Fe-C, consider plain carbon steels, put in alloying elements and finally to
select some particular classes of steel for examination.
R. H.Wagoner MSE 661 – Steel Section 1, Page 10
Table 1-2 – Common Classifications of Irons and Steels
Class Distinguishing Features Typical Uses Source of Strengthening
Cast Iron More than 2% C Pipes, valves, gears, Ferrite-pearlite structure
and 1 to 3% Si engine blocks as affected by free graphite
Plain-carbon Principal alloying Structural and machine Ferrite-pearlite structure i f
steel element is carbon parts low carbon; quenching and
up to 1% tempering if medium t o
high carbon
Low-alloy Metallic elements High strength Grain refinement, precipitation,
steel totaling up to 5% structural and and solid solution if low
machine parts carbon; otherwise quenching
and tempering
Stainless At least 10% Cr; Corrosion resistant Quenching and tempering i f
steel does not rust piping and nuts and <15% Cr and low Ni; otherwise
bolts; turbine blades cold work or precipitation
Tool steel Heat treatable to Cutters, drill bits, Quenching and tempering, etc.
high hardness dies
and wear resistance
After: Mechanical Behavior of Materials, N.E.Dowling, Prentice-Hall, 1999, pp.55, 56
Table 1-3 – Typical Irons and Steels
Principal Alloying Elements, Typical % by Weight
Description Identification UNS No. C Cr Mn Mo Ni Si V Other
Ductile ASTM A395 F32800 3. 50 - - - - 2 - -
cast iron
Low-carbon AISI 1020 G10200 0. 2 - 0. 45 - - 0. 2 - -
steel
Medium-carbon AISI 1045 G10450 0. 45 - 0. 75 - - 0. 2 - -
steel
High-carbon AISI 1095 G10950 0. 95 - 0. 4 - - 0. 2 - -
steel
Low-alloy AISI 4340 G43400 0. 40 0. 8 0. 7 0. 25 1. 8 0. 2 - -
steel
HSLA steel ASTM A588-A K11430 0. 15 0. 5 1. 1 - - 0. 2 0. 05 0.3 Cu
Martensitic AISI 403 S40300 0. 15 1 2 1. 0 - 0. 6 0. 5 - -
stainless steel
Austenitic AISI 310 S31000 0. 25 2 5 2. 0 - 2 0 1. 5 - -
stainless steel
Precipitation 17-4 PH S17400 0. 07 1 7 1. 0 - 4 1. 0 - 4 Cu
hardening 0.3 (Nb+Ta)
stainless steel
Tungsten AISI T1 T12001 0. 75 3. 8 0. 25 - 0. 2 0. 3 1. 1 18 W
high-speed
tool steel
18 Ni ASTM A538-C K93120 0. 01 - - 5 1 8 - - 9 Co, 0.7 Ti
maraging steel
After: Mechanical Behavior of Materials, N.E.Dowling, Prentice-Hall, 1999, pp.55, 56
R. H.Wagoner MSE 661 – Steel Section 1, Page 11
The Strength of Iron – Effect of Temperature
At normal atmospheric pressure, iron, in equilibrium, is BCC below 910°C, FCC
between 910°C and 1390°C and then BCC up to its melting point of 1536°C. Hence, at
normal ambient temperatures and below, iron is BCC. Its mechanical properties, as
determined by testing in tension at room temperature and at typical strain rates of 10
-3
to 10
-4
sec
-1
, are functions of purity, grain size and dislocation content. The yield
strength is also temperature dependent, increasing with a reduction in temperature as
shown in Figure 1-2 (W.C. Leslie, Met. Trans. 3, 9, 1972.) These results were obtained
with iron to which Ti had been added to getter the interstitial elements. ( For example, the
carbon that combines with titanium is removed from solid solution. It has, therefore been
“got” by the titanium which explains why the verb “to getter” was coined.)
Figure 1-2: Dependence of yield stress on temperature for pure iron. (After Leslie:
Metall. Trans. 1972, vol. 3:9.)
Above 300 K, dislocation glide can occur relatively easily so that the magnitude of the
macroscopic yield stress is controlled by the effects of long range internal stress fields on
dislocation motion. Therefore, the temperature dependence of the yield stress in this
0
1 0
2 0
3 0
4 0
5 0
6 0
7 0
8 0
9 0
100
110
0 100 200 300 400 500 600 700 800 900 1000
- 200 - 100 0 100 200 300 400 500 600 700
0
100
200
300
400
500
600
700
T
r
u
e

s
t
r
e
s
s
,


k
s
i
Temperature, K
Temperature, deg. C
M
P
a
Thermal
activation
dominant
Peierls
dominated
behavior
εεε ε=0 . 1 6
εεε ε=0 . 0 8
εεε ε=0 . 0 4
εεε ε=0 . 0 2
εεε ε=0. 002
Lower
Yield
Stress
Fe-0. 15%Ti
gs = ASTM 5-6
de/dt = 0.00025/s
R. H.Wagoner MSE 661 – Steel Section 1, Page 12
temperature realm is due to the temperature dependence of the shear modulus of iron
which decreases gradually as the temperature is increased and vice versa.
From 250 K downwards, the variation of the macroscopic yield stress with temperature
increases linearly and much more rapidly than the variation above room temperature. The
rapid increase in the flow stress at low temperatures appears to occur because the atomic
displacements involved in the movement of individual dislocations by glide becomes
difficult very rapidly. As the yield stress increases, the ductility decreases, and, if the
strain rate is also increased, dislocation glide is replaced by twin formation and brittle
fracture by cleavage can occur. The temperature where brittle behavior replaces ductile
behavior depends upon the strain rate. Dislocation glide can still occur at temperatures as
low as 4.2 K when the strain rate is small.
Peierls Stress – Explanation of Temperature Effect
In order to understand why BCC iron shows this two-stage temperature dependence, one
must refer to the “Peierls Stress”, sometimes called the “Peierls-Nabarro Stress.” The
Peierls Stress is the stress required to move a dislocation through a perfect crystal lattice.
It arises from the strength of the local atomic bonding, the crystal structure of the lattice,
and the characteristics of the local dislocation configuration (slip plane, direction, burgers
vector).
If the Peierls stress is high (particularly relative to thermal energy, i.e. temperature), then
it represents the principal obstacle to dislocation motion. I such cases the dislocations are
termed “hard” because they remain rigidly linear, lying in the crystalline directions the
represent the lowest energy. (They are also “narrow” meaning that their cores are very
local, in order to minimize their energy in view of the very sharp energy well they reside
in.)
If the Peierls stress is low relative to other stresses or temperature, then the dislocation is
termed “soft” because the dislocation lines are curved according to other obstacles. (The
cores in this case are more diffuse, or “wide.”) Thedislocations are not constrained to be
rigid rods lying in special crystallographic directions. In this case, it is the long-range
obstacles in the crystal that represent the most serious strengthening agents. Such
obstacles might include grain boundaries, other dislocations, and second-phase particles.
It turns out that the role of crystal structure on the Peierls stress can be understood
relatively easily, without needing to invoke a complex calculation:
• For highly-directional or strong crystals, the Peierls Stress is very high at most
practical temperatures, the dislocations are hard, and the materials themselves are
usually strong and brittle. Examples include Si, Ge or Al2O3.
• FCC materials, such as Au, have wide, soft dislocations at nearly all temperatures.
These materials are strengthened primarily by long-range obstacles and dislocations
R. H.Wagoner MSE 661 – Steel Section 1, Page 13
readily adapt shape depending on their surroundings. These materials can be soft
and ductile, or they can be hardened by long-range interactions.
• BCC materials, contrary to the other two cases, have a Peierls stress that is highly
temperature dependent. At low temperatures, Peierls stress is very high, the
dislocations are hard and narrow, and the material is strengthened primarily by the
crystal structure, and less so by long-range obstacles. At higher temperatures, the
Peirls stress becomes similar to FCC materials, the dislocations become wider and
more flexible. The principal strengthening mechanism at higher temperatures
become long-range obstacles.
It may be inferred from this discussion that BCC materials may undergo a transition from
ductile-to-brittle and from weak-to-strong at some characteristic temperature. In real
alloys, this transition (as discussed later) will depend on alloy content and other factors,
but the origin should be understood to lie with the Peierls stress and its behavior as a
function of temperature.
Role of Solutes in Iron
The foregoing description of the strength of BCC iron at low temperature became clear
only after Leslie recognized that the presence of impurities (whether intentional or not)
confounds the picture. This is because small amounts of additions change the flow stress
or iron considerably, thus making it difficult to reveal the role of temperature on the BCC
iron itself.
For example, zone refined iron of 99.999% purity can be produced in small quantities but
it has no commercial use because it is so soft. Typical numbers show yield and ultimate
tensile strengths of 7000 psi (48 MPa) and 24000 psi (165 MPa) with 40% elongation.
Commercially-pure “iron”, for example “Armco Iron” has typically: 0.012% C, 0.02%
Mn, 0.05% P, and 0.025% S. These modest amounts of impurities provide significant
strengthening without serious, adverse effects upon ductility. The typical yield and
ultimate strengths for Armco Iron are 20000 psi and 45,000 psi, with an elongation of
28%. Thus, a total alloy content of less than 0.1% changes the yield strength by a factor
3, the ultimate tensile strength by a factor of 2 and the elongation by 50%!
With this result in mind, it is clear that an understanding of the role of solutes in iron is
critical for understanding the mechanical properties of commercial alloys.
Interstitial Solutes - Description
See also Chapter 2 in Leslie's book and chapter 1 in "Principles of Heat Treatment of
Steel by G. Krauss, ASM, 1980.
FCC crystals have octahedral interstitial sites at coordinates 0.5, 0.5, 0.5 and the
equivalent 0.5,0,0 positions in the unit cell, Figure 1-3. These are symmetric sites; i.e. in
the hard sphere-central force model for metallic crystals, the centers of these sites are
R. H.Wagoner MSE 661 – Steel Section 1, Page 14
equidistant from the centers of the nearest neighboring atoms. Using 2.52 angstroms for
the diameter of FCC iron atoms ( 1 angstrom =

10
-10
meters) we find that 1.04 angstroms
is the diameter of a sphere that will just fit in the octahedral site. We do this by noting
that if the lattice parameter is “a” then the length of the diagonal of a face of the unit cell
is a 2 . This is also equal to 2D
Fe,
where D
Fe
is the diameter of an iron atom that is
assumed to behave as a hard sphere. Thus we can evaluate “a” and then the diameter of
the octahedral hole which is equal to ( a - D
Fe
). The tetrahedral sites are at 0.25, 0.25,
0.25, for which we find the diameter of the hole to be 0.27 Angstroms. Because of this
difference in size, the interstitial solutes C and N ( of which C has a radius of about 0.8
angstroms and N one near 0.7 angstroms) fit much more easily into the octahedral sites in
the FCC structure than in the tetrahedral sites.
Figure 1-3: Interstitial sites in FCC and BCC iron. In FCC, the octahedral sites are
preferred, in agreement with a simple hard-sphere calculation of hole size. In BCC, the
octahedral sites are also preferred, even though a hard-sphere calculation reveals smaller
holes. Why?
OCTAHEDRAL TETRAHEDRAL
BCC
a 3 2
a / 2
a / 2
a 5 4
a 3 2
a
FCC
a 3 4
a / 2
a / 2
a / 2
Metal Atoms Interstitial Sites
R (A) D(A)
Fe: 1.26A 2.52 A
C: 0.8 A 1.6 A
N: 0.7A 1.4 A
d
o
c
t
=
0
.
3
9

A
(
n
o
t

s
p
h
e
r
i
c
a
l
)
d
o
c
t
=
1
.
0
4

A
d
t
e
t
=
0
.
7
2

A
d
t
e
t
=
0
.
5
6

A
R. H.Wagoner MSE 661 – Steel Section 1, Page 15
BCC sites are also octahedral and tetrahedral, but the octahedral site is not symmetric
because two of the neighboring atoms are closer to the center of the site than the others.
The sites are at the coordinates 0.5, 0.5, 0 (and the equivalent 0.5,0,0 ) and 0.5, 0.25, 0,
respectively. Using the same diameter as before for an iron atom we find interstitial hole
radii of about 0.19 and 0.36 Angstroms. Even though the octahedral holes are smaller
than the tetrahedral holes the interstitial solutes reside in the octahedral holes because
then they only have two nearest neighbors to displace markedly.
Effects of Interstitials
Solubility - The first consequence is that the solubility of C and N is greater in FCC iron
than in BCC iron because the interstitial holes are larger in the FCC structure, even
though the hard sphere model predicts the FCC structure to be the more close-packed.
We will see that this is a very major consequence that has far reaching effects in ferrous
metallurgy.
Strain - Dislocations create elastic strain fields in the crystal around them and generate
major atomic displacements at and near their cores where the displacements cannot be
described elastically. Consequently, because the interstitial solutes, C and N, are larger
than the interstitial sites within the crystalline matrix they prefer to occupy sites along the
dislocation cores and also those matrix sites that are located in places where the elastic
strain field of the dislocation helps to overcome the size misfit.
Segregation - When BCC iron is annealed and slowly cooled, the interstitial solutes are
able to diffuse to and occupy, preferentially, those sites that they prefer in the elastic
strain fields. Thus the solute is not randomly dispersed throughout the iron but is
segregated at preferred sites around the dislocations and along dislocation cores.
Furthermore, this is true for dislocations of any character because the octahedral sites in
the BCC structure are non symmetric so that the solutes cause tetragonal distortions by
displacing two nearest neighbors more than the other four. This enables strong
interactions with the shear strain fields of screw components as well as with the strain
fields due to the edge components of dislocations.
Interactions with Dislcoations - The interactions between interstitial solute elements
and dislocations in BCC iron at ambient temperatures are powerful. When the material is
treated so that segregation takes place as, for example, by aging at temperatures that
permit the solutes to diffuse, dislocations are pinned and many become immobile because
moving the dislocation away from the segregated solute requires a very large stress.
When a tensile test is conducted on a sample of such material at room temperature, most
of the existing dislocations do not glide at yield. When plastic deformation begins it does
so only at a few particular places within the steel where there are localized stress
concentrations that are able to activate or create dislocation sources. ( Leslie used Ti
gettered iron in an attempt to minimize the influence of interstitial solutes upon the yield
R. H.Wagoner MSE 661 – Steel Section 1, Page 16
stress) Consequentially, at yield there is a low density of mobile dislocations. However,
the sample is forced to extend at the strain rate imposed by the testing machine so these
dislocations must glide rapidly and this requires a high stress.
When the dislocation sources (e.g. Frank-Read sources and double cross slip sources) are
activated they generate new dislocation line and the density of mobile dislocations
increases rapidly. This means that the rate of glide necessary to maintain the strain rate
drops and the applied stress needed to produce that rate also drops. The result is a yield
point phenomenon of an upper and lower yield stress and localized yield that causes
Luders bands to form and spread along the sample. In tensile tests conducted at higher
temperatures, serrated yielding occurs because the solute elements are able to diffuse fast
enough to “catch up” dislocations and pin them again. Consequently the stress required to
maintain the strain rate fluctuates up and down like a series of mini yield point
phenomena.
Strain Aging, Blue Brittleness - The material is subject to strain aging and blue
brittleness. Strain aging refers to a process in which the dislocation density is first
increased by plastic strain during a cold work operation and then the material is held
(aged) at temperatures high enough for carbon and nitrogen to diffuse but not high
enough for extensive recovery to take place. The resulting combination of pinned
dislocations and the greater density of forest dislocations raises the flow stress
significantly. Strain aging can be detrimental because of the loss of ductility that occurs
as a consequence. Nitrogen diffuses more rapidly than does carbon at room temperature
and it has a greater solubility in iron at this temperature. If the nitrogen content is high
enough, strain aging and loss of ductility can arise after prolonged aging at room
temperature. Effects due to carbon arise when the temperature is in the range of 100-200
Celsius. Steel so treated can acquire a bluish oxide; hence the term “blue brittleness”.
The solubility of carbon in BCC iron is small ( about 0.02 wt. % maximum) and depends
on what phase is formed when the solubility limit is exceeded. The equilibrium
precipitate is graphite but the kinetics of the nucleation and growth processes of forming
graphite is very sluggish. Metastable carbides are formed first when supersaturated solid
solutions of Fe-C are aged. The precipitate that forms above 150 to 200°C is an
orthorhombic, metastable carbide, Fe
3
C, that is called cementite. From room temperature
to 200°C, another transition phase has been observed ( Honeycombe, page 8). This is an
hexagonal carbide that contains less carbon than is in cementite and is called epsilon
carbide.
Expressions for the solubility are (from Leslie's book):
for equilibrium with graphite...log(c) = 7.81-5550/T
for equilibrium with cementite....log (c) = 6.38-4040/T
where T is in degrees Kelvin and c is in part parts million by weight.
The interstitial solutes diffuse very rapidly in BCC iron so that static and dynamic strain
aging can occur at temperatures as low as 60 deg. C. (Dynamic strain aging refers to
strain aging that occurs while the sample is being deformed). This is caused by their
R. H.Wagoner MSE 661 – Steel Section 1, Page 17
small sizes and also because the diffusion process does not involve vacancies so that each
solute atom usually has an empty site next to it into which it can jump. Honeycombe
presents some diffusion coefficients for solutes in iron in his book, Steels-Microstructure
and Properties, page 7. It is important to note that the diffusion coefficients vary
exponentially with -1/T and that the activation energy for diffusion (of substitutional
solutes also) is smaller in BCC iron than in FCC iron, so that diffusion can be more rapid
in BCC iron than in FCC iron even though the latter is at a higher temperature.
Substitutional Solutes
The majority of solutes in iron occupy substitutional sites. (This is because most
elements are much larger than C or N and the substitutional sites are also much larger.)
These are symmetric sites so that substitutional solute strengthening is not as effective,
for similar concentrations, as interstitial strengthening. Nevertheless, because greater
concentrations can be dissolved, substitutional strengthening at, and above, room
temperature can be significant and of practical use. It is not clear what controls the degree
of strengthening, but both size misfit and electronic differences appear to contribute. Co
and Cr are not effective while P and Pt have strong effects. Ni , Si and Mn have
significant effects.
When the iron is free of interstitial solutes (either because of purification or because the
solutes are scavenged by alloying additions such as Ti, which reduces the dissolved
solute concentration by combining with solute atoms to form second phases) the effects
of substitutional elements are more clearly seen. Interestingly, experiments with these
materials have shown that it is also possible for substitutional solutes to cause static and
dynamic aging. Because these solutes diffuse much more slowly than do interstitial
solutes, the aging temperatures are higher and recovery processes occur also. However,
when interstitial solutes are present in solution, as is most commonly the case, then the
aging effects due the interstitial solutes are dominant.
Slip in BCC Iron
When dislocations glide in BCC iron during and after general yield at ambient
temperatures, they produce wavy slip bands. ( General yield refers to plastic deformation
that occurs across the cross section of the sample not to localized micro yield processes
that correspond to strains of the order of 10
-6
) The dislocations are narrow and are not
pairs of well separated partial dislocations because the stacking fault energies are high.
Cross-slip is evidently easy.
The wavy slip bands and prolific cross slip indicates that the plastic deformation of iron
is dominated by screw dislocations because dislocations of other character cannot cross
slip. Slip has been identified on the {110}, {112} and {123} families of planes which, in
combination with Burgers vectors parallel to <111>, gives rise to 48 slip systems. Iron is
therefore ductile and formable at room temperature and above.
R. H.Wagoner MSE 661 – Steel Section 1, Page 18
As the temperature is lowered, the stress required to make the dislocations glide becomes
very large (see Figure 1-2) and the material shows a transition from ductile to brittle
behavior when deformed at high strain rates. This transition is commonly identified by
impact tests wherein notched bars are broken by impact and the energy expended in
breaking the samples is measured. The transition from ductile to brittle fracture is
associated with a marked reduction in the amount of energy expended. The fracture
process is transgranular cleavage on {100} planes. A material that has a ductile to brittle
transition temperature has obvious limitations for use below that critical temperature.
Iron-Carbon Phases
Phase Diagram - The equilibrium binary Fe-C phase diagram, Figure 1-1, depicts the
phase fields that involve graphite and the solid solutions of the allotropes of iron.
However, graphite does not ordinarily form in commercial steels unless the steel is
seriously overheated. This is because there exists metastable carbide phases that can
nucleate and grow much more rapidly than can graphite when supersaturated Fe-C solid
solutions are aged.
The most important of these carbides is Fe
3
C, which is called cementite. If the sulfur and
phosphorus content are very low, Fe
3
C can decompose into graphite and iron in about 8
hours at 700°C. Graphitization, as this type of formation of graphite is called, is
undesirable because it has an adverse effect on the mechanical properties of the steel.
Fortunately, graphite does not form during the treatments normally given to steels.
Cementite is found in steel artifacts that are many hundreds of years old. Hence the
relevant diagram for practical purposes is the metastable diagram on which cementite
replaces graphite.
The iron-carbon, metastable phase diagram is a valuable tool when one is considering the
nature of steels. Provided it is borne in mind that the details e.g. the exact carbon
composition at the eutectoid point and the exact temperature of the eutectoid isotherm
can be changed by the presence of manganese and residual elements, it can be applied
readily for considering plain carbon steels. Many of the phase transformations in alloy
steels can also be considered using the binary diagram, but then it is important to evaluate
the effects due to the alloying elements in more detail as their concentrations are
increased because these effects can be pronounced and important. The iron carbon
metastable phase diagram is particularly important with respect to ferrous metallurgy. It
should be consulted so frequently that it becomes burned into one’s memory.
Phases - alpha ferrite is the BCC solid solution of carbon in iron that exists in the low
temperature range; austenite is the FCC solid solution, cementite is the carbide and
delta ferrite is the BCC solid solution that can exist at high temperatures. When the term
ferrite is used alone, it is usually understood that it is the low temperature phase
otherwise the ferrite is qualified with the appellation delta. Earlier we observed that
epsilon carbide can precipitate from supersaturated ferrite. However, this metastable
phase is not shown in figure 3 which is really the iron cementite phase diagram.
R. H.Wagoner MSE 661 – Steel Section 1, Page 19
Nomenclature - It is customary in the USA to refer to the eutectoid isotherm as the A
1
line, the boundary between the austenite and austenite/ferrite phase fields as the A
3
line
and the boundary between the austenite and austenite/cementite phase fields as the A
cm
line. Then, for any particular alloy, the temperatures corresponding to the intersection
points with these lines of a vertical line drawn through the composition of the alloy are
called the A
1
point etc. Because the temperatures at which the corresponding
transformations begin during heating and cooling depend upon the rates of heating and
cooling, the effects of heating and cooling are indicated by using c for chauffage and r for
refroidissement i.e. A
c1,
A
r1
. These are related to the kinetics of the phase
transformations that occur during heating and cooling whereas the points on the diagram
refer to equilibrium..
Phase field sizes - Note the extent of the gamma austenite field compared to that of the
alpha ferrite field and reflect on the importance of this with respect to the existence of
steel. Many of the microstructures that can be produced in steels, and therefore many of
the properties that can be tailor-made, depend upon the fact that when austenite is cooled
it can become supersaturated and therefore undergo phase transformations as it seeks
equilibrium. Were this not so, the wealth of microstructures that can be produced by heat-
treating steel would be severely curtailed. It follows that many functions now performed
by steel would not be performed and world would be a rather different place.
Other Alloying - When other alloying elements are added to Fe-C alloys the details of the
phase diagram change. As we will see, these changes can be quite large; however, for
plain carbon steels the changes due to the relatively low quantities of Mn, Si etc. are not
great, so the binary diagram can be used as a reasonable guide to the constitution of these
steels. This is still true for low alloy steels, but it is important to remember always that
there will be differences from the binary diagram. Steels can be classified with respect to
the eutectoid composition: a eutectoid steel has the eutectoid composition, a
hypereutectoid steel has more carbon than a eutectoid steel and a hypoeutectoid has less.
Brief Overview of Strengthening
For purposes of this course, the strength of a metal or alloy may be considered to be
decomposed into three general parts, which we will treat additively even though they are
not necessarily independent as this would imply. A simply breakdown might look like
this:
σ
flow
= σ
o
(intrinsic) + ∆σ
micro
(microstructural) + ∆σ
strain
(strain hardening) (1-4)
Each of these terms is complex, depending on several factors. These may be briefly (and
incompletely)thought of as follows:
σ
o
may be thought of as the intrinsic strength of a “pure” metal or alloy in is softest state,
i.e. with an optimum number of dislocations, the fewest obstacles to dislocation motion,
and so on. This flow stress will depend on the crystal structure of the material, the
R. H.Wagoner MSE 661 – Steel Section 1, Page 20
temperature, the available slip system types, the nature of the metallic bonding, and so
on. Metallurgists often consider these to be unchangeable properties except via alloying,
and except for the role of temperature and slip system, little about this aspect will be
discussed in this course..
∆σ
micro
depends on a variety of changes that can be made to the microstructure of the
basic material, including alloying, additions of second-phase particles (via precipitation,
mechanical alloying or other means), additions of solute atoms (via alloying, solubility
reactions), and distributions of these second-phase particles or solutes. In addition, the
scale of the microstructure, including grain size, is important to the flow stress, on top of
the absolute numbers and sizes. Since these are “tailorable” properties, they are of
considerable importance in producing steels for given applications, and these
mechanisms will be discussed, albeit more briefly than in other courses devoted to
metallurgical mechanisms. For now, it is sufficient to note that the flow stress increased
for smaller grain sizes and for smaller spacing of hard, second-phase particles.
∆σ
strain
depends primarily on the mechanical working or strain that is applied to the
material. The main mechanism for strain hardening is by way of dislocation
multiplication, which causes dislocations to be not only the mechanisms of deformation,
but also the obstacles to it. This category of hardening is very important because this
aspect is a principal governor of the “ductility” of the metal, without which strength is of
little use for practical applictions.
Each of these aspects of hardening will be discussed at a later stage of this course, but it
is important to keep them in mind during the descriptive sections which follow.
SECTION 1 - CONCEPTS AND STUDY QUESTIONS
Important concepts
Definitions of steel, plain-carbon steel, alloy steel.
Principles of steelmaking and how they affect the presence of elements in the final
product. In plain-carbon steels:
Fe – the principal element of steel
C – closely adjusted amount to affect mechanical properties in a variety of ways
Mn – in virtually all steels to form MnS preferentially to FeS, the latter of which
collects at grain boundaries and is either melted or very soft at hot-rolling
temperature, producing “hot shortness.”
Al,Si – these elements appear in coarse, widely-separated particles via entrainment of
slag during production, and in much finer particles from killing
Rare earths – used for sulfide shape control in very small quantities
Tramp elements - such as Cu or Pb, from recycling
Ore elements – mainly sulfur, but many other traces could be found
R. H.Wagoner MSE 661 – Steel Section 1, Page 21
Fe-C phase diagram, equilibrium vs. metastable one, phases, temperatures, compositions
should be memorized (Fe-rich end only). Alternate names of phases.
Steel cleanliness, killing, mini-mills, continuous vs. batch casting
Breakdown of cast structure (dendritic grains) via hot rolling.
Hot rolling vs. cold rolling.
Role of crystal structure on slip systems, Peierl’s energy, low-temperature behavior.
Size and shape of interstitial holes on carbon solubility and lattice strains.
Study Questions
1. The lattice parameter of Fe is 3.6468 angstroms at 916 deg. C and 2.8664 angstroms at
20 deg. C. Calculate the diameters of spheres that would just fit in each case. What is
the more important consequence of the this result? (Also calculate the volume of a box
that would fit among the Fe atoms.)
2. Discuss the characteristics of dislocation glide in Fe.
3. Which interstitial sites in Fe are not surrounded symmetrically by nearest neighbors?
Explain why the existence of non-symmetrical sites is important.
4. When significant quantities of carbon and nitrogen are dissolved at iron, tensile stress-
strain curves at certain temperatures show serrated flow and Luders bands. Strain aging
and blue brittleness can occur. How are these aspects connected?
5. List all of the kinds of factors that determine the flow strength of an iron specimen,
with reference to mechanisms.
6. What is steel?
7. What is a plain carbon steel?
8. What raw materials are used in pig iron production? What are the products and by-
products of a blast furnace iron making process?
9. Why is aluminum or silicon added to steel before casting? Explain the effects clearly.
10. Draw the iron-rich end (up to 2% C) of the metastable iron-carbon phase diagram.
G Label the axes properly
G Show temperatures and concentrations of special points and lines
G Show alternate names and labels of each phase (i.e. "beta" and "wustite" may be
the same phase, but both should be shown.)
R. H.Wagoner MSE 661 – Steel Section 1, Page 22
11. Why is Mn added to plain-carbon steel?
12. What causes the upper yield point in plain carbon steels? How can this be avoided?
13. Why does the fracture toughness of a metal depend on the size of particles?
Physical Metallurgy of Steel
Class Notes and lecture material for MSE 661 (formerly 651.01)
Notes revised by: Robert H. Wagoner, Smith Chair
Notes originally compiled by: Glyn Meyrick, Professor Emeritus
Web installation by: Wei Gan, Graduate Research Associate
Last revision date: 9/30/02
SECTION 2 – NUCLEATION AND GROWTH PHASE
TRANSFORMATIONS
Observations in Pure Iron
Before proceeding to plain-carbon steels, let us consider the allotropic transformation
between the BCC and FCC phases that occurs in nearly pure iron when the temperature is
cooled below about 910 Celsius. A number of investigators have studied this
transformation as a function of the rate of cooling from the high temperature at which the
iron was initially FCC. Their work has been recently reviewed by Borgenstam and
Hillert, Met and Mat. Trans A, 27A, 1501, 1996.
The following classifications can be made based on the imposed cooling rate, which is
shown approximately for each regime. Cooling plateaus corresponding to these various
transformations have been observed in iron with a few parts per million of carbon and
also in Fe-0.01 wt. % C.
• 1 K/sec: When the cooling rates are small, of the order of a few degrees per second,
the temperature of the iron falls below 910 degrees and the iron becomes super
cooled. Then BCC crystals nucleate and grow when the degree of super cooling is
sufficient to provide the necessary driving force . Nucleation is heterogeneous on
grain boundaries and at grain boundary triple junctions. The beginning of the
transformation can be detected experimentally by observing the rates of change in
the temperature of the iron because of the effects due to the evolution of heat as the
transformation occurs. For small cooling rates the degree of undercooling attained
at the onset of the transformation is small and the total transformation is completed
at high temperatures by diffusion controlled nucleation and growth processes that
produce an equiaxed array of grains separated in general by high angle grain
boundaries.
• 1000 K/s: The degree of undercooling at which the transformation begins,
increases initially as the rate at which the FCC iron is cooled is increased but then
remains constant at a plateau value. At the same time the intercrystalline
boundaries of the BCC phase become blocky and faceted and the composition
R. H. Wagoner MSE 661 – Steel Section 2, Page 2
remains uniform. These are characteristics of the massive transformation and the
BCC phase is called massive ferrite. For the case of 0.01C, the massive ferrite
plateau begins near 800 degrees at a cooling rate of several thousand degrees
Celsius per second. The actual temperature of the plateau varies with the grain size
of the FCC phase, being smaller the larger the grain size.
• 10,000 K/s: At still higher cooling rates above 10
4
K/sec, another plateau is
observed at a lower temperature that the one described previously. At this plateau,
the ferrite begins to form as acicular plates which emanate from original grain
boundary positions and also form within grains. Such a morphology is usually
called Widmanstatten (see Figure 2-1), hence the phase is sometimes identified as
Widmanstatten ferrite. It is generally agreed that the equiaxed and acicular ferrite
forms by diffusion controlled processes, although a combination of diffusion and
diffusionless mechanisms may also be in play.
• 100,000 K/s: Cooling rates greater than 10
5
K/sec avoids these transformations and
then another plateau occurs at a lower temperature still. The transformation that
begins at this plateau is still from FCC to BCC but it now occurs so rapidly that it is
not effected by diffusion. It is called a martensitic transformation and the product
is called martensite, whether this represents a new phase or only a microsctructure.
For pure iron, the phase is ferrite (which may also be called martensite because of
the tranformation characteristics) which comprises bundles of narrow, lath shaped
crystals in which the dislocation density is high. (Thus, the microstructure and
hardness are distinguishable from normal ferrite such as that produced at slow
cooling rates, even though the phase is the same.) This is often called lath
martensite.
• Even higher: It has been reported that yet another plateau occurs in the latter
material at even higher rates of cooling. At this fourth plateau that occurs at about
700 K the transformation is again FCC to BCC but the product consists of twinned
plates. This is called lenticular martensite or plate martensite.
The diffusionless processes by which these last two plateaus and transformations are
produced are called martensitic. Martensite or martensitic are the general names given to
a transformation that produces a change in crystal structure in a displacive manner and
does not require diffusion. The products of the transformations in these cases are called
lath martensite and lenticular or plate martensite. In each case the transformation begins
when the plateau is reached but the amount of the sample which transforms depends on
the temperature to which the sample is cooled not on time at that temperature, unlike
diffusion controlled nucleation and growth transformations.
Lenticular martensite has not been observed to form in very pure iron. Also, no one has
succeeded in cooling FCC iron to below the lath martensitic start plateau without forming
martensite. That is, a 100 % FCC phase cannot be retained by rapid cooling to room
temperature.
R. H. Wagoner MSE 661 – Steel Section 2, Page 3
Figure 2-1: Widmanstatten side plate formation in a quenched Fe-0.2C alloy. Grains A
and B, separated by a grain boundary, have orientations that favor growth into different
austenite grains. Electron micrograph from an extraction replica.Magnification, 7500*.
Courtesy of R.N. Caron, Olin Corp., New Haven, CT ((From: G. Krauss: Steels: Heat
Treatment and Processing Principles, ASM International, 1990, p. 40.)).
Plain Carbon Steels - Austenitization
The mechanical properties of steel are controlled by controlling its microstructure and the
nature of the phases within that microstructure. This is accomplished by choosing
appropriate chemical compositions and thermo-mechanical treatments. The final
microstructure of a steel depends upon its composition and its thermo-mechanical history.
This history can involve heat-treatments alone, mechanical treatment alone or
combinations of the two. Heat-treatments can change the microstructures of steels by
processes such as recovery, recrystallization, grain growth and phase transformations. In
order to appreciate these effects let us begin with the subject of phase transformations in
plain-carbon steel.
Austenitization converts a steel into the FCC solid solution called austenite when the
steel is heated into the austenitic phase field and held there long enough. It is the first step
in the majority of heat treating processes and the nature of the austenite that is obtained is
important because it has a direct effect upon the microstructures produced during
subsequent treatments in which the austenite is transformed into other phases and the
steel is returned eventually to ambient temperature. Austenitization also precedes many
hot working processes used to form steel products because it occurs when the steel is
R. H. Wagoner MSE 661 – Steel Section 2, Page 4
heated to the hot-working temperatures. In our discussion of the allotropic
transformation of FCC iron to BCC iron we paid very little attention to producing the
FCC phase because in pure iron and dilute solid solutions a homogeneous polycrystalline
aggregate is easily and rapidly formed when a sample is heated above 910 deg. C. This is
not necessarily so in the case of steels because the starting material is usually a mixture of
phases of different compositions which means that long range diffusion is necessary to
attain homogeneity.
Inspection of the metastable Fe-C phase diagram, Figure 1-2, enables us to estimate the
range of temperatures from which one could be selected for the austenitization process.
These temperatures lie in the austenite phase field. When the temperature is chosen, the
time has then to be selected. The optimum combination of temperature and time depends
on the objectives of the total heat treatment process and upon the microstructure of the
steel before austenitization. We recognize that the steel is initially at ambient
temperature and has a microstructure that depends upon its past history. Commonly, it
will comprise a mixture of cementite and ferrite. There could be free ferrite grains,
discrete particles of the carbide embedded in ferrite grains and there could be regions of
pearlite in which the two phases exist as a lamellar mixture. ( It is important to
understand that cementite and ferrite are phases; pearlite is not a phase it is the name of a
microstructure.) Furthermore, if the austenitization heat -up time is long, the
microstructure could change somewhat before austenitization begins. The heating rate is
an important part of the overall practical process, however, the most important changes
that occur during austenitization do so above the A
1
line and in the austenite phase field.
First, suppose we have a hypoeutectoid steel. When the temperature exceeds the
eutectoid isotherm, the carbides wish to dissolve but the ferrite cannot accept the carbon
residing in the carbides. Hence the first part of the transformation has to be the formation
of some austenite. This occurs by a nucleation and growth process. The nucleation events
are heterogeneous and take place on ferrite grain boundaries and on ferrite/cementite
interfaces. When some austenite forms it provides a sink for the carbon in the adjacent
carbides and the simultaneous growth of austenite and dissolution of carbides takes place,
and, when the temperature exceeds the A
3
line, the austenite devours any remaining
ferrite. Until these transformations are completed, the rate controlling parameters are the
nucleation kinetics and the rates of diffusion of solutes in austenite. Because the
nucleation process is heterogeneous, fine grained ferrite and finely spaced lamellae in
pearlite cause the transformation to occur more rapidly than it would in a coarse
microstructure. Furthermore, the finer the scale of the initial microstructure then the finer
the austenitic grain size will be when the carbides have just all dissolved.
At this point it is important to recognize that the purpose of an austenitization treatment
is usually to produce austenite so that it can then be transformed into something else such
as a mixture of ferrite and pearlite. For producing the most desirable mechanical
properties in the final product it is usually necessary that the austenite be fine-grained. Its
grain size will depend upon the nucleation rate which will in turn depend upon the
availability of nucleation sites and also upon the temperature at which the nuclei form.
When austenitization is complete, grain growth can take place and will do so at rates that
R. H. Wagoner MSE 661 – Steel Section 2, Page 5
are greater the higher the austenitization temperature. Thus it is generally desirable that
the austenite be transformed as soon as possible unless, of course, hot rolling or hot
forging is being done..
The end of a typical austenitization process does not usually correspond to the time at
which the carbides have just dissolved because, although the steel is austenite, its
composition is inhomogeneous because the carbon concentration at and near the positions
originally occupied by the carbides exceeds the concentration elsewhere. Further time is
required for the austenite to homogenize. The diffusion rate of carbon determines this
time in a plain carbon steel. Manganese, however, diffuses substitutionally and therefore
much more slowly than does the carbon. Compositional variations (segregation) of
manganese in an ingot can persist within the steel and cause microstructural effects such
as banding ( i.e. a cyclic variation in the local volume fractions of pearlite and ferrite with
position) in subsequent heat treatments. We note that the range of diffusion is
proportional to the square root of the product of time and the diffusion coefficient. This
can be used as a guide for the selection of the time required for the composition to
approach homogeneity when the scale of the initial compositional fluctuations is known.
It follows from the previous comments that the kinetics of producing homogeneous
austenite can be retarded considerably in steels that contain alloy carbides (e.g.
molybdenum carbides) because substitutional diffusion is much slower than interstitial
diffusion.
We have noted that an important microstructural change to be considered during
austenitization is grain growth. Some austenite grains continue to grow and, necessarily,
others shrink and disappear causing a decrease in the number of grains per unit volume.
Typically, the average grain size increases with temperature at a rate that is proportional
to the square root of the time at the temperature. Usually a large austenitic grain size is
undesirable because it leads to a coarse microstructure when the austenite is transformed
back into the microstructure in which it will be used. Sometimes the austenitization heat
treatment is performed only to create austenite and is immediately followed by the rest
of the heat treatment . On other occasions mechanical work such as forging or rolling is
carried out while the steel is austenitic. These processes can be accompanied by recovery,
recrystallization and grain growth. A valuable effect derived from using aluminum to kill
steels is relevant here. Aluminum forms fine particles of nitrides and oxides which
impede grain boundary migration below 2000°F and thereby helps to keep the austenitic
grain size small. Thus it is said to be a grain refiner (This is an example of metallurgical
English. The aluminum-rich particles certainly do not free the steel of impurities. They
impede grain growth and resist grain coarsening. Presumably, the lack of coarse grains
makes the steel more refined in the same way that young ladies are presumed to be more
refined than young men.)
Plain-Carbon Steels – Decomposition of Eutectoid Austenite
The array of microstructures (and thus mechanical properties) that are obtainable in steels
by various heat-treatments, depends on the manner in which austenite transforms when it
R. H. Wagoner MSE 661 – Steel Section 2, Page 6
is cooled to ambient temperatures. In order to appreciate how these microstructures are
produced it is important to understand the various modes by which austenite can
decompose. Indeed, this is imperative if one wishes to further develop and improve
steels.
After this description of the austenitization process in hypoeutectoid steels, the extension
to eutectoid and hypereutectoid steels should be apparent; see figure 4.
Figure 2-2: Austenitizing rate-temperature curves for commercial plain carbon eutectoid
steel. Prior treatment was normalizing from 1610 F (875 C); initial structure, fine
pearlite; second curve, final disappearance of pearlite; third curve, final disappearance of
carbide; fourth curve, final disappearance of carbon concentration gradients.
Transformations in Steel of the Eutectoid Composition - The mode of cooling from the
austenitization temperature is a very important parameter to consider when discussing the
transformation of austenite. A particularly important variable is the rate of cooling. It is
customary, for obvious reasons of convenience, to describe cooling rates as "slow",
"furnace cooling", "air-cooling", "water-quenching" etc. Such descriptions are not very
1550
1500
1450
1400
1350
1300
1 10 100 1000 10,000
Austenite and
residual
carbide
Pearlite
Ae
1
840
820
800
780
760
740
720
Time, sec
T
e
m
p
e
r
a
t
u
r
e
,

F
T
e
m
p
e
r
a
t
u
r
e
,

C
Homogeneous
austenite
Austenite
with carbon
inhomogeneities
99.5 %
Austenite
0.5 %
Austenite
Pearlite and
austenite
and residual carbide
R. H. Wagoner MSE 661 – Steel Section 2, Page 7
precise because the actual cooling rate that a piece of steel undergoes will depend on
particular conditions: it can depend upon its environment, the dimensions of the piece of
steel, and will usually vary from place to place within the sample. Also, the cooling rate
at any particular site within a sample will vary as the temperature at that site changes
unless special precautions are taken to ensure that the rate is constant. Air cooling means
simply that the steel is allowed to cool naturally in still air. Quite different
microstructures can be produced in different forms of the same steel e.g. a half inch
round bar versus a four foot diameter coil of rolled sheet.
If a piece of a plain carbon steel is austenitized and then cooled it will attempt to
transform into a mixture of ferrite and cementite according to the phase fields indicated
on the phase diagram. Inspection of the equilibrium compositions of these phases at the
temperature of the eutectoid isotherm leads to the conclusion that a transformation to
ferrite and cementite must require atomic diffusion and that the growth rates of the new
phases will depend primarily on the rate of diffusion of carbon which has to undergo long
range diffusion. The kinetics of such a transformation is therefore a function of time and
temperature.
During continuous cooling, various parts of the sample transform at different times and
hence at different temperatures. When the cooling rates are slow enough to allow
diffusion to occur readily, the transformation is completed at high temperatures and
occurs over a small range of temperature. This is the situation to which “furnace cooling”
and “slow cooling” refer; the steel is close to its equilibrium state as its temperature
drops. Under these conditions the lever rule and the phase diagram provide reasonable
estimates of the volume fractions of various phases to be expected in the final
microstructures. As the cooling rates are increased, the transformation to ferrite and
cementite begins and ends at lower and lower temperatures and the steel is no longer
close to equilibrium as it cools. Quantitative predictions from the phase diagram such as
the volume fraction of free ferrite to be present in the final microstructure are no longer
possible. Because of these complications, it is sensible to consider, first, what happens
during isothermal transformations.
Isothermal Transformations, Descriptive - Strictly speaking, the term isothermal
transformation of austenite implies that the transformation takes place at a constant
temperature and, also, that none (or negligible amounts) of the transformation take place
before the austenite arrives at this temperature. This is generally difficult to achieve in
practice, especially when the transformation is rapid and the sample of steel is large.
Initially, we will suppose that the samples that we are considering are so small that the
temperature is always essentially uniform throughout them and that temperature changes
can be made instantaneously. This supposition is made now to enable us to concentrate
on the nature of the transformations and to defer until later considerations of changes that
might occur while the samples are being cooled to the isothermal transformation
temperature.
Let us suppose that an experiment has been conducted in which many samples of a plain
carbon steel of the eutectoid composition (e.g. a 1080 steel) were austenitized in the same
R. H. Wagoner MSE 661 – Steel Section 2, Page 8
manner, instantaneously cooled to, and held for various times at, selected temperatures
below the eutectoid isotherm. Let us suppose also, that the microstructural changes that
occurred within them were monitored. At this stage we shall not concern ourselves with
how the changes are monitored, we simply suppose that they are. What are the results of
such an experiment?
• ∆T<100K: During the rapid quench to a transformation temperature within 100
deg. C of the isotherm, each sample remains a polycrystal of austenite. While the
sample is at the transformation temperature, processes occur which eventually
produce recognizable crystals of ferrite and cementite. The time taken for this to
happen is called the incubation period.
From a practical point of view, the magnitude of the incubation period that is
determined experimentally, depends on the resolution of the technique used to
monitor the process because the minimum size of a particle of a new phase that can
be detected becomes smaller as the resolution increases. Because of this, let us
arbitrarily define the end of the incubation period as corresponding to a volume
fraction transformed of 0.5%. One might similarly define the end of the
transformation as a transformed volume fraction of 99.5%.
The incubation period is temperature dependent. It is large near the A
1
line and
decreases rapidly as the transformation temperature is reduced. This is because the
magnitude of the thermodynamic driving force for the transformation increases as
the degree of undercooling below the A
1
line increases. The transformation begins
at grain boundaries where particles of cementite and ferrite form adjacent to each
other. These particles grow and the boundary advances into one of the grains,
acting as a boundary between the untransformed austenite ahead of it and the region
that has transformed behind it. As the interface moves, the particles of ferrite and
cementite develop into plates so that the transformed volume comprises a lamellar
array of alternating plates of these two phases. These events occur at various places
on the grain boundaries and at triple junctions. The lamellar microstructure is
called pearlite, Figure 2-3.
It is important to note that pearlite is not a phase; the phases in it are ferrite and
cementite. Pearlite is the name of the microstructural constituent that consists of
alternating lamellae of the two phases. In each colony of pearlite, the lamellae are
mostly parallel and are frequently curved. The spatial orientations of lamellae from
colony to colony are not the same. Later, neighboring colonies of lamellae,
differently oriented (in space), join together and continue to advance into the
austenite. When the transformation occurs at low degrees of undercooling such
groups of colonies become bounded from the austenite by an approximately
spherically curved boundary and are called pearlite nodules. Eventually, as the
nodules continue to grow, they impinge on one another and finally the samples are
converted into an assembly of pearlite colonies. As the temperature at which the
transformation is occurring is decreased, the nucleation process becomes so rapid
R. H. Wagoner MSE 661 – Steel Section 2, Page 9
that all of the grain boundaries are consumed very early in the transformation and
the pearlite nodules are in contact along the areas where the boundaries existed.
The average inter-lamellar spacing in the colonies depends upon the temperature at
which the pearlite was formed, decreasing with a decrease in the transformation
temperature. It is customary to refer to this by using terms such as "fine" or "coarse"
pearlite. The spacings can vary over the approximate range of 1 micrometer to 0.1
micrometers.
Figure 2-3: Pearlite in a furnace-cooled Fe-0.75C alloy. Magnification, 500x. (From: G.
Krauss: Steels: Heat Treatment and Processing Principles, ASM Int., 1990, p. 20.)
• Lower temperatures: As the temperature at which the transformation occurs is
decreased further, both the incubation time and the inter-lamellar spacing decrease.
Then, near 550 °C, the sample does not transform entirely to pearlite, instead parts
of the microstructure comprise non-lamellar aggregates of small crystals of ferrite
and cementite. At lower temperatures, the incubation period begins to increase
again, and pearlite is no longer formed. Instead, the microstructure is made up
entirely of fine aggregates of ferrite and cementite. These microstructures are
called bainite. Examples are shown in Figures 2.4 and 2.5. Bainite structures will
be discussed later in this section.
• Quenching. The next important observation made in this experiment is that when
samples are quenched to below a particular temperature (about 220°C in the present
instance) it is no longer possible to retain all of the original austenite. This is
entertaining considering that the incubation time for forming bainite had been
increasing as the temperature fell and it was getting easier and easier to preserve the
R. H. Wagoner MSE 661 – Steel Section 2, Page 10
austenite phase. As the samples are quenched below this critical temperature, fine
crystals of a BCT phase appear. The transformation that produces this new phase is
called a martensitic transformation. It is an athermal transformation in plain carbon
steels because it occurs during cooling and begins when the sample is cooled below
a particular temperature; the amount of the new phase that is formed depends on
the temperature to which the sample is cooled rather than the time it spends at that
temperature. The temperature at which the martensitic transformation starts is called
the M
s
temperature. We recall that the martensitic transformation occurs in
relatively pure iron which contains very little carbon. When rapid quenching
suppresses the diffusion processes necessary for the formation of cementite in the
steel and the temperature continues to fall, the free energy can be reduced by
transforming the iron solvent from the FCC structure. The iron tries to become BCC
but is distorted to BCT by the supersaturation of carbon that remains in solid
solution. This statement will be elaborated upon in the future.
Figure 2-4: Microstructure of upper bainite: (a), 0.8C steel, transformed 20s at 400°C,
x670; (b) 0.34C steel transformed at 400°C, x330; (c) 0.8C steel transformed at 450°C
(TEM). (d) 0.8C steel transformed 5s at 450°C (TEM) (Ref. 3, p. 116.)
R. H. Wagoner MSE 661 – Steel Section 2, Page 11
Figure 2-5: Microstructure of lower bainite (a) 0.8C steel transformed 30s at 300°C,
x260; (b) 0.8C steel transformed at 300°C, x260; (c) 0.8C steel transformed 6 min at
250°C (TEM), (d) 0.6C steel transformed at 300°C (TEM). (Ref. 3, p. 119.)
In ferrous metallurgy it is customary to identify the phase that is produced in the
steel as it is quenched below the M
s
point by the name martensite. We will follow
this custom, but it is important to note that the martensitic transformation is not
limited to steel. It is a common mode by which materials can transform their crystal
structures, e.g. brasses, titanium alloys, stabilized zirconia and cobalt. Hence, in
materials in general, the term martensite may not necessarily refer to a phase that
occurs in steels.
The amount of martensite that is formed during quenching, increases as the
temperature to which the samples are quenched decreases. The transformation
ceases when the temperature falls below the M
f
temperature, where the samples of
eutectoid steel are mostly martensite with a small amount of untransformed
austenite dispersed within them. This untransformed austenite is called retained
austenite. The amount that is present in a quenched sample depends upon the
R. H. Wagoner MSE 661 – Steel Section 2, Page 12
composition of the steel. In plain carbon steels, increasing the carbon content tends
to increase the amount of retained austenite but only to a few percent. Retained
austenite is more important in alloy steels because many of the alloy steels have M
f
temperatures that are lower than the temperatures of the usual cooling baths, e.g.
water or oil at room temperature. Martensite and Martenisitic transformations will
be discussed in more detail in the Section 3 of these notes.
When the samples are held at the temperatures to which they have been quenched, further
microstructural changes can occur. Austenite can transform to bainite and the martensite
can decompose into ferrite and carbides. Details of the latter changes will be discussed
later. However, it is appropriate to note, now, that the BCT martensite is a phase that does
not appear on the phase diagram. This is because it is a metastable phase that readily
transforms when held at temperatures above a couple of hundred degrees Celsius.
Summary of Eutectoid Phase Transformations - It should now be reasonably easy to
recognize the main changes associated with the decomposition of the austenite in larger
samples, continuously cooled at various rates. When the rate of cooling is slow enough
for the transformation to begin and end just below the eutectoid isotherm the product will
be coarse pearlite. As the cooling rate increases, finer and finer pearlite will appear and
the spacings throughout the sample become more variable reflecting the fact that different
colonies form at different temperatures. Mixed microstructures of fine pearlite and bainite
then become possible. For large samples in which different parts cool at different rates
the microstructures will vary correspondingly and one can expect to find pearlite, bainite
and martensite in the same piece of steel.
TTT Diagrams
Before extending the description of eutectoid phase transformations to hypoeutectoid and
hypereuctectoid compositions, it is useful to digress for a moment and introduce the
method for presenting the experimental kinetic information graphically. We will then
return to a more thorough description of the mechanisms and foundation for what appears
in the kinetics of the nucleation and growth reactions already introduced.
We restrict our attention to the decomposition of austenite by reactions that occur at
constant temperatures. Many experimental investigations of isothermal transformations in
steel have been made. The results have been published in the form of TTT diagrams
which depict the Transformations as a function of Time at Temperature. (Such diagrams
are alternatively called I-T diagrams for Isothermal Transformation.)
The diagrams, an example for eutectoid steel which is shown in Figure 2-6, have
temperature as the ordinate and the logarithm of time in seconds as the abscissa. The
times at which transformations begin and end at various temperatures are plotted on the
diagram. A typical diagram is produced by the experiment described previously, in
which small samples of the steel are austenitized to produce samples of austenite of the
same grain size. The samples are small so that the temperature can be varied rapidly and
temperature gradients minimized. However, cooling rates to the isothermal temperatures
R. H. Wagoner MSE 661 – Steel Section 2, Page 13
vary from experiment to experiment and from laboratory to laboratory so that particular
diagrams, even if determined on identical steels, can show differences due to
microstructural changes that occur during cooling. In one method of determining a TTT
diagram, the samples are austenitized, quenched to a selected temperature, held at that
temperature for various times and then rapidly quenched to room temperature or below.
The progress of any transformation is then measured by means of quantitative
metallography and the data assembled to produce the diagram.
Let us suppose that Figure 2-6 represents the true TTT diagram for an AISI 1080 steel,
i.e. it depicts data from truly isothermal transformations. A horizontal line marks the M
s
temperature which is the temperature at which martensite begins to form during cooling.
The diffusion controlled transformations that occur above this line are described by a pair
of "C" shaped lines that mark the beginning and end of the transformations. The products
of the transformations are ferrite and cementite (denoted as “F” and “C” in the
appropriate areas on the diagram, with “A” representing austenite.).
Usually these curved lines present above the M
s
line are not projected below it, as has
been done for your convenience on Figure 2-6. However, it should be noted that unless
the M
s
temperature is so low that diffusion is suppressed, ferrite and cementite can still
form below the M
S
temperature from the untransformed austenite as aging is continued.
Because the martensitic transformation in plain carbon steels is athermal, the martensite
forms during the cooling process and the amount that forms depends upon the
temperature to which the sample is cooled. Sometimes other horizontal lines are plotted
to depict temperatures at which 50% of the sample is martensite etc. Then the
temperature below which no more martensite forms is sometimes shown as the M
f
temperature.
Inspection Figure 2-6 reveals that at small degrees of undercooling below the A
1
line, the
transformation takes a long time to begin. As the degree of undercooling is increased the
incubation time decreases rapidly even though the transformation is occurring at lower
temperatures where diffusion is less rapid. Eventually the incubation time reaches a
minimum, t
m
and then begins to increase again. The part of the “C” curves at this
minimum incubation time is called the nose of the TTT diagram. The product of the
transformation at temperatures above the nose is pearlite, hence the labeling of the first
line above the nose as the “P
s
” (for “pearlite start”) and the final line as ““P
f
” (for
“pearlite finish”). Bainite is formed below the nose so the same lines below the nose are
labelled “B
s
” (for bainite start) and “B
f
” for “bainite finish). Near the nose, mixtures of
both microstructures are formed.
R. H. Wagoner MSE 661 – Steel Section 2, Page 14
Figure 2-6: A TTT diagram (or I-T diagram) for eutectoid plain-carbon steel (AISI
1080). (After Ref. 7., p. 16)
Extension to Hypoeutectoid and Hypereutectoid Compositions
Hypoeutectoid Compositions - As these steels are cooled from the austenitizing
temperature they pass through the alpha + gamma phase field before the A
1
line is
passed. Consequently, the first decomposition product that forms, provided the cooling
rate is not too rapid, is the proeutectoid alpha phase, Figured 2-7 and 2-8.
During slow cooling, which allows adequate time for the diffusion processes to occur, the
ferrite nucleates heterogeneously on the austenite grain boundaries and then grows along
these boundaries and also into the austenite grain interiors forming equi-axed grains of
ferrite embedded in the austenite. No preferred orientation relationships exist between the
two phases. These microstructural changes involve the diffusion of the carbon ahead of
800
700
600
500
400
300
200
100
0
T
e
m
p
e
r
a
t
u
r
e
1 MIN
0.5 1 2 5 10 10
2
10
3
1 HOUR
10
4
10
5
10
6
1 DAY 1 WEEK
Time - Seconds
A
A
s
A
F + C
11
32
38
40
40
41
43
50
55
57
H
a
r
d
n
e
s
s
-
R
C
οοο ο
C
A
M
s
M
50
M
90
50%
A
+
F
+
C
Type: 1080
Grain size=6
T
austenitize
=899
οοο ο
C
P
s
B
s
P
f
B
f
R. H. Wagoner MSE 661 – Steel Section 2, Page 15
the ferrite/austenite boundaries so that the growth is controlled by long range diffusion in
the austenite.
When the temperature reaches that of the A
1
line, the microstructure will consist of a
mixture of equiaxed ferrite and austenite grains. The composition of the austenite and its
homogeneity will also depend on the cooling rate. Very slow cooling rates allow the
composition with respect to carbon to be uniform and to approach the eutectoid
composition at the A
1
line. In such a case, the austenite is now a eutectoid steel and it
transforms as described previously, as cooling continues.
As the cooling rates are increased, the austenite becomes inhomogeneous in composition
and there is more of it present when the A
1
temperature is reached than would be
indicated by applying the lever rule to the phase diagram because the steel can no longer
remain close to its equilibrium state as it cools. Proeutectoid ferrite continues to form as
the temperature falls below the A
1
line. Then pearlite appears and grows into the
austenite.
Figure 2-7: 1045 steel bar. Structure is fine lamellar pearlite (dark) and ferrite (light).
(From: Metals Handbook Ninth Edition, Volume 9, ASM Int., 1985, p.188.)
As the temperature at which the proeutectoid ferrite forms becomes lower and lower the
morphology of the ferrite changes. Growth along the grain boundaries becomes markedly
easier than growth normal to them. Therefore, films of ferrite form at the sites of the
former austenite boundaries. Then, as the cooling rate increases further, ferrite side-
plates emerge from these films and project into the austenite grains. These side-plates are
wedge-like in shape and the broad faces comprise flat areas separated by ledges. The flat
R. H. Wagoner MSE 661 – Steel Section 2, Page 16
areas are interfaces across which special orientations exist. The growth process now
involves the nucleation and sideways migration of the ledges that separate the flats. These
side-plates are called Widmanstatten plates. The microstructure on a polished and etched
sample resembles feathers where the spine lies along the prior boundary and the quills
project at oblique angles away from the boundary.
Figure 2-8: TTT diagram for 1050 hypoeutectoid steel. (Ref. 7, p. 15)
Eventually, the under-cooling can create a thermodynamic driving force that is large
enough to cause ferrite to nucleate within the austenite grains as well as on the
boundaries. These nuclei also grow into plates. The microstructure of a hypoeutectoid
steel that is cooled rapidly but at a rate that still permits the austenite to transform into
ferrite plus pearlite will contain the Widmanstatten ferrite and fine pearlite.
The volume fraction of pearlite present when the austenite has transformed will exceed
that predicted by the appropriate tie-line on the phase diagram because the formation of
pearlite begins before the equilibrium volume fraction of free ferrite has had time to
800
700
600
500
400
300
200
100
0
T
e
m
p
e
r
a
t
u
r
e
200
400
600
800
1000
1200
1400
Estimated
Temperature
1 MIN
0.5 1 2
5 10 10
2
10
3
1 HOUR
10
4
10
5
10
6
1 DAY 1 WEEK
Time - Seconds
A
A
s
A
F + C
18
23
28
32
36
42
48
52
62
H
a
r
d
n
e
s
s
-
R
C
ο
F
ο
C
I - T Diagram
A
M
s
M
50
M
90
S
S
S
S
50%
A
f
A+F
A+F+C
AISI 1050
0.50%C, 0.91%Mn
gs = 7-8
Austenitized at 819 deg.
C
R. H. Wagoner MSE 661 – Steel Section 2, Page 17
form. This is an important point to bear in mind if one is trying to estimate the carbon
concentration in a plain carbon steel by observing its microstructure after cooling.
Eventually, when the cooling rates become fast enough, bainite and martensite begin to
form and, if the sample can be cooled fast enough, all of the diffusion-controlled
transformations can be suppressed and only the martensitic transformation occurs.
Hypereutectoid steels - In this case the proeutectoid phase is cementite, Figure 2-9. The
transformations and the resulting microstructures are very similar to those described
above with the cementite replacing the ferrite. Widmanstatten plates of cementite occur
but do not grow as easily as the ferrite, so are often not seen in plain-carbon steels, which
seldom have much more than the eutectoid composition of carbon.
Figure 2-9: Pearlite and grain-boundary cementite in an hypereutectoid steel. Cementite
appears white.
R. H. Wagoner MSE 661 – Steel Section 2, Page 18
Continuous Transformation (C-T) Diagrams.
Isothermal transformation diagrams provide valuable information about the kinetics of
the transformation of austenite and are useful tools for considering heat-treatments. In
practical heat-treatments, however, where the size of the part being treated is usually
significant, the temperature and its rate of change vary with position within the part being
heat treated. Also, there are heat treatments in which the austenite is cooled
continuously. The real TTT diagram (i.e. the diagram that describes transformations in
samples which cannot be quenched extremely rapidly) is not the same as the true
isothermal diagram.
This is indicated in Figure 2-10. A true isothermal start line or “C” curve is shown as
well as a cooling path( dotted line ) down to an isothermal transformation temperature.
This cooling curve can be approximated by a series of isothermal steps linked by vertical,
instantaneous temperature changes as is shown by the stepped line. Inspection of this
sketch shows that as the temperature falls through the transformation range below T
0
,
each of the horizontal steps represents a fraction of the incubation period corresponding
to that temperature.
Figure 2-10: Conceptual view of a continuous cooling curve as a series of isothermal
transformations periods.
In the 1930’s, Scheil proposed that these “fractional incubation times” could reduce the
time required for the transformation to begin during the final isothermal stage. This has
been considered further by I.A. Wierszykkowski in Met. Trans. 22A, 993, (1991).
During a truly isothermal transformation that takes place at a temperature T
i
the
transformation begins when the time t = t
i
i. e. when t/t
i
= 1. All of the isothermal steps
from a to b in the series drawn on the sketch, are too short for the transformation to have
begun at any of the temperatures concerned. Thus at point ,b, the austenite is still
austenite. If the isothermal stage begins at b, then the transformation will begin after an
additional time of ∆t
x
We see that the real TTT curves will be displaced to the right on
the diagram. ∆t
x
is smaller than the true isothermal transformation time for that
temperature because some of the “pre-precipitation” microstructural changes will have
Continuous
Isothermal
T
log (t)
T
o
∆∆∆ ∆t
i
∆∆∆ ∆T
i
R. H. Wagoner MSE 661 – Steel Section 2, Page 19
already occurred. We are forced to conclude that even though a particular steel will
possess a particular true TTT diagram, it has a variety of real diagrams in practice
because of the influence of different rates of cooling which are themselves dependent on
the sieze of the piece of steel and the nature of the quenchant used.
This has led to the creation of diagrams that predict the modes of transformation that take
place during cooling. There are two main types of continuous cooling diagrams. The first
type uses axes that show the temperature of the steel on the ordinate axis and the
logarithm of time in seconds along the abscissa, Figure 2-11. Then lines are drawn on the
diagram to show the cooling path in terms of the variation of temperature with time. If we
consider one such cooling path and travel along it , downwards from the austenite, we
encounter the start of a transformation. The location of this point is marked on the
diagram and joined to similar points on other cooling paths to create curved lines that
denote the initiation of the decomposition of the austenite. Similar lines which show the
progress of the transformation in terms of the % completed are often appended between
the start and finish curves. Then the diagram is divided into areas according to the nature
of the product of the transformation. Diagrams of this type are determined experimentally
by means of dilatometry.
When a steel part is austenitized and cooled, the heat is extracted from the free surface. In
order for the center of the part to cool, heat has to be conducted to the free surface. The
actual rates of cooling depend upon shape, size, thermal conductivity and the rate of
extraction of the heat. For example, if a cylindrical bar is quenched in a bath of cold
water, the surface cools the most rapidly and the center the most slowly. Thus, different
parts of the steel will cool at different rates. To use a continuous diagram of the kind we
are discussing requires a knowledge of the cooling rates at various positions.
The second type of diagram retains temperature as the ordinate axis but calibrates the
abscissa in terms of diameters of steel bars cooled in air, oil and water respectively,
Figure 2-12. These quenching media extract heat from the surface of a bar at different
rates with the water quench being the most rapid. They are methods of cooling that are
commonly used in practice. This statement is not intended to imply that all quenching
oils extract heat at the same rates, so the diagrams provide guidelines but should not be
regarded as quantitatively applicable. There are also variations in chemical composition
among steels of the same designation which can change details of the transformation.
With those caveats, the diagrams show the progress of the transformation at the center of
bars of various diameters when they are cooled at those rates. It should be noted that this
microstructure will not necessarily be found throughout the bar.
R. H. Wagoner MSE 661 – Steel Section 2, Page 20
Figure 2-11: A continuous cooling transformation (C-T) diagram of the time-temperature
type. This is for an alloy steel that exhibits separated pearlite and bainite curves, the so-
called “bay.” This will be discused later, under “Alloy Steels.”
Austenitizing temperature
Martensite
Martensite
Martensite
ferrite
Martensite
perrite
pearlite
Ferrite
pearlit
Austenite to bainite
Austenite to martensite
A
u
s
te
n
ite
to
fe
r
r
ite
A
u
s
t
e
n
i
t
e

t
o

2
1
0
0


F
/
h
r
2
1
0
0


F
/
h
r
5
4
0
0
0

F
/
h
r
1
5
0

F
/
h
r
4
0

F
/
h
r
.
AC3
AC1
T
e
m
p
e
r
a
t
u
r
e
,

d
e
g
.

F
1600
1400
1200
1000
800
600
400
200
0
5
1
0
2
*
1
0
5
*
1
0
1
*
1
0
2
2
*
1
0
2
5
*
1
0
2
1
*
1
0
3
2
*
1
0
3
5
*
1
0
3
1
*
1
0
4
2
*
1
0
4
5
*
1
0
4
1
*
1
0
5
2
*
1
0
5
5
*
1
0
5
Transformation time, seconds
p
e
a
r
l
i
t
e
R. H. Wagoner MSE 661 – Steel Section 2, Page 21
Figure 2-12: Continuous cooling transformation (C-T) diagram of the cooling rate –
temperature type.
Kinetics – Overview
The formation from austenite of pearlite, ferrite (hypoeutectoid compositions) and
cementite (hypereutectoid compositions) upon cooling below 727 deg. C is controlled by
nucleation and growth kinetics. In the next heading, we will consider these reactions, the
kinetics, and the resulting morphology in more detail. However, a quick overview may
be helpful first, along with a contrast with bainite and martensite formation. (Bainite is
not easily characterized, having both characteristics of a nucleation-and-growth
transformation and of the diffusionless martensitic transformation.)
Nucleation and Growth – Pearlite, Cementite, Ferrite – Only a few concepts need to
be kept in mind to understand the qualitative nature of these products. First, the
nucleation rate of a new phase, whether homogeneous (seldom) or heterogeneous (usual),
follows a simple law something like:
J A. exp−
∆G

kT
j
(
,
\
,
(
(2-1)
B
A
P
START
FINISH
M
F
Ac
3
Ac
1
10 %
50 %
90 %
0.1 0.2 0.5 2
5
10
20
50
100
200
500 1000 2000 1 5
10
10
20
20
50
50
500
500
200
200
100
100
150 300
150 300
BAR
DIAMETER mm OIL
mm WATER
AISI 1040, 0.40 C, 0.70 Mn
Austenitized at 850 C
COOLING RATE AT 750 deg. C, deg. C per minute
1000 500 200 100 50 20 10 5
2 1
900
800
700
600
500
400
300
200
100
0
C
mm AIR
R. H. Wagoner MSE 661 – Steel Section 2, Page 22
where J is the nucleation flux of particles of the new phase, that is, the number of such
particles formed per unit area (at an interface, for example) per unit time. ∆G* is the free
energy needed to create a critical-sized nucleus (i.e. one that can grow). It is clear that
this flux varies very rapidly as the undercooling is increased, that is, as ∆G* changes
from zero at the eutectoid isotherm to a negative number at lower temperatures. Over
this fairly narrow temperature range, the KT term is only slowly changing, so ∆G*
dominates. However, as the temperature falls further, eventually the KT term begins to
dominate and in fact the factor A also depends on temperature, so the nucleation flux
goes through a maximum at some critical undercooling.
The main consequence of Equation 2-1 is that nucleation will be very slow near the
isotherm and will increase rapidly as the transformation temperature is reduced.
However, at lower temperatures still, the number of nuclei will decrease. A second
consequence of Equation 2-1 is that nucleation rate will be increased if there are more
sites for nucleation. Thus, if there is more grain boundary area in a sample (i.e. smaller
grains), then there should be more nuclei.
The growth phase of a nucleation and growth process is usually controlled by diffusion,
although the details of the exact mechanism, species being diffused, and diffusion
distances differ depending on the exact transformation. Nonetheless, simple results from
solutions to the diffusion equation may be used to understand the principal consequences.
The first and most important is as follows:
X k Dt (2-2)
where X is the “diffusion distance”, i.e. a characteristic distance of how far the diffusing
species can move in a give time, D is the diffusion coefficient and is the specified time. k
is a simple proportionality constant. In order to understand how Equation 2-2 influences
the morphology of the product phases, one more equation is required, the dependence of
the diffusion coefficient on temperature:
D D
Q
kT
o


j
(
\
,
exp (2-3)
where D
o
is a constant and Q is a characteristic energy required to move the diffusing
species through the medium.
The consequence of Equations 2-2 and 2-3 is clear: growth of the new product phases
(from the nuclei produced according to Equation 2-1) will take place faster at higher
temperatures. This assumes that the driving force for the reaction is of secondary
importance (as it will be if the growth is diffusion controlled).
Considering the nucleation and growth reactions together, it is clear that at temperatures
near the isotherm, nucleation will take place slowly, whereas the growth will be rapid.
Therefore, we should expect to see large particles of the second phase(s), at large
distances from one another. At lower temperatures, the nucleation rate is increased, so
R. H. Wagoner MSE 661 – Steel Section 2, Page 23
there will be many more particles to grow, but the growth rates will be slower. Under
these conditions, the second phase particles will be much greater in number but much
smaller. It is the balance between nucleation and growth that determines much of the
morphology of the product phases for such transformations. For production of pearlite,
we expect to see a coarse structure when transformed at temperatures near the isotherm
and a fine structure at lower temperatures, in agreement with experimental observations.
Later in this section, the kinetics of nucleation and growth transformations will be
analyzed more quantitatively.
Bainite and Martensite.
So far, we have been considering only the upper portion of the TTT diagram for a
eutectoid steel. In this region that extends from the nose to the A
1
line, the product of
isothermal transformations is pearlite. Another type of microstructure is formed during
isothermal transformations below the nose. In this temperature range, the lamellar
product is replaced by mixtures of discrete particles of cementite in ferrite. Details of
these microstructures vary with the temperature and also vary from steel to steel. In
general, the product formed near the nose is called upper bainite and that near the
martensitic start temperature is called lower bainite. It has often been stated that upper
bainite looks like feathers in optical micrographs while lower bainite is more acicular. It
is not easy to distinguish between them. In practice it is wise not to identify the
microstructure unequivocally ( particularly from a single photomicrograph) without
additional supporting information. At this stage it is worthwhile for the reader to spend a
few hours examining microstructures of plain carbon steel as shown in the optical
photomicrographs that are presented in the ASM atlas of microstructures, eighth and
ninth editions.
Numerous studies have been conducted by optical microscopy and by transmission
electron microscopy. As a result, the following major characteristics have been identified.
• In upper bainite, the ferrite comprises bundles or sheaves of fine, elongated grains
that resemble laths. The cementite particles lie on the long boundaries between the
individual laths.
• In lower bainite the ferrite crystals that form the sheaves appear more like needles
and plates. Particles of cementite form both at the boundaries between the grains
and within the crystals of ferrite.
It is also observed that when the transformation to lower bainite is conducted in samples
that have been previously metallographically polished, surface relief is generated because
the flat surface becomes rumpled and tilted during the transformation. Surface relief is
uncommon in transformations that only involve diffusion; it occurs generally during
martensitic transformations.
An understanding of the mechanism(s) by which the bainitic microstructures form has
been actively sought for decades. Such details are really beyond the ambit of these notes,
but it is interesting to consider some generalities of the subject. Opposing viewpoints
have been presented and hotly debated by respected scientists. One issue concerns
R. H. Wagoner MSE 661 – Steel Section 2, Page 24
whether or not there is a contribution from a displacive or martensitic mechanism i.e.
diffusion creates the carbides and the ferrite forms martensitically. Proponents of the
viewpoint that there is such a contribution, have considered that the surface relief that
forms when lower bainite is created is significant. The M
s
point for the initiation of
martensite is raised as the carbon content of the austenite is reduced. Thus, if the carbon
content at various locations is reduced by the precipitation of cementite then the M
s
temperatures for these locations would be raised and the allotropic change accomplished
martensitically. Such a mode of transformation has been dubbed a diffusion paced
martensitic mode.
Others believe that the transformation is a diffusion controlled nucleation and growth
process in which the crystals grow by the lateral propagation of ledges and the sheaves
form because of sympathetic nucleation of new crystals at the boundaries of the prior
crystals. Articles supporting both viewpoints have been published in the technical
literature during the past five years so one can hardly say that the situation has been fully
resolved to everyone’s agreement but it seems fair to suggest that the nucleation and
growth description is most favored.
As more instrumental techniques for studying the bainitic microstructure are created and
improved more complexities are revealed. Some examples of these microstructural
details have been presented recently by Fang et al. in Met and Mat Transactions, 27A,
1535, (1996). They interpret their observations as showing that the needles or plates
within the sheaves of lower bainite are built up of smaller units. They conclude that the
bainite plates consist of subplates which consist of subunits which are made up of sub-
subunits which consist of (you guessed it!) sub-sub-subunits. This conclusion, in part,
concerns interpretations of topographical observations made using the relatively novel
technique of scanning, tunneling microscopy and its validity awaits confirmation.
As the temperature of the austenite being quenched falls below the Ms temperature, a
transformation of structure begins to take place while the temperature is falling. This
marks the beginning of the martensitic transformation. A martensitic transformation can
be defined as a mechanism for changing crystallographic structure that does not require
atomic diffusion. In most steels, this transformation occurs very rapidly during the
quench and the fraction transformed depends on the temperature to which the sample is
cooled, not on the length of time spent at that temperature. We note here that the quench
rates that are of practical importance are those obtainable by plunging the steel into liquid
baths such as iced brine; the effect of these treatments are under discussion here.
Eventually, the transformation ceases at the M
f
temperature.
It is customary to call the phase produced by the martensitic transformation, martensite. It
is important to recognize that the transformation rarely goes completely to 100%. A
small amount of the microstructure in a eutectoid plain carbon steel can be retained
austenite. The amount of retained austenite that is found in steels varies with
composition tending to be more prevalent in steels containing high contents of carbon
and other alloying elements. Obviously, if the M
f
temperature is below that of the
quenching bath and room temperature, austenite will be retained. For the sake of
R. H. Wagoner MSE 661 – Steel Section 2, Page 25
completeness we might as well note also that there are some steels in which it is possible
for a martensitic transformation to occur as an isothermal transformation. Hence, the
definition of a martensitic transformation does not include the word quenching. If a
sample of a steel is partially transformed to martensite by being quenched to a
temperature above the M
f
temperature and then held at that temperature, the
untransformed austenite can decompose to form bainite.
Kinetics of Ferrite Nucleation in Hypoeutectoid Steels
Suppose that samples have been austenitized and rapidly quenched to and held at
temperatures below the A
3
point and above the A
1
line. At these temperatures the steel
will tend to transform to a mixture of austenite and ferrite because it is supersaturated
with respect to the ferrite which means that the free energy change for forming a unit
volume of ferrite from the austenite becomes negative (this refers to the bulk or chemical
free energy). When the magnitude of this free energy change is large enough to
compensate for other positive terms, such as the energy of creating new surfaces, the
nucleation flux of ferrite crystals becomes significant.
The start of the transformation is controlled by nucleation because the transformation is
controlled by nucleation and growth. The nucleation flux depends on the free energy of
formation of a critically sized nucleus. Experimental observations show that the ferrite
nucleates heterogeneously on austenite grain boundaries. Let us suppose that lens shaped
embryos form and each one eliminates an area of the grain boundary. Let the volume be
V, the grain boundary area eliminated be A
b
, the surface area of the particle be A
n
, the
surface energy of the ferrite/austenite interface be γ per unit area, γ
b
be the specific grain
boundary energy and suppose that there is no strain energy involved. Then, the free
energy of formation of the embryo is:
∆G V. ∆G
v
+ A
n
. γ − A
b
. γ
b
and the critical value is the maximum value for a given shape.
Note that the critical value decreases as the magnitude of the (negative) free energy of
formation of ferrite increases. This quantity is proportional to the under cooling below the
A
3
line. (See the next heading for a calculation of how the free energy varies with
undercooling.) Inspection of the phase diagram shows that for a particular temperature of
transformation, the under cooling increases as the carbon content decreases because of
the slope of the A
3
line. Hence, the nucleation flux is enhanced and the ferrite forms
earlier as the carbon content is reduced in steels transformed at the same temperature.
The incubation time for the formation of proeutectoid ferrite continues to decrease for a
given steel as the transformation temperature decreases below the A
1
line. Also, at the
lower temperatures, morphological changes such as the formation of Widmanstatten side
plates of ferrite occur. Furthermore, nucleation of ferrite plates begins to take place
within the austenite grains.
Some time after the ferrite begins forming the pearlite transformation starts. The
magnitude of the time interval between the ferrite start curve and the pearlite start line
R. H. Wagoner MSE 661 – Steel Section 2, Page 26
decreases as the temperature is lowered until either the two lines meet the ordinate axis or
they merge at the nose of the diagram. The curves approach each other more and more
closely as the temperature decreases because nucleation of pearlite becomes easier due to
the increased driving force at the lower temperatures and also because of the carbon
concentration immediately in front of the growing ferrite . Once a pearlite colony is
nucleated it can grow more rapidly than the ferrite because the growth rate of the latter is
controlled by long range diffusion. A consequence of this is that the final microstructure
of the steel can often contain a smaller volume fraction of proeutectoid ferrite than that
which would be predicted by applying the lever rule to the phase diagram.
The martensitic transformation attempts to convert the FCC iron to the BCC allotrope.
The product is BCT because of the carbon incorporated in the crystals. The martensitic
phase can then logically be regarded as an extension of the ferrite and therefore the
driving force for its formation derives from the undercooling below the A
3
line. Because
the martensite is a metastable phase, the undercooling required for its formation is larger
than that required for the formation of ferrite.
The M
s
temperature is determined by the degree of super cooling necessary to provide an
adequate driving force. In order to attain the high undercooling, the cooling rate must be
fast enough so that a transformation that takes place by diffusion controlled mechanisms
doesn’t have time to occur. When this is so, the transformation begins when the M
s
point
is passed. Thus the first reason for an increase in the M
s
temperature as the carbon
content decreases, is because the A
3
point moves to higher temperatures thereby
increasing the amount of undercooling for a given transformation temperature. Also, the
degree of supersaturation of the martensite with respect to carbon decreases as the total
carbon content decreases. The reduction in the amount of carbon in the martensite
reduces the strain energy associated with the misfit of the carbon atoms with the
interstitial sites. This reduces the required driving force and hence the necessary
undercooling. For these reasons the M
s
temperature increases as the carbon content falls.
In the case of hypereutectoid steels, similar comments can be made with respect to the
proeutectoid cementite phase. However, the martensite can still be regarded as an
extension of the ferrite, and the austenite undercooling is determined by the extrapolation
of the ferrite plus austenite/austenite phase boundary below the A
1
line. Thus, the
magnitude of the undercooling at a particular transformation temperature decreases as the
carbon content rises. Simultaneously, there is an increase in strain energy within the
martensite due to the increasing carbon conten. These two effects combine to depress
the M
s
point further.
Kinetics of Pearlite Formation – Zener’s Analysis
The initiation process in the transformation that produces pearlite is the formation of
neighboring nuclei of ferrite and cementite on a grain boundary in the austenite. Suppose
that first of all a nucleus of cementite forms on the boundary. It will be rich in carbon
which it has obtained from its immediate surroundings. This reduction in the carbon
R. H. Wagoner MSE 661 – Steel Section 2, Page 27
content encourages the nucleation of ferrite next to the nucleus of cementite. If this
occurs and the cooperative process continues, adjacent nuclei of alternating ferrite and
cementite are formed. These can then grow by a relatively short range diffusion process
in which carbon diffuses parallel to the reaction front from in front of the growing ferrite
to the growing cementite and the diffusion distance does not increase with time as would
be the case for a continuous precipitation process. This is a plausible process that leads to
the lamellar product. The colony can add more plates by additional nucleation or by
branching events in which one plate, say a cementite plate, bifurcates into two.
Consider the steady state nucleation flux equation:
J A. exp−
∆G

kT
j
(
,
\
,
(
which relates the nucleation flux of particles of a new phase to the free energy of
formation of a critically sized nucleus, the absolute temperature and Boltzmann's
constant. There is no critically sized nucleus at the A
1
line because all three phases can
coexist-exist in equilibrium at that temperature. The nucleation flux is zero because there
is no chemical free energy available to provide the energy required to form an interface
between a nucleus and the matrix. The chemical free energy of formation of ferrite and
cementite from austenite becomes negative at temperatures below the A
1
line, and, as the
temperature drops, nucleation becomes possible. To illustrate the essence of the process,
let us consider a simpler case; one in which a single phase, a, transforms to another
single phase, c, and for which ∆G
v
is the chemical free energy change for forming a unit
volume of the new phase. Let us relate this quantity to the temperature.
Let G
a
and G
c
be the molar free energies of the two phases and let H and S refer to the
molar enthalpy and entropy respectively. Then G
a
= H
a
-TS
a
and G
c
= H
c
- TS
c. The
free energy for forming one mole of c from one mole of a:
∆G = G
c
- G
a
∆G is zero when T = T
e, where Te
is the temperature at which a and c coexist in
equilibrium. Using this fact and neglecting any changes in S and H caused by changes in
the temperature, it is easily shown that
∆ ∆ ∆

∆ G = G = T H
T
T
or
H
T
e e
1−
j
(
,
\
,
(
j
(
,
\
,
(
,
where ∆t is the undercooling. Division by the molar volume of the new phase gives ∆G
v
.
Note that when T>T
e
then ∆H is negative and vice versa.
Thus we see that as the temperature of the austenite is decreased below the A1 line, the
absolute magnitude of the free energy change driving the nucleation processes in the
transformation increases. This suggests that critically sized nuclei become smaller and
smaller as the temperature falls and that the nucleation flux should increase because of
R. H. Wagoner MSE 661 – Steel Section 2, Page 28
the decrease in the magnitude of the free energy of formation of a critically sized nuclei.
The free energy of formation of the critically sized nucleus is in the exponential term of
the equation for the nucleation flux and its changes dominate the pre-exponential term
(which contains the diffusion coefficient) in the temperature range above the nose. Of
course, the nuclei to which we are referring, form at heterogeneous sites on the grain
boundaries and the grain boundary energy that is thereby released augments ∆G. The fact
that the driving force for the phase transformation to occur increases as the temperature
falls, provides an explanation for the observation that the incubation period for the
transformation to pearlite decreases as the temperature falls. This is a common
observation for a solid state precipitation reaction that occurs by nucleation and growth.
Now we should recognize that the pre exponential factor contains the diffusion
coefficient of the rate controlling diffusion process by which the transformation takes
place. This gets smaller as the temperature falls and eventually the incubation period
should increase again as cooling continues.
If the pearlite reaction continued to take place as the temperature is continually
decreased, the decrease in the diffusion rates would become eventually become dominant
and slow down the beginning of the transformation. Also, to determine, experimentally,
that the transformation has begun, it is necessary that the products of the transformation
be large enough to be observed. Usually this means that the new phases have grown
considerably larger than the sizes of critical nuclei when they are visible. The growth
rates are influenced by diffusion rates and the driving force derived from the free energy
change.
At temperatures near the eutectoid isotherm the driving force is small and the growth
rates are correspondingly small. However, as the temperature is decreased the growth
rate first increases and then decreases as the diffusion rates become dominant.. These
effects lead us to predict that the kinetics of the transformation to pearlite would be
described by "C" shaped TTT diagram. In the particular case that we are considering, a
“C” curve is observed. However, the situation is more complicated than implied above
because at the nose of the TTT diagram there is a change in the mode of transformation
from that which produces pearlite to one which forms ferrite and cementite in the non-
lamellar, bainite microstructure. This transformation also involves nucleation and growth
controlled by diffusion. Hence it, too, should have kinetics described by “C” curves.
Evidently the pearlite reaction is faster above the nose of the TTT diagram while the
bainite reaction is faster below it. The “C” curves describing the two reactions merge at
the nose.
The growth rate of a pearlite colony depends on the rate of diffusion of carbon. This
diffusion process is required to move the carbon from the volume of austenite that is to
become ferrite into the volume that is to become cementite. The diffusion distances are of
the order of the inter-lamellar spacing. Metallography shows us that the inter-lamellar
spacing is largest in pearlite that is formed at the highest temperatures and that it
decreases as the temperature at which the transformation takes place is decreased. Why?
R. H. Wagoner MSE 661 – Steel Section 2, Page 29
Consider Figure 2-13 to represent a portion of pearlite in which the spacing, S, extends
from the centers of two neighboring cementite lamellae. For simplicity, curvature of
interfaces is neglected.
Figure 2-13: Schematic of the austenite-to-pearlite tranformation of extent δx, in time δt.
Suppose the transformation interface advances a small distance δx. The change in the
free energy, per unit area of the interface, is given by:
∆G ∆G
v
.δx +

S
.δx
where the first term on the right hand side contains the chemical free energy released per
unit volume of pearlite formed and the other term contains the interfacial surface energy
created.
At a particular temperature, the total free energy change varies with the spacing of the
lamellae. As the spacing decreases, the interfacial area created increases. There is a
critical spacing at which the free energy change becomes zero. Pearlite of smaller spacing
will not form because it would raise the free energy of the system.
Let the critical spacing be given by S
min
. Then S
min


∆G
v
∆G
v
is, of course, negative. Hence ∆G

S
δx −

S
min
δx. When δx 1, ∆G refers to a
unit volume of the pearlite and then ∆G 2γ
1
S

1
S
min
¦
¦
¦
¦
¦
¦
.
The growth of the pearlite in the direction shown, requires carbon to diffuse laterally at,
or ahead of the interface. A number of treatments of this process have been published.
We will consider one of the original treatments, (Zener, TAIMME Vol. 167, 550, 1946).
It is not a detailed rigorous treatment but it serves well to illustrate the nature of the
problem; many subsequent analyses have sought to elaborate on Zener’s work.
S
α
γ
Fe
3
C
δx
R. H. Wagoner MSE 661 – Steel Section 2, Page 30
Experimental studies show that the growth rates of pearlite colonies at fixed temperatures
are constant and independent of time. This observation tells us that the diffusion process
is a steady state process. For diffusion across an area, A, due to a concentration gradient
∆C
∆X
we can write the flux according to Fick’s first law:
J −D. A.
∆C
∆X
Let us suppose that the rate of growth is such that the interface moves into the austenite at
u cms. per sec. As the interface moves, the carbon composition in the volume that
becomes ferrite, changes from C
γ
( the original composition of the austenite) to C
α
. The
compositions refer to volume. The width of the ferrite lamella is some fraction of the
spacing S. Let us make it S.a, where a is a number less than one. Then, neglecting
differences between the specific volumes of austenite and ferrite, the amount of solute
that was transferred during one second of growth is
u. S. a. C
γ
− C
α
( )
This is the flux of solute, Hence, omitting the negative sign which merely indicates that
diffusion occurs down the concentration gradient, we have:
D. A.
∆C
∆X
u. S. a. C
γ
− C
α
( )
The true form of the concentration gradient is unclear. However, if we assume that the
concentration is C
α
at the interface between the ferrite and cementite, and Cγ ahead of
the center of the ferrite lamella, then an approximate expression is:
∆C
∆X

C
γ
− C
α
( )
b. S
j
(
,
\
,
(
where b is another number less than unity.
The next problem we face is what to do with the area A. We have two extremes; the
diffusion flux might be restricted to the boundary between the austenite and the pearlite,
or it could be occurring over some distance ahead of this interface. Let’s suppose the
latter is the case, What can we chose for A? Who knows? If diffusion does, indeed, take
place in the austenite some distance ahead of the interface, then it seems reasonable that
the distance will be smaller, the smaller the spacing becomes. Zener chose to make this
distance approximately equal to S. Using all these approximations we have:
u ≅
D. S. C
γ
− C
α
( )
S. a. C
γ
− C
α
( )
. b. S
,
¸
,
,
]
]
]
]

D
a. b. S
R. H. Wagoner MSE 661 – Steel Section 2, Page 31
If the thickness of a ferrite lamella is much greater than that of a cementite plate, then a.b
is about a half.
This approximate treatment indicates that the growth rate at a given temperature is
proportional to D and inversely proportional to S, a result that can be predicted
intuitively. An important aspect of this relationship is that if it had been derived by means
of a rigorously correct method it would still be a relation in which u is some function of
the variables D and S. The point of this remark is that a theoretical description that
enables one to evaluate u when S and D are known, is not particularly enlightening. It
would be much more useful to be able to predict both u and S when only D is known. The
meaning of this should become clearer as we continue with Zener.
In order to evaluate u and S explicitly, when D is known, we need another equation
relating the three variables. In an attempt to obtain one, Zener noted that the rate of
migration of an interface is often expressed as the product of a mobility and a driving
force. Hence he wrote:
u ∝ ∆G
.
Then, because u is proportional to D and inversely to S we find that
u ∝
D
S
∆G
Recal l i ng t hat
∆G 2γ
1
S

1
S
min
j
(
,
\
,
(
and subst i t ut i ng we obt ai n
u
K. D
S
.2γ .
S
min
− S
S. S
min
j
(
,
\
,
(
where K is a constant.
However, u and S cannot be separately evaluated. In this kind of a situation it is common
to search for some extremum principle. Zener proposed that the spacing that would occur
would be that which would maximize u. There is no obvious reason why this should be
so, but it is a reasonable suggestion because the colonies that grew the fastest should
make up most of the pearlite at the end of the transformation and experimentally
determined spacings should belong to them. Assuming this to be so, we can differentiate
u with respect to S and equate the result to zero. Then, after a little algebra, we find that
the spacing is twice as big as the minimum possible spacing.
Now the magnitude of the chemical free energy change for forming a unit volume of
pearlite is proportional to the undercooling below the equilibrium temperature, ∆T=(Te-
T). Hence we may write:
∆ ∆ G q T
v
.
where q is a constant.
Because ∆G
v


S
min
then
S

q. ∆T
R. H. Wagoner MSE 661 – Steel Section 2, Page 32
Although this equation is not quantitatively correct, it is of the correct form because the
average spacing in pearlite varies inversely with the undercooling.
A number of theoretical descriptions of the growth of pearlite have been published since
this early treatment. References to these works can be found in Honeycombe’s book. In
general, they attempt to refine Zener’s work by trying to take more parameters into
account e.g. considering curved interfaces and the effects of alloying elements but the
difficulty of separating S and u explicitly remains.
Pearlite has been the subject of many investigations in which the characteristics of it’s
formation are compared to the predictions of early theories of nucleation and growth
kinetics. It has also been the subject of crystallographic orientation studies. These studies
are interesting and relevant to the topic of phase transformations, however, a knowledge
of specific details is unnecessary for our purposes here.
Consequences on TTT Diagrams
The mechanisms discussed above may be used to make predictions on how TTT
diagrams will vary depending on several quantities.
Carbon content: the role of increasing carbon content in plain-carbon steels will delay
the transformation of austenite to pearlite (or austenite to proeutectoid ferrite or
proeutectoid cementite) because of two factors:
• The nucleation, and possibly to some extent the growth of product phases, will be
slower at a given temperature because carbon is an austenite stabilizer. That is,
adding carbon tends to expand the gamma phase field (see the phase diagram,
Figure 1-1 for verification), thus reducing the driving force to form pearlite.
• The growth of the product phases will be inhibited significantly because more
carbon must be moved (via diffusion) for a given mole or volume of product
material.
Thus, we would expect that the additions of carbon to move the TTT curves to the right,
exactly as may be observed in Figure 2-14.
Another prediction from nucleation rates is that austenite of smaller grain sizes will
transform to product phases more rapidly because the presence of additional sites of
nucleation. This prediction is also correct, as shown in Figure 2-15, which have as the
main variable the austenite grain size (although the differing contents of Mn cannot be
excluded).
A more direct verification of the grain size effect appears in Figure 2-16. In this case, the
“hardenability” of a steel is carefully measured as a function of austenite grain size, with
the results shown. “Hardenability” and its measurement will be defined more fully in the
next section, but it is sufficient to know at the moment that is simply a measure of the
minimum time to start the austenite-to-pearlite transformation (i.e. the time to reach the
“nose” on the TTT diagram).
R. H. Wagoner MSE 661 – Steel Section 2, Page 33
Figure 2-14: Effect of carbon content on TTT diagrams.
800
700
600
500
400
300
200
100
0
T
e
m
p
e
r
a
t
u
r
e
200
400
600
800
1000
1200
1400
Estimated
Temperature
1 MIN
0.51 2
5
10 102 103
1 HOUR
104 105 106
1 DAY
1 WEEK
Time - Seconds
οοο οF οοο οC
I - T Diagram
M90
S
S
A
As
F + C
A
+
F
A
+
F
+
C
800
700
600
500
400
300
200
100
0
T
e
m
p
e
r
a
t
u
r
e
200
400
600
800
1000
1200
1400
1 MIN
0.51 2 5 10 102 103
1 HOUR
104 105 106
1 DAY
1 WEEK
Time - Seconds
A
As
F+C
οοο ο
F
οοο ο
C
I - T Diagram
Ms
A
+
F
+
C
A+F
Af
90%
800
700
600
500
400
300
200
100
0
T
e
m
p
e
r
a
t
u
r
e
200
400
600
800
1000
1200
1400
Estimated
Temperature
1 MIN
0.51 2 5 10 102 103
1 HOUR
104 105 106
1 DAY
1 WEEK
Time - Seconds
A
As
A
F +
C
οοο ο
F οοο ο
C
I - T Diagram
A
Ms
M50
M90
S
S
S
50%
Af
A+F
A+F+C
800
700
600
500
400
300
200
100
0
T
e
m
p
e
r
a
t
u
r
e
200
400
600
800
1000
1200
1400
Estimated
Temperature
1 MIN
0.51 2 5 10 102 103
1 HOUR
104 105 106
1 DAY
1 WEEK
Time - Seconds
A
As
A
F + C
οοο ο
F
οοο ο
C
I - T Diagram
A
M50
M90
S
S
S
S
50%
A+F+C
Ms
800
700
600
500
400
300
200
100
0
T
e
m
p
e
r
a
t
u
r
e
200
400
600
800
1000
1200
1400
Estimated
Temperature
1 MIN
0.51 2 5 10 102 103
1 HOUR
104 105 106
1 DAY
1 WEEK
Time - Seconds
A
As A
F + C
οοο ο
F
οοο ο
C
I - T Diagram
A
M50
M90
S
S
S
S
50%
Ms
S
A+F+C
Ms
S
M50
S
R. H. Wagoner MSE 661 – Steel Section 2, Page 34
Figure 2-15: Effect of grain size on TTT diagrams. Notice also the differing Mn
contents.
800
700
600
500
400
300
200
100
0
T
e
m
p
e
r
a
t
u
r
e
200
400
600
800
1000
1200
1400
1 MIN
0.5 1 2 5 10 102 103
1 HOUR
104 105 106
1 DAY 1 WEEK
Time - Seconds
A
As
A
F + C
26
33
35
33
38
42
50
55
65
H
a
r
d
n
e
s
s
-
R
C
οF
οC
I - T Diagram
A
Ms
M50
M90
50%
Af
A+F
A+F+C
Type: 1060, 0.63C, 0.87Mn
Grain size=5-6
T
austenitize
=816
οοο ο
C
T
e
m
p
e
r
a
t
u
r
e
800
700
600
500
400
300
200
100
0
200
400
600
800
1000
1200
1400
Estimated
Temperature
1 MIN
0.5 1 2 5 10 102 103
1 HOUR
104 105 106
1 DAY 1 WEEK
Time - Seconds
A
As A
F + C
18
23
28
32
36
42
48
52
62
H
a
r
d
n
e
s
s
-
R
C
οF
οC
I - T Diagram
A
Ms
M50
M90
S
S
S
S
50%
Af
A+F
A+F+C
1050, 0.5C, 0.91Mn
Grain size=7-8
T
austenitize
=910
οοο ο
C
800
700
600
500
400
300
200
100
0
T
e
m
p
e
r
a
t
u
r
e
200
400
600
800
1000
1200
1400
1 MIN
0.5 1 2 5 10 102 103
1 HOUR
104 105 106
1 DAY 1 WEEK
Time - Seconds
Ae1 A
F + C
21
27
300
31
35
45
62
H
a
r
d
n
e
s
s
-
R
C
οF
οC
I - T Diagram
A
Ms
M50
50%
Ae3
A+F
A+F+C
1045/1050, 0.47C, 0.57Mn
Grain size=50% 8, 50% 5
T
austenitize
=843
οοο ο
C
R. H. Wagoner MSE 661 – Steel Section 2, Page 35
Figure 2-16: Hardenability (time to nose in TTT diagram) as a function of austenite grain
size.
SECTION 2 - CONCEPTS AND STUDY QUESTIONS
Important concepts
Distinctions and similarities between continuous cooling and isothermal transformations.
Why are both kinds of cooling used in creating transformation diagrams?
Phase transformations and microstructures in pure iron: gamma to alpha, equiaxed
grains; blocky/faceted grains – massive; widmanstatten ferrite; martensite. Kinetics and
characteristics of each.
Principles of nucleation and growth. Course vs. fine structures – why?
Stages of austenitization.
Transformations in eutectoid steels: coarse/fine pearlite, bainite, martensite. Incubation
times, fractional transformation concept for C-T diagrams. Temperatures in eutectoid
steel where such transformations occur isothermally.
Appearance of transformation products.
Upper vs. lower bainite. Similarity to tempered martensite. Difficulty of resolving such
fine structures.
1 0
1 4
1 8
2 2
2 6
0 0 . 2 0 . 4 0 . 6 0 . 8 1
Carbon content (%)
B
a
s
e

h
a
r
d
e
n
a
b
i
l
i
t
y
,

D
i
C

(
m
m
)
(0.10 mm)
(0.025)
(0.036)
(0.05)
(0.07)
Grain size
R. H. Wagoner MSE 661 – Steel Section 2, Page 36
Differences of hypoeutectoid and hypereutectoid phase transformations from eutectoid
compositions.
TTT diagrams for plain-carbon steels as function of carbon content and prior austenite
grain size. Why?
C-T diagrams of two types, how to construct conceptually from TTT diagrams.
Zener’s analysis of pearlite formation. Thermodynamics and kinetic factors. Qualitative
results and assumptions.
Study Questions
1. Estimate the amount of proeutectoid ferrite at 20 deg. C in a well-annealed and very
slowly cooled sample of AISI 1050 steel. What changes would you expect to see in the
actual amount present if samples were prepared with faster and faster cooling rates from
the annealing temperature? The annealing temperature is well within the austenite phase
field.
2. Samples of AISI 1080 steel were heat treated to produce certain microstructures. The
microstructures were:
a) Almost 100% fine crystals of white martensite.
b) Colonies of fine, uniformly spaced pearlite lamellae.
c) Colonies of pearlite lamellae in which the lamellae spacing varied with position in
the sample.
d) A mixture of coarse pearlite embedded in patches of lower bainite.
e) Equiaxed ferrite grains containing a dispersion of cementite globules.
f) Fine grained, acicular ferrite containing very fine needles and platelets of cementite.
Describe the heat treatments that could have been used. Include explanations for your
choices.
3. Sketch a TTT diagram for a 1080 steel. Explain, in general, the reasons for its shape
and then explain, in detail, the processes that control the shape of the diagram below the
A1 line but above the nose of the diagram.
4. Sketch a TTT diagram for a 1040 steel and explain the reasons for any differences it
has compared to the one for 1080 steel.
5. Two experimenters separately determined (experimentally) a TTT diagram for a 1080
steel. Both experimenters used samples of the same steel and the same equipment. The
two diagrams were similar in form, but one had a nose that was significantly closer to the
ordinate axis (which corresponded to the time 0.1 sec) than the other. Offer explanations
for this difference.
R. H. Wagoner MSE 661 – Steel Section 2, Page 37
6. Describe and explain the microstructural changes that occur as a function of time
during an austenitization treatment given to a 1040 steel that had previously been hot
rolled at 1000 deg. C and air cooled.
7. A small sample of 1080 steel was fully austenitized at 1650 deg. F (890 deg. C) and
then rapidly quenched to 550 deg. F (288 deg. C) and held there for one hour. IT was
then up-quenched to 1310 deg. F (710 deg. C) held there for 5 minutes and rapidly
quenched into ice brine. What is its final microstructure and why?
8. A small sample of a 1035 steel was austenitized above the A3 line, very rapidly
quenched to 1370 deg. F (743 deg. C) and held there until equilibrium pertained and then
very rapidly quenched in ice brine. A similar sample of the same steel was similarly
austenitized, rapidly quenched to 1100 deg. F (593 deg. C) and held there until in
equilibrium and finally quenched in ice brine. Compare in detail the microstructures
obtained. If you predict there to be any differences, explain why.
Physical Metallurgy of Steel
Class Notes and lecture material for MSE 661 (formerly 651.01)
Notes revised by: Robert H. Wagoner, Smith Chair
Notes originally compiled by: Glyn Meyrick, Professor Emeritus
Web installation by: Wei Gan, Graduate Research Associate
Last revision date: 9/30/02
SECTION 3 – MARTENSITIC TRANSFORMATIONS
Background
As discussed in Section 2, rapid quenching of plain-carbon steel (indeed, even pure iron)
produces a special phase transformation, martensitic one, that does not depend on
diffusion to proceed. While we have discussed some of the attributes of the product
martensite structure, in this section we proceed in more detail to describe what the
martensite phase looks like, how it is obtained, and how its properties can be modified by
subsequent treatment. We also include some standard heat treatments.
A note is required here: the product of a martensitic transformation is called martensite,
whether a special phase or only a special microstructure is involved. This is confusing in
steels containing normal amounts of carbon, where the product martensite is indeed a
new phase, with a body-centered tetragonal microstructure. In pure iron, however, the
product is a very fine microstructure of normal bcc ferrite.)
The quench rates that are required in most steels to suppress the nucleation and growth
reactions (i.e. the nose of the TTT diagram) are obtainable by plunging the steel into
liquid baths such as iced brine. The Ms temperature must be reached with at least some
remaining austenite to transform to martensite. When the M
f
temperature is reached, the
martensitic transformation has proceeded to completion (although this seldom means that
100% of the austenite has been transformed to martensite), but all austenite that can be
transformed, has been transformed.
The Martensite Phase in Carbon-Containing Steel
Earlier we discussed experiments in which iron and dilute iron carbon binary alloys were
quenched at rates above 10
5
K/sec. At these rates there is insufficient time for the
structural change to be produced by diffusion. However, the samples became BCC.
Evidently the iron knows another way to change its structure and, when prevented from
doing it by means of diffusion, does it without using diffusion. Instead, it does so by
means of a displacive transformation by which it is possible to displace the atoms from
FCC positions to BCC positions without requiring them to hop from site to site by
R. H. Wagoner MSE 661 – Steel Section 3, Page 2
diffusion. The transformation is called martensitic. The displacements associated with the
transformation in iron are illustrated in Figure 3-1. These show what is called the Bain
strain. This is a strain proposed by Edgar Bain to describe the distortion of a tetragonal
cell in the FCC crystal into the BCC of the martensite.
Figure 3-1: Concept of the Bain strain. In the upper figure, a BCT cell is shown
superimposed on two FCC cells. The BCT cell is removed below and is distorted along
the cell edges to produce a BCC cell.
Now, when we turn to the case of a plain carbon steel and quench its austenite too rapidly
for any change to be wrought by diffusion, the iron atoms still wish to adopt BCC
positions. Cementite can only form when diffusion is possible because one out of four
atoms in it is C, but the martensitic change is available for the iron. The possibility arises
that the FCC solid solution can change to a BCC solid solution. This can happen if the
free energy released is negative even though the transformation produces an intermediate
metastable state. We find that it begins to happen when ∆T becomes large enough, i.e. the
temperature falls to below the M
s
temperature.
The martensite in pure iron is the stable phase (i.e. bcc ferrite), even though it contains
numerous defects such as dislocations and boundaries. Because the carbon content is
unchanged during the formation of martensite in a plain carbon steel, the phase is
[001]
γγγ γ
[001]
ααα α’
[100]
γγγ γ
[100]
ααα α’
[010]
ααα α’
[010]
γγγ γ
a
o
a
o
/ 2
a
c
R. H. Wagoner MSE 661 – Steel Section 3, Page 3
metastable and is a supersaturated solid solution if its carbon content exceeds the
solubility in ferrite.
Inspection of Figure3-2 illustrating the Bain strain shows that the carbon atoms that were
previously randomly distributed in the octahedral sites of the austenite now occupy a
particular selected set of sites within the martensite. When the carbon content exceeds
about 0.2 wt.% the carbon atoms produce a significant tetragonal distortion so that the
structure becomes BCT. The c/a ratio of the tetragonal unit cell increases as the carbon
content Increases, Figure 3-3 . To a good approximation, the variations in c and in a are
linear, with the c value increasing at a greater rate than the rate at which the a parameter
decreases, as the carbon concentration increases.
Figure 3-2: Stucture of carbon-containing martensite. (After Bain and Paxton, p. 35.)
c-axis
a-axis
Range of positions for iron atoms
Probable positions for carbon atoms
R. H. Wagoner MSE 661 – Steel Section 3, Page 4
Figure 3-3: Dependence of martensite BCT lattice parameters on carbon content.
Martensite Microstructures
The microstructure of martensitic steel is complex and varies in detail from steel to steel.
Examples are shown in Figure 3-4. Details depend upon the M
s
temperature. In general
lath martensite forms in the high temperature range and lenticular or plate-like
martensite forms at lower temperatures. For plain carbon steels this translates into a
dependence of the microstructure on carbon content because increasing the carbon
content lowers the M
s
temperature.
Martensitic crystals in low carbon steels are fine laths which contain a high density (
>10
11
cms./cm.
3
) of dislocation line. As the carbon content is increased, crystals shaped
like plates appear. Twins begin to replace the dislocations within the plates so that high-
carbon martensite is composed mainly of twinned plates. The transformation is also
associated with an appreciable volume increase because it replaces the close packed FCC
structure with a BCT structure of smaller density. Because of this, the transformation
creates residual stresses and strains related to the specific volume change in addition to
the strains due to the misfit of the interstitial solute atoms. At high carbon levels these
stresses can become so severe that the material cracks as the martensite forms. The
cracks can range from small micro cracks that require microscopy to be detected, to large
cracks easily visible to the unaided eye.
The crystals of martensite vary from fine laths in low carbon steels, to plates in high
carbon steels. The lath morphology is such that the laths form in packets of nearly
parallel units; the individual units are often too small to be resolved in an optical
microscope. The crystals have habit planes that tend to lie close to the {111} planes in the
austenite. Orientations of {557} have been reported. Because these indices have two
c-axis
a-axis
c/a = 1 + 0.045 (wt% C)
3.05
3.00
2.95
2.90
2.85
2.80
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
Carbon, %
L
a
t
t
i
c
e

d
i
m
e
n
s
i
o
n
,

a
n
g
s
t
r
o
m
R. H. Wagoner MSE 661 – Steel Section 3, Page 5
identical digits, there are 12 non parallel variants. However, they occur in groups of three
with each group clustered about a (111) with only 16 degrees between members of a
group. Because there are only 4 non parallel variants for {111}, lath martensite appears to
be more orderly than does plate martensite.
Figure 3-4: Fe-0.8C alloy quenched from 1100°C: a, optical micrograph x200; b, thin-
foil EM showing twinning in martensite laths. (From Honeycombe, p. 95.)
The habit plane of the plate martensite in carbon steel varies from {225} for steels with
0.92 wt.% C to {259} for 1.78 wt. % C. There are 12 variants for the former and 24
variants for the latter. Although each martensitic crystal could, in principle, choose any of
these variants there is a tendency for certain variants to occur in groups. This is because
the shape and volume change associated with the transformation are resisted by the
austenitic matrix. This resistance encourages the choice of those variants which, in
combination, provide the minimum macroscopic shape change.
Lath martensitic crystals contain a dense population of dislocations that are associated
both with the transformation itself and with the strains generated by the volume change.
Plate martensite is much less dislocated; instead, the plates contain arrays of very fine
twins. Excellent examples of martensitic microstructures can be found in chapter 3 of
Krauss' book.
Shape Change from Martensitic Transformation
An important characteristic of the martensitic transformation is that it produces surface
relief on a metallographically polished surface, as shown schematically in Figure 3-5.
This is due to the displacive nature of the transformation. When a portion of the austenite
is converted to the other crystal structure, its physical shape is also changed. The total
strain resulting from a martensitic transformation is the result of the shape change
produced by the crystallographic change itself combined with post-transformation strains
caused by creating the martensitic crystals within the austenite matrix. Experimental
studies have shown that the transformation shape change transforms planes into planes
R. H. Wagoner MSE 661 – Steel Section 3, Page 6
and straight lines into straight lines but only special planes remain unrotated. Hence,
portions of the free-surface of the sample become tilted from the previous orientation as
the martensite forms.
Figure 3-5: Habit planes and surface relief of unconstrained martensite transformation.
The total change in the shape caused by a martensitic transformation is the result of the
shape change associated with the transformation alone and the strains arising from the
interaction between the martensite and the matrix. It is usual to separate the two by
calling any strains that do not refer to the transformation of the structure, post-
transformation strains. An important result of the shape change is that previously flat
surfaces become tilted and rumpled during the transformation.
Studies of the formation of plate martensite have shown that the general shape change
can be described as being produced by a homogeneous strain, i.e. a strain which does not
change the degree of an equation. This conclusion has been reached because planes are
tilted but remain planes, straight lines (such as a scratch on the surface of the sample)
become kinked but remain made up of straight segments. There are distortions but these
are attributed to the post-transformation strains, which are not relevant to the actual phase
transformation mechanism. During the transformation, a crystal of plate martensite in
contact with austenite at the habit plane clearly shares the same interface; hence, we
conclude that if we were to describe the transformation by a homogeneous strain then the
strain would have to leave at least one plane (the habit plane) of the austenite unchanged.
Thus the strain that describes the transformation of austenite to martensite is a
homogeneous invariant plane strain.
Theoretical attempts to determine the strain have led to the phenomenological theory of
the martensitic transformation. The objective of this theory is to determine a strain that
would produce both the crystallographic change and the shape change simultaneously
when applied to austenite; it does not directly concern itself with the mechanism by
M
Habit
Plane
Free surface
Old position
New position of scratch
Old position of scratch
New position
Old position
γγγ γ
Unconstrained martensitic
transformation forming a plate
γγγ γ
New position
R. H. Wagoner MSE 661 – Steel Section 3, Page 7
which the transformation takes place. This is why it is a phenomenological theory. Let us
consider some elements of this theory.
The first thing that the strain must produce is the conversion of the FCC unit cell into the
BCT unit cell. This is usually referred to as the lattice strain. We have already met a
strain that will accomplish this. It is the Bain strain introduced previously. It is obtained
directly from the lattice parameters of the corresponding cells. The numerical values of
the components of the Bain strain vary with carbon concentration because of the variation
in the c/a ratio of the martensite, Figure 3-3.
First we consider an origin of an orthogonal coordinate system located on a corner of the
tetragonal cell drawn in the FCC lattice in Figure 3-1 such that the x, y and z axes lie
along the edges of the unit cell. The Bain strain causes strains of e
1
, e
2
and e
3
along
these axes but leaves the angles between them unchanged. One can describe the
coordinates of any other point by x, y and z. Then if the lattice is deformed so that the
coordinates become x’, y’ and z’ then
X e X X
Y e Y Y
Z e Z Z

= + =

= + =

= + =
( )
( )
( )
1
1
1
1 1
1 1
1 1
η
η
η
(3-1)
These equations may be written in matrix form conveniently, and we also change
notation from X, Y, Z to the equivalent X
1
, X
2
, X
3
:
X
X
X
X
X
X
1
2
3
1
2
3
1
2
3















=




















η
η
η
(3-2)
where η’s are called the distortions. This may be written, in even more compact form, as
follows:
′ [ ]
= [ ][ ] ′ = X B X or X BX , (3-3)
For ferrous martensite, η
1
and η
2
are equal and greater than one; η
3
is less than one.
The actual values depend upon carbon content (Figure 3-3).
The change in shape produced by applying such a strain to a sphere is shown in the
Figure 3-6. It is clear that the “before” and “after” shapes now have surfaces that
coincide along two circles. This exercise shows that the Bain strain does not leave any
plane in the sphere undistorted or invariant. Thus, whereas it can produce the lattice
change it cannot describe the martensitic shape change. Therefore, it is necessary to seek
additional strains which can change the shape but not the structure of the crystal.
R. H. Wagoner MSE 661 – Steel Section 3, Page 8
Figure 3-6: Conception combination of transformations to reproduce key features of the
Fe-C martensitic phase transition.
Such strains are shears produced by slip or by twinning. Particular shear strains, P, have
been found which when combined with B produces a total strain, PB, that creates the
martensitic crystal structure and leaves at least one plane undistorted. Usually it is found
that when this P is determined the undistorted plane is not invariant but has been rotated.
To take care of this, a rotation R is included to finally produce an invariant plane strain.
The total transformation strain can then be written as
T =RPB
Although the theory is phenomenological, it is interesting to note that when it
successfully describes a transformation, the martensitic crystals are found to be internally
twinned by twins that correspond with the twinning shear used for P. The
z
x
Bain Strain
Invariant
Lengthens
Shear strain, P
Shortens
1. Let the inner sphere be expanded isotropically
by e
1
=e
2
=e
3
. This is a simple size change.
All lines and planes in the inner sphere become
lines and planesin the outer sphere. None are
distorted
2. Apply the Bain Strain, B, to the sphere. Use
typical values of e
1
=e
2
and e
3
for Fe+C
(depending on C content) so that the result is
BCT, not BCC: 0.12, 0.12, 0.02. (Note: in the
BCT cell, c>a, but aγ->c and aγ/√2->a.) The strain
produces the ellipsoid shown. There are common
lines to both, one of which is shown. No plane is
invariant or distorted.
3. Inspection shows that some lines are
lengthened and some shortened. This can be
reversed by a shear strain, which shortens some
vectors and lengthens others.
Thus, there is acombination of strains PB such
that at least one plane is undistorted. Often this
plane is rotated as a rigid body. That rotation can
be reversed using a pure rotation R. The total of
these transformations, T, would thus be
T = R P B
z
x
y
Inner sphere
expands to
outer sphere
R. H. Wagoner MSE 661 – Steel Section 3, Page 9
phenomenological theory has been successful in ferrous alloys e. g. binary Fe-Pt alloys,
but it is only partially successful in carbon steels because it does not predict the variation
of the habit planes observed. This may well be a consequence of the post transformation
strains.
Effect of Carbon on Ms
The martensitic transformation attempts to convert the FCC iron to the BCC allotrope.
The product is BCT because of the carbon incorporated in the crystals (because there is
not time to move it out by diffusion). The martensitic phase can then logically be
regarded as an extension of the ferrite and therefore the driving force for its formation
derives from the undercooling below the A
3
line (which represents the equilibrium
between bcc and fcc iron). Because the martensite is a metastable phase, the
undercooling required for its formation is larger than that required for the formation of
ferrite.
The M
s
temperature is determined by the degree of super cooling necessary to provide an
adequate driving force. In order to attain the high undercooling, the cooling rate must be
fast enough so that a transformation that takes place by diffusion controlled mechanisms
doesn’t have time to occur. When this is so, the transformation begins when the M
s
point is passed. Thus the first reason for an increase in the M
s
temperature as the carbon
content decreases, Figure 3-7, is because the A
3
point moves to higher temperatures
thereby increasing the amount of undercooling for a given transformation temperature.
Figure 3-7: Dependence of the martensite start temperature on carbon content.
Another way to think about it: the degree of supersaturation of the martensite with
respect to carbon decreases as the total carbon content decreases. The reduction in the
amount of carbon in the martensite reduces the strain energy associated with the misfit of
the carbon atoms with the interstitial sites. This reduces the required driving force and
hence the necessary undercooling. For these reasons the M
s
temperature increases as the
carbon content falls.
In the case of hypereutectoid steels, similar comments can be made with respect to the
proeutectoid cementite phase. However, the martensite can still be regarded as an
1 0.6 0.06
wt % C
Ms
1000
400
600
F
R. H. Wagoner MSE 661 – Steel Section 3, Page 10
extension of the ferrite, and the austenite undercooling is determined by the extrapolation
of the ferrite plus austenite/austenite phase boundary below the A
1
line. Thus, the
magnitude of the undercooling at a particular transformation temperature decreases as the
carbon content rises. Simultaneously, there is an increase in strain energy within the
martensite due to the increasing carbon content. These two effects combine to depress the
M
s
point further.
Hardenability of Steels
Hardenability refers to the ability to obtain martensite during cooling from the austenite
phase field. Thus, it is really an inverse measure of the speed with which pearlite is
formed at the nose of the TTT diagram. A simple conceptual measure would be the time
to the nose of the TTT diagram, or alternatively the time to form 50% pearlite at a give
temperature. However, these measures are not used in practice because of the difficulty
of measuring a true TTT diagram and because the practical application of hardenability is
the production of martensite as opposed to the avoidance of pearlite (although they are of
course intimately related). For practical heat-treatment, the C-T diagram is preferred and
we can anticipate that similar constructions yielding various amounts of martensite will
be useful.
Hardenability has little to do with the martensite transformation itself, or the strength of
any of the products. This concept should be contrasted clearly with hardening, which
may occur by a variety of mechanisms. Indeed, a material with high hardenability may
be much softer than one with low hardenability.
When hardenability is important in the material selection process, it will be useful to have
a method of quantitatively comparing the hardenability of steels. How can this be done?
The Problem
Consider a cylindrical bar of steel that has a radius R and a length that is several times
larger than R (so that end-effects can be eliminated from consideration). Now suppose
that this bar is austenitized at T
a
and quenched in a particular quenching medium, say
water at 20°C. After the quench, the bar is sectioned perpendicularly to its length and
hardness measurements are made on the cross section at locations that span the diameter
of the bar.
The hardness value at a particular location reflects the nature of the microstructure at that
location: for the steels that we are considering, the hardest microstructure produced by
continuously cooling austenite is that corresponding to nearly 100% martensite, the
softest microstructure is equiaxed ferrite grains and colonies of coarse pearlite, (the
spheroidized microstructure doesn’t form in these treatments). The hardness varies from
the softest to the hardest condition as the microstructure varies between these two
extremes.
It is important to recognize that the value of a single, particular hardness measurement in
this experiment has no direct significance with respect to hardenability. It is the variation
R. H. Wagoner MSE 661 – Steel Section 3, Page 11
of the hardness with position that is significant because this provides information about
the variations of microstructure. Similar microstructures in different steels (e.g. 100%
martensite) can have quite different hardness. The data obtained are used to create a
graph of hardness versus position in the bar, Figure 3-8.
Figure 3-8: Typical hardness distributions across quenched bars of various diameters for
several steels.
Provided that the radius R and the hardenability of the steel are both appropriate, the
microstructure near the surface can contain the maximum amount of martensite that can
form in the steel when it is quenched to 20°C. The microstructure at the center of the bar
can contain ferrite plus pearlite. Therefore, the variation of the hardness with position
shown on Figure 3-8 reveals the variation of microstructure.
When this experiment is carried out for a variety of steels using bars of the same size, the
same heat treatments and the same quenching medium, a collection of graphs are
obtained. Then we define a “depth of hardening” as the magnitude of the distance from
the outside surface to the position where the hardness is less than that very near the
surface by a specified percentage. When the specified hardness percentage is chosen
(corresponding to some microstructure), the graphs can be used to rank the steels in an
order corresponding to their depths of hardening. The smaller the specified decrease in
hardness, the larger will be the % of martensite in the microstructure. In any case, greater
the depth of hardening the greater the hardenability.
20
30
40
50
60
H
a
r
d
n
e
s
s
,

H
R
C
4140
1040
2 in. (50mm)
4 0
4 5
5 0
5 5
6 0
H
a
r
d
n
e
s
s
,

H
R
C
2 in. (50mm)
4 in. (100mm)
4 1 4 0
4 1 4 0
R. H. Wagoner MSE 661 – Steel Section 3, Page 12
If one is to use the depth of hardening approach one must choose the “specified hardness
percentage decrease”. Should it be 15%, 40%, or what? Let us consider the
microstructures. When a bar of steel is austenitized and quenched so that it is cooled
below the M
f
temperature, it is possible for it to be almost entirely transformed to
martensite or to have a rim of martensite at its surface and various mixtures of martensite,
cementite and ferrite within. Typically, the microstructure varies from almost fully
martensitic through various mixtures that include bainite and pearlite. The thickness of
the martensitic rim can be measured by quantitative metallography if a definition of
thickness can be agreed on (i.e. what is the rim and what isn’t).
Many experiments have shown that for the standard etchants, such as nital, that attacks
intercrystalline and interphase boundaries, the microstructure near 50% martensite is the
most easily revealed and recognized. For this reason (and also because 50% is a nice
stopping point between 100% and 0%!), the 50% martensite microstructure is often
chosen to define the depth of hardening or the critical microstructure.
The rate of variation of the hardness with microstructure is most rapid where the 50%
martensite mixture is located, which means that the errors based on hardness distributions
will be minimized in this vicinity. For these reasons, the 50% microstructural mixture is
most readily located experimentally and, therefore, the thickness of the martensitic rim
usually refers to the location at which the microstructure contains 50% martensite.
Similarly the depth of hardening is the depth at which the hardness corresponding to 50
% martensite is located on a cross section; this is the place where the rate of decrease in
hardness with position is the greatest. Hence, if two steels are quenched in the same
medium and so examined, the steel with the larger depth of hardening is the steel with the
greater hardenability.
Note: The term “depth of hardening” under discussion here should not be confused with
the term “case depth” in carburized parts such as axles and gear teeth. The case depth is
usually specified in terms of a specific hardness value such as that depth at which the
hardness is RC 50.
Conceptual Measures of Hardenability
With the foregoing discussion in mind, we can introduce briefly the standard ways that
are usually considered for specifying hardenability. In the next heading, we consider
more quantitative measures and the methods for obtaining them in practice. The three
main conceptual ways of specifying hardenability are as follows:
• Depth of hardening – as described above, a series of steel cylinders of identical
dimensions can be quenched identically and the depth at which a microstructure of
50% martensite is obtained. (Alternatively, if the hardness corresponding to a
microstructure of 50% is known, then hardness measurements can be used to locate
the proper depth. However, this “critical hardness” will depend on the steel
chemistry, grain size, and other factors.) The deeper the location of this
microstructure, the more hardenable the steel. This concept has several problems,
the main one being that it may be difficult to reproduce exactly the same quench at
R. H. Wagoner MSE 661 – Steel Section 3, Page 13
various sites. Also, it may not be possible to obtain the desired microstructure with
a given bar diameter, in which case there is no method to compare results for
differing bar diameters.
• Critical diameter - When bars of the same steel but with different diameters are
quenched in the same quenchant, Figure 3-9, it is found that there is one particular
diameter for which a microstructure of 50 % martensite is found at the center of the
bar. This diameter is called the critical diameter. It is the diameter of a bar that has
a depth of hardness (as defined above) equal to its radius. Thus, if several steels are
compared, the steel with the greatest hardenability will possess the largest critical
diameter.
• Critical cooling rate - The nature of the microstructure of a given steel at any
location within a quenched bar is mainly dependent upon the rate at which that
location cooled. If we make the approximation that any effects due to stresses
generated by thermal contraction are negligible, then we can expect all locations
that had the same rates of cooling to have the same microstructures. On this basis,
we can refer to a critical cooling rate as that cooling rate that will produce the 50 %
martensitic microstructure. The critical cooling rate decreases as the hardenability
of the steel increases.
Figure 3-9: Hardness profiles for a series of cylindrical bars of various diameters
for an identical steel and quenching medium.
Quantitative Measures of Hardenability
Whereas these kinds of observations clearly allow different steels to be ranked in order of
their hardenability, they do not provide a quantitative measure of the property. A
2.5
2.0 1.5 1.0 0.5 0 0.5 1.0 1.5 2.0 2.5
6 5 4 3 2 1 0 1 2 3 4 5 6
60
55
50
45
40
35
30
Distance from center (cm)
Distance from center (in)
302
392
513
R
o
c
k
w
e
l
l

C
567
critical hardness
H
V
N
Hardness
R. H. Wagoner MSE 661 – Steel Section 3, Page 14
knowledge of the critical diameters in different quenchants is of practical use but,
because the performance characteristics of nominally similar quenching media can vary
with time and place, they do not provide an absolute value for the hardenability. This
problem was investigated by Grossman and his colleagues 50 years ago with the
objective of describing the hardenability by a parameter that is independent of the quench
bath (“Elements of Hardenability, M.A. Grossman, ASM, 1952). There has been a great
deal of work done since, but the early work established the basic concepts and the
terminology.
The Quenching Process
Let us consider the quenching process more closely. When a hot bar of steel is immersed
within the quenching fluid, heat is extracted from the surface of the bar and the
temperature of the surface is decreased. This creates a temperature gradient in the bar.
The rate of flow of heat from the interior of the bar, down the temperature gradient,
depends upon the conductivity of the steel. The combined effects of the rate at which heat
is extracted and the rate at which it is conducted to the surface from within, determines
the rates of cooling at various locations within the bar.
The cooling rates at the surface of the sample being quenched vary with time during the
quench and thus with the actual temperature of the surface. For example, in experiments (
see “The Hardenability of Steels, by C.A. Siebert, D.V. Doane and D.H. Breen, ASM
1977) in which half inch diameter bars of nickel-5%Si, containing embedded
thermocouples, were heated to 850°C and immersed in water, the surfaces of the bars
almost immediately became separated from the water by a blanket of steam. While this
separation existed, the temperature at the centers dropped to about 650°C at a relatively
slow cooling rate. Experimentally determined values ranged up to 50 degrees per second.
In the next phase, liquid water made contact with the sample and steam bubbles formed
rapidly. These bubbles rose from the surface due to gravity and convection turbulence.
During this stage the cooling rate rose to over 300 degrees per second as the temperature
dropped to about 450°C at the center of the bar. At this stage the nucleation and growth
of steam bubbles ended and the rate of cooling became dominated by convection in the
coolant. This caused a rapid decrease in the cooling rate which decreased to less than 50
degrees per second during the time taken for the temperature to drop another 100 degrees.
Evidently, the heat transfer process during quenching in water is not a simple one. It is
interesting to note that it is fortuitous that the greatest cooling rates for water happen to
occur at temperatures corresponding to the nose of TTT diagrams for low to medium
carbon steels. It is fortuitous because these temperature ranges are those through which
the steel must usually be cooled fastest in order to avoid diffusion controlled phase
transformations. Should it have happened in practice that the cooling rate was small in
these temperature ranges, quench hardening of steels and the martensitic transformation
in steels might never have been discovered!
R. H. Wagoner MSE 661 – Steel Section 3, Page 15
The theoretical maximum heat removal rate from a quenched bar occurs if the quenchant
immediately lowers the temperature at the surface of the bar to that of the bath and holds
it there throughout the cooling process. The quenching power of a quenching medium is
described by a coefficient H which describes the ability to transfer heat across the
specimen/bath interface. Its value for the ideal maximum case above ∞. Of course, such
a quenching process is unobtainable practically and is defined as an ideal quench.
Although this quench is unobtainable it is closely approached by quenching into
vigorously agitated, iced brine. Table 3-1 presents some typical H values for various
quenching media.
Table 3-1: Typical values of quenching severity parameter H. (After Honeycombe, p.
156)
Quenching Medium
Agitation Oil Water Brine
None 0.25-0.30 1.0 2.0
Moderate 0.35-0.40 1.2-1.3
Violent 0.8-1.1 4.0 5.0
The Grossman Test
The principles of the Grossman Test have already been introduced in the foregoing
discussion. A series of cylindrical bars of identical material are quenched under identical
conditions to obtain results as illustrated in Figure 3-9.
The bar with a critical diameter, Do, is defined as one that under these conditions obtains
a microstructure of 50% martensite at its center. In order to find such a bar, each of the
test bar must be sectioned, polished, etched, and examined metallographically to reveal
the one that is closest to exhibiting the desired microstructure at its center. This is the
procedure for the standard Grossman test.
Alternatively, if for a given steel the hardness of a 50% martensite microstructure is
known, let us call this the “critical hardness”, then hardness profiles across sections of the
various bars will be sufficient to find the critical diameter. In Figure 3-9, we have
assumed that the critical hardness for a given steel chemistry and grain size is known, and
the hardness profiles are sufficient to identify a critical diameter of a bit less than 2.0
inches. (The next discrete bar diameter is a bit too small, so perhaps an interpolation
between the two diameters would improve the precision a bit.)
The problem with Do as a measure of hardenability is that it clearly will depend on the
effectiveness of the quenching medium, in the same way that the depth of hardness will
depend on how fast heat is removed from the surface of a bar. Thus, while Do’s
measured under identical conditions will be useful rankings of hardenability, it will be
impossible or difficult to compare values obtained at different sites or for differing
quenching media.
R. H. Wagoner MSE 661 – Steel Section 3, Page 16
In order to remove the interference of the quenching effectiveness from the hardenability
measure (which we want to be a property of the material along), Grossman introduced a
term called the ideal diameter. The ideal diameter, Di, is defined as the critical diameter
(i.e. the bar diameter that has 50% martensite at its center ) measured followeing an ideal
quench.
The ideal diameter as a measure of hardenability removes the medium from the definition
entirely, because the surface cooling rate is perfect. The cooling rates throughout a bar
during an ideal quench depend only on the heat capacity and heat conductivity of the
material. These are also material properties, so the hardenability defined in this way is
indeed a material property. Fortunately, most austenitic steels have very similar heat
capacities and heat conductivities, so these variations can often be ignored.
The question remains of how to obtain the ideal diameter, Di. In practice, the
effectiveness of a quench can be measured in a variety rates and an effective H value
(severity of quench parameter) can be obtained. One method might be to compare
cooling rates measured inside a bar (from thermocouples, for example) and compare
them for various quenchants and for an ideal quench (which can readily be solved for
analytically). Once H is known, standard conversions can be used to transform any H
and Do to and equivalent ideal diameter. Such a conversion figure is shown in Figure 3-
10.
Figure 3-10: Conversion chart relating the ideal diameter Di to the severity of quench, H,
and critical diameter Do. (Honeycombe, p. 157)
It should be noted that under a few reasonable assumptions, Di is equivalent to a critical
cooling rate. The relationship is implicit in the multiple abscissa of Figure 2-12,f or
0
1 0
2 0
3 0
4 0
5 0
0 1 0 2 0 3 0 4 0 5 0 6 0 7 0
D
0

(
m
m
)
D
i
(mm)
1 0 5 2
1
0. 8
0. 4
0. 2
0. 1
0. 01
H =
R. H. Wagoner MSE 661 – Steel Section 3, Page 17
example. The main errors in such a direct correlation lies in the fact that the heat transfer
out of bar is not truly a constant throughout the quench. As discussed earlier, the
conditions in the bath adjacent to the hot bar depend on the bar temperature. Steam can
be generated, shrouding the bar from the liquid water at higher bar temperatures, for
example. Also, the real quenching experiments may involve impediments or accelerators
of martensitic transformation such as thermally-induced stresses (that may vary within
the bar and throughout the quench). Finally, and probably much less important, the heat
capacity and conductivity in the bar may depend on temperature and phase changes, such
that the cooling rate may not be known perfectly.
The Jominy End-Quench Test
The Grossman test is time-consuming to perform. Not only must many bars be machined
from a given steel, metallography must be performed on each of them, and the quenchant
must be characterized carefully. A much simpler alternative, the Jominy end-quench test,
Figure 3-11, is considerably more convenient for routine measurements.
Figure 3-11: Schematic of the Jominy end quench test.
The Jominy test maintains all factors except material constant. It involves quenching the
end of a bar of particular dimensions in a specified manner. Typically, the test bar is one
inch in diameter and four inches long. It is austenitized and placed in a jig that suspends
it vertically above a half-inch diameter water pipe which points upwards. The water
pressure in the pipe is set (prior to the test) such that when a fast-opening valve is
activated, the jet of water emitted from the pipe rises to a height of 2.5 inches above the
end of the pipe. (The flow rate and other parameters are carefully specified.) When the
Water spray (24
o
C)
Jominy specimen
Mounting fixture
1/2”
4”
1”
Flat ground
along bar
Rockwell C
hardness
tests
1/2”
R. H. Wagoner MSE 661 – Steel Section 3, Page 18
test bar is put in place, its lower end is 0.5 inches above the end of the pipe. The valve is
opened immediately and the water sprays onto the end of the bar. When everything is
correct, the water only contacts the end of the bar and then sprays outward and downward
resembling an umbrella. Meanwhile, the end of the bar becomes dark and the boundary
between the dark and red-hot steel can be observed to move upwards. It is kind of
entertaining to watch.
When the end-quench is complete, a flat surface, 0.15 inches deep is ground along the
length of the bar. This has to be done with care lest any microstructural changes are
induced by the heat generated in the grinding process. Then Rockwell C hardness
measurements are made at intervals of one sixteenth of an inch along the bar. The results
are plotted as hardness versus distance from the quenched end. See, for example, Figures
3-12 and 3-13.
Figure 3-12: Hardenability curves for five steel alloys containing 0.4% carbon. (after
McCallister)
The assumption inherent in the Jominy geometry is that heat loss from the water-
quenched end is much more rapid that from the cylindrical surface in contact with air.
This is certainly true, particularly in view of the large diameter of the cylinder and the
relative heat conductivity of steel as compared with air. Thus, the surface temperature of
270 170 70 31 18 9 5.6 3.9 2.8 2
o
C/s
490 305 125 56 33 16.3 10 7 5.1 3.5
o
F/s
in.
0 10 20 30 40 50 mm
Distance from quenched end
Cooling rate at 700
o
C (1300
o
F)
in.
60
50
40
30
20
0 0.25 0.5 0.75
1 1.25 1.5 1.75 2
H
a
r
d
n
e
s
s
,

H
R
C
4340
4140
8640
5140
1040
100
80
50
P
e
r
c
e
n
t

M
a
r
t
e
n
s
i
t
e
R. H. Wagoner MSE 661 – Steel Section 3, Page 19
the cylinder in nearly the same as inside the bar the same distance from the quenched
end, and the hardness near that surface can be used to represent the entire section. In
effect, the geometry is idealized as a 1-D heat flow problem.
Figure 3-13: Hardenability curves for similar alloys with differing carbon contents.
(after McAllister)
Numerous investigations have been carried out to correlate the results of a Jominy test
with other measures of the hardenability, such as the ideal diameter from the Grossman
test or the critical cooling rate (both to obtain 50% martensite). Of the three, the cooling
rate seems to have the clearest definition, with the “critical” part being the only fuzzy
aspect.
One can imagine generating a clear picture of cooling rates within both the Jominy end
quench bar as a function of position from end (and time!). This could be done either
analytically or experimentally. A similar exercise could be applied to Grossman bars of
various sizes, again obtaining actual cooling rates as a function of radius (and time!).
Once completed, and verified by metallography on real specimens, conversion figures
can be generated to allow mapping among the three measures of hardenability.
270 170 70 31 18 9 5.6 3.9 2.8 2
o
C/s
490 305 125 56 33 16.3 10 7 5.1 3.5
o
F/s
in.
0 10 20 30 40 50 mm
Distance from quenched end
8660
(0.6 wt% C)
8660
(0.4 wt% C)
8660
(0.3 wt% C)
8660
(0.2 wt% C)
Cooling rate at 700
o
C (1300
o
F)
in.
60
50
40
30
20
0 0.25 0.5 0.75
1 1.25 1.5 1.75 2
H
a
r
d
n
e
s
s
,

H
R
C
R. H. Wagoner MSE 661 – Steel Section 3, Page 20
In Figure 3-14, the diameters of bars that have the same cooling rates at their centers (and
several other positions) as those at particular distances along a Jominy bar are plotted
against the Jominy distance. Curves for two different quench severity’s are presented.
The graph enables the determination of the distance along a Jominy bar of the location at
which the cooling rate would be the same as at the center of the bar. The hardness at the
center of the bar can now be obtained from the results of the Jominy test.
The Jominy distances involved are called Jominy equivalent cooling rates or Jominy
equivalent conditions, J
ec
. Lets consider anr example of their use. Suppose a particular
part is to be manufactured and heat treated in a currently available process that uses a
particular quenching medium and the hardness value for particular locations within the
part are specified.
An appropriate steel is to be chosen. One procedure is to manufacture a number of the
parts using a relatively shallow hardening steel such as 8620 or 1040 and also make
Jominy test bars of the same heat of steel. These parts along with the Jominy bars are heat
treated in the production facilities. Following the heat treatment, selected parts are
sectioned and the hardness at the critical locations is measured. Meanwhile the Jominy
bars are tested to determine the end-quench data. Then by comparing the hardness
values measured on the parts with the end quench data, the Jominy distances
corresponding to these hardnesses can be found. These are the J
ec
’s for the particular
locations; i.e. these locations cooled similarly to the corresponding positions on the
Jominy bars. It is important that the test bars and parts are exposed to the same thermal
history.
Finally, end quench data available for various steels produced for selection on the basis
of hardenability can be examined to choose one which will produce the specified
hardnesses at the measured J
ec
’s. This procedure requires the test steel and those of the
members of the group from which the selection is to be made to have similar thermal
conductivity so that the cooling rates are similar at similar Jominy distances. When steels
of suitable hardenability are identified the final choice is made taking into account other
factors. For example, it is usual to choose a steel with the lowest satisfactory carbon
content and of the minimum hardenability necessary for the job. This reduces the risk of
problems that might arise with respect to quench cracking and distortion if the quenching
process is vigorous.
R. H. Wagoner MSE 661 – Steel Section 3, Page 21
Figure 3-14: Cooling rates for various Jominy distances and Grossman bar diameters.
(after McCallister.)
170 70 31 18 9 5.6 3.9
o
C/s
305 125 56 33 16.3 10 7
o
F/s
0 10 20 30 mm
Distance from quenched end
Cooling rate at 700
o
C (1300
o
F)
D
i
a
m
e
t
e
r

o
f

b
a
r

(
m
m
)
Center
1/2 R
3/4 R
Surface
in.
(b)
4
3
2
1
0
0
0.25
0.5 0.75 1 1.25 in.
100
75
50
25
0
D
i
a
m
e
t
e
r

o
f

b
a
r

(
i
n
.
)
170 70 31 18 9 5.6 3.9
o
C/s
305 125 56 33 16.3 10 7
o
F/s
0 10 20 30 mm
Distance from quenched end
Cooling rate at 700
o
C (1300
o
F)
D
i
a
m
e
t
e
r

o
f

b
a
r

(
m
m
)
3/4 R Center
1/2 R
Surface
Center 1/2 R
3/4 R
Surface
(a)
4
3
2
1
0
0 0.25 0.5 0.75 1 1.25 in.
100
75
50
25
0
D
i
a
m
e
t
e
r

o
f

b
a
r

(
i
n
.
)
Mildly agitated
water quench
Mildly agitated
oil quench
R. H. Wagoner MSE 661 – Steel Section 3, Page 22
Tempering Reactions in Carbon Steel
The heat treating process that is applied to martensite in order to endow it with better
toughness is called tempering. A tempering treatment consists of heating the steel to a
selected temperature below the A
1
line, holding it there for a certain time and then
cooling it back down to ambient temperature. The transformations that occur and the
property changes that are produced, depend upon the tempering time and temperature.
The general aim is to raise the toughness of the steel. Because this is achieved at the cost
of a reduction in hardness and strength, the process is always a compromise and is
selected according to the intended application. A great deal of research has enabled us to
understand much of what goes on, but there are still some uncertainties with respect to
transition carbides.
Stages of Tempering
On the basis of the research into the tempering process, it has become customary to
divide the process into stages that occur in different temperature regimes. It should be
recognized that this division is a convenient way of describing tempering reactions; the
particular temperature ranges should be considered as guidelines and, obviously,
neighboring stages overlap. For example, see Figure 3-15 for typical tempering curves
for plain-carbon steels. There is nothing to indicate a crisp delination of these
microstructural “stages.”
Stage 1 - Stage 1 occurs at temperatures below 200 °C. Here, carbon atoms diffuse
throughout the martensite and segregate at dislocations and intercrystalline boundaries.
Because the M
s
temperature for carbon steels is higher than room temperature, it is
possible for these segregation processes to begin when the martensite is first formed,
especially if the cooling rate below the M
s
point is slow. In this case, one refers to the
changes as being produced by auto tempering . or steels of less than about 0.2 wt.% C no
carbides form in this stage.
When the carbon content is greater than 0.2 wt. %, very fine particles of a transition
carbide nucleate and grow within the martensite. This carbide is richer in carbon than
cementite and is described by the formula Fe
2.6
C. It is metastable. It has a hexagonal
structure and is referred to as epsilon carbide. Several researchers have claimed to have
observed the presence of another metastable carbide in high carbon steel. This one has
been christened eta carbide. It is orthorhombic Fe
2
C. There is, however, still uncertainty
because the experimental confirmation of these claims is not easy. It is generally
accepted that epsilon carbide definitely forms. When it does, the martensite retains some
degree of tetragonality because it still contains more carbon in solid solution than would
ferrite; there are no changes in the morphology of the martensite crystals. Thus the first
stage of tempering involves the segregation of carbon to various defects in the
microstructure and the conversion of the martensite to low carbon martensite plus epsilon
carbide when the total carbon content is high enough.
R. H. Wagoner MSE 661 – Steel Section 3, Page 23
Figure 3-15: Tempering curves for plain-carbon steels. (Krauss, p. 208)
During stage 1 tempering, there are changes in the physical properties such as electrical
resistivity which can be used to monitor the progress of the changes. However, there is
not much of a reduction in the hardness; indeed, it can increase slightly for steels of
medium to high carbon levels. Obviously, these changes involve diffusion and are
therefore time dependent. A common practice is to use tempering times of one hour so
much of what follows will refer to the changes produced by tempering treatments one
hour long .
Stage 2 - The second stage of tempering occurs between 200 and 300°C. It is not very
important in plain carbon steels because it refers directly to the conversion of retained
austenite to bainite. The quantities of retained austenite in carbon steels are not large.
However, they can be large in alloy steels especially those for which the M
f
temperature
is below 0°C; stage 2 is important when there are significant quantities of retained
austenite included within the martensite.
1 0
2 0
3 0
4 0
5 0
6 0
7 0
200 400 600 800 1000 1200 1400
100 200 300 400 500 600 700
H
a
r
d
n
e
s
s
,

R
o
c
k
w
e
l
l

C
Tempering temperature, deg. F
Tempering temperature, C
As-
quenched
0. 1- 0. 2%
0. 2- 0. 3%
0. 3- 0. 5%
0. 4- 0. 5%
0. 5- 0. 7%
0. 7- 0. 11%
Carbon Content
R. H. Wagoner MSE 661 – Steel Section 3, Page 24
Stage 3 - Cementite begins to precipitate out from solid solution at about 300°C. This is
the beginning of stage three. The precipitates nucleate on heterogeneous sites such as
intercrystalline boundaries. Because the martensitic crystals form with specific
orientation relationships with the austenite crystal in which they grow, they do not grow
across austenite grain boundaries. In effect, the prior austenite grain boundaries are
retained as surface defects.
At the beginning of stage three, the heterogeneous nucleation process at the sites of the
prior austenite grain boundaries tends to produce sheets of cementite particles. The
matrix loses tetragonality and becomes BCC ferrite as its carbon content drops because of
the diffusion of carbon to these growing carbides .
If the steel is sufficiently rapidly heated to the tempering temperature then no transition
carbides form. If transition carbides were present due to a previous tempering treatment
or because they formed during a slow heat-up, then they dissolve during stage three. The
phases present become ferrite and cementite. When tempering is carried out at
temperatures near 300 °C, the ferrite crystals remain acicular during the times usually
used. Because the diffusion coefficient of carbon increases with an increase in the
temperature, the cementite particles are able to grow more rapidly and to decrease in
number per unit volume, as higher tempering temperatures are used.
The hardness after tempering becomes increasingly smaller; an effect which becomes
more and more significant as the tempering temperature becomes higher. Eventually the
temperature becomes high enough to cause the dislocation content to decrease in a
similar manner to the changes in the dislocation content of cold-worked metals during
recovery processes. At about 600°C, the ferrite grains are able to grow and become
equiaxed. These morphological changes in the ferrite grains are similar to those that
occur during recrystallization in cold worked ferrite. Because of this, these later
processes have been said to comprise a fourth stage of tempering, but that name has also
been assigned to secondary hardening processes that can occur during the tempering of
alloy steels so to avoid confusion we won't use it for carbon steels. Instead, we will
consider all of processes that occur above three hundred degrees Celsius as belonging to
stage three.
These various stages offer a convenient way of regarding the tempering reactions.
However, it is important to recognize that the temperatures that have been quoted to
delineate the stages are merely guidelines. Obviously the stages overlap because the
dominant type of change and the rate thereof depends upon temperature and time and
does not change discontinuously. It is also worthwhile pointing out a common error
made by students new to this subject when they describe tempering reactions in small
samples that have been rapidly up-quenched to high tempering temperatures such as 600
°C. The error is to describe the transformations that take place as those of stages 1,2 and
3 occurring sequentially. Unless the heat-up time is slow enough, epsilon carbide will
not form at this temperature and pearlite not bainite can form from retained austenite
because the tempering temperature is above the position of the nose of the TTT diagram.
R. H. Wagoner MSE 661 – Steel Section 3, Page 25
Standard Heat Treatments
The microstructure of a steel depends upon its prior thermo-mechanical history. For
example, a hot rolled plain carbon steel can consist of equiaxed ferrite grains and pearlite
nodules. If the steel does not have the microstructure desired for its application or for the
next process that is to be performed on it, then the microstructure is changed by heat
treating. A number of standard processes exist. These are briefly described below.
Normalization - This heat treatment is performed to refine grain size and to obtain a
carbide size and distribution that provides a favorable starting point for subsequent heat
treatments; it is usually done to hypoeutectoid steels. It comprises austenitization at 100
to 150 °F above the A
c3
line followed by an air cool. A standard rule of thumb is to
austenitize the steel for 1 hour per inch of thickness. This rule of thumb has been used for
many years. However, because computing has become easier, empirical and semi-
empirical studies are being made to determine soak times more accurately. It is
important to recognize that the microstructural details produced by normalization will
depend on the size of the part; a thin part cools more rapidly in still air than does a thick
part.
Full Anneal - This process is used to make the steel relatively soft and to endow it with
good machinability. The microstructure is large-grained ferrite and coarse pearlite.
Austenitization is conducted at a high temperature to produce a large grained austenite
and to dissolve all the carbides. The steel is then cooled slowly enough to allow the
desired microstructure to form by the transformation of the austenite at temperatures near
to the A
1
line.
Isothermal Annealing - This has essentially the same aim as the full anneal. It is more
efficient for small parts because furnace cooling can be replaced by transferring the parts
between constant temperature baths containing molten salts. The steel is then usually
protected from scaling and oxidation by the molten material into which it is placed. These
salts are various mixtures of compounds such as alkali metal hydroxides and nitrates that
are chosen for particular temperature ranges. Baths of molten lead were used effectively
for this purpose before concerns about the dangers of lead to human health became
influential.
Spheroidization - In the spheroidized microstructure the carbides are discrete globules
embedded within the ferritic matrix and also resident on grain boundaries. Steels with this
microstructure are in their softest state and are most easily mechanically formed. It is
often part of overall treatments given to high carbon tool steels that are formed into
particular shapes before being hardened. It can be produced by cooling very slowly from
100°F above the eutectoid isotherm or by a long isothermal treatment just below it.
Spheroidization can also be obtained by cyclic processing in which the temperature of the
steel is cycled above and below the A
1
line.
R. H. Wagoner MSE 661 – Steel Section 3, Page 26
Process Annealing and Stress Relieving - This terms normally mean some form of
treatment in which the steel is heated to and held at a temperature below the A
1
line.
Recovery processes occur which removes or diminishes residual stresses that might be
present from previous treatments. Recrystallization can also be produced in cold worked
parts.
Quench and Temper - The purpose of these heat treatments is to produce parts in which
a desired hardness is obtained, either throughout the part or in the vicinity of its surface.
It involves austenitization, quenching to room temperature or below and reheating to and
holding at a tempering temperature below the A
1
line. The first part of the process is
intended to form martensite. The tempering temperature and time are selected to produce
the required hardness.
Excellent properties can be obtained by this procedure. However, there are serious
limitations for plain carbon steels. This is because the transformations that occur by
diffusional processes are not easily avoided during cooling. Consequentially, very severe
quenching rates are necessary and even then only small samples can be converted to
martensite throughout. The rapid quench rates are undesirable because there is no
opportunity for stress relief during cooling; warping and distortion are more likely to
occur during rapid quenching. These stresses superimposed on the transformation stresses
due to the volume expansion that occurs when martensite forms, can help to cause
quench-cracking. The situation is that carbon steels of the correct microstructure could
perform adequately in many applications, but it is often not possible to produce the
microstructure because the part cannot be cooled fast enough. This is the first reason for
the development of low alloy steels in which one of the roles of the alloying elements is
to slow down diffusional transformations to enable martensite to be produced during
slower cooling rates.
Martempering - A modification of the quench process that relieves the thermally
induced stresses initially generated by thermal contraction is called martempering,
Figure 3-16. In this process, the part is first quenched rapidly until its temperature has
fallen below that of the nose of the TTT diagram but is still above the M
s
point. It is held
at that temperature to relieve stresses by becoming of a uniform temperature and then
finally cooled through the M
s
point. This process is possible when the bainitic
transformation takes much longer to begin than the changes associated with the nose of
the diagram. Note, also, that the steel can be relatively slowly cooled from the isothermal
hold and still form martensite which serves to minimize distortion further.
Austempering - As we discussed earlier, bainite can have useful hardness and toughness
values . When they are adequate for the application, the formation of bainite can be
chosen instead of martensite. To allow this, the steel is quenched to a temperature below
the nose and held there until the bainitic transformation is completed, Figure 3-16. This
avoids the problems associated with the rapid volume changes during the martensitic
transformation. The process is termed austempering.
R. H. Wagoner MSE 661 – Steel Section 3, Page 27
Figure 3-16: TTT diagram with Martemper and Austemper heat treatments shown
schematically.
SECTION 3 - CONCEPTS AND STUDY QUESTIONS
Important concepts
Martensite as a phase and a microstructure.
The role of carbon in martensite: hardness, lattice parameter, Ms, hardenability.
Bain strain, transformation strains, post-transformation strains. Invariant planes.
Lath martensite vs. plate martensite.
Origin of surface relief in martensitic transformation. Role of habit plane.
800
700
600
500
400
300
200
100
0
T
e
m
p
e
r
a
t
u
r
e
1 MIN
0.5 1 2 5 10 10
2
10
3
1 HOUR
10
4
10
5
10
6
1 DAY 1 WEEK
Time - Seconds
A
A
F + C
11
32
38
40
40
41
43
50
55
57
H
a
r
d
n
e
s
s
-
R
C
οοο ο
C
I - T Diagram
M
s
M
50
M
90
A
+
F
+
C
Austemper
Martemper
R. H. Wagoner MSE 661 – Steel Section 3, Page 28
Basic definition of hardenability. Contrast with hardness.
Methods for characterizing hardenability and measuring it.
Definition of depth of hardening, critical diameter, ideal diameter, Jominy distance.
Characterization of quench severity.
Hardening curves. Effect of C and alloying.
Conversion of one hardenability measure to another.
Stages of tempering. Definitions. Microstructural changes.
Standard heat treatments. Reasons for each.
Study Questions
1. It is required that a microstructure of 50% tempered martensite and 50% equiaxed
ferrite grains be produced in a plain carbon steel. What steel would you choose and how
would you produce the microstructure? Make sure that you include your reasoning.
2. Discuss the microstructural changes and the variation in the hardness of martensitic
AISI 1060 as a function of time during tempering at 600 deg. C.
3. Describe and explain the effects of carbon content on the Ms temperature in plain-
carbon steels.
4. What changes occur in the crystallographic unit cell of martensite in iron as the carbon
content of the iron is increased from 0%. Why?
5. What is the Bain strain in ferrous metallurgy? Why is it not sufficient by itself to
describe the shape change during a martensitic transformation?
6. Suppose you have a plate of 1080 steel that has the dimensions: 6 inches by 6 inches
by 3 inches. Suppose also that you have a refractory brick made mostly of alumina that is
6 inches by 6 inches by 6 inches. You place the steel on top of the brick so that the 6x6
face of the plate is in contact with the brick. This combination is placed in a furnace at
900 deg C and left there until the plate becomes homogeneous austenite. The
combination is then removed from the furnace, placed on a concrete floor with the steel
uppermost. Immediately, a jet of ice-cold water is sprayed on the top surface of the steel
and maintained until everything is cold. Discuss, with explanations, the microstructural
characteristics you would expect to find in the plate.
7. What is the effect of austenite grain size on hardenability? Why?
R. H. Wagoner MSE 661 – Steel Section 3, Page 29
8. Why is martensite tetragonal instead of cubic? How does this aspect affect its
mechanical properties?
9. How is the quantitative measure of hardenability, Di, defined? Explain the
experiment that might be used to determine it and any analytical procedure needed to
correct or calibrate.
10. Describe Stage I, Stage II, and Stage III tempering in plain-carbon steels.
11. How does the austenitizing temperature affect further heat treatment?
Physical Metallurgy of Steel
Class Notes and lecture material for MSE 661 (formerly 651.01)
Notes revised by: Robert H. Wagoner, Smith Chair
Notes originally compiled by: Glyn Meyrick, Professor Emeritus
Web installation by: Wei Gan, Graduate Research Associate
Last revision date: 10/6/02
SECTION 4 – MECHANICAL PROPERTIES AND
MICROSTRUCTURE
Background
So far, the heat treatment and microstructures of plain-carbon steels have been described
and presented, without significant development of the resulting mechanical properties.
Before proceeding to the complications of alloying elements in steels (beyond the ones
found in plain-carbon steels), it will be worthwhile to begin addressing mechanical
properties of steels, particularly their relationship to microstructure (and therefore to
thermo-mechanical treatments that produce these microstructures).
Before considering steels at all, it is important to understand the fundamental principles
that govern the mechanical properties of all metallic systems. Many of these
relationships cannot be quantified precisely because of the many complicated interactions
in real materials, but the general trends can be understood to good advantage. Without
such fundamental knowledge, the inherent tradeoffs in alloy design and heat treatment
cannot be appreciated or understood.
In the first part of this chapter, dealing with fundamentals, we will attempt to answer or
address the following questions:
• What are strength, ductility, and toughness?
• What is the trade-off among these measures and why?
• How are these macroscopic measures related to microstructure and mechanisms?
In order to answer these questions, we shall consider tensile tests and how each of these
quantities are derived (at least approximately) from them.
In the next section, we shall consider definitions of strength, ductility, and toughness
based on tensile tests. Then, we will see if we can understand the tradeoffs among each
of these desirable properties.
Macroscopic Definitions of Strength, Ductility, Toughness
Consider Figure 4-1, a standard tensile test plotted first in terms of engineering stress-
strain (which measures are proportional to load and extension, respectively) and then in
terms of true stress and true strain (which are measures that relate to the material itself
rather than to external load carrying capacity of a component). Note: In order to
illustrate the linear elastic region of the test, the slope of this has been divided by a large
number, perhaps 10 to 100. If the graph were reproduced realistically, the linear region
would appear closely parallel to the vertical axis and little or no elastic strain could be
distinguished.
Figure 4-1: Tensile test results represented in engineering and true stress and strain, with
common measures taken from tensile tests.
The basic quantities are defined as follows:
σ
e
= P/A
o
= Engineering stress
e = ∆l/l
o
= Engineering strain (sometimes called elongation)
where
A
o
= Initial cross-sectional area
l
o
= G = Initial gage length
∆l = l-l
o
= Change in gage length
Engineering stress has units of force per area and engineering strain is dimensionless
(mm/mm, for example). Engineering strain is often presented as a percentage by
multiplying by 100.
X
Strain
S
t
r
e
s
s
e
f
, total elongation (%)
e
u
, uniform elongation (%)
UTS, ultimate
tensile strength
σσσ σ
y
, yield
strength
Engineering Stress-Strain (σσσ σ
e
-e)
True Stress-Strain (σσσ σ−−− −εεε ε)
It is clear that engineering stress and strain do not adequately describe what is going on
inside the material, because the measures depend on an original area or gage length, even
though both of these change throughout the test. (This means that a change of
engineering strain 10% at the start of a test involves more material stretching than a
change of 10% near the end of the test, where the real gage length is longer.) To take
care of this problem, we introduce real or true strain, an increment of which refers to an
infinitesimal extension per unit of current length. Note: It is now necessary to limit
consideration to a small change of extension to insure that the current length is constant
and well-known. By assuming that the incremental strain over the current gage length is
uniform

, we can write mathematically, then, that the true strain increment is: dε = dl/l
(not dl/lo). We can express the total true strain as a simple integral:
ε ε
ε
ε
= = =





 ∫ ∫
d
dl
l
l
l
o o
l
l
o
ln (4-1)
Similarly, the real or true stress refers to the load divided by the current cross sectional
area:
σ =
P
A
(4-2)
By some manipulation and by noting that the volume doesn’t change during plastic
deformation of metals (see below), it is possible to obtain the following relationships
among true and engineering stress and strain.
ε ε
σ σ ε σ σ ε
σ σ σ σ
= + = −
= = −
= + = +
ln( ), exp( )
exp( ), exp( )
( ), /( )
1 1
1 1
e e
e e
e e
e e
(4-3)
Note: Volume conservation. In order to understand consequences of the conservation of
volume during plastic deformation, consider a simple right parallelepiped corresponding
to a simple tensile specimen, Figure 4-2.
l1
l2
l3
∆V = 0
l'
3
l'
2
l'
1
Figure 4-2: Schematic deformation of a tensile specimen.
The engineering and true strain in each direction are as follows:


As will be shown later in this section, the assumption that the tensile specimen deforms uniformly holds
only up to the uniform elongation, approximately where necking begins. Once necking begins, there is no
simple way to obtain true stress and strain data from the engineering stress and strain data, although other
experiments and experience show that the true stress-strain curve continues to rise, up to very large strains,
as shown in Figure 4-1.
e l l
l
l
e l l
l
l
e l l
l
l
1 1 1 1
1
1
2 2 2 2
2
2
3 3 3 3
3
3
= − =
= − =
= − =
'
'
'
'
'
'
, ln
, ln
, ln
ε
ε
ε
(4-4)
With some simple algebra, it may be easily shown that when l
1
l
2
l
3
=l
1
’l
2
’l
3
’ (i.e. when the
volume is unchanged) that
0 = e
1
+ e
2
+ e
3
= ε
1
ε
2
ε
3
(4-5)
Strength: The strength of an alloy is usually defined by either the yield strength (σ
y
)
(Figure 4-1) or the ultimate tensile strength (σ
uts
), depending on the application for the
material. Either measure refers to the load that must be applied to a given size structural
member to deform it plastically (i.e. permanently). For building construction
applications, where no plastic deformation can be tolerated, usually the yield strength is
most important. For automotive crash applications, where the strength of the material is
to be known, usually the ultimate tensile strength is more pertinent. Note that the
difference between these measures is a measure of the strain hardening of the material.
The larger the difference between σ
uts
, and σ
y
, the larger the strain hardening ability of a
given material.
Ductility: Ductility refers to the ability of a material to withstand plastic deformation
without failure, i.e. while still resisting at least some load. It is most important in two
contexts: 1) the ability to form a material to make a useful part, and 2) the ability to
absorb energy, such as in the automotive crash application mentioned above. Ductility
has two standard measures from a tensile test, just as strength does. The first of these,
total elongation (e
f
or e
t
), refers to the engineering strain that a material can withstand
while still maintaining a load. Thus, a 1” long part with a total elongation of 30% can be
stretched to 1.3” without fracture. The second measure, uniform elongation (e
f
), refers to
the elongation up to the ultimate tensile strength (that is, the maximum load point in the
tensile test). This measure is useful first because it is often a close approximation to the
total elongation, but in general can be measured more accurately. Second, for forming
applications, the deformation become nonuniform (by necking) after the uniform
elongation, and this cannot usually be tolerated. Therefore, the uniform elongation is a
measure of the useful ductility of a material for forming.
Toughness: Toughness refers to the ability of a material to absorb energy during plastic
deformation. It is thus related to both strength and ductility. Of course, as in the case of
either stress or strain, there must be some normalization to the size of the specimen.
(Clearly, a large specimen can absorb more energy that a small one!) We can start with
the definition of an increment of work (fdx, where f=force, dx = increment of
displacement). When applied to a tensile test, the force is the force applied to the end and
the dx is an increment of extension (or length change) at the end where the force is
applied. Then, the total work absorbed by the specimen is
W P dl
l
l
=

1
2
(4-6)
In order to normalize this to our standard measured quantities, we replace P and l with
their equivalents in terms of stress and strain (either definition):
W P dl A l de A ld
l
l
e o o
e
e
= = =
∫ ∫ ∫
1
2
1
2
1
2
σ σ ε
ε
ε
(4-7)
By noting that the volume of the tensile specimen is A
o
l
o
(original) or Al (at any time)
and that both of these are identical because the volume doesn’t change by plastic
deformation, the normalized work definition (per unit volume of material) is obtained:
W
vol
de d
e
e
e
= =
∫ ∫
σ σ ε
ε
ε
1
2
1
2
(4-8)
The work per unit volume is a standard measure of “toughness” of the material. For an
idealized, non-strain hardening materials, the integrals could be approximated W/vol =
σ
y
ε
f
, (where ε
u
is some idealized strain at which the true stress drops suddenly from its
constant value to u). This quantity may further be approximated for a real material by
σ
f
ε
u
where σ
f
is the flow stress, which is sometimes defined as the average of the yield
stress and ultimate tensile strength, and the true strain corresponding to ultimate tensile
strength is taken as an approximation of the ductility.
Considere Criterion
As shown above, toughness clearly depends both on strength and ductility. However,
although strength and ductility are apparently independent, it has been shown repeated by
numerous experiments on many metallic systems that they are often inversely related, at
least approximately. This experience is shown schematically in Figure 4-3.
Figure 4-3: Conceptual plot ductility vs. strength for alloys.
D
u
c
t
i
l
i
t
y

(
%

u
n
i
f
o
r
m

e
l
o
n
g
a
t
i
o
n
)
Yield Strength (ksi)
30
20
10
50 100 150 200
The obvious question is, why can’t we have both strong and ductile alloys at the same
time? In order to understand the relationship, we must understand what limits the
ductility of alloys. First, let us restrict our thinking to the tensile test and furthermore to a
material which is quite ductile, that is one that undergoes large plastic deformation
without intervention of brittle fracture by another mechanism. Under these conditions, a
stress-strain plot like Figure 4-1 is appropriate, rather than a sudden loss of load-carrying
capacity because of a crack.
Necking is the localization of strain in a neck which occurs near the end of a tensile test.
It is important to know when this localization begins. The most widely accepted estimate
of the uniform elongation is attributed to Considere
*
. The onset of necking occurs at the
point of maximum load (or engineering stress) according to this view. Note that the
necking phenomenon is a result of the competition between work hardening (σ always
increases with increasing ε) and the reduction of cross-sectional area because of
continuing deformation. That is to say, even though the material continues to get harder,
the decreasing area makes the load tend to decrease and eventually a maximum load is
reached.
The maximum load point (i.e. where dP=0) is readily available on load-elongation curves
or engineering stress-strain curves, but also may be found purely by knowing the true
work hardening law, that is the work-hardening. In terms of true stress and strain, the
load P is equal to σA and dP may be written
dP d A dA Ad = ( ) = + σ σ σ (4-9)
Setting dP equal to zero (to find the maximum load point) and rearranging gives:
dA
A
d
= −
σ
σ
(4-10)
In order to put this into more familiar quantities we note again that the volume of the
specimen remains constant under plastic deformation, so that Al=constant, or
d Al Adl ldA ( ) = = + 0 (4-11)
dA
A
dl
l
d = − = − ε (4-12)
Therefore, we can substitute –dε for dA/A and obtain one of the common forms of the
Considére Criterion:
d
d or
d
d
σ
σ
ε
σ
ε
σ = = , (4-13)

*
A. Considere: Ann. Ponts Chaussees, 1885, Vol. 9, pp. 574-775.
The second form of these is most common. It states that the maximum load point of a
tensile test may be computed solely by knowing the form of the strain hardening law.
The maximum load point (which is also identified experimentally as the onset of
necking) occurs when the slope of the true stress-strain curve just equals the stress at that
point.
Quantitative Effects of Strain Hardening
One of the simplest equations representing strain hardening is attributable to Hollomon

:
σ ε = k
n
(4-14)
where n is called the work hardening rate or exponent and k is the strength coefficient.
Surprisingly enough, this simple equation works reasonably well for a variety of steels
and thus we can consider the implications of this on strength, ductility, and toughness.
Let us start with ductility, i.e. the uniform elongation. With the aid of the Considere
Criterion, we can find out the necking point if we know k and n. We simply note that
d
d
n
ln
ln
σ
ε
= (4-15)
and therefore that the uniform elongation is simply equal to n:
ε
u
n Hollomon material = ( ) (4-16)
The ultimate tensile strength of a Hollomon material is easily obtained once the uniform
strain is known. At this point ε=n, the true and engineering (UTS) stresses are
σ σ = =

kn or kn e
n
UTS
n n
, (4-17)
While the Hollomon equation is not intended to represent the early part of the stress-
strain curve accurately (there is no kink at the elastic-plastic transition, for example) it is
possible to get some kind of estimate of the yield strength by substituting a small strain,
let us say 0.005 (consisting of perhaps 0.003 elastic strain and 0.002 offset strain). Using
this approximation, the yield stress may be estimated as follows:
σ σ
y
n
k ≈ = ( ) ( . ) . 0 005 0 005 (4-18)
Finally, the toughness of a Hollomon material may be estimated by integrating the true
stress-strain curve up to the uniform strain point (thus ignoring the post-uniform region):
W
vol
d k d
k
n
n
n
n
n
n
= = =
+
∫ ∫
+
σ ε ε ε
0 0
1
1
(4-19)


J. H. Hollomon: A.I.M.E. Trans., 1945, Vol. 162, p. 268
While the Hollomon hardening law is convenient for finding the uniform elongation, it
isn’t very realistic at small strain, where the elastic-plastic transition takes place. A
modification captures this transition better and also separates more clearly a change of
yield stress from a change of work hardening.
σ σ ε = +
o
n
k (4-20)
An Exercise to Examine the Role of Yield Strength
The problem: For a material obeying a hardening law as shown in Equation 4-20, and
having k=75ksi and n=0.25, find the ductility, toughness, yield strength and ultimate
tensile strength for σ
o
=0, 20, 40, 60, 80, and 100ksi. (Assume that yield occurs at a strain
of 0.002.)
The yield stress is readily found. Assuming that an offset strain of 0.002 approximates
the yield strain, the yield stress is simply:
σ σ
y o
n
k = + ( . ) 0 002 (4-21)
The remaining quantities of interest require first a calculation of the uniform elongation,
which requires solution of Considere’s Criterion, Equation 4-13. Applying this to the
new hardening law (Equation 4-20) gives:
d
d
nk k
n
o
n
σ
ε
ε σ σ ε σ = = = +
−1
(4-22)
or, more compactly:
0
1
= + −

σ ε ε
o
n n
k nk (4-23)
Unfortunately, Equation 4-22 is transcendental, that is, it cannot be solved by algebraic
means. (This is the price one pays for trying to be realistic!) While trial-and-error and
direct substitution methods can be used, they have little generality. Instead, let’s use a
very powerful and general method, the Newton or Newton-Raphson Method.
Cconsider a general function, f, of a single variable, x, for which we seek the x value that
satisfies
f(x) = 0 (4-24)
For example, the right-hand side of Eq. 4-23 is just such a function. The solution method
is shown graphically in Fig. 4-4. For each trial solution, starting with an arbitrary one of
x
o
, a recursive formula based on Figure 4-4 is generated. For a trial solution of x
o
, the
next trial solution is
x x
f x
f x
o
o
o
1
= −
( )
( )
'
(4-25)
The next iteration is simply:
x x
f x
f x
2 1
1
1
= −
( )
( )
'
(4-26)
and this process is continued until the error on f(x) reaches some tolerance or an
estimated error on x reaches some tolerance:
f x tolerance
n
( ) < (4-27)
x x tolerance
n n
− <
−1
(4-28)
Fig. 4-4: Graphical interpretation of Newton's method
For our particular equation to be solved, Equation 4-23, and using σ
0
=0 and a starting
value x
o
=0.1, the next six trial solutions for x are x=0.1, 0.171, 0.229, 0.249, 0.250,
0.250, 0.250. For x
6
=0.250, the estimated error on x (as defined in Equation 4-28) is less
that 10-10. (Note: for the particular value of σ
0
=0, the new equation becomes simply the
Hollomon law, Equation 4-14) and we know that for this case e
u
=n (Equation 4-16), i.e.
0.25, as we just solved for.
To complete the exercise, once we know e
u
or ε
u
, we need to find the other required
quantities. Following the procedures for the Hollomon law, we obtain the following
relationships:
f’(x
o
)
f’(x
1
)
x
o
x
1 x
2
f
x
f(x
o
)
Toughness, or work/volume:
W
v
k d
k
n
o
n
o u u
n u
= +
( )
= +
+





+
σ ε ε σ ε ε
ε
0
1
1
(4-29)
Ultimate tensile strength:
UTS k
o u
n
= + σ ε (4-30)
With these results in mind, Table 4-1 contains the summary results for the calculation,
and Figures 4-5, 4-6, 4-7 and 4-8 present these results graphically.
Table 4-1: Strength, Ductility, Toughness vs. Sigma-0 or Yield Strength
Sigma-0 Unif. Eng. Uniform true WORK UTS Yield ( . 002)
0. 00 0. 28 0. 25 10. 6 68. 1 15. 9
20. 00 0. 19 0. 18 10. 4 82. 0 35. 9
40. 00 0. 14 0. 13 10. 1 97. 4 55. 9
60. 00 0. 11 0. 10 9. 8 113. 8 75. 9
80. 00 0. 09 0. 08 9. 4 130. 9 95. 9
100 0. 07 0. 07 9. 1 148. 5 115. 9
Figure 4-5: Effect of Sigma-0 on other ductility.
Effect of Sigma-0 on Uniform Elongation
0 . 0 0 0
0 . 0 5 0
0 . 1 0 0
0 . 1 5 0
0 . 2 0 0
0 . 2 5 0
0 . 3 0 0
0 2 0 4 0 6 0 8 0 1 0 0 1 2 0
Sigma-0 (ksi)
U
n
i
f
o
r
m

E
l
o
n
g
a
t
i
o
n
k=75 ksi, n=0.25
σσσ σ
o
as shown
sigma-0 Unif. Eng. Uniform true WORK UTS Yield (.002)
0.00 0.28 0.25 10.6 68.1 15.9
20.00 0.19 0.18 10.4 82.0 35.9
40.00 0.14 0.13 10.1 97.4 55.9
60.00 0.11 0.10 9.8 113.8 75.9
80.00 0.09 0.08 9.4 130.9 95.9
100 0.07 0.07 9.1 148.5 115.9
Figure 4-6: Hardening curves for various values of Sigma-0.
Figure 4-7: Ductility as a function of yield stress, constant strain hardening.
σσσ σ=σσσ σ
o
+ke
n
Hardening (Various Sigma-0)
0 . 0 0
2 0 . 0 0
4 0 . 0 0
6 0 . 0 0
8 0 . 0 0
1 0 0 . 0 0
1 2 0 . 0 0
1 4 0 . 0 0
0 . 0 0 0 0 . 0 5 0 0 . 1 0 0 0 . 1 5 0 0 . 2 0 0 0 . 2 5 0 0 . 3 0 0
Engineering Strain
E
n
g
i
n
e
e
r
i
n
g

S
t
r
e
s
s

(
k
s
i
)
0
2 0
4 0
6 0
8 0
100
Sigma-0 (ksi)
DUCTILITY
0
0 . 0 5
0 . 1
0 . 1 5
0 . 2
0 . 2 5
0 . 3
0 . 0 0 2 0 . 0 0 4 0 . 0 0 6 0 . 0 0 8 0 . 0 0 1 0 0 . 0 0 1 2 0 . 0 0 1 4 0 . 0 0
Yield Stress (ksi)
U
n
i
f
o
r
m

E
l
o
n
g
a
t
i
o
n
k=75 ksi, n=0.25
σσσ σ
o
=related to yield stress
Figure 4-8: Toughness as a function of yield stress, constant strain hardening.
Summary: As can be readily seen from these calculations and plots, increasing yield
strength reduces ductility and reduces toughness. Although increasing yield stress
increases the UTS, the important ratio between the two, UTS/Yield decreases as the yield
strength increases.
Extension to Flawed Bodies
The tensile test, considered above, is sufficient to understand the role of strain hardening
and yield strength on mechanical properties of homogeneous materials. In fact, these
same relationships extend to flawed or cracked bodies, but the reason for this is not as
quantifiable or as easy to understand.
We consider here only the broadest element of fracture mechanics, with many of the
aspects ignored. In general, we can consider two flaw limits, one is a planar, infinitely
sharp crack (Figure 4-9a) and the other is a very blunt crack (Figure 4-9b). For the case
of the sharp crack and a purely elastic body, the stress near the tip may be represented as
follows:
σ θ
π
σ π = = f
K
r
K a
applied
( ) ,
2
(4-31)
Where K is the stress intensity factor and f(t) is a function that we shall ignore, thus
assuming that the stresses are radially symmetric around the crack tip. Note that for the
idealized sharp crack, the local stress approaches infinity as the location approaches the
crack tip. In practice, cracks are not ideally sharp, even for very brittle glasses, but this is
a reasonable approximation for our consideration. (It is also certainly true that neither
WORK / VOLUME (Ludwik)
9
9 . 2
9 . 4
9 . 6
9 . 8
1 0
1 0 . 2
1 0 . 4
1 0 . 6
1 0 . 8
0 . 0 0 2 0 . 0 0 4 0 . 0 0 6 0 . 0 0 8 0 . 0 0 1 0 0 . 0 0 1 2 0 . 0 0 1 4 0 . 0 0
Yield Stress (ksi)
W
o
r
k
/
v
o
l
u
m
e

(
k
s
i
)
k=75 ksi, n=0.25
σσσ σ
o=related to yield stress
stress nor strain approach infinity, as predicted by Equation 4-31, but the differences can
be small enough to ignore in some cases.)
Figure 4-9: Extreme material crack configurations.
At the other extreme, a blunt crack has a similar long-range field but the local stress at the
tip approaches a maximum value at the tip of
σ σ
ρ
max
= 2
applied
a
(4-32)
Starting from these two extremes, let’s attempt to develop a simple model of fracture that
is capable of predicting the trends in fracture toughness related to material properties
such as strain hardening. Start by assuming that the strain distribution ahead of the blunt
crack follows that of the sharp-crack / elastic one, that is
ε
σ π
π
σ σ
( ) , r
a
E r E
a
r
Ar where A
a
E
applied applied applied
= =




= ≡

2 2 2
1
2
1
2
(4-33)
Sharp crack
a a
ρ
Blunt crack
σ θ
π
σ π = = f
K
r
K a ( ) ,
2
σ σ
ρ
max
= 2
applied
a
where we have simply divided the stress in Equation 4-31 by Young’s modulus, E, to
obtain strain, and have ignored the angular dependence of the crack tip field (thus
assuming that it is radially symmetric). The inverse relationship (r in terms of ε) is:
r
a
E
A
applied
=






=
− −
2
2
2 2 2
σ
ε ε (4-34)
To correct this for a blunt crack and a finite maximum allowable local strain (ε*), limit
our attention to distances in front of the crack tip where the strain is less than ε*.
(Equivalently, we can picture translating the strain distribution if Equation 4-32 such that
the strain ε* occurs just at the crack tipe. Furthermore, in view of the result for the blunt
crack, Equation 4-32, a local fracture criterion based on a strain of ε* is equivalent to a
local stress criterion of σ*, where σ*=f(ε*) and f(ε) is the strain hardening law.) The
assumed situation for the model is shown in Figure 4-10.
Figure 4-10: Hypothesized strain distribution in front of a blunt crack. The blunt crack
origin is taken to be at r* whereas the sharp crack origin is at the origin.
Given the strain distribution, we want to calculate the plastic work that is done in the
plastic zone (between r* and r
y
) as the applied stress is raised sufficiently to just meet the
local fracture criterion, i.e. either σ* or ε*. For a given element of material subjected to a
final strain of ε
final
, the work done per volume of that element for a material hardening law
represented by Equation 4-20 is similar to Equation 4-29:
ε
σ π
π
σ
( ) r
a
E r E
a
r
applied applied
= =




2 2
1
2
S
t
r
a
i
n
Distance from Crack Tip
ε*
ε
y
r*
r
y
W
v
k d
k
n
o
n
o final final
n final
= +
( )
= +
+





+
σ ε ε σ ε ε
ε
0
1
1
(4-35)
To get the work (not just work per volume element), we must integrate over the plastic
volume near the crack tip. To simplify this, we assume again that the strain field is
radially symmetric, but only in front of the crack tip, Figure 4-11.
Figure 4-11: Plastic zone configuration envisioned in simplified fracture model.
Each semi-annular area in the plastic zone has a volume of dv = πrdr. To find the total
work done in raising the external stress to the critical value, we integrate over the volume
as follows:
W
W
v
dv rdr
k
n
rdr
v
o final
r
r
final
n
r
r
y y
= = +
+




∫ ∫ ∫
+
σ ε π ε π
* *
1
1
(4-35)
but the final strain in each volume element depends on its position in the plastic zone, r,
according to Equation 4-33:
ε
σ σ
final
applied
final
applied
final
r
E
a
r
Ar where A
a
E
( ) , =




= =

2 2
1
2
1
2
(4-36)
With this substition into Equation 4-35, the integral is ready to be evaluated:
W Ar rdr
kA
n
r rdr
o
r
r
n n
r
r
y y
= +
+

+ − +
∫ ∫
σ π π
1
2
1 1
2
1
* *
( )
(4-37)
which is
W B r r C r r
y y
n
n
= −






+ −








3
2
3
2
3
2
3
2
* * (4-38)
where: B
A
C
kA
n n
o
n
= =
+ −
+
2
3
2
1 3
1
σ π π
,
( )( )
We need only find expressions fo r
y
, r*, σ
applied
(final) in order to perform the necessary
calculations. These are readily obtained by the assumptions in our model and the
r
p
=r*-r
y
a
ρ
assumed form of the hardening law. The final applied stress is just enough to produce the
critical local stress σ* at the crack tip. But σ* depends on our choice of a fracture strain
(or vice-versa):
σ σ ε ε
σ σ
* * , *
*
= + ( ) =





o
n
o
n
k or
k
1
(4-39)
With σ* known, the relationship to the applied stress is known for a blunt crack of a
given configuration (i.e. length and tip radius, Figure 4-11) from Equation 4-32:
σ
σ ρ
applied
final
a
=




*
2
1
2
(4-40)
Then the inner cutoff radius for the plastic zone is known from Equation 4-34 and
Equation 4-39:
r A A
k
o
n
* *
*
= ( ) =







− 2 2 2
2
ε
σ σ
(4-41)
The outer cutoff radius (transition to elastic deformation) occurs when the stress drops to
the yield stress, σ
o
:
r A A
E
y y
o
=
( )
=







2
2
2
2
ε
σ
(4-42)
These equations are sufficient to evaluate the work, the final applied stress for
propagation, and the stress distribution in front of the crack in terms of material
properties (E, k, n, ε*or σ*) and crack properties (ρ, a).
Assuming hardening law values similar to the ones that we adopted for the tensile test
(k=75,000psi,n=0.25), other plausible steel parameters (E=30x106psi, ε*=1), and some
arbitrary crack dimensions (ρ=0.01”, a=0.1”) it is possible to investigate the role of
changing various parameters on the material toughness. Figures 4-12 to 4-14 show a few
of these results, starting from the standard hardening law.
Figure 4-12 shows the calculated stress ahead of the blunt crack tip. While the stress at
the tip increases rapidly for increasing yield stress (really σ
o
), the extent of the plastic
zone decreases dramatically. Clearly, the area under the curve, proportional to the work,
is much less for stronger materials. (Note that because of the semi-circular nature of the
plastic zone, much less material is deformed close to the crack trip, where the stresses are
highest, than at the larger radii.)
Figure 4-13 shows the calculated work as a function of yield stress (really σ
o
). As
expected, the work done decreases rapidly with increasing flow stress.
In Figure 4-14, the applied stress to attain σ* at the crack tip is shown. This is a direct
consequence of Equation 4-32 for a blunt crack with a fixed tip radius.
Figure 4-12: Stress distribution ahead of a blunt crack for various values of σ
o
, the yield
stress (see Equation 4-20).
Stress Ahead of Blunt Crack
0
1 0 0 0 0
2 0 0 0 0
3 0 0 0 0
4 0 0 0 0
5 0 0 0 0
6 0 0 0 0
7 0 0 0 0
8 0 0 0 0
9 0 0 0 0
1 0 0 0 0 0
0 0 . 0 2 0 . 0 4 0 . 0 6 0 . 0 8 0 . 1 0 . 1 2 0 . 1 4
Distance from Crack Tip (inch)
S
t
r
e
s
s

(
p
s
i
)
σσσ σ
o
= 8 0 , 0 0 0 ps i
σσσ σ
o
= 6 0 , 0 0 0 ps i
σσσ σ
o
=4 0 , 0 0 0 ps i
σσσ σ
o
=20, 000psi
σσσ σ
o
=15, 000psi
σσσ σ
o
=1 0 , 0 0 0 ps i σσσ σ
o
= 7 , 0 0 0 ps i
σσσ σ
o
= 5 , 0 0 0 ps i
Yield stress = σσσ σ
0
, k=75,000 psi,
n = 0 . 2 5
Crack length = 0.1in, Tip radius=0.01in
Young's modulus = 30x106psi
Figure 4-13: Calculated work to propagate a blunt crack. (See model.)
Figure 4-14: Applied stress to obtain ε* at the tip of a blunt crack.
WORK TO PROPAGATE BLUNT CRACK
0 . 0 0
0 . 0 5
0 . 1 0
0 . 1 5
0 . 2 0
0 . 2 5
0 . 3 0
0 . 3 5
0 . 4 0
0 . 4 5
0 . 5 0
0 2 0 0 0 0 4 0 0 0 0 6 0 0 0 0 8 0 0 0 0 1 0 0 0 0 0 1 2 0 0 0 0
Yield Stress (psi)
W
O
R
K

(
p
s
i
-
i
n
)
Yield stress = σσσ σ
0
, k=75,000 psi,
n = 0 . 2 5
Crack length = 0.1in, Tip radius=0.01in
Young's modulus = 30x106psi
APPLIED STRESS TO PROPAGATE BLUNT CRACK
0
5 , 0 0 0
1 0 , 0 0 0
1 5 , 0 0 0
2 0 , 0 0 0
2 5 , 0 0 0
3 0 , 0 0 0
0 2 0 0 0 0 4 0 0 0 0 6 0 0 0 0 8 0 0 0 0 1 0 0 0 0 0 1 2 0 0 0 0
Yield Stress (psi)
C
r
i
t
i
c
a
l

A
p
p
l
i
e
d

S
t
r
e
s
s

(
p
s
i
)
ield stress = σσσ σ
0
, k=75,000 psi,
= 0 . 2 5
rack length = 0.1in, Tip radius=0.01in
oung's modulus = 30x106psi
Applied stress =
Yield stress line
(brittle behavior to
the right of
i nt er sect i on)
Microstructural Origins of Strength
As outlined briefly in Section 1, the flow stress of a metal or alloy may be considered
conceptually to consist of three additive terms:
σ
flow
= σ
i
(intrinsic) + ∆σ
micro
(microstructural) + ∆σ
strain
(strain hardening) (4-43)
In view of the results for a hardening law of the type in Equation 4-20, it is useful to lump
the first two terms into one that is assumed not to depend on strain and to maintain the
third term which contains all of the strain-dependent parts:
σ
flow
= σ
o
(strain-independent) + ∆σ
strain
(strain hardening) (4-44)
where σ
o
= σ
i
+ ∆σ
micro
.
The intrinsic term was discussed in Section 1 in terms of the Peierl’s stress, but ∆σ
micro
has not been treated except to list the kinds of microstructural features that may
contribute to it.
Pin Model: The key to understanding ∆σ
micro
lies in the idea of dislocation pinning by
foreign objects which interact with the dislocation. The extremes of pinning behavior are
illustrated in Figure 4-14.
Figure 4-14: Pinning of dislocations, two extreme cases.
Hard Pin / Soft Dislocation
φ
φ ∼ 0
Soft Pin / Hard Dislocation
φ ∼ π
L L
The critical angle at which dislocations can break away from the pin, φ
c
shown on Figure
4-14, is a property mainly of the pin strength for a given base metal (such at iron in
steels). The critical shear stress acting on the dislocation, necessary for the break away,
may be found as follows:
τ β
φ
c
Gb
L
=




cos
2
(4-45)
where β is a constant between 0.5 and 1.5 depending on the character of the dislocation
(0.5=edge, 1.5=screw, intermediate configurations are intermediate), G is the shear
modulus, b the burgers vector, and L the spacing between pins. For ideally “hard” pins
and assuming that an average value of β is 1.0, the “hard pin” model gives”
τ
c
Gb
L
= (4-46)
Equation 4-45 or 4-46 shows that the closer the distance between pins, the stronger the
alloy will be. On the other hand, the strength of a precipitate decreases, for example, for
very small sizes. This happens because it is no longer necessary for the dislocation to
loop around the small precipitate, but it can instead either bypass it by a cross-slip or
climb mechanism, or it can cut through the precipitate itself via a “shearing mechanism.”.
Therefore, at very small precipitate sizes, the strength of the alloy decreases because of
shearing, even though the spacing between particles may be very small. Above some
critical precipitate size (maybe of the order of 100’s of angstroms), the pins are
essentially “hard” and the only important variable is the spacing between them. Once
this threshold size is reached, there will be a continued decrease in strength as the
particles become larger (and thus further apart). Figure 4-15 illustrates this concept
schematically.
Figure 4-15: Schematic representation of the effects of increasing particle size on the
strength of an alloy.
Y
i
e
l
d

s
t
r
e
n
g
t
h
Particle Radius
∝∝∝ ∝ 1/r
∝∝∝ ∝ r1/2
Peak Strength
Orowan
by-pass
Shearing
mechanism
Strain-Hardening: Strain hardening can be understood in terms of the pin model
described above. In fact, dislocations are “hard pins” in the classical sense: they serve as
hard obstacles for perpendicular dislocations trying to cut through them.
In order to understand strain hardening, we must cite a well-known relationship that has
been measured many times:
σ σ ρ = +
o
k (4-47)
where ρ is the dislocation density (in dislocations per cm
2
, for example). This empirical
relationship is easily understood in terms of the hard pin model by considering a regular
array of cross-cutting dislocations, Figure 4-16.
Figure 4-16: A moving dislocation encountering a regular array of cross-cutting
dislocations.
By considering a periodic cell of four dislocations at the corners, it is cleared that each
dislocation lies in four cells, so there is a net of one dislocation per cell having an area of
L2. Thus, the relationship between dislocation spacing and density is
ρ
ρ
= = 1
1 1
2
L
or L , (4-48)
It is only required to substitute for L in the hard-pin model (Equation 4-46) to obtain
σ
β
β ρ
crit
Gb
L
Gb = = (4-49)
and we can take into account the other kinds of hardening unrelated to dislocation
obstacles with a constant additive term:
L

L

ρ
ρ
= = 1
1 1
2
L
or L ,
σ σ ρ β
crit o
k where k Gb = + = , (4-50)
Comparison of Equation 4-49 with Equation 4-47 shows that this simple view of
dislocation interactions is sufficient to obtain the observed dependence. However, it is
difficult to find a satisfactory model to predict the net rate of dislocation multiplication
during straining. By some geometric constructions, the dislocation density should be
proportional to the strain, however this view ignores the dislocation annihilation that
occurs, particularly at grain boundaries. If dislocation density were proportional to strain,
the Equation 4-49 would take on our empirical strain hardening form, Equation 4-20,
with an n value of 0.5. That is much higher than is usually observed, particularly in
steels, although n in some 70/30 brasses approaches 0.5.
It is best at the current development of understanding, to simply measure the strain
hardening law in a tensile test and thus to use the empirical form thus obtained.
Hall-Petch Relationship: The Hall-Petch relation relates the yield stress ( the lower yield
stress for samples that show a yield point phenomenon) linearly to the inverse of the
square root of the average grain diameter, i.e.
σ σ = +

o
kd
1
2
(4-51)
where k is a proportionality constant and d is the grain diameter. A few common values
for k are as follows: steel: 0.3 MNm
-3/2
, Armco iron: 0.6 MNm
-3/2
, Molybdenum: 1.8
MNm
-3/2
, Aluminum: 0.07 MNm
-3/2
, Copper: 0.11 MNm
-3/2
. Plots of this emperical
behavior are shown in Figure 4-17.
0
100
200
300
0 2 4 6 8 1 0 1 2 1 4
T
r
u
e

S
t
r
e
s
s

(
M
P
a
)
Fe
70/ 30
Brass
Cu
Al
RT
d
-1/2
- (Grain Size)
-1/2
(mm
-1/2
)
Fig. 4-17: Yield stress – grain size relationship at room temperature for several metals.
(After N. Hansen: Polycrystalline Strengthening, Metall. Trans. A, 1985, Vol. 16A, pp.
2167-2190)
It should be noted with reference to Figure 4-17, or the k values presented above, that
grain size strengthening can be an important effect in steels, but is much less significant
in aluminum or copper alloys. Presumably, this differences arises because of the high
stacking fault energy in bcc metals. In turn, this discourages separation of dislocations
into partials dislocations separated by high-energy stacking fault. Whole dislocations can
cross-slip to evade obstacles, hence the observation of “wavy slip” in steels. On the other
hand, fcc metals like copper have low stacking fault energies and the dislocations readily
dissociate into partial dislocations. Once this is done, it is very difficult for the partial
pair to change planes and thus “planar slip” is observed for high-stacking fault energy
metals. How this translates precisely to the differences observed for Hall-Petch
coefficients is beyond the scope of this course.
A simple model of grains pinning dislocations would lead to an inverse relationship
between grain diameter and strength. In fact, the data is often conflicting and scattered,
particularly for weak grain-hardening materials, so that the power on d of –1/2 or –1 is
not unequivocal. However, there are several models to explain the –1/2 power in the
Hall-Petch equation, the simplest and best-known of which is based on the ideas of
dislocations pile-ups, Figure 4-18.
Figure 4-17: Dislocation pile-up at a grain boundary.
The length of a pile-up is related to the number of dislocations present, n, and the applied
shear stress, τ
applied
, as follows:

L

D

L
Gbn
applied
=
− τ π ν ( ) 1
L
Gbn
applied
=
− τ π ν ( ) 1
(4-52)
or, the number of dislocations in a given length equal to one-half of the grain diameter
(i.e. L=D/2) can be solved for directly:
n
D
Gb
applied
=

2
1 τ π ν ( )
(4-53)
The grain boundary is assumed to provide a local impediment to dislocation motion
equivalent to some local critical stress t*, while the stress equivalent to the force pushing
on the first dislocation is simply n times tapplied. (That is, all of the forces felt by the
individual dislocations sitting in tapplied is resisted by the single obstacle. The total
force at the front is n times the force felt by each dislocation individually.) Thus
τ τ
τ π ν
*
( )
*
= =

n
D
Gb
applied
applied
2
1
2
(4-54)
or, rewritten to solve for the applied critical stress to pass dislocations
τ τ
τ
π ν
applied applied
n
Gb
D
*
*
( )
= =







2
1
1
2
(4-55)
Note that we now have the observed dependence of observed flow stress on grain
diameter. When other hardening mechanisms are considered and put into a term that
does not depend on grain size, the overall relationship is obtained:
σ σ
τ
π ν
= + =








n kD where k
Gb
o
1
2
1
2
2
1
,
*
( )
(4-56)
There are other models leading to the Hall-Petch relationship, but this one is simplest and
most often cited. The main difficulty with this one is that planar pile-ups are very rare in
steels, because of the high stacking fault energy, ease of cross-slip and hence wavy slip.
On the other hand, while the conditions for pile-ups are better in fcc metals, for these the
Hall-Petch coefficient is quite small.
Mechanical Properties of Plain Carbon Steels.
Ferrite Pearlite steels: Plain carbon steel is very often used when it consists of a
mixture of ferrite and pearlite. The yield and ultimate strengths of these steels are
controlled by the mechanical properties of the ferrite and cementite and microstructural
details concerning how they are mixed together. We have already reviewed the behavior
of iron when it is deformed in a tensile test. We have also noted the effects of solutes
dissolved in iron. In particular, when the interstitial solutes, C and N, are allowed to
diffuse they segregate at and around the dislocation lines. At room temperature this
segregation can prevent most dislocations from moving under applied stresses so that
yield requires the activation of sources at the sites of stress concentrations or the creation
of new dislocation line. This gives rise to the yield point phenomenon in which the iron
shows an upper and lower yield point and a Luders extension. At higher temperatures,
serrated yield can occur. Another consequence of the interactions between the interstitial
solutes and dislocations is the increase in strength and the corresponding reduction in
ductility arising from strain aging. Ferrite is a BCC iron solid solution and it can behave
as described above.
The flow stress in polycrystalline ferrite ( this is the lower yield strength if a yield point
phenomenon occurs) is dependent on dislocation density and grain size. Taking these
contributions to be additive, we can write the yield stress as:
σ
y
= αµρ
1
2
+ md
−1
2
where α and m are material constants, d is the grain diameter, ρ is the dislocation density
and µ is the elastic shear modulus. The grain size dependence arises because the grain
boundaries act as obstacles to the gliding dislocations.
Cementite is a hard and brittle phase. It tends not to deform plastically at ambient
temperatures. It can crack. Hence, yield in a pearlitic microstructure is associated with the
movement of dislocations in the ferrite. When dislocation sources are activated in the
ferrite plates the glide paths of the dislocations are limited by the cementite plates
because they cannot slip in the carbide. Instead they are forced to pile up at or near the
ferrite/cementite interfaces. The first consequence of the lamellar microstructure of
alternating ductile and hard, brittle plates is an increase in the yield stress of the material
compared to that of the ferritic matrix because the inter-lamellar spacing is usually much
smaller than the proeutectoid ferrite grain size. We expect that the increase in the yield
stress due to the restriction of dislocation glide paths in the ferrite lamellae should be
larger the finer the spacing in the pearlite. This expectation turns out to be true.
Stress concentrations are created at the ends of pile ups where the dislocations meet the
ferrite carbide interfaces and eventually the carbides can be broken. This occurs more
rapidly when the ferrite plates are thicker because more dislocations can participate in an
individual pile-up for a given applied stress. The generation of cracks in the cementite
plates can initiate general fracture which implies that coarse pearlite should fail before
fine pearlite because the cementite plates are expected to crack sooner when the ferrite
lamellae are thicker and, also, the cracks will be longer and therefore more able to
propagate. This prediction is also true; fine pearlite is both stronger and more ductile
than coarse pearlite.
A large tonnage of low carbon steel, containing 0.15 wt.% C and below is used in sheet
form. Slow cooling from hot rolling processes or annealing heat treatments leaves these
steels in the form of mixtures of ferrite and pearlite but the ferrite dominates the
mechanical properties because it comprises most of the steel and is the most continuous
phase. These hot rolled steels can have yield strengths of 25 to 45 K.S.I (172-310 MPa)
and ultimate tensile strengths of 40 to 65 K.S.I (276-448 MPa). The yield strength can
rise to 100 K.S.I. (690 MPa) as the carbon content of the steel is increased because the
properties of pearlite eventually become dominant. When the carbon level becomes high
enough for the pearlitic microstructure to influence the properties, the morphological
details such as pearlite spacing also affects the yield and ultimate strength levels as
shown in the sketches provided with this text.
Impact Properties of Ferrite Pearlite Mixtures: The mechanical properties discussed
above are those measured during standard low strain rate tensile or compression tests.
The response of materials to high strain rates is also important. This is particularly true
for BCC metals which, unlike FCC metals, can have a ductile to brittle transition
temperature under impact loading. An indication of the response to impact loading is
provided by the Charpy impact test in which a notched, square-section bar is broken by a
swinging pendulum. Measuring the height of the pendulum at the beginning of the test
and that to which it rises after breaking the sample, enables the energy consumed in
breaking the bar to be calculated. When the bar breaks in a brittle mode the energy
consumed is small. In this case the bar breaks into two pieces and the pendulum swings
onwards and upwards. When the fracture occurs in a ductile fashion, a large amount of
energy is consumed by the deformation processes that precede fracture as well as those
that accompany fracture. Then the distance traveled by the pendulum after the fracture is
smaller. Very ductile samples do not break and the pendulum is stopped. A crafty trick
often done by certain professors I know, while teaching a lab course in which Charpy
impact tests are being performed, is to sneak a sample of annealed, austenitic stainless
steel into a group of brittle samples that break easily, then to watch them being tested
and to chuckle like mad at the surprise of the person performing the tests, when the
pendulum comes to a sudden halt with a loud clunk.
The nature of the fracture and the amount of energy consumed in an individual test
depends upon the temperature of the specimen at the time it was tested. It is standard
practice to perform Charpy tests over a range of temperatures. When considering the
results of these kinds of tests, it is wise to bear in mind that microstructural features other
than the those of the ferrite and pearlite can affect them. For example, a common
ingredient in a carbon steel is MnS which is formed when the alloying element, Mn,
combines with the residual impurity, S. The MnS is present in the form of inclusions in
the steel and the morphology of these inclusions depends upon prior history. When the
steel has been processed by hot rolling, the inclusions can be elongated and strung out in
the rolling direction and are called, therefore, inclusion stringers. The important point is
that the interface between the MnS and the steel is not strong. If the steel being tested
happens to have stringers that lie parallel to and near to the notch, the fracture will not be
representative of the steel itself, but of the interfaces between the inclusions and the steel.
Burns and Pickering (JISI 202, 899, 1964) studied the effects of the volume fraction of
pearlite on the impact properties of carbon steel
%

E
L
O
N
G
A
T
I
O
N
1100 1300
I.T. TRANS F
Temperate at which
austenite transformed
F
15
5
80
100
120
140
160
180
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Fine
Coarse
alpha+pearlite
0
10
20
30
40
50
60
100 120 140
160
180 200
σσσ σ
u
KSI
% r e
or
%R
<=D
N
in area
I.T. 1300 F
I.T.
Isothermally
transformed
I.T. 1200 F
Reduction in area
% Elongation
σσσ σ
u
= Ultimate strength
-80
-40 0 40 80
Impact
energy
ft/ lbs
200 0.11 C
0.2 C
0.31 C
0.41 C
0.8 % C
C
Charpy impact energy ft. lbs
VS
Temperature of impact as a
function of % C in
ferrite/pearlite steels with
low inclusion content.
After: Honeycombe, p.167
0
5 0
1 0 0
1 5 0
2 0 0
- 1 5 0 - 5 0 5 0 1 5 0
Test Temperature (C)
I
m
p
a
c
t

E
n
e
r
g
y

(
J
)
0.11% C
0.2% C
0.31% C
0.41% C
0.49% C
0.6% C
0.69% C
0.8% C
1 joule = 0.74 ft-lb
After: Callister, p.323,324
0 3 6 9 12 15
350
300
250
200
150
100
160
140
120
100
80
60
40
0 0.2 0.4 0.6 0.8 1.0
1200
1100
1000
900
800
700
600
500
400
300
M
P
a
p
s
i

x

1
0
3
Percent Fe
3
C
Y
i
e
l
d

a
n
d

t
e
n
s
i
l
e

s
t
r
e
n
g
t
h
Composition (wt.% C)
B
r
i
n
e
l
l

h
a
r
d
n
e
s
s

n
u
m
b
e
r
Tensile
strength
Brinell
hardness
Yield strength
Pearlite
+ ferrite
Pearlite
+ Fe
3
C
(a)
0 3 6 9 12 15
0 0.2 0.4 0.6 0.8 1.0
Percent Fe
3
C
Composition (wt.% C)
Pearlite
+ ferrite
Pearlite
+Fe
3
C
(b)
80
60
40
20
0
120
100
80
60
40
20
0
D
u
c
t
i
l
i
t
y

(
%
)
I
z
o
d

i
m
p
a
c
t

e
n
e
r
g
y

(
f
t
-
l
b
f
)
Izod
impact
energy
Area
reduction
Elongation
0 3 6 9 12 15
0 0.2 0.4 0.6 0.8 1.0
Percent Fe
3
C
Composition (wt.% C)
(a)
280
240
200
160
120
80
320
B
r
i
n
e
l
l

h
a
r
d
n
e
s
s

n
u
m
b
e
r
HRC
35
30
25
20
100
90
80
70
60
HRB
R
o
c
k
w
e
l
l

h
a
r
d
n
e
s
s
Fine
pearlite
Coarse
pearlite
Spheroidite
0 3 6 9 12 15
0 0.2 0.4 0.6 0.8 1.0
Percent Fe
3
C
Composition (wt.% C)
(b)
80
70
60
50
40
30
90
D
u
c
t
i
l
i
t
y

(
%
A
R
)
20
10
Spheroidite
Coarse
pearlite
Fine
pearlite
in which the amount of the MnS present and its morphology was controlled in order to
minimize its influence on the results. They found that 0.11 wt.% carbon steel had a flat
upper shelf ( i.e. the energy vs. temperature plot was horizontal and at a high energy
value) where the fracture was ductile and a large amount of energy was consumed. As the
temperature of the test was decreased this steel showed a transition from a ductile to a
brittle fracture that occurred rapidly near -40°C. At lower temperatures the plot of energy
consumed vs. the temperature of the test became horizontal again, forming the lower
shelf. As the carbon content was increased, the energy associated with the upper shelf
decreased and the transition from ductile to brittle fracture took place at higher
temperatures. The temperature range occupied by the ductile to brittle transition
increased. The eutectoid steel, which is entirely pearlitic, shows a gradual transition of
the energy consumed as a function of temperature. Presumably, the gradual transition at
the high pearlite levels occurs because even though the cementite plates crack, the total
fracture path is tortuous in comparison to cleavage in the ferrite. Cleavage fracture in
ferrite is a brittle fracture process in which the fracture path follows particular
crystallographic planes in each grain, forming large, relatively flat facets.
Until approximately 20 years ago, pearlitic steels were commonly chosen as structural
steels because of the useful strength levels that can be attained in them. Unfortunately,
the use of medium to high carbon steels in structures such as bridges and grain elevators
created the practical difficulties encountered in joining components made from these
steels which were therefore commonly bolted and riveted together. Their use in such
structures is not really desirable because the low toughness and high ductile to brittle
transition temperatures can cause any failures that occur to be catastrophic. These days,
alternative ways are used to get high strength while maintaining good toughness and a
low ductile to brittle transition temperature, e.g. through the development of weldable,
high strength low alloy steels.
In hypereutectoid steels the proeutectoid constituent is cementite. In steels cooled at rates
which permit the formation of pearlitic microstructures, the proeutectoid phase has time
to nucleate on and grow along the austenite grain boundaries. When this occurs the steel
becomes less tough because the thin continuous film of cementite on the prior austenite
grain boundaries becomes the most continuous phase and dominates the properties. The
makers of Damascus swords in the middle ages, used steels with over 1 wt.% carbon but
produced hard, tough blades by working the steel below the A
cm
line. This destroyed the
continuous films of cementite. Apparently European blacksmiths didn't know about that
and couldn't use such high carbon steel. In modern times the undesirable effects of
continuous cementite films along prior austenite grain boundaries can be important in
case hardening treatments in which high carbon levels are created near the surfaces of
parts that are then quenched and tempered. When such parts are broken it is common to
find a region of the fracture which appears to be intergranular because the crack
propagated along the proeutectoid cementite films.
Bainite and Martensite: The lamellar microstructure of pearlite has a strong effect upon
its toughness, with the toughness increasing as the spacing decreases. In bainite there is
no lamellar structure. Instead, it is an array of sheaves of very fine laths, needles or
platelets in hypoeutectoid steels and the fine carbides therein, are discrete particles. As a
consequence, bainitic microstructures are generally tougher than pearlitic
microstructures. Because of the fine scale of its microstructure bainite is also harder than
pearlite in steels of the same carbon contents.
Martensite is a supersaturated solid solution in which the interstitial solutes interact
strongly with dislocations. It is also composed of very fine crystals with many
intercrystalline boundaries; it can be highly dislocated or twinned. For these reasons, it is
the strongest form of a carbon steel. For example, the 0.2% proof stress is 120 K.S.I. (828
MPa) for martensite containing 0.1% C and 240 K.S.I. (1.65 GPa)for 0.4% carbon.
Correspondingly, the Rockwell C hardness number increases from about 30 to 65 with a
change in the carbon level from 0.1 to 0.6%. There is little change in the hardness for
further increases in the carbon content. The hardness of freshly formed martensite is
directly dependent upon its hardness content. This is shown in the appended figure. Also
shown on this figure are the data for air cooled austenite(curiously labeled pearlitic) and
spheroidized steels. Such steels consist of a polycrystalline array of ferrite crystals in
which discrete globules of cementite particles are embedded. This microstructure
represents the softest state of a carbon steel. A martensitic, plain carbon steel can be very
hard and very strong. Unfortunately, freshly formed martensite ( i.e. as-quenched) lacks
toughness and is intolerant of small flaws which can initiate brittle fracture. For this
reason, freshly quenched martensite is not knowingly used in practical applications. It is
heat treated further in order to impart to it with some toughness. These heat treatments
are generally compromising treatments because to gain toughness one has to sacrifice
some strength. These additional heat treatments are called tempering treatments and are
the subject of the next discussion.
Tempering Reactions in Carbon Steel
The heat treating process that is applied to martensite in order to endow it with
better toughness is called tempering. A tempering treatment consists of heating the steel
to a selected temperature below the A
1
line, holding it there for a certain time and then
cooling it back down to ambient temperature. The transformations that occur and the
property changes that are produced, depend upon the tempering time and temperature.
The general aim is to raise the toughness of the steel. Because this is achieved at the cost
of a reduction in hardness and strength, the process is always a compromise and is
selected according to the intended application. A great deal of research has enabled us to
understand much of what goes on, but there are still some uncertainties with respect to
transition carbides.
On the basis of the research into the tempering process, it has become customary to
divide the process into stages that occur in different temperature regimes. It should be
recognized that this division is a convenient way of describing tempering reactions; the
particular temperature ranges should be considered as guide lines and , obviously,
neighboring stages overlap.
Stage 1 occurs at temperatures below 200 °C. Here, carbon atoms diffuse throughout the
martensite and segregate at dislocations and intercrystalline boundaries. Because the M
s
temperature for carbon steels is higher than room temperature, it is possible for these
segregation processes to begin when the martensite is first formed, especially if the
cooling rate below the M
s
point is slow. In this case, one refers to the changes as being
produced by auto tempering . For steels of less than about 0.2 wt.% C no carbides form
in this stage. When the carbon content is greater than 0.2 wt. %, very fine particles of a
transition carbide nucleate and grow within the martensite. This carbide is richer in
carbon than cementite and is described by the formula Fe
2.6
C. It is metastable. It has a
hexagonal structure and is referred to as epsilon carbide. Several
After: Bain & Paxton, Alloying Elements in Steels, 2
nd
edition, p.37
1100
1000
900
800
700
600
500
400
300
200
100
0.20
0.40
0.60 0.80
1.00 1.20
10
20
30
40
50
60
65
H
a
r
d
n
e
s
s
,

R
o
c
k
w
e
l
l

C
H
a
r
d
n
e
s
s
,

V
i
c
k
e
r
s
Martensitic structure (quenched)
Pearlite structure (air cooled)
Spheroidized carbide structure
Carbon, %
researchers have claimed to have observed the presence of another metastable carbide in
high carbon steel. This one has been christened eta carbide. It is orthorhombic Fe
2
C.
There is, however, still uncertainty because the experimental confirmation of these claims
is not easy. It is generally accepted that epsilon carbide definitely forms. When it does,
the martensite retains some degree of tetragonality because it still contains more carbon
in solid solution than would ferrite; there are no changes in the morphology of the
martensite crystals. Thus the first stage of tempering involves the segregation of carbon
to various defects in the microstructure and the conversion of the martensite to low
carbon martensite plus epsilon carbide when the total carbon content is high enough.
During stage 1 tempering, there are changes in the physical properties such as electrical
resistivity which can be used to monitor the progress of the changes. However, there is
not much of a reduction in the hardness; indeed, it can increase slightly for steels of
medium to high carbon levels. Obviously, these changes involve diffusion and are
therefore time dependent. A common practice is to use tempering times of one hour so
much of what follows will refer to the changes produced by tempering treatments one
hour long .
The second stage of tempering occurs between 200 and 300°C. It is not very important in
plain carbon steels because it refers directly to the conversion of retained austenite to
bainite. The quantities of retained austenite in carbon steels are not large. However, they
can be large in alloy steels especially those for which the M
f
temperature is below 0°C;
stage 2 is important when there are significant quantities of retained austenite included
within the martensite.
Cementite begins to precipitate out from solid solution at about 300°C. This is the
beginning of stage three. The precipitates nucleate on heterogeneous sites such as
intercrystalline boundaries. Because the martensitic crystals form with specific
orientation relationships with the austenite crystal in which they grow, they do not grow
across austenite grain boundaries. In effect, the prior austenite grain boundaries are
retained as surface defects. At the beginning of stage three, the heterogeneous nucleation
process at the sites of the prior austenite grain boundaries tends to produce sheets of
cementite particles. The matrix loses tetragonality and becomes BCC ferrite as its carbon
content drops because of the diffusion of carbon to these growing carbides . If the steel is
sufficiently rapidly heated to the tempering temperature then no transition carbides form.
If transition carbides were present due to a previous tempering treatment or because they
formed during a slow heat-up, then they dissolve during stage three. The phases present
become ferrite and cementite. When tempering is carried out at temperatures near 300 °C,
the ferrite crystals remain acicular during the times usually used. Because the diffusion
coefficient of carbon increases with an increase in the temperature, the cementite
particles are able to grow more rapidly and to decrease in number per unit volume, as
higher tempering temperatures are used. The hardness after tempering becomes
increasingly smaller; an effect which becomes more and more significant as the
tempering temperature becomes higher. Eventually the temperature becomes high enough
to cause the dislocation content to decrease in a similar manner to the changes in the
dislocation content of cold-worked metals during recovery processes. At about 600°C,
the ferrite grains are able to grow and become equiaxed. These morphological changes in
the ferrite grains are similar to those that occur during recrystallization in cold worked
ferrite. Because of this, these later processes have been said to comprise a fourth stage of
tempering, but that name has also been assigned to secondary
70
65
60
55
50
45
40
35
30
10
1
10
2
10
3
10
4
10
5
300
400
500
600
700
1 min 1 h 1 day
Tempering time (s)
R
o
c
k
w
e
l
l

h
a
r
d
n
e
s
s
,

H
R
C
B
r
i
n
e
l
l

h
a
r
d
n
e
s
s

n
u
m
b
e
r
205
o
C
(400
o
F)
315
o
C
(600
o
F
)
4
2
5
o
C
(8
0
0
o
F
)
5
3
5o
C
(1
0
0
0o
F
)
Q
10
20
30
40
50
60
70
80
200
400
600
800 1000
1200
Tempering temperature, C
100
200
300 400 500 600 700
Tempering temperature, F
1.2% C steel
0.80 % C steel
0.35 % C
steel
H
a
r
d
n
e
s
s
,

R
o
c
k
w
e
l
l

C
After: Bain & Paxton, p.38
After: Callister--page329
After: Krauss: Steel: Heat Treatment and processing principles, p205,206
1 0
2 0
3 0
4 0
5 0
6 0
7 0
200 400 600 800 1000 1200 1400
100 200 300 400 500 600 700
H
a
r
d
n
e
s
s
,

R
o
c
k
w
e
l
l

C
Tempering temperature, F
Compiled from data of
Luerssen and Greene (108)
Engel (112)
Fletcher and Cohen (100)
Bain (113)
Lindsey and Roff (114)
Crafts and Lamont (115)
Wellauer (116)
And Own Data
Tempering temperature, C
As-
quenched
10- 20C
20- 30C
30- 40C
40- 50C
50- 70C
70- 110C
0
20
40
60
80
100
0 200 400 600 800 1000 1200 1400
I
z
o
d

o
r

V
-
n
o
t
c
h

C
h
a
r
p
y

a
t

r
o
o
m

t
e
m
p
e
r
a
t
u
r
e
,

f
t
-
l
b
Tempering temperature, F
0.40 % C
V-notch Charpy
0.50 % C
Izod
A
f
t
e
r

A
S
M

M
e
t
a
l
s

H
a
n
d
b
o
o
k
,

v
o
l
.
4
,

9
t
h

e
d
i
t
i
o
n
,

p
.
7
2
hardening processes that can occur during the tempering of alloy steels so to avoid
confusion we won't use it for carbon steels. Instead, we will consider all of processes that
occur above three hundred degrees Celsius as belonging to stage three.
These various stages offer a convenient way of regarding the tempering reactions.
However, it is important to recognize that the temperatures that have been quoted to
delineate the stages are merely guidelines. Obviously the stages overlap because the
dominant type of change and the rate thereof depends upon temperature and time and
does not change discontinuously. It is also worthwhile pointing out a common error
made by students new to this subject when they describe tempering reactions in small
samples that have been rapidly up-quenched to high tempering temperatures such as 600
°C. The error is to describe the transformations that take place as those of stages 1,2 and
3 occurring sequentially. Unless the heat-up time is slow enough, epsilon carbide will
not form at this temperature and pearlite not bainite can form from retained austenite
because the tempering temperature is above the position of the nose of the TTT diagram.
Effects of Tempering on the Mechanical Properties: During stage one, the hardness
change after an hour of tempering is not great. But as the temperature is increased the
hardness begins to fall rapidly and continues to do so as the temperature increases.
Examples of this behavior are presented. It is tempting to believe that as the hardness
falls the toughness rises. In general this is true, but in detail it is not. When carbon steels
are tempered at temperatures near the beginning of the onset of stage 3, say 350°C or
500°F, there is often a reduction in the Charpy impact energy. This is traditionally known
as 500 degree embrittlement. The drop in toughness seems to be associated with the
formation of sheets of plate-like cementite particles along boundaries and can be
aggravated by the segregation of elements such as As, Sb, and P. Tempering at this
temperature is normally avoided. Similar phenomena also occur in low alloy steels, as we
will see later. The tensile strength of a carbon steel is related to hardness and falls as the
hardness is reduced. Clearly, the choice of tempering conditions depends on the
importance of the relevant properties.
The advantages gained by heat treatments that use austenitization, quenching to
martensite and then tempering the martensite are that the desired microstructure can be
obtained either throughout the part or at selected locations. Obtaining the desired
microstructure means that the proper mechanical properties can be obtained where
required. All of this is true provided that the part being heat treated can indeed be
converted to martensite when the austenite is cooled. If it can not be so transformed then
the required microstructure cannot be produced. This brings us to the concept of
hardenability which concerns the question of how much of a steel bar can be converted
into martensite by a specific quench. This question is of central importance in quench
hardening heat treating processes. The TTT diagrams of plain carbon steels show us
clearly that the production of martensite requires very rapid cooling rates. These cooling
rates are so fast that it is not possible to transform large parts into martensite. Thus, the
excellent properties provided by tempered martensite cannot be obtained in thick
components made of carbon steel simply because they cannot be cooled fast enough. The
reason is that plain carbon steels have low hardenability. We define the hardenability of
a steel as that property that enables it to avoid transformations that occur by diffusion
and, instead, to transform to martensite when cooled from the austenite phase field. A
steel with a low hardenability must be cooled very rapidly. As the hardenability
increases, the necessary cooling rate decreases. Note that the hardenability refers to the
ability of the steel to transform to martensite; it does not refer to the magnitude of the
hardness of the martensite.
Because of the very useful and desirable properties that can be achieved with tempered
martensitic microstructures, research into methods of increasing the hardenability of
carbon steels became an important endeavor many years ago. It was found that profound
effects can be obtained by adding alloying elements to steel. This is the first reason for
the existence of low alloy steels. Often the alloying elements that are added to control the
hardenability of the steel do not markedly improve the mechanical properties obtainable
in the tempered martensite in comparison with those of similarly tempered martensite in
plain carbon steel; they do, however, increase the hardenability so that the martensite can
be produced in the first place. Before we consider the roles of alloying elements in steel,
we will complete this description of heat treating processes by summarizing some heat
treating practices.
Physical Metallurgy of Steel
Class Notes and lecture material for MSE 661 (formerly 651.01)
Notes revised by: Robert H. Wagoner, Smith Chair
Notes originally compiled by: Glyn Meyrick, Professor Emeritus
Web installation by: Wei Gan, Graduate Research Associate
Last revision date: 11/21/06
SECTION 5 – ALLOYING ELEMENTS AND
SURFACE HARDENING
Alloy steels are those steels that contain minimum specified amounts of alloying
elements in addition to those found in plain carbon steels. The list of elements that are
important in alloy steels, in addition to those that we have already considered in plain
carbon steels, includes Cr, Ni, Mo, Nb, Ti, W, Co, Cu, V, Si, Al and B. These elements
produce a variety of changes in the behavior and properties of steel. Because the
equilibrium condition of the steel is now described by phase diagrams applicable to the
Fe-C-x (etc.) systems, the alloying elements have important effects on the details of
phase transformations in steel. These effects concern both the kinetics of the
transformations and, because additional phases not present in plain carbon steels can be
created, the type of transformation. Alloying elements also have pronounced effects on
phenomena such as recovery, recrystallization and grain growth.
AISI Designations for Alloy Steels
The AISI/SAE standard low alloy steels are listed in ASM handbooks and in numerous
texts and journals. These steels are identified according to a numerical code in which the
final pair (trio if there are five digits) of digits describes the carbon content as is so for
plain carbon steels and the first two numbers indicate the principal alloying elements .
The standard steels are manganese steels with code numbers 13xx; molybdenum steels
described by 40xx numbers; molybdenum and chromium steels which use 41xx numbers;
molybdenum, chromium and nickel steels assigned 43xx numbers and chromium steels
which use 51xx numbers. Other variations of compositions have 6xxx and 7xxx etc. The
series 2xxx and 3xxx refer to steels which contain nickel as the principal alloying
element. Some of the other combinations of alloying elements, however, are more
effective at raising the hardenability than is nickel alone and provide a given increase at a
lower cost. Hence, nickel steels are not used as much as they used to be. Indeed, these
days they are often omitted from some tables of standard low alloy steels. When the
hardenability is of prime importance for the application of an alloy steel, it is possible to
obtain steels which are manufactured with closely controlled compositions so as to ensure
R. H. Wagoner MSE 661 – Steel Section 5, Page 2
that a specified hardenability is obtained. Such steels are identified by adding the letter H
to the numerical code.
Other elements that are added to standard alloy steels are V, B and Pb. Vanadium is a
very strong carbide former and forms fine carbides that along with the aluminum
compounds in killed steels, inhibit grain growth during austenitization. This is effective
because its carbides can persist at temperatures close to 2000°F which is much higher
than typical austenitization temperatures used for standard low alloy steels. When
present, lead resides in steel in the form of small globules of lead; its solubility in ferrite
is very small. It is added because it improves the machinability of the steel. Boron is
added to increase the hardenability. This is the only reason that it is added. When boron is
present in a standard alloy steel, the code designation is amended by inserting the letter B
between the second and third digit e. g. 50B44.
Table 5-1 summarizes the various classes of alloy steels.
Classification of Alloying Elements
The purpose of this section is to attempt to divide the long list of alloying elements into
groups that have similar effects. The properties of carbon steels depend upon the
properties of the individual phases that comprise the steel, the volume fractions of these
phases and the manner in which they are arranged within the total microstructure. Two
very important phases that can exist in steel are the FCC and BCC solid solutions that are
called austenite and ferrite. The first method of classifying alloying elements is based on
the effect of the alloying element on the austenite and ferrite phase fields on binary phase
diagrams of iron and the alloying element. This classification divides the alloying
elements into two main groups called austenite stabilizers and ferrite stabilizers.
Austenite Stabilizers - If we examine the metastable iron carbon binary phase diagram
we see that adding carbon to iron favors the existence of austenite over ferrite. This is so
because it decreases the temperature of the austenite to alpha ferrite transition ( the A
3
point ) and also increases the temperature of the austenite to delta ferrite transition. Thus
carbon expands the gamma field and is therefore called an austenite stabilizer. Those
elements that behave similarly are also called austenite stabilizers. When the binary phase
diagrams with iron of these elements are examined, Figure 5-1, it is found that they can
be sub divided into two groups. In one group the addition of the alloying element
depresses the FCC to BCC transition temperature towards room temperature before other
phases appear. Examples of such elements are nickel and manganese. In contrast, in the
iron carbon system the austenite phase field is limited by the appearance of cementite.
(Figure 5-2 shows the effect of third elements on the Fe-C diagram.) Nitrogen and
copper act similarly in that they expand the gamma field. Nitrogen forms a nitride phase
that limits the austenite phase field whereas the phases in the iron copper system are solid
solutions.
Other austenite stabilizing elements are Pt, Ir, Os, and Ru.
R. H. Wagoner MSE 661 – Steel Section 5, Page 3
Table 5-1: AISI-SAE Designations for Common Carbon and Low Alloy Steels. (N.E.
Dowling, “Mechanical Behavior of Materials,” 2
nd
edition, Prentice Hall, 1999)
Designation
1
Alloy Content, % Designation Alloy Content, %
Carbon Steels Nickel-moly
steels
10XX Plain carbon 46XX Ni 0.85 or 1.82;
Mo 0.25
11XX Resulfurized 48XX Ni 3.50; Mo 0.25
12XX Resulfurized and
rephosphorized
15XX Mn 1.00 to 1.65 Chromium steels
50XX(X) Cr 0.27 to 0.65
Manganese steels 51XX(X) Cr 0.80 to 1.05
13XX Mn 1.75 52XXX Cr 1.45
Molybdenum steels
40XX Mo 0.25 Chromium-
vanadium steels
44XX Mo 0.40 or 0.52 61XX Cr 0.6 to 0.95; V
0.15
Chromium-molybdenum
steels
41XX Cr 0.50 to 0.95; Silicon-
manganese steels
Mo 0.12 to 0.30 92XX Si 1.40 or 2.00;
Mn 0.70 to 0.87;
Nickel-chromium-
molydenum steels
Cr 0 or 0.70
43XX Ni 1.82; Cr 0.50
or 0.80; Mo 0.25
47XX Ni 1.45; Cr 0.45; Boron steels
2
Mo 0.20 or 0.35 YYBXX B 0.0005 to 0.003
81XX Ni 0.30; Cr 0.40; Mo
0.12
86XX Ni 0.55; Cr 0.50; Mo
0.20
87XX Ni 0.55; Cr 0.50; Mo
0.25
94XX Ni 0.45; Cr 0.40; Mo
0.12
Notes:
1
Replace “XX” or “XXX” with carbon content in hundredths of a percent, such as AISI 1045
having 0.45% C, or 52100 having 1.00% C.
2
Replace “YY” with any two digits from earlier in table to
indicate the additional alloy content.
R. H. Wagoner MSE 661 – Steel Section 5, Page 4
Ferrite Stabilizers: The second main group consists of those elements that favor ferrite
over austenite. Again, these can be divided into two sub-groups. In one, the FCC field is
restricted and separated from the BCC field, in which alpha and delta ferrite fields merge
together, by a two phase area. The FCC area is called a gamma loop . Examples are Cr,
Ti, V, Mo, Si and Al.
The second sub-group restricts the gamma field by lowering the delta ferrite transition
temperature and raising that of alpha ferrite while also producing other phases. Examples
are B, Nb, Zr and Ta.
When the quantities of alloying elements present in a steel are small of the order of a few
weight percent or less, it is usually possible to consider the equilibrium states of the alloy
steel by means of the Fe-C binary phase diagram while, of course, recognizing the
diagram is an approximation. This becomes unreliable as the alloy content increases.
Then ternary diagrams or pseudo-binary diagrams are employed for Fe-C-X systems.
These difficulties increase as the number of different alloying elements added to the steel
increases. Therefore, parameters that are important in heat treating processes such as
austenitization temperatures and the eutectoid composition etc. are determined directly by
experiments.
Alloying elements can also cause additional phases to appear within the steels. These
phases include alloy carbides and intermetallic compounds. Because cementite is a very
important phase in carbon steels it follows that effects of alloying elements on cementite
are also important. With this in mind, it is customary, also, to divide the alloying
elements into groups according to their propensity to form carbides.
Carbide Formers: These elements either do not object to the presence of cementite, e.g.
Mn dissolves in cementite, or they form carbides themselves, e.g. Ti. They are ranked
according to the stability of their carbides in the iron matrix.
Moderate carbide formers....Cr, Mo
Strong carbide formers...W, Ta, Nb
Very strong carbide formers... V, Ti, Zr
When significant amounts of these elements are present, the situation with respect to
microstructural constituents can become complicated. Steels can possess cementite,
mixed carbides i. e. cementite containing dissolved alloying elements, and a variety of
alloy carbides in which there is little iron. Nitrogen tends to prefer to reside in the
carbides giving rise to carbonitride compounds. It can also combine with some alloying
elements such as aluminum to form nitride precipitates.
Graphite Stabilizers: Some alloying elements dissolve in ferrite and thermodynamically
oppose the formation of cementite. This encourages the formation of graphite and so
these elements are called graphite stabilizers. Examples are Ni, and Si. In fact, silicon is
so effective that it enables the formation of graphite from the melt in cast irons and is,
therefore, a particularly important alloying element in those materials.
R. H. Wagoner MSE 661 – Steel Section 5, Page 5
Figure 5-1: Effect of various alloying elements on phase stability in iron alloys.
MP
A
4
δδδ δ
δδδ δ + M
Melt
γγγ γ + M
δδδ δ +++ + γγγ γ
A
3
ααα α
γγγ γ
ααα α +++ + γγγ γ
Ni, Mn
γ γ γ γ

-

s
t
a
b
i
l
i
z
e
r
s
MP
δδδ δ
A
4
A
3
γγγ γ + M
γγγ γ
γγγ γ +++ + ααα α
ααα α
ααα α + C
γγγ γ + C
M + C
Melt
C, N, Cu
MP
A
4
A
3
γγγ γ
δδδ δ
Melt
δδ δ δ + M
γγγ γ +++ + ααα α(( ( (δδδ δ))) )
ααα α
Si, Cr, Ti, V, Mo, Al
αα αα

-

s
t
a
b
i
l
i
z
e
r
MP
A
4
A
3
ααα α + C
γγγ γ + C
M + C
Melt
δδδ δ + M
γγγ γ + M
Nb, Zr, Ta
δδδ δ
γγγ γ
ααα α
ααα α +++ + γγγ γ
δδδ δ +++ + γγγ γ
R. H. Wagoner MSE 661 – Steel Section 5, Page 6
Figure 5-2: Effect of third alloying elements on the iron-carbon phase diagram.
2800
2600
2400
2300
2000
1800
1600
1400
1200
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Carbon, %
T
e
m
p
e
r
a
t
u
r
e
,

F
T
e
m
p
e
r
a
t
u
r
e
,

C
Carbon steel, 0% Mo
4% Mo
2% Mo
7% Mo
1500
1400
1300
1200
1100
1000
900
800
700
600
2800
2600
2400
2300
2000
1800
1600
1400
1200
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Carbon, %
T
e
m
p
e
r
a
t
u
r
e
,

F
T
e
m
p
e
r
a
t
u
r
e
,

C
Carbon steel, 0% Ti
1500
1400
1300
1200
1100
1000
900
800
700
600
0.3%
Ti
0.5%
Ti
0.6%
Ti
0.7%
Ti
0.8%
Ti
1%
R. H. Wagoner MSE 661 – Steel Section 5, Page 7
Figure 5-2 (continued): Effect of third alloying elements on the iron-carbon phase
diagram.
2600
2400
2300
2000
1800
1600
1400
1200
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Carbon, %
T
e
m
p
e
r
a
t
u
r
e
,

F
T
e
m
p
e
r
a
t
u
r
e
,

C
Carbon steel
0.35% Mn
1500
1400
1300
1200
1100
1000
900
800
700
600
2.5%
9%
4%
6.5%
0.35% Mn
2.5% Mn
4% Mn
6.5% Mn 9% Mn
2600
2400
2300
2000
1800
1600
1400
1200
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Carbon, %
T
e
m
p
e
r
a
t
u
r
e
,

F
T
e
m
p
e
r
a
t
u
r
e
,

C
Carbon steel
0% Cr
1500
1400
1300
1200
1100
1000
900
800
700
600
19%
Cr
15%
Cr
12%
Cr
5%
Cr
R. H. Wagoner MSE 661 – Steel Section 5, Page 8
Effects of Alloying Elements on Hardenability
There are many applications for which the properties of tempered martensite are
particularly suitable and, indeed, often necessary for the application to be successful.
Unfortunately, these desirable properties can only be obtained in components that can
first be converted to martensite. This imposes severe restrictions on the use of plain
carbon steels because of the generally low hardenability of these steels. When it is
desired that martensite forms throughout the part (this is usually called through
hardening), rapid quenching throughout is necessary which restricts the sizes of parts that
can be successfully treated. Very high quench rates bring about the risk of distortion and
cracking and this can pose serious practical problems.
Carbon steels are generally of low hardenability because the diffusion controlled
reactions by which austenite can transform can occur rapidly in them. Hardenability can
be increased if the rates at which these transformations take place can be reduced.
Research has shown that this can be achieved by adding appropriate alloying elements.
With the exception of cobalt, which has a negative effect, all alloying elements that are
added to steel, raise the hardenability provided that they are dissolved in the austenite.
This is an important condition. The effects alloying elements have on the hardenability
of steels are the primary reason for the development of the AISI/SAE standard low alloy
steels
The elements that are present in standard low alloy steels primarily for increasing the
hardenability are:- Cr, Mo, Ni, Mn and B. Mo and Cr are also carbide formers. Thus
they can retard the rates of tempering reactions and cause alloy carbides to form during
tempering of martensite. Another important role of Mn in general is to scavenge sulfur
by forming MnS so as to reduce the likelihood of iron sulfide forming at grain
boundaries. This is particularly important when the steel does not have a very low sulfur
content as was generally so before the advent of modern steel-making technology.
Although the presence of manganese sulfides in steel can adversely effect the toughness
of the steel, its effects are not usually as serious as those due to the presence of iron
sulfides on grain boundaries. Boron is added only for hardenability control. The standard
low alloy steels are usually completely austenitized during hardening heat treatments
except when a very strong carbide former, such as V, has been added so that its carbides
can inhibit grain growth and thereby refine the austenitic grain size.
The relative ranking of several common alloying elements in terms of hardenability is
shown in Figure 5-3, and the following equation is often used to relate Di (Grossman
ideal critical diameter) to alloy content (in weight percentages):
D
i
= D
iC
x 2.21(%Mn) x 1.40(%Si) x 2.13(%Cr) x 3.275(%Mo) x 1.47(%Ni) (5-1)
where D
iC
= ideal critical diameter for given grain size and carbon content (no other
alloying. This “base hardenability” is shown in Figure 2-16. Equation 5-1 is reproduced
from Honeycombe, 2
nd
edition, p. 161.
R. H. Wagoner MSE 661 – Steel Section 5, Page 9
In order to consider the effects of alloying elements on the hardenability of steels it is
necessary to consider the effects that they have on phase transformations in steel. The
first of these concerns the effects on austenitization; then there are the effects upon the
formation of proeutectoid ferrite, pearlite and bainite; then lastly, of course, are the
effects on the martensitic transformation itself. Let’s look at these effects in sequence.
Austenitization: Let us now define the "eutectoid temperature" as the lowest
temperature at which austenite can be formed. For ternary Fe-C-X alloys, in which the
alloying element is X, both the eutectoid temperature and the eutectoid composition can
vary from the values that apply to the binary Fe-C system. For small quantities of the
element referred to as X, the changes are usually small, except for elements (like
vanadium) that are very strong carbide formers and very strong ferrite stabilizers. Some
examples, shown in the appended sketches, demonstrate the modifications to the
austenite phase field caused by selected alloy additions.
Figure 5-3: Effect of common alloying elements on hardenability. Honeycombe, 2
nd
ed.,
p. 163.
There are three main consequences of these effects upon austenitization.
• When the steel contains alloying elements that are both ferrite stabilizers and
moderate to strong carbide formers, the austenitizing temperatures may have to be
higher than those used for treating plain carbon steels of similar carbon contents.
• If the starting microstructure contains alloy or mixed carbides, longer austenitizing
times can be required to produce homogeneous austenite Consider the dissolution of
1
1.5
2
2.5
3
3.5
4
4.5
5
0 0.4 0.8 1.2 1.6 2
Alloy Content (%)
H
a
r
d
e
n
a
b
i
l
i
t
y

M
u
l
t
i
p
l
y
i
n
g

F
a
c
t
o
r
Mo Mn
Cr
Ni
Si
R. H. Wagoner MSE 661 – Steel Section 5, Page 10
cementite. When the FCC phase forms in contact with the cementite, carbon can
diffuse into it and the carbide begins to dissolve. Continued dissolution is paced by
the rate at which carbon can diffuse away from the carbide/austenite interface. Then,
when the carbide has just dissolved, there will be a local, carbon-rich volume of
austenite where the carbide had existed. Continued diffusion of the carbon in the
austenite is necessary to obtain homogeneity. The rate of diffusion of the interstitial
solute, carbon, is much greater than those of substitutional elements. Carbon diffuses
rapidly even at temperatures just above the Fe-C eutectoid temperature of 727°C
(1333°F) so homogeneity is obtained in short times.
In contrast to the case of the dissolution of cementite, the dissolution of an alloy
carbide, e.g. a molybdenum or tungsten carbide, both the dissolution process and
attaining solute inhomogeneity will take much longer because tungsten and
molybdenum diffuse much more slowly than does carbon. Indeed, this can directly
affect the quantity of austenite in the steel. As an example of this effect, let us
suppose that the steel contains a few wt. % of tungsten in the form of tungsten
carbides. As the carbides dissolve, the adjacent austenite becomes enriched with
tungsten while much of the carbon diffuses away. Because tungsten is a ferrite
stabilizer and restricts the range of austenite stability it is possible for ferrite to form
again in the tungsten rich volume. Thus complete austenitization is paced by the
diffusion of the tungsten. This can be important in steels with high tungsten and
molybdenum contents because such steels must be austenitized at high temperatures.
Examples occur in tool steels.
• When very strong carbide formers are present, a similar story would be told if
homogeneous austenite were required. However, the alloying elements (Nb, V, Ti)
which form carbides that are stable at temperatures well above the normal A
1
line of
a carbon steel, are not added to low alloy steels with the intent of increasing the
hardenability. They are added because of the effects due to their carbides. Indeed, if a
steel that contains fine vanadium carbides is austenitized at temperatures at which the
carbides do not dissolve, the effect of the vanadium is to reduce the hardenability.
This is due to two reasons. First of all, carbon increases the hardenability when it is
dissolved in austenite . Thus, if some of the carbon is not in solution because it is
combined with the vanadium, then a decrease in the hardenability is produced by the
addition of the vanadium. Second, during the austenitizing process the migration of
the austenite grain boundaries is resisted by the carbides. This produces a finer grain
size than for the case of a homogeneous austenite formed in similar austenitizing
treatments. Thus, when the steel is cooled again, the rate of heterogeneous nucleation
on grain boundaries is enhanced which shortens the time required to initiate the
decomposition of austenite and the hardenability is reduced accordingly. This
example demonstrates that if one is considering the effects of alloying elements on
the hardenability of a steel, it is imperative to consider how the element is
incorporated in the steel. If it is not dissolved within the austenite then it cannot
directly raise the hardenability, but it might, indirectly, lower it.
R. H. Wagoner MSE 661 – Steel Section 5, Page 11
Proeutectoid Ferrite Formation: The SAE/AISI standard low alloy steels are
hypoeutectoid steels. Even though many are intended for use at high strength levels in the
tempered martensitic condition, high carbon contents are usually unnecessary because the
increase in the martensitic hardness produced by carbon occurs most rapidly in the
hypoeutectoid range. Therefore, proeutectoid ferrite is the first phase to form when
austenite is transformed at temperatures above the "nose" on the TTT diagram. The
kinetics of the transformation of austenite to proeutectoid ferrite is retarded, in general,
by the presence of both austenite stabilizers and ferrite stabilizers. This is probably due to
effects such as solute drag on migrating interphase boundaries and the rate of diffusion of
carbon.
An example of the effect of Mo on the growth of proeutectoid ferrite is provided by
results that have been obtained by thermo-ionic microscopy. In this technique, the growth
of ferrite is observable in-situ within the microscope while the sample is undergoing the
transformation. It is possible to photograph the process. Observations on the rates of
thickening of lens shaped grain boundary allotriomorphs of ferrite growing in Fe-0.11 wt.
% C and Fe-0.11 wt. % C-1 wt.% Mo have shown that the half-thickness of the ferrite
lens increases linearly as a function of the square root of time (Kinsman and Aaronson,
transformation and hardenability in steels, Climax Molybdenum, p. 39, 1967). This is
consistent with a process that is controlled by long range diffusion of carbon and, indeed
is quantitatively so for the iron carbon alloy. The slopes of the lines decrease when Mo is
added, showing that the growth rate is slower. Clearly Mo retards the transformation and
the fact that there is still a linear dependence on the square root of time suggest that the
Mo interacts with carbon and reduces the rate at which it can diffuse away into the
austenite. Another possible contribution could be segregation of the molybdenum to the
ferrite/austenite interfaces which could reduce the migration rates of those interfaces.
The beneficial effects due to boron occur when the boron is in solid solution in the
austenite. The solubility is very low. Consequently, numerical values describing it have
been difficult to determine with good precision. It has been estimated to lie between
0.001 and 0.005 wt. %. Boron is chemically reactive with other alloying elements in
steel, particularly carbon and nitrogen. When compounds are formed with other alloying
elements the amount dissolved can be reduced. The usual practice is to add boron within
the range of 0.0005 to 0.003 wt. %. Also, the steels are fully killed with aluminum. In
addition, small quantities of titanium and zirconium, which react strongly with nitrogen,
are added to help to scavenge the nitrogen and to keep more of the boron in solution. One
might question this practice and ask why one shouldn’t just add more boron.
Unfortunately, this is not a solution. If more than 0.003 wt. % boron is added, the
hardenability begins to decrease again and, at about 0.004 wt. %, iron boride precipitates
on the austenite grain boundaries and causes a reduction in the final toughness.
Boron increases the hardenability primarily by delaying the start of the transformations to
ferrite and pearlite. It does not seem to affect the time to complete these transformations.
This has led to explanations of the boron effect in terms of the interaction of boron with
austenite grain boundaries. It has been suggested that boron segregates at the grain
boundaries because it is able to reduce the grain boundary energy and that this reduction
R. H. Wagoner MSE 661 – Steel Section 5, Page 12
in energy makes the boundaries less effective as heterogeneous nucleation sites. An
alternative explanation suggests that borocarbides, {Fe
23
(C,B)}
6
, precipitate at the grain
boundaries and inhibit the nucleation of ferrite and cementite thereon. Perhaps both
phenomena participate. It is generally accepted that the austenite grain boundaries are
affected in some manner because the transformations begin at the boundaries and the
effect of boron is greater in fine grained austenite than it is in coarse grained austenite.
The increase in the hardenability due to additions of boron is greatest for steels of low
carbon contents. There is no effect at the eutectoid composition; the hardenability of
hypereutectoid steels is decreased by boron additions. A valuable characteristic of boron
is that it increases the hardenability without lowering the M
s
temperature. This is
valuable because lowering the M
s
temperature causes martensite to form at lower
temperatures which increases the likelihood of quench cracking occurring and can
increase the amount of retained austenite.
Pearlite formation: When pearlite forms during the transformation of austenite in
standard low alloy steels, the carbide in the pearlite is essentially cementite. The
cementite can contain some dissolved carbide forming alloying elements which replace
iron but it forms preferentially to the alloy element carbides, e. g. chromium carbide,
even though the formation of the latter would provide a greater reduction in the free
energy of the system. This is because it can grow much more rapidly due to the much
greater rate of diffusion of carbon compared to the rates of diffusion of substitutionally
dissolved elements. Nevertheless, the rates of nucleation and growth of pearlite in the
alloy steels are smaller than those in plain carbon steels.
When graphite stabilizers, e. g. Ni and Si are dissolved within the austenite it is
energetically favorable for them to remain in solution. This means that they have to
diffuse away from the carbide as it grows and they do so slowly because they are
substitutionally dissolved. One expects both the start and finish lines of the pearlite
transformation to be affected.
Mn partitions to the cementite but it also slows down the reaction. Cr and Mo, when
present alone, interact more strongly with pearlite formation than the elements mentioned
previously. They cause significant shape changes to the start and finish lines on
isothermal transformation diagrams. These changes are probably mainly due to the
reduction in the rate of diffusion of carbon caused by interactions with the dissolved Cr
and Mo, but solute drag may also effect interphase mobility. When Mn, Ni, Mo, and Cr
are all present together in the same steel, their combined effects are dramatic. In some
steels, the effects on proeutectoid ferrite and pearlite formation are so strong that the start
and finish curves for these transformations are separated from those of the bainite
transformation by a region in which the austenite remains in metastable equilibrium for
long times. In these cases the TTT diagrams show the two types of transformation
separated by a "bay". This confirms earlier statements herein that the nose in plain
carbon steel TTT diagrams happens to be where transformations taking place by different
mechanisms merge. The effects of several elements are sketched in Figure 5-4.
R. H. Wagoner MSE 661 – Steel Section 5, Page 13
Figure 5-4: Schematic effect of various alloying elements on TTT diagrams.
A.I.S.I 4340 m
o
=0.33 %, Cr = 0.80 %
N i = 1.79 % Mn = 0.78 %
M
s

o
C = 539-423 (%C) -
30.4 (%Mn)- 17.7(%Ni)
- 12.1 (% Cr) - 7.5(%Mo)
Example of empirical relation
T
Logt
10 secs
10
4 secs 1 week
“BAY”
Bs
Bf
ααα α+Fe
3
C
ααα α+Fe
3
C mixed
pearlite
γγγ γ
R
3
R
1
M
s
T
Logt
plain C Steel Hypoeutectiod.
e.g. 1040
T
Logt
Nickel
added.
Curves
moved.
Shapes
not
affected
much.
T
Logt
Mn added
Curves --->
pearlite finish
curve changed
T
Logt
Cr added
Curves-->
Both pearlite
startand finish
curves
changed
Changes occur
to R3,R1 and Ms.
Roted in the text
but not considered
in these sketches
R. H. Wagoner MSE 661 – Steel Section 5, Page 14
An interesting example of thermo-mechanical processing can be carried out with steels
that possess bays in their TTT diagrams. They can be austenitized and cooled to a
temperature within the bay and then deformed by rolling or forging while still austenitic.
This increases the dislocation density present in the austenite prior to the formation of
martensite or bainite when the steel is finally cooled. The combination of the effects due
to the additional dislocation line introduced during the plastic deformation with those due
to the formation of martensite can produce significant improvements in mechanical
properties. This type thermo-mechanical treatment is referred to as ausforming. Steels
that have useful properties when ausformed are often called ausforming steels.
Bainite Formation: In general, the kinetics of the reactions that form bainite are
retarded by alloying elements. Presumably, the effects arise for reasons similar to those
mentioned for the transformations that occur at higher temperatures. The bainitic
transformation in steels that have a bay in the TTT diagram, however, takes place more
rapidly than the transformations that occur at higher temperatures. This suggests that the
larger driving force associated with the greater degree of undercooling at which the
bainite forms negates the retarding effects somewhat, because, in these steels, bainite can
form more rapidly than pearlite.
Examples of the effects of alloying elements on TTT diagrams are shown in the
accompanying sketches. Appropriate changes are also found in continuous cooling
diagrams.
Martensite Formation: In a few instances, it is possible for alloying elements to cause
the martensitic transformation to take place isothermally. In such a transformation, no
martensite forms while the austenite is cooled to the temperature chosen, but it begins to
form after an incubation time while held at that temperature. In this case, the kinetics of
the transformation can be described by “C” shaped curves on a temperature-time plot.
An example of an isothermal martensitic transformation has been observed in an Fe-0.6
wt. % C -6 wt. % Mn alloy, in which austenite can be retained after quenching and then,
while the temperature is held constant, martensite begins to form. Each of the martensitic
crystals grows rapidly; it is the initiation process that is time dependent. Isothermal
martensitic transformations are rare and do not occur in standard low alloy steels, where
the transformation is athermal. Therefore the M
s
temperature remains an important
parameter.
Because austenite stabilizing alloying elements expand the realm of austenite and, in
particular, reduce the minimum temperature at which the austenite is stable, they are
expected to lower the start and finish temperatures for the martensitic transformation in
order to attain the necessary degree of undercooling. This is so in practice and C, Ni and
Mn have strong effects. The carbide formers, Mo and Cr, also lower the M
s
temperatures even though they are ferrite stabilizers and would be expected to enhance
the driving force for the transformation. Presumably, these alloying elements cause
changes make the transformation more difficult. It is not obvious what these changes are.
It is possible that the initiation sites are affected. Solid solution strengthening of the
R. H. Wagoner MSE 661 – Steel Section 5, Page 15
austenite and changes to its shear modulus which might resist the necessary shape change
may contribute. Nevertheless, all five elements lower the martensitic start temperature.
Extensive experimental data has been accumulated for particular families of alloy steels
in order to assess the effects of composition on the start of the transformation. Using
such data, a number of empirical equations have been published which relate the M
s
temperature to the percentages of C, Mo, Cr, Mn, and Ni in the steel by means of
multiplication factors. Several different ones have been published. Such empirical
equations can be of practical use in material selection but they provide little help towards
elucidating the fundamental reasons for the particular effects.
While we are on the subject of the martensitic start temperature, we should now
recognize that steels with significant concentrations of alloying elements that depress the
temperatures at which martensite forms, can retain large amounts of austenite when
quenched to room temperature. If more martensite is desired, the steel must be cooled
below room temperature. If this is not done before the tempering treatment then the
decomposition of the retained austenite which will occur during tempering, can have
significant effects on the final microstructure and properties.
Deformation-Induced Martensitic Transformation: This is a suitable time to
introduce the M
d
temperature. It has been observed that the M
s
temperature can be
raised by the application of stress to the steel sample. This is because a crystal that forms
during a martensitic transformation has a different shape than the volume of steel from
which it formed. Therefore, if there is help available to achieve this shape change, the
transformation will be accomplished more easily. If it can occur more easily, then it will
require less thermodynamic driving force for it to occur which means that the necessary
degree of undercooling below the A
3
line will be smaller. Hence, we deduce that if an
applied stress is locally oriented so that it helps the transformation to produce the shape
change it will allow martensite to form at a higher temperature than that which pertains to
the unstressed condition.
In general, tensile stresses are found to be more effective than compressive stresses, but
rolling can have appreciable effects. The highest temperature to which the M
s
temperature can be raised by applied stresses is defined as the M
d
temperature. When
this temperature lies above room temperature and the M
s
is below room temperature, it is
possible to retain the austenite at room temperature and then to form some martensite by
working the metastable austenite at room temperature. This can be important in highly
alloyed steels such as stainless steels.
Figure 5-5 illustrates the dramatic effect of deformation-induced martensite on the stress-
strain curve in type 301 stainless steel, and also illustrates that higher temperatures (in
this case, from deformation heating) inhibit the production of martensite. Note the
extremely large tensile strains that can be attained (up to 80% engineering strain) and the
high volume fraction of martensite (up to 80%). How is this combination possible in
view of the strength-ductility tradeoffs discussed extensively in Section 4?
R. H. Wagoner MSE 661 – Steel Section 5, Page 16
Figure 5-5: Engineering stress-strain curves for several stainless steels, with
corresponding deformation temperatures and martensite fractions. (V. Talyan, R. H.
Wagoner, J. K. Lee: Metall. Mater. Trans. 1998, A, vol. 29A, pp. 2161-2171).
1000
800
600
400
200
0
0 0.2 0.4 0.6 0.8
0
30
60
90
120
150
εεε ε === = 10
-3
/s
301
201
304L
301
201
304L
Engineering strain
E
n
g
i
n
e
e
r
i
n
g

s
t
r
e
s
s

(
M
P
a
)
%

M
a
r
t
e
n
s
i
t
e
Type 301
Temperature
10
-1
/s
10
-2
/s
10
-3
/s
10
-3
/s
10
-2
/s
% Martensite
10
-1
/s
Engineering strain
%

M
a
r
t
e
n
s
i
t
e
T
e
m
p
e
r
a
t
u
r
e

(
C
)
60
40
20
0
-20
-40
-60
0 0.2 0.4 0.6
0
20
40
60
80
100
120
% Martensite
Temperature
R. H. Wagoner MSE 661 – Steel Section 5, Page 17
The Effects of Alloying Elements on Tempering Reactions
The reactions that take place when martensite is tempered involve solute segregation and
clustering at crystal defects, precipitation from solid solution, particle coarsening and
grain growth. In general, alloying elements dissolved in the martensite reduce the rates at
which these processes occur. Also, alloying can decrease the M
s
temperature and give
rise to significant quantities of retained austenite when the quenchant is at or near room
temperature. Accordingly, the decomposition of this austenite during tempering can have
important effects and therefore stage two tempering processes can become important,
unlike the case for hypoeutectoid plain carbon steels.
As quenched, the steel will have a smaller hardness than would be predicted from the
carbon concentration of the steel because of the crystals of retained austenite distributed
within the martensite. Then, when the austenite transforms to bainite during tempering in
the appropriate temperature range, the net hardness can increase. The austenite
stabilizers, Mn and Ni, are effective elements for lowering the martensitic start
temperature. In the following discussion of the effects of alloying elements upon
tempering reactions in martensite we will neglect effects due to retained austenite.
Obviously, this is being done so that we can focus on martensitic reactions.
The various effects are shown schematically in Figure 5-6 and selected systems are
illustrated with real data in Figure 5-7.
Graphite formers and Mn: Si and Ni partition to ferrite and avoid being incorporated
in cementite. This means that the rate of growth of the carbide particles depends on the
rate at which these elements can diffuse out of the way. They do so by substitutional
diffusion with the result that the growth rates of the cementite particles are retarded in
comparison to the case of rate control by the diffusion of carbon.
Mn can be incorporated into the carbides and does not need to diffuse away. However, it
still retards the growth rate, possibly by reducing the diffusion rate of carbon. At high
tempering temperatures where the growth of ferrite crystals becomes important, the
solutes can impede the migration rate of the ferrite/ferrite interface by solute drag. Solute
drag refers to an effect that can arise due to the tendency of solutes to segregate at or
away from intercrystalline boundaries. Segregation lowers the effective surface energy
and hence the driving force for boundary migration.
Carbide Formers: These elements retard the rates of the transformations that occur
during tempering. Equivalent effects require smaller amounts of solute the greater the
affinity of the solute for carbon. For example, 1 wt. % Cr, 0.5 wt. % Mo and 0.15 wt. %
V have similar effects. At these alloying levels, tempering times of 0.5 to 1 hour produce
carbides that consist primarily of cementite due to the comparatively high carbon content
and also because the rate of diffusion of the carbon is so much faster than those of the
alloy elements. The retardation in the carbide growth rate suggests that the solutes
diminish the rate at which carbon can diffuse in the steel. Alloy carbides that are rich in
alloying elements can form instead of cementite in more concentrated alloy steels at
R. H. Wagoner MSE 661 – Steel Section 5, Page 18
high tempering temperatures. When this happens their growth rates are controlled
directly by the diffusion rates of the (substitutional) solutes.
Figure 5-6: Effect on various alloying elements on hardness following tempering.
S
c
m
n
1200 F
R
c
1200 F
R
c
1200 F
1200 F
R
c
Carbipe formers retard
in small amounts
1% Cr=
0.5%Mo=0.15%V
Larger amounts
cause secondary
hardening
Austenite
stabilizers
and graphite
formers
Retard softening
effect of Si >Mn
>Ni
Strong C.F’s move peak to high T
12% Cr
2
0.5
0
800 600
600
800
R. H. Wagoner MSE 661 – Steel Section 5, Page 19
Figure 5-7: Effect of alloying elements on tempering (a,b — Mo)
Tempering temperature, F
10
20
30
40
50
60
400 600 800 1000 1200
200
300 400 500 600 700
H
a
r
d
n
e
s
s
,

R
o
c
k
w
e
l
l

C
0.35 %C
0.50 %Cr
5% Mo
2% Mo
0.5% Mo
0% Mo
Q
H
a
r
d
n
e
s
s
,

R
o
c
k
w
e
l
l

C
10
20
30
40
50
60
Tempering temperature, C
200 300 400 500 600 700
400 600 800 1000 1200
Q
5% Mo
2% Mo
0.5% Mo
0% Mo
0.35% C
2% Cr
Tempering temperature, F
Tempering temperature, C
R. H. Wagoner MSE 661 – Steel Section 5, Page 20
Figure 5-7: Effect of alloying elements on tempering (c — Mn, d - Cr)
15
20
25
30
35
40
45
50
55
60
Tempering temperature, C
100 200 300 400 500 600 700
Tempering temperature, F
H
a
r
d
n
e
s
s
,

R
o
c
k
w
e
l
l

C
200 400 600 800 1000 1200
Q
0.40 to 0.45% C
SAE T1340
(1.75% Mn)
SAE 1040 (0.75% Mn)
10
20
30
40
50
60
Tempering temperature, C
200 300 400 500 600 700
Tempering temperature, F
H
a
r
d
n
e
s
s
,

R
o
c
k
w
e
l
l

C
12% Cr
4% Cr
2% Cr
0
.
5
%

C
r
Carbon steel
400 600
800
1000 1200
Q
R. H. Wagoner MSE 661 – Steel Section 5, Page 21
Figure 5-7: Effect of alloying elements on tempering (e — Si, f - Ni)
20
25
30
35
40
45
50
55
60
65
Tempering temperature, C
Tempering temperature, F
H
a
r
d
n
e
s
s
,

R
o
c
k
w
e
l
l

C
200 400 600 800 1000 1200
Q
1 Hr.
0.50 to 0.55 %C
3.8% Si
2.3% Si
1.3% Si
0.5% Si
15
20
25
30
35
40
45
50
55
60
Tempering temperature, C
100 200 300 400 500 600 700
Tempering temperature, F
H
a
r
d
n
e
s
s
,

R
o
c
k
w
e
l
l

C
200 400 600 800 1000 1200 Q
0.40 to 0.45% C
SA 2340
(3.5% Ni)
SAE 1040
100 200 300 400 500 600 700
R. H. Wagoner MSE 661 – Steel Section 5, Page 22
Hardness measurements provide a convenient method of following the progress of
tempering reactions because these reactions change the hardness of the steel. The
hardness in hypoeutectoid steels is usually gradually reduced by tempering because the
carbon initially dissolved in the martensite forms precipitates of carbides causing the
martensite to become softer and to lose its tetragonality. Eventually, when the
precipitation of cementite is complete, continued tempering causes the average size of
the carbides to increase in order to reduce the total area of the ferrite/carbide interphase
boundary. This process reduces the number of carbides per unit volume which causes the
spacing between them to increase. It is the spacing between these carbides that
determines how effective they are at hindering the glide of dislocations. Therefore, the
steel softens continuously.
At tempering temperatures of 800°F and above, plain carbon, and many low alloy steels
continue to soften because of the coarsening of the carbide population and, also, because
of ferrite grain growth. However, in alloy steels in which significant amounts of alloy
carbides can form, the new carbides replace the coarsening cementite-rich carbides with
a new, fine array of alloy carbides. Note that the rapidly diffusing carbon enables
relatively pure cementite to form first because the nucleation and growth of the alloy
carbides is paced by substitutional diffusion. When the dense array of fine alloy carbides
forms, the gliding dislocations meet a higher density of obstacles to slip and the hardness
of the steel is increased. This phenomenon is called secondary hardening and is clearly of
practical importance because it provides another method of obtaining a microstructure of
high hardness and toughness. The particular temperatures at which secondary hardening
occurs depends on the species of carbide that forms and increase as the stability of the
carbide increases.
It is important to recognize that this discussion of the effects of alloying elements applies
to situations in which the alloying elements were initially in solution in the austenite.
When this is not so, important changes can arise. Let us consider an example to predict
the kind of changes that can occur. We select an AISI 1050 steel to which 1 wt.% V has
been added and which has a microstructure that contains vanadium carbides. First, let us
austenitize the steel at a temperature that is high enough to transform the steel into
homogeneous austenite. Then suppose that the steel is successfully transformed to
martensite during quenching. Because the vanadium and the carbon are dissolved in the
martensite, the events described earlier take place during tempering.
Now suppose we repeat the experiment but this time the steel is held at a temperature that
is not far above the A
3
point during the austenitization treatment. At this temperature the
ferrite is converted to austenite but the vanadium carbides are not dissolved so the
austenite contains them when it is quenched. The first effect that we predict is a change in
the hardenability of the steel. The hardenability will be decreased in comparison to that of
the previous case because the vanadium is not dissolved in the austenite and, also, there is
less carbon in the austenite because some resides in the undissolved carbides. The
hardenability will also be reduced because the grain size of the austenite will be smaller
than in the previous case due to the lower temperature of austenitization and because the
R. H. Wagoner MSE 661 – Steel Section 5, Page 23
vanadium carbides inhibit grain growth. In fact, the hardenability could now be so low
that we may fail to produce a uniformly martensitic sample. That could certainly cause
problems if a fully martensitic component was a required result of the heat treatment.
However, let us suppose that we are fortunate and that the sample is successfully
transformed into martensite. What then? The hardness of the as-quenched martensite will
be lower than that of the martensite produced in the first experiment because it has less
carbon dissolved in it. Also, it is possible for a reduction in the hardness to arise due to
auto-tempering that could occur because of the increase in the M
s
temperature due to the
absence of Vanadium and the reduction of carbon in the austenite. When such samples
are tempered, the transformations that occur will be very similar to those that would
occur in a plain carbon steel that contains less carbon than a 1050 steel. Secondary
hardening would not take place.
Embrittlement During Tempering
The aim of a quench and temper treatment is to produce a microstructure that has a
particular set of mechanical properties. The microstructural changes that take place in
low alloy steels during tempering usually cause a reduction in the hardness of the steel
unless, of course, secondary hardening occurs. The decrease in hardness is usually
accompanied by an increase in the toughness as shown by the results obtained during
impact tests. However, there are occasions when this is not so and, instead, the toughness
falls. This reduction in the toughness raises the ductile to brittle transition temperature
and can reduce the energy absorbed during fracture at and above room temperature.
There are two types of embrittlement. One is called tempered martensite embrittlement. It
is also referred to as 500°F embrittlement and one step embrittlement. The other is
simply called temper embrittlement or, perhaps preferably, two step embrittlement. The
one step and two step terminology is less confusing.
Let us review some pertinent aspects of fracture in metals. First, from the macroscopic
point of view, a very ductile fracture is obviously associated with extensive plastic
deformation. Most of this plastic deformation precedes the fracture itself, and creates
distortion such as elongation, bending and reduction in thickness. The work done during
the gross deformation accounts for the high values of the energy absorption in the total
failure. In contrast, a very brittle fracture takes place without the production of any
obvious distortion and smaller amounts of energy are absorbed during the failure.
In the preceding paragraph, ductile and brittle failures were described from the
macroscopic point of view. On the microscopic scale the ductile failure mode produces
characteristic fractographic features visible at magnifications above a few hundred times.
These features are called dimples. They are small craters or holes bounded by raised
ligaments and their presence demonstrates that the crack propagated by a process that
involved a ductile failure mechanism. Studies of the ductile fracture process under a
tensile stress, has shown that micro voids nucleate and grow in the metal ahead of the
crack. The crack advances through this array of micro voids by tearing the metal
ligaments that separate neighboring voids, leaving behind the tell-tale dimples on the
fracture surfaces. The nucleation of the micro voids is frequently easier at small included
R. H. Wagoner MSE 661 – Steel Section 5, Page 24
particles in steels; then it is common to observe the particle sitting in the dimple when the
fracture is examined with a large enough resolution. The process of micro void
coalescence is a ductile process and the fracture is therefore a ductile fracture. However,
it is not necessarily true that a fracture that proceeds by micro void coalescence, will be
associated with extensive, macroscopic plastic deformation. It could be initiated at a
reduction in section or some other stress concentrator. In these cases, there that can be
little indication from the macroscopic view, that the failure took place mechanistically by
a ductile mode.
Fractures which occur by cracks that do not propagate by micro void coalescence can
occur in metals. BCC and CPH metals can break by cleavage. The initiation of a crack in
a cleavage fracture may involve some localized plastic flow but the crack is sharp and
advances by material separation at its end. If plastic deformation does occur ahead of the
crack then the crack can become blunt and stop. Here we have an inherently brittle
process on the microscopic and macroscopic scales. Cleavage cracks tend to follow
particular crystallographic planes, typically {100}in BCC crystals. Adjacent parts of the
crack can advance along different, but parallel planes. These adjacent parts then become
connected on the final fracture surface by steps that are called cleavage steps. Fractures
that occur by cleavage and by micro void coalescence usually produce distinctive
characteristic topographic features on the fracture surfaces that enable them to be
distinguished and readily identified. A cleavage failure usually takes place rapidly and
consumes less energy than does a ductile failure.
Because cleavage cracks follow particular crystallographic planes, the fracture paths pass
through grains and the fracture is called transgranular. Intergranular fractures during
which the cracks follow the grain boundaries within the material can also take place.
These fracture surfaces are characteristically smooth and comprise curved and flat facets
that clearly reveal the intercrystalline boundary path followed by the cracks. Again, there
is usually little or no micro void coalescence involved. The energy consumed during the
fracture is small because the crack follows a path across which there is weak cohesion.
This mode of failure is brittle on both microscopic and macroscopic scales and can occur
in FCC metals as well as BCC and CPH metals.
We have described three modes of fracture that can occur when metals are broken. It is
important to realize that a metallic component can fracture by more than a single mode.
Fracture surfaces can show regions of micro void coalescence and regions of cleavage.
As the percentage of the former decreases so does the energy consumed during the failure
and the metal becomes less tough.
The toughness of steels is of great practical importance. It is to be expected that some
components that are made of steel will contain small cracks or similar flaws. The
tolerance of the material to defects decreases as the toughness goes down and
catastrophic failure can occur under “normal loading conditions”. If a material that would
normally be expected to have adequate toughness under certain conditions develops a
microstructure that permits crack propagation in a brittle mode, it is said to have become
embrittled. As we said earlier, this can happen during tempering.
R. H. Wagoner MSE 661 – Steel Section 5, Page 25
One Step Embrittlement: This form of embrittlement can occur in standard low alloy
steels that are heat treated for high strength levels, e.g. quench and tempered 4140. It can
also occur in plain carbon steels that contain phosphorus. A typical treatment that
produces it is austenitization, quenching to martensite and tempering in the region of
450°F to 550°F. See Figure 5-8. (The old name of One Step Embrittlement was “500
Degree Embrittlement.”) The embrittlement manifests itself by raising the ductile to
brittle transition temperature and by reducing the amount of energy absorbed during the
impact in comparison to those pertaining to samples tempered immediately above and
below this range.
This range of tempering temperatures corresponds to the beginning of stage three
tempering in which sheets of cementite needles and platelets precipitate heterogeneously
on interlath boundaries. Sheets of cementite particles can also be deposited from retained
austenite when this austenite transforms during the tempering treatment. In general terms,
invoking the principle that the properties of the whole sample are dominated by those of
the most continuous phase, one might expect a deterioration of toughness with these
microstructural changes. This seems to be true when the steels are relatively free of P and
N because the fracture paths follow the sheets of cementite. Furthermore, the
embrittlement is effected during the tempering time required to produce these
microstructures. These times are typically about an hour. The fracture paths are
transgranular and the failure mode is often a mixture of cleavage and microvoid
coalescence. However, when the steel contains high amounts ( say 0.03 wt. %) of P, the
embrittlement is associated with intergranular failure along prior austenite grain
boundaries. This suggests that the embrittlement is also associated with a diminution of
cohesive strength across prior austenite grain boundaries due to the segregation thereon
of impurities within the steel as well as with precipitates of cementite.
Two Step Embrittlement: This is a loss of toughness because of the occurrence of
intergranular failure along prior austenite grain boundaries. At first sight, it is very
similar to the intergranular mode of failure in one step embrittlement, but it does not
afflict plain carbon steels. It arises in alloy steels containing Ni, Cr and Mo that are
converted to martensite, tempered at 600°C (note the switch to Celsius) and then slowly
cooled past 350°C. The cooling rate must be very slow, as is unfortunately the case in
practice when heat treating very large turbines made from steels containing Ni, Cr, Mo
and V. If the steel can be cooled rapidly from the tempering temperature then it is not
embrittled. Indeed, this form of embrittlement is reversible by reheating to 600°C and
then cooling rapidly. Unfortunately, that cannot be done with very large components.
Even when it is possible to cool fast enough to avoid the problem, the embrittlement
demon lies in wait.
If the susceptible steel is heated into the 375 to 500°C range subsequently and held there,
it will become brittle. Hence the term two step embrittlement. The kinetics of the
embrittlement process exhibits “C” curve behavior during isothermal aging at various
temperatures. The nose of the curve, where the rate of embrittlement is most rapid,
occurs typically at about 550°C. At this temperature significant embrittlement can occur
R. H. Wagoner MSE 661 – Steel Section 5, Page 26
in times of the order of an hour. At 375°C several hundred hours are required. This is
particularly unfortunate because this temperature range is often encountered in practical
applications and, eventually, flaws in the steel can become super critical as the toughness
gradually falls with time.
Figure 5-8: Schematic views of one-step and two-step embrittlement.
A great deal of research has been devoted to understanding two step embrittlement. Much
of this work has employed Auger spectroscopy to analyze the composition of fracture
surfaces immediately after breaking the sample, in situ, within the apparatus. As a result
Cool very
slowly
past
Temper 600 C
Austenitize
Temper
DBTT
Impact temperature
I
m
p
a
c
t

e
n
e
r
g
y
ONE STEP
TWO STEP - version 1
Austenitize
600 C Temper
Reheated TO 350 C
TWO STEP - version 2
Q
u
e
n
c
h
Austenitize
T
e
m
p
e
r
a
t
u
r
e
Time Time
T
e
m
p
e
r
a
t
u
r
e
R. H. Wagoner MSE 661 – Steel Section 5, Page 27
of this work, there is general agreement that the loss of toughness is due to the
segregation of tramp elements in the alloy steels. Elements that are known to cause
embrittlement are P, Sb, Sn and As. In addition, the alloying elements Ni, Mn, Si and Cr
seem to exacerbate the problem. It appears that these elements, particularly Ni, encourage
the segregation. Molybdenum is beneficial. It retards the embrittlement.
Surface Hardening Treatments in Steel Components
Most of our discussion about quench hardening and tempering heat treatments has so far
been concerned with through hardening components. There are many applications of
steel for which through hardening is unnecessary, but for which a high hardness at and
near the exterior surface is important. Obvious examples are those in which wear
resistance and dent resistance is important, e. g. gears and bearing shafts. Other
examples include power transmission shafts subject to bending and torsion actions that
create much higher mechanical stresses at the surface regions than in the interior. In such
instances it is desirable to produce the hard strong microstructures at and near the exterior
surfaces. Such microstructures can be unnecessary in the interiors. Indeed, they are
often undesirable. For example, although hard, wear resistant surfaces can be required
for a shaft it can be also important that the less severely loaded interior be tough and
resistant to crack propagation. These requirements have led to the development of a
variety of commercial practices aimed at surface hardening components.
Flame and Induction Hardening: One way to produce a microstructure at the surface
of a component that is different from the microstructure in the interior, is first to produce
the interior microstructure by a uniform heat treatment applied to the whole part and then
to selectively heat treat the surface only. In these treatments the composition of the steel
is uniform; the heating and cooling is not uniform. The removal of heat from a quenched
bar is partially controlled by the rate at which it is conducted through the steel to the
cooled surface. The relatively low thermal conductivity of steel is an enemy when one is
trying to through harden bulk pieces. It becomes an ally, however, when one only wishes
to austenitize a region near the surface because it enables surface regions to be
austenitized using high rates of energy input without the interiors being significantly
affected.
Probably the oldest technique for surface hardening is flame hardening in which heat is
applied directly to the surface concerned using torches that burn gases such as oxygen-
acetylene mixtures. In passing, it should be noted that steel plate is often cut by means of
flame torches which can alter the microstructure and properties of the steel near the cut
surface. The objective of the flame hardening process is to austenitize the steel at and
near the surface and then to remove the flame and rapidly quench the work to produce
martensite and, finally, to use the flame to temper the martensite. Many variations on the
theme can be envisioned ranging from hand held torches wielded by workers of various
skill levels and experience to automated ignition, burn and quench assemblies, e. g. a
rotating shaft within a surrounding, stationary array of burners.
R. H. Wagoner MSE 661 – Steel Section 5, Page 28
The use of high frequency, oscillating, magnetic fields induced in the surface layers of
steel parts is an important method of surface hardening. This is achieved by using
appropriately shaped induction coils made from hollow, water-cooled copper tubing that
are placed near the surface ( or even selectively near parts of a surface). The oscillating
field produced by the induction coils induces electrical eddy currents in the steel. These
eddy currents are produced within a certain depth of the outer surface called the skin
depth. This depth decreases as the oscillating frequency is increased. The eddy currents
cause Joule resistance heating in the skin depth which rapidly raises the temperature of
the steel at the surface. Then, after the appropriate time, the power is interrupted and the
part is quenched. The process is readily automated. It is more adaptable than flame
heating because a wide variety of coil configurations are possible; for example, the
interior wall surface of cylindrical tubes can be induction hardened. Usually, surface
hardening by localized heat treatment techniques is applied to steel components that have
uniform compositions. The selection of the steel must be made with this in mind so that
the appropriate carbon content for the required hardness is chosen; often medium to high
carbon steels are used.
The use of induction heating is not restricted to producing changes close to the surface.
For example, the mechanical engineering analysis of an axle might specify a particular
minimum tensile yield strength at the 3/4 radius position from the center of the shaft.
This translates into a certain hardness that is to be produced there in the tempered
condition and thereby specifies the minimum carbon content for the steel. The 3/4 radius
position mentioned in this example can be relatively distant from the outer surface e. g.
in a 3 inch diameter axle. In this case, the choice of the steel will also include
hardenability considerations. Induction heating for longer times and in the lower
frequency ranges enables the appropriate austenitization to be achieved. However, in
many cases when the needed depth of hardening is quite shallow, the induction heating
time is completed within a few seconds (lengthy heating cycles allow time for heat
conduction into the workpiece). This means that the austenitization time is very short
which can have significant effects on the microstructural changes.
We note first that coarse microstructures such as massive carbides widely dispersed in
large grained ferrite are not suitable as a starting microstructure because the time required
to produce relatively homogeneous austenite is too long. A high temperature excursion
can compensate by raising the diffusion rates but this increases the risk of obtaining a
large austenitic grain size. Refinement of the initial microstructure such as by
normalization is helpful. An interesting observation that has been made is that the
hardness of martensite manufactured in treatments using rapid heating and short
austenitization times exceeds that obtained in the same steel by conventional through
hardening heat treatments. The increases can be about 5 points on the Rockwell scale.
This has not been completely explained, but is probably related to refinements in the
comparative scales of the microstructures.
In short time induction heating processes, the depth to which the steel is austenitized is
small. Shape changes due to thermal expansion and transformation of structure are
accommodated by plastic flow in the hot metal. Then, when the rim transforms to
R. H. Wagoner MSE 661 – Steel Section 5, Page 29
martensite during cooling it tries to expand but the core on which it sits is relatively cool
and resists plastic flow. The result, often, is that the surface is in compression. This can
endow the part with an increased resistance to fatigue because fatigue usually requires
tensile stresses. During treatments in which the heating depth is greater, the situation can
be reversed and the surface can be put into tension which is not good when fatigue
resistance is important.
Case Hardening: In the methods described above, locally different heating and cooling
rates are used to produce different microstructures in particular locations in the steel
components of uniform compositions. It is possible to produce similar kinds of effects by
means of through heating treatments given to steels in which the composition varies
locally, i. e. from the surface to the interior. Surface hardening refers to all treatments
that are intended to make the surface of articles harder than the material in their interiors.
If this is done then the article is encased ( completely or partially in selected places) in a
hardened layer. Case hardening would appear then, to be another term for surface
hardening. This is true, but I have separated it here because it has been my experience
that when the term case hardened is used, much more often than not it refers to a surface
hardening process that depends on modifying the composition near the surface. In these
processes the case depths are relatively shallow, unlike the case of an induction hardened
automobile axle. The method employs solid state diffusion to enrich the surface with
carbon, nitrogen or both.
Carburizing: Because the factor that has the greatest effect on the hardness of
martensite is its carbon content, increasing the concentration of carbon at the surface of a
component provides a means of endowing the surface with a greater hardness than the
interior. The process by which this is done is called carburizing. It is possible because
austenite can accept up to 2 wt. % carbon in solution. It is done by exposing austenite to
an appropriate environment that maintains the concentration of carbon at the surface of
the steel at a chosen high value while carbon diffuses into the steel. The eventual case
depth will then be a function of the surface concentration, temperature and time. When
the carburization is completed, the steel is quenched to render the case martensitic and is
then tempered. The quenching process can be done at the end of the carburization time
or the part can be cooled down and then re-heat treated for case hardening. We will
consider a gas carburizing process to illustrate some of the principles.
An important reaction is
2CO = CO
2
+ C (5-2)
If we consider a particular mixture of the gases and a piece of austenite at equilibrium at
a given temperature and a total pressure of one atmosphere, then the carbon would be
dissolved in the austenite so that the C in the equation would correspond to the carbon
concentration in the austenite that is in equilibrium with the particular mixture of gases.
If the CO content is increased, the reaction will move to the right increasing the
components on the right hand side, and, in particular the carbon concentration in the
austenite. Similarly, if the austenite was put in oxygen it would be decarburized as CO
R. H. Wagoner MSE 661 – Steel Section 5, Page 30
and CO
2
are created. The variation of carbon concentration in austenite as a function of
the partial pressures of CO and CO
2
is shown on page 404 of Darken and Gurry’s
Physical Chemistry of Metals.
When a piece of low carbon austenite is placed in a gas mixture for which a greater
carbon content in the austenite corresponds to equilibrium, the reaction will proceed to
the right until the required carbon concentration is established at the surface of the steel.
This happens rapidly provided there are adequate quantities of the gas mixture available
and it is continually being mixed. The result is that a chosen carbon concentration can be
established and maintained at the surface. Meanwhile carbon diffuses down the
concentration gradient into the steel. The diffusional flux into the steel is slow enough
under the usual conditions for the chemical reaction at the surface to maintain the
composition there.
Practically, carburization is not carried out using pure CO and CO
2
mixtures. Typical
commercial atmospheres use mixtures of natural gas, which provides methane, and
nitrogen which acts as a diluent. Active gases in the mixtures are CO, CO
2
, H
2
, H
2
O,
O
2
and CH
4
. The appropriate mixtures generate and maintain chosen carbon levels.
Carburization temperatures are usually in the range of 850 to 950 Celsius. Carburization
times depend upon the desired case depth. A common definition of the case depth is that
depth below the surface at which the hardness is RC 50 in the final quenched and
tempered part . Thus, hardnesses a few points higher are required in the as-quenched part
which enables the carbon level at that location to be predicted. Typical case depths are
less than a tenth of an inch and are often fractions of that. I have seen specifications
calling for case depths of 0. 002-0. 006 inches: this is simply a statement that implies
neither approval or disapproval.
The time required for the carburization treatment can be estimated from the appropriate
solution of Fick’s second law.
C
x
− C
s
C
0
− C
s
= erf
x
2 Dt






(5-3)
Where the subscript s refers to the surface of the austenite, x refers to distance into the
austenite and the suffix 0 refers to the initial base steel being carburized. Other symbols
have their usual meanings. Such solutions have been cast in the form of: case depth =
K(t)
1/2
and tabulated results published.
Typical concentration profiles at various times and temperatures and case depth results
appearin Figure 5-9 and Table 5-2.
The microstructures in the carburized case depend upon the details of the temperature -
time history. Examples can be found in the metal handbooks and in Krauss’ book. A
major difference exists in samples that were quenched directly from the carburizing
furnace compared to samples that were first cooled and then re-austenitized and
R. H. Wagoner MSE 661 – Steel Section 5, Page 31
quenched. In the latter case, the re-austenitization of the ferrite produced on cooling from
the carburization furnace produces a fine grained austenite in contrast to that which exists
after prolonged carburization because the re-austenitization temperature is lower, usually
below the A
cm
line. In this case, the steel matrix is austenitized and the proeutectoid
cementite forms into globular particles that are retained as such, dispersed in the
martensite, when the samples are quenched. The microstructure so formed comprises
fine crystals of martensite that is about a hard as it can be and a dispersion of
proeutectoid particles that is beneficial for wear resistance. In contrast, the austenitic
grain size at the end of the carburization time is large. Often the part is cooled below the
A
cm
before it is quenched. This reduces the thermal shock by decreasing the
temperature difference between the sample and the quenching medium. However,
proeutectoid films can grow along the austenite grain boundaries during this stage and
these films are retained along the prior austenite grain boundaries during quenching.
Many of the martensitic plates are bigger because the grains in which they form are large
and, also, the retained austenite is less finely dispersed.
In some applications the steel that is carburized has a relatively low bulk carbon content.
Many have about 0. 2 wt. %. This usually means that the microstructure within the core
is a mixture of ferrite and pearlite with ferrite being the dominant phase. As an extreme
example, the metal parts which have raised letters on their end faces and used to be used
in typewriters to print letters by striking a dyed ribbon were often made of carburized
(shallow case depth) AISI/SAE 1010. On the other hand, there are many applications for
which strong cores are necessary as well as hard cases. Thus, the steel chosen will have a
hardenability large enough for the core to become martensitic or bainitic during the
quench. Krauss gives the examples of carburizing steels shown in the accompanying
figure. Others appear in the handbooks.
Fatigue is an important consideration for many case-hardened and heat-treated
components, which are often used for rotating, load-bearing applications. Such
components undergo large changes in stress over many cycles during their lifetimes, as
well as being subjected to wear at the surface. Experience has shown that grain
refinement by one or multiple cyclic heat treatments is very effective in increasing the
resistance to fatigue crack nucleation and propagation, Figure 5-10.
Nitriding Steels: The carburizing process requires that the parts be quenched. The
quenching action and martensitic transformations always create the risk of distortion,
warping and cracking. Case hardening can also be done via the formation of numerous
precipitates of nitrides near the surface
of the steel. This process is called nitriding and, in contrast to carburizing, is performed
while the steel is largely ferritic not austenitic. Thus, the process does not involve rapid
temperature changes and allows excellent dimensional control.
Typically, the steel to be nitrided is a medium carbon steel that is initially quenched and
tempered at temperatures from 540°C to 700°C. The high tempering temperature is
intended to minimize microstructural changes that might occur during the subsequent
R. H. Wagoner MSE 661 – Steel Section 5, Page 32
nitriding processes that are carried out at lower temperatures. These temperatures are
typically 500 to 525°C. Nitriding is conducted in controlled atmospheres that use the
decomposition of ammonia to generate nascent nitrogen on the surface of the steel. This
nitrogen diffuses into the steel and also combines with the iron at the surface to form iron
nitride. The reaction is NH
3
= N + 3H
Nitriding steels contain alloying elements that form nitrides so that the nitrogen that
diffuses in forms numerous very fine precipitates. Aluminum is used for this purpose; e.
g. 1. 3 wt. % Al. The case depth, in this case the depth of the precipitate band, is
shallow, usually less than half a millimeter, even though nitriding times can exceed a 100
hours. During single stage, nitriding treatments in which a single nitriding atmospheric
composition is maintained, a “white layer” is formed. (So-called because in standard
etchants for steels it is unattacked. ) This layer is iron nitride and is hard but it can crack
and spall. When this is unacceptable the layer is removed by surface grinding.
Alternatively a two stage nitriding process is used in which, after the first stage, the
atmospheric conditions are changed so that iron nitride no longer forms at the surface and
the existing layer is removed because its nitrogen content dissolves into the steel.
Because the two processes described are successful, combined processes have been
developed in which both carbon and nitrogen are simultaneously diffused into the surface
These are carbonitriding and nitrocarburizing and are done to both austenite and to
ferrite.
R. H. Wagoner MSE 661 – Steel Section 5, Page 33
Figure 5-9: Carburizing profiles vs. time and temperature. (Krauss, p. 290-1)
Distance, 0.001 in.
0
0.4
0.8
1.2
1.6
2
0 1 2 3 4 5
Distance, mm
C
a
r
b
o
n

c
o
n
c
e
n
t
r
a
t
i
o
n
,

w
t
%
0 40
80 120
160
1700 F
D=Const.
925 C
t = 2
t = 4
t = 8
t = 16
Distance, 0.001 in.
0.4
0.8
1.2
1.6
C
a
r
b
o
n

c
o
n
c
e
n
t
r
a
t
i
o
n
,

w
t
%
0
2
0 1 2 3 4 5
Distance, mm
0 40 80 120 160
1800
1700
1600
R. H. Wagoner MSE 661 – Steel Section 5, Page 34
Table 5-2: Calculated case depths. (Krauss, p. 292)
Case Depth Calculated by the Harris Equation
Case depth (in.)
Time, t after carburizing at(a):
(h) 1600 F 1650 F 1700 F
2……………… 0. 025 0. 030 0. 035
4……………… 0. 035 0. 042 0. 050
8……………… 0. 050 0. 060 0. 071
12…………….. 0. 061 0. 073 0. 087
16…………….. 0. 071 0. 084 0. 100
20…………….. 0. 079 0. 094 0. 112
24…………….. 0. 086 0. 103 0. 122
30…………….. 0. 097 0. 116 0. 137
36…………….. 0. 108 0. 126 0. 150
1. Case depth = 0.025 t for 927
0
C (1700
0
F); 0. 021 t for 899
0
C (1650
0
F);
0. 018 t for 871
0
C(1600
0
F). For normal carburizing (saturated austenite at the steel
surface while at temperature).
Figure 5-10: Effect of grain size on fatigue life.
Double reheat
Single
reheat
Direct quench
8620 steel
4-pt bend
M
a
x
i
m
u
m

s
t
r
e
s
s
,

k
s
i
Cycles to failure
10
3
10
4
10
5
10
6
10
7
100
200
300
R. H. Wagoner MSE 661 – Steel Section 5, Page 35
Section 5 – Study Questions
1. What are the principal ways of classifying alloying elements in steels.
2. Some alloying elements are known as “carbide formers.” What does this term mean?
Beyond this, some elements are called very strong, strong or moderate carbide formers?
What is meant by this, and which are the important elements of each category?
3. Describe the effects of very strong carbide formers on hardenability.
4. Describe the mechanisms by which alloying effects affect hardenability.
5. What five elements are added to improve hardenability? Of the four elements which
can be added in appreciable quantities (>1%), which is the most powerful hardenability
agent and which is weakest? Describe the AISI system for labeling steels containing
these elements as the primary additions.
6. Why might one add large quantities of Ni (10%) to steel?
7. Why is Si added to cast irons?
8. Describe the effect of alloying elements on the Ms temperature, depending on their
classification.
8. Define the Md. How does the presence of Md sometimes affect the forming behavior
of alloys steels?
10. Why does Stage II tempering become more important in alloy steels?
11. What is meant by “secondary hardening?” Show what is meant and outline the
mechanism.
12. Contrast the mechanisms and heat treatment of one-step and two-step embrittlement.
How do these relate in terms of mechanism to blue brittleness?
13. What are the three principal kinds of fracture? Describe the fracture mechanisms and
show or describe the microscopic characteristics you would expect to see in each case.
14. Describe case hardening by carburizing in some detail. Why would such a treatment
be preferred over other hardening mechanisms?
15. What are the principal advantages and disadvantages of nitriding as compared to
carburizing as case-hardening mechanisms?
16. What are the three principal kinds of fracture? Describe the fracture mechanisms and
show or describe the microscopic characteristics you would expect to see in each case.
Physical Metallurgy of Steel
Class Notes and lecture material for MSE 661 (formerly 651. 01)
Notes revised by: Robert H. Wagoner, Smith Chair
Notes originally compiled by: Glyn Meyrick, Professor Emeritus
Web installation by: Wei Gan, Graduate Research Associate
Last revision date: 11/22/02
SECTION 6 – SPECIAL STEELS
Ultrahigh Strength Steel
As a general guide, yield strengths of 690 -1380 MPa (100 -200 KSI) are considered to
be high, and those above 200 KSI and up to 300 KSI are very high. Steels with strengths
near and above the upper end of the latter range are called ultrahigh strength steels. 100
KSI ≈ 689. 5 MPa. To obtain high strengths it is usually necessary to heat treat the steel
to convert it to tempered martensite. Some standard low alloy steels such as 4130, 4140,
4330 and 4340 in the tempered martensitic condition can have yield strengths above 200
KSI. Greater strengths are obtainable by modifying the compositions of the standard
steels. High strengths can also be produced by combining plastic deformation processes
with heat treatment in what are called thermo-mechanical treatments. Examples of
applications which ask for very high yield strengths are aircraft landing gear, submarine
hulls and high pressure chambers.
Because high and ultrahigh strength steels have very poor toughness when converted to
martensite it is necessary to temper them. Temperatures as high as 300°C can be
required. This is necessary in spite of the fact that the tempering reduces the yield
strength: the objective is a compromise that gives acceptable strength and toughness and,
hopefully, high enough values of both for successful application. While it is true that
harmful residual stresses can be reduced during such tempering treatments and the
toughness can be improved, practical problems arise because one step tempered
martensite embrittlement lurks dangerously nearby. The search for ways by which these
standard alloy steels might be modified in order to reduce the risk of one step tempered
martensite embrittlement has been the subject of much research in the past few decades.
Modified Standard Alloy Steels.
It is generally accepted that one step tempered martensite embrittlement is associated
with the formation of sheets of carbides at the beginning of stage 3 tempering and, also,
the segregation of some solute elements such as phosphorus at prior austenite boundary
sites. The carbides form by nucleation and growth processes and thus the time required
for their appearance within the microstructure can be increased if the kinetics of the
transformation can be retarded. When appropriate alloying elements that might
R. H. Wagoner MSE 661 – Steel Section 6 Page 2
22
accomplish this are considered, graphite stabilizers come immediately to mind because
they retard the rates of nucleation and growth of cementite. Additions of Si have been
found to be effective for this purpose. It is able to delay these reactions sufficiently for
useful residual stress relief and concomitant toughness improvement to be obtained
during tempering at 300°C before harmful quantities of carbides form. In general, the
toughness of the decomposition products of austenite is always improved by refining the
austenitic grain size. Also, in the present case there is an added advantage to having a
small austenite grain size. When the grain size of the austenite is reduced, the area of the
prior austenite grain boundaries is increased. Thus contributions to embrittlement from
solute segregation at these interfaces should be reduced. Following this reasoning,
vanadium is often added in order for it to form carbides and nitrides which restrict grain
growth during austenitizing.
Many of the modified alloy steels have trade names e. g. INCO 300M, which is 0. 35-0.
45 % C, 0. 65-0. 9 Mn, 1. 45-1. 8 % Si, 0. 7-0. 95 % Cr, 1. 65-2 % Ni, 0. 3-0. 45 % Mo,
0. 05 % V with P and S less than 0. 025 %.
Here we have V to lower the austenite grain size and thereby improve toughness in the
final product; Mn, Mo, Si, Ni and Cr to increase hardenability; Si to strongly retard
cementite formation which enables tempering to be done while minimizing the risk of
embrittlement; Mo helps to defend against the segregation, of tramp elements at grain
boundaries. Ni is not a desirable element with respect to intergranular failure by temper
embrittlement because it tends to promote segregation. If, however, this form of
tempered martensite embrittlement is indeed avoided, then the Ni that is dissolved in the
matrix has a strong, favorable effect on toughness during transgranular fracture.
Minimizing P and S reduces segregation. Modern steel making techniques produce
cleaner steels than those made earlier and this also improves the fracture toughness.
Even when very high strength steels are successfully treated to take advantage of the
strength of tempered martensite while remaining free from one step tempered martensite
embrittlement, they are still not very tough and they do not tolerate flaws. Generally, it is
to be expected that large structures will contain flaws such as pre-existing cracks or sites
at which cracks can form more easily than elsewhere. The fracture toughness of a
material that contains such a flaw determines the critical applied stress at which the flaw
will propagate. The lower the toughness the lower the stress for a given flaw to
propagate. Typically, cracks propagate by cleavage below the DBTT and by micro void
coalescence above it. For the ductile failure mechanism, scanning electron microscopic
studies have demonstrated that inclusions in steels are often involved in the formation of
the voids. When the fracture surfaces are studied, small inclusions are found inside many
of the dimples. This indicates that the interface between the inclusion and the matrix is
one across which there is weak bonding so that it is a site at which a void can more easily
nucleate. When a steel contains inclusions they can often dominate the impact toughness
even when other favorable microstructural changes are made. Thus as the yield strengths
of the steels are increased to higher and higher levels it becomes increasingly important
to control the inclusion content and shape. S and O must be reduced. Additions of rare
earth elements such as Ce are beneficial for making very small spherical oxide inclusions.
R. H. Wagoner MSE 661 – Steel Section 6 Page 3
33
Precipitation Hardened Steels
The strengthening mechanisms used in the modified standard alloy steels described above
are those provided by the martensitic transformation. It is unfortunate that the tempering
process is necessary because it tends to reduce the strengths initially achieved. This led
people to consider whether or not secondary hardening processes during tempering could
be effective enough to produce ultrahigh strength steels. One example is a group of steels
called maraging steels which were so named because the martensitic transformation was
used to obtain a saturated solid solution in which precipitation hardening was
subsequently obtained by aging the steel at a suitable temperature.
These steels were first developed in the 1960's and have been used in aircraft landing
gear. They contain a rather high content of Ni which makes them expensive. This
partially explains why they have not been as widely used as their properties would imply
that they should be.
A typical example is Mar M 300 for which the yield stress can reach 350 KSI (2. 4 GPa).
It contains: . 03 % C max. , 0. 1 % Mn max. , 0. 1 % Si max. , 18 % Ni, 5 % Mo, 9 %
Co, 0. 6 % Ti, 0. 1% Al.
A typical heat treatment is as follows:- austenitization at 850°C, cool to room
temperature and age for 3 hours at 480°C. (See Figure 6-1.) Yield strengths can be as
high as 350 KSI with 6 % elongation in tension. During austenitization essentially all the
alloying elements are dissolved. Hence the steel has a high hardenability, and is air-
hardenable in thick sections. Note that because the steel contains less than 0. 03 % C,
the diffusion controlled transformations that have to be avoided in order to form
martensite those that form ferrite and those that involve precipitation of intermetallic
compounds among the iron and the alloying elements. These compounds grow by means
of substitutional diffusion which is a sluggish process compared to interstitial diffusion.
Because the martensite has a low carbon content it is relatively soft, typically in the
range of about RC 25-30. This means that it is machinable so that parts can be easily
manufactured after the martensitic transformation has occurred. The martensite has the
low-carbon, lath type of microstructure and contains many dislocations and interfaces.
These become important during aging process because they are catalytic sites for
heterogeneous nucleation of the intermetallic compounds. During aging, many small
precipitates of Ni
3
Mo and FeTi form. These precipitates provide the barriers to
dislocation glide that are necessary for the high strength. Also, because they nucleate
heterogeneously on dislocations, they pin much of the dislocation density present
initially which prevents recovery from taking place during the aging process. Recovery,
of course, would reduce the yield stress. The cobalt provides some solution
strengthening and it has been suggested that the cobalt-rich matrix undergoes short range
ordering which would also contribute to strengthening.
Vacuum melting techniques are used to minimize P, O, S and N; this is necessary to
ensure that the toughness is acceptable. Finally, it is interesting to note that these steels
R. H. Wagoner MSE 661 – Steel Section 6 Page 4
44
have been found to have good resistance to stress corrosion cracking.
Figure 6-1: Heat treatment and mechanical properties of precipitation-hardening steel.
850
T
C
Austenitization
MAR M 300
heat treatment
3 Hrs
Time
Age
Fabricate
480 C
10 Ni---2Cr
1Mo, 8 Co, 0.19 C
Tempering curve
Secondary
hardening
steels
200 400 600
Tempering temperature C
1.Hour
H
A
R
D
N
E
S
S
C
H
A
R
P
Y

E
N
E
R
G
Y
80
40
160 180 200
σσσ σ
y
(0.2% offset) KSI
Test at 27 C
Same alloy
480 C tempering
treatment
Arrows show
results of
increasing
temering time
R. H. Wagoner MSE 661 – Steel Section 6 Page 5
55
The maraging steels contain very little carbon and the precipitation reactions described
above involve substitutional elements that form intermetallic compounds; carbides are
unimportant. In contrast, the phenomenon of secondary hardening that can occur during
tempering of martensite in appropriate steels is associated with the precipitation of fine
carbides that are rich in alloying elements such as Mo and Cr. Because the tempering
treatments that cause secondary hardening are carried out at higher temperatures than
those used in the heat treatments of low alloy and modified low alloy steels, improved
toughness can be expected. This can be advantageous provided the secondary hardening
precipitation processes are able to produce ultrahigh strengths. Much research has been
devoted towards this objective and a class of steels containing Ni, Co, Cr, and Mo has
been developed during the last 25 years, in which secondary hardening yields very high
strengths and good toughness. At this stage, in this document, we are familiar with the
general roles and effects due the Ni, Cr, Mo and carbon but what about the cobalt? The
principal effects discovered during this research are: cobalt additions cause an increase
in the M
s
temperature in contrast to the reductions caused by the other elements; it
dissolves mainly in the matrix and provides some solution strengthening and,
importantly, it accelerates the kinetics of the nucleation and growth of the alloy carbides
responsible for secondary hardening. Let us consider some examples.
Speich et al. in Met Trans. , 4, 303 (1973), describes an investigation in which they
studied 19 steels that had compositions that were variations around a basic mixture of 10
wt. % Ni, 2 wt. % Cr, 1 wt. % Mo, 8 wt. % Co and 0. 12 wt. %C and which contained
small amounts of Si and Mn and trace amounts of Al, N, S and P. The highest carbon
content was 0. 19 %. This is rather smaller than the carbon levels in some of the
modified xx40 standard steels because low carbon content promotes weldability. Typical
heat treatments were austenitization at 845°C and quenching into water after which
samples were tempered at temperatures that spanned the range from 100-700°C for
various times. Based on the results obtained in these experiments, detailed experiments
were conducted by tempering in the range of 480-510°C which spanned the secondary
hardening peak.
It is important to remember that the microstructural changes produced during tempering
are functions of time and temperature, Figure 6-1. First, we consider the effects of
temperature at fixed tempering times of one hour. The investigation showed that
initially, as the tempering temperature was increased, either there was little variation in
the hardness or some softening occurred due to the precipitation of cementite during
tempering below the secondary hardening range. Then when the temperature became
high enough, typically in the 420-480°C range, the cementite dissolved and the
precipitation of numerous very fine particles of (Mo,Cr)
2
C alloy carbides began. These
microstructural changes caused secondary hardening to begin and, as the tempering
temperature was increased further, the hardness rose to a peak and then declined again.
A notable aspect of these results is that the rate at which the hardness decreased as the
temperature increased, when the temperature was above the value that gave peak
hardness, was rapid, more rapid than its rate of increase with temperature for
temperatures that were below. We will refer to this again when we consider the reasons
for overaging later. The height of the peak in the hardness versus tempering
R. H. Wagoner MSE 661 – Steel Section 6 Page 6
66
temperature, increased with increasing carbon content because the volume fraction of the
carbide phase increased accordingly. It also increased with an increase in the
molybdenum content, reaching a maximum at 1%. Chromium additions shifted the
tempering curve to slightly lower temperatures but this effect saturated out at about 2 %.
Cobalt did not partition to the alloy carbides yet its presence in the alloy raised
considerably the peak hardness and also the yield strength that could be obtained in a
given tempering time. Presumably, this is due primarily to the fact that Co enhances the
kinetics of the transformation. Yield strengths of over 180 KSI ( 1240 MPa) were
obtained.
The observations described above, showed that a hardness peak is obtained when
tempering is carried out for fixed times of one hour at different temperatures. This is, in
general, similar to the usual behavior of precipitation hardenable alloys, for which aging
at a constant temperature causes a gradual increase in hardness with time until a peak is
reached. Then the hardness decreases from the peak value as time elapses while aging is
continued. Ordinarily during this process in precipitation hardenable alloys the toughness
decreases as the hardness rises and then increases again as the hardness decreases. An
interesting observation made in the work being described here for secondary hardening
was that when the Charpy impact energy and yield strength of the samples were both
determined as a function of tempering time, the curve obtained by plotting impact energy
values against the yield strength values was an inverse “C” shaped, as shown in the
appended figures. This curve shows that as tempering in the secondary hardening range
begins, both the strength and toughness initially increase. This is in contrast with the
usual behavior of materials in which the toughness generally decreases as the material
becomes stronger. Eventually, as the tempering time grows longer, overaging begins and
the strength begins to fall while the toughness continues to increase, as in the usual
behavior. Consequently, a tempering treatment can be chosen to obtain an optimal
combination of strength and toughness. The authors concluded that nickel was beneficial
with respect to the toughness. This work led to many subsequent investigations of
similar steels which has resulted in the commercial development of steels such as:
AF1410. . . 0. 16 % C, 10 % Ni, 13. 8 % Co, 2 % Cr, 1 % Mo, 0. 16 % Mn and trace
amounts of Ti, Al, S, P with O and N in parts per million, and AerMet 100 which was
developed in the early 1990’s and has compositions near 0. 16 % C, 11 % Ni, 3 % Cr, 1.
2 % Mo, 14% Co, and small amounts of Mn and Si.
These steels are vacuum melted and manufactured with careful attention to cleanliness so
that they contain low amounts of impurities. Recent publications describing the
microstructural phenomena in these steels are by Raghavan and Machmeier in Met Trans.
24A, Sept. 1993, page 1943 and in Met Trans. 27A, Sept. 1996, page 2511.
The strength of these steels is controlled by the inherent flow stress of the matrix, the
dislocation density and sub cell size, the grain size and the effects due to the precipitate
population. The initial microstructures of the martensites are controlled by the particular
heat treatments and the M
s
temperature of the steel. It is generally true that the tendency
for martensite to adopt a twinned plate morphology, increases as the M
s
temperature
R. H. Wagoner MSE 661 – Steel Section 6 Page 7
77
decreases. Conversely, the tendency for the formation of dislocated lath martensite
increases as the M
s
temperature increases. Autotempering reactions occur more readily
the higher the M
s
temperature is. Lath martensite predominates in AF 1410 while
partially twinned plate martensite predominates in the Aermet 100. Secondary hardening
is produced in the martensite of both types of steel by the precipitation of a high density
of M
2
C carbides during tempering. Here the M stands for metallic elements. The ratio
of metal atoms to carbon atoms is 2: 1 but the metal atoms in the carbides include Mo, Cr
and Fe; the precise composition of the carbides depends on the total composition. For
example, if the molybdenum composition in the alloy is raised, then the molybdenum
content of the carbides is increased.
The austenitizing treatments used in the investigations referred to above did not produce
completely homogeneous austenite. Instead, the austenite contained some MC and
M
23
C carbides that were retained in the martensite after quenching but they did not
participate in the tempering reactions. When these steels are quenched to room
temperature it is usual for them to contain a small amount (1-2 %) of retained austenite.
Refrigeration eliminates the retained austenite. Combinations of tempering times and
temperatures that allowed cementite to form but which did not produce secondary
hardening e. g. for temperatures below about 427°C, produced plates of cementite on
plate, lath and twin boundaries in the martensite. Then in the temperature range of 450-
485°C secondary hardening occurred. Samples tempered at 510°C rapidly overaged.
There seems to be general agreement that the secondary hardening peak occurs when the
optimum population of M
2
C is generated. Here optimum refers to the particle size and
spacing. Furthermore, the more recent investigations show that at peak hardness, the
carbides are needle-shaped and coherent with the matrix and that the sudden, rapid
softening at the higher temperatures is caused by a transition from coherency to
incoherency, rather than carbide coarsening due to Ostwald ripening. This description is
based on details of the TEM observations in which characteristic contrast effects
indicated that for the peak hardness the particles were surrounded by coherency strain
fields. This contrast was absent around carbides that formed at 510°C.
It appears to be agreed that the passage of dislocations through the precipitate array is by
Orowan looping of dislocations around the particles. When looping occurs in the absence
of strain fields around precipitate particles, the dominant microstructural parameter that
effects the stress required to force a dislocation to glide through the array is the inter-
particle spacing. The observation that overaging sets in rapidly when coherency is lost
and before there is any significant change in this spacing suggests that the interactions
between the strain fields of the carbides with those of the dislocations are an important
contribution to the strengthening effects of the carbides. This allows us to begin to
understand the effects due to the variations in the composition of the steel on the
secondary hardening curves. For example, increasing the molybdenum content to 1 %
increases the molybdenum content of the carbides. This changes the lattice parameters of
the carbides because of the different atomic sizes of its constituents. Because the
magnitude of the coherency strain depends upon the differences in the lattice parameters
R. H. Wagoner MSE 661 – Steel Section 6 Page 8
88
of the two phases it will change and therefore so will the magnitude of the resistance to
the glide of dislocations due to the coherency strain fields.
R. H. Wagoner MSE 661 – Steel Section 6 Page 9
99
High Strength Low Alloy Steels (HSLA).
There are many applications of steel for which the steel is not heat treated to produce
hardening via the martensitic transformation. Instead, it is used directly as supplied.
Naturally, the successful use of the “as manufactured” steel depends on how well the
manufacturing process produces material in the appropriate form and with the desired
properties. Examples of the form of the steel are plates, sheets and bars. Plain carbon
steel sheet and plate is easily produced by hot rolling wherein the steel is rolled while
austenitic and then cooled. After natural cooling these steels have microstructures that
consist of mixtures of proeutectoid ferrite and pearlite. The mechanical properties
depend upon details of the microstructure as has already been described. For many
applications e. g. ships, gas and oil pipe lines and oil-drilling platforms, it is essential
that the steel has good formability and good weldability as well as reasonably high
strength and toughness. Increasing the carbon content of the steel increases the strength
but decreases the weldability because it raises the hardenability, changes the properties of
the transformation products in the heat affected zones and enhances residual stresses in
the vicinity of the weld. For good weldability it is desirable to keep the carbon contents
low. This is also desirable for good toughness. Unfortunately, reducing the carbon
content in the interests of improving weldability and toughness decreases the volume
fraction of pearlite in the steel; this reduces the yield strength.
What can be done to solve this dilemma?
Contemplation of the relevant factors reveals that both the yield stress and the toughness
of low carbon steels can be increased by refining the ferrite grain size. The final grain
size of the proeutectoid ferrite in hot rolled products depends upon the grain size of the
austenite when it transforms during cooling and, also, on the rate of cooling after the
final rolling pass. Recrystallization takes place in the austenite during hot rolling.
Recrystallization can occur in semi-killed, low carbon, unalloyed steel at temperatures as
low as 760°C. It is not possible to hot roll the standard low carbon steels in the range of
900-1000°C and also produce fine grained austenite because dynamic recrystallization
takes place during rolling. This means that unless the cooling process is accelerated by,
say, water cooling, the final product of such treatments has coarse ferrite grains which
lowers the strength and toughness.
Because of these limitations on plain carbon steel, a new class of low carbon steels has
been developed in the last few decades. The objective has been to produce a material
with good strength, toughness, formability and weldability. The carbon content is kept at
or below
0. 25 wt. %. These steel are called high strength low alloy (HSLA) steels. Sometimes
they are referred to as microalloyed steels because they do not contain large amounts of
alloying elements.
Before we see what kind of steels these are, let us ask how we would go about designing
a steel which will have these properties when manufactured in the physical form ( e. g.
R. H. Wagoner MSE 661 – Steel Section 6 Page 10
10 10
rolled sheet or plate) in which it is intended to be used? The first thing that must be
recognized is that the properties of the ferrite are the important properties because the
ferrite dominates the contributions from the small volume fraction of pearlite that might
be present. Thus we must consider the factors that control the properties of ferrite. These
are:
• grain size; fine grain size raises toughness and yield stress,
• dislocation content; a high density of dislocations raises the flow stress,
• precipitates; a distribution of finely spaced precipitates provides precipitation
hardening,
• solutes, elements dissolved in the ferrite provide solution strengthening.
Question: How do we produce a fine grain size in the ferrite and still use high rolling
temperatures to form the austenite?
First we must find a way of inhibiting dynamic recrystallization and grain growth in the
austenite because fine-grained austenite provides many more heterogeneous nucleation
sites for the proeutectoid ferrite than does coarse grained austenite. Then, provided grain
growth can be controlled, the transformation during cooling should yield a fine-grained
ferrite. It is well known that refinement of the austenitic grain size can be done by
putting particles of second phases into the austenite, so we need to arrange that suitable
particles are in the steel during the hot working process. If these particles are also
inherited by the ferrite then they will also impede ferrite grain growth. O. K. Has this
been done? Yes.
It was recognized that the logical candidates for these second phase particles are carbides
and nitrides. Ti, Nb, and V form carbides and nitrides in steel; Al forms nitrides. Also,
any excess amounts of these elements that are left in solution in the ferrite, will contribute
to solution strengthening. In practice, the highest temperatures experienced by steel
during the rolling process are those existing at the beginning of the process. At these
temperatures, above 1000°C, the austenite containing small quantities of the alloying
elements mentioned is mostly a solid solution in which the alloying elements are
dissolved. Then, as the temperature drops, precipitation of carbides and nitrides occurs
and the progress of recrystallization and grain growth is impeded. These elements are
very effective for this purpose and only small additions are required. As an example of
the effect, we note that a 30 % reduction in thickness at 815°C in a semi-killed, low
carbon steel is accompanied by 10 % recrystallization, while for the same amount of
recrystallization to occur when 0. 03 wt. % Nb has been added, a 50 % reduction is
required at 925°C.
These HSLA steels are produced to mechanical property specifications, e. g. 950A sheet
and strip has specifications of a yield strength of 50 KSI(345 MPa) and a tensile strength
of 70 KSI (483 MPa). The actual contents of Nb etc. are not specified. The yield and
ultimate strengths of the high strength low alloy steels range from 40 to 80 KSI and 60 to
95 KSI respectively. Typically the steels contain 0. 25 wt. % or less of carbon and 1 to
1. 65 wt. % Mn. The Mn helps to controls any sulfur and provides solid solution
strengthening.
R. H. Wagoner MSE 661 – Steel Section 6 Page 11
11 11
Leslie points out an interesting fact in his book. The use of small additions of Nb to
carbon steel in order to strengthen the ferrite was patented in 1936. But it was just a
scientific curiosity then because of the cost of Nb and, more importantly, because there
was no real commercial demand for such steels. 20 or so years later the price of Nb
dropped and also the need for pipe line steel appeared. Economics drove the renewal of
interest. The buck demands respect!
Question: What other microstructural characteristics could contribute?
An additional way of strengthening ferrite is to increase its dislocation density. How
might we accomplish that without cold work? Previously, when we considered the
microstructural characteristics of ferrite we noted that equiaxed ferrite forms in slowly
cooled austenite. This morphology indicates that the ferrite forms at relatively high
temperatures while the steel is being cooled. When this occurs, the dislocation density
within the grains is of the order of 10
8
cms/cm
-3
, which is not particularly high from the
point of view of strengthening. However, if we contrive to make the ferrite form at low
temperatures, then the ferrite morphology changes from equiaxed grains to acicular
grains and, because the ferrite is not exposed to the higher temperatures during cooling,
recovery processes are suppressed. Such acicular ferrite can contain a dislocation density
as high as 10
10
cms per cm
3
. Acicular ferrite can be made to form by cooling the steel
rapidly. However, it is more desirable to avoid this necessity and to arrange for the
acicular ferrite to form during air cooling.
For this to happen, the ferrite formation has to be retarded. This can be done
economically by adding Mn and Mo. An example is : C 0. 08 %, Mn 2. 18 %, Si . 1 %,
Mo 0. 38 %, Nb 0. 07 %, Al 0. 02 % and N 0. 007 % (this composition is in wt. %0.
These alloy additions refine the austenitic grain size and also cause the beginning of the
transformation to be retarded because the ferrite start curve on the TTT diagram is
displaced to longer times. Therefore, the austenite decomposes at lower temperatures as
it cools, yielding fine grained, acicular ferrite and a greater dislocation density. It turns
out that the ferrite is often then supersaturated with respect to Nb so that precipitation
can occur if the steel is reheated. An interesting discovery that was made when these
dislocated acicular ferrites were produced, was that the yield point phenomenon
disappeared. Instead, the steel began to yield gradually with a rounded stress strain curve
and then work hardened rapidly. This suggests that the microstructure produced in this
manner, contains a significant population of mobile dislocations. For steels of the
composition above, yield can occur at 70 KSI in samples in the hot rolled, air cooled,
condition. For comparison, the lower flow stress in a 0. 08 % C plain carbon steel is
about 25 to 30 KSI.
A logical extension of the acicular ferrite case is to arrange to produce low carbon
bainite. To do so, Mo and B are introduced to form a bay in the TTT diagram; the carbon
level is raised to increase the strength and Nb is added to control the austenitic grain size.
The result is to separate the ferrite and pearlite curves from the bainite curves so that the
R. H. Wagoner MSE 661 – Steel Section 6 Page 12
12 12
transformation to bainite is easily achieved, as is clear from the TTT diagram sketched in
Figure 6-2.. Such steels can yield in the range 65-120 KSI.
Figure 6-2: An alloy steel TTT diagram with a bay.
In summary, HSLA steels are low carbon steels with various combinations of small
(micro-alloying) additions of Nb, V, Ti, Mo, and B. They have increased in popularity
during the last quarter of a century. However, versions of such steels existed before,
where the strength was provided by solution strengthening and solute clustering. USS
Corten steel has the composition:
% C, 0. 35 % Mn, 0. 5 % Si, 0. 4 % Cu, 0. 1 % P and 0. 8 % Cr.
One of the principal reasons that it has been used is that it is an example of a weathering
steel in which the combination of the Cu and P affects the rusting process so as to
produce a tightly adherent, reddish patina that protects the steel and, so some people
claim, looks pretty.
Question: Can we make use of precipitation hardening?
Grain refinement is a desirable way of strengthening these types of steel because this
also improves toughness. Often precipitation hardening is used in addition. In order for
precipitates to be effective obstacles to dislocation motion, they must be closely spaced.
This means that the precipitates used for refining the ferrite grain size are of little use for
precipitation strengthening, because they are large and their spacing scales with the grain
size. This is fine for anti-recrystallization purposes because the scale of the
microstructure is larger but finer spacings are necessary for additional strengthening due
to precipitation hardening. How can we bring this about? One obvious approach is to
seek additional alloying elements that will be retained initially in solid solution in the
ferrite and then enable a distribution of fine precipitates to be produced by aging. Copper
additions such as those in Corten steel can do this. We will consider another type of
reaction first, however, because it turns out that the solubility of the niobium and
vanadium carbides are greater in austenite than in ferrite. Thus if the matrix composition
of the austenite in which the grain refining precipitates have formed is inherited by the
ferrite, then the ferrite is supersaturated and another population of fine precipitates can
ααα α+ Fe
3
C
Bainite
log t
A
1
B
s
M
s
T
R. H. Wagoner MSE 661 – Steel Section 6 Page 13
13 13
occur as ferrite forms or as it is aged. This occurs more readily for vanadium additions
than for niobium.
A summary of relative contributions of various hardening mechanisms in to steels
appears in Figure 6-3.
Figure 6-3: Yield strength as a function of grain size in low-carbon steels.
Contributions of various other strengthening mechanisms are indicated. (after Krauss, p.
137.)
An interesting mode of precipitation occurs in these alloys. It has been observed in other
alloy systems since its was discovered in HSLA steels. It is called interphase
precipitation. Its major characteristics are that the precipitation occurs on the interface
between the ferrite and the austenite and the final morphology is regularly spaced sheets
of small particles within the ferrite. Very small particles, of the order of 5 NM, can form
in closely, spaced arrays that are effective strengtheners. Figure 6-4 illustrates the
process.
The precipitation mechanism begins after a ferrite grain has nucleated in the steel as it is
cooling from the rolling process. Heterogeneous nucleation of the carbide phase then
occurs along the austenite/ferrite interphase boundary. The particles of carbide e. g. VC,
pin the boundary which stops migrating, and then the carbides grow. As they grow they
deplete the carbon concentration in the austenite at the interface. When the concentration
drops as indicated in the sketches, the driving force for forming ferrite is increased.
ASTM Grain Size Number
Y
i
e
l
d

s
t
r
e
n
g
t
h

(
M
P
a
)
d
-1/2
(mm
-1/2
)
Mild steels
HSLA
steels
Ferritic
grain-size Solid
solution
Matrix strength
k
s
i
∆∆∆ ∆Y
d (microns)
0 4 8 10 12
600
400
200
0
0
5
10
15
0
25
50
75
5 10 25 100 250
R. H. Wagoner MSE 661 – Steel Section 6 Page 14
14 14
Eventually it becomes large enough to overcome the restraint of the precipitates on
austenite/ferrite interface and allow the ferrite to grow. . Then, as the interface advances,
the carbon gradient ahead of it builds up again until nucleation of the carbides can occur
again and the process repeats itself. The observations of the phenomenon suggest that the
boundary "un-zips" from the sheets of precipitates so that its forward migration is
actually accomplished by the sideways movement of boundary ledges.
Figure 6-4: Schematic diagram illustrating the unzipping process.
The fact that copper rich precipitates can form in copper bearing steels was mentioned at
the beginning of this section. In the early Seventies there was a great deal of interest in
the nature of the precipitation of copper from iron and low carbon steel primarily because
of fundamental aspects of phase transformations. There has been continued interest,
however, as copper bearing HSLA steels have been further refined. Alloys have been
developed which have very low carbon content, over 1 wt. % Cu for precipitation
hardening, additions of substitutional elements such as Mn, Ni, Mo and Cr for
hardenability along with Nb, Al and N for grain size control. An example is A710 which
is described in the 1995 ASTM book of standards. A particular example of this group is
HSLA-80 which is so designated by the US navy for ship building. It is appropriate to
end the section by returning to this topic to consult recent work.
% C
Interface
%C
G
Growth
Carbides
Magnification
Growth
Nucleation
Break
away
Uzipping
mechanism
Net
Uzipping
ααα α
γγγ γ ααα α γγγ γ
ααα α γγγ γ
γγγ γ
ααα α ααα α
R. H. Wagoner MSE 661 – Steel Section 6 Page 15
15 15
An example of HSLA-80 is: 0. 05 wt. % C, 0. 5 wt. % Mn, 0. 009 wt. % P, 0. 002 wt.
% S, 0. 28 wt. % Si, 1. 12 wt. % Cu, 0. 88 wt. % Ni, 0. 71 wt. % Cr, 0. 2 wt. % Mo, 0.
035 wt. % Nb, 0. 02 wt. % Al, 0. 009 wt. % N. A typical A710 is very similar.
Detailed studies of the phase transformations in these alloys both during continuous
cooling and during isothermal aging have been described recently by S. W. Thompson,
D. J. Colvin and G. Krauss, Met. & Mat. Trans. 27 A, 1557 (1996) and S. W.
Thompson and G. Krauss ibid. page 1573.
Inspection of the compositions of the steels shows that austenitic grain size control is
effected by the niobium, aluminum, nitrogen and carbon. The S and P contents are low
to minimize undesirable effects due to inclusions and segregation. Leaving aside
consideration of the copper for the moment, inspection of the remainder of the
composition leads one to conclude that these alloying elements will be in solid solution.
It would be reasonable to predict that if the austenite is cooled sufficiently slowly, the
austenite to ferrite transformation should produce predominantly relatively equiaxed
grains with normal dislocation content. This is so in practice. These authors refer to this
microstructure as polygonal ferrite. However, they also observe that even with this low a
carbon content, small pockets of austenite become enriched in carbon and transform to
upper and lower bainite and martensite. More rapid cooling reduces the amount of
polygonal ferrite and increases the amounts of Widmanstatten ferrite side-plates and
acicular ferrite. These authors also observe small equiaxed ferrite crystals which they
call granular ferrite. Some martensite associated with a little retained austenite can form
although the bainitic transformation is suppressed. Their results enabled them to present
a continuous cooling diagram. The complexities of the microstructural characteristics are
impressive, but the ingredients for good properties are there; fine ferrite grain size,
acicular, dislocated ferrite etc. Let us now turn to the precipitation reactions.
It was discovered many years ago that supersaturated binary alloys of Cu in BCC Fe
could be obtained by solution treating and quenching. Isothermal aging of the metastable
solid solutions caused precipitation of copper. Initially, metastable BCC copper
precipitates appear within the ferrite, presumably because they match the ferrite
structurally and can form coherently which is favorable for nucleation. These
precipitates grow to about 4 NM and transform in-situ to the FCC structure. In the HSLA
alloys under discussion, the copper is in solution when the steel is austenitic. If the
austenite is rapidly quenched so as to form martensite, the degree of tetragonality is small
because of the small amount of carbon in solution and the copper is essentially in solution
in a BCC host. Thompson and Krauss show that when the quenched steel is aged at 455,
565 and 675°C FCC copper precipitates form. They also show, as previous work had
indicated, that copper precipitation can also take place during slower cooling treatments.
Previous work had indicated that the copper precipitation in appropriately cooled
austenite began near 675°C. Accordingly Thompson and Krauss compared the
morphology of the copper precipitates produced during isothermal aging at 675°C with
their morphologies when formed during continuous cooling experiments. The results
showed that the precipitates produced in the isothermal aging experiments were
predominantly located at dislocations, indicating that in these materials, precipitation
from a supersaturated single phase at this temperature was nucleated heterogeneously. In
R. H. Wagoner MSE 661 – Steel Section 6 Page 16
16 16
contrast, detailed transmission microscopic studies established that when the copper
precipitated during continuous cooling of the austenite, the dominant mode of
precipitation was interphase precipitation at interphase boundaries between ferrite and
austenite. There is a useful consequence of this. The precipitates are not restricted to be
located at dislocations in the ferrite and may form in successive finely spaced lines of
precipitates that are effective at precipitation strengthening.
R. H. Wagoner MSE 661 – Steel Section 6 Page 17
17 17
Interstitial Free Steel
Now that we have discussed HSLA steels which have low carbon contents, it is an
appropriate time to mention interstitial free steel. These are steels in which the
concentrations of interstitially dissolved solute elements are very small. This is
accomplished by the dual processes of taking care to reduce the carbon and nitrogen
levels during the manufacture of the steel and by scavenging these elements from solid
solution by adding small quantities of niobium, aluminum and titanium. One of the
original reasons for producing such steel was that it was discovered that reducing the
carbon concentration improved the quality of enamel coatings. The formability is also
improved. Today, interstitial free (IF) steels represent the state of the artfor sheet steels
which have exceptionally good deep drawing properties. The properties of the final
product are determined by the chemistry of the steel and the particular thermo mechanical
history. These two determine the nature of the ferrite and the nature of the second phase
particles present. An example of such a steel is discussed by Rauf and Boyd in
“Microstructural Evolution During Thermo Mechanical Processing of a Ti-Nb
Interstitial-Free Steel Just Below the A
r3
Temperature” Met. Trans. 28A,1437, 1997.
The steel that they describe had the following composition in wt. %: C 0. 0038, N 0. 003,
S 0. 006, P 0. 006, Mn 0. 17, Si 0. 008, Ti 0. 018, Ni 0. 019, Al 0. 043. Note that it is
very rich in iron but is still called a steel. Perhaps now is a good time to re-read the
section from the fist page that followed the question “What, then, is a steel?” The final
microstructures comprised ferrite grains in which there was a dispersion of a variety of
fine particles of nitrides, sulfides and carbides. The nature of the ferrite matrix depends
upon the characteristics of the austenite at the end of the hot rolling process and the
manner in which it is cooled. As we should expect, the basic behavior is essentially that
which was described for iron on pages 8 and 9. Equiaxed ferrite forms when the cooling
rates permit the ferrite to form at high temperatures. In the language of IF steel such
ferrite is called polygonal ferrite. Increasing the cooling rate eventually suppresses the
polygonal ferrite because the austenite then transforms by the massive transformation that
creates the characteristic blocky, massive grains.
R. H. Wagoner MSE 661 – Steel Section 6 Page 18
18 18
Dual Phase Steels
(See: M. S. Rashid: Ann. Rev. Mater. Sci, 1981, vol. 11, pp. 245-266.)
This is another example of a class of steels that has been developed with the object of
tailor-making certain desirable properties for particular applications. Steel sheet is
commonly formed into complicated shapes for use in a variety of applications, as in the
automobile industry for example. Good formability and high strength are desirable for
production reasons and for weight reduction so as to economize on fuel use by the final
product. A summary of desirable properties is:
• high ultimate tensile strength for a good strength to weight ratio;
• no yield point phenomenon because that can cause non-uniform strains during the
forming process;
• a low initial yield strength, a rapid work hardening rate and a high ratio of
theultimate tensile strength to the initial yield strength.
These properties give rise to uniform straining, eliminate or minimize the occurrence of
plastic instabilities and produce high yield strength in the deformed regions of the final
part because of strain hardening.
If we consider these properties with respect to the behavior of plain carbon ferrite
/pearlite sheet we note that this material has some short comings. Slowly cooled, hot
rolled, low carbon, plain carbon steel sheet usually deforms via upper and lower yield
points followed by Luders band propagation. This causes non uniform straining and
localized thinning during forming. Because the Luders extension is localized, the plastic
strain rate within the band is high which causes the flow stress to be relatively high. This
reduces the magnitude of the ratio of the ultimate tensile strength to the lower yield
stress. When the necessary strength levels of the steel sheet require the presence of
significant volume fractions of pearlite colonies, difficulties arise because of local failure
mechanisms associated with the formation of cracks and pores in and adjacent to carbide
lamellae. What can be done then?
Question: How can we eliminate the yield point phenomenon?
• Temper rolling. This is a process wherein the steel sheet is passed through rolls that
bend it and re-flatten it in order to introduce fresh, unpinned dislocations which can
then glide during forming and thus eliminate the upper yield point . This process has
been used for years. The problem with it is that it uses up some of the available
ductility before the forming process can properly begin and it tends to reduce the
practical UTS/YS ratio.
• Because the yield point phenomenon is caused by carbon and nitrogen interstitial
solutes, removing these elements from the steel would eliminate the problem.
However, this would lower the strength and would not be a useful solution unless
other modifications can be found to raise the flow stress and we’ve already
considered high strength low alloy steels.
R. H. Wagoner MSE 661 – Steel Section 6 Page 19
19 19
• When the dislocations in ferrite are pinned by the interstitial solutes, the pinning
forces are so strong that the dislocations cannot break away unless the local stress is
very high. Thus we need some way to produce local stress concentrations within the
steel. In essence, we need to be able to modify its microstructure so that either some
clean dislocations exist in the as-produced material or to generate residual stress
concentrations that can be superimposed upon the applied stresses so that slip is
activated. These stress concentrations will have to be distributed throughout the steel
in order for them to cause macroscopically uniform plastic flow. This would be a fine
solution if we can find a way to do it.
Question 2: Assuming that we will be able to solve question one we now have to ask
question 2. How do we create a low yield stress, a rapid work hardening rate and a high
tensile strength?
Figure 6-5: Schematic diagram stress-strain behavior with a yield point phenomenon and
without.
The first thought that comes to mind is that the dominant matrix phase should be ferrite
because it is ductile above the DBTT. To obtain a high rate of work hardening during
plastic deformation it is necessary to arrange for a rapid rate of increase in the dislocation
density within the deforming grains. One way to lower the yield stress is to increase the
grain size but that would adversely affect the ductility and lower the tensile strength.
Fine grained ferrite is still desirable. A low applied stress for yield will arise if we can
introduce residual stress concentrations that help to initiate slip. Thus we find that we
find a single way to answer both questions. Because we want to accomplish all this
during the production of the steel it makes sense to ask what can be done via the phase
transformations that occur during heat treatments.
The production of ferrite/pearlite mixtures has already been ruled out. This leaves
ferrite, bainite and martensite. Consider martensite. When some martensite forms in a
steel which has a significant carbon content, it is hard and it generates localized strains
around it due to its change in shape and volume. Such incompatibilities are generally
disadvantageous in heat treatments designed to fully transform the steel because of the
risk of the occurrence of distortion and quench-cracking. Now, however, we are
searching for a way to produce residual stresses. If we can find a way to generate
numerous small, separated, volumes of martensite dispersed throughout a ferritic
microstructure we might have a solution our problem. The way to produce such a
Not wanted
σσσ σ
εεε ε
Desired
Behavior
σσσ σ
εεε ε
R. H. Wagoner MSE 661 – Steel Section 6 Page 20
20 20
microstructure is first to produce a mixture of fine ferrite and austenite grains and then to
quench it to convert the austenite to martensite. Thus the steel must be held in the
alpha/gamma two phase field and then cooled such that the austenite transforms to
martensite (see Figure 6-6). If some more proeutectoid ferrite forms during the cooling it
should have no particularly adverse effects; the formation of pearlite is undesirable. The
final microstructure should then comprise a mixture of ferrite and martensite.
Figure 6-6: Intercritical heat treatment to obtain a dual-phase structure.
The general procedures described above have been successfully carried out. Because the
product is a mixture of martensite and ferrite, the steels have been dubbed dual phase
steels. A purist might complain that ferrite/pearlite steels also contain two phases- ferrite
and cementite- and that a better name might be ferrimart or some such concoction.
However, there we are.
In order to ensure that the martensitic or lower bainitic transformation does occur during
natural cooling it is necessary to augment the hardenability. This can be done by adding
Mn and Si. An example is 0. 1 % C, 0. 5 % -0. 9% Si and 1 to 3 % Mn. In a typical
process the sheet is passed through a continuous annealing furnace where it is heated into
the alpha/ gamma range and then cooled again. Sometimes it is reheated to temper the
martensite, but normally this is not done because it tends to anchor clean dislocations,
reduce stress concentrations and lower the UTS which defeats the whole purpose. In
some processes, the continuous annealing is preceded by cold rolling. In this case, the
deformed ferrite recrystallizes during heating. With appropriate control of the rate of
heating and the degree of cold work, a fine grained mixture of ferrite and austenite can be
obtained. The time at the high temperature has to be short to retain these fine grains, so
the austenite is inhomogeneous. Typically the aim is to produce about 15 to 20 %
austenite.
T
RT
Fe x Wt %C Time
M
s
T
Rc
3
Rc
1
BAY
log t
T
R. H. Wagoner MSE 661 – Steel Section 6 Page 21
21 21
Another production method is to finish the hot rolling stage so that 80 to 90 % ferrite
forms and to allow the remaining austenite to transform later while the steel is being
formed into a coil. For this to be successful, the steel must have suitable continuous
cooling transformation characteristics. Alloying to produce a bay in the TTT diagram,
creates a window so that the metastable austenite can be coiled before the martensite or
bainite begins to form e. g. 0. 06 % C, 0. 9 % Si, 1. 2 % Mn, 0. 6 % Cr and 0. 4 % Mo.
Note that these compositions are similar to some of those of HSLA steels so the
properties of both are being combined.
Deformation Behavior and Microstructure: The general microstructural constituents
that can be present in dual phase steel after production are: a fine-grained ferrite matrix in
which there is a relatively low dislocation content because of the high temperature
excursion; some proeutectoid ferrite that formed from the austenite during cooling and is
therefore more dislocated than the other ferrite; martensite or lower bainite whose
microstructures depend upon the carbon content and the M
s
temperature. These
martensitic and bainitic volumes are associated with residual stresses be cause of the
transformation shape changes. Typically, the steels also contain some retained austenite
and finally, when V or Nb are present there can be carbonitride particles embedded
within the matrix.
When these kinds of steels are deformed at room temperature, the precise behavior will
depend on the microstructural details. Here, we consider the general types of behavior.
The first objective to be satisfied by these steels is for them to yield smoothly and at low
stresses. This is successfully achieved by the processing that has been described.
Because the Ms temperatures for these steels are high, auto tempering should lock many
of the dislocations generated during the transformation. The relative importance of
residual stresses, stress concentrations and clean dislocations is not clear. It is clear,
however, that the yield point phenomenon is absent. Consequently, the steel yields at
relatively low stresses and work hardens rapidly.
Initially, the plastic strain occurs in the ferrite that existed before cooling began because
this is the softer ferrite due to its low dislocation content. When that ferrite has work
hardened so that its flow stress is the same as the yield strength of the ferrite that formed
during cooling, all the ferrite deforms. The retained austenite usually transforms to
martensite early during the deformation. The associated shape change contributes to the
dimensional strain ocurring during the deformation and the transformation produces more
of the hard phase. Electron microscopic studies have shown that eventually the
martensite begins to deform and fracture. Voids open up at the martensite/ferrite
interfaces and general failure ensues if the deformation is continued. The tensile strength
varies approximately linearly with the % martensite or lower bainite. Typical results are
80 KSI for 10 % by volume rising to 110 KSI at 30 %. Elongations of about 20 % are
obtained.
During the tensile deformation of ferrite/pearlite steels, the lower yield strength is
typically 2/3 to 3/4 of the tensile strength. In contrast, the ratio in dual steels is 0. 5-0. 6.
R. H. Wagoner MSE 661 – Steel Section 6 Page 22
22 22
After the Luders extension is completed the stress strain curve for ferrite/pearlite steels
can be reasonably well described (except near yield) by the Hollomon equation
σ = Kε
n
(6-1)
where K is a constant and n is called the strain exponent. The equation relates the true
stress to the true strain. By taking logarithms of both sides, we see that this equation
predicts a linear relation between logσ and logε with a slope n. This linear relationship is
frequently, well obeyed as is shown by the curves marked alpha/P in the following
sketches. Dual phase steels do not obey this kind of equation. Examples of the behavior
that has been observed are shown, schematically in Figure 6-7.
Figure 6-7: Hardening law anomalies for dual-phase steels.
-2 -1
logεεε ε
5
4.6
log σσσ σ
d(logσσσ σ)
d(logεεε ε )
n
0.1 0.3
vol. fraction mart
of lower bfff
u
80
110
ααα α+++ +ρρρ ρ
ααα α+++ +ρρρ ρ
R. H. Wagoner MSE 661 – Steel Section 6 Page 23
23 23
Wrought Stainless Steels
Iron is not mined in its elemental form. It is extracted from ores in which it exists mainly
combined with oxygen in oxides. After being extracted it is then used to manufacture
steels. All the steels that we have considered so far are generally subject to degradation
by forming oxides and hydroxides, either by oxidation when heated or by corrosion
unless protected from environments that contain water and oxygen. Hence, once made
and put into use in steels, iron attempts to return to its former state by forming rust or
scale. Usually, with the occasional exception such as weathering steels, the corrosion
product is not effective protection for the steel below it. Consequently the steel
continues to rust as time passes. Oxidation occurs readily when carbon steel is heated.
The oxide scale spalls off and oxidation continues as long as the steel is hot. This
behavior along with other corrosion processes such as dissolution in dilute acids, imposes
obvious limitations on the use of carbon and low alloy steels.
Many methods of combating rusting and corrosion have been and are still being used.
These include protective coatings such as paint, plastic and tin that are intended to isolate
the steel from the corrosive agent, sacrificial coatings such as zinc that corrodes
preferentially, and the application of electrical potentials designed to provide protection
by inhibiting the electrochemical reactions. Alloying elements such as silicon,
molybdenum and copper are helpful for increasing lifetimes at ambient temperatures;
weathering steels have already been mentioned. As the effects of alloying additions on
the properties of steels became increasingly understood, it was discovered that Cr
dramatically improves the corrosion and oxidation resistance of steel. Cr oxidizes and
forms a thin, transparent, tightly adherent layer of oxide on the surface of the steel that
prevents or minimizes further corrosion. Steels to which sufficient Cr has been added for
this purpose, form the family of steels called stainless steels. The tonnage of stainless
steels used annually is only a small fraction of that of carbon and low alloy steels. Its
range of specialized uses, however, is very wide, spanning from knives and forks through
brewing to aerospace applications.
According to the US steel Co. handbook, the AISI has defined a steel to be stainless if it
contains more than 4 wt. % Cr. Most standard types, however, contain rather more
chromium than four % so, for our purposes, we will consider stainless steels to be steels
that contain at least 11. 5 wt. % Cr. They are further classified by being divided into
four main groups, viz. martensitic, ferritic, austenitic and precipitation-hardening
stainless steels. The fourth group contains stainless steels in which strengthening is
provided by precipitates of a second phase. There are minor classifications such as
duplex stainless steels which contain significant quantities of both austenite and ferrite.
The four main types of stainless steel are further identified by a mixture of numerical and
letter codes, see for example the ASM handbook, edition 9, volume 3.
Because the most important alloying element in stainless steel is Cr, it is logical to begin
a discussion of the physical metallurgy of these materials with a consideration of the
binary Fe-Cr system.
R. H. Wagoner MSE 661 – Steel Section 6 Page 24
24 24
As the binary phase diagram, Figure 6-8, shows the introduction of the ferrite stabilizing
Cr forms a gamma loop in which resides those combinations of temperature and
composition for which austenite is the stable phase.
Figure 6-8: Iron-chromium phase diagram. (after Honeycombe, p. 251)
This phase field is separated from the ferrite phase field by the two phase field. When the
Cr content exceeds 12%, it is possible for ferrite to exist at all temperatures at which the
binary alloys are solid. The binary system also contains the intermetallic, Fe-Cr, sigma
phase. This is tetragonal and has a unit cell that contains 30 atoms. This phase diagram
shows that for some compositions, the solid solutions of the alloy can be transformed into
austenite by heating them into the gamma loop. It also shows that when the austenite is
cooled it must transform to ferrite to attain equilibrium. We infer from this that if the
cooling process is conducted in a manner that precludes this transformation from
occurring by diffusion controlled nucleation and growth processes then a martensitic
transformation is possible. Hence, the existence of martensitic stainless steels.
The effect of other various elements on the gamma loop are shown in Figure 6-9.
Stainless steels are formulated primarily to resist corrosion (hence “stainless”) by the the
addition of significant quantities of chromium. The desirable concentration of chromium
is related to how corrosion-resistant the alloy is intended to be and the intended
environment in which it will be used. There are three broad divisions of chromium
2000
1800
1600
1400
1200
1000
600
Fe 10 20 30 40 50 60 70 80 90 100
Chromium (wt %)
T
e
m
p
e
r
a
t
u
r
e
,

C
1800 C
L
820 C
910 C
1505 C
1539 C
γγγ γ
ααα α
ααα α +++ + σσσ σ ααα α +++ + σσσ σ σσσ σ
ααα α +++ + γγγ γ
ααα α +++ +
ααα α +++ +
850 C
R. H. Wagoner MSE 661 – Steel Section 6 Page 25
25 25
content that can be delineated. In general, it is thought that a concentration of 10-12% Cr
is sufficient for normal atmospheric conditions at normal temperatures. A higher
concentration of approximately 18-20% is sufficient for the modest acidity and
temperatures usually encountered in food processing, or for quite high temperature and
normal atmospheric exposure. Higher concentrations of chromium are required in special
alloys intended for very aggressive environments and much higher temperatures.
We consider the broad classes of stainless steels in no particularly logical order.
Figure 6-9: Gamma loops formed in various binary systems of iron. (after Krauss, p.
353.)
C
Ni
Mn
Cr
W
Mo Si
V
Al
C
Mn
Ni
3000
2600
2200
1800
1400
1000
0 2 4 6 8 10 12
Alloy addition, weight percent
T
e
m
p
e
r
a
t
u
r
e
,

F
R. H. Wagoner MSE 661 – Steel Section 6 Page 26
26 26
Martensitic Stainless Steels
Martensitic stainless steels are used in the quenched and tempered condition. The term
quenched is somewhat of a misnomer because with such large alloying element contents
parts made from these steels are often air-hardenable, i. e. the martensite forms when the
part is removed from the austenitizing furnace and simply allowed to cool down in still
air. Quenching in oil is sometimes used for thick sections. In either case, tempering is
performed soon after the martensite has formed, especially when there is significant
carbon present, in order to avoid delayed cracking. The binary Fe-Cr diagram indicates
the compositional limits of those alloys that can be rendered into martensite. Because the
Cr is to be kept as high as possible for anti-corrosion purposes, austenite stabilizers such
as carbon, manganese and also nickel are added to expand the austenite phase field.
These additions enable martensite to be produced in 11. 5 wt. % Cr steels. As was true
for the martensitic steels considered previously, the carbon dissolved in the martensite
has a significant effect on its hardness.
When the carbon content becomes significant, the formation of chromium carbides
becomes possible and the pertinent phase diagram becomes the ternary Fe-Cr-C diagram.
An example of a section at 0. 1 wt. % C is shown below. The complexity is obvious.
Two important additional things can be recognized. First, a variety of chromium carbides
can form. Also, certain heat treatments can form mixtures of ferrite and austenite in
which cooling could convert the austenite to martensite. Thus we can have fully
martensitic microstructures and ferrite/martensite microstructures. This implies that the
classification of a particular steel as martensitic or ferritic cannot be done solely on
composition but must also recognize the microstructure existing during use.
Figures 6-10 and 6-11: Complex phase stability of Fe-Cr-C.
The martensitic steels are given 4xx numbers in the AISI numbering system, as
summarized in Table 6-1. An example is 410 which has 0. 15 % C, 1% Mn, 1 % Si, 11.
5-13. 5 % Cr. However, the ferritic steels which as we will soon see are mainly ferrite,
are also described with 4xx numbers. 430 designates a ferritic steel while 431 refers to a
martensitic steel. It is not, therefore, immediately obvious that a 4xx steel is martensitic
or ferritic.
A
r
b
i
t
r
a
r
y

T
e
m
p
e
r
a
t
u
r
e

(
d
e
g
.

C
) L+ααα α
ααα α+++ +γγγ γ+++ +C1
ααα α+C
c
ααα α+
C2
ααα α+C1+C2
ααα α+C1
ααα α+C1
t. %C
Fe+11%C
γγγ γ
ααα α+++ +γγγ γ
ααα α+++ +γγγ γ
1600
1500
400
1300
200
1100
1000
00
00
00
0 10 20 0
L
L +
ααα α + K1
K
0
=M
3
C
K1=M23C6
K2=M7C3
hromium (%)
T
e
m
p
e
r
a
t
u
r
e

(
C
)
ααα α + K
0
ααα α + K2
ααα α + K1 + K2
γγγ γ
ααα α +++ + γγγ γ
ααα α
Cc = Fe3C
C1 = (Fe,Cr)23C6
C2 = (Fe,Cr)7C3
Fe-11%Cr section of the ternary diagram
e-0.1%C section of the ternary diagram
R. H. Wagoner MSE 661 – Steel Section 6 Page 27
27 27
Table 6-1: Compositions of selected Type 400 Martensitic Stainless Steels. (After
Krauss, p. 381)
AISI ---------------------Nominal composition, %----------------------
type no. C Mn Cr Ni Others
403 0. 15 max 1. 0 11. 5-13
410 0. 15 max 1. 0 11. 5-13
416 0. 15 max 1. 2 12-14 … 0. 15S min
420 0. 15 min 1. 0 12-14
431 0. 20 max 1. 0 15-17 1. 2-2. 5
440A 0. 60-0. 7 5 1. 0 16-18 … 0. 75 Mo max
440B 0. 75-0. 9 5 1. 0 16-18 … 0. 75 Mo max
440C 0. 95-1. 2 0 1. 0 16-18 … 0. 75 Mo max
Let us consider the main roles of alloying elements in martensitic stainless steels with
respect to properties other than corrosion resistance:
• The extent of the austenite phase field is usually controlled by the competition
between the austenite stabilizers, Mn, C, N and Ni and the ferrite stabilizer Cr.
• During the decomposition of the austenite by diffusion controlled processes, the
high concentration of Cr strongly retards the transformations and moves the nose of
the TTT diagram to the right, increasing the time for the start of the transformations
to over 300 seconds, Figure 6-12. This endows the steels with high hardenability.
Indeed, many sections are air-hardenable. Another consequence of the sluggish
rates of the diffusion controlled transformations is that martempering processes are
readily conducted.
• The carbon endows the martensite with hardness and strength. The Cr contributes
additional hardening.
• In a similar manner to its effects on the diffusion controlled reactions involved in
the transformation of austenite, Cr strongly retards the rates of the processes that
occur during tempering. One consequence of this is that stress relief can occur and
residual stresses can be removed during tempering, before serious softening takes
place. This is obviously of practical use.
The tensile yield stress and the tensile ultimate stress for samples of a Type 410
martensitic stainless steel are depicted as a function of tempering temperature in Figure
6-13. It shows that secondary hardening due to the precipitation of chromium carbides
can occur. The tempering temperatures were one hour in each case. These are fine
properties and some enhancement is possible through additional alloying. Unfortunately,
tempering in the vicinity of 885 °F (475°C), which corresponds to the initiation of the
mild secondary hardening peak, produces hardening but reduces the impact toughness
severely as is shown in the corresponding sketch. This embrittlement appears to be
related to the beginning of the secondary hardening process in which sheets of carbides
precipitate on boundaries. Besides this embrittlement problem, the DBTT's of these
R. H. Wagoner MSE 661 – Steel Section 6 Page 28
28 28
steels are around or barely below room temperature which means that they are not useful
in cryogenic applications.
Figure 6-12: Isothermal TTT diagram for Type 410 stainless steel, austenitized at 980
deg. C (1800 deg. F) (After ASM Stainless Steels, p. 17.)
1 10 100
10
3
10
4
10
5
10
6
45 HRC
A + F + C
F + C
A + F
A
A
M
s
M
50
M
90
Approximate
A
start
1 min 1 h 1 day
1
75 HRB
81 HRB
87 HRB
90 HRB
H
a
r
d
n
e
s
s
Time, s
800
(1470)
600
(1110)
400
200
0
(30)
T
e
m
p
e
r
a
t
u
r
e
,

C

(
F
)
50%
(0.5)
Type 410
R. H. Wagoner MSE 661 – Steel Section 6 Page 29
29 29
The effect of austenitization temperature is also shown in Figure 6-13 by comparing the
left hand side (T
γ
=925 deg. C) and the right hand side (T
γ
=1010 deg. C).
Figure 6-13: Effect of tempering on mechanical properties of Type 410.
Brinell hardness Brinell hardness
Tensile strength
0.2% yield strength
Proportional limit
Reduction of area
Elongation in 2 in.
Izod Izod
El in 2 in.
RA
PL
0.2% YS
TS
0
400
(205)
800
(425)
1200
0 400
(205)
800
(425)
1200
(650)
Tempering temperature, F (C)
ΤΤΤ Τ
γγγ γ
= 925 deg. C
600
500
400
300
200
280
240
200
160
120
80
40
80
60
40
20
0
120
80
40
(55)
0
B
r
i
n
e
l
l

h
a
r
d
n
e
s
s
k
s
i

(
M
P
a
)
P
e
r
c
e
n
t
E
n
e
r
g
y
,

f
t
-
l
b

(
J
)
Type 410
Tempering temperature, F (C)
ΤΤΤ Τ
γγγ γ
= 1010 deg. C
Type 410
R. H. Wagoner MSE 661 – Steel Section 6 Page 30
30 30
Ferritic Stainless Steels
In the classical sense, ferritic stainless steels have been considered to be those that
possess Cr contents that would place them to the right of the gamma loop on the binary
Fe Cr phase diagram. The binary alloys of that composition would, indeed, be ferritic to
the melting point. However, this ignores the effects of carbon and, as the Fe Cr 0. 1 C
diagram shows, mixtures of austenite and ferrite can form in these steels at high
temperatures. Because of the high hardenability, the austenite can transform to
martensite during cooling. Most of the commonly designated ferritic steels can behave in
this manner, particularly those containing 17-18 % Cr. From the practical point of view,
a better definition of a ferritic stainless steel is a steel that contains more than 12 % Cr
and is used in a state in which its microstructure is mainly ferrite plus carbides.
Common grades of ferritic stainless steels and their compositions are shown in Table 6-2.
Table 6-2: Compositions of AISI Type 400 Ferritic Stainless Steels.
AISI ----------------Nominal Composition, %----------------
type no. C Mn Cr Others
430 0. 08 max 1 16. 0-18. 0
430F. 0. 12 max 1. 25 16. 0-18. 0 0. 6Mo max
430F Se 0. 12 max 1. 25 16. 0-18. 0 0. 15Se min
446 0. 20 max 1. 5 23. 0-27. 0 0. 25N max
Ferritic stainless steels have very good corrosion resistance in many environments. In
particular, they are generally not susceptible to stress corrosion cracking, which austenitic
stainless steels are. However, their mechanical properties are not very good. Being BCC
(which are similar to martensitic grades that are not quenched, Fig. 6-14) they have a
DBTT (ductile-to-brittle transition temperature). In the common cases where the steel
contains N, P, S and C, the DBTT can be higher than room temperature. AISI 446 (27 %
Cr ) has a DBTT around 150°C. Very pure steels have much better toughness but most
normal, commercial alloys have a transition temperature near room temperature.
Furthermore, because the grain size cannot be refined by heat treatment, large grains
produced in the heat affected zones at welds can reduce the local strength. For these
reasons, ferritic stainless steels have not been very popular for structural applications. An
exception is 405 (11. 5-14. 5 % Cr, 0. 08 % C, 1 % Mn, 1 % Si, 0,04 % P 0,03 % S, 0. 1-
0. 3 % Al) has been used in the petroleum industry.
The inherent problems of the BCC structure and irreversible grain coarsening are
exacerbated by the existence of three forms of embrittlement that can arise during
particular heat treatments.
Sigma phase embrittlement: When ferritic stainless steels are aged for very long times
between 540-750°C the sigma intermetallic phase is precipitated and the steel becomes
brittle. The embrittlement is removed by heating the steel above 800 °C where the sigma
phase is dissolved. This remedy is of little use if the heat treatment cannot be done.
Fortunately, the kinetics of the precipitation process is very slow; several hundred hours
R. H. Wagoner MSE 661 – Steel Section 6 Page 31
31 31
are necessary for the embrittlement to occur. Hence it doesn't happen during the usual
times of cooling or welding. It is important if the component is intended to operate in the
relevant temperature range.
Figure 6-14: Toughness vs. test temperatures of tempered of quenched Type 410
stainless steel. (After ASM Stainless Steels, p. 39.)
475 °°° °C or 885 °°° °F embrittlement: This is a loss of impact toughness that can be caused
by holding the steel at temperatures between 400°C and 540 °C . The embrittlement is
associated with the precipitation of a coherent chromium-rich phase throughout the
grains. It can be cured by reheating the steel above 800 °C and cooling. Quenching is
unnecessary because this form of embrittlement requires long times to form.

High temperature embrittlement: When ferritic steels containing the usual contents of
C, N, P and S are heated above 1000 °C and then cooled naturally as for example during
welding, they can become severely embrittled and also lose their corrosion resistance
near grain boundaries. The phenomenon is similar to sensitization in austenitic steels( to
be considered later) and is considered to be caused by precipitation of Cr carbides and
nitrides on grain boundaries.
-200 0 200 400 600
125
100
75
50
25
0
400 200 0 -200
0
50
100
150
200
Type 410
A
B
C
Temperature, F
I
m
p
a
c
t

e
n
e
r
g
y
,

f
t
-
l
b
f
Temperature, C
C
h
a
r
p
y

I
m
p
a
c
t

e
n
e
r
g
y
,

J
T
=
6
6
5

C
,

R
C
=
2
4
T
=
5
9
5

C
.

R
C
=
3
0
Tempering
Temperature,
Hardness:
T
=
7
9
0

C
,

R
B
=
9
5
R. H. Wagoner MSE 661 – Steel Section 6 Page 32
32 32
The first two forms of embrittlement are of no great concern unless the steel can
experience them during use. They do not arise otherwise and do not occur during
welding because of the very sluggish rates at which the transformations occur. In
contrast, however, the high temperature embrittlement has been a very serious problem
for years. It has been of particular concern with respect to welding. In this case we
already have the problem that there is no way to refine the grain size if it becomes large
in the heat affected zone. This, combined with the embrittlement and the localized loss
of corrosion resistance, has been very troublesome.
There have been three main methods used to combat this problem. The first is to pay
stringent attention to the steel making process in order to ensure a very low interstitial
solute content. The second is to add small amounts of very strong carbide and nitride
formers e. g. V, Nb and Ti to scavenge these elements. These two methods can be
successful . However, as the Cr content increases so does the minimum acceptable level
of interstitial solutes. At 35 % Cr, the interstitial level can not be reduced low enough by
these methods. In these cases the third method is via the addition of other substitutional
solutes. It has been found that small additions of Al, Cu, Pt, Pd, Ag, V in the range of 0.
1 -1. 3 % can improve the as-welded ductility and corrosion even in the presence of
relatively high quantities of interstitial solutes.
Austenitic Stainless Steels
This group of steels are iron-chromium alloys to which enough Ni and Mn have been
added to prevent the formation of significant amounts of ferrite. Typical composition and
grades are shown in Tables 6-3 and 6-4. In fact, it is the balance between gamma-
stabilizers and alpha-stabilizers that determine to a large extent the important chemistry
of stainless steels. For this reason, alloying elements are often classified in terms of
chromium equivalents and nickel equivalents on this basis, and ranges of these two
quantities typically determine the type and usage of the resulting alloy, Figure 6-15. The
rules for chromium equivalence and nickel equivalence are as follows:
Schaeffler diagram. Effect of alloying elements on the basic structure of Cr-Ni stainless
steels (Schneider and Climax Molybdenum Co. , Foundry Trade J. 108,562,1960).
Cr equivalent =
(Cr) + 2(Si) + 1. 5(Mo) + 5(V) + 5. 5(Al) + 1. 75(Nb) +1. 5(Ti) + 0. 75(W) (6-1)
Ni equivalent = (Ni) + (Co) + 0. 5(Mn) + 0. 3(Cu) + 25(N) + 30(C) (6-2)
Different alloys comprise austenites of different stability as reflected by them having
different M
s
temperatures. Accordingly, they can be subdivided into two groups
depending on whether or not the austenite can be induced to transform martensitically
when deformed at room temperature, i. e. when the M
d
temperature is above room
temperature. In these groups we have stable austenite which does not transform and
metastable austenite which does form some martensite.
R. H. Wagoner MSE 661 – Steel Section 6 Page 33
33 33
Figure 6-15: Effect of alloying elements on the chemical structure of stainless steels, the
Schaeffler Diagram. (After Schneider and Climax Molybdenum Co., Foundry Trade J.,
1960. 108, p. 562.)
Table 6-3: Compositions of AISI Type 200 Austenitic Stainless Steels.
(After Krauss, p. 372.)
AISI ---------------Nominal Composition, %---------------
type no. C Mn Cr Others
201 0. 15 max 7. 5 16-18
202 0. 15 max 10. 0 17-19 0. 6Mo max
Table 6-4: Compositions of Selected AISI Type 300 Austenitic Stainless Steels.
(After Krauss, P. 359.)
AISI ---------------Nominal Composition, %---------------
type no. C Mn Cr Ni Others
301 0. 15 max 2. 0 16-18 6. 0-8. 0
302 0. 15 max 2. 0 17-19 8. 0-10
304 0. 08 max 2. 0 18-20 8. 0-12
304L 0. 03 max 2. 0 18-20 8. 0-12
309 0. 20 max 2. 0 22-24 12-15
310 0. 25 max 2. 0 24-26 19-22
316 0. 08 max 2. 0 16-18 10-14 2-3 Mo
316L 0. 03 max 2. 0 16-18 10-14 2-3 Mo
321 0. 08 max 2. 0 17-19 9-12 (5x%C) Ti min
347 0. 08 max 2. 0 17-19 9-13 (10x%C) Nb - Ta min
25
20
15
10
5
0
0 5 10 15 20
25 30 35
40
Chromium equivalent (%)
N
i
c
k
e
l

e
q
u
i
v
a
l
e
n
t

(
%
)
Austenite
Austenite
+
δδδ δ - ferrite
δδδ δ -ferrite
Austenite
+martensite
+ δδδ δ ferrite
Martensite
+ δδδ δ ferrite
Martensite
Austenite +
martensite
R. H. Wagoner MSE 661 – Steel Section 6 Page 34
34 34
Consider the room temperature tensile stress-strain curves that are typical for tests on the
two austenitic steels, 301 and 304, Figure 6-16. That of 304 shows the "normal"
behavior of an FCC polycrystal while that of the 301 has an increased work hardening
rate caused by the impedance to slip provided by the formation of martensite on the
active slip planes. The compositions are shown in Table 6-4
The M
d
temperature for a 304 steel is usually just below room temperature. This means
that if the steel is refrigerated and then deformed, martensite can form in 304 and change
the stress-strain curve accordingly.
A major reason for alloying so as to create an austenitic steel is to take advantage of the
greater ductility offered by the fcc crystal structure, and to eliminate the DBTT, thus
obtaining good impact toughness over a wide range of temperatures. These steels are
generally of good toughness and good corrosion resistance. Their tensile strengths can be
varied by varying the grain size and the dislocation density. However, there are a couple
of negative aspects that must be recognized (beyond cost, which will be a decided
disadvantage because of the high cost of nickel).
Figure 6-16: High strain-hardening rate for unstable Type 301 versus stable Type 304,
with measured martenite percentage. (After V. Talyan, R. H. Wagoner, J. K. Lee:
Metall. Mater. Trans. A, 1998, vol. 29A, pp. 2161-2171.)
1000
800
600
400
200
0
0 0.2 0.4 0.6 0.8
0
30
60
90
120
150
εεε ε === = 10
-3
/s
301
304L
301
Engineering strain
E
n
g
i
n
e
e
r
i
n
g

s
t
r
e
s
s

(
M
P
a
)
%

M
a
r
t
e
n
s
i
t
e
% Martensite
Stress
R. H. Wagoner MSE 661 – Steel Section 6 Page 35
35 35
Sensitization: When held at high temperatures and then slowly cooled through 600 °C
these alloys can lose their corrosion resistance in the near vicinity of grain boundaries and
become less tough. Because the loss of corrosion resistance is restricted to narrow strips
at the grain boundaries, deep grooves are formed. Whole grains can "fall out ". A similar
phenomenon can occur in ferritic steels but it appears to have been more troublesome for
austenitic steels. The treatment is said to have sensitized the steel. It has been
established that the loss of corrosion resistance is associated with the precipitation of
chromium carbides on the grain boundaries. As the temperature falls past 600 °C , these
carbides nucleate heterogeneously on grain boundaries and grow for a while as the steel
cools further. The chromium required for this growth must be supplied by substitutional
diffusion which occurs slowly and more slowly as the temperature decreases. As a result,
the chromium concentration near the boundaries on which the carbides grow is severely
depleted. This has been confirmed by modern electron micro-analytical techniques.
Hence a grain boundary has, on either side of it, a thin layer of steel in which the
chromium content is not high enough to provide protection. Severe localized corrosion is
the result.
Attempts to cure this problem when it is not easily possible to avoid the heat treatment
that causes it, involve compositional modifications. The required modification is
obvious. Eliminate the carbon and no carbides will precipitate. Elimination is not
presently feasible but a reduction is. For example, modified 304 called 304L has 0. 03 %
C max. specified. An alternative approach is to try to combine the carbon into
precipitates that do not easily dissolve at high (>950 °C) temperatures. For this purpose
small additions of Nb, V and Ti are added to sequester the carbon. Obviously these two
approaches can be combined. These modifications are similar to those employed in the
case of ferritic steel.
Stress corrosion cracking: Stress corrosion cracking is a process in which cracks form
and grow into a material that is simultaneously experiencing a stress and is in an
appropriate corrosive environment. The stress need not be applied externally; residual
stresses due to prior processes can be effective. The corrosive media are generally
specific for the particular material. They are not necessarily particularly aggressive from
the point of view of general dissolution and corrosion. Austenitic stainless steel is
particularly susceptible to aqueous media that contain chlorine ions. In general, the
process can take place at ambient temperatures and will occur more rapidly as the
temperatures are raised. The cracks tend to branch frequently and propagate through the
grains. Eventually sheets can be cracked completely through. Stress corrosion cracking
is a serious problem for austenitic steels but ferritic stainless steels are immune to it.
To avoid this problem when it is desirable to use austenitic stainless steel, it is necessary
to minimize residual stresses and to ensure that the environment does not contain those
species such as chlorine that are known to be particularly potent for stress corrosion
cracking. Often this is not easy to do. Although the phenomenon has been known for
many years, austenitic steel components (as well those made from numerous other alloys)
continue to fail in the field from stress corrosion cracking.
R. H. Wagoner MSE 661 – Steel Section 6 Page 36
36 36
Precipitation Hardenable Stainless Steels
Under appropriate conditions it is possible to produce precipitation of intermetallic
compounds that can cause precipitation hardening in stainless steels. The precipitates
involve the alloying elements Mo, Ni, Ti and Al. There are three main classes of
precipitation hardenable alloys : austenitic PH steels, martensitic PH steels and semi-
austenitic PH steels, Table 6-5.
Table 6-5: Compositions of Selected Precipitation-Hardening Stainless Steels.
(After Krauss, p. 236.)
AISI Nominal Composition, %
type no. Trade name C Cr Ni Mo Cu Al Ti Other
Martensitic grades
635 Stainless W
(U. S. Steel) 0. 06 16. 75 6. 25 0. 2 0. 8
630 17. 4 PH
(Armco Steel) 0. 04 1 6 4. 2 3. 4 0. 25Cb
…. . PH 13-8 Mo
(Armco Steel) 0. 04 12. 7 8. 2 2. 2 1. 1
Semiaustenitic grades
631 17. 7 PH
(Armco Steel) 0. 07 1 7 7. 1 1. 2
633 AM 350
(Allegheny Ludlum) 0. 05 16. 5 4. 25 2. 75 0. 10N
Austenitic grades
600 A-286
(Allegheny Ludlum) 0. 05 14. 75 25. 25 1. 3 0. 15 2. 15 0. 30V, 0. 005B
Members of the first group have compositions that produce very stable austenite with
respect to martensite. An example is Allegheny-Ludlum ‘s A-286: 14. 75 % Cr, 25. 25 %
Ni, 0. 05 % C, 1. 45 % Mn, 1. 3 % Mo, 0. 15 % Al, 0. 3 % V, 2. 15 % Ti . The heat
treatment is relatively simple: solution treat for 1 hour at 1000 Celsius, water quench and
age at 730 Celsius. Yield strengths of 190 KSI can be obtained but these type of steels
are not widely used.
The martensitic precipitation hardenable stainless steels are similar to maraging steels.
An example is the PH 13-8 Mo type. Here the PH indicates precipitation hardenable, the
13 denotes the chromium content (12. 7 %), the 8 the nickel content (8 % ) and the Mo
indicates that it contains molybdenum (2. 2 %). The alloy has a low carbon content and
contains some aluminum. In this case the solution treatment produces homogeneous
austenite that transforms to martensite on cooling. Precipitation occurs when the
martensite is aged at 450-650 Celsius.
The semi-austenitic precipitation hardenable stainless steels are so named because in the
annealed or solution treated condition they generally contain 5-20 % delta ferrite which
persists. The metallurgy and heat treatment schedules for these steels are quite complex.
The general procedure is as follows. After production the steel annealed at the Mill
(called the Mill anneal) which puts the steel into its semi-austenitic state. It is then
R. H. Wagoner MSE 661 – Steel Section 6 Page 37
37 37
shipped to the customer who fabricates it into particular products. These products are
then given a heat treatment known as conditioning. Conditioning involves heating to a
particular temperature and maintaining the temperature for a specified time and air -
cooling. The object of the conditioning treatment is to remove the microstructural effects
of the fabrication process e. g. forming deformation and to adjust the composition of
the matrix solid solution so that its M
s
temperature is raised to close to room temperature
from its earlier lower value. This is typically achieved by precipitating chromium and
carbon from solid solution. Conditioning enables the steel to be transformed into
martensite by refrigeration. Finally, the martensite can be precipitation hardened by an
aging treatment.
The chemistry of these steels is chosen with the aim of balancing the ferrite and austenite
stabilizing elements so as to control the M
s
temperature. The steel must not be too stable
as austenite because then the M
s
temperature will be too low for convenience. On the
other hand, if the M
s
temperature in the steel after the mill anneal is too high, martensite
might be produced during transit, specially during winter. An example of such a steel is
Armco 17-7 PH, 17 % Cr, 7% Ni, 1. 2 % Al, 0. 5 % Mn, 0. 04 % N, 0. 3 % Si and 0. 07
% C for which a mill anneal is, using Celsius, 1053 + or - 15; conditioning is at 953 + or -
8 for 10 minutes and an air-cool. The martensite is produced by refrigeration to -73 and
precipitation occurs during an hour at 510°C.
R. H. Wagoner MSE 661 – Steel Section 6 Page 38
38 38
Tool Steels
Tool steels are obviously those steels from which tools are made. There are many such
steels ranging from relatively simple plain carbon steels to complex, highly alloyed
steels. This is a reflection of the enormous variety of operations that are performed by
tools. To begin a survey of tool steels, let us first consider the range of operations in
which tools are used. We begin with procedures in which items or other material are
moved. For these, we require tools such as screw drivers, wrenches, pliers, spades, post-
hole diggers, bull-dozer blades etc. . Then there are numerous operations in which
materials are cut or fractured; for these we require knives, axes, saws, scissors, mower
blades, farming implements, chisels, razor blades and jack-hammers and tools for
drilling, broaching, reaming, milling and lathes. Then there are the processes in which
materials are formed into different shapes by deformation in hot or cold working
operations such as stamping, forging, rolling and extrusion. These processes involve the
forming machines themselves and various cold and hot work dies that interact directly
with the work pieces. And, lest we forget, we need tools with which to make tools.
It is clear that there is an enormous range of applications for tool steels. For some tools,
such as a chisel, a simple plain carbon steel can be perfectly adequate. In other cases
such as high speed machining, during which the cutting tool can become red hot, a much
more sophisticated steel is necessary. A factor that is common to tools is that they are
used on some type of work piece. Usually the shape or form of the work-piece is
changed by plastic deformation, material removal or fracture. Therefore, there is a
common requirement for tools. The work-piece should have a minimum effect on the
form of the tool in order for the tool to do its job successfully and for it to have a
reasonable lifetime. With this background in mind let us list those properties of materials
that should be important with respect to selecting or designing a material for a tool.
(1) Strength. In general the tool should be stronger and harder than the work piece.
(2) Edge retention. In many applications the tool material must be capable of being
sharpened and then be able to resist becoming blunt rapidly.
(3) Wear resistance. Wear removes material from tools and thereby changes their shapes.
This property often determines the useful lifetime of the tool.
(4) Shock resistance. In many operations tools are loaded rapidly and must therefore be
capable of sustaining stresses created during shock loading.
(5) High temperature stability. During operations conducted at elevated temperatures
such as hot forming and high speed cutting it is important that the tool material retain its
properties at those temperatures rather than rapidly undergoing microstructural changes
that cause undesirable changes in its properties.
Now let suppose that we have materials that possess some or all of these properties. It is
then necessary that tools can be manufactured from them. Obviously, tools made of a
material with superior properties can be used to make tools from materials of inferior
properties (referring, of course, to relevant properties), but how are the first kind of tools
made? Diamond is an example of a material that has excellent properties from the view
point of a variety of tools. For example, it is extremely hard. It is used in a number of
R. H. Wagoner MSE 661 – Steel Section 6 Page 39
39 39
abrasive tool applications but it does not lend itself easily to being manufactured into
monolithic tools. In contemplating this situation, the value of the fact that the properties
of steels can be changed by heat treatments becomes evident. This enables tools to be
fabricated into shapes close to their final physical form from steels that are initially easily
formed and/or machined. Then when they have been made, they can be heat treated in
order to endow them with the properties required for operation as a tool. It is this
versatility that accounts for the wide use of steel to make tools. There are many different
kinds of tool steel. This is partly because of the many different applications of tools. For
any particular application, the steel chosen should have adequate properties to perform
properly and also be an economical choice. Usually the steel is delivered in the annealed
or spheroidized condition; the customer is responsible for subsequent heat treatments.
Inspection of the list of desirable properties leads us to conclude that in many cases the
heat treatment will involve austenitization, cooling rapidly enough to form martensite and
then a tempering treatment. Often, then, the tool or the working parts of it will consist of
tempered martensite. When the desired properties can be obtained by using plain carbon
steels then they are the most economical choice. Otherwise alloy steels are required. Let
us consider the main alloying elements that we might predict to be important in tool
steels.
a) Hardenability is an important consideration because it controls the sizes that can be
heat treated throughout or the depth of case hardening. The hardenability also dictates
the necessary quench rates. This is particularly important because the incidence of
quench cracking, distortion and dimensional change can be minimized by reducing the
rate of cooling required to produce martensite. This suggests the use of Mo, Cr and Mn.
b) A high carbon concentration dissolved in the austenite is required to obtain tempered
martensite that has a high hardness. Furthermore, wear resistance at the surface of the
tool is enhanced by the presence of hard particles of second phases. The obvious second
phase is a carbide. The stable alloy carbides coarsen more slowly than does cementite
and are therefore more effective than cementite at higher temperatures. We predict the
use of carbide formers such as Cr, Mo, W and V and sufficient carbon to provide these
alloy carbides while also endowing the martensite with enough hardness. When the most
important role of the alloy carbides require that they be finely distributed within the steel,
then they should precipitate during tempering as opposed to being inherited from the
austenite. Therefore, it is to be expected that heat treatments involving high
austenitization temperatures and long times will be required to form austenite in which
the appropriate amount of the relevant alloying elements is dissolved. This creates the
risk of excessive grain growth during austenitization. As always, a favorite hedge against
that problem is grain boundary pinning by very strong carbides such as vanadium
carbide.
c) Very often it is the properties of the material at the surface of the tool that determine
the performance of the tool. These can deteriorate rapidly unless the material resists
oxidation and decarburization. In this respect, we predict that Cr should be an effective
alloying element.
R. H. Wagoner MSE 661 – Steel Section 6 Page 40
40 40
We conclude that Cr, Mo, W, V, C and Mn should be important alloying elements in tool
steels.
There is a large number of different tool steels. At first, one might attempt to organize
them by classifying them chemistry or by their intended application. Unfortunately, there
does not seem to be an easy method whereby they can neatly classified. Often many
different steels can be used for the same application, so a classification in terms of use is
not clear. Although, some steels can be grouped together according to their
compositions, many steels contain the same alloying elements and, again, it is often
possible for a major element in one steel to be partially or entirely
R. H. Wagoner MSE 661 – Steel Section 6 Page 41
41 41
AISI classification and nominal compositions of major tool steels
Identifying elements, %
UNS
AISI No. C Mn Si Cr V W Mo Co Ni
Water-hardening tool steels
W1 T72301 0. 60-1. 40(a) … … … … … … … …
W2 T72302 0. 60-1. 40(a) … … … 0. 25 … … … …
W5 T72305 1. 10 … … 0. 50 … … … … …
Shock-resisting tool steels
S1 T41901 0. 50 … … 1. 50 … 2. 50 … … …
S2 T41902 0. 50 … 1. 00 … … … 0. 50 … …
S5 T41905 0. 55 0. 80 2. 00 … … … 0. 40 … …
S6 T41906 0. 45 1. 40 2. 25 1. 50 … … 0. 40 … …
S7 T41907 0. 50 … … 3. 25 … … 1. 40 … …
Oil-hardening cold-work tool steels
O1 T31501 0. 90 1. 00 … 0. 50 … 0. 50 … … …
O2 T31502 0. 90 1. 60 … … … … … … …
O6(b) T31506 1. 45 0. 80 1. 00 … … … 0. 25 … …
O7 T31507 1. 20 … … 0. 75 … 1. 75 … … …
Air-hardening, medium alloy cold-work tool steels
A2 T30102 1. 00 … … 5. 00 … … 1. 00 … …
A3 T30103 1. 25 … … 5. 00 1. 00 … 1. 00 … …
A4 T30104 1. 00 2. 00 … 1. 00 … … 1. 00 … …
A6 T30106 0. 70 2. 00 … 1. 00 … … 1. 25 … …
A7 T30107 2. 25 … … 5. 25 4. 75 1. 00 (c) 1. 00 … …
A8 T30108 0. 55 … … 5. 00 … 1. 25 1. 25 … …
A9 T30109 0. 50 … … 5. 00 1. 00 … 1. 40 … 1. 50
A10(b) T30110 1. 35 1. 80 1. 25 … … … 1. 50 … 1. 80
High-carbon, high-chromium cold-work steels
D2 T30402 1. 50 … … 12. 00 1. 00 … 1. 00 … …
D3 T30403 2. 25 … … 12. 00 … … … … …
D4 T30404 2. 25 … … 12. 00 … … 1. 00 … …
D5 T30405 1. 50 … … 12. 00 … … 1. 00 3. 00 …
D7 T30407 2. 35 … … 12. 00 4. 00 … 1. 00 … …
Low-alloy special-purpose tool steels
L2 T61202 0. 5-1. 10 (a) … … 1. 00 0. 20 … … … …
L6 T61206 0. 70 … … 0. 75 … … 0. 25 (c) … 1. 50
Mold steels
P2 T51602 0. 07 … … 2. 00 … … 0. 20 … 0. 50
P3 T51603 0. 10 … … 0. 60 … … … … 1. 25
P4 T51604 0. 07 … … 5. 00 … … 0. 75 … …
P5 T51605 0. 10 … … 2. 25 … … … … …
P6 T51606 0. 10 … … 1. 50 … … … … 3. 50
P20 T51620 0. 35 … … 1. 70 … … 0. 40 … …
P21 T51621 0. 20 1. 20 (Al) … … … … … … 4. 00
Chromium hot-work tool steels
H10 T20810 0. 40 … … 3. 25 0. 40 … 2. 50 … …
H11 T20811 0. 35 … … 5. 00 0. 40 … 1. 50 … …
H12 T20812 0. 35 … … 5. 00 0. 40 1. 50 1. 50 … …
H13 T20813 0. 35 … … 5. 00 1. 00 … 1. 50 … …
H14 T20814 0. 40 … … 5. 00 … 5. 00 … … …
H19 T20819 0. 40 … … 4. 25 2. 00 4. 25 … 4. 25 …
Tungsten hot-work tool steels
H21 T20821 0. 35 … … 3. 50 … 9. 00 … … …
H22 T20822 0. 35 … … 2. 00 … 11. 00 … … …
H23 T20823 0. 30 … … 12. 00 … 12. 00 … … …
H24 T20824 0. 45 … … 3. 00 … 15. 00 … … …
(continued)
(a) Available with different carbon contents. (b) Contains graphite. (c) Optional
R. H. Wagoner MSE 661 – Steel Section 6 Page 42
42 42
AISI classification and nominal compositions of major tool steels
Identifying elements, %
UNS
AISI No. C Mn Si Cr V W Mo Co Ni
Tungsten hot-work tool steels (continued)
H25 T20825 0. 25 … … 4. 00 … 15. 00 … … …
H26 T20826 0. 50 … … 4. 00 1. 00 18. 00 … … …
Molybdenum hot-work tool steels
H42 T20842 0. 60 … … 4. 00 2. 00 6. 00 5. 00 … …
Tungsten high-speed tool steels
T1 T12001 0. 75(a) … … 4. 00 1. 00 18. 00 … … …
T2 T12002 0. 80 … … 4. 00 2. 00 18. 00 … … …
T4 T12004 0. 75 … … 4. 00 1. 00 18. 00 … 5. 00 …
T5 T12005 0. 80 … … 4. 00 2. 00 18. 00 … 8. 00 …
T6 T12006 0. 80 … … 4. 50 1. 50 20. 00 … 12. 00 …
T8 T12008 0. 75 … … 4. 00 2. 00 14. 00 … 5. 00 …
T15 T12015 1. 50 … … 4. 00 5. 00 12. 00 … 5. 00 …
Molybdenum high-speed tool steels
M1 T11301 0. 80(a) … … 4. 00 1. 00 1. 50 8. 00 … …
M2 T11302 0. 85-1. 00(a) … … 4. 00 2. 00 6. 00 5. 00 … …
M3, class 1 T11313 1. 05 … … 4. 00 2. 40 6. 00 5. 00 … …
M3, class 2 T11323 1. 20 … … 4. 00 3. 00 6. 00 5. 00 … …
M4 T11304 1. 30 … … 4. 00 4. 00 5. 50 4. 50 … …
M6 T11306 0. 80 … … 4. 00 2. 00 4. 00 5. 00 12. 00 …
M7 T11307 1. 00 … … 4. 00 2. 00 1. 75 8. 75 … …
M10 T11310 0. 85-1. 00(a) … … 4. 00 2. 00 … 8. 00 … …
M30 T11330 0. 8 … … 4. 00 1. 25 2. 00 8. 00 5. 00 …
M33 T11333 0. 90 … … 4. 00 1. 15 1. 50 9. 50 8. 00 …
M34 T11334 0. 90 … … 4. 00 2. 00 2. 00 8. 00 8. 00 …
M36 T11336 0. 80 … … 4. 00 2. 00 6. 00 5. 00 8. 00 …
Ultrahard high-speed tool steels
M41 T11341 1. 10 … … 4. 25 2. 00 6. 75 3. 75 5. 00 …
M42 T11342 1. 10 … … 3. 75 1. 15 1. 50 9. 50 8. 00 …
M43 T11343 1. 20 … … 3. 75 1. 60 2. 75 8. 00 8. 25 …
M44 T11344 1. 15 … … 4. 25 2. 00 5. 25 6. 25 12. 00 …
M46 T11346 1. 25 … … 4. 00 3. 20 2. 00 8. 25 8. 25 …
M47 T11347 1. 10 … … 3. 75 1. 25 1. 50 9. 50 5. 00 …
(a) Available with different carbon contents. (b) Contains grapite. (c) Optional
After: "Tool Steels" by G. Roberts, G. Krauss, and R. Kennedy, ASM
International, 5
th
edition, pp8-9, 1998
R. H. Wagoner MSE 661 – Steel Section 6 Page 43
43 43
replaced by another element without creating much of a change in the properties of the
steel. The AISI classifies tool steels into seven major categories. These are: water-
hardening tool steels, cold work tool steels, hot work tool steels, high speed tool steels,
shock-resisting tool steels, mold tool steels and special purpose tool steels. Here we see a
classification by a mixture of use and heat treatment. These seven major classes are then
subdivided into numbered classes ranging from the 100 class to the 600 class. I can see
no simple way deducing the details of the classification. For selecting and identifying
tool steels one should consult manuals on the subject. A fine reference is the ASM’s
publication “Tool Steels”, 4
th
edition, by George A. Roberts and Robert A. Cary. What
I propose to do here is to select a number of particular tool steels and consider their
compositions, treatments and properties in the light of the fundamental relations between
composition and phase transformations that we have examined hitherto.
Water hardening tool steels: These steels are so named because they have relatively
low hardenability so that they are hardened by being austenitized and quenched in water.
They are basically plain carbon steels and are also called carbon tool steels, They contain
from 0. 6 wt. % to 1. 4 wt. % carbon with a manganese content of usually 0. 25 wt. %.
Some of them contain small amounts of chromium or vanadium and a few contain both
vanadium and chromium and, therefore, the group is sub-divided according to such
details. The additional alloying elements, however, always comprise a fraction of a wt.
%. Examples are:
W 1. . . 0. 6% to 1. 6 % C, 0. 25 % Mn and 0. 25 % Si
W 5. . . 1 % C, 0. 25 % Mn and Si and 0. 5 % Cr
W7. . . . Same as W 5 except that the Cr level is 0. 35 % and 0. 2 % V is present.
There are some applications of carbon tool steels for which they are used when having a
pearlitic microstructure. Most commonly, however, the intended microstructure in the
tools made from these steels is tempered martensite where it is needed and softer, tougher
microstructures elsewhere. Indeed, it is probably true to say that this class of tool steels
is widely used because the relatively low hardenability of the steels enables water-
quenching to produce tools as cores of fine pearlite encased within hard martensite.
Typically, the working hardness in the case is in the range of HRC 58-65 and the core
hardness is in the range of HRC 38-43. The hardness of the tempered martensite is
primarily controlled by the carbon dissolved in the austenite prior to the quench. As we
have seen, the hardness increases approximately linearly with carbon content in freshly
formed martensite until about 0. 6 wt. % C; thereafter it varies slowly and is essentially
constant by 0. 8 %. Because of this, we recognize that hypereutectoid tool steels are not
selected for use because their high carbon contents endow the martensite with additional
hardness but because the volume fraction of carbides in them is greater and this provides
increased wear resistance. Therefore, it is unnecessary for all of the carbon to be
dissolved in the austenite during austenitization.
The full story of the heat treatment of carbon tool steels begins with the forging process
that produces the desired stock e. g. bars. Forging is carried out at temperatures well in
the austenite phase field but below those which would cause dissolution of the vanadium
R. H. Wagoner MSE 661 – Steel Section 6 Page 44
44 44
carbides present in some steels. The steel is air cooled after forging. Then it is annealed,
sometimes after a normalization treatment, to produce a suitable microstructure for
delivery to the customer. The purpose of the annealing treatment is to enhance the
machinability of the steel, eliminate any residual stresses and to provide a microstructure
that will respond uniformly to subsequent hardening heat treatments. The desirable type
of microstructure that achieves this purpose is either a finely spheroidized microstructure
or a mixture of fine spheroidite and pearlite. It is important to avoid the formation of
films of cementite along austenite grain boundaries or a coarse spheroidal microstructure.
These microstructures are produced by heating to and holding at temperatures within the
austenite/cementite two phase field followed by slow cooling
( maximum rate 50°F per hour) to 1000°F. The steel is protected from excessive scaling
and decarburization during annealing, often by being embedded in inert substances such
as sand.
For hardening, these steels are austenitized above the A
1
line but below the A
cm
line.
The microstructure before the quench comprises austenite grains in which proeutectoid
cementite particles and possibly vanadium carbides are embedded. These carbides
restrict grain growth during austenitization and are retained within the martensite when
quenching is complete. The austenitization temperatures are high enough for the Cr be
dissolved. This is desirable partly because it then is helpful in resisting oxidation and
decarburization. However, these degradation processes should also be minimized by
using protective atmospheres or baths during heat treating. Another important
contribution of the dissolved chromium is that it retards the rate of tempering reactions
which enables some residual stress relief to be accomplished without too much softening.
Finally, the chromium that becomes incorporated in the cementite improves the wear
resistance provided by the carbides.
The microstructure in the case of carbon tool steels austenitized as above and then water-
quenched to room temperature is not 100 % martensite even very near the surface.
Retained austenite is usually present because the M
s
temperature drops below room
temperature as the carbon content rises. For example, samples of W1 steel austenitized at
1450°F begin to transform to martensite at approximately 400°F and will continue to
transform, if continuously cooled, until the temperature is about -250°F where there is
typically 1 % retained austenite present. If the quench is to room temperature the
hardness of the mixture of martensite and retained austenite is typically above HRC 65
and roughly 90 % of the microstructure is martensite.
It is time now to elaborate on the martensitic transformation in these steels. When we
first broached the subject of the martensitic transformation in plain carbon and the
standard low alloy steels, I defined it as usually being an athermal transformation during
which the quantity of martensite produced was dependent on the temperature to which
the steel was cooled and not on the time spent at that temperature. This was to emphasize
the difference between its kinetics and those of diffusion controlled nucleation and
growth transformations. For hypoeutectoid plain carbon and low alloy steels this
definition is fine because it is close to the truth and, also, for practical purposes,
quenching to room temperature generally means continuously cooling to below the M
f
R. H. Wagoner MSE 661 – Steel Section 6 Page 45
45 45
temperature. It is, however, a definition that is not exactly true. Let us consider the
behavior of W1 in greater detail. If a sample is quenched to room temperature, the
transformation to martensite is incomplete with typically 12 % of the sample being
retained austenite. If the sample is immediately cooled further, the transformation
continues as if the cooling was continuous. If, however, the sample is held at room
temperature for significant times, some of the austenite does transform. For example,
Cohen in “Retained Austenite”, Trans. ASM, Vol. 41, 35, (1949), shows that another
1% or so of martensite forms in an hour and about 2 or 3 % in 50 hours. During this
isothermal hold, the remaining austenite also becomes more resistant to the athermal
transformation because when cooling is resumed martensite does not form until the
temperature has decreased on the order of a hundred degrees and much less of it forms.
These subtleties become important now. If it is desired to transform as much austenite as
possible then cooling should not be interrupted for significant delays. It is well known
that tool steel articles can suffer delayed cracking if stored for periods of time after being
quenched. Very likely, the volume changes produced by the continuation of the
transformation during storage contributes to this delayed cracking phenomenon. It is
particularly important that such steels be tempered immediately after being quenched.
The phenomenon of delayed cracking was mentioned in the foregoing paragraph. This is
related to quench cracking which refers to cracking that occurs during the quenching
process. The risk of quench cracking increases as the carbon content of the austenite
increases and as the severity of the quench increases. It is also true that quench cracking
becomes more likely to occur as the austenitic grain size increases. This is another
reason that vanadium is added to the higher carbon steels to refine the austenitic grain
size.
Cold Work Die Steels: This is a very important family of tool steels. The steels are
used in a wide variety of applications involving tool and die work. Of these, the
compositions that require quenching in oil are probably the most commonly used but
when larger parts are used or when smaller dimensional changes are desirable during heat
treatment, compositions that enable air hardening to occur are used.
As was summarized by Roberts and Cary, the major properties of these steels are:
ααα α+Fe
3
C
wt. % C
Fe
T
Tough core
Low hardenability
Tempered
martensite
Austenitization temperatures
why intercritical at %C > .8 ?
γ
R. H. Wagoner MSE 661 – Steel Section 6 Page 46
46 46
1) small dimensional changes ( called movement) during hardening, hence small
finishing work required;
2) high as-quenched hardness, due primarily to carbon dissolved in the austenite;
3) high hardenability when austenitized at low austenitization temperatures;
4) reasonably good resistance to quench-cracking in intricate shapes;
5) good retention of a keen cutting edge during use.
These are good properties, however, the steels are limited to cold work because they
soften when heated and have no “red-hardness” i. e. they do not retain hardness when
red-hot.
Examples are:
01 0. 95 % C 1. 2 % Mn 0. 25 % Si 0. 5 % Cr 0. 2 % V 0. 5 % W
07 1. 2 % C 0. 25 % Mn 0. 25 % Si 0. 6 % Cr 0. 2 % V 1. 6 % W 0. 25 % Mo
We see that the amounts of alloying elements are not high and that most of them are
carbide formers and, if dissolved in the austenite, will raise the hardenability. These
steels typically undergo forging processes that are conducted around 1000 to 1100
degrees Celsius and are then annealed around 850 before being austenitized.
Austenitization temperatures are in the range of 840 degrees where some of the alloy
carbides except the vanadium carbides dissolve. The vanadium carbides and the low
austenitization temperatures resist grain growth. When the Mn content is high, the
permissible austenitization temperature is at the lower end of the range because Mn is an
austenite stabilizer. The ability to austenitize the tool at relatively low temperatures
reduces the dimensional changes that occur subsequently. Any elastic shape changes that
occur because of differential expansion/contraction while the temperature is changing,
are eliminated when the temperature of the tool is uniform. If the stresses associated with
these changes are large enough to cause plastic deformation before the temperature
becomes uniform then the shape changes are not recovered and are superimposed on
those due to the phase transformations. Reducing the austenitization temperature is
helpful for retaining prior dimensions. One should note that because of the high carbon
content and the effects of the dissolved alloying elements, quenching to room
temperature usually leaves a significant quantity of retained austenite in the steel,
typically around 10 % by volume.
Tempering is typically carried out below 300 Celsius, so the principal role of dissolved
alloying elements is then to retard softening while residual stresses are relieved.
When even greater dimensional stability than that which can be obtained in intricate
parts during quenching in oil is desirable, the logical thing to try is to increase the
hardenability so that the martensitic transformation will occur during very slow cooling
rates that eliminate serious temperature gradients in the work piece. This leads us to air
hardenable tool and die steels in which greater amounts of Mo and Cr are present. The
increased amount of Cr also adds the benefit of increased resistance to oxidation and
decarburization. An example is A2 which contains:
R. H. Wagoner MSE 661 – Steel Section 6 Page 47
47 47
1 % C 0. 6 % Mn 0. 25 % Si 5 % Cr 0. 25 % V and 1 % Mo
Now, in order to raise the hardenability by means of the increased alloy content it is
necessary to dissolve more of the alloying elements into the austenite. These steels are
thus austenitized at higher temperatures than are the oil hardening types. Typically they
are austenitized at about 1000 Celsius, air cooled and tempered. Heating to the
austenitization temperature is carried out in stages to minimize distortion. It is common
to heat to 750 - 800, hold at that temperature and then to continue up to the final
temperature. The intermediate hold allows some of the phase transformations to be
performed which reduces the time required at the highest temperature. This is beneficial
because the effects due to decarburization etc. are more serious the longer the steel is at
the elevated temperature, even though it is usually packed in protective materials.
Another reason for the intermediate hold is that it provides a means of bringing the times
spent at the maximum temperature by the surface and the interior closer together so that
their microstructures are more similar.
With the increase in the dissolved alloy content comes an increase in the volume fraction
of retained austenite because of the reduction in the M
s
temperature. 18 % is typical.
This retained austenite can transform during tempering. Indeed chromium carbides can
form within it thereby reducing its carbon and chromium content such that its M
s
temperature is raised above its former value causing martensite to form during cooling
after tempering ( similar to conditioning in stainless steels). Double tempering might
then be necessary. Sometimes A2 is refrigerated to reduce the amount of retained
austenite within it.
High chromium, high carbon, cold-work tool and die steels.
So far we have considered cold work tool steels that are of relatively low alloy content.
Because chromium and chromium carbides give these steels good properties it follows
that the effects of increasing the amounts of these two elements are worth examining.
This results in the high chromium, high carbon, cold work steels. An example is D2:
1. 5 % C 12 % Cr 0. 8 % Mo 0. 6 % V 0. 3 % Mn 0. 25 % Si.
It is clear from this composition that the steel will be rich in alloy carbides and have good
resistance to oxidation. In fact the alloy content is so great that there are very significant
departures from the behavior of carbon steels that are important to note. The as-cast
microstructure of the steel following its manufacture, consists of primary grains
surrounded by a eutectic mixture. As solidification begins, marked coring occurs as Cr
and C are rejected from the solid. It is irrelevant whether the primary nuclei are delta
ferrite or austenite; the eutectic constituents that solidify are austenite and carbides.
Austenite transforms to ferrite as the casting cools and it is possible for martensite to
form. Forging serves not only to produce the steel in suitable forms but also modifies the
microstructure creating a more uniform mixture of matrix and carbides which, in turn,
endows the material with more uniform properties. There is, however, a tendency for
artifacts such as carbide stringers to arise. Forging is carried out from about 1000°C to
R. H. Wagoner MSE 661 – Steel Section 6 Page 48
48 48
1100°C. Higher temperatures are not used because the eutectic constituent can melt.
These steels have lower thermal conductivity than low alloy steels so that a pre-heat soak
and slow heating rates are necessary. Also, very slow cooling rates are needed to avoid
the formation of martensite.
The forged steel is annealed before it is machined into tools. A typical annealing process
is to heat to about 875°C in a controlled atmosphere, hold for two hours and then to
slowly cool to 775°C, hold there for 2-4 hours and air-cool. The final microstructure is a
mixture of large and small alloy carbides in a ferrite matrix.
Again the tools are hardened after they have been machined to shape. To ensure
minimum dimensional changes and to avoid scaling, heating is done slowly in baths or
protective atmospheres. Typically the tools are preheated to and soaked at an
intermediate temperature and then heated to about 980°C. Some of the steels are oil-
quenched but the molybdenum in D2 makes it air-hardenable in large sizes. The actual
final “austenitization” temperature is quite important because, of course, complete
austenitization does not occur . The steel will be austenite plus alloy carbides and the
volume fractions of these phases will vary with the temperature. More of the carbides
dissolve the higher the temperature that is used. This increases the concentration of
solute dissolved in the austenite which lowers the M
s
temperature. This, in turn, causes
an increase in the volume fraction of retained austenite which reduces the as-quenched
hardness.
The variation of hardness during tempering is also dependent upon the austenitization
temperature because of the effects of the composition of the austenite. In general, the
rate of softening is low because the progress of the tempering reactions is restrained by
the high alloy content. Furthermore when the higher austenitization temperatures are
used, secondary hardening can arise due to carbide precipitation and the formation of
martensite in the “conditioned” retained austenite. The “conditioning refers to the rise in
the martensitic start temperature of the retained austenite caused by the precipitation of
carbides therein during tempering.
aust 2050 F
aust 1850 F
65
60
55
50
45
40
35
400 600 800 1000
1200
1400
R
c
F
R. H. Wagoner MSE 661 – Steel Section 6 Page 49
49 49
We have considered the main types of cold work tool steels but there are variations of
these. For example, special wear resistant cold work die steels. What do we need for
improved wear resistance? The answer is: hard, wear resistant carbides and lots of them.
Vanadium and tungsten carbides are harder than chromium and iron carbides, so the an
improvement should be obtained by replacing some of the chromium carbides with
these. Hence D7:
2. 3 % C 0. 5 % Mn 0. 5 % Si 5. 25 % Cr 4. 75 % V 1. 1 % W 1. 1 % Mo
Hot work die steels: There are many applications in which the work piece is hot. For
example, many of the parts that are made from aluminum alloys are made by hot
working operations. During hot working operations the dies are often cooled by water or
air blasts but the tool still becomes hot, particularly at its surface in contact with the
work-piece. Commonly temperatures can range above 1000°F. The following are some
of the important properties that should be possessed by hot-work tools steels.
1) They should resist being deformed at the working temperatures. The main reason that
high carbon, low alloy tool, steels are unsuitable for hot working processes is that
although they can be produced to have high hardness and strength at room temperature,
they soften rapidly at elevated temperatures. It is true that the hot work steels that are
used also soften when heated, but they do so slowly and retain enough strength to have a
significant working life. For this to be so, the steels have medium to high contents of
alloy element.
2) They should be shock resistant. This implies a resistance to mechanical and thermal
shocks, both of which are encountered in service. Good notch toughness is desirable to
avoid catastrophic crack propagation. To satisfy this requirement most hot work tool
steels have low to medium carbon contents e. g. 0. 25-0. 6 wt. %. The increase in
toughness has to be a compromise of course because it is necessary to maintain adequate
hardness and strength and also to avoid rapid wear or erosion. Wear rates tend to
increase as toughness increases.
3) As is usually the case for tools, it is desirable that dimensional changes and warping
are minimized when they are heat treated. This is particularly true when the tool is
intended to produce a part as close as possible to its required net shape as is the case for
dies. An obvious reason is that this will reduce the need for subsequent machining and
thereby the cost of the tool. Hot work steels can develop a protective surface film during
heat treatment. If, however, this film is removed during grinding processes required to
adjust dimensions after heat treatment then the useful life of the tool will be less. The
most common method of minimizing dimensional changes during heat treating is to use
high alloy contents that raise the hardenability so that rapid temperature changes are
unnecessary. This is beneficial with respect to undesirable effect caused by thermal
shocks and thermal gradients.
4) Tools used in hot work operations experience cyclic variations in the surface
temperature as they heat up and cool down. These temperature excursions generate
R. H. Wagoner MSE 661 – Steel Section 6 Page 50
50 50
cyclic stresses with the result that thermally induced fatigue is very common. This
phenomenon generally manifests itself by the development of a network of fine cracks on
the surface of the tool. The cracks are usually along the traces of grain boundaries and
grains can even be pulled out of the surface. The phenomenon is known as “heat
checking” or “craze cracking”. This is a common practical problem and its occurrence is
often the limiting factor in the life of a die.
The hot-work tool steels can be divided into three broad classes: those in which Cr is the
principal alloying element, those which contain W or Mo as the principal alloying
element and those which contain about equal quantities of Cr and W. Most of them
contain V which is combined with carbon and is not taken into solution during the normal
heat treatments.
Examples
S7 0. 5 % C 0. 7 % Mn 0. 3 % Si 3. 25 % Cr 1. 4 % Mo
H11 0. 4 % C 0. 92 % Si 5. 1 % Cr 1. 5 % Mo 0. 4 % V
H14 0. 4 % C 0. 3 % Mn 1 % Si 5 % Cr 0. 25 % V 5 % W
An interesting aspect of the metallurgy of hot work die steels is the use of master
tempering curves. An example is shown below and refers to an H11 steel that was
austenitized at 1850°F, air-cooled and tempered at various temperatures and times. The
diagram was constructed from the experimental data as a plot of hardness ( HRC) versus
the parameter T(20 + logt ) where T is °F + 460 and t is the tempering time in hours. The
hardness measurements were, of course made at room temperature; typical hardness
values of 44-50 are used for working tools. The curve is used to estimate the
combination of temperatures and times for which the tool can be used before it becomes
too soft. For example, for a hardness of HRC 50 a horizontal line drawn across the
diagram at that hardness intersects the curve at a particular value of the abscissa
parameter. Then knowing the value of T( 20 + logt) it is possible to determine the
various combinations of t and °F for which the hardness will be at least HRC 50.
R. H. Wagoner MSE 661 – Steel Section 6 Page 51
51 51
High Speed tool steels
Cold and hot work dies are tools that are designed to form the work piece into a desired
shape and size. In contrast, the basic function served by the tool during many machining
processes is to change the form of the work piece by cutting pieces (chips or swarf) from
its surface. The tool is in contact with the work piece and the two move relatively. The
cutting process generates heat which can raise the temperature of the working part of the
tool. Accordingly, liquid coolant streams are commonly used in machining processes to
lengthen the tool life by minimizing heat induced deterioration of its properties. Heat is
created at greater rates the faster the machining is carried out and this imposes a limit on
production rates. An increase in the rate at which a particular machining process can be
performed is economically desirable. Faster machining rates are possible if the cutting
tools can be made of a material that resists softening when heated. Tool steels which
possess this property are called high speed, tool steels. Their major function is cutting
and their major property is hot hardness, i. e. they retain a high hardness when heated to
a red heat. Typically a HRC of 52 at 1000°F and HRC 48 at 1100°F can be achieved.
The first thing that comes to mind when contemplating making a steel hard is the
martensitic transformation. So the steel should contain carbon and be of such a
composition that it can be converted largely to martensite by a heat treatment. In order
for it to have good wear resistance and to maintain a good cutting edge it should contain
a high volume fraction of alloy carbides. Then for it to have “red hardness” undesirable
changes in the microstructure should not occur or should occur slowly when the tool is
hot. This means that the rates of tempering reactions and carbide coarsening processes
should be small at the temperatures concerned. Indeed, some secondary hardening during
use would be useful provided overaging is not rapid. The alloying elements that are
suggested by these requirements are moderate to strong carbide formers that have low
rates of diffusion in ferrite. This suggests elements like W, Mo and V. These elements
along with cobalt are the important alloying elements in high speed tool steel. Tungsten
was the first major element explored for this class of steel and the classic 18-4-1 high
speed steel emerged around 1910. This steel is capable of cutting while at a visible red
heat. Its composition is:
0. 75 % C 0. 3 % Mn 0. 3 % Si 18 % W 4 % Cr 1 % V 0. 7 % Mo
R
c
T ( 20 + log(t) )
28,000 36,000
55
50
R. H. Wagoner MSE 661 – Steel Section 6 Page 52
52 52
In general, high speed tool steels can be grouped as tungsten based, molybdenum based,
and molybdenum-tungsten based. Vanadium and cobalt are the other important alloying
elements. Although the original tool steels were tungsten based, each now has a
molybdenum counterpart which has come to be more widely used in this country as a
result of strategic materials limitations during wars. We may simply note that the steels
are rich in these elements and the metallurgy of the various alloys is quite similar. One
should note that tungsten is roughly twice as heavy as molybdenum when comparing
weight percentage compositions. Some other examples are;
T4 The 18-4-1+5 % Co
M2 0. 8 % C 5 % Mo 6 % W 2 % V 4 % Cr
M10 0. 9 % C 0. 3 % Mn 0. 3 $ Si 4 % Cr 2 % V 8 % Mo
In general, these steels have high levels of carbide formers which produce excess
carbides in the tempered martensite; i. e. the martensitic transformation occurs in
austenite that contains alloy carbides so that these carbides are inherited by the
martensite. Because of the high content of ferrite stabilizers, austenite isn’t formed after
holding at temperatures suitable for low alloy steels. For example heating to 1650°F and
quenching doesn’t produce martensite because austenite doesn’t form. When the steel is
heated to within 150°F of its melting point, austenite forms and dissolves sufficient
carbon and other alloying elements so that it can then be hardened to HRC 63 or higher
by cooling in air in thick sections. Secondary hardening by the precipitation of fine alloy
carbides occurs during tempering and during use.
Microstructural constituents and treatments: The cast steel is extremely cored due to
the segregation that occurs during solidification. It is irrelevant whether or not delta
ferrite or austenite is the nature of the initial nuclei because the first portion to solidify is
relatively lean with respect to its alloy content and is ferrite at room temperature.
Embedded within it are some carbides and usually an amount of lamellar pearlitic-like
constituent. During the solidification process, the enrichment of the liquid results in the
formation of a skeleton-like, eutectic constituent in the final volumes to freeze. The first
step after casting is to break down the eutectic network and to produce an array of ferritic
grains containing alloy carbides. This is accomplished by hot forging or rolling and
annealing at 1650°F. Typically these steels contain 0. 25 -0. 3 volume fraction of
carbides.
The carbides have been extracted (i. e. the ferrite has been dissolved away using
appropriate chemical reagents) and identified (e. g. by electron diffraction) as follows:
M
6
C. . . . . . . . . . . . . . . . . . . Fe
3
W
3
; Fe
4
W
2
C. Cr, V and Co dissolves in the
former and Mo replaces the tungsten in the
latter in Mo steels.
M
23
C
6
. . . . . . . . . . . . . . . . Chromium rich, dissolves Fe, V, Mo and W
MC. . . . . . . . . . . . . . . . . . . . . . . Vanadium rich
M
2
C. . . . . . . . . . . . . . . . . . . . . . . Tungsten or molybdenum rich carbide
R. H. Wagoner MSE 661 – Steel Section 6 Page 53
53 53
The matrix contains some of the alloying elements in solid solution and very little carbon.
Roughly half of the chromium content is in solution which is beneficial for corrosion
resistance. It is possible for intermetallic compounds to form when the total carbon is
low e. g. Fe
3
W
2
.
The heat treatment used to harden these steels first produces a mixture of austenite and
alloy carbides. Initially the microstructure is an array of ferrite grains and alloy carbides.
When the austenite begins to form the microstructure contains austenite, ferrite and
carbides. Both the ferrite grains and the carbide particles restrict the grain growth in the
austenite. Then M
23
C
6
begins to dissolve and the dissolution is complete when 2000°F
is reached during slow heating. Then partial solution of the other carbides occurs until
the austenite typically contains about 7-12 % alloy carbides of the types MC and M
6
C.
On cooling there is an immediate tendency for the carbides to re-precipitate. Some
precipitation occurs, especially on austenite grain boundaries. However, the rates of
diffusion of W and Mo are very low and it requires very slow cooling for significant
carbide precipitation to occur on the grain boundaries and within the grains. More rapid
cooling yields bainite and then martensite plus carbides plus retained austenite. A typical
tempering response is indicated in the sketch.
The starting material is 60-80% BCT martensite, 15-30 % FCC austenite and 5-12 %
MC and M
6
C. It is very hard and also very brittle with much residual internal stress.
The usual tempering range is 950 to 1100°F but for the sake of completeness we will note
that three stages of tempering have been observed.
1) Up to 270°C, BCT martensite transforms to BCC plus epsilon carbide.
2) In the range 300-400°C BCC plus epsilon carbide goes to BCC plus cementite.
3) From 400- 560°C, BCC plus cementite changes to BCC with less cementite and the
formation of alloy carbides occurs.
While these transformations are taking place at the higher temperatures in the martensite,
precipitation of alloy carbides also occurs in the retained austenite which is also
supersaturated. This reduces the concentration of dissolved alloying elements in the
austenite which raises the M
s
temperature. This causes more martensite to form when
the tool is cooled after being tempered. Often double or even multiple tempering is
necessary to eliminate the retained austenite. The red hardness is due to the reluctance of
the carbides to coarsen because of the slow rates of diffusion and also because secondary
hardening can occur while the tool is hot.
Once a high speed tool steel has been hardened by such a heat treatment they cannot be
re-hardened by a straight forward repetition of the same heat treatment because
65
1000 1200
o
F
R
c
R. H. Wagoner MSE 661 – Steel Section 6 Page 54
54 54
discontinuous grain growth occurs. During the heating process used to austenitize the
steel, the small acicular BCC crystals transform to austenite but their intercrystalline
boundaries are pinned by fine carbides. These carbides restrict grain growth until high
temperatures are reached whereupon , as the carbides dissolve, some crystal boundaries
become unpinned before the others. Then a few grains are able to grow rapidly at the
high temperatures and consume their neighbors. This process leads to exaggerated grain
growth in which a few grains become very large. The effect is inherited in the tempered
martensite that is formed subsequently causing very poor toughness. The elimination of
this tendency requires special treatments such as long term cyclic heat treatments.
This completes our dash through part of the kingdom of tool steels.

Sign up to vote on this title
UsefulNot useful

Master Your Semester with Scribd & The New York Times

Special offer: Get 4 months of Scribd and The New York Times for just $1.87 per week!

Master Your Semester with a Special Offer from Scribd & The New York Times