Dissimilar metal welds (DMW): the challenges

Dissimilar metal welding (DMW) is one of the most critical and most misunderstood topics in modern process and power plants. Every point where a carbon steel pipe meets a 316L flange, where a low-alloy boiler tube meets an austenitic superheater, or where a 2.25Cr-1Mo header meets a nickel-based transition piece is really a compromise zone — two metallurgical worlds forced to coexist. Filling the joint is the easy part. The real problem is predicting what different chemistries, different thermal expansion behavior and a microstructure that keeps changing at service temperature will do to each other over the long run. This article covers the real field challenges of DMW: filler metal selection, dilution, carbon migration, fusion line discontinuities, and the metallurgical reasons why high-temperature applications move to nickel-based filler metal.

1. Why DMW is not a "normal" weld

When you join two metals of the same type, the weld metal, the heat-affected zone (HAZ) and the base metal all come from broadly the same alloy family. Expansion, melting and transformation behavior are close to each other. In DMW, at least two different base metals — and usually a third chemistry, the filler metal — mix in the same molten pool. The result is not a single "weld" but several sharply separated micro-regions: the ferritic side, the transition zone formed by the filler metal, the austenitic side, and the narrow but decisive fusion lines between them. Each region shows different hardness, different toughness and different corrosion behavior. The classic carbon steel–304/316 joint demands a careful procedure for exactly this reason. Both metals weld well on their own, but their joint can reveal brittleness that neither shows alone.

2. Ferritic–austenitic and low-alloy–stainless joints

The most common DMW family is a ferritic or low-alloy steel joined to an austenitic stainless. The ferritic/low-alloy side — carbon steel, C-Mn, 1.25Cr-0.5Mo, 2.25Cr-1Mo, mostly ferritic-bainitic or tempered bainitic in the normalized-and-tempered or quenched-and-tempered condition — has a body-centered cubic structure with low thermal expansion and high thermal conductivity. The austenitic side (304, 316, 321, 347) has a face-centered cubic structure with markedly higher thermal expansion and lower thermal conductivity. Join these two worlds and the weld metal chemistry sits on a very fine balance that decides which phase forms during solidification.

The dominant risk in this joint is not solidification (hot) cracking, as many people assume. The real danger is that dilution from the carbon steel side drives the weld metal to form martensite, producing a hard and brittle interface. Hot cracking is secondary, and it belongs more to similar-metal austenitic welds: choose the wrong filler metal, leave the austenitic weld metal without enough primary ferrite, and the solidification cracking risk rises. The correct filler metal forms primary ferrite during solidification. Ferrite dissolves more sulfur and phosphorus than austenite, so it prevents continuous low-melting films at the grain boundaries and makes the microstructure crack-resistant. This ferrite balance matters in DMW too, but the primary fight is against the martensitic interface.

3. Filler metal selection: 309L and nickel-based fillers

Filler metal selection is the single most important decision governing the life of a dissimilar joint. When you join carbon steel or low-alloy steel to austenitic stainless, the classic and correct choice is the E/ER 309L group. Among its variants, 309LMo adds molybdenum, so on joints to 316/316L it preserves the Mo level in the weld metal after dilution and keeps pitting and crevice corrosion resistance intact; its contribution to strength is negligible. 309LSi raises silicon to improve wetting and puddle fluidity — its benefit is about workmanship and weld pool behavior rather than dilution or strength. The high chromium and nickel content of 309L (roughly 23Cr-13Ni) keeps the weld metal in an austenitic-ferritic balance despite dilution from the carbon steel side, so it retains toughness without forming martensite. 308L (roughly 20Cr-10Ni) is not adequate for this job, and putting the two numbers side by side shows why: when 308L is diluted with carbon steel, the chromium and nickel equivalents drop far enough to reach the martensite field, and a hard, brittle interface appears. 309L also has a service temperature limit. As a high-ferrite austenitic weld metal, it is prone to embrittlement under long-term exposure at continuous service temperatures above roughly 500 °C; section 9 explains the mechanism behind that limit.

As service temperature rises, or as the expansion mismatch between the two sides grows, the choice moves to nickel-based filler metals: ERNiCr-3 (commonly known as Alloy 82), ENiCrFe-3 (Alloy 182) and the ERNiCrMo types. The composition of this family also corrects a common misconception. ERNiCr-3 contains roughly 18–22% chromium and ENiCrFe-3 roughly 13–17%, so chromium — a strong carbide former — is abundant in these weld metals as well. The advantage of nickel-based filler metal is not the absence of chromium. The first advantage is diffusion kinetics: carbon has markedly lower diffusivity and solubility in a nickel-based face-centered cubic matrix, so the carbon activity difference and the diffusion rate between the ferritic side and the weld metal both fall, and the carbon migration band develops more slowly. The second advantage is thermal expansion. Averaged over 20–100 °C, the expansion coefficient of nickel-based weld metal is about (13.3–13.9) × 10⁻⁶ /°C; averaged over 20–500 °C it is around 15 × 10⁻⁶ /°C. On the same basis, carbon/ferritic steel is about (11.7–12) × 10⁻⁶ /°C over 20–100 °C, and austenitic 304/316 about (16.6–17.2) × 10⁻⁶ /°C. Nickel-based weld metal therefore sits between the two sides: near room temperature it stays closer to the ferritic side, and around 425–500 °C (ferritic ~13.7 / nickel-based ~15.1 / austenitic ~17.6 × 10⁻⁶ /°C) it sits almost exactly in the middle. In services sensitive to stress corrosion cracking, such as nuclear and pressurized water environments, the 82/182 family has largely been replaced by ERNiCrFe-7A / ENiCrFe-7 (Alloy 52/52M/152). The practical split is this: use the ERNiCr-3 type for high-temperature creep transitions, and the 52/152 type in water chemistries that carry PWSCC risk.

4. Dilution: the invisible design variable

Dilution measures how much of the weld metal comes from melted base metal, and in DMW it sets the weld chemistry directly. High current, narrow grooves, deep penetration and heavy single-pass deposition all raise dilution, and the root pass is usually where dilution peaks. Typical ranges by process are roughly: GTAW 10–20%, SMAW 20–30%, GMAW 20–35%, SAW 30–60%. The practical threshold is equally clear: 309L weld metal tolerates about 25–30% dilution from the carbon steel side without martensite, but once dilution passes the 30–40% band, the composition crosses into the martensite field on the Schaeffler diagram and a hard band appears along the fusion line.

Excessive dilution from the carbon steel side lowers the chromium and nickel equivalents of the 309L weld metal and pushes the microstructure into the martensite field of the Schaeffler diagram. The choice of diagram matters here. The DeLong diagram does not show the martensite field; it focuses on ferrite number (FN) prediction in the near-austenitic region. The WRC-1992 diagram predicts ferrite number more accurately than DeLong, but on its own it does not show the martensite field either. To read dilution-driven martensite risk, use either the Schaeffler diagram or the martensite boundary lines Kotecki added to WRC-1992. In practice, high dilution means a hard, crack-prone region in the root pass. The field methods for managing dilution are: deposit the root pass at lower heat input, apply buttering, control electrode angle and weave, and switch to nickel-based filler metal at critical transitions. A good procedure does not leave dilution to chance — it constrains it inside the WPS through heat input and position variables.

5. Buttering and the transition piece logic

For high-risk dissimilar joints, the most reliable approach is not to butt-weld the two metals directly, but to butter one side first. For example, clad the groove face of the ferritic base metal with nickel-based filler metal, apply the post-weld heat treatment (PWHT) that the ferritic material requires to the buttered ferritic component if needed, and then make the actual joint between the austenitic side and the buttered layer. This approach brings two major benefits. When a nickel-based butter layer is used, the direct ferritic–austenitic contact that triggers carbon migration disappears — note that buttering with 309L does not remove the migration interface, it only moves it. And the required PWHT can be applied to the ferritic side alone, without exposing the austenitic base metal to excessive heat. Power plants build their permanent transition joints on exactly this logic: the shop makes them under controlled conditions and leaves only similar-metal joints for the field.

6. Carbon migration, decarburized and carburized bands

The most insidious damage mechanism in DMW is carbon migration. When ferritic or low-alloy steel stays in contact with a high-chromium austenitic metal at high temperature for a long time, carbon migrates from the low-chromium side to the high-chromium side because of the chemical potential difference. The driving force is not an "attraction" but a thermodynamic activity difference: chromium lowers the activity of carbon in solution, so a carbon activity gradient forms between the two sides, and carbon diffuses toward the high-chromium side to close that gradient. The result is a thin decarburized band on the ferritic side adjacent to the fusion line, and a carbide-rich carburized band on the austenitic/filler metal side. The decarburized band softens and loses creep strength; the carburized band hardens and embrittles. Together the two bands form a weak plane where creep cavities nucleate in service.

Carbon migration is diffusion controlled. It advances exponentially with temperature (Arrhenius) and, in the early stage, roughly with the square root of time (band thickness ~√(Dt), parabolic). That parabolic relation is an idealized diffusion approximation. Under long exposure the carburized band saturates with chromium carbides, the carbides coarsen and chromium depletion progresses on the ferritic side, so growth deviates from parabolic and slows down. Even so, a small rise in temperature accelerates migration significantly. In high-temperature DMW, the most effective countermeasure is to use nickel-based filler metal, in which carbon diffusivity and solubility are low, and so reduce both the carbon activity difference and the diffusion rate at the interface from the start.

7. Thermal expansion mismatch and service damage

Near room temperature the thermal expansion coefficient of austenitic stainless is about 40–45% higher than that of ferritic steel: averaged over 20–100 °C, ferritic/carbon steel is about (11.5–11.7) × 10⁻⁶ /°C while austenitic 304/316 is about (16.6–17.2) × 10⁻⁶ /°C. Always state which range these averages cover, because the two coefficients converge somewhat as temperature rises: over 20–500 °C, carbon steel averages ~13.9 and 304 averages ~18.4 × 10⁻⁶ /°C; the absolute difference falls from 5.5 to ~4.5 and the ratio from 1.47 to 1.32. So saying "the gap widens further at service temperature" is wrong. The difference persists across the whole service range; what actually grows as ΔT increases is the accumulated total free expansion difference (the α·ΔT product), and therefore the stress generated at the interface. The quantity that drives the damage mechanism is that product, not the coefficient difference itself.

The difference looks trivial on paper, but it produces real stress at the interface on every heating and cooling cycle. Through start-up, load change and shutdown cycles, the joint sees a repeated shear stress caused by the two metals expanding and contracting at different rates. Combined with low-cycle fatigue and creep-fatigue interaction, this initiates cracks near the fusion line. This is the second major value of nickel-based filler metal: because its expansion coefficient is close to ferritic steel, the peak thermal strain moves away from the ferritic fusion line weakened by carbon migration and shifts to the more ductile, more creep-resistant austenitic–weld metal interface. Even so, the designer should place the DMW point in a region with the lowest possible stress and thermal gradient. Even the best filler metal cannot rescue a bad location.

8. Fusion line problems and Type II boundaries

Most DMW failures concentrate not in the middle of the weld metal but in the narrow fusion line between the ferritic base metal and the weld metal. Chemistry changes rapidly between base metal and weld metal here, creating a "transition zone" and a set of special grain boundaries that go with it.

The most characteristic feature of this transition zone is incomplete mixing in the physical sense. Immediately on the weld metal side of the fusion boundary lies a thin layer where the base metal melted and solidified without mixing with the filler metal: the unmixed zone. Chemically this layer is base metal — on the ferritic side it has carbon steel chemistry — and it can transform to martensite on rapid cooling. Beyond it, in the partially mixed zone, filler metal and base metal mix in locally varying proportions. The macro-segregation morphologies known in metallography as islands, peninsulas and beaches are products of this region. These structures are locally high in dilution and therefore contain high-hardness martensite, and they rank among the leading crack-initiation sites in DMW. The martensitic band and the Type II boundary described below are products of exactly this region.

The structures known in the literature as Type I and Type II boundaries differ in orientation, and that distinction is critical. Type I boundaries run almost perpendicular to the fusion line — they are the continuation of solidification grain boundaries and epitaxial growth. The real risk lies with Type II boundaries, which run parallel to the fusion line in a band immediately adjacent to it — typically a few tens of micrometers away, generally closer than 100 µm. Do not quote a single point value; it creates false precision. Under long-term high-temperature exposure (thermal aging) the band can also migrate away from the fusion line. The formation mechanism is known: the ferritic base metal undergoes an allotropic transformation (δ/α→γ) while the weld metal does not; at high temperature a γ boundary forms parallel to the fusion boundary and stays in place on cooling. For this reason the Type II boundary does not appear in every DMW — the base metal must show an allotropic transformation and the weld metal must solidify as primary austenite. In weld metals that solidify with primary ferrite, such as 309L, its formation is suppressed. Once weakened by carbon migration, these boundaries offer a ready path for creep and stress corrosion cracking. Treating both Type I and Type II boundaries as "parallel" is metallurgically wrong.

The second classic fabrication problem in this region is hydrogen. Hydrogen has high solubility and low diffusivity in austenitic and nickel-based weld metal. Hydrogen that passes from the weld metal into the ferritic HAZ is trapped at the fusion line and, together with the hard martensitic band, produces delayed cracking. The countermeasures: use low-hydrogen, dry consumables; keep electrode and wire baking and storage discipline; maintain preheat without interruption throughout buttering; apply hydrogen bake-out where needed; and delay the final examination by at least 48 hours.

This region demands special attention during examination. Take hardness traverses at close intervals along the fusion line, identify hardness spikes, and examine the transition zone by metallographic replication where necessary. Radiography and ultrasonic examination catch volumetric discontinuities, but austenitic and nickel-based weld metal is anisotropic and coarse-grained, so conventional UT suffers beam skewing, high scattering noise and a poor signal-to-noise ratio. Without DMW-specific probe selection (low frequency, TRL/dual-matrix, low-frequency angled longitudinal wave), a DMW-specific procedure and a representative calibration block, the result is not reliable. PAUT and TOFD face the same constraints. No volumetric method reveals microstructural degradation such as carbon migration; for that you need destructive sampling or in-situ metallographic replication.

9. Why nickel-based filler metal in high-temperature DMW

Once service temperature enters the creep range, the choice for DMW moves almost always to nickel-based filler metal. In DMW the limit is always set by the weaker side: creep becomes governing above roughly 370–425 °C on the ferritic/low-alloy side, whereas for austenitic stainless steels that limit is markedly higher (~510–550 °C). The creep behavior of the joint is therefore governed by the ferritic side.

There are three separate and mutually reinforcing reasons to move to nickel-based filler. The first is carbon migration: carbon diffusivity and solubility are low in a nickel-based matrix, so the carbon activity difference and diffusion rate between the ferritic side and the weld metal fall, and the decarburized/carburized band develops far more slowly. The second is thermal expansion compatibility: the expansion coefficient of nickel-based filler metal (~13.5 averaged over 20–100 °C, ~15 × 10⁻⁶ /°C averaged over 20–500 °C) lies between the ferritic and austenitic metals. As a result, the peak thermal strain moves away from the ferritic fusion line weakened by carbon migration and shifts to the more ductile, more creep-resistant austenitic–weld metal interface. The third is microstructural stability, and it has a name: the δ-ferrite in a high-ferrite austenitic weld metal such as 309L transforms into sigma phase under long exposure in the roughly 550–900 °C band, causing loss of toughness and embrittlement. This is a different mechanism from 475 °C embrittlement, which attacks the ferritic phase around that temperature. Nickel-based weld metal is markedly resistant to sigma formation, and the established industry rule of thumb reflects it: above continuous service temperatures of roughly 425 °C (800 °F), move from 300 series filler metal to nickel-based filler.

That said, nickel-based filler metal carries its own penalties, and an honest picture has to name them. ERNiCr-3/ENiCrFe-3 weld metals are susceptible to ductility-dip cracking (DDC) and to solidification cracking. DDC is grain-boundary driven and specific to fully austenitic weld metals; it demands tight heat input and weave discipline. The puddle is sluggish and viscous, which raises the tendency for lack of fusion (LOF) and calls for extra care in joint preparation and pass technique. Alloy 182 is known for grain boundary carbide precipitation under long exposure and for PWSCC in pressurized water environments. Nickel-based filler metal also does not prevent Type II boundary formation; its advantage lies on the carbon migration and thermal expansion side. Its thermal conductivity is low, so heat input management and interpass temperature control become more critical, and its cost is markedly higher. Taken together, nickel-based filler metal is the dominant engineering choice for permanent DMW transitions in boilers, superheaters and high-temperature piping — but the selection must account for the service environment and for the weaknesses of the consumable itself.

10. Preheat, PWHT and procedure control

Thermal management in DMW cannot be considered one-sidedly. The ferritic/low-alloy side is a hardenable metal and requires suitable preheat based on carbon equivalent, thickness and diffusible hydrogen content, to prevent hydrogen-induced cold cracking. The austenitic side, by contrast, tolerates high interpass temperature poorly. State the risk correctly, though: preheat temperatures on the order of 100–250 °C sit far below the sensitization band (roughly 425–815 °C), so preheat alone does not cause sensitization. Sensitization is mainly relevant in unstabilized, non-low-carbon austenitic grades (304, 316), and depends on the total time spent in the 425–815 °C band. In the L grades (304L/316L) and stabilized grades (321/347) that risk is largely eliminated; there the governing constraints are the interpass temperature limit (typically ~150 °C), sigma phase precipitation and heat-input-driven distortion.

PWHT is even more delicate. The ferritic side needs stress relief, but the PWHT temperature for ferritic material (typically 700–760 °C) falls inside the sensitization and sigma formation band for the austenitic side, and nickel-based weld metal can also form unwanted precipitates in that band. The buttering plus local PWHT approach exists precisely to resolve this conflict. Fix all these variables in the WPS, verify them by PQR, and define hardness and toughness acceptance criteria explicitly. Of all welding, DMW is the class that can least afford to leave anything to chance in the procedure.

From the field

At a refinery, the field welds joining a carbon steel line to a 316L header cracked along the fusion line and leaked within a few months of start-up. At first glance the welds looked clean, and the radiographs met the acceptance criteria. Then a hardness traverse found local values above 400 HV in the narrow, heavily diluted band on the weld metal side of the fusion line — the unmixed and partially mixed zones. That is the clear signature of a 308L-type filler metal diluted with carbon steel and transformed to martensite. We also recorded secondary hardness peaks in the carbon steel HAZ, but those came from a separate mechanism: quench hardening driven by heat input and cooling rate, which nobody should confuse with dilution. The acceptance limit is what gives the measurement meaning. For sour service joints, NACE MR0175/ISO 15156 typically requires ≤250 HV, and Cr-Mo fabrication codes commonly cap hardness around 241–248 HB. 400 HV sits far above both. The fix was to switch the filler metal to 309L, cut heat input in the root pass to limit dilution, and butter the ferritic groove with nickel-based filler metal at critical transitions. The lesson was plain: in DMW, an acceptable radiograph does not mean acceptable metallurgy. On these joints the most valuable examination tool is often not the film but a hardness traverse along the fusion line. An inspector who makes a habit of mapping hardness here catches most late failures early.

Related standards

ASME BPVC Section IX — welding procedure (WPS/PQR) qualification and welder performance qualification. The governing paragraph for DMW is QW-424: a joint between two different P-Numbers is assessed separately under the QW-424 rules, even when each P-Number has been qualified on its own. QW-283 applies to buttering/overlay, and ASME Section II Part C is required for consumable definitions. Filler metals are classified by F-Number; A-Numbers are defined only for ferrous weld metal chemistry, so nickel-based filler metals have no A-Number.
EN ISO 15614-1 — qualification by welding procedure test: arc and gas welding of steels and arc welding of nickel and nickel alloys (aluminium falls under EN ISO 15614-2). In DMW both base metal groups must be covered; base metal groups are assigned according to ISO/TR 15608.
EN ISO 15609-1 — preparation of the welding procedure specification (WPS).
EN ISO 9606-1 — welder qualification test (steels).
EN ISO 9606-4 — welder qualification test: nickel and nickel alloys. This is the part that applies to DMW welds made with nickel-based filler metal.
EN ISO 3834 — quality requirements for welding; selecting the appropriate quality level for critical joints such as DMW.
EN ISO 5817 — quality levels for imperfections in welds.
EN ISO 6520-1 — classification and definitions of imperfections in welds.
EN ISO 2553 / AWS A2.4 — welding symbols and their representation on drawings.
Filler metal standards: for stainless (austenitic) filler metals, AWS A5.4 (covered electrodes) and AWS A5.9 (wire/rod), with ISO equivalents EN ISO 3581 (covered electrodes) and EN ISO 14343 (wire/rod) — e.g., ER309L → ISO 14343 G/W 23 12 L. For nickel and nickel-based filler metals, AWS A5.14 (bare wire electrode/rod, e.g., ERNiCr-3) and AWS A5.11 (covered electrodes, e.g., ENiCrFe-3), with ISO equivalents EN ISO 18274 (wire/rod, e.g., ERNiCr-3 → ISO 18274 S Ni 6082) and EN ISO 14172 (covered electrodes). For flux-cored wires, AWS A5.22 (stainless flux-cored) and AWS A5.34 (nickel-based flux-cored); FCAW is common in DMW, so do not skip these two. A5.4/A5.9 do not cover nickel-based filler metals, and that distinction matters in DMW filler metal selection.
Plant-specific construction and inspection codes (boiler, pipeline) complete the picture for DMW selection.

Türkiye-specific note

In Türkiye (Turkey), the EN ISO standards listed above are published with a TS EN ISO prefix — use that prefix when searching national catalogs. The technical content is identical; only the designation differs.

Take it to the field: To keep the welding standards, procedure logic and field control steps in this article in your pocket — completely offline and free — take a look at the Doawise Welding Eng Guide app.

At DoaWise we carry out welding engineering and examination services to international standards. On dissimilar metal joints we tie every decision to procedure — from filler metal selection to dilution control, from carbon migration risk to PWHT strategy — and produce recordable, auditable results. Our aim is that every DMW point in the field is not just a weld that passes today, but an engineering solution that stays reliable for the whole service life.