J. Phys. Chem. Solids
ON
THE
Pergamon Press 1958. Vol. 6. pp. 213-222.
MECHANISM
OF
ON 0.
FLINT
OXIDE
FILM
FORMATION
ZIRCONIUM and J. H. 0.
VARLEY
Atomic Energy Research Establishment, (Received
Harwell, Berks.
27 June 1957)
Abstract-Some experiments are described that indicate the extent to which the initial surface condition of zirconium controls both thermal and anodic oxidation of the metal. Some observations of the effect of ultraviolet light are also reported. The interpretation suggested for these observations is in terms of the possible variation in electronic work function and defect state of the oxide depending upon the physical condition of the metal surface. These variations will affect the efficiency of anodic film formation, and a qualitative scheme of possible mechanisms whereby film growth is controlled is described.
1. INTRODUCTION EXPERIMENTS to determine the details of the mechanism of oxide film growth on zirconium during thermal oxidation have been carried out by CHIRIGOS and THOMAS.(‘) Using the marker technique introduced by PFEIL,@) they observed that thermal oxidation proceeds by oxygen (anion) migration through the oxide film towards the metal/metal-oxide boundary. These observations have been confirmed in other studies.(ss4) Recently the authors have reported an experiment in which ion migration in zirconia at room temperature under comparatively low electric fields has been studied by chemical detection.c5) Interpretation of the results again indicates that, at least for small electric fields, oxidation proceeds by anion migration through the zirconium oxide. The details of film growth under anodizing conditions are not necessarily the same, since high electric field strengths are used in these circumstances, and the ratio of the mobilities of the metal cations and oxygen anions is not independent of field strength across the oxide for high fields, as it is for low fields. The larger ionic charge on the metal cations may allow oxide growth to occur by both anionic and cationic migration under high field conditions, particularly if the difference in the activation energies for movement of cation and anion defects amounts to only a few tenths of an electron volt. 213
It has long been recognized that the anodizing behaviour of metal surfaces depends upon the physical and mechanical state of the initial surface. In this paper experiments are reported in which a study has been made of the variations in oxide growth on zirconium metal under different surface preparations. Both thermal growth and anodic growth have been studied. 2. EXPERIMENTAL
DETAILS TIONS
AND
OBSERVA-
A number of specimens, 5 x 1 x 0.02 cm, was cut for test from Kroll metal zirconium sheet produced by Murex Ltd. The surface condition as received was bright rolled. A nominal analysis gave hafnium 0.5-3.0, carbon 0.15, oxygen 0.8, and nitrogen 0.04 per cent. For the determination of film thickness (thermal and anodic) a series of reference strips of zirconium was anodized by the standard technique described in Section 2.2.2. After passing through the initial incubation period, the totally immersed strip was anodised for 2 min at 1 V and then withdrawn 1 cm and the process repeated at 1 V increments up to 230 V. No staining due to electrolyte creep was observed, a demarcation line only forming between each lTcm band due to high current density at the solution line. In this way strips having a range of interference colours, in well-defined bands, equivalent to formation
214
0.
FLINT
and
potentials of l-23OV or thicknesses of 20-4600 A,($) and showing four orders of colour, were prepared. Above 230 V the colour of the film was greenblack, and it was difficult to discriminate fine differences in colour. 2.1 Incubation Period Study When a specimen of zirconium is anodized, there is a period of variable duration in the initial stages of the process-the incubation periodwhen the voltage drop across the cell remains constant for a fixed current density. At the end of this period film growth starts, and the voltage required to maintain the same current density increases, not necessarily in a linear manner, as the oxide film thickness increases. The incubation period varies with the initial surface condition of the zirconium metal with its protective oxide film. Some study of the factors affecting this variation has been made with the following results. There is considerable variation in the incubation period of cold-rolled, degreased, sheet specimens of zirconium, which is thought to be due to variations in surface hardening produced in the final reduction stages during cold rolling. Some confirmation of this was obtained when a batch of zirconium which exhibited a long incubation period (- 20 min) in the “as-received” cold-rolled condition, had a very short incubation period (- 20 set) after the surface layer had been removed with a mild fluoride pickle. The importance of a distorted surface layer (Beilby layer) is also supported by the following experiments. If zirconium with a long incubation period is ground dry with an abrasive such as emery or Carborundum, as the grinding time is increased up to 2 min the incubation period is progressively reduced to a minimum value ( N 30 set). However, dry grinding for longer periods than N 2 min and to fine (400) mesh abrasive, restores the long incubation period (N 20 min). Dry grinding for very long periods (- 5 min), to produce a “burnt” finish on the metal surface to be anodized, results in an incubation period of more than 2 hr. On the other hand, zirconium metal which is prepared by wet grinding down to fine (400) mesh abrasive under water always results in short incubation periods of N 20 set, (cf. 20 min for similar material dry ground).
J.
H.
0.
VARLEY
Once film growth does start, the rates of increase of voltage with time for a given current density are in the approximate ratio of 2 : 1 for wet-ground and dry-ground material respectively. The voltage/time rate for pickled material was half-way between the rates for wet- and dryground metal for the particular batch of material chosen. The greater voltage/time rate for wetground metal surfaces has been observed on two samples of metal from different batches, which showed different absolute rates. While these results may not be exhaustive, inasmuch as the anodizing voltage/time rates during oxide film growth vary considerably between different batches of zirconium metal similarly prepared, nevertheless it is held that the ratio of the rates under different surface preparations is approximately constant. In addition, it is possible to prepare zirconium metal in such a manner as to give long (dry grinding) and short (wet grinding or pickling) incubation periods. 2.2 Conjoifgt Therm& and Anodic Oxidation To elucidate further the mechanism of oxide film growth on zirconium metal, the formation of thermal films and of anodic and thermal films in juxtaposition has been studied on metal surfaces prepared in the different ways outlined in the previous section. The observations will now be given for the several experiments in turn. 2.2.1 The zirconium
thermal formation
of oxide films
on
In these experiments oxide films were grown on zirconium sheet by heating in air at 300°C for various times on a hot plate. The oxide thicknesses were estimated from the interference colours produced. The metal sheet was prepared by degreasing first in acetone and then either dry grinding, or pickling the surface. “Aswet grinding, received” metal, degreased, was also oxidized. The results showed that metal in the “as-received”, dry-ground, wet-ground, and pickled conditions oxidized thermally at progressively lower rates. The very low,rate for a pickled metal surface was similar to that quoted for pickled van Arkel zirconium.(‘) Fig. 1 shows the oxide thickness/time relation at 300” C for the four types of surface preparation. According to the theory for the growth of very thin films at constant voltage@) there should
MECHANISM
OF
OXIDE
10 20
0
FILM
30
40
50
FORMATION
60
70
60
ON
90
ZIRCONIUM
215
100 110 120
Time, min
FIG. 1. Thermal oxidation of zirconium in air, at 301)” C.
be a linear relation between film thickness and the logarithm of the time required to form the film. The results are therefore replotted in this manner in Fig. 2. - - -
evidence on thermal growth of wery thin films (< 250 A) is not entirely conclusive through lack of sufficient experimental evidence. Electrondiffraction observations on these thin films, formed on “as-received” metal, showed that a sharp diffraction pattern developed at about the film thickness corresponding to the point 0 in Fig. 2.
in--r -
loo-
-
i
4 A.R. =as receivec D.G. dry ground
=
I’
.c E 3 z
-
lO--
-
-
\ . \ LV.G ‘,
i
O_ c
‘2
0.006
-L1
1
:, 0.. 01.
$7
0
0.014
0
-1
A
LI
FIG. 2. Relationship between log t and reciprocal thickness.
It will be observed that there is differential growth behaviour when the oxide films are very thin, but that after a certain critical thickness has been reached, further thermal growth is the same for both wet- and dry-ground metal surfaces. The
2.2.2 Anodic filmgrowth on a thermal oxide substrate In these experiments thermal films were formed on specimens of zirconium metal, the surfaces of which had been prepared by (a) dry grinding and (b) wet grinding. The subsequent growth of anoditally formed films on these oxidized specimens was then studied as a function of surface preparation for different thicknesses of the thermally formed film. Anodic oxidation was carried out in aqueous ammonium borate (20.6 g/l.) at 20” C, at a constant current density of 2.5 mA/cm2, using a platinum cathode. (a) Dry-ground material. The anodizing voltage/ time curves obtained on dry-ground metal are shown in Fig. 3 for various thicknesses of the preformed thermal film. The slight curvature in the voltage/time relation for a specimen with no thermal film is enhanced as the thickness of the thermal film is increased. The voltage/time rate is increased at high voltages under prolonged initial thermal treatment and approaches the value obtained on wet-ground material (- 20 V/min). The increase in film thickness, determined by changes in interference colours, did not occur until the voltage across the specimen had reached that value corresponding to the equivalent formation potential of the thermally preformed film.
216
0.
FLINT
and
J
H.
The total charge passed during the time when the formation voltage increased to an arbitrary voltage of 70 V, that is to say, from the end of the incubation period to the instant when 70 V was reached, was sensibly constant to within N 10 per
60
1
2
3
4
5
6
Time,
7
8
9
IO
11
min
FIG. 3. Anodic growth on thermal substrate material.
; dry-ground
cent and was independent of the thickness of the thermal oxide previously formed. The total charge passed during the time was 2-23 times that which is calculated as necessary to form a 70 V anodic film assuming that all the current is film forming. To complete film growth additional charge was passed at the constant finishing voltage (70 V) for 2 min. The final film thickness, as judged from the colour, was close to, but slightly greater than, that expected for a 70 V film in all cases. Additional charge was also passed during the incubation period at the start of anodizing. The surface was dry ground in such a manner
1
0
4.
3
2 Time,
FIG.
VARLEY
as to produce a relatively short incubation period N 24 min. Thermal treatment did not completely remove this incubation period, but reduced it to between N 20 set and 1 min 20 sec. (b) Wet-ground material. The anodizing-voltage curves for wet-ground material are shown in Fig. 4; again, each curve is for a given thickness of the preformed thermal film. Comparison with the results of Fig. 3 for dry-ground metal reveals significant differences in anodizing behaviour resulting from the details of initial surface preparation. The anodizing voltage necessary to maintain the constant current density rises quickly to the value corresponding to the equivalent voltage across the thermally formed film of given thickness. The extrapolated portions of the anodizing-voltage curves generally intersect at approximately the voltage equivalent to the thickness of the thermally formed film; the equivalence factor has been estimated at 20 AjV.@) The curvature of the voltage/ time curves is opposite in sign to that obtained on dry-ground metal. The time taken to grow the anodically formed film on wet-ground metal to the same arbitrarily chosen finishing voltage of 70 V (1400 A), is with the forshorter by a factor of N 2 compared mation time on dry-ground metal. Furthermore, the total charge passed during anodizing decreases as the thickness of the preformed thermal film on wet-ground metal increases. Thus, for wet-ground surfaces, the anodizing film growth efficiency remains more nearly constant and independent of the thickness of the preformed thermal film. For dry-ground metal the anodizing efficiency decreases as the thickness of the preformed thermal
70
0
0.
min
Anodic growth on thermal substrate;
wet-ground
material.
MECHANISM
OF
OXIDE
FORMATION
FILM
film increases. In the latter case, therefore, appreciable electronic charge must flow through the thermal film to establish the necessary potential drop across the film which will promote further (anodic) film growth. This necessary total electronic charge must be increased as the thermal film thickness is increased. The voltage/time rate at high voltages on wetground metal, when anodic film growth is taking place, is approximately double that of dry-ground metal provided that the heating time for thermal film formation in the latter case is not too long nor
ON
ZIRCONIUM
217
initial anodic film; finally the voltage increases with time at a nearly constant rate once the final anodic film begins to grow. There is one important difference, however, from the results obtained for anodic growth on thermal films formed at 320” C on dry-ground material with no preformed anodic film. This is that the time to complete the formation of the 70 V film during the final anodizing decreases with increasing total thickness of the preformed anodic plus thermal film. The total charge passed thus decreases similarly to the observed anodizing be-
80 70 60 50 40 30 20 10 0
1
2 Time,
3 min
FIG. 5. Final anodic film growth on anodic/thermal
the temperature too high. If these conditions are not observed, the differences between wet- and dry-ground surface preparation are reduced, and the growth rates approach the value of the wetground material. 2.2.3 Anodic growth on compound thermal/anodic oxide substrates In these experiments an anodic film was first formed to some chosen formation voltage. The specimen was then thermally oxidized so that the total oxide film was thickened. Finally, the oxide film was grown further by anodizing to the final arbitrary finishing thickness of - 1400 a under a final anodizing potential of 70 V. (a) Dry-ground material. The final anodizing voltage/time curves are shown in Fig. 5. It will be observed that the preformed anodic and thermal films both tend to retain their identity. The thermal films were formed at 460” C in this case. Thus the final anodizing voltage rises quickly to the voltage of the anodic film formed first; then the voltage is established slowly over the thermally formed film as in the experiments on metal with no
oxide substrate; dry-ground
material.
haviour of wet-ground material with preformed thermal films. This difference is ascribed to the higher temperature (460” C) which was used to produce subsequent thermal film growth on an oxide film which had already been formed anodically. (b) Wet-ground material. The final anodizing voltage/time curves are shown in Fig. 6. These curves are similar to those obtained on wetground material having only preformed thermal films (Fig. 4). There is a rapid increase of anodizing voltage up to approximately the equivalent formation voltage of the composite film (anodic plus thermal). The latter are in rough agreement with the values of the potential at the intersections of the extrapolated linear initial and final parts of the curves. The curvature in the voltage/time relation is again opposite in sign to that observed on dryground material. This curvature occurs where the potential across the thermal film is being established. Once the equivalent formation potential has been established across the composite anodic/ thermal film, further anodic growth begins, resulting in the ultimate nearly linear part of the
16OOr
801
1400
70
t200m
/
,
60
1000
~50
800.
40
600
30
400
20
200-
10
Time, FIG.
h. Final
anodic
film growth
on anodic
~Iol~age~tirn~ relation, up to the chosen finishing voltage of 70 1’. The total charge passed again decreases with increasingcomposite thicknessof the preformed films. 2.3 The ,Yffect of Ultra-violet Lighton Oxidation Kutes Some study has been made of the growth bchaviour of oxide films on zirconium metal prepared by wet and dry grinding. It was found that, pYoz@ll’edthe metal surface was ~I~~lrniil~~ted during the very early stages of fiIm growth : (a> growth
(b) growth
was y~?~~.~ef~on
These observations hold whether the oxide grown anodically or thermally. The results shown in Table 1.
was arc
min
thermal
oxide
substrate;
wet-ground
materiul
3. DISCUSSION
‘I’he proposed explanation of these experimentat observations is based on the supposition that, whereas at low tield strengths across the wide the growth occurs predominantly by anion movement, at high field strengths growth occurs both hy anion and cation movement under favourable circumstances. However, experimental conditions which Favow the evolution of oxygen gas and the transport of electrons across the oxide, in addition tn the growth
Of the oxide
impede
R~o~‘~rn~nt, SO that
anion
by ion absorption,
migratiol~ will contribute Thus, anion movement trons which are freed from the osygcn 0xicle:solution
mrty
only the cation to oxide growth. is impaired when clcc-
interface
are
trapped
vacancies
;LS they move acrOss the w&k.
vacancies
will have
t!leir
etkctivt.
ions at the at
anion
‘l’hc anion
positive
charge
MEC~ANrSM
OF OXIDE
FILM
reduced with the result that their mobility under the action of the external field will be less. Under strong fields, the mobility of an ion defect, which is proportional to its successful jump frequency, v, depends upon the exponential factor exp (XeBlkZ’), where z is the effective valency of the defect and F the field strength. It follows that a smaller value of a resulting from electron trapping at anion vacancies will seriously reduce the anion contribution to oxide growth. The activation energy controlling the jumping of anions into neighbouring vacancies in the absence of an external fieid will be reduced if the adjacent anion vacancy has trapped an electron. This effect will counteract the reduced mobility of the anion defect. At high field strengths it is considered that the resultant effect will be to reduce the anion mobility. Two possible cases must therefore be considered : (1) If the surface condition at the oxide/ solution interface is such as to favour the reaction O--- -+ O+Ze at low field strengths, then many electrons will pass through the existing oxide film and the anion vacancies will become saturated with trapped electrons. As the negative charge develops in the oxide film, the passage of electronic current across the film will become more difficult and the external field nec,essary to maintain a constant current density must be increased. In addition the “double layer” of trapped negative charge in the oxide and compensating positive charge at the metal/oxide interface will set up an internal field acting in the same direction as the external field. Consequently ionic defect movement across the metal/oxide barrier will be enhanced and film growth will occur. At constant current density the ratio of ionic to electronic current will increase. If all the anion vacancies are saturated, only the cation migration can contribute towards oxide growth. The negative space charge induces a positive charge at the metal surface in contact with the oxide. The resultant internal field reduces the energies of electrons close to the metal/oxide interface below the Fermi energy of the electrons in the metal. A high-resistance barrier to further electron conduction across the interface_ from oxide to metal is thus set up 3s electrons become trapped. (2) If the initial surface condition does not favour the reaction O- - +O f2e at low fields, then
FORMATION
ON ZIRCONXUM
219
electronic flow across the oxide will be restricted and will constitute a smaller fraction of the total constant anodic current density. In such a case the (high) field established across the existing oxide film will maintain a current which is predominantly ionic. Oxide growth will thus occur by both cation and anion vacancy migration in this event since no appreciable electronic current will have passed for any length of time across the existing oxide film; ~onsequentIy few anion vacancies will have trapped electrons. The growth rate will therefore be approximately twice as great as in the first case described above, provided that the normal contributions to growth from cation and anion movement are nearly equal. On this model the experimental observations can be satisfactorily described in a qualitative manner. It is suggested that the atomic structure of dryground metal will be more disordered in its surface layers than will that of wet-ground metal. Also “as-received” metal will be surface hardened to a degree depending upon the final cold-rolling conditions. This surface hardening will modify the electronic work function between the metal and the metal oxide. Thus the efficiency of the reaction O--+0+&, which will depend upon the work function, will differ according to the state of the metal surface. Furthermore, it is possible that the more disordered metal oxide on the surfacehardened material will contain many more anion vacancies than in undistorted oxide, so that electronic conduction through the oxide may occur via the resulting electron defect band, rather than via the electronic conduction band of the more perfect oxide. A low electronic resistance may thus be expected if there is severe surface disorder, and this is consistent with the prolonged incubation period observed in this case. Ultimately, under the low field strength which maintains the constant total current density across the disordered oxide, a redistribution of anion vacancies will occur and film growth will then begin. The voltage/time curve will show a positive curvature, as the field strength will increase with the anion vacancy redistribution. The film growth rate on surface-distorted metal will be low, since the anion vacancies will be rendered relatively immobile by electron trapping. Oxide films will grow faster on the more perfect metal basis of wet-ground and pickled metal surfaces
220
0.
FLINT
and
than on metal prepared by dry grinding or in the “as-received” condition. The experimental observations of composite anodic/thermal film growth can then be interpreted as follows. Anodic growth on dry-ground material takes place via the mobility of positive ions only. Anodic growth on wet-ground material occurs by the motion of both cation and anion’ defects and is therefore faster than film growth on dry-ground metal. Thus the growth of composite films on dryground metal occurs by: (a) the formation of an anodic film through the movement of cation defects only; (b) the growth of a thermal film by anion defect movement, resulting in the formation of the thermal film below the anodic film; and (c) the growth of a final anodic film above the existing films by cation defect movement only. During this stage the anodizing voltage necessary to maintain a given constant current density will first rise quickly to the formation potential of the first anodic film (a). Then the voltage will rise slowly as the potential is established across the thermal film which, formed from distorted metal, is a relatively good electronic conductor. Electron trapping at anion vacancies will result in the formation potential across the thermal film (b) ultimately being established. During the final anodic film formation growth by cation defect, migration will occur along with electron trapping at newly formed anion vacancies, and the voltage will rise to the final film-formation potential. There are now two points to be considered. First, why do the anodic and thermal films formed consecutively on dry-ground metal retain their identity inasmuch as the voltage/time relationship during final anodizing indicates that this is so? During the formation of the first anodic film on dry-ground metal, electrons are trapped at anion vacancies formed at the metal/oxide interface during growth, and film formation occurs by movement of cation defects across the film. After the anodic film has been formed and the external voltage removed, there will be a readjustment of positive charge at the metal/oxide interface and throughout the oxide to neutralize the electronic space charge arising from trapped electrons in the oxide. Subsequent thermal film formation will occur by the production of fresh anion vacancies at the
J. H.
0.
VARLEY
metal/oxide interface which will move through the oxide towards the oxide/air interface. In this way fresh (thermal) oxide will be formed beneath the anodic film by the effective movement of oxygen ions towards the metal. If the anion defects in the initial anodic film have trapped electrons and are themselves trapped by interaction with cation defects then the anodic film will retain its identity and during the final anodizing the thermal oxide, formed beneath the first anodic film, must first be brought to the same physical state (electrons trapped at anion vacancies plus positive ion defect distribution) as the anodic film. Prolonged heating during thermal film formation, or heating at a high temperature, will reduce the extent of the deformed surface region so that under these conditions final anodic film growth will show voltage/time characteristics which approach those observed on less distorted wetground or pickled surfaces. The second question to be considered is: Why should the oxide film thickness be nearly the same when an anodic film is formed to the same finishing voltage, independent of whether the mechanism of anodic oxidation involves the movement of cation defects (dry-ground metal) or both anion and cation defects (wet-ground metal)? The theory of the formation of thin oxide films at suitably low temperatures under strong electric fields(“) indicates that below some critical temperature the growth rate effectively becomes negligible once the oxide film has formed to a given thickness which is uniquely related to the final formation potential. The linear growth rate, dt~/dt,is given by: &/&
=
cOnSt
, e-rafkT
. ,ze
VbixkT. (1)
The constant is proportional to the number of active sites per unit volume on ‘the metal surface which feed defects into the oxide, to the volume of oxide produced per ion moved, and to v,,, the ion defect vibrational frequency. W is the height of the energy barrier controlling the frequency of successful jumps into the oxide across the metalioxide barrier , 2 is the valency of the ion, Y the potential across the film of thickness X, and b is the half jump distance of the rate-controlling jump process which is considered to be the excitation of ion defects across the metal/oxide interface. Experjmentally, the growth rate of anodically formed
MECHANISM
OF
OXIDE
FILM
films becomes small in a time which is N 100 set at the final formation potential. It has been found that a ZrO, film N 1400 A thick doubles itself in 17 hr at 70 V. The average value of dx/dt, the growth rate, after its initial decay at the formation potential is thus N lo-lo cm see-l. Taking the constant in equation (1) to be lo4 cm set-l (cf. CABRERAand MoTT(~)), it follows that exp (- W/kT+zeVb/xkT)
21 10-14.
The relation between the formation potential V and the corresponding thickness of the anodic film x is thus: x ~{xeVb/(w-32kT)).
(2)
If now two processes are taking place simultaneously, namely cation and anion transport, during anodic film*formation under favourable conditions (e.g. wet-ground surfaces), then equation (1) will be modified to contain two exponential terms on the right-hand side. If the growth rate becomes small at approximately the same film thickness, for a given voltage, as in the case of anodic film formation by cation transport only, then, given that the coefficients of the exponential terms are of the same order of magnitude, it is a necessary condition that:
de
x
-=
eV
l
w,-32kT
=
z&a w,-32kT
(3)
where the subscripts c and a indicate cation and anion parameters respectively. Zirconium oxide has a crystal structure equivalent to a distorted calcium fluoride structure. If it is assumed that the cation defects are interstitial metal ions, then b, N b,2/2, the anions moving by lattice-vacancy migration. Also z, = 4, z, = 2, and 32 kT N 0.8 eV (T 2 300” K), whence (23/*w,--w,) N (23/2- 1)0.8. Experiment indicates that dxjdv N 20 A/V; taking b, N I.3 A this gives W, N 1 eV, so that w, N 1.36 eV. The value chosen for b, has been calculated assuming a calcium fluoride structure and a density of 6 g/cm3 for ZrO,; 6, is then the half distance between adjacent O-- ions. The calculated difference of 0.36 eV between cation and anion energies (w, > w,) is sufficient to imply a small contribution to thermal film growth at temperatures around 300” C. The ratio of cation
FORMATION
ON
ZIRCONIUM
221
to anion contribution as a result of this difference is N 7~ 10-4, so that thermal growth of oxide on zirconium in this temperature range is essentially anionic in nature. It remains to attempt to describe the effects of ultraviolet light on ZrO, film growth. Experimentally, it is observed that ultra-violet light accelerates the formation of oxide films on wetground material, but impedes their formation on dry-ground material. This result is independent of the method of formation; it is the same for thermal and anodic growth. The other important feature is that these effects are not observed if the film is not illuminated during the early stages of formation, when it is very thin. This indicates that the ratecontrolling step lies in the barrier to ion (defect) migration set up at the metal/oxide interface, as in the MOTT theory. To account for the observations it seems necessary to postulate that the absorption of photons plays a dual role. Thus: (1) Absorption increases the population of electrons in excited states above the Fermi energy in the metal. This effectively reduces the electronic work function of the metal; consequently the double-layer field set up by electron distribution at the metal/oxide interface will be smaller. Ion defects will move across the barrier more easily the smaller it is and oxidation will be faster. In addition the electron availability at the oxide/air interface will be increased in the case of thermal oxidation and hence, as CABRERAhas already explained, (l”*ll) oxidation will be enhanced by the increased field. (2) Absorption of photons within the oxide produces excitons or free electrons and positive holes. As a result, electrons can become trapped at anion vacancies, so reducing the effective charge and hence the mobility of these defects. This effect will tend to reduce the oxidation rate. If the work function barrier between metal and oxide is high, as has already been suggested is the case for relatively undistorted metal surfaces (wetground), then an effective reduction of this barrier will have a significant effect and will tend to increase the oxidation rate. If the work function barrier at the metal/oxide interface is low, then reduced mobility of anion defects will be more important and will tend to lessen the oxidation rate.
0.
222
FLINT
and
This would account for the increased and decreased thermal oxidation rates on wet- and dryground metal surfaces, respectively. Moreover strong photon absorption will limit the effect to the early stages of film growth. In the case of anodic film growth, the increased rate in the early stages of formation under ultraviolet illumination is possibly due to excitation of electrons on oxygen ions at the metal/oxide interface assisting the production of anion defects, i.e. increasing the jump rate of oxygen ions into available sites on the metal surface. The reduction in barrier height discussed above will have a similar effect. The formation of anion vacancies may be more difficult in the case of less distorted metal surfaces than for surface-hardened metal. In the latter case, it has already been suggested that the building up of space charge within the oxide is a prerequisite to film growth. The release of some of this space charge by photon absorption, possibly indirectly by interaction with excitons or positive holes, will reduce the efficiency of film formation. This description may account for the observed increased and decreased anodic oxidation rates on wet- and dry-ground surfaces, respectively. 4. SUMMARY
The experimental results qualitatively on the suggested
can be described basis that:
(1) The electronic work function between surface-hardened zirconium and its attendant oxide is lower than that between more perfect metal and oxide. (2) In the case of a relatively undistorted metal surface, anodic oxidation proceeds by anion and cation defect migration.
J.
H.
0.
VARLEY
(3) The electronic conduction through the oxide on relatively distorted metal is high. This results in a space charge forming within the oxide by electrons trapped at anion vacancies. The anion contribution to anodic film formation is thus severely reduced, and oxidation is essentially cationic. (4) The electronic space charge in the oxide induces a positive charge at the metal surface in contact with the oxide. This has two effects: (a) An internal field in the same direction as the external field enhances cation migration across the metal/ oxide barrier. The effective activation energy for cation migration is thus reduced. (b) The energy levels of electrons in the vicinity of the metal/ oxide interface are so reduced relative to the Fermi energy of the electrons in the metal that for a sufficiently large space charge a high-resistance barrier is set up to electron penetration into the oxide.
Acknowledgements-The authors wish to thank Dr. H. M. FINNISTONand Mr. F. P. CLARKEfor their interest in this work. REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.
CHIRIGOSJ. and THOMASD. E. W.A.P.D. 53 (1952). PFEIL L. B. r. Iron Stl. Inst. 119, 501 (1929). THOMAS D. E. W.A.P.D.-T-186 (1954). FLINT 0. Unpublished work. FLINT 0. and VARLEY J. H. 0. Nature Load., 179, 145 (1957). CHARLE~BYA. Acta Met. 1, (1953). CHIRIGOSJ. and THOMASD. E. W.A.P.D. 98 (1953). CIIARLESBYA. Proc. Phvs. Sot. B66. 317 (1953). CABRERAN. and MoT~N. F. Rep. grogr.‘ Phyi. 12, 163 (1948-49). CABRERA N., TERRIEN J. and HAMMON J. C.R. Acad. Sci. Paris, 224, 1558 (1947). CARRERAN. Phil. Mug. 40, 175 (1949).