Sindo Kou - Welding Metallurgy

Figure 3.17Effect of sulfur on YAG laser welds: (a) 304 stainless steel with 40 ppm sulfur; (b) 304 stainless steel with 140 ppm sulfur.

Source : Limmaneevichitr and Kou [34].

Sahoo et al. [35] and McNallan et al. [36] calculated the surface tension of liquid metals based on thermodynamics. Figure 3.18shows the surface tension of liquid iron as a function of temperature and the sulfur content [37]. For pure Fe, ∂γ/∂T is negative at all temperatures. For sulfur‐containing Fe, however, ∂γ/∂T can be positive at lower temperatures, which is consistent with the surface tension measurements by Sundell et al. [38]. Based on the surface tension data in Figure 3.18, Pitscheneder et al. [37] calculated Marangoni convection in stationary steel weld pools. Zacharia et al. [39] showed that computer simulations based on a positive ∂γ/∂T for liquid steel at all temperatures can over predict the pool depth.

Figure 3.18Liquid iron with various levels of sulfur: (a) surface tension; (b) temperature coefficient of surface tension.

Source : From Pitscheneder et al. [37]. Welding Journal, March 1996, © American Welding Society.

Kou and Sun [23] conducted bead‐on‐plate welding of Wood's metal as a physical simulation of the effect of the Lorentz force on weld penetration. Since the Wood's metal has a very low melting of about 70 °C, it is possible to use a heated Cu rod in contact with its top surface as a heat source for welding. Figure 3.19shows the transverse cross‐sections of the welds. The transverse cross‐section of the weld is essentially a semicircle. However, with a 75 A current going through to the Cu rod into the weld pool (no arcing and negligible resistance heating), the weld becomes much deeper. This physical simulation proves the Lorentz force deepens the weld penetration.

Figure 3.19Welds in Wood's metal produced under the influence of: (a) buoyancy force: (b) Lorentz force.

Source : Kou and Sun [23]. © TMS.

The direction of Marangoni flow on the weld pool surface can be identified by using tracers that float on the surface. Figure 3.20shows a bead‐on‐plate weld of a 304 stainless steel containing a low sulfur level of 40 ppm. To avoid keyholing, the YAG laser beam was defocused to 6 mm in diameter (conduction mode). Two dark patches of slag that formed by themselves on the weld pool surface are visible on the weld top surface. The smaller one on the left was pushed to the trailing edge of the weld pool surface during welding. The larger one on the right, on the other hand, was pushed to the leading edge. Thus, in both cases, Marangoni flow was outward along the weld pool surface toward the pool edge, consistent with the expected outward Marangoni flow in the absence of a significant amount of surface‐active agent.

Figure 3.20Bead‐on‐plate weld of 304 stainless steel with 40 ppm sulfur made by conduction‐mode laser beam welding at 3 kW and 7.62 mm/s. Dark slag patches at the leading and trailing portions of the pool boundary indicate outward surface flow during welding.

Source : Chao and Kou [34].

Heiple's theory represents a major milestone in welding science and should be verified, especially the reversal of Marangoni flow by a surface‐active agent. The computer simulation of Sun and Kou [23] was the first theoretical verification of the reversed Marangoni flow and deepened weld penetration caused by a surface‐active agent. Although this was significant, the most direct verification is still flow visualization, that is, to actually see the flow going inward along the pool surface and downward along the pool axis when the surface‐active agent is present. Since liquid metal is opaque to visible light, visualization of fluid flow below the weld pool surface is difficult to do.

Limmaneevichitr and Kou [40] conducted physical simulation of weld‐pool Marangoni flow using a transparent pool of molten NaNO 3(307 °C melting point). The pool was hemispherical in shape with a concave free surface that was heated at the center by a defocused CO 2laser beam as illustrated in Figure 3.21. NaNO 3has a ∂γ/∂T = − 0.056 dyne/cm/°C. Since its transmission range is from 0.35 to 3 μm, NaNO 3is opaque to CO 2laser (10.6 μm wavelength) just like a metal weld pool is opaque to an arc. By using a thin sheet of He‐Ne laser (red light), either vertical [40] or horizontal [32], to illuminate tiny tracer particles suspended in the pool, the flow pattern in the pool can be revealed clearly.

Figure 3.21Visualization of Marangoni flow using laser light‐cut technique: (a) vertical light sheet [40]; (b) horizontal light sheet [32].

Source : Limmaneevitchitr, Kou, Wei. Welding Journal, May 2000 and December 2011, © American Welding Society

.

Figure 3.22a shows the flow pattern induced by a CO 2laser beam of 2.5 W power and 3.2 mm diameter [40]. The narrow band above the pool surface is from the mirror image of the flow lines near the pool surface, and the arrows in the band indicate the directions of flow at the pool surface. Two counterrotating cells are, in fact, the two intersections between the donut‐shaped flow pattern and the meridian plane of the pool. The outward surface flow is much faster than the inward return flow, which is typical of Marangoni convection. As the beam diameter is reduced, convection grows stronger and penetrates deeper.

Figure 3.22Verifying effect of surface‐active agent on Marangoni flow using NaNO 3: (a) outward flow; (b) inward flow in a NaNO 3pool containing 2 mol% C 2H 5COOK as a surface‐active agent.

Source : Limmaneevitchitr and Kou [40, 41]. Welding Journal, May 2000 and November 2000, © American Welding Society.

It is worth noting that in conduction‐mode (no keyholing) laser beam welding, the pool surface can be concave due to Marangoni convection and surface tension [32] and, in fact, this has been shown to be the case experimentally [42] and by computer simulation [43]. The concave NaNO 3pool surface in Figure 3.22, however, is just a coincidence – that is, the melt wets the container wall and the meniscus makes the pool surface concave.

Limmaneevichitr and Kou [41] added 2 mol% of C 2H 5COOK to the NaNO 3pool as a surface‐active agent and reversed the direction of Marangoni flow. C 2H 5COOK reduces the surface tension of NaNO 3significantly. Its effect on ∂γ/∂C is −22 dyne/cm/mol% [44]. Since the heating of the CO 2laser beam decomposes C 2H 5COOK and makes it ineffective, the surface tension is now higher near the center of the pool surface instead of lower as in the case of the pure NaNO 3pool. As shown in Figure 3.22b, the flow is now inward along the pool surface and downward along the pool axis, thus confirming Marangoni flow is indeed reversed and deeper penetrating.