Failure of Steam-Water Heat Exchangers

B. F. Brown, U. S. Naval Research Laboratory

From: B.F. Brown, Failure of Steam-Water Heat Exchangers, Metals Engineering Quarterly, Vol 1 (No 1) Feb 1961 p 68–74

Abstract: Gross wastage and embrittlement were observed in plain carbon steel desuperheaters in five new Naval power plants. The gross wastage could be duplicated in laboratory bomb tests using sodium hydroxide solutions and was concluded to be caused by free caustic concentrated by high heat flux. The embrittlement was shown to be caused by the flow of corrosion-generated hydrogen which converted the cementite to methane which nucleated voids in the steel. A thermodynamic estimate indicated that a small amount of chromium would stabilize the carbides against decomposition by hydrogen in this temperature range, and laboratory tests with 2-14% Cr steel verified this.

Keywords: Boiler tubes; Corrosion environments; Sodium hydroxide; Warships

Material: Fe-0.13C (Nonresulfurized carbon steel)

Failure type: Hydrogen damage and embrittlement


Except to those directly concerned with the service failure, a report of analysis of such a failure is apt to make prosaic reading unless the failure was spectacular, or the analysis required brilliant metallurgical sleuthing, or in the process of analysis new basic information of general import to the field of metallurgy was uncovered. These characteristics are rare in the average failure analysis, and they do not apply to the one reported here.
Even to those not involved directly, however, failure analyses may provide useful secondary services by pointing out gaps in fundamental metallurgical information of present practical need. (If those who perform uncommitted research in metallurgy made an occasional failure analysis, or occasionally read a report of one, some of these investigators might recognize new and fertile areas of fundamental importance.) Additionally, a failure analysis report may be useful to remind management of the breadth of capabilities of research establishment which may be required to analyze expeditiously a reasonably simple service failure. Finally, the present report may serve to remind anyone who may need such reminding that a large proportion of service failures submitted to metallurgists involve either corrosion or mechanical metallurgy or both — areas of metallurgy that are emphasized in few academic curricula and that, somehow, unfortunately fail to arouse enthusiastic pursuit by many students of metallurgy.

Service Environment

The subject service failures occurred in boiler desuperheaters in five modern Naval power plants of compact construction and having therefore high fluxes through heat-transfer surfaces. These desuperheaters consist of steel tubing submerged in the water of the steam drum. Superheated steam, at 950 F in most of these power plants, enters one end of the tubing and, during its passage, transfers heat through the tube wall to the water of the steam drum, which in most of these was at 550 F. The effect of this passage is to bring the superheated steam down to approximately the temperature of the water in the steam drum, and if the heat flux through the tube wall is sufficient, it causes boiling on the water side.
One of the service failures is shown in Fig. 1. The steam inlet flange is seen to the left and the exit flange to the right. This particular unit illustrates the two failure aspects observed in all the units, namely extensive corrosion in a few hundred steaming hours, and embrittlement to a degree which led to breaking off the inlet flange during disassembly - a degree of embrittlement uncommon in the low carbon unalloyed steel of which these desuperheaters are made.
Figure 1

Fig. 1  Failed desuperheater unit (about eight feet in length).

The failure analysis is then divided into searches for (1) the cause of the rapid corrosion and (2) the cause of the embrittlement.

Failure Analysis

Cause of Corrosion

The corrosion products on the water-sides of all tubes consisted largely of loose, poorly-adherent magnetite (as identified by x-ray diffraction) together with crystals of elemental copper and, in some instances, some Fe2O3 (which probably formed after shut-down). Only thin, tightly-adherent films of magnetite were observed on the steam-sides.
These desuperheaters were operating in boiler water said * to contain enough free caustic (plus some phosphate) to bring the pH to 11; rapid corrosion to form magnetite would hardly be expected in such solutions. Furthermore the generating tubes of these same boilers, exposed to the same water at the same temperature, did not suffer the same attack. A clue to the cause is the general pattern of the attack in the desuperheater of Fig. 1, where the corrosion is most severe at the inlet end, with little attack in the central zone where the circulation of water in the steam drum is high, but severe again near the downstream end of the first pass (far right-hand side). The attack is also more severe near the tube support plates than on either side. Judging from appearance of several desuperheaters in various stages of deterioration, the advance of the corrosion attack was toward the steam exit end. It thus appears that the more severe attack occurs (a) where the steam is hotter, and (b) where circulation is poorer.
One might speculate that boiling on the desuperheater surface might damage by thermal shock the magnetite film depended upon for protection against corrosion, but this can hardly have been the case: the generating tubes of the same boiler certainly experienced thermal shocks certainly no less severe without exhibiting the same rapid corrosion. A second effect of boiling is concentration of the boiler-water chemicals. For reasons not well understood, the magnetite produced in highly caustic solutions tends to be nonprotective. Whether this involves the dissolution of the iron as the soluble FeOOH- complex, which later transforms to magnetite, or whether simply the magnetite which forms in caustic solutions is physically defective, is not known. But in either case the primary causative of the abnormally high corrosion rate is postulated to be free caustic.

Cause of Embrittlement.

We now turn to the search for the cause of the embrittlement. Tensile ductility and Charpy impact data were not taken on the failure metal for the simple reason that the qualitative evidence for embrittlement was convincing enough. The fracture of Fig. 1 is suggestive. In a more severely deteriorated unit from another boiler with 1100 steaming hours (less than 2 months), the inlet zone was so badly embrittled that pieces of the remaining metallic material could be broken out by hand.
The analysis of the metal from a unattacked area of a typical desuperheater is shown in Table 1.

Table 1   Analysis of Tube from an Unattacked Area of a Desuperheater

The microstructure of such steel in the unattacked areas is the typical ferrite plus patches of fine, rather spherical carbides which one commonly sees in low carbon steels. The microstructure of the remaining metal under the severely attacked areas, however, is quite different. A typical example is shown in Fig. 2.
Figure 2

Fig. 2  Microstructure of center of attacked tube wall. 250×

Most of the carbides have disappeared, and there is extensive void formation. Most surprising of all, when one recalls that the unattacked steel shows 1 or 2 ppm hydrogen content, the steel under the severely attacked areas in most desuperheaters was found to contain more than 100 ppm of hydrogen, and in one case, 268 ppm. The clue to the disappearance of the carbides was provided by a mass spectrographic analysis of gas released by (1) drilling the embrittled metal under mercury, and (2) dissolving the embrittled metal in aqueous cupric chloride. The analysis showed that about one-third of the gas was methane, the remainder mostly hydrogen. The hydrogen released by corrosion had decomposed the cementite.
The mechanical properties of low carbon steel are not drastically deteriorated by hydrogen in solution, and the embrittlement noted here is attributed largely to the extensive internal “notching” provided by the crack-like voids, which may be considered similar in their effect on ductility to the graphite flakes in gray cast iron. The cause of the embrittlement, then, is not so much the absorption of hydrogen as the formation of the crack-like voids. Evidence will be given later to support the postulate that these voids are formed because of the formation of methane, not because of the large amount of hydrogen per se.
The methane is formed by the reaction
Fe3C + 2H2→ CH4 + 3Fe
which is sometimes referred to as the methanation reaction. The heat treater who has used methane to carburize steel may be surprised at first to see the reaction going in the other direction. The reason for the reversal is the temperature effect; estimating from the free-energy data selected by Kubaschewski and Evans, the temperature of reversal is about 875° for equilibrium conditions. Below this temperature, the lower the temperature, the higher will be the driving force for the reaction to go to the right, and it should be higher at room temperature than at boiler temperatures. But a mild steel tube, periodically refilled with HCl plus a trace of phosphorus until it had corroded one-third through the tube wall at room temperature, presumably with a fairly high hydrogen flux through the wall toward the abraded outer surface, showed no signs of void formation. A clue to the reason for this is suggested in Fig. 3, which shows the microstructure of a severely corroded steam inlet flange from a desuperheater. This is cast steel having, originally, a microstructure of pearlite plus grain-boundary ferrite and coarse Widmanstatten ferrite. Voids have formed in the grain-boundary ferrite, and the adjacent pearlite is beginning to decompose, with the carbon diffusing to the void to form methane. It is noted in all these embrittled specimens that there are very many fewer voids than there were originally carbide particles. Carbon must therefore diffuse a considerable distance, and the positive temperature coefficient of diffusion of carbon would thus be expected to tend to offset the negative temperature coefficient of the energetics of the methanation, so that at a sufficiently low temperature this reaction will not take place to any important extent.
Figure 3

Fig. 3  Section of attacked cast steel flange showing voids surrounded by ferrite. 500×

The postulated mechanisms for the corrosion and embrittlement of the present service failures are summarized schematically in Fig. 4. Rapid boiling concentrates the caustic at the surface of the tubes causing accelerated corrosion of the iron, which forms nonprotective magnetite deposits; the cathodic partial-reaction corresponding to the anodic oxidation of iron is the reduction of hydrogen, which dissolved in the steel; rapid corrosion means high hydrogen levels in the steel, and high temperatures permit high diffusivity of carbon to a void where, with the ubiquitous hydrogen, it can react to generate methane. Under the energetics of the methanation reaction, the voids grow to cracks readily visible under the microscope, and these serve as embrittling internal stress-raisers.
Figure 4

Fig. 4  Schematic summary of mechanisms of deterioration.

Laboratory Tests

Several tests were performed in the laboratory to serve as a guide check on some of the above. Ideally, these tests should have involved boiling, but because of the complexity of apparatus required to perform such tests at high temperatures and, therefore, high pressures, a simplified bomb system was used instead. Rods 2 in. long and $\genfrac{}{}{0.1ex}{}{1}{2}$in. diam were drilled to provide a longitudinal cavity of about 0.3 in. diam, leaving tube walls about 0.1 in. thick. These rods were of annealed plain carbon steel. Concentrated solutions were prepared in such proportion that at the same dilution, all would show room-temperature pH of about 11. These solutions included a solution with a high ratio of NaOH to Na3PO4, a high ratio of Na3PO4 to NaOH, and, in a separate series, 40% NaOH of reagent grade.
The solutions were introduced into the bombs, the end were closed with pressure-fit plugs, and the closures were welded shut. The bombs were then inserted into a furnace at 950 F** for 24 hr, removed, sectioned, and examined metallographically. The essential findings are shown by Fig. 5a and 5b.
Figure 5

Fig. 5  (a) Section through wall of plain carbon steel bomb after 24-hr exposure at 950 F to concentrated high free-caustic solution. Corroding surface, at left, shows absence of carbides and presence of voids. (b) Same as (a) except solution was high in phosphate and low in caustic. 100×

Figure 5a shows complete decomposition of the pearlite and cracking toward the corroding surface in the solution containing a high concentration of free caustic. Figure 5b shows the absence of this attack in the high-phosphate solution. This was a failure analysis, not a research and development project, so the corrosion rates were not measured, but subjectively there appeared to be a much thicker layer of oxide in the specimen of Fig. 5a than that of Fig. 5b.
A test of the concept that it is the methane, not the hydrogen, that caused the embrittlement was made indirectly by selecting a steel that should have carbides thermodynamically stable to hydrogen.

Selection for Specific Environment

Estimates were made (from the data tabulated by Kubaschewski and Evans) of the standard free energy of the methanation reactions for Fe3C and for Cr23C6, and from these a “rule-of-mixtures” estimate indicated that at 500 K a carbide containing roughly 60% of chromium might be stable to hydrogen. Using the distribution ratio of Hultgren and co-workers for chromium in carbide: chromium in ferrite of 28:1, one is led to conclude that slightly more than 2% Cr (in a low-carbon steel) could reasonably be expected to stabilize the carbides against methanation. Conveniently, there is available a commercial low-carbon steel containing $\mathrm{2-}\genfrac{}{}{0.1ex}{}{1}{4}$% Cr.
Bombs similar to the ones just described were made of the $\mathrm{2-}\genfrac{}{}{0.1ex}{}{1}{4}$% Cr steel and were tested with the high free-caustic solutions. Again corrosion rates were not measured, but subjectively the specimen of Fig. 6 had a heavy scale on the inside comparable with that on the specimen of Fig. 5a, except that the carbides were intact throughout, but thinner scale than on the specimen in Fig. 5b.
Figure 6

Fig. 6  Same as Fig. 5a except steel containing about $\mathrm{2-}\genfrac{}{}{0.1ex}{}{1}{4}$% Cr. 100×

Although the relatively low chromium content might reasonably be expected to stabilize against methanation, it is far too low to be expected to protect against normal corrosion. Also, in high pH solutions chromium is taken up into a soluble complex, so that it would not be expected to prevent corrosion due to concentrated caustic. Its role thus is not to prevent corrosion but to prevent the catastrophic embrittlement of the uncorroded steel.
The arguments above suggest an even simpler solution to the cracking problem where strength is not an important consideration: use a “steel” with no carbon in it. To test this, a bomb was made of Armco ingot iron and the experiment for the specimen shown in Fig. 5 was duplicated, with the finding that no cracking was observed.
Nickel and nickel alloys are often used to contain NaOH solutions at elevated temperatures, but the data available indicated that for conferring protection to steel against attack by free caustic, the nickel content would probably have to be so high that the steel would be austenitic and therefore susceptible to unrelated stress-corrosion cracking. This implies that the alloy approach is a less desirable one than the water chemistry approach suggested by the phosphate tests, if adequate control can be maintained under the less-than-ideal conditions under which these power plants are required to operate.


The analysis, then, is concluded. The cause of the rapid corrosion, high free-caustic; the cause of the embrittlement, formation of microcracks by methane generated by hydrogen reacting with carbides in the steel. These are not proven with the rigor one expects in a research project, as distinguished from a failure analysis, but they appear to provide simple and adequate explanations of all known facts. They also carry a clear-cut implied solution: avoid high free-caustic as by the balanced phosphate method, which should obviate both the gross corrosion and the embrittlement caused by the corrosion-generated hydrogen.
It is not suggested that high phosphate treatments will guarantee freedom from all ills in high heat-flux systems. A review of the literature reveals that we know too little of the complex phosphate chemistry of iron at high temperatures to be sure that we will not some day encounter some ugly surprises in such systems.
To avoid ending this report on such a negative note, it is suggested that we can apply some of the arguments above to explain (a) why caustic-embrittlement cracking occurs in boilers, and (b) why we can get away with cathodic protection of mild steel in sea water and wet soil for years without causing embrittlement. The actual caustic cracking as illustrated by micrographs in the literature looks very much indeed like the incipient stages of the sort of cracking observed in this present study, that is, cracking by methane formation, which could be prevented by balanced phosphate treatment (as is practiced in some quarters) or by the use of a low chromium steel, as suggested by the present experiments. In the case of cathodic protection, hydrogen is being reduced on the protected steel surface at a rate that may be comparable to the rates experienced in the present failures. The diffusivity of the hydrogen is certainly adequate to move it throughout the steel; but at ambient temperature the carbon cannot move to the inclusion, high-angle grain boundaries, or whatever special site is required to permit reaction with the hydrogen and therewith the nucleation and growth of the damaging voids. Even if the carbon could diffuse, the energy available to activate the reaction is so limited at room temperature that this reaction, like the room-temperature reaction between charcoal and hydrogen, would be expected to be very slow.


The author wishes to acknowledge with appreciation the metallographic analyses of the desuperheaters by Mr. A. J. Edwards, the conduct and metallographic analyses of the “bomb” tests by Mr. R. Newbegin, the chemical analyses by Mr. D. Walter, and helpful discussions with Drs. M. C. Bloom and A. J. Pollard. The work was supported by Code 651 of the Bureau of Ships.

Related Information

D.N. French, Failures of Boilers and Related Equipment, Failure Analysis and Prevention, Vol 11, ASM Handbook, ASM International, 1986, p 602–627
R.J. Franco, Failures of Heat Exchangers, Failure Analysis and Prevention, Vol 11, ASM Handbook, ASM International, 1986, p 628–642
Hydrogen Damage and Embrittlement, Failure Analysis and Prevention, Vol 11, ASM Handbook, ASM International, 2002, p 809–822


This brings up the subject of two of the difficulties under which the failure analyst labors: The total failure conditions are seldom known, and what little information is known is apt to be transmitted through one or many intermediaries, often verbally. Failure condition information should be sought directly from the operators by the analyst himself, at the failure site, where possible.
The reason the methanation reaction might be expected to occur so high above the reversal temperature may be attributed only partly to inaccurate thermodynamic data - the more important reason is undoubtedly that in these experiments thermodynamic equilibrium was not achieved.