Surface temperature of a 2 in. Ti target during DC magnetron sputtering Open Access

Magnetron sputtering is one of the most important techniques for the deposition of thin films. The history and development of sputtering have been extensively described and explained in a number of publications.^1–6 However, as mentioned in these and other review papers, the complexity of the plasma-electromagnetic field-target surface interactions and various other aspects of the deposition process are not completely understood.⁵ Sputtering is the ejection of atoms by the bombardment of a target surface by incident energetic particles, mostly ions produced in the plasma. The magnetic field in a magnetron cathode traps electrons close to the target, increasing the ionization of the gas and, therefore, the ion concentration. This increase in ionization increases the flux of ions incident on the target surface and, therefore, increases the sputtering rate and the consequential deposition rate. The form of the magnetic field produces a nonuniform erosion racetrack in the target surface. This nonuniform erosion has several consequences, target material utilisation is typically less than 30%, the target temperature (T_T) is not uniform, the depth of the racetrack, and the age of the target can change the gas discharge characteristics, as well as the distribution of the sputtered material. Various authors have derived models describing the ratio between the number of ejected atoms and the number of incident projectiles, the sputter yield, as a function of energy, the angle of incidence, and gas ion-target arrangements, among other things.⁷ Many of these models include the surface binding energy of the target atoms as a way to describe the sputtering of different materials, as well as oxides and nitrides. However, none of these models include the temperature of the target and its expected effect on the strength of the surface barrier for sputtered particles.⁸ However, it has been reported that the sputtering yield of C by Ar ions increased by almost an order of magnitude as the graphite target temperature was increased from 925 to 1525 °C.⁹ Other studies found that at lower temperatures, 75–520 °C, the increase in the sputtering yield of Mo, W, and Ta targets by Ar ions was only 26%–39%. Nevertheless, some papers have reported a decrease in the sputtering yield as the target temperature was increased.¹⁰ 

It has often been reported that something like 80%–90% of the electric power supplied to a sputter target is transformed into heat, and consequently, good target cooling is necessary to avoid melting. Such target heating can strongly limit the power that can be used to deposit low melting point materials, such as Bi, Cd, In, Pb, Sn, Se, etc.

If the flow of the cooling water is reduced, the target can be made to operate at elevated temperatures. It is known that the sputter yield increases linearly with power and does not, necessarily, depend on the target temperature. However, if the temperature comes close to the melting point Tm (approx. 0.8 Tm), then so-called thermal spikes can be formed in the target surface with a corresponding strong increase in the sputtering yield. Such a “hot target sputtering” process has recently been studied and reported by several research groups.^11–17 A silicon target operated close to its melting point gave a deposition rate that augmented from about 0.05 nm/s at RT to 1.4 nm/s at 1400 °C, an increase of 28 times. High target temperature studies of Ni showed that when the target was allowed to heat up to 720 °C there was a 15%–25% increase in the deposition rate, probably because the temperature was above the Curie temperature for Ni, there were also modifications of the microstructure of the deposit.^18,19 Studies of DC magnetron sputtering of hot or cold Ti targets led to the conclusion that the differences were related to different degrees of gas rarefaction of the gas close to the hot and cold targets.^20,21

Additionally, radiation from the target can, under some conditions, strongly affect the temperature of the substrate or depositing film. Studies have shown that two components are involved in substrate heating: a fast contribution, which is attributed to collisional plasma processes, and a slower one, which is attributed to IR emission from the heated sputtering target surface.²² Measurements showed that the IR radiation from a Ti target corresponded to 15%, 3%, and 76% of the substrate heating using DC magnetron, pulsed DC, or HiPIMS (High Power Impulse Magnetron Sputtering) sputtering, respectively.²³ 

Almost all of the experiments involving hot targets have been limited as emissivity and the angular distribution of the IR emission from the target had to be assumed.^14,17 Similarly, the measured temperatures were normally taken as the maximum value, and the temperature distribution over the surface of the target was not known.

More recently, various papers have been published about reactive magnetron sputtering of several metallic oxides and nitrides using hot metallic targets. Such studies have shown that the film structure and deposition rate strongly depend on the target temperature used.^24–26 

In the present work, we measured the surface temperature of a Ti target by direct contact with a type K thermocouple. The radial distribution of the temperature was measured for three applied powers and as a function of the argon gas pressure. It is known that during the preparation of titanium nitride coatings, the nitrogen in the reactive gas mixture can chemically react with the target and can result in target poisoning, which is known to decrease the deposition rate. To attempt to provide useful information about this chemical reaction, we measured the target temperature versus the concentration of nitrogen in the gas mixture.

A 2 in. water-cooled MAK magnetron cathode was mounted facing vertically upwards in a vacuum chamber connected to a diffusion pump with a cooled baffle and a combination of roots booster and mechanical pump. The system base pressure was <2.7 × 10⁻⁴ Pa (<1 × 10⁻⁶ Torr).

In MAK sputtering sources, a permeable magnetic “keeper” is attached to the bottom of the target, such that the target is clamped and coupled to the magnetic field of the cathode without any additional clamps. A thermal contact paste is provided with the MAK cathode and is used to ensure a good thermal and electrical connection between the target and the water-cooled part of the cathode, see Fig. 1. The inner and outer diameters of this water-cooled section of the cathode are approximately, 16 and 37 mm. Therefore, only part of the back-face of the target is in contact with the cooling. The company MeiVac, Inc. recommends that the maximum power used with the 2 in. cathode be less than 1 KW (DC) or 400 W (RF).

Most sputtering experiments were carried out at 2.7 Pa (20 mTorr) using a total gas flow of 60 SCCM. A type K thermocouple was manufactured from 1.27 × 10⁻⁴ m (0.05 in.) wire, which was threaded inside a 1 mm diameter doble-bore ceramic tube. The chromel and alumel wires were connected to a PFA insulated extension cable, which passed through a ¼ in. glass tube fixed to a 25KF flange using Torr Seal epoxy. The ceramic tube was mounted such that its weight meant that the thermocouple tip was in good contact with the surface of the 2 in. × ¼″ titanium target mounted on the MAK cathode. As can be seen in the schematic drawing in Fig. 1, the pivot mount of the ceramic tube was mounted on the end of a 4 in. linear feedthrough, such that the thermocouple could be pulled across the surface of the target. The thermocouple was maintained in any given position until the recorded temperature was stable, typically 10–20 s, and it could be lifted away from the target surface to return it to the center of the target for each experiment. The thermocouple and the digital multimeter were not connected to the sputtering system ground, but were floating at the cathode potential, and therefore the voltage difference between the alumel and chromel wires (the thermocouple reading) was completely independent of the cathode potential. Additionally, the difference between the potential of the sputtering target (∼300 V) and the same potential measured via the thermocouple wires was measured using the same DMM (Digital Multi-Meter), which had an input impedance of 10⁷ Ω. The idea behind this measurement was to detect whether there was a good contact between the probe point and the target surface. Assuming that the resistance of the thermocouple wires was <5 Ω, a contact resistance of ∼30 Ω would produce a potential difference of ∼1 mV.

The lateral target surface temperature measurements were performed as a function of the applied power, with gas pressures of 1.47 Pa (11 mTorr), 2.7 Pa (20 mTorr), and 4.0 Pa (30 mTorr). The same lateral temperature measurements were carried out as a function of the nitrogen to argon concentration (0%–9%) at a fixed plasma power, gas pressure, and total gas flow. Finally, under the same discharge conditions, the maximum temperature in the racetrack was measured for nitrogen concentrations from 0$ to 28% and then back down to 0%. The power supply used for this work could only be used at values of constant power, the configuration of constant voltage or current was not available; however, the discharge current and voltage were recorded for all experiments.

In general, the lateral distribution of the temperature was not symmetric on each side of the center of the racetrack. The lateral temperature data were analyzed using the “bigaussian” peak function in the ORIGINLAB software. When a good fit is achieved, this function gives the base, the center position, and height of the distribution, as well as the left- and right-hand values of the widths at half maximum, FWHM.

The surface temperature of the target and the erosion of the surface are both a result of the ionic bombardment of the target. Now it was observed, and it is well known that the attrition of the area close to the racetrack is larger than at the center or edge of the target. Therefore, the ion flux close to the racetrack is greater than at the center or edge of the target, and a similar variation in the surface temperature was observed. A complication of this model is that the thermal input to the surface of the target could diffuse laterally and through the target to the cooled back-face of the target, as well as heating the argon gas. So, the lateral variation of the surface temperature almost certainly depends on the target material, the design of the cooling system, and the gas pressure. The topography of the target was measured using a shape tracer; see Fig. 2. It is difficult to quantify the racetrack profile and compare that with the shape of the temperature distribution, but it is clear that the asymmetry of both forms is less steep toward the center of the target than toward the outer edge.

Optical observation and photographs demonstrated that the thermocouple assembly did not appear to significantly disturb the magnetron gas discharge.

Deposition experiments were carried out on pieces of silicon at a target-to-substrate distance of 6 cm as a function of the nitrogen concentration used for the temperature experiments.

X-ray photoelectron spectroscopy analyses were performed in an ultra-high vacuum (UHV) system, the Scanning XPS microprobe PHI 5000 VersaProbe II, with an Al K_x-ray source (hν = 1486.6 eV) monochromatic with a 200 μm beam diameter and a MCD analyzer. The surface of the samples was etched for 10 min with 2 kV Ar⁺ at 0.5 μA/mm². The XPS spectra were obtained at 45° to the normal surface in the constant pass energy mode (CAE) E0 = 117.40 and 11.75 eV survey surface and high-resolution narrow scan. The peak positions were referenced to the background silver 3d5/2 photopeak at 368.20 eV, having a FWHM of 0.65 eV, and C 1 s hydrocarbon groups at 285.00 eV central peak core level position.

The potential difference between the target and ground measured directly and measured via the thermocouple wires was <0.2 mV for all of the nonreactive studies, indicating good physical contact between the two parts. In this way, we could assume that the probe temperature was the same as the target surface.

Figure 3 shows the lateral variation of the surface temperature of the Ti target as a function of the applied DC power. For a 2 in. target, the applied powers correspond to power densities of approximately 1.0, 2.2, and 4.1 W/cm². Figure 4 shows that thse variation of surface temperature with the applied power was slightly less than linear, but the maximum temperature was almost 840 °C. The data in Fig. 3 show that both the width of the temperature distribution and the position of the maximum temperature changed with the applied power.

Figure 5 shows that both the half width at full maximum (sum of the left- and right-hand side widths) and the position of the maximum temperature changed almost linearly with the applied power. In particular, the position of the maximum temperature moved toward the center of the target, and the width of the hottest part of the surface of the target was enlarged.

The lateral distribution of the target temperature as a function of Ar gas pressure for an applied power of 37 W is shown in Fig. 6. For gas pressures greater than 2.7 Pa (20 mTorr), the width of the distribution did not vary, and the maximum temperature occurred at 30 mm from the edge of the target. Repeated experiments at pressures close to 1.47 Pa (11 mTorr) showed that the width of the distribution did not change at lower gas pressures, but the maximum temperature moved approximately 1 mm closer to the edge of the target.

In relation to gas rarefaction, it is often supposed that the main contribution to the heating of the gas is the energy transfer by collisions between sputtered atoms and gas atoms, but some recent work has included the gas pressure.^27,28 One of those studies showed that the gas temperature did not vary for gas pressures >2.5 Pa for medium plasma powers but was notably lower at 1.5 Pa because the cross section of the elastic scattering of a sputtered atom with a gas particle was significantly lower. The position and dimensions of the temperature distribution versus gas pressure probably reflected the size and form of the gas discharge and the gas temperature since the discharge is the source of the Ar ions that are incident on the target and cause the heating of the surface.

The lateral variation of the temperature of the target surface was measured as a function of the nitrogen content in the gas mixture, the applied power was 47 W, and the total gas flow and pressure were kept constant at 60 SCCM and 2.7 Pa (20 mTorr), respectively. A second series of experiments were performed where the thermocouple was placed in the hottest part of the racetrack and the maximum temperature was recorded as the N2% was increased from 0% to 28% and was then reduced to 0%. At least 20 s were used for each measurement to allow the target temperature to stabilize. The average potential difference between the thermocouple probe and the target was <10 mV for the Ar + N₂ studies. The highest voltages were measured for the center and edge sections of the target, probably due to the existence of a TiNx layer. Even though the voltage was higher than in the no-reactive case, the small value indicated that there was still good physical contact between the two parts.

The lateral distribution of the surface temperature for a nitrogen content in the discharge gas from 0% to 9% can be seen in Fig. 7. Again, the lateral distributions were analyzed using the Bigaussian function in ORIGINLAB, and the results are given in Fig. 8. The lateral position of the maximum temperature, Xc, is shown, and the FWHM of the distribution is given by width. The position of T_max moved 4 mm closer to the center of the target as the N₂% was increased to 5% and then returned by 2 mm as the N₂% was increased to 9%. The width of the distribution of the target surface temperature increased by approximately 3 mm as the N₂% was changed from 0% to 9%.

In the experiments up to a N₂ concentration of 9%, it was seen that the maximum temperature tended to decrease as the concentration was increased, but the details were not clear. Figure 9 shows the maximum temperature of the racetrack and the discharge voltage as the N₂ concentration was increased to 28% and then decreased to 0%. The gas pressure, total gas flow, and discharge power were kept constant at 2.7 Pa (20 mTorr), 60 SCCM, and 50 W. The maximum temperature was checked, and the position of the thermocouple was adjusted for each N₂ concentration. It can be seen that both the maximum temperature and the discharge voltage increased as the N₂ content increased from 0% to 2%. For higher concentrations, the discharge voltage was almost constant, while the temperature gradually decreased by more than 20 °C. The Ti target was probably in the poisoned mode for all N₂ concentrations >2%. As the N₂ concentration was decreased to close to 6%, the discharge voltage showed little change, but as the temperature increased to close to 500 °C, both parameters returned to their initial values when the N₂ flow was turned off. The observed variations are similar to those reported by other groups in relation to some gas discharge characteristics during reactive sputtering.²⁹ 

The results of XPS analysis of films are given in Table I for samples prepared using a gas pressure of 2.7 Pa (20 mTorr), a total gas flow of 60 SCCM, and an applied DC power of 47 W. The films had a fairly high oxygen content of about 20%, probably because the base pressure of the sputtering system was only 1 × 10⁻⁶ Torr. It can be seen that 20% of the nitrogen in the gas mixture was sufficient to form a deposit of approximately stoichiometric titanium nitride. The vacuum system and gas lines had been checked for leaks, but as mentioned earlier, the system base pressure was close to 2.7 × 10⁻⁴ Pa. The substrate was not heated, and the coating thickness was typically about 100 nm, with a deposition rate of approximately 3–4 nm/min. Under these conditions, the oxygen content of the films was close to 20% throughout the whole thickness of the deposit.

The existence of the somewhat surprisingly high temperatures on the surface of the target implies several important consequences for both reactive and nonreactive sputtering. Obviously, the maximum temperature limits the plasma power that can be used for many targets of materials that have melting points lower than, or close to, 900 °C, but such high temperatures may also strongly affect the gas temperature and pressure near the target or the temperature of substrates. Similarly, the sputtering yield at such temperatures is not known for the majority of the materials, and during reactive sputtering, it can be expected that high temperatures change the chemical reactions between the target material and the reactive gas.

Even though water-cooled magnetron cathodes are well designed and it is unusual to observe hot water coming out of the cooling system, we have demonstrated that the surface temperature of a Ti target mounted on our MAK 2 in. cathode during DC sputtering was not uniform and could be very hot. At plasma power densities of 1.0, 2.2, and 4.1 W/cm² and Ar gas pressure of 2.7 Pa (20 mTorr), the temperature difference at the outer edge and the center of the target was not more than 200 °C, but the corresponding racetrack temperatures were approximately 430, 630, and 835 °C. It was found that as the plasma power was increased, the maximum temperature increased, the FWHM of the maximum temperature zone in the racetrack increased, and its position moved outward.

The maximum temperature, FWHM, and position of the hottest area did not vary for gas pressures greater than 2.7 Pa (20 mTorr). At a lower pressure of 1.47 Pa (11 mTorr), the width of the hot zone did not change, but the maximum temperature increased by about 100 °C and moved closer to the center of the target by approximately 1 mm. This may be related to a reduction in the thermal conduction from the target to the gas because of a more notable gas rarefaction in front of the target at lower pressures.

The maximum temperature as a function of the concentration of nitrogen in the gas mixture was quite complex, particularly as the concentration was increased and then decreased. The width of the hottest area of the target increased as the nitrogen concentration increased. In reactive sputtering, it is known that the target metal reacts with the reactive gas used, but normally it is assumed that such a reaction is driven by the plasma and not the temperature of the target surface.

Almost all studies of sputtering using a hot target assume that the target temperature is uniform and try to measure the temperature using optical techniques, assuming the value of the emissivity of the target.

This study was carried out using a partially used Ti target, in the future, we plan to continue this study by including a new Ti target and other materials such as silver, aluminum, copper, and graphite.

The authors would like to acknowledge financial support of the PAPIIT Project No. IG101123, of Dirección General de Asuntos del Personal Académico, Universidad Nacional Autónoma de México (DGAPA-UNAM).

The authors have no conflicts to disclose.

Carlos Ramos: Data curation (equal); Investigation (equal); Resources (equal). Daniela S. Jacobo-Mora: Data curation (equal); Investigation (equal); Methodology (equal). Julio Cruz: Data curation (equal); Methodology (equal); Writing – review & editing (equal). Stephen Muhl: Conceptualization (lead); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal).

The data that support the findings of this study are available within the article.