RNL-TM-3866 ot YR o ; w0 / | CORROSION AND MASS TRANSFER CHARACTERISTICS OF NaBF,—NaF (22-8 mole %) IN HASTELLOY N J. W. Koger THIS DOCUMENT CONFIRMED AS UNCLASSIFIED gl\yISION OF CLASSIFICATION DATE__'_1e/iclHz GISTRIBUTION OF THIS BOCUMEKT 15 UNLAETED RIDGE NATIONAL LABORATORY IO CARBIE R TS R T This report was prepared as an account of work sponsored by the United States Government. Neither the United States nor the United States Atomic Energy Commission, nor any of their employees, nor any of their contractors, subcontractors, or their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness or usefulness of any information, apparatus, product or process disclosed, or represents that its use would not infringe privately owned rights. A} ) ‘&) e} +y - ORNL-TM~-3866 Contract No. W-7405-eng-26 METALS AND CERAMICS DIVISION CORROSION AND MASS TRANSFER CHARACTERISTICS OF NaBF,—NaF (92-8 mole %) IN HASTELLOY N J. W. Koger - . NOTICE This report 'was prepared as an account of work "} sponsored by the United States Government. Neither the United States nor the United States Atomic Energy Commission, nor any of their employees, nor any of their contractors, subcontractors, or their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, com- pleteness or usefulness of any information, apparatus, .| product or process disclosed, or represents that its use would not infringe privately owned rights. - -~ OCTOBER 1972 NOTICE This document contains information of s preliminary nature and was prepared primarlw for internal use at the Oak Ridge National ; Laboratory. It is subject to revision or correction and therefore does not represent a final report, A OAK RIDGE NATIONAL LABORATORY Oak Ridge, Tennessee 37830 - operated by = UNION CARBIDE CORPORATION for the . U.S. ATOMIC ENERGY COMMISSION ™ R7CES BISTRIBUTION OF THIS DOCUMENT IS GNLIVHTED < iii CONTENTS Page : Abstfact Gt e e e s e e e e e e s e e e e e e e e e e e e e s 1 Introduction « . ¢ ¢ &+ « ¢ ¢ ¢ ¢ 4 v e v e e e e e e e e e e e 1 Experimental Procedure . « « « « « o o o o o o o o ¢ o o & o o o . 6 Equipment . . « ¢ o ¢ ¢ v v o s 0 e e e e e e e e e e e e e e 6 Salt Preparation . . . « ¢ ¢« ¢ ¢ ¢ o 0 ¢ o o o s s s s s s e s 8 AN2lySEeS8 « 4 ¢ o s o o s s s s s s % s e s e s s e e e s s e e 11 Operation and Resulfs Gt e e e s s s e s e ee e e e e e e 11 Thermal Convection Loops NCL-13, NCL-13A, and NCL-14 . . . . . 11 LoOp NCL=17 & & &« ¢ o o s s o o o o o s o s s o o s o s s o o 20 LOOP NCL=20 &+ v v v v v v o o v o v v o oo e w e e e e 2] FCL-1 Pump LOOD &+ « « « + o « o o o o o o o o o o o o o o » o+ o 28 RUNS 1 8Nd 2 « o v v o o v e e e e e e e e e e . 32 RUL 3 . v v o o o o o o o o o o o o o o s s e e e 0 e o 37 ‘ Installation and Testing of Cold Finger During Run 3 . .. 38 ) RUN 4 « 4 o o + o o o o o o o o s o s o s s s o oo o« 40 : PP . RUD 6 &« v ¢ o o o o o o o o o o s o o o s o o o o o o« b5 RUN 7 « o « o o o o o o o o o o s- o o o s o o s o o+ o o 48 Salt ChemiStTY « « &« o & « o o o o o o o o o o o o o o o o o s s o= 33 Purification . . « v « ¢ ¢ « o ¢ o o o s o o o o o s o o = s s I3 Analytical Chemistry . . . . « « ¢« ¢« ¢ ¢ o ¢ ¢ o o o . . .. 54 Discussibn.. « e ..;‘. T . 7 Theory Q e e e e 0 e C e e e e e e e e ... « « o« o 54 Equations . . . . . . s e e e e e s e e e e e e e e 61 Kinetics . . e v e e e C et e s e s e s e s e e 62 ‘Solid-State Diffusion Control . . . . . . . .‘.‘. .« +« + - 066 Solution Controlling . .';'. T Y Experimental Data . . . . . Gt e e e e s e s s e v s e . . 68 - ; Diffusion Calculations . « « o + o o o « o o v o s oo o oo 10 SUmmAYY « o o« o o ; P e e e e e e e e e n e e e s .. 76 COI‘ICluSiOHS s 8 ® s o & 7 e s e s e s s s s 2 s .o . s 8 e | v e s » 78 g ACkDOW].edgment -® 8 8 & 8 & & ¥ 9 8 B2 8 & " » & 4 T v+ s e e = e 2 79 wi) 1) «)) wi CORROSION AND MASS TRANSFER CHARACTERISTICS OF NaBF,—NaF (92-8 mole %) IN HASTELLOY N | J. W. Koger ABSTRACT A series of corrosion experiments in thermal convection and pump loops designed to test the compatibility of fused NaBF,—8 mole % NaF with Hastelloy N has shown the extreme effect of impurities on the system mass transfer.” The diffi- " culty in keeping the melt sufficiently pure was also illus- trated. Kinetic considerations showed that the mass transfer process is controlled either by solid-state diffusion or by the solution rate, depending on the amount of impurities allowed to enter the salt. The mass transfer behavior is very similar to that for the Cr-UF,4 corrosion reaction (material removal in the hot section and deposition in the cold section), the differences being that alloy constituents other than chromium participate in the process and that the reaction rates are different. Because of these differences, the effective diffusion rate controlling the hot leg attack is larger than that obtained in typical diffusion experi- " ments for chromium in Hastelloy N. Titanium-modified Hastelloy N with less chromium and iron shows a greater resistance to attack in a fluoroborate salt with 500 ppm oxide impurity than does standard Hastelloy N. Its corrosion rate is doubled by increasing the oxide concentration to 1500 ppm. Average corrosion rates were low for all systems tested. -~ : ~ INTRODUCTION. ‘The successful development and operation of the molten salt reactor ‘experiment (MSRE) have led to the present program for_ development of a molten salt breeder reactor (MSBR) In the MSRE the heat transferred from the fuel salt to the coolant salt was. rejected to an air-cooled radiator. In the MSER a secondary‘coolant ‘will be required to remove heat fromltne fuel in the primary heat exchanger and transport this - heat to supercritical steam at quite low temperatures. This secondary coolant should have low viscosity and density and high heat capacity and thermal conductivity to permit use of écceptable heat exchangers, coolant pumps, and steam generators. The melting point must be low enough to meet the heat transfer temperature requirements, and the vapor pressure must be low at the temperature of‘operation. The salt must also be easily remelted, with‘no precipitation of high-melting compounds during cboling. The coolant must be commercially available in high purity, and the price should not be prohibitive. In nuclear systems the coolant must be stable under the radiation that it encounters. We _'also must consider the result of accidental mixing of the coolant fluid and steam and select a codlant in which the effects of this mixing will _bé minimiied. Last but not least,-the coolant must be compétible,fiith the container materials .of the system. _ | _ It now appears that the best choice for the MSBR secondary coolant - is the eutectic mixture of 8 mole % NaF in NaBF4, with a meltihg point! of 385°C (725°F) as shown in Fig. 1. The salt is quite inexpensive lc. J. Barton, L. V. Gilpatrick, H. Insley, and T. N. McVay, MSR Program Semiann. Progr. Rept. Feb. 29, 1968, ORNL-4254, p. 166. 1000 \ 900 ~ ORNL-DWG 67-9423A 700 —— N ‘ 600 - \ 500 _ R TEMPERATURE (°C) 400 - — — 300 200 —1—L — NaF 20 40 60 80 Nt'JBF4 ' NaBF, (mole %) : Fig. 1. The System NaF-NaBF,, . 7] C e wl (< $0.50/1b). ‘At elevated temperatures the fluoroborates show an appreciable equilibrium pressure of gaseous BF3; however, at. the proposed maximum temperature of the MSBR secondary coolant (621 C) the 2 pressure is only 252 torr. ‘The equilibrium pressure above a melt of NaBF,—8 mole 7 NaF is given as a function of temperature by- = 0 _ . ' o o log Ptorr'-'9°024 5920/T( K) . Table 1 contains some of the pertinent physical properties for this mixture at temperatures of interest.,‘Other possibilities'for secondary coolants, sdEh'as'fluorides,_chlorides and liquid metais,'are discussed 3 l+ 5 elsewhere.® Before our compatibility tests with the‘sodium fluoro- borate‘mixture, little was known or reported about its corrosive behavior in the molten state. _ Because of its appreciable vapor.gressure at temperatures of interest, BF 3 oorrosion is of importance'also. In corrosion experiments6 with gaseous BFj3, it was rapidly attacked by traces of moistnre to give hydroxyfluoboricsacid (HBF 30H) and HF. Also BF3; and glass reacted at an' appreciable rate just above 200°C. 'Underrthe conditions of those experiments, BF3; did not appreciably attack a wide variety of metals or alloys examined at temperatures up to 200°C. 2S Cantor, J. W. Cooke, A S. Dworkin, G D. Robblns, R. E. Thoma, and G. M. Watson, Physical Properttes of'MbZten Salt Reactor Fuel, 0RNL-TM—2316 (August 1968). ol : , ; W. R. Grimes, "Molten-Salt Reactor Chemistry," Nch Appl Pechnol. 8: 142 (1970) . “J. W. Koger and A. P. Litman, Compattbtlzty of Hastelloy N and Croloy 9M with NaBFy-NaF-KBF, (90-4-6 mole %) Fluoroborate S&Zt ORNL-TH-2490 (April 1969). ‘53, W. Koger and A. P. Litman, Compatzbzltty of Fused Sodtum Fluoroborates and BF3 Gas with HasteZZoy N Alloys, ORNL-TM-2978 - "~ (June 1970). ®F. Hudswell, J. S. Nairn, and K. L. Wilkinson, "Corrosion Experi- ments with Gaseous Boron Trifluoride,” J. Appl. Chem. 1: 33336 (1951). Table 1. Some Pfoperties,of theAMixtfire NaBF,—8 moie % NaF Approximate_meiting_point, °C . Vapor pressure at 621°C, torr Density,a g/cm3: at t°C ‘at 621°C at 538°C at 455°C Viscosifiy,b centipoise; at T°K at 621°C at 538°C at 455°C Heat Capacj.ty:c ., = 0.360 cal g~! °c"! ‘Thermal Conductivity:d e O & bl 384 - 252 p = 2.252 — 7.11 x 1071 1.82 1.87 1.93 0.0877 exp(2240/T) at 621°C:K = 0.0039 W cm~! °c~! at 538°C:K = 0.0041 W cm™}! °c~!? at 455°C:K = 0.0043 W em™} °C™! Latent Heat of Fusion = 31 cal/g #s. Cantor, MSR Program Semtann. Progr. Rept. Aug. 31, 1969, ORNL-4449, p. 14. bS. Cantor, MSR Program Semiann. Progr Rept. Aug 31, A. S. Dworkin, MSR Program Semiann. Progr. Rept. Feb 29 1968, ORNL-4254, p. 168. d ORNL-4449, . 92. J. W. Cooke, MSR Program Semiann. Progr. Rept. Aug 31, 1969, 1969, 1) € ‘-“‘) a) u) w) All metal components of the MSRE in contact with molten salt were made of Hastelloy’ N (formerly called INOR-8). Two decades of corrosion. testingt’_15 and e)cperiencew""19 with-the MSRE have demonstrated the excellent compatibility of‘Hastelloy N and graphite with fluoride salts containing LiF, BeF,, UF,, and Tth. Hastelloy N, perhaps with some modification of composition, is quite likely to be the primary contain- ment material for MSBR. Thus, it was of great interest to the molten salt program to determine ‘the compatibility of the fluoroborate salt mixture with Hastelloy N and related alloys. Of Special interest is Vtemperature-gradient mass transfer, which must always be considered " where corrosion in a heat exchanger is possible.l Corrosion and deposi— tion processes ‘in flowing nonisothermal systems are interdependent and :each exerts considerable influence over the extent ‘and characteristics of the other. Therefore,'a performance analysis of a nonisothermal system must consider these processes as complementary and equal in sig— nificance to the overall system behavior. This paper is an up-to-date, open-ended report on these studies. Hastelloy N is the trade name of Cabot Corporation for a nickel— base alloy containing 16% Mo, 7% Cr, 5% Fe, and 0.057% C L. S. Richardson, D. C. Vreeland, and W. D. Manly, Cbrrosron by Molten Fluorides: Interim Report for prtember 1962, ORNL -1491 (April 20 1953) , G. M Adamson, R. S. Crouse, and W D ‘Manly, Interzm Report on Corrosion by Alkali-Metal Fluorides: Wbrk_to Muy 1, _1953 0RNL-2337 (March 20 1959). . _ G. M. Adamson, R. S. Crouse, and W. -D. Manly, Interzm Report on Corroszon by Ztrcontum-Base Fluorzdes 0RNL-2338 (Jan. 3, 1961) . 'y, B. Cottrell, T. E. Crabtree, A. L. David, and W. G. Piper, - Disassembly and Postoperative Examination of the Atrcraft Reactor Emperzment 0RNL—1868 (April 2, 1958) . _ 124, . Manly, G. M Adamson, Jr., J. H. Coobs, J. H. DeVan, D. A. Douglas, E. E. Hoffman, and P. Patriarca, Aireraft Reactor EwpertmenteMEtaZZurgtcaZ Aspects, ORNL-2349, pp. 2-24 (Dec. .20, 1957). 134. b. Manly, J. H. Coobs, J H. DeVan, D. A. Douglas, H. Inouye, P. Patriarca, T. K. Roche, and J. L. Scott, '"Metallurgical Problems in Molten Fluoride Systems, Progr. Nucl Energy Ser., IV 2: 164-79 (1960). EXPERIMENTAL PROCEDURE Equipment In corrosion‘studies the thermal convection loop represents an intermediate stage of sophistication and complexity between simple capsule tests and a fulléscale engineering pump loop experiment. It is particularly suited to small-scale tests that involve temperature_ gradient mass transfer. The flow of the liquid is caused by its varia- tion in density with temperature. The development of a modified thermal convection loop has permitted important strides in ohtaining basic'cOrro— sion information; The thermal convection loop used in this work and shown in Fig. 2 permits unrestricted access to specimens and salt at any time without significantly disturbing 1oop operation or introducing' air contamination. Access is provided by'twin ball valve arrangements atop both the hot and cold legs of the loops. The molten salt sampling r'device illustrated in Fig. 3 was used in our thermal loops and can also W. D. Manly, J. W. Allen, W. H. Cook, J. H. DeVan, D. A. Douglas, H. Inouye, D. H. Jansen, P. Patriarca, T. K. Roche, G. M. Slaughter, A. Taboada, and G. M. Tolson, Fluid Fuel Reactors, pp. 595604, James A. Lane, H. G. MacPherson and F. Maslan, eds., Addison Wesley, Reading, Pa., 1958, J. H. DeVan and R. B. Evans III, "Radiotracer Techniques in the Study of Corrosion by Molten Fluorides,' pp. 557—79 in Conférence on Corrosion of Reactor Materials, June 4-8, 1962, Proceedings Vol. II, Internat10na1 Atomic Energy Agency, Vienna, 1962, ®H. E. McCoy, 4n Evaluation of the Molten-Salt Reactor Emperzment HusteZZoy N Surveillance Specimens — First Group, ORNL-TM-1997 | ' (November 1967). "H. E. McCoy, An Evaluation of the Molten-Salt Reactor Emperzment Hastelloy N Surveillance Specimens —-Second Group, ORNL-TM-2359 (February 1969). 1%y, E. McCoy, An Evaluatton of the MbZten-SaZt Reactor Empertment Hastelloy N Surveillance Specimens — Third Group, 0RNL-TM—2647 (January 1970) § H. E. McCoy, An Evaluation of the Molten-Salt Reactor Empérzment' Hastelloy N Surveillance Speczmens — Fourth Group, ORNL~TM-3063 (March 1971). #} *) . w) ( ORNL-DWG 68-3987 STANDPIPE =y} s BALL VALVES !g! _ TWIN I - © 3 / | | 1N ' P coamsmere | |~ HEATERS 1 ‘ 30in. ' INSULATION CORROSION SPECIMENS- ——— SAMPLER , . FREEZE : N VALVES " OFLUSH o TANK - , DUMP AN ' TANK " Fig. 2. MSRP Natural Circulation Loop and Salt Sampler. ORNL-DWG 68-2796 . SWAGE LOCK FITTING BUCKET - v CONNECTION SAMPLE ' ' REMOVAL SECTION : O VACUUM " CONNECTION - PERMANENT 7 LOOP ASSEMBLY~ BALL VALVE LFB PUMP MAXIMUM LIQUID LEVEL MINIMUM LIQUID LEVEL Fig. 3. Molten Salt Sampling Device. be used in pumped 1oofis. Typical thermal convection loops in operation are shown in Fig. 4. A forced convection loop was also used to evaluate the fluoroborate salt mixture and will be described later. Salt Preparation The salt for these tests was processed by the Fluoride Processing Group of the Reactor Chemistry Division. Very pure (> 99.9%7) starting materials were evacuated to about 380 torr, heated to 150°C in a vessel -lined with nickel, and then held for about 15 hr under these qonditions. PHOTO 75125A ¢ FLUSH TANK A Fig. Thermal Convection"Ldops in Operation. 10 If the rise in pressure was not excessive (indicating no volatile impur- ities), the salt was heated to 500°C while still'Under vacuum and agitated - for a few hours with bubbling helium. It was then transferred_to the £fill vessel and from it forced into the loops with‘helium pressure. In the .case of the pumped loop the salt was transferred initially into a dump tank and then into the loop tubing.- The hot portion of each thermal convection loop was heated by sets of clamshell heaters, with the input power controlled by silicon controlled rectifiers (SCR units) and-the'temperature‘controlled by a Leeds and Northrup Speedomax H series 60 type CAT controller.. The loop tempera- 'tures were measured by Chromel-P vs Alumel thermocouples spot welded to the out31de of the tubing, covered first by quartz tape and then by stain— less steel shim stock. ‘Each loop was degreased with ethyl alcohol heated to 150°C under - vacuum to remove moisture, and leak checked before filllng with salt. All lines from the fill tamk to the loop that were exposed to the fluo- roborate salt were of the same material as the loop and were cleaned and tested in the same manner as the loop. All temporary line connections were made with stainless steel compression fittings. ‘ - Each loop was filled by heating it, the salt pot, and all connecting lines to at least 530°C and applying helium pressure to the salt pot to force the salt into the loop. Air was continuously blown on the freeze valves leading to the dump and flush tanks to provide a positive salt seal. Tubular electric heaters controlled by variable autotransformers _heated the cold—leg portions. Once the loop was filled the heaters were turned off, “and the proper temperature difference was obtained by remov1ng some insulation to. :expose portions of the cold leg to ambient air. The -first charge of salt was circulated under a small temperature difference (20°C) for>24_hr and dumped. This flush removed surface oxides and-other possible impurities. The loops were then refilled with new salt and put into operation. A helium cover gas_under_slight posi- tive pressure (about 5 psig) was maintained over the salt in the loops ~ during operation. . »n *) " ) ffand results ‘of all the tests are reviewed. 11 ~ Analyses o Severallmethodsican.be used;to;obtain;quantitative data froma temperature-gradient mass transfer’experiment relating to the kinetics of mass transfer, the . thermodynamics.of the process, or both. The most obvious and easiest to obtain are weight change measurements on specimens and liquid analyses. The weight changes allow the calculation of mass transfer rates in both legs, We may also analyze specimens and loop tubing with a microprobe to determine composition gradients, x-ray fluorescence to determine the surface composition, and spectrochemistry to determine overall composition. Standard metallographic examination is also helpful to determine the extent of attack or void formation in the hot leg or the amount of deposit in the cold leg. " OPERATION AND RESULTS . The results of tests of the fluoroborate salt mixture in two ther- mal convection loops and in a ‘series of capsule tests have already been" reported.?%s2} In addition, four other thermal convection‘loops and one pumped“loop‘have'been'operated. Thése latter tests are described below, -4 Thermal Convection Loops NCL-13, NCL—13A and NCL-14 Thermal convection loops NCL—13 rand NCL—14 (constructed of standard Hastelloy,N)_were started at the same time under the identical conditions of 605°C maximum_temperature and a}temperature~difference of 145°C. .. . | Theseioperatingrconditions,were,thosefipropoSed for the coolant circuit ~of the_MSRE.A,puring circulation each loop (0.75-in.-OD X 0.072-in.-wall tubing) contained about 2.7 kg of salt that contacted 1740 cm’® (270, in.?) 2°J W, Koger and A. P. Litman, Compatzbzlzty of Hastelloy N and iCroZoy M with NaBFu-NuF-KBFg (90-4 6 moZe %) Fluoroborate saZt ORNL-TM-2490 (April 1969). 213, W. Koger'and A. P.: Litman, Compatzbtltty of‘Fused Sodium Fluoroborates and BF3 Gas. with Hastelloy N Alloys, ORNL-TM-2978 (June 1970) ‘ 12 of surface and traveled 254 cm around the harp. Typical flow under the above conditions was 7 ft/min. Loop NCL-13 contained standard Hastelloy N - specimens, while loop NCL-14 had titanium-modified Hastelloy N suspended in the salt stream. The compositions of these alloys are given in - Table 2. The modified alloy is being‘considered because of its superior ‘mechanical properties.under radiation’'at elevated temperature. ‘Table 2. Composition of Hastelloy N Alloys Content, wt % Allo , ‘ T y Ni Mo Cr Fe Si Mn Ti Standard Hastelloy N L 70 17.2 7.4 4,5 .0,6 0.54' 0.02 Titanium-modified Hastelloy N 78 13.6 7.3 < 0.1 < 0.01 0.14 0.5 The weight changes measured for the,specimens‘in NCL-13 and NCL—14 showed an increase in mass transfer rate between 3500 and 4300 hr of ' egpcsute to the salt (Fig. 5). This was accompanied by perturbations in salt composition. Analyses of the circulating salts from these loops showed that the oxide content increased to above 2000 ppm (from initially less than 1000 ppm), and the nickel and molybdenum contents exceeded 100 ppm (from below 25 ppm). Also, the chromium and iron contents increased "nbrmally"-with time, as shown in Fig. 6. After 4700 hr of operation, the helium gas regulator that provided - the ovérpressurerto NCL-13 failed and caused a surge of gas to the loop, stopping the salt flow. Circulation of the salt could'notfbe resumed until a vacuum was pulled on the loop, which we believe removed a gas pocket. Shortly after circulation was restored, an electrical short, which eventually burned out a heater, occurred and heated the bottom of the hot 1eg to 870°C (1606°F). This disrupted the flow and caused a loss of BF3; from the loop, which changed the salt COmposition and plugged all.thelgas 1ines. The loop was drained of alllsqlt,'and plugged lines were replaced or unplugged. Other necessary repairs to the loop were made, and the loop was filled with new salt. The loop was then designafed as NCL-13A. | I P ~ ORNL-DWG 68-12031A 6 COLDEST SPECIMENS | - ) . - | o * ORNL-DWG 68-13M7 4 /)/A 460 °C W - a0 —— “owe_ A . o Fe | 2 ] A : ] '/ _ . ‘ ,/‘ T - 1 a _e—o—""" | : § 0G5, - T T | 1300 ~N \A\ T~—o o , : Cr o ; A e : 81 -2 < \'\ = g. z RN - - 4 ,. . — ' 200 , X : o Ty | : ‘ A\ | é - = -6 , - HOTTEST SPECIMENS £ | N 607°C \\ o //// -8 |— © TITANIUM MODIFIED HASTELLOY ; 100 N-NCL~-44 o A STANDARD HASTELLOY = \ —40 (— N-NCL-13 | : - 10 o -2 ‘ 0 ' 3 o 1 2 3 4 5 6 (x10%) o 2 - 3 4 5 6 (x103) - N “TIME (hr) ‘ | ' TIME (hr) Fig. 5. Weight ChangerVérSus Time for Standard Fig. 6. Average Concentrations of Iron and and Titanium-Modified Hastelloy N Specimens in Chromium in the Fluoroborate Salt in NCL-13 and . NCL-13 and NCL-14 Exposed to Fluoroborate Salt at = NCL-14. - Various Temperatures. €T 14 After new fluoroborate salt was added to loop NCL-13A, circulation could not be achieved. We coficluded that previous‘overheatiflggbf the loop had caused BF3 to evolve and had made the salt much richer in NaF. This in turn caused the higher melting NaF to segregate in the loop. To bring the composition back to normal, we added BF; gas to the melt through the surge tank. The initial amounts of gas were dissolved in the salt, and additions continued until a slight overpressure built up over the salt. (No devices were available to measure the amount of BF; gas added.) - Circulation of the salt was then attemptéd and was successful. Only a small temperature difference (55°C) was obtained at first. Further addi- tions of BF3 as before lowered the temperature of the salt in the cold leg and increased the temperaturefdifference to about 125°C. Since more ~BF3 did not cause immediate changes and did not appear to dissolve in the salt, the additions of gés were stopped. . The loop has since operated satisfactorily. Loops NCL-13A and NCL-14 have now operated for two and three years, fréspectively. Figure 7 gives the weight changes of specimens at various temperatures as functions of operating time. The overall corrosion rates at the maximum temperature, 605°C, are 0.5 and 0,7 mil/year for loops NCL-13A and NCL-14, respectively. The attack is general, and actual . measurements of specimefls show changes in thickness agreeihg with those _calculated from weight changes. We will now discuss the circumstances ‘that have affected the mass transfer conditions. The air leaks in loops NCL-13A and NCL-14 occurred at various mechan- ical joints in the system, such éé ball valves, level probes,'and pressure gages, all located in the cover gas spaée above the circulating salt. We had gaified extensive experience with similar mechanical joints before testing fluoroborate salt mixtures and had encountered_no_leakage problems. Therefore, the frequency of leaks in.theée two fluoroborate loops strongly suggests that the BFg atmosphere in these.tests'has déteriorated our mechanical seals. We are currently investigating the chemical nature of ‘the BF3 to determine ény posSible contaminants which might have caused such deterioration. Despite the known broblems'offiair leakage, these loops have opefated for over two years with the fluoroborate‘salt, and the overall corrosion rate is within acceptable limits. (\ N NY) ) -} WEIGHT CHANGE (mg/cm?) 15 ORNL-DWG 69-4763B 502°C ~522 532 547 557 572 0 2000 4000 ‘6000 8000 10,000 {2,000 14,000 16,000 TIME (hr) : ORNL-DWG 69-12625RA 20 . 465°C 10 00 470 . 480 OO I —- \ o—e-s 520 o € “v\‘\\mhi‘\\‘\ ' | L , . £ -10 N o ' No——0-0 540 W : = g I - . B e -2 —— g3 565 o \ ' w Lz =30 580 ~40 — _ \‘f-fi* 607 -50 : : ' | O 5000 - 0,000 , 15,000 © 20,000 25,000 30,000 TIME OF OPERATION {hr) 18,000 Fig. 7. Weight Change Versus Time for Standard Hastelloy N Specimens' in NCL-13A and Titanium-Modified Hastelloy N Specimens in NCL-14 Exposed to Fluoroborate Salt at Various Temperatures. 16 We also used loop NCL-14 to determine the corrosion properties. of pure Cr, Fe, Mo, and Nickel 280 (commercially pure nickel with Al,03 added as a grain refiner) at 600°C‘ifi”the‘fluoroborate mixture., As a consequence of the air leaks discussed above, the oxide éontent of the salt in loop NCL-14 was at an!abnormaliy high 1éve1 and indicated fairly strongly oxidizing condifions. We.initialiyrexamined the pure metal spécimens after a 234-hr exposure. On removal, we found thét'the chro- mium and iron specimens had completely'disintegfated. 'The molybdenum specimen showed no weight change, and the nickel specimen showed a slight weight gain. The cHromium'and iron specimens’were-replaced, and a second experiment was conducted for 23 hr. Again, the molybdenum specimen did not change weight, and the nickel specimen showed a slight weight gain. Very little of the chromium specimen remained, afid the weight loss of the iron specimen was 40 mg/cm2 (70_fiils/year). Close ‘examination of the chromium épecimen showed it to be coated with~é black reaction layer about 0.015 in. thick. Laser spectroscopy of the outer surface of the black material showed iron and'bqron as major conStituents and sodium and chromium as minor constituents. However, x-ray diffrac- tion of the bulk coating showed it to be over 90% NaBF,. This blgck coating has been observed on Hastelloy N specimens eprsed to impure NaBF4,—8 mole % NaF in the past'22~but in these cases it was too thin or too adherent to remove and identify. | ‘ Attack of Hastelloy N by impure fluoroborate salt tends to be non- Selectlve, that is, all components of Hastelloy N suffer oxidation, and surfaées of the alloy appear to recede uniformly. However, in this experiment with pure metals, only the chromium and iron were noticeably attacked, the chromium more than iron. We cbnclude, therefore, that the pure chromium and iron specimens in this system inhibited the attack of the nickel and molybdenum specimens., The inhibition apparently is much less when chromium and iron are present as alloying additions in Hastelloy N, presumably because their chemical activity in the alloy.is too low or the rate at which. they can enter the salt is too slow, being limited by solid-state,diffusion\in.the alloj. When the - 22J W. Koger and A. P. Litman, MSR Program Semfdnn. Progr. Rept. Feb. 29, 1968, ORNL-4254, p. 225. T » N ) 17 oxidation pbtential of the salt is reduced however,,as would be the case in a purer fluoroborate salt attack of Hastelloy N becomes more : selective, and the buffering effect of chromium may become more impor-- 'tant. In loop tests of nickel-molybdenum alloys containing Fe, Nb, or - W, DeVan observed23 that the. concentration of these elements in FLINAKZ“ salt was lowered by the presence of chromium in the alloy. These obser- - vations suggest that the incorporation of an active metal such as chro- ‘mium would effectively inhibit attack in a Hastelloy N—fluoroborate circuit. _ ‘ Figure 8 shows optical'and scanning electron micrographs of speci- mens from the hot and cold leg of NCL-13. These specimens were exposed to NaBF,—8 mole % NaF for 4180 hr at 580 and 476°C. The hot-leg specimen lost -3.3 mg/cm?, and the cold-leg specimen_gained-3.2 mg/em?. The sur- face of the hot-leg specimen, as seen in Fig. 8 (left) has been attacked nonuniformly and is heavily contoured. Figure 8'(right) shows that the ;,cold-leg specimen is coated with a discontinuous depoSit which is essen- 5tia11y analyzed as Hastelloy N slightly enriched in nickel and molybdenum. “The dumped salt from NCL-13 was chemically analyzed and the results are given in Table 3. The Cr, Fe, Ni Mo, O, and H»0 contents of the ‘salt all showed large increases., Increases in the nickel and molybdenum _content of the salt often indicate the onset of stronger oxidizing con- ditions, usually due torincreased water or oxygen content. Thus.the most obvious explanation of the-composition changes'and'increased weight changes was contamination by water, which produced HF, which-in turn ‘ attacked all the constituents of the COntainer material. This was sub- stantiated by the discovery of a faulty helium line common to both loops. This line admitted air. Loop NCL—14 initially operated about 3500 hr with the oxide content in the salt between 500 and 1000 ppm. During that period the weight , changes were small and- predictable as a function of time. Wet air then » 237, R, DeVan, Efféat of.AZZaycng Addttzons on Corrosion Behavtor - of Nickel-Molybdenum Alloys in, Fuaed FZuorzde thtures, 0RNL-TM—2021 Vol. I, pp. 3839 (May 1969) 2“Com.position' NaF-LiF-KF-UFu (11 2-45 3- 41 0—2 5 mole %) 18 Fig. 8. Hastelloy N Specimens from NCL-13 Exposed to NaBF,~8 mole % - | NaF for 4180 hr. Left: in hot leg at 580°C; weight loss 3.3 mg/cm?. . Right: in cold leg at 476°C; weight gain 3.2 mg/cm®. Top: optical micrographs at 500X, etched with glyceria regia. Bottom: scanning electron micrographs at 3000x (left) and 1000x. | o . " ) "N » ) 19 ‘Table 3. Chemical Analyses of Fluoroborate Salt in NCL-13 . Content, ppm | Content, wt % Sample €t ~Fe N Mo Q0 Na B F Salt before loop .. 19 223 28 <10 459 21.9 9.51 68.2 operation e : _ - . | Dumped salt circulated 348 650 95 125 3000 21.7 9.04 67.0 for 4700 hr ' ’ : e . camewin contact with the salt through the defective gas line, andrthe water content and corrosion rate increased dramatically. During the ~next 5000 hr the oxide content of the salt decreased, and the corrosion rate again decreased as a function of time. The maximum rate of weight loss:during this 5000-hr period was about 1.5 X 1072 mg cm™ 2 hr™! (0.65 mil/year). This corrosion rate is double that measured during the earlier operating period. , : | ' . The concentration offiimpurities in the salt in loop NCL-14 is given as a'fonction of.operating'time in Fig. 9. The.second.change_in corrosion rate, at some tine between 11,000 and 13,000 hr, was due to a.leaking standpipe ball valve. .Ihis last intake offimpnrities illustrates a con- tinuing problem of detection (other than the obvious after-the-fact increase of impurities and corrosion rate), since the loop overpressure did not change significantly. Apparently moisture from the air effuses into the relatively moisture-free athoSPhere,of'the'loop (equalization =~ of partialfpressure) without the need for a'gradient-in overall pressure. In this last corrosion rate increase period the analyzed oxide content of the salt increased from 1300 to 4500 ppm.l The_ball valve was repaired, and the loop continued to Operate.‘ As we have shown, abrupt increases in the slope of the weight change curves (increased rates of gain and loss) indicated changes in the oxida- tion potential of the salt, which were caused by inanvertent air leaks into the system.r These slope changes were accompanied by increases in :Cr,rNi Fe, and Mo contents in the salt, which would indicate general 'attack of the alloy.' Under conditions where.external impurities did not 20 ORNL-DWG 68—11768AR 3000 2000 1000 450 400 350 250 CONCENTRATION (ppm) 200 150 100 50 0 2000 4000 6000 - 8000 _"10,000 12,000 14,000 ' TIME (hr) ' ' - Fig. 9. Variation of Impurity Contents with Time in NaBFw—B mole 7 NaF Thermal Convection Loop NCL-14. enter the loops, chromium’ was the only element removed from the alloy, and corrosion rates during these periods fell to as low as 0.08 mil/year, - Loop NCL-17 Loop NCL-17, constructed of commercial Hasteiloy N with‘removable specimens in each leg, is being operated to determine the efiect of steam inleakage in a fluoroborate salt—Hastelloy N circuit The maximum tem- perature is 607°C with a temperature difference of 107°C. ThlS experiment has run for over 10 000 hr and is continuing in order to provide informa- tion on the immediate and long—range corrosion of the system after steam injection. The loop was operated for 1050 hr, the specimens were removed and welghed and steam was forced into the flowing salt system through a . 5 21 16-mil hole in a. closed 1/4—in.-Hastelloy N tnbe, simulating a leaking heat exchanger.‘ Steam was forced.into the system until the pressure of the cover gas began to increase, thus indicating that Nno more steam was soluble in the salt. Figure 10 shows thelweight changes and temperatures for the specimens in NCL-17. As usual, the changes depend on temperature, and the rates decrease mith time after an initialrapid rate. The corro- sion rate appears to have stabilized, and nd”plugging conditions have been noted. Figure 11 shows the specimens from the loop 424 hr after exposure to steam. The specimens with the etched‘appearance have lost weight, and those that'are darkened'have gained;meight. | For the last 3500 hr of operation, as shown in Table 4, ths concen- trations of impurity ‘constituents in the salt (Fe, Cr, Ni, Mo, and 0) remained relatively_constant. The present chromium level is abou: 320 ppm. The results of this test show that the Hastelloy N system ORNL-DWG 70-4936A 30 ‘ l L STEAM . 20 lNJECTION : ‘ - " ¢ 527-493°C : l .&."__-. .- : 10 - ‘ : ' ' fi’” ~ ~ | —tosaac| g o 11 -;O-—_‘l . - ' | ‘ on e —— E O e B & -10 t‘\\.- - IR | ~~Jeseoc| Z R T~ | Tesmec). 2—20 > -l \\ - 8. \\2:..\0 | \-"‘ . i 2 \.\ . \ | -30 . . ' e 582°C o ' o e ~ To593°C - ~40 7 ‘ \\\\\\ | T : Ye607°C - =50. o '4ooo 2ooo 3000 4000 5000 6000 7000 8000 9000 ' ‘ TIME OF opznmom (hr) Fig. 10 Weight Change Versus Time for Hastelloy N Specimens in NCL-17 Exposed to NaBFu—fi mole % NaF at Various Temperatures. Fig. 11. Hastelloy N Specimens from NCL-17 Exposed to NaBF,—8 mole % : , NaF. Specimens were exposed to salt for 1050 hr before steam injection ? - and for 424 hr after steam injection. The specimens on the left were in the hot leg, and those on the right were in the cold leg. Specimens with the etched appearance lost weight, and the darkened specimens gained , ;" weight. _ . n »n a) +) 23 Iahle 4.A$Copgentration~of~lmpuritigg,in,NCL-l7 Salt g id ik Tt ;'.'.l.'i rSalt Circulation "-Concentration”in'Salt,;ppm Time f-:f L —————— , “(hr) Crr TFe = Ni - Mo 0 48 56 211 - 8 5 1179 82 204 <5 <5 630 11180 . Added steam 1182 109 222 51 71 1400 1468 243 303 7% 56 1900 1973 259 31 - 9 8 2880 320 375 11 12 3600 4324 310 356 15 3 3600 6239 317 362 11 < 2 3500 containing fluoroborate salt can tolerate inleakage of steam and, once the leak is repaired can continue to operate without exten31ve damage even if the salt is not repurified . ' Figure 12 shows micrographs of the specimens in the hottest and- coldest positions. ‘The attack seen in the hot leg is general as expected;for'impurityficontrolled mass transfer processes. Micrometer measurenents showla loss“of'about'l'mil'from specimen surfaces at the makimum’temperature position.a In Fig; 12(b) we see a large amount (about 2 mils) of deposited material. This material has been analyzed with an electron beam miCroprobe, and Table 5 gives the results for the sub- strate and the deposit. These results show that very little chromium has depOsited.' The nickel and molybdenum-concentrations'reached maxima -of 74 and 56 ppm, respectively, 288 hr after the steam addition, decreased and quickly leveled off to about 10 .ppm each. From the microprobe evi- dence much of the nickel and molybdenum apparently deposited in the cold leg, although'not in the Same ratio'aslthey exist'in;Hastellofi N. . One may conclude that under highly oxidizing conditions, during which large ‘weight changes occur (NCL~14 during ‘the air 1eak and NCL-17 after steam inleakage), most of the changes are attributable to the movement of the } Y-98296 Y-98297 Fig. 12. Microstructure of Standard Hastelloy. N Exposed to NaBF,—8 mole % NaF in NCL-17 for 3942 hr. Steam was injected into salt after 1054 hr. (a) 607°C, weight loss 26.8 mg/cm®, as polished; ) (b) 495°C, weight gain 14.8 mg/cm?, as polished. A relatively spongy surface layer on this sample is about 2 mils thick. 500x, ' ' 1 ") 25 Table 5. Microprdhe Analysis of Specimen in Coldest Position. - (493 ’C) in Loop -NCL-17 Exposed to Fluoroborate Salt for fi:' 4000 hr* Steam Injected into Salt’ 1000 hr , . AR after Beginning of - Run 1 L Composition,afwt % Elemeht S E Substrate | Deposit oML ST 73 34.8 Mo L 143 30.6 e 6.8 . <0.5 Fe 3.8 | 2.9 P - o a0P Undetermined — N30 8Corrected for absorption, secondary fluorescence and atomic number effects. bThin layer close to sample surface. 'normally stable nickel and molybdenum. Any‘increased chromium removal from the hot leg is noted only as an increase in the chromium concentra- tion in the salt. The phosphorus found in the deposit was apparently from a phosphate impurity in the steam._- ‘ | Figure 13 ‘shows optical and scanning electron micrographs of speci- mens from the hot and ¢old leg of NCL—17.- The specimens at 593 and 500°C were exposed to the salt for 1050 hr, and then steam was injected into the loop. The specimens then remained in the loop for an additional 9000 hr. The total weight loss of the hot—leg specimen was 40 9 mg/em?, ‘and the weight gain of the cold leg was 16.0 mg/cm . 1In Fig. 13 (top, left) we see that the surface has ‘receded uniformly, although a Widmanstatten " precipitate was left in relief [Fig..13 (lower 18ft)]. Figure 13 (top right) __shows a dark deposit on the cold-leg Specimen, and the deposit exhibits a —layered structure ‘when viewed from above [Fig. 13- (lower right)]. . - ) - Fig. 13. Hastelloy N Specimens from NCL-17 Exposed to NaBF,—8 mole % : NaF for 10,050 hr (1050 hr before Steam Addition). Left: in hot leg at | | 593°C; wei§ht loss 40.9 mg/cm?. Right: in cold leg at 500°C; weight gain - '16.0 mg/cm“. Top: optical micrographs at 500%, etched with glyceria regia L Bottom: scanning electron micrographs at 1000x, 27 Loop NCL-20 In loop NCL-20, d%figiructed of standar&&figstelloy N with removable specimens in each leg, the fluoroborate coolant salt has circulated for 12,000 hr. The selt used was especially pure (see Table 6). The loop is being eperated at temperatufe conditions very near those proposed for the meximum_(681°c) and minimum (387°C) ealt-metal temperature (primary heat exchanger and steam generetor,.respectively) of the MSBR secondary circuit. Fofced air eoolihg (as opposed to ambient air cooling used on other thermal—convection'leops) is used on the lower half of the cold | leg,,and the practical opefating temperatures obtained were 687°C maxi- mum and about 437°C minimum: a temperature difference of 250°C. This difference is thought to be the largest obtained at ORNL in a molten-salt thermal-convection loop, and the maximum temperature is the highest for fluoroborate salt in a loop. Table 6 gives the corrosion results for this experiment. The corrosion rates are the lowest yet attained with a fluoroborate mixture and confirm the importance of salt purity on com- patibility with structurel_materials.‘ The low concentrations of chromium and oxide in the salt are'eiso indicative'of low corrosion rates. Figure 14 shows weight changes of the specimens as functions of time and temperature. Table 6. Weight‘CHenge of Specifien at 687°C and Impurity ~ Content of Salt of NCL-20 ?;?; | _W%ig?zmggss .;,Cizzzzzfiztgige. Chemical Analysis, ppm | P (mils/year)™ Cr . Fe 0 0 T S 58 . 227 - . 500 . 626 ,_0.3 - . _.0.19 90 198 . - 550 1460 0.7, ~ 0.18 99 110 590 2586 1.0 0.15 109 104 470 3784 1.7 017 - 131 91 670 aAssuming uniform loss. 28 ORNL-DWG 71-8059 S WEIGHT CHANGE (mg /cm?) O L ,,,fi - g v 460°C COLDEST SPECIMENS _| 660°C 685°C HOTTEST SPECIMENS . | | | | | O 2000 4000 6000 8000 10,000 12,000 14,000 16,000 TIME OF OPERATION (hr) Fig.'14 Weight Changes of Removable Standard Hastelloy N Specimens in Flowing Sodium Fluoroborate (NCL-20). / / [/ 1] L 3 FCL-1 Pump Loop The MSR-FCL-1 forced-circulation loop operated-for over 11,000 hr (7 runs) to test the compatibility of standard Haételloy N with NaBF,—8 mole % 'NaF coolant salt at temperatures and flow rates similar to those that ‘existed in the MSRE coolant circuit. Salt velocity in the 1/2-in.-0D X 0.042-in,-wall Hastelloy N loop waé nominally 10 fps and was limited by the use of an available pump. The salt in the loop was heated by electri- cal resistance and cooled by a finned, air-cooled heat exchanger. A - schematic of the 100p is given in Fig. 15. Figure 16 is a typical tem- perature profile of the system; this profile was based on a computer plot of the temperature information from the test loop and on heat transfer - calculations that estimaterthe'inner wall temperatures. The maximum and minimum salf temperatures of the loop are 588 and 510°C. Hastelloy N corrosion test specimens were exposed to the circuiating salt at three temperatures, 510,.555, and 588°C. A typical set of specimens is shown schematically in Fig. 17. One of the main drawbacks to this 1oop was that the salt had to be drained from the loop and the loop tubing had to - . " ") GAS LINES TO PUMP OPERATING PRESSURE 18 psio LFB PUMP Fig. Corrosion L | 950 °F 29 ORNL-DWG 6B-2T97R SALT SAMPLE LINE AJUSTO SPEDE 5 hp MOTOR et e ———— = OIL LINES TO PUMP METALLURGICAL SPECIMEN ‘ {030 *F RESISTANCE HEATED SECTION. METALLURGICAL SPECIMEN BLOWER ICA lsnggé:'figssc - HEATER LUG (TYPICAL) — /‘J RESISTANCE HEATED SECTION — 950 *F - FREEZE i VALVE — : ' ' })—.- AIR REYNOLD'S NUMBER ~3400 l SUMP I C VELOCITY ~7 ft/sec Fig. 15. Schematic of MSR-FCL-1. FLOW RATE - 3 gpm ‘ ORNL-DWG £%-253% | FIRST | - | SECOND | ' FINNED l |HEATER | | HEATER |_ COOLING COIL 10 1 - - - i | N P PUMP ) i ) METAL COUPON LOCATION } ‘ :— LIQUID TEMPERATURE 1300 3 { === INNER WALL TEMPERATURE L] 1200 +— ?": 1 . ; ! 100 ! e & b1l ™ . = | l ] “‘-—_-h:‘fl-— % 900 [— +—rt 4 BT F 800 —t Lo i 0 700 |— } o o €00 i i |- . .0 10 .20 0 % - &0 50 .60 ~ LENGTH (f) 16.,vTemperature Profileiof.Molten-Salt Foréed—Coqyéction Loop, MSR-FCL-1, at Typical Operating Conditions. 30 ORNL-DWG €8-4294 Fig. 17. Specimen in Molten Salt Corrosion Forded—ConVection Loop. be cut to remové the specimens. This, of course, necessitated welding to replace the specimens and resulted in a fairly long turnaround time | between runs. | | ‘ | After 659 hr of shakedown operation under various conditions and the initial 2082 hr of operation at design conditions a bearing failed in the LFB centrifugal pump. The pump rotary element was removed, and a new bearing was installed. To take advafitage'of this unscheduled “downtime, portions of the loop piping and the corrosion specimens were: removed for metallographic ekamination and weight change measurefients.- This first operating period was designated run 1. It was then decided: that the loop would be shut down each 2000 hr for pump-bearing replace- ment and specimen.examination. , | The 6perating times and corrosion rates'fo:_the first five runs are summarized in Table 7, and Fig. 18 shofis the weight changes of specimens as measured after each run. The details of each run willlbe discussed in the following pages.‘ It is nbted‘that the mass transfer rates were higher in the pump loop than in thermal-convection loops circulating | - salt of éqfial purity. Thus,_a‘velocity effect is suspected., o | \ ) wh " 31 Table 7. Average Weight Change of Specimens in MSR-FCL-1 Total Salt | " Average Specimen’weigfit Corrosion Rate at R Exposure Time, hr Change at Indicated - 588°C Assuming un Tem erature mg/cm? Uniform Loss,? mils/year Design Isothermal P &/ ch 2 Y Conditions Conditions 510°C 555°C 588°C Overall Incremental 1 2082 - 659 +1.1 -2.7 -11.1 1.6 1.6 2 4088 667 +2.5 —4,2 -17.3 1.4 1.2 3 6097 . - 667 +3.1 5.2 —21.0 1.2 0.7 4 8573 1018 +2.1 -5.8 -23.1 0.9 0.3 5 8995 1068 - +2.5 —7.9 —31.9 1.2 . 7.2 aBased‘on total salt exposure time. ORNL-DWG 7{-4935R 20 10 950°F I -10 1030°F ~-20 WEIGHT CHANGES (mg/cm?)- -30 1090°F o . 200 4ooo . 6000 8000 10,000 12,000 L - "SALT EXPOSURE TIME (hr) : B 'Fig;'18 Welght Changes of Removable Standard Hastelloy N Specimens in Flowing Sodium Fluoroborate in Pumped LoopiMSR-FCL-l. A 1-mil/year corrosion rate’ is included for comparison.,' o : o 32 " Runs 1 and 2. Figure 19 shows micrographs of specimens from each location after _2082 and 4088 hr (run 2) of design operation. In (a) and (b) it is clear ~ that the-attaek at 588°C, the highest'salt temperature,foccurred primarily 'at'the intersections of the grain boundaries with the surface. At 555°C as the weight changes indicated, much less attack has occurred [(e) and (d)]1. . For 510°C, the lowest salt temperature, a metallic deposit is visible [(e) and (f)] along the surface in the as-polished condition. This deposit occurs under steady-state conditions and was shown by elec- tron microprobe analysis to have. the approximate composition of Hastelloy N; no salt-like deposit was observed. Figure 20 shows the surface of the specimens at 588 and 510°C. Note the smoothness of the surface of-the specimen that has been attacked and the deposit on the specimen in the cold region. | . Samples of loop tubing were periodically removed and exanined metal- lographically for comparison with the weight-change specimens. The behavior was similar, as shown in Fig. 21. The attaekdconcentrated at 'the-interseotions of the grain bonndariesiwith-the surface, and deposits nere-seen in the cold region,' The tuBing had.a much_Smaller grain size . = than the specimens; from this one might.anticipate‘a_greater corrosion rate for the tubing. No difference could be seen between tubing exposed under the heaters'and tubing at the same temperature but not under the heaters. | Examination of the pump (Fig. 22) and the pump bowl (Fig. 23) did not disclose any gross contamination; however, small amounts of a green materiai were found on the pump bowl at the sait level, After the bowl was washed with water to remove the salt, very fine, whisker-like, green crystals were also found in the bottom of the pump bowl (Fig;.Qé). large piece of mixed green material and white salt (< 1 g total weight) was analyzed and identified; Table 8 shows the composition.' These , results are equivalent to 55 wt %Z NaBFy,, 30 wt % Naachs, and smaller o amounts of iron, nickeI; and molybdenum fluorides. The presence of the | green corrosion product and the level of chromium concentration in the | - ‘salt indicate that we reached the saturation concentration of chromium kusj & " Y-95551 Fig. 19, StandardiHastelloy'N SpeCimehs from MSR~-FCL-1, Exposed to . NaBF ,—8 mole % NaF for the Indicated Times and Temperatures and for an . Additional 659 to 667 hr at Varying Conditions. (a) 588°C, 2082 hr, weight loss 12.5 mg/cm?, etched With glyceria regia, 500%x. (b) 588°C, 4088 hr, weight loss 18 mg/cm?. (¢) 555°C, 2082 hr, weight loss i 3.0 mglcm (d) 555°C, 4088 hr, weight loss 4 mg/cm?. (e) 510°C, 2082 hr, weight §ain 2. 2 mg/cm , as polished (f) 510°C, 4088 hr, weight gain 4 mg/cm . ‘ 91 Surface of Standard Hastelloy Spécimens‘Eqused in Loop Condi- st 12,5 mg/cm? at °C. 2 (b) Deposit on specimen that gained 2.2 mg/cm® at 510 ions (and 659 hr at Varying (a) Smooth surface on specimen that lo 45%. Fig. 20. MSR-FCL-1 for 2082 hr at Design Condit tions). 588°C. 1 Exposéd to (a) Tub ing exposed at from MSR-FCL- - ing Standard Hastelloy N Loop Tub % NaF for 2082 hr at Design Cond exposed at 588° 510°C. 21. NaBF,—8 mole Fig ing itions. 500x, egia. (b) Tub lar Etched w c ith glycer As polished. PHOTO 96311 at .Fig. 22, Pump'Impéller from Loop MSR-FCL-1. This impellerloperated in NaBF,—8 mole % NaF for 659 hr under variable conditions and 4088 hr of ‘normal operation at 510°C. PHOTO 96313 " ») 5 ; ' ' Fig. 23. Pump Bowl from Loop MSR~FCL-l. The pump operated in ' NaBF4—8 mole Z NaF for 659 hr under variable conditions and 4088 hr of normal operation at 510°C. 36 Fig. 24. Inside of Pump Bowl from Loop MSR~FCL-1 After Being Washed with Water to Remove Salt, Showing Crystalline Residue. The pump operated in NaBF4~8 mole % NaF for 659 hr under variable conditions and 4088 hr of normal operation at 510°C. B - Table 8. Chemical Composition of Green Deposit Removéd from the Pump Bowl After Run 2 ' Element - Concentration, wt % Cr | 6.14 Fe - 0.26 Ni j 0.30 Mo - 0.022 Na o 21.6 B - . 4.56 -~ F . ' 57.0 o & - 37 corrosion product in the pump bowl during the test. No indication of fluoride deposit or plugging in any other part of the system has been seen. Run 3 'Figure 25 shows micrographs'of specimens in MSR-FCL-1 after more ‘than 6000 hr salt exposure (3 runs). Uniform attack is seen on the specimens exposed to salt at 588 and 555°C, with the rougher surface seen on thé higher temperature samplo. In (c) the mounting material remqved some of the deposit on the'cold-leg specimen, but portiomns can still be seen. | Visualrexamination of the pump and pump bowl after run 3 showed no obvious corrosion. None'of the green Na3CrF5 oorrosion product that had been observed previously was present. Y-97618 [Y-97616 (a) - " . (b) (c) Fig. 25. Standard Hastelloy N Specimens from MSR-FCL-1, Exposed to NaBFy—8 mole % NaF for 667 hr under Variable Conditions and 6097 hr at Design Conditions. As polished, 500x. (a) Exposed at 588°C, weight loss 21.4 mg/em?. (b) Exposed at 555°C, weight loss 4.7 mg/cm ‘(c) Exposed at 510°C weight gain 5.0 mg/cm - The light material away from the surface of the sample is part of the surface deposit that became separated from the sample during mounting. 38 Installation and Testing of Cold Finger During Run 3 During run 3 a cold-finger corrosion product trap, shown in Fig. 26 d2% in sodium fluoroborate and similar in design to one previously use test loop PKP-1, was inserted into thefisalt in the pump bowl in an attempt to induce preferential deposition of corrosion products from the salt | circulating in the loop. Such deposition had been observed on the cold finger in loop PKP-1. Gross.deposition'is always possible in a temperature- gradient system, and we feel that this eold—fingefrdevice is a likely com? bonent to prevent this problem. o The cold finger-ueed in MSR-FCL-1 is a closed-end nickel cylinder, 1 3/4 in. long X 3/8 in. OD, with a 0.070-in.-thick wall. It is cooled by an argon—water mixture injected into it and then discharged to the atmosphere. The metal wall temperature is measured and recorded by twe 0.020-in.-0D sheathed, ungrounded Chromel vs Alumel thermoeouples inserted in two 1-1n.—deep axial holes (0.023 in. diam) in the 0. 070—1n.-thick wall of the cold finger. 25), N. Smith, P. G. Smith, and R. B. Gallaher, MSR Program Semiann. Progr. Rept. Feb. 28, 1969, 0RNL—4396, p. 102, Fig. 26. Cold-Finger Corrosion Product Trap for MSR-FCL-1. d *h had been foun 39 " Eight tests were ‘made in fihich'the-cold fingerlwas inserted into the salt in the pump bowl of the loop and cooled to temperatures ranging from 493 to '140°C, = The duration of the tests ranged from 1 5 to 5.3 hr. The temperature of the salt in the pump bowl was 510° C. In contrast to the deposits of material containing Nagchs, which d?® on a cold'finger'in°PKP-1'loop"at metal wall tempera- tures of 400, 460, and 477°C no significant deposit of any kind was seen on the cold finger tests. in loop MSR-FCL-l Even in the three tests where indicated wall temperatures were below the salt liquidus ‘temperature (385°C), the surface of the cold finger was essentially clean, as visually obser#ed'fihen‘it7Was withdrawn into a sight glass. 0ccasionally, small patches (1/8 to 1/4 in. across) of white material ‘estimated to be a few mils thick were seen on the surface. The cold finger was then installed in the PKP- 1 fluoroborate loop -n an attempt to duplicate the previous cold—finger test results of that loop. The chromium concentration of the fluoroborate salt in loop PKP-1 1is about 500 ppm,'while that in MSR-FCL-1 is about 250 ppm. Two “tests were run in which the indicated cold-finger wall temperature was " about 150°C (salt'temperature'in"the'pump bowl was 548°C). Test times were 1-and 4.5 hr. No deposition on the cold'fingerewas seen after withdrawal into a sight glass. Tests at higher temperatures were not made because of temperature control difficulties with the cold finger.\ Since the cold finger"preVious1y used in the PKP-1 loop.contained . grooves on the outside Surface, the lower half of the surface of the MSR-FCL-1 cold finger was scored with file marks about 0 010 in. deep, and a third test run ‘at about 150°C lasting 671/2 hr was made in the PKP-1 loop. ‘In this run a deposit was obtained. Generally, the entire :surface'of the'éold”finger was covered'With‘a'white deposit,NWith‘an overlay of bright green’ ‘material on the lower half of the cold finger. The cold finger was allowed to ‘stand overnight under a helium atmosphere, and by this time the deposit,had begun to separate from the cold-finger surface. While the cold finger was moved for photographing, the entire +26p, B. Gallaher and A. N. Smith, MSR Program Semiann. Progr. -Rept. - Aug. 31, 1969, 0RNL—&449, PpP. - 74—75.,; 40 depositrspalled off, leaving_a'clean metalrsurface. The total deposit weight was 1.63 g, of which_l;26 g was mostly green material (complete separation was not possible). Chemical_analysis_of'the green material disclosed 2.93 wt % Cr, 1.03 wt % Fe, 370 ppm Ni, and < 500 ppm Mo, with the remainder:Na, B ~and F, Stoichiometric calculations indicated 11 mole % Na3CrF5, 4 mole Z NagFeFe, and the remainder a mixture of. NaBF, and NaF Subsequent operation of the cold finger at the same condltions in | 'loop MSR-FCL-1 did not produce a deposit. On the basis of these tests we concluded that a wetting problem exists with the salt-and that a - grooved or roughened surface is required to obtain a depositl -We then tried various designs to improve adherence of the deposit to the .cold. finger and to'lessen the_prohability of accidental removal during with- drawal from the pump bowl. Run 4 .After run 4 when the pump had been removed, we found approx1mately 50 g of intermixed white and green material on the pump liner above the liquid level Analysis of'a small portion of this material showed about equal amounts of NaF and NaBFq, with 4300 ppm Cr, which we assume to be in the form of Na3CrF5. We do not know the conditions that caused the dep091t10n of the corrosion product in the pump liner. | Speclmens in the coldest part of the system gained weight during the first three runs‘and then lost a small amount of weight during run‘4. Thickness measurements of these specimens disclosed that the upstream end had reduced in thickness compared with the middle and downstream end. Specimens at other locations in the loop (555 or 588 C) showed similar but smaller thickness variations. Figure 27 shows a specimen at 588°C after run 4 and clearly illustrates the‘heavier attack at the | upstream end. Changes in specimen thickness‘are‘Summarized,in Table 9, and theylare-quite consistent with the changes calculatedffrom_weight 'measurements. Figure 28 shows micrographs of specimens from MSR-FCL-l after run 4 (9600 hr total exposure to the fluorohorate mixture). The left micro- graph shows the surface of the specimen~at~the'hottest'position (5889C),d o » 41 Fig. 27. Hastelloy N Specimen 4 from MSR-FCL-1 Exposed to NaBF,—8 mole % NaF at 588°C for 9600 hr (Through Run 4). Flow is to the left. ‘ Table 9. Average Thickness Change of Specimens After 9600 hr Total Exposure to NaBF,—8 mole 7% NaF in MSR-FCL-1 Specimen Average Thickness Temperature Change (°c) (mils) 588 ‘ ' ' —-1.25 555 . : -0.75 510 R +1.0% aDiscounting upstream end of sample, thickness increase is +2.5 (see text). Y-99704 Temperature, *C 588 . _ : 555 - 510 . Weight Change, mg/cm? -21.3 i : —4.6 ~0.1 Fig. 28. Hastelloy N Specimens from MSR-FCL-1 Exposed to NaBF,—8 mole % NaF for 9600 hr (4 Runs). 42 Average weight losses of the specimens are 21.3 mg/cmz. ‘Microprobe analysis_disglosed.arconcentrafion gradient of chromium and iron for a ‘distance of 0.6 mil. ‘There is less than 1.0 wt % Cr and only 1.0 wt Z Fe 0.4 mil into the Specifien. This'depletéd area is represented by;thé darkened area_extending'from the'surface for approximafely'O.S mil. The center micrograph shows the surface of a specimén at 555°C. The average weight loss of'ffieserépecimenswas 4.6@g/cm2. Microprobe analysis of these specimensvdisclosed only a small concentration gradient for a dis- tance of O.i mil. This follows from the small amount of attack seen on the.specimen;"The micrograph to the right shows the deposit on the sur- face of the spéciméns at 510°C. The average weight loss of the specimen is.O.i mg/cm?, but thiS“overall'loss stems from a great deal of corrosion on the leading edge'and is not indicative of the whole speéimen. Micro- probe analysis of.this surface showed a large amount of iron and a little more nickel than usual. This situation existed for a distance of 0.5 mil, after which the conéentrations of both elements approached that of the matrix. The average deposit is only 0.2 mil thick; thus, it appears that portions of the deposit have diffused in. Prior evidence in deposit study had disclosed similar fifidings. Run. 5 Run 5 was terminated after 422 hr of operation at design conditions ‘when the pump bearings failed. This failurg allowed the pump impeller to contact the impeller houéing, abrading the impeller. Metal particles from the impeller were seen in the salt left in the pump bowl after dumping.' Acéordingly, the pump was removed from the loop, andlat the same time we removed'the corrosion test specimen assemblies. This allowed fis to clean the loop piping by flushing withideminerélized water. A total of 3.31 g of Hastglloy N particles.was recovered from thesloop.v The loop_Was dried bj air purging, and the corrosion specimen assemblies and pump were reinstalled. A ni¢ke1 filter with 40-um pores was'also installéd in the dump tank line to filter any metal particles from the salt being returned to the loop. | | | _ During the scheduled loop shutdown at the end of run 4, we had enlarged the access port in the pump bowl from 1/2- t6.3/4-in._diam 43 (maximum size possible’in'the pump bowl) to provide more flexibility in the design and operation of cold-finger devices. With the enlarged access port we were able to insert a 5/8-in.-0D.cold finger with circum- ferential grooves that had been previously used in another circulating fluoroborate loop (PKP-1) and on which significant deposits of salt con- taining Na3;CrF¢ had ‘been obtained. Several tests were made during run 5 with this cold finger at tem- peratures between 150 and 420°C. The cold.finger was inserted below the liquid surface in the pump bowl'for:periods of 5 to 6 hr for each test. The salt temperature_in the'pump bowl was approximately 516°C. Small deposits with a light-green'color'(indication of corrosion product Na3CrFg) were-occasionally seen on the cold finger when it was withdrawn into the sight glass above therpump'bowl. Howeuer, such deposits usually dropped off immediately. Apparently any slight vibration is enough to knock most of the deposited material loose before the cold finger can be .pulled all the way out of the pump bowl. The weight changes of Hastelloy N specimens in loop MSR-FCL~1l are compared in Table 7 for each of five test runs. The corrosion rate | decreased steadily over the first four runs and, at the conclusion of run 4 “had dropped to 0.3 mil/hr at the point of maximum temperature (588°C). However, the corrosion rate during run 5 was higher than in .any of the preceding runs. Although the increased rate of attack during run 5 is undoubtedly assoclated with operating difficulties encountered during the run, we have not pinpointed the specific causes. Increased corrosion is normally accompanied by an increase in impurities in the salt; however, we did not see any significant impurity changes during run 5. Because of the nature of the last pump-bearing failure, we were not able to take a salt ' sample from the loop'before shutdown=in.our normal manner.‘ Thus a sample 'was_taken from the dump tank after the loop had been shut down. However, because of dilution-there'was no assurance that.the'sample was represen- tative of the salt that had circulated in the loop. ' 'Figure 29 shows specimens 2 (555°C) 4 (588°C), and 8 (510°C) at the end of rum 5. Note the change in appearance of specimen 4 between run 4 (Fig. 27) and run 5 (Fig. 29). The etched appearance of specimen 4 after " 'PHOTO 77947 Fig. 29. Hastelloy N Specimens from MSR-FCL~1 Exposed to NaBFy;—8 mole % NaF for 10,000 hr Total (Through Run 5). From top to bottom the temperatures were 555, 588, and 510°C. 9% by 45 run 5 1s typical of the appearance of Hastelloy N specimens in thermal convection loop NCL-17 (see previous section, Fig. 13) after extreme oxidizing conditions had been created by intentional steam inleakage and large amounts of mass~transfer had occurred. Run 6 The sixth run was abruptly terminated after 1241 hr, when oil from ‘a broken pump cooling oil line ignited on contact with the pump bowl and ‘adjacent piping. We'extinguished the fire and turned off all loop ‘heaters, which allowed the salt to freeze in the loop piping. Figure 30 shows the pump bowl and pump with no apparent damage after the fire. The material in the beaker is salt and corrosion products left in the pump bowl. | We replaced the electrical wiring, thermocouples, and service piping necessary to melt the salt in the loop piping and transfer it to - the drain tank. This was the first time that we had attempted to melt salt in the entire loop circuit,:although salt had been melted in the cooling'cdil without difficulty on several occasions. Normally when the loop is placed in standby condition, the salt circulation is stopped, | Fig. 30. Standard Hastelloy N Pump from MSR-FCL-1. Impeller exposed to NaBF,—8 mole % NaF 1319 hr; baffles exposed 11,371 hr. Outside of the pump was exposed to oil fire. 46 the main loop heaters are turned off, and sufficient heat is supplied to the loop to maintain the loop circuit above the salt liquidus tem- perature, 385°C. During the attempted melting, the main loop piping (1/2~in.-0D x 0.042-in.~wall Hastelloy N) ruptured at a point_near the U-bend in the loop and adjacent to one of the metallurgical specimen holders, as shown in Fig. 31. After the rupture, we froze the salt in the loop as quickly as possible by turning off all loop heaters. How- ever,,approximately 77 g salt was lost from the loop. Before the rupture the salt'in the bowl and cooling coil section of the main loop had been melted without difficulty,.the main loop piping had‘reached'temperatfires varying between 371 and 482°C around the loop circuit, and the drain line temperatures were above 426°C. ORNL-LR-DWG 64T40R METALLUREGY : SAMPL HOkva POWER SUPPLY A (MAIN POWER} _ (DETAIL 1600-amp BREAKER % =) METALLURGY 74 SAMPLE 1 LFB PUMP il 10 kva POWER SUPPLY {COOLER PREHEAT) Fig. 31. Molten-Salt Corrosion Testing Loop MSR-FCL-l and Power Supplies Showing Points of Rupture in Runs 6 and 7. (. e ”» 47 In reviewing the loop melting procedure to determine the cause of the rupture, we concluded that uneven thicknesses of thermal insulation caused considerable variation in the temperatures. In partiéular, the temperatures under the clamshell emergency heaters (where‘the thermal insulation was about 1 in. thick, compared to about 2 in. on the remainder of the tubing) were about '80°C below temperatures in the more heavily insulated sections of tubing. Thus, the sélt in the ruptured area melted but was not able-toiexpand‘properly because of frozen salt plugs under the emergency'heaters. Another problem was the difficulty in controlling ‘the rate .of heatup of the main loop by resistance heating. Even though 'thé lowest setting on the‘ioop heater was used, we still had to turn off the heater supply power for short periods to maintain temperature control. We removed -the metallurgical Specimen'éssembly and ruptured section of tube and replaced it with new tubing. Figure 32 shows the ruptured section of tubing. Measurements of the outside diameter of the remaining loop tubing were made énd'compared with the original outside diameter to determine if excessive permanent strain had occurred at other locatioms. | Fig. 32. Tube Ruptured During Salt Melting after Run 6,ALdop 48 Generally all measurements were in the range of 0.500 to 0.505 in., as compared to the original diameter of 0.500 in. The tubing that surrounded the specimens was removed after run 6 (11,371 hr) and examined by both optical metallography and scanning electron microscopy. Figure 33 shows the tubing exposed to'nalt at 588 _ ‘and 555°C. 1In the nptical micrograph of Fig. 33(a) the surface roughening (some salt is still in place) can be seen, and the attack at the grain‘ .boundaries is noted. In the scanning electron micrograph we can see the ~delineation of the grains. Figure 33(b) shows much less attack at 555°C. Figure 34 shows the tubing ddwnstream of the cooler in the coldest posi- ‘tion (510°C) of therloop. The deposit is easily seen in the optical -m1crograph and is seen in the scanning mlcrograph to be in the form of nodules. The upper portion of Fig. 34(c) shows a side view of the nodu- '-lar deposits. ‘ After the oil fire that ended run 6 and the subsequent rupture of the tubing during the attempt to melt the salt from the loop, the corro- sion specimens-were‘removed, weighed, and examined. Surprisingly, weight was lost by all specimens. The weight loss rates found for specimens exposed to the salt at 588 and 555°C were (assnming uniform losses)labout 1.9 and 0.5 mils/year, respectively, for the 1240 hr of run 6. The corro- sion rate ratio for these specimens (severnl timeS'greater>1oss at 588°C ‘than at 555°C) was fairly consistent with past findings."However, in earlier runs, specimens at 510°C had gained weight. .For tnis time period specimens at this-temperatnre 1evelnlost-ma£efia1 at the rate of 1.9_mils/year, the same as for specimens at 588°C. This weight loss anomaly and the overall higher weight losses are attributed to the problems encountered during this run and probably atteét to a local high temperature in the coldet_section alongrwith oxidizing impurities in the salt. A salt analysis showed small increases in chromium, iron, and nickel contents, all of which reflect the increased corrosion rates. "Run 7 After repairs we returned the loop to design conditions for the start of run 7, circulating the same NaBF,—8 mole % NaF coolant salt used since the start of loop operation. C. 49 — e 105076 -} Hastelloy N Tubing from.MSR—FCL~1 Exposed to NaBF,—8 mole % Left: optical micrographs, 500x. Right: scanning Top: at 588°C and bottom at 555°C during Fig. 33. NaF for 11,371 hr. ‘electron micrographs, 1000X design operation. ¥ 1 | i 1 j i 50 Y-105978 (c) ‘ Fig. 34. Hastelloy N Tubing from MSR-FCL-1 Exposed to NaBF,—8 mole % NaF at 510°C for 11,371 hr. (a) Optical micrograph, 500x. (b) Scanning electron micrograph, 1000%. (c) Scanning electron micrograph, 3000X, taken with the specimen tilted to give a side view. 1 51 During run 7 while repairs were .under way on the loop high-temperature \protective instrumentation, an operating error resulted in a portion of the loop tubing- being heated above 1100°C.‘ This caused one failure of the loop tubing adjacent to one of the electrical power input lugs (see Fig. 31) and also about 12 in. downstream. Part of the salt in the ‘loop'(about 2 1iters) was discharged into the 1oop secondary containment. ‘ Temperatures were immediately reduced and the facility was shut down. At about 1100°C the strength of Hastelloy N is very low, and the salt vapor pressure approaches 1000 psi. The salt boiled locally, and over- heating could have reduced'the tube strength until rupture occurred. Local melting of the tubing was also efidenced. At the time of failure the loop had accumulated 10,335 hr of operation at design conditions (101 br during run 7). The tube failure is shown in Fig. 35. Following the rupture'of the_loop-tubing that ended run 7, the corrosion test specimens were again removed for weighing and metallurgi- cal examination. The specimens adjacent to the ruptured area (hottest section under normal conditions, 588°C) were badly warped and partially fusedzto the tube wall. Thus, only metallographic analysis was possible. The specimens'in the intermediate zone (555°C) appeared to be unharmed. The average weight loss for these specimens was 1.3 mg/cm? in the 101 hr “of run 7. Expected 1oss for these specimens in 100 hr would be approxi- ,mately 0.1 mg/cm . Thus, if we attribute the balance of the corrosion to just the period of the temperature excursion (to about 1000° C), the loss rate was 1.2 mg/cm® in 2.5 min or 10,000 mils/year (10 in. /year). Weight losses were also found for the specimens downstream of the . cooler. Although the specimens were adjacent, one specimen lost 0.3 mg/cm2 and the other 3.0 mg/cm®. Again the severity-of the over- heating was'evident. Certainly a large part of these losses may be attributed to moisture contamination of the salt and air oxidation after the loop rupture. Nonetheless, the corrosion rate was quite high. The loop is now being modernized and prepared for additional service. PHOTO 78452 r VR 2 i HEATER LUG A cs Fig. 35. Tube Rupture Caused By Overheating During Run 7, Lobp MSR-FCL-1. ah 53 " SALT CHEMISTRY - | " "Purification With the realization that the fluoroborate salt is easily contami- nated with oxygen and water-type impurities and that corrosion increases with the amount of these'impurities;‘we“have attempted, in close coopera~- ~tion with R. F. Apple and A. S. Meyer, of the Analytical Chemistry -'Division, to purify the fluoroborate in a manner that could be adapted to a‘large system. | In ‘the method that gave favorable résdlts,-a‘mixture of He, BF;,, and HF (purified with fluorine) was pASSed through the fluoroborate at 454°C. A schematic of the process is given in Fig. 36. The gas mixture was passed for 15 hr into a Hastelléy N vessel containing the impure (about~2000'ppm H,0) fluoroborate salt (2.7 kg) at 480°C. 'The'exit gas entered a solutionfof'lo_v61,% pyridine in methyl alcohol. At the com- pletion of the purification process titration of this solution with " Karl:Fischer reagent to the dead-stop end point indicated that about 600 -ppm-H,0 fidsfremoved;'“Chémicalfanalysis;discIosed'that the oxygen content Hécreaéed from-1700 to 1400 ppm"and‘that.the H,0 content changed ORNL-DWG €9-42547R - . 3 h." " BF, 35 CIT\- J/min | KARL FISCHER 60 cm>/min . ‘REAGENT 125 crns/min A END POINT INDICATOR 000 - 90% METHYL ALCOHOL NoBF, — NoF (92-8 mole %) 10% PYRIDINE 450 ~ 510° C ‘ORIGINAL = 2000 ppm H0 . AFTER 15 hr SPARGING = 300 ppm -H,0 Fig;;36. System for Removing Water from Coolant Salt. sh from 2000 to 300 ppm. Changes in the BFj; content of the salt appeared to be minimal. Further work is planned using a nickel vessel to elimi- nate corrosion products resulting from the reaction of HF with Hastelloy N. Analytical Chemistry Over the last few years, we have been-concerned‘dver.the behavior of water in molten fluoride salts. Water in these salts has implications for both corrosion and tritium removal. In corrosion work, we have long - recognized the need for analyses that would allow us to distinguish B between water and (1) compounds that contain H ions, (2) compounds which contain 0%~ ions, and (3) HF (highly corrosive). In the past, a complementary indication-of mass transfer (besides weight changes and metal analysis of the salt) was the analysis for'HZO,ahd 0, in the salt. On the basis of our tritium injection experiment?’ and work in the Analytical Chemistry and Reactor Chemistry Divisions, we now feel that the results of the water determinations included many other constituents ~of the salts, including oxides. Thus, a resulfi_indicating a large amount of water really only showed a corresponding amount of oxide. The lack ~of watervifi the salt is completely reasonable because all the hydfogen in impurity cofipcunds that enter the salt would eventually cause oxida- tion of the metal wall and would in turn be reduced to hydrogen gas, . which would diffuse out of the loop (the corrosion equations will be discussed later). However, increased mass transfer (weight losses and ‘gains by specimens) by virtue of water impurity‘is still accompanied b& an increase in the oxide content of the salt;' Thus; the oxide content is still somewhat indicative of the mass transfer of the loop system. DISCUSSION “Theory Metal corrosion by salt mixtures consists of oxidation of metal constituents to their fluorides, which are soluble in the melt and hence 273. W. Koger, MSR Program Semiann. Progr. Rept Feb. 28, 1971, ORNL-4676, pp. 210-11. 1 . r 55 do not form a protective film.. This electrochemical-type attack is » therefore limited only by the’thermodynamic potential for the oxidation reaction and is selective, removing the least noble constituent, which - _ in the case of Hastelloy N is chromium. Possible corrosion reactions involving impurities in the melt may be written as 2HF + Cr = H, + CrF,,. NiF, + Cr < Ni + CrF,, Fer + Cr = Ni + CrF,. . Although the reaction product is written as Cer, the actual species in _the melt is unknown. Chromium(II) is possibly unstable in the fluoroborate melt, since the only corrosion product isolated and identified is Naachs. Table 10 illustrates the thermodynamic stability of the fluoride compounds in question. Table 10. Relative Thermodynamic Stabilities of Fluoride Compounds S S | cL - Standard Free Energy . - L o , Most Stable of Formation Element Fluoride " (kcal/g-atom F) - Compound at 800°C at 600°C Structurel metals Cr N ‘Cer ’ =12 :—77d | | Fe FeFy ~66 69 N ~ NiF» - =59 —63 | Mo MoFs - -s7.. -58 Diluent sales ~ Li - .. LiF - ~120 125 Na ' NaF -110 =115 ‘XK . KF. - =108 . -113 Be. " BeF2. —103 - . -100 .zr . .- .2ZrFy. . 92 .. 96 : "B BFg . —86 . 88 Active salts U T URy =92 94 Th " "ThFy —99 e -102 56 - In an isothermalmsystem, immediately after immersion,io the fused salt system and before thermodjnamic;equilibrium-has been fully estab- - : ” lished, the electrode potential of the metal has a much higher electro- negative value than the redox potential of the surrounding medium. : As ' ~ a result, reduction of the oxidizer commences.which takes place along . with the corrosion of the metal. As the corrosion products accumulate the metal's electrode'potential shifts in the positive direction, whereas the redox potential moves toward the'negative.side; Equilibrium occurs. when therpotentials are eqoel. Since we are dealing with corrosion in a nonisothermal system (i. e.,-a heat exchanger), the mechanism of interest to us is temperature-gradient mass transfer. A schematic ofi this process is glven in Fig. 37. In a loop ‘that contains salt with UF, and no corrosion products, ‘all points of the loop initially expe- " rience a loss of chromium by the reaction 2UF, + Cr = CrF2 + 2UFj3. As the corrosion—product concentration in the salt is”increeseo, eqoi--‘- librium with respect to the corrosion reaction is eventually reaehed at the point of lowest temperature of the system. At regions of higher temperature, a driving force for the corrosion reaction stiil exists. Thus, the corrosion-product concentration will continue to increase and the equilibrium temperature will begin to increase from the coldest temperature. At this stage, chromium is returned to the walls of the coldest point of the system. The rise in oorrosion-product conoentra- tion in the circuleting salt continues until the amount of chromium returning to the malls exactly balances the amount of chromium entering the system in the hot-leg regions in the same interval.‘ Under these conditions, the two positions of the loop at equilibrium with the salt, ‘which are termed the "balance points," do not shift measurably with time. Thus, a quasi—steadf—state situatiom is eventually achieved, whereby chromium is transported at very low rates and under conditions of a ? fixed chromium surface concentration at any given loop position. At this point the chromium removal.is controlled by the solid-stete diffu~ s sion rate of chromium in Hastelloy N. Table 11 illustrates the small . | \N-j 57 . .. ORNL-DWG 67-6800R HOT SECTION DIFFUSION TO SURFACE: ' SOLUTE ESCAPE THROUGH NEAR-SURFACE LIQUID LAYER , '_—--— ) o DIFFUSION INTO BULK LIQUID | — TRANSPORT T TO COLD PORTION OF SYSTEM COLD SECTION e — — [ X SUPERSATURATION —_ = — NUCLEATION — o GROWTH TO STABLE CRYSTAL SIZE .. _0OR o = 'SUPERSATURATION AND- DIFFUSION _ -~ " THROUGH LIQUD | NUCLEATION ANd GROWTH - - ON METALLIC WALL — —— - — OR DIFFUSION INTO WALL : Fig. 37. Temperature-Gradient Mass Transfer. 58 Table 11. Chromium Transfer from Hastelloy N with Surface Concentration Reduced to Zero | Diffusion Coefficienta‘ Depth of Temperature : D Time Chromium Gradient - (°c) 2o (hr) | . (cm”“/sec) . (cm) (mils) | | | x 1073 650 2 x 1071 4,0000 1.5 0.6 | | 16,000 3 - 1.2 | | . 256,000 (30 years) 12 ; 4.7 850 2 x 10712 4,000 12 4.7 ' - 16,000 - 18 7.1 256,000 130,years) 70 27 By surface activity measurements and microprobe traces after 250~ and 500-hr exposures. i : bChromium concentration within 95% of initial concentration in Hastelloy N (about 7 wt %). , amount of chromium that will be removed if oxidizing conditions are mini- mal and solid-state diffusion is controlling. - | There should'be-Veryrlittie corrosion of Hastelloy N by pure molten NaBF, or NaF, such as'is seen.fOr LiF or Ber; thus we.attribute the | mass transfer to other sources. We have shown several examples'of' inadvertent air (moisture) inleakage?and~thesconcomitant increase in mass transfer. This was also iilustratedkiniNCL-l7; where_steam was added. The corrosion reactions in the sodium fluoroborate salt mixture have not been well established but can be expected to be reasonably represented by the following equations:' "H20 + NaBF, = NaBF30H + HF, NaBF 30H == NaBOF, + HF, 6HF 4+ 2Cr + 6NaF = 2Na3;CrFg + 3H2 (and similar reactions with nickel and iron), ¥ 30; + 4Cr = Zgr203, Cr,0; + 3NaBF, + 6NaF <= 2Na3CrFs + 3NaBOF, "FeF, + Cr= CrF, + Fe (and similar reactions with nickel and molybdenum fluorides), N 59 "LTheinext‘result'is'thatfwater reacts with sodium fluoroboraterto produce hydrogen fluoride and exide ions or an oxygen—containing complex " 'in the salt. The hydrogen fluoride reacts with the constituents of _Hastelloy N to bring'metal ions into solution in the 'salt and to release hydrogen. It is not known if the oxide in solution can. corrode Hastelloy N, but oxygen in air can react*w1th Hastelloy N to produce metal oxides. ' These metal oxides dissolve in the melt to form metal ions and oxide ions in solution. Metal dions in solution in the salt react with and diSplace less noble constituents of the Hastelloy N. Highly oxidizing conditions (presence of HF) can bring all the constituents into, solution. As the salt becomes less oxidizing the more noble'metals in solution react with 1ess noble alloy constituents until only chromium ions are present in the salt in important. quantlties. This movement of chromium is governed primarily by the rate of diffusion of chromium in the Hastelloy N. Assuming that HgO (or HF) causes the mass transfer in sodium fluoro- “borate and on the basis that 1 mole of H20 is required to corrode 1 mole of metal we have calculated the amount of water required to cause known amounts of corrosion on several of our systems (Table 12) For NCL~17 in the 1054-hr operation before steam addition, 41 mg Hgo would have been required to cause the 137 mg metal removed leaving 75 pPpm ox1de in the salt. At the end of 10,000 hr- after steam addition 23 g of metal had:been'removed " This would have required-6.3 g H20 and would have resulted in 13,800 ppm oxide in the salt. Since we have never measured | this much oxide in the salt, this material must deposit. The amount of water required to produce the corrosion seen in NCL~20 is small but has continued to increase with time. This can mean several things. One, there is a constant inleakage in our test systems or, two, there are impurities present in the salt which continually cause corrosion.f‘The first possibllity is quite plausible because we have seen faulty gas B regulators, which can leak undetected for quite a long time.' As for the fiseCOnd possibllity, the Fe ¥ concentration in the salt has decreased over 120 ppm, which would result in the removal of only 350 mg of chromium from the Hastelloy N. This still leaves .about 600‘mg.that=we might "attribute to water-caused corrosion. .60 Table 12, Calculated Amount of Water Necessary to Remove Experimentally Determined Amount of Metal in Different Loop Systems Total Material Corrosion Rate H20 Required Oxide Time HfiimgzgdAgizz At E:Z;::zn of To Remove Total Resulting From (hr) - ‘A Given Time Temperature Material . . React%on (mg) . (mils/year) _ (me) - (mg) NCL-17 1,0542 137 0.22 41 200 239 3,320 19.4 996 . . 5,000 424 4,150 13.6 1,245 6,000 663 4,700 9.9 1,410 - 7,000 1,474 6,090 5.8 - 1,827 9,000 2,888 7,200 3.5 2,160 - 12,800 4,756 - 9,960 S 2.9 2,988 ~ 15,000 6,030 12,700 - 2.9 3,810 19,000 9,006 14,400 2.2 4,320 - 21,600 10,178 . 23,000 3.1 6,900 35,000 | | | NCL-20 3,898 349 | 0.17 o 105 | 525 6,173 615 0.19 ~ 185 925 8,501 923 - 0.20 o211 1,400 NCL-13A | 958 1,380 1.6 414 - 2,070 2,225 1,890 0.95 - 567 | 2,850 3,545 2,340 0.74 672 | 3,360 5,339 3,030 0.64 909 4,550 11,680 5,300 0.51 1,590 . 8,000 14,495 7,030 0.54 2,109 — 10,550 16,843 11,500 0.76 3,450 17,250 | FCL-1 | 2,741 11,790 1.7 3,540 ' 17,700 4,747 17,690 1.5 5,310 | 26,500 6,756 21,620 1.3 . 6,486 . 32,000 9,600 23,580 ~0.96 7,074 " 35,000 10,000 35,370 1.4 ' 10,611 53,000 11,472 42,250 1.4 12,675 63,500 7 ‘aBefore steam addition, all other times of loop NCL-17 after steam addition. : ' ' | ' B 6l _NKEquetions The experiméntalJWeight?changeslof specimens in the hot and cold portions of the polythermal systems are functions of time at a given temperature and are expressed by the equation W= at? , | . D where _ , o S L L : , W= weignt‘chsnge_in milligrams-ner‘square centimeter of surface‘ . area, - a‘= eonstant, t = time,in'hours, b = kinetic time constant. N y _ The actual weight loss per unit. time is a function of temperature and is represented by the Arrhenius relation 3‘;’ Aema o where '%%fi='w918ht change in miiligrams per square centimeter of surface area per unit time, - ‘= constant, ' o A R = gas constant, T = absolute temperature in °K, Q = activation energy. From Fick's second law of diffusion, one can develop the’ equation e, - _AM = weight of material transported, - where Az = internal peripheral area of loop tubing, p, = density of original alloy, 62 initial chromium concentration, equilibrium constant at the balance point, equilibrium constant at the absolute temperature T, diffusion coefficient of chromium in alloy, and o o ™ n time. This equation describes the mass transfer that can occur in a temperature- gradient system. The use of equations of this type is given in a paper by Evans, Koger, and DeVan.?2® Of importance is the fact that at the balance point no material is transported at temperatures above the | balance point material is transported away, and at temperatures below ' the balance point material is brought in. The equation was originally considered only for_diffusion—controlled processes such as the corrosion reaction between chromium and UFy, but we see in our plots of weight change' versus position that balance points exist even when solid-state diffusion is not controlling (Figs. 38, 39, 40, and 41). Thus the same mass transfer theory holds for our impurity-controlled reactions. _ The material deposited in the cold leg during the normalfoperationv of the loop is metallic, generally nickel and molybdenum.- Plugging- type deposits, usually resulting when the solubility of a metal fluoride in the salt is exceeded, have been found to be mixed or complex fluorides such as NajCrFg, NaFeF3, NaNiF;. .This report does not include discussions of condition of saturation of any components. Most of the material retained in the salt is chromium with some iron, both in the form of fluorides. ' | | Kinetics According to prevailing concepts of reaction mechanisms, a reaction such as the corrosion of a solid material by a fused salt occurs as a sequence of steps. The combined effect of these steps determines the net rate of the reaction. However 1if, as is common, one step is \ 28R. B. Evans III, J. W. Koger, ‘and J. H. DeVan, Cbrroszon in Pbe- thermal Loop Systems. II. A Solid-State thfuscon Mechanism With and thhout quuzd lem Effects, 0RNL-4575 Vol 2, (June 1971). C, » 63 ML-m--NTT?. . LOOP ¥4~ MODIFIED SPECIMENS oo 065 02 W0 seress eS| T s o HEATED AND INSULATED o _ o L " 600 w : -2 8 5502 | g 3 800 WEIGHT CHANGE (mg/om?) 3 -4 | -‘ v ' e Sl o .o 82T - W22 . 4 HHW -8 . 2044 W . s 3603w . 4448 N -9 . s SOIGw * 2w ¢ 7600Mm -.o. . . . m" . -H -2 - 13 -4 - sl S Q L 2 0 40 50 & -0 L 0 - DISTANCE FROM MOTTEST PORTION OF THE UPPER CROSSOVER {in) ok vhoohe e o UPPER COLD LEG BEND LOWER BEND HOT LEG CROSSOVER {VERTICAL} CROSSOVER (VERTICAL) Fig. 38. Weight Change as Function of Position (Temperature) for ‘Titanium-Modified Hastelloy N Exposed to Sodium Fluoroborate Salt in - . Loop NCL-14, ' o . - . - : O -8 -10 WEIGHT CHANGE (mg /cm?) ! @ -2 -18 -20 ORNL-DWG 71-8062 J ! t B l'-*HEATED AND INSULATED —i ?\x Y 3 X ‘3¥‘» \\ 4 N/ L .y S =A / / ( THERMAL CONVECTION LOOP NCL-13A A f 550 500 - 450 | [ a \\ _ _ [ - ° . ! \ : | v 958 br \ { T w2225 hr ‘ : [ -® 3545 hr | \ { | . 45339bh | , T S AaYWe80h | T T T T {014,495 hr | L essa3nr L | | ” * ' ! -t - i J | ! i |- ! ] | 0 10 20 30 40 50 60 70 80 90 100 DISTANCE FROM HOTTEST PORTION OF THE UPPER CROSSOVER (in.) t“ I - . i | _|..+_mfi,,{..*|.___ ..| UPPER COLD LEG BEND LOWER BEND HOT LEG CROSSOVER . {(VERTICAL} - CROSSOVER (VERTICAL) Fig. 39. Weight Change as Function of Position (Temperature) for Hastelloy N Exposed to Sodium Fluoborate Salt in Loop NCL-13A. ) . TEMPERATURE (°C) ORNL-DWG T1-8060 |=——HEATED AND INSULATED 30 600 o - < e 550 § S e o £ & w 500 W Q = z2 L [ - - o ‘ 450 ; THERMAL CONVECTION LOOP NCL-17 @ W = . ® 239 hr o 424 hr 4 663 hr 41474 hr | 1)ME AFTER v 2888 hr ) v 4756 hr STEAM INJECTION ‘ e 6030 hr e 9006 hr ¢ 10,478 hr ¢ 1054 hr BEFORE STEAM O 0 20 30 40 50 60 70 80 20 100 DISTANCE FROM HOTTEST PORTION OF THE UPPER CROSSOVER (in.) L ol [ [ | I _ I P | | i UPPER COLD LEG BEND LOWER BEND HOT LEG ' CROSSOVER (VERTICAL) CROSSOVER (VERTICAL) ‘Fig. 40. Weight Change as Function of Position (Temperature) for Hastelloy N Exposed to Sodium Fluoborate Salt in Loop NCL-17. %9 65 ORNL-DWG 71-8061 AIR-. : COOLED-I-iHEATED AND INSULATED SECTION | 450 & 500 5 & < : £ 550 - e g T - 600 Ul < HERMAL CONVECTION LOOP NCL-20 & o = | | Z I- 650 9 o | = o 3784 hr | ® 6059 hr 700 a 8387 hr A 11,883 hr -20 : 0 10 20 30 40 50 60 70 80 = 90 100 DISTANCE FROM HOTTEST PORTION OF THE UPPER CROSSOVER (in.) L L - b | el 1 ~ T T =T . i UPPER COLD LEG BEND LOWER BEND - HOT LEG CROSSOVER {VERTICAL) CROSSOVER (VERTICAL) Fig. 41. Weight Change as Function of Position (Tenperature) for Hastelloy N Exposed to Sodium Fluoborate Salt in Loop NCL-~20. significantly slower than the others, the—:ate of the entire reaction is governed by the slowest step. It is convenient to break up the dissolution process in the hot leg _ 1nto four separate steps: 1. diffusion of the attacked elements through the matrix of the alloy to the surface, | 2, migration of the oxidizing species from the salt to the surface, 3. oxidation of the metal to a dissolved fluoride,‘ o 4. migratlon of the oxidized metal to ‘the flowing salt. In cases where the slow step involves awdiffusion;proceSS‘(volume diffusinn‘in a selid), the reecfion is tefmed &iffusionecontrolled and is govefned by the laws of diffusion kinetics.‘ Alternately; if a boundary-region process‘(solutiqnj constitutes the slow step, .the rate ofithe,overallsreactien,is_determined-by the kinetics of that process. 66 The rate of formation of the corrosion product fluoride should be included in this analysis, but experimental data'relating to its formation and the effect of impurities are lacking. | _ If. the diffusion step is rate controlling, then no velocity effect " would be seen. However, if step 2 was slowest and was controlling, the process would be velocity dependent. At a certain critical velocity (possibly even that of the salt inm natural'circulatien loops),.the pathv for step 2 would become short and step 3 would contrel. Above this critical velocity, no velocityrdependence would again be seen. Step 4, 1ike step 2, would be‘velocity dependent. Thus, experifients in systems with different salt velocities could help to determine the rate-controlling mechanism.‘ | In this work, we will consider the possibilities of solid-state diffusion and solution control. "80lid-State Diffusion Control A form of Fick's second law can be used for the weight loss in this case = AC /Dt , (4) where W = weight change, C = concentration change of diffusing substance, D = solid state diffusion coefficient, t = time, A = constant. 1f Eq. (&) applies, b of Eq.‘(l) is theoretically equal to 1/2. The diffusion coefficient follows the Arrhenius relationship .D = D, exp(-Dg/RT) , | | 6y where Do is constent and Dg is the effective activation energy for diffusion in the alloy. O 4 67 If we differentiate Eq. (4) and substitute Eq. (5) for D, we obtain aw AC BTt ext , dt -'-'i— ¥ fl/t ’EXP(“'DSIZRT) » shoWing‘thét‘the theoretical activation energy for the corrosion reaction is half that for diffusion: Q. = Dg/2 theo . o ® Solution Controlling - d?? to describe solution attack The following type of equation is use controlled’by,1iquid 520 * 15°C, oss where | C = concentration of Fe + Cr in Hastelloy N, g/cm?, D = diffusion coefficient, cm?/sec, t = time, sec., By using the weight changes and the combined iron and chromium concentration for the loop specimens, almost identical D values were -obtained for the standard and modified Hastelloy N. This is quite significant, since the iron content of titaniumemodified Hastelloy N is negligible. Thus D, or in reality a rate constant Kl’ is independent of the concentration of chromium, iron, and other diffusing species. The weight change, however, is quite sensitive to the total concentration of the migrating elements. Confirmation of this hypothesis is evident when K1 is compared with D, 1in Hastelloy N 31»32 ~ For loops NCL~13 and C - 31y. R. Grimes, G. M., Watson, J. H. DeVan, and R. B. Evans, "Radio- tracer Techniques in the Study of Corrosion by Molten: Fluorides, pp. 559-74 in Conference on the Use of Radioisotopes in the Physical - Setences and Industry, September 6-—17, 1960, Proceedings, Vol. III, International Atomic Energy Agency, Vienna, 1962. - 32R. B. Evans III, J. H. DeVan, and G. M. Watson, SerBthfuston of . Chromzum in Ntckel Base AZZoys, 0RNL—2982 (January 1961). 71 Table 14. Rate Expressions for Material Deposited from Coolant Salt ~ NaBFy—S8 mole 7 NaF Oxide Weight Change | ' Tempereture Content Iempera- Constants? . ' Dependence Material ture of Salt °c) _ (ppm) a b A Q, cal/mole Ti-modified 500 460 0.0085 0.6 9 x 1077 8,240 Hastelloy N 482 0.0069 0.6 | 527 0.0056 0.6 Standard 500 460 0.17 0.33 4x10°% 14,400 Hastelloy N 482 0.13 0.33 527 0.088 0.33 | Ti-modified 1500 460 0.007 0.65 - 2.9 x 10 12 28,800 Hastelloy N 482 0.0052 0.65 527 0.0016 0.65 qConstants for equation W atb; where W = weight change, mg/cmz; = time, hr, | bconstants for equation gt = A eXp(+Q/RT),. NCL-14 at 607°C Ky =5 X 10—_12 cm?/sec, while the literature value for D, at 607°C is 5 x 10 '* cm®?/sec. The appropriate Arrhenius relation constants are Do = 6.068 x 10 ° cm?/sec and Q = 41.48 kcal/mole. The effect of impurity oxide content in fluoroborate salt on weight losses of the Hastelloy N alloys at 555 and 604°C is given in Fig. 42, The data show that titanium—modified alloys lose less weight than the . standard Hastelloy N for exposure in fluoroborate salt containing 500 ppm sz-impurity‘at temperatures between 555 and’ 604°C. However, at the same tempetetufes the titaniumrmodified alloy exposed to fluoroborate salt containing 1500 ppm oxide shows a greater weight 1oss than either ..alloy exposed to the 500 ppm oxide salt. B The actual -average weight loss rate,wes_calcnlateq.for several alloy=-salt combinations and_plntted in Fig. 43 against the inverse of ~ the absolute temperature. . Also included in Fig. 43 are weight loss rates calculated from 72 ORNL —DWG €9-763R g T T TN ' : 'Sl4n1fl | 605°C 5 ,‘/ ‘ 7 7 £ ¥ 05 °C 7 605 ° o 7 ”‘y,—’ly’ E 7 555°c . L 2l o= / o4 mil ———— S “’,av’ 1 A mi g ‘ 5°C 4 S . e // I N O ©5655°C i _ J - ‘ i z o - i . : - * 05 L2 0.025 mil /’s85°C /1 T | b | 7 © Ti-MODIFED HASTELLOY N . © STANDARD HASTELLOY N 0.2 —— {500 ppm OXIDE IN SALT | 500 ppm oxnm»: IN SALT - | ] . i ‘ r ye0r ' 0.1 L | _* _ 10° 2 5 10% 2 TIME (hr) Fig. 42. Weight Loss for Haételloy N Alloys in Sodium Fluoroborate Coolant Salts. — ' S ' = (Coécs)_/ADt/fl , where. W= weight loss in milligrams per square centimeter of surface area, ‘D = the diffusion coefficient in square centimeters'per'Second, t = time in seconds, Cg and C, = the concentration maintained at surface and the initial uni- form concentration, respectively, in milligrams per cubic centimeter. 73 ORNL-DWG €9-762R2 ¢ TYPE 304L STAINLESS STEEL 4 4 Ti-MODIFIED HASTELLOY N 0 ¢ STANDARD HASTELLOY N x CALCULATED-ALLOY CONTAINS 7% Cr + .CALCULATED-ALLOY CONTAINS 8% Cr ® 4 ¢ FUEL SALT 4 ¢ FLUOROBORATE COOLANT SALT TEMPERATURE (°C) o2 700 650 - 600 550 5 2 ‘ ' 4073 | D (cmPsec) .cqu . 3;10‘“‘4 ‘ppm -OXIDE Pl 23:10“4 +1500 E : 4 [84x10715— — <4500 s 2 .' 45 — %\ __45 + 5.0*‘0-‘ N 4500 0 e ox0 P = 5 2 107> _ ‘ - 098 102 106 440 444 148 122 126 - 1000y | | Fig 43. -Arrhenius-Type'Plotffor Corrosion .of Materials in Fluoride Salts. ' - ' . - For 'this calculation we assumed C = 0, ‘that all'weight loss was the removal of chromium, and that this removal was controlled by ‘the diffusion of chromium to-the surface. The chromium dlffusion coefficient32 in Hastelloy N Was:usedain?the calculation. | - _~An.extrapolation-of thc.graphs of?Fig} 43 shows that both fluorobo- rate.salts (500 ahd 1500‘ppmrokidc impurity)oarevmore‘aggreSSiVe than the fuel :salt to ‘the Hastelloy N- alloys. ‘Also, it appéarsfithatrideotical rates would ‘be found for the type 304L stainless. steelrfuel salt combination ' and ‘the titanium-modified Hastelloy N fluoroborate . salt (500 ppm oxide impurity). 74 It would be expected, if the solid-state diffusion mechanism controlled, that the calculated rates of Fig. 43 based on diffusion - coefficients would be identical to the-experimental rates. However, results of only the titanium-modified Hastelloy N —fuel salt system coincided with thercalculated values. Incidentally, fuel salts were used in the earlier experiments'that showed_that the rate was controlled by the diffusion coefficient of chromium in the alloy.’® Thus, we have shown that another mechanism besides chromium solid-state diffusion is controlling the process for several of the metal-salt systems investigated ~here. ‘The weight losses of the specimens exposed to the fluoroborate salt _ and the value calculated for 600°C and 4000 hr are compared in Table 15. It is noted that the experimental losses are greater than those calculated. As has been discussed earlier, the impurities in the fluoroborate salt appear to be very important because of the formation_of‘HF, which is strongly oxidizing and will.attack all the constituents of the Hastelloy N alloys. However, in the salt'containing 500 ppm oxide impurity the titanium-modified alloy showed smaller weight losses than the standard alloy, again because of its lack of iron. 1In the salt with 1500 PpPM oxide impurity more HF was available for attack and the weight losses were higher, since measurable amounts of nickel and molybdenum were removed from the alloy. An increase in nickel and molybdenum concentrations r-is often noted in the salt analyses under these conditions. The amount of attack is approximately doubled. Because of the large amounts of nickel and molybdenum in-the alloys very little differences in the weight losses of the standard and titaniumrmodified.alloys will be noted when ‘large amounts of HF are available for attack. It is also believed that this HF attack is responsible for the mixed kinetics noted earlier. %34, R. Grimes, G. M. Watson, J. H. DeVan, and R. B. Evans, "Radio- tracer Techniques in the Study of Corrosion by Molten Fluorides," PP. 55974 in Conference on the Use of Radioisotopes in the Physical - Setences and Industry, September 6—17, 1960, Proceedings, Vol 111, International Atomic Energy Agency, Vienna, 1962 L4 75 'fiTable”IS.' Effect of Impurities on Hastelloy N Corrosion by Fluoroborate Coolant Salt for 4000 hr at 600°C a ight Loss entration ¢ Material Weig Concentration Oxide (mg/em®) (ppm) Ti-modified Hastelloy N - f5.7 ) 1500 B Standard Hastelloy N 2.9 | . 500 - Ti—modified Hastelloy N - | - 1.9 . 500 Calculated ‘ "0,2 Using Doy = 2.0 X 10 15 szlsec and Cg = 0,07 - oo With reSpect-to'the deposition process, it is obvious that the calculated rate includes only the weight changes that occur from the ‘deposition on the surface. The amount of the diffusion into the metal is not measured. However, during the straight-line portion of deposition with ‘time, on which the kinetic calculations are based it was theorized that the rate at which the metal deposited is the same as the rate at which it diffuses into the container wall. In time the weight gains would continue, the amount of deposited material on the surface would hremain essentially constant, and a chromium gradient would appear in - the colder sections. All experimental findings do support these '.suppositions. ' The kinetic calculations for the weight—gain portion of the mass - transfer show a mixed control between solid-state diffusion and. solution rate. Analogous to the behavior in the hot leg, in the cold leg the . - weight gains could be first solution-rate controlled and later, as the 'process slows down, be controlled by the diffusion. Thus, the overall "effect would be mixed control As theorized earlier ‘it still appears fi‘that the action in the cold leg controls the mass transfer process for ‘,‘the entire system. In liquid metal systems, the simple approach to the mass transfer problem 15 the statement that the weight loss rate is a function of the difference in the solubility of the dissolved metal at the temperatures 76 of interest.“_The'proportionality constant may be relafed_to the diffusion coefficient of the metal in the.liquid if diffusion is the controlling .step or may be related to another factor concerning,Some process:occurring at the_inteffaee if the rate of solution is controlling. However, many additional ‘models have been required to explain all the experimental findings.v~The most ticklish problems involve the possible effect of oxygen on.the solubility of certain metals in”liqmid metals, the velocity effect; end the downstream effect. ‘Similar problems eXist‘in'the fused fluoride'selt systems. Inrthe'salt systems, we can make the.general statement that the corrosion rate in fuel salts is a functiom of the chromium and iron contents of the alloy. In the fluoroborate salts, the rate is a function of the alloy composition and the impurity level of the salt. The proportionality constant for_bbth cases is a function of temperature amd can be found experimentally. As in liquid metal systems, the rate controlling mechanisms are not entirely clear cut but are- functions of many variables. - Although the'corrosion ef these systems has been widely diseussed, ié should be pointed out that the weight loss ‘rates are quite low. SUMMARY Our data on mass transfer in the fluoroborate sait mixture points out that impurities such as water and.oxygen strongly affect theffluoride oxidation fiotential of this salt. However, we do not yet completely know the chemical forms takem by these impurities upon entering the salt. Corrosion rates show a marked increase following an air or steam leak into the fluoroborate salt. Once the leak has been detected and corrected, the'corrosien rete dfbps shafply. Analyses of tfie salt shows afi increase in oxide level as the corrosion fate'incfeases; the explanation for this behavior is that moisture reacts with the salt to produce HF and comprnds (oxides and hydroxides) that analyze as exides_by thexpresent techniques. The HF causes rapid oxidation of the containment material and is used up in the.proeess. The compounds that ensiyze as water and oiygen remain but have only a minor effect on cott6sioh rate. Kinetic eomsiderations showed that the mass transfer process is between solid-state diffusion 0 77 control and solution control, probably depending on the amount of impurities allowed to enter the salt. The mass transfer behavior is very similar to that seen for the chromium-UF, corrosion reaction, the difference being that other alloy constituents (beside chromium) participate in the process and.the,reaction rates are different. We have also shown the difficulty.in keeping the fluoroborate salt mixture free- . of moisture-type impurities but feel that this can be done by exercising sufficient care. Except for certain periods of the fluoroborate tests, the overall ‘corrosion rates have been relatively small, .Thermal-convection loop . NCL-14 has operated successfully for over two years and the pump loop MSR-FCL-1 nearly one_year with the fluoroborate mixture. Comparison of the rates experienced by NCL-13A, NCL-14, and MSR-FCL-1, all circulating the sodium fluoroborate-mixture,'indicates a velocity effect on the mass transfer (Table 16). We believe that the effect of velocity on corrosion in this system is a function of declining importance as the purity level of the salt improves. The impurity effect per se has been discussed in detail previously. Because of the’importance of the impurities on the corrosion of Hastelloy N by fluoroborate, efforts have begun on methods of purifying the salt. Also to be considered are better methods for analyzing ‘and identifying the impurities. At some purity level, perhaps 200 to 400 PP oxide as analyzed solid-state diffusion of chromium in the Hastelloy N will likely control corrosion, as it does in the fuel salts. This is most important, since it suggests that an entire MSBR of either the one~ or two-fluid variety can be operated with none of the | main circulating channels suffering more than a few tenths of a mil per year corrosion attack ‘Experimentalpproof of this is one of;our major near—term goals. T 78 able 16. Comparison of Corrosion Rates for Standard Hastelloy N in MSR Systems After More Than 5000 hr Operation s o Teosiey o, M o | | (°c) (°c) (mils/year) . NCL-13A Coolant® 605 145 0.1 0.6 NCL-14 - Coolantd 605 145 0.1 0.55 MSR-FCL-1 ~ Coolant® 588 78 10 1.2 NCL-16 .~ Fuell 705 170 0.1 0.04 NCL-15A Blanket© 675 55 0.1 0.03 NCL-18 - Fertile- 705 170 0.1 0.05 Fissile - 2NaBF ,—NaF (92-8 mole %), 1000 ppm oxide. bLiF~Ber—UF4 (65.5-34, 0-0.5 mole %), < 200 ppm oxide. CLiF—BeF2—ThF, (73-2-25 mole Z), < 200 ppm oxide. d1,1FBeF,~ThF,—UF;, (68-20-11.7-0.3 mole %), < 200 ppm oxide. CONCLUSIONS The corrosion rate is a function of the. alloy composition and impurity level of the salt. Because of the corrosive nature of the BF3 in an impure state and because it is present in appreciable quantities over the melt, leaks in mechanical portions of the systems can essily'occur. Kinetic considerations disclose that the weight change'rates are controlled by a mixture of solid-state diffusion and solution rate. The effective diffusion rate that controls the hot-leg attack is larger than the volume diffusion coefficients obtained in typical diffusion experiments. The mass transfer behavior (weight losses in hot leg, weight gains in cold leg, steady balance points) is very similar to that seen for the chromium-UF, corrosion reaction. 79 6. Titanium-modified Hastelloy N with less chromium and iron shows a greater resistance to attack in a fluoroborate salt with 500 ppm oxide impurity than standard Hastelloy N. 7. The corrosion rate of titanium-modified Hastelloy N in fluoroborate salt with 1500 ppm oxide impurity is double that found in salt with 500 ppm oxide. 8. Corrosion rates were relatively low for all systems tested. ACKNOWLEDGMENT It is a pleaSuré to acknowledge that E. J. Lawrence supervised ‘construction and operation of the thermal convection loops and was responsible for the weight change measurements of all the corrosion specimens. _We'also_recbgnize P. A, Gnadt and W. R. Huntley of the Reactor Division for operation and design of‘the pump loop. We thank A. P, Litman for his invaluable assistance through part of this work. We are also indebted to H. E. McCoy, Jr., J. H. Devan, T. S. Lundy, "and J. V. Cathcart for constructive review of the manuscript. Special thanks are extended to the Metallography group, especially -H. R. Gaddis, C. E. Zachery, H. V. Mateer, T. J. Henson, and R. S. Crouse .and_to the Analytical Chemistry Division, particularly Cyrus Feldman and H. W. Dunn, the Graphic Arts Department, the Metals and Ceramics. Division Reports Office, and particularly S. Peterson, for invaluable assistance.' (3) (10) (2) . “]. - J. J. S. 81 ORNL-TM-3866 - INTERNAL DISTRIBUTION - - (113 copies) . - Central Research Library | ORNL — Y-12 Technical Library Document Reference Section Laboratory Records Department Laboratory Records, ORNL RC ORNL Patent. Office _ MSRP Director's Office (¥-12) G. M. Adamson, Jr. o J. L. Anderson R. F. Apple C. F. Baes E. Beall S. Bettis F. Blankenship . G. Bohlmann o E. Boyd B. Briggs Cantor V. Cathcart B. Cavin C. Cole S. Crouse L. Crowley L. Culler E. Cunningham - M. Dale H. DeVan R. DiStefano J. Ditto W, D.‘éf,m LT P. Eatherly R. Engel | Feldman - E. Ferguson = - M. Ferris “ | P, Fraas, =~ = " H Frye,; Jr. R. Gaddis 0. Gilpatrick A. Gnadt J. Gray R. Grimes Grindell . Guymon Harms Haubenreich E. F. E. G. R. S. J. 0. N. R. J. F. J. J. S. H. W. C. D. L. (3) (10) R, E. .J. R. R. “Helms Henson Hill Huntley Inouye R. ’_}'J. . W. I‘. S. J. I. S E. L. R. Vol E. J. Kasten ‘Keyes Koger Krakoviak Kress Laing Lawrence Lundin Lundy . MacPherson Manning Martin Mateer MeCoy, Jr. .. - McHargue McNabb E. S. , L., H. L. McNeese Meyer - - Moore : Neill Nicholson = Patriarca = M. B. C'o Perry Pollock Savage . Scott L. Scott H. Shaffer M. Slaughter Sood. . - E. Thoma B. Trauger M. Watson - M. R. E. C. , Young P. Weinberg Weir, Jr. Whatley White Young E. Zachary 82 et —_— EXTERNAL DISTRIBUTION o - o' (38 copies) - i AEC DIVISION OF REACTOR DEVELOPMENT AND TECHNOLOGY, Washington, DC 20545 A. R. DeGrazia : ' : ° J. E. Fox ' ‘ N. Haberman (2) T. W. McIntosh J. F. Neff M. Shaw AEC DIVISION OF REACTOR LICENSING, Washington, DC 20545 (3) P. A. Morris AEC DIVISION OF REACTOR STANDARDS, Weshington, DC 20545 (2) E. G. Case .AEC DIVISION OF SPACE NUCLEAR SYSTEMS Washington, DC 20545 A. P, Litman AEC MSBR PROGRAM, Washington, DC 20545 (2) Program Manager | AEC OAK RIDGE OPERATIONS, P. O. Box E, Oak Ridge, TN 37830 Research and Technical Support Division : : ' ' - ~AEC-RDT SITE REPRESENTATIVES, Oak Ridge National Laboratory, P. 0 Box X, Oak Ridge, TN 37830 | D. F. Cope K. Laughon TECHNICAL INFORMATION CENTER, Office of Information Services, P 0. Box 62, - 0ak Ridge, TN 37830 (2) Manager (17) Manager (for transmittal to members of ACRS) UNIVERSITY OF TENNESSEE Department of Nuclear Engineering, Knoxville,_ TN 37916 H. G. MacPherson