e 3 4yys5L D358073 & g AEC RESEARCH AND DEVELOPMENT REPORT W : Fasacn, ‘flQ'L{‘M - g INTERIM REPORT ON CORROSION BY ALKALI-METAL FLUORIDES: WORK TO MAY 1, 1953 (A1) w]sh'-samea - report, name with report and o nbrary -will-mange a:,_loan ~ ORNL-118 (6-97) NATIONAL LABORATORY operated by UNEON CARBIDE CORPORATION r for the L TOMIC ENERGY COMMISSION c”‘“w Tg"!'fl#grrfl E 'H'fl-' Il"w “’fifigflgflfl.filr" r!ll'§ ? #....'r'.f" F'fl:f’ _""=='gf\ h: - r rrTET -u----'.- e —-— A =, | ofios e - - fp'!fl"{gc_} M = lf ?JF“ This document consists of 44 pages. Copy 7.6 of 222 copies. Series A, Contract No. W-7405-eng-26 METALLURGY DIVISION INTERIM REPORT ON CORROSION BY ALKALI-METAL FLUORIDES: WORK TO MAY 1, 1953 G. M. Adamson R. S. Crouse W‘ D. Mdnly DATE ISSUED MAR £ 01333 OAK RIDGE NATIONAL LABORATORY Oak Ridge, Tennessee operated by UNION CARBIDE CORPORATION for the U.5. ATOMIC ENERGY COMMISSION SECRET” TR 3 4456 03548073 & CONTENTS Summary .. Introduction .... Experimental Methods............... EQUIPMENt .uisisimissmonsessspnssvsonsy Procedure.....cccceeeevrennnn. Results and DiSCUSSION vuuviiivireeriverieeererressensenns Screening Tests ....coceriiviennsnranns Corrosion of Inconel .... Standard Inconel Serles Effect of Fluoride Composmon sipvisias Oxide Removal Procedures.... Effect of Hot-Leg Temperature ..... Crevice Corrosion .... . .. Mechanism of Corrosive Afluck Corrosion of Type 316 Stainless Steel Effect of Temperabure: ... uciiainamosohiiis s sieisiasiie Plug Identification... = il et i Effect of Fluoride Composmon ST rerrmmEre — o MR 11 11 11 21 22 22 24 28 28 28 28 FOREWORD This report reflects the work done to May 1, 1953, on corrosion by alkali-metal fluorides. |t is realized that some of the interpretations will appear to be naive when compared with those which have been developed from subsequent investigations. The purpose of issuing the report is to incorporate the data into a permanent record. INTERIM REPORT ON CORROSION BY ALKALI-METAL FLUORIDES: WORK TO MAY 1, 1953 G. M. Adamsen SUMMARY In connection with the search for container ma- terials suitable for use in high-temperature molten- fluoride reactor systems, the corrosion of several metals by various fused fluoride mixtures was studied in thermal-convection loops made from the metal being tested. In these tests the temperature drop was fairly large, but the velocity of the liquid was low and not subject to control. The fluoride mixture used in most of the study contained 43.5 mole % KF, 10.9 mole % NaF, 44.5 mole % LiF, and 1.1 mole % UF4, and is referred to as "“fluoride 14, In a series of screening tests Inconel, or similar high-nickel alloys, was shown to be the most promising container material. It is called to the reader’s attention that the data reported are from loops that were operated for only short times, 500 hr, and with impure fluoride mixtures. 1 relatively More recent data are reported elsewhere,’ The lack of purity of the mixtures was an important consideration in the tests re- ported; however, the purification processes have since been perfected by the Materials Chemistry Division. The more recent work' shows that mass transfer is of more importance than it was con- sidered to be at the time of the reported tests and also that the rate of mass transfer varies greatly for different fluoride batches. With fluoride 14 an average figure for the maxi- mum microscopically visible attack on Inconel was 0.012 to 0.015 in. after 500 hr of circulation. The attack was in the form of small voids which did not connect with the surfoce or with each other. It is shown that for the times used here the attack was caused primarily by the reduction of fuel impurities by chromium metal from the container and by the leaching of sufficient chromium to reach equilibrium concenfrations in reactions with the fluorides, primarily with UF ;. In contrast to the effects of impurities, the depth and type of attack varied only slightly when changes were made in the major com- ponents of the various fluoride mixtures studied. 16 M: Adamzom; R. 5: Crovse: ond We Doiddnly; Interim Report on Corrosion by Zirconium-Based Fluo- ride Mixtures, ORNL=2338 (to be issued). R. S. Crouse W. D. Manly The only effect definitely established was that, in every case, mixtures containing uronium fluoride caused more attack than did similar mixtures con- taining no uranium. When the loop was operated at a fluid temperature of 1650°F, the attack was slightly deeper thon it was ot 1300°F. The od- dition of zirconium hydride to the fuel almost com- pletely eliminated the attack. acted as inhibitors. It was found that type 316 stainless steel might be considered for a container for molten fluoride mixtures that do not include potassium fluoride. With such stainless steels and zirconium fluoride— Alkali metals alse base mixtures, the maximum depth of attack was about the same as that found with Inconel; how- ever, more material was mass-transferred and cold- leqg deposits were found. When olkali-metal-base mixtures were circulated in type 316 stainless steel, plugging occurred in a relatively short time; the plugging was probably caused by the formation of a compound similar to K NaCrF . INTRODUCTION A major problem connected with the development of circulating-fluoride-fuel reactars for aircraft propulsion at ORNL? has been that of corrosion, since these reactors will involve high-temperature operation and the use of corrosive liquids. There- fore the Experimental Engineering Section of the Aircraft Nuclear Propulsion Division, the ANP Reactor Chemistry Section of the Materials Chem- istry Division, and the ANP Materials Section of the Metallurgy Division were assigned the joint task of finding a material that could be fabricated into thin-walled tubes and that could be used in a molten-fluoride reactor system in which the maxi- mum temperature of the fluid is 1500°F and the temperature drop is 400°F. This interim report deals with some of the metallurgical phases of the problem, The high temperature, considerable temperature drop, and dynamic flow conditions under which the ZA. F. Fraas and A. W. Savolainen, ORNL Aircrafi Nuclear Power Plant Designs, ORNL=1721 (Nev. 10, 1954). v power reactors will operate all have an effect on corrosion and therefore had to be included in the study. A dynamic test is the only type of cor- rosion test that will incorporate these variables and permit a study of mass transfer. When this work was started, pump loops were not feasible, since reliable high-temperature fluoride pumps and seals had not then been developed. A loop em- ploying thermal convection for the driving force that hod been developed by the General Electric Company® and that had been modified by the Ex- perimental Engineering Section of the ANP Di- vision at ORNL for corrosion testing with sodium® was simplified and odopted for the preliminary stages of the investigation. The thermal-convection loop could be operated with a hot-zone temperature of 1500°F with o reasonable temperature drop and was fairly simple to fabricate. Its main disadvantages were the low and fixed flow rates and the size of the system. Standard loops comprise about 10 ft of pipe and contain about 35 in.? of liquid, which made them obviously unsuitable for conducting a large number of screening tests. Thus the thermal loops were used for an intermediate test. The preliminary screening work was done with static and seesaw tests,*r® whereas the final testing would utilize high-velocity pump loops. Since the thermal loops were an intermediate step, the data obtained from them are not to be considered as design data. Such data can be obtained only from systems that more closely reproduce the proposed operating conditions. However, thermal loops have proved extremely useful in investigating corrosion mech- anisms and in ascertaining relative effects of system temperature ond temperature drop. The program was divided into two parts: o search, consisting in o series of screening tests based upon data obtained in preliminary capsule tests,” for the best material for containing the fluorides and a search for a method to render E'L. F. Epstein and C. E. Weber, in Use of Molten Sodium as a Heat Trans fer Fluid, TID-70, p 59 (Jan. 1951). 4R.B. Day, Examination of Thermal Convection Loops, memorandum to E. C. Miller, April 12, 1951, L. §. Richardson, D. C. Vreeland, and W. D. Manly, Fggac;sion by Molten Fluorides, ORNL-1491 (March 17, 6p. C. Vreeland, E. E. Hoffman, and W. D. Manly, Nucleonics 11(11), 36=39 (1953). Incone! acceptable as a container material. Ef- forts were made to understand the chemical re- actions and corrosion mechanisms both portions of the program. All loops were operated for a standard time of 500 hr. The composition of the fluoride mixture pro- posed for use as a reactor fuel was changed several times during the period of this investi- Because of economic reasons and limited time, it was necessary to switch to o new fuel be- fore work could be completed on the old one, which resulted in incomplete corrosion information being obtained for many of the proposed fuels. The majority of the work reported herein was carried out with fluoride 14 (composition given in Table 1), the first composition proposed for a circulating- fuel reactor. Considerable work was also done with zirconium fluoride~base fuels, fluorides 27, 30, and 44. Details of this work are given ina separate re.'porf.] involved in gation. EXPERIMENTAL METHODS Equipment The major equipment used in this investigation was the thermal-convection loop. The advantages and disadvantages of this equipment are discussed in the “Introduction.’”” Figure 1 shows the con- figuration which has been used as a standard, except for a few minor modifications to some of the early loops used in the screening series. These early loops were filled through a flanged line into the side of the surge tank rather than through a Swagelok fitting in the top. The loops were assembled by Heliarc welding with an inert- gas backing. Figure 2 shows a partly assembled loop, its support, and the location and method of attaching both the heaters and the thermocouples; Fig. 3 shows some loops in operation and the location and type of insulation used. The loops were heated by six sets of Hevi-Duty tubular-furnace heating elements. Each set of elements was 6 in. in length and 11{‘ in. in inside diameter. The heaters were centered on the pipe and separated by means of ceramic spacers in- serted in the ends. The heaters were connected in two parallel circuits, and the power was supplied from saturable-core reactors. This arrangement provided for propertional control rather than on-off control. k. UNCLASSIFIED ORNL-LR-DWG 35691 3'-"2 in. =5 VENT LINE—= /-'SPfl.RK PLUG SWAGELOK FITTING—mp. o, fi’| /ELECTHI(‘AL CONNECTOR 1L I SURGE TANK ——== I | 41{'2 o THERMOCOQUPLE WELL Lt § * 2%z in. ] ! i X 2 12 ] | x 5 . --——1"."2...in' SCHED-4O - 26 J”B n. SEAMLESS PIPE ————= 1 3 K | §— 4 10 Y 75deg ELECTRICAL CONNECTOR YE Fig. 1. Standard Thermal-Convection Loop. UNCLASSIFIED PHOTO 21343 Fig. 2. Partly Assembled Thermal-Convection Loop. | UNCLASSIFIED ' PHOTO 21341 Fig. 3. Thermal-Convection Loops in Operation. During the filling of the loop, auxiliary heating was needed. The loop was heated by passage of an electric current through the pipe itself, and the expansion pot was heated by a coiled 1500-w Calrod heater. Once the loop was filled, these auxiliary heat sources were turmed off and the clamshell heaters were turned on. Johns-Manville Superex insulation was used and was applied as preformed semicylinders with a wall thickness of 3 in. The two halves were wired together, and the cracks were filled with Superex cement. When the loops were operating ot 1500°F, the surface temperature of the insulation was about 200°F, The major experimental difficulties encountered were concerned with temperature measurement, |t was necessary to measure the maximum and the mini- mum wall temperatures, and it was desirable to know the temperature at several intermediate points, The temperatures were measured with Chromel-Alumel thermocouples and were recorded on 12-point Brown Electronik instruments. Ina dynamic system the question arose as to what temperatures should be measured and used for control. It was realized that saturation and reaction rates in the liquid would depend upon the temperature of the fluorides, that corrosion rates would depend upon the inner- wall temperature, and that the outer-wall temper- ature would be the easiest to measure. The di- ameter of the loops was so small that an inserted thermocouple well would change the flow con- ditions; also, since the flow was laminar, it was possible that the center fluid temperature would be considerably lower than the outside fluid Thermocouple wells could not be drilled into the pipe walls, because they were too thin. For these reasons, the only temperature that could reasonably be measured was the outside wall temperature. To obtain the outside wall temperature, ond one that would check the inner temperature reasonably well, the thermocouple beads were welded to the pipe wall with a con- denser discharge welder and then covered with a layer of Sauereisen cement. The maximum temper- ature was measured }’2 in. down the top horizontal pipe, while the minimum temperature was meas- ured 2 in. above the joint at the bottom of the cold leg. Both these thermocouples were under the layer of insulation. Other thermocouples, on ex- posed sections, were covered by a layer of as- bestos tape. Before these precautions were taken, temperature, temperature variations of 50 to 60°F were common, Even with the above precautions the temperatures should be regarded as being only relative and not exact, The power input to the loops was controlled by Leeds & Northrup Micromax instruments operating with Chromel-Alumel thermocouples, which were located between the first and second heaters and were welded on the vertical hot-leg section about 7 in. below the joint. A duplicate thermocouple was always placed on the opposite side of the pipe for use as a check and as a spare for control. Procedure One of the main advantages of the thermal- convection loops was their simplicity of operation. Since these loops had no seals or moving parts, once filled they required very little attention, Dur- ing filling, precautions were necessary to exclude air and moisture, but those precautions would be required with any corrosion test, The following steps were necessary to place the loop in operation, and some of them will be discussed in detail: 1. machining parts, 2. preliminary degreasing of all parts with a solvent degreaser, 3. Heliare welding of all bottom plug, 4. final degreasing and rinsing, 5. welding in either bottom plug or Swagelok fitting, 6. testing for leaks with an air pressure and soap bubble technique and by pumping to @ 50-u vacuum, 7. attaching heaters and thermocouples and mounting the loop on a stand, 8. insulating, 9. connecting to operating line, 10. drying by heating to a minimum temperature of 500°F and holding under a vacuum of at least 50 p for 30 min, 11. tilling, 12. operating for 500 hr, 13. sectioning and sampling for examination. The purity of the fluorides varied considerably, since they had to be obtained from a single large production facility rather than from o facility set up especially for the corrosion test program. No standard or fixed production procedure was used, and since the production and handling of such joints except the mixtures was a new art, many changes were made from batch to batch. In general, fluoride 14 was made by the following procedure. The sodium fluoride, lithium fluoride, and petassium fluoride were mixed, melted under a vacuum, and cooled to room temperature. The uranium fluoride wos added; then the batch was remelted and mixed by stirring with a stream of argon. The zirconium-base fuels were produced under the supervision of W. R. Grimes of the Materials Chemistry Division and were more highly purified and more uniform. The dry ingredients were mixed and then evocuated several times to remove all the moisture possible. After the mixture was melted, it was held at 1500°F and treated with hydrogen for 1 hr to reduce all oxides and oxy- fluorides of uranium to UO,; it was then treated with hydrogen fluoride for 2 hr to convert the UO, to UF ,. The last step was to purge the mixture of the excess hydrogen fluoride. This procedure has since been modified! for the fuels used in the main investigation of zirconium fluoride—base mixtures. Both types of fluoride mixtures were transferred from the production vessel to storage cans. When a loop was to be charged, the mixture was trans- ferred by pressure from the storage can to the loop through o B:t'in' nickel or Inconel line. All con- nections were made with Swagelok fittings. A spark plug probe was used to indicate when the loop was full. A sompler was placed in the trans- fer line but was not used for the majority of the loops discussed in this report. The distribution of impurities in the fluorides after circulation had to be determined, which made it necessary that the loops not be drained but that the fluorides be allowed to freeze in place. After cooling, the loop was cut up into approximately 6-in. lengths, with six 2-in. sections used for chemical and metallographic study. The location of these sections (x) is shown in Fig. 1. After the loops were sectioned, the pieces were turned over to J. P. Blakely, of the Materials Chemistry Division, who was responsible for removing the samples and obtaining the required analyses. Since no solvent had been developed for re- moving the frozen fluorides from the sections, they were melted at around 1125°F in a helium atmos- The sections were inspected visually, and if any unusual phases were present, additional phere, sections were cut and submitted for a petrographic examination. After the fluorides had been removed, they were ground and submitted for chemical analysis, and the small pipe sections were ex- amined metallographically. Check samples of the fluorides were obtained from the loop by drilling, and they showed that no changes in fluoride com- position or wall structure had token place during the short melting period. If any layers were dis- covered on the pipe wall during the metallographic examination, a sample was submitted for an x-ray diffraction study, RESULTS AND DISCUSSION The investigation consisted in a series of screen- ing tests to determine the most promising container materials for circuloting fluoride mixtures, a de- tailed study of Inconel as a container material, and a short, incomplete study of stainless steels. Most of the observations made and conclusions drawn in this report were based upon only one or two runs. Many of them were checked in seesaw tests, but this should not be regarded as com- plete confirmation. As mentioned previously, the loop tests were designed as an intermediate test between static testing and pump loop testing. All the conclusions were based upon loops operated for the relatively short time of 500 hr. For these reasons the data presented must be regarded as tentative and not be used as design data for high- velocity or high-temperature-gradient systems, |t is possible that there will be an increase in cor- rosion in going to high-velocity, high-temperature- differential pump loops, as was the case in going from the static tests to the thermal-convection loops. Various fluoride mixtures were used in this work. Their nominal compositions’ and number desig- nations are given in Table 1. Because of pro- duction variables, actual compositions may vary slightly from these figures. Screening Tests The initial effort in this program was the oper- ation of a series of loops to determine which ma- terials were the best suited for use in a plumbing system for high-temperature fluorides. The ma- jority of materials tested were those that showed promise in static corrosion tests carried out by 7C. J. Barton, Fused Salt Compositions, ORNL CF- 53"1']29 {Jdn- 15‘- TQSS). Table 1. Molten Fluoride Compositions* Flvoride UFs NaF KF ZrF, LiF BeF, No. Weight % Mole % Weight % Mole % Weight % Mole % Weight % Mole % Weight & Mole % Weight % Mole % 12 1.7 1.5 59.1 42,0 29.2 46,5 14 7.8 1.1 10.3 10.9 56.1 43.5 25.8 4.5 17 12,6 2.0 39,4 47.0 48.0 51.0 21 10.7 3.8 1.8 4.8 25.9 50.1 61.6 41,3 24 14.8 36,0 10.2 18.0 75.0 46.0 2 10.7 3.8 13.8 36,6 7.3 14.0 68.3 45.6 27 10.9 4.0 16.7 46.0 72.4 50.0 30 11.4 4.0 19.0 50,0 69.6 46.0 3] 20.1 50,0 79.9 50,0 32 21.4 52.0 78.6 48,0 35 54,2 57.0 45.8 43.0 44 18.6 6.5 20,5 63.5 60.9 40,0 *Sae C, J. Barton, Fused Salt Compositions, ORNL CF-53-1-129. groups under the supervision of D. C. Vreeland in the Metallurgy Division® and F. Kertesz in the Materials Chemistry Division. Fluoride 14 was the liquid circulated in these loops. The data from the loops operated as part of these screening tests are given in Table 2. A bar graph showing the relative times in which plugging oc- curred with the various materials is shown in Fig. 4. Typical photomicrographs from the top portion of the hot legs from four of these loops are pre- sented in Fig. 5. It was apparent from these data that the material selected must be one which would not cause plug- ging in the loop. The depth of corrosion was necessarily of secondary consideration. Nickel and Monel, which had been rated as the most likely containers in the static tests, had to be rejected because the thermal loop tests revealed excessive mass transfer of metal from the hot-leg surface to the cold leg. Figure 6 is o photograph of a mass of nickel crystals that had collected on a small flat sample inserted in the bottom of the cold leg of a nickel loop. From the data obtained for the alloys tested, it was apparent that the nickel-base alloys would moke the best containers for the fluorides. All other alloys tested either caused plugging in less than 500 hr or showed excessive mass transfer, Inconel, a commercial alloy, was deemed suffi- ciently resistant to fluoride corrosion to warrant its selection as a material to receive odditional attention. Since materials vorying widely in composition, such as the various stainless steels, iron, and nickel, all showed plugging or at least severe mass transfer, it was obvious that more than one metal or compound must be involved. In the iron and nickel loops the deposits were metallic, where- as in the stainless steel loops they were non- metallic. With the iron and nickel loops the mass transfer was possibly caused by very small changes in solubility with temperature. Changes of only a few ppm would be enough to account for the amounts of metal transferred. A material shown to be unsatisfactory in these tests should not necessarily be excluded for use with other fluoride systems. force in a thermal-convection loop is very small, the loop is quite easily plugged but may require only Since the driving slight changes in loop design, operating technique, or fuel composition to operate for 500 hr. In another part of this report it is shown that type 316 stainless steel loops can be oper- ated if the fuel composition or the minimum temper- ature is changed. Also, large isothermal pump loops of type 316 stainless steel were operated Table 2. Results of Container Material Screening Tests Loop Loop Migve. of Reason for Metallographic Examination 3 Circulation g Chemical Analysis No, Material (hr) Termination Hot Leg Cold Leg 40 410 9 Loop plugged 0.010 in. even removal; ne pitting or Metallic deposit with oxide Fe decreased; Cr increased intergranular ottack; tronsformed particles surface 43 410 12 Loop plugged Surface rough, some grains removed, Metallic deposit with non- All materials varied probably even removal; no inter= metallic erystals; visible granular attack metallic crystals in hot horizontal section 48 430 8 Loop plugged Rough and pitted; some removal Thin metallic layer with in- Cr increased; all others clusions varied 46 lzett iron 46 Loop plugged Surface rough with considerable even Many metal crystals on Large increase in Fe; remova | surface decrease in Ni 104 Nickel 500 Scheduled Mo pitting or penetration; 0.009 in. Metallic erystals in all Fe and Cr unchanged; Ni even removal sections increased in cold leg 107 Nickel 1000 Scheduled No pitting or intergranular attack; Heavy metallic crystal Fe decreased; Ni increased 0.010 in. even removal deposits slightly 341 Mone | 31 Leaked Approximately 0.010 in. even removal Metallic deposit 3465 Nimenic 75 500 Scheduled Intergranular pitting 0.008 o 0.013 in. Thin metallic deposit Cr increosed, Fe decreased, Ni varied 210 Inconel 500 Scheduled Subsurfoce holes from 0.010 to 0.015 in., Intermittent deposited layer Cr increased, Fe decreased, mainly in grain boundaries Ni and U varied 211 Inconel 524 Scheduled Subsurface holes mainly in grain Metallic-appearing deposit Cr increased; Fe and Ni boundaries, 0.004 to 0.008 in. 0.0005 in, decreased 0l Table 2 (continued) Loop Loop ¥ i 08 Reason for Metallographic Examination Circulation Chemical Analysis Ne. Material Termination Hot Leg Cold Leg (he) 219 Inconel 480 Heater failed Heavy primarily intergranular voids Light metallic deposit with Cr increased, Fe and U de- 0.005 to 0.013 in. o nonmetallic layer on top creased, Ni varied 229 Inconel 500 Scheduled Moderate to heavy primarily intergranular Deposit that was at least Cr increased; Fe and U de- pitting to 0.018 in, partially nonmetallic creased 227 Inconel 500 Scheduled Moderate to heavy primarily intergranular Thin metallic layer Cr increased, Fe decreased, pitting 0,006 to 0.016 in. Ni and U varied 112 Type 316 58 82 Loop plugged Rough surface with intergranular attack Rough uneven layer Cr increased, Fe decreased, up to 0.008 in. Ni and U varied 120 Type 316 55 62 Loop plugged Rough and pitted surfoce with inter- Thin deposit with both Cr and U increased, Fe de- granular attack 0.004 to 0.012 in. metallic and nonmetallic creased, Ni varied crystals adhering 127 Type 43 Loop plugged Heavy intergranular attack up to Metallic deposit on wall, Cr increased, Fe decreased, 316 ELC sS 0.011 in.; grains spongy with some attached crystals Ni and U varied 251 Type 310 5S 75 Loop plugged Very heavy intergranular attack and Nonmetallic deposit with a Cr increased, Fe decreased, general pitting 0.008 to 0.015 in.; thin metallic loyer Ni varied large grain growth 252 Type 310 S5 368 Loop plugged Very heavy intergranular pitting 0.018 to Heavy deposit; at least a All analyses varied 0.025 in, partial plug of metal in cold leg 275 Type 347 55 39 Loop lecked Severe intergranular attack 0.008 to Metallic deposit 0.0005 to Cr increased, Fe decreased, and plugged 0.013 in.; grain growth 0.001 in. thick Ni increosed 276 Type 347 SS 125 Loop plugged Considerable intergranular attack 0.002 Surface rough with thin de- Cr increased, Fe decreased, to 0,004 in. posit layer U varied, Ni constant UNCLASSIFIED ORNL-LR-DWG 35692 e00 500 £ o o | | | | | 1 TIME FOR PLUGGING (hr) 3 o RN INCONEL R\ INCONEL LC 316 30 [ [ STAINLESS STEELS CONTAINER MATERIALS IRON MNIMONIC Flgl 4. Loops Made of Various Materials. Times for Plugging of Thermal-Convection with alkali-metal fluorides. The lack of time and manpower made it necessary to concentrate efforts on the most promising materials and leave the others for future programs. Corrosion of Inconel Standard Inconel Series. — As is pointed out in the preceding section, Inconel was the material that appeared most promising in the screening tests. When the investigation was started, fluoride 14 was the mixture proposed for use as a reactor fuel. It soon became apparent that fluoride 14 could not be used because isotopically separated lithium was not available. The zirconium fluoride fuels were then proposed but could not be ob- tained immediately in sufficient quantity for test- ing. Therefore the corrosion study was continued with the Inconel-fluoride 14 system. The maxi- mum fluid operating temperature was set at 1500°F, as in the reactor designs then being considered. The cold-leg temperature was about 1300°F. A time of 500 hr was arbitrarily selected as the standard operating period. These conditions were established os standard and were used as a base W the study of the variables reported; Table 3 gives the results obtained. Figures showing samples taken from the hot legs of the loops are presented in several sections of this report: loop 229 is shown in Fig. 7, loop 227 in Fig. 9, loop 219 in Fig. 12, and loop 210 in Fig. 5. The loops were operated over a considerable time span with fuel from various batches that showed considerable variation in chemical composition. Chemical analyses of the fluorides after the loops had been operated all showed an increase in chromium and a decrease in iron content. The depth and in- tensity of attack found in the hot legs of the loops checked fairly well, except that the attack in loop 211 was slightly less thon that in the other loops. The following detailed description of loop 229 is presented as typical of the standard loops. Photomicrographs of various sections showing the distribution of attack around the loop are pre- sented in Fig. 7. Sections 3 and 4 show that once the attack started it proceeded quite rapidly. The maximum depth of penetration increased only slightly in moving up the hot leg from section 3 to section 1, but the concentration of voids in- creased considerably. In the upper hot-leg section more general attack was found. |t can be noted that the deepest penetration and the first sign of attack always occurred in the grain boundaries but that voids occurred frequently within the grains of the metal. The data obtained by analyzing the fluorides from the sections shown in Fig. 7 are given in Table 4. As was true in about 75% of the loops, the amount of both chromium and iren found in the cold leg was slightly more than that found in the hot leg. No systematic variation in uranium could be found. In all loops the chromium concentration increased during the run, while the iron decreased. When nickel was present as an impurity, its con- centration also decreased, but for these loops the original nickel impurity was usually low. Effect of Fluoride Composition. = Effect of Dif- ferent Mixtures. — During the progress of the ANP work a variety of fluoride mixtures have been pro- posed as suitable fuels or coolants. With the ma- jority of these fuels only exploratory corrosion tests were carried out. In every case, the change from one fuel or coolant to another was made for reasons other than comrosion. 11 ¢l UHCLASMFIED T 762 wh - _-':+ F Yy - 1 "_'_o_--‘..fi4 von SR E A ARS8 3008 g % s p o K oy Ay e TE a8y ' e Pt .f:' } 'J{:-.r' » -;‘-“1..‘!'”"...T\\:f:.;“t;"l'l_-P.‘- t_..o.q -.$ .-_',!' - $ - = . _"‘."H e : pe & w® PN .' o e »nst w Ty : i‘.‘ " .:.l'..“ - ‘:..4 s L wh N e - l" cew 8> W3 . M e e T TR R ] I ' » o S Ple ., o o N St . ‘..:?\I".t‘ 3 7 cede o " "" - 0 G w v - l‘ .. '.‘ * .'::1'.-". X % a2 ! o et ' { . - - ¢ t' j . ; L ?‘\-: - c“‘ e . . o .. :l“ v A .'lun I r f . e g % INCONEL, 500 HOURS UMCL ASSIFIED T 768 | % - & Lo Lok 4 Ve L % i‘! 316 STAINLESS STEEL, B2 HOURS EEEEFERBER CORR. I 1 EEEEFERBERE T 50X | 2 EEE] : T IHETIE! s0% | P e UNCLASSIFIED T 12327 NICKEL, 500 HOURS UNCLASSIFIED T 682 e 21N 440 STAINLESS STEEL, 12 HOURS oo kB INCHES [& 8 8 B FEE Fig. 5, Typical Hot-Leg Sections from Thermal Loops Made of Various Materials After Circulating Alkali-Metal Fluorides. 250X, Reduced 39%. -wifh caption) UNCLASSIFIED T-2824 INCHES Fig. 6. Nickel Crystals Mass-Transferred in o Thermal Loop Circulating Fluoride 14 for 500 hr at 1500°F, - with caption) - Table 3. Standard Inconel—Fluoride 14 Loops Loop CTiml'! :’_f Reason for Metallographic Examination il Aot irculation ST emica alysis No. (hr) Termination Hot Leg Cold Leg % 210 500 Scheduled Heavy concentration of small Very thin surface layer Cr increased, Fe decreased, holes in band from 0.010 in. U and Ni varied in to a maximum of 0,015 in, fluorides; Cr leached from from surface; mainly inter- wall granular but some in grains 211 524 Scheduled Moderate to heavy concentra- Metallic-appearing loyer Cr increosed, Fe decreased, tion of subsurface voids 0.0005 in. thick Ni remained low in fluorides 0.004 to 0,008 in.; pri- marily intergranular but some in grains 219 480 Heater foiled Moderate to heavy concentra- Thin metallic-oppearing Cr increased, Fe decreased, tion subsurface voids 0.005 layer with nonmetallic U constant, Ni varied in to 0.013 in.; primarily inter- particles as layer on top fluorides; Cr leached from granular wall 227 500 Scheduled Moderate to heavy subsurtace Thin metallic-appearing Cr increased, Fe decreased, voids 0.006 to 0.016 in. layer Ni and U varied in fluorides 29 500 Scheduled Moderate to heavy subsurface Thin layer ot least partially Cr increased, Fe decreased, voids up te 0.018 in, and averaging 0.008 in. nonmetallic Ni and U unchanged in fluorides; Cr leached from wall 13 vl UGL ABILF ET T Mani e ) . o : .'- I i 3 & ; . & T i a I: & g A'r el '.' & " . . " . - STANDARD BEFORE CIRCULATION o " -t = . A'I - " : o - B 1 - "‘".. .“ - b , ~ . b : ol . x ..- " LOWER VERTICAL HOT LEG SECTION 4 Fig. 7. Variation in Attack Around a Standard Loop. 250X. » e W E§ wenk i (To=EFEEPERRER ~wENE uatssuren || P _..‘{.. -.“fl' A L - . ' ..- ' ‘ -" 1* 8 TOP OF VERTICAL HOT LEG SECTION { L aasriiD | | T e HOT HORIZONTAL LEG SECTION 7 MIDDLE OF VERTICAL HOT LEG SECTION 3 » - ' " LOWER COLD LEG SECTION 8 Reduced 61%.-wirh caption) iR FEEFENARD R FEEFBLERD ~oiiE] EE i i1 — Table 4. Analysis of Fluorides from Loop 229 Section U (%) F (%) Fe (ppm) Cr (ppm) Ni (ppm) 1 4.18 43.7 40 2095 <20 2 4.16 42.3 140 2170 <30 7 4,18 42,2 130 2900 <20 9 4.13 43.0 65 3635 <20 10 4.20 42,2 210 3500 <30 14 4.19 42.1 200 3100 <30 Original botch 4,31 42.4 770 660 20 Table 5 lists the corrosive effects found in The presence of these layers showed that even Inconel loops after they had circulated various fluoride mixtures (see Table 1 for compositions). The variation in attack is shown by the typical hot-leg sections presented in Figs. 8 and 9. These loops were operated not necessarily to give a minimum amount of corrosion but to be as nearly comparable as possible. In the case of fluoride 30, it was shown that lower corrosion rates were possible with changes in the production procedure, and it seemed likely that similer reductions could be made with the other compositions. From the data in Table 5 it is apparent that the maximum depth of attack did not vary greatly with the different fluoride mixtures. An average maxi- mum penetration of about 0.010 in. was typical. The variation in the intensity of attack and in the size of the voids in the metal resulting from cor- rosion found with different mixtures was thought to be caused mainly by differences in batch purity. The mechanism of attack appeared to be the same for all these fluoride mixtures and is discussed in detail in the section ‘‘Mechanism of Corrosive Attack.” This single mechanism was evidenced by both the similarity in the voids and the similor variations in chemical analysis. As listed in Table 5, thin altered layers were often present on the cold-leg surfaces. Efforts by x-ray diffraction and spectroscopy to identify these layers failed. |t seemed that various metals had been deposited on the cold-leg surface but that their identities were lost through diffusion; the resulting changes in the base metal were not large enough to be picked up by the analytical or diffraction techniques used. Any nonmetallic particles found were usually identified as UO.,,. with Inconel some mass transfer had taken place. Effect of Uranium Fluoride. ~ Although variations in depth of attack were fairly small, a close study of the results showed one systematic trend: in all cases the ottack was both deeper and more in- tense in the loops which circulated uranium-bearing fluoride mixtures than in loops which circulated similar mixtures without uranium fluoride. The data are summarized in Table 6, and photomicro- graphs of hot-leg sections from some of the loops are given in Figs. 8 and 9. Effect of Reducing Agents. — In static tests, the alkali metals and zirconium were found® to inhibit corrasion by fluoride 14. A program was then set up to check several reducing agents under the dynamic conditions found in the thermal- convection loops. Zirconium hydride was substituted for zirconium metal in the loops because it is more stable at room, or slightly elevated, temperatures and does not absorb as many gases. |t decomposes just be- low the loop operating temperature to give very fine particles of pure zirconium metal and nascent hydrogen. zirconium metal concentration of 0.5% was added to the expansion pot of a loop, and the fluorides were charged in over the zirconium hydride. The zirconium hydride addition to fluoride 14 almost completely eliminated the corrosion of The first photomicrograph in Fig. 10 shows the hot leg of loop 225. Some surface rough- ness was found, but none of the usual subsurface voids were present. An intermittent metallic- appearing layer was found on the surface of the Enough zirconium hydride for a final Incenel. 15 91 Table 5. Corrosive Effect of Yarious Fluoride Mixtures on Inconel Loop Fluoride Metallegraphic Examination Chemical Analysis Comments No. No. Hot Leg Cold Leg of Fluorides 78 12 Moderote subsurfoce voids to 0.013 in. No attack; nonmetallic layer 0.0002 in. Cr increased slightly; Fe Run for 1000 hr ond Ni low and constant 217 17 Mederate subsurface voids to 0.013 in.; Heavy metallic-appearing layer, up to Cr high originally but still mainly concentrated in grain 0.001 in. increased; lorge decrease boundaries in Fe; smoll decrease in U 221 21 Maderate to light subsurface voids; Thin nonmetallic layer Some Cr increase; Ni and U maximum penefration, 0.007 in. decreased 226 26 Widely scotterad voids, almost en- Deposit with both metallic and nen- U and Ni decreased slightly; Hydregen-cleaned; tirely in grain boundaries; maximum metallic phases 0.001 in. thick Cr increased; Fe high, fuel from lob penetration, 0.010 in. dropped to only half 227 14 Moderate amount of small subsurface Thin metallic layer Cr increased; Fe decreased; voids from 0.006 to 0.016 in. Mi and U constant 230 24 Light, widely scattered voids usually Ne deposit Cr increased slightly; Fe in grain boundaries; maximum pene- showed large decrease tration, 0.009 in. 264 27 Moderate subsurface voids; penatra- No deposit Large Cr increase; Ni and Fe tion, 0.005 te 0.009 in. decreased; U increosed 246 32 Light, scattered subsurface voids; No deposit Cr increased; Ni ond Fe mainly in grain boundaries; 0.003 te decreased 0.008 in. 262 35 Moderate subsurfoce voids; maximum Scattered nonmetallic deposit Large Cr increase; Ni high penetration, 0.009 in.; general eriginally but now low; Fe attack, 0.004 in. decreased slightly 277 3 Light to moderate subsurfoce voids Thin metallic-oppearing layer Cr increased slightly; Ni and 0.0015 to 0.005 in.; mainly in grain Fe decreased boundaries 283 30 Mederate to heovy subsurface voids to Layer of metallic-appearing crystal Large increase in Cr; Ni and As melted 0.015 in.; deepest attack inter- Fe content high and de- granular creased 282 30 Moderate subsurface voids to 0.009in.; MNo deposit Ni low and Fe moderate; Cr Prepared in grophite general attack te 0.006 in. increased; Fe decreased and hydrogen-treated {1 Fig. URCLASSIFIED T 409 FLUCRIDE 42 FLUORIDE 7 8. Cotrosion of Inconel by Various Fluoride Mixtures. 250X, FLUORIDE 14 FLUORICE 35 Reduced 43.5%._1« ith caption) UNCLASSIFIED 8l UMCLASSIFIED T 1334 e o UNCL .F\:'JIFILLT‘ T 1233 FLUCRIDE 32 FLUORIDE 24 Fig. 9. Corrosion of Inconel by Various Zirconium Fluoride —-Base Fluoride Mixtures, 250X, Reduced 40%. -with caption) Table 6, Corrosive Effect of Uranium Flueride Addition Loops Circulating Fluoride Mixtures Containing Uranium Loops Circulating Fluoride Mixtures Containing Ne Uranium luerid i Loop Fluoride ot-Lag Aask Loop Fluoride HotoLag Abtask No. No. Ne. No. 227 14 Moderate subsurface voids; 78* 12 Meoderate; maximum, 0.013 in. maximum, 0.016 in. 217 17 Mederate, in grain boundaries; 262 35 Moderate; maximum, 0.009 in, maximum, 0.013 in. 226 26 Widely scattered in grain 230 24 Light, widely scattered in boundaries; maximum, grain boundaries; maximum, 0.010 in. 0.009 in. 264 27 Mederate; moximum, 0.009 in, 246 32 Light and scattered in grain boundaries; maximum, 0.008 in. 282 30 Moderate subsurface voids; 277 31 Light to moederate; moximum, maximum, 0.009 in. 0.004 in, *QOperated for 1000 hr. hot leg and in some of the grain boundaries near the surface. The second photomicrograph in Fig. 10 shows the cold leg of a duplicate loop (242). A well-bonded layer on the surface filled up what fabrication cracks were present. Unsuccessful attempts were made by both x-ray diffraction and spectrographic means to identify these layers. When the fluorides from these loops were analyzed, it was found that iron, nickel, and chromium had decreased in concentration during the circulation. The zirconium inhibited the corrosion by reduc- ing the impurities in the fuel before they reached the chromium. It was thought that the surface layers were products of the reduced impurities and excess zirconium metal and that they could be eliminated if the fluorides were treated with zir- conium hydride in the transfer pot and filtered into the loop. The fluoride 14 itself remained essentially unchanged when low concentrations of zirconium hydride were added. The fuel from these loops was examined petrographically by Hoffman, & of the Materials Chemistry Division, but she failed Bp. & Hoffman, Examination of Thermal Convection Loops with Added NaK or Zirconium Hydride, memo- randum to W. R. Grimes, Oct. 10, 1952. to find any reduced fuel compunents.? When similar additions were made to fluoride 27, some reduction of the fuel did take place. Titanium hydride was similarly odded to loop 250. The addition of 0.5% titanium hydride de- creased the corrosion to a maximum penetration of 0.0025 in. but produced a thin metallic layer, identified by R. M. Steele by x-ray diffraction as NijTi. Under a high magnification (750X) it was revealed that the layer was actually made up of several phases imposed one upon another, as can be seen in Fig. 11, a photomicrograph of the hot leg of loop 250. A similar layer was also found in the cold leg. On the basis of static tests, the alkali metals also appeared to be effective corrosion inhibitors, Therefore 0.5% of sodium-potassium alloy was added to loop 224, in which fluoride 14 was circu- lated. The maximum depth of attack measured 0.010 in., and although a reduction in intensity of Mote recent work by improved techniques has shown that the zirconium hydride works primarily by the forma- tion of reduced uronium compounds, which change the equilibrium conditions in the corrosion reactions. Since metallic uranium will be formed, which alloys with the container woll, the addition of zirconium hydride is not a practical method for retardation of corrosion. 19 | UNMCLASSIFIED T 1764 g |n E o ] 1 INC‘;‘ES T 1 s 4 ' - e oor HOT LEG UNCLASSIFIED | | T 12333 1. | 26 = k. | ' . COLD LEG Fig. 10. Effect of Zirconium Hydride Additions on Fluoride Corrosion in Inconel. 250X. Reduced 25%. [SUoEr with caption) 20 UNCLASSIFIED | T-2488 . Fig. 11, Hot Leg of Inconel Loop After Circulating Molten Fluorides with Small Addition of Titanium Hydride. 750X, with caption) attack was noted, it was less than had been ex- pected. It is possible that a more efficient mixing of the sodium-potassium alloy and the fuel would result in a greater improvement, It is also possible that some of the sodium-potassium alloy was vaporized while the fuel was being charged. A similar addition, but more carefully controlled, definitely caused a reduction in depth of attack with fluoride 30. When the fluorides from the loop to which the sodium-potassium alloy had been added were analyzed, the maximum chromium con- centration found was 520 ppm, which was additional evidence of decreased corrosion. No fuel reduc- tion products® could be identified by spectro- graphic studies. Since corrosion occurs by the leaching of chro- mium from the alloy, it was reasoned that adding chromium metal to the fuel before it was placed in the loop might reduce the attack in one of two ways: the attack might be slowed down in accord- ance with the law of mass action, or the corrosive impurities in the fuel might be reduced before the fuel was charged into the loop. Several attempts were made to add chromium metal to the fluorides both before and after they were charged into the loop. It proved surprisingly difficult to add chromium metal to these mixtures, and the results obtained were erratic and not re- producible. Oxide Removal Procedures. — In any corrosion test the method of cleaning the metal surface prior to testing is one of the most sensitive vari- ables. It is desirable to clean the surface without changing the basic corrosion mechanism or corro- sion rate. Many of the early failures with the sodium loops were caused, or at least influenced, by the cleaning cycle.* The standard cleaning cycle chosen for use on the loops consisted in degreasing and inspecting; making all welds with a helium backup to reduce oxidation; degreasing; and inspecting with both swab and borescope. |t was realized that oxides, whether initially present or formed during welding, would not be removed by this procedure. The work discussed in this section was an investigation of several possible methods for removing these oxides without etching or otherwise changing the surface. None of these methods were adopted as a standard procedure, A series of loops was cleaned by passing dried hydrogen through them while they were heated to above 1800°F. The loops were held at this tem- perature for 1 hr or until the dew point of the exit hydrogen was —60°F, In most cases the dew point 21 was about —100°F at the end of 1 hr. The corro- sion results from loops cleaned by this methaod were quite erratic, and considerable grain growth usually occurred, The size of the loops, varia- tions in wall thickness, and variations in type and thickness of insulation made it extremely difficult to heat a loop uniformly; therefore temperature gradients of less than several hundred degrees were difficult to obtain. With these temperature gradients it was necessary to heat most sections of the loops into the range of rapid grain growth (above 1850°F) to get the colder sections to the temperature (1800°F). The resulting large grains caused variation in attack and made comparisons difficult. Because of the difficulties in control, this method was not considered as being adaptable either to the loops or to the large engineering components of a reactor, The results obtained from the operation of loops with sodium and sodium-potassium alloy, as well as cleaning tests carried out by the group under D. C. Vreeland, indicated that these metals would remove the oxides without attacking the tube wall.4'® Vreeland showed that, with small tem- perature drops, temperatures in excess of 1450°F were required for cleaning but that the time was not critical. Several loops were filled with sodium- potassium alloy, which was allowed to circulate for 4 hr with the hot leg at 1600°F and the cold leg at 1500°F. The sodium-potassium alloy was drained while the locap was still hot, The erratic variations in corrosion were traced to variations in the oxides left in the loop after cleaning, The oxides were left as deposits in cold sections of the loop and as a solid deposit when the loop was drained. The oxides from the weld scale had been converted to alkali-metal oxides with low solubilities. This cleaning method is feasible where a cold bypass or filter may be used to re- move the oxides, but not for the thermal loops. Effect of Hot-Leg Temperature. — Two loops were circulated with fluoride 14 at temperatures other than the standard temperature of 1500°F, Loop 218 was circulated at a hot-fluid temperature of 1300°F and loop 222 at a hot-fluid temperature of 1650°F, Photomicrographs of typical hot-leg sections from both loops and from standard loop cleaning mD. C. Vreeland, Experiments on Removing Welding Scale from Inconel, memorandum to W, D. Manly, Aug. 7, 1952. 22 229 are presented in Fig. 12. Some increase in depth of attack was noted with increasing tempera- ture: from a maximum of 0.010 in. at 1300°F to 0.015 in. at 1500°F and to 0.018 in. at 1650°F. At the higher temperature fewer voids were found, but they were much larger and showed more tendency to migrate to the grain boundaries. Con- siderable grain growth also was noted. Crevice Corrosion. — Although the accepted mechanism of corrosion in the fused fluorides would not be expected fo lead to crevice corrosion, a direct test was desired, inasmuch as crevices may be present in the hot portion of a reactor plumbing system where tubes join to headers. In order to maoke such a test, two crevices were built into loop 223. The pipe was sawed through in two places near the center of the hot leg, and the cut ends ot each place were secured by a sleeve welded over the ocutside of the pipe. Thus, at each end of the section which had been cut out, a crevice opening into the fluorides existed be- tween the sleeve and the outside of the pipe wall, The maximum attack found in the upper crevice was from 0.006 to 0.008 in., while that on. the in- side of the pipe wall, from the same area, was from 0.004 to 0.007 in. Although the increase in maximum penetration was small, the attack was much more intense inside the crevice. While the attack was not so severe in the lower crevice, the same relationships to that on the inside of the pipe wall were found. As an additional check on crevice corrosion, the top welds from standard loops 219 and 229 were examined. The welding techniques used had resulted in an incomplete weld penetration, and hence crevices were present on the inside of the loops. The increase in corrosion in such a crevice is apparent from Fig, 13, a photomicro- graph of the top weld of loop 219. The maximum penetration of the wall in this area was 0.003 in., while in the crevice the attack was 0.006 in. deep and several times as intense, Similar results were obtained with loop 229. These results indicated that crevice corrosion by the fused fluorides was not a serious problem. The most likely explanation for the small increase in corrosion found in the crevices is that the cleaning was inadequate. Some oxidation had taken place during welding, and the oxides were not removed during cleaning. Helium was used as a backup gas during welding, but purging of i~ LOOP 229, {500°F £C EEE] ] IM’."PIIES EERBE . EREEF | 2 UNCLASSIFIED T 12330 LOOP 222, 1650°F Fig. 12, Effect of Hot-Leg Temperature on Fluoride Corrosion in Inconel. 250X. Reduced 39%. ' with caption) E [asox | EEE] T |NCJHE5 FEEFPERERER, e air was not complete, The oxides in the crevice reacted with the fuel and caused the in- crease in attack in the crevice. In the loop itself, sufficient fuel and agitation were present to dilute any effect caused by the weld scale on the pipe surfaces, Mechanism of Corrosive Attack. — The theory that was developed during this work to explain the mechanism of corrosive attack by fluoride salts must be regarded as tentative, even though much evidence is available to support it. Considerable work has been carried out, principally with zir- conium fluoride—base fuels, to confirm the theory; the results will be presented in another report. To determine what changes took place in the pipe material, concentric samples were drilled from the inside of both the hot and cold legs of several loops. Chemical analyses of the drillings from loop 229 are presented in Table 7. These data show that chromium was leached from the hot-leg wall to @ depth beyond the maximum depth at which the attack was visible under the micro- In the cut taken between 0.025 and 0.030 in, deep, the chromium content was still not so high as in the external, or standard, sample. scope. 24 UNCLASSIFIED | T-1802 42 CH Fig. 13. Increase in Attack in a Crevice of an Inconel Loop. 250X, Figure 14 is a plot of the change in chromium con- centration with depth. With one exception all the points lie quite close to a straight line. From other evidence it appears that the solubility of chromium in the fuel in these temperature ranges is between 3000 and 3500 ppm. The chromium concentrations for loop 229 were in this range, but for most loops they were approximately 2000 ppm. Any chaonges in the iron and nickel content of the drilled hot-leg samples were small. The nickel percentage did change, but mainly because the chromium was depleted. Except in the first two samples, both the iron percentage and the iron- In other loops no variation was found even in the surface samples. In the surface samples from the cold leg the iron to-nickel ratio remained constant. content was higher, as was expected, since iron and nickel were deposited out of the solution. The depth to which the alkali metals were found in the drillings was surprising. While the maximum attack extended to 0,018 in., traces of alkali metals were found in the 0.020- to 0.025-in. cut. Enough sample was not available from these drillings for fluoride analysis, so the form in which Table 7. Analyses of Pipe Wall —~ Loop 229 Total Depth Others® Layer Fe Ni Cr Ratio Na® K2 Li® u? Total No. “::_‘;’ % (% (%) Fe/Ni (%) (%) %) (%) (%) (%) Hot Leg — Section 3 ] 0.003 6.73 81.7 8.2 0.083 0.15 0.5 0.3 c 1.34 98.9 2 0.006 6.83 80.9 9.4 0.085 0.15 0.3 0.15 c 1.19 98.9 3 0.010 6.99 797 110 0.088 0.08 0.2 0.1 c 1.23 99.4 4 0.015 6.98 78.1 13.1 0.089 0.06 0.1 0.08 c 1.40 100.0 5 0.020 6.85 77.8 142 0.088 0.04 0.08 0.06 c 1.42 100.4 6 0.025 6.91 76.8 154 0.090 Trace Trace Trace c 1.47 100.6 7 0.030 6.87 76.4 15.6 0.089 ¢ 1.66 100.5 8 External 6.90 76.5 16.0 0.090 c 1.36 100.8 Cold Leg ~ Section 11 1 0.003 8.46 73.8 166 0.114 ¢ c c c 1.45 100.3 2 0.006 8.5 746 165 0.110 c c c c 1.64 100.9 3 0.010 7.82 744 167 0,105 c c c c 1.75 100,7 4 0.015 763 745 16.6 0,102 c c c c 1.79 100.5 5 0.020 7.72 744 16,5 0.104 c c c c 1.64 100.3 6 External 7.62 746 167 0.102 c c c c 1.65 100.6 A5pectrographic determination. b1otal of Ce, Cu, 5i, Al, Mg, Mn, and Ti (spectrographic analyses). Not present. UNCLASSIFIED ORML-LR-0WG 35693 w 20 3 [ & Ly iy 1S - - o = o = ol —~ o e | O 2EO | u—e—see L OOP 273 D 5 — LOOP 229 | 3 £, | o 10 20 a0 40 DEPTH FROM SURFACE (mils) Fig. 14. Change in Chromium Cencentration in Wall of an Inconel Thermal-Convection Loop. the metals are present is not known. The ratio of the alkali metals, as shown by these analyses, varied greatly both from sample to sample and from that in the fuel (Na, 1; K, 10; Li, 4). Further- more, the total amount of the alkali metals, on either a weight or a molecular basis, was much less than the amount of chromium removed. It is possible that the presence of alkali metals in the deeper samples resulted from poor sampling, either through nonconcentric drilling or from runout. During metallographic examination of the hot-leg sections, an area was noted near the surface that etched differently from the remainder of the sample. In this area carbides were no longer visible. This layer extended slightly deeper than the visible 25 attack and followed the line of the attack. A hot- leg somple from a standard loop was submitted to the International Nickel Company for study of this layer, It was found that apparently the carbon had not been leached out with the chromium but had gone back into solution in the metal, The carbon content of the surface layer and that of the re- mainder of the pipe were the same, !’ While examining some sections from a loop, W. C. Tunnell noted that the inner pipe surface had become magnetic. Some chemical samples drilled at about this same time were then tested, and it was found that the magnetism decreased with depth but that a few magnetic particles were still present in the last cut. The surface layer in the cold leg was also magnetic. The presence of these magnetic loyers was also confirmed metallographically; when the sample was covered with a colloidal dispersion of iron and made the core of an electromagnet, the iron collected on the magnetic portions of the alloy. Figure 15 shows the magnetic areas of the hot and cold legs of loop 229. The heavy black line is the area where the magnetism begins; it is heavy be- cause the particles from the unaffected surface moved to this areq, The presence of magnetic layers on the hot-leg surface was additional confirmation that chromium had been leached out., When chromium is added to nickel, the Curie point is lowered quite markedly, An alloy with 8% chromium has a Curie point at about rocom temperature, while that of pure nickel is at 665°F.'2 The addition of iron to nickel raises the Curie point, and presumably it would do the same in these alloys. The Curie temperature for commercial lnconel is approximotely —40°F, Since the Curie point of the specimens which had been exposed to fluorides was above room temperature, the change must have been caused by the removal of chromium, In the examination of hot-leg sections one strik- ing fact is opparent. In spite of the depth to which the attack extends, there is wvery little elongation of the voids in this direction; all voids are nearly spherical. With normal intergranular attack, the wvoids are shaped like worm holes, ”E. N. Skinner, International Nickel Co., Inc., per- sonal communication to W. D. Maenly, 12M£*ra!s Handbook, 1948 ed., p 1046, American Society, for Metals, Clevelond. 26 with definite elongation in one direction. In an effort to follow the course of separate holes, a sample was examined metallographically, reground, and then re-examined at depths increasing by 0.00025 to 0.0005 in. From this study it was apparent that at least the majority of the holes did not connect to the surface or to each other. This conclusion was substantiated by a vacuum leak test on a section from which the outer, unaffected area had been removed by machining. Tests with dye penetrants produced the same results. In view of the above facts, W. D. Manly proposed a mechanism to explain the formation of the holes. It had been established that chromium was selec- tively leached from the alloys by the fluorides and thot the chromium at the surface of the pipe was reploced by a diffusion process. In a dif- fusion process that is unidirectional, there will be a change in the density ond/or shape of that portion of the specimen in which the composition is changed. Since diffusion in metals is believed fo occur by migration of vacancies from the surface, it may produce a concentration of vacancies in the lattice that exceeds the ‘‘solubility limit"" for vacancies at the test temperoture, If this limit is exceeded, the vacancies will ‘‘precipitate’’ from the lattice, collect at discontinuities, and grow to visible voids whose size will be a function of time and temperature, Such formation of voids has been observed in many bimetallic diffusion couples, and it has been demonstrated that identi- cal effects can be obtained in metal-liquid and metal-gas systems in which similar diffusion phenomena occur,” The removal of the chromium from the Inconel surface takes place by several chemical reactions. Since these are primarily chemical problems and are still under study by W, R, Grimes and F, F, Blankenship in the Materials Chemistry Division, they will be discussed only very briefly, From the dota developed, the reactions listed below appear to be the most probable ones for removing the chromium: (1) 3Ni ™ + 2C0 =22Cr**" 4 3Ni0 (2) 3Fe*? + 2CI9==2Cr"** 4 3Fe0 (3) 2HF + NiF==NiF, + H, (4) 3UF, + Cr==CF, + 3UF, The metallic iron and nickel formed in these re- actions are deposited on the cold-leg surface and UNCLASSIFIED T 3151 EEE T IHB?ES T 1 EEFEERER 2 > 50X | | 2 b = HOT LEG h j UNCLASSIFIED T 12332 : B B : T INCHES 1 o 2a COLD LEG Fig. 15. Magnetic Areas on Sections from the Hot and Cold Legs of an Inconel Loop. 250X. Reduced 19%. th caption) 27 diffuse into the base metal. Reaction 4 becomes more important with loops operated for longer times than those in this study, since it is thought to be the principal reaction for the mass transfer of chromium. It does, however, account for some of the attack, since equilibrium is established in a relatively short time ond requires the addition to the fluorides of between 1000 and 2500 ppm of chromium, depending upon the temperature, Corrosion of Type 316 Stainless Steel It was pointed out in the section on "‘Screening Tests” that all tests in which fluoride 14 was cir- culated in a stainless steel loop were terminated because the loop became plugged. Since stainless steels are such well-known engineering materials, a few tests were carried out to see whether the corrosion could be prevented by slight modifi- cations in operating conditions or fluoride compo- sition, Effect of Temperature. — A series of type 316 stainless steel loops were operated at different temperatures to ascertain whether plugging could be prevented by a slight increase in the cold-leg temperature. The data from these loops (presented in Table 8) showed that the plugging was tempera- ture sensitive. Loop 123, which was operated with a minimum cold-leg temperature of 1500°F, was the only one to be operated for 500 hr, A similar loop, with the same hot-leg temperature but with a slightly lower cold-leg temperature, became plugged in 91 hr. The longer time for the loop with a hot-leg temperature of 1300°F to be- come plugged was probably caused by a lower corrosion rate, In the loop that was operated for 500 hr a layer up to 0,001 in, thick was transferred to the cold- leg surface. This layer was identified by dif- fraction and spectrographic methods as being iron with some chromium in solution. Such a layer would not cause plugging but could interfere with proper heat transfer, Plug ldentification. — The nature of the plugs in the type 316 stainless steel loops is not definitely known. Although the standard temperature used for melting out the fluorides for examination was about 100°F below the minimum cold-leg tempera- ture, all the material in the loop flowed out. Loops were also sectioned longitudinally without dis- closing an apparent plug. D. C. Hoffman, of the Materials Chemistry Division, made a petrographic 28 examination of several of these loops before the fluorides were melted out and found'® that the only substance likely to form o plug was K ,NoCrF , which has a physical appearance very similar to the fuel and would be difficult to detect visually. This compound melts at around 1850°F, but it would probably separate from fluoride 14 ot about 1500°F under the conditions existing in the loops. Crystals of this compound might form as dendrites, which could collect together in o tangled structure similar to a log jam. Without being o complete plug, such a mass could furnish enough resistance to stop circulation in the loop. This is known to be the mechanism operating with metallic plugs. The few crystals holding such a mass to the side of the loop could be broken during freezing and remelting, and thus might not have interfered when the fluorides were melted out for examination. Analysis of the fluorides melted from these loops showed that the chromium content was con- siderably higher in the cold leg, where the plugging presumably occurred. This result would be ex- pected if the plugs were formed by KzNuCrFfi. Effect of Fluoride Composition. — Many of the early large-scale engineering tests were carried out with type 316 stainless steel loops containing fluoride 12. Inconel parts for the loops were not available at that time, and the use of fluoride 12, which contains no uranium, required no accounta- bility. Later, several loops were operated with zirconium fluoride~base fuels in order to test the proposed theory of plugging, Data obtained from these loops are presented in Table 9, and repre- sentative hot-leg sections are shown in Fig. 16. The compositions of the fluoride mixtures are given in Table 1. Of the four mixtures included in Table 9, fluoride 14 was the only one that caused plugging. The proposed theories of plugging are substantiated by the fact that the zirconium-base fuels, which contained no potassium, did not form plugs. The corrosion rate in the two loops with fluoride 12 was probably too low for enough material to be furnished to form a plug. Figure 17 shows a section from the hot leg of the type 316 stainless steel loop in which NaK was added to fluoride 12. Comparison with the Bp, e Hoffman, ANP Quar, Prog. Rep.’Mnch 10, 1953, ORNL-1515, p 136. 6 Table 8. Effect of Temperature on Corrosion of Type 316 Stainless Steel by Fluoride 14 Loop Hot=Leg Temperature (°F) Initial Cold-Leg Temperature (°F) Time of Circulation (hr) Reason for Termination Metallographic Examination Hot Leg Cold Leg Chemical Analysis 121 120 125 123 117-a* 117