Variation of Surface Tension of Liquid Metal with Fluorides in Tungsten Inert Gas Welding

Abstract

Real-time measuring and obtaining surface tension of liquid metal during the arc welding process is critical for studying the behavior of the weld pool, such as Marangoni flow and flow velocity. The dynamic variation of surface tension of weld pool during tungsten inert gas welding process with and without fluoride flux was measured by weld pool oscillation sensing system, and the effect of fluoride flux on surface tension and pool behavior was analyzed. The results show that the surface tension gradient of liquid metal can convert from a negative to a positive value in a critical arc time. The flow velocity of liquid metal and critical arc time significantly increased and decreased, respectively, with fluoride flux. The variation of surface tension, flow velocity, and critical arc time with fluoride flux was induced by arc temperature increasing.

1. Introduction

Surface tension is an important thermophysical parameter of liquid metal, which plays an important role in controlling grain growth, phase transition, and interface/surface motion [1,2]. Temperature field and chemical composition can affect the surface tension. Positive surface tension gradient causes the metal on the molten pool surface to flow to the center, resulting in deep and narrow weld, while negative surface tension gradient causes the metal on the molten pool surface to flow outward, resulting in shallow and wide weld [3]. Reliable measurement of the surface tension of liquid metal is of great importance in improving thermal processing of metals, such as casting and welding [3].

Currently, there are several measurement methods available to measure the surface tension of liquid metal, including the maximum bubble pressure method, the sessile drop method, the pendant drop method, and the oscillating drop method [4,5]. Cao et al. [6] and Jian et al. [7] developed an oscillating drop method and a maximum bubble pressure method to measure the surface tension of liquid metal for practical applications in casting and metallurgy, respectively. However, the effect of the welding arc has limited the use of these measurement methods in the study of the surface tension of the weld pool during the welding process. There exists strong interaction between the welding arc and the weld pool during welding, and there is a non-uniform distribution of temperature in the weld pool [8]. Under such a condition, the liquid metal in the weld pool is in a non-equilibrium state, which makes it difficult to use conventional methods for real-time measurement. Hence, there is a great need to develop a reliable method to in-situ measure the surface tension of the liquid metal in a weld pool in order to understand the fundamental mechanisms involved in the welding and joining.

Tungsten inert gas welding is an arc welding method that has been widely used in the joining of metals [9]. The principle of the TIG welding is that the heat from the electrical arc between the non-consumable tungsten electrode and workpiece causes the melt of metals to form a pool of liquid metal, i.e., the weld pool. The TIG welding technique has many advantages over conventional welding methods, including high quality and low cost, while the small penetration depth has limited the applications of the technique. To increase the penetration depth during the TIG welding, an activating flux of oxides or fluorides has been used to cover the surface of the workpiece prior to welding. Such a method has been referred to as activating flux TIG or A-TIG welding [10]. Tseng and Hsu et al. [11] have reported that using A-TIG welding can produce 2–4 times the penetration depth of TIG welding. The underlying mechanisms for the increase in the penetration depth in the A-TIG welding remain elusive. It has been suggested that there are two potential mechanisms for the increase in the penetration depth: the arc constriction and the change of the surface tension gradient. However, there are contradictory reports about the effect of the arc constriction even with similar welding conditions. Lowke et al. [12] believed that this controversy is due to the welding arc easily affected by welding parameters, such as arc length, welding speed, and the coating density of activating flux. Heiple et al. [13] believed that the change of the surface tension gradient is the main mechanism. Traidia et al. [14] and Wang et al. [15] using the numerical simulation method confirmed that active elements, such as oxygen and sulfur, can alter the pattern of the Maragnoni flow and heat transfer and increase the penetration depth. However, the temperature coefficient of surface tension used in numerical simulation is obtained in thermal equilibrium, which is in contrast to the non-equilibrium state of the weld pool. Also, strong interaction between activating flux and arc plasma during the welding process has been neglected in the numerical simulation. Hence, understanding the effect of flux on the surface tension during the TIG welding process has become a key issue to further study the mechanism of penetration increase and optimize flux formulations in A-TIG welding.

This study is aimed at the effect of fluorides on the surface tension of liquid metal during the TIG welding process. A weld pool oscillation measurement system was applied to measure the dynamic surface tension of liquid metal with and without fluoride flux during the welding process. Then, the variation of surface tension, pool behavior, and arc behavior were analyzed, and the mechanism of penetration increased with fluorides was discussed.

2. Sensing System

A simple laser-vision sensing system, as shown in Figure 1, has been developed by Shi et al. [16,17] to measure the oscillation frequency of a weld pool. Briefly, the laser beam from a laser of 500 mW in power strikes onto the surface of a weld pool and reflects off the surface. The reflected beams are recorded by a high-speed camera. To limit the effect of the arc plasma on the image quality, a filter matched with the wavelength of the laser is used. The sampling rate is 1000 frames/ms since the typical oscillation frequency of a weld pool is less than 250 Hz. For detailed information, see the works by Shi et al. [16,17].

3. Experimental

The pulsed tungsten inert gas (P-TIG) welding was used to induce surface oscillation of the weld pool. A tungsten electrode of negative polarity was used for the welding torch. The base metal was a stainless steel SUS 304 of 10 mm in thickness. The activating fluxes were MgF₂, CaF₂, BaF₂, and NaF. The stationary P-TIG welding was performed. The arc time was in a range of 5 to 30 s to vary the heat input to the base metals.

Table 1. Welding parameters.

Parameter	Value	Unit
Material (stainless steel 304 L)	10	mm
Density (stainless steel 304 L)	7.93	g/cm2
Arc length	3	mm
Average current	136	A
Peak current (Ip)	220	A
Base current (Ib)	80	A
Duty rate (σ)	40%	-
Pulse frequency (f)	3.5	Hz
Welding time	5–30	s
Welding speed	0	mm/s
Shielding gas (Ar)	7	CFH

lists the welding parameters used in this work.

After the welding, the weld beads were sectioned with the surface normal parallel to the welding direction. An optical microscope was used to measure the dimensions of the weld beads.

4. Results and Discussion

4.1. Size of Weld Beads

Figure 2 shows typical cross sections of the weld beads produced by the TIG welding with and without fluorides at two arc times of 10 and 15 s. Using the activating flux of fluorides during the TIG welding increases both the penetration depth and the bead width.

From the optical images shown in Figure 2, we can measure the penetration depth and the bead width. Figure 3 shows the temporal evolution of the geometrical dimensions of the weld beads. Both the penetration depth and the bead width increase with arc time, independent of the flux used. The MgF₂ flux causes the largest increase in the penetration depth, and the NaF flux causes the smallest increase in the penetration depth. It is interesting to note that there are two stages for the temporal evolution of penetration depth, and there is a linear relationship between the weld width and the arc time. For each stage, the penetration depth linearly increases with the increase in the arc time, and the increase rate is calculated and given in Figure 3. Generally, fluorides do not cause significant change in the increase rate for the same stage. There is a change in the increase rates from one stage to the other stage. The transition times for the change of the penetration depth from one stage to the other stage are 22.5, 22.5, 20, 17.5, and 13.5 s for the without flux and with NaF, BaF₂, CaF₂, and MgF₂, respectively. The smaller the transition time, the larger the effect of the fluoride on the penetration depth and the bead width.

4.2. Oscillation Frequency of Weld Pool

Figure 4 shows the optical images of the laser-line pattern at different times during the TIG welding. It is evident that the laser-line pattern appears periodically. The periodical appearance of the laser-line patterns suggests that the surface of the weld pool experiences oscillation at a finite frequency. From the images of the laser-line pattern, the frequency was calculated via the brightness value image-processing algorithm, which is detailed in Shi et al. [16,17].

Figure 5a shows the time domain oscillation signal of the weld pool. Converting the brightness value in the time domain to that in the frequency domain via the fast Fourier analysis, we can obtain the power spectral density (Figure 5b). The oscillation frequency of liquid metal is related to the surface tension. The surface tension could be calculated by the power spectral density oscillation signal. Figure 6 shows the variation of the oscillation frequency with arc time with and without activating flux.

4.3. Surface Tension of Liquid Metal in Weld Pool

Generally, the dispersion relationship between the angular frequency, ω, and the wavelength, λ, of the surface oscillation can be expressed as [16,17]

$${\omega }^2=\left(gk+{\displaystyle \frac{\gamma }{\rho }}{k}^3\right)\tanh \left(\mathrm{k}\mathrm{h}\right) with k={\displaystyle \frac{2\pi }{\lambda }}$$

(1)

for a weld pool of constant temperature and the liquid metal of constant surface tension and density. Here, γ is surface tension, ρ is density of liquid metal, and h is the depth of the weld pool.

Due to the ratio of penetration depth to bead with is in the range of 0.2–0.54, the Equation (1) can be approximated as [16,17]

$${\omega }^2={\displaystyle \frac{\gamma }{\rho }}{k}^3 or f=5.84{\left({\displaystyle \frac{\gamma }{\rho }}\right)}^{\frac{1}{2}}{\left(W\right)}^{\frac{-3}{2}}$$

(2)

where W is the diameter/width of the weld pool. Note that the surface tension calculated from Equation (2) is an average surface tension, <γ>, of the liquid metal in a weld pool since the distribution of the surface tension of the liquid metal in the weld pool is non-uniform.

Figure 7 shows the results of our experiment; it shows variation of average surface tension with arc time with and without fluorides activating fluxes during the TIG welding. R-squared of Figure 7a–e are 0.96, 0.94, 0.98, 0.97, and 0.96, respectively. In general, the variation of average surface tension can be divided into two stages, dependent on the activating flux. In stage I, the average surface tension linearly decreases with the increase in the arc time, and in stage II, the average surface tension linearly increases with the increase in the arc time. There exists a transition time at which the behavior of the average surface tension changes from stage I to stage II.

According to Figure 7a, the minimum value of the average surface tension of the steel is 0.88 mN/m during the TIG welding, which is close to 0.92 mN/m measured by the sessile drop method at 1978 K. Such a small difference suggests that the method used in this work is applicable in the measurement of the surface tension of liquid metal in a weld pool. The transition times are 21.75, 21.75, 20, 17.5, and 12.5 s for the without flux and with the NaF, BaF₂, CaF₂, and MgF₂ flux, respectively, which are compatible with the transition times determined from Figure 3a. Such a trend suggests that the average surface tension plays an important role in determining the penetration depth. The activating fluxes of fluorides except NaF can decrease the transition time for the average surface tension of the steel from stage I to stage II. The stronger the activating effect, the smaller is the transition time.

Generally, the average temperature of the liquid metal in a weld pool is a function of the heat input, arc time, thermal conductivities of base metals and the liquid metal, and the flow of the liquid metal in the weld pool. Using numerical simulation, Traidia et al. [14] and Wang et al. [15] studied the temperature distribution of the weld pool in the P-TIG welding and found that the average temperature, T, of the liquid metal in a weld pool increases linearly with arc time, t. Thus, the change of the average temperature of liquid metal in a weld pool is proportional to the change of the arc time to a certain extent. This is to say that there is d<γ>/dT ∝ d<γ>/dt. According to the results shown in Figure 7, there exists a transition temperature of T_c at which the average surface tension is minimum. For T < T_c, d<γ>/dT < 0, and for T > T_c, d<γ>/dT > 0.

4.4. Flow Motion of Liquid Metal in Weld Pool

Surface tension plays a critical role in determining the flow direction of the liquid metal in the weld pool. Generally, two types of flow patterns occur in the weld pool surface, which depends on the temperature coefficient of surface tension. Outward flow: liquid metal near the surface of the weld pool flows towards the edge of the weld pool; or inward flow: liquid metal near the surface of the weld pool flows towards the center of the weld pool. The weld pool surface was observed by a high-speed camera (sampling rate = 200 fps), and flow motion behavior was measured by visualizing the displacement of trace particles (AlN). The trajectory of trace particles and average flow velocity were extracted via the tracer particle images algorithm developed by Wang et al. [18]. Due to the flow motion of the pool surface under pulsed current mode being extremely complex, direct current (136A) with the same average current as pulsed current mode was applied in this experiment.

Figure 8 and Figure 9 show snapshots of weld pool evolution and flow trajectory of tracer particles, average flow velocity without and with CaF₂ flux, respectively. As can be seen from the figure, two types of flow patterns occurred in different arc time stages. In stage Ⅰ, the surface flow is outward. As time progresses, the outward flow gradually converts to inward flow. The average flow velocity in stage Ⅰ is higher than stage Ⅱ, and CaF₂ flux can obviously increase the average flow velocity. Furthermore, the transition time at which flow pattern changed is compatible with transition times determined in Figure 7a,e. This phenomenon indicates that the temperature coefficient of surface tension must be changed in different stages. For stage Ⅰ(d<γ>/dT < 0), the outward flow gives rise to the heat from the arc is transferred to the edge of the weld pool, leading to the formation of weld beads of shallow depth and large width; and for stage Ⅱ(d<γ>/dT > 0), the inward flow gives rise to the heat from the arc is concentrated to the center of the weld pool, resulting in the formation of weld beads of large depth and narrow width.

Zhao et al. [19] have reported that a small number of active elements, such as sulfur, oxygen, and selenium, placed on the surface of a liquid metal can cause a change in the coefficient of temperature gradient of surface tension from a positive value to a negative value, resulting in the direction change of the Marangoni convection. Also, active elements of sulfur, oxygen, and selenium have been detected in the weld beads, suggesting the migration of the active elements into the weld pool during welding. However, fluorides or fluorine have not been found in weld beads; in other words, fluorides likely have no effect on the chemistry of liquid metal. Thus, fluorides or fluorine elements have no direct effect on the change of the coefficient of temperature gradient of surface tension and increase flow velocity in the weld pool.

4.5. Effect of Coating Density on Average Surface Tension of Liquid Metal

Figure 10 shows variations of the average surface tension and the penetration depth with coating density for the fluorides of NaF, BaF₂, CaF₂, and MgF₂ with an arc time of 10 s. For comparison, the average surface tension of 1.186 mN/m of the liquid metal without activating flux under the same welding conditions is also included in Figure 10. It is evident that fluorides cause a change in the average surface tension of the liquid metal. The average surface tension of the liquid metal with the use of NaF in the TIG welding is approximately the same as that without activating flux. The average surface tension of the liquid metal first decreases slightly, then increases to the average surface tension of 1.186 mN/m of the liquid metal without activating flux, and finally remains unchanged with the increase in the coating density. Due to such a small change in the average surface tension, the penetration depth remained approximately the same with the increase in the coating density. Using CaF₂ and BaF₂ in the TIG welding causes the decrease in the average surface tension and then approaches plateau with the increase in the coating density, while the penetration depth increases with the increase in the coating density and approaches plateau with further increase in the coating density. The average surface tension of the liquid metal with the use of MgF₂ in the TIG welding decreases first, then increases, and finally remains unchanged with the increase in the coating density. Similar to the TIG welding with the use of CaF₂ and BaF₂, the penetration depth increases with the increase in the coating density and approaches plateau with further increase in the coating density. The TIG welding with the use of MgF₂ has the largest increase in penetration depth. Such a trend reveals the important role of fluorides in controlling the average surface tension of the liquid metal in the weld pool and the penetration depth.

As pointed out above, there is no fluorine detectable in weld beads. This behavior suggests that the mechanisms for the fluorine-induced decrease in the average surface tension are likely different from the TIG welding with the use of oxygen and sulfur elements. Further study is needed.

4.6. Effect of Fluoride on Arc Behavior

It has been reported that the temperature of the welding arc and the weld pool are dependent on the arc size and the arc voltage and determine the geometrical dimensions of weld beads in the TIG welding. Both the arc shape and the arc voltage for the TIG welding were synchronously monitored with a high-speed camera (sampling rate = 1000 fps) and a digital data acquisition system (sampling rate = 1000 Hz). Figure 11a shows the effect of fluorides on the arc profile in the peak current period for the coating density of 2.0 mg/cm² at an arc time of 10 s. The welding arc consists of an inner core of the arc plasma and an outer flame. To limit the effect of the outer flame on the analysis of the arc plasma, the optical image captured by the high-speed camera was processed by the binarization algorithm, and the contour of the arc plasma was extracted. Figure 11b shows the profile of the arc plasma without and with the activating fluxes of fluorides. The fluorides cause the expansion of the arc plasma, and no arc column contraction was found during the welding process.

4.7. Effect of Fluoride on Arc Voltage

Figure 12 shows the temporal evolution of the arc voltage in the peak current without and with fluorides. The fluoride of MgF₂ causes the largest increase in the arc voltage of 1.49 V over that without activating flux. There are increases of 0.91, 0.82, and 0.41 V in the arc voltage for the activating fluxes of CaF₂, BaF₂, and NaF, respectively. According to Figure 11 and Figure 12, there is no correlation between the increase in the arc voltage and the expansion of the arc plasma. This trend suggests that the increase in the arc voltage is likely associated with the increase in the potential/voltage in the polarity region (cathode and/or anode).

The weld arc consists of a cathode, anode, and arc column, as shown schematically in Figure 13. The total arc voltage of V can be calculated as

$$V=V_a+V_b+LE$$

(3)

where V_a and V_b are the potential drops in the anode and cathode regions, respectively, and LE (L is the arc length, and E is the magnitude of electric field intensity in the arc column) is the potential drop in the arc column. V_a and V_b are virtually independent of the arc length, and the magnitude of the electric field intensity in the arc column can be calculated as

$$\mathrm{E}={\displaystyle \frac{dV}{dL}}$$

(4)

Figure 14 shows the variation of the arc voltage with the arc length without and with fluoride. A linear relationship exists between the arc voltage and the arc length.

Table 2. Magnitude of electrical field intensity and the increase in average arc voltage in the anode region for an arc length of 3 mm.

	Without Activating Flux	MgF2	CaF2	BaF2	NaF
Average arc voltage (V)	12.6	14.09	13.51	13.22	12.94
Potential drop in the polarity region (V)	10.44	12.23	11.65	11.42	11.32
Potential drop in the anode region (V)	-	1.79	1.21	0.98	0.88
E (V/mm)	0.72	0.62	0.62	0.6	0.54

lists the magnitude of the electric field intensity in the arc column and the potential drop in the anode region for the TIG welding without and with fluoride. It is evident that the magnitude of the electric field intensity in the arc column for the welding without fluoride is higher than that with fluorides, and the potential drop in the polarity regions without fluoride is less than that with fluorides. Comparing the arc profile shown in Figure 11 with the magnitude of the electric field intensity listed in Table 2, we note that there is a one-to-one relationship, i.e., the expansion of the arc column leads to a decrease in the magnitude of the electric field intensity. Such a result suggests that fluoride reduces the magnitude of the electric field intensity in the arc column and increases the potential drop in the anode region. Also, the increase in the average arc voltage with fluoride in the TIG welding can be attributed to the increase in the potential drop in the anode region.

In the arc column, the electric field is dependent on charge carriers (cations, anions, and free electrons). Increasing the number of charge carriers can cause a decrease in the local electric field. For traditional TIG welding, the sources of charge carriers consist of the base metal, tungsten electrode, and shielding gas. With a layer of fluoride on the surface of liquid metal, charge carriers can also be generated from the ionization of the fluoride vapor.

Table 3. Physical properties of fluorides.

Fluoride	First Ionization Energy (kJ/mol)	Melting Point (K)	Boiling Point (K)	Lattice Energy (kJ/mol)
MgF2	737.3	1534	2533	2978
CaF2	589.8	1675	2733	2651
BaF2	502.9	1641	2533	2373
NaF	495.8	1266	1968	930

lists the first ionization energy of fluorides. The first ionization energy of fluorides is smaller than that of Fe (762 kJ/mol) and Mn (717.3 kJ/mol) in the base metal and Ar (1520.6 kJ/mol) in the shielding gas. Both free electrons and cations can be easily generated from fluoride during the TIG welding, suggesting that the number of charge carriers in the arc column with fluoride is more than that without fluoride, which leads to a decrease in electric field and the expansion of the arc column.

Due to the fact that no arc constriction has been observed, the constriction of the anode spot is not the dominant mechanism for the increase in the potential drop in the anode region. The dominant mechanism is the dissociation of fluoride $M{F}_n\to M^{n+}+n{F}^{-}$ (M = Mg, Ca, Ba, and Na). The dissociation energy is correlated with the bonding energy. The higher the bonding energy, the higher the melting temperature/lattice energy. Higher arc energy density is needed to form ions from fluoride of higher lattice energy/melting temperature, which can cause an increase in the potential drop in the anode region and arc temperature. According to the lattice energies listed in Table 3, the welding arc for the TIG welding with MgF₂ has the largest temperature, followed by that with CaF₂, BaF₂, and NaF. Hence, we can conclude that the increase in penetration depth, flow velocity in the liquid metal, and decrease in critical time for which the surface tension temperature coefficient changed with fluoride is caused by an increase in arc temperature.

5. Conclusions

Using a weld pool oscillation sensing system, we measured the average surface tension of the liquid metal without and with fluorides in the TIG welding. The following briefly summarizes the results obtained.

(1) The surface tension gradient of liquid metal and Marangoni flow type in the weld pool converts to a transition temperature. When the molten pool temperature is higher than the transition temperature, the surface tension gradient changes from positive to negative, and the inward flow on the molten pool surface changes to outward flow. The value of surface tension under transition temperature is the lowest.

(2) Fluorides can significantly increase the flow velocity of the weld pool and reduce the critical arc time. Fluorine causes the weld pool surface to flow inward and the bead to be deeper and narrower.

(3) Increasing arc temperature with fluorides during the welding process is the main mechanism for the variation of surface tension, increasing flow velocity, and penetration depth.