*Article* **Properties of Tool Steels and Their Importance When Used in a Coated System**

#### **Bojan Podgornik \*, Marko Sedlaˇcek, Borut Žužek and Agnieszka Guštin**

Institute of Metals and Technology, SI-1000 Ljubljana, Slovenia; marko.sedlacek@imt.si (M.S.); borut.zuzek@imt.si (B.Ž.); agnieszka.gustin@imt.si (A.G.)

**\*** Correspondence: bojan.podgornik@imt.si; Tel.: +386-1-4701-930

Received: 29 January 2020; Accepted: 10 March 2020; Published: 12 March 2020

**Abstract:** The introduction of new light-weight high-strength materials, which are difficult to form, increases demands on tool properties, including load-carrying capacity and wear resistance. Tool properties can be improved by the deposition of hard coatings but proper combination and optimization of the substrate properties are required to prepare the tool for coating application. The aim of this paper is to elaborate on tool steel substrate properties correlations, including hardness, fracture toughness, strength and surface quality and how these substrate properties influence on the coating performance. Results show that hardness of the steel substrate is the most influential parameter for abrasive wear resistance and load-carrying capacity, which is true for different types of hard coatings. However, high hardness should also be accompanied by sufficient fracture toughness, especially when it comes to very hard and brittle coatings, thus providing a combination of high load-carrying capacity, good fatigue properties and superior resistance against impact wear. Duplex treatment and formation of a compound layer during nitriding can be used as an additional support interlayer, but its brittleness may result in accelerated coating cracking and spallation if not supported by sufficient core hardness. In terms of galling resistance, even for coated surfaces substrate roughness and topography have major influence when it comes to hard ceramic coatings, with reduced substrate roughness and coating post-polishing providing up to two times better galling resistance.

**Keywords:** tool steel substrate; coatings; hardness; fracture toughness; load-carrying capacity; wear

#### **1. Introduction**

In forming applications of modern metallic materials, including die casting, stamping, forging and rolling, tool lifetime is limited due to very demanding working conditions. These include mechanical, thermal and impact loading [1,2]. Under such complex working conditions, comprising high contact stresses as well as abrasive and adhesive wear [3–5], tool surfaces are attacked by different wear and fatigue mechanisms [6–8]. By the current demands on reducing mass and size of components, lowering fuel consumption and CO2 emission, increasing recycling and improving overall strength and safety [9,10]—especially when talking about transportation and energy sector— tools are exposed to new and more severe requirements and demands. This is related to design of the tool, selection of material and heat treatment [11], and surface engineering, where substrate preparation is essential [12].

Increased demands mean strengthened requirements in terms of the tool material properties, including temperature resistance, strength, shock and fatigue resistance, impact and sliding wear resistance, etc. Ductility and fracture toughness are among the main tool properties essential for the majority of forming applications and the influencing of tool resistance [13,14]. Properties of the tool core material, i.e., tool steel, depend on its chemical composition and production process, but primarily on the heat treatment parameters. Heat treatment defines final microstructure and corresponding

properties. In general, steel properties, mainly focused on strength and hardness are provided by a well-defined heat treatment process, consisting of an austenitization treatment with a subsequent quenching and a multiple tempering [13,15]. Typically, a trade-off between toughness and hardness is required [16]. On the other hand, through optimized heat treatment procedures, involving vacuum hardening, selection of tempering and austenitizing temperature, and inclusion of deep cryogenic treatment [17], fine-grained microstructure with homogeneous carbides distribution can be obtained [15]. Thus, improved toughness and fatigue resistance is obtained while maintaining high strength and hardness [18].

The typical property used for the planning of heat treatment for tool steels is hardness. High hardness is related to abrasive wear resistance and resistance to plastic deformation. However, beside hardness there are also other material properties, which are based on the application and surface engineering techniques, which become more important as we use tools that are more complex. These include fracture toughness, compressive and bending strength, creep and wear resistance, machinability, etc. [2]. Different material properties are defined and determined by different standards and test methods. However, each standard and test method uses specific test specimens with different geometries. Different geometries relate to different heat treatment conditions, and thus in microstructure deviation [19] and problematic properties correlation. On the other hand, circumferentially notched and fatigue-pre-cracked tensile bar (CNPTB) specimens used for measuring fracture toughness of more brittle materials, i.e., tool steels have been found as the best alternative, allowing determination and correlation of many different properties [20]. The advantage of the CNPTB specimen is in its radial symmetry and uniform microstructure through the whole volume.

Introduction of light-weight high-strength materials, being very difficult to form also sets more demanding properties requirements on tool properties, especially its surface and wear resistance [21]. One way of improving wear properties of the tool is application of different heat and diffusion treatments, i.e., plasma nitriding, thus modifying surface microstructure and properties [22,23]. Another way, proven in cutting tool applications is deposition of wear resistant coatings [24]. However, limited load-carrying capacity, adhesion and topographical characteristics of the substrate restrict successful application of hard wear-resistant coatings on forming tools. Thus, proper combination and optimization of the substrate properties are needed in order to prepare tools for coating deposition and provide improve performance of the tool [23–27].

The aim of this research work, carried out by using CNPTB test specimen configuration was to determine correlations between different tool steel properties, including fracture toughness, hardness, compressive and bending strength, wear resistance and surface quality and how these substrate properties influence on the coating performance.

#### **2. Materials and Methods**

#### *2.1. Materials*

Two tool steels being the most common in forming and tooling industry have been included in this investigation. First one, aimed at studying the effect of heat treatment parameters as well as additional plasma nitriding on the tool steel substrate properties, their correlation and influence on the load-carrying capacity and sliding wear resistance was conventional AISI H11 type hot-work tool steel (H11), produced through casting, electro slag re-melting, forging and annealing. The second one was high fatigue strength P/M cold work tool steel (PM-CW), aimed at determining the influence of hardness and fracture toughness as well as surface preparation on the load-carrying capacity as well as impact and galling wear resistance when coated with different type of hard coatings. Chemical composition of the investigated tool steels is given in Table 1.


**Table 1.** Chemical composition of the investigated tool steels (wt %).

#### *2.2. Heat Treatment and Coatings*

Heat treatment of hot work tool steel (H11) included vacuum heat treatment performed in Ipsen VTC 324-R horizontal vacuum furnace (Ipsen, Kleve, Germany). The specimens machined from soft annealed material were preheated to 850 ◦C and then progressively heated at 10 ◦C/min to the austenitizing temperature of 990–1000 and 1030 ◦C, respectively, soaked for 20 min and quenched in N2 gas flow at a cooling speed of 3 ◦C/s. After quenching specimens were double tempered for 2 h. First tempering was always performed at 540 ◦C, immediately followed by second tempering at six different temperatures, varied from 550 to 630 ◦C.

Specimens planned to investigate the effect of heat treatment parameters on the load-carrying capacity and sliding wear were further plasma nitrided in a Metaplas Ionon HZIW 600/1000 reactor (Metaplas Ionon, Bergisch Gladbach, Germany), surface polished and coated. Plasma nitriding was performed at 540 ◦C for 20 h using different gas mixtures. One group of test specimens was treated in 25% N2:75% H2 gas mixture, resulting in diffusion zone of about 260 μm and approx. 5 μm thick top compound layer. Another group was treated in 5% N2:95% H2 gas mixture, providing ~230 μm thick diffusion zone without any compound layer. After nitriding all specimens were polished to a mirror-like finish (*R*<sup>a</sup> = 0.1 μm) and coated with the commercial TiN/TiB2 nanocomposite multilayer coating. Coating consisted of a primary TiN monolayer, a multilayer zone (TiN/Ti-B-N) with a lamella thickness of 85 nm and a top TiB2 overcoat. TiN/TiB2 coating with total thickness of ~2 μm and hardness of 3000 HV was deposited by a bipolar-pulsed glow discharge PACVD. Processing temperature was 530 ◦C and pressure 200 Pa.

In the case of PM cold work tool steel (PM-CW) three groups of vacuum heat treatment parameters were used and combined with the process of deep cryogenic treatment (Table 2). First group (A1) providing maximum hardness was quenched from high austenitizing temperature (1130 ◦C) and triple tempered at low tempering temperatures (520/520/490 ◦C). In order to increase fracture toughness but still maintain hardness above 64 HRC second group (A2) was austenitized at 1100 ◦C and tempered at 500 ◦C (500/500/470 ◦C). The last group (A3) was hardened from 1070 ◦C and triple tempered at increased tempering temperature (585/585/565 ◦C), thus providing high fracture toughness. In the cases when vacuum heat treatment was combined with deep cryogenic treatment—DCT (groups B1–B3), DCT was performed immediately after quenching and followed by a single 2 h tempering. DCT consisted of a controlled immersion of the test specimens in liquid nitrogen for 25 h (Table 2). After heat treatment specimens were mirror polished (*R*<sup>a</sup> = 0.10 μm), sputter cleaned and coated. Investigation included three representative PVD coatings. These were monolayer TiAlN coating (~3300 HV), AlTiN/TiN multi-layer coating (~3500 HV) with lamellas thickness of ~50 nm and ~80 nm, respectively, and (Ti,Si)N nano-composite coating (~3800 HV). All three coatings were about 2 μm thick and deposited by magnetron sputtering process at the substrate temperature of ~450 ◦C. Details of the deposition process are provided in Reference [28,29].

The effect of substrate roughness on galling performance was evaluated by preparing A1 group of specimens with four different procedures; coarse grinding and polishing with a 20 μm industrial polishing paste (A1-1; *R*<sup>a</sup> = 0.5 μm), coarse grinding and double polishing with 20 μm and 10 μm polishing paste (A1-2; *R*<sup>a</sup> = 0.3 μm), fine grinding (A1-3; *R*<sup>a</sup> = 0.15 μm) and polishing (A1-4; *R*<sup>a</sup> = 0.1 μm). Afterwards specimens were coated with commercial TiN monolayer and W-doped DLC multilayer coating [30].



#### *2.3. Mechanical Properties*

For each material CNPTB specimens [20] (Figure 1) were machined and used for further investigation. Fracture toughness was measured by pre-cracking CNPTB specimen under rotating-bending loading (400 N, 4500 cycles). After pre-cracking specimens were subjected to tensile load at the cross-head speed of 1.0 mm/min until fracture. Measuring the load at fracture (*P*) and diameter of the fractured area (d) fracture toughness is calculated according to the Equation (1) [31,32]. Details of the CNPTB specimen and fracture toughness measurement technique are given in Ref. [20]. For each group of treat treated specimens at least 12 samples were characterized in order to provide reliable results.

$$K\_{\rm Ic} = \frac{P}{d\_0^{3/2}} \cdot \left( -1.27 + 1.72 \frac{d\_0}{d} \right) \tag{1}$$

**Figure 1.** (**a**) Circumferentially notched and fatigue-pre-cracked tensile bar (CNPTB) specimen and (**b**) extraction of 4-point bending, load-carrying capacity, compression and sliding/impact wear test specimens.

Rockwell-C hardness measurements, performed circumferentially (×3) on each CNPTB specimen, were carried out on Willson-Rockwell B2000 machine (Buehler, Esslingen, Germany) and then average value calculated.

One half of the fractured CNPTB specimen was used to machine φ 10 mm × 12.5 mm cylinder for compression test (Figure 1) and φ 18 mm × 8 mm disc for sliding and impact wear testing. Compression tests were performed according to ASTM E9-09 standard [33] and used to determine yield strength, maximum compression strength and strain hardening exponent. Strain hardening exponent was defined between yield and maximum compression stress on a log-log plot of true stress-true strain. The other part of the fractured CNPTB specimen was used to prepare 4-point bending test specimen (φ 5 mm × 60 mm) or load-carrying capacity test specimens (φ 10 mm × 60 mm) (Figure 1). After high-speed machining, bending and load-carrying capacity test specimens were ground and polished (*R*<sup>a</sup> = 0.1 μm). 4-point bending tests at room temperature were performed according to ASTM E290-09 standard [34], using support span of 40 mm and load span of 16 mm.

Effect of heat treatment on the machinability was analyzed by measuring surface roughness of as machined 4-point bending specimens. High-speed machining involved pre-turning with standard cutting inserts (Sandvik-Coromant DNMG11 R0.4, Sandviken, Sweden) at a cutting speed of 100 m/min, depth of cut of 0.3 mm and feed rate of 0.12 mm/rev, followed by final turning (VBMT 16 04 cutting

inserts) at the same cutting speed, depth of cut of 0.2 mm and feed rate of 0.08 mm/rev. Each specimen was machined with a new cutting insert and surface roughness analyzed (ISO 4287:1997 standard [35]) by Alicona InfiniteFocus G4 microscope (Alicona, Raaba, Austria).

#### *2.4. Load-Carrying Capacity and Wear Testing*

Load-carrying capacity was evaluated by load-scanning test rig, Figure 2a. The test configuration consists of two crossed cylinders (10 mm) which are sliding against each other at fixed speed, but progressive loading. Thus, each position of the wear scar corresponds to a unique load without any loading history [36]. In this investigation coated tool steel cylinder was loaded against polished WC cylinder (*R*<sup>a</sup> = 0.05 μm, 2200 HV), using dry sliding conditions, room temperature, fixed sliding speed of 0.01 m/s and normal load in the range of 400–4000 N (*p*<sup>H</sup> = 2.8–6.1 GPa). Load-carrying capacity is determined by defining critical loads at which first cracks in the coating are observed and when coating starts to flake [4,36].

Sliding wear tests were done under dry sliding conditions using ball on disc contact configuration and reciprocating sliding (Figure 2b). Polished 20 mm diameter Al2O3 ball (*R*<sup>a</sup> = 0.05 μm) was used as a counter material in order to simulate abrasive wear mode and focus all the wear on the investigated disc material. Test were done under normal room conditions (RT and 45% RH) and elevated temperature of 150 ◦C applying different contact conditions; loads corresponding to contact pressure between 800 and 1300 MPa and sliding speeds between 0.01 and 0.1 m/s, obtained by changing oscillating frequency from 1 to 15 Hz. All tests were performed up to the total sliding distance of 100 m (up to two hours), with the average coefficient of friction being analyzed and wear volume measured using 3D confocal microscope.

Impact wear tests were performed on servo hydraulic fatigue testing machine, with the coated disc being repetitively impacted against a WC ball (φ 32 mm; Figure 2c). Testing machine is position controlled during testing, which includes continuous monitoring of the impact force. Impact wear tests were performed at the frequency of 30 Hz, initial impacting distance of 0.5 mm and maximum impacting load of 5.5 kN (*p*<sup>H</sup> = 3.5 GPa). New WC ball was used for each test and testing specimens cleaned with ethanol. Adhesive wear of the WC ball was prevented by applying a thin layer of lithium grease on the disc surface.

**Figure 2.** Load-carrying capacity and wear testing setups; (**a**) load scanner, (**b**) reciprocating sliding wear test, (**c**) impact wear test.

Effect of substrate roughness and surface quality on resistance against galling was also evaluated by a load-scanning test rig (Figure 2a). In this case tempered austenitic stainless steel (ASS) cylinder (AISI 304, 335 HV, *R*<sup>a</sup> = 0.2 μm, φ 10 mm) was used as a moving counter cylinder. Galling tests were performed dry, using sliding speed of 0.01 m/s and normal load from 20 to 1300 N. Results were then evaluated by analyzing wear tracks after sliding and determining critical loads for galling initiation and gross galling formation [36,37].

#### **3. Results**

#### *3.1. Tool Steel Substrate Properties Correlation*

Diagram displaying fracture toughness and hardness of vacuum heat treated AISI H11 type hot work tool steel as depending on the tempering and austenitizing temperature is shown in Figure 3. Fracture toughness obtained by hardening from 1000 ◦C followed by double tempering at 630 ◦C was 87 MPa√m. It was reduced to less than 30 MPa√m by reducing tempering temperature to 550 ◦C. Hardness, on the other hand, increased from 40 HRC to almost 50 HRC. Further hardness increase is provided by increased Si content and austenitizing temperature. In the case of low Si content, increase in austenitizing temperature to 1030 ◦C results in about 5% higher hardness (up to 52 HRC) and for high Si content even more, especially at low tempering temperatures, up to 54 HRC. However, for low Si hot work tool steel increase in austenitizing temperature also provided higher fracture toughness, being between 45 and 115 MPa√m, while for high Si content fracture toughness has been reduced at elevated austenitizing temperature, ranging between 25 and 80 MPa√m, as shown in Figure 3.

**Figure 3.** Effect of Si content and heat treatment temperatures on fracture toughness and hardness of AISI H11 hot work tool steel.

Yield and ultimate compression strength of the investigated hot work tool steel (AISI H11) are between 1200 and 1850 MPa, and 1450 and 2130 MPa, respectively. Similar to hardness, ultimate compression strength and yield strength are increasing when increasing austenitizing temperature (for 5%) and Si content, but decreasing with tempering temperature, as shown in Figure 4a. On the other hand, material ductility being analyzed by measuring strain hardening exponent was found mainly independent on the austenitizing and tempering temperature. For the tempering temperatures up to 610 ◦C it shows more or less constant value, 0.45 for low Si content and 0.4 for high Si content.

In the case of bending test, maximum and yield strength at the austenitizing temperature of 990 ◦C changed from 1860 and 3260 MPa to 2600 and 4550 MPa, respectively, when reducing tempering temperature from 630 to 550 ◦C. Further increase was obtained by increasing austenitizing temperature to 1030 ◦C. In this case and low Si content, yield and maximum bending strength reached peak values (tempering at 550 ◦C) of 2700 and 4780 MPa, respectively, and even up to 2800 and 4950 MPa, respectively, for high Si content (Figure 4b).

When analyzing correlations, strong correlation between hardness and strength of tool steel substrate has been observed. In agreement with well-established correlations [38,39] compression and bending strength increase linearly with hardness, but dropping digressively with fracture toughness, as shown in Figure 5. On the other hand, strain hardening exponent has no direct correlation with hardness but it shows rising trend with increased fracture toughness (Figure 6).

**Figure 4.** (**a**) Compression and (**b**) bending strength tempering diagram.

**Figure 5.** Strength vs. (**a**) hardness and (**b**) fracture toughness correlation.

**Figure 6.** Strain hardening exponent vs. fracture toughness correlation.

Surface roughness analysis of machined 4-point bending and load-carrying capacity specimens revealed deteriorated surface quality with higher average roughness and intensified tearing component when increasing tempering as well as austenitizing temperature [39]. As shown in Figure 7, lower average roughness values (*R*a), lower kurtosis (*R*ku) representing less sharp surface profile and zero skewness (*R*sk) indicating symmetric profile, with all indicating improved machinability and better surface quality of tool steel substrate, are obtained when increasing hardness (above 45 HRC) and having fracture toughness below 60 MPa√m. However, when material becomes too hard, above 50 HRC or too tough (>80 MPa√m) surface quality quickly deteriorates (Figure 7) and becomes too rough for coating deposition [40].

**Figure 7.** Effect of (**a**) hardness and (**b**) fracture toughness on surface quality.

In terms of tribological properties, coefficient of friction for AISI H11 tool steel was found largely independent on the austenitizing and tempering temperature, displaying average value of about 0.75, which is well in agreement with many tribological investigations on tool steels. On the other hand, wear volume was found to increase with tempering temperature and being dependent also on austenitizing temperature. Furthermore, the form and rate of increase was dependent on the contact conditions used, as shown in Figure 8. In the case of low load-low sliding speed and high load-low sliding speed (Figure 8a) conditions wear volume shows linear increase with tempering temperature, with the higher austenitizing temperature giving faster increase rate and higher wear, especially for high loads. By increasing the sliding speed (low load-high sliding speed and high load-high sliding speed—Figure 8b) effect of austenitizing temperature on wear has been reversed. In these cases, wear shows exponential increase with tempering temperature but drop in values for higher austenitizing temperature, which is more pronounced in the low tempering range (Figure 8b). Furthermore, the best wear resistance and the lowest wear was observed for mid-tempering range between 570 and 590 ◦C. This indicates that for low sliding speeds higher fracture toughness obtained by lower austenitizing temperature is dominating over hardness, while for high sliding speeds hardness prevails.

**Figure 8.** Wear of AISI H11 tool steel as a function of heat treatment temperatures; (**a**) low sliding speed and (**b**) high sliding speed case.

When it comes to friction under abrasive wear conditions, steady-state coefficient of friction for tool steel was found largely unaffected by fracture toughness and hardness when operating within working hardness range (42–52 HRC). On the other hand, as shown in Figure 9, wear volume and wear rate, defined as wear volume divided by load and sliding distance show direct dependency on hardness and fracture toughness. Under the abrasive wear conditions wear rate increases with fracture toughness and is reduced with hardness, with the hardness being found as the most influencing parameter. For the best abrasive wear resistance tool steel hardness should be above 48 HRC (strength above 1900 MPa) and fracture toughness below 55 MPa√m, although not too low (Figure 9).

**Figure 9.** Effect of (**a**) hardness and (**b**) fracture toughness on hot work tool steel wear resistance.

#### *3.2. Thermo-Chemical Treatment vs. Coating*

Based on Archard law [41] abrasive wear resistance of materials is in general dependent on their hardness. Hardness increase and thus better wear resistance of steels can be achieved by increasing austenitizing and decreasing tempering temperature [42]. However, even higher surface hardness is provided by thermo-chemical processes, i.e., nitriding and deposition of wear resistant coatings [12]. As shown in Figure 10 plasma nitriding in 5% N2:95% H2 gas mixture, providing nitrided surface without compound layer and a hardness of 1100 HV0.05 reduces wear of hot work tool steel by about 25%. However, although operating under abrasive wear mode and wear being concentrated within the nitride zone of just 250 μm [4] steel core microstructure and properties, especially hardness and fracture toughness play an important role in terms of surface wear resistance. For lower austenitizing temperature with the hardness in the range of 47–50 HRC and fracture toughness above 35 MPa√m wear rate of nitrided surface increases as the hardness is reduced (higher tempering temperature). On the other hand, increase in austenitizing temperature, providing higher core hardness (51–53 HRC) but much lower toughness (<30 MPa√m; Figure 3) resulted in about 30% higher wear of the investigated AISI H11 tool steel, both at room and high temperature sliding. In this case wear resistance of plasma nitrided tool steel has been found more or less unaffected by core hardness, but mainly defined by reduced fracture toughness, which is further escalated by nitriding and formation of hard brittle surface zone [43].

Although nitriding improves wear resistance of metallic surfaces [44] it cannot match wear resistance provided by PVD and CVD coatings. As shown in Figure 10, coating of tool steel by hard protective TiN/TiB2 coating can improve surface wear resistance by two orders of magnitude. Furthermore, when load is carried entirely by the top coating and substrate deformation is within elastic range wear rate of the coated surface becomes independent on the substrate preparation and properties, including heat treatment temperatures and/or plasma nitriding used [4].

**Figure 10.** Effect of heat treatment, plasma nitriding and coating deposition on wear resistance of hot work tool steel.

In the case of coated surface load-carrying capacity is the primary requirement. As shown in Figure 11, it doesn't depend just on core hardness but also on the subsurface properties determined by the eventual thermo-chemical process [4]. In the case of plasma nitriding, producing few microns thick compound layer (25% N2:75% H2) on the hot work tool steel surface, increase in tempering temperature and corresponding drop in core hardness of less than 5 HRC led to about 30% lower critical loads for TiN/TiB2 coating cracking and flaking or spallation. Even though compound layer may provide additional load support [45], it is quite brittle and thus its cracking resistance dependent on the core hardness. Any crack starting in the compound layer will propagate directly into the substrate but mainly into and through the top coating [46], Figure 12. By avoiding formation (5% N2:95% H2) or removing the intermediate compound layer load-carrying capacity drops further. However, it is influenced by the combined effect of fracture toughness and core hardness. For low hardness values (50 HRC) and fracture toughness above 30 MPa√m (austenitizing @1000 ◦C) increase in fracture toughness obtained by higher tempering temperatures prevails over the reduction in core hardness providing better coating resistance to cracking and load-carrying capacity. Positive effect of core fracture toughness is present also for lower toughness values (<30 MPa√m) but it fades away as soon as core hardness drops to about 50 HRC, as shown in Figure 11.

**Figure 11.** Dependency of load-carrying capacity on substrate heat treatment and plasma nitriding conditions [4]. Reprinted with permission from [4]. Copyright 2015 Elsevier.

**Figure 12.** Crack pattern in TiN/TiB2 coated and plasma nitrided hot work tool steel with intermediate compound layer.

#### *3.3. E*ff*ect of Substrate Hardness and Fracture Toughness on Coating Performance*

Coatings represent only one part of the coated system and require proper substrate giving sufficient support. Thus, beside coating type its properties and performance greatly depend on the substrate, mainly its hardness and fracture toughness. Hardness provides resistance to plastic deformation while fracture toughness is responsible for hindering crack initiation and propagation. As shown in Table 3, by combining different heat treatment parameters (tempering and austenitizing temperature) with eventual deep cryogenic treatment different combinations of fracture toughness and hardness for cold work tool steel can be obtained.


**Table 3.** Fracture toughness and hardness of P/M cold work tool steel using different heat treatment procedures.

Load-carrying capacity results for three different coatings (monolayer, multilayer, nano-composite) obtained for different core hardness vs. fracture toughness values are shown in Figure 13. In the case of a TiAlN monolayer coating deposited on the hardest tool steel substrate (66 HRC; Group A1) first signs of coating cracking are observed at the critical load of about 3.1 kN. Increase in fracture toughness from 6 to 10 MPa√m at the same time resulting in reduced substrate hardness (66→64 HRC; Group A2) leads to loss in load-carrying capacity, reducing critical load for coating cracking for about 10%, down to 2.8 kN. Further drop in hardness (60 HRC; Group A3), in spite of providing high fracture toughness values results in additional 10% drop in load-carrying capacity (*L*<sup>C</sup> <2.5 kN). However, when maintaining high hardness level (above 65 HRC) any increase in fracture toughness, obtained by combining conventional heat treatment with deep cryogenic treatment [17,27] will provide better crack initiation and propagation resistance and thus higher load-carrying capacity (Group B1).

In the case of AlTiN/TiN multilayer coating with improved cracking resistance, substrate hardness above or equal 64 HRC provides comparable load-carrying capacity, regardless of the *K*Ic/HRC ratio and level of the fracture toughness obtained (Figure 13). However, with the drop in substrate hardness below 60 HRC (Group A3) critical loads for the beginning of coating cracking are reduced, indicating about 20% lower load-carrying capacity. This time increase in fracture toughness (Group B3) provides some load-carrying capacity improvement, although not reaching the same level as with the harder substrates.

**Figure 13.** Load-carrying capacity of different coating types as a function of substrate heat treatment and properties.

The most marked effect of substrate properties on load-carrying capacity is found for the most brittle and hardest (Ti,Si)N nano-composite coating. Again, the best load-carrying capacity is provided by the hardest substrate, but very low for too soft one, irrespectively of the fracture toughness level obtained. On the other hand, fracture toughness becomes important for the intermediate working hardness values, as shown in Figure 13. Combination of high fracture toughness (>12 MPa√m; Group B2) and working hardness of about 64 HRC guarantees load-carrying capacity similar to hardest substrates but at greatly improved fatigue resistance.

Another coating property, altered by substrate properties is its impact wear resistance. In agreement with high load-carrying capacity the best impact wear resistance for different types of hard coatings is achieved when applied on steel substrate with the highest hardness (Group A1; Figure 14). In this case coatings are removed through abrasive wear mechanism, without any evident coating delamination or cracking. Even the smallest drop in hardness, even though accompanied by increased fracture toughness (Groups B1 and A2) leads to increased coating impact wear and beginning of coating delamination. However, if hardness of about 64 HRC is paired with the fracture toughness above 12 MPa√m (Group B2) coating impact wear can be reduced for up to 30%, closely matching harder but more brittle substrate case. Improved fracture toughness retards crack initiation and propagation while high hardness guarantees high load support. Further rise in fracture toughness, meaning drop in hardness below 60 HRC (Group A3 and B3) has clear negative effect. Low load-carrying capacity of the steel substrate results in excessive deformations and thus in high impact wear of the coating. However, for the most brittle and the hardest coatings (i.e., (Ti,Si)N) substrate hardness is found as the dominant factor in terms of impact wear resistance. In this case, the steel substrate with the highest hardness is the most suitable (66 HRC; Group A1). Use of any softer substrate results in coating cracking and flaking with wear exceeding coating thickness (Figure 14).

**Figure 14.** Impact wear volume for different coating types and substrate pre-treatments.

#### *3.4. E*ff*ect of Substrate Roughness*

If tool is coated its galling resistance depends on the material to be formed and coating type, as well as on the substrate roughness and any additional surface treatment, i.e., post-polishing [25,47]. In the case of hard nitride-based coatings, i.e., TiN, which generally show lower galling resistance than tool steels [48], substrate roughness is even more important. For rough, coarse ground substrates (A1-1 & A1-2) coating gives higher friction (0.4) and about 25% lower galling resistance as indicated by reduced critical loads for stainless steel transfer, as compared to uncoated tool steel (Figure 15). However, by reducing roughness of the substrate (A1-3 and A1-4) coated surface provides comparable resistance to galling and material transfer. Even more, when post-polished (A1-3 + post-polishing) to Ra values of 0.1, use of coated surface can provide up to two times higher resistance to galling. On the other hand, low friction and excellent galling resistance against ASS are obtained by low-friction DLC coating, regardless of the substrate roughness and post-polishing procedure (Figure 15), with the critical loads for galling exceeding 1 kN even under dry sliding [48].

**Figure 15.** Effect of substrate roughness and post-polishing of coated surface on galling resistance.

#### **4. Conclusions**

Effect of steel substrate properties on coating performance can be summarized in the following conclusions:


Workshop practice recommendations:


**Author Contributions:** Conceptualization, B.P.; methodology, B.P.; validation, B.P., M.S. and B.Ž.; formal analysis, M.S. and B.Ž.; investigation, M.S. and A.G.; writing—original draft preparation, B.P.; supervision, B.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Slovenian Research Agency (research core funding No. P2-0050 and research project L2-9211).

**Acknowledgments:** Authors would like to acknowledge help from Miha Cekada from Institute Jozef Stefan for ˇ the deposition and supply of coatings.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Distribution of the Deposition Rates in an Industrial-Size PECVD Reactor Using HMDSO Precursor**

**Žiga Gosar 1,2, Denis Ðonlagi´c 3, Simon Pevec 3, Bojan Gergiˇc 3, Miran Mozetiˇc 4,5, Gregor Primc 4,5, Alenka Vesel 4,5,\* and Rok Zaplotnik 4,5**


**Abstract:** The deposition rates of protective coatings resembling polydimethylsiloxane (PDMS) were measured with numerous sensors placed at different positions on the walls of a plasma-enhanced chemical vapor deposition (PECVD) reactor with a volume of approximately 5 m3. The plasma was maintained by an asymmetric capacitively coupled radiofrequency (RF) discharge using a generator with a frequency 40 kHz and an adjustable power of up to 8 kW. Hexamethyldisiloxane (HMDSO) was leaked into the reactor at 130 sccm with continuous pumping using roots pumps with a nominal pumping speed of 8800 m<sup>3</sup> h−<sup>1</sup> backed by rotary pumps with a nominal pumping speed of 1260 m3 h<sup>−</sup>1. Deposition rates were measured versus the discharge power in an empty reactor and a reactor loaded with samples. The highest deposition rate of approximately 15 nm min–1 was observed in an empty reactor close to the powered electrodes and the lowest of approximately 1 nm min–1 was observed close to the precursor inlet. The deposition rate was about an order of magnitude lower if the reactor was fully loaded with the samples, and the ratio between deposition rates in an empty reactor and loaded reactor was the largest far from the powered electrodes. The results were explained by the loss of plasma radicals on the surfaces of the materials facing the plasma and by the peculiarities of the gas-phase reactions typical for asymmetric RF discharges.

**Keywords:** HMDSO; PECVD; deposition rate; uniformity of deposition; polymerization; organosilicon thin films

#### **1. Introduction**

Many materials should be coated with a thin protective layer to provide an adequate surface finish and stability in harsh environments [1–5]. A variety of techniques have been proposed, and a few have also been commercialized [6–10]. One technique for depositing compact and hydrophobic films similar to polydimethylsiloxane (PDMS) is plasma polymerization. A suitable monomer is provided and partially dissociated and ionized under plasma conditions [11,12]. The radicals adhere to the surface of any object exposed to the plasma and form a thin film. The structure and composition of the coating depend on the type of precursor, plasma parameters and specifics of the discharge used for sustaining gaseous plasma [13–17]. The growth kinetics is complex and difficult to control because of the large number of radicals formed in the gaseous plasma. An early report of the kinetics was presented by Bourreau et al. [18]. The authors used different sources to deposit protective coatings rich in silicon oxides: silane (SiH4), hexamethyl disiloxane (HMDSO) and tetraethoxysilane (TEOS). They correlated the evolution of the coverage

**Citation:** Gosar, Ž.; Ðonlagi´c, D.; Pevec, S.; Gergiˇc, B.; Mozetiˇc, M.; Primc, G.; Vesel, A.; Zaplotnik, R. Distribution of the Deposition Rates in an Industrial-Size PECVD Reactor Using HMDSO Precursor. *Coatings* **2021**, *11*, 1218. https://doi.org/ 10.3390/coatings11101218

Academic Editors: Alessio Lamperti and Alessandro Patelli

Received: 26 August 2021 Accepted: 30 September 2021 Published: 5 October 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

with the deposition kinetics and compared the growth rates. The profiles were independent of the substrate temperature or the deposition rate when silane was used as a precursor. In the case of organic precursors, however, the deposition rate decreased with an increase in the deposition temperature. They found the adsorption–desorption phenomena to be important factors for the coverage evolution. At low deposition temperatures, the film growth rate was sensitive to ion surface bombardment and resulted in a non-conformal deposit even in compounds with high surface mobility.

Theirich et al. [19] studied the gas-phase reactions in HMDSO/O2 mixtures and pressures between 20 and 70 Pa. Plasma was characterized by mass spectrometry and infrared spectroscopy. They found the film homogeneity dominated by the precursor content and its spatial distribution in the gas or plasma phase. Three reactive intermediate species were proposed to act as a precursor for silica-like film growth, all having a mass of 148 Da, so the authors concluded that further work should be performed to distinguish between the radicals.

In their classic paper, Hegemann et al. [20] studied the deposition rate and threedimensional uniformity of capacitively coupled radio-frequency (RF) plasma useful for depositing protective layers using HMDSO as a precursor. The deposition rate increased with monomer gas flow, whereas it was independent of pressure. Large differences in the deposition rates at different positions of the samples were reported, as well as the influence of the dimensions of the samples on the growth kinetics. In another paper [21], the same group investigated the deposition rate in symmetrical and asymmetrical electrode configurations and found that the deposition rate depended on the so-called reaction parameter (power input per gas flow of the monomer).

More recently, Ropcke's [22] group performed a detailed characterization of the HMDSO plasma by optical emission spectroscopy (OES) in the visible spectral range and infrared laser absorption spectroscopy (IRLAS). They used a plasma reactor of a rather large power density (discharge power per volume of the discharge chamber) of the order of 100 W per liter. They managed to derive the concentrations of the various stable and unstable plasma species, which were found to be in the range between 1017 and 10<sup>21</sup> m<sup>−</sup>3. They also studied the influence of the discharge parameters, such as power, pressure and gas mixture, on the molecular concentrations. Based on the construction principle of the reactor, the plasma generation was characterized by a certain degree of inhomogeneity with different temperature zones, i.e., hottest, hot and colder zones. This complexity was characterized by the multiple molecular species, including the HMDSO precursor and products in the ground and excited states existing in the plasma.

Plasma-enhanced chemical vapor deposition (PECVD) technique for the deposition of protective coatings from HMDSO was commercialized decades ago despite the experimentally observed non-homogeneities and instabilities, which may lead to inadequate properties of the deposited films. Recently, Gosar et al. [16] reported that the composition of the deposited films depended on the time-evolution of the plasma parameters, although the discharge parameters (power, pressure, flow rate, pumping speed) remained fairly constant. The time evolution was explained by the drifting plasma parameters, which was detrimental to the quality of the protective films, especially where a rather high power density was used to sustain the gaseous plasma. At low discharge powers, however, the properties of the deposited films were not time dependent. The quality of the films is a crucial parameter in the industrial application of the PECVD technique using HMDSO, so many industrial reactors operate at a very low power density to minimize the risk [23]. On the other hand, the low power density results in a poor deposition rate, as explained by the above-cited authors.

The problem of plasma non-uniformity and the resultant deviations of the film thickness from the desired value in large plasma reactors may be suppressed by rotating samples upon plasma processing [24]. This is a standard solution in commercial reactors for depositing protective coatings in batch mode. The samples are mounted on planetaria and moved through zones with different plasma parameters. The relatively long treatment time

(several minutes in commercial plasma reactors) ensures a reasonable coating thickness and uniformity. Still, the problem arising from plasma inhomogeneities is not solved, so there is a need to develop configurations of plasma reactors with deposition rates that are as uniform as possible throughout the entire reactor.

Commercial reactors for the deposition of the protective coatings using the HMDSO as the precursor may be upgraded if the non-uniformities are known and understood. Several groups have already reported the non-uniformity in plasma parameters, but only a few have measured the deposition rates in different parts of the plasma reactor [12,13,20]. The present paper provides measurements of the deposition rate performed with several sensors mounted in selected positions within a large plasma reactor. The deposition rates for an empty and a fully loaded reactor were measured to reveal the influence of the samples on the non-uniformity of the deposition rates.

#### **2. Materials and Methods**

#### *2.1. Plasma-Enhanced Chemical Vapor Deposition Reactor*

The industrial PECVD reactor useful for the deposition of PDMSO-like coatings was presented in detail in our previous paper [25]. The reactor has a cylindrical shape with a diameter of 1.9 m and a height of 1.8 m. During the deposition, the reactor was pumped with two roots pumps with a total nominal pumping speed 8800 m3 h–1, backed by two rotary pumps of a total nominal pumping speed 1260 m3 h–1. Before the deposition, in order to get the base pressure as low as possible (around 0.02 Pa), the reactor was also pumped with two diffusion pumps with a total pumping speed 35,000 L/s. HMDSO was the only gas that was introduced into the plasma reactor. It was introduced through a calibrated flow controller. The pressure was measured with a Pirani gauge. At the HMDSO inlet of 130 sccm (cm3/minSTP), which is the standard flow rate used in mass production, the pressure was about 4 Pa. Plasma was characterized by optical emission spectroscopy (OES) AvaSpec-Mini4096CL (Avantes, Apeldoorn, Netherlands) near one of the powered electrodes as shown in Figure 1.

**Figure 1.** Cross-section of the cylindrical PECVD reactor with the position of the pump ducts, powered electrodes (E1, E2), HMDSO inlet, sensors for deposition rate measurements (S1−S8), OES lens, optical fiber and OES spectrometer.

An asymmetric capacitively coupled RF discharge was used for sustaining gaseous plasma. The discharge was powered by an RF generator (PE II 10K, Advanced Energy, Denver, CO, USA) operating at 40 kHz and adjustable power between 1 and 8 kW. A couple of powered electrodes were mounted close to the pump duct. The area of each electrode was approximately 0.4 m2. The area of the grounded electrode (housing) was approximately 16 m2. The ratio between the areas of the powered and grounded electrodes was

approximately 40. Therefore, the plasma was sustained by an asymmetrical capacitive coupled RF discharge, and the gradients in the plasma parameters were expected.

The HMDSO inlet was provided through vertically oriented grounded metallic tubes, as shown in Figure 1. The tubes were positioned close to the grounded walls of the plasma reactor. They had small holes separated by 15 cm. The precursor was thus introduced into the reactor unevenly.

#### *2.2. Sensors of the Deposition Rate*

Eight sensors were fixed on the sidewalls of the plasma reactor (BDS-MF, Arzuffi, Vallezzo Bellini, Italy) for the real-time monitoring of the deposition rate, as shown in Figure 1 (marked with S1 to S8). The sensor S1 was positioned on the rough grid, which separates the discharge chamber from the polycold pump duct, which was not used in this experiment. A photo of the sensor S1 is shown in Figure 2a. Other sensors were fixed on the chamber walls on the grounded housing.

**Figure 2.** (**a**) Fixation of the sensor S1 and (**b**) the photo of a sensor mounting.

Each sensor essentially consisted of a single-mode optical fiber, which was cleaved and exposed to the processing chamber on one side, while being connected to an appropriate opto-electronics signal integration system on the other side. Opto-electronics signal integration system launched light into the fiber, while acquiring and processing back-reflected optical power from cleaved fiber end. Since the deposited PDMSO-like layer has a different refractive index than vitreous silica, the back-reflectance from the cleaved fiber end changed during the PDMSO deposition. This change was correlated with the change in thickness of the deposited material. The correlation was obtained by an appropriate calibration and processing of acquired signals. One such sensor was already used in our previous work [26], where the deposition rates measured with such sensor in real time were the same as those measured with time-consuming post-deposition surface analysis such as atomic force microscopy (AFM) (Solver PRO, NT-MDT, Moscow, Russia), X-ray photoelectron spectroscopy (XPS) (TFA XPS Physical Electronics, Münich, Germany) and time-of-flight secondary ion mass spectrometry (ToF-SIMS) depth profiles (ToF-SIMS 5 instrument, ION-TOF GmbH, Münster, Germany).

Figure 2b shows a photo of an optical fiber sensor fixed on the aluminum holder, which was fixed on the wall of the plasma reactor.

#### *2.3. Optical Emission Spectroscopy (OES)*

An optical lens was mounted in the PECVD reactor (Figure 1) and connected with optical fiber through optical feedthrough to a standard low-resolution optical spectrometer Avantes AvaSpec-Mini4096CL (Avantes, Apeldoorn, Netherlands). The spectrometer measures light emission spectra. The device is based on AvaBench 75 symmetrical Czerny Turner design with a 4096-pixel CCD detector with a focal length of 75 mm. The range of measurable wavelengths is from 200 nm to 1100 nm, and the wavelength resolution is 0.5 nm. The spectrometer has a USB2.0 interface, enabling high sampling rates up to 150 spectra per second. Signal-to-noise ratio is 300:1. Integration time is adjustable from 30 μs to 50 s. At integration times below 6.5 ms, the spectrometer itself performs internal averaging of spectra before transmitting them through the USB interface. The spectrometer was connected to the process computer via USB. The integration time was set to 5 s.

#### **3. Results and Discussion**

Plasma in the empty discharge chamber was characterized by OES. Here it should be stressed that an empty chamber means that there are no samples and no planetaria (sample holders) inside the reactor. A typical OES spectrum is shown in Figure 3. The spectrum consists of Balmer series of radiative transitions of H atoms from excited states to the first excited state. The next prominent spectral feature arises from the relaxation of the CH radicals with the bandhead at 431 nm. Other features are marginal. The OES indicates partial dissociation of the precursor molecules, but otherwise, it does not provide any additional significant information. Other radicals are also in the reactor, but their emission is marginal. More interesting is the intensity of the spectral features versus the discharge power. Figure 4 shows quite linear curves. The emission intensity depends on the electron density and temperature as well as the density of radicals in the ground state, and the dependence is not trivial. Still, the behavior of the lines in Figure 4 indicates either more extensive dissociation of the precursor molecules or higher electron density/temperature or both at higher power. This observation is expected, considering that the optical lens for acquiring spectra was mounted just next to the powered electrode.

**Figure 3.** An optical spectrum of the plasma at the discharge power of 5 kW and 130 sccm of HMDSO.

Figure 5 shows the measured deposition rate versus the discharge power. Interestingly enough, the deposition rate is rather constant in the broad range of powers from approximately 2 to 7 kW. This observation is not correlated with data in Figure 4, which shows a gradual increase in the emission intensity. This paradox can be explained by a fact already reported for small experimental systems [16]: only moderate dissociation of the precursor is sufficient for a reasonable deposition rate. Extensive dissociation of the precursor leads to the formation of various radicals that do not stick to the sample surface but are pumped out from the system; therefore, in cases where large power densities are used for sustaining plasma in HMDSO, the deposit does not resemble PDMS but rather silica. Detailed study of the transition from polymer-like films to films rich in silicon oxides was reported in [16]. The power density used in this study was at least 10 times lower than

the power density needed for such full transition; however, there are still mild transitions, towards films richer in silicon, that can affect the deposition rates seen in Figure 5.

**Figure 4.** The intensity of the Hα and CH lines at 656 nm and 431 nm as a function of the discharge power at 130 sccm HMDSO.

**Figure 5.** The deposition rate versus the discharge power in the empty reactor.

Both Figures 5 and 6 indicate large differences in the deposition rate at different locations ranging from 1.6 to 14.7 nm min–1. The deposition rate is the largest for sensor S1. This sensor was placed on the grid between the electrodes, as shown in Figures 1 and 2. The highest deposition rate is on the surface, where it is not needed because the radicals at the position of S1 are likely to be pumped away from the system. The high deposition rate indicates a high density of radicals that are capable of forming the protective coating. According to the state-of-the-art, such radicals are partially dissociated HMDSO molecules, including those found at the mass of 148 Da [19]. In the empty chamber, these radicals are denser or more concentrated at the position near the pump ducts than anywhere else in the system, as revealed in Figures 5 and 6.

Figure 6 shows the thickness of the coating obtained from the sensors' signals versus the treatment time for the empty plasma reactor. One can observe almost perfectly linear behavior, which indicates excellent stability of plasma parameters during the deposition of the protective coatings. The stability may be a consequence of the appropriately low pressure in the reactor, which prohibits instabilities that may appear because of the cluster formation [27] and thus the loss of radicals useful for the deposition of the protective coating.

**Figure 6.** The thickness of the deposited films derived from the sensors' signals (points) with linear fits (lines) versus the plasma treatment time in an empty reactor at a power of 4 kW at 130 sccm HMDSO. In the inset figure, a deposition rate is presented with the height of the column at a sensor position.

Examining Figure 5 and compared to Figure 1, one observes the next largest deposition rate at sensors S2 and S8, which were located a bit farther from the pump ducts. In fact, sensors S2 and S8 were located between the gas inlet and the powered electrodes, as shown in Figure 1. The possible reasons for favored deposition rate at these positions will be discussed later in this report.

The deposition rates at the position of sensors far from the electrodes are lower but still reasonably high. For example, Figure 5 reveals the deposition rates of about 6 nm min–1 for the sensors S4, S5, and S6. Conversely, sensors S3 and S7, which were placed close to the gas inlet but away from the powered electrodes, show a poor deposition of approximately 2 nm min–1.

The distribution of the deposition rate in the plasma reactor provides a qualitative model of the gas kinetics that allows the most reasonable degree of fragmentation of the precursor molecules. The injected HMDSO molecules do not interact with the solid materials but should be partially dissociated to radicals with a reasonable sticking coefficient. The plasma density far from the powered electrodes in the reactor used for these experiments is only on the order of 1014 m–3 [23]. Such a low density of electrons does not enable immediate dissociation to useful fragments. This may explain the poor deposition rates detected by sensors S3 and S7, located close to the gas inlet but away from the powered electrodes. The molecules should be allowed a prolonged residence time in the weakly ionized gaseous plasma to dissociate into useful radicals. The residence time will be estimated later in this paper. The injected precursor molecules enter the plasma reactor with a significant drift velocity but quickly thermalize (assume the random motion after a few elastic collisions). The motion is then governed by diffusion, i.e., it is random. The molecules suffer numerous collisions with plasma electrons while diffusing from the source (gas inlet) to the position of the sensors S4, S5, and S6. The gas at the position of these sensors is thus reasonably well dissociated, which favors the deposition on the surfaces far away from the electrodes. As mentioned above, the residence time of the

injected molecules is too short to cause significant deposition at the positions of sensors S3 and S7.

Sensors S2 and S8 are as close to the gas inlet as S3 and S7, but Figure 5 indicates a deposition rate several times higher at S2 and S8 compared to S3 or S7. This paradox may be explained by the larger residence time of molecules striking the surface of the sensors at positions S2 and S8, but the variation of the plasma density versus the distance from the powered electrode may be more important. The asymmetric capacitively coupled RF discharge is characterized by an oscillating sheath next to the powered electrode. Since the frequency of these oscillations is rather low (the RF generator operates at 40 kHz), the electrons oscillate within the sheath and gain energy enough for a rather extensive dissociation and ionization of the gaseous molecules within the oscillating sheath [28]. Therefore, the dissociation of the precursor molecules is more extensive next to the electrodes than in the bulk plasma far away from the powered electrodes. As a result, the deposition rate at the sensors S2 and S8 is favorable despite the proximity of the gas inlet.

The radicals stick to surfaces of any material facing plasma; therefore, the deposition rate as determined by the sensors located in the reactor according to Figure 1 should be lower if the reactor is additionally loaded with samples. To study the influence of samples on the deposition rate, samples were mounted on the planetaria, as shown in Figure 7. About 250 medium-sized, approximately 40-cm-long samples, which represented about 100% of the total chamber capacity, were evenly distributed inside the chamber. The height and the diameter of the planetaria were 160 cm and 55 cm, respectively, and the distance between axles was around 60 cm. The planetaria were spinning at a speed of 6 rpm. The deposition rate measurements were repeated with sensors located at the same positions as in the empty chamber. The results are shown in Figure 8. The highest deposition rate was observed for the sensors S2 and S8. These sensors are located between the gas inlet and the powered electrode (Figure 1). The deposition rate at the positions S2 and S8 are about an order of magnitude greater than at any other position except near the pump ducts. The presence of samples in the plasma reactor, therefore, influences the deposition rate significantly. Not only is it lower than in the empty reactor (compare Figures 6 and 9), but a reasonably large deposition rate is observed only in the region close to the electrodes (S2, S8, and S1). Elsewhere, the deposition rate is below 1 nm min<sup>−</sup>1.

**Figure 7.** A photo of the fully loaded chamber with samples mounted on planetaria.

**Figure 8.** The deposition rate versus the discharge power in reactor loaded with samples.

**Figure 9.** The deposition rate versus the discharge power in reactor loaded with samples.

The very low deposition rate at S4, S5, and S6, as observed in Figure 8, is explained by the loss of radicals on the surfaces of the samples. As discussed above, the plasma density away from the electrodes is low, so the loss of radicals useful for depositing protective coating cannot be balanced by production because of electron-impact dissociation. Conversely, the deposition rate close to the powered electrode (sensors S2 and S8) remains reasonably high because of the higher electron energy in the oscillating sheath.

The ratio between the deposition rate in an empty reactor and a full reactor is shown in Figure 9. The highest ratio of 10–20 is observed for sensors positioned far from the electrodes. This observation was already explained by the loss of radicals on the surface of the samples. However, the ratio is much lower for the sensors positioned close to the powered electrodes. For sensors S2 and S8, the ratio is approximately 3 for the lowest power of 1 kW and only 2 for the highest power of 7 kW. The power-dependence of the ratio is explained by the fact that the electron energy in the vicinity of the powered electrodes is much higher than far from the electrodes, so a significant fraction of injected HMDSO molecules get dissociated and thus contribute to the film growth.

The upper discussion reveals the crucial role of the residence time of molecules in the plasma reactor. Gaseous molecules diffuse in the plasma reactor because the random velocity is much higher than the drifting from the gas inlet to the pump ducts. The drift velocity of gaseous molecules at the entrance to the pump ducts can be calculated if the effective pumping speed at that position is known. The effective pumping speed depends on the nominal pumping speed of the roots pumps and the conductivity of any vacuum elements mounted between the roots pumps and the plasma reactor. The conductivity is difficult to determine, but one can also determine the effective pumping speed from the measured gas flow and pressure inside the reactor by considering the constant mass flow:

$$p\_1 \ S\_1 = p\_2 \ S\_2. \tag{1}$$

Here, *p*<sup>1</sup> is the atmospheric pressure, *S*<sup>1</sup> is the gas flow as measured by the flow controller, *p*<sup>2</sup> is the measured pressure in the plasma reactor, and *S*<sup>2</sup> is the effective pumping speed at the grid which separates the plasma reactor and the pump ducts. Taking into account the measured values, i.e., *<sup>p</sup>*<sup>1</sup> = 105 Pa, *<sup>S</sup>*<sup>1</sup> = 130 cm3/min = 2×10−<sup>6</sup> m3 s–1, *<sup>p</sup>*<sup>2</sup> = 4 Pa, one can estimate the effective pumping speed as:

$$S\_2 = p\_1 S\_1 / p\_2 = 0.05 \text{ m}^3 \text{ s}^{-1} = 180 \text{ m}^3 \text{ h}^{-1}.\tag{2}$$

As calculated from Equation (1), the effective pumping speed is an order of magnitude lower than the nominal pumping speed of the roots pumps. This observation may be explained by the deviation of the real pumping speed of the roots pumps from the nominal value (the latter is just the maximum pumping speed at optimal conditions) and the limited conductivity of vacuum elements mounted between the plasma reactor and the roots pumps.

There is a negligible pressure gradient throughout the plasma reactor, because the conductivity is orders of magnitude greater than the effective pumping speed. The crosssection of the plasma reactor is a product of the reactor diameter and height, i.e., *A* = 3.5 m2. The gas drift velocity from the source to the pump ducts is:

$$
\upsilon = S\_2/A = 0.014 \text{ m s}^{-1}.\tag{3}
$$

This value is orders of magnitude lower than the random velocity due to the thermal motion of the molecules, which is:

$$
\overline{\upsilon} = \sqrt{\frac{8kT}{\pi m}} = 200 \text{ m s}^{-1}.\tag{4}
$$

In Equation (4), we considered the room temperature (*T* = 300 K) and the HMDSO mass *m* = 162 Da. By considering the distance between the gas inlet and the grid separating the reactor from the pump ducts of *l* = 1 m, one can estimate the average residence time of gaseous molecules as:

$$
\pi = l/\upsilon = 80 \text{ s}.\tag{5}
$$

The residence time as calculated from Equation (5) is an averaged value taking into consideration the simple calculations. Because the random velocity as calculated from Equation (4) is orders of magnitude higher than the drift velocity as determined from Equation (3), the residence time is spread broadly from the value calculated using Equation (5), and thus it should be taken just as an estimation. In any case, the residence time is long enough to assure for numerous collisions with plasma electrons. The large residence time is the reason for the rather large deposition rate at any position far from the gas inlet in the empty reactor. The maximal deposition is observed on the grid near the pump ducts (sensor S1) in the empty reactor. The radicals entering the pump ducts are likely to have been created well before reaching the grid.

Plasma reactors are useful only when the coatings are deposited on various products mounted on the planetaria. Technologically relevant results are presented in Figure 8. The deposition rate at sensor S1 (mounted on the grid near the pump ducts) is moderate at

about 2 nm min–1, which is favorable from the technological point of view. Still, a significant fraction of the radicals useful for the thin film deposition is pumped out from the reactor. However, the major deficiency of the plasma reactor is the poor deposition rate at any other position. Despite the long residence time of gaseous radicals, the deposition rate is poor because of the loss of radicals on the samples placed on the planetaria. The only useful part of the reactor, when loaded with samples, is at positions S2 and S8, so close to the powered electrodes. The discharge configuration in this reactor is, therefore, inadequate. The configuration with electrodes placed opposite to the pump duct should be better.

No sensor was placed on a powered electrode because it would heat significantly. Still, according to the measured deposition rates and according to the above discussion, it is reasonable to assume the large deposition rate on the powered electrodes. In fact, the electrodes should occasionally be etched in chemical baths to remove the excessive deposits. The extensive deposition of thin films on the electrodes and thus loss of radicals for coating the samples is a major drawback of the reactor used in this study. The problem could be minimized using symmetric discharge, but it is often not feasible as in our PECVD reactor.

Despite the large dissipation of the deposition rate, the composition of the deposited films remains similar for all films at the positions of different sensors. Figure 10 represents the composition of the films as deduced from XPS survey spectra. The measurements were performed in the reactor loaded with samples. The concentration of carbon is close to 50 at.%, while the concentrations of oxygen and silicon is between 25 and 30 at.% for all samples. The small variations in the composition may be attributed to the accuracy of the XPS technique or to actual variation in the composition, but because the differences are marginal it is possible to conclude that the stoichiometry of the deposited films does not vary significantly between different positions in the plasma reactor.

**Figure 10.** Concentrations of carbon, oxygen and silicon of deposited films at different positions in the PECVD reactor as deduced from XPS survey spectra.

#### **4. Conclusions**

Many commercial plasma reactors for the deposition of thin films from organic precursors using the PECVD technique suffer from non-uniform deposition rates. Moving the products to be coated by placing them on planetaria enables reasonable coating uniformity, but the efficiency is poor, because a significant fraction of the precursor radicals used as building blocks of the protective coatings are lost by adsorption on the powered electrodes and/or by pumping out from the reactor. An attempt was made to measure the deposition rates at various locations inside an industrial reactor powered by a capacitively coupled RF discharge. The plasma reactor had a volume of approximately 5 m3. The maximum deposition rate for an empty reactor was measured on a grid near the pump ducts. The next

highest rates were measured close to the powered electrodes, but a reasonable deposition rate was also observed far from the powered electrodes or the pump duct. The observation was interpreted by the formation of radicals useful for the deposition of the thin films throughout the reactor. The average residence time of approximately 80 s ensured a reasonably large production rate, despite the very low electron density in the plasma away from the oscillating sheaths next to the powered electrodes. Loading the reactor with numerous samples caused a significant difference in the deposition rates. Not only were they lower, but the distribution changed significantly. The deposition rates far from the powered electrodes dropped by more than an order of magnitude for a fully loaded chamber. Deposition rates above about 1 nm min–1 were only observed close to the powered electrodes. These observations indicate the need for modification of the discharge configuration in the industrial plasma reactor for depositing protective coatings from HMDSO precursor using the PECVD technique.

**Author Contributions:** Conceptualization, R.Z., D.Ð. and S.P.; methodology, R.Z., Ž.G. and D.Ð.; software, G.P. and D.Ð.; validation, D.Ð., B.G. and A.V.; formal analysis, A.V., Ž.G. and B.G.; investigation, R.Z., G.P, and Ž.G.; resources, Ž.G. and D.Ð.; data curation, M.M., S.P. and B.G.; writing—original draft preparation, M.M.; writing—review and editing, A.V.; supervision, M.M.; project administration, G.P.; funding acquisition, D.Ð. and M.M; All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Slovenian Research Agency, grant number L2-1835 (Innovative sensors for real-time monitoring of deposition rates in plasma-enhanced chemical vapor deposition (PECVD) systems) and research core funding grant number P2-0082 (Thin-film structures and plasma surface engineering).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland www.mdpi.com

*Coatings* Editorial Office E-mail: coatings@mdpi.com www.mdpi.com/journal/coatings

Disclaimer/Publisher's Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Academic Open Access Publishing

www.mdpi.com ISBN 978-3-0365-8621-2