EPRI-NASA Cooperative Project on Stress Corrosion Cracking of Zircaloys

This report presents the results of EPRI RP 455, a 30-month program of research aimed at developing improved understanding of the stress corrosion cracking (SeC) mechanism considered responsible for pellet-cladding interacti~n (PCI) nuclear fuel failures. PCI failures originate at the fuel side of the Zircaloy cladding, and therefore, the major objective of EPRI Project 455 was to define the sec mechanism of Zircaloy in environments relevant to those at the inside sUI'face of operating fuel cladding. The work was conducted in complementary prograDis at SRI International, NASA-Ames Research Center, and, under subcontract, Ar(:onne National Laboratory and Massachusetts Institute of Technology. The results of examinations of the inside surfaces of irradiated fuel cladding from two power reactors showed that Zircaloy cladding can be exposed during service to a number of potentially aggressive substances. Zircaloy samples were screened at reactor operating temperatures for susceptibility to sec induced by such substances. The most aggressive were iodine, cadmium, and iron-contaminated cesium. Detailed studies were made of iodine-induced sec of well-characterized samples of Zircaloy sheet and tubing. The results of exper ments on pressurizeu, cladding-g~ade tubes indicated that a threshold stress must be exceeded for iodine sec to occur. The threshold s~ress was sensitive to the microstructure of the Zircaloy and probably to in-reactor irradiation, but was not affected by the presence of an oxide film or by iodine concentration within the range studied. lbe existence of a threshold str~ss indicates that crack formation probably is the key step in iodine sec. A detailed investigation of the crack formation process showed that the cracks responsible for sec failure nucleated at locations in the metal surface that contained highe~ than average concentrations of alloying elements and impurities. A four-stage model of iodine sec is proposed based on the experimental results and the relevance of the observations to FeI failures is discussed.

Results of Press\~ization Tests on Unflawed Specimens of 7AH1I-S Zircaloy-2 Exposed to Iodine at 590 K  . , BLANK AGE .  The incidence of PCI failures can be kept acceptably low by uSing the very s low rates of local and overa~l powe r change recommended by nuclear fuel vendors .
However, this solution is expeu i ve because of the substantial cost s associated with the operation of the pl ant a t partial capacity throughout the long periods while the power is being slowly increas ed. Thus , there is ar. important need to understand and improve the sec behavior of Zircaloy cladding so that increased operational flexibility Rn duced costs can be achieved while maintaiuing the frequency of fuel fa!l'" ~~ an acceptably low level. The research program described in this . ,)( wa~ i ntended to meet that need. The r e s ul ts of these research programs and our interpretation of them are presented in de tai l i n this report. The remainder of this sect ion summarizes the major findings . To allow easy cross correlation with the deta iled descr iptions given later i n t he report, t he s ummary is divided into subsections that correspond to the major report sections.

Characteriza tion of Deposits on the Inside Surfaces of LWR Fuel Cladding (Section 2)
This study was undertaken to generate information on the nature of the chemical envir orunent to wh i ch the inSide surfaces of LWR fuel cladding a·', exposed during reactor service.
Scanning elec tron microscope examinations were made in the ANL hot cell s of the deposits on the inside surfaces of power reactor cladding from two sources. One lot of cladding was from some f uel rods that had fa iled by PCI (Maine Yankee), whereas the s econd lot of claddi ng (H.B. Robinson) wa s f r om unfa i led rods.
The predominant types of deposits found on t he i ns i de surfaces of Ma ine Yankee cladding were: • Sharply defined stripes that were formed opposite cracks in the fuel . The stripes consisted of a band of particles of various sizes and compositions, presumed to represent such compounds as CS 2 U0 4 (or similar Cs-U-O compounds), Cs 2 0, CsI, Cs 2 Te and mi xtures of them, as wel l as particles containing manufacturing impurit ies . On either Side of t he central band was a parallel band of tiny nodules (-1-2 ~) that were intimately attached to the surface and contained Cs, U, and Te.
• Heavy deposits of a ceramic-like material that contained mainly U and Cs (and pres ably oxygen) and were located opposite pelletpellet interfaces • Occasional circul r mound deposits that appeared to consist of the same ceramic mate ial as t he pe llet-pellet interface deposits.
• Pieces of fuel bo ded to the c l adding by the same Cs-U ceramic material.
The likely mechanism of tra sport of fission products from the fuel to the Maine It was not possible to identify a single substance as the active agent in PCI f ailurps. Candidates considered likely were iodine (released from CsI by radiolys is), cadmium and possibly iron (ill liquid cesium). Iodine was chosen to be the active agent in most of the laboratory studies of SCC discussed below because its aggressiveness was more reproducible than that of the other substances.

1-3
Characterization of Experimental Materials (Section 3) SRI Im.ernational was responsible for procurement, distribution , and characterization (If the Zircaloy sheet, plate, and tubing needed for the experimental phases TWenty-eight of the sheet, plate, and tubing products were screened to determine their optical microstructures, hardnesses, and basal pole figures. As anticipated, the microstructures of the thicker products generally were coarser than those of the thinner products and the thinner products generally were harder. The basal pole figures of the tubing materials showed the familiar ±30 o texture. The reduction 6chedule intended to reproduce this texture in sheet and ylate was successful as was the schedule designed to generate a high basal pole density in the sh~et normal direction. Hnwever, the schedule tha t was intended to generate a high basal ,pole density in the long transverse direction was found to be unsuccessful.
Tbr01h-thickness texture variations could be detected in all products but were parti~ularly noticeable in the thickest plate materials.

Addit.onal characterization studie s on the tubing materials included
Biaxial tensile tests at -600 K that indicated yield stresses ranging from 210 to 490 MPa and burst stresses from 370 to 550 MFa, dependent primarily on alloy heat treatment.
*Zir loy 'Inelastic Deformation Project, with General Electric (NP 500), Stanford Uni rsity (NP 507) and MIT. +rbe ublng was from the special lots procured by EPRI for RP 251, 249, 355, and 507, and also used by NRC-RSR in its Zircaloy ProgrBm. 1-4 • Measurements of hoop residual stresses at r oom temperature that indicated predominantly compressive re s idual s tres ses i n the i nner part of the tube wall and resultb~ in peak residua l stress e s timates of ~ 100 MFa in all three tubing material s.
• Profilometry measur ements and SEM examinations of the topography of the tubing surfaces that indicated rougbnesses in accordance with the manufacturer ' s specifications and r evealed the presence of c onsiderable numbers of large s econd-phase particles and inclusions in and near the inside surface which proved to be important in the crack f ormation stage of iodine SOC ( see below).
Crack Pormation in Zircaloys Exposed t o Iodine (Section 4) 'The processes involved in the formation of crack~ in Zircaloy samples stressed in iodine at reactor fuel c ladding operating temperatures were studied using a new technique. A spherical indenter was used to produce a small, well-defined area of tens il e strain i n a sample surface expos ed t o i odine. The s ample s urface was cleaned by a r gon ion nlilli ng and the topogra phy and chemical composit ion of the small s t rained area were charactel'i zed by scanning e l ectron microscopy.
Preliminary experiments were conducted on stress· relieved Zircaloy-4 sheet samples.
Subsequently, more detailed i nvestigations were ~onduc ted on samples of stressr elieved Zircaloy-4 and Zircaloy-2 tubing and annealed Zircaloy tubing at 630 K and 590 K, whi ch are typical claddi ng temperatur~s during reactor operation. In t he absence of iodine, cracking was observed only under the most severe mechanic,ll c onditions and was probably attri butable t o cr e ep-r upt ure f a i l ure . When iodine was present , cracks were formed under much milder mechanical conditions but not in the comp lete absence of strain. Therefore, we attributed the cracks f ormed in the presence of i od i ne to sec.
Two disti .ctly different types of iodine-induced cracks were observed in all four ~ater1al s tested. One type, classif ied as small cracks, was ob~erved in all s amples strained in iodine, and in occasional samples relatively large cracks were also observed. The small c racks generally s eemed to form at microstructural inhomogeneities (such as certai n grain boundaries i n recr ystal l ized material) that woul d be expected to be associated with large local stress concent r ations in a mater i a l undergoing plastic flow. These cracking sites were not associated with any 1-5 detectable chemical inhomogen, ei ties in the me tal. The frequency of occurrence of the swall cracks increased with increasing iodine pressure but was not very dependent on strain (indenter load ) or time under load. We did not find either an lodine pressure or a plastic strain below which cracks did not form. The lowest imposed values of these variables in the present tests were 0.03 Pa 12 and ~ 1% strain. In experiments with pressurized tubes we found s~~ilar small cracks at stresses well below the threshold stress for iodine-induced failure. Therefore, it appears likely that these c:racks do not propagate, The larger cracks were observed in about 30% of the samples and diffe red from the small cracks in that they were consistently associated wi th local concentr~ ions of impurities or alloying elements. We presume that the erratic appearance of these larger cracks was related to the probability of there being a site containing a susceptible chemical inhomogeneity within a randomly located ~ 1 mm 2 area of sample surface (i.e. , the area of the s trained region produced by the randomly positioned indenter), Because of their relatively infrequent appearance, we were unable to determine whether any of the experir'.ental variables affected the f requency or size of the larger cracks. In experiments with pressurized tubes, the larger cracks appeared only at stresses above the threshold stress for iodineinduced failure. Therefore, it appears likely that propagation of this type of crack leads to iodine see failure.
In preoxidized specimens, met al cracks always were associated with oxide cracks, which suggests that mechanical fracture of the oxide is a necessary precursor to the formation of a crack in the metal. However, iodine seemed to be able to penetrate very thin oxide fi l ms in the a bsence of any applied stress.
Mechanical cracking of oxide films a t ~ 600 K wa s briefly studied . Oxide cracking was detectable at a tensile strain of ~ 0. 4 percent and the number of cracks in the oxide increased systematically at higher strains. The frequency of cracks depended on oxide thickness--oxides around 2 ~ thick were cracked more severely than either thicker or thinner oxides.

1-6
The resu lt s of this study of iodine-induced crack formation indicate that local concentration of alloying elements and impurities in the metal surface playa key role in the crack formation precess. Although the origin of these local chemical inhomogeneities is not certain, it seems possible that they are formed in tubing as a result of surface contamination picked up during tube reduction.
Iodine-Induced Failure of l nternal l y Pressurized Zircaloy Tubes (Section 5) Tube pressurization experiments were conducted on three types of unirradiated, nuclear cladding grade Zircaloy tubing and one type of irradiated Zircaloy fuel cladding in inert and iodine-containing environments at -600 K. The objective of these tests was to learn more about the threshold conditions required for iodine see failure of thin-walled Zircaloy tubes.
The results of tests on p~e flawed and unflawed specimens of unirradiated stressrelieved Zircaloy-4 exposed to iodine at 630 ± 5 K (a typical PWR cladding service temperature) indicated that hoop stress was a more useful parameter than either stress intensity or diametral (hoop) strain for predicting iodine-induced failure.
The data supported a relationship between hoop stress and failure time that strongly suggested a t hreshold hoop stress for iodine see at about 300 MFa . At stresses just above this threshold, i~line-induced failures occurred in much shorter times and at much smaller strains than creep-rupture failures. The iodine see and creep rupture failure times and failure strains converged at higher stresses and became similar at stresses close to the burst stress.
The behavior of unirradiated stress-relieved Zircaloy-2 specimens pressur ized in the presence of iodine at 590 ± 5 K (a typical BWR cladding service temperature) was qualitativel y similar to that of the stress-relieved Zircaloy-4 specimens at ~6 30 K. The threshold stress for iodine see of the stress-relieved Zircaloy-2 was -330 MFa and the failure times at stresses above threshold were somewhat longer than for the stress-relieved Zircaloy-4, perhaps because of the different test temperatures. The tests on unirradiated stres s -relieved Zircaloy-2 also indicated that decreasing the iodine concentration from -6.0 to -0.06 mg 12 per cm 2 of Zlrcaloy surface tended to increase the failure time but did not significantly increase the see threshold stress.

1-7
Annealed (recrystallized) unirradiated Zircaloy-2 specimens showed a threshold wtress for iodine sec of about 280 MFa in tests at 590 ± 5 K. The threshold stress was not affected by a preoxidation treatment that formed a 1.5-~ thick oxide film. The annaaled Zircalo~_was found to be susceptible to iodine sec only wi thin a narrow range of stresses j·~bove. tw.,rellliOI t\t. j}:lgher stresses, ductile failures were observed in specimens stressed in the presence of iodine, and it was concluded that crack blunting d u" t·o plastic flow suppressed iodine sec at strain rates above about 3 x 10-5 sec-i. Some of t he specimens tested in the susceptible range of stresses failed in periods 01 only a few minutes. Because the annealed Zircaloy-2 exhibited both a smaller threshold stress and shorter failure time~, it was considered more susceptible than the stress-relieved Zircaloy-2 to iodine sec in tube pressurization tests at '" 590 K.
The results of a limited number of tests on specimens of irradiated Zircaloy-4 fuel cladding were qualitatively consistent with the trends in behavior exhibited by the unirradiated, stress-relieved Zircaloy-4 specimens. The threshold stress for the irradiated cladding at 630 ± 5 K appeared to be about 200 MPa--'" 30 percent less than the threshold stress for the unirradiated Zircaloy-4 and a much smaller fraction of the yield stress . In addition, the failure times for the irradiated specimens seemed to be shorter. Tests on preflawed speCimens of the irradiated Zircaloy-4 suggested that K for iodine sec of irradiated Zircaloy at reactor Iscc operating temperatures was not dramatically smaller than the values of 9 to 10 MPa'm i / 2 that have been obtained for unirradiated Zircaloy.
Although the results suggested that irradiation significantly increased the susceptibility of Zlrcaloy-4 to failure by iodine sec, proof of the deleterious effect of in-reactor service and a proper assessment of the severity of any such effect must await the results of additional experiments on irradiated and unirradiated specimens from the same tubing lot. Such experiments are being undertaken by ANL as part of an ongoing EPRI project (RP 1027, "Characterization of Irradiated Zircaloys").
Fractogrephic observations generally were indicative of a transgranular mode of sec crack growth in both irradiated and unirradiated materials, although some 1-8 features suggestive of intergranular growth were observed in unirradiated annealed Zir caloy-2 specimens. Corrosive attack of Zircaloy by the test environment was apparent on the fracture surfaces, and the inside surfaces of the specimens were pitted . The stress corrosion cracks in specimens tested close to the threshol d stress showed little indication of branching , but highly branched cracks we r e observed in specimens tested at stresses well a bove t he t hreshold stress .
The results of these pressurization tests on thin-walled Zircaloy tube specimens generally seemed more consistent with the concept th t the threshold stress is an intrinsic property of the material/environment system related to the formation of an sec crack t han with the alternative i dee that the threshold st res s is the minimum stress that is capable of maintaining a critical rat e of plastic strain.

SCC Screening Tests in Several Environments (Section 6)
The SCC susceptibility of stress-relieved Zircaloy-4 was screened in a number of environments using different test methods . M os t of the screening tests were performed at NASA-Ames Research ~enter us ing sheet and plate specimens but the aggressiveness of cesium and, to a l esser extent , cadmium was also investigated at SRI in t es ts on t ubing s pe cimens .
The gases H 2 , 1 2 , Br 2 , and C1 2 and the liquid metals Cs and Cd were shown to be capable of promoting subcritical crack growth in stress-relieved Zircaloy-4 at fuel cladding service temperatures . Further, evidence was found that 1 2 , Br 2 , and Se are all capable of severely corroding Zircaloy-4 at reactor operating temperatures in the absence vf stress. The metals Se, Sn, Te, Sb, and Ag did not cause crack extension 1n statica lly loaded specimens. It was difficult to compare t he aggressiveness of the substances that caused subcrltical crack growth because t he dif f ereing natur es of the c r acki ng processes and t he differ ing char a cteristic s of the various substanc es necessitated the us of different test methods .
The halogens were the only substances that caused crack extension in statically loaded s pe c imen s a t ~ 630 K. They s eemed to be s imi lar t o each other in their aggressiveness toward Zircaloy and their presenc resulted in very s imila r / 1-9 fracture surfaces that exhibited a mixture of cleavage and ductile rupture. The threshold stress intensities (K ) for the halogens were estimated to be about Iscc 9 MPa· m 1/2 at ~ 630 K, and i odine-induced cracking proved to be essentially immune to the presence of o~ygen.
Tn contrast, cadmium and cesium were aggressive toward Zircaloy at cladding service temperatures only when steps were taken to reduce the oxygen contents of the test enVironments to very small values. Furthermore, cesium and cadmium embrittled Zircaloy only in rising stress tests and cesium had to be contaminated with a trace of iron before embrittlement was observed. K for cadmium cracking at 633 K Iscc was estimated to be about 20 MPa · m I / 2 , but this value was believed to be an overestimate. In tests on smooth specimens, cadmium and iron-contaminated cesium generally caused embrittlement at stresses just above the yield stress, although on two occasions much smaller failure stresses were observed in the iron-contaminated cesium environment.
Gaseous hydrogen proved to be much more aggressive toward Zircaloy at ambient temperature than at cladding operating temperatures. Hydrogen was capable of promoting subcritical crack growth in statical ly loaded specimens at room temperature, but an alternating stress was required to obtain crack extension at ~ 570 K.
Because the halogens promoted slow crack growth in statically loaded specimens and seemed to be insensitive to the presence of contaminants, the results of the screening tests could be interpreted as indicating that the halogens are more aggressive towards Zircaloy-4 than the other elements that were screened. However, in rising load tests both Cd and Fe-contaminated Cs caused embrittlement under conditions where 12 was not aggressive.

Discussion and Conclusions, (Section 7)
A model of iodine SCC of thin-walled Zircaloy tubes was developed based on the experimental results. The model consists of four steps, namely oxide penetration, crack formation, crack propagation, and cladding rupture.

1-10
Iodi ne is capable of penetrating very thin oxide films in the absence of an applied stress but a strain threshold of a fAW tenths of one percent must be exceeded to mechanically fracture the oxide before iodine can penetrate films with thicknesses in t he micrometer range. The number of oxide cracks increases systematically at higher strains and depends somewhat on oxide thickness . Because the threshold stress for iodine SCC of annealed Zirca10y in tube pressurization experiments is wel l above the stress required t o f r a cture t he oxide , oxide penetration is not believed to be the key step i n i odine SCC.
Crack formation in the metal surface occurs beneath cracks in the oxide if the applied stress exceeds a threshold value . Crack formation, which we regard as t he cri tical step in the overall iodine SCC process, occurs at spots in the metal surface where t here are locally high concentrations of alloying elements and impurit i e s . The exact mechanism of crack formation is uncertain, but rapid chemical reaction to form br i ttle mixed met al iodides may be involved. The t hreshold stre s s for crack formati on depend s on microst ructure and probably a lso on irradiation, but is insensitive t o iodi ne concentration in the range 6 to 0.06 mg/ cm 2 Zircaloy surf ace .
As the crack nuclei grow, the stress requir ed f or their continued propagation falls .
Initiall y , the applied stress must be large enough to maintain the net section s tress above the crack formation threshold stress, but when the cracks are deeper than 10 0 to 200 ~, conditions at the crack tip become dominant and continued propagat ion at lower net section stresses becomes possible. In stress-relieved materials, prop gation occurs by transgranu1ar cleavage and ductile tearing, whereas recrystallized materials sometimes exhibit intergranu1ar propagation. The cracks propagate on a surface perpendicular to the hoop stress and become more branched at larger stresses. The crack growth rate depends on stress and microstructure and probably also on irradiation, temperature and iodine concentration. The role of iodine in crack propagation is uncertain. It seems most probable that the pr esence of i odine at t he crack tip leads to a reduction i n t he strength of metal-metal bonds .
Crack propagation continues until the stress state i n the uncracked l i gament exceeds an ins tability cri terion . Shear f ailure then occurs on a s urface inclined at about 1-11 45° to the h.oop stress. At applied stresses near the burst s t ress, instability occurs when the net section stress exceeds the burst stress. At smaller applied stresses, the net section stress at instability approaches three times the yield stress.
The most important observations with res pect to PCI f~ilure are as follows: • Significant quantities of s everal potentially aggressive substances can reach the surface of f uel cladding under normal design conditions.
• Iodine stress corrosion cracks form at specific sites in the metal surface.
• A stress threshold exists for iodine sec below which Zircaloy tubing is immune to failure. The general nature of the deposits on the inside of the cladding can be seen in the opt ical photograph, Figure 2-1, which shows the inside surface of the cladding after the rod had been slit open and t he fuel removed. The deposits were of three types: (a) sharply defined linear stripes, which occurred opposite cracks in the fuel; (b) concentrations of more m' :sive deposits at the locations of pell et -pellet i nterfaces; a nd (c) Ci rcular, mound-like deposits at random locations. In addition, ther e were occasional chips of f ue l that strongly adhered to the cladding.
Small pieces of cladding were examined in a scanning electron microscope (SEM).
The SEM was a Model Ul made by ETEC (Hayward, California) and modified to control alpha as well as beta-gamma radioactivity. The SEM was equipped with an energy di spersive x-ray analyzer (Kevex Corp., Burlingame, California) with a gold collimator, as described by Wolff and Wolf (2-4), which served to reduce the background r adiation reaching the detector, thus allowing analyses of samples with greater level~ of act1vity. Samples of cladding au l arge as 0.5 cm x 0 . 5 cm could be examined, although occasional ly they we r e too radioactive for the Kevex de tector and had t o be cut into smaller piee .s for elemental analyses .
The Kevex energy dispersive analyzer could analyze the x-ray fluorescence arising from elements heavier than Na. Oxygen could not be analyzed, but it was assumed t be present in all of the surface l~yers because of their ceramic or crystal ine appearance and because of the strong avidity for oxygen of the elements present (Zr , U, Cs). The sensitivity of the instrument was such that 1 to 5% of an element could be detected. The 20 keY electron beam used probabl y penet rated of t he order of 1 ~ into the surface being examined; therefore, the analyses of thin surface Width of photo correspond~d to 1.5 cm of cladd ing length . The vertical band of deposits occurred at a pellet in terface . The thin line deposits were opposi te fuel cracks .

2-3
features included some of the substrate material. Thp. shades of gray observed in SEN photos are established by the relative electron emissions from the surfaces and are generally not the same as would be recorded by an optical photograph of the same surface. The rods were sectioned and samples were prepared in hot cells.
Caution was exercised to avoid contamination of sample surfaces, but occasional features were observed that were obviously introduced in the hot cell.
The SEM examination was concentrated on the distinctive features seen in Figure   nodules (~l to 2 ~ diameter and -1 ~ apart) closely attached to the substrate, which was presumably a surface coating of Zr0 2 on the Zircaloy. The nodules contained mainly Zr (pa rtly observed due to penetration of the electron beam through the nodules), and some U, Cs, and Te. The large dark particle near the center of the photo showed only U on analysis and was apparently a chip of U0 2 fuel stuck to the surface. The almost white, angular particles contained U and Cs and wero probably a cesium uranate (possibly CS 2 U0 4 ). The lower one-third of the photo was part of the region separating two bands of nodules. Particles observed in those bands included pieces of U0 2 , particles of cesium uranate that were sometimes crystalline and sometimes amorphous, occasional swall cubic crystals that were found to be CsI (by comparison of the x-ray :Uuorescence spectrum with that from a standard CsI crystal), and some amorphous-looking deposits (typical of that near the bottom, center of Figure 2-3) that contained Cs and Te and were presumed to be CS 2 Te. Occasionally small amounts of Fe, Mn, Cu, and Cl were observed in some of the amorphous deposits.

2-6 _
Th e deposits found opposite pellet-pellet interfaces (type b) generally contained larger amounts of material and were much more radioactive than the linear deposits found opposite fuel cracks . As r~+ed above, intense peaks of CS 137 activity were observed a t pellet-pellet interfaces in gamma scans of the cladding. A massive deposit of the type found opposite 'pellet-pellet interfaces is shown in Figure 2 Those mounds were found to consist of essentially the same material a s tho massive deposi ts, showing the same brittle fract ure and duplex s tructure, with greater cesium concentrations next to ~he cladding.
A piece of fuel adhering to the cln1ding is shown in ~' igure 2-4. The fuel was distinguishable by its grain size (~2 ~) and structure; analysis showed only U. Fuel c~ip s were apparently cemented to the cladding by some of the Cs-U ceramic material, which was observed at the base of the chip. Near the base of the ceramic material, as well as near t he base of the circular mounds, small bands of nodules, similar to those in the stripes, were often observed.
In some regions the small dot-like features shown in  Width of photo = 270 pm.

2-11
U, Cs, and Te. The small dots in Figure 2-8 seem to be particles of U0 2 that were reacting wit h the zr0 2 surface layer.

H.B. Robinson Rod
Cladding from the H.B. Robinson reactor was studied as part of the NRC-LWR safety program . Two pieces were obtained from a rod designated F-7, which had been irradiated to an average burnup of 28 GW d/ MTU at an average power of 230 W/cm (7 kW/ft) (2-5). Axial gamma scans showed no peaks of Cs activity and fission gas release in those rods was small « 1%).
SEM examination Showed t hat the inner surfaces of the cladding were relatively clean and free of deposits. Figure 2-9 shows the general appearance. The majority of the surface was covered with a dark gray zirconium oxide. There were occasional lighter gray patches of various sizes and shapes, s ome of which were several millimeters in longest dimension, where apparently fuel pellets or pieces of fuel had pressed against the cladding. Also, the surface was sprinkled with small crystals. None of the prominent features found on the Maine Yankee cladding (stripes opposite fuel cracks, pellet-pellet interface deposits, or Circular mounds) were observed on these samples of H.B. Robinson cladding. Also, no fission product s were detected in any of the features or particles on the surface except for an occasional trace of cesium.
The only elements generally observed were uranium, in the light grey patches and small particles, and of course Zirconium. (In one small spot a stain containing iron was noted, which may have been fo rmed after the rod was opened. In another spot a particle presumed to be alumina was found embedded in the surface .)

2-13
The type of light gray surface seen just to the lower left center of Figure 2-10 is shown at higher magnification in Figure 2-11. The small, light crystals contained mainly U and were apparently grains of U0 2 fuel. The worm-like material was Zr0 2 containing variable proportions of U0 2 • The detailed structure of those worm-like features indicates they were caused by tha reaction of U0 2 with Zr0 2 • The U0 2 apparently became incorporated into the Zr0 2 which resulted in a faster local oxidation.
The horizontal white :ine in Figure 2-9 is shown at higher magnification in Presumably that structure was originally a scratch on the cladding surface made during fuel rod loading. The scratched area apparently oxidized somewhat more rapidly than the unscratched areas due to the interaction of U0 2 particles with the fresher Zircaloy surface.
The majority of the surface was covered with somewhat undulating, black zir conium oxide, sprinkled with occasional white particles of U0 2 • Figure 2-13 shows such a surface, which is interesting because of the cracks in the surface zirconia layer that were associated witp intimately attached U0 2 particles. The white particles were rich in uranium and were probably grains of U0 2 • TIlose fused to the sur fece were apparently in the process of diffusing into the zirconia, and we presume that the diffusion process gradually resulted in the formation of the worm-like material .
Thus, the H.B. Robins~ cladding had no significant depOSit s of fission products.
The U0 2 apparently reacted with the cladding in spots where there was good conta ct .
Wh ere pellets or pieces of fuel were pres sed against the cladding, some bonding occurred . The bonds generally broke so as to l eave some U0 2 on the cladding when the rod was opened; however, occasionally the zirconia surface oxide wa s pu led away from the cladding metal. Wllen small particles of U0 2 were in intimate contact with the cladding, they apparently could react with its surface oxide , which resulted in an a ccelerated oxidation.

MECHANISM OF FORMATION OF DEPOSITS
The de posi~s on Main Yankee cladding appeared to have been formed mainly by vapor transport. Tha is, he substances deposited were first released from the fuel and transport ed through the gas to the cladding. The release of volatile fission products occurs during fuel restructuring. There is some evidence that the fractional release of fission products is the same as that of the fission gases (Xe and Kr) (2-6). Accordingly, we shall assume that the release of volatile fission products (and manufacturing impurities) from the U0 2 fuel occurs by mechanisms similar to those of the fission gases so that the fraction of volatile fission products relea ed is the same as that of fissi on ga ses . That assumption is supported by th Maine Yankee rods examined (10 to 13% fission gas release and many deposits) compared with the Robinson rod « 1 % gas and very few deposits). Another important factor is the volatility of the fission product, which, in turn , is a function of its chemical state in the system. Thus in our model, fission product release is related to the fraction of fission gas released and the chemical states of the fission products.
The fission gas release fraction for the Maine Yankee rods studied was about 10 to 13%, a relatively large value tt~t was related to the relatively high temperatures r eached in the fuel C~ 1900 K at the center (2-3>]. Thus, ~i sslon product compounds with significant vapor pressure at ~ 1900 K would be expect~d to be releaSed to t he same fract10n as the permanent gas fission products. Presumably the fraction r eleased from the hotter fuel center was larger than that from the cooler periphery. The volatile compounds would be expected to travel radially from the hotter regions , t hrough open porosity in the fuel, along fuel cracks or pellet interfaces and f inally deposit on the cladding, ~l&e coolest surface available . There was no evidence (2-3) for axial migration, indicating that once the depOSits formed, they were not significantly volatile at cladding temperatures . In particular, CS 137 a ctivity was concentrated at pellet interface positions and no significant amount was observed at the lower end of undefected fuel rods in gamma scans made before t he rods W0re opened. This suggests that there were no significant amounts of liquid Cs in the Maine Yankee rods. The released Cs was apparently all combined in solid forms.

2-17
Chemical States The behavior, especially the volatilities, of the fission products will depend on their chemical states in the fuel, which will be governed in turn by the oxygen potential in the system. Fuel pellets initially have a stoichiometry of about UO a • 002 • As the uranium is fissioned, the oxygen becomes partitioned among the fission products, those with greater avidity for oxygen forming oxides and others going into elemental s t ates, on the basis of equilibrium thermodynamics. The free energies of formation of the oxides are shown in Figure 2-14. Th ordinate of that figur e gives the oxygen potential at which an element will be oxidized. From the inventory of fission products, one can estimate the oxygen potential expected in the system. The method has been applied to fast reactor fuels and is described in Reference 2-10. LWR fuels differ from fast reactor fuels in that the spectrum of fission products for U0 2 includes more oxygen-avid elements and also because the Zircaloy cladding is a strong oxygen getter. The rate of uptake of oxygen by the cladding iR determined by the kinetics of formation of oxide films on zirconium, which are parabolic for moderate burnup (2-11) . We estimate that at a burnup of The chemical states* of the fission products and the amounts of oxygen combined by them are shown in Table 2-1. Certain fission products are very oxygen-avid and will certainly be oxidized (Item 2 of Ta.ble 2-1). They (plus the oxygen taken by the cladding) consume 0.027 o/u uni ts of oxygen at 1.5% burnup. The total amount of oxygen made available because of U fission is 0.030 o/u ~mits at *The chemical states of the fission products in U0 2 fuel are not well known. A first assessment as to whether an element would be expected to be oxidized or not will be made on the basis of the data in Figure 2-14 (which is an iterative process of estimating the oxygen potential and amounts of fission products oxidized). However, the data of Figure 2-14 refer to the elements and oxides in pure states. When the fission products are incorporated in U0 2 they may have different oxidation states that are formed at somewhat different oxygen potentials (see Reference 2-10). The assumptions u~ed in Table 2  Thl! high-terr.,>erature branches of the cu rves for Te02 ' CS:z0' 8aO, CdO, and SrO represent format ion of the sol id oxides from the gaseous elements at one atmosphere. This column gives the number of oxygen atoms reacted per fission product metal atom to give the oxide in the first COllWO. Note tbat in CS 2 U0 4 each U initially bas two oxygen atoms and only one (4) ° per Cs is needed.
RE represents the rare earth elements.
2-20 thatburnup. The difference (0.003) is available to be applied to the formation of the next higher oxides (Figure 2-14), which would appear to be Cs 2 0 at lower temper a tures and Mo0 2 at higher t emperatures. However, the situation is somewhat more complex. Because of the stability of the halides of Cs and Rb, (which wi ll be treated together with Cs), it is assumed that Br and I will react with Cs in preference to oxygen . Also it is assumed that Te and Se will react with Cs in preference to 02 [ as is the case for K compounds (2-8) J . Another complication arises in that Cs can form the relatively stable CS 2 U0 4 • Figure 2-15 shows the oxygen potentials for formation of CS 2 U0 4 vapor at 10-2 atmospheres.* Also the stoichiomet r y of the U0 2 is a function of the oxygen potential of the system.
Blackburn  Since elemental M o will remain, some of the oxygen of the U0 2 • 002 will be used to oxidize Mo and the stoichiometry of the fuel will decrease accord ingly . This leads to the conclusion that the oxygen potential of the system is buffered by the Mo/Mo0 2 couple. *It is assumed thllt the partial pres s ure of Cs in the fuel rod at these burnups is about 10-2 atm. That assumpti on is based on the ~rgument that no liquid Ca was pres ent; hence, the pressure must have been les s than the vapor pr essure of Cs (~JO-1 .7 atm) at cladding temperatures (~ 650 K) and Cs seemed to be vapor t ransported throughout the rQds.

2-22 Volatilities
The fission product compounds, as indicated in As indicated earlier, the release of even the very volatile noble gases (Xe, Kr) does not occur during normal rod operation even to relatively large burnup. That i s, the kinetics of release of fission products must be a controlling factor. We will assume that the r elease of volatile fission products is comparable to that of the fission gases and the fraction released will be the same for all the compounds in Classes I and II. The pressure of atomic Cs expected is less than 1% of the Cs 2 0.  (1)
(3)Esttmated to be comparable to the halides. Bee text. 2-24 We believe that the formation of deposi ts such as CsI and Cs 2 Te opposite fuel cracks resulted from several events--release of fission products from the fuel center when its temperature exceeded a threshold value, reaction of released atoms in the gas phase to form stable molecules, transport in the gas along fuel cracks, and deposition on the cladding (the coolest local surface available). In addition to those compounds, one would expect to find Cd, Sb, Sn, Ag, and possibly Pd, some 01 which were observed. The low fission yields of those elements makes their detection difficult. The presence of the trace impurities observed is in accord with the model because the impurities observed have sufficient volatilities to be vapor transported as elements (Mn, Cu) or compounds (Fe-Cl, or possibly Fe-Te).
The small crystals containing Cs and U (presumably as Cs-U oxide) probably were formed from particles of U0 2 that had broken away from fuel pellets, adhered to the cladding, and reacted with a cesium-bearing vapor species to form the Cs-U oxide.
The small nodules arranged in bands opposite fuel cracks and occasionally at the base of other deposits on the cladding appeared to be sites of greater zirconium oxidation, which was likely enhanced by the p esence of the Cs-U-Te material that was always associated with nodules. We speculate that some volatile compound of Cs, U, Te, (and possibly 0) exists that carries those elements to the cladding.
That compound then reacts with the surface zirconium oxide on the cladding and locally accelerates the cladding oxidation to form the observed nodules.
A model for the formation of the more massive Cs-U ceramic depoSits must conSider the transport of Cs, 0, and U. We propose the following model, which is illustrated in Figure 2-16. When the center of the fuel became hot enough to release fission products, gaseous Cs 2 0 was evolved. It was transported to the gaps between pellets, and then to the cladding at pellet interfaces, where it deposited to form type b deposits. The depOSit, according to this model, was ori ginally rich in Cs 2 0, but at some time it came in contact with the fuel. The Cs 2 0 melted and dissolved some U0 2 to form a high~r melting soJ id that solidified into the observed depOSit. Core I fell substantially below their threshold; yet we obse rved r eaction layers similar to theirs, indicating there is some important factor not included in their correlation . We believe that fuel temperature is more fundamental than linear power i n view of the Maine Yankee Core I experience (2-3). That is, the release of f i s sion product s from fuel probably occurs by mechani sms that are thermally activated and therefore the temperature attained by the fuel i s a primary variable.

2-25
The temperature of the fuel depends on the linea r power but also on the geometry of the fuel rod, espe(~ally the fuel-cladding ga p a nd associat ed gap thermal conduct i vity . Therefore, a correlation of fuel tefllperature and burnup with the formation of reaction layers may give a better defined threshold. The H.B. Robinson rod ope rating conditio~8 fell on the threshold line drawn by Bazin. No reaction layers were observed in t hat rod. Anoth er relevant paper is the Gl report of the examination of some BWR rods (2-19).
Those investigators observed some fuel-clad bonding and a reaction layer in rods that had operated at maximum power levels of 420 W / cm (12.6 kW/ft) to burnups of 34    In addition, other substances found or potentially present in the depos its may be embrittling to Zircaloy. For example, the semimetals (Te , Se, Sb, As) and B group metals (Ag, Cd, In, Sn, Pb) a re embrittling to ste 1s, whereas Cd and Ag embrittle Ti alloys . They may also be embrittling to Zircaloy.
As part of the assessment of the i :nportance of these substances as PCl-active a gents, we should consider whethel' the quantities that reach the cladding are suffic ient to embrittle it. Estimates for some elements are presented below. It was found that the major concentrations of fission product deposits occurred at locations such as pellet-pellet interfaces and oppOSite fuel cracks, where analysis indicates that stresses and strains also are concentrated. The refore, the estimates below ale of the amounts of some fission products and impurities that might be encountered at such locations in the cladding surface.
I odine I n the fuel rod enVironment, iodine (and bromine, which will act like iodine) is mainly combined as CsI because of the large availability of Cs. CsI alone does not cause SCC of Zircaloy under fuel rod operating conditions. However, in the inten~e r adiation flux in a reactor, CsI undergoes radiolytic decomposition and releases iodine (2-26), and that combination can cause SCC of Zircaloy (2-27).

2-30
An estimate of the amount of iodine that might be available at the inner clndding surface can be made as folloWD , The fission yield of iodine is about 1 atomic percent.
Therefore, at 1 percent burnup (9500 MWd/MTU), there will be about 0.8 mg I generated in a typical U0 2 fuel pellet (assumed to be 1.2 cm diameter by 1.2 cm long).
We further assume that about 10% of that iodine is released from the fuel because of a temperature rise in the fuel. (Fission gas release of about 10 to 13% was observed in Main Yankee rods and it is reasonable to expect that a similar fraction of volatile substances in the fuel--CsI in this case--woold be released . See   Ingots were rotary forged, extruded, and cold worked to fabricate the tubes. Although detailed tube fabrication schedules were not provided by Sandvik, the procedures used were reportedly typical of those used for the manufacture of commercial nuclear fuel cladding and included a final cold reduction of about 60 percent prior to heat treatment and straightening (3-1). Accordingly, the Schedules C, J, and K used for the O.B-mm sheets also included a final cold reduction of 60 percent . Unfortunately, an error was made when ~VCA &caled the s chedules for the 3.2-mm sheets and the final cold reduction in th i~ case was 40 percent rather than the desired 60 percent. It was not considered feasible to include a 60 percent cold reduction in the rolling schedules used for the l2.7-mm plates; hence a warm reduction of 60 percent was used as shown in Table A-I.

PRIMARY CHARACTERIZATION
To allow cross correlation of the results of experiments on sheet, plate, and tubing, we evaluated certain key characteristics of the experimental materials.
Ingot compositions were obtained from TWCA and most of the products were subjected to a primary characterization screening cons isting of • Optical metal lography (3 orthogonal sections) • Vickers diamond pyramid hardnesses (3 orthogonal loading directions) • Basal texture determination.  hardness and basal texture. Table 3-1 identifies the produc t s that were screened and also indicates those whose compositions were analyzed after fabricat ion. Detailed information on the results of the primary characterizat i on studies can be found in Appendix A. A summary of the observations is presented below.

Ingot and Product Composit i ons
The chemical compositions of the :I ;our ingots used to fa bricate the tube and f l at products listed in Table 3-1 are presented in Tables A-4 to A-7. The contents of the alloying elements and major impurities were al l in accordance with normal nuclear cladding specifications, (ASTM B353-77) and no Significant differences were found between the chemical compositions of the ingots used for th~ sheet and plate products and those used for the tubing.
At SRI' s request, TWCA analyzed the three tubing products and an arbitrary but representative group of the sheet and plate products. These analyses were restricted to the interstitial impurities oxygen, nitrogen, and hydrogen, which are the only elements whose contents are likely to change significantly during fabrication (3-2).
The data obt ained are presented in Table A-8. Although the interstitial impurity level s generally i ncreased slightly during fabrication, the final contents were within normal nuclear specifications in all cases . It was concluded that no significant diffe ences existed between t he chemical compositions of the tube products and those of the corresponding sheet and plate products.

Basal Textures
Basal pole figures for the 28 products that were subjected to the primary characterizat ion screening are presentej in Appendix A as Figures A-l to A-28 . Each of these basal textures was obtained at the midplane (sheet and plate products ) or midwall (tube products) using a technique similar to that described by Holland .
Values of t he texture parameter f, defined as Schedule J, which was intended to produce sheet and plate materials with basal textures similar to those of tubi ng , was generally successful in a qualitative sense; however, with a few exceptions, the Schedule J sheet and plate textures are somewhat less intense than those of the tubes. Schedule K, which was int ended to produce a symmetrical basal pole distribution centered in the plate normal di rection, also was generally successful, although some of the Schedule K materials have textures that show two near-normal peaks rather than one . Also, the l2.7-mm Schedule K plates s how much les s intens e textures than the thinner materials , presumably because of the difference in working temperat ure.
Schedule C was not successful in producing the desired high basal pole intensity in the plate transverse direction. The textures of the Schedule C materials were similar to those of the Schedule J materials for sheets and plates of all thicknesses. No explanat ion emerged for the present lack of success with Schedule J which previ ously had been used successfully by Lee (3)(4).  Although no other products were investigated in detail, the computer output indicated that the nature of the texture variations in the Schedule J and Schedule C sheets and plates and in the tubes was qualitatively similar to that illustrated in Figure 3-2. Therefore, we can conclude that the basal pole orientations of all those materials was more strongly radial (normal) near the surface t han indicated by themidwa11 (midplane ) pole figures presented in Appendix A.

Biaxi al Tensile Tests on Tubing
The biaxial stress-strain behavior of the three lots of tubing was evaluated with a simple but tedious test method. Fir~t, the diameter of a length of tubi ng was accurately measured. Next, the tubin~ length was heated to the desired test temperature* and pressurized with helium to a predetermined pressure at a rate of about 100 kPa/sec. After reaching the predetermined level, the pressure was i mmediately reduced to a small value, the temperature of the specimen wa s returned to room temperature, and the new diameter was measured. The entire procedure was repea ted for increasing maximum pressure levels until the specimen burst. The averaged data obtained from duplicate tests are shown in Figure 3-3 and Table 3-2.
As expec t ed, the fully annealed Zirca10y-2 tubing (Sandvik lot number 7AH11-S) is considerably weake r and more ductile at 590 K than the stress-re1i ved Zirca10y-2 tubing (7AHIl-H). More surprising is the observation that the struss relieved Zircaloy-4 tubing (7FD11) i s detectab1y weaker and less ductile at 630 K than the stress-relieved Zircaloy-2 tubing (7AH11-H) at 590 K. The difference in test temperature probably accounts for the small strength different i al. However, the lower ductility of the 7FD11 tubing most likely is due to its smaller wall thickness *The Zirca1oy-2 tubings were tested at 590 ± 5 K (a typical BWR cladding service temperature) and the Zircaloy-4tubingwas tested at 630 ± 5 K (a typical PWR cladding service temperature). As will become apparent later, these test temperatures were used conSistently throughout the program.  Curves are smoothed averages of dupl icate tests. See Table 3-2 for var iab ility. (which would tend to red uce non~~1form strain), because t he higher test temperature used for the Zi.rcaloy-4 tubing 10uld normally be expected to enhance ductill ty.

Res idual Stresses in Tubing at Room Temperature
Estimates of the residual stresses in the tubes were obtained by cutting several ,... l-cm··long rings rp .. ldomly from each tubing lot , mf,asuring the out sid e diameter Do, s litting t he ring s axially using a diamond saw , and measuring the new diameter (where E and ~ are Young's Modulus and Poisson's ratio, respectively, and t is the wall thickness of the tube) was 'Jsed to estimate the magnitude of C1 , the residual r hoop stress. The mean data obtained in 5 t ests on each tubing are shown in Table 3-3. (1) Mean of five detel'tlinations and range of oboerved variation.

3-12
All the t :ll'es contain considerable residual hoop stresses that probably orig 1 ate f r om tube st~aightening, which is the final fabrication step and is performed after ann ealing. The behavior of rings of the 7FDll tubing was noticeably more consistent tha n that of rings of either 7AH11-S or 7AHII-H, indicating that the residual hoop stress in the 7FDll tubing was less variable along the length of the tube. The ring diameter increased w·h~n the ring was slit axially for all three materials, ind icating that the residual stresses in the inside part of the tube wall were predominantly compres s ive, whereas the residual stresses nearer to t~e outside s urface were predominantly tensile. Howover, these experiments do not allow us to conclude that the residual hoop stres at the. in.side surface is compressive.   r emove air .lnd water vapor before starting the gas !low. Adjusting the position of the sample relative to the indenter a llowed several dimples to be produced (unde r the same or different conditions a s desired) in a single tube length .
The a pparatus used f or sheet samples is s hown in Figure 4-2. The loading system described above was uRed to press the indenter into the center of a lO-mm~di ameter disc sample. When the sample was under load, it pres sed against a gold O-ring gasket, Which in turn pressed against a s upport ring in the base of the molbydenum enVironmental chamber . Argon gas carry t ng 12 flowed through the environmental . . chamber, which was heated to the desired test temperature . Thus , t he dimple tha t developed on t he side of t he disc sample remot e from t he indenter was e xposed to the c ombined action of an Ar-1 2 gas mixture and a biaxial tensile stress .
Af t er the test, the samples were sectioned to recover the dimpled areas, which were then cleaned of surface oxide and corrosion products by argon ion milling. Milling wa s accomplished using an RF Bias DC Sputter Etch System (Material s Research Corp., Orangeburg, New York) at 820 volts, 300 watt s in an Ar press ure of 0.3 Pa. Under those conditions , material was removed from the sample s~rf ace at approximately 1 #J1fl per hour. In addition to cleaning off corros ion product s, Ar-ion millir.g served to r emove oxide films that had been preformed on some of the s amples. The outlines of cracks that had been formed in the surface oxide due to the local s train generated by the indenter were etched into the sample surface during the ion milling step. Therefore, the relative locations of oxide and metal cracks could be observed.   An initial serie s of tests was made with st ress-relieved Zircaloy-4 s heet samples.
In these tests the temperature was 570 K and the indenter load was 50 kg, which was sufficient to produ ce a dimple about 2 mm in diameter and 50 to 100 ~ high.
The tensi l e plastic strain near the dimple center was found to be 3 to 5 percent.
The sam ples were expo:,ed for times ranging from 5 to 200 minutes to eith er Ar or to AI' containing 12 at a partial pressure of 40 Pa, The sample surfaces were oxidized at 670 K for 3 hours to fo rm a thin, black oxide film prior to the tests.
No cracks were found in samples tested wi thout 12 present. In samples exposed to 1 2 , two kind s of cracks were observed in the metal surface . Small cracks were found in all samples exposed to iodine. In occasional samples, relatively large cracks were also observed . These larger cracks were associated with relatively large local concentrations of impurities (AI, Si, Ti) or alloying additives (Fe, Cr). Metal cracks of both t ypes were found to be associated with cracks in the OXid e . were typical of regions COI1~?; . :!.r.~ higher-than-normal concentrations of alloying additives and impurities. Similar features were also observed in samples that had not been exposed to 1 2 , but there were no cracks associated with those features 4-7   since there was no iodine. The distinctive shape resulted from the ion milling process--apparently the regions with unusually high concentrations of alloying elements and impurities were more resistant to ion milling than the Zr matrix.

4-9
Grubb (~-4, 4-5) observed that Cd can cause embrittl ment of Zircaloys; accordingly, we made a few tests in which pieces of Cd were placed on the surface of samples in an argon gas stream . The behavior in these tests was indistinguishable from that observed in tests made with Ar alone. In other te~ts, a stick of Cd was rubbed against the surfaces of samples (either preoxidized or as-received) before placing them in the test apparatus. In those samples many surface cracks were formed. Apparently Cd caused cracks to form after it had been rubbed into the Zircaloy surface, but not when it was simply placed on the surface. We presume that the small amounts of oxygen present during the tests were sufficient to form films that prevented reaction of Cd with Zr except when the Cd was forced into intimate contact with the Zr metal by abrasion. No further tests were made with Cd because its aggressiveness toward Zircaloy was not reproducible in the present type of test.
The pa rameters vari d and the values used in the tests are given in Table 4-1 .
Two types of cracks in the metal surface were again observed: small cracks that occurred fairly routinely and a smaller number of relatively large cracks that were always associated with local concentrations of impurities in the metal surface.   ( 2 )AS received.
Oxidized 100 hours at 770K--oxi de film _ 5 pm thick, determined by weight-gain measurements. (1) (1) (2) (1) The results reported ~bove ere obtained with as-received specimens, presumably with an air-formed oxide film of the order f te s of nanome ters thick. The as-received samples also suffered general co rosion', in addition to crack formation in environments containing odine. Appa rently, iodine s omehow penetrated the thin oxide films during the tests. Since Zircaloy cladding in fuel rods has on its inside surface an oxide film of the orde r of micrometers thick, indenter tests were also run to investig&te how thicker oxide films influence crack initiation.
Samples with three different oxide film thicknesses that were stressed under the same conditions of iodine pressure, load, time, and temperature are shown in      observed after ion milling of surfaces that had not been exposed to iodine.

4-19
~I mechanical properties of the material (in contrast to the case of annea cd Zircaloy, which had been heat treated at 840 K). In some samples the depths of the cracks we re estimated by repeated ion milling and SEM examination. Most of the cracks appeared to be about 10 ~ deep but a few cracks penetrated into the metal at least twice that far.
Usually, no cracks were observed in samples indented in the absence of iodine.
However, samples subjected to high indenter loads (40 and 50 kg) and long times under loads (1000 minutes) in the ~bsence of iodine exhibited small surface defects that tended to be hole~ rather than cracks. We believe that these defects indicate the initiation of ductile creep-rupt~re failure.  and that attack by iodine generated A II~ and Fel, (probably ZrI. also), which were volatile at the test temperature and coated nearby surfaces.

Stress-Relieved Zircaloy-2 Tubing (7AHII-H)
A series of 28 tests was performed with stress-relieved Zircaloy-2 tubing. The values of the test parameters are given in Table 4-1.
In this material, as in the others, two kinds of cracks appeared--small ones that were randomly distributed and occa~ional larger ones associated with concentration: of impurities or alloying additives. No cracks were observed in the absence of iodine.   Effect of Oxide 1hickness. The frequencies of oxide cracks were measured on films o various thiclmess. The thinnest oxide studied was the air-formed film on " asreceived" materials, which was estimated to be less thpn 0.1 IJIll thick. The other oxides were formed as described above. The dependence of oxide cracking frequency on ox de thickness at constant strain and temperature is shown in Figure ~-20.
Our interpretation of the results is indicated by the dashed lines in the figure.
During the thermal oxidation of Zircaloys, the nature of the oxide film changes when it reaches a thickness 01 2 to 3~. The thicker films are less protective .
The change in oxide cracking frequency between thi n films and thick films shown in The chemistry of the metal-environment system evidently plays a dominant role in i od ine-induced crack initiation. Apparently, iodine reacts more readily with regions cont aining a few percent of certain elements than with zirconium itself.
Since iodine was found in material near cracks even after surface layers had been ion-milled away, iodine probably was incorporated into the lattice in these regions .
Perhaps the structure was modified by the presence of the foreign elements, allowing iodine to be accommodated more easily than in zirconium. In the presence of zirconium at these temperatures, iodine will react to form gaseous ZrI 4 , which ,will in t urn react to form solid i odides, such as ZrI o • 9 ' ZrI 1 • 6 and others (4-7

4-32
Iodine partial pressure clearly would affect the rate of such reactions and this may explain why the frequency of cracking in constant time tests depends on iodine pressure.
The present results are consistent with the crack formation mechanism suggested by Cox and W ood . Because zr0 2 is thermodynamically very stable with respect to iodine, penetration of the oxide is a necessary first step in the cracking process. At reac e',·r operating temperatures, penetration of oxide films a f ew micrometers in thickness appears to require that the material be strained a few tenths of one percent, whereas iodine seems to be able to penetrate thinner oxide films in the absence of an applied stress. The mechanism involved in the latter case is not known. Novak and Rolfe (5-1) showed that the growth of a sharp, deep flaw in an elastically stressed specimen exposed to a corrosive medium was determined by the magnitude of the elastic stress field at the flaw tip (as described by the opening mode stress intensity K I ) rather than by the n~inal tensile stress applied to the specimen. From studies on sec in maraging steels, they concluded (5-1) that flaw growth would not occur unless KI exceeded a threshold value termed K ISCC Novak and Rolf's conclusion has since been verified for many cases of sce and in some instances, values of K measured in the laboratory have aided in designing rscc against SCC fajlure. However, application of the K concept t o SCC of fuel cladlscc ding is problematical because the preexisting surface flaws present in nuclear grade Zircaloy tubing are very tiny (see, f0r example, Figure 3 -5). This limits the magnitude of the stl"eSS intensity that can be generated at the flaw tip and increases the likelihood that a parameter other than stress intensity will control the onset of sec. In fact, it has often been observed in other metal /er.vironment 5-1 systems that environ~ent-aRsisted growth of very small flaws can proceed at nominal stress inten~ities well below K and that str~ss is a more useful see Iscc threshold criterion thaI stress intensity (5-2).
The objective of the work described in this section of the report was to determine whether stress, stress intensity, or some other parameter such as strain or strain rate could be used as a threshold criterion for se~ failure of Zircaloy tubing.
As explained earlier, iodine was chosen as the test environment because of its consistent and reproducible aggressiveness t oward Zircaloy . Internal pressurization was selected as the test method because it allows accurately known states of general stress and local stress intensity to be generated in tubing specimens . A particular objective of the work was to define at least a preliminary threshold criterion for iorline-induced failure of irradiated Zircaloy tubing, since relatively little published information is available on this aspect of the see behavior of irradiate~ cladding (5-3). Pressurization experiments on irradiated power reactor cladding were therefore conducted under subcontract in the hot cells at Argonne National Laboratory (ANL).* A parallel series of tests ~as conducted at SRI on unirradiated material to help set the test cond tiona for the hot cell tests and to investigate the effects of mlterial, environment, and test variables.

MATERIALS AND TEST METHOD
The tests at SRI were conducted on stress-relieved Zircaloy-4 tubing (7FDll) at 630 ± 5 K PWR cladding service temperature) and on stress-relieved and annealed Zir caloy-2 tubi gs (7AHII-H a nd 7AHII-S, respectively) at 590 ± 5 K (BWR cladding service t mperature). The metallurg cal haracteristics, dimensions, and compositions of those three tubing lots are discussed in Section 3 and Appendix A. The tests at ~~L were conducted at 630 K on stress-rel ieved Zircaloy-4 (7FDll) and on irradiated Zircaloy-4 cladding from the 1.B . Robinson reactor (HBR Zircaloy-4).
Unf ortunately, no in ormation as available regarding the fabrication history or *This work formed the basis for an expanded project at ANL to characterize the mechanical and see behavior of power reactor cladding, RP 1027.

5-2
metallurgical characteristics of the HBR cladding, but we presume that it was metallurgically similar to the 7FDlI Zircaloy-4 which had a similar inside diameter and wall thickness. The fast neutron fluence to which the HBR Zircaloy-4 specimens had been exposed was estimated to be about 3.6 x 10 2 1 neutrons/cm 2 , based on the position of the test specimens in the fuel rod and the known peak flux.
The unirradiated test specimens were about 13 cm l ong and the irradiated specimens were 10 cm to l~ cm long. All the Zircaloy-2 specimens were tested in the unflawed  Figure 5~2 illustrates the shape and dlmensions typical 01 actual machined preflaws, based on examinations of polyvinyl acetate replicas, supplemented by the results obtained from two specimens that were serially sectioned.
All unirradiated specimens were prepared for testing by success ively soaking them with met hanol, acetone, and ethanol, followed by etching for 10 seconds in a solution consisting of 45 parts of 70 % HN0 3 45 part of 3% H202' and 9 part s of 48% HF.
After etching, the specimens were thoroughly rinsed twice in deionized water and then dried with an air blast . Measurements of inside and outside diameter were made at several locations prior to testing. One of the 7AHII-S Zircaloy-2 speciments was oxidized before testing for 5 hours at 770 K to form an oxide film ~ 1.5 ~ thick. All the other unirradiated specimens were tested as etched. Most of the HBR Zircaloy-4 specimens were prepared for testing by removing the fuel, cleaning the tubing lengtns, and machining the preflaws (when used).

5-6
the reservoir pressure to be determined at any time. The specimen was pressurized from the reservoir and the specimen pressure was monitored as a function of time by the pressure transducer (BLH Type DHF) and the chart recorder (H.P. Model 7100B).
Gage No. 2 provided a means of accurately calibrating the response of the pressure recording system. When the recorded pressure was calibrated against Gage No. 2 at a system pressure of ~ 10 MFa, the recorded pressure and Gage No. 2 reading were essentially identical over the entire working range of the system.
The test specimen was connected to the pressurization system via specially modified stainless stee l fittings (Swagelok) and wa s located in the center zone of a threezone electric furnace (Lindberg Hevi -duty). To minimize the pressurized volume in the hot parts of the system, small-bore tubing was used fo r the inlet and outlet lines and a Vycor volume-displacing mandrel was placed inside the test specimen.
Safety features included a sealed stainless steel chamber (which completely enclosed all of the hot parts of the test system) as well as tho usual overpressure releases, gage snubbers, etc.
To perform a tes t , the test specimen was loaded with 250, 25, 2.5 , or 0 mg of iodine (~6, 0.6, 0.06, 0 mg of iodine per cm 2 of xposed Zircaloy surface), a volumedisplac ing mandrel was inserted, and the specimen was connected to the pressure systum. Before a test was begun , the speci men was successively pressurized (with high purity helium) to ~ 7 MPa and vented a total of at least 10 times at room temperature and twice at ~ 380 K to reduce the partial pressures of air and water vapor inside the specimen. Also, the sealed stainless steel chamber was evacuated and backfi lled with helium several times to provide a protective atmosphere outSide the specimen. After the final evacuation, the chamber was backfilled to a slightly positive helium pressure, and the specimen was brought up to the desired test temperature (590 ± 5 K or 630 ± 5 K). Then the specimen was pressurized with high purity helium using one of the t hree types of pressure-time histories shown in  Creep of the test specimens during long-term tests resulted in detectable pressure decreases. However, because the change of specimen volume was small compared with the total volume of pressurized gas, the pressure decreases were never greater than about 3% of the initial pressure.
Fractographic observati ons were made 0 some of the failed specimens. To obtain samples suitable for scanning electron microscope (SEM) examination, a ring section containing the failure region was cut from the failed specimen. The ring was cut axially so that the failure region was at the apex of one of the two half rings .
Then the failure section was broken open by bending the half ring that contained the failure site. In addition, a few unirradiated specimens were transversely sectioned through the failure region and examined by optical metallography.

RESULTS
7FDll Zircaloy-4 The results obtained in the pressurization tests at 630 K on the unirradiated 7FDll Zircaloy-4 tubing are presented in Table 5  (Z)Specia ens identified with an asterisk we re tes ted by ANL, other specimens tested by SRI.

5-10
where a is t he maximum de pth of the machi ne d def ect (measured opt i cally after completi on of t he tes t) and w is t he wall thickness ( = Ro -Ri)' The estimated uncertainty in the values of 0 is ±6%. net The nominal mode I stress intensity, K , at the pressure P f associated with the Inom pr esence of a machined preflaw of depth a was obtained rom the calibration curve shown in Figure 5 whereas t he experimental f laws are quit e blunt. Therefore the actual stress intensities gener ated in t he s peci mens must be l ess than the est imates gi ven in Table 5-1.
• Similar result s were obtained in tests at SRI and ANL on unflawed specimens (specimens Zr-4-l through Zr-4-4) and for tests on specimens prefl awed by SRI (specimens Zr-4-5 through Z4-4-8).  • 7FDll Zircaloy-4 was notch insensitive at 630 K as indicated by the observation that the values of anet for preflawed specimens at failure generally were larger than the values of a nom for unflawed specimens at failure.
The subsequent group of three Type 3 tests (specimens Zr-4-11 through Zr-4-13 in Table 5-1) was used to define the creep-rupture behavior of preflawed specimens of 7FDll Zircaloy-4 at ~ 630 K in the absence of an aggressive environment. As expected, the data in Table 5 The presence of a preflaw effectively eliminated the ballooning and resulted in failure strains that conSistently fell in the 3 to 5 percent range.
With the exception of specimen Zr-4-10, all of the preflawed specimens failed at the preflaw location. Therefore, the values of a and K for the preflawed net Inom specimens should provide a more meaningful description of the strE ' S state leading to f -ilure than a However, since none of the K values mee t the ASTM-recomnom Inom . * mended acceptance criteria (5-5) , their significance is questionable. As indicated *ASTM Recommend ed Test Method E-399-72 defines the specimen dimensions required to ens ure that at a given nominal stress intenSity, the crack tip conditions in a material of giv~n yield stress close ly approximte a linBar elastiC, plane strain state. The maxim" um stress intens~ties that could be generated in the present preflawed tubing specimens under conditions that would be considered acceptable by ASTM were -3 WPa·m 1 / 2 for 7FDll specimen and -5 MPa'm 1 / 2 for H.B. Robinson specimens, the precise values depending on the actual flaw size.

5-14
in Table 5-1 , the value of a for s pecimen Zr-4-10 (the preflawed specimen that nom failed remote from the pref law) i~ considered to provide a more useful description of the stress state associated with failure than either a 01' K • net Inom Specimens Zr-4-14 through Zr-4-23 all contained preflaws and all were pressurized to failure at ~ 630 K after they were loaded with 250 mg of i odine. A combination of test types w s us ed in tests at both SRI and ANL. As shown in Table 5-1 , the data obtained at SRI and ANL are completely consistent for s pecimens preflawed by SRI. Mor.eover , the behavior of the specimen that was prefl::.wed by ANT. (Zr-4-23) was indistinguishable from that of two other specimens tested by ANL that had been preflawed by SRI(Zr-4-21 and Zr-4-22 ). Thus, the slight differences in the methods used to preflaw and test specimens at SRI and ANL, which were due to the limitations of the equipment available in the ANL hot cells, did not result in any detectable changes in the behavior of 7FDll Zircaloy-4 test specimens.
The smallest values of a , K ,and € that were associated with iodine-induced net Inom f failur~ wi thin the maximum test durat i on of 168 hours were 305 MPa, 6 . 4 MPa·m 1 / 2 , and 0.3 percent, respectively. Again, none of the values of K associateG with I nom specimen failure sati s fy the acceptance criteria recommended by ASTM and so these data should be interpreted with caution.
Two types of sr.e failure~ were observed in the pref lawed specimens Zr-4-14 through Zr-4-23. The majority of specimens showed pinhole failures (termed See-l in Table   5-1), but two showed short axial splits (See-2 in Table 5   All of the preflawed specimens te~~&d by SRI in the presence of iodine t hat are included in Table 5-1 filed at the preflaw. However, another specimen that contai ned a 50-~ deep preflaw failed under one of the pressure fittings after 20 hours at the relatively low nominal hoop stress of 260 MPa (a ~ 280 MPa). SEM fracnet tography ~vealed the large calcium-rich inclusion shown in Figure 5-10, which may have provided a more severe stress concentration than the comparatively blunt machined preflaw. This result was not included in Table 5-1 becaus e the stress state in the tubing beneath the pressure fittings cannot be accurately estimated .
A final group of four unflawed specimens of 7FDll Zircaloy-4 (Zr-4-25 through Zr-4-28) was tested under similar conditions to those used for the preflawed specimens to determine the conditions needed to promote iodine see in the absence of an artlficial preflaw. The smallest hoop stress at which failure was 0 served in tes t times up t o 100 hours was 305 MFa and the smallest value of strain at failure was 1.5 percent. Pinhole failures and short axial splits were both observed in these tests but no 1()tailed fractography was performed.

HBR Zircaloy-4
The results of the tube pressurization tests at ANL ~n the irradiated HBR Zircaloy-4 are summarized in Table 5-2. This cladding present ed difficulties because of its ovality, which presumably resulted from creepdown during service. The HUR tubing ,"

FIGURE 5-10 A LARGE CALCIUM -RICH INCLUSION AND POSSIBLE SCC INITIATION SITE AT THE INNER EDGE OF THE FRACTURE SURFACE OF A 7FD11 SPECIMEN
The lower part of the micrograph is the tube 10.

5-20
sections were found to be out of round by up to 250 ~, which made it difficult to obtain a good pressure seal. In fact, two specimens that were prepared for testing were never pressurized because a leak-tight seal could not be achieved when the fi ttings were tightened to the maximum torque available in the hot cell. Moreover, those s~~cimens that could be satisfactorily sealed showed a marked tendency to develop leaks at or near the seals during testing. These leaks prov&d to be due to sec failures in the highly s~r~ined regions in the tubing beneath the pressure fittings. We believe that the ovality of the tubing caused an unusually severe stress state in these regions, thus resulting in rapid sec. Because of these problems, only one of the first three specimens tested generated useful quantitative information on the conditions leading to iodine-induced sec failure (see Table 5-2).
To alleviate the problems associated with the ovality of the HBR tubing, the ends of subsequent specimens were annealed and rounded using a specially modi f i ed Jacobs lathe chUCk. None of the three specimens prepared in this way developed leaks at or near the fittings during testing.
The smallest value of stress associated with the failure of an HBR Zircaloy-4 speCimen containing iodine was 200 MPa. The same stress level also resulted in the fa ilure of specimen HBR-6, which was tested without removing the fuel or adding iodine . However, since iodine was detected on the fracture surface of speCimen HBR-6 and fractography indicated an sec failure, it is likely that residual iodine and iodides r9maining in the system from previous tests were responsible.
Because &nly one pI'eflawed specimen (HBR-l) failed at the preflaw, only one value of K was obtained that was associated with a failure. This value (7.9 MPa·m 1 / 2 ) Inom was found to be marginally unacceptable in terms of the ASTM-recommended acceptance criteria . The smallest values of failure strain observed in these tests were 0.25 percent (for a preflawed specimen that failed remote from the preflaw) and 0.4 percent (fOl' an unflawed specimen).
SEM fractography of several of the specimens revealed that the details of the fracture surfaces generally were almost totally obscured by thick iodine-containing layers that were presumed to be corrosion products. The failure near the pressure 5-21   (2)Values identified with a dagger t are associated with specimen failure. seals in specimen HBR-3 was found to be better preserved, perhaps because that reg ion was partially protected from the test environment. Figure 5-11 is a typical view of one of the failures in specimen HBR-3 and shows the mixture of the transgranular cleavage and plastiC tearing tha~ is often associated wi th iodin~ see of Zircaloys . Note that the appearance of the fracture surface in Figure 5-11 is cimilar (although better preserved) to that of the unirradiated specimen in    The results of the pressurization tests at 590 K on the unf1awed 7AH11-S Zirca10y-2 tubing specimens are presented in Table 5    Zircaloy-4 and 7AHII-H Zircaloy-2 ,~bing lots, the annealing temperatur~ used for the 7AHII-S Zircaloy-2 was high enough to permit the thermal growth of a substantial oxide film without degrading the mechanical properties of the tubing. Specimen Zr-2S-17 was therefore oxidized f or 5 hours at 770 K in oxygen to form a -1.5-~-thick oxide. Then 250 mg of iodine was placed in the specimen and it was pressure-tested to failure to investigate whether the relatively thick oxide film affected the susceptibility to iodine seC. As indicated in Table 5

Comparison of Possible Threshold Criteria for Iodine SCC of Zircaloy Tubes
The data in Table 5-provide convincing evidence that hoop stress is a more useful parameter than nominal stress intensity for predicting the conditions that will lead to iodine-induced failure of Zircaloy tubing under internal pressurization . It might be argued that stress is a more useful predictive parameter than stress intensity in the present experiments because of the blunt character of the machined preflaws. Ho vever, d1fficultjes have also been encountered in using stress intensity to predict the times to iodine-induced failure of speCimens containing fatiguesharpened preflaws of different sizes (5)(6). Therefore, the deficiences of K Inom as a predictive parameter are likely due to the departures from a~ ideal linear elastic situation that exist under the present test conditions. As mentioned earlier, none of. the present K measurements satisfy the empirically Inom based acceptance criteria recommended by ASTM , which provide a check of whether or not the test under consideration was performed under conditions leading to an acceptable approximation of a linear elastic, plane strain stress state. In all the present tests, the preflaw depth was not sufficiently larger than the plastic zone associated with the preflaw to achieve an acceptable linear elastic, plane strain approximation. Practical experience (5-2~ and theoretical considerations (5-7, 5-8) suggest that the K threshold conc .. pt. l;\r''M~s down in this small-Iscc fl aw regime and is replaced by a thr~~hold stre3S that mu~'~ be ~y.ceeded for stress corrosion cracking to occur. Under these Circumstances, it i s e :~pected that the minimum value of K at which stress corrosion cracking can be induced will Inom decrease with decreaeing flaw depth because of the increasin~ departure from linear *As explained earlier, O'net is considered to be a more meaningful measure of the hoop stress than O'nom for preflawed speCimens that failed at the preflaws.

5-30
e lastic conditions. It is therefore not surprising that the values of K for Iscc i odine-SeC of Zircaloys that have been deduced from experiments on thin-walled tubing s pecimens containing fatigue-sharpened preflaws (5 -6) and nat ural flaws (5 -9) are much smaller than the value of 9 to 10 MPa.m 1 / 2 that has been obtained at reactor opera ting temperatures in val i d, fracture mechanics type tests (5-10, 5-11, and s ee Sect i on 6 ). Table 5·-2 shows that the irradiated s pecimen HBR-l failed at a ~ominal stress intensity of 7.9 MPa·m 1 / 2 • Because of the high yie l d str s s of the HBR cladding (-600 MPa according to reference 5-12) , t h is test was only marginal l y inval id i n t erms of the ASTM acceptance criteria ( 5-5) at a stress intensity of 7 . 9 MPa . m 1 / 2 and was va l id at the lower values of stress intensity (2 . 4 to 3.3 MPa · m 1 / 2 ) t hat did not promote iodine sec failure in a total exposure period of over 50 hours.
Thus, the resul t of the test on specimen HBR-l seems to ~~dicate that K for Iscc iodine sec of irradia ted Zircaloy a t reactor operating temperature s is not dramatically lowe r than the va lues that have b een reported for unirradiated Zircal oy (5-10,

5-11) •
The data in Tab es 5-1 and 5-3 also offer convincing evide·.lce that a threshol d stress rat her than a threshold strain must be exceeded to induce iodine sec of int e r n~lly pressurized Zirc~loy tubing. Compare, for example, the results of the tes t s on specimens Zr-4-16 (an incrementally pressurized preflawed specimen) and Zr-4-20 (a preflawed specimen held at constant pressure). The test results are consistent with the idea of a stress threshol d in that both specimens fa : l ed after about 20 hours at a hoop stress of 315 MPa, even though specimen Zr-4-16 was exposed for 72 hours at lower stresses . The same test results are not consistent wi th a threshol d strain concept. The strains at failure should be nearl y identical if attainm~nt of a critical strain is a s ufficient criterion for the onset of sec, whe r eas the actual failure strains differ by a f~c tor of three . Similarly, the behavi or of specimens Zr-2H-4 through Zr-2H-13 (Table 5-3) suggests a threshold stress at about 320 MPa, whereas t he fai l ure strain values for the same specimens show a minimum at intermediat e failure times that is inconsistent with the behavior that would be expected if the onset of sec occurred at a critical strain.

5-31
Although it seems possible to conclude t hat attainment of a certai~ critical strain i s not a sufficient condition for iodine see of Zircaloy tubing, all four of the materials studied exhibited significant plastiC strains at failure. Therefore, we cannot exclude the possibility that attai nment of a strain threshold is a necessary but not sufficient condition for the occu~rence of iodine see.   Table 5-1, the decreased failure time in iodine environments is accompanied by a decrease of failure strain of up to a factor of 10 for both preflawed and unflawed test specimens.
The creep rupt ure behavior of 7AHI1-H Zircaloy-2 was not investigated but its performance in iodine-containing environments at -590 K is qualitatively similar to that of the 7FDll tubing at -630 K (compare Figures 5-13 and 5-15). However, the 7AHll-H material shows a somewhat larger threshold stress (-330 MPa) and generally longer failure times than the 7FDll material. These effects may be at tributable either to t he difference of the chemical compositions of the two materials (differences in other metallurgical characteristics were found t o be very slight--see Section 3 and Appendix A) or, more .likely, to the difference of test temperature.    to 2.5 mg 12 (-6.0 to 0.06 mg 12 per cm 2 of Zircaloy surface) tends to increase the failure time at a stress close to the threshold, but does not significantly increase the threshold itself. An increase in failure times with decreasing icdlne concentration 1n the same range has also been reported by .

5-32
Although additional data clearly are required, the behavior of the irradiated HBR Zircaloy-4 specimens ( Figure 5-14) suggests a similar trend to that of the unirradiated 7FDll Zircaloy-4 specimens ( Figure 5-13) at the same test temperature. Based on the limited number of results available, the threshold stress for iodine sec of the HBR tubing at 630 K appears to be about 200 MPa, which is about 30 percent less than the threshold stress for the unirradiated 7FDll tubing. The threshold stress also is a much smaller fraction of the yield stress for the irradiated material (0.33 for HBR tubing compared with 0.75 for 7FDll tubing). Moreover, the failure strains at stres ses just above threshold were lower in the HBR specimens and the time to failure for specimen HBR-l (0.53 hours) was much less than that for specimen Zr-4-l8 (2.9 hours) at the same net stress. Tests on 7FDll in the hot cell gave results that were indistinguishable from those obtained in out of cell tests (see Table 5-1). Thus, the differences between the behavior of the 7FDll and HBR specimens cannot be attributed to differences in test procedure and we can conclude that the irrndiat ed HBR tubing is markedly more susceptibl e to iodine sec at 630 K than the unirradiated 7FDll Zircaloy-4 tubing. Although the greater susceptibility of the HBR material is probably a direct effect of neutron irradiation during service, no definite conclusion can be reached in the absence of information on the susceptibility to iodine sec of unirradiated tubing from the same lot as the HBn cladding.
Unfortunately, such uni r radiated lnaterial is not available for testing.
The behavior of the 7AHIl-S Zircaloy-2 ( Figure 5-16) Q~ffe re d from that of the other three materials in that iodine sec was c onsistently observed only within a very narrow range of stresses. No failures occurred below 282 MPa, which appears to be the threshold stress for iodine sec. Stress corrosion cracking ~as observed at stresses between 282 MFa and 307 MPa, but all four specimens te ted at stresses greater than 307 MFa failed by ductile processes even though iodine was present in 5-37 the test environment . We believe that crack blunting due to rapid creep suppressed i odine see in these specimens. Support for this theory is provided by the results of the crack initiation st udies on 7AHll-S Zircaloy-2 reported in Section 4 of this report . Those studies showed that iodine-induced stress corrosion crack nuclei became wider rather than deeper when increasing local stresses were applied for a constant time. Suppression of iodine sec at high strain rates has been reported by others (5-3 , 5-13, 5-14) and is observed in many other metal/enviroru Jnt systems .
The present results suggest that a strain rate of about 3 x 10-5 sec-1 will s uppress iodine see of Zircaloy-2 at 590 K, in reas onable agreement with published data One of the two 7AHI l-S spec imens that showed long-time sec failures just above the threshold stress was t he preoxi dized s pecimen. Therefore, we can concl ude that the presence of a 1 . 5 -~ oxide film has little or no effect on the threshold stre ss f or iodine-induced see of 7AHIl-S Zircaloy-2 tubing at 590 K but may have an effect on the fa i lure t ime.
Since t he onl y difference between the 7AHll-S and 7AHl l -H lots of Zircaloy-2 t ubing was the f inal heat treatment temperature, a comparison of the behavior of those two tubing lots provides an i ndication of the ef fect of microstructure on the 5-38 suscepribility of Zircaloy-2 to iodine sec. The test data in Tables 5-3 and 5-4 and Figures 5-15 and 5-16 indicate that • The threshold stress for iodine sec of the annealed tubing was slightly lower than that for the stress relieved material.
• The shortest failure times for stress-relieved specimens were nearly two orders of magnitude larger than the shortest failure times for annealed specimens at stresses just above the iodine sec stress thresholds.
• Stress-relieved specimens showed smaller failure strains than annealed specimens in tests at stresses just above the iodine sec stress thresholds.
Thus two out of three indic~tors of relative sec susceptibility favor the stressrelieved 7AHll-H material over the annealed 7AHll-S tubing. It can be concluded that Zircaloy-2 tubing becomes more susceptible to iodine sec in internal pressurization tests as the heat trea~ment temperature of the tubing is increased from 770 K to 840 K and the microstructure changes from a stress-relieved to a recrystallized condition.
It might be crgued that failure strain is the most relevant indicator of sec susceptibility as far as the fuel c ladding app11cation of Zircaloy tubing is concerned because cladding is subjected to a fixed displacement type of loading (imposed by the fuel) in reactcr s~rvice. However, the larger failure strain of the 7AHll-S tubi ng is mainly due to the fact that the iodine sec stress thres hold is above the yield stress of this material. It is likely that the failure strain of 7AHll-S tubing would fall as the yield stress was increased by radiation hardening and that the lower threshold stress and failure time of 7AHll-S Zircaloy-2 would then make it more prone than 7AHll-H tubing to iodine-induced failure even under fixeddisplacement loading.

Origin of the Threshold Stress
In many instances of the stress corrosion cracking of reactive metals that form highly protective oxides, it is observed that a certain minimum strain rate must be sustained to continually rupture the oxide formed at the crack tip so that see can continue (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15). In some cases ~5-l5), sec threshold stresses have been shown

5-39
to be the smallest stresses that could generate the required rates of strain.
The data provide mixed support iu~· this idea. For example, the average diametral (hoop) strain rates in specimens Zr-ZH-7 through Zr-2H-12, which were tested at stresses just above threshold, were all about 3 x 10-8 sec-1 based on the failure strains and times to failure; this is consistent with a critical s train rate concept. However, if attainment of this strain rate were a sufficient condition for iodine SCC of Zircaloy-2 at 590 K, the stress threshold for the 7AHll-S Zircaloy-2 would have been much lower than that actually observed because the average stra in rat es in specimens of this material at stresses just above threshol d were up to 100 times faster than 3 x 10-8 sec-I. Moreover, the behavior of specimlms Zr-2S-l5 and Zr-2S-l6 (Table 5-4), for which the average strain rates differ by more than two orders of magnitude, is Il\uch easier to explain in terms of a threshol d stress thau in terms of a threshold strain rate.
The present observations seem less consistent with the idea of a critical strain rate than with the concept that the threshold stress is the minimum stress that is capablr of forming a propagatable crack in Zircaloy exposed to iodine, as suggested by Ranjan and Smith (5)(6)(7)(8). However, if a critical strain rate is required for iodine SCC of Zircaloy tubing, the present data indicate that it is material-and/or temperature-dependent and has the following values: $ 3 x 10-8 sec-1 for 7AHll-H a~ 590 K; $ 10-7 sec-1 for 7AHll-S at 590 K; $ 2 X 10-7 sec-1 for HBR at 630 K; and $ 10-7 for 7FDll at 630 K. Although the embrittlement; oi Zircaloys by hydrogen charging and subs equent !)recipitation of hydrides "has been widely studied, comparatively little attention has ,. ,. .

6-1
been paid to the possibility that gaseous hydrogen environments may be capable of promoting slow crack growth in Zircaloys with low internal hydrogen contents. Accordingly, the study described below was undertaken.

Experimental Procedures and Results
"Off the shelf" annealed Zircaloy-4 plate (12.7-mm thick) obtained from TWCA was Hydrogen-induced slow crack growth was studied in precracked compact-tension specimens prepar ed from the annealed Zircaloy-4 plate. Variables investigated were applied stress intenSity, hydrogen pressure, and temperature (within the range 296 to 343 K). The results obtained in these tests, which were conducted in accordance wi th ASTM procedures,* are summarized in Figures 6-1 through 6-3.
No subcrit1cal crack growth was observed at temperatures of 393 K and above in the experiments on the annealed Zircaloy-4. Tests on stress relieved Zircaloy-4 specimens at 573 and 658 K confirmed that under static loads and at high hydrogeL pressures, subcritical crack growth did not occur at reactor operating temperatures.
However, under cyclic loadi ng (1 Hz, tension-tension, R = 0.1), crack growth was observed at 573 K at hydrogen pressures less than 66 Pa . Interestingly, the crack velOCity at constant nominal peak stress intenSity f ell with increasing hydrogen pressure as shown in Figure 6

Discussion
The form of the crack velocity vers us stress intensity curves obtained at low temperature ( Figure 6-1) is similar to that observed in many other cases of s tres s corrosion cracking. A threshold stress intensity must be exceeded to obtain s low crack growth, and a diffusion-limited plateau (Stage 2) in the rate of crack growth is encountered at higher stress intensities. The threshold stress intensity decreased with increasing hydrogen pressure, whereas the Stage 2 growth rate increased approximately with the square root of the hydrogen pressure (Figura 6-2).
Increas es of test temperature up to 343 K at constant gas pre ssure increased both the threshold stress and the Stage 2 growth rate ( Figure 6-3). Fractographic observations did not completely clarify the nature of the rate controlling pr ocess, but they did demonstrate that hydride formation is involved in hydrogen-induced slow crack growth. Figure 6-6 shows that the process is very crystallographic with many cle~ved facets and unidirectional slabs of zirconium hyd ride on the fracture surface. Moreover, the cross sect i on in Figure 6-7 shows that the entire f racture surface was covered with a 10-~-thick layer, which was identified as ZrH by x-ray diffraction.

6-10
Hyd ride formation also plays a key role in the crack growth behavior at higher temperatures. Figure 6-8 is a cross section through the midsection of a stressrelieved specimen that was statically loaded in hydrogen at ~73 K. A ball of hydrid e, which in this case proved to be Zr H 2 , has f ormed at the crack tip and has completely blunted it . The res ultin~ decrease in stres s intensity is the probable r~ason why no s low crack growth occurred in the statically loaded specimens at the higher test temperatures . pressure is decreased. mechanical fatigue ca n crack t he zirconium hydri de at the crack tip a s f a st as the hydride is formed. Theref ore, T. n increase i n crack velocity occurs with a decrease i n hydrogen partial pressure.

Summary
St a tically loaded precracked pecimens of Zircaloy-4 exhibit subcritical crack gr owth at stress intensities ~ 20 MPa.m 1 / 2 when exposed to hydrogen gas ~t ambient temperature. The threshold stress intensity increases with decreasing hydrogen pressure and increasing temperature. The Stage 2 crack growth rate increases with both hyd rogen p ressure and temperature . The activation energy of the process that cont rols the ra te of Stage 2 cra cking at near-ambient temperatures is about 15.5 kJ/ mole and the fracture surfaces are completely covered with a thin layer of ZrH.
Hydrogen-induced slow crack growth in statically loaded specimens is suppressed at react or operating temperatures , apparently because of crack blunting associated with the formation of ZrH 2 at the crack tip . However, subcritical crack growth is obs erved in cyclically loaded, precracked specimens exposed to low hydrogen pressures . The cr a ck veloc ity i n thi s r e gime incr e ases wit h dec r ea s i ng hydrogen pre ssure, probably becaus e crack blunting by hydride formation oc curs les s readily at low hydrogen pr e s sures . 6-11 S A·419 7 ·88 Other workers (6-10) have claimed that K must be less than Iscc 1 MPa'm 1 / 2 to explain the behavior of thin-wall tubing. It was considered essential to firml y establish the value of K under valid plane strain linear elastic con-Iscc ditions to provide a sound basis for understanding iodine-induced crack growth i n fuel cladding. Therefore the study described below was undertaken. In addition to determining K for iodine SCC, slow crack growth tests were conducted in Iscc bromine and chlorine (present in the fuel r od environment to a lesser extent than iodine), in zirconium iodide, and in the combined environments of iodine with iron, iodine with oxygen, and iodine wit h cadmium. A paper describing the results obtained has been accepted for publ i cat ion (6-11).

Experimenta l Procedures
The 9912-4B lot of stress-relieved Zircaloy-4 plate wa s used for this investigation.
The fabricati on history and metallurgical characteristics of this l2.7-mm-thick plate material are described in Section 3 and Appendix A.

6-13
Wedge-opening loaded, compact tension (WOL-CT) fracture mechanics specimens 1.25 cm thick were cut from tho !Zircaloy-4 plate so that the crack would propagate in the rolling direction on a plane norma . ... to the plate width. The specimens were of 8ufficient thickness to ensure that the ASTM guidelines for plane strain conditions would be met at all values of stress intensity less than about 18 MPa·m 1 / 2 • In accordance with ASTM recommended procedures , specimens were fatigue-precracked in air and a wedge was driv~n into the inIt i al machined notch to produce a fixed cra~!, opening displacement (COD). The wedge was made of a molybdenum alloy known to be chemically inert to halogens. The speCimens were cncapsulatej in i ndividual 3.8-cmdiameter Pyrex capsules . After the appropriate environment had been introduced (iodine, bromine, chlorine, zirconium i odide, or a combined environment of iodine and iron, iodine and oxygen, or iod ine and cadmium), the capsules were sealed and placed in a furnace at 633 ± 5 K for various lengths of time. After removal from the furnace, the specimens were air-cooled and then reloaded .0 determine the actual final load.
The load on the specimen at 633 K was esti mated to be 75% of the load at room temperature. This factor is based on the re uced material modulus at the test temperature and the difference between the thermal expansion coeffiCients of molybdenum and zirconium. Specimens were fatigue cracked post test 1n order to protect the region of stress corrosion crack growth and then were failed by overload. The final value of str~ss intensity during the sec part of the test was calculated from the estimate~ load at temperature, and from the location of the stress corrosion crack tip, Which was det ermined by eAamining the fract ure surface of the failed specimen. An average crack growth r ate for each test was determined by dividing the total length of environment-induced slow crack growth observed on the fracture surface by the total time the speCimen had been expos' ed to the test environment.
Detailed fractography was conducted using a scanning electron microscope.

Results and Discussion
Stress corrosion crack growth in Zircaloy-4 was observed at approximate PWR fuel cladding operating temperature (633 K) in all the environments investigated. Table   6-1 summarizes the tests and lists the chemical substances and concentrations (mg  of atress int~nsity at the termination of each test (estimated to be accurate to about i O. 5 MPa'ml/~) , and the observed average crack growth rate . As can be seen , the behavior of Zircaloy-4 was someNhat similar in all environments. Differences do exis t, however , when the influence of each environment is compared separately and fracture topography is also conSidered.
In an iodine environment at ~, 633 K, the value of K appears to be approximately and cold-worked Zircaloy-2). As discussed in detail in Section 5 , the l~wer values of "K " reported by Videm and Lunde (6-9) and Kreyn s et al (6)(7)(8)(9)(10) probably are lscc attributable to departures from plain strain linear elastic conditions in their thin-walled tubing specimens. Since crack propagation in thin-walled tubes occurs a t nominal s tress intensities that are well be low the true value of K , i t seems lscc cl~r.r that the onset of iodine-induced crack growth in Zircaloy cladding tubes is controlled by a parameter other than stress intensity, and indeed t he evidence presented in Section 5 suggests that stress is the key parame··er .
The fractUl'e surface of a specimen exposed under load to an environment containing iodine is shown in Figure 6-9. The gene ra l topography of this surface is typical of that observed in all halogen-containing environments investigated in this study.
Two rather unusual features are evident in the region of stress corrosion crack growt. First, the fracture surface is extremely rough with deep troughs extending i nto the speCimen, perpendicular to the plane of macr oscopic crack growth. These troughs probably are evidence of preferred cracking on or near the basal planes.
The second unusual feature is the highly irregular stress corrosion crack front .
Cracking appears to ~~v e occurred along preferred channels ahead of the macroscopic crack front leaving microscopic ligaments of unfailed material behind. Again the preferred crack paths appear related to the texture of the material. by Vid wand Lundo (6-9). A discus s ion of the origin of thi s discrepancy is presented in Section 5 . It eems probable that recrystallized material 0 the kind studied by Videm and Lund (6)(7)(8)(9) 19 more susceptible to intergranular cracking than str~ss-rolieved matorial but the underlying reason for this effect of microstruct ure is not understood.
Chemical interaction between iodine and zirconium in the absence of stress was observed on the machin d notch surface of the specimen . The stres s corrosion crack growth behavior of stress-relieved Zircaloy-4 in t he combined environments of iodine plus iron, iodine plus oxygen, and iodine plus cadmium appears to be ::Iimila)' .' to that observed in an iodine environment alone.
Although no stress corrosion crnck growth was observed in either iodine plus iron or iodine plus oxygen at 9 MPa 'm 1 / 2 , crack growth in iodine at this level was extremely slow. Thus , this stress intenSity is considered to be the threshold stress intenSity. A similar value of K probably exists for the three combined Iscc environments investigated. Qualitatively the data suggest that iron retards the crack growth rate in an iodine environment. In contrast, results reported by other workers indicated that the presence of iron wool increased the probability of stress corrosion cracks in an iodine environment . This difference suggests that iron may enhance crack initiation (as proposed in Section 4) but reduces the crack growth rat e once tQe crack has been initiated. Oxygen does not appear to have a significant effect OD iodine stress corrosion cracking . Cadmium alone, in either the liquid or vapor phase has recently been observed to cause embrittlement of Zircaloy (6-12, 6-13). In the present study, no evidence of a major synergistic influence of cadmium is indicated.

6-19
The fracture mode observed in the three combined environments of iodine plus iron, iodine plus oxygen, and iodine plus cadmium was identical to that observed in iodine alone with fail ure occurring by cleavage combined with ductile rupture.
Th~ presence of oxygen does have a definite effect on the chemical reaction of iodine with zirconium. When the environmental capsules were opened in laboratory air, in those originally containing only iodine the zirconium iodide that had formed on the fracture surface began to hydrolyze giving off hydrogen iodid and forming a white zirconi um oxide that masked the fracture surfac~ detail. When oxygen was originally present in the capsule, subsequent exposure of the fract u r~ ~urfaces to air did not reRult in hydrolysis. The surfaces remained free from heavy O A~ e and hydrogen iodide was not evolved. Oxygen originally present in th capsule appeared to hinder the formation of a heavy layer of zirconium iodide by rapidly converting zirconium iodide to a thin layer of zirconium oxide as the crack tip propagated.
In tests in an environment initially containing zirconium iodide (ZrI 4 ) and no fr ee iodine, average crack growth rates at comparable stress intensities are essentially identical to those observed for an environment initially containing iodine alone .
Additionally, the primary fracture mode of zirconium iodide-induced stress corrosion cracking of Zircaloy-4 is transgranular cleavage interconnected by ductile rupture, 'identical to that observed in iodine. Free iodine should be converted to zirconium iodide at the temperature of these tests. In zirconium iodide no evidence was found of gross surface pitting on the machined notch surface as was observed in an iodine environment ( Figure 6-11). The similarity in stress corrosion behavior suggests that cracking in an iodine environment occurs through the formation of zirconium iodide and free iodine is not necessary.
The aggressiveness of bromine and chlorine toward Zircaloy-4 at -633 K is similar to that of iodine. The average crack growth r~tes at comparable final s ress intensities are similar for all three halogens, ~s seen in Table 6-1. Like iodine, the value of K for bromine appears to be about 9 MPa·m 1 / 2 • Wood (6-5) and Iscc Cox and Wood (6-7)have investigated the influences of bromine and chlorine, respectively on the crack growth behavior of Zircaloy-2 at room temperature. In 6-20 bromine plus air at 296 K, the observed value of K was approximately 14 ~· M l / 2. l scc In moist chlorine at 296 K, K w s observed to be a pproximately 28 MPa · m 1 / 2 • I scc Thus both halogens app ar to become more aggressive at higher tempera tures .
Crack growth in bromine and chlorine occurs by transgranular cleavage interconnected by ductile rupture as shown in Figures 6-12 and 6-13, respectively. Thus the SCC f a il ure modes in the three halogen environments are qualitatively identical .
However, long-term exposure of the fracture surface to bromine suggests that bromine r eac ts chemical ly with Zircaloy more rapidly than do iodine and chlorine . The ngu lar features of the bromine-induced fracture become rounded and the cleavage facet s become more difficult to identify. Additionally, bromine produced heavy pi tting on the prefatige crack surface . Such pits did not appear on the stress corrosion cracked region and so must have either occurred before the onset of cracking or are associated with an oxide fi l m. The pits due to bromine are much rounder and smoother than those prod uced by iodine.
In an effort to establish wheth er or not subcritical crack growth can occur in :~ . ,stress-relieved Zircaloy-4 exposed to an inert environment, a WOL-CT specimen was l oaded to an initial stress intensity of approximately 25 ~tPa.ml/2 at room temperature. The specimen was placed in a capsule, which was evacuated, sealed, and raised to a temperature of ~ 633K. The s pecimen was held for four weeks and then removed and examined for the exi stence of subcritical crack growth . No perceptible crack growth was observed at a calcula 'ced final stress intensity of approximately 17 MPa. m 1 / 2 • Therefore, t he crack growth observed in the halogen environments at lower values of stress intensity must be due to the aggressive nature of the environments and is not the result of purely mechanical instability at the crack tip.  Did not appear to "wet" the s urface; NPCG 9 . 5 Did not appear to "wet" the s urface; NPCG 8 Did not appear to "wet" the s urfac ; NPCG 9 Did not appear to "wet" the surface; Cd foil used to obtain Cd vapor NPCG

11
Did not appear to "wet" the surface; lots of Cd vapor deposition on cooling; prepumped to a ha r d vacuum before sealing; NPCG

11
Did not appear to "wet " the s urface; specimen submerged in liquid Cd ; NPCG 5 Did not appear to "wet" the surface; specimen submerged in l iquid Cd; NPCG

12-13
Did not appear to "wet" the surface; special 34755 wedge and machin~~ mating surface to obt ain increa~ing COD upon going to temperature; NPCG (1 ) NPCG signifies no percepL i ble crack growth . tbe outer surfac had two pha ses ; one having a 1:3 and the second a 2:3 ratio of Zr to Cd. A Similar analysis of the interior region indicated a 1:10 r a t io of Zr to Cd . Thi s evidence s uggests that corrosion of Zircaloy is likely in many fiss10n product environment s in the absence of an a pplied s tres s .
Because embri ttlement by cadmium at similar test t emperature!! :las been reported by ot her workers (6-12,6-13), a s pecial effort was made to induce cadmium cracking in WOL-cT s pecimens of stress-relieved Zircaloy-4. However, as s hown in Table 6-2, al l tests in cadmium environment s failed to promote cracking . No perceptible crack growth was observed in vapor-phase cadmium, in liquid-phase cadmium, wi th cadmium rubbed onto the surface of the Z1 r caloy , in a cadmium environment with very low oxygen, and in a special expanding-mated-wedge test.
The major differences between tho test method used by Grubb and 10rgan (6-13) and that used here a re: (1) the exist ence of a significant tensile strain rate in th e rising-load tests used by  and (2) the use of a getter (6)(7)(8)(9)(10)(11)(12)(13) to reduce the oxygen partial pressure to ve ry low values. Both these differences could affect the extent to which the ca dmium can penetrate the surface oxide fil m on the Zircaloy . TenSile deformation will cause continuous cracking of the oxide and a low oxygen partia l pressure will decrease the rate at which the oxide cracks can be r e paired. Thus, the presen t results s uggest that oxide f ilm penetration is an extremely important step in cadmium-induced c racking of Zircaloy.

EMBRITTLEMENT BY CESIUM IN RISING-LOAD TESTS
Cesium is an abundant fission product and embrittlement of Zircaloy by liquid cesium has been reported by other workers (6-8 , 6-14 , 6-15). However, at t empts (including those reported in the previous subsection) to induce embrittlement at reactor operating temperatures (573 to 633 K) were not successful and the comparative work of Wood, et al. (6)(7)(8) sugges ted that cesium is much less aggressive than i odine toward Zircaloy .
In other cases of liquid met al embrittlement (LME) it i s frequently observed t hat the aggressiveness of the molten metal is very sensitive to its composition (6-16).

6-26
Si nce the effect of composition variations bad not be~n investigated for the ce s ium-Zircaloy syste , it seemed poss ibl that under some circumstances cesium might be much more aggress i ve tcward Zircal oy than the previous work had indicated . The lack of any information on the influence of the oxygen content of the cesium was considered to be of par ticular concern because this variable will strongl y af t ct the ease wi th which the cesium can pen tra te the pr otective oxide film . We therefore undertook the limited investigation des cribed be low . A complete description of this work has been pub l is hed elsewhere (6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17).

Expe ri mental Procedures
The materia l studied was the 7FDll lot of st r ess -re lieved Zircaloy-4 tubing . The cha racteris t ics uf thi s tubin!" are discu!lsed in Sect on 3 and Appendix A. Tost specimens were prepared by cutting -l 2 . 7-mrn lengt hs fr()1II the tubing , degreasing wi t h acetone , rins ing wi th e thyl alcohol , a nd drying in wa rm air.
The t ests were conduc ted by diame trally compressing the s pecimens at a constant di splacement rate in one of the controlled environment chambe r s shown in Figure 6-15.
The t est environment was e ither argon, ail' (room temperature only ) or liquid cesiun-.  (1) (2)A fresh batch of 99.95% cesium was used in each of t he six test series.

Discussion
We anticipated that Zircaloy-4 might be more susceptible to embrittlement by cesium containing less dissolved oxygen since oxide film repair w()uld be more difficult in a low-oxygen environment. Although the results 0:' th~ Series 1 tests suggest that cesium does indeed become more aggressive toward Zircaloy-4 following calcium deoxidation, a more systematic study will" be required to unequivocally prove that oxygen content is an important variable. The results of Series 2.1 and 3.1 (Table   6-3) show clearly that Zircaloy-4 behaves in a ductile manner when loaded in diametral compression in deoxidized, uncontaminated cesium. Since we do not know at what stage the cesium used in Series 1 became contaminated, we cannot completely rule out the possibility that the transition from ductile behavior in Series 1.1 to brittle behavior in Series 1.2 was associated with a chvnge in the contaminant content of the cesium rather than with a reduction of dissolved oxygen content.
The results of Series 3 show that contamination of calcium-deoxidized cesium with a small amount of iron is accompanied by a transition from ductile to brittle behavior in tests at ~ 313 K.
In test series 4, 5, and 6, freshly contaminated, deoxiuized cesium consistently caused embrittlement in tests at both 313 K and 573 K. Iron was also present in the cesium environments that caused embrittlement in Series 1 and 2. However, the effects of contaminating the cesium with chromium alone or aluminum alone were not investigated, so it is not known whether these contaminants contributed to the 6-34 A s econd possible explanation was recently advanced by Douglass (6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20), who s uggested that the observed behavior could readily be understood if a chemical reaction between iron and zirconium yielding a s paringly soluble reaction product was involved in the embrittlement process . Addition of iron to a fresh batch of cesiwn would p r omote embrittlement until the solubility limit of the Zr-Fe reaction product was reached. Thereafter, the cesium would no longer serve as a sink for the r eaction product, which therefore would remain undissolved in the reaction zone to stifle further'reaction and inhibit embrittlement. In addition to being consistent with the behavior observed in the Zircaloy test chamber and in the stainless steel test chamber that had been partially lined with Zircaloy, this explanation is attractive in t hat it also provides a possible reason why a reversion from brittle to ductile behavior was not observed in Series 1 or 6. In those series, the only zirconium encounte red by the iron-contaminated cesium was the test s pecimen s . Thus, satu r a ti on of t h e cesium wi t h a Zr-Fe rea cti on pr oduct mi ght require many more tests in t he s tainl ess s t eel chambe r t ha n i n the Zircaloy chamber, where the iron-contaminated cesium was exposed to a much larger surface area of zirconium.
Further experimentation will be required to determine whether or not embrittl ement of Zircaloy-4 by deoxidized, iron-contareinated cesi~ is affec t ed by strain rat e or test temperature. Based on the four tests conducted in Series 1.2, the plastic strain at the onset of cracking does not seem to be very sensitive to test temperature but may increase somewh~t with increasing strain rate. The load and displacement at the onset of cracking appeared to be red1.lced by the presence of notches as would be expected in view of thei~ streRs and strain concentrating effect.
Any future work in this area should also consider the possibility that embrittlement by iron-contaminated cesium could be a purely chemical phenomenon. All specimens that showed brittle behavior in the present tests were e xposed to the iron-contaminated cesium for a prolonged period at -650 K during the calcium deoxidation treatment. Therefore , we cannot dismiss the possibility that the unstressed rings were embrittled during the deoxidation t reatment, that is, prior to mechan i cal loading at the final test temperature.
~::: 'uy Zircaloy-4 exhibits brittle behavior in tests in liquid cesium at 313 K and 573 K if the cesi um has a low oxygen content and is contaminated with i ron. In diametral compression tests on smooth, Zircaloy-4 rings that were exposed to iron-contamina ted cesium at -650 K for 18 hours before t esting, brittle fracture sometimes was initiated at tensile strains lass than 0.2 percent in tests at 573 K. It is no~ clear whether i ron is the primary embrittling agent and cesium is merely a vehicle for the iron, or whethel' cesium is the primary embrittling agent and the role of iron i s related to oxide film penetration.

FltfBRITl'LEMENT BY CADMIUM IN RISING LOAD TESTS
As noted earlier, embrittlement 01 Zircaloy by cadmium has been reported by other workers (6-12, 6-13) but was not observed in our statically loaded WOL-CT screening tests. We speculated that this difference in behavior might be due to the difference 6-36 between our static tests and the continuously loaded tensile tests that had resulted in embrittlement (6)(7)(8)(9)(10)(11)(12)(6)(7)(8)(9)(10)(11)(12)(13) . As a preliminary test of this theory, four diametral compression tests were made on l2 .7 -n~-10ng rings of 7FDll Zircaloy-4 tubing partly immersed in cadmium at ~ 593 K. The experiments were conducted using the Zircaloy-4 chamber shown in Figure 6-l5(b). The first ring was heated at ~ 593 K in contact with the cadmium for 2 hours prior to the test and showed no embrittlement. The second specimen was heated for 26 hours before testing and showed severe embrittlement of a type similar to that illustrated by the Type 2 curve in Figure 6-16. The specimen was found to be noticeably oxid 'zed after the test, which suggested that it had gettered dissolved oxygen from the cadmium during the 26 hour exposure period prior to the test. The final two tests showed no embrittlement but subsequent examination r e vealed that all the cadmium had reacted with the Zircaloy walls of the test chamber to form Zr-Cd intermetallics so that neither specimen had been exposed to t he molten metal.
The result s of the diametral compression tests encouraged us to believe that both a r ising stress and a small dissolved oxygen content in the cadmium were prerequi- beam thickness, as shown in Figure 6-19. An interesting feature is the large amount of plastic deformation that accompanied the crack growth. The third type of curve showed essentially elastic or slightly plastic loading followed by an abrupt drop in l oad to a very low value. The concomitant c r ack extension is shown in Figure 6-20. This crack growth process showed very little plastic deformation adjacent to the crack. However, when ~he surface c~ack ng was discontinuous, small plastic bridge regions were obaerved, as shown in Figure 6-21.
An exact value of the threshold stress intensity K was difficult to obtain. Iscc As seen in Figures 6-19 ' 6-21, crack branching dissipates much of the energy and leads to questionable values of K I • Using the load at failure and the initial flaw length, estimates of the initial stress intensities were made. The spread in i nitial stress intensity values was ~ 16 to 24 MPa.m 1 / 2 , with the average value being ~ 20 MPa · m 1 / 2 • This value was obtained from specimens that were exposed to a low oxygen partial pi"eSSUre and were either totally submerged in liquid cadmium or were deformed into the liquid cadmium upon loading. No fallures were obser"~d for s p~cimens tested in cadmium vapor.

,,'
Th fr ture surface of a specimen fractured in a liquid cadmium environment at ~ 633 K is very complex. Several features appear frequently and therefore can be assumed to be c hal'~cteristic of cadmium-induced failures . One feature often seen is cleavage. Figure 6-22 shows that cleavage regions have occurred at the interface of the fatigue precr, ck and the stress corrosion crack. The dominant feature on the fractu e surfaces of specimens that have failed in cadmium is fluted regions of the type shown in Figure 6-23. At higher magnification t he cleaved and plastically deformed ar~as of the flutes can be seen more clearly, as illustrated in Figure 6-24 . Cadmium-induced failure generally appears to follow a transgranular path; however, in one instance evidence of intergranular failure was observed, as shown in Figure 6-25. Optical metallography of fracture surfaces using polarized illuminati on confirmed that intergranula. ' features were present but were much less common than transgranular features.
In summary, cadmium stress corrosion cracking of stress-relieved Zircaloy-4 will /' . occur in rising load tests if the oxygen partial pressure is low and if the specimen 6-39  The threshold stress intensities (K ) for the halogens were estimated to be I s cc about 9 MPa·m 1 / 2 at 633 K and iodine-induced cracking proved to be essentially immune to the presence of oxygen.
In contrast, cadmium and cesium were aggressive toward Zircaloy at fuel cladding operating temperatures only when s t eps were taken to reduce the oxygen content s of the test environments to very small values. Furthermore , cesium and cadmium seemed to require a riSing load type of test before embrittlement occurred and cesium had to be contaminated with a trace of iron. K for cadmium cracking Iscc 6-44 at 633 K was estimated to be about 20 MPa·m 1 / 2 , but this value was believed to be an overestimate. In tests on smooth specimens, cadmium and iron-contaminated cesium generally caused embrittlement at stresses just above the yield stress, although on ~wo occasions much smaller failure stresses were observed 1n the ironcontaminated cesium environment.
Gaseous hydrogen proved to be much more aggressive toward Zircaloy at ambient temperature than at fuel cladding operating temperatures. Hydrogen was capable of promoting subcritical crack growth in statically loaded s pecimens at room temperature but an alternating st res s was required to obtain any crack extension at 573 K.
The present evidence suggests that the halogens are more conSistently aggressive toward Zircaloy at reactor operating temperatures than the other elements screened.
Thus , the halogens were the only substances that caused cracking in the statically loaded WOL-CT tests and were relatively insensitive to the presence of contaminant s. Iodine was therefore selected as the environment for the detailed studies reported in Sections 4 and 5. However, it cannot be concluded that 12 is more aggressive than, say, Cd in the particular conditions that prevail inside an operating nuclear fuel rod. For instance both Cd and Fe-contaminated Cs caused embrittlement of 7FDll Zircaloy-4 tubing in diametral compression tests at temperature,s -600 K. In contrast, when a specimen was similarly tested in the pr esence of 1 2 , no evidence of embrittlement was observed. This result, which js probably a consequence of the sensitivity of iodine-induced SCC to both strain r~te and mod e of loading  shows that under s ome circumstances, Cd and Fe-contaminated Cs are more aggressive toward Zircaloy than the halogens. Consequently, the results of the present screening tests do not allow a positive identification of the chemical agent responsible for PCI cladding failures. However, when the present dat a are combined with the availability considerations reported in Section 2, we can tentatively conclude that the chemical substances most likely to be involved in PCI failure are 1 2 , ~d, and perhaps Fe-contaminated Cs.

Oxide Penetration
Oxide cracking apparent ly is an e~sential first st e~. Although we have observed the formation of pits in unst~essed, unoxidized Zircaloy exposed to iodine (which indicates that iodine is capable of penetrating thin, air-formed oxide films), these isolated pits do not correlate with the locations of iodine-induced crack nuclei in the metal. Mo over, iodine apparently cannot penetrate oxide f ilms with thicknesses in the pange relevant to the inside surface of irradiated fue l cladding (~ 0.5 ~) unless the oxide is ruptured mechanically. We therefore believe that before the oXide film can be penetrated during iodine-induced SCC, the cladding must attain a sufficient level of strain to mechanically rupture the oxide film.
Experiments mentioned ~ Section 4 show that ~ l-~-and ~ 3-~-thick oxides (formed by thermal oxidation in dry oxygen) are detectably cracked when Zircaloy samples are subjected to a total (elastic plus plastic) str~in of ~ 0.4% at reactor operating temperatures. This strain would be reached at a s~ress of a~out 300 MPa in a perfectly elastic sample of Zircaloy. However, that i s an upper limit for the cladding stress required for oxide rupture because plastic floVl and creep usually arti possible at smaller stress levels. For example, it was found that a short time hoop stress of only -200 MPa was required to generate a hoop strain of 7-1 0.4% in annealed Zircaloy-2 cladding at 590 K. Similarly, the data in Table 5 At ~igher strains, the number of cracks in the oxide increases systematically, a s shown in Figure 4-19 and the number of cracks formed at a given strain depends on the oxi~q thickness (Figure 4-20). Thus, the amount of metal surface exposed to the environment depends both on strain and on the nature of the oxide film.
The threshold stress for iodine SCC of the annealed Zircaloy-2 tubing was about 280 MPa, whereas oxide cracking could be detected at a stress of -200 MPa. Moreover , the threshold stress was completely unaffect ed by preoxidation. We therefore believe that although oxide film fracture is necessary to allow contact between the metal and the aggressive environment , penetration of the oxide film does not automatically imply that SCC will occur. That is, oxide film penetration is not the critical step in the overall iodine-SCC process.
Our information on oxide pene tration a ddresses the behavior of oxide films formed by thermal oxidation in dry oxygen at a pressure of about 0.1 MPa. We have no data on the strains required to rupture oxide films formed under the low oxygen potential condit i ons that prevail inside an cperating nuclear fuel rod. Howev er, there is no reason to believe that the resistance of such films to mechanical rupture is s ignificantly superior to that of oxides formed at higher oxygen pressures. In f act, some of the observations reported in Section 2 suggest that the oxide films fornled on the cladding in-reactor are l" t her imperfect. We therefore doubt that oxide film penetration is a critical step in PCI failure of nuclear fuel cladding.

Crack Formation
Cracking of the protective oxide film exposes a small a rea of Zircaloy metal surface in the metal-oxide i~telface to the iodine-containing environment. If the applied stress is sufficiently large, stress corrosion crack nucleation then occurs at certain favored sites in the exposed metal surface.

7-3
We think that the threshold stresses observed in the pressurizatiOfi tests (Section 5) are associated with crack formation because the max i mum stress for continued growth of a preexisting stres s corrosion crack is expected to decrease with increasing crack depth. Therefore, we regard crack formation as the critical step in the iodine-induced failure of internally pressurized Zircaloy tubes.
Our tube pressurization data indica t e that stress rather than strain or e' r~= s intensi ty is the key parameteI in crack formation, as discussed in Sect ion 5. The present data indicate that threshold stress depends on microstructure (f.nnealed is inferior to stress relieved) and probably also on irradiation (irradiat d is inferior to wlirradiated) but is not sensitive to iodine concentration in the range 6 to 0.06 mg per square centimeter of Zircaloy surface.
The indenter studies discussed in Section 4 indicate that two populations of crack nuclei are formed in unirradiated Zircaloy. Small nonpropagating cracks form at stresses well below the threshold stress for iodine-induced failures of internally pressurized tubes. These sma 1 cracks generally seem to be intergranular and occur at sites such as certain grain boundaries in recrystallized material where the local stresses generated during plastic deformation are exceptionally large.
Crack nuclei of the second type are larger than those of the first type , g " erally seem to be trans gran'llar, and always occur at sites in the metal surface that contain higher-than-normal concentrations of alloying e lements or impurities. These crack nuclei are observed only in specimens stressed at level s above the threshold s tress for iodine-induced failure of internally pressurized tubes. W e therefore believe that the threshold stress is the stress required to form crack nuclei at chemical inh~mogeneities and that their propagation causes iodine-induced failure.
The exact mechanism that results in the formation of cracks at loca l concentrations of impurities and alloying elements is not currently known. In Sec~ion 4 we suggested a mechanism based on the formation of brittle mixed metal iodies. This mechanism probably can account for many of the experimental results but it is not obvious why the threshold stress for recrystallized Zircaloy-2 should be lower 7-4 than that for stress-relieved Zircaloy-2 from the same tubing batch. Perhaps interdiffusion at the higher annealing temperature used for the recrystallized material increases the effective size of the chemical inhomogeneities. Also, the mechanism suggested in Section 4 does not provide an i mmediate explanation for the effect of irradiation. Possibly oxygen is picked up by the local inhomogenei ties during long-term in-reactor exposure and thi ' makes them more susceptible to embrit tlement by iodine. Another possibility is that the small crack nuclei that are formed at lower stresses but do not propagate in unirradiated Zircaloy are capable of growth in irradiated material, perhaps as a cons equence of intense local stresses a ssociated with di s location channeling and flow localization.
Whatever the exact mechanism, crack formati on is a rapid process. Crack nuclei were observed after times under load above the threshold stress as short as 5 minutes. This suggests that the time to f ailure in pressurized tube experiments may be associated chiefly with the time required to grow the largest crack nucleus formed in the s pecimen to a sufficient size for final rupture of the tube to occur.
In that case, the very consistent times to fai lure observed in the tube pre~suri za tion tests (see, for example, Table 5-3) must indicate that the largest crack nuclei formed in ~ 3 cm lengths of tubing (the uniformly stressed region in our tube press uri zation s pecimen) are always of very similar size . This is not unreasonable if the number of nucleation sites in such tube lengths is very large, as the result s of the indenter tests (Section 4) suggest.

Crack Propagation
Once a stress corrosion crack has been initiated, crack growth will occur if the stres s remains sufficiently large and the specimen continues to be exposed to a sufficiently aggressive environment. In the present work, an iodine concent r ation of 0.06 mg I2 per cm 2 Zircaloy was found to be more than sufficient to al low the stres s corrosion process to proceed; the results of Busby et al. (7-2) indicate that the i odine concentration threshold is less than 0.005 mg/cm 2 • Crack propagation occurs on a surface a pproximately perpendicular to the principal applied stress ( Figure 5-9) and more crack branching occurs at higher stress levels.

7-5
The results of our tube pressur~zation tests on preflawed specimens suggest that the minimum value of nominal hoop stress for continued crack growth initially falls linearly with flaw siz e in such a way that the net section stress remains essentially equal to the crack formation threshold stress. Similar behavior has been reported by videm and Lunde ~7-3), who showed that the pr~sence of fatigue-sharpened flaws up to 100 ~ in depth did not reduce the net section stress threshold for iodine sec of annealed Zircaloy-2 at 613 K.
At larger crack sizes, one would expect that the minimum nominal stress for continued crack growth a i would be related to the stress intensity threshold K If the failure times in the pressurization tests depend mainly on th e time required to propagate crack nuclei of almost constant size (aR suggested in the previous subsection), the data in Section 5 suggest that the crack growth rate is more 7-6 -.rapid in recrystallized Zircaloy-2 than in stress-relieved Zircaloy-2 and probably increa~es with iodine concentration, irradiation, and an increase of testing temperature from 590 to 630 K. The effects of testing temperature and iodine concentration probably relate to the availability of iodine at the crack tip, whereas the effects of irradiation and heat treatment probably reflect changes in the deformation behavior of the metal in the vicinity of the crack tip.
The mechanism by which iodine promotes intergranular fracture and transgranular cleavage in Zircaloy is not understood at present nor do we clearly understand the distinction between the two crack propagation paths. Iodine-induced crack propagation at stress intensities above K in fracture mechanics specimens is always Iscc transg ranul~r (7-1, and see also Section 6). Perhaps propagation tend s to be in~e~granular in re~rystallized Zircaloy tube specimens because the low yield stress and small cross section combine to relax the stress state at the crack tip to such an extent that transgranular failure (which probably requires the expenditure of more energy because of the involvement of plastic processes) becomes very dif ficult . In both intergranular and transgranular crack propagation, we presume that the presence of iodine at the crack tip in some way weakens the interatomic bonding sufficiently to permit the local stress to break Zr-Zr bonds.

Cladding Rupture
Stress corrosion crack propagation continues until plastic instability occurs, res ulti~g in rapid, ductile fracture of the uncracked ligament ahead of the crack, usually on a shear surface ( Figure 5-9). We observed two modes of final failure-pinholes, which were more common at low stresses, and short axial splits, which were more common at high stresses.

7-7
Two criteria have been proposed previously to define the conditions required for cladding rupture. Videm and Lunde (7-3), use a critical apparent s t ress intensity criterion, whereas Kreyns, et al.  prefer a critical net section stress, which they regard as equal to the ultimate tensile strength (UTS) of the cladding. Neither of these criteria is entirely satisfactory.
The presence of X-marks on the outside surface of cladding spec~nens pri or to failure in internal pressurization sec tests and at failure sites in t est reactor experiments  indicates that the uncracked ligament becomes fully plastic before final failure. The use of a linear elastic stress intensity failure criterion is clearly inappropriate in this situation. The UTS criterion is probably a good approxima~i on at high stress levels where the stress corrosion crack does not penetrate very far into the metal before plastic instability occurs, but it seriously underestimates the net sectiQn stress req~ired for instability at low stress levels because the plastic constraint associated with deep cracks is ignored. As Cottrell (7-6) has pointed out, plastic constraint in a deeply notched tensile member can rai se the apparent yield strength of the uncracked ligam~nt by about a factor of 3. Thus, if we assume that in the present case plastic instability immediately follows net section yielding, ductile failure of the tube will occur when the net section stress reaches approximately 3a (where cr is the yield strength of the tube material) in y y a deep-notch, low-stress situation. Therefore, the net section stress required for fiual cladding failure should rise from a value close to the cladding UTS for shallow cracks to a value close to three times the yield strength for deep cracks.

RELEVANCE OF THE RESULTS TO PCI FAILURE
The results of power-ramping experiments in test reactors (7-1, 7-7) provide convincing evidence that PCI failures are due to the combined action of stresses in the cladding associated with the power ramp and t e chemical environment to which the inside cladding surface is exposed. That is, PCI failures occur under conditions where neither the stress applied to the cladding nor the chemical environment would a lone cause failure. The general picture of the mechanism of PCI failure that is begi nning to gain wide acceptance is as follows. As the power increases, the temperature at the fuel cente r rises and causes the fuel to expand. Fuel expansion generates tensile strains and stresses in the cla dding that are locally intensified at the inside cladding surface at locations such as pellet-pellet il:terfaces and fuel chip s and over pellet cracks. The protective Zr0 2 film is ruptured at s uch locations, exposing the underlying metal to a corrosive environment that either has been released previously f rom the fuel or is rel eased during the ramp as a result 7-9 of the temperature increase in the fuel center. If the local stress/strain state is severe enough and the environment is corrosive enough, and if both stress and environment are sustained long enough, a small stress corrosion crack will form at the inside surface of the cladding and grow until it penetrates the cladding wall.
In the context of this model of PCI failure, the most important observations of the present work are: • Significant quantities of a number of fis s ion products and fuel impurities can reach the cladding of power reactor fuel under some operating conditions (Section 2) and several of these substances can embrittle Zircaloy (Section 6) • Crack nuclei form at specific sites in the metal surfnce when stressed Zircaloy is exposed to iodine (Section 4) • A stress threshold exists for iodine sec of Zircaloy at reactor operating temperatures below which the metal is immune to SCC failure (Section 5).
The significance of each of these observations is amplified below.

The Chemical Environment
The availability of sufficient concentrations of embrittling substances is one of the prerequisites for the occurrence of PCI failure and if contact between the cladding and the embrittling substances could be prevented, PCI failure would be suppressed . Our results indicate that several embrittling substances can pot entially reach the cladding in sufficient concentrations to promote failure.
Therefore, to be successful, any remedial measures must be broad-based rather than specific. That is, they must be effective in preventing contact between the cladding and several different fission products and impurities. One way in which this could be accomplished would be by s uppr~ssing the release of the aggressive substances, either by keeping fuel center temperature below the levels at which f ission products are released or by interfering with the processes involved in the release of fission products and impurities from the fuel. Another approach would be to place a physical barrier of some sort between the fuel and the cladding, e.g., a coating on the cladding. Both concepts are currently being investigated. For instance, the CANLUB graphite coating system is believed to be successful in part because it preve~ts fission products from reaching the cladding (7-8). elements were observed about one-third of the samples ~xamined . The area of strain produced and examined in each sample was typically ~ 1 mro 2 • Therefore, there must have been one large crack for every ~ 3 mm 2 of Zircaloy surface. Thus , at the relatively large strains used in the indenter tests, oxide cracking exposed about 30 SCC initiation sites per square centimet9r of Zircaloy surface . Since the area of th e insi de surface of a O. l-mro length of Zircaloy cladding is ~ 3 mm 2 , we can tenta ti vely conclude that t he re ig likely to be ~ 1 crack ini ti a tion s i te wi thin an axial d ist ance of about ±0.05 mm from every pellet-pellet interface .
Thi s crack site denSity is in reasonable agreement with observations made on s everely ramped test reactor fuel rods. In such rods, up to ~ 10 cracks are somet imes observed at a Si .151e pellet-pellet interface. At a denSity of 0.3 site/ mm 2 , 10 cracks would be expected at one pellet-pellet interface if t he region of strain concentra tion ( r idgi ng) a ssoci ated with that int erf ace extended f or an axial di stance of about ±0.5 mm, which is of the order of magnitude observed in profilometer 7-11 measurements on ridges in severely r amped fuel rods. Thus , the data are not inconsistent with the idea that t he s ~es at which PCI cracks form are the same sites as those at which we observed initiation of iodine-induced stress corrosion cracking ih unirradiated Zircaloy tubings.
It has yet to be confirmed that the initiation of PCI failure in power reactor fuel rods occur5 at sites in the surface of the cladding that contain abnormal concentrations of, alloying elements and impurities. In project EPRI RP 829 incipient defects from power reactor fuel rods are being examined to add r es s that question (7)(8)(9). In the interim it is worthwhile pursuing method s of eliminating t hes e sites to impI'ove the PCI resistance of cladding. In an EPRI funded follow-on program at SRI Int ernational, the effect on see susceptibility of state-of-the-art s urfa ce tr eatments that could be applied to homogenize or modify the inside surface of commercial tubing will be evaluated prior to test reactor demonstration of improved PCI resistance.
Although this remedy appears promising, it 1s poss i ble that features other than chemical inhomogeneities in the metal s urf a c e are important, and so it is appropriate to proceed wi th caution. The literature reports that see by iron iodide and aluminum iodide is more severe than by iodine alone (7-1). That res ult may indicate that elements transported to the metal surface through the vapor are effective in promoting crack initiation, which means that crack initiation in met al halide environments may not require the presence of a chemical inhomogeneity in the metal surface. Therefore, in addition to investigating the origin of chemical inhomogeneities in Zircaloy and developing ways of eliminating them it would be highly desirable to clarify the reason why iodine stress corrosion cracks preferentially initiate at such inhomogeneities.

The Stress Threshold
Our experiments with pressurized tubes (Section 5) show that a stress threshold exists for iodine sec of Zircaloy at reactor operating temper atures below which the Zircaloy is immune to iodine sec and above which the failure time decreases systematically with increasing stress. If PCI failure exhibits similar 7-12 characteristics, the present results suggest it should be possible to opt i mi ze the PCl resistance of the Zircaloy cladding by modifying its microstructure.
PCl f ailures are not encountered if rates of power change are kept very small.
Although this may be because fission products are not released from the fuel during slow power ramps, it is also clear that a small rate of power change will result in the generation of small cladding stresses because creep of the cladding and fuel can largely accommodate the thermal strains if the strain rate is slow. Thus, the effectiveness of vendor-recommended power-change restrictions in suppressing PCl failure ~rovides some support for the idea that a threshold stress for PCl failure does exist. W hether or not there is a PCl threshold, any steps that reduce local stress concentrations in the cladding should reduce the incidence of PCl failure und er given operating conditions. Hence, lubricants that prevent fuel-cladding bond ing, chamfered fuel pellets that reduce ridging at pe llet-pe llet interfaces, and ductile fuel-cladding interlayers ( " barriers") that reduce local stresses by p1asticly deforming should all be somewhat effective in alleviating PCl.
Three aspects of iodine-induced failure of Zirca10y that are particularly relevant to PCl failure require further work before optimum operating procedures can be properly defined. First, additional data are required to define the stress threshold for irradiated Zi1."ca10y more exactly and to position the stress versus failure-t ~me envelope at stresses above the threshold. Second I the effE~(!t of iodine concentration on the stress threshold and failure time for irradiated Zirca10y should be studied, particularly at very small iodine concentrations. Finally, a damage accumulation law* is needed at stresses above the threshold so that the behavior observed in constant stress laboratory tests can be related to the variable stress situation encountered in fuel cladding service. These needs are being addressed under EPRl Projects RP 1027 and RP 971 (7-9).
*That is, we need to know how prior exposure at one stress above the threshold affects the life of a specimen that is subsequently exposed at a different stress. ':'-13 ~.
If iodine sec and PCI behavior ,are similar, one approach to optimizing the PCI r esistance of cladding would be to change the microstructure so as to maximize both the thres~old stress and the time to f~ilure at stresses a bove threshold.
As shown in Section 5, unirradiated stress-relieved Zircaloy has a higher threshold stress than recrystallized Z1rcaloy and also shoV{a longer failure times. It would be desirable to confirm the superiority of a stress-relieved microstructure with tests on irradiated material, and the data being generated on RP 1027 should resolve the question of the optimum microstructure in the irradiated condition (7-9).
In the fuel cladding application, Zircaloy tubing is s ubjected to a constant displacement type of loading. Since SCC failure requires that the applied stress be sustained above a crack size dependent minimum value, any attempt to optimize the cladding characteristics must consi er the stress relaxation characteristics of the cladding (which determine the stress-time history under fixed displacement loading) as well as the threshold stress and failure time. Present evidence suggests that the crystallographic texture of the cladding is one of its key characteristics.
In particular, there is evidence that a 0 0 basal texture is more resistant to iodine SOC than a ±30 0 texture (7-10). As texture also affects stress relaxation behavior, the effects of texture variations on iodine SCC and stress relaxation will be studied in detail during the follow-on to the present program.  Table 3-1 in Section 3 of this report. The tables and illustrations in this appendix provide details of (1) the reduction s chedules used by TWCA to fabricate the flat products (Tables A-I -I to A-28). A discussion of the significance of these results is presented in Sect ion 3 together with additional metallurgical character ization information on some of the experimental materials that was generated during the program.
A-I (l)Seo Table 3-1 for additional details.
A-9  and Appendix A) were forcee onto stainless steel or aluminum tapered mandrels that had been chilled to liquid nitrogen temperature. The sample was stres 3ed by heating it and the mandrel to a temperature in the range 590 K to 750 K. Stressed samples were exposed to iodine vapor for various times from 40 to 200 hours and were periodically examined for cracking.
It proved to be very difficult to achieve iodine sec by this technique. In all 122 samples were test~d. Early experiments at a temperature of 590 K were unsuccessful; hQnce I the test conditions were made increasingly severe. A few cracks v·· "e fi.nally obtained in specimens that had bee notched longitudinally, stressed with stainless . steel mandrels, and heated to 750 K in an argon gas stream containing iodine at a pressure of 300 Pa (~2 torr). The tests were run on batches of seven samples.
In one batch, t wo samples out of seven failed in 41 hours, and in another batch, five out of seven samples failed in 94 hours. The sporadic incidence of cracking may indicate that the susceptibility of the samples of tubing is variable.
~fter the tests, the samples were examined in a scanning electron microscope (SEM).
. :1 The surfaces were found to be coated with a layer of corrosion product that contained patches of acicular crystals of FeI 2 • In some areas of the outside surface of a sample that suffered sec, there were cracked protuberances that seemed to have been formed by a reaction that caused localized swelling. An example is shown in Analys is sh owed only Zr.

8-5
proces ses . Occasiona lly a feature like the one in Figure D-5 was observed.
Analysis of such features showed only Zr .
These stressed ring tests show that under the conditions of the tests, iron was transferred from the stainless steel mandre l to the Zircaloy by the iodine. In addition, Fe was found in the regions of the fractur~ that occurred by cleavage but was not found in the ductile regions. It seems likely that the iror contributed in some way to the occurrence of cleavage; however, it can also be argued that the iron was deposited on the freshly formed surfaces after cracking had occurred. 8-6