Vapor–plasma plume investigation during high-power fiber laser welding

The technology of fiber lasers and their applications has shown recent rapid progress [1–10]. Since the new generation of high-power fiber [11], and later thin disc [12] lasers with high brightness radiation, was invented, a great interest in applying these lasers in deep penetration metal welding technology has been observed. Obvious advantages over the high-power CO₂-lasers traditionally used for these purposes are the absence of laser-induced welding plasma absorption of the short wavelength radiation and the possibility of essentially increasing the laser output power and penetration depth. As the coefficient of the inverse bremsstrahlung absorption of laser radiation by metal vapor is proportional to the second power of wavelength [13], the vapor in case of solid-state laser welding absorbs about 100 times less energy. Therefore, instead of hot plasma plume formation, as encountered in CO₂-laser welding [14], the vapor in this case forms only a weakly ionized, relatively cold plume with a temperature between 3000 and 6000 K [15, 16]. Such a plume itself should not influence the laser radiation and the welding process. However, as was shown in some previous investigations [17–20], due to such a low temperature of the medium, hot metal vapor, which flies out of the keyhole, can condense again and form a cloud of small metal clusters with theoretically predicted diameters of 10–100 nm [21, 22]. It is evident that a cloud consisting of such particles can scatter and absorb laser radiation and thus influence the welding quality.

In this work a comprehensive investigation of the welding plume during different processes of low-alloyed steel deep penetration welding with a 20 kW fiber laser was carried out. The aim of the investigation was to describe the state of the medium above the keyhole as well as to determine the interaction between the plume and the high-power laser radiation as well as the influence of this interaction on the weld quality. Hereby, special attention was paid to the effect of metal vapor condensation and scattering by small condensed particles over the metal surface. Some experiments were carried out based on the measurement of optical transparency of the welding plume medium. The average particle size, the spatial distribution of the particle concentration and the attenuation of the high-power fiber laser beam by the welding plume were estimated.

The result of high-speed video observation (see figure 1) shows that the welding plume has two different parts with different dynamics, temperature, brightness and geometrical form.

Zoom In Zoom Out Reset image size

The lower part, with about 5 mm height, has a bright glow and oscillates in the welding plane in both forward and backward directions with frequencies of up to several kilohertz. The upper part has considerably weaker emission and can reach up to 50 mm height. It moves slowly up from the metal surface, and the geometrical form of this part corresponds to the fiber laser beam caustic.

In order to determine the welding plume state, measurement of its emission spectra from both the lower and the upper parts was carried out. The measurements were made during welding in ambient air atmosphere as well as with a shielding gas supply (Ar, He). The characteristic form of the lower part emission spectra is shown in the figures 2(a)–(c). Besides continuous spectra emission, all spectra have a number of bright spectral lines, which correspond to the neutral Fe I electron transitions. Also, a sharp decrease of the light intensity was observed to occur with the shielding gas supply. In comparison with air welding, the average intensity is about 20 and 80 times lower in the case of Ar and He supply, respectively. At the same time, the number of registered spectral lines is also reduced.

Zoom In Zoom Out Reset image size

Figure 3 shows a dependency of the welding plume temperature (lower part) calculated with the Boltzmann plots method on the incident laser power. For the given temperatures the electron density values were calculated from the perfect gas law and the Saha equation based on the first-order ionization only. It can be seen that even with a maximal laser power of 20 kW the welding plasma temperature is lower than 4500 K and the electron density is not higher than 10¹⁵ cm⁻³. The maximal ionization degree of the metallic vapor was obtained as about 10⁻³. This is to say that the lower part of the welding plume corresponds to a low ionized gas with much lower temperature than in case of the CO₂-laser welding.

From the above given parameters of the welding plasma state the inverse bremsstrahlung coefficient can be calculated for the ytterbium fiber laser wavelength (1.07 μm) [13].

$$\begin{eqnarray} &&{k}_{\mathrm{ib}}\approx \frac{{\nu }_{\mathrm{c}}}{c}\frac{{n}_{e}{e}^{2}}{{m}_{e}\omega }\approx 3.3\times 1{0}^{-4 1}\frac{{n}_{e}^{2}}{{T}^{3/2}}\sim 0.6~{\text{m}}^{-1} \end{eqnarray} \tag{ 1 }$$

where ν_c is the collision frequency, c the speed of light, and ω the laser radiation frequency.

From the obtained absorption coefficient value it can be concluded that the welding plasma during the deep penetration metal welding with high power and high brightness solid-state lasers has practically no influence on the laser radiation and on the weld quality, even with a laser power of up to 20 kW.

Zoom In Zoom Out Reset image size

Figure 4 shows the dependences of the welding plume temperature on the plume height, where the temperature was measured in two different ways. In the lower part (h < 7 mm) it was calculated from the Boltzmann plots in the case of air- and Ar-shielded welding. As mentioned above, the Ar supply decreases the plume temperature near the metal surface.

Zoom In Zoom Out Reset image size

However, starting from the height h ≈ 5 mm the temperature stabilizes at some constant level, which is equal for both air and Ar cases. In spite of this high temperature level (T ≈ 2500 K), the temperature measurement in the higher plume area with this method was impossible, as the intensity of the line emission becomes lower than the continuum radiation level. Therefore, in order to measure the temperature of the upper plume part the spectral pyrometry method was used. The description of the measurement technique was reported in [23]. From figure 4 (h > 5 mm) it can be seen that in the upper plume part the temperature also does not fall down to zero, but remains at a quite high almost constant level at heights of up to 30–40 mm.

Based on the obtained data as well as on the results of the high-speed video observation, the conclusion was made that only the lower part of the plume can be referred to the traditionally called laser-induced plasma, i.e. to the jet of weakly ionized metal vapor. The upper part has another nature. Its emission is caused by the heated or evaporated small metal condensed particles, when they come under the influence of the laser radiation. That is why the geometrical form of the upper part corresponds to the laser beam caustic.

Zoom In Zoom Out Reset image size

This supposition was experimentally confirmed by video recording of the welding plume with an opposite light source of high brightness (see figure 5). From this result, the area which absorbs the rear illumination light can clearly be seen behind the laser beam (on the right side). This absorption is apparently caused by the condensed particle cloud, which was found out to definitely correlate with the appearance of the upper part plume emission.

Consequently, it was shown that the welding process can be influenced not by the lower part (laser-induced plasma), but by the upper part (condensed particles) of the plume. This part consists of the small condensed metal particles and absorbs the fiber laser radiation. A numerical study of this absorption effect as well as the investigation of the particle parameters are presented below.

In order to describe the extinction of laser radiation by the small condensed particles, an optical transmission measurement of the welding plume medium was made using probe 1.3 μm wavelength IR-radiation from a 1 mW power diode laser source. The probe radiation was collimated in the 0.5 mm diameter beam and transmitted parallel to the metal surface through the welding plume medium during the welding process (see figure 6).

The measured probe signal showed that just from the starting point of the high-power laser operation (t = 0 s) the medium transparency decreases in average down to 6% and oscillates during the welding process in a mean-square deviation band of about ±3%. The frequency spectrum of the probe signal oscillations (up to 2–3 kHz) corresponded to the vapor–gas ejection dynamics, which was observed during the high-speed video recording and welding plasma emission registration.

Zoom In Zoom Out Reset image size

For the numerical description of the probe signal attenuation a value of the relative extinction $Q=\frac{{I}_{0}-I}{{I}_{0}}$ was defined, where I₀ is the initial probe beam intensity.

As was already reported in [24], using the measured extinction signals and some mathematical approximations, the full spatial distribution of the extinction coefficient at wavelength of 1.3 μm was rebuilt (see figure 7).

Zoom In Zoom Out Reset image size

The measurement of the condensed particle parameters was based on the multi-wavelength method of particle sizing in dusted mediums and aerosols [25, 26]. The ratio between the extinction values at two different wavelengths of the probe radiation is independent of the particle concentration as well as of the optical path of the probe beam. In the approximation of mono-size particles, this ratio is a function of the particle average size and determined only by the relation between two corresponding extinction efficiencies [27].

The principle of the experiment was almost the same as was described before, however, in this case a laser driven broadband light source (cw-mode radiance about 10 mW (mm² nm sr)⁻¹ in the wavelength range 170–850 nm) and a spectrometer were used instead of the probe laser source and the photoreceiver, respectively. The probe light was similarly collimated with an optical system with a 1 mm diameter beam and collected behind the welding plume with an optical fiber into the spectrometer. For the probe light registration two different spectrometers with a spectral range in the near UV-(range I: 198–432 nm) and visible (range II: 381–588 nm) ranges were used.

Zoom In Zoom Out Reset image size

We supposed that all particles in the plume have some average diameter a₀, i.e. the size distribution function is N(a) = δ(a − a₀). In this case the logarithmic extinction function can be written as [28]:

$$\begin{eqnarray} &&\ln [1-E(\lambda )]=\frac{\pi {a}_{0}^{2}}{4}{Q}_{\mathrm{ext}}({a}_{0},\lambda )\int _{L}{C}_{n}\hspace{0.167em} \mathrm{d}l \end{eqnarray} \tag{ 2 }$$

where Q_ext(a₀,λ) is the extinction coefficient, which can be precisely calculated for every given value a₀ and λ from the Mie-theories [29] if we suppose that the particles consist of pure iron. As the integral in the right-hand side of (2) does not depend on the radiation wavelength λ, the extinction efficiency ratio is:

$$\begin{eqnarray} &&\frac{{Q}_{\mathrm{ext}}({a}_{0},{\lambda }_{1})}{{Q}_{\mathrm{ext}}({a}_{0},{\lambda }_{2})}=\frac{\ln [1-E({\lambda }_{1})]}{\ln [1-E({\lambda }_{2})]}. \end{eqnarray} \tag{ 3 }$$

In figure 8 the typical forms of the probe radiation spectra without welding plume (I₀) and during the welding process (I_t) are shown. The difference between the upper and the lower curves shows the extinction of the probe radiation, which in this case is much higher than in the case of IR-radiation. The spectra of the transmitted probe light also have some absorption lines, which were attributed to the Fe neutral atoms (metal vapor).

Figure 9 represents some values of the average particle size a₀, which were obtained from equation (3) by means of measurement of the right-hand side (for the wavelength couple λ₁ = 300 nm, λ₂ = 400 nm) and the respective graphical solution at different plume heights. All measurements were made when the probe beam was transmitted directly through the plume axis (r = 0). The errors represent the mean-square deviation band of the values during the welding process. In general, the average size values varied in a quite broad range from 60 to 110 nm depending on the measurement point as well as on the welding process parameters.

Zoom In Zoom Out Reset image size

Nevertheless, there was qualitative coincidence between the spectra of the logarithmic extinction function ln[1 − E(λ)] and the spectra of the extinction efficiencies for given values of a₀ ≈ 80 nm.

Zoom In Zoom Out Reset image size

Figure 10 shows the measured logarithmic extinction function in two spectral ranges and the extinction efficiency calculated for three different particle size. As the integral in (2) is unknown, all functions were normalized at a wavelength of 300 nm. So good coincidence was observed not only for the wavelength of calculation λ₁, λ₂, but also in the rest of the spectral range.

Using the obtained a₀ values, the concentration of the condensed particles in the plume can be received. As we supposed only mono-size particles, the spatial distribution functions, which were obtained before, should be adequate independently of the probe light spectral range. The only difference can be the amplitude of the Gaussian functions A(h), which acts as a constant factor. In this case it follows from (4) and (2):

$$\begin{eqnarray} &&{K}_{\mathrm{ext}}=\frac{1}{\sqrt{\pi }}\frac{A(h,\lambda )}{w(h)}{\mathrm{e}}^{-\frac{{r}^{2}}{w(h)^{2}}}=\frac{\pi {a}_{0}^{2}}{4}{Q}_{\mathrm{ext}}({a}_{0},\lambda ){C}_{n}(r). \end{eqnarray} \tag{ 4 }$$

Hence, the spatial distribution function of the condensed particle concentration can be obtained as:

$$\begin{eqnarray} &&{C}_{n}(r)=\frac{1}{\sqrt{\pi }}\frac{A(h,\lambda )}{w(h)}\frac{4}{\pi {a}_{0}^{2}}\frac{1}{{Q}_{\mathrm{ext}}({a}_{0},\lambda )}{\mathrm{e}}^{-\frac{{r}^{2}}{w(h)^{2}}}. \end{eqnarray} \tag{ 5 }$$

The result of the calculations made with (5) for an average particle size a₀ = 77 nm and a plume height h = 10 mm is shown in figure 11. As follows from the calculation, the particle concentration can reach up to 10¹⁰ cm⁻³.

In order to verify the obtained result, the particle cloud parameters (size, concentration) were used to recalculate the extinction coefficient value at a wavelength of 1.3 μm. This coefficient was already obtained before by means of direct measurements with the probe 1.3 μm wavelength laser beam (see figure 7). Accordingly, we can compare two values, of which the first one was measured without any concretization of the absorption nature and the second one was calculated from the condensed particle parameters. In this case we have: Q_ext(a₀ = 77 nm λ = 1.3 μm) ≈ 0.0889 and K_ext (λ = 1.3 μm) ≈ 4.9 m⁻¹. At the same time, the corresponding K_ext value from figure 7 (h = 10 mm, r = 0) turned out to be about 3.8 m⁻¹. Hence, good correspondence between two differently measured values was noted.

Zoom In Zoom Out Reset image size

On the one hand this fact confirms the correctness of the condensed particle parameter measurement. On the other hand it indicates the truth of the supposition that the measured extinction in the welding plume is mainly caused by absorption and scattering at the small condensed particles.

Using the obtained distribution K_ext(r,h) the attenuation of the high power fiber laser beam, which propagates through the whole plume from the welding head down to the metal surface (see figure 6), can be calculated:

$$\begin{eqnarray} &&{E}_{\mathrm{A}}=1-\exp \left [\int \nolimits _{5}^{\infty }{K}_{\mathrm{ ext}}(r=0,h)\hspace{0.167em} \mathrm{d}h\right ]\approx 0.0 7 5 \end{eqnarray} \tag{ 6 }$$

where the integration is made only in the upper part of the welding plume (see above). Obviously, the extinction coefficient in this case had to be recalculated for the wavelength of the fiber laser radiation (1.07 μm). Considering the received particle parameters, this could be made with either the exact Mie theory or with the Rayleigh approximation. As before, it was also supposed that the condensed particles consist of pure iron, i.e. the complex refractive index was taken as m₁ = 3.25 + 4.38i and m₂ = 3.41 + 5.19i for the wavelengths of 1.07 μm and 1.3 μm, respectively.

Based on this approach, the average extinction of the ytterbium fiber laser radiation was estimated to be about 12%. This means that the welding plume can considerably change the fiber laser beam parameters and the beam power at the metal surface can be essentially lower than the output laser power. However, it is well known that for the deep penetration welding process not the average beam power, but the temporal power stability during the process plays a major role. In [30, 31] it was shown that even small oscillations of the laser beam power can lead to resonance melt pool vibrations and consequently cause spattering. As follows from the high-speed video observations the dynamics of the probe beam extinction corresponds to the dynamics of the vapor–gas ejection from the keyhole. Therefore, the fiber laser beam extinction has probably the same dynamics in an up to several kilohertz frequency range [32].

It should also be noted that all calculations concerning the influence of the particles on laser radiation were made without consideration of the inverse radiation influence. Owing to the high incident power density the particles can be heated up and evaporated inside the laser beam caustic. Consequently, the extinction coefficient will be decreased. To consider this effect it is necessary to know the ratio between evaporation and condensation rates as well as the gas dynamics in the condensed particle cloud area. Therefore, more detailed investigations of the condensed particles (size and density distribution, chemical composition etc.) as well as of the condensation process on the whole is necessary.

The welding plume during the deep penetration fiber laser welding of low-alloyed steel plates was investigated. It was shown that the plume consists of two different parts with different natures of origin.

The lower part has about 5 mm height and is formed by the laser-induced weakly ionized jet of the metal vapor from the keyhole. Due to the low temperature (<4500 K), ionization degree (<10⁻³) and free electron density (<10¹⁵ cm⁻³) (the maximal values correspond to 20 kW laser power) the lower part has no influence on the fiber laser radiation and on the welding process. The inverse bremsstrahlung absorption coefficient is less than 0.6 m⁻¹. The lower part is strongly influenced by the ambient atmosphere type. Ar (and especially He) shielding gas cools down the lower part and breaks the light emission.

The upper part reaches up to 50 mm height and is formed by the emission of the small condensed particles, when they are evaporated by the fiber laser beam. The average temperature of the medium remains practically constant in the whole upper part (about 2500 K) and almost does not depend on the shielding gas type and laser power. The measured extinction coefficient (at 1.3 μm wavelength) in the upper part is about several m⁻¹, i.e. much higher than the inverse bremsstrahlung absorption coefficient in the lower part.

Using the multi-wavelength method and supposing a pure Fe-composition, the condensed particles were measured to have an average size of (80–100) nm and a number density on the plume axis of about 10¹⁰ cm⁻³. The recalculated value of the extinction coefficient (at 1.3 μm wavelength) for the particles with the measured parameters was found to be close to the measured one, i.e. absorption and scattering by the condensed particles are the main nature of the measured probe signal extinction.

With the obtained spatial distribution of the extinction coefficient (recalculated for 1.07 μm) the integral attenuation of the high-power fiber laser beam was calculated when it propagated through the whole plume height. The integral value was about 12%. Taking into account the temporal characteristics of the measured extinction it was concluded that the plume can implement the modulation of the laser beam power in a frequency range of up to several kilohertz. Therefore, the vapor condensation effect can be a reason for spattering and instability of the welding process and requires investigation in more detail (particle evaporation and oxidation processes).

This work was partially performed in cooperation with the Laser Centre of National Research Nuclear University MEPhI, Moscow, Russia. The contributions were made by Viktor Petrovskiy, Sergey Uspenskiy, Alexander Streltsov, Andrey Vilyanskiy and Andrey Kuznetsov. Considerable help was also provided by Marco Lammers.