Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
, AMSTERDAM
A critical review of previous work on the a/,!? transition la vitesse de chauffage soit suffisamment rapide. Lea
in pure uranium is presented. It is shown that the deux theories actuelles concernant lea textures sub-
j3 -+ a change is diffusional at slow cooling rates and sistant dans I’uranium cy sont examinees B la lumihe
martensitic during fast cooling and that the OL+ /l des conclusions ci-dessus et ces 2 theories n’apparais-
change is always diffusional except in the very special sent pas satisfaisantes. Une autre explication eat
case of the transformation of martensitic OLwhich proposee en se basant sur lea effets probables de la
transforms by reversal of the primary shear if the grosseur du grain /i’ sur la transformation B Tt LY.
heating rate is sufficiently rapid.
The two current theories of remnant textures in
n uranium are examined in the light of the above Es wird ein kritischer tfberblick iiber die Arbeit
conclusions and both found to be unsatisfactory. gegeben, die bisher auf dem Gebiet des a//? Ubergangs
An alternative explanation is proposed based upon in reinem Uran geleistet worden ist. Es wird dabei
the probable effects of /l grain size on the ,8 * o( gezeigt, dass es sich bei dem /? + n ifbergang urn
transformations. einen Diffusionsvorgang handelt, wenn langsame
Abkiihlungsgeschwindigkeiten verwendet werden. Da-
gegen ist der Wechsel martensitisch bei schnellen
Une revue critique des recherches effect&es anterieure- Abkuhlungsgeschwmdigkeiten. Der (Y + p Wechsel
ment sur la transformation a//l de l’uranium eat dagegen ist immer ein Diffusionsvorgang. Ausge-
present&e. On montre que la transformation /I + LY nommen ist der spezielle Fall der Umwandlung vom
a lieu avec diffusion aux lentes vitesses de refroidisse- martensitischen ‘x, welches in primiiren Schichten
ment et par transformation martensitique aux grandes umschlagt, wenn die Erhitzungsgeschwindigkeit hin-
vitesses de refroidissement. La transformation o( + ,Cl reichend gross ist. Die beiden geltenden Theorien iiber
procede toujours par diffusion sauf dans le cas t&s Resttexturen in oc-Uran wurden im Hinblick auf die
special ob l’uranium a &tB forme par transformation obigen Ergebnisse geprtift. Es wurde gefunden, dass
martensitique. Dana ce cas, il y a transformation beide unbefriedigend sind. Es wird versucht, auf der
(Y + /l par inversion du cisaillement primitif dti It la Basis des wahrscheinlichen Einflusses der ,%Korngriisse
transformation martensitique j3 + a It condition que auf die ,Cl$ x Umwandlung eine Erkllirung zu geb en
38
THE&e/f? PHASE TRANSITION IN PURE URANIUM 39
flitting position and to indicate those aspects types have been discussed thoroughly else-
to which future investigation could be directed where 1).
most fruitfully. We shall not discuss each Other modes have been recognised in other
previous contribution separately or chrono- systems. The bainite reaction in steels involves
logically. Instead, each aspect of the problem both shear and diffusion but this is possible
will first be considered in simple theoretical only in a multicomponent system. The so-called
terms and those experimental observations massive transformation 2) is diffusionless macro-
which are consistent with the theoretical pre- scopically but involves small scale atomic
dictions and with each other indicated. The re-arrangement. This mechanism is possible in
contradictory results will then be examined in pure uranium but at the present time there is
a little more detail and their value assessed. insufficient evidence to justify an attempt to
In discussions of phase changes it is important distinguish between pure martensitic reactions
to define certain terms. In many earlier papers and this mode. Consequently, we shall confine
on the transformations in uranium the term the discussion to diffusional and martensitic
“nucleation and growth” has been used to modes whilst realising that the latter may
describe a transformation mechanism in which involve some microscopic atomic re-arrange-
product nuclei grow by a diffusion-type process. ment .
It has been long reoognised that this term is
unsatisfactory because all mechanisms involve 2. The /3 --f a Transformation
nucleation and growth. Some previous authors Previous authors have put forward the
have also considered nucleation and growth as following conflicting conclusions concerning the
two separate steps and implied that a particular mechanism of the /3 --f LXchange in pure uranium
nucleation mechanism can be combined with (using the present terminology): (1) that it
the same or a different growth mechanism. occurs by diffusional decomposition under all
This approach is unsound. The process of conditions of cooling and undercooling, (2) that
nucleation, if it exists at all +, comprises growth it is martensitic under all conditions and (3) that
of domains of product phase from a size at both modes can be realised depending upon the
which they are unstable to a size at which they cooling rate. In order to establish which is the
are capable of continued existence. In other correct view we shall first present some general
words, nucleation is merely a convenient name theoretical considerations.
for the initial stages of growth. Thus, reactions
must be defined in terms of the growth mecha- 2.1. THEORETICAL CONSIDERATIONS
nism. We shall use the term “diffusional” to Diffusional decomposition produces equili-
describe a transformation in which growth of brium 01 whose free energy, G,, is lower than
the regions of product occurs by the capture of that of ,6, CT,,at all temperatures below TE, the
single atoms from the parent lattice by means equilibrium transition temperature. Hence, this
of thermally activated jumps across the inter- mode is thermodynamically possible at all
face. A martensitic transformation is defined as degrees of undercool. From kinetic consider-
one in which growth proceeds by the systematic ations it follows that it is least likely to occur
co-ordinated shear of large numbers of atoms at very low temperatures at which diffusion is
in the parent phase through small distances too slow to yield an appreciable reaction rate.
relative to their neighbours such that the parent In contrast, martensitic reaction gives a product,
lattice is transformed into that of the product, LX’, whose free energy, Gg,, is greater than G,
and which does not depend on atomic inter- owing to the elastic strain energy arising from
change by diffusion. The characteristics of these the shape change and coherency and also owing
t In many cases there is reason for believing that to the retention in supersaturated solid solution
nuclei may be preformed. of trace impurities. A temperature TO exists,
40 J. BURKE AND P. H. DIXON
necessarily below TE at which G,, =Gp, above Cd) The 01produced by quenching is typical of
which martensitic decomposition is not possible. the martensitic 0~’ found in alloys 3).
In practice, additional supercooling to M, is
(4 Butcher 10)quenched a specimen of uranium
required to provide sufficient excess free energy
after subjecting it to a heat treatment
to overcome the energy barrier to transfor-
known to produce a single crystal of /3 in
mation. However, once M, has been attained,
dilute alloys. The orientation of almost all
the very low activation energy for martensitic
of the oc crystals were consistent with those
growth permits of appreciable reaction rates
to be expected from the martensitio decom-
even at low temperatures. Consequently, there
position of a single crystal of uranium-
is a finite range of undercooling, and thus of
chromium.
cooling rates for continuous cooling transfor-
mation, within which transformation must be (f) Bement and Wallace ii) found that a graph
diffusional; martensitic transformation is most of hardness as a function of cooling rate
likely at drastic quenching rates. In the para- through the ,8 + (Y change exhibited a
graph above, (2) is incompatible with this sharp change in slope at a cooling rate of
conolusion; and of the other two clearly (3) is about 20” C per second. They suggested,
the more likely. on the basis of this rather circumstantial
evidence, that this rate is a critical cooling
rate above which the reaction is totally
2.2. EXPERIMENTAL EVIDENCE IN FAVOUR OF martensitic.
TWO REACTION MECHANISMS
The theoretical conclusion has been well 2.3. EVIDENCE IN SUPPORTOF A MARTENSITIC
substantiated experimentally in dilute alloys REACTION ONLY
in which it is firmly established that ,8 decom- Several previous investigators have concluded
position is diffusional at temperatures high in that the reaction is martensitic at all temper-
the OLrange and is martensitic at low temper- atures and cooling rates, a conclusion which is
atures a-6). The following experimental results at variance with theory and the results (a) to (c).
on the pure metal are in accordance with the This view was first put forward by White s),
hypothesis of diffusional LYformation at slow who noted that the lowest temperature at which
cooling rates : diffusion-controlled transformation occurs in
uranium-chromium alloys increased as the
(a) Single crystals of 01can be produced by the
chromium content decreased. Hence he sug-
phase change technique 7). gested that at zero chromium this temperature
(b) The 01 produced b y slow cooling has the would coincide with TE, leaving shear as the
metallographic characteristics of that ob- only mode. This was based upon an extra-
tained in alloys by isothermal transfor- polation to zero chromium content from only
mation near to TE ; increasing the rate of three compositions and must be regarded as
cooling refines the 01 grain size 7). highly speculative.
Butcher 10) also thought that the reaction was
(0) When a specimen is partially transformed
always martensitic. His crystallographic studies,
to ,5’by passing it slowly through a thermal
whilst proving convincingly the existence of a
gradient, the a grains adjacent to the a/#?
martensitic mode, can give no information as
boundary grow into the /? when the direc-
to transformation during slow cooling because
tion of travel is reversed 9).
the work was carried out on drastically quenched
The following observations are in accordance samples.
with the hypothesis of martensitic transfor- Jepson et ~1.4) obtained a TTT diagram for
mation at fast cooling rates: pure uranium consisting of a single C curve
THE ol&/!? PHASE TRANSITION IN PURE URANIUM 41
in contrast to the two C’s found for dilute Suffice it to note here that whilst orientation
alloys. They concluded that only one mechanism relationships are an unavoidable characteristic
is possible in pure uranium and, in view of of reaction by the martensitic mode, they are
Butcher’s results lo), that this is shear. However, possible also for diffusional reactions, e.g. the
their diagram was based upon insufficient ex- formation of Widmanstiitten precipitates.
perimental points to be entirely convincing. The
possibility of discontinuities in the curve cannot 2.4. EVIDENCE IN SUPPORT OF DIFFUSIONAL
be excluded. Also the technique was incapable TRANSFORMATION ONLY
of detecting “reaction start” times less than
The view that the reaction is always diffusion-
20 seconds. It is unfortunate that no metallo-
controlled was expressed by Duwez 16), who
graphic examination of the dilatometric samples
found that the transformation temperature
was carried out. We tentatively suggest that
decreased progressively as the rate of cooling
this curve is that associated with diffusional
increased. This interpretation was based upon
transformation and that M8 was undetected.
older ideas that martensitic reactions are
Klepfer and Chiotti 12) found that on cooling
athermal and insuppressible. Isothermal, time-
to a temperature just below TE, the resistivity
dependent shear reactions are now recognised,
increased isothermally over a period of time to
a constant value ; further isothermal increases notably in uranium alloys, and these can be
were noted on stepped quenching until a temper- suppressed by increasing the cooling rate. Thus,
this piece of evidence loses its significance.
ature 15" C below TE was attained below which
Lehr and Langeron 7) produced single crystals
no further isothermal changes occurred. The
of a-uranium by the phase-change technique
sum of the isothermal changes was said to be
and concluded that a forms by diffusion. The
that expected for complete transformation of
generalization of this conclusion to all cooling
p to 01. Hence, it follows that the amount of
31 formed depends upon the temperature. This conditions cannot be sustained since the experi-
is a feature of athermal martensitic reaction ment was necessarily restricted to very slow
and it was concluded that the transition is rates of cooling.
martensitic even at very small undercooling.
However, the isothermal martensitic reaction 2.5. SUMMARYOFTHE,!? -+CXTRANSFORMATION
in dilute uranium alloys goes to completion at
To summarize, most of the reliable experi-
a fixed temperature 396) and has no athermal
mental observations are consistent with the
characteristics. Evidently, the conclusion is
theoretical prediction that the @ -+ N change
based upon doubtful ground. Moreover, attempts
in pure uranium is diffusional at slow cooling
to reproduce these results by the present
rates and martensitic at fast cooling rates.
authors 13)and others 14J5) have been unsucoess-
ful. Much importance must not be placed upon
this work until it has been thoroughly re- 3. The OL+ /? Transformation
examined. There is very little direct evidence relating
Finally, it has been reported by several to the transformation even in alloys and present
authors that vestiges of preferred orientation in ideas are largely based upon interpretation of
n-uranium persist through an 01-p--6(cycle 17-20). circumstantial observations e.g. the effect of
One explanation is that there is a reversible heat treatment on textures. In view of this the
crystallographic relationship between LYand B discussion which follows is largely theoretical.
through the transitions. The conclusion from We shall consider two cases (a) transformation
this, that the reactions must, therefore, be of (Y (cc’) previously formed martensitically from
martensitic cannot be sustained. We shall B and (b) the more general case of transfor-
discuss remnant textures more fully in Section 4. mation of equilibrium 0~.
42 J. BURRE AND P. H. DIXON
(a) THE TRANSFORMATIONOF MARTENSITICIX’ some indication of whether shear is ever likely
Martensitic reactions produced by cooling are may be obtained from general theoretical con-
crystallographically reversible on heating, start- siderations. The rate of nucleation of a phase
ing from a temperature, As, above TO. There is determined by the temperature-dependent
is no a priori reason for placing As above or activation energy for the formation of a nucleus,
below TE. In many systems reversibility is not A(T), and the activation energy for the transfer
observed because of the intervention of temper- of atoms across the interface, Q, assumed
ing reactions during heating which destroys the independent of temperature. A(T) for shear is
geometric relationships in the low temperature greater than that for the corresponding diffusion-
shear product. In pure uranium t,he tempering controlled reaction but QS (for shear) is very
takes the form of thermal relaxation of lattice much less than QD (for diffusion).
distortion and loss of coherency. If AS is below For the diffusional transformation 1-3--f Lx
TE, DC’may transform to 16 by simple reversal during cooling, decreasing the temperature
of the /? + (x shear provided that the heating decreases both A(T) and the rate of diffusion.
rate is fast enough to prevent thermal relaxa- These two effect’s are in opposition and thus the
tion. If relaxation does occur 01’ degenerates to reaction rate exhibits ‘C-curve behaviour, the
a, the transformation of which is considered rate of transformation being negligible at low
below. If AS is above TE, a further condition temperature. Owing to the small value of Qs,
must be satisfied for a shear back to b, viz. that the martensitic reaction is capable of appreciable
the heating rate above TE must be sufficient velocities even at low temperatures where it is
to prevent diffusional nucleation of B which the preferred mode. For the heating transition
becomes thermodynamically possible at any a + P, raising the temperature above TE
temperature above TE. decreases A(T) for the diffusional reaction and
This type of reverse shear has been observed accelerates diffusion. The two effects are additive
in alloys. Holden 21) noted that upquenching and the reaction rate increases rapidly with
a partially transformed U-Cr alloy caused the
01’ plates to disappear. There is also indirect
evidence of it in pure uranium. Klepfer and
Chiotti ii) found that rapid cycling through
the a/,!? transition raised the temperature of the
onset of thermal twinning in the a-phase. This
symptom of progressive lattice hardening is
indicative of the retention of lattice defects
through the phase changes which is possible
only if both reactions are martensitic.
temperature ; ‘C’-curve kinetics cannot be pro- very special case of martensitio oc’, which
duced. The difference is illustrated by a hypo- probably undergoes simple crystallographic
thetical TTT diagram for transformation as reversal of the ,!l +- LYshear under oon~tions
shown in fig. 1. The lines BD and o(n are the of rapid heating.
incubation periods for the formation of ,6 from
oc and 01 from ,!3respectively by diffusion. The 4. Remnant Textures
shear reaction to ,!l is not thermodyna,mically A detailed account of the considerable work
possible until a temperature XscBit some way on textures in LXuranium has been given recently
above Ts, is attained, and the incubation line by Butcher 24) and it is unnecessary to repeat it.
for #? formation by shear, &z, is as shown and Here, we shall confine attention to those aspects
the corresponding line for oc formation is 0~~. and theories which need re-examination in the
At T > TE the gradient of ps is approximately light of the conclusions in the previous sections.
equal to -Qs and that of ,8n to -Qn. Since In this connection, the important result is that
Qo> Qs, /& must always be steeper than ,8n preferred orientation in a may persist after
and the two curves can never intersect. It treatment in the /3 range, It is clear from
follows that diffusional transformation is always Butcher’s work that all the important factors are
the preferred mechanism. not as yet unequivooally identified. McDonell19)
A further possibility that has been suggestedza) reported that as the temperature and time of
is that very rapid heating would enable the ,8 treatment were increased the extent of
6s line to be attained before all the 01had had retained texture decreased. Donze and Faivre 26)
time to transform to & and hence that the last working with single crystals, concluded that
parts of a may transform martensitically. We only temperature was important. The effect of
feel that this is unlikely for several reasons ; the rates of heating and cooling has not been
(a)the activation energy for the formation of a systematically studied, but McDonelli9) showed
martensiti~ nucleus is greater than that for that quenching gave a more random structure
diffusional transformation and so the latter will than air cooling.
always be energetically more favourable at Two theories have been advanced to explain
temperatures where diffusion rates are appre- this phenomenon. The first, due to Butcher 26),
ciable; (b) it is clear from fig. 1 that the rate of is based upon the supposition that the (L:-+ /3
the diffusional reaction is always greater than change is marte~itic. The retention of texture
that of the shear and the disparity increases through an LY4 p -+ & cycle is inevitable if the
with increasing superheat. Thus, the greater cooling rate is sufficiently rapid to give marten-
the superheat (i.e. rate of heating) the more sitic IX’, because then both structural changes
favourable is the diffusion mechanism. involve specific orientation relationships. The
Experimentally, the most direct evidence basic assumption of the theory is inttonsistent
relating to this transformation is the resistivity with the conclusion in Section 3 that the a + /3
work of Klepfer and Chiotti. Their interpretation transition is never martensitic. Furthermore,
that the reaction is athermal and therefore, McDonell’s 19) result that quenching increased
martensitic, is at variance with the above con- randomization also appears to be incompatible
clusions. However, for the reasons given earlier with this theory. It should perhaps be em-
we consider this work to be unreliable. Further- phasized here that orientation relations~ps are
more, Barwood and Butcher 23) in a dilatometric not a feature exclusive to martensitic reactions,
study found no athermal characteristics. so that the persistence of textures provides no
information relative to the mechanism of the
3.3. SUMMARY, 01 -+ ,d TRANSFORitIATION phase changes.
We conclude that the Al-+ /3 transformation The alternative explanation, advanced by
is diffusional at all heating rates except in the McDonell 19) and independently by Donze and
44 J. BURKE AND P. H. DIXON
Faivre 14Js), supposes that domains of the It is considered that the evidence available
original 01 exist in the @ phase after transfor- at the present is inadequate to support this
mation and serve as nuclei for re-transformation. theory and much further work is necessary
Donze and Faivre 25) found that 01single crystals before it can be regarded as proved.
were preserved in the original orientation after
heating into the @ range provided a critical 4.1. A TENTATIVE ALTERNATIVE EXPLANATION
temperature was not exceeded; this temperature OF TEXTURE RETENTION AND OF THE
uranium to 14” C of superheat for metal con- A tentative explanation of Donze’s thermal
taining 150 ppm carbon. analysis results and of remnant texture is
Subsequently 14,27),it was shown from cooling possible on the basis of the conclusion of
curve studies of polycrystals that the temper- Section 3 that the 01--f B change is diffusional.
ature of the start of p --f 01 transition at a In the light of general experience of phase
constant cooling rate was depressed by super- changes and of energetic considerations, it is
heating in the p range, from 665” C after 2” C reasonable to assume that both the 01--f ,8 and
superheat to 655’ C after 80” C superheat. ,8 + OLchanges will be predominantly grain-
According to Donze 14), increasing the superheat boundary-nucleated +. Increasing the time or
decreases the size of the (x embryos and thus temperature of the @ anneal prior to cooling
greater underoooling of the b --f 01 change is will coarsen the ,5 grain size and, because of the
necessary to raise them to the status of nuclei. reduced area for nucleation, retard 01formation
The stability of the embryos was explained by during cooling, i.e. necessitate increased under-
supposing that the interface is an array of cooling.
dislocations, which is geometrically stable al- Remnant textures imply that crystallographic
though thermodynamically unstable. The pre- relationships exist in both OL--f B and /3 --f 01
sence of alloying elements stabilizes B by pinning transitions. Orientation relationships are pos-
these dislocations and thereby retarding growth sible in diffusional transformations ; Widman-
of the embryos during cooling. statten figures are well known examples. These
Objections to this theory are: (a) at the arise because, by matching specific planes in
cooling rates used (5” C/hour) in the thermal the two phases, it is possible to obtain a low
analysis work, 01formation must be diffusional energy interface and minimise the energy barrier
(Section 2). A dislocation interface must thus to nucleation. Crystallographically related grain
move by a process analogous to climb and it is boundary precipitates can also exist 29) but in
difficult to imagine solute atom pinning having this case the product will be related to only one
a significant effect. The martensitic growth of of the contiguous grains (except in unique
a would be more susceptible to this effect and boundary orientations). Thus during heating
so it is important to examine how solutes affect only a fraction ++ of the LXgrain orientation will
the p --f OL change at various cooling rates ; be transmitted to the B phase and only a
(b) if notwithstanding the above argument, fraction of this amount will be passed back to
pinning does retard the growth of o(, it should a due to grain boundary nucleation from /?
also retard the opposite movement of the inter- during cooling. Increasing the time or temper-
face. Hence, alloys should retain larger 01 ature in the ,5 induces grain growth, which by
embryos than pure uranium and require less virtue of the disappearance of some /? grains,
undercooling to initiate 01 formation; this is
contrary to observation 27); (c) this inter- t This has been demonstrated in dilute uranium
alloys 28).
pretation of the thermal analysis results ignores
tt Considerably less than half because the most
the influence of p grain size on the kinetics of favourable nucleation sites are multi-grain junctions sO)
the /3 + a transformation. and because, statistically, some nuclei will be random.
THE &2$/f? PHASE TRANSITION IN PURE URANIUM 45
destroys the ‘memory’ of more and more of the A.E.R.E. Harwell, which helped to clarify the
original OLgrains, and thus leads to a reducbion ideas here presented.
in the remnant texture.
References
l) B. A. Bilby and J. W. Christian, Inst. Met.
5. Future Work
Monograph and Report Series No. 18 (1955) 121;
The gaps in our knowledge of the trans- H. K. Hardy and T. J. Heal, ibid, p. 1.
formations are obvious from this review. By 2) D. Hull and R. D. Garwood, ibid, p. 219
way of summary, we briefly suggest a number 9 D. W. White, J. Metals 7 (1955) 1221
of lines of investigation which should prove 4, M. D. Jepson, R. B. Kehoe, R. W. Nichols and
G. F. Slattery, Proc. of the 2nd Geneva Conf.
particularly fruitful : 6 (1958) 42
(4 A detailed kinetic and metallographic study 5, Mlle J. Beaudier, G. Cabane and P. Monturat,
of the TTT diagram for the /I --f OLchange Rev. MBt. 58 (1961) 3
9 P. H. Dixon and J. Burke, Inst. Met. Symposium
(in pure uranium) and of the effects of
on Uranium and Graphite (1962)
thermal history. P. Lehr and J. P. Langeron, Rev. M&. 4 (9)
‘1
(b) A continuous cooling curve study to obtain (1958) 829
a proper estimate of MS and the critical 9 F. G. Foote Progress in Nucl. Energy Series V,
cooling rate, and an indication as to whether 1 (1956) 105