Corrosion of stainless steels and corrosion protection strategies in the semiconductor manufacturing industry: a review

2.1.1 Moisture adsorption on material surfaces

When the equipment is in contact with the atmosphere, moisture, impurities, and gas molecules are adsorbed on its surface (Figure 1) (Saito 1997). Before the semiconductor process, the chambers and pipelines are pumped down to ensure a vacuum inside the equipment. The pumping process can be divided into the following stages: (1) A vacuum pump expels the air inside the equipment. (2) Some gases are adsorbed on the surface at atmospheric pressure and then slowly desorbed under vacuum. H₂O dominates the process, and the desorption rate depends on various gases’ surface binding energy, temperature, and surface coverage. (3) Some gases are dissolved in the material during manufacturing and slowly diffused in a vacuum. The process is mainly H₂, and the emission rate depends on material properties, temperature, gas solubility and diffusion rate. (4) Extremely small amounts of gas can penetrate the vacuum chamber walls and enter the chamber. The emission rate depends on the material’s properties and the chamber walls’ thickness (Weston 1975).

Figure 1:

Vacuum systems of semiconductor manufacturing industry. (a) Percentage of vacuum system gas types; (b) sequence of pumping.

Figure 1 shows that moisture is a major component of pumping down and is not easily dislodged by desorption on the material’s surface. Most high-purity corrosion gases (such as HCl, HBr, etc.) must overcome the bond energy to react with stainless steel; therefore, the corrosion rate is small. However, when the moisture adsorption is large on the stainless steel surface, some corrosion gases dissolve, and electrochemical corrosion occurs (Fine et al. 1995; Lequien and Moine 2021). Therefore, moisture can significantly contribute to stainless steel corrosion. Therefore, moisture adsorption and desorption on the material surface is a condition that must be considered in the gas delivery system.

The moisture adsorption can be physical and chemical. Physical adsorption occurs under van der Waals forces between the molecules’ surface, while chemical adsorption involves the chemical bonding of molecules to surface atoms. When sufficient energy is obtained, the adsorbed molecule desorbs under vacuum, called the adsorption energy. Under ideal conditions, the adsorption energy (desorption activation energy) is a single value. However, the adsorption energy is not a single value since the adsorption state is complex due to material surface roughness, surface chemistry, etc. (Figure 2). Molecules with adsorption energy below 100 kJ/mol desorb rapidly at room temperature (Saito 1997).

Figure 2:

Moisture adsorption model: (a) ideal surface, (b) actual surface.

The moisture adsorption capacity is affected by various factors, including partial pressure, surface properties of the specimen and the gas flow rate containing moisture. Siefering and Whitlock (1994) reported that the moisture adsorption capacity is significantly related to the moisture partial pressure. Under constant ambient pressure, the partial pressure of moisture changes by varying the moisture concentration, and experimental results are shown in Figure 3.

Figure 3:

Isobaric adsorption lines for moisture molecules at different temperatures and partial pressures (Siefering and Whitlock 1994). Reprinted with permission; copyright 1994 American Vacuum Society.

More moisture is absorbed on the material’s surface when the partial pressure of moisture is high and the temperature is lower. The surface moisture coverage of the material was 2.2 × 10¹⁴ mol/cm² at a moisture concentration of 357 ppb and a temperature of 273.15 K (Figure 3). At this point, it is sub-monolayer adsorption, where multiple adsorption layers exist on stainless steel surfaces (EP) at higher partial pressures. For example, seven monolayers of moisture are formed on stainless steel surfaces at a moisture concentration of 100 ppm in argon (Siefering and Whitlock 1994).

Siefering proposed a semi-empirical model of the Langmuir relationship to fit the results obtained with the following equation.

where PH2O is the atmospheric partial pressure of moisture, γ1, γ2, γ3, γ4 are positive fitting constants, γ5, γ6 are fitting constants, and T is the absolute temperature. The fitted curves enable extrapolation of surface adsorption at different temperatures and different partial pressures of the moisture, as shown in Figure 4.

Figure 4:

Semi-empirical model fitting lines for moisture molecules at different temperatures and partial pressures (Siefering and Whitlock 1994). Reprinted with permission; copyright 1994 American Vacuum Society.

Surface properties, including roughness, treatment processes and chemistry, significantly impact moisture adsorption. Dylla et al. (1993) found that the gas emission rate increased with the surface roughness factor (the ratio of true surface area to geometric surface area), i.e. the content of adsorbed moisture molecules increased, attributed to the increased number of adsorption sites on moisture molecules.

Tsuji et al. (2017) described the moisture adsorption process on the pipe’s inner wall and calculated the amount of moisture adsorption by integral calculation. Figure 5 shows a delay time between the moisture-injected and outflowed sides. This is the time required for the moisture adsorbed on a specimen surface to reach saturation. The amount of moisture adsorption is related to the surface state of different materials during this time interval.

Figure 5:

Schematic diagram of the moisture delay adsorption process (Tsuji et al. 2017). Reprinted with permission; copyright 2017 Japan Society of Applied Physics.

After different surface treatments, the initial phase exhibits the same adsorption rate, as shown in Figure 6. The final stage of moisture adsorption significantly differs due to the surface properties, i.e. the smoother surface properties absorb less moisture.

Figure 6:

Moisture adsorption capacity of different surfaces (Tsuji et al. 2017). Reprinted with permission; copyright 2017 Japan Society of Applied Physics.

2.1.2 Moisture condensation in the corrosive gases

When the stainless steel surface is exposed to atmospheric or residual moisture from various valves, moisture adsorption is inevitable. Moisture condensation occurs under certain conditions on the surface as moisture adsorption reaches a critical value.

There is a correlation between the moisture molecules’ condensation on the surface and the dew point temperature of the gas mixture (Tsuji et al. 2017). Moisture condensation occurs once the dew point temperature is above the ambient temperature. Conversely, the moisture molecules remain in a gaseous state. The dew point temperature is calculated from the moisture partial pressure by the following equation (Kiang 1981);

where PH2O is the moisture partial pressure (mmHg), and T is the dew point (K).

When the dew point temperature of the gas mixture drops, moisture condenses on the surface after mixing with HCl, HBr, HF, etc. The equation to calculate the dew point temperature for different gases is as follows (Kiang 1981);

where T is the dew point temperature (K), PH2O is the partial pressure (mmHg) of the moisture, PHCl, PHF and PHBr are partial pressures (mmHg) of HCl, HF and HBr, respectively.

For ideal gaseous conditions, the partial pressure and concentration are converted. Therefore, plotting moisture concentration versus dew point temperature for different gaseous environments is feasible.

Significant differences exist between dew point temperatures of different gases at ambient temperature. Corrosive gases can significantly increase the dew point temperature of a gas mixture. The higher the dew point temperature, the greater the condensation tendency for moisture. HBr, HCl, and HF gases can significantly increase the condensation tendency of moisture, with HBr being the greatest (Figure 7).

Figure 7:

Dew point temperatures of different gas mixtures at an total pressure of 1 atm and moisture partial pressure of 0.01 atm.

In the gas delivery system, there are cases of replacement cylinders, maintenance pipelines and damaged valves, resulting in contact between the inner surface of the gas pipelines and the atmosphere. The moisture in the gas condenses on the material surface, when the temperature is different between the material surface and the gas mixture. Figure 8 depicts that a tiny amount of moisture is mixed with HBr upon introducing high-pressure HBr, forming a high-propensity gas mixture for moisture condensation, resulting in secondary condensation in the gas delivery system.

Figure 8:

The adsorption behaviour of moisture in gas delivery system.

2.1.3 Corrosion mechanism of moisture interaction with stainless steel

Lange’s Handbook of chemistry suggests that corrosion gases have extremely large solubility in moisture. Corrosive gases have a high propensity to dissolve in moisture (Fine et al. 1995). Many corrosive gases are dissolved in the condensate moisture, forming a supersaturated acidic solution, generating electrochemical effects to accelerate the corrosion reaction. Thus, the corrosion products in the form of particles contaminate high-purity gas, reducing the wafer quality.

Taking an H₂O and HCl mixture as an example, water molecules are adsorbed on the stainless steel surface as a monolayer. The water molecules condense when the dew point temperature is higher than the ambient temperature, gradually forming a precursor film (Figure 9a). As moisture condensation proceeds, droplets tend to form at depressions, defects or inclusions (Tsuji et al. 2017). Due to the high affinity of HCl gas for water, it dissolves into droplets, forming a supersaturated acidic and a chloride ion concentrated electrolyte (Figure 9b), inducing a rapidly developing droplet corrosion (Figure 9c). In addition, a certain amount of gas is dissolved into the precursor film, causing certain corrosion on the material’s surface. The droplet volume gradually increases as the water molecules condense, while the film thickens, prompting the metal ions dissolution and accelerating the corrosion rate (Figure 9d). Under the influence of gravity and gaseous fluids, condensation droplets move and stagnate along the specimen, thus, accelerating the corrosion reactions. The high-purity gas is contaminated. When HCl concentration in the condensed droplets increases, the surface tension gradually decreases, flattening the condensed droplets. In the presence of high water content, it may continuously concentrate the solutes in the precursor film on the material surface over time, gradually evolving into a thin liquid film morphology (Figure 9f) (Askey et al. 1993; Lequien and Moine 2021; Mizushima 2007; Weissenborn and Pugh 1996).

Figure 9:

Corrosion mechanism of H₂O and HCl gas mixtures.

Lequien and Moine (2021) proposed that the corrosion morphology can be significantly altered by intermingling moisture with the HCl gas. A distinct droplet corrosion pattern characterises the initial stage of corrosion, and traces of micro-droplet formation can be observed. The droplet corrosion morphology disappears and transforms into a liquid membrane-covered corrosion morphology (Figure 10). The above corrosion mechanism was verified.

Figure 10:

Corrosion morphology of 316L (EP) with a mixture of H₂O and HCl: (a) 720 h, (b) 1848 h, (c) 4224 h (Lequien and Moine 2021). Reprinted with permission; copyright 2021 John Wiley and Sons.

Tomari (1997) dried SUS316L (EP) in the air to adsorb sufficient moisture to the specimen surface and exposed it to HCl gas. Granular deposits development, observed by SEM, revealed that higher atmosphere temperatures resulted in more serious corrosion. SUS316L is a stainless steel for semiconductors produced following the standard of semiconductor manufacturing industry (SEMI F20).

The corrosion of stainless steel is linked to the moisture concentration in the gas mixture (Fine et al. 1995; Smudde et al. 1995). As shown in Figure 11, in H₂O and HBr mixture, the corrosion rate increases exponentially when the moisture concentration exceeds a critical value (10 ppm). The typical corrosion morphology is the presence of a thin layer of granular deposits in the central part, surrounded by a circular area of 5–10 μm, consistent with the shape of droplet corrosion. EDS analysis determined the deposited crystals in the central part of the feature to be FeBr₃-χH₂O, and elemental bromine (Br) was detected in the corrosion pit wall below the crystals.

Figure 11:

Corrosion morphology of 316L (EP) with hydrogen bromide gas containing different moisture concentrations: (a) 0.2 ppm, (b) 10 ppm, (c) 100 ppm, (d) 1000 ppm, (e) enlarged view of area λ (Fine et al. 1995). Reprinted with permission; copyright 1995 IOP Publishing.

2.1.4 Corrosion mechanism in welded joints of gas pipelines

Welding is inevitably used to connect gas pipelines in gas delivery systems. However, welded joints are vulnerable to corrosion, so clarifying their corrosion mechanisms in the semiconductor manufacturing industry is necessary. The austenitic stainless steel weld zone metal (WM) is a cast iron structure (Cui and Lundin 2005) consisting of an austenitic with 2–10 % δ-ferrite. The welded joints are divided into weld, fusion and heat-affected zones (HAZ), each with different characteristics, as shown in Figure 12 (He and Xing 2019).

Figure 12:

Microstructure of welded joints by TIG: (a) heat affected zone, (b) weld (He and Xing 2019). Reprinted with permission; copyright 2019 IOP Publishing.

Cui and Lundin (2007) deposited 316 on the base metal 304 by shielded metal arc welding (SMAW). Subsequently, a 3 % FeCl₃ solution was used for immersion testing. The initial corrosion attack is in austenite instead of ferrite. Cleslak et al. (1982) investigated the chromium (Cr) and nickel (Ni) segregations within dendrites in austenitic weld metal, suggesting that solidification produces Cr and Ni-depleted zones in the crystal core, thus, reducing the corrosion resistance of austenite.

In the welded joint, the HAZ is an obvious weak corrosion site due to the sensitization of austenitic stainless (Lu et al. 2020; Trigwell and Selvaduray 2005). The sensitisation of weldments refers to the rapid carbon diffusion to grain boundaries in the temperature range of 450 and 850 °C, precipitating the Cr-rich carbides (M₂₃C₆) near the grain boundaries (Kumar et al. 2015). Cr concentration reduces to well below 12 wt% due to the precipitation of M₂₃C₆ near the grain boundaries, known as the Cr-depleted zones. At this point, the corrosion resistance at the grain boundary is reduced and prone to intergranular corrosion (Kumar et al. 2015). The susceptibility to intergranular corrosion depends on the Cr concentration variation and the width of the Cr-depleted zone (Zhang et al. 2013).

In addition, the molten zone shrinks during the cooling and solidification of pipe welding, and the base material adjacent to HAZ produces residual tensile stresses to resist deformation, further increasing the corrosion susceptibility of the heat-affected zone.

Smudde et al. (1995) and Fine et al. (1995) described that the corrosion rate of 316L (EP) welded specimens increased with moisture content in HBr gas through gas mixture exposure experiments, as shown in Figure 13.

Figure 13:

Corrosion morphology of 316L (EP) welded specimens: (a) 0.5 ppm, (b) 10 ppm, (c) 100 ppm, (d) 1000 ppm (Fine et al. 1995). Reprinted with permission; copyright 1995 IOP Publishing.

The heat tint appears after welding stainless steel in the joint area, and its colouring depends on the welding conditions (Łabanowski and Głowacka 2011). After welding, a thicker oxide layer is formed on the surface with colour variation due to the light interference effect. During the oxide layers formation, the Cr content in the material decreases, thus, reducing the corrosion resistance of the welded joint (Fuertes et al. 2017). Removing the heat tint in the semiconductor manufacturing industry is necessary. The removal of heat tint can be divided into three stages: (1) Removal of oxide film by mechanical grinding, (2) the removal of Cr-depleted surface via chemical cleaning or electrolytic polishing, and (3) the improvement in corrosion resistance by passivation, found in ASTM A380/A380M and ASTM A967 (Łabanowski and Głowacka 2011).

When the thermal source acts on the weld material surface, it experiences a sharp heat-up. Without considering the vaporization of elements in the melt pool, the stainless steel melt pool soon boils (Zacharia et al. 1991), as shown in Table 3. However, the cooling effect accompanying evaporation limits the temperature of the melt pool; therefore, the welded joint exhibits partial vaporization and uneven distribution of elements.

Block-Bolten and Eagar (1984) suggested that the higher percentage of manganese (Mn) in the stainless steel results in faster vapourization, with Mn being the main element in the melt pool. Figure 14 illustrates that Mn has the highest vapour pressure from a thermodynamic viewpoint, and vaporization occurs preferentially. The composition of melt pool vapour is examined. In the initial melting stage, the Mn element dominates the gas phase composition, and after a sustained period, the Fe element gradually rises.

Figure 14:

Vaporization of Mn elements in the process of welding. (a) Curves of temperature versus vapour pressure for different elements, (b) curves of melt pool vapour composition versus time (Block-Bolten and Eagar 1984). Reprinted with permission; copyright 1984 Springer.

Trigwell and Selvaduray (2005) found a blue–brown discolouration in the HAZ region associated with the Mn deposition near the weld after evaporation. In 316L stainless steel, the vapour pressure of Mn is higher than other elements, and it tends to evaporate during the welding process and is subsequently deposited in the HAZ adjacent to the weld, accelerating the corrosion.

Hattori et al. (1994) reported Mn deposits downstream of the weld centre, resulting from post-vapour deposition of Mn during welding. Subsequently, experiments were performed on stainless steel welded joints with different Mn contents. They also proposed a mechanism to deposit Mn after vapourization to accelerate corrosion. Figure 15 shows a surge of Mn content downstream of a high Mn stainless steel weld, where the peak was at 5 mm. Mn coexists with iron (Fe), Cr and Ni oxides in the form of MnO₂. MnO₂ results from oxidizing extremely fine Mn particles by residual moisture or oxygen on the surface. The corrosion rate is further accelerated by H₂O, produced by the MnO₂ reaction with HBr.

Figure 15:

Elemental composition of the upper and downstream surfaces of high manganese stainless steel after welding (Hattori et al. 1994). Reprinted with permission; copyright 1994 Japan Society of Applied Physics.

2.2.1 Thermal ageing behaviour of stainless steel

Nowadays, the austenitic stainless steel manufacturing technology is excellent. However, owing to the limitations of non-equilibrium rapid solidification of the manufacturing process (Hong et al. 2002), a slight amount of ferrite is inevitable in austenitic stainless steels (Elmer et al. 1989; Li et al. 2021a,b; Warren et al. 2015). Ignoring these ferrites in the semiconductor manufacturing industry with ultra-high purity requirements is not recommended, even if the ferrite content is extremely low. Moreover, austenitic stainless steel welds have duplex microstructures consisting of an austenite substrate with approximately 5–12 % ferrite (Chandra et al. 2012). Ferrite in welds can effectively avoid thermal cracking (Chandra et al. 2012; Gill et al. 1989; Lee et al. 2018). However, when stainless steel is under long-term service at 280–330 °C, ferrite undergoes structural transformation, increasing the strength and hardness, decreasing the plasticity and toughness, and thermal ageing embrittlement (Lin et al. 2019; Sahu et al. 2009). Microstructural changes due to thermal ageing can facilitate corrosion (Was et al. 2011).

Microstructural changes in thermal ageing at 280–330 °C are characterized by unstable ferrite decomposition (spinodal decomposition) and the G phase precipitation (Lee et al. 2018). The unstable decomposition of ferrite is characterised by Fe-rich (usually noted as α) and Cr-rich (usually noted as α′) zones. BCC structure’s iron-rich and Cr-rich phases have similar lattice constants (Chandra et al. 2012). The difference in Cr content between α and α′ phases increases with thermal ageing. The α-phase contains a small amount of Cr and is susceptible to corrosion. Figure 16 illustrates the unstable decomposition morphology of ferrite with an orange peel appearance (Jeong et al. 2022). The G phase is a intermetallic nickel silicide that correlates well with Ni. The G phase precipitates in the ferrite phase during thermal ageing, with a lattice constant four times that of ferrite, as FCC intermetallic silicides. Figure 17 shows the G phase precipitation morphology and composition (David et al. 1996; Kong et al. 2023). The G-phase is preferentially nucleated at α and α′ phase boundaries and acts synergistically with ferrite spinodal decomposition. Ni are repelled by α and α′ regions, diffusing and aggregating in the middle of α and α′ phases, forming clusters, precipitating into G-phase (Jeong et al. 2022; Kong et al. 2023).

Figure 16:

TEM analysis results of ferrite in the 316L: (a and b) untreated, (c and d) ageing at 400 °C for 5000 h, (e and f) ageing at 400 °C for 10,000 h, (g and h) ageing at 400 °C for 20,000 h (Jeong et al. 2022). Reprinted with permission; copyright 2022 Elsevier.

Figure 17:

TEM analysis results for ferrite at 400 °C for 20,000 h: (a) HRTEM image, (b) resultant FFT taken in (a), (c) STEM/EDS mapping, (d) compositional line scan profile (Kong et al. 2023) (no permission is required).

Since the sluggish thermal ageing rate, behaviour under service conditions is studied by accelerated methods. The materials’ structural transformation (thermal ageing mechanism) occurring under service and accelerated thermal ageing temperatures are consistent (Lin et al. 2019; Pumphrey and Akhurst 1990). Chung and Chopra (1986) demonstrated the effectiveness of accelerated thermal ageing, which is generally considered a thermally activated process in stainless steel and obeys the Arrhenius Formula. Temperature and time for accelerated thermal ageing are determined in the laboratory to simulate the thermal ageing behaviour at service temperature. Although the activation energy is a function of temperature, the thermal ageing activation energy is approximated as a constant over a suitable temperature (Lin et al. 2019).

The structure variation due to spinodal decomposition possesses the highest embrittlement rate at 475 °C, known as the 475 °C embrittlement phenomenon (Sahu et al. 2009). The thermal ageing characteristics of structure variation are transformed when the temperature is higher than 475 °C.

Thermal ageing structure transformation of stainless steel at 500–850 °C is characterized by carbide precipitation (Trillo 1997). Carbon remains in austenite in a supersaturated state, and when heated to an appropriate temperature for a sufficient period, this supersaturated carbon element precipitates as a carbide. As shown in Figure 18, there is a decrease in Cr content at grain boundaries. Ding et al. (2009) described that precipitate phases at grain boundaries are chromium-rich carbides (M₂₃C₆) by thermodynamic calculations and TEM.

Figure 18:

Element distribution of stainless at 650 °C under different ageing time: (a, b) 0 h, (c, d) 1 h, (e, f) 5 h, (g, h) 48 h (Wang et al. 2021). Reprinted with permission; copyright 2021 Elsevier.

The temperature range for M₂₃C₆ precipitation during thermal ageing is 500 °C to 850 °C (Trillo 1997). In addition, the precipitation kinetics of M₂₃C₆ is associated with the stainless steel chemical composition and processing histories (Allahyari et al. 2005; Ding et al. 2009; Li 2014; Li et al. 2021a,b). Figure 19 shows that the corrosion resistance of stainless steel is significantly reduced after thermal ageing, and the longer time result in worse corrosion resistance (Wang et al. 2021).

Figure 19:

Corrosion morphology of specimens immersed in 1 M hydrochloric acid for 7 days after the different thermal ageing time at 650 °C: (a) 0 h, (b) 1 h, (c) 5 h, (d) 48 h (Wang et al. 2021). Reprinted with permission; copyright 2021 Elsevier.

The precipitation time for carbides varies at different temperatures. The carbide precipitation times of stainless steels at 500 °C to 850 °C are summarized in the literature (Allahyari et al. 2005; Ding et al. 2009; Li 2014; Li et al. 2021a,b). The higher the temperature, the more rapid the diffusion and the shorter the time for carbide formation, as shown in Figure 20.

Figure 20:

Statistics and fitting of carbide precipitation times for stainless steels at different temperatures (Allahyari et al. 2005; Ding et al. 2009; Li 2014; Li et al. 2021a,b).

2.2.2 Vaporisation behaviour of corrosion products at high temperatures

The reaction mechanism between corrosive gases and metals in a high-temperature vacuum environment significantly differs from that in a conventional atmospheric environment. Under conventional corrosive environments, corrosive species dissolve in water to produce ionic conductors, resulting in electrochemical corrosion. However, under a high-temperature vacuum environment, water exists in trace amounts of gas phase molecules, depicting a weak accelerating effect on corrosion, and the overall reaction is predominantly chemical. Taking the corrosion of stainless steel in HCl gas as an example, the reactions are as follows:

Metal halides are extremely volatile, and their vapourization at high temperatures and low pressures must be considered (Speight 2005).

Ihara et al. (1981, 1982a, 1983 investigated the corrosion behaviour of metals in pure HCl gas at high temperatures and 1 atm in a confined stationary space. Figure 21 indicates that the corrosion behaviour of Fe/Cr/Ni in HCl gas at high temperatures is determined by the formation and sublimation of ferrous chloride. Below a certain temperature, corrosion products are constantly attached to metal surfaces and do not undergo vaporisation. When the temperature exceeds a certain level, corrosion products undergo vaporization. The gaseous reaction products leave the surface, and almost no corrosion product adheres to the specimen surface. The middle-temperature range is a transition zone; some corrosion products undergo vaporization, while others adhere to the surface. The reaction of Cr with HCl at 800 °C undergoes complex weight changes. It is speculated that gaseous CrCl₃ is supersaturated in the chamber during the final reaction stage, resulting in redeposition on the specimen surface.

Figure 21:

Curves for thermogravimetric experiments in a confined stationary space: (a, b) pure Fe, (c) pure Ni, (d) pure Cr (Ihara et al. 1981, 1982a, 1983). Reprinted with permission; copyright 1981, 1982a, 1983 Elsevier.

However, the interaction between Fe, Ni and Cr must be considered for alloys such as stainless steel. When the percentage of Ni content in the alloy increases, it could significantly inhibit the product vaporisation of the alloy (Ihara et al. 1982b), as shown in Figure 22. Therefore, elements in stainless steel influence the corrosion product vaporisation.

Figure 22:

Curves for thermogravimetric experiments in a confined stationary space: (a) pure Fe, (b) Fe–12Ni, (c) Fe–20Ni (Ihara et al. 1982b). Reprinted with permission; copyright 1982b The Japan Institute of Metals and Materials.

The influence of the vaporization behaviour of HCl gas is more prominent in the gas mixture. Chlorosilane gas streams are frequently used at high temperatures (>500 °C) in semiconductor, polysilicon and fumed silica sectors. The chlorosilane gas streams mainly consist of hydrogen, tetrachlorosilane and HCl. The main reaction equations are as follows;

Aller et al. (2015, 2016 reported that the corrosion of 316L stainless steel completely differs in chlorosilane gas streams at different temperatures. When the temperature exceeds ferrous chloride’s solid gas transition temperature, the vaporization phenomenon occurs, prompting the corrosion products to be rarefied and easily dislodged. With the increase in HCl concentration in the chlorosilane gas stream, the protective properties of corrosion products sharply decrease (Figure 23).

Figure 23:

Corrosion morphology of a gas mixture (SiH₄ and HCl) with Fe–Cr alloy at 550 °C for 100 h: (a) 0.014 mol/L HCl, (b) 0.025 mol/L HCl, (c) 0.057 mol/L HCl (Aller et al. 2016). Reprinted with permission; copyright 2016 IOP Publishing.

When the HCl content in the chlorosilane gas stream is minimal and no vaporisation occurrs, the corrosion products are composed of dense FeSi after etching by the chlorosilane gas stream (Figure 24), protecting the material and slowing down the corrosion. HCl in the chlorosilane gas stream is a key factor affecting the corrosion resistance of stainless steel.

Figure 24:

Corrosion morphology of a gas mixture (without HCl) with Fe–Cr alloy at 550 °C for 100 h (Aller et al. 2016). Reprinted with permission; copyright 2016 IOP Publishing.

3.2.1 Mechanical polishing (MP)

Mechanical polishing is a surface treatment with micro-abrasives and soft tools that reduces surface roughness and enhances gloss. Three processes are involved in micro-abrasive particles and the products’ surfaces; sliding, ploughing, and cutting. After mechanical polishing, a work-hardened layer is formed on the material’s surface, the grain size is reduced, and a Beilby layer appears, as shown in Figure 26 (Kaneko and Sato 1990). The Beilby layer is a amorphous surface layer formed during mechanical polishing, with thickness ranging from a few nanometres to a micron. The surface roughness in semiconductor manufacturing is more demanding, and a fine surface finish is required. A lower surface roughness value indicates a smaller difference in height between the raised and depressed surfaces; thus, impurities, debris and pollution cannot hide on the material surface, reducing the specific surface area and slowing down the corrosion rate (Yu 2017). To meet these requirements, electrolytic polishing is used to modify the stainless steel surface in the semiconductor manufacturing industry.

Figure 26:

Schematic diagram of the Beilby layer after mechanical polishing.

3.2.2 Electrolytic polishing (EP)

Electrolytic polishing (EP) forms a smooth surface by selective dissolution of microscopic projections on the anode in an electrolytic cell (Hernando et al. 2012; Lee 2000). Figure 27 illustrates that the metal for polishing at the anode undergoes a dissolution reaction, and the inert electrode undergoes a hydrogen evolution reaction.

Figure 27:

Equipment of electrolytic polishing.

Decreased surface roughness and increased surface gloss during electrolytic polishing are attributed to the selective dissolution of concave and convex areas on the material surface. Therefore, the electrolytic polishing process can be further divided into the anodic smoothing process to reduce surface roughness and the anodic brightening process to increase surface gloss (Li et al. 2005, 2008; Łyczkowska-Widłak et al. 2020; Zhao et al. 2002). The schematic diagram of the two processes is shown in Figure 28.

Figure 28:

Mechanisms of electrolytic polishing: (a) anodic smoothing, (b) anonic brightening.

The anode smoothing process is a macro-level modification. At the anode, dissolved metal ions constantly enter the adjacent solution and the generation rate of metal ions is higher than that of migration by diffusion into the solution. Various metal ions gradually accumulate on the anode surface. Eventually, a viscous liquid film structure is formed between the electrolyte and metal surface, and the reaction is controlled by diffusion. The viscous liquid film (diffusion layer) is the Jacquet layer. The higher viscosity of the polishing solution facilitates the formation of Jacquet layers. The diffusion rate of metal ions into the electrolyte in the Jacquet layer depends on the concentration gradient between the liquid film and electrolyte, i.e., the current remains constant with increasing voltage. The overall reaction rate depends on diffusion current density. The Jacquet layer is thicker in metal surface depressions and thinner in bumps. The thinner the Jacquet layer, the greater the concentration gradient, resulting in a higher current density. Generally, Jacquet layers are approximately 20 to 80 μm thick. When the surface roughness is within the above range, the selective dissolution of the metal surface occurs (Hickling and Higgins 1952; Landolt 1987; Magaino et al. 1993; Mohan et al. 2001; Uhlig 1940; Yang et al. 2017).

The anodic brightening process is a micro-level modification inducing a mirror glossiness on the metal surface and further reducing the surface roughness. Glossiness is the metal surface’s ability to reflect light. On an ideally smooth surface, the light reflection angle is equal to the light incidence angle, whereas multi-directional scattering occurs on a non-glossy metal surface. More light is reflected when the light is less scattered, and then the surface gloss is high. A surface can exhibit mirror-quality glossiness if the unevenness of the polished surface is less than the wavelength of visible light. In the Jacquet layer, electrolytic polishing generally results in smoothing, not brightening. After the anodic smoothing, the surface roughness of the metal is approximately 0.01–0.1 µm, attributed to incomplete structures near the grain boundaries. During the micro-modification phase, continuously maintained at a high potential, the anode surface generates a thin film. The created film layer is located between the Jacquet layer and the metal surface, distributed at approximately the same thickness. The film allows the crystalline incomplete grains to be preferentially dissolved, removing the microscopic unevenness of the polished surface, further reducing the surface roughness and achieving an anodic brightening on the metal surface (Alkire and Cangellari 1983; Beck 1982; Grimm and Landolt 1994; Grimm et al. 1992; Russell and Newman 1986; West et al. 1992; Williams and Barrett 1956).

Electrolytic polishing is recommended to be only suitable for refining low-roughness surfaces (Lee 2000). To reduce surface roughness, the surface must be mechanically polished before the electrolytic polishing process. Electrolytic polishing usually reduces the roughness by 50–80 %; therefore, the effect of mechanical polishing on surface roughness cannot be ignored. The lower the initial surface roughness after mechanical polishing, the smoother the final polished surface under the same electrolytic polishing parameters, as shown in Figure 29.

Figure 29:

Effect of initial surface roughness on electrolytic polishing.

Electrolytic polishing is sensitive to non-metallic inclusions in stainless steel (Lowery and Roll 1998; Sutow 1980). Electrolytic polishing usually attacks these non-metallic inclusions, with sulphide inclusions being the primary attack location. Sulphide inclusions act as anodes in the electrolytic polishing and are selectively removed. Several pits remain on the surface in the presence of a high amount of sulphide (Figure 30). Therefore, efficient melting technology can reduce the volume percentage of elemental sulphur, generally limiting it to 0.003 % and below (Collins 1997).

Figure 30:

Surface morphology after electrolytic polishing: (a) 0 min, (b) 15 min, (c) 25 min (Sutow 1980). Reprinted with permission; copyright 1980 John Wiley and Sons.

Gibb’s free energy calculations result in a preferred reaction order: MnS < Fe < Al₂O₃. Oxide inclusions in the electrolytic polishing act as cathodes without dissolution. However, due to the Fe dissolution, the oxide inclusions also fall off the metal surface. After refining and cold working of stainless steel, plastic deformation layers are formed with particle contamination and residual gas adsorption on the metal surface. Degreasing and cleaning are commonly used in semiconductor manufacturing to remove contamination, and baking removes adsorbed gases from surfaces. Moreover, residual stresses and amorphous layers (Beilby layers) are present on the surface after mechanical polishing. Electrolytic polishing techniques can effectively remove these residual stresses and Beilby layers while forming a dense passivation film on the surface (Watanabe et al. 1989).

Figure 31 shows a 10 nm work-hardened layer on the metal surface before electrolytic polishing. Electrolytic polishing removes contamination and work-hardened layer from the original surface, achieved by stripping the material surface layer by layer. Finally, the surface after electrolytic polishing is similar to the internal organisation of the substrate. Hardness experiments have effectively verified the surface stripping process.

Figure 31:

Schematic diagram of the surface stripping process in electrolytic polishing.

Electrolytic polishing creates a dense and thin passivation layer, significantly better than the passivation film formed by air and nitric acid. Lee and Lai (2003) and Awad et al. (2012) found that Fe dissolved before Cr and Ni, and the Cr content of the passivated film was significantly increased during the electrolytic polishing process (Figure 32), inferring that the main component of the film was Cr₂O₃, which can improve the corrosion resistance of stainless steel substantially.

Figure 32:

Characterisation of the surface composition of electrolytic polishing: (a) XPS depth profiling, (b) surface SEM/EDS characterisation (Awad et al. 2012; Lee and Lai 2003). Reprinted with permission; copyright 2012 Elsevier; reprinted with permission; copyright 2003 Elsevier.

3.2.3 Electrochemical buffing (ECB)

Baba and Sato (1990) proposed that mechanical and electrolytic polishing can create a finer surface, further reducing surface roughness and changing the surface stress state, known as electrochemical buffing (ECB). Figure 33 depicts the ECB technology can achieve lower surface roughness and can handle 316L ultra-high purity components to obtain a bright, contamination-free, micro-crack-free surface (Tsuji et al. 2017).

Figure 33:

Surface morphology of different surface treatments: (a) bright annealing, (b) electrolytic polishing, (c) electrochemical buffing (Tsuji et al. 2017). Reprinted with permission; copyright 2017 Japan Society of Applied Physics.

ECB can significantly reduce surface roughness, change the metal surface’s stress state, and eliminate defects caused by cold working, as shown in Table 5.

The ECB mechanisms was described by Baba and Sato (1990) and Tsui et al. (2007). During the electrolytic process, the metal surface dissolves, and a passivation film is subsequently formed. Due to the grinding of the abrasive particles, the passivation film is stripped from the surface bumps, leaving the bumps in a dissolved state again, thereby dissolving the metal. Figure 34 illustrates that the iterative electrochemical and mechanical processes between the workpiece and grinding edge shorten the polishing time and smooth the bumps on the metal surface, finally polishing the surface to a mirror finish. The average and maximum surface roughness is further reduced.

Figure 34:

Equipment and mechanisms of ECB.

A distinct advantage of ECB over EP exists is that ECB creates a fine grain structure on the stainless steel surface due to the synergistic effect of electrolytic and micro polishing with viscoelastic abrasives. Shirai et al. (1996) and Ohmi et al. (1996) found that the α-phase appeared on the ECB-treated stainless steel surface, as shown in Figure 35. In addition, Baba and Sato (1990) revealed the appearance of an amorphous structure on the ECB surface via high-energy electron diffraction and that the amorphous structure positively affects corrosion resistance.

Figure 35:

Structural differences between ECB and EP surface crystals: (a) ECB surface, (b) EP surface (Ohmi et al. 1996). Reprinted with permission; copyright 1996 American Vacuum Society.

3.3.1 Passivation of stainless steel

Stainless steel can form a passivation film in the atmosphere for excellent corrosion resistance (Zhao et al. 2020). The structural denseness of the passivation film, chemical composition, and other factors determine its corrosion resistance (Ge et al. 2003; Wang et al. 2019). Nowadays, certain methods make obtaining passivation film structures with excellent properties possible. The higher the percentage of Cr in the film’s chemical composition, the higher the corrosion resistance (Wallinder et al. 1998). In the semiconductor manufacturing industry, however, stainless steel has to meet corrosion resistance requirements and requires significant attention regarding moisture absorption and desorption (Tomari et al. 1990). This review highlights methods on improving passivation film performance (Table 6).

The passivation film of stainless steel is composed of Cr₂O₃ and Fe₂O₃. When the Cr percentage increases, the corrosion resistance improves significantly (Ohmi et al. 1998; Shirai et al. 1994). The reasons can be divided into two parts: (1) When the temperature is below 170 °C and above 400 °C, the Gibbs free energy change in Cr₂O₃ reacting with corrosion gases is significantly higher than Fe₂O₃, as shown in Figure 36a. The Cr₂O₃ stability in the corrosive gases environment is better. The higher Cr₂O₃ percentage in the film demonstrates the film’s better corrosion resistance. (2) The passivation film (Cr₂O₃) exhibits excellent moisture release properties, i.e., minimal moisture remains on the Cr₂O₃ after purging. Water molecule adsorption is an important factor affecting the corrosion rate; therefore, increasing the Cr₂O₃ percentage contributes to reducing the effect of the condensation water in accelerating corrosion.

Figure 36:

Moisture sensitivity of material composition: (a) curves of Gibbs free energy, (b, c) curves of moisture release of different specimens surfaces (Shirai et al. 1994). Reprinted with permission; copyright 1994 IEICE.

Tomari et al. (1990) suggested that obtaining excellent surface passivation films on 316L (EP) stainless steel was possible by dry passivation treatment. The passivation film’s ability to release moisture is significantly improved compared to 316L (EP), and the dry passivation treatment is strictly temperature dependent, with excellent hydrophobic properties achieved in the appropriate interval, as shown in Figure 37. In addition, the passivation film is exposed to 35 % HCl solution, and the corrosion resistance is evaluated by observing the hydrogen generation time. As shown in Figure 38, the material’s corrosion resistance increases significantly in the corresponding temperature interval, i.e., the lower the moisture concentration in dry passivation treatment, the better the film’s corrosion resistance. Composition analysis of passivation films prepared at different moisture concentrations indicated that at lower moisture concentrations, the Cr percentage in the passivation film is high; thus, the corrosion resistance is higher (Tomari 1995).

Figure 37:

Curve of moisture adsorption capacity versus temperature (dry passivation treatment) (Tomari et al. 1990). Reprinted with permission; copyright 1990 Elsevier.

Figure 38:

Effect of different moisture concentrations during preparation on the corrosion resistance of passivated films: (a) curves for hydrogen generation time, (b) curves for the percentage of chromium in passivation films prepared with different water concentrations during preparation (Tomari et al. 1990). Reprinted with permission; copyright 1990 Elsevier.

Therefore, the lower the moisture content of dry passivation treatment, the higher the Cr content of the prepared passivation film and the better the corrosion resistance. The decrease in corrosion resistance at excessively high temperatures is attributed to the microstructural changes caused by thermal ageing.

Shirai et al. (1996) and Ohmi et al. (1996) demonstrated the successful formation of passivation films without Fe elements on the ferritic stainless steel (FS9) and austenitic stainless steel (316L) surface. The percentage of Fe elemental on the surface is zero, and the chromium-to-oxygen ratio is 2:3 (Figure 39), suggesting that the passivation film composition is predominantly Cr₂O₃. Meanwhile, the Cr₂O₃ passivation film exhibits excellent corrosion resistance, superior to nickel-based alloys. Figure 40 shows no corrosion after 120 h of exposure to HBr (20 cc/min) gas at 120 °C.

Figure 39:

XPS depth profiles of iron-free passivation film: (a) 316L (EP), (b) Passivation (Cr₂O₃) (Ohmi et al. 1996). Reprinted with permission; copyright 1996 American Vacuum Society.

Figure 40:

Corrosion morphology of stainless steel with different surface treatments after 120 h of exposure to HBr (20 cc/min) gas at 120 °C: (a, b) 316L-EP, (c, d) HASTELLOY-EP, (e, f) FS9-EP-Cr₂O₃, (g, h) 316L-ECB-Cr₂O₃ (Shirai et al. 1994). Reprinted with permission; copyright 1994 IEICE.

Stainless steel forms excellent passivation films by oxidation. Fluorinated passivation films can also be prepared using F₂ (Maeno et al. 1992a,b; Miki et al. 1990). Fluorinated passivated films demonstrate excellent corrosion resistance. The equation for the reaction between F₂ and stainless steel is as follows;

where the Fe_xF_y is a mixture of FeF₂ and FeF₃. Here, there is an excess of fluorine in the interstitial and lattice positions of the fluorinated film and between the fluorinated film and stainless steel. Since the solid-gas transition temperature of FeF₃ is much lower than that of FeF₂, the Fe_xF_y has a loose structure, unsatisfactory corrosion resistance and hydrophobic properties, requiring post-treatment. Transformation of FeF₃ to FeF₂ by surface modification is done according to the following equation;

Miki et al. (1990) reported that when the modified fluorinated films were exposed to a high-pressure chlorine environment (Cl₂, 490 kPa, 25 °C, 72 h) and a hydrogen fluoride environment containing moisture (HF: H₂O: N₂ = 5: 1: 94, 98 kPa, 25 °C, 72 h), no corrosion is observed (Figure 41).

Figure 41:

Corrosion morphology of the fluorinated film: (a, b) Cl_2, 490 kPa, 25 °C, 72 h, (c, d) HF: H₂O: N₂ = 5: 1: 94, 98 kPa, 25 °C, 72 h (Miki et al. 1990). Reprinted with permission; copyright 1990 Elsevier.

3.3.2 Coating technology of stainless steel

Nickel alloys exhibit excellent corrosion resistance but cannot be used widely due to their high expense. Therefore, nickel is deposited on the stainless steel surface in the semiconductor manufacturing industry by electroless plating for excellent corrosion resistance. Nowadays, the widest application of electroless nickel is the Ni–P coating. The electroless nickel plating allows a uniform coating on irregularly shaped workpieces, uniformly coating the entire surface and depositing into depressions and internal areas (Narayanan et al. 2006; Sarret et al. 2006).

Electroless Ni–P coatings’ microstructure and corrosion resistance depends primarily on phosphorus content (Shozib et al. 2022). The microstructure of low phosphorus coating comprises a mixture of amorphous and microcrystalline, with relatively high mechanical properties. High phosphorus coatings have high corrosion resistance due to their homogeneous amorphous microstructure; however, their mechanical properties are relatively weak (Gu et al. 2005; Sahoo and Das 2011). Electroless Ni–P coatings with 1–5 wt% phosphorus (low phosphorus) are crystalline, while electroless Ni–P coatings with 6–9 wt% phosphorus (medium phosphorus) are a mixture of crystalline and amorphous structures and electroless Ni–P coatings with 10–13 wt% phosphorus (high phosphorus) are amorphous. With the increase in phosphorus content, the structure of the coating changes gradually from crystalline to amorphous. The amorphous structure of nickel–phosphorus solid solution is extremely homogeneous, without any dislocations, twins, grain boundaries, and other defects. The surface oxide film also does not show faults and other defects (Shozib et al. 2022). The high phosphorus coating possesses an extremely high toughness that makes it difficult to fracture, while the non-crystalline surface oxide film is produced and repaired quickly once damaged. The oxide film is a phosphate film, preventing corrosion and therefore enhancing the corrosion resistance of the coating. However, this coating is susceptible to corrosion by strong oxidizing acids. The amorphous structure of nickel–phosphorus is initially deposited on grain boundaries and then slowly spreads in all directions, as shown in Figure 42, gradually forming a cellular morphology with time, densely and compactly aligned with each other (Guo et al. 2017).

Figure 42:

Morphology of electroless Ni–P: (a) surface, (b) cross-section (Guo et al. 2017). Reprinted with permission; copyright 2017 John Wiley and Sons.

Figure 43 shows unavoidable pores on the coating surface; therefore, the corrosion medium diffuses from the surface to the substrate through meandering pores. At this point, there is a potential difference between the coating and the substrate, with a larger cathode (coating) and a smaller anode (exposed substrate), resulting in an extremely rapid corrosion rate.

Figure 43:

Schematic diagram of the corrosion failure mechanism of electroless Ni–P.

Electroless Ni–P is sensitive to temperature. When held at lower temperatures, the amorphous structure transforms from a higher energy sub-stable amorphous structure to a lower energy sub-stable amorphous structure, called structural relaxation. It can reduce the internal stress in the coating, increasing the hardness, precipitating hydrogen, reducing the porosity and improving corrosion resistance. At a higher temperature, the coating atoms gain enough energy, enhancing the diffusion ability, overcoming the potential barrier of amorphous transformation into crystals and generating a new crystal phase (Ni₃P) crystal defects (grain boundaries and segregation), and reducing corrosion resistance (Georgiza et al. 2013; Zhang et al. 2008).

Electroless Ni–P can be applied on aluminium alloys, steel, and polymers, etc. (Shozib et al. 2022). However, even after the coating fails in the semiconductor manufacturing industry, stainless steel remains stable compared to other substrates. Therefore, considering safety, cost and semiconductor profit, stainless steel is a suitable substrate in the semiconductor manufacturing industry.

In semiconductor manufacturing, Ni–F coatings on stainless steel exhibit excellent corrosion resistance in fluorinated environments. Ni–F coatings are suitable in fluorinated gas environments. However, directly depositing Ni–F on stainless steel surfaces is not conducive; therefore, priority is given to the electroless Ni–P and subsequent modification by fluorination (Ohmi 1994).

Maeno et al. (1992a,b) described that the composition of the Ni–F coating is NiF₂. In a fluorinated environment, the phosphorus in the Ni–P coating is oxidized by the fluorine gas to PF₅. PF₅ is volatile in a fluorinated environment and detaches from the coating, as shown in Figure 44, while the fluorine combines with nickel to form a stable NiF₂ structure, according to the following reaction equation;

Figure 44:

Conversion of phosphorus to fluorine in electroless Ni–P: (a) XPS depth profile, (b) FTIR spectroscopy (Maeno et al. 1992a,b). Reprinted with permission; copyright 1992 IOP Publishing.

The corrosion resistance of the coating substantially improves by fluorination. No corrosion is observed after exposure to high-pressure chlorine and wet HF gas (Figure 45).

Figure 45:

Corrosion morphology of the Ni–F coating: (a, b) Cl_2, 2 kg cm⁻², 25 °C, 168 h, (c, d) HF: H₂O: N₂ = 5: 1: 94, 1 kg cm⁻², 25 °C, 72 h (Maeno et al. 1992a,b). Reprinted with permission; copyright 1992 IOP Publishing.