1 Introduction

Interest in molten salts as heat transfer fluids and thermal storage media in concentrated solar power (CSP) plants has increased [1]. One of the most commonly used salts in CSP is molten nitrate salts in heat transfer and storage applications. Salt based on alkaline nitrates (Solar Salt, 60 wt.% NaNO3/40 wt.% KNO3) is the most used medium [2]. This salt is characterized by high density, high specific heat, low melting point, high thermal stability, low viscosity in its molten state, and low vapor pressure [3, 4], as well as a competitive price [1]. However, according to several experiences with this fluid in CSP environments, solar salt can cause significant corrosion in the steel components of the power plant [5]. This fact is closely related to the nitrite, chloride and sulfate impurities contained in industrial grade salt formulations [1, 6] and also with the high operating temperature and chemical properties of the salt itself [7, 8]. Corrosion degradation of materials by molten salts in CSP plants is under study [5]. Ferritic-martensitic steels have been used for industrial applications due to their cost, despite having worse properties than other materials, such as stainless steel, as they are prone to localized corrosion under severe conditions [9, 10]. One of the methods used to avoid this phenomenon is protective coatings [11]. The use of high temperature corrosion resistant coatings on ferritic-martensitic steels would be an important advance for this type of material. This solution not only would help to overcome the corrosion problems of using ferritic-martensitic steels at these temperatures, but also would allow the CSP industry to improve the LCOE (Levelised Cost of Energy) by reducing O&M (Operating and Maintenance) costs [12]. At present, results on the effect of protective coatings on steels against corrosion caused by molten nitrate salts are very limited. Some results are related to the effect of aluminising [13, 14] and multilayer coatings based on nickel and aluminium [15]. Dorcheh and Galetz [14] proposed aluminium slurries as a solution for the corrosion process suffered by P91 and 304 steels in molten nitrates. They performed isothermal corrosion tests at 600 °C under flowing laboratory air for up to 2500 h. Their results were promising and they concluded that the coating needed to be optimised to improve its behaviour under cyclic conditions. To protect low-alloy steels against corrosion caused by molten solar salts, the multilayer PVD NiV-Al metallic coating has been used, resulting in a significant reduction of corrosion rates and an improvement in the durability of the developed coatings and the underlying steel substrate [15]. Audigié et al. [16] also proposed slurry aluminide coatings deposited on ferritic-martensitic P92 steel, performing isothermal tests at 550 °C and 580 °C for 2000 h. These authors obtained promising results and observed the formation of numerous protective FeAl species in the outer layers of the coated substrates. Fähsing et al. [17] to improve the molten salt corrosion resistance of VM12 and T91 ferritic-martensitic steels proposed the use of Cr diffusion coatings and Al-Si slurry coatings. These authors carried out corrosion tests on molten nitrates with different Cl and SO42− contents at 560 °C for 1000 h and obtained promising results, although longer exposure times are required to confirm their suitability. Other authors have studied the use of protective coatings on steels that operate in contact with molten carbonates and chlorides. In this respect, Gómez-Vidal and Morton [18] assessed the employment of SiO2 coatings on SS347 working in contact with eutectic mixtures of carbonates and chlorides (K2CO3-Na2CO3, K2CO3-Na2CO3-Li2CO3, NaCl-LiCl and NaCl-KCl) at temperatures that ranged between 600 °C and 750 °C. These types of coatings were able to protect the substrate during short-term tests, but longer tests are needed to evaluate the chemical stability of the system.

A variety of techniques are used for the deposition of coatings, such as chemical vapor deposition (CVD), physical vapour deposition (PVD), electrochemical deposition, plasma spraying and the sol-gel process [19]. Among these, sol-gel coatings present numerous advantages, several of the most important being [19, 20]: they generally present low processing temperature; they may be deposited on complex shapes and can be produced without the need for machining or melting; and finally, they are formed by ‘green’ coating technologies, i.e., the compounds used do not introduce impurities into the end product as initial substances. One of the most commonly used sol-gel protective coatings is formed by oxides, such as TiO2 [21], ZrO2 [22], SiO2-ZrO2 [23] and ZrO2-CeO2 [24]. Within these, yttrium-stabilised zirconia seems to be a great option owing to its properties [25]. Moreover, sol-gel solutions can be deposited by: spin coating, sputter coating, electrodeposition, dip coating, etc. [26, 27]. The dip coating technique allows the preparation of uniform coatings by simply controlling the removal rate and has a low cost [28]. Sol-gel coatings offer advantages, such as [19, 20]: low processing temperature; they allow deposition on complex shapes, can be produced without machining or casting… etc. Sol-gel protective coatings formed by oxides, such as TiO2 [21], ZrO2 [22], SiO2-ZrO2 [23] and ZrO2-CeO2 are widely used [24]. Within these, yttria-stabilized zirconia seems to be a great choice due to its properties [25].

In this study, zirconium-based coatings dip-deposited on P91 and uncoated P91 substrate were tested in contact with molten nitrate salts at 500 °C up to 2000 h, and the results obtained for coating resistance were compared.. Finally, the coatings were tested up to failure, and the locations were analyzed from the morphological and compositional point of view.

2 Materials and methods

2.1 Materials

2.1.1 Steel samples

The substrate used was 9Cr-1Mo ferritic-martensitic steel (ASME Grade P91/T91), one of the most widely used steels in engineering. P91 due to its high LCF life, could have a great interest in its application in high temperature systems [29]. P91 contains a higher percentage of Cr (~9%) and a composition, of 0.10% C, 0.25% Si, 0.40% Mn, 9.00% Cr, 0.95% Mo, 0.20% V, 0.07% Nb, and 0.04% N in weight percentage. The substrates were fabricated with a size of 20 × 10 × 2 mm3 and were surface polished

2.1.2 Manufacture of the sol–gel solution

The yttria-doped zirconium sol-gel solution was prepared by mixing a solution of N-propoxide/1-propanol/H2O/HNO3 with a solution of yttrium acetate/2-propanol/H2O/HNO3 both with a molar ratio of 1/15/6/1. The mixture was mechanically stirred for 30 min, and then finished by adding H2O followed by mechanical stirring for 1 h [30].

2.2 Coating deposition and characterisation

The materials in sol phase, before gelation, were used to coat P91 samples 20 mm long, 10 mm wide and 1.5 mm thick using a KSV-Nima dip coater. Before starting the coating process, the substrates were ground and polished (with sandpaper, up to 800 grains cm−2 and polishing cloths up to 3 μm), and ultrasonically cleaned with acetone and de-ionised water, for 5 min. To dip-coat the samples, a withdrawal speed of the substrate equal to 25 mm·min−1 was used. The coated substrates were dried at 100 °C, 60 min and subsequently sintered at 500 °C, 120 min, with a heating and cooling velocity of 3 °C min−1. This heat treatment promotes the densification of the film without unsuitable degradations. The surface preparation, withdrawal rate and heat treatment were selected on the basis of previously published studies [30, 31], owing to the good thickness uniformity they provide.

The coated substrates were then evaluated by weight difference and characterized by scanning electron microscopy (SEM) with a JEOL® JSM-820 equipment and by X-ray diffraction (XRD) with a PANalytical® equipment (model X′Pert PRO MRD). The SEM allows to study of morphology and chemical composition of the surface and corroded samples after the isothermal immersion test. In the SEM of the cross sections of the samples, these were mounted on a phenolic resin (Buehler ®) and ground with 240, 320, 400, 600, 1000, 2400, and 4000 silicon carbide with distilled water. XRD allows phase analysis of the coating layer deposited on the samples. The XRD patterns obtained were compared with the standards compiled by the Joint Committee on Powder Diffraction and Standards (JCPDS) to identify the phases.

2.3 Preparation and characterisation of the salt mixture

Solar salt (60 wt% NaNO3/40 wt% KNO3) was used as a corrosive agent, employed in solar power plants [32]. The eutectic NaNO3/KNO3 was prepared from the above compounds (purity >99%) supplied by BASF® and Haifa® respectively. From the starting salts, the impurity level of the prepared mixture was of 130 ppm Cl and 60 ppm SO42−. The impurity level is a parameter that highly influences the properties and corrosive action of the salt (the higher the impurities, the higher the corrosive behaviour) [6].

The presence of impurities in the prepared binary salt influences its melting point and decomposition temperature. The melting point of the salt was determined by differential scanning calorimetry (DSC) with a DSC-Q20 calorimeter using a hermetically closed aluminum crucible and the thermal stability was measured by thermogravimetric analysis (TGA) with a SDT-Q600 equipment using an open platinum crucible. In both crucibles, 10 mg of sample were used in each case, in an inert nitrogen atmosphere of 50 ml·m−1 and at a heating velocity of 10 °C-min−1. Before the analysis, both instruments were calibrated with indium.

2.4 Corrosion study of the coated samples

Isothermal immersion tests were carried out by the gravimetric method. The coated and uncoated samples were immersed in 50 ml of the molten salt to a depth of about 3.5 cm in alumina crucibles.

The crucibles together with the solid salt mixture were placed inside an electrical chamber (Carbolite®) until the required temperature was reached in an air atmosphere. The corrosion temperature assay was selected on the basis of previous TGA results. Samples immersed in the molten salt were weighed at 0, 24, 48, 72, 72, 168, 250, 500, 750, 1000, 1250, 1500, 1750 and 2000 h. Three samples were extracted from each system analyzed (P91 coated and P91 uncoated) for each working time, cooled slowly and rinsed with hot distilled water to remove salt residues. Subsequently, they were dried and weighed (average of five weight values). The formula (Eq. 1) used to calculate the mass increase with time is.

$$\frac{{\Delta m}}{{S_0}} = \frac{{m_f - m_i}}{{S_0}}$$

where mi is the initial mass of the specimen, mf is the mass of the sample at the selected time and S0 is the initial area of the specimen.

After 2000 h of testing, three samples from each system were extracted from the binary salt for examination and microstructural analysis, as described in Section 2.2.

3 Results and discussion

3.1 Characterisation of the coated samples

To acquire the surface morphologies and the phases formed, coatings prepared under the conditions described in Section 2.2 were subjected to SEM-EDX and XRD. Figure 1a, shows that the morphology of the ZrO2-Y2O3 coating is uniform and homogeneous, with a thickness of 463 nm. As previously reported [33], the coating grows by forming coral-like agglomerations that are embedded within trenches, while a diffusion zone, with a depth of ~315 nm, appears as consequence of the thermal treatment to which the coatings are subjected. This diffusion zone was also confirmed by a compositional analysis carried out using energy-dispersive X-ray spectrometry (EDX) (see Fig. 1b). The EDX analysis showed the higher atomic percentage of Zr, Y and O in the upper layer (Spectrum 1: 7.39%, 1.05%, and 38.37%, respectively), which corresponds properly to the deposited coating, and its gradual reduction as you go deeper into the sample (Spectrum 3: 1.94%, 0.32% and 3.56%, respectively). The results also reveal the outer diffusion of Fe (48.33 % in Spectrum 1), which possibly indicates the presence of Zr-Fe species.

Fig. 1
figure 1

Cross-section SEM micrographs and EDX analysis (in atomic percentage) of the coated P91 specimens before testing: (a) x500; (b) x10000

Figure 2 shows the XRD analysis of the coatings deposited on P91 steel. XRD results suggest the presence of a tetragonal phase as the main phase responsible for the properties of the system [34]. Additionally, the analysis reveals the presence of Fe-Zr peaks, which correspond to the species formed in the previously identified diffusion layer. Carrasco-Amador [35] and Chęcmanowski [36] also suggested the formation of these kinds of species, which improve the protective properties of the coated system [37].

Fig. 2
figure 2

XRD analysis of the coated samples before testing

It is important to highlight that unlike the results reported in [33], the coated samples in this study presented a high uniformity and homogeneity, which as previously reported in [30] clearly indicates the importance of the substrate surface preparation (polishing vs roughing).

3.2 Characterisation of the salt mixture

As described above, the impurity level is a parameter that highly influences the properties of the salt [6]. This is why the melting point and decomposition temperature of the prepared salt were determined by DSC and TGA, respectively. These parameters helped determine the heat treatment to which the salt would be subjected, and mainly to select the temperature of the corrosion tests. This temperature corresponded to the maximum working temperature of the salt.

Thus, Fig. 3a shows the curve obtained by DSC, where three peaks are detected. The first of the peaks (132.1 °C) corresponds to the α–β transformation [38]. The second represents the melting point of the salt mixture, 226.4 °C, while the peak appearing at 214.7 °C during the cooling stage corresponds to the freezing point of the salt. Figure 3b shows the results of the thermogravimetric study, in which the temperature of maximum stability corresponds to the temperature at which the weight loss of the salt mixture is first detected. That is, the sample started to degrade at 535.1 °C (weight loss percentage around 0.1%), [39]. The maximum operating temperature of the Solar Salt® is 565 °C [40], and a percentage weight loss (around 3%) [39], which explains the differences with the temperature established for this study. The temperature obtained at this percentage of weight loss for the tested salt was also shown in Fig. 1, being 599.8 °C, similar to that obtained in other studies [39]. These differences in the maximum operating temperature of Solar Salt® are attributable to the presence of impurities [41]. Therefore, the binary salt prepared is stable at 500 °C, the temperature selected for the corrosion tests.

Fig. 3
figure 3

Characterisation of the prepared salt mixture: (a) calorimetric curve; and (b) thermogravimetric curve

3.3 Corrosion study of the samples

3.3.1 Gravimetric analysis

During the corrosion test, gravimetric analysis of the samples was carried out. For this purpose, the weight of the samples introduced into the molten salt mixture for 0, 24, 48, 72, 168, 250, 250, 500, 750, 10 100, 1250, 1500, 1750, and 2000 h was determined.

Figure 4 represents the variation of the gravimetric mass with time in the P91 coated and uncoated samples immersed in the molten nitrate salt at 500 °C. The results correspond to the weight of the specimens after cleaning with water and drying. The figure shows that P91 in contact with the molten nitrate salt increases its weight as reported by other authors [14, 17, 42]. The weight gain value showed increases and decreases, which seems to indicate the existence of detachments along the test.

Fig. 4
figure 4

Gravimetric results for uncoated and coated P91 specimens immersed in Solar Salt at 500 °C

The weight change curve of the samples reveals that uncoated P91 does not show a protective action in contact with molten nitrate salts, since the oxide flakes formed on its surface are prone to buckling and blistering [14]. While the coated P91 specimens have higher corrosion resistance (3. 5 times less mass increase than uncoated P91), after 500 h of testing. After 1000 h of testing, the difference in weight gained between the coated and uncoated samples is smaller (coated P91 shows 1.75 times less mass gain than uncoated P91). These weight losses are a consequence of the detachment of non-adherent corrosion layers and were also observed by other authors [42]. At the end of the test, the uncoated P91 sample achieves an average weight gain of 0.3980 mg·cm-2 and the coated sample with a mean weight loss of 0.0280 mg·cm-2

Likewise, the weight change curves also show that up to 72 h there is an initial mass loss in the coated P91 (see Fig. 4). This mass loss may be due to a partial dissolution of some coating components in molten nitrate salts during the initial period of the test. Between 72 h and 2000 h of testing the weight gain of the coated steel remains practically unaltered, which indicates the stability of the system during the test. Agüero et al. [43] also found an initial mass loss in slurry aluminide coatings, this being attributed to un-diffused slurry residues (‘bisque’).

3.3.2 Metallographic and composition investigations

Scanning Electron Microscopy coupled with EDX (SEM-EDX) for elemental analysis was performed in order to characterise the sample at high resolution. Furthermore, XRD analyses were carried out to compositionally identify the tested samples.

Uncoated samples

The spallation of oxide scales and corrosion products in the uncoated P91 samples, to which the mass losses observed in the gravimetric analysis are attributed, was also observed by SEM. As Fig. 5 indicates, cross-section micrographs clearly show the formation of non-adherent corrosion layers, these leading to the above-mentioned spallation, which, as stated by Audigié et al. [42], occurs randomly throughout the entire test. This randomness in the spallation process makes it difficult to compare the obtained results with those obtained by other authors.

Fig. 5
figure 5

Cross-section micrographs of uncoated P91 specimens after 2000 h in contact with Solar Salt at 500 °C using a back-scattered electron detector at various magnifications: (a) x1500; and (b) x2000

EDX analysis makes it possible to clearly identify the multi-layered structure of these non-adherent corrosion layers and their elemental composition. As shown in Fig. 6, the corrosion layers are composed by iron oxides on the outer side and iron-chromium spinels on the inner side. XRD analysis confirmed the elemental composition of these corrosion products, consisting of Fe2O3 and Fe3O4 in outer layer and FeCr2O4 in the inner layer (see Fig. 7). The iron oxides that constitute the outer layer are known for their low adherence, mainly in the case of Fe2O3, and lower protective behaviour compared to other oxides, such as Al2O3 [44].

Fig. 6
figure 6

Lineal cross-section EDX analysis of the uncoated P91 specimens after 2000 h in contact with Solar Salt at 500 °C

Fig. 7
figure 7

XRD analysis of the uncoated P91 specimens after 2000 h in contact with Solar Salt at 500 °C

These results are in-line with those obtained by Fäshing [17], Dorcheh [14], Audigié [42] and Gurr [15], who also found these multi-layered structures at higher test temperatures (560–600 °C). However, given the difficulty of distinguishing Fe3O4 and FeCr2O4 phases from XRD results due to overlapping peaks, authors such as Audigié et al. [40] proved the presence of Fe3O4, but did not confirm the presence of FeCr2O4, attributing it to the high solubility of chromates in nitrate molten salts [45].

Coated samples

Fig. 8 shows the SEM cross-section micrographs of the coated samples after 2000 h of testing. SEM micrographs show the good behaviour of the samples after being in contact with molten salt during 2000 h at 500 °C, especially when compared to uncoated samples. One aspect of particular note is the diffusion layer that emerges from the punctual EDX analyses performed on the cross-section (see Fig. 8). This seems to indicate that an inner diffusion of the coating’s components into the substrate takes place. In this diffusion layer the atomic percentage of Zr decreases from 14.60% in the outer zone (see Spectrum 1 in Fig. 8) to 0.65% in the inner zone (see Spectrum 6 in Fig. 8). It is also important to highlight the outer diffusion of Fe in the system, which reaches atomic percentages of around 70% in Spectrum 2, while Cr does not seem to play an important role. This diffusion layer can be observed more clearly in the lineal analysis performed on the same sample (see Fig. 9).

Fig. 8
figure 8

Cross-section SEM-EDX analysis of a coated P91 specimens after 2000 h in contact with Solar Salt at 500 °C

Fig. 9
figure 9

Lineal cross-section EDX analysis of a coated P91 specimen after 2000 h in contact with Solar Salt at 500 °C

Thus, according to the EDX punctual and lineal analyses, Zr reaches a depth of about 5.18 μm. This diffusion layer can be considered as one of the main causes of the good behaviour of the coated samples. Considering the outer diffusion of Fe, the diffusion layer is essentially expected to be composed of Fe-Zr species.

Figure 10 shows the XRD analysis of the tested samples, in which the peaks corresponding to the substrate and ZrO2-Y2O3 coating are identified. Additionally, species constituted by the interaction between Zr and Fe are included, which correspond to the diffusion zone previously detected by SEM-EDX. Moreover, peaks corresponding to the binary salt are detected in the analysis, which indicates the presence of salt residues on the surface of the tested samples.

Fig. 10
figure 10

XRD analysis of the coated P91 specimens after 2000 h in contact with Solar Salt at 500 °C, including the Fe-Zr and substrate peaks overlapping

As reported by Sharma [46], it is difficult to perform the identification of Fe-Zr species given the high content in Fe, since this causes the Fe-Zr peaks to overlap those corresponding to the substrate (see Fig. 10). The degree of deviation from the substrate peaks depends on the percentage of Zr, which in this particular case seems to be low (around 8–14% [46]), which is in agreement with those values obtained by EDX. These results would clearly support the presence of the diffusion layer constituted by Zr-Fe species.

Diffusion in sol-gel coatings in air atmospheres has been widely studied by several authors. Czerwinski and Szpunar [47] affirmed that CeO2 sol-gel coatings reduce the oxidation rate in Cr2O3-forming steels because Ce4+ cations form pairs with cationic vacancies in the oxide grain boundaries, thus blocking these fast ionic paths. Baron et al. [36] also proposed that a similar mechanism occurs in sol-gel ZrO2-Y2O3 coatings. These authors affirmed that in air a thin Fe2O3 layer is formed near the coating layer, which allows ZrO2 to react with this iron oxide and form pairs with vacancies due to Coulombic

$$3ZrO_2\mathop { \to }\limits_{2Fe_2O_3} 6O_O + \left( {3Zr_{Fe}^\blacksquare + V_{\,\,\,\,\,Fe}^m} \right)$$

Although there is formation of a new vacancy of iron, there are three zirconia cations available to form pairs with the vacancies of iron from the oxide. Consequently, the mobile vacancy concentration in the oxide decreases due to the formation of such pairs. Once this protective scale is formed, the ionic diffusion is slowed, decreasing the oxidation rate [36]. Chęcmanowski and Szczygiel [35] also took into account this mechanism to explain the protective behaviour of ZrO2-base sol-gel coatings deposited on FeCrAl alloys.

In view of the above, it may be confirmed that the same mechanism is occurring in the ZrO2-Y2O3 system assessed in this study, to an even higher degree considering the results obtained by EDX and XRD, where Fe-Zr pairs have been clearly detected. Thus, Fig. 11 shows the potential corrosion mechanism followed by the coated P91 samples, where both coating and diffusion layers are shown, including also the diffusion of the main species involved in the mechanism.

Fig. 11
figure 11

Corrosion mechanism followed by a coated P91 specimen throughout a 2000-hour test in contact with Solar Salt at 500 °C

4 Conclusions

This study assessed the protective effect of ZrO2–3%molY2O3 sol-gel coatings on 9Cr-1Mo ferritic-martensitic steel (ASME Grade P91/T91) when working in contact with Solar Salt for CSP applications. To this end, isothermal corrosion tests were performed at 500 °C for 2000 h in order to compare the behaviour of both coated and uncoated P91 samples. Apart from a gravimetric analysis, a microstructural and compositional study was performed on the samples using SEM – EDX and XRD analysis at the end of the tests.

Corrosion results demonstrated the good behaviour of the assessed coating system and its distinct protective properties. The weight-change curve of the coated specimens compared to the uncoated ones reveals the beneficial influence of the coating on corrosion resistance of P91 in contact with molten nitrate salt. After 2000 h of testing, uncoated samples showed a weight increase of 0.3980 mg·cm-2, while the weight gain of the coated samples remained practically unaltered, which indicates the stability of the system throughout the entire test. Regarding the microstructural and compositional study, spallation of oxide scales and corrosion products was observed in the uncoated samples, presenting a multi-layered structure with non-adherent corrosion layers consisting of Fe2O3 and Fe3O4 in the outer layer and FeCr2O4 in the inner layer. The microstructural and compositional study performed on the coated samples revealed the maintenance of the coating layer and the presence of a 5.18 μm-thick diffusion layer. This diffusion layer mainly consists of Fe-Zr species, which are well known for reducing the ionic diffusion in the coated system and thus improving its protective properties.

The results reveal the possible suitability of the proposed system as a potential solution to the severe corrosion problems that currently affect industrial tanks and pipes operating in contact with Solar Salt.