Optical Evidence of Compositional Fractioning between Plasma‐Condensed and Melt Pool Matter

Control of multicomponent alloys during welding is challenging because it lacks a real‐time understanding of composition. The optical emissions of plasma formed during laser‐induced metal welding correlate with the composition of particles ejected from the melt pool. Plasma emissions observed in this study contain large iron, manganese, chrome, and copper signatures, which match the composition of emitted particles. Particles recovered closest to the melt pool exhibit a core–shell morphology that is composed of iron‐manganese‐chrome intermetallic cores within copper shells. Particles collected farther from the melt pool, do not share this core–shell morphology, though similar elemental compositions are observed. The correlation between plasma optical emissions and particle composition can be used to predict the composition of the melt pool, allowing for real‐time welding and sintering control.


Introduction
There is presently an opportunity to quantify composition through plasma emission signatures.3] The emission signatures employed are typically Raman-or X-Ray-based.Both techniques are suitable but are difficult to employ in industrial settings because control feedback is interrogated on the milliseconds time scale.
Optical emissions provide a unique opportunity [2] to gain real-time knowledge of the weld by correlation of the material removed from the melt pool to the solidified condition of the metal.What remains to be understood is the form and composition of the plasma and resultant particles; understanding those factors would allow a mass balance to be done around the melt pool.
With such a mass balance, it would be possible to predict the composition of the weld from the plasma emissions and predict the material properties of the solidified metal.
Spatter (ejecta) from a melt pool has been studied in reasonable detail and its source is understood as the rapid vaporization of metal, pushing melt material violently outward.This phenomenon has been found to be especially common during keyhole mode laser welding. [4]The trajectory and frequency of ejecta have been recorded, analyzed, and modeled. [5]The material from this process is compositionally indistinguishable from the melt pool itself. [6]Due to the ease of particle collection and observation, spatter is the most well-documented of the various mechanisms for material loss during laser welding or melting.As spatter is compositionally similar to the base metal, what remains to be understood is the atomically selective removal of material by the plasma plume.
The region in the gas plume where condensation occurs has been studied, as seen in the work by Shcheglov et al. [7] who calculated what the size of condensed particles should be as a function of height in the plume, starting at 2.5 mm above the surface.Vishnyakov et al. [8] go into further detail about the nucleation of condensate particles, dividing it into two regimes: nucleation and coalescence.Nucleation events occurring under supersaturation and ionization conditions quickly consume the supersaturated vapor, at which point the mechanism of particle growth changes to the coalescence of other particles.The chemical composition of these condensed particles is rarely reported, with focus instead being on the composition of the vapor, given that it represents a compositional change in the weld pool.
Condensate rings on the welding surface have not been studied extensively.One of very few published studies-perhaps the only one-noted weld condensate rings that were reportedly observed on each side of a laser weld track. [9]Ma et al. did not, however, investigate the particle composition on the welding surface, instead collecting particles from the plume a few millimeters above the surface. [9]16][17] Mathematical models for predicting the physical attributes and behaviors of these phenomena have been investigated. [18,19]30][31][32][33][34][35] This work presents an initial look at correlating the optical emissions of the weld plasma to that of the composition of particles ejected from the melt pool.Specifically, we find that particles removed from the melt pool have atomic percent level differences in the composition of alloy elements when compared to the originating metal.The most interesting result of this initial work is that while the emission height of the particles influences the physical form (core-shell/continuous) of the particle, the overall composition of the particles matches well the ratio of emitted plasma peaks.This correlation suggests the real-time optical control and feedback loops are possible for laser-driven metal welding.

Results
The goal of this study is to determine the feasibility of measuring weld composition through optical spectroscopy.Five iterations of each shot condition were used to build up a statistical picture of performance.For a 100 (10 by 10) shot matrix testing laser power levels from 100-300 W, five iterations led to a 5000-point data set for subsequent trend analysis.The following results summarize the optical and electron characterization of the plasma plume and collected particles.
A high-speed video still of the experimental flash is shown in Figure 1a, with the laser creating a melt pool, a plume of matter condensed from the plasma, and a plume of ejected particles from the melt pool.In this work, only particles condensed from the plasma plume are considered as it appears to be a gap in the literature.The light collection optics are shown in Figure 1b.This involves two methods for particle collection, one along the surface of the metal, and another vertical holder to catch particles in the plume directly before they fall back to the surface.Additionally, the spectrometer is aligned such that it observes the melt pool by what is marked the collection volume in Figure 1b.A photo of the setup is shown in Figure 1b with a high-speed still image of a melt pool and plume.Figure 1a shows the plasma is separated into the matter near the melt pool (seen as the dot in the bottom of the image) and the matter reignited by the laser beam (the mushroom cloud above the pool).
As the near-pool light was desired, collection optics were put in place to collect light only from the melt pool to 3 mm above the metal surface.An example spectrum of the near-surface plasma is shown in Figure 1c, which consists of the emissions of the surrounding argon gas as well as constituents of the steel substrate.The change concerning time of each of these signatures is seen in Figure 1d.
As can be seen in the time-change of the atomic emission signatures, most of the metal-based peaks follow the same trajectory of an initial rise of prevalence followed by a slow decay.This is true for all but the manganese primary state and secondary copper peaks.Both of these elements diminish more quickly during the middle portion of the shot, but show an increased prevalence when compared to the other metal constituents at late times during the shot.As both copper and manganese showed a relative increase during later times in the shot, the particulate rings were compositionally analyzed for these compounds.To systematically investigate the behavior of the laser melt process over a range of processing powers, the laser was set to fire a sequence of shots with increasing power from 100 to 300 W for 5 ms bursts.An inset array of optical images of the resulting welds is shown in Figure 2a.The size of the melt pool in these images was quantified and plotted in Figure 2a.During analysis, three distinct regions were observed.The first was a region of shock damage (also termed thermal damage, resulting from being proximate to the heated zone, but not directly illuminated), [15,16] followed by two distinct sets of condensate rings.As the particles in the condensate rings were smaller than the resolution of the optical microscope, they were collected for electron imaging and EDS, as shown in Figure 2b.
An example of the condensed particles is shown in Figure 2c.The halo observed around the bright core in the SEM image indicates the presence of two distinct densities.Subsequent EDS mapping confirmed this by showing an iron-nickel-chromemanganese core with a copper shell, as seen in Figure 2d.
Further transmission electron microscopy (TEM) analysis showed that the particles were composed of an intermetallic core and a crystalline copper shell as seen in Figure S1, Supporting Information.Additional analysis showed that the appearance of optical rings of particles was caused by two different condensation morphologies.Particles that condensed near the melt pool laid down in a perfect layer of particles, while those farther from the edge of the melt pool were linked up in small chains of particles as seen by comparison of the morphologies in Figure S2a and b, Supporting Information, and shown in Figure S3, Supporting Information.Since the material falling around the melt pool was observed to originate 1-3 mm above the melt pool, efforts went to collecting and characterizing material projected farther from the melt pool.
Considering the plumes of ejected matter appear reasonably circular around the melt pool, Figure 3a, particles were collected radially.This approach differs significantly from the more filterbased collection approaches seen in the literature.Figure 3b shows a "tower" that was 3D printed with adjustable posts to allow for the collection of matter originating from the melt pool while simultaneously preventing matter from falling back onto the posts.This design allowed for particle collection from heights of 0.1-3.5 cm above the melt pool and radii of 0.3-3.5 cm from the epicenter of the melt pool.Slots for TEM grids and chips of silicon for SEM were printed into the adjustable posts.The TEM grids and Si chips were subsequently analyzed for collected particles.Figure 3c contains an SEM image of a long, chain-like aggregation of particles.Lastly, EDS mapping was also performed on these particles, and is presented in Figure 3d.In comparison to the EDS maps from Figure 2d, these particles appear much more uniform in composition, indicating they do not share the same core-shell structure.
As seen in Figure 3c, the trend toward chain-like particle agglomeration continues with distance from the melt pool.Matter collected at centimeter scales of distance from the melt pool showed distinct chain-like patterns of agglomeration, which is distinct from the layer-like deposition near the melt pool's edge.These agglomerates may be extensions of the chains observed farther out from the melt pool, with regards to the plasma-condensed particles.Subsequent EDS mapping of particle composition showed a distinct difference in the matter collected at centimeter scales of distance from the melt pool when compared to that of the matter collected closer to the melt pool (at millimeter scales of distance), shown in Figure 3d.
Once compositional data were determined from the EDS maps, results were correlated to the optical spectroscopy.
The integrated EDS signals of the collected particles versus the base metal are presented in Table 1.The condensate rings around the melt pool are termed as "mm-scale" and the particles collected 1 cl above the melt pool are termed "cm-scale" Theses data show particles with enhanced proportions of manganese, copper, and chrome, while having a reduction in iron, nickel, molybdenum, and silicon content.It should be noted that the measurement error of EDS for the base metal is 1 percent and 5 percent for the particles.The particles show a composition, in agreement with measurements taken of the plume.

Discussion
As seen in Figure 1c, laser-welding-induced plasma emits light, which corresponds to the alloy elements of the base steel.The ratio of the plasma peaks from the alloy elements are at a similar intensity to that of iron, suggesting that the resultant condensed matter will have a composition heavy in alloy elements when compared to the base metal.In contrast, Figure 1d shows that matter condensed from plasma near the melt pool exhibits strong manganese and copper signals, which maintain their intensities for longer than the other atomic emissions.This suggests that copper and manganese compositions are greatly enhanced, along with the possibility of a core-shell-like particle morphology.
These plasma data are corroborated by the EDS analysis, shown summarized in Table 1.Matter collected at all heights above the melt pool had a higher ratio of copper, manganese, and chrome when compared to the base metal, correlating directly to the plasma data.These results suggest that the ratio of the metals would be similar between material collected at different heights.
The interesting morphology of the particles collected near the melt pool (i.e., an intermetallic core inside a copper shell) correlates loosely with the time decay of the plasma signals presented in Figure 1d.The persistence of the copper and manganese signals, however, suggests the prevalence of their atoms in the resultant particles, likely in a core-shell structure.It is a  hypothesis for future work that the core-shell morphology represent matter condensed from the plasma a single time while the fully integrated particles collected further out from the plume represent matter which may have been reionized by the laser.This is hypothesis will be explored in future work.
The agglomeration results of the particles show a compelling deposition continuum, where near-melt pool particles are deposited in monolayers, followed by short chains at distances of hundreds of microns, followed by long chains at the centimeter scales, as presented in Figure S2, Supporting Information and Figure 3.The change from monolayers to short chains also explains the two distinct condensate rings of particles surrounding the melt pools (Figure S3, Supporting Information).The agglomeration of particles forms a predictable pattern as the length of chains increases with distance from the melt pool.
Future work will revolve around the explanation of the core-shell particles near the melt pool, but continuous particles collected farther from the melt pool.Current hypotheses include the remelting of particulate matter by the laser light and the different origination pathways (i.e., matter ejected as a liquid and matter recondensed from a plasma).Both avenues require further testing to establish accurate prediction models of final elemental composition of welded materials.

Conclusions
There is a strong correlation between the plasma emission spectrum of a melt pool in laser welding and the composition of the resultant ejected particles.The particles removed from the melt pool are enriched in certain alloy elements, indicating a resultant decrease of alloy elements in the weld.Future work will involve mass balancing around the melt pool and removed matter.It is anticipated that correlating plasma emissions to the composition of the melt pool may allow the development of real-time weld qualification.By qualifying weld performance in real-time, it would be possible to make in-line improvements to the condition of welding and laser-based additive manufacturing.

Experimental Section
Sample Preparation: Steel samples of 316 alloy were procured from McMaster-Carr (Product: 9083K229, see www.mcmaster.com/9083K22/for composition details).Each sample set was compared to steel with the same processing parameters.The samples employed were either 4 mm thick bar stock (2.54 cm per side sections) or shim stock with a 500 μm thickness and the same lateral dimensions.
The samples were polished to a mirror finish (AE10 nm roughness Ra basis) in-house and potted in polystyrene epoxy for later ease of use, sample control, and standardization of surface heights.The "pucks" of potted steel were then issued unique identification numbers to maintain control of the sample and to link information to processing parameters.The samples were cleaned with solvent sonication and kept covered in a laminarflow hood to minimize the addition of foreign particles to the workpiece surface.
Sample Measurement Conditions: The laser used in this study was a IPG photonics ytterbium fiber laser Model YLR-300/3000-QCW-MM-AC-Y12, with a 1070 nm emission wavelength, a focal distance of 6 cm, a power range from 100 to 3000 W, and an emission time from 1 to 50 ms.The electronic gain to on/off the laser was measured by the vendor to be 26 þ 5 microseconds.The laser beam spot when focused on the workpiece had a radius of 45.23 microns.The beam profile was a Gaussian-like intensity distribution with the intensity fit parameter determined by the manufacturer to be 1.6 mm*mrad.The Gaussian center was 13% skewed off the center Gaussian axis.The laser was enclosed in an ambient pressure argon environment and kept below 20 ppm water and oxygen during use.The vendor-supplied software was employed to fire the laser and control the emission parameters.The emission pulse was set at 5 ms single pulses with 5 s between each shot to allow the metal to cool.The powers employed in this study varied from 100 to 300 W. The metal samples were affixed to polymer stands ensure that any laser light breaking through the metal would not damage the X-Y stage below.
Characterization Tools and Parameters: The optical spectroscopy of the plasma emissions was taken on an Ocean Insights HDX spectrometer fitted with both Ocean Insights and Thorlabs components.The plasma emission light was collected with a 2 inch fused silica lens and then collected by a 0.5 inch fused silica collimating lens with an subminiature connection version A (SMA) connection.The spectrometer is specified to have a resolution of 1.0 nm in the configuration for this work.The total resolution experimentally determined for the light collection setup was 1.35 nm full-width-at-half-max.
The optimal angle for light collection was experimentally determined to be 17 degrees.The spectrum was evaluated from 55 to 0 degrees, with intensity being the optimized parameter no spectral change was observed.The light was fed into 6 foot-long 22 NA solarized fiber with a 600 micron diameter.The light was passed through a hot mirror to inhibit any stray laser light before arriving at the gas-blocked feedthrough out of the laser enclosure.The light was then fed into the HDX spectrometer and recorded (5-50 ms integration times, 3-pixel box car average, no backgrounds removed to minimize writing time) with vendor-supplied software (OceanView v.2.0.8).Spectral data was compared to the National Institute for Standards and Technology's (NIST) data for optical emissions by atom.The spectral data was first deconvolved and then the peak centers and widths from NIST were used to identify the corresponding atom.
High-speed video of the laser shot was collected with a Photron Inc.'s FASTCAM Mini AX100 540 K-C-32 GB camera with a vendor-supplied lens, Nikon -24-85 mm Micro/Macro Zoom Lens F/2.8.The shots were recorded from outside the protective laser glass of the welding enclosure at about 2 feet from the laser head and an angle of about 17 degrees from the horizontal.The data was captured at 4000 frames-per-second and was analyzed with the vendor-supplied program PFV4 (v.4.0.5.0).
Scanning electron microscopy (SEM) was performed on a JEOL IT-800 equipped with an Oxford X-Max energy dispersive spectrometer (EDS).Variable accelerating voltages ranging from 5 to 30 kV were used.Scanning transmission electron microscopy (STEM) was done on a JEOL ARM200CF aberration (Cs) corrected microscope operated at 200 kV using a JEOL Centurio EDS detector with a solid collection angle of 0.9 sR.

Figure 1 .
Figure 1.a) A high-speed video still of the melt pool and matter plume.b) A photograph of the laser head, workpiece, and flash collection optics with a scaled inset of plasma plume as captured by high-speed video.The red-marked area shows the approximate collection volume.c) the emission spectra of the emitted plasma near the melt pool.d) The signal intensity of each plasma emission (per element) versus time into the laser pulse.

Figure 2 .
Figure 2. a) Diameter of the melt pool, shock damage, and condensate rings versus laser power with an inset montage of the optical images of the melt pools and condensate ring(s) of nanoparticles for 5 ms shots from 100 to 300 W. b) An image of particle collection from condensed particles near the melt pool.c) An SEM micrograph of an example condensate particle.d) The 2D EDS map of the particles showing a core-shell structure with a high composition of copper.

Figure 3 .
Figure 3. a) Photograph of a cloud of ejected matter from a melt pool.b) Photograph of the towers fabricated to collect matter from 0.1 to 3.5 cm above the melt pool and from 0.3 to 3.5 cm away from the melt pool.c) SEM image of the particles collected by the towers.d) EDS mapping of the collected particles.

Table 1 .
The integrated EDS signal of particle compositions collected at mm and cm scales of height from the melt pool and the percent change from the original 316 steel substrate.Signal normalized to total metal composition.