Metallogr. Microstruct. Anal.

DOI 10.1007/s13632-016-0319-4

CLASSIC PAPER

The Role of Hydrogen and Other Interstitials in the Mechanical

Behavior of Metals
(1959 Edward De Mille Campbell Memorial Lecture)

Alexander R. Troiano

 American Society for Metals 1960

At this time each year, our society pauses in its multitude of hydrogen, oxygen, and boron. It is quite likely that others
activities to pay homage to the memory of Edward De Mille under certain limited conditions, may also reside in the
Campbell—a scholar and teacher of great skill and under- interstices of the host lattice.
standing, but above all, a man of indomitable courage. Of these, carbon has received the most attention; where
Throughout all but the first two years of his professional life, the behavior of nitrogen has generally been considered to
Professor Campbell worked under the handicap of total be analogous to that of carbon. Oxygen and boron, stran-
blindness. That is, a handicap for one of less determination. gely enough, have been almost virtually ignored as inter-
Using the eyes of his students and others, in addition to stitials. Recently, hydrogen has come under close scrutiny,
teaching and administrative duties, he contributed some 77 although its influence on the mechanical behavior of steel
papers to the scientific literature, more than half of which has been well-recognized for years.
dealt with a correlation of the chemical, physical and The interstitial alloying elements have in common a
mechanical properties of steel. high diffusion rate relative to that for the substitutional
It is with a deep sense of responsibility and humility that elements, but can vary widely among themselves in other
I assume the privilege of joining you in honoring the characteristics, such as solubility, tendency to form com-
memory of Professor Campbell and of becoming a member pounds, crystal structure, etc. In fact, three of these are
of that distinguished roster of past Campbell Memorial gases at ambient temperatures and pressures.
Lecturers. It is our desire to discuss the role of hydrogen as an
It has long been appreciated that interstitial alloying interstitial alloying element in the mechanical behavior of
elements will, in small quantities, exert a potent influence metals, and then to examine certain corollary influences of
on the mechanical behavior of alloys. The elements nor- some of the other interstitials, either observed or
mally considered to be interstitial are carbon, nitrogen, anticipated.
An effort of this type involves the ideas, activities, and
cooperation of many people. As I stand here on this
platform, I am flanked by my students, my colleagues,
This is the Thirty-fourth Edward De Mille Campbell Memorial
Lecture, presented by Alexander R. Troiano, Chairman, Department and my family, all of whom have contributed beyond
of Metallurgical Engineering, Case Institute of Technology, measure. I am grateful for their assistance. It is an added
Cleveland. The lecture was presented November 4, 1959 during the pleasure to acknowledge the encouraging and enlightened
Forty-first Annual Convention of the American Society for Metals, sponsorship of this entire investigation by The Aeronau-
held in Chicago. This text is reproduced without change from the
Campbell Memorial Lecture as delivered before the American tical Research Laboratory, Wright Air Development
Society for Metals. Center, United States Air Force, under Contract AF
33(616)-3431.
Reprinted from Transactions of American Society for Metals, vol. 52, That hydrogen can severely embrittle steel and several
54–80 (1960). Copyright 1960 by American Society for Metals,
Metals Park, Ohio. At the time of the original printing, the author,
other metals is well-known [1–7], and one of the more
Alexander R. Troiano was Chairman, Department of Metallurgical intriguing aspects of this embrittlement is its sensitivity to
Engineering, Case Institute of Technology, Cleveland. strain-rate and temperature. Briefly, the embrittlement is

123
 Metallogr. Microstruct. Anal.

enhanced by slow strain-rates and elevated temperatures

[8–13]. These are mutually dependent variables and may be
perturbated over a fairly wide range of strain rate and
temperature.
To differentiate this type of ductility loss from the more
conventional behavior, where high strain-rates and low
temperatures enhance brittle behavior, it is generally
referred to as ‘‘slow strain-rate embrittlement’’ and will be
so termed in this lecture. This slow strain-rate and tem-
perature sensitivity strongly suggests that the phenomenon
is under the control of and paced by the diffusion of
hydrogen.
All brittle failures involve the initiation and propagation
of cracks. Under the conditions of relatively slow strain- Fig. 1 Schematic representation of failure characteristics of a
rate controlled by diffusion of an interstitial, the times for hydrogenated high strength steel
crack initiation and propagation become highly significant
and readily measurable over broad operating ranges of time
and temperature. An analysis of this type of failure will
then, involve a measurement of the time for crack initia-
tion, or more specifically—the crack incubation period, and
a determination of the rate and mechanism of crack prop-
agation to failure.
Sustained load tests at different initial stress levels and
test temperatures offer definite advantages in crack initia-
tion and propagation studies as compared to tests where the
load and stress are continuously increasing, often at rela-
tively high rates such as in a conventional tensile or torsion
test.
A systematic study of this type has been in progress at
Case Institute of Technology; where sustained load or
stress rupture tests were conducted on uniformly hydro-
genated steel, titanium, and other alloys under various
conditions of time and temperature [6, 14–16]. Fig. 2 Static fatigue curves for various hydrogen concentrations
The method of examination was simply to load statically obtained by baking different times at 300 F. Sharp-notch specimens.
230,000 psi strength level [14]
a series of notched tensile specimens at different stresses
and wait for failure. limit—that is, a stress below which failure will not occur
Figure 1 typifies the kind of sustained load to failure for an indefinite period of time.
curves obtained. This behavior is sensitive to hydrogen concentration as
There are a number of significant features that should be shown in Fig. 2, where it may be seen that all delay failure
noted here: parameters; notch strength, rupture time, and static fatigue
limit increase with decreasing hydrogen concentration.
1. The notch tensile strength may be less than normal and
Also note that even after 24 h at 300 F there is still a
directly reflects the loss of ductility due to hydrogen.
substantial stress range, of the order of 60,000 psi over
2. Delayed failure may occur over a wide range of
which delayed failure will occur. In an unnotched speci-
applied stress.
men, full recovery of the ductility as measured by the
3. There is only a slight dependence of the time to failure
reduction of area can be attained in less than 20 h at 300 F
upon the applied stress.
[17], yet delayed failure will occur after 24 h or longer of
4. Perhaps, the most significant characteristic of this
baking time at 300 F. Indeed, the notch tensile strengths
stress rupture relationship lies in the fact that there is a
on these same curves indicate complete recovery between 7
minimum critical value below which failure does not
and 12 h at 300 F, yet delayed failure will still occur over
occur.
a wide range of applied stresses. Thus, it is apparent why
These are, in essence, static fatigue curves, and the conventional laboratory tests may not detect the suscepti-
lower critical stress may be considered a static endurance bility to delayed failure.

123
Metallogr. Microstruct. Anal.

Fig. 3 Static fatigue curves for specimens of different notch sharp-

ness. Baked 0.5 h at 300 F [14]

The increase in lower critical stress with decreasing

hydrogen concentration suggests that crack initiation is
controlled by a combination of hydrogen and stress. With Fig. 4 Cracks observed in hydrogenated notched specimens sec-
decreasing hydrogen concentration, more stress is neces- tioned after static loading. Longitudinal section at 9100. Top
sary to initiate delayed failure. The concept is supported by specimen has 0.001-inch notch radius and bottom specimen has
experiments on specimens of varying notch severity. 0.010-inch notch radius [14]
The variation of the lower critical stress with notch
severity is shown in Fig. 3. It is evident that the static
fatigue limit rises as the notch severity (radius) decreases.
The unit stress at the root of the notch is high and
undoubtedly is sufficient to cause some localized plastic
flow.
All of this, strongly suggests that a critical combination
of stress state and hydrogen concentration must be
attained to initiate a crack. Additional support for this
hypothesis is obtained from an examination of the position
of crack initiation as a function of the notch severity.
Sharply notched specimens are observed to crack slightly
below the notch bottom as indicated in Fig. 4, and less
sharp notches more deeply. These results agree qualita-
tively with the concept that the most severe triaxial stress
state exists below the notch surface, and that the position Fig. 5 Schematic representation of failure characteristics of a
of maximum triaxiality moves into the specimen as the hydrogenated high strength steel
notch severity is decreased; thus correlating the location
of crack initiation with the position of maximum triaxial
stress. The kinetics of crack initiation and propagation can be
We are now ready to consider specifically the initiation followed by electrical resistance changes [18]. A typical
and propagation of cracks leading to failure. Let us first resistivity-time curve for sharply notched specimens is
direct our attention to the events occurring prior to fracture shown in Fig. 6. The existence of an incubation period for
in the static fatigue range. The cracks shown in Fig. 4 and crack initiation and a period of apparently controlled crack
the appearance of the fracture surfaces in Fig. 5 indicate propagation are apparent.
that failure occurs by a process of crack initiation and The existence of an incubation period is a matter of
relatively slow propagation. considerable importance. For example, as we have pointed

123
 Metallogr. Microstruct. Anal.

Fig. 7 Variation of incubation period with applied stress and

hydrogen concentration corresponding to different baking times at
300 F. Sharp-notch specimens [14]

Fig. 6 Typical resistivity-time curve for sharp-notched specimens

to the strain gradient and the diffusion coefficient and
out, in hydrogen embrittlement the fracture ductility inversely to the absolute temperature.
decreases with decreasing strain-rate. Of course, this frac-
ture ductility is dependent upon both initiation and propa-
gation of a crack; nevertheless, it is evident that, for a given
incubation period, a slow strain-rate test will be more likely We know that the diffusion coefficient varies exponentially
to detect embrittlement than a high strain-rate test. If the with the absolute temperature.
strain-rate is so high that the specimen fractures in less time
D ¼ AeQ=RT
than the incubation period, then hydrogen embrittlement
will not be detected. Thus, again it is apparent why con- Thus, if the first initiation of a crack (incubation time) is
ventional tensile tests may not detect the susceptibility to dependent upon the diffusion of hydrogen, then a plot of
delayed failure [19]. the log of the ratio of the incubation time to the absolute
The incubation period is strongly dependent upon temperature should vary in linear fashion with the recip-
hydrogen content but is little influenced by applied stress as rocal of the absolute temperature [21]. The results shown in
indicated in Fig. 7. The incubation period is interpreted in Fig. 8 indicate that such a relationship does exist and an
the following manner [14]. Just prior to testing, the activation energy of 9200 cal per mole is in good agree-
hydrogen is uniformly distributed throughout the specimen. ment with several of the previously reported values for the
The proper or critical concentration of hydrogen for failure diffusion of hydrogen in alpha iron for this test temperature
does not exist in the region of high triaxiality because the range and for hydrogen embrittlement.
specimen does not fail on loading; but since the load or The failure process may be divided into 3 distinct stages,
stress does not change, the incubation time must be that as indicated by the resistivity curves:
time necessary for the hydrogen to concentrate in this
1. The incubation period.
region. It has been demonstrated that the driving force for
2. A period of relatively slow crack propagation.
diffusion is an activity gradient for the diffusing element
3. Catastrophic failure with extremely rapid crack
[20]. In this case the activity gradient must arise from the
growth.
nonuniform stress state, since the hydrogen is uniformly
distributed. When sufficient hydrogen has concentrated in An interpretation of the incubation period has been
this region, a crack will be initiated. presented. Let us turn our attention to the other two stages.
If the incubation time is dependent primarily on the The appearance of the fracture as well as the resistivity
diffusion of hydrogen, then it should be simply related to curves clearly differentiate between these latter two stages;
the temperature in terms of hydrogen diffusion. For stress- that is, the area of slow crack growth and that of catas-
induced diffusion there is a simple relationship in which trophic failure as indicated in Figs. 5 and 9. From obser-
the drift velocity of the interstitial is directly proportional vations of this type, the average slow-crack-propagation

123
Metallogr. Microstruct. Anal.

Calculations of this type show that the fracture stress

was constant and approximately equal to the fracture stress
of the nonembrittled material. This was true over a wide
range of hydrogen concentration and applied stress. This
simply means that stage 3 is merely a high strain rate test
and thus, there is no manifestation of hydrogen embrittle-
ment [14].
Let us now focus our attention on Stage 2… slow crack
propagation. The constancy of the fracture stress just
considered above really pinpoints the role of hydrogen
diffusion. The average hydrogen concentration is not suf-
ficient to propagate a crack. In other words, crack propa-
gation must await a localized increase in hydrogen
concentration in front of the crack. This merely means that
the crack progresses slowly at a rate controlled by hydro-
gen diffusion to a region near the crack tip.
Thus, the implication is that crack propagation is a
discontinuous process and, in fact, consists of a series of
crack initiations. The most severe triaxial-stress state will
Fig. 8 Ratio of incubation time to absolute temperature versus arise just slightly in advance of the crack. When the critical
reciprocal absolute temperature. Applied stress—150,000 psi [21] hydrogen concentration is attained, a small crack forms and
rate can be calculated by taking the difference between grows through the hydrogen-enriched region, thus joining
incubation time and rupture time and dividing this into the the previous crack. Further crack growth must await dif-
area associated with slow crack propagation. fusion of hydrogen to the new region of high stress state.

Fig. 9 Fracture surfaces of hydrogenated sharp-notched specimens. Dark areas indicate extent of hydrogen-controlled crack propagation [14].
Fracture stress hours baking time applied stress (psi)
Fracture Stress Hours Baking Time Applied Stress (psi)
(a) 330,000 7 175,000
(b) 324,000 12 200,000
(c) 342,000 12 250,000
(d) 325,000 18 225,000

123
 Metallogr. Microstruct. Anal.

Fig. 11 Resistance increase as a function of time for uniformly

hydrogenated specimens tested in the stress range of delayed failure
Fig. 10 Resistance increase as a function of time for a hydrogenated [21]
specimen tested in the stress range of delayed failure

Resistivity curves taken at room temperature, (see external specimen notch plus the microscopic concentra-
Fig. 10), give only a slight hint of discontinuous growth. tion of the array are not sufficient to produce localized
However, similar curves at lower temperatures where the fracture. Hydrogen is attracted to the stress field of the
diffusion rate of hydrogen is retarded yield clear evidence array and when the combination of stress and hydrogen
in support of this hypothesis as shown in Fig. 11. It is concentration in the lattice at the tip of the array is suffi-
apparent from the abrupt resistance increases that the crack cient, localized fracture occurs. The crack propagates until
propagates in a discontinuous manner. After each step of stopped, probably by passing out of the influence of the
instantaneous growth a plateau of constant resistance is hydrogen rich region and then the sequence of stress-in-
present. The plateaus of constant resistance are then the duced diffusion and localized fracture must be repeated,
‘‘secondary’’ incubation periods. As the temperature is giving rise to the discontinuous propagation.
lowered the plateaus become longer and the individual Since the brittle failure is brought about by a series of
crack extensions become larger, or in other words, fewer crack initiations, the factors involved in controlling the
discontinuous units of crack propagation are needed to incubation time should be highly significant parameters of
achieve failure. Finally, at -50 F, the ultimate in dis- the basic mechanism. The initiation of a hydrogen induced
continuous propagation is attained. That is, the first crack crack is dependent on two factors.
propagates instantaneously through the specimen [21].
1. The stress induced diffusion of hydrogen producing an
The hydrogen induced failure process is considered to
appreciable hydrogen segregation in a localized region.
be essentially a normal fracture on which has been super-
2. The basic influence of hydrogen on the material
imposed the embrittling influence of hydrogen. The
causing localized cracking.
external specimen notch produces a local stress concen-
tration. Plastic flow occurs at the base of this notch and In terms of 1 above, (the stress induced diffusion of
generates fracture embryos in the form of blocked dislo- hydrogen) a basic component of the incubation time is the
cation arrays. These arrays which act as microcracks can be relation between the hydrogen content and the stress nec-
produced by dislocation blocking at grain boundaries, by essary to initiate a crack. We have repeatedly made refer-
dislocation coalescence on a nonslip plane, and by for- ence to this combination of hydrogen and stress. Let us
mation of a shear stress field, such as occurs in the triaxial now re-examine it with a view to separating and evaluating
region beneath a notch, which opposes the stress field of the specific influence of each. It has been demonstrated that
the slipped dislocations [22–26]. the crack incubation time is relatively insensitive to the
At a given value of applied stress, below the notch applied stress, implying that the process is dependent pri-
tensile strength, a stable configuration of dislocation arrays marily on the development of a critical hydrogen
should exist. However, the stress concentration due to the concentration.

123
Metallogr. Microstruct. Anal.

There is however, a minimum stress necessary to cause

a critical accumulation of hydrogen less than this will
develop some segregation but not the critical amount. For
dilute solutions, the equilibrium distribution of solute
atoms can be described in terms of a Boltzmann distribu-
tion function which has been utilized to analyze solute
segregation [28].
C ¼ Co eU=KT
where C = interstitial concentration at any point, Co = -
average interstitial content, U = interaction energy
between the interstitial and the stress field, K = Boltz-
mann’s constant, and T = absolute temperature.
Fig. 12 Effect of hydrogen content (log current density) on the The stress field about the dislocation array which serves
ductility of 4340 steel tested at -321 F [29]
as the fracture embryo should produce general effects on
hydrogen which are governed by this equation. The con-
centration (C) of interstitial hydrogen about the array
At room temperature, as a result of stress-induced dif- should be greatest where the interaction energy (U) is a
fusion during a mechanical test, the hydrogen content at the maximum. This is at the region of maximum triaxiality
point of crack initiation is not known. Thus, in determining where the fracture is first initiated. Thus, the analysis of the
the stress-hydrogen relation to produce a crack, it is nec- critical stress for failure can be restricted to this region.
essary to test under conditions where the hydrogen con- Considering only hydrostatic components, the interac-
centration is not altered during the test. Such conditions are tion energy can be expressed as
attainable at low temperatures where the diffusion rate of U ¼ pDv
hydrogen is essentially nil.
Normal tensile tests were conducted at liquid nitrogen where p is the average hydrostatic stress,
temperature (-321 F) and the reduction in area was used

¼ 1=3 rxx þ ryy þ rzz
as the parameter to measure embrittlement. The hydrogen
concentration was varied by regulating the charging current and Dv is the change in volume produced in the lattice by
density. the hydrogen. rxx ; ryy and rzz are mutually perpendicular.
Figure 12 shows the influence of hydrogen content on The initiation of the crack is dependent on the devel-
the ductility at -321 F, as indicated by the logarithm of opment of a critical hydrogen content (CC). At stresses
current density [27]. Of course, it would be highly desir- below the static fatigue limit no cracking occurs indicating
able to have specific hydrogen contents plotted on the that less than the critical concentration is developed; above
abscissa, and although such values are available, certain the lower critical the concentration is greater than the
problems associated with a precise quantitative interpre- critical amount. At the lower critical, C = CC, and the
tation of their significance have not yet been resolved. distribution law applies;
Nevertheless, for the argument presented here the results
Cc ¼ Co eUm =KT
clearly indicate that the relationship between hydrogen and
stress necessary to initiate a crack is primarily dependent where Um represents the interaction energy corresponding
on the hydrogen content. Below some critical quantity of to the position of maximum hydrostatic stress.
hydrogen no embrittlement occurs, but when this critical Notches exist in all metals, whether they be metallur-
hydrogen content is attained catastrophic embrittlement gical or mechanical, and their influence on a dislocation
takes place. The initiation of a hydrogen-induced crack array constitutes an important component of the theory of
above some threshold stress is therefore dependent on the fracture. Since the influence of a notch on a dislocation
development of a critical hydrogen content, rather than a array resides primarily in its ability to concentrate the
critical relationship between both hydrogen and stress. The macroscopic stresses acting on the embryo, its action on
relative insensitivity of the incubation time to the applied the failure parameters is quite predictable and, in the case
stress is understandable on this basis. What then, is the role of external machined notches, amenable to experimental
of stress? Presumably, the stress in the failure range analysis [29].
influences the process primarily through its ability to pro- For a given initial hydrogen content (Co) and test tem-
duce a critical amount of hydrogen grouping in the region perature (T), the maximum interaction energy (U) should
where a fracture embryo exists. be constant and independent of notch geometry at the lower

123
 Metallogr. Microstruct. Anal.

critical stress. As the notch acuity increases the applied

stress required to develop a given interaction should
decrease,
ln Co =Cc ¼ U/KT
ln Co =Cc KT ¼ U = Constant


U ¼ 1=3 rxx þ ryy þ rzz Dv

and since Dv is constant and independent of notch

geometry


then rxx þ ryy þ rzz ¼ Constant
Due to the limitations inherent in elasticity theory, the


stress state, rxx þ ryy þ rzz , at the tip of the dislocation
array cannot be exactly determined. However, this is not


necessary since the term rxx þ ryy þ rzz can be
expressed as the product of the sum of the three principal
stresses resulting from the external notch and a constant
(W) which determines the additional stress concentration
on a microscopic scale due to the dislocation array. Thus, it
follows:


rxx þ ryy þ rzz notchðWÞ ¼ constant
Since W is essentially constant and

Fig. 13 Lower critical stress (static fatigue limit) as a function of the
rxx þ ryy þ rzz notch can be stated in terms of the stress concentration factor calculated by relaxation methods. Hydro-
applied stress (rA) at the lower critical limit and an genated 4340 Steel [29]
external notch stress concentration factor (Kf),
attained. Plastic deformation at the root of a notch will
rA Kf ¼ constant lower the stress concentration factor. Thus, for a given
and applied load in the range of small plastic deformation, the
stress concentration factor will be greater as the yield
Kf ¼ constant/rA
strength increases simply because there will be less plastic
A plot of the inverse lower critical stress as a function of flow.
the stress concentration factor should be linear and On the basis of the proposed distribution law, the
extrapolate through the origin as indicated in Fig. 13 [29]. interaction energy associated with the critical quantity of
The stress concentration factor Kf is defined as the ratio hydrogen is constant and independent of strength level of
of the maximum normal stress to the applied stress. Thus it the material. Thus the sum of the principal stresses


was necessary to know the maximum values of the stresses rxx þ ryy þ rzz is constant and independent of strength
in the notch specimens which have been subjected to only level. However, this sum is dependent on both applied load
small amounts of plastic deformation. The particular types and stress concentration factor, and thus can be a constant
of notches were employed for which complete stress with increasing yield strength only if the applied load, i.e.,
analyses are available in the literature [30, 31]. the lower critical stress, decreases. This is, of course,
Now since the relationship is linear and extrapolates exactly what has been observed. The lower critical stress
through the origin as predicted by the derived equation, the decreases as the strength level of the material increases and
assumptions become quite credible. That is, that the dis- explains why the high strength steels are more sensitive to
tribution law applies, that there is an interaction energy hydrogen induced failure. This concept has been developed
which at the critical stress for failure is constant and in more rigorous form and will appear elsewhere shortly
independent of notch geometry and that hence, the notch [29].
functions merely to enable the critical stress state to be This, then, is the action of hydrogen in steel. How far
attained at different applied stresses. may we generalize this behavior? Slow strain-rate hydro-
In determining the critical stress state, the yield strength gen embrittlement has been observed in titanium and its
is the limiting parameter since it effectively provides a alloys displaying alpha [32], alpha plus beta [33–38] and
‘‘cut-off’’ for the maximum value of stress which can be beta [38] structures. Indeed, not only has slow strain rate

123
Metallogr. Microstruct. Anal.

Fig. 15 Ductility at fracture as a function of strain rate in a hydrogen

Fig. 14 Ductility at fracture as a function of strain rate in hydrogen charged and uncharged 72% Ni–28% Fe Alloy [47]
charged and uncharged nickel [47]

embrittlement been observed in all three structures, but

also the same brittle delayed failure characteristics under
sustained load already described for steel have also been
obtained in almost perfect analogy including incubation
time and crack propagation [37–39].
Other materials in which low strain rate hydrogen
embrittlement has been definitely observed include Vana-
dium [40–43], Columbium [44] and indirect evidence
indicates its presence in Molybdenum [45] and possibly
Tantalum [46]. All of the metals considered thus far in this
lecture are body-centered cubic with the exception of alpha
titanium which is hexagonal close packed.
It has generally been accepted that face-centered cubic
metals will not display slow-strain rate hydrogen embrit-
tlement. However, recent tests in our laboratories have
revealed this phenomenon in nickel and several nickel base
alloys when heavily charged with hydrogen [47]. Fig- Fig. 16 Log strain rate versus blue brittle temperatures for alpha
titanium alloys containing the indicated interstitial contents [51]
ures 14 and 15 indicate the influence of strain rate on
hydrogenated nickel and a nickel–iron alloy. There can be
no doubt that the phenomenon is more prevalent than
generally realized. the straining and aging will occur during the test when the
Let us now examine the situation for some of the other strain rate and temperature are appropriate to impart suf-
interstitials. Actually, the oldest evidence of slow strain ficient mobility to the particular interstitial involved.
rate embrittlement involves the so-called blue-brittle In titanium alloys slow strain rate embrittlement has
behavior of steel, which is merely a manifestation of definitely been correlated with strain-aging embrittlement
spontaneous strain aging embrittlement caused by carbon for a number of interstitials including carbon, nitrogen and
and nitrogen, and was first clearly demonstrated 35 years oxygen as well as hydrogen [38, 49–51]. The strain-rate
ago [48]. sensitivity of most of these interstitials is summarized in
At this point, the path of speculative logic leads to the Fig. 16. For several of the higher strength hydrogenated
conclusion that strain aging embrittlement is only another titanium alloys, both the slow strain-rate and strain-aging
facet of slow-strain rate embrittlement or vice versa; that is, embrittlement have, in turn, been experimentally related

123
 Metallogr. Microstruct. Anal.

Fig. 17 Static fatigue characteristics of a high temperature alloy

steel. Sharp-notch specimen, 270,000 psi strength level [52]

(with respect to time and temperature) to brittle failure in a

sustained load test [38].
Now, if you will go along with my earlier generalization Fig. 18 Resistance increase as a function of time for high temper-
that (1) strain-aging embrittlement is merely a manifesta- ature alloy steel in the stress range of delayed failure [52]
tion of slow strain rate embrittlement, and that (2) slow
strain rate embrittlement involves discontinuous crack
propagation to brittle failure as exemplified by a sustained Figs. 17 and 18. Here we have the typical static fatigue
load test; then, the delayed failure or static fatigue type of curve (Fig. 17) complete with crack incubation time and
failure should not be restricted to hydrogen, but should be a static endurance limit. In addition, the fracture appearance
much more general phenomenon and possible for other indicates the process of slow crack propagation. Figure 18
interstitials not to mention substitutional elements; a con- shows the now familiar resistivity curve, conclusively
sideration of which is beyond the scope of this lecture. demonstrating that the process involves a crack incubation
More specifically, one might expect brittle delayed period followed by discontinuous crack propagation. Thus
fracture in steel caused by carbon and nitrogen if the it becomes evident that interstitials, other than hydrogen,
conditions are appropriate. Let us examine briefly what we can bring about brittle delayed failure with a finite crack
mean by appropriate conditions. First, it is quite apparent incubation period and discontinuous crack propagation in a
that the particular interstitial involved must have reason- perfect image of the behavior observed for hydrogen in
able mobility. For hydrogen this occurs near ambient steel at lower temperatures.
temperatures. For carbon or nitrogen, equivalent tempera- A rationalization of this behavior in terms of current,
tures should lie in the range 700–1000 F. It is also nec- general brittle fracture theory is not apparent. It is difficult
essary that the yield strength be above the threshold value to visualize that a dislocation pile-up which raises the flow-
necessary to attain the critical interaction energy as stress to the fracture value will give rise to discontinuous
described earlier. crack propagation of the type observed.
With these considerations in mind, a steel was selected Indeed, no satisfactory explanation for strain-aging
which did not appreciably lose its yield strength at tem- embrittlement has yet been presented. With regard to the
peratures at which sufficient carbon mobility may be yield point phenomenon in strain-aging, the Cottrell
expected, that is up to 1000 F. Other than to permit the mechanism of dislocation locking by condensed atmo-
attainment of these conditions, the composition of the steel spheres of foreign atoms provides a reasonably adequate
is of little consequence. This is true, at least within the explanation; but there is nothing in this mechanism which
confines of our present knowledge. However, there is no will account for the loss of ductility in strain-aging. Dis-
doubt that there may be substantial interaction effects location locking will only bring about a rise of the flow
between different interstitials, substitutionals, and disloca- stress with an attendant increase in strength; but strength
tion arrays. and ductility are not, per se, reciprocal properties. In
Sustained load tests were conducted at elevated tem- addition, strain-aging embrittlement always occurs in a
peratures [52]. That brittle delayed failure can occur higher temperature range than the yield point phenomenon.
without hydrogen is vividly demonstrated by the data of To be sure, there may be some overlapping. However, there

123
Metallogr. Microstruct. Anal.

Fig. 19 Density of states in hydrogen-free and hydrogenated nickel

and nickel base–chromium–iron alloys

is no simple device in theory for increasing the locking of

dislocations by raising the temperature.
We offer for your consideration and future evaluation,
the thought that the fracture strength or, if you wish, the
cohesive strength of the lattice is lowered by the segrega-
tion of interstitials in the lattice at the region near the tip of Fig. 20 Hydrogen embrittlement and catalytic activity versus com-
the dislocation array. This is considered to be an uncon- position of Fe–Ni and Ni–Cu alloys [47]
densed atmosphere and as such permits the array to be a
more active fracture embryo and allows a greater average
concentration of the interstitial over a larger volume. Insofar as hydrogen is concerned, the lowering of the
In this connection, it is interesting to note that, based on fracture strength as a basic influence of hydrogen has, in
the best values of the interaction energy U for carbon in a- recent years, been the basis of several theories of hydrogen
iron, the condensation temperature is near 700 F. This is embrittlement [53–55] as opposed to the earlier theories
about the top of the temperature range for a carbon yield founded on the build-up of disruptive hydrogen pressure
point phenomenon and near the bottom of the range for [56]. With one exception [57] most of the newer theories
strain aging or blue brittle behavior. Similar calculations involve the surface adsorption of hydrogen thru the pre-
for hydrogen in a-iron, give a condensation temperature cipitation of hydrogen gas on a crack or lattice imperfec-
near -150 F. Again, this is near the top of the temperature tion surface, which lowers the surface energy necessary for
range for the hydrogen yield point but about the bottom of extension of the crack.
the range for hydrogen embrittlement. At least, for hydrogen it is interesting to speculate how
The implication is strong and intended that not only is this true fracture strength of the lattice may be lowered.
delayed failure the result of the lowering of the true frac- The electrons of the hydrogen atoms in solution in a
ture strength of the lattice; but also, slow strain rate transition metal will enter the d bands of the metallic cores
embrittlement and strain-aging embrittlement likewise [58] as shown schematically in Fig. 19. In addition, the
result from the same phenomenon. The further implication repulsive forces determining the interatomic distance of
is that, under appropriate conditions, this loss in lattice transition metals such as iron, cobalt, nickel, etc. are due to
cohesion can be brought about by any of the interstitials or the overlapping of their d bands. Thus, it may be expected
combination of such. We anticipate that studies of the that an increase of the electron concentration of these bands
influence of other interstitials, particularly oxygen, from produces an increase in the repulsive forces between the
this point of view will be of intense interest and most metallic cores, or, in other words, a decrease in the cohe-
fruitful. sive strength of the lattice.

123
 Metallogr. Microstruct. Anal.

In attempting to evaluate this hypothesis, it is interesting deleterious since they tend to destroy the homogeneity of
to note that only transition elements have thus far been the mass.’’
hydrogen embrittled. In addition, as an example of the
influence of alloying, this repulsive force should be
lessened when either iron or chromium is added to References
nickel. The electron concentrations of iron and chromium
1. R.W. Buzzard, H.E. Cleaves, Hydrogen Embrittlement of Steel:
are lower than that for nickel, thus, a nickel–chromium– Review of the Literature. United States, National Bureau of
iron alloy also has a 3 d band which is less full than Standards, Circular 511 (1951)
nickel (see Fig. 19). Also when the d shell of a transition 2. C.E. Sims, The Behavior of Gases in Solid Iron and Steel. Gases
in Metals (American Society for Metals, Cleveland, 1953),
metal is not nearly full, the importance of repulsive
pp. 119–198
‘‘exchange’’ forces between the ions decreases [59]. 3. D.P. Smith, Hydrogen in Metals (University of Chicago Press,
Thus, for alloys of nickel plus transition elements of Chicago, 1948)
lesser atomic number, the increase of energy due to the 4. C.A. Zapffe, C.E. Sims, Hydrogen embrittlement internal stress
and defects in steel. Trans. Am. Inst. Min. Metall. Eng. 145,
addition of hydrogen to the alloys becomes lessened. Of
225–259 (1941)
course, when sufficient iron and/or chromium are added 5. A.W. Dana, F.J. Shortsleeve, A.R. Troiano, Relation of flake
to nickel to cause the gamma ? alpha transformation, formation in steel to hydrogen microstructure and stress. Trans.
this trend is reversed. Am. Inst. Min. Metall. Eng. 203, 895–905 (1955), Journal of
Metals, vol. 7
According to this hypothesis, additions of copper to
6. R.P. Frohmberg, W.J. Barnett, A.R. Troiano, Delayed failure and
nickel should initially increase the propensity to hydrogen hydrogen embrittlement in steel. Trans. Am. Soc. Metals 47,
embrittlement and then decrease it to zero as the 3 d band is 892–925 (1955)
progressively filled. This concept was tested by examining 7. E.P. Klier, B.B. Muvdi, G. Sachs, Hydrogen embrittlement in an
ultra-high-strength 4340 steel. Trans. Am. Inst. Min. Metall. Eng.
the reduction in ductility caused by hydrogen embrittle-
209, 106–112 (1957), Journal of Metals, vol. 9
ment in several selected iron–nickel and nickel–copper 8. P. Bastien, P. Azou, Influence de l’amplitude et de la vitesse des
alloys [47]. The dependence on composition of the deformations plastiques sur la segregation de la hydrogene dans
hydrogen embrittlement of these selected alloys is indi- le fer et les aciers. Comptes Rendues 232, 69–71 (1951)
9. N.J. Petch, P. Stables, Delayed fracture of metals under static
cated in Fig. 20. It appears that the embrittlement first
load. Nature 169, 842–843 (1952)
increases and then decreases as the 3 d band of iron–nickel 10. J.D. Hobson, J. Hewitt, The effect of hydrogen on the tensile
and nickel–copper alloys is progressively filled. properties of steel. J. Iron Steel Inst. 173, 131–140 (1953)
This behavior is strikingly similar to that of the activity 11. J.B. Seabrook, N.J. Grant, D. Carney, Hydrogen embrittlement of
SAE 1020 steel. Trans. Am. Inst. Min. Metall. Eng. 188,
of iron–nickel and nickel–copper catalysts for the hydro-
1317–1321 (1950), Journal of Metals, vol. 2. Discussion, Journal
genation of styrene as shown by Fig. 20 [60, 61]. The of Metals, vol. 3, 1951, 558–560
significance of the similarity lies in the fact that the cat- 12. J.T. Brown, W.M. Baldwin Jr., Hydrogen embrittlement of steels.
alytic action of these alloys is caused by the absorption of Trans. Am. Inst. Min. Metall. Eng. 200, 298–304 (1954), Journal
of Metals, vol. 6
the electrons of hydrogen by the 3 d band of the alloys,
13. T. Toh, W.M. Baldwin Jr., Ductility of Steel with Varying Con-
lending further support to the concept presented. centrations of Hydrogen. Stress, Corrosion, Cracking and
I should like to conclude this, the 34th Annual Campbell Embrittlement (Wiley, New York, 1956)
Lecture, by direct reference to a few words written by a 14. H.H. Johnson, J.G. Morlet, A.R. Troiano, Hydrogen, crack ini-
tiation and delayed failure in steel. Trans. Metall. Soc. Am. Inst.
man who had been without sight for most of his career, but
Min. Metall. Eng. 212, 526–536 (1958)
most certainly not without vision. 15. R.D. Daniels, R.J. Quigg, A.R. Troiano, Hydrogen embrittlement
In the broadest sense, we have been largely concerned and delayed failure in titanium alloys. Trans. Am. Soc. Metals 51,
this morning with the variation in concentration of a solute 843–860 (1959)
16. A.R. Troiano, Delayed failure of high strength steels. Corros.
and its influence on the behavior of the base solvent. In this
Natl. Assoc. Corros. Eng. 15, 207–212t (1959)
connection, almost half a century ago, Professor Campbell 17. H.H. Johnson, E.J. Schneider, A.R. Troiano, The recovery of
had this to say [62], and I quote: embrittled cadmium plated steel. Iron Age 182, 47–50 (1958)
‘‘If iron be considered a solvent and the chemical 18. W.J. Barnett, A.R. Troiano, Crack propagation in the hydrogen-
induced brittle fracture of steel. Trans. Am. Inst. Min. Metall.
constitution of steel be regarded as a study in solubili-
Eng. 209, 486–494 (1957), Journal of Metals, vol. 9
ties, we would then find, as in ordinary aqueous solu- 19. H.H. Johnson, A.R. Troiano, Crack initiation in hydrogenated
tions, that the physical properties of the solvent are most steel. Nature 179, 777 (1957)
profoundly modified by the nature and concentration of 20. L.S. Darken, Diffusion of carbon in austenite with a discontinuity
in composition. Trans. Am. Inst. Min. Metall. Eng. 180, 430–438
those substances which are in solution, while substances
(1949)
which are merely mechanically mixed or in suspension 21. E.A. Steigerwald, F.W. Schaller, A.R. Troiano, Discontinuous
exert comparatively little influence. Substances which are crack growth in hydrogenated steel. Trans. Metall. Soc. Am. Inst.
in excess of what can be held in solution are often Min. Metall. Eng. 215, 1048–1052 (1959)

123
Metallogr. Microstruct. Anal.

22. C. Zener, The Micro-Mechanism of Fracture, Fracturing of 41. B.W. Roberts, G.W. Sears, P.D. Zemany, Hydrogen embrittle-
Metals (American Society for Metals, Cleveland, 1948), pp. 3–31 ment of vanadium by catalytic decomposition of water with
23. N.F. Mott, Fracture in metals. J. Iron Steel Inst. 183, 233–243 manganese. Trans. Metall. Soc. Am. Inst. Min. Metall. Eng. 212,
(1956) 71–72 (1958)
24. A.N. Stroh, A theory of the fracture of metals. Adv. Phys. 6, 42. W. Rostoker, The Metallurgy of Vanadium (John Wiley and Sons,
418–465 (1957) New York, 1958)
25. A.H. Cottrell, Theory of brittle fracture in steel and similar 43. W.M. Baldwin Jr., The Metal Molybdenum, Discussion, p. 279
metals. Trans. Metall. Soc. Am. Inst. Min. Metall. Eng. 212, 44. E.S. Tankins, R. Maddin, Effect of grain size, strain rate and
192–202 (1958) temperature on the yield point of columbium. Trans. Metall. Soc.
26. Fracture, Proceeding of conference held April 1959, sponsored by Am. Inst. Min. Metall. Eng. 215 (1959) (to appear)
National Academy of Science. MIT Press and John Wiley, 1959 45. A. Kramer, W.M. Baldwin Jr., Investigation of the fracture of
Library of Congress Catalogue #59-14116 metals. ONR Report, April 1958, NR-049/9-19-56
27. A.E. Schuetz, W.D. Robertson, Hydrogen absorption, embrittle- 46. J.H. Bechtold, Tensile properties of annealed tantalum at low
ment and fracture of steel. Corrosion 13, 437–458 (1957) temperatures. Acta Metall. 3, 249–254 (1955)
28. A.H. Cottrell, B.A. Bilby, Dislocation theory of yielding and 47. P. Blanchard, A.R. Troiano, La Fragilisation des Métaux par
strain aging of iron. Proc. Phys. Soc. Sec. A 62, 49–62 (1949) L’hydrogene. Influence de la Structure Cristallographic et Elec-
29. E.A. Steigerwald, F.W. Schaller, A.R. Troiano, The role of stress tronique. Presented before Societé Francaise de Metallurgie,
in hydrogen induced delayed failure. Submitted to Metallurgical October 1959. To be published in Revue de Metallurgie
Society of the American Institute of Mining and Metallurgical 48. E. Maurer, R. Mailander, Zur Frage der Blausprodigkeit. Stahl
Engineers for publication Eisen 45, 409–423 (1925)
30. D.N.G. Allen, S.R. Southwell, Relaxation methods applied to 49. H. Margolin, Yield point in polycrystalline titanium. Final
engineering problems, XIV. Plastic straining in two dimensional Report, Watertown Arsenal Laboratory 1954, Contract DA-30-
stress-systems. R. Soc. Lond. Ser. A Philos. Trans. 242, 379–414 069-ORD-1217
(1949) 50. F.D. Rosi, F.C. Perkins, Mechanical properties and strain aging
31. J.A. Hendrickson, D.S. Wood, D.S. Clark, The initiation of brittle effects in titanium. Trans. Am. Soc. Metals 45, 972–992 (1953)
fracture in mild steel. Trans. Am. Soc. Metals 50, 656–676 (1958) 51. H.S. Gurev, W.M. Baldwin Jr. Research on strain aging effects in
32. N. Makrides, W.M. Baldwin Jr. High temperature brittleness in titanium. WADC Technical Report 59-223 (May, 1959)
titanium alloys. WADC Technical Report 57-251 (March, 1957) 52. T.E. Scott, Research in Progress, Ph.D. Thesis. Case Institute of
33. H.M. Burte, E.F. Erbin, G.T. Hahn, R.J. Kotfila, J.W. Seeger, Technology, Cleveland
D.A. Wruck, Hydrogen embrittlement of titanium alloys. Metal 53. F. Bastien, P. Azou, Effect of hydrogen on the deformation and
Prog. 67, 115–120 (1955) fracture of iron and steel in simple tension, in Proceedings of the
34. E.J. Ripling, Hydrogen embrittlement in a commercial alpha–beta First World Metallurgical Congress, American Society for
titanium alloy. Trans. Am. Inst. Min. Metall. Eng. 206, 502–503 Metals (1951), pp. 535–552
(1956), Journal of Metals, vol. 8 54. N.J. Petch, Lowering of the fracture stress due to surface
35. R.I. Jaffee, D.N. Williams, The effect of composition on the adsorption. Philos. Mag. Ser. 8, 1, 331–335 (1956) also N.
hydrogen embrittlement of alpha–beta titanium alloys. Trans. J. Petch, Delayed fracture of metals under static load. Nature 169,
Am. Soc. Metals 51, 820–841 (1959) 842–843 (1952)
36. D.N. Williams, F.R. Schwartzberg, R.I. Jaffee, The effects of 55. F. de Kazinczy, A theory of hydrogen embrittlement. J. Iron Steel
microstructure and heat treatment on the hydrogen embrittlement Inst. 177, 85–92 (1954)
of alpha–beta titanium alloys. Trans. Am. Soc. Metals 51, 56. C. Zapffe, Discussion of metal arc welding of steel, by S.A.
802–819 (1959) Herres, Transactions, American Society for Metals 39, 191–192
37. R.D. Daniels, R.J. Quigg, A.R. Troiano, Hydrogen embrittlement (1947)
and delayed failure in titanium alloys. Trans. Am. Soc. Metals 51, 57. J.G. Morlet, H.H. Johnson, A.R. Troiano, A new concept of
843–860 (1959) hydrogen embrittlement in steel. J. Iron Steel Inst. 189, 37–41
38. R.J. Quigg, A.R. Troiano, Hydrogen embrittlement in steels, (1958)
titanium alloys, and several face-centered cubic alloys. WADC 58. R.E. Norberg, Nuclear magnetic resonance of hydrogen into
Report 59-172 (April 1959) R. J. Quigg, Ph.D. Thesis, Case palladium wires. Phys. Rev. 86, 745 (1952)
Institute of Technology, Cleveland 59. N.F. Mott, H. Jones, The Theory of the Properties of Metals and
39. R.D. Daniels, R.L. Harmon, A.R. Troiano, The influence of Alloys (Dover Publication, New York, 1958)
hydrogen on delayed failure in titanium alloys. Wright Air 60. R.H. Griffiths, J.D.F. Marsh, Contact Catalysis (Oxford Univer-
Development Center, WADC Technical Report 57-30 (February, sity Press, London, 1957)
1957) 61. J. Chem. Phys. 63(4), 451–453 (1959)
40. B.W. Roberts, H.C. Rogers, Observations on mechanical prop- 62. E.D. Campbell, Discussion to ‘‘Physical properties of heat-treated
erties of hydrogenated vanadium. Trans. Am. Inst. Min. Metall. carbon steel’’. J. H. Nead, Trans. Am. Inst. Min. Metall. Eng. 53,
Eng. 206, 1213–1215 (1956), Journal of Metals, vol. 8 235–237 (1915)

123