TR e essenon 3 445k D3505Lc H ORNL-2338 UC-25 = Metallurgy and Ceramics INTERIM REPORT ON CORROSION BY ZIRCONIUM-BASE FLUORIDES G. M. Adamson R. S. Crouse W. D. Manly OAK RIDGE NATIONAL LABORATORY operated by UNION CARBIDE CORPORATION for the U.S. ATOMIC ENERGY COMMISSION ORNL-2338 UC-25 = Metallurgy and Ceramics Contract No. W-7405-eng-26 METALLURGY DIVISION INTERIM REPORT ON CORROSION BY ZIRCONIUM-BASE FLUORIDES G. M. Adamson R. S. Crouse W. D. Manly DATE ISSUED JAN 3 1961 OAK RIDGE NATIONAL LABORATORY Qak Ridge, Tennessee operated by UNION CARBIDE CORPORATION for the MARTIN MARIETTA ENERGY SYSTEMS LIBRARE e N 3 445L 0350512 y FOREWORD The corrosion data reported herein are an extension of the data given in the com- panion report Interim Report on Corrosion by Alkali-Metal Fluorides: Work to May 1, 1953 (ORNL-2337) but concemn zirconium fluoride—base mixtures, principally fluoride 30. The companion report should be used in conjunction with this one for background information, which will not be repeated here. The period covered by this report is approximately from July 1952 to June 1956. CONTENTS FOr@WOIT ...cccovisiossasssssssssssssssusassnasssiosess sinsunssisssssosnansnnasssissssssbsassss A DS ACT oottt ConClUSIonS .oy s R e T S Expaiimantol Methoths cicms i rimim i i s i P SR I T b v civR isean v bemab SR e s by Equipment .....ocovoivniieiciiciienan,, Procedure ....c.cooievnnnne Results and Discussion......cuiiiamnwmmisnasms Fluoride 30 in Inconel Loops......ccoivencrvnciinnicierire e Standard Loops ...... R R Dperating Time s Effect of Temperature.........cccoiiieciiiesiesicssisssssssssesessssasesessssnesnns NEMPOrature LD .. ovovsrres ramismsmssusssmesnsssssasssrmnssanssonsppaseenpassmiss st stass assiss Additives . OO0 WL v n N NN B — O — Ratio of Surface Areu 10 Lnop Volume SR R A e BB bR Oxide Removal Procedures .............. Heating Methods .................. Loop Size and Shape......c.cc.c...... e s e T A T e G SRR T TR A Change in Analysis of inner Prpe Wail Effect of Uranium Concenfration......c.......... Barren Fluoride Mixtures .............. Screening Tests of Possible Container Materials.......... Staindess Staals . onininn s iR SR BB VPO cmes e snansnesmas o msmontns sensos s HTeE e eSS FRS PR B WS S i Nickel .. Inconel X T S R S S PR G e A e TR Mnlybdenum und Ntoblum Hastallay B . .. ot i i e i Bt Special Alloys .............. Special Fuel Mixtures ................ Alkali-Metal-Base Mixtures Containing Trivalent Uranium...... Reactions and Mechanisms .....cccocovivvvevenvnnnnnn. W LI WK M O B — 00 =] L W ~ ~ 5588 E&&EED Ln O~ INTERIM REPORT ON CORROSION BY ZIRCONIUM-BASE FLUORIDES G. M. Adamson R. S. Crouse W. D. Manly ABSTRACT (rhe zirconium fluoride—base fluoride mixture NaF-ZrF -UF , (50-46-4 mole %), referred to as “‘fluoride 30," was circulated in Inconel thermal convection loops for periods varying from 500 to 5000 hr and ot a hot-leg tempera- ture of 1500°F. companying the circulation of zirconium fluoride—~base mixtures in Incenel, The purpose of this program was to develop an understanding of the corrosion mechanism ac- The attack in the Inconel loops was in the form of subsurface voids formed by selective leaching of chromium | from the alloy. After 500 hr of operation the voids were found to depths of about 10 mils, with the depth continuing to increase at a rate of abaut 4 mils per 1000 hr of operation. The effects of such variables as time, hot-leg temper- ature, temperature drop, fluoride purity, loop size and shape, and inhibitors on the depth of carrosion were studied. It wos found that the attack was reduced when a portion of the uranium wos present in the trivalent state, but the results were not reproducible. ] — Also, o few tests were carried out in loops constructed from nickel, stainless steels, iron, Hastelloy B, molybdenum, and nicbium. A limited amount of work was done on Inconel loops circulating alkali metal~bose mixtures (NaF, LiF, KF, UF,) with portions of the uranium in the trivalent state. Reduced ottacks were found, but disproportienation and production problems require additional study. CONCLUSIONS From the data presented in this report it may be concluded that molten mixtures of NaF, ZrF ,, and UF, may be circulated for several thousand hours in low-velocity nonisothermal Inconel systems. After 2000 hr of operation the corrosion depth in such a system will be deeper than desirable but will still not be excessive. A depth of about 0.010 in. will be reached in 500 hr; after this the depth will increase at a rate of 0.004 in. per additional 1000 hr of operation. The corrosion of the Inconel takes place by selective leaching of chromium metal from the hotter surface of a nonisothermal system, The removed chromium is reploced on the surface by diffusion from the center, but since no inward diffusion takes place, voids surface of the metal. The principal removal reaction is QUF‘i + Cr— EUF:3 + CtFy form under the accompanied by the reverse reaction to form chromium metal in the cold leg. With short operating times a large portion of the attack takes place by enough chromium being leached from the wall to reach the equilibrium concen- tration necessary for the mass transfer reactions to take ploce, It is known that other mass transfer reactions may occur when the uranium is not present, and these may also be of secondary importance when uranium is present, Neither the nature nor the location of the rate-controlling step has been determined. The diffusion of chromium to the hot-leg surfoce has been shown not to be controlling. With present production techniques for the fluoride mixtures, the nickel, iron, and hydrogen fluoride impurities are reduced to levels where they no longer play a vital part in the corrosion mechanisms; however, if they are present in the zirconium fluoride—base mixtures, they will cause attack in the loop hot legs. The most important variable controlling the depth of corrosion is the maximum wall temper- ature. The wall temperature is more critical than the maximum average bulk fluoride tempera- ture. While wall temperature is important, it must be a relative temperature; appreciable attack is found only at the hottest point in a loop no matter what the actual temperatures are, Variables of secondary importance are uranium concentration, loop size and shape, and temperature drop. The replacement of a portion of the UF; content of a batch with UF,, either in the production procedure or by the addition of reducing agents, will provide reduced depths of attack. The UF, has only a limited solubility in zirconium fluoride— base mixtures, so all the uranium cannot be present in the trivalent form. These mixtures are difficult to control during production, ond dis- proportionation of the UF; may occur, causing the formation of hot-leg deposits. While Inconel is on acceptable material for present reactors, it will not be adequate for future larger reactors. With zirconium fluoride— base fluoride mixtures at 1500°F, niobium, mo- lybdenum, and Hastelloy B show promise as future reactor materials. Some reduction in depth of attack may be obtained in alloys similar to Inconel but with reduced chromium concentration. The stainless steels did not plug with these fluoride mixtures, as they did in the case of the alkali-metal-base fluoride mixtures, but they still are not as good as Inconel. The addition of UF, fo alkali-metal-base fluoride mixtures also resulted in a reduction in depth of attack in Inconel loops. The solubility of UF, is higher in this system, but mixtures of UF, and UF , were still necessary. While this mixture shows promise for future use, production and disproportionation problems remain to be solved, EXPERIMENTAL METHODS Equipment Most of the loops used for this study were slightly modified from those originally de- veloped;! the modifications are discussed in detail below (“‘Loop Size ond Shape’). The expansion pot was reploced by a length of pipe of the same size as that used for the loop itself. The fill line and spark plug probe entered the sides of this pipe through Swagelok fittings. A typical loop is shown in Fig. 1. Most of the loops were constructed from IPS ¥%-in. sched=10 pipe, but, if this wos not available, IPS ]fz-in. sched-40 pipe was used as an altemate, No changes were made in the auxiliary equipment. Procedure The operating procedure remained essentially the same os that used for the alkali-metal-fluoride 6. M. Adamson, W. D. Manly, and R. S. Crouse, Interim Report on Corrosion by Alkali-Metal Fluorides: Work to May 1, 1953, ORNL-2337 (Mar. 20, 1959). work,! except for the addition of a cleaning step, which consisted in circulating another batch of fluorides for 2 hr before the loop was filled with the test mixture. All the molten fluorides used in this study were mixed and purified by the ANP Chemistry Section of the Materiols Chemistry Division.? The purification procedure included treatment with both hydrogen fluoride and hydrogen goses at elevated temperature followed by o prolonged stripping operation with hydrogen and helium, During the latter part of this study the batch size was increased, from 50 |b, to 250 |b; however, for handling and use, these batches were subdivided into 50-lb batches. The batching down was carried out while the mixture was molten, and the usual steps were taken to avoid contamination. Once a pot was filled, it was held under a positive helium pressure at all times. Helium gas was allowed to leak to the atmosphere whenever a connection was made or broken. During the actual transfer to a loop, helium pressures were maintained in both the loops and the fill pot. The rate of transfer was controlled by adjusting the difference between these pressures, The liquid level in the loop was controlled by the location of the outlet from the transfer line. When o signal was obtained on the spark-plug probe, which was located above this level, the helium pressure in the loop was increased and the excess liquid was transferred back to the fill pot. After the loop reached operating temperature, any excess liquid resulting from expansion was blown back to the fill pot. A sample of the original liquid was trapped in an enlarged section of the transfer line for chemical analysis. After the loop had operated for the desired time, the power was turned off and the liguid allowed to freeze in place. The loops were then sectioned as shown in Fig. 2, The material was melted out of the 2-in. sections in an inert-atmos- phere fumace; the pipe sections were sent for metallographic examination, and the fluorides for chemical analysis, Sections 1A, 2A, and 4A were stored in a dry box as reserve samples, The ends of the other sections were covered with tape., The loop sections were stored for three months; then, if they had not been used, they were sent to salvage. *E. F. Joseph et al,, Aircraft Nuclear Propulsion Fluoride Fuel Preparation Facility, ORNL CF-54-6-126 (June 1, 1954). UNCLASSIFIED PHOTO 21341 Fig. 1. Thermal Convection Loops. —— ve {-in. RESERVE METALLOGRAPHIC SAMPLE 2-in. METALLOGRAPHIC SAMPLE 4-in. DRY-BOX SAMPLE (CHEMICAL) 2-in. METALLOGRAPHIC SAMPLE 6-in. CHEMICAL SAMPLE, SPECIAL LOOPS ONLY UNCLASSIFIED ORNL-LR-DWG 48763 5| €-in. CHEMICAL SAMPLE, SPECIAL LOOPS ONLY 4| 2-in. METALLOGRAPHIC SAMPLE 2| 5-in.DRY-BOX SAMPLE (CHEMICAL) 3| 2-in. METALLOGRAPHIC SAMPLE f!f"e in, st N Fig. 2. Location of Metallographic and Chemical Samples. Standard loops. The amount of attack was determined by metallographic exomination. Uncertainties in the original pipes, in cleaning the surfaces after operation, and in determining losses from extesmol oxidation prevented the use of weight loss data. The depth of attack was determined by the depth of the deepest void formation found metallographi- cally in a traverse of an entire circumference of a section. Other than classifying the intensity as moderate, heavy, etc., no attempt was made to determine the amount of the attack. Typical sections of each type are shown in Fig. 3. Wall thicknesses were measured microscopically, and it was shown that no large, even surface remcval type of attack was taking place; however, smoll amounts less than the normal variation in pipe wall thickness would not be measurable. For UNCLASSIFIED T 5183 Fig. 3. Light; () moderate; (c) heavy, 250X. Reduced 29%. i ’ ’ 1 vy = - < - * o? ' 4 e - " ' e - . w | )- . / —{\“ { - " rg e & = s . e “ | T e o e J ) ok 4 ) " r'- 2 d' S o . s B L . " Leot] . g AT Ve - \l‘ Fe 3 i 3 o0b il o | T'T';*'p g ..-"‘A:h L B - r [ L& i, S — e e R el « : . bk e T e i - = - 3 - ¢ N | : . 1 - - _‘E_L.. Photomicrographs lllustrating Various Intensities of Attack as Reported for Fluoride Corrosion. each loop, at least two sections from the hottest area were examined to determine the maximum attack. The fluoride mixtures used in this study are referred to by number, identified in Table 1, and are RESULTS AND DISCUSSION Fluoride 30 in Inconel Loops Standard Loops. — A considerable number of loops were operated under what were considered to be standard conditions. These loops circu- lated a zirconium fluoride mixture for 500 hr with a hot-leg temperature of 1500°F and a tempera- ture drop of about 225°F, general groups, They fall into two The first group were operated UNCLASSIFIED T 5185 BRI o - . - 5 a i " . ! 3 — o k-] " {a) Table 1. Composition of Molten Fluoride Mixtures Fluecride UFai NaF ZrF4 Ne. Weight % Mole % Weight % Mole % Weight % Mole % 27 10.9 4.0 16.7 46.0 72.4 50.0 30 11.4 4.0 19.0 50.0 69.6 46.0 31 20.1 50.0 79.9 50.0 44 18.6 6.5 20.5 53.5 60.9 40.0 either as preliminary loops to outline the problems to be encountered with this mixture or as control loops for some of the early studies. These loops were not precleaned, and close control of con- ditions had not yet been established., The second group were run as controls for later studies, and all were precleaned. While some scatter in re- sults still exists in this second group, the reproducibility, especially during the fluoride production, was much better than was found in the early loops. The data for the first group are tabulated in Table 2 and those for the second group in Table 3, It is apparent from Table 2 that a variation in maximum depth of attack from 5 to 18 mils was from supposedly duplicate loops. With many of these loops the iron and nickel impurity concentrations of the original fluorides were also high, but the maximum depth of attack did not vary in a systematic manner with these concen- trations. Another cause of variation in attack was an undetermined amount of hydrogen fluoride left in the fluoride mixture from the purification process. This hydrogen fluoride would react with the loop wall, producing iron and nickel fluorides which in turn would reduce the chromium, When this group of loops were operated, no suit- oble method waos available for measuring the found hydrogen fluoride concentration. When a method became available for measuring the hydrogen fluoride content of the mixtures, it was found to be necessary to increase the stripping time after purification to reduce the hydrogen fluoride to acceptable levels. The control loops tabulated in Table 3 were those in which the controlled fluoride batches were circu- lated., While considerable variation is still found in the maximum depths of attack, it is not as great as in Table 2, The loops in Table 3 are tobulated approxi- mately in the order in which they were filled or in order of time. |f this table is divided into quarters and average moximum ottacks calculated for each quarter, it may be noted that the depths of attack are higher at the end, that is, that the attack is graduclly increasing with time, The average depth for the entire table was 9.1 mils, while the attack for the four quarters averaged 7.0, 9.2, 9.6, and 10.5 mils, respectively. As a comparison, the average in Table 2 was 10.1 mils. The poorer control of production with the large batches was responsible for some of the increase but not for all of it, since the trend was a gradual one, Neither the other causes of this gradual increase nor the cause of the occasional very deep attacks has been determined. The increased purging time and closer control in the second group have made the ottack more consistent but have not resulted in a large average reduction in maximum depth of attack. Since the average chromium content in the fluorides after operation of the loops in Table 2 is higher than after operation of those tabulated in Table 3, there oppears also to be a reduction in intensity of attack over the entire loop with the controlled batches. In an effort to determine whether air contami- nation during filling or operation could be re- sponsible for the variation in ottack, o series of loops were operated with fluoride batches which had been deliberately contaminated. Loops were operated under three conditions: (1) with helium pressure only during filling, (2) with no protective atmosphere combined with a small air leak, and (3) with a l-liter volume of air bubbled through the batch before transfer. The loop with no protective atmosphere and a slow leck developed an attack of 21 mils in 500 hr, which was double Toble 2. Data from Preliminary Control Loops? Fluoride Mixture Anclysis of Flueride Mixture i Eluoiide Barch Aftost Nickel (ppm) Iron (ppm) Chromium (ppm) Uranium (%) o Ne. No. tnnty Reptiei) e M DiEm - A0y B i See e 234 27 Lak Light 5 530 <30 4100 600 1460 1900 8.3 9.2 236° 27 Lab Moderate 10 610 <30 2900 1000 1070 2400 7.9 9.2 244 27 Moderate 9 140 <20 540 100 <20 1450 8.2 5.0 245 27 Moderate 12 340 <20 340 100 75 1500 8.4 9.3 2610 30 Lab Moderate 6 <20 150 950 273 30 Moderate 10 <20 60 860 175 160 1200 8.6 2.0 272 30 Moderate 16 720 30 520 200 45 1350 8.3 7.0° 283 30 d Heavy 15 1100 40 740 100 100 2800°¢ 4.6 5.1 280 30 EE-58 Heavy 8 85 <20 430 50 <20 1000 8.6 9.7 287 30 EE:50 Heavy 13 30 20 520 60 <20 1450 8.8 9.1¢ 282 30 e Moderate 9 40 <20 520 30 <20 1100 8.9 8.9 289 30 EE-59 Heavy 8 95 20 620 60 230 900 8.8 9.5 295/ 30 EE-63 Heavy 6 145 <20 625 30 <20 1100 8.8 9.4 296 30 EE-63 Heavy 8 40 20 470 45 90 1050 8.9 9.1 284 30 d Heavy 12 500 <20 1000 40 1450 1400 9.0 9.2 310 30 R-110/ Heavy 9 40 <20 265 50 <20 900 8.6 9.0 307 30 e Heavy 5 90 <20 235 30 <20 1100 9.0 8.8 298 30 EE-68 Heavy 9.5 40 <20 550 25 100 1100 8.6 8.9 343/ 30 EE-106 Heavy 18 75 <20 210 50 890 1500¢ 9.0 8.9 324% 30 EE-92 Heavy 15 60 <20 c 45 250 1300 8.6 8.9 345/ 30 EE-113 Heavy 6 40 <20 65 45 120 600 9.0 9.2 326 44 EE-114 Heavy 750 <20 500 65 250 1100 14.4 14.0° 381 44 EE-140 Heavy 10 <20 <20 175 120 50 700 14.6 12.0° 387 44 EE-145 Haovy 16.5 <20 <20 395 90 135 700 15.3 15.5¢ 10 (av) 1190 (av) PAll loops circulated o zirconium fluoride mixture for 500 hr with a hot-leg temperature of 1500°F and a cold-leg temperature averaging 1300°F, Loop treated with hydrogen. “Considerable spread was found in the individual values. 4p: melted. “Made in grophite. Loop cleaned with flucride mixture 31, Table 3. Data from Contrel Loops” Fluoride Mixture Analysis af Fluoride Mixture Lnnp EE Attack Na. Fluaride e — e i Nickel (ppm) Chremium (ppm) Iren (ppm) Uranium (%) HF® No Batch Intensity Depth (mils) ¥ b Before After Befare After Before After Before After 352 30 19 Moderate 5.5 <20 <20 50 450 110 60 8.8 9.4 353°¢ 30 e Heavy 3 <20 <20 50 600 130 75 8.8 9.1 3.6 3609 30 19 Heavy 4.5 <20 <20 100 500 100 70 8.7 9.0° 3.7 380/ 30 141 Heovy 7 <20 <20 95 700 185 90 8.7 8.9 382 30 155 Heavy 10 <10 <10 70 800° 65 90 8.6 8.6 2.2 383 30 150 Heavy 9 <10 <10 115 900 95 %0 9.0 B.5 2.1 384 30 150 Heavy 12 <20 <10 75 950 130 85 8.4 8.6 1.1 421 30 150 Heavy 5 <10 <10 100 800 110 300¢ B.8 8.6 3.0 435 30 160 Heavy 8 <10 <10 &5 400 40 70 8.6 8.8 2.5 440 30 158 Heavy g <10 <10 60 500 40 60 8.6 9.2 442 30 155 Heavy 10 <10 <10 80 800 150 60 8.5 8.6 4.6 4449 30 158 Heavy 8 120 <10 90 400 85 70 8.9 9.5 6.7 469 30 162 Heavy 8 25 15¢ 90 700% 55 70 8.5 8.8 0.9 462 44 203 Heavy 10 as <10 35 400 50 45 13.9 13.7¢ 2.8 463 44 203 Heavy 10 i5 <10 30 475 50 45 13.2 13.5° 2.5 540 30 188-1 Heavy 1 10 15 40 520 30 80 8.5 8.6 2.1 545 30 198 Heavy n 40 <10 70 400 65 70 9.5 8.8 1.7 571 30 232-8 Heavy 5 25 25 55 900¢ 165 e B.6 B.6 554 30 188-7 Heavy n 30 <10 80 600 90 60 8.5 B.6 2.7 570 30 232-8 Moderate 9 20 25 75 850 70 100 8.2 8.8 1.2 577 30 228-12 Heavy 9 40 40 45 850 85 80 B.9 9.2 0.5 586 44 239-3 Heavy 12 6 <10 50 800° 45 90° 14.0 13.7 0.5 607 44 241 Heavy 12 50 50 65 700 95 80 14,0 14.3 0.1 608 44 241 Heavy g 25 50 50 700 80 75 14.1 14.1 0.1 609 30 223-1 Heavy 7.5 <1 20 50 e 145 40 8.9 9.1 0.7 610 30 223-1 Heavy 1 <1 20 50 1000 145 50 8.9 8.7 0.7 651 30 513-4 Heavy 9 10 e 110 800 75 70 8.3 8.4 1.0 684 30 246-1 Heavy 11 15 10 55 200 255 50 8.9 9.1 1.2 685 30 246-1 Heavy 14 5 5 60 700 30 40 8.9 9.0 1.3 686 30 246-1 Heavy 9 15 10 80 775 60 15 8.8 9.0 1.5 698 30 434-R Moderate 10 9 10 40 800 105 55 B.6 8.9 1.5 700 30 246-4 Heavy 13 5 25 45 625 120 65 8.9 9.1 1.9 2.1 (av) 675 (av) %A1l loaps circulated a zirconium flucride mixture for 500 he with a hot-leg temperature of 1500°F and a cold-leg temperature averaging 1300°F, bRelative readings on Solubridge ofter bubbling 2 liters of helium through bath and then through boric acid. “No cleaning. dNu trap. “Individual results vary, ICurved loop. —— e R e e e e e ——E that found in the control loops. The other two loops developed maximum depths of attack within the usual spread. However, in loop operation there is only a remote possibility that develop- ment of a leak and failure of the protective gas system would occur simultaneously. When the gradual increase in attack became apparent, loops 684, 685, and 686 were filled on the same day from the same batch of fluorides. The operators were watched and were extremely careful to avoid any operation that could permit contamination, The attack in 500 hr in these loops was 11, 14, and 9 mils, respectively. Since a variation was still found, these loops indicated that the trouble was not from careless operators or from the batch, Changes in the batches could have caused the grodual increase but not the wide and unpredictable variations. As yet, no satisfactory explanation has been offered for these variations. To illustrate the distribution of attack around the loop and to support the procedure of basing conclusions primarily on the sample from the top of the hot leg, loop 833 was sectioned in con- siderable detail as shown by Fig. 4; the metallo- graphic data from this loop are given in Table 4; Fig. 5 shows typical photomicrographs from the various loop sections. These data confirm the fact that the maximum aottack was found at the top of the hot leg. Very little attack was found in the hot horizontal leg, and a gradual increase in depth was found in moving up the vertical hot leg. This increase was more gradual than was found in o similar examination of a loop in which alkali-metal-base fluorides were circulated, but otherwise the two loops were similar.! The attack depth decreased rapidly above the heated area, and this will be discussed in more detail below (‘“‘Effect of Temperature'’)., The three samples from under the upper half of the top heater checked each other very well, Operating Time. — Since thermal loops have no moving parts or mechanical seals, they are ideal for long-time operation. The limiting feature in such loops is the life of the heoting elements, which can be replaced during operation only with Table 4. Metallographic Examination of Sectioned Loop 833 Section Mo, Metallegraphic Notes 1 Heavy surface pitting to 0.5 mil with heavy intergranular subsurface veoids to 12 mils 1A Heavy surface pitting to 0.5 mil with heovy intergranular subsurface void formation to a depth of 11 mils 2 Heavy surface pitting to 0.5 mil with heavy intergranular subsurface voids to 9 mils m N O ot B W Light, shallow surface roughening Moderate surface pits to 0.5 mil Light, shallow surfoce roughening Moderate surfoce pits ta 0.5 mil Some as section 6 Moderate to heavy general subsurface void formation to a maximum depth of 5 mils 9 Heavy surface pitting to 1 mil with intergronulor subsurfoce voids to 6.5 mils 10 Some as section 9 11 Heavy surface pitting ta 1 mil with heavy general subsurface voids to 3 mils 12 Heavy surface pitting te a depth of 1 mil UNCLASSIFIED ORNL-LR-DWG 48764 & 7 Sin __ © T-8870 (8] . ~ {in. 8 1al fin. T-8871 i r-ss7z|'| 5™ T 1 4in, 2in. f .L 5] T-8880 T-8873|2] 2in. * 5in. Y T-8874 (19| in, : 6in J 4] T-8879 T-8875 (101 lin. ; j 4in 5IIT'I. | : 3] T-8878 : T © 12 M ¢ b &z = | P N M 1 %‘.l i T M - Fig. 4. Location of Metallographic Sections from Loop 833. 10 *%GE pesnpay *gfg dooT punody 3ooiy Ul UONDMDA °G 614 n __ = Lo s | o - = o = . b= x 2l o] V| wi0 *i o] [=ie] | 9989944 4 : NG ;_ "1 i, _n....f . ; 3 . e s .p/\?-.\.d.-r.,.ru. » ..r ._._ I SIHINI i FLE0-1L AISISSYIONN SLEB-L aIHISSY TINN " e o t T Ead “.fl_.fl.. wid i Cd ° i . - m_“ — i m 3 . !/ % . T £ 3 ’ o~ ‘. ] .‘ % T - N . g = = . " l‘.a - - N o i ,...u._ ..fi ’ - NO ” 11 3 e i e o 2 =1 5 o o =T —T - m Liee-1 n FLEE-L @ oSSy IINA Q3 HIS5% TINN ha m I“.. y TWL - > — - g0 a_nl_ - ™ -\ 7 Tio | €0 v ’ . . ., e = 1 T0 e - ® . - r o o i e a 2! ) . ‘ ' - e o 9288 | R - e - @ e 1 rrral 4 s @ 1 ’ i = ' .. r - U e . . 6188 Gl ). - . - . % 5 = » ’ I ' y i ik 4 6.88 |t mond Jo s r e - » # . - 5| - e b it ~ NOILD3HIO hese] .-y 2 : v & - w rr =1} ~ .’ o b * » P 88 |z ] z 2 : . & fiwi Bi8E-1 X -~ s L ( m T @ 1a3HESYIDNN & . LT QIASSYTIONN rn 3 1 K ~, o £.B8 |2 L i o il O vig | - v eps . 0ses8 i ) 3 . 2.8871 1 . . e8| i ¢ 5 _LI¥i v T8 ¢ v ; 0.88 e ) % [0 = ", s . . % . > nnnm_ P g . oo . ¢ . X T 550" ] \ (550" o] - [Te] - 3 T o - # oo | ..\-v L_..r. o : 555" =] - i 7 Thftyl i 2 % ', Wl ey pol e :m 3 kiR L 84 _.l.-.,._.fl_dm ¥ ol X ISR T % 2] -u IR tb.:.n..._. ..firw..._...._....l...f..fl..,..._.m...si_w. . ..d L] ml uees a3 IssYIDINN .h. SO AL RN A T T e L Q3NISs¥ 1IN0 R ~ ru._ TM._ = * r Wi | . OrR X © fi - ¢ ’ ; Y oo s w0 i " .' - ®. ’ 410 70 o . : "o g -. 7 el e -~ 8 N - — | N 1T o a1 - & j Yy . = ., T - ? - o o0 555" : “-n i ? .o . ¥ - 169 r oo - 1 . - ¢ - . " . $. ° 5] ] 4 w3 - . sk - x '. ® S a | | = - I s " ‘l‘j i 71 W > b = | 3] ¥ g L . e | o ] & 0988~ | o i Ve Y o e 5 1488 @ |a3l4ISSYIINN e .- . o a3 AISSYIONN r ra fml A Im.u > - (oo | ; s ¥ ! e e 110 - g E Tia | Tio | e & - - - i . o] s T . “ » . . 2 -~ : 1 oY _ . - n.\ - i { ‘ e * ® : . P J o] z 0o T".n 0LEE-1 ©@ 1a31I55YIINN S9L8P OMO-HT-TNYO Q3141SS¥ITINN 11 some difficulty., In this program a considerable number of loops were operated for 1000 hr or longer, To determine how the maximum depth of attack varied with operating time, several series of tests were made with various fluoride batches. These data are tabulated in Table 5 and plotted in Fig. 6. As shown on the plot, considerable spread was apparent in the data; however, if they are separated into the various fluoride production batches, a series of nearly parallel curves is found. In each case a few points do not fall on the curve; these are thought to be another ex- ample of the unexplained occasional wide varia- tions found in the control loops. With all these curves, after about 300 hr the increase in depth of attack is linear and is at a Table 5. Effect of Operating Time on Corrosion Attack EE 0 ti Locp PBTQ e Attack Average Cr Content . Batch Time ) Fluoride Mixture No. No. (hr) Intensity Depth (mils) (ppm) After Operation 263 30 100 Moderate 4 1000* 27 from 50-1b batch containing 720 ppm Ni 264 30 500 Modercote 9 1300 27 from 50-1b batch containing 720 ppm Ni 265 28 100 Moderate 5 1000 27 from 5-1b batch containing 1380 ppm Ni 266 31 500 Moderate 15 1800 27 from 5-1b batch containing 1070 ppm Ni 267 100 Moderate 3 200 27 pretreated 268 500 Moderate 13 1750 27 pretreated 274 35 1000 Heavy 11 1400 30 297 68 10 Light 1 230 30 300 68 50 Meoderate 2 575 30 305 68 250 Heavy 9 1200 30 298 68 500 Heavy 9.5 1100 30 306 68 1000 Heavy Q 1400 30 299 68 2850 Heavy 18 1500* 30 318 88 100 Heavy 4 900 30 317 88 350 Heavy 8 1200 30 345 113 500 Heavy 6 600 30 327 113 1000 Heavy 10 550 30 328 113 2000 Heavy 7 250 30 329 113 3000 Heavy 23 600 30 344 113 5000 Heavy 27 350"~ 30 381 140 525 Heavy 10 44 394 149 1000 Heavy 17 44 12 Taoble 5 (continued) EE Operating Attack A Lo Batch Time verays Cr Conta‘nt Fluoride Mixture Ne. b, (hr) Intensity Depth (mils) {(ppm) After Operation 459 162 500 Heavy é 550 30 plus 0.2% ZrH, 413 150 1000 Light 7 550 30 plus 0.2% ZrH, 414 150 2000 Heavy 11 85 30 plus 0.2% Zer 340 113 3000 Heavy 13!5 300 30 plus 0.25% Zer 431 160 10 Light 2 300 30 432 160 50 Heavy 3 400 30 433 160 100 Moderate 4.5 425 30 434 160 250 Heavy 8 600 30 435 160 500 Heavy 8 400 30 436 160 1000 Heavy 14 700 30 445 160 1500 Heavy 13 525 30 437 160 2000 Heavy 16 600 30 450 160 2500 Heavy 11 600* 30 448 173 1000 Heavy 10 600 30 540 188-1 500 Heavy 11 520 30 535 188-1 1080 Moderate 10 475 30 501 188-7 100 Heavy 3 700 30 554 188-7 500 Heavy 11 600 30 463 203 500 Heavy 10 475 44 464 203 1000 Heavy 15 450* 44 570 232-8 500 Moderate 9 850 30 567 232-8 1000 Light 9 800 30 568 232-8 2000 Moderate 12 800 30 *Wide wvariation in individual values. rate of about 4 mils per 1000 hr, A single loop rate of mass transfer is not strongly dependent was operated for 5000 hr and the relationship still held. This continuing increase in depth of attack with operating time is attributed to mass transfer. It is obvious that if this variable is to be studied in a thermal loop, operating times of from 1000 to 2000 hr are a necessity. Since the curves for the different batches are parallel, the upon the original batch. A series of photomicro- graphs representing various test durations are shown in Fig. 7. |t may be seen that the voids grow in size with time and show a tendency to concentrate in the grain boundaries, The rapid attack during the initial stages was sensitive to the nature of the batch. The attack UNCLASSIFIED ORNL=-LR-DWG 4B87E6E 3‘3' | 1 O IMPURE -68 AND 88 (4mils/ 1000 hr) X 160 (4mil=/1000hr ) TO 5000 hr e |3 AT 27 mils A 188-198 b s 25 ————— = — AVERAGE 5 sl | o 232-8 > T @ Y-in. TUBING * e [ A ZTHE'O-E‘:;C' S | 20 - il /‘_,-/ z : pr z /..-" E | 1( / 5l | 1 i ‘ o -1 > | r...a-"‘ w L1 ‘_‘...-*‘- 7 a ~ Jo= " N 10 =¥ —— (e} f _._..-"";HE f? _'___4—-""1: » L < i 0 0 500 1000 1500 2000 2500 3000 3500 4000 4500 OPERATING TIME (hr) Fig. 6. Variation in Depth of Attack with Operating Time. during this time was caused by the system crystals found to explain all the attack. The reaching equilibrium; it was affected by such things as impurities from any source and by the dissolving of sufficient chromium to reach equi- librium concentrations. An interesting point in the tests was the average chromium content found in the fluorides after operation of the loop. While the depth of attack increased with time, the chromium concentration in the fluorides did not. The concentration was established during the first stage of the attack and then remained constant. In some long-time loops, metallic chromium crystals were found in the trap at the bottom of the cold leg. This deposit appeared first as a ring made up from fine dendritic crystals sticking to the trap wall just above the liquid-solid interface found in the trap. It continued to grow until an entire disk was formed, as shown in Fig. 8. In many 1000- and 2000-hr loops, no crystals at all were found, and in none of the loops were sufficient metallic trap deposits were identified as metallic chromium by x-ray diffraction and spectrographic studies, With the depth of attack varying with time as shown in Fig. 6, diffusion of chromium within the pipe wall could possibly be the rate-controlling step. |f this were true, the depth of attack would be independent of the solutions and so should be the same after circulating two fluoride batches for 250 hr each as after circulating one batch for 500 hr. However, loop 281 developed a ftotal depth of attack of 17 mils with two batches circu- lated for 250 hr each, while loop 280 developed an attack of 8 mils with a single 500-hr circulation of another portion of the same batch, Additional evidence that the controlling step is not in the wall is provided in loops 333 and 334, Loop 334 circulated a batch that had been previously circulated in loop 333. An attack to 5 mils was found after 500 hr in loop 334, while in loop 333 an attack to 14 mils was found after only 240 hr. Gl UNCLASSIFIED UMCLASSIFIED 1 T-9195 T-3701 PRk ReRE TEs: pasl o e - oot ! . s I {\ S [20R ’ . i -, - '.r‘ 3 e s » 1.-!“h -"/ 20H g J." tir'l" W 202, - - w = - “VgT | @ € - & - _'i‘_' = ‘:- " _g_ 2 | | 2 s L {::“-\ . - - ___I [ el _ stk h . S S m : o o0 I o = ; R ot P = 008 "\\'“ < % T Yir. 008 "——1 S ;j \ . ."; L - ‘ . -2, _:‘ &, . ! L.00s . f - - o 010 | I . 010 - S A {'I.l" 9, Lot . I/ [+ fASS o1l 5 & ’ J '\ -, ’ LOIE : —_— - { ol2 _'- \ & - \"., -“'!T g 5 f - 013 | —, ; 013 ‘ [ \ 1 {{]} r‘ o4 [b} \ H_/ . A ole UMNCLASSIFIED UNCLASSIFIED T.5194 i = 002 | |00z | .003 .003 w ol w w b L x o o = = D08 D06 5 2 .007 | . s |oor ® oos| @ . & 4 o W oo8 - i - 1009 & : e " . * 009 ® & . gio 2 010 T S - , - . I ol 5 ol =P - 0iz it ® oIz 013 ..) o'l h ‘ ° i 214 | (d) e . a ! 014 Fig. 7. Effect of Operating Time on Attack. (a) 10 hr; (k) 100 hr; (c) 1000 hr; (d) 3000 he. 250X. Reduced 17%. UNCLASSIFIED T-5211 Fig. 8. Metallic Deposit Found at Top of Cold Trap from Loop 344, The attack in loop 334 was higher than would be predicted from the curves, but the original batch was of poor quality, and the 240 hr in loop 333 does not appear to be sufficient time for it to reach equilibrium. When the fluoride mixture was transferred to loop 334, 225 ppm of iron was still present. Other evidence that chromium diffusion is not the limiting factor during the longer operating times is presented below (‘‘Change in Analysis of Inner Pipe Wall*). These tests show that while the chromium is continually removed from the inner surface of the pipe wall, the wall is not being continually depleted. Chromium is removed fairly rapidly during the early stages of rapid attack, but with the decrease in removal rate the chromium diffuses to the surface as fast as it is removed and possibly even at a slightly higher rate. If the chromium diffusion rate was the limiting factor, the surfoce would be continually depleted, or at least remain constant, instead of the concentration evening out. Effect of Temperature. — For all this work, the loop operating temperature was measured from thermocouples spot-welded to the pipe wall several inches above the top heater. As long as the temperature is measured in the same manner and at the same spot in all loops, the measure- ments are comparable, but they may be only relative. The development of the methods of installing the thermocouples and the choice of locations were discussed in the previous prGrt.1 With the techniques used on the loops several errors are known to be present. The temperatures were measured on the outer pipe wall surface above the heaters. While the bulk fluoride temper- 16 ature would change little, if at all, in this dis- tance, a difference in wall temperature would be found. Under the heaters the driving force is inward, and therefore the walls are hotter than the fluoride; above the heaters, the reverse is true. The temperature of interest is the maximum temperature at the interface between the liquid and the pipe wall, but for practical reasons it is the temperature of the outer wall that is measured, and this at only a few spots. Thus there is no assurance that the maximum interface temperature is being measured. Two attempts were made to determine the maximum wall temperature under the heaters, Thermocouples were attached to the walls of these loops by various methods, and several different radiation shields were provided for the thermocouple beads., In addition, several thermo- couples were buried in the wall, As was ex- pected, a spread in temperatures was obtained, but for both loops the maximum wall temperature appeared to be between 150 and 175°F above the measured standard loop operating temperature. This temperature difference was larger than was expected, but since so much work hod already been done with the temperature above the last heater used as the so-called maximum, it was decided to complete the work without making a change in the technique. The thermal loop data are suitable only for direct comparison to give trends and for studies of mechanisms and are not intended as reactor or loop design data; therefore such o decision was reasonable, When this work was started, considerable discussion had occurred as to what temperatures should be measured and how. Considerable difficulty had been encountered with failure, by breaking loose, of thermocouples under the heaters; installation there was more difficult, and such failure could easily produce high values. For these reasons, it was decided to measure the temperature just above the heaters. |In retrospect this appears to have been a mistake, since wall temperature has now been shown to be the most important variable aond the error in measuring with this method is larger than was thought; so it would now appear desirable with each loop to actually measure the temperature under the heaters. The data showing the variation in depth of attack with variations in hot-leg temperature are tabulated in Table 6 and summarized in Table 7. Table 6. Dota from Loops Operated with Various Hot-Leg Temperatures Loop EE Batch Hot-Leg Temperature Temperature Drop Operoting Time Attack No. No,? (°F) (°F) (hr) Intensity Depth (mils) 275 35 1650 125 500 Moderate 11 273 22 1500 500 Moderate 10 286 59 1300 175 500 Heavy 9 287 59 1500 500 Heavy 13 289 59 1500 500 Heavy 8 288 59 1650 500 Moderate 12 314 73 1250 500 Heavy 3 318 88 1500 100 Heavy 4 Ky} 88 1650 150 100 Heavy 7 350 113 1200 150 500 Heavy 3 345 113 1500 500 Heavy 6 390 141 1200 140 500 Heavy 5 380° 141 1500 500 Heavy 7 439 158 1300 160 500 Heavy 4‘6 440 158 1500 200 500 Heavy 9 441 158 1650 150 500 Heavy 8 537 188-1 1200 170 500 Heavy 6 538 188-1 1300 170 500 Heavy 4 539 188-1 1400 200 500 Heavy 5% 540 188-1 1500 200 500 Heavy 1 562 232-8 1200 150 500 Heavy 7 575 232-8 1200 150 500 Heavy 8 563 232-8 1350 185 500 Heavy 6 570° 232-8 1500 500 Moderate 9 5719 2328 1500 500 Heavy 5 586 239-3° 1500 200 500 Heavy 12 581 239-3° 1600 215 714 Heavy 14 351 119 1650 120 1000 Moderate 8 391 141 1200 150 1310 Heavy 7 392 141 1400 180 2000 Heavy 8 393 141 1600 185 1900 Heavy 13 449 158 1250 160 2000 Heavy 6 533 188-1 1250 165 1060 Moderate 4 534 188-1 1350 175 1125 Heavy 5 535 188-1 1500 215 1080 Moderate 10 536 188-1 1600 220 1150 Heavy 9 578 239-3€ 1200 170 1500 Heavy 8 5772 239-3° 1350 190 1500 Heavy 5 583 239-3° 1350 200 1500 Heavy 5 584 239.3° 1500 220 1500 Heavy 15 580 239-3° 1500 1500 Heavy 12 585 239-3° 1600 250 1500 Heavy 18 Elyoride 30 except where indicated. bCurvad locp. c ].5 in. sched 40. dy .fi in. sched 10, ®Fluoride 44. Table 7. Summary of Effect of Hot 013 E PR ‘ . . (g) oi4 (b} T . - r o4 | UNCLASSIFIED UNCLASSIFIED[ | 001 T-4487 | 001 002 | | .00z | o o =] o Q {7 | b ¢ | T T INCHES 1 1 .Y ® 8 F_ N e ® o I IHCI;!EE s @ \ > ot AN Vi i . . 00T N f . \. q W [ocor = - e ——— o - ; o8 . & ‘: 008 o0 | [LOI2 ) : ’ 6 - o2 - T a \ - £ 1 (¢) _ . sl () o T Fig. 10. Changes in Appearance of the Attack with Increasing Hot-Leg Temperature. (a) 1300°F; (b) 1400°F; (c) 1500°F; (d) 1650°F. Note how voids grow in size. 250X. Reduced 17%. would be found in bulk fluoride temperature between samples B and 1A; however, some differ- ence would be found in wall temperature. The depth of attack has been reduced by one-half in this 1-in, distance. This rapid decrease in attack was confirmed with loop 268. The top of the hot leg of this loop was sectioned longitudinally, and the attack decreased from a heavy 12 mils to a light 4 mils in a distonce of 0.75 in. In another attempt to show the correlation between depth of attack and heater location, loop 456 was operated with an unheated, but well-insulated, area near the top of the hot leg. The maximum depth of attack in the unheated area was 3 mils, while attacks to 9 mils were found in the heated areas on both ends of this zone. The change in depth was quite rapid at both ends. The effects of the unheated area must be complex, since the 9-mil attack below the gap was deeper than would be found in a comparable area in a control loop. It is possible that the gap in the heaters had an effect upon the forced convection currents found by Hamilton,® These currents could affect both the laminar layer and the wall temperature. The importance of location is also shown clearly in another loop. Loop 259 was part of the temperature study and had an Inconel thermocouple well suspended in the middle of the hot leg. Fluoride 27 was circulated in this loop, and after 500 hr a typical attack to 8 mils was found on the loop wall. In a section directly opposite this area, the thermocouple well showed only a light attack to @ maximum depth of 1 mil. The distance between the thermocouple well and the loop wall was only 0.2 in., but all the attack had been concentrated on the hotter surface, All these tests illustrate the fact that the maximum wall temperature is a more critical variable than average bulk fluoride temperature is, When a temperature difference is present between two surfaces, the attack will be concen- trated on the hotter surface, even if the colder one is at a level where attack is normally encountered, Temperature Drop. — The thermal loop is not ideal for studying the effects of temperature drop in @ dynamic corrosion test, Any change in 3I:J'. C. Hamilton, F. E. Lynch, and L. D. Palmer, The Nature of Flow of Ordinary Fluids in a Thermal Convection Harp, ORNL-1624 (Feb. 23, 1954). 20 temperature difference between the two legs is also a change in flow driving force, and so must be reflected as a change in velocity. With a standard loop design and heater location, any change in temperature drop requires a change in heat input and therefore a change in maximum wall temperature, Since in a thermal loop the Reynolds number is already very low, it was speculated that the changes in flow would not have a big effect on the laminor layer and there- fore that changes in temperature drop could be studied, ot least in an approximate manner. |t has now been shown in the pump loops that velocity is not a critical varioble, The varia- tions in wall temperature would, however, still cause errors, For this series of tests the temperature drops in the loops were varied by placing various amounts of insulation on the cold leg. An attempt was made to increase the drop by air-cocling the cold leg, but with the standard loop design, sufficient heating was not available to maintain the hot-leg temperature at 1500°F. The long-time loops in this series were terminated by a building power failure after they had been circulated for varying periods. For comparative purposes, small corrections were made in the depths of attack to compensate for the variations in time. In all cases the corrections, and therefore any errors, were smaller than the normal spread in results. As shown by the data tabuloted in Table 8 and summarized in Table 9, a decrease in depth of attack was found with smaller temperature drops. This was found with both the short- and the long- time loops. While these dota seem to show an effect, they cannot be considered as proof, since with the lower temperature drops, lower wall temperature would be found. This would also reduce the attack, and the two variables cannot be separated from these data, |t was shown previously that while wall temperature was critical it was relative in that a temperature drop was also necessary, mation that increased temperature drop will cause increased attack. A conflict is apparent in the data in Table 9. An increase in attack was found for each time series with an increase in temperature drop; however, the increases were independent of time, Thus these data indicate that the depth of attack does not increase with time and therefore conflict with the results presented above (‘‘Operating This would be indirect confir- Table 8. Data from Loops Operated with Various Temperature Drops Average Chromium _ EE Batch Goldilsg Temperature Dperating Sy S Attack Ne. Ne. Condition thop it Operation : Depth (°F) (hr) Intensity : (ppm) (mils) 422 155 Wrapped 130 500 750* Heavy 5 442 155 Control 200 500 800 Heavy 10 546 198 Insulated 70 500 400 Heovy 4 547 198 Wrapped 140 500 400 Heavy 6% 545 198 Control 200 500 400 Heavy i 423 155 Wrapped 140 2000 650 Heavy 6 447 173 Insulated 65 1000 350 Moderate 4 448 173 Control 200 1000 600 Heavy 10 446 173 Cocled 220 1000 500 Heavy 12 510 198 Insulated 85 1820 450 Heavy 6 (5))** 530 198 Wrapped 130 1640 1700* Moderate 7 (7)** 531 198 Control 220 1400 350 Moderate 9% (10})** 532 198 Cooled 235 1410 475 Heavy 10 (1)** 677 246-1 Insulated 70 14 675 Light 3 678 2461 Insulated 95 50 700 Mederate 2J 679 246-1 Insulated 100 100 750 Moderate 4 680 246-1 Insulated 70 100 725 Moderate 4 682 246-1 Insuloted 80 500 800 Moderate 5% 684 246-1 Control 220 500 900 Heavy n *Wide variation in individual values. **The attacks in the loops in this series were adjusted to 1600 hr; the corrected values are shown in parentheses. Time''), This conflict cannot be blamed on batch variation, since portions of both EE batches 155 and 198 were operated for different times with the same temperature drop and in both cases this independence with time was noted. A series of loops with insulated cold legs were therefore operated for varying times, The data for this series are also tabulated in Table 8 and do show an increase in depth with time. The attack in the 500-hr insulated loop is only half that found in the control loop, confirming the data reported above. The fact that 3 mils attack was found after only 14 hr with a low temperature drop shows that prolonged cleaning times should be avoided. Additives. — |n the discussion on the control loops it was pointed out that the depth of attack seemed to increase with the presence of iron or nickel in the fluorides, but because of other variables a definite correlation could not be 21 Table 9. Summory of Changes in Depth of Attack with Variations in Temperature Drops Operating Time Depth (mils) of Attack at Temperature Drops of (hr) 65 to B5°F 130 to 140°F 200 to 220°F 1 500 4%, 5% 5,6, 6% 10, 11, 1 1000 4 12, 10 1 1 1600 54 7 10%, 11 2000 & made, The wvariation of attack with iron and nickel concentrations was also noted in the original alkali metol work.! To establish a better correlation, several loops were therefore operated with controlled odditions of nickel fluoride. The nickel fluoride was added to the batch in the tronsfer pot in order to produce a more uniform distribution, While the batch was being agitated with helium, some of the nickel fluoride reacted with the pot walls, resulting in smaller additions than had been planned. A Micro Metallic filter was used in the transfer line to prevent transfer of any metallic nickel. The data from these loops are tabulated in Table 10, From these data, on increase in depth of attack over that of a control loop was found in each of the 500-hr loops to which nickel fluoride had been added. Since the amount of nickel actually added is not known, an exact comparison cannot be made, but the depth of attack does increase with increasing additions. The increases ore not as large as was expected, with an addition of 1000 ppm of nickel fluoride causing a doubling in the depth of attack, Since some of the chromium in these loops was picked up in the charge pot, the fluorides were partially saturated before being circulated. To compare these loops with control loops the hot-leg attack would have to be increased sufficiently to balance the chromium picked up in the pot. With present production techniques for the fluoride mixtures, the varia- tions in nickel and iron found in the original batches are not large, and fram this work it does not appear that they are the cause of the erratic attack. Typical hot-leg sections from one of these loops and from the 500-hr loops with the other additions discussed below are presented in Fig. 11, Portions of the same fluoride batch 22 were used in loops 454 and 455 to determine the effect of nickel fluoride on the rate of mass transfer, The longer-time loop developed an attack of 16 mils, which is an increase of only 1 mil in 1335 hr; however, some increase was found in the number and size of the voids, While it is possible that nickel fluoride will not increase the rate of mass transfer, it should react to produce chromium fluoride, which is shown below to increase this rote, No satisfactory explanation has been found for this inconsistency. Typical hot-leg sections showing the effect of this addi- tive and others on mass transfer are shown in Fige. 12, Additions of both chromic and chromous fluoride were made in a manner similar to that described for the nickel fluoride. The data for these loops are also tabulated in Table 10, 1t has been shown by work in the Materials Chemistry Division that when CrF, is odded to fluoride mixtures with a zirconium fluoride base it is reduced to CrF,, the latter being the only valence state that has been found in such mixtures, The addition of CrF, should reduce the attack found in the first stage by providing some of the chromium required to reach equilibrium, With CrF, this effect would also be found, but it would be offset by an in- crease in attack caused by the extra fluorine atom, The data in Table 10 do show a reduction in depth of attack when CrF, was added to a loop operated for 500 hr, However, in a loop operated for 1710 hr the depth was slightly deeper than was found in control loops operated for similar times; thus, since the depth at 500 hr wos re- duced, the rate of mass transfer increased. For these loops the data would indicate a rate of 1 mil per 100 hr rather than the 0.4 mil found with the control loops. Table 10. Variation in Depth of Attack with Impurity Additions Loop aih o Metal in Additive (%) Opsrating Time Depth of Attock No. No. Added By Analysis (hr) (mils) 289 59 Control 500 8 290 59 NiF, 0.05 0.03 500 10 291 59 NiF, 0.2 0.05 500 11 442 155 Control 500 10 418 155 NiF2 0.5 0.18 500 17 419 155 NiF, 0.5 0.18 500 17 466 173 Control 500 8 448 173 Control 1000 10 454 173 NiF, 0.4 0.11 500 15 455 173 NiF, 0.4 0.13 1835 16 452 173 CrF4 0.3 0.14 500 9 453 173 CrF, 0.3 0.14 2000 22 545 198 Control 500 11 512 198 CrF, 0.5 0.26 500 6 513 198 CrF, 0.5 0.27 1710 17 296 63 Control 500 8 294 63 Graphite rod 500 13 674 246-4 Graphite tube 1000 23 675 246-4 Graphite tube 1000 28 312 73 Cr powder 1.0 0.20 500 7 323 92 Cr floke 0.5 0.21 500 4 324 92 Control 500 15 335 92 Inconel turmings 500 11 After 500 hr, batches to which CrF, had been added developed the same depth of attack found in the control loops. It is possible that the presence of CrF, aond the accompanying free fluorine atom caonceled each other. A definite increase in amount and depth of attack was noted for these loops at 2000 hr; the rate during the period from 500 to 2000 hr was 0.9 mil per 100 hr, In both cases the addition of chromium fluoride resulted in an increase in the rate of mass transfer, This appears to conflict with the statement above (see '‘Operating Time'') that the rate of mass transfer seems to be independent of batch purity; however, the additions made here are much larger than would be found in normal batch variations. In some of the early loops, chromium metal powder and Inconel turnings were added to the fluorides in the charge pot. The chromium did result in a small decrease in attack, but metallic layers were found on the cold-leg surface. A small improvement appears to have been obtained from the Inconel turnings, but the results are not definite, From the time data obtained with the 23 Ve UNCLASSIFIED T-11074 (@) T-5347 'l . - .‘ . = Y "noe o . .'\\ .'. ;q:“;’.- ': » .-".. ¥ ’ - 5 /,/ ., .y " 0. % ¢ ‘. N v ‘ . § t _.. - o v :_-’ .. ‘. .. al ' ' " .d o ""‘“._ - 3 ‘ = . t( H-"' a ‘ . 4’ S UNCLASSIFIED] UNCLASSIFIED 343 1 & INCHES | 1 EEREE *_ 2 - - b . " a 4 1 0l N v ! i W\ o014 UNCLASSIFIED T-5145 Fig. 11. Effect of Additives on Attack in 500 hr Operation. (a) CrFE; (b) CrFB.‘ (c) Nin; (d) stoandard loop. 250X. Reduced 17.5%. T4 UNCLASSIFIED UNCLASSIFIED ] T-9685 T-9683 N LT - ‘ U & - = i - 3 “ q b N - - : '.p 'L g e e i = . I" . “ f‘- T — \ 3 » -l e X = : v .: \ : ‘. . R 4 3 -, 1, x ¥ - = ‘- ’- = -!' - .. / : .- '- i - g , for gl UL SO B : g KT o & " i L T SR - * .. % T LS Tioy = Lui— : 5 ~ T8 e el - . 5 . A o [&] ) 2y 4 ( § % . = - ' A g E A z > . ‘ - ‘ . . .-. o g T 3 e g e s % - ‘ o r’ ‘ - 5 ke 1 - i - " » - - - 0.02 - . 1 ! [ il -. |o.02 ] : ' . ’ P - .8 Lt - . ! -l - I . \ = 0.03 .08 (a) s gl T 4 (b) . g : . UNCLASSIFIED T-9482 tat & i i x - . 23 . -h s ¥ » ' . \ i h . Ky, .. : . o ~ . . - i " - i N A o ST B e S o | . ! RS i VoEk | . . s e . e 4 N T Pl 'r = \ . 1 L . = Luf— - - S ® - s i - g 5 » [ - . 3 s " r r ; - i z r A » : n, e . z l ol I- ey 1-‘\ 5 . B i e 3 »r 5 : . 0.02 R »—v : / o 0.02 - : — . & . : ; ta, . g 3 —— s . | ‘. " s ; “ : Fig. 12. Effect of Additives on Mass Transfer Reaction in 2000 hr Operation. () CrFT‘ (b) CrFa; (c) Nin; (d) standard loop. 100X. Reduced 17.5%. loops, it was shown that about 200 hr was re- quired for the impurity and saturation reactions to be completed, so it appears that fairly long treatment times would be necessary for equi- librium. For these batches 24 hr was the longest treatment time used. A group of loops was operated with graphite present in the hot legs, The purpose of these loops was to determine whether graphite could be used in an Inconel-fused-salt system. With loop 294 a h-in. graphite rod was suspended in the center of the upper portion of the hot leg, while with loops 674 and 675 tubular inserts with internal diameters the same as the diameter of the Inconel pipe were used, The tubular inserts were about 6 in. long ond were located just below the hottest portion of the loop. With each of these loops a definite increase in the depth of attack in the Inconel was noted. A deposit was present on the surface of the graphite and pene- trated it for a considerable depth, The photo- micrograph in Fig. 13 shows the surface of the insert from loop 294 and shows the deposit extending for a depth of 17 mils, On the surface the deposit is in the form of a thin film of metal surrounding nonmetallic particles, Another material that was added to the fluorides was zirconium hydride, It is now known that the function of this material was to convert a portion of the UF, to UF,; therefore this material will be discussed in the section on the reduced uranium fluoride mixtures (“'Special Fuel Mixtures'’). Ratio of Surface Area to Loop Volume., — One difficulty in comparing results from various corrosion tests is the wide variation between the tests in the surface area to volume ratios. In isothermal static tests it is easy to define and figure such ratios; however, with dynamic sys- tems it is difficult to determine what the signifi- cant parameters are, let alone measure them. The significant surface area is the one where the rate-controlling step takes place and could be the cold-leg surface as well as the hot-leg surface, The thermal loop is not an ideal rig in which to study variations in these ratios, since they cannot be varied independently. Any variation in the ratio of surfoce aorea to volume also results in variations in rate of heating and in temperature distribution, which in tum affect the wall temper- atures and velocities, JUNCLASSIFIED | ¥ 79681 i -h-l_ n W z ) | lo.02 | | | « | B Fig. 13. Metallic Deposit on Surface of Graphite Rod Suspended in Hot Leg of Loop 294. 100X. 26 For this discussion the ratio of the heated surface to the total loop volume was used, Since each loop was made from a single size of pipe and all loops are the same length, such ratios are dependent only upon the pipe diameter. The ratios would vary in value depending upon what is considered as the critical surface, but such variations would cancel out in comparing various loops with the same length of critical surfaces ond would depend only upon differences in the diameters. A series of otherwise identical loops were constructed from tubing ond pipe of various diameters, The data from these loops are tabu- lated in Table 11, After 500 hr the loops with the low ratios of surface area to volume show the most attack. This is as would be expected, since the chromium required by the batch was obtained from a smaller surface., With the longer-time loops, no definite conclusion may be drawn, as shown by the data summarized in Table 12, After 1000 hr, less attack is found in the loops with the low ratios than in those with high ratios; while this is still true after 2000 hr, the differ- ences are smaller rather than having continued to increase, |f the time from 500 to 2000 hr is con- sidered, the differences in rate are within the vsual spread; for the loops from ]é-in. tubing the rate is about the same as those obtained from the time curves in Fig. 6. Oxide Removal Procedures. — While much of the available commercial tubing contains internal surface scale and dirt, the pipe used in this study was supposed to have been annealed and pickled. In most cases this specification was met, but some loops were found in which heavy surface Table 11. Effect of Ratio of Heated Surfoce Area to Total Loop Volume on Depth of Attack Ratio of Heated Surface Loop EE Batch Pipe Size* e late Vb Operating Time Attack No. No. (in.2/in.3) (hr) Intensity Depth (mils) 349 19 lin. T 1.7 500 Moderate 9 359 119 Y%in. T 3.9 500 Heavy 4 352 19 Y in. P 2.3 500 Moderate 5% 362 19 %oin T 3.9 500 Heavy 5 549 232-8 Yin, T 3.9 2000 Heavy 147 550 232-8 % in. P 2.6 500 Moderate 9 551 232-8 % in. T 3.9 1000 Heavy 14 565 232-8 iy T 1.7 1000 Heavy 7 566 232-8 Vs T 1.7 2000 Heavy 12 567 232-8 % in. P 2.6 1000 Light 9 568 232-8 % in. P 2.6 2000 Moderate 12 569 232.8 % in. P 2.3 2000 Heavy 1Y, 570 232-8 % in. P 2.6 500 Moderate 9 160 217-5 Yin. T 3.9 3000 Moderate 18 529 217-5 Yin T 3.9 1000 Heavy 12 *P = pipe (IPS); T = tubing (outside diameter). 27 Table 12. Summary of Corrosion Results from Loops with Various Ratios of Heoted Surface Area to Loop Volume Attack Rate (mils per 100 hr of operation) for i i geya) BE Baiseh Surface to Volume Ratio (in.2/in.’) of (hr) Ne. 3.9 2.6 2.3 1.7 0-500 119 0.9 1.1 1.8 0-500 232-8 1.9 1.8 500-1000 232-8 1 0 10002000 232-8 0.05 0,3 0.5 500-2000 232-8 0.3 0.2 10003000 217-5 0.3 layers were present on the walls before operation. opparent but they are not systematic. |f aony con- Since these loops had been previously degreased with a solvent degreaser, all loose dirt and grease had been washed out. The layer was identified as being primarily chromium oxide. Several procedures were tried in an effort to remove these oxide |oyers before the loops were filled and operated, The loops designated “'First Ser.es’’ in Table 13 were operated in connection with the original cleaning or oxide removal study made shortly after this work had started, As shown by the data, no large differences in depth of attack were found with any of the methods studied. Although none of these loops showed a reduction in attack from the usual control loops, a cleaning step was still thought to be desirable to elimi- nate erratic results caused by the occasional heavily contaminated loops. The surest and simplest method of oxide removal appeared to be the prior circulation for short times of another batch of fluorides. This method was then adopted as standard procedure, When the gradual increase and poor reproduci- bility in the depths of attack were noted, cleaning variables were one of the suspected causes and were checked again. The results obtained in this second study are reported as ‘‘Second Series’ in Table 13. Since cleoning variables represent attack due primarily to the reduction of impuri- ties, they should show up more in short-time loops; therefore the loops in this series were operated for only 250 hr. These loops again show no apparent variation in depth of attack with the various oxide removal steps. Some variations are 28 clusion were to be drawn, it would have to be that the lowest attack was in the loops that received only a careful degrease and visual check with no oxide removal operation. Heating Methods. — Both to reduce the number of variables and for ease in operation, it was desirable to use only a single heating method. From previous work it was known that clamshell resistance-wound heaters and saturable-core reactors for control were a satisfactory combi- nation. Consequently, these heaters were adopted for this study, but, since it was shown previously that the attack was heavier and deeper under the heaters, a study of the effect of heat flux on corrosion was desirable. In addition, some corrosion theories were proposed in which the heating method was a variable. To answer these questions it was necessary to operate some loops with a different method of heating, A method of heating by means of the electrical resistance of the pipe wall was used during the cleaning operation and was then modified for use during operation. To concentrate the heating in the hot leg, so that the heating area and dis- tribution would remain constant, one electric terminal was connected to the middle of the hot leg and the other one connected in parallel to both ends of this leg. Because of the short path, the resistance was low and very high currents were required to give the required power. The data from the loops heated in this manner are tabulated in Taoble 14, and a typical hot-leg section is shown in Fig. 14. The depths of attack obtained with this second heating method Table 13. Variations in Depth of Attack with Oxide Removal Procedures Loo EE Batch Attack H: o Oxide Removal Method ey TR First Series 237 R-56 Hot sodium 7 234 R-43% None Light 5 236 R-44¢ Dry hydrogen Muderate' 10 231 R-514 Hot NaK Moderate 9 244 144 None Moderate 9 293° 63 None Haivy 7 295 63 Molten fluorides Heavy 6 296 63 None Heavy 8 353 119 None Heavy 3 352 N9 Molten flucrides Moderate 5% Second Series® 722 249-1 HNO, + HF Heavy 8 737 249-4 HNO, + HF Heavy 8 723 249-1 Dry hydrogen Heavy 9 724 249-1 Dry hydrogen Heavy 8 725 249-1 Molten fluorides Heavy 8 726 249-1 Melten flucrides Heavy 8 727 249-1 None Heavy 7 728 249-1 Neone Heavy 7 732 514-1 Machined surface Heavy 7 733 514-1 Machined surface Heavy 4 739 514-1 HNO, + HF Heavy 9 740 514-1 HNOB +HF Heavy 6 742 514-1 Dry hydrogen Moderate 8 743 514-1 Molten fluorides Moderate 6 744 514-1 Molten fluorides Heavy 6 745 514-1 None Heavy 6 746 514-1 None Moderate 7 “Fluoride 27. bOx'rdizad hot leg. “Operated for 250 hr. 29 Table 14. Depth of Attack in Loops Heated by Electrical Resistance of Hot-Leg Wall Loo EE Batch Time Attack NO-P No. (hr) S o Intensity Depth (mils) 685 246-1 500 Clamshell Heavy 14 684 246-1 500 Clamshell Heavy 11 686 246-1 500 Clamshell Heavy 9 618 246-1 500 Resistance Heavy 10 619 246-1 1000 Resistance Heavy 13 703 246-1 1000 Resistance Heawvy 15 736 249-1 1500 Resistance Heavy 18 UNCLASSIFIED T-3'|5U 0ol a2 Wy T " v . x .’!.:‘E o.' b ¥ o Te*. Br.yr et L002) « 4 re. . ® AL J:-.- - . 003 n . @ e > " 2 / 5 , R i | = . i - < ~ ¢ \_ ’ - .008 e 1 007 oA ¢ 1 2 e % . 8 : P v, o0t . - - i 2 -, ‘ J s o 010 ' ° / \ - ol : R R -l 012 - = o - = .013 £ " o ' 3 o - 014 Fig. 14. Appearonce of Attack in Loop Heated by Electrical Resistance of Wall. 250X. are within the spread of depths from the control loops. Some effort was spent by the Experimental in developing a gas-fired Engineering Section and the maximum fluoride temperature occur in the same vicinity and in which no hot spots are present, While direct comparisons cannot be made at this time, no larger differences in attack This furnace capable of heating the loops. method would be especially advontageous for higher temperature loops, since at 1500°F the heating capacity of the electrical equipment is being approached. As yet, a gas furnace has not been built in which the maximum wall temperature 30 have been found with the preliminary gas-heated loops. While it has been shown that deeper attack may be found under a heater, the attack does not depend upon the method of heating. The differ- ences are caused by the higher temperatures, either in the wall or, more likely, in the fluid boundary layer adjacent to the walls, When a reactor system is compared with a thermal loop, two differences in heating are found. In the reactor the heat is generated within the liquid, and so the wall is at a slightly lower temperature than the liquid. In addition, tempera- ture cycling may be present and may be quite ropid. As yet, no method has been devised by which a loop may be internally heated, so that the first difference will have to be checked in the radiation tests. A series of tests were set up to determine whether the second difference — temperature cycling — increased the rate of mass transfer, The temperature cycle used for these loops was periodic and, to avoid excessive wall temper- atures, of fairly long duration. Figure 15 shows the hot-leg temperature pattern of a typical loop. The corrosion data for this series are tabulated in Table 15, Within the accuracy of the data this slow cycling did not affect the depth of corrosion. |t appears that the cycling averages the attack over the entire temperature range rather than effectively increasing the differences, and it has already been shown that the depth of attack is not affected greatly by changes in hot- leg temperatures, It is possible that a more rapid cycling could increase the attack by effectively increasing the temperature drop, but this cannot be confirmed in the present equipment. Loop Size ond Shape. — As discussed in the companion report,' the size ond shape of the loops were arrived at primarily from physical considerations such as sizes of availoble heaters and ease of operation rather than from flow and considerations. As corrosion the corrosion UNCLASSIFIED ORNL-LR-DWG 48768 LA TIME (hr} Fig. 15. from Thermal-Cycled Loop. Hot- ond Cold-Leg Temperoture Patterns Table 15, Effect of Temperature Cycling on Hot - By i ‘ .‘ e L ma . e ; s E . ‘ L.00%) . 9 - ; 210 . & Q9 ‘ " 010 ., - - P |ane | \ 0l o : . e NG . O8] J \\\' s =8 |ois - " 4 013 ' . P 3 il 2 - Q e ! < (¢) Fig. 19. Growth of Voids Formed by Barren Fluorides with Increasing Time of Operation. (a) 500 hr; (b) 1000 hr; (¢) 2000 hr; (&) 3000 hr. 250X. Reduced 17%. plugged in relatively short times, Nickel loops operated for 500 hr, but metallic crystals were found, and Hastelloy B loops failed by cata- strophic oxidation. Only the high-nickel alloys similar to Inconel looked promising. Since the properties of the zirconium fluoride—base mixtures are considerably different from those of the alkali-metal-base mixtures, many of these tests were repeated with the zirconium mixtures. In UNCLASSIFIED DRNL-LR-OWG 48771 |4 — ' — —I— 2 + 1 - | J | v ¢ ‘T_ / 3 “ ! / o E - —t 4 S e - 4 : | g:} ; A-—A LAB BATCH E o6 4 L | ¥ AREBATCH —| o ! » X © EE 134 I | i « EE 210-5 e q -0 2 t o al 1= __:, 4 b | | | | Q | - L i el —r L o 500 1000 1500 2000 2500 3000 OPERATING TIME (hr) Fig. 20. Increase in Depth of Attack with Operating Time for Barren Flueride Mixtures. (@) Fig. 21. UNCLASSIFIED T-4045 l“"r . . general, these tests are much more favorable than the previous ones were, Stainless Steels. — All stainless steel loops operated satisfactorily for 500 hr when zirconium fluoride—base mixtures were circulated in them, as shown by the data tabulated in Table 21. The hot-leg attack in type 316 stainless steel loops was similar in appearance (Fig. 22) and to about the same depth as was found in Inconel loops. As would be expected from solubility studies, the iron and nickel concentrations of the fluorides remained low and the chromium increased during operation. The disadvantage of this alloy is the metallic deposits that were always found in the cold legs (Fig. 22). Attempts to identify the deposit by spectrographic and diffraction studies were not successful, but indications are that iron and chromium are both present, When the loops were fabricated from a 400 series stainless steel, the nature of the attack changed. The hot-leg surfaces were rough, with entire crystals being removed, and no subsurface voids were present, Large deposits of dendritic metallic crystals were found in the cold legs. Typical cold-leg sections with types 410 and 446 stainless steel are shown in Fig. 23. The cold-leg deposit in a type 410 loop analyzed 53.4% Fe-4,3% Cr=1% Ni-12,9% Zr., In a type 446 loop the crystals were 35.7% Fe-7.6% Cr-1% Ni-13.8% Zr. The zirconium is not part of the crystals but is from unseparated fluorides. These data indicate that both iron and chromium can be UNCLASSIFIED T-4130 VL . 4 -) ppR- e |oos) \ v 1 . > B L.008 % | ". Comparison of Attack in Loops Operated with Two Barren Batches and with a Precirculated Batch. (a) Loop 337, operated with two batches; (b) loop 338, operated with a precirculated batch. 250X. Reduced 29%. 40 Ly Tflblfl 21. Data from Stainless Steel Loops Analysis of Fluoride Mixture Loop Alloy Batch Time Metallographic Examination Nickel Chromium Uranium No. Type No. (hr) Hot Leg Cold Leg (ppm) (ppm) (ppm) Before After Before After Before After Before After 128 316 R-50% 500 8 mils, intergranular Crystals 320 <20 <20 1600° 665 110 7.50 9.3 133°€ 316 500 11 mils Deposit 200 20 35 1300 2100 150 8.80 9.0 134 316 52 500 7 mils, heavy; rough Deposit 30 <20 110 Bflflb 260 60 8.81 9.1 surface 403 430 1199 500 Even Deposit <20 <20 75 825 230 b 8.99 8.8 138 410 141 500 Very rough Layer and crystals® <20 <10 95 800 185 120 9.14 8.8 137 446 141 500 10 mils, voids; rough Layer and crystuls'f <20 <20 90 800 155 80 8.53 8.8 142 430 158 500 7 mils, light® Some deposit and <20 <10 30 SGUb 55 b 8.80 9.0 crystals “Fluoride 27. Individual results varied. “Cleaned only by degreasing. d3.0 ppm HF. ®Crystals 53.4% Fe, 12.9% Zr, 4.3% Cr. ICrystals 35.7% Fe, 13.8% Zr, 7.6% Cr. &L arge grains and no carbides in hot leg. 42 L v 8 - - A l; - ‘\ | T - - * - ¥ . x, = . . '\::—.’"" i » - » " o ; k NN o P - 1 - - - & - - L /" " .. \\- " p - o % : b .- 2o N\ . J - b % i E + ¥ S ¥ s ¢ \ . - - e - - - 1)4 - = * - wg . " - ."‘. . . = . @ o F A « . i r T - 4 - ) - . (a) -, : - - - e Fig. 22, (note thin metal deposit). 250X. Reduced 15%. Corrosion of Type 316 Stoinless Steel by Fluoride 30. \ ) S UNCLASSIFIED T-2992 . b le o ! 250X UNCLASSIFIED [ ] T-2995 o o o ! T 1flcrE5 3 J . B o o o 4, 250% LY o = (a) Hot leg; (k) cold leg UNCLASSIFIED INCHES gy @0 LOOP 138 - SECT. 10 () LOOP UNCLASSIFIED INCHES T-4435 2 3 137 - SECT. 10 Fig. 23. Deposits Found in Cold Legs of 400 Series Stainless Steel Loops. (2) Type 410; (b) type 446. mass-transferred in these systems, and while the loops operated without plugging for 500 hr, the rate of mass transfer is much too high for use of these alloys. Izett lron. — The single instance of a loop plugging with zirconium fluoride—~base mixtures was found with lzett iron (low-carbon high-purity iron) loop No. 47. This loop plugged in 39 hr by the formation of balls of dendritic metal crystals (Fig. 24). The hot-leg attock was an even re- moval, but the depth could not be determined, since considerable oxidation of the outer surface also occurred. Nickel. — Two *“‘A’" nickel loops operated for 500 hr with very little hot-leg attack. The wall surfaces were polished, but the thicknesses were still within commercial tolerance, so at least no large amount of metal was removed. Neither intergronular nor void type of attack was found, as shown in Fig. 25. A few metallic crystals were found in the cold leg of loop 108, but none were found in loop 110, In both loops, the nickel content of the fluorides remained very low. Inconel X. — Attacks of 22 and 19 mils were found after circulating flvoride 44 for 1000 hr in loops constructed from Inconel X (15% Cr, 9% Fe, 2.5% Ti, 0.6% Al, 11,0% Mn, 1.0% Cb, bal Ni), This is slightly deeper than would be found in comparable loops of Inconel, but the attack is similar in nature, as shown in Fig. 26. No definite deposit was found in the cold leg of either loop; however, with loop 170, an area on the cold-leg surface |1 mil deep etched differently UNCLASSIFIED T-11282 ONE INCH Fig. 24, Metal Deposit Token from Cold Leg of on lzett Iron Loop. from the remaining material. The fluorides in- creased in aluminum and chromium concentration but not in titanium, Molybdenum and Niobium. — Although, because of fabrication difficulties, it does not seem likely that they can be used immediately for reactors, melybdenum and niobium show considerable 43 UNCLASSIFIED T-4233 B L 7 LOO% | . " \...&. Fig. 25. Hot Leg of a Nickel Loop. 250X. Reduced 33%. promise of future use as containers for molten fluorides, Low corrosion rates have been found with both metals. Both these metals are difficult to weld and must be protected against oxidation, In attempting to operate these loops, many failures occurred from cracks in the original tubing and from welding stresses. The molybdenum loops were about half size, being 15 in, in each leg, and were made from 3af‘-in. tubing. The niobium loops were full-size loops but were constructed from ‘fz-in. tubing, The corrosion data from these loops are tabulated UNCLASSIFIED[ | T-T655 m - 5 : s e - - o las . -.' 5 .‘. IAMSS e T Y (e .{'. i ‘oW t -'hl', .: r_’-'_'. o ‘-. _I-:I=-I u ‘, " " & S . o8 " i - * = -;‘l‘ g-. =y . l s - u'.“ - C"t“- ¥ 3 /=] t W f LT E _\. . ™A v & . W - "‘ 008 | : . o =1 3 -JI '\.‘ wot ] Pl ! « b i - - v L- k L Loos| - e s » ‘- ‘ 4 - 010 | . ¢ * lau . - ‘ ¥, loia 8 k'l . - 215 = -, & "-. . \, o Fig. 26. Hot Leg of an Inconel X Loop. 250X. Re- duced 37%. in Table 22, and typical photomicrographs of the hot legs are shown in Figs. 27 and 28. No evidence of attack or of any deposit was found in the molybdenum loops. The surface was rough ond pitted, but as shown in Fig. 27, the same condition was present in the original tubing. Difficulty was encountered with the analytical determination of molybdenum in fluorides, but it is obvious that the solubility is very low. These loops were not operated with in-line traps, so a final judgment on mass transfer should be re- served until this has been done, Table 22. Corrosion in Molybdenum and Niobium Loops Loop Alloy Time Maximum Attack Metallographic Examination Mo or Nb After No. (hr) (mils) Hot Leg Cold Leg (ppm) 180 Molybdenum 250* None Few pits No ottack or deposit <1 184 Molybdenum B43* None MNe attack or deposit 15 185 Molybdenum 1000 Mone Mo attack or deposit 11-40 1003 Moly bdenum 424* MNone No attack or deposit 250 1002 Molybdenum 2000 None No attack or deposit 12-65 760 Miobium 1000 ]5 Light intergranular No attack or deposit 761 Niobium 695+ % Light infergianulor No atfack or deposit 15 762 Hishivi 1000 Y Light intergranuler Niobium erystals; no <1-270 deposit *Operation terminated becouse of leak. 44 Sy »y N> . }r.?:_:\(“.‘. e - .-'\ . '-_ ffa.'pp_r" i?: St e e ,2 % \*'-l . LL:\Q & Vom > b ey Tt AT WL ‘&.-{- '..’hI. i - UNCLASSIFIED T-10372 T T . INCHES JB F; o I 1 1 P = i o lo B T [ - - > - o . T T 2 5 |§ INCHES |S E 1 1 = o o - o & UNCLASSIFIED [ ] UNMCLASSIFIED T-8238 Fig. 27. Effect of Operating Time on Corrosion of Molybdenum by Fluoride 30. (a) As received; (b) 250 hr; (c) 1000 hr; (4) 2000 hr. 250X. Reduced 17%. UNCLASSIFIED T-8358 R N e & . . A | | Jes R i fi* | ( i = » 'i y d 3= < A ; > . ; . A = .I,J \ .f' ) { J A 4 ¥ [, STy - X { p {--,_, a5 A B 4 : Vel R S R \%IJ [ i I :. .,_"‘q-J i . ¢ TS~ ] 23 C AN 4 Fig. 28. Hot Leg of Niobium Loop After 1000 hr with Fluoride 30. 250X. Reduced 33%. No evidence of attack was found in the hot legs of the niocbium loops; however, possible crystals were found in one of the cold legs. Further work will be required to resolve this discrepancy. Hastelloy B. — Another material that was investigated was Hastelloy B (28% Mo, 5% Fe, 0.12% max C, bal Ni). In the original screening studies,' two such loops failed in short times by catastrophic oxidation. The first two loops in this series failed in a similar manner, but the difficulty was remedied by replacing the Sauer- eisen cement normally used to protect thermo- couples with metallic shields. Hastelloy B is another material showing promise for use with the fluorides, as shown by the data in Table 23, A few voids and cracks to shallow depths were found in both legs, with the cracks also being found in the as-received pipe. Very little, if any, increase in attack was found with an increase in hot-leg temperature from 1500 to 1650°F or with an increase in operating time to 2000 hr (Fig., 29). While the rate of attack was low, some occurred, since in several loops a few dendritic crystals of nickel were found in the cold legs. When Hastelloy B is fabricated, some mo- lybdenum is vaporized; therefore there was o possibility that a nickel-enriched layer on the surface was causing the mass transfer. The internal surfaces of several loops were machined to eliminate any such layer. The attack found in these loops was essentially the same as in the normal loops. 46 The major disadvantage with Hastelloy B is the large amount of mass transfer found when molten sodium or sodium-potassium alloy is circulated in it. With the presently proposed reactors, it would be highly desirable for o metal to have corrosion resistance to both fluorides and alkali metals, Special Alloys. — |t has been shown that the only metal leached from Inconel was the chromium; therefore, in an effort to reduce the attack, alloys were made with decreasing chromium concen- trations. These alloys were vacuum-melted and extruded by the ORNL Metallurgy Division and then drawn into 'I:-'z-in. tubing by the Superior Tube Company. Fluoride corrosion data from these loops are tabulated in Table 24, and representa- tive hot-leg sections are shown in Fig. 30, In each case when the chromium in these alloys was reduced from the original 15% to about 5%, the depth of attack was reduced by a factor of 3. The reduction was found when the chromium was replaced with either iron or molybdenum, When the chromium concentration was reduced to 10%, erratic results were found, Results from another series of special loops ore tabulated in the lower portion of Table 24, These loops were operated to determine whether variations in carbon content had an effect on corrosion. Loops 366 and 367 were fabricated from vacuum-cast metal and contained only 0.01% carbon. The attack in both these loops was essentially the same as that found in control loop 359 made from commercial alloy containing 0.07% carbon (Fig. 31). Special Fuel Mixtures In the last report,’ it was pointed out on the basis of only a few tests that the addition of zirconium hydride to alkali-metal-base mixtures showed promise of reduced depths of attack, but that unidentified layers had been found on the hot-leg surfaces. This type of work has con- tinved, with similor additions being made to zirconium fluoride~base mixtures, A large in- vestigation into the basic mechanisms in such a system has also been carried out by the ANP Chemistry Section. In two series of tests, various amounts of zirconium hydride were added to portions of the some large batch of fluoride 30. The zirconium hydride was added to the fluorides in small Table 23. Corrosion Data from Hastelloy B Loops Loop Time Heat Metallographic Examination Nickel (ppm) Molybdenum No. (hr) Treatment Hot Leg Cold Leg Before After (‘:;:; Comments 378 94* Rough surface No deposit <20 No attack on welds 140 500 Aged 1950°F 2 mils, light; rough 20 10 1 No attack on welds 153 394* 1 mil, heavy; rough 1.5 mils, intergranular cracks 20 20 154 1000 1 mil, moderate; rough 1.5 mils, cracks; rough <20 <10 157 736** Aged 1.5 mils, moderate Rough <10 <10 1 156 0 1.5 mils; rough 1.5 mils; rough 158 2000 2 mils, heavy; rough 2 mils, heavy <20 1 548 2000 1 mil, moderate 1.5 mils, intergranular 60 <20 162 1000 3 mils, moderate; rough 2 mils, intergranular 15 25 15 1650°F 161 1000 2 mils, moderate; rough 2 mils, intergranular 25 25 10 1650°F 163 1000 Aged 2 mils, moderate 2 mils, moderate 35 3-135 10 167 1000 2 mils, heavy 1.5 mils, moderate 40 <20 <10 Flueride 44 164 1000 Aged 2 mils, moderate 2 mils, moderate 35 <20 <10 174 1000 3 mils, heavy; rough 1.5 mils; rough 35 20-60 Metal crystals 159 2000 2 mils Rough 20 20 <] 186 1000 3.5 mils, heavy; rough 2.5 mils; rough; crystals 15 <1=-70 <5 1650°F 187 1000 4 mils, heavy 1.5 mils; rough; few crystals <20 <10 <20 1650°F 769 500 1.5 mils; large voids 1.5 mils, voids; metal in trap 5 10 15 Machined internal surface 770 1000 1 mil, heavy; rough No deposit 10 70 Machined internal surface 771 1500 2 mils, maderate 2 mils, moderate 10 <10 Machined internal surface Ly *Operation terminated because of leak. **Operation ferminated because of pawer failure. UNCLASSIFIED 6 EE 0o o o ¥ T - - o = . / e .. e T By e ’ N - . z L3 L AN . g o -\-\-:-":. s i 1 oo © & . =i & o - e i Bl o - - 1 2 - o e o I o X ma T - - v o P o Oy ‘ug v T TR T D AT LS “%: . W .;'h‘ * g = o o= " x B ' ; . = oo8 L ® | - e =) % F g ? i 3 @ g = s - . fo - e .L. e B T 3 ’ & Ty i > L e [ 1 y . ., . - acs¥ - |\ i A - - 1 c @ 1'% . L - oig & - o b= dn ¥ % * . - 2 mBev i ol i v € ' = oy [ 1. z gr o * i - PR e L Loz | et = . F 3, - i 4 e - . 1 > ) ors \ o . . 2 & s - | (a) - / D14 UNCLASSIFIED T-6085 o Y !_-a.._a-- - » oo ked 3 ud \.\' :‘l -‘h.l.- Ll - P e (Gt A A AT AP Sl A i J e i e " ' 4 SR . v . § 1A . P . - L] ¥ L hy = b " ‘ i - b -y A& . F » oie e UMEL ASSIFIED T-6084 o EE o o L T T INCHES 1 1 & Fig. 29. Effect of Time and Temperature of Operation on Attack in Hastelloy B. (a2) As received; (&) 500 hr at 1500°F; (c) 1000 hr at 1500°F; () 2000 hr ot 1500°F; (&) 1000 hr at 1650°F. 250X. Reduced 30%. 43 Table 24. Corrosion Data from Modified Inconel-Type Alloys D Alloy Composition (%) Time Attack Chrom.iurn Con::entrariun No. Cr Fe Ni Other (hr) Intensity Depth (mils) 3 Tiuondes ppm) 520 176 68 759 1000 Heavy 13, 1000 522 9.8 14.5 75.6 8239 Heavy 8 580 523 9.8 14.5 75.6 1000 Heavy 12 500 524 4.8 19.8 76.1 647° Moderate 3 400 525 4.8 19.8 76.1 460 Heavy 3 300 526 9.8 6.7 83.4 1000 Moderate 13 650 527 9.8 6.7 83.4 1000 Heavy 15 650 521 6.0 10.1 74.4 Mo, 9.8 1000 Light 2 450 472 6.0 10.1 744 Mo, 9.8 500 Light 3% 300 529€¢ ~15 V75 76 1000 Heavy 12 550 366 15.0 7.3 78.1 C, 0.01 500 Moderate 3% 950 367 147 7.2 78.1 C, 0.0 500 Heavy 3% 850 359¢ ~15 ~TS T8 C, 0.07 500 Heavy 4 850 aDperution terminated because of leck. bOpercfion terminated because of power failure. CCommercial Inconel. (a) Fig. 30. Typical Hot-Leg Sections from Loops of Modified Inconel-Type Alloys. (b) 4.8Cr-19.8Fe-76.1Ni. 250X. Reduced 21%. UNCL ASSIFIED T-5455 sk B E | T ] (5) UNCL#S?FIED o EE] ‘o o - T 1MC?ES EREE (a) 6Cr-10.1Fe-74.4Ni-9.8Mo; 49 UMCLASSIFIED T-4517 Fig. 31. (k) low-carbon Inconel, 0.01% C. 250X. Reduced 30%. charge pots, and after being agitated at 1300°F by helium, the mixture was transferred to the loops through grade G Micro Metallic filters, The metallographic data from these loops are tabu- lated in Table 25, with the chemical data in Table 26. These data show that a gradual re- duction in depth of attack was found as the zirconium hydride concentration was increased to about 0.7%, With larger hydride additions, layers of increasing thickness were found on the surfaces of the hot and cold legs, as shown in Fig. 32, The deposits attained greater thickness in the hot legs. With the use of a special microspark spectrographic method, it was found that the thicker layers were predominantly zirconium. This so this metal may have also been present and not detected, The shallow attack found in many of these loops is not the usual subsurface void type of attack, These areas are holes which are open to the surface and may have originated as scratches or pits in the original pipe. After operation, the chromium fluorides to which zirconium hydride had been added was always much lower than in similar tests without the addition. The chromium gradu- ally decreased with hydride additions up to about technique was not sensitive to uranium; content of 50 Comparisen of Attack in Commercial and Low-Carbon Inconel. UNCLASSIFIED T-4575 (5) sid (a) Commercial Inconel, 0.07% C; 0.5% but remained low and constant with further additions. affected by the zirconium hydride additions. As the hydride was increased above 0.5%, notice- able amounts of uranium were reduced in the charge pots and left on the filters during transfer. With a 2% hydride addition, half the uranium was lost. With the higher additions, small amounts of uranium were also lost during operation of the The uranium concentration was also loop. It has been found by the ANP Chemistry Section that the action of the zirconium hydride in re- ducing the attack is not from purifying the so- lutions, as was first thought, but is by the reduction of a portion of the UF, to UF,. Attempts were made to determine the UF, content of the series both before and after operation; however, the results are in such doubt that they are not being reported. From these results it does appear, however, that in every second case the UF, content was reduced to values below 1% in 500 hr of operation. A comparison of the depths of attack from loops 413 and 414 shows that the addition of small amounts of hydride does not stop mass transfer. Between 1000 and 2000 hr the rate was the same as that found from the time curve Table 25. Metallographic Results from Loops in Which Mixtures Containing ZrH, Were Circulated Loop ZrH, Added EE Batch Operating Time No. (%) No. (hr) Hot Leg Cold Leg 248 0.6 19 500 Moderate attack to 4 mils Deposit 308 0.5 68 500 Light attack to 2‘5 mils Deposit 413 0.2 150 1000 Light attack to 7 mils 414 0.2 150 2000 Heavy attack to 11 mils 469 0 162 500 Heavy attack to 8 mils No deposit 459 0.2 162 500 Moderate to heavy attack to 6 mils Deposit 470 0.5 162 500 Light to moderate attack to 3 mils Deposit 460 0.9 162 500 Thin deposit Deposit 471 2.0 162 500 Deposit to 1 mil Deposit 800 0 513-4 500 Heavy attack to 10 mils No deposit 787 0.4 513-4 500 Moderate attack to 5 mils No deposit 788 0.5 513-4 500 Scattered light deposit; pitting Scattered deposit 2 mils 789 0.6 513-4 500 Pitting 1 mil Scattered deposit 790 0.7 513-4 500 Intermittent deposit; pitting Deposit 1],-5 mils 791 0.8 513-4 500 Deposit; pitting 1% mils Deposit 792 0.9 513-4 500 Deposit 1 mil Thin layer UNCLASSIFIED 1 T L. INCHES IS E I 1 1 = = E (a) UNCLASSIFIED T-11064 (&) Fig. 32, Deposits in Hot Legs of Loops with Zirconium Hydride Additions. (a) 0.8% Zer; (&) 2% ZrHQ (note slight increase in thickness), 250X. Reduced 29.5%. 51 Table 26, Chemical Results from Loops in Which Mixtures Containing Zer Were Circuloted e ZiH, Added Uranium (%) Nickel (ppm) Chromium (ppm) Iron (ppm) No. (%) Before ~ After Before After Before After Before After 248 0.6 8.0 2.0 1590 <20 40 350 710 100 308 0.5 8.9 8.7 <20 <20 30 100 55 40 413 0.2 8.6 8.9 <10 <10 170 200 150 45 414 0.2 8.5 8.7 <10 <10 130 250 85 50 469 0 8.5 8.9 25 10 90 700 55 70 459 0.2 8.6 8.6 40 <10 75 200 %0 70 470 0.5 7.5 7.0 10 40 30 20 50 70 460 0.9 54 5.1 35 <10 20 35 65 60 471 2.0 4.0 4.1 <10 15 60 45 55 45 800 0 8.6 8.5 20 40 105 850 30 20 787 0.4 8.8 8.5 15 30 85 110 95 45 788 0.5 9.3 7.5 3 270 15 65 50 85 789 0.6 8.3 6.6 15 <1 35 60 75 65 790 0.7 6.6 6.2 40 15 25 90 30 85 791 0.8 6.5 6.3 70 <] 35 65 75 90 792 0.9 6.7 6.3 30 <1 35 40 45 85 (Fig. 6). The actual depths are less in both Difficulty was encountered in making batches cases, showing that some reduction in attack had occurred in the early stages. While the rate of mass transfer is the same, the chromium concen- trations of the fluorides are much less with the additions. This low value did not increase in the 1000 hr of operation even with the mass transfer, While zirconium hydride will reduce the depth of attack, it does not reduce the rate of mass tran sfer. Unless the compositions are con- trolled very closely, uranium will be lost from the system both before and during operation. |f such a system were to be used, the deposits in both legs would have to be studied in detail. After it was found that the formation of UF, was responsible for the reduction of attack with zirconium hydride additions, work was started on batches with a portion of the uranium present as UF,. The program consisted of both corresion loops and a study of solubilities and equilibriums in such systems, The most striking feature of this study was the inability to obtain reproducibility in the results. 52 containing known and predetermined amounts of UF, ond in developing a suitable method of analyzing such mixtures, The hot-leg attack was replaced by thin de- posits in the first three loops, which circulated batches made under carefully controlled con- ditions in the laboratory (Table 27). Two of these loops operated for 500 hr and the other for 2000, These loops all operated with low total uranium concentrations ond therefore high UF, to UF, ratios. The hot-leg layer in the 2000-hr loop was identified as zirconium by the micro- spark spectrographic procedure, To show the layers were formed during operation and not during cooling, a loop was drained aofter 500 hr of operation and while still at temperature. This loop also showed no attack and the presence of a zirconium layer, The solubility of UF, in the zirconium fluoride— base mixtures is not sufficient to provide the concentration necessary for reactor operation; therefore it would be necessary to use mixtures £S Table 27. Corrosion Data from Loops with Zirconium Fluoride—Base Mixtures Containing UF, EE Analysis of Fluoride Mixture Loop Time Metallogrophic Examination ThE Batch h Mickel (ppm) Chromium (ppm) Iron (ppm) Uranium (%) u (%) No. No. (hr) Hot Leg Cold Leg Before After Before Afrer Before After Before After Before After 438 Lab 500 Deposit 10 10 10 30 95 60 2.79 3.0 a 443 159 500 0.4-mil deposit Intermittent deposit 20 10 20 40 75 60 5.8 5.3 1.5 457 F-2 2000 Layer,? 0.5 10 5 10 80 120 75 2.8 2.6 1.6 mil; rough; no attack 473 F-4 500 Madrrme, 5 Occasional crystals 10 20 30 125 40 45 B.7 8.3 1.2 mils 488 B-85 2000 Modplerafe, 7 Thin deposit 15 10 60 90°¢ 40 60 B.7 8.8 2.0 mils 491 F-5 500 Heavy, 7 mils Deposit 30 20 70 180 30 50 12.8 13.2 2.5 492 F-5 2000 Heavy, 15 mils 1-mil deposit 10 10 35 250 40 40 14.2 13.2 1.8 0.7-2.8 5{13‘f B-89 500 annsitb No deposit 40 15 30 35 8O 60 2.4 2.5 2ud 1.0 508 207 500 No attack 0.5-mil deposit 10 c 260 60 50 100 10.4 9.7 4.3 0.6-3.3 627 426RF 1000 Heavy, 8 mils 1-mil deposit 15 20 50 200 40 40 9.0 8.9 1.2 0.4 628 427RF 1000 Heavy, 10 mils Thin Cr deposit 10 20 45 225 35 100 8.7 8.7 0.8 0.2 633 420RF 500 Heavy, 10 mils Intermittent deposit 25 20 100 575 65 35 11.8 11.4 1.4 0.7 634 422RF 500 Heavy, 8 mils Deposit 35 20 65 300 40 60 B.8 B.7 1.3 0.5-1.4 635 424RF 1000 Moc!?rntm 7 Thin deposit 60 20 100 200 BO 50 8.9 9.7 2.0 0.4-1.2 mils 636 430RF 1371° Mndrrula, 9 Mo deposit 15 20 40 250 BO 50 8.4 9.1 0.6 0.3 mils 683 423RF 500 Mndrram, 7 Rough; no deposit 25 1-45 140 225 25 40 9.0 8.4 1.4 0.6-1.0 mils 690 432RF 1000 Heavy, 10 mils T-mil attack; no 40 20 190 275 40 40 7.7 8.3 1.3 0.4 deposit 781 371 500 Mo:_lgerute, 9 Deposit 7 40 75 700 40 55 9.5 9.6 0.1 0.1 mils 782 ar2 500 Heavy, 9 mils No deposit 10 25 75 390-525 50 85 9.8 10.2 0.3 0.1 784 373 500 Heavy, 7 mils No depesit 20 25-185 60 200 45 70 10.7 11.2 1.5 0.4=1.0 785 372 1500 Mnt_ifmte, 12 Deposit 10 10 75 305-485 50 85 9.9 11.2 0.1 0.1-0.4 mils 786 373 1500 Heovy, 9 mils No deposit 35 20 60 200-600 40 95 10.6 2.6=-11.6 Ll 0.4 a All UF3. bZir:unium layer. “Individual results varied. dLnt:p drained while hot. Operation terminated because of power failure, of UF, and UF,. The next group of loops (473- 508 and 781-786 in Table 27) operated for varying times with mixtures containing various proportions of UF; to UF, and produced erratic results, The attack was reduced, but not elimi- nated, and the results were not consistent. Cold- leg deposits were found in almost all of these loops. The attack in the 2000-hr loop is much lower than would be found in an all-UF, loop, but drawing any conclusion concerning mass transfer from these data would be dangerous. An actual reduction in mass transfer rate would, however, be partially confirmed by loop 457, which operated with an all-UF ; mixture. The loops (627690 in Table 27) operated with batches made in the production facility, ond very little, if any, reduction in attack was found in these loops. Analyses of these batches have shown them to contain much less UF, than had been predicted. With both these loops and those discussed above, the attack in most cases does large increase in the chromium concentration of the fluorides. Even small ad- ditions of UF, must markedly lower the chromium solubility level. The attack must be predomi- nantly from mass transfer, No additional loops are being operated at this time with zirconium fluoride—base mixtures con- taining UF,. What is causing the production difficulties and why it has not been possible to not cause @ obtain corrosion results comparable with those obtained with the hydride addition, has not been explained, Both the production ond the dis- proportionation problems would have to be over- come to allow the use of such mixtures, Alkali-Metal