High temperature molten salt corrosion behavior of aluminide and nickel-aluminide coatings for heat storage in concentrated solar power plants

Accepted Manuscript High temperature molten salt corrosion behavior of aluminide and nickel-aluminide coatings for heat storage in concentrated solar power plants

P. Audigié, V. Encinas-Sánchez, M. Juez-Lorenzo, S. Rodríguez, M. Gutiérrez, F.J. Pérez, A. Agüero PII: DOI: Reference:

S0257-8972(18)30652-2 doi:10.1016/j.surfcoat.2018.05.081 SCT 23528

To appear in:

Surface & Coatings Technology

Received date: Revised date: Accepted date:

23 March 2018 24 May 2018 26 May 2018

Please cite this article as: P. Audigié, V. Encinas-Sánchez, M. Juez-Lorenzo, S. Rodríguez, M. Gutiérrez, F.J. Pérez, A. Agüero , High temperature molten salt corrosion behavior of aluminide and nickel-aluminide coatings for heat storage in concentrated solar power plants. Sct (2018), doi:10.1016/j.surfcoat.2018.05.081

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT High temperature molten salt corrosion behavior of aluminide and nickel-aluminide coatings for heat storage in concentrated solar power plants P. Audigié (1,*), V. Encinas-Sánchez (2), M. Juez-Lorenzo (3), S. Rodríguez (1), M. Gutiérrez (1), F.J. Pérez (2) and A. Agüero (1)

RI

PT

(1) Instituto Nacional de Técnica Aeroespacial, Área de Materiales Metálicos, Ctra. Ajalvir Km 4, 28850, Torrejón de Ardoz, Spain (2) Surface Engineering and Nanostructured Materials Research Group, Complutense University, Av. Complutense s/n, Madrid, Spain (3) Fraunhofer Institute Chemische Technologie Energetic Systems, Joseph-von-Fraunhofer-Str.7, D-76327 Pfinztal, Germany.

SC

(*) corresponding author: [email protected]

NU

Abstract

MA

Sprayed slurry aluminide and nickel-aluminide coatings deposited by means of electrodeposition and slurry application to 9 wt.% Cr P91 alloy were studied to mitigate

D

molten salt corrosion in concentrated solar power plants. Both coatings were tested

PT E

isothermally at 580ºC in contact with the Solar Salt (60% wt.% NaNO3, 40 wt.% KNO3) under static and dynamic conditions. Uncoated P91 showed considerable mass gains in both conditions and there was evidence of extensive spallation on both cases. Mass loss

CE

was therefore also measured after removing the corrosion products by chemical etching

AC

so that the corrosion rate could be better estimated. P91 developed a complex, fast growing multilayered oxide scale which included Fe2O3, Fe3O4 and NaFeO2 in molten nitrates whatever the test conditions. All the coated systems and in particular the nickelaluminide coating in contact with the Solar Salt up to 1000 h performed much better than the uncoated material as they exhibited lower weight variations and no evidence of significant spallation. The aluminide coating developed a thin Na ferrite scale as shown by SEM-EDS and XRD after testing under static conditions. On the tested nickelaluminide coating NiAl2O4 was detected only by XRD, so it is not possible to establish

ACCEPTED MANUSCRIPT if it resulted from the oxidation of the residual undiffused material left after heat treatment, or to a thin layer formed on top of the coating after exposure to molten salt. Interdiffusion between the coating and the substrate also occurred in the nickelaluminide coating whereas the aluminide coating formed at high temperature remained

PT

quite stable both in composition and morphology. KEYWORDS

RI

Aluminide coatings; nickel-aluminide coatings; steels; corrosion; molten nitrates;

AC

CE

PT E

D

MA

NU

SC

thermosolar plants

ACCEPTED MANUSCRIPT 1. INTRODUCTION In the last decades renewable energies, and in particular concentrated solar power (CSP) plants have experienced a growing interest due to the increased demand of clean energy. Efforts are also being carried out worldwide in order to increase power plants thermal

PT

efficiencies and to reduce electricity costs [1]. Indeed, electricity from CSPs is still very expensive when compared with that obtained from traditional fossil fuels and other

RI

renewable sources to produce energy, in particular with photovoltaics. Thermal energy

SC

storage (TES) is still a key issue as CSP are subjected to intermittency of the source (predictable night periods and unpredictable adverse weather conditions). The most

NU

widely used and studied system is based on molten salts as TES fluid [2]. Current state-

MA

of-the-art molten salt receivers use an eutectic mixture of sodium and potassium nitrates, the so-called Solar Salt (60% wt.% NaNO3, 40 wt.% of KNO3) which demonstrated considerable advantages until now. Furthermore, the Solar Salt is

PT E

D

inexpensive ($0.7/kg – $1.5/kg) [3-5] and its freezing point is ~214ºC, but the upper operating temperature is limited to 580º C due to salt decomposition and corrosion of

CE

the base materials [6].

Higher operating temperatures promote faster corrosion of the base materials (carbon,

AC

ferritic and stainless steels) used in tubes and tanks that are in contact with these molten-salt mixtures [7]. The corrosion resistance of steels under these conditions depends on the formation of a protective oxide scale rich in Cr, which is similar to what happens during oxidation in high temperature gaseous atmospheres. However, an important difference when using molten salts is that chromium compounds are soluble in these salts and prevents the formation of a protective oxide scale (passivation) [8]. This results in non-protective and/or fast-growing oxide formation and in the increment of material degradation due to higher corrosion rates.

ACCEPTED MANUSCRIPT An alternative to prevent molten salt corrosion can be provided by using coatings. Aluminide coatings are shown good corrosion properties in molten nitrates [9, 10] as well as in other molten salts like carbonates mixtures [11, 12]. For instance in molten K/Na/Li carbonates at 650ºC, the aluminide reacted with Li and formed a thin layer of corrosion-resistant LiAlO2 [13]. Gonzalez-Rodriguez et al. [14] reported that a

PT

minimum of 25 at.% of Al in a FeAl intermetallic was required to form a continuous

RI

LiAlO2 scale in a Li/K carbonate melt at 650ºC. Moreover, slurry aluminide coatings

SC

are a low cost alternative that allows uniform coating of internal surfaces. Recent studies [9, 10] have demonstrated the good behavior of these coatings in molten Solar

MA

mechanism is still not well identified.

NU

Salt at 600ºC and 580ºC by forming a thin layer of NaAlO2, but the protection

Nickel-base alloys are generally much more resistant to molten nitrate corrosion than iron-base alloys [15] but also significantly more expensive. Increasing the nickel

PT E

D

concentration in these alloys improves the corrosion resistance in a molten nitrate-nitrite salt [8]. In fact, Ni based intermetallics and coatings have been investigated to mitigate molten salt corrosion. For instance, Tortorelli et al. [16] showed that Ni3Al had a

CE

moderate corrosion rate in NaNO3-(-KNO3)-Na2O2 environments while NiAl performed

AC

relatively well. Olson and collaborators [17] indicated that a Ni-electroplated coating on Incoloy-800H had the potential to significantly improve the corrosion resistance of the alloy in molten fluoride salts while Agüero et al. [12] reported that a NiAl coating deposited by thermal spray was attacked shortly after the beginning of the test in molten carbonate mixture (62 mol.% Li2CO3 / 38 mol.% K2CO3) at 700ºC. Although others studies have already reported the fabrication of hybrid Ni-Al coating and its resistance in different high-temperature atmospheres [18, 19], no studies of Ni aluminide coatings exposed to molten nitrates have been found in the literature.

ACCEPTED MANUSCRIPT In this work, sprayed slurry aluminide and nickel-aluminide coatings deposited by means of electrodeposition and slurry application to 9 wt.% Cr P91 were characterized in the as-annealed state. Then, both systems as well as the uncoated P91 were isothermally tested at 580ºC, in contact with the Solar Salt under static and dynamic

PT

conditions.

RI

2. MATERIAL AND METHODS

SC

2.1. Materials

Sample coupons (20 x 10 x 3 mm) of P91 alloy were machined from tubular sections

NU

from Vallourec, ground with P180 SiC papers and degreased in ethanol in an ultrasonic

MA

bath before coating or testing. The alloy composition is given in TABLE I. 2.2. Coatings

D

Both the aluminide and the nickel-aluminide coatings were applied on all sides of each

PT E

specimen.

The aluminide coating was applied by spraying with an environmentally friendly Cr+6 free, water based Al slurry developed by INTA. The slurry was produced by mixing Al

CE

powder (Ø: 5 µm, 99 wt.% obtained from Poudres Hermillon) with water and a

AC

proprietary binder constituted by a mixture of inorganic compounds. The slurry was homogenized by magnetic stirring and an average slurry thickness of 100 μm was sprayed with a Sagola spraying gun. After spraying, all the coated samples were left to dry under laboratory air during 3 h. Then, a heat treatment was realized at 1050ºC for 35 min under argon to produce the aluminide coating. The nickel-aluminide coating was produced in a two-step process with an electrodeposition of a 15 μm nickel layer followed by the same Al slurry spraying process described above. After spraying, all the coated samples were left to dry under

ACCEPTED MANUSCRIPT laboratory air during 3 h. Subsequently, a diffusion heat treatment was performed under nitrogen flow at 700ºC for 10 h to produce the diffused coating. For both coatings after heat treatment, undiffused slurry residues (“bisque”) were removed by slightly grinding.

PT

2.3. Corrosion Testing All the coated and uncoated samples were tested by immersion in the Solar Salt at

RI

580°C under static and dynamic conditions. The Solar Salt is a eutectic mixture of 60 %

SC

NaNO3 – 40 % KNO3 with industrial quality provided by BASF and Haifa Chemical respectively. The compositions are shown on TABLE II.

NU

Before beginning the test, the salts were kept in the oven at 110ºC for 2 h in order to

MA

remove humidity, and then cooled down to room temperature.

Static Test: each specimen was fully immersed in a 30 ml alumina crucible with a

PT E

D

suitable amount of salt and placed in a Carbolite muffle furnace at the corresponding temperature, which was constantly measured during the tests. Samples were removed from the furnace at different time intervals to be weighted. To do that, samples were

CE

extracted from the crucible, left in laboratory air during a few minutes to cool down and

AC

then placed in hot distilled water (≈ 80ºC) in order to remove the salt residues. As the salts slightly evaporated at the test temperatures, crucibles were replenished with salts as needed.

Dynamic Test: each specimen was introduced in a sample holder within a previously reported rig [20] in contact with the salt mixture. This rig, patented under the reference code WO2016102719 [21], ensured a continuous re-circulation of molten salt in contact with the studied specimens. The samples were placed inside an alumina chamber which

ACCEPTED MANUSCRIPT avoids the contact between samples and the device material. The test linear velocity was set to 0.2 m/s, and was selected on the basis of the flow velocity in areas near tube bends and valves in current CSP plants (0.2 m/s – 0.5 m/s), which correspond to zones where high levels of corrosion are found. The samples were placed into the facility so that the flow was parallel to their large surfaces. This allowed to simulate the most

PT

common attack of the molten salt flow in CSP plants when the salt circulates through

SC

RI

steel tubes.

Descaling: in order to accurately measure the corrosion rate of uncoated P91, a

NU

descaling process was undertaken to measure the material loss rather than the scale growth. After taking out samples from the furnace and cleaning them with hot water,

MA

some of them were chemically etched after 1000 h of test to determine the corrosion kinetics. A solution of 50% HCl and 3.5g hexamethylenetetramine was used during 10

D

min at room temperature. It was demonstrated that this solution remove all the corrosion

PT E

products without affecting the base material [22]. Samples were weighted after

CE

corrosion products removal.

3. RESULTS AND DISCUSSION

AC

3.1. As-deposited coatings The aluminide coating formed at 1050°C is a thick (~180 µm) crack-free layer with FeAl as the only intermetallic phase on top of an Al interdiffusion zone rich in AlN precipitates, resulting from the reaction with N present in P91 (Figure 1.a) [23]. The coating exhibits a relatively low Al content (18-20 wt.%) at the surface and contains a Cr concentration of 4-6 wt.%. Between the intermetallic coating and the Al-substrate interdiffusion zone, Kirkendall porosity develops due to faster Fe outward diffusion

ACCEPTED MANUSCRIPT when compared with Al inward diffusion. The X-ray diffraction pattern confirmed the presence of the single-phase FeAl in the as-annealed state (Figure 1.b) as all the observed peaks correspond to the FeAl phase (ICDD 01-073-8033, cubic structure with a=2.889 Å). The nickel aluminide coating produced at 700ºC exhibits a much higher Al

PT

concentration (~44 wt.%) and a multilayered microstructure including several Ni-Al,

RI

Ni-Al-Fe and Ni-Fe-Cr intermetallic phases (Figure 2.a). The outermost layer is a thick

SC

Ni2Al3 phase (~38 μm with the composition 56Ni-44Al in wt.%) with very fine grains close to the surface, as well as larger grains deeper in the coating. X-ray measurement

NU

confirmed the existence of the Ni2Al3 phase as all the observed peaks correspond to the ICDD 03-065-9699 hexagonal Al3Ni2 phase (lattice parameters a=b=4.0282 Å and

MA

c=4.8906 Å) (Figure 2.b). Below, three other thin layers can be observed: a 2.5 µm thick NiAl (63Ni-37Al in wt.%) phase and two very thin Ni-Al-Fe layers with the

D

composition 70Ni-27.5Al-2.5Fe and 80Ni-15Al-5Fe (wt.%) were observed evidencing

PT E

the Fe outward diffusion from the substrate into the coating. Very low Al content was detected close to the initial surface and three Ni-Fe-Cr layers with the composition

CE

73Ni-24.5Fe-2.5Cr, 45Ni-50Fe-5Cr and 8Ni-84Fe-8Cr (wt.%) were successively observed with decreasing the Ni concentration while increasing the Fe and Cr

AC

concentrations to finally reach the substrate composition of Fe-9Cr (wt.%). Very fine black features were observed within the top zone of the coating. EDS point analysis indicated an increase in C concentration but since these features are so small, it is difficult to conclude if they are precipitates or pores. Moreover, remains of undiffused slurry (“bisque”) were found on top of the coating on some areas. Contrary to the nickel-free aluminide coatings for which the bisque easily spalled, the one on the nickelaluminide was surprisingly adherent to the coating. Ni and Al were detected by EDS on

ACCEPTED MANUSCRIPT this bisque indicating that Ni reacted and diffused within the Al particles in contact, resulting in Ni aluminide particles within the bisque (figure 2.a), despite a relatively low heat treatment temperature (700ºC). A similar behavior was observed by Bouchaud et al. [24] when a slurry aluminide was applied to a pure Ni substrate with a ceria selective diffusion barrier and can be explained by the interconnection of the Al particles present

RI

PT

in the slurry and the substrate.

SC

3.2. Gravimetric results after molten salt exposure

The hot corrosion behavior of both coatings as well as that of P91 was examined in the

NU

Solar Salt at 580°C under static and dynamic conditions up to 1000 h, and the gravimetric results are shown in Figure 3. All the coated samples behaved well in both

MA

conditions when compared with uncoated P91. Particularly, the nickel aluminide coating evidenced an excellent corrosion resistance in both conditions after 1000 h. It

D

showed very low negative mass changes (~ -0.3 mg/cm2) which were attributed to the

PT E

detachment of undiffused slurry residues. Indeed, as mentioned before, some slurry residues were inadvertently not removed prior to the beginning of the test as they were

CE

found to be very adherent to this particular coating. The aluminide coatings also performed well in molten nitrates up to at least 1000 h (0.06 mg/cm² in static conditions

AC

and 0.8 mg/cm2 in dynamic conditions) but showed higher weight variations than the nickel-aluminide coating. Besides, no significant difference of mass changes for the coated samples was found between static and dynamic conditions. In contrast, uncoated P91 exhibited a high degree of corrosion demonstrated by the large mass variation in molten nitrates under static and dynamic conditions with evidence of an important degree of spallation. For instance, the specimen mass change of P91 after 1000 h at 580°C was equal to 23.2 mg/cm² in static molten nitrates and ~ -

ACCEPTED MANUSCRIPT 7mg/cm2 in dynamic molten nitrates. In order to better estimate the metal loss and to be able to calculate the corrosion rate of P91, one uncoated sample exposed for 1000 h to each test condition, was descaled by chemical etching and the corrosion rates 𝑉𝐶 were obtained by comparing the final weight after descaling and the initial weight before testing, using the following equations:

PT

∆𝑊 𝑊𝑓 − 𝑊𝑖 = 𝑆0 𝑆0

(1)

∆𝑊 𝐾 𝑉𝐶 = | |× 𝑆0 𝑡𝜌

SC

RI

(2)

Where Wi and Wf are respectively the initial and final weights (g), S0 is the initial

NU

metallic surface in contact with the corrosive fluid (cm2), K is a constant to express the

MA

final result in mm/year (8.76 x 104), t is the exposure time (hours) and ρ is the alloy density (g/cm3) [25]. The exposure time and the P91 density used for the calculations

D

were 1000 h and 7.76 g/cm3 respectively.

PT E

The results gathered on TABLE III revealed that the test conditions (static or dynamic) in the molten nitrates atmosphere did not affect the metal loss as no significant variations were observed by this method. Indeed, the mass changes of the samples with

CE

the scales after 1000 h exhibited different values as shown in Figure 4, but the same

AC

weight change and consequently the same corrosion rate of 0.21 mm/year was estimated after descaling. This result clearly indicates that under molten corrosion conditions, the gravimetric results of uncoated samples without descaling cannot even used in a comparative manner as spallation takes places in a random manner. Nevertheless, the employed formula to calculate the corrosion rate based on the metal loss rate, assumes that the rate follows a linear law. More exposure time is required to establish if the corrosion process follows a different law (parabolic, logarithmic, para-linear, etc.), so the calculated rates can only be used as estimations. Once the corrosion rate law is

ACCEPTED MANUSCRIPT established after longer exposures, better calculations will be undertaken employing the corresponding equation.

3.3. Microstructures after exposure to molten nitrates 3.3.1. Under static conditions

PT

Figures 5 and 6 show the microstructures and the XRD measurements of the uncoated

RI

P91 and the aluminide and nickel-aluminide coatings respectively, after 1000 h at

SC

580°C in static molten nitrates.

NU

3.3.1.1. Uncoated P91

Figure 5.a shows the corrosion products of the uncoated P91 after 1000 h at 580ºC in

MA

molten nitrates under static conditions. A very thick stratified oxide scale formed, in particular when compared with that formed on the coated samples (Figure 5.b.c).

D

Indeed, P91 developed complex fast-growing multi-layered oxides with little cohesion,

PT E

as already reported in various studies [9, 10, 15]. Very high degree of spallation could be observed and therefore, only remains of oxide were found. After 1000 h at 580ºC in

CE

molten nitrates, the observed ~30 μm thick oxide layer was covered by a thin layer rich in Na and Fe as well as O, whereas the bulk of the oxide was rich in Fe, O and Cr with

AC

some Na rich layers. The content of Fe and Cr oscillated as a function of the scale thickness. This indicates the repetition of cycles in which Cr rich spinels formed, as observed by others in ferritic steels exposed to steam [26, 27]. Cr diffused from the bulk alloy until the spinel could not be sustained due to continuous depletion of Cr, therefore resulting in the formation of Fe rich oxides. The X-ray diffraction pattern after testing, confirmed the presence of the NaFeO2 as well as Fe2O3 (Figure 6.a) suggesting that Na released from the molten salt can be a

ACCEPTED MANUSCRIPT dominating factor of accelerated corrosion. The Na-rich dark zones found within the oxide can also be NaFeO2. The formation of such multilayered structure may be related to the modification in the salt composition. Basic dissolution of the oxide scales seem to be the main corrosion mechanism. This corrosion mechanism is usually observed at high temperature hot corrosion (type I) [28]. However, in this study, as the samples are

PT

immersed in the salt, results might change.

RI

Na2O formed according to the following reaction which is thermodynamically favored

SC

at any temperatures: 5 𝑁𝑎𝑁𝑂2 → 3 𝑁𝑎𝑁𝑂3 + 𝑁𝑎2 𝑂 + 𝑁2 (𝑔). As Na2O is getting consumed to form NaFeO2 under static conditions, the local NO3/NO2 activity might

NU

increase and after the consumption of NO3/NO2, the local Na2O activity increases again. The alternation in acidity in the corrosion front might promote the dissolution of

MA

different species provoking the layered structure observed in the scale. Other hypotheses are the penetration of the Na-containing melt through cracks or pores, or outward

D

diffusion of Cr and Fe through these Na rich layers. The formation of NaFeO2 was

PT E

observed before in other studies [7, 9, 16] and typical for high temperature ferrous alloys in NaNO3 although NaFeO2 was generally expected to form at temperatures

CE

above 600ºC according to Bradshaw and Goods [29]. However, Tortorelli et al. [16] calculated the equilibrium phases for Fe-Na-K-N-O and Al-Na-K-N-O as a function of

AC

PO2 and PN2 using the SOLGASMIX program and reported that NaFeO2 should be stable in molten nitrates at 507 and 677°C. NaFeO2 has its origin in the reaction between trace amounts of sodium oxide in the molten nitrate Na2O and hematite Fe2O3 as Na2O + Fe2O3  2NaFeO2. No corrosion products containing K were found likely because the basicity of NaNO3 is higher than that of KNO3 [30] and also because KNO3 has been found slightly more stable at 580ºC compared to NaNO3, and the eutectic KNO3-NaNO3 [31].

ACCEPTED MANUSCRIPT Moreover, the substrate under the scale has been nitrided, and as a result, chromium nitride has formed both inter- and intra-granularly. It is clearly evidenced on the other FESEM image of Figure 5.a and substrate nitriding was already demonstrated by the same authors in [10] after 1500 h at 580°C. However its probable effect on the

PT

mechanical properties has not yet been established.

RI

3.3.1.2. Aluminide coating

SC

According to the gravimetric results, the behavior of the aluminide coating appeared quite similar in static and dynamic conditions. Indeed, in static molten nitrates, as

NU

already shown by the authors [10, 32] there was no significant difference between the morphology of the tested coated specimens and that of the as deposited coatings (Figure

MA

5.b). Since this coating was formed at high temperature (1050°C), the resulting microstructure did not appear to have changed, maintaining the same FeAl phase at the

D

top according to the XRD measurement (Figure 6.b). Concentrations of 16 wt.% and 5.8

PT E

% (wt.%) of Al and Cr respectively were detected by EDS. Therefore, little Al was consumed after 1000 h which explains the gravimetric results and the protective

CE

behavior of the coating. No evidence of substrate attack could be seen but an oxide scale (~7 µm thick) developed on its surface (Figure 7). EDS analysis indicated O (33.1

AC

wt.%), Al (31.7 wt.%), Fe (18 wt.%) and Na (6.4 wt.%), which could correspond to NaFeO2 and/or NaAlO2. However other than peaks corresponding to FeAl, very low intensity peaks assigned to NaFeO2 and Fe2O3 oxides were detected by XRD (Figure 6.b). The FESEM image of the scale (Figure 5.b), shows at least two different colored zones within the oxide which is in agreement with two different oxides detected by the XRD analysis. Given its morphology the scale could be composed of several oxides.

ACCEPTED MANUSCRIPT Since this coating was heat treated at 1050ºC, it is also possible that a thin Al2O3 layer formed at this high temperature, prior to coating exposure. According to Tortorelli and coworkers, NaAlO2 or NaFeO2 formed on aluminiumcontaining surfaces are not very adherent to the exposed material but are stable in the nitrate salt environment at 580ºC [16]. In this case, the scales appeared to remain

PT

attached given the low variations in weight, as well as in the composition and

RI

morphology of the coating which appear to lose very little Al. In addition, according to

SC

the same authors, iron-aluminides need an initial Al concentration greater than about 35 at.% to perform relatively well in molten nitrate salts. In this study, the aluminide

NU

exhibited 30 at.% of Al (18 wt.%) close to the surface and demonstrated a good corrosion resistance against molten nitrates up to 1000 h. This suggests that a slight

MA

reduction of the Al content limit established by Tortorelli et al. can be achieved for applications of at least short durations. Longer exposure tests are being carried out to

PT E

D

corroborate this hypothesis.

3.3.1.3. Nickel-aluminide coating

CE

Figure 5.c exhibits the microstructure of the nickel-aluminide coating after 1000 h at 580ºC in static molten nitrates. As shown by the gravimetric results, the coating

AC

performed very well. The SEM observation of the sample after 1000 h did not lead to a visible degradation of the surface which validates the good corrosion resistance of the coating. The only significant microstructural change observed when comparing with the initial state, was that the Al-rich Ni2Al3 phase thickness was reduced while the NiAl large grains grew. Indeed, the thickness of NiAl phase in the initial state was ~2 μm while the one after 1000 h was close to 6 μm. Very low interdiffusion occurred between the coating and the substrate as the total coating thickness was almost the same than the

ACCEPTED MANUSCRIPT one in the as-annealed state (~ 40-45 μm). No further inward Al diffusion was evidenced as this element was detected only up to a depth of 42 μm as in the asdeposited sample. No scale could be observed by SEM on top of the coating on which some undiffused slurry residues could still be seen. X-ray diffraction analysis revealed the existence of NiAl2O4 and Al2O3 which could result from the oxidation of the bisque

PT

still attached to the coating including the observed NiAl particles. This residual oxides

RI

may improve the corrosion resistance by acting as a barrier. XRD analysis also

SC

confirmed the presence of the coating phase Ni2Al3 and exhibited very low intensity peaks corresponding to NiAl that could be present in the residual bisque, as the

MA

3.3.2. Under dynamic conditions

NU

thickness of Ni2Al3 does not allow to detect the inner NiAl phase.

Figures 8 and 9 show the microstructures and the XRD measurements of the uncoated

PT E

dynamic molten nitrates.

D

P91, the aluminide and nickel-aluminide coatings, respectively after 1000 h at 580°C in

CE

3.3.2.1. Uncoated P91

Figure 8.a. shows the corrosion products of the uncoated P91 after 1000 h at 580ºC in

AC

molten nitrates under dynamic conditions which seem more aggressive conditions than the static molten nitrates according to the initial gravimetric results (Figure 3.b) in the sense that a higher degree of spallation took place. The SEM observation confirmed the spallation as only a residual ~ 10 μm thick oxide was observed on top of the substrate. This oxide developed some cracks and consisted in a multilayered structure as the one formed under static conditions. The EDS analysis evidenced that the outermost layer of 2-3 μm thick was rich in Na, Fe and O while the ~7-8 μm thick inner layer was rich in

ACCEPTED MANUSCRIPT Fe, Cr and O. XRD analysis (Figure 9.a) revealed the presence of the sodium ferrite NaFeO2 as well as the iron oxides Fe2O3 and Fe3O4 confirming that the same corrosion products formed on the P91 whatever the test conditions. However, nitriding occurred in the substrate as it happened in the static conditions but it took place in a greater extent with an affected zone thickness of 55 µm (~11-18 µm under the static

PT

conditions). Both inter and intra-granular chromium nitrides formed as clearly

RI

evidenced in the Figure 8.a. This behavior can be explained by the higher degree of

SC

spallation under dynamic conditions, as the contact of the molten nitrates directly with the substrate is increased. Below the oxide, a Cr concentration of 7 wt.% was detected

NU

by EDS instead of 9 wt.% in the bulk. This suggested that a Cr depletion zone appeared due to the Cr consumption to form both the oxide and the nitrides.

MA

These observations seem to indicate that the oxide formed on P91 is not protective at all as the metal loss proceeds at essentially the same rate independent of the spallation

PT E

D

degree.

3.3.2.2. Aluminide coating

CE

Figure 8.b. shows the microstructure of the aluminide coating after 1000 h under

AC

dynamic molten nitrates. The coating exhibited the same microstructure than in the initial state, maintaining the same FeAl phase at the top according to the XRD measurement (Figure 9.b). Concentrations of 19 % and 5.6 % (wt.%) of Al and Cr respectively were detected by EDS. Therefore, similarly to what happened under static conditions, little Al was consumed after 1000 h to probably form an oxide which explains the protective behavior of the coating. However, by comparing the results between static and dynamic conditions, the gravimetric results showed that the aluminide coating had a weight gain higher of an order of magnitude after test under

ACCEPTED MANUSCRIPT dynamic conditions than under static conditions. No evidence of substrate attack could be seen but a very thin scale (~2-3 µm thick instead of 7 μm after static molten nitrates) developed on its surface. This layer was rich in Fe, Al, O and contained 2 wt.% of Na detected by EDS. But, the XRD analysis did not reveal the presence of oxides, probably due to their very low thickness. Similarly to what happened under static conditions and

PT

according to the EDS results, NaFeO2 or NaAlO2 could develop on top of the coating

RI

and protect the material against corrosion. However, if the oxide was not very adherent

SC

to the coating, spallation can take place or the metallographic preparation may affect the

3.3.2.3. Nickel-aluminide coating

NU

oxide scale explaining why a thinner layer was observed after dynamic molten nitrates.

MA

Figure 8.c shows the microstructure of the nickel-aluminide coating after 1000 h at 580ºC in dynamic molten nitrates. In a general way, the sample did not evidence

D

significant microstructural changes except the extent of the interdiffusion between the

PT E

coating and the substrate. This was evidenced by the thickening of the different layers as the NiAl and Ni-Al-Fe zone thicknesses increased whereas the Ni2Al3 one decreased

CE

or nearly disappeared as observed in some areas of the sample. Moreover, at the initial state, the nickel-aluminide coating was a 45 μm coating with the Ni2Al3 phase as the

AC

main and larger phase while after 1000 h in dynamic molten nitrates the coating clearly showed various binary, ternary and even quaternary phases from the surface to the substrate being the following: 60Ni-40Al, 65Ni-35Al, 78Ni-28Al-2Fe, 85Ni-12Al-3Fe, 78Ni-4Al-16Fe-2Cr and 74Ni-24Fe-2Cr (wt.%). After 1000 h the coating had an average thickness of 58 μm which demonstrated that interdiffusion occurred in a greater extent than in static conditions at in principle, the same temperature. As it is difficult to maintain a constant temperature in the dynamic test, the interdiffusion between the

ACCEPTED MANUSCRIPT coating and substrate could be affected and its temperature instability can explain this difference. XRD analysis confirmed the presence of the intermetallic Ni2Al3 and NiAl (Figure 9.c) and indicated the formation of the NiAl2O4 oxide. This oxide was not observed by SEM when the cross-section was analyzed. But an EDS mapping of the surface coating displays the combination of Ni, Al and O in some “nodule-like” areas

PT

confirming its occurrence (Figure 10). Some Na (2 wt.% in average) could also be

RI

measured by EDS. Very likely, these nodules correspond to bisque residues. Longer

SC

exposure tests will be performed to confirm if NiAl2O4 formation occurred either on top of the coating during the salt exposure or correspond to the oxidation of the residual

NU

bisque. Moreover, the quantity of the residual bisque as well as its oxidation during the molten salt exposure may explain why the weight change after 1000 h under dynamic

MA

conditions was higher than the one after the same duration under static condition.

D

4. CONCLUSION

PT E

Sprayed slurry aluminide and nickel-aluminide coatings deposited by means of electrodeposition and slurry application to 9 wt.% Cr P91 alloy were isothermally tested

CE

at 580ºC in contact with the Solar Salt under static and dynamic conditions. Uncoated P91 showed considerable mass gains in static molten nitrates and there was evidence of

AC

extensive spallation on both cases. The uncoated material developed a complex, nonprotective and fast growing multilayered oxide scale which included Fe2O3, Fe3O4 and NaFeO2. Mass loss was measured after removing the corrosion products by chemical etching so that the corrosion rate could be better estimated and the results indicated that the metal loss rate is the same under dynamic and static conditions. The main difference observed between the dynamic and static results was the higher degree of oxide spallation under dynamic conditions which led to a higher substrate nitriding degree.

ACCEPTED MANUSCRIPT All the coated systems performed much better than the uncoated material up to at least 1000 h at 580ºC as they significantly exhibited lower weight variations. No microstructural changes occurred in the aluminide coating between the initial state and after 1000 h in both conditions as the coating formed at high temperature. A thin layer rich in Al, was highlighted on top of the coating explaining its protective behavior. The

PT

nickel-aluminide had the better corrosion behavior amongst the studied systems. Very

RI

low negative weight changes were obtained after 1000 h in molten nitrates under both

SC

static and dynamic conditions which were attributed to the loss of the residual bisque. An oxide rich in Ni and Al was observed on top of the coating as demonstrated by

NU

SEM-EDS and seems to be NiAl2O4 according to the XRD measurements. This oxide either corresponds to the oxidation of the residual bisque or to a thin layer formed on

MA

top of the coating after exposure to molten salt. Interdiffusion between the coating and the substrate also occurred in the nickel-aluminide coating as Ni2Al3 reduced while

PT E

D

NiAl increased.

Finally, from an industrial point of view, both coatings have benefits. The lower Al aluminide coating formed in a slurry one-step process that allows to easily coat internal

CE

tube surfaces but the heat treatment was performed at high temperature whereas the

AC

higher Al nickel-aluminide coating was produced after a heat treatment at low temperature but needed a two-step process including a Ni electrodeposition.

ACKNOWLEDGEMENTS

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 686008 (RAISELIFE). We

ACCEPTED MANUSCRIPT acknowledge its support and we also thank all the members of the Metallic Materials

AC

CE

PT E

D

MA

NU

SC

RI

PT

Area at INTA for technical support.

ACCEPTED MANUSCRIPT List of references [1]

D. Kearney, U. Herrmann, P. Nava, B. Kelly, R. Mahoney, J. Pacheco, R. Cable, N. Potrovitza, D. Blake, and H. Price, Journal of Solar Energy Engineering, 125, pp. 170-176 (2003). H. Benoit, L. Spreafico, D. Gauthier, and G. Flamant, Renewable and Sustainable Energy Reviews, 55, pp. 298-315 (2016).

C.K. Ho, M. Carlson, P. Garg, and P. Kumar, Journal of Solar Energy

RI

[3]

SC

Engineering, 138, pp. 051008-1 (2016). [4]

PT

[2]

N. Lorenzin and A. Abanades, International Journal of Hydrogen Energy, 41,

NU

17, pp. 6990-6995 (2016).

H.K. Ho, Solar Energy, 152, pp. 38-56 (2017).

[6]

R.W. Bradshaw and S.H. Goods, Accelerated Corrosion Testing of a Nickel-

MA

[5]

[7]

PT E

Albuquerque, (2001).

D

Base Alloy in a Molten Salt, SAND2001-8759, Sandia National Laboratories,

S.H. Goods and R.W. Bradshaw, Journal of Materials Engineering and Performance, 13, 1, pp. 78-87 (2004). J.W. Slusser, J.B. Titcomb, M.T. Heffelfinger, and B.R. Dunbobbin, Journal of

CE

[8]

[9]

AC

Metals, pp. 24-27 (1985). S.A. Dorcheh and M.C. Galetz, Solar Energy Materials & Solar Cells, 146, pp. 8-15 (2016). [10]

P. Audigié, N. Bizien, I. Baraibar, S. Rodriguez, A. Pastor, M. Hernandez, and A. Agüero. Aluminide slurry coatings for protection of ferritic steel in molten nitrate corrosion for concentrated solar power technology, in International conference on concentrating solar power and chemical energy systems (SOLAR

ACCEPTED MANUSCRIPT PACES 2016), AIP Conference Proceedings, 1850, Edited by AlObaidli, A; Calvet, N; Richter, C., Published by AIP Publishing 2017. [11]

A. Agüero, M.C. Garcia, R. Muelas, A. Sanchez, F.J. Pérez, D. Duday, M.P. Hierro, and C. Gomez, Materials Science Forum, 369-372, pp. 759-766 (2001).

[12]

A. Agüero, F.J. Garcia de Blas, M.C. Garcia, R. Muelas, and A. Roman, Surface

C.S. Ni, L.Y. Lu, C.L. Zeng, and Y. Niu, Journal of Power Sources, 261, pp.

RI

[13]

PT

and Coatings Technology, 146-147, pp. 578-585 (2001).

[14]

SC

162-169 (2014).

J.G. Gonzalez-Rodriguez, M. Cuellar-Hernandez, M. Gonzalez-Castaneda, V.M.

NU

Salinas-Bravo, J. Porcayo-Calderon, and G. Rosas, Journal of Power Sources, 172, 2, pp. 799-804 (2007).

M. Spiegel and J. Mentz, Materials and Corrosion, 65, 3, pp. 276-281 (2014).

[16]

P. Tortorelli, P.S. Bishop, and J.R. DiStefano, Selection of corrosion-resistant

MA

[15]

D

materials for use in molten nitrate salts, Technical Report Oak Ridge National

[17]

PT E

Lab., TN (USA), (1989).

L. Olson, K. Sridharan, M. Anderson, and T. Allen, Journal of Nuclear

[18]

CE

Materials, 411, 1-3, pp. 51-59 (2011). Z.D. Xiang, S.R. Rose, and P.K. Datta, Scripta Materialia, 59, pp. 99-102

AC

(2008). [19]

L. Zheng, X. Peng, and F. Wang, Corrosion Science, 53, pp. 597-603 (2011).

[20]

G. Garcia-Martin, M.I. Lasanta, V. Encinas-Sanchez, M.T. de Miguel, and F.J. Pérez, Solar Energy Materials & Solar Cells, 161, pp. 226-231 (2017).

[21]

Universidad Complutense de Madrid, Apparatus and method for testing materials and fluids at a high temperature, such as salts, metals, alloys or glass in a melted state, WO2016102719, June 30, 2016.

ACCEPTED MANUSCRIPT [22]

Standard ISO8407:2009, Corrosion of metals and alloys. Removal of corrosion products from corrosion test specimens, AENOR, 2009.

[23]

A. Agüero, V. Gonzalez, and R. Muelas, Submitted to Surface and Coatings Technology (2018).

[24]

B. Bouchaud, B. Rannou, and F. Pedraza, Materials Chemistry and Physics, 143,

F.J. Ruiz-Cabañas, C. Prieto, R. Osuna, V. Madina, A.I. Fernandez, and L.F.

RI

[25]

PT

pp. 416-424 (2013).

[26]

SC

Cabeza, Solar Energy Materials & Solar Cells, 157, pp. 383-392 (2016). J.W. Quaddakers and P.J. Ennis, Materials for Advanced Power Engineering I,

[27]

NU

123, (1998).

A. Agüero, V. Gonzalez, P. Mayr, and K. Spiradek-Hahn, Materials Chemistry

MA

and Physics, 141, pp. 432-439 (2013).

R.A. Rapp, Corrosion Science, 44, pp. 209-221 (2002).

[29]

R.W. Bradshaw and S.H. Goods, Corrosion of alloys and metals by molten

D

[28]

PT E

nitrates, SAND2000-8727 Sandia National Laboratories, Livermore C.A., (2001).

A. Garche, Encyclopedia of Electrochemical Power Sources 2, 264 (2009).

[31]

T. Bauer, D. Laing, U. Kröner, and R. Tamme, Sodium Nitrate for High

CE

[30]

AC

Temperature Latent Heat Storage. in The 11th International Conference on Thermal Energy Storage, Effstock 14-17 June 2009, in Stockholm, Sweden. [32]

A. Agüero, P. Audigié, S. Rodríguez, V. Encinas-Sánchez, M.T. De Miguel, and F.J. Pérez, Protective Coatings for High Temperature Molten Salt Heat Storage Systems in Solar Concentration Power Plants, in International conference on concentrating solar power and chemical energy systems (SOLAR PACES 2017), Santiago de Chile, AIP Conference Proceedings.

ACCEPTED MANUSCRIPT

Tables TABLE I. Alloy composition (wt. %) Fe

Cr

N

Ni

Si

V

S

Mo

C

Mn

W

Nb

P91

Bal.

8.6

0.04

0.08

0.40

0.20

0.006

0.95

0.10

0.50

1.80

0.08

AC

CE

PT E

D

MA

NU

SC

RI

PT

Alloy

ACCEPTED MANUSCRIPT TABLE II. Purity (%) and composition of NaNO3 and KNO3 (ppm) Element

Cl

Na

Ca

Mg

Fe

Cu

99.5

190

-

<5

<1

<1

< 0.5

99.7

150

150

13

5

3

0.5

AC

CE

PT E

D

MA

NU

SC

RI

PT

NaNO3 KNO3

Purity (%)

Zn

Pb <1

2

ACCEPTED MANUSCRIPT TABLE III. Weight changes and corrosion rates of the uncoated P91 after testing and descaling the samples exposed 1000 h at 580°C with the Solar Salt under static and dynamic conditions. Solar Salt Static Dynamic After testing

∆𝑊

(mg/cm2)

23.2

-7.0

(mg/cm2)

-18.6

-18.6

𝑉𝐶 (mm/year)

0.21

0.21

𝑆0 ∆𝑊 𝑆0

AC

CE

PT E

D

MA

NU

SC

RI

PT

After descaling

ACCEPTED MANUSCRIPT List of figures caption Figure 1. a. Microstructure (IZ for Interdiffusion Zone) and b. XRD measurement of the as-annealed aluminide coating, (Al concentration is in wt.%).

PT

Figure 2. a. Microstructure and b. XRD measurement of the as-annealed nickel-

SC

RI

aluminide coating, (concentrations are in wt.%).

NU

Figure 3. Weight changes of the studied coated and uncoated materials in Solar Salt at

MA

580ºC under a. static conditions and b. dynamic conditions.

D

Figure 4. Comparison of the weight changes of the uncoated P91 exposed 1000 h in

PT E

molten nitrates after testing and descaling.

CE

Figure 5. Microstructure after 1000 h at 580ºC in static molten nitrates of the a.

AC

uncoated P91, b. aluminide coating and c. nickel-aluminide coating.

Figure 6. X-ray diffraction patterns after 1000 h at 580ºC in static molten nitrates of the a. uncoated P91, b. aluminide coating and c. nickel-aluminide coating.

Figure 7. EDS mapping of the aluminide coating after 1000 h at 580ºC in static molten nitrates.

ACCEPTED MANUSCRIPT Figure 8. Microstructure after 1000 h at 580ºC in dynamic molten nitrates of the a. uncoated P91, b. aluminide coating and c. nickel-aluminide coating.

Figure 9. X-ray diffraction patterns after 1000 h at 580ºC in dynamic molten nitrates of

RI

PT

the a. uncoated P91, b. aluminide coating and c. nickel-aluminide coating.

SC

Figure 10. EDS mapping of the surface of the nickel-aluminide coating after 1000 h at

AC

CE

PT E

D

MA

NU

580ºC in molten nitrates under dynamic conditions.

ACCEPTED MANUSCRIPT

18-20 % Al

FeAl

Kirkendall voids

a.

NU

01-073-8033 FeAl

4000

3000

MA

Intensity counts

RI

SC

AlN

PT

IZ

D

2000

PT E

1000

0

b.

20

30

40

50

60

70

80

90

100

110

120

2θ(º)

CE

Figure 1. a. Microstructure (IZ for Interdiffusion Zone) and b. XRD measurement of

AC

the as-annealed aluminide coating, (Al concentration is in wt.%).

ACCEPTED MANUSCRIPT

bisque Al2O3 Black features Ni and Al rich particles

44 % Al

Ni2Al3 1

PT

NiAl

2

3

RI

1. 63Ni-37Al 2. 70Ni-27.5Al-2.5Fe 3. 80Ni-15Al-5Fe

SC

a. 4

4. 73Ni-24.5Fe-2.5Cr 5. 45Ni-50Fe-5Cr 6. 8Ni-84Fe-8Cr

5

NU

03-065-9699 Al3Ni2

4000

MA

Intensity counts

6

3000

0

20

CE

b.

PT E

1000

D

2000

30

40

50

60

70

80

90

100

110

120

2θ(º)

Figure 2. a. Microstructure and b. XRD measurement of the as-annealed nickel-

AC

aluminide coating, (concentrations are in wt.%).

ACCEPTED MANUSCRIPT

25

P91 P91+NiAl P91+Al

15

10 5

0 250

500

750

Time (h)

4

NU

P91 P91+NiAl

2

MA

0 250

500

-2 -4

750

1000

D

Δm/S (mg/cm2)

P91+Al

0

SC

-5 a. Solar Salt - 580ºC - Static

1000

RI

0

PT

Δm/S (mg/cm2)

20

PT E

-6

-8

CE

b. Solar Salt - 580ºC - Dynamic

Time (h)

Figure 3. Weight changes of the studied coated and uncoated materials in Solar Salt at

AC

580ºC under a. static conditions and b. dynamic conditions.

ACCEPTED MANUSCRIPT

As tested 23.2

Descaled

20

PT

10

RI

0

-7.0

-10

-20

SC

Weight variations (mg/cm²)

30

-18.6

NU

Static

-18.6

Dynamic

Figure 4. Comparison of the weight changes of the uncoated P91 exposed 1000 h in

AC

CE

PT E

D

MA

molten nitrates after testing and descaling.

ACCEPTED MANUSCRIPT

Na-rich layer NaFeO2

RI

PT

Nitriding

NaFeO2

NU

SC

a. P91

MA

FeAl

PT E

D

Kirkendall voids

AC

CE

b. P91+Al

Ni2Al3

Ni and Al rich particles within the bisque

NiAl Ni-Al-Fe (+/-Cr) Fe-Ni-Cr

P91

c. P91+NiAl

Figure 5. Microstructure after 1000 h at 580ºC in static molten nitrates of the a. uncoated P91, b. aluminide coating and c. nickel-aluminide coating.

Intensity counts

ACCEPTED MANUSCRIPT 200

NaFeO2 Fe2O3 NaNO3

150

100

0 30

40

50

60

70

80

100

SC

750

120

2θ(º)

NaFeO2 Fe2O3 FeAl

NU

1000

110

MA

Intensity counts

1250

90

RI

20

a. P91

PT

50

500

0 30

40

50

60

70

80

90

100

110

120

2θ(º)

AC

CE

PT E

20

b. P91 + Al

D

250

Figure 6. X-ray diffraction patterns after 1000 h at 580ºC in static molten nitrates of the a. uncoated P91, b. aluminide coating and c. nickel-aluminide coating.

SC

RI

PT

ACCEPTED MANUSCRIPT

Figure 7. EDS mapping of the aluminide coating after 1000 h at 580ºC in static molten

AC

CE

PT E

D

MA

NU

nitrates.

AC

CE

PT E

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

Figure 8. Microstructure after 1000 h at 580ºC in dynamic molten nitrates of the a. uncoated P91, b. aluminide coating and c. nickel-aluminide coating.

AC

CE

PT E

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

Figure 9. X-ray diffraction patterns after 1000 h at 580ºC in dynamic molten nitrates of the a. uncoated P91, b. aluminide coating and c. nickel-aluminide coating.

SC

RI

PT

ACCEPTED MANUSCRIPT

Figure 10. EDS mapping of the surface of the nickel-aluminide coating after 1000 h at

AC

CE

PT E

D

MA

NU

580ºC in molten nitrates under dynamic conditions.

ACCEPTED MANUSCRIPT

AC

CE

PT E

D

MA

NU

SC

RI

PT

RESEARCH HIGHLIGHTS  The uncoated P91 developed a complex, non-protective and fast growing multilayered oxide scale which included Fe2O3, Fe3O4 and NaFeO2.  All the coated systems performed much better than the uncoated P91 up to at least 1000 h at 580ºC.  The nickel-aluminide had the better corrosion behavior amongst the studied systems.  The aluminide coating developed a thin Na ferrite scale whereas the nickelaluminide coating developed a NiAl2O4 oxide.