Temperature-Dependent Reflectance of Copper with Different Surface Conditions Measured at 1064 nm

Abstract

In order to fully understand and to eventually overcome the instabilities in laser welding of copper using 1064 nm wavelength lasers, it is important to get a deeper understanding about the interaction of the incident beam with the copper surface during the process. One of the main parameters determining this behavior is the temperature-dependent absorptance. However, the existing datasets only show little consistency between the different sources, do often not include the important absorptance step at melt temperature, or they are only calculated. This article thus presents an experimental approach to measure the reflectance of electro tough-pitch copper (Cu-ETP) from room temperature until beyond the melting point with an integrating sphere setup. The setup was developed with particular attention to measurement accuracy up to high temperatures. The reflectance using a 1064 nm laser under an incident angle of 12° has been measured for four different sample groups, which are untreated rolled copper, mechanically polished copper, oxide reduced copper and resolidified copper. The measurement results reveal significant differences in the temperature-dependent reflectance between the sample types and, for some types, also between the samples of the same type. This study investigates the origins of the most significant features in the reflectance curves and their potential impact on processing copper using NIR lasers.

1 Introduction

Processing of copper by lasers is challenging, particularly when using near-infrared (NIR) lasers with their disadvantageous absorptance characteristics. Being the second best metallic conductor for electricity and heat after silver (Ref 1), copper is a preferred material for many electrical and thermal applications. But while the availability of green and blue lasers has significantly facilitated reproducible laser welding of copper, the same process remains challenging when using NIR lasers, which are widely used in the industry. The main difference with respect to the applied wavelength is the absorption. While the absorptance of copper for green and blue lasers decreases upon melting, the absorptance for near-infrared lasers increases abruptly (Ref 2), leading to a sudden increase of the absorbed laser power and thus a narrow process window if predetermined process parameters are applied. A small process window alone does not necessarily lead to instabilities, but while the absorptance of copper for green and blue lasers at room temperature is in the region of 40% or even more (Ref 2), the absorptance of copper for near-infrared lasers is usually only 5% or even less. Oxides or other surface impurities with their substantially higher absorptance thus have a significantly larger impact on the absorbed power of near-infrared lasers and lead to substantial shifts of the already narrow process window. The constant shifts of the narrow process window make it difficult to control the welding process. In order to achieve a more stable process, it is hence important to investigate which process strategies can increase the width or reduce the shifts of the process window. This can be done by simulation studies that focus on the exact relation between the temperature-dependent absorption and the process behavior.

However, reliable simulation studies that focus on the effects from temperature-dependent absorption of real samples also require reliable absorptance data over a sufficiently wide temperature range, which generally can be obtained from direct measurements or derived from reflectance values. These data are still difficult to obtain from literature over a wide temperature range for the combination of copper and a wavelength of 1064 nm. Values for the absorptance can either be measured directly, or they can be derived from reflectance values, since copper is not transmitting any light except for the case of ultra-thin coatings. A significant proportion of the literature on laser copper interactions focuses on reflectivity, which is, however, only of limited use for the simulation of welding processes. Different from reflectance values, reflectivity values only represent the reflected fraction of light from a single interface. This might not be problematic as long as perfectly pure and smooth copper surfaces are regarded, since they indeed only reflect once from a single interface. But copper materials for industrial processing applications, however, are usually not perfect and involve a combination of oxide layers, other surface impurities, surface roughness and other defects. The light is thus no longer reflected only once at a single interface and the amount of reflected light is no longer equal to the reflectivity. Instead, reflectance or absorptance values are required then.

In the following, a brief overview of publications is given that present reflectance or absorptance data for copper at room temperature in the wavelength range from 1000 nm to 1100 nm, including works reporting on reflectivity or absorptivity. Duley (Ref 3), for example, calculated the reflectance for copper at room temperature over the range from 300 nm to 1100 nm from measured literature values of n and k.

Various authors reported on calorimetric measurements at room temperature, in which they used different approaches to determine the absorptance of copper. Quimby, Bass et al. (Ref 4) measured the absorptance of diamond turned or mechanically polished samples of OFHC copper at 1.08 µm wavelength at vacuum conditions. Engler, Ramsayer et al. (Ref 5) irradiated an untreated and thus slightly oxidized SE-Cu58 sheet with a 1064 nm processing laser, which was defocused to avoid surface melting. Moalem, Witzendorff et al. (Ref 6) conducted measurements according to ISO 11551:2003 with 1064 nm radiation on freshly polished copper samples. Zhu, Lu et al. (Ref 7) determined the absorptance of bare, unpolished copper wires at a wavelength of 1064 nm by optical diffraction measurements, in which they measured a temperature increase of up to more than 100 K in response to millisecond laser pulses of 5 J and 9 J.

Other authors, instead, used reflective measurements to determine the reflectance or absorptance of copper at room temperature. For example, Hass (Ref 8) presented measurement results for the reflectance at room temperature for a freshly evaporated coating of copper for wavelengths from 0.22 µm to 10 µm. Lehman, Livigny et al. (Ref 9) measured the reflectance of a copper mirror under 45° by shifting it into or out of a 1064 nm laser beam of either p- or s-polarization and measuring the power of the reflected and the direct laser beam with two identical detectors. Lee, Watkins et al. (Ref 10) used a reflectance spectrometer to measure the reflectance of laser cleaned copper samples at 500 nm to 1100 nm. Helm, Schulz et al. (Ref 11) used a spectrometer to measure the reflectance of untreated and laser structured copper samples from 400 nm to 1100 nm, and Hummel, Schöler et al. (Ref 12) used an identical spectrometer to determine the absorptance spectrum for Cu-ETP from 300 nm to 1200 nm. Bergström et al. (Ref 13) used an integrating sphere of 150 mm diameter in a double beam setup to determine the absorptance of commercially pure copper (99.8% purity) with a surface roughness Sa of 0.16 µm for a wavelength of 1064 nm and at an incident angle of 8°. In their study, which also included other nonferrous alloys, they found that the measured values for engineering grade surfaces have considerably higher absorptances than already published values for pure, polished surfaces.

Moreover, some authors investigated the temperature-dependent interaction of light with copper by employing diverse approaches such as Drude theory, electron-phonon collision, and calorimetry. Ujihara (Ref 14) calculated the reflectivity of solid copper over temperature up to the melting point for the wavelengths 0.69 µm, 1.06 µm and 10.6 µm, based on the Drude theory and the theory of electron-phonon collision. Siegel (Ref 15) extended Ujihara’s calculations for the liquid metal by using the Drude theory and the Percus-Yevick collective coordinate theory of simple liquids. He presented discrete reflectivity values for the solid copper at melting temperature and for the liquid copper at 1.0 times, 1.5 times and 2.0 times the melting temperature. Walter (Ref 16) used the electron-phonon collision frequency, the plasma frequency and the temperature-dependent resistance to calculate the reflectance of liquid copper at its melting temperature. He further demonstrated that the results of Ujihara (Ref 14) deviate significantly from experimental values. Xie, Kar et al. (Ref 17) used the Hagen-Ruben relationship and the temperature-dependent d.c. resistivity to calculate the absorptivity for copper at 1064 nm from 300 K to 1900 K. Quimby, Bass et al. (Ref 4) also presented their results for a calorimetric measurement on a diamond turned copper sample, that they conducted in a furnace under high vacuum conditions at room temperature and temperatures of 473 K, 673 K, 873 K, and 1073 K. After the temperature cycle, the measured absorptance of the sample at room temperature was increased with respect to before heating. Kohl, Kaufmann et al. (Ref 2) calculated the absorptivity over wavelength for solid copper at 293 K and for liquid copper at 1873 K from literature values for the optical properties at the respective temperatures. Blom, Dunias et al. (Ref 18) presented the temperature-dependent absorptance of copper for 1064 nm from about 250 K to 2000 K.

The aforementioned studies reveal substantial differences with respect to the absorptance, and most of the temperature-dependent results do not originate from measurements. Already for polished copper samples at room temperature, the absorbed or reflected fraction of light in the given literature differs by more than 1 percentage points, sometimes even within a single study. A difference of the absorptance from 1% to 2%, as reported within a single study, would double the absorbed laser power. The results for the untreated copper samples with their inherent imperfections such as slight oxidation layers even differ by up to several percentage points between the studies. Here, the absorbed laser power would even triple, if the highest reported absorptance of more than 6% is compared to the lowest one of 2%. Only two of the temperature-related studies in the literature survey present absorptance data that rely on measurements. One of the two relies on calorimetry, while the absorptance in the other study is calculated from measured temperature-dependent optical properties from the literature. Both studies do not provide any information on the height of the absorptance step due to the phase change at melting temperature, which is known to be crucial for the stability of a laser welding process.

The lack of reliable measurements over a temperature range sufficient for laser melting simulations requires the conduction of new measurements, where calorimetry is challenging when it comes to the combination of precision and temperature-dependence. Reliable information for copper on the absorbed or reflected fraction of laser light at 1064 nm, that covers the temperature range from room temperature up to the liquid copper at melting temperature, is evidently missing. Since the overall absorptance of copper is so low, reflectance measurements would need to be precise in order to avoid extreme relative errors for the resulting absorptance. On the other hand, calorimetric measurements face the problem that they require an irradiation induced temperature increase, which in turn leads to a manipulation of the absorptance. Therefore, a precise calorimetric measurement requires that the irradiated power is high enough to provide a sufficient measurement resolution but at the same time stays low enough to not excessively influence the measured absorptance. At elevated temperatures, especially when the sample temperature approaches the melting point, this might become a problem that even leads to unintended melting of the sample. Additionally, according to ISO 11551:2019, the interval to measure the temperature drift before the irradiation shall be at least 30 seconds and the cooling interval after the irradiation shall be at least 200 s. This means that the environment of the sample would need to stay at each temperature of interest for about 4 min to measure the respective absorptance value, resulting in long measurement times at elevated temperatures and thus allowing even slow processes to have significant effects on the surface characteristics.

In this study, an integrating sphere has been used for temperature-dependent reflectance measurements of copper, which require a higher relative precision than calorimetric methods but at the same time involve fewer challenges to achieve a high precision. According to literature, the expected values for the reflectance of copper are 95% or even higher. Hence, even small relative deviations in the measured reflectance can lead to a strong relative uncertainty in the absorptance, which governs the welding process. It is therefore important to measure the reflectance as precisely as possible.

2 Experimental Setups and Methods

Temperature-dependent reflectance measurements were conducted in an integrating sphere, that allowed to heat the sample and to freely adjust the incident angle of the measurement laser. The measurements were conducted with a 1064 nm laser source on electro tough-pitch copper (Cu-ETP or CW004A), which is widely used in electrical applications, from room temperature to above the melting point. To not completely oxidize the samples, the measurements were conducted in a nitrogen atmosphere, which is also advantageous for the production of high-quality welds. Since industrial applications usually do not use freshly polished copper with a perfect surface, copper samples with different surface conditions, namely untreated, reduced, mechanically polished and resolidified surfaces, have been included.

2.1 Optical Measurement System

Figure 1 shows a sectional view of the experimental setup. The central element is an integrating sphere with inner diameter of 500 mm, which is mounted rotatable around the axis of the sample holder to allow any desired incident angle of the laser on the sample. The inner surface of the sphere is coated with barium sulfate (BaSO₄) for improved light integration. The sphere has 6 ports: A sample port for sample delivery, a reference beam port, a sample beam port, a viewport for process observation with a camera, a gas port for the supply with shielding gas and a measurement port. A continuous wave fiber laser (not depicted) delivers the 1064 nm radiation through an optical fiber to the collimation optics at the top of the sphere. The laser beam leaves the collimation optics with a power below 0.4 W, which is low enough to not significantly heat the sample. The already linearly polarized beam is then again polarized by a linear polarizer with a guaranteed extinction ratio higher 10⁵ in order to stabilize the polarization and to prevent reflectance fluctuations due to uncontrolled polarization shifts. The polarization stabilized beam then arrives at the optical chopper wheel, which rotates at a fixed frequency of 6.5 Hz to switch through the three different phases of measurement:

• 1. Background signal measurement, twice per rotation: The chopper wheel blocks sample beam path and reference beam path to enable the background signal measurement.

• 2. Reference signal measurement, once per rotation: Mirror 1 on the chopper wheel reflects the laser beam into the reference beam path to enable tracking of changes in the incident laser power or in the behavior of other setup components.

• 3. Sample signal measurement, twice per rotation: One of the two cutouts in the copper blade allows the laser beam to pass straight through the sample beam port to the sample in the sphere.

Fig. 1

Section view of the integrating sphere setup

During the sample signal measurement, the s-polarized sample beam hits the sample at an incident angle of 12°. The reflected fraction of the beam is collected by the sphere coating, which integrates the light flux by multiple reflections within the sphere. In the case of unpolished samples, the surface structure from the sheet rolling process is aligned parallel to the incident plane of the laser to avoid that the main directions of the reflections include any ports of the sphere.

During the reference signal measurement, the reference beam is reflected by mirror 1 on the chopper wheel to mirror 2, which reflects the beam through one of the two cutouts in the chopper wheel to mirror 3. Mirror 3 reflects the beam through the reference beam port onto mirror 4 in the sphere, which reflects the beam onto the coating of the sphere, in proximity of the position to which the sample reflects the main proportion of the sample beam.

During the sample signal measurements as well as the reference signal measurements, a proportion of the integrated light flux arrives at the measurement port below the baffle. The geometry of the baffle is optimized to block direct illumination from all positions on the sphere coating to which the sample can directly reflect or scatter. This improves the independence of the measurement signal from the angle under which the sample beam is reflected or scattered. Light that reaches the measurement port is scattered by the ground glass diffusor into a black anodized extension tube, which serves as a spatial filter that absorbs light of larger 3° with respect to the tube axis. This helps to reduce blue shifts of the transmission window of the bandpass filter behind the tube. The transmission window has a central wavelength of 1064 nm and a full width half maximum of 8 nm. It blocks most of the heat radiation, which else would exceed the laser signals by factors up to larger 10 and thus decrease the available measurement resolution. Finally, the transmitted light arrives at the photodiode, which detects the sample signal, the reference signal and the background signal. A transimpedance amplifier converts the diode current into a voltage, which gets measured by a measurement card at a sampling rate of 50 kHz.

During the background signal measurement, no laser light enters the integrating sphere. Since the bandpass filter cannot block the heat radiation at the laser wavelength without impeding the measurement, the regularly updated background signal measurement is additionally required for the complete elimination of all background signals in the measured sample signal and reference signal.

2.2 Temperature Control System

The sample in Fig. 1 rests on the heating element out of silicon nitride (Si₃N₄). The sample and the maximum temperature position of the heating element are both aligned with the position of the sample beam. The heating element has a length of 75 mm, a width of 14.7 mm and a thickness of 4.4 mm, while the copper samples for the melting experiments have a length of 49 mm, a width of 30 mm, and a thickness of 1 mm. The larger width results in overhanging edges that do not have direct contact with the heating element and therefore remain as a solid frame when the center of the sample is already melting. This frame keeps the melt flat in the sample plane despite the cuprophobic properties of the heating element it is lying on. It thus allows measurements with a defined incident angle but without the risk of melt-contaminations, as observed in pre-experiments when using a metallic base plate.

Preliminary tests with a ratio pyrometer and the copper samples of this study showed variations of the measured fusion temperature of up to 600 K, which is why a type S thermocouple was employed for the temperature measurement instead. The high electrical conductivity of the sample allows to individually connect the two individual wires of the thermocouple only by the sample surface, as shown in Fig. 1. This configuration brings the material transition as the actual measurement position right to the surface of the sample. Consequently, the agreement between the measured temperature and the actual temperature of the sample surface is substantially better than in the standard configuration, in which a weld bead directly interconnects the thermocouple wires. Also, the surface contacting problem is reduced to either measuring the correct temperature or measuring no temperature at all. The distance between the thermocouple wires and the laser spot is 9 mm, what leads to a negative deviation of the measured temperature but is necessary to avoid any relevant interference of the wires with the reflectance measurement.

The closed-loop temperature control of the sample was realized by a microcontroller. Depending on the current sample temperature, which the controller received from the thermometer once per second, it adjusted the input power for the heating element through a phase-angle control by commanding a solid state relay. The 100 Hz cutting frequency of the phase-angle control was high enough to keep the temperature of the sample sufficiently stable at any preset value.

2.3 Experimental Procedures

To prevent measurement errors due to changes in environmental variables or due to possible degradations of the setup, the initial reflectance of each sample was individually compared to that of a reference mirror. To do so, the average value of a 60 s measurement on the sample without thermocouple wires was compared to the combined average of a pre- and a succeeding 60 s measurement on the reference mirror in place of the sample. The reference mirror was an ion beam sputtering low loss laser mirror (short: IBS mirror) with a theoretical reflectivity of 99.999837% for s-polarized light at the wavelength of 1064 nm (guaranteed 99.98%). The mirror specification error was therefore < 0.02%.

For the temperature-dependent reflectance measurement, the thermocouple wires were contacted to the already referenced sample on the heating element. After the sphere was closed, it was flushed with nitrogen (N5.0) for at least one hour before the heating experiments to prevent oxidation during heating. Flushing was continued throughout the experiments until the sample was back at room temperature, since the sphere could not be hermetically sealed and thus required a slight overpressure to prevent gas contamination. The measurements started with a segment of at least one minute at room temperature. The average signal of this segment was used in the later evaluation to calibrate the measurement with the predetermined initial reflectance of the sample. In the second segment, the temperature control heated the sample up to the predefined sample temperature, kept the temperature for a predefined duration and then shut down the power supply. In some measurements, the power supply was manually shut down to prevent the sample from collapsing due to complete melting. Repetitions of the reflectance measurements were conducted with the four different sample types from Table 1.

Table 1 Description of sample types

In addition to these regular measurements, a cyclic experiment was conducted, through which the reversibility of reflectance changes in a specific temperature window was investigated. For this experiment, the temperature control heated an untreated sample to about 1073 K, kept the temperature for a specified duration, then let the sample cool down to 853 K and kept the temperature again for a specified duration. From there the temperature control heated the sample and let it cool down two more times to the respective temperatures with the respective holding times and then let the sample cool down back to room temperature.

Further, the following spectroscopic measurements were conducted on selected samples to get additional information on possible chemical changes of the sample surface as a result of the conducted reflectance measurements: Energy-Dispersive X-ray spectroscopy (EDX), Nuclear Reaction Analysis (NRA), Rutherford Backscattering Spectroscopy (RBS) and Raman Spectroscopy

2.4 Signal processing

Figure 2(a) shows a raw signal segment of 50,000 data points, equal to one second, as it was recorded by the measurement card. Three levels are clearly distinguishable: The background signal at about 0.11 V, the reference signal at about 1.24 V and the measurement signal at about 1.76 V. Since the chopper wheel had two cutouts but only one mirror, the reference signal only appears half as often as the other two signal types.

Fig. 2

(a) 1 s raw signal as recorded, (b) separation of background, reference and sample signal, (c) corrected and averaged reference and sample signal with the resulting ratio ${Q}_{\mathrm{sample}2\mathrm{ref}}$

In the first step of signal processing, the processing routine suppressed all negative edges from the sample signal to the background signal. Then, in predefined distances from the beginning of each remaining data block, the routine selected a predefined number of consecutive data points as background data (orange), reference data (yellow) and sample data (purple), as displayed in Fig. 2(b), and transferred each data type to the corresponding columns of a data matrix. The distance of selected data to slopes or to the bump in the background signal was sufficient to exclude any undesired influence on the results. The bump is a result of reflections from a counterweight to mirror 1, that balanced the chopper blade and still passed through the incident beam when the reference beam path got already cleared. Since the first gray data block in Fig. 2(b) is incomplete, the preceding shift from suppressed to unsuppressed data is missing in the processed signal segment. Consequently, the whole data block was ignored to avoid any error potential due to shifted starting points.

In the next step, the routine summarized the data points of every 0.5 s interval (equals 25,000 data points) by one mean value per data type. This reduced the noise and the required data processing capacity for the following steps significantly, while the update rate was still double as high as the 1 Hz from the thermometer. Then, the routine calculated the corrected reference signal ${U}_{\mathrm{ref}}\left(t\right)$ from the reference raw signal ${U}_{\mathrm{ref},\mathrm{raw}}\left(t\right)$ and the concurrent background signal ${U}_{\mathrm{backgr},\mathrm{raw}}\left(t\right)$, leading to

$${U}_{\mathrm{ref}}\left(t\right)={U}_{\mathrm{ref},\mathrm{raw}}\left(t\right)-{U}_{\mathrm{backgr},\mathrm{raw}}\left(t\right)$$

(1)

The routine calculated the corrected sample signal ${U}_{\mathrm{sample}}\left(t\right)$ from the sample raw signal ${U}_{\mathrm{sample},\mathrm{raw}}\left(t\right)$ likewise, which resulted in

$${U}_{\mathrm{sample}}\left(t\right)={U}_{\mathrm{sample},\mathrm{raw}}\left(t\right)-{U}_{\mathrm{backgr},\mathrm{raw}}\left(t\right)$$

(2)

The ratio of the two results

$${Q}_{\mathrm{sample}2\mathrm{ref}}\left(t\right)=\frac{{U}_{\mathrm{sample}}\left(t\right)}{{U}_{\mathrm{ref}}\left(t\right)}$$

(3)

represents the uncalibrated equivalent to the reflectance of the sample $R\left(t\right)$. The resulting curves are shown in Fig. 2(c). Finally, the routine calculated the calibrated reflectance of the sample

$$R\left(t\right)=\frac{{Q}_{\mathrm{sample}2\mathrm{ref}}\left(t\right)}{{Q}_{\mathrm{mirror},\mathrm{avg}}}*{R}_{\mathrm{mirror}}$$

(4)

with Q_mirror,avg being the averaged ratio of the cleaned signals for the IBS mirror and R_mirror being the known reflectance of the mirror. The routine plotted the resulting reflectance of the sample over the corresponding temperature of the same time stamp to show the temperature-dependent reflectance of the sample $R\left(T\right)$.

3 Results and Discussion

3.1 Results

Figure 3(a) shows the measurement results for 3 untreated copper samples, which were cut from the same copper sheet of 100 µm thickness. Two of the untreated samples were placed on a polished sample to increase the overall copper thickness and thus avoid rupture of the melt. The specifications of all sample types are summarized in Table 1. All three untreated samples show a similar behavior with two substantial reflectance increases. The measurements start with a reflectance between 94.74% and 94.94%. After a short initial decrease, the reflectance rises to between 95.60% and 95.77% at a temperature of about 670 K. From there it decreases again, until the second substantial increase takes place from values between 94.95% and 95.21% at about 860 K to values between 96.76% and 96.91% at about 1050 K. Above this temperature, the measured values do not show a clear trend until the reflectance drops due to melting. The plot only shows two reflectance drops, because the sample without support of a polished sample was not sufficiently heated for melting. Only one of the reflectance drops leads to available reflectance values for liquid copper, since the melt film ruptured for one of the samples despite the support of a polished sample. The two reflectance drops happen at a measured temperature of 1190 K and 1220 K. The available values for the liquid copper show an average reflectance of 93.66% at a temperature of 1349 K.

Fig. 3

Recorded reflectance over temperature for (a) 3 measurements on separate untreated copper samples, (b) 6 measurements on separate polished copper samples, (c) 3 measurements on separate reduced samples, (d) 7 measurements on resolidified copper, some of them conducted consecutively on the same sample

Figure 3(b) shows the measurement results for 6 polished samples as they are described in Table 1. They originate from a 1 mm thick sheet of Cu-ETP to avoid rupture during polishing and to avoid rupture of the melt during the measurements as it happened among the untreated samples. The measurements show one substantial reflectance increase up to temperatures of around 600 K, which is followed by an approximately linear decrease up to the melting temperature. The measurements start with an initial reflectance between 96.57% and 97.70%. After a short initial decrease, the reflectance rises to values between 97.96% and 98.27% at about 600 K. From there, the reflectance decreases approximately linear until reaching values between 96.60% and 96.95% at the melting point, with measured temperatures between 1270 K and 1320 K. The reflectance in the liquid melt lies between 93.40% and 93.94% at measured temperatures between 1251 K  and 1376 K, depending on the sample.

Figure 3(c) shows the measurement results for 3 reduced samples as they are described in Table 1. These samples were only heated to shortly below their melting temperature to avoid rupture of the samples. Their reflectance principally shows an approximately linear decrease from the initial temperature to the maximum temperature, with a temporary but significant drop followed by a rebound back to the initial linear decrease. For an unknown reason, the reflectance values of one measurement start to deviate persistently by approximately + 0.13 percentage points above 620 K and later by up to about + 0.35 percentage points at temperatures above 1000 K. The values of the other two measurements, however, are generally in good agreement with each other. The measurements start with an initial reflectance between 98.90% and 98.96%. The reflectance drop starts between 740 K and 800 K and reaches its minimum value of 96.13% to 96.88% at temperatures between 950 K and 1050 K. After the rebound, the reflectance continues to decrease until it reaches a final reflectance of 97.13% at 1077 K for the one and 96.94% at 1189 K for the other of the two agreeing measurements. The deviating measurement has a final reflectance of 97.36% at 1130 K.

Figure 3(d) shows the measurement results for the resolidified samples as they are described in Table 1. Up to 3 subsequent measurements were conducted on the same resolidified sample, depending on how many measurements the melt film allowed without bursting. This resulted in 7 measurements distributed over 4 different samples. The general course of the reflectance is an approximately linear decrease, followed by a dip with rebound and then a tendentially decreasing reflectance until the reflectance drop due to melting. The measurements start with an initial reflectance between 97.41% and 98.70%, which shows no dependence on the initial reflectance of the polished sample or the number of conducted resolidification cycles. Despite the different initial values, the reflectance in all 7 measurements decreases at a similar gradient, ranging between −2.92 percentage points and −3.68 percentage points per 1000 K, with a tendency of steeper gradients for lower initial reflectance levels. The reflectance dip starts at above 800 K and reaches a minimum of between 93.37% and 96.49% at around 1000 K. At the end of the following rebound, the reflectance has a more uniform reflectance of 96.80% to 97.15% at temperatures around 1200 K. All samples show reflectance values between 96.60% and 96.97% when they melt at temperatures between 1302 K and 1344 K. The reflectance values for the melt lie between approximately 93.4% and 93.8% at measured temperatures of above 1305 K for one of the measurements and between 1340 K and 1355 K for the 6 other measurements.

Figure 4(a) shows an overlay plot for all samples of (a) the measured temperature over time and (b) the measured reflectance over temperature. Most of the samples were heated at the maximum power of the heating element. However, the untreated sample which was not supposed to melt was heated at a lower and defined heating rate of 8 K/s. This leads to a significantly increased number of data points for the reflectance over temperature during heating but does obviously not affect the course of the curve in a significant extent. An even lower heating rate of 2 K/s applies to all reduced samples. Their data points in the plot of the reflectance over temperature are therefore that close to each other that the individual data points can only hardly be distinguished.

Fig. 4

(a) Overlay of temperature curves for all sample types, (b) overlay of reflectance measurements for all sample types

When comparing the different sample types, the highest average initial reflectance applies to the reduced samples with an average value of 98.93%, followed by 98.12% for the resolidified samples, 97.01% for the polished samples and 94.85% for the untreated samples. The spread for the initial reflectance is the lowest for the reduced samples with 0.06 percentage points, followed by the untreated samples with 0.20 percentage points, the polished samples with 1.13 percentage points and the resolidified samples with the greatest spread of 1.29 percentage points. The reflectance over temperature shows the most similar behavior between the reduced and the resolidified samples. Both sample types start with a reflectance decrease of comparable slope over a wider temperature range, followed by a reflectance dip with rebound within the range of 800 K to 1200 K. Unlike for the reduced samples, the rebound for the resolidified samples typically results in a significantly increased reflectance compared to the values before the beginning of the dip. On the other side, the polished and the untreated samples all show, after a rather short initial decrease over some tens of kelvins, a substantial increase of reflectance up to a temperature in the region of 600 K. From there, the polished and the untreated samples show a reflectance decrease of a similar gradient as the other sample types. The main difference in the general behavior between the polished and the untreated samples is the second substantial reflectance increase for the untreated samples in the temperature range of 860 K to 1050 K. The polished samples instead continue the same linear decrease function as the reduced samples, only without the reflectance dip and the following rebound. Despite the often significant differences of the initial reflectance, all measurements in Fig. 4 (b) end in the same corridor of reflectance over temperature after reaching a temperature of about 1200 K, except for the single deviating measurement on reduced copper. The reflectance after the reflectance drop at melting temperature is comparable for all successfully melted samples but shows substantial scattering within each measurement. Therefore, the mean value for the reflectance of liquid copper just above melting temperature of 93.68% delivers a more representative value for the reflectance of molten copper. It was calculated from the mean values of all measured reflectance values after completion of the reflectance drop and before the beginning of the not depicted rise of the reflectance upon resolidification. Summarized, except for the dips in the cases of the reduced and the resolidified samples, the negative slope sections for all samples show a comparable reflectance gradient. The main difference between the different samples is the occurrence and magnitude of reflectance increases in the approximate temperature ranges from 400 K to 600 K and from 900 K to 1200 K.

Figure 5 shows the measurement results for the cyclic experiment in which the temperature control changed the temperature of an untreated copper sample multiple times between 1073 K and 853 K before letting the sample cool back down to room temperature. Each target temperature was kept constant for a specific duration. Since the measurements were conducted with a previous version of the measurement setup, the exact measurement values are not directly comparable with the ones for the other experiments. Major differences are the calibration with a gold mirror instead of the more accurate IBS mirror, a less advanced compensation of transverse effects due to heating of the measurement setup and a different aging condition of the sample. The experiment is nevertheless suitable to analyze the basic effect that leads to the observed reflectance increase above 800 K for the untreated sample type. The reflectance in Fig. 5(c) starts with a short decrease from its initial value, followed by an increase up to a temperature of about 600 K. After another reflectance decrease up to a temperature of about 800 K, the reflectance increase of interest is clearly visible. From about 900 K, the reflectance starts to decrease again until the temperature is kept constant at 1073 K. Despite the constant temperature, the reflectance rises and continues to rise when the temperature is already decreasing, until the measured temperature falls below about 950 K. Then the reflectance decreases and continues to decrease even when the temperature stays constant at 853 K, until the temperature exceeds a measured temperature of about 1000 K again. From there it rises again until the measured temperature falls below 950 K another time. The choice of temperature holding times of 1 minute at 853 K and 10 min at 1073 K led to the results in Fig. 5(c), in which the reflectance over temperature in the repetitions follows exactly the path of the previous cycles. After the third holding time at 853 K, the reflectance increases again during final cooling. Figure 5(a) and (b) show the corresponding reference respective sample beam signals over temperature. They demonstrate that the reflectance decrease at 853 K does not result from cooling of the setup but from the sample surface, since the reference signal continues to rise.

Fig. 5

Threefold heating and cooling cycle with (a) reference signal, (b) sample signal and (c) reflectance, with arrows indicating the process direction

The spectroscopic measurements with energy-dispersive X-ray spectroscopy (EDX) and Raman spectroscopy did not reveal any significant differences between an untreated sample without previous heating and an untreated sample that was previously heated in nitrogen to 1073 K within the integrating sphere. The results of the nuclear reaction analysis (NRA) for carbon, nitrogen and oxygen showed a substantial decrease of carbon from 75.76e15 atoms/cm² to 9.85e15 atoms/cm² and of nitrogen from 7.33e15 atoms/cm² to 0.21e15 atoms/cm². However, the presence of oxygen stayed nearly unchanged with 23.15e15 atoms/cm2 for the untreated sample and 22.29e15 atoms/cm² for the heated sample. The Rutherford backscattering spectrometry (RBS) did not show any significant difference between the samples.

3.2 Discussion

3.2.1 Interpretation of Reflectance Measurements

The results show that a universally applicable reflectance over temperature, that applies to all types of copper samples, does not exist. Instead, not only the initial value at room temperature but also the development of the reflectance during heating is strongly related to the specific sample conditions. The measured reflectance values for samples of the same type all show the same trends at comparable temperatures, but their exact reflectance at a given temperature depends on their exact initial conditions and may differ significantly before a temperature of about 1200 K is reached. Above this temperature, no significant differences between different samples were noticeable anymore. Therefore, the goal of this discussion is not to find a generalized equation for the temperature-dependent reflectance, but to discuss the general trends that were observed for the different sample types, identify important features and try to find the reasons for them.

Since all reflectance measurements were conducted in nitrogen atmosphere, no strong oxidation effects were expected. Likewise, no significant reduction was expected for the already reduced samples and the polished samples, which should be almost oxide free. Consequently, the reduced and the polished samples were expected to show a behavior close to that of pure copper within the purity of Cu-ETP. Resolidified copper should theoretically show a similar behavior since it is formed from the melt of Cu-ETP in the nitrogen atmosphere. The measured initial reflectance for the reduced samples is highly uniform, with a spread of only 0.06% over all 3 samples. The reflectance develops almost linearly from the beginning, at least up to 700 K, and continues in the direct extension of this path after the reflectance dip with rebound. The reflectance of the polished samples joins the same path over temperature after reaching 600 K. From 1200 K at the latest, the reflectance of all sample types runs, within small limits, along this line. This is therefore considered the baseline for the reflectance of pure copper. As can be expected after a uniform reflectance just before the melting point, the reflectance for the liquid copper after the reflectance drop does not show any dependence on the initial surface conditions, at least within the scope of the samples. The fluctuations of the value might result from dynamics in the melt surface.

The reflectance measurements of all samples deviate from the baseline in certain temperature ranges at solid state, which are specific to the initial surface condition and the existing ambient conditions. Unexpectedly, the polished samples do not start with the same initial reflectance as the reduced samples but at lower values, that also show a significantly larger spread. Both are indicators of surface imperfections. The reflectance of all polished samples increases from shortly above room temperature until it homogenizes among all polished samples and aligns with the discussed baseline of copper. A similar effect can be observed at similar temperatures for the untreated copper. The reflectance of these samples starts with the lowest initial reflectance and shows a similar behavior until about 650 K, only at a generally lower level of reflectance. It is still too low to join the baseline at this temperature, but it starts to decrease parallel to it. According to the manufacturer, the untreated copper was treated with the organic compound benzotriazole as oxidation inhibitor. The evaporation of benzotriazole cannot be finally confirmed as the reason for the first reflectance peak of the untreated copper, but the temperature of the reflectance increase in the measurements is in good agreement with the evaporation temperature of benzotriazole at 623 K. Also, the decrease of carbon in the results of the nuclear reaction analysis (NRA) from 75.76e15 atoms/cm² to 9.85e15 atoms/cm² and the decrease of nitrogen from 7.33e15 atoms/cm² to 0.21e15 atoms/cm² rather support than contradict this assumption. The polished samples were not treated with benzotriazole during or after polishing, but the presence of residues from, for example, the diamond suspension on their surface cannot be completely excluded, despite cleaning in an ultrasonic bath with isopropanol, rinsing with fresh isopropanol, and drying with a hairdryer thereafter. The evaporation of organics is therefore also a possible explanation for the increase of the reflectance until shortly above 600 K for the polished samples, although a final proof is missing.

The reflectance curves for the polished samples do not exhibit any further indications for changes of the surface condition until the melting point. But the reduced, the resolidified and the untreated samples all show a significant reflectance increase that starts at measured temperatures between 850 K and 1050 K. In the case of the untreated samples and the resolidified samples, the difference in the starting temperature is similar to the difference between the temperatures at which the two sample types melt. Since the melting temperature of copper is a physical constant, this indicates that the difference in the recorded temperatures is only the consequence of different temperature measurement errors and that therefore the same temperature applies. In the case of the reduced and the resolidified samples, the reflectance increase is preceded by a reflectance dip that is significantly more pronounced than the decrease of the baseline of copper. The spectroscopic measurements, including the nuclear reaction analysis (NRA), did not reveal significant changes of the oxygen content. But the threshold temperature at which the reversible reflectance decrease transitions into the reversible reflectance increase in the reflectance cycle in Fig. 5 agrees well with the threshold temperature at which the reflectance drop transitions into the rebound for the other samples. The evident reversibility of the reflectance increase and decrease in Fig. 5, along with the fact that the only available reaction partners are the sample and the surrounding gas, only allows the conclusion that the reflectance changes are caused by reduction and oxidation of the sample. According to the Ellingham diagram for copper and its oxides (Ref 19), the oxygen partial pressure for the equilibrium of

$$4\mathrm{Cu}+{\mathrm{O}}_{2}=2{\mathrm{Cu}}_{2}\mathrm{O}$$

(5)

at a temperature of 1050 K is given with values between 10⁻³ Pa and 10⁻⁴ Pa. Due to the position of the thermocouple wires, the actual threshold temperature in the laser spot was most likely even higher than the measured temperature, what further increases the derived oxygen partial pressure. The integrating sphere was not equipped to measure the residual oxygen partial pressure, but the given values appear reasonable if considering the nominal purity of the delivered gas. The results strongly suggest that the intensified decrease of the reflectance for the reduced and the resolidified samples before 1000 K were caused by oxidation and that the following increase for all samples but the polished ones results from the reduction of oxides. The oxidation below the equilibrium temperature might also be the reason why the resolidified copper starts with lower and less uniform initial reflectance values than the reduced copper: while the reduced copper was heat treated in a specialized oven that provides a feature for quick cooling within the sealed volume, the resolidified sample cooled down on the heating element in the constantly flushed but not sealed integrating sphere. Therefore, it is possible that it was subject to oxidation, similar to the sample in Fig. 5.

In the end, the reflectance of all samples basically follows the baseline of copper, but this behavior is in some cases superposed by other effects. With regards to the process stability in laser welding, the reduced copper seems to be the best option at first look, since it has the most uniform starting values for the reflectance. It also follows the baseline of copper without any indications of chemical reactions, except for the reflectance dip with rebound, which is likely to be less pronounced if the sample is heated quickly enough. However, for most processing applications the absorptance is of interest, for which the initial value of the reduced copper is the lowest among all the samples. Relative to the derived absorptance value of 1.04% to 1.10%, the spread of 0.06 percentage points already scales to about 5.6%. The spread of the initial reflectance values of the untreated copper is larger than for the reduced copper, but if the 0.2 percentage points are related to the derived absorptance of 5.26% to 5.06% this spread is only 3.9% and thus smaller. This clearly demonstrates that the suitability of the highly reflective copper materials for processes like laser welding can only be reasonably estimated from the absorptance point of view. Also, while the differences between the samples seem to be small when looking at their reflectance values, the average initial absorptance of 5.15% for the untreated samples is 4.81 times as high as the average initial absorptance for the reduced samples of 1.07%.

3.2.2 Data Quality and Transfer to Simulations

The measured reflectance values at room temperature in this study fall within the range that could be expected from the values for similar surfaces from the literature overview. The temperature-dependent measurements of this study in the solid phase agree best with the values of the diamond turned copper sample from Quimby, Bass et al. (Ref 4), who used an calorimetric approach. While their value at room temperature corresponds to a reflectance of 99.19%, the reduced samples of this study have an average reflectance of 98.93%. At the highest temperature in Quimby, Bass et al. (Ref 4), their value corresponds to about a reflectance of 97.5%, while the average reflectance in this study is about 97.0%. The results of this study thus deviate from theirs on average by −0.26 percentage points at room temperature and by −0.51 percentage points at 1070 K. Regarding the absorbed power, the absorptance in this study is 132% of their value at room temperature and 120% of their value at 1070 K, which is a good agreement compared to how the results of the other studies deviate from each other. Quimby, Bass et al. do not present any values for liquid copper. The average reflectance value for liquid copper of 93.68% in this study shows the best agreement with the calculated value of about 93.9% for liquid copper at 1873 K from Kohl, Kaufmann et al. (Ref 2). Transferred to the absorptance, the value of this study is 103.6% of their value and thus approximately equal. In a previous study (Ref 20), in which the temperature-dependent absorptance from Blom, Dunias et al. (Ref 18) was incorporated into a simulation model, it was found that the laser power to melt the sample in the simulations was 1.7 times lower than in corresponding welding experiments. Since the absorptance values from Quimby, Bass et al. (Ref 4) are lower than the ones in Blom, Dunias et al. (Ref 18), the best agreement with Quimby, Bass et al. is a promising indication that the measurement results of this study come close to the actual temperature-dependent absorptance of copper.

Before using the results of this study for simulations, some aspects regarding the setup and the experimental procedure should be taken into consideration. In general, the measurement results of this study have demonstrated a good reproducibility with respect to the reflectance; at the temperature of 1200 K, the maximum spread for the reflectance between all measurements is less than 0.6 percentage points. Also, the results did not show a relevant dependence of, for example, noise from the current value. However, the remaining absorptance decreases with increasing reflectance, which means that relative errors with respect to the absorptance value tend to grow substantially as the reflectance increases. Therefore, the measurement results for the reduced and polished samples have, on average over the whole temperature range, a lower reliability for the prediction of the exact absorbed laser power than the measurement results for the untreated samples.

Further, 2 effects were identified, that might lead to deviations of the reflectance measurements in this study from the actual reflectance during, for example, a welding process. One of these effects is a thermal drift of the reflectance values, which is not related to the actual reflectance but to an unintended temperature increase of the setup. The other one is the time-dependence of the reflectance decrease or increase due to oxidation and, respectively, reduction of the sample surface.

The effect of time-dependence for the reflectance decrease or increase due to oxidation or reduction can be observed in the experiment in Fig. 5. The longer the sample stayed in the respective temperature region, the stronger was the total increase or decrease of the reflectance. Therefore, the velocity, at which the sample is heated, can have a substantial effect on the extent of surface oxidation or reduction until a specific temperature is reached. In the measurements of this study, the samples were heated relatively slowly at heating rates of only some kelvins per second. However, when heating the sample surface to its melting temperature within milliseconds or even faster, as it is not unusual in laser welding of copper, the effect of oxidation or reduction might be significantly less pronounced than in the measurements of this study. In some cases, such as the already reduced sample, oxidation might even be negligible and as a result, neither dip nor the subsequent rebound of the reflectance between 800 K and 1000 K would occur.

Figure 6(a) demonstrates the effect of the thermal drift at the example of a measurement for a polished sample. The reflectance generally increases with decreasing temperature during cooling, what agrees with the literature. But when the temperature has almost returned to room temperature, the trend changes and the reflectance decreases by another 0.2% to 0.4%, although the concurrent temperature decrease is nearly negligible in comparison to the previous decrease. Figure 6(b) shows the same measurement over time, where the increase in reflectance turns into an exponential decrease that is typical of cooling processes. As the sample is heated, the temperature of the integrating sphere, the fitted components and the contained gas volume inevitably increase as well. The constant gas exchange through flushing with nitrogen mitigates the temperature increase but cannot fully prevent it. The increase of the setup temperature makes the measured reflectance drift to higher values. The direct determination of the drift magnitude with respect to time or sample temperature is challenging, since the temperature-induced reflectance drift cannot be measured without changing the sample temperature as well. In additional experiments, the heating element was shifted out of the sample beam and a mirror was placed as sample replacement on the bottom of the sphere. This configuration was meant to allow similar heating conditions of the setup while maintaining an approximately constant sample temperature, but the reflectance drifted into the opposite direction and demonstrated that the reflectance drift strongly depends on the sample position. Therefore, it seems to be more promising to estimate the expected reflectance drift through laser melting experiments and corresponding simulations. These simulations should include the measured reflectance curves of the used samples, in which a temperature-dependent linear reflectance drift compensation is calibrated until simulation and experiments match. In order to avoid substantial estimation errors due to the different time-dependence of the redox-reaction between reflectance measurement and laser melting experiment, this procedure should be conducted with samples that neither oxidize nor reduce during heating.

Fig. 6

(a) Measured reflectance over temperature for a polished sample from before heating until cooled back to room temperature and (b) the same process, but plotted over time

Besides the two effects on the reflectance data, the temperature measurement has two regimes, which can be referred to by the unmolten regime and the molten regime. When heating in the unmolten regime, the measured temperature is always lower than the temperature in the laser spot, since the thermocouple wires cannot be placed in the center of the laser spot, where the temperature of the sample is highest. Hence, the sample melts at measured temperatures below the theoretical melting temperature. This temperature deviation can be used to recalibrate the temperature for the solid sample, providing a good approximation with respect to the position of the reflectance measurement. However, as the sample melts, the melt pool will usually grow quickly and reach the thermocouple wires, which is when the diagrams of reflectance over temperature tend to show a shift to the direction of the theoretical melting temperature. As this happens, the thermocouple wires do no longer measure in the unmolten regime but in the molten regime. This includes a permanent change of the contact conditions between the sample and the heating element in the whole molten area. The molten regime results in a more homogenous temperature distribution, which is why the resolidified samples show the lowest deviation of the measured temperature from the theoretical melting temperature upon melting.

Different processes can lead to slightly different reflectance curves of the reflectance over temperature, even if the same surface conditions apply. Therefore, the measurement results of this study are presented in their original version. However, the above discussion of the effects can be used to adjust the results with respect to the process to be simulated.

4 Conclusions

In this study, the reflectance of electro tough-pitch copper was experimentally measured over temperature for different surface types, which were categorized as untreated, reduced, polished, or resolidified. The measurements were conducted in nitrogen atmosphere and covered the temperature range from solid copper at room temperature to liquid copper just above the melting point. The detailed measurement results revealed significant differences of the reflectance over temperature for the different sample types but also typical features for the temperature-dependent reflectance of each specific group. Most of the striking features could be explained and their significance for processes like laser welding was discussed. However, the measurements revealed a temperature-related reflectance drift of the integrating sphere setup as an error source that could not be eliminated completely. Also, experiments showed that the effect of redox reactions on the reflectance depends on the heating rate, what should be considered when applying the data in simulation models that involve significantly faster heating. Simulations that implement the measurement results in combination with accompanying experiments for comparison might help to characterize these error sources more in detail and to perform the suggested compensation strategies.