Catching the Culprits in Stainless Steel Corrosion

During the past 10 years, a number of publications have suggested that 316L ultralow-manganese stainless steel alloys (<0.05% Mn) offer superior corrosion resistance to more conventional low-manganese 316L alloys (<0.5% Mn) and standard 316L alloys (<2.0% Mn).^1-15 Several publications additionally claim that, during welding, the manganese evaporates from the weld pool and redeposits in the heat-affected zone, primarily downstream of the weld pool.^5-9

It is possible that this redeposited manganese is the cause of pitting corrosion in as-welded 316L alloys that are exposed to gaseous atmospheres containing halogens (i.e., fluorine, chlorine, bromine) and minute amounts of water vapor. The evaporation of manganese is supported by its higher vapor pressure than other metals in the alloy including iron, chromium, nickel and molybdenum.

Although the pitting corrosion argument has been widely published, convincing data and a meaningful thermodynamic-based discussion have not been provided to date. In this study, we investigated the corrosion behavior of ultralow- and low-manganese 316L alloys in different gaseous environments and sought to determine:

• Which elements evaporate during orbital autogenous welding of the 316L alloy test samples with different concentrations of manganese?
• How does a low-humidity (100 ppm H₂O), halogen-containing environment affect as-welded test samples during short-term (24 hr) and long-term (28 day) exposure?
• How does post-weld passivation affect corrosion resistance of samples exposed to a low-humidity, halogen-containing environment?

We performed the majority of the tests on two low-manganese and two ultralow-manganese VIM/VAR 316L SS alloys, with limited testing on a standard 316L AOD alloy (Table ).

Tube sections machined from bar stock or cut from electropolished (EP) tubing were orbitally welded into longer sticks. During welding, we collected and analyzed the weld fumes for chemical composition.

Tube sections 1.00 in. (25.4 mm) long, with 0.250 in. (6.4 mm) outside diameter (OD) and 0.035 in. (0.89 mm) wall thickness were machined from bar stock or cut from tubing, depending on product form. Alloy D tubing had a wall thickness of 1.0 mm. We squared the ends of the tubing samples using a Swagelok facing tool.

All tube sections were cleaned using an industrial alkaline detergent (10% strength) by fixturing the test specimens in a sieve and immersing them in an ultrasonic bath at 71°C for 15 minutes. The wash step was repeated twice with fresh solution. We then flushed the specimens with deionized (DI) water and placed them in a beaker of DI water with ultrasonic agitation for 10 minutes. Then we placed them in fresh DI water and repeated the wash step until the rinse water remained clear after rinsing. The specimens were then dried in an oven at 110°C for 30 minutes.

Sections machined from bar stock were electropolished, passivated in nitric acid, then rinsed and cleaned in an aqueous ultrasonic bath. The alloy D and E tubing were electropolished in the as-received condition.

We used several tube sections of each alloy to determine the welding conditions for producing welds according to Swagelok specifications. The root weld width equals approximately twice the wall thick-ness, causing it to fall in the range 0.052-0.088 in. (1.34-2.26 mm). Eleven tube sections of each alloy were welded into 11 in. (27.9 cm) long sticks containing 10 orbital welds each. We prepared three sticks for each of the four VIM/VAR alloys.

During welding, we flowed argon purge gas from the downstream tube section through a short piece of Tygon tubing into a non-fritted glass impinger. A solution of 15 mL of 2% nitric acid (HNO₃) in the impinger collected the weld fume particles. During welding, the inside diameter (ID) purge gas from the entire stick was passed through the acid solution. The impinger and hose were rinsed thoroughly, and the rinse water was added to the trap solution. Each stick has a separate trap solution.

To extract weld fume deposits from the inner surface of the weld sticks, each was heated in a convection oven at 48°C, capped at one end with a PTFE cap, filled with 5% HNO₃, and soaked for 15 minutes. This extract, together with the rinse water from the interior walls of each stick, was added to the trap solution. That solution was diluted with DI water to a final volume of 20 mL and analyzed by inductively coupled plasma atomic emission spectroscopy (ICP-AES) for iron, chromium, nickel, manganese and molybdenum against a reagent blank solution.

The results show that none of the alloys emitted detectable levels of chromium, nickel or molybdenum during welding (Fig. 1 ). We determined the amount of iron and manganese per weld by dividing the total measured amounts by 10 (the number of welds per stick). The lower detection limit of the analysis method was 0.4 µg per element for each of the trap solutions, which amounts to 0.04 µg per weld. The reproducibility of the weld fume results is very good for sticks of the same material.

1. None of the alloys emitted detectable levels of chromium, nickel or molybdenum. The amounts of iron and manganese in each stick were very reproducible between the three samples.

To determine whether any iron or manganese was leached from the tube walls and reacted with dissolved weld fume deposits, we performed extraction tests on as-received alloy D and AOD alloy E tubing. These tests showed that small amounts of iron were leached from unwelded tubing. We estimate that the amount of iron shown in Figure 1 is maximally 10% higher than the amount of iron traceable to weld fume deposits only. The extraction tests on the unwelded tubing showed that manganese was not extracted during the nitric acid soak. Hence, the manganese amounts shown in Figure 1 are traceable to weld fume deposits only.

The low- and ultralow-manganese 316L VIM/VAR alloys emitted similar amounts of iron during autogenous orbital welding. Compared to the amount of iron, a relatively small amount of manganese was emitted from the low-manganese alloys A and B. The amount of manganese emitted from the ultralow-manganese alloys C and D was below the detection limit of the ICP-AES analysis technique. The amount of iron emitted during welding increased when higher power levels were necessary to obtain welds meeting the specification for weld bead width (Fig. 2 ).

2. To meet the specifications for weld bead width, higher power levels became necessary, which drove up the amount of emitted iron per weld (normalized to weld bead width at tube ID).

Exposure to corrosive gases

To measure the effects of corrosive gases in the presence of low moisture concentration, we prepared tube sections as described above, except that the individual tube sections for this test had a length of 1.5 in. (38.1 mm). The sections were welded to Swagelok specifications into 3.0 in. (76.2 mm) long test samples using a gas mixture of 95% argon and 5% hydrogen (H₂) as the purge and shield gas. The flow rate of the high-purity gas mixture (2 ppm O₂) through the tubes was 12.5 ft³/hr (354 L/hr). Orbital weld speed was 10 rpm. Ceriated tungsten electrodes 0.08 in. (2.0 mm) in diameter were used to weld with an arc gap of 0.03 in. (0.76 mm).

We chose a corrosive gas flow bench for exposing specimens to a relatively low-moisture environment (100 ppm) containing 5% of the corrosive gas, either chlorine (99.997% pure) or hydrogen chloride (99.995%) and nitrogen (99.999%). Gas flows were controlled by electronic mass flow controllers, which were protected with in-line filters.

In the corrosive gas flow bench, the dry carrier gas (nitrogen) is split into two streams and used either as dry diluent gas or as the carrier gas to deliver moisture. The system is also plumbed to provide total system dry-down by introducing dry nitrogen via a purge assembly located between the corrosive gas source cylinder and its gas regulator.

We produced the desired moisture condition by flowing the dry nitrogen through a permeation chamber containing a 20 cm long permeation device, calibrated to deliver 2 ng H₂O/min per centimeter of device length at 100°C. The permeation device releases a given amount of moisture according to temperature — independent of flow rate. The corrosive gas stream was maintained at 5% of the total gas volume and all gas pressures were set at 20 psig (1.3 bar).

The test chamber consisted of a horizontal 12 × 1.5 in. (305 × 38.1 mm) OD Pyrex glass tube, connected to Swagelok Ultra-Torr fittings at either end. Specimens were positioned in the chamber by removing a fitting at one end and sliding in or out a Pyrex glass cradle that supported the specimens in a horizontal position. The moisture was monitored online using a digital hygrometer located immediately prior to corrosive gas introduction. A second hygrometer monitored moisture downstream from the chamber during system dry-down.

3. After welding, the samples were cut open to investigate corrosion damage.

We performed corrosion testing on as-welded and post-weld passivated tubular samples of alloys A-E. After welding, the 3 in. (76.2 mm) long samples were cut open (Fig. 3 ) to facilitate characterization of the internal surfaces following exposure to the corrosive gas mixtures. The samples were exposed to the flowing HCl-containing and Cl₂-containing gas mixtures for 24 hours and 28 days, respectively.

Following the 24-hour exposure to the flowing, HCl-containing and Cl₂-containing gas mixtures, the low-manganese samples showed some discoloration just downstream of the weld zone. However, the discolored band was barely noticeable. It was even more difficult to detect any discoloration in the ultralow-manganese samples. The AOD alloy E sample developed a more pronounced yellow-brown band.

We then characterized the sample surfaces by scanning electron microscopy (SEM). All samples showed the presence of submicron particles, primarily downstream of the weld zone. EDS of the particles indicated the presence of chlorine. Identification of metal constituents was inconclusive because the small particles were in intimate contact with the alloy surface.

After the particles had been rinsed off with distilled water, we examined the tube samples by SEM for signs of corrosion, but none was found. No brown bands were observed in any of the post-weld passivated samples, and SEM examination found no particles or signs of corrosion.

The presence of a yellow-brown band on the exposed AOD alloy E sample may be a result of the evaporation and redeposition of manganese. The manganese content of AOD alloy E was 1.57%, and that of alloys A and B was only ~0.3%, leading to the expectation that a substantially larger amount of manganese would evaporate and redeposit during the welding of AOD alloy E. Also, the number and density of the observed particles were highest for the AOD alloy E samples. The brown band appears to be related to the larger number of chlorine-rich crystals observed downstream of the weld zone.

The deposits of neither iron nor manganese led to any signs of corrosion in the alloy samples exposed for 24 hours. This finding suggests that, under 24-hour test conditions, the presence of manganese in the weld fume deposits did not lead to corrosion of the low-manganese alloy samples.

Following the 28-day exposure of welded tubes of alloy A, B and D to the HCl-containing gas mixture, all samples showed a band of discoloration upstream and downstream of the weld zone. The bandwidth was ~0.4 in. (10.2 mm) on either side of the weld zone. SEM analysis showed that the bands consisted of individual particles ~1-10 µm in size. The areas between particles appear to have been severely etched during exposure, and showed signs of pitting corrosion. In locations where the particles had been rinsed off with distilled water, the substrate surface appeared to be similarly etched and pitted (Fig. 4 ).

4. SEM analysis of low-manganese (A and B) and ultralow-manganese (D) alloys revealed a similar pattern of pitting corrosion downstream of the weld zone.

The samples were also characterized by SEM in areas farther downstream of the weld zone, ~1.0 in. beyond the weld. Very small (0.1 µm diameter), isolated particles were detected. The surfaces between the particles appeared featureless and resembled the surfaces of as-welded (unexposed) tubes. The extremely small size of the particles made it impossible to determine whether they were associated with any signs of corrosion.

The most significant finding of the 28-day corrosion test is that both ultralow-manganese (D) and low-manganese (A and B) alloys experienced comparable attack on surfaces in close proximity (0.4 in.) to the weld zone (Fig. 4 ). Particles with similar morphology formed in the three alloys and, upon rinsing with distilled water, caused remnants to stay behind. The surfaces of the alloys, both between the particles and at locations where the particles had been located (prior to rinsing), appeared etched and pitted.

Based on the SEM observations, all three alloys showed signs of similar, significant corrosion in locations where the discolored bands had been observed. No such corrosive attack was observed in locations beyond the bands. Hence, it is likely that corrosion was caused by the presence of iron weld fume deposits. SEM analysis of the samples did not provide any evidence that the presence of manganese in the weld fume deposits accelerated or exacerbated corrosive attack.

Based on the results from the weld fume analysis and the corrosion tests, we determined that:

• During autogenous orbital welding of 316L SS tubes, the molten weld pool reaches temperatures high enough for measurable quantities of alloy constituents to evaporate. As long as the manganese concentration of the alloys is sufficiently low, more iron evaporates than manganese. The emitted metals redeposit on the colder tube surfaces adjacent to the weld zone.
• The iron-rich surface film adjacent to the weld zone serves as an initiation site for corrosive attack on the tubes. Most likely, iron reacts with hydrogen chloride or chlorine to form iron chloride. In the presence of moisture, the chloride hydrates. This reaction liberates hydrochloric acid that subsequently etches the surface of the alloy.
• Because the low- and ultralow-manganese alloys released about the same amount of iron during welding, and because the corrosion tests led to very similar observations in all the alloys, it is believed that the iron-rich weld deposits adjacent to the weld zones are responsible for the observed corrosion behavior of welded samples made from either type of alloy.