Thermochemical interactions between yttria‐stabilized zirconia and molten lunar regolith simulants - Journal of the American Ceramic Society - 2024 - Yu - Thermochemical interactions between yttria‐stabilized zirconia and.pdf

Received: 10 January 2024

Revised: 12 March 2024

Accepted: 16 March 2024

DOI: 10.1111/jace.19821

RESEARCH ARTICLE

Thermochemical interactions between yttria-stabilized

zirconia and molten lunar regolith simulants

Kevin Yu

Jamesa Stokes

Bryan Harder

Lorlyn Reidy

Katherine T. Faber

Division of Engineering and Applied

Science, California Institute of

Technology, Pasadena, California, USA

NASA Glenn Research Center,

Cleveland, Ohio, USA

NASA Marshall Space Flight Center,

Huntsville, Alabama, USA

Correspondence

Katherine T. Faber, Division of

Engineering and Applied Science,

California Institute of Technology,

Pasadena, CA, USA.

Email:

ktfaber@caltech.edu

Funding information

NASA Space Technology Graduate

Research Opportunity, Grant/Award

Number: 80NSSC22K1184; NASA

Aeronautics Research Mission

Directorate’s Transformational Tools and

Technologies Project

Abstract

Oxygen produced through in-situ resource utilization (ISRU) is critical to main-

taining a permanent human presence on the lunar surface. Molten regolith

electrolysis and carbothermal reduction are two promising ISRU techniques for

generating oxygen directly from lunar regolith, which is primarily a mixture

of oxide minerals; however, both processes require operating temperatures of

1600

◦

C to melt lunar regolith and dissociate the molten oxides. These condi-

tions limit the use of many oxide refractory materials, such as Al

and MgO,

due to rapid degradation resulting from reactions between the refractory mate-

rials and molten lunar regolith. Yttria-stabilized zirconia (YSZ) is shown here

to be a promising refractory oxide to provide containment of molten regolith

while demonstrating limited reactivity. This work focuses on corrosion studies

of YSZ powders and dense YSZ crucibles in contact with molten lunar maria

and highlands regolith simulants at 1600

◦

C. The interactions between YSZ and

molten regolith were characterized using scanning electron microscopy/energy

dispersive spectroscopy, X-ray diffraction, and electron backscatter diffraction.

A FactSage thermochemical model was created for comparison with the exper-

imental results. These combined analyses suggest that lunar maria regolith will

degrade the YSZ faster than the lunar highlands regolith due to the lower viscos-

ityofthemariaregolith.Thefeasibilityoflong-termmoltenregolithcontainment

with YSZ is discussed based on the YSZ powder and crucible results.

KEYWORDS

anorthite, coarsening, diffusion/diffusivity, electron microscopy, lunar regolith, yttria-

stabilized zirconia

INTRODUCTION

In 2019, NASA announced the Artemis program, with a

long-term goal of establishing a sustainable human pres-

ence on the lunar surface.

This has led to a renewed

focus on in-situ resource utilization (ISRU) technolo-

gies, which are techniques to extract and process extra-

terrestrial resources to minimize the costs and mass

of launch vehicles for lunar missions.

Specifically, the

production of oxygen by ISRU processes has garnered

significant attention due to oxygen serving as an oxi-

dizer in chemical propulsion systems.

The top 10 meters

of the lunar surface is a layer of granular oxide min-

erals and glasses that is known as the lunar regolith.

J Am Ceram Soc.

2024;1–13.

wileyonlinelibrary.com/journal/jace

YU et al.

This layer is approximately 40% oxygen by weight, mak-

ing it an ideal feedstock for oxygen production because

its abundance eliminates the need for extensive resource

prospecting.

4,5

There are several high-temperature ISRU processes

in development that target oxygen generation from

regolith, including carbothermal reduction,

6–8

hydrogen

reduction,

9–11

molten salt electrolysis,

12–14

and molten

regolith electrolysis (MRE).

15–17

Of these, MRE is an attrac-

tive option since it does not require additional reactants

that need to be resupplied from Earth, such as carbon (for

carbothermal reduction), hydrogen (for hydrogen reduc-

tion), or salt (for molten salt electrolysis).

In MRE, a

cathode and anode are inserted into molten lunar regolith,

producing molten metals at the cathode surface and oxy-

gen at the anode. In order to fully melt and dissociate the

lunar regolith, MRE requires a high operating temperature

(1600

◦

C), which presents challenges for selecting appro-

priate refractory containment materials.

Pure oxygen is

generated within an MRE reactor, so materials in contact

with the regolith must not oxidize under these operating

conditions, which further limits the materials that can be

used for containment.

Platinum group metals have traditionally been used

for containment of molten aluminosilicate-rich slags due

to their relative stability under corrosive and oxidizing

conditions.

In addition to their high cost and scarcity,

these metals also form eutectics with the molten silicon

produced in MRE, precluding their use as a containment

material for MRE.

Kim et al. investigated the use of

iridium as an anode for MREs in different compositions

of slag; the iridium is only used at the anode and does

not contact the molten silicon produced at the cathode,

avoiding the formation of eutectics between iridium and

silicon.

However, the authors found that the composition

of the slag, and its subsequent optical basicity, had a sig-

nificant effect on the corrosion rate of iridium. The more

acidic slag (higher in silica) resulted in a roughly order-

of-magnitude reduction in the corrosion rate of iridium

compared to the more basic slag (higher in calcia). Based

on these results, variations in the composition of molten

lunar regolith should result in differences in the corrosion

rates of containment materials.

Standish et al. investigated refractory ceramics, includ-

ing alumina, zirconia (calcia-, magnesia-, and yttria-

stabilized), and magnesia plates, in contact with molten

lunar regolith simulants at 1600

◦

C and concluded that

alumina exhibited the least degradation.

However, the

porosity of the ceramics used in that study varied sig-

nificantly, making it difficult to compare the degrada-

tion between ceramics. Rosenberg et al. studied crucible

materials for carbothermal reduction of lunar regolith

simulants at 1650

◦

C, including alumina, magnesia, yttria-

stabilized zirconia (YSZ), yttria, and tantalum carbide;

they concluded that yttria and YSZ were the only suitable

materialsforcontainingmoltenlunarregolithsimulants.

Both of the aforementioned studies did not investigate

the degradation mechanisms of these refractory ceram-

ics in contact with molten regolith simulants and focused

only on evaluating these materials from an engineering

perspective, thereby motivating this investigation.

The current study sought to elucidate the degradation

mechanisms of YSZ in contact with two representative

lunar regolith simulants and evaluate the ability of YSZ

to contain these molten simulants. YSZ was selected due

to its high melting point and common usage as a crucible

material. The thermochemical equilibrium state of YSZ in

molten lunar regolith simulants was determined through

YSZ powder experiments. YSZ crucible experiments fur-

ther extended these results by introducing microstructural

effects to the system. Ultimately, this study should inform

and guide future material design for molten regolith

containment applications.

EXPERIMENTAL PROCEDURES

AND MATERIALS

2.1

Lunar regolith simulants

Two lunar regolith simulants were selected to represent

the two geologically distinct regions of the lunar surface—

the lunar highlands and the maria. The lunar highlands

are brighter regions of the lunar surface and are compo-

sitionally dominated by feldspar, while the lunar maria

are the darker regions of the lunar surface and are mostly

comprised of basaltic lavas and pyroxene.

LHS-1 (Space

Resource Technologies) was selected as the highlands

regolith simulant and JSC-1A (NASA) was selected as the

maria regolith simulant.

24,25

The compositions of both

regolith simulants are summarized in Table

1.TheSiO

content of both simulants is similar, with JSC-1A hav-

ing a higher concentration of FeO and MgO and a lower

concentration of CaO and Al

compared to LHS-1.

2.2

Powder experiments

YSZ (11 wt%/6.3 mol% Y

) powder (Hermann C. Starck)

was mixed individually with each regolith simulant on a

ball mill in ethanol using 10 mm cylindrical YSZ media for

16 h. Four mixtures were produced: two at a 25:75 molar

ratio of simulant to YSZ and two at a 50:50 molar ratio

of simulant to YSZ. Approximately 1 g of the resulting

mixed powders was uniaxially pressed into 12.7 mm diam-

eter pellets. The pellets were placed in platinum crucibles

and heat treated at a ramp rate of 10

◦

C/min to 1600

◦

Cin

air and held at that temperature for 3 h. After the heat

15512916, 0, Downloaded from https://ceramics.onlinelibrary.wiley.com/doi/10.1111/jace.19821 by California Inst of Technology, Wiley Online Library on [14/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

YU et al.

TABLE 1

Manufacturer-provided compositions of lunar regolith simulants in wt%.

24,25

Species

SiO

FeO

MgO

CaO

TiO

JSC-1A (maria)

48.7

11.0

9.2

10.6

15.3

1.6

2.8

0.8

LHS-1 (highlands)

51.8

2.7

1.6

12.9

26.9

0.6

2.9

0.8

treatments, the pellets were air quenched using a cool-

ing fan, reaching room temperature after approximately 10

min. The samples were then cross-sectioned in half, with

one half polished for scanning electron microscopy (SEM)

analysis. The other half was crushed and ground in an alu-

minamortarandpestleforpowderX-raydiffraction(XRD)

analysis.

XRD analysis (Cu K-

source, D8 Advance, Bruker) was

performed on the starting YSZ powder, both regolith sim-

ulants prior to mixing, the YSZ/simulant mixtures, and

the final reground, heat-treated YSZ/simulant mixtures.

SEM (1550VP FESEM, ZEISS) with a backscattered elec-

tron (BSE) detector was used to observe thermochemical

changes in the heat-treated mixtures. For more accu-

rate compositional analysis, electron-probe microanalysis

(EPMA)(JXA-iHP200F,JEOL)wasperformedontheheat-

treated samples using standards for Si, Al, Ti, Ca, K, Na,

Mg, Fe, Mn, Zr, Hf, and Y. Oxygen content was obtained

based on stoichiometry, assuming that all species were

fully oxidized.

2.3

Crucible experiments

YSZ (14 wt%/8.2 mol% Y

) crucibles (40 mm OD

40 mm tall) were obtained from Edgetech Industries. Each

crucible was loaded with approximately 30 g of LHS-1

or JSC-1A, filling half of the available volume. Alumina

lids, coated with calcia-stabilized zirconia (ZYP Coatings),

were used to prevent the regolith simulants from spilling

over the top of the crucibles. Both crucibles were heat

treated in an alumina tube furnace under flowing argon.

The samples were ramped at 2

◦

C/min to 1600

◦

C, held

for 3 h, and then cooled back to ambient temperature

at 2

◦

C/min. Both regolith simulants melted into a glassy

phase during the heat treatment and were bonded to the

YSZ crucibles. The entire assembly was mounted in epoxy,

cross-sectioned to the center of crucible, and polished for

SEM analysis. Optical microscopy (VHX-2000, Keyence)

was performed on the polished cross-sections to observe

macroscopic (

1 mm) phenomena of the YSZ, solidified

regolith simulants, and the YSZ/simulant interface. SEM

and energy dispersive spectroscopy (EDS; X-Max EDS,

OxfordInstruments)wereusedtoassessthecompositional

and microstructural changes at the YSZ/simulant inter-

face. Electron backscatter diffraction (EBSD; HKL EBSD)

was employed to evaluate the phases and grain sizes of

YSZ at the YSZ/simulant interface. MTEX, an EBSD analy-

sis software package, was used for segmentation and grain

size analysis of the EBSD maps to provide statistical data

on grain sizes.

2.4

FactSage thermochemistry model

A FactSage model was created to thermodynamically pre-

dict possible reaction phases and estimate Zr solubility at

1600

◦

C in both molten regolith simulants using the FToxid

database.

Y was not included in this analysis since it is

not optimized in the FToxid database. Both regolith sim-

ulant systems were modeled using the oxide chemistry

provided in Table

1. The Equilib module was used to find

the Zr solubility in both regolith simulants. The amount

of ZrO

was added to the regolith compositions as a vari-

able term, with tetragonal ZrO

set as a formation target.

FactSage calculations then provided the amount of ZrO

needed to form the tetragonal ZrO

phase. The Equilib

module was also used to identify any phases, other than

a slag phase, that might precipitate from the regolith sim-

ulant compositions. Finally, the viscosity module was used

with the melt database to estimate the viscosity of both

regolith simulants at 1600

◦

RESULTS AND DISCUSSION

3.1

Powder experiments

XRD spectra for the starting regolith simulant and YSZ

powders, the mixed YSZ/simulant powder (50:50 molar

ratio), and the heat-treated YSZ/simulant powder (50:50

molar ratio) are presented in Figure

1. Both regolith sim-

ulants contain anorthite (Ca

) as the dominant

crystalline phase based on the starting regolith simulant

XRD spectra. The initial YSZ powder is primarily a mix-

ture of the cubic and tetragonal zirconia crystal structures.

After ball milling to mix the YSZ and simulant powders, a

small anorthite peak (28

◦

) was observed in both mixtures,

indicating that the powders were well mixed. The lower

intensity of this anorthite peak is attributed to the lower

atomic scattering factor of anorthite’s constituent cations

(Ca, Si) compared to that of YSZ’s cations (Y, Zr).

After heat treatment and quenching, the YSZ/simulant

XRD patterns no longer contain the anorthite peak at 28

◦

YU et al.

FIGURE 1

(A) X-ray diffraction (XRD) spectra of LHS-1 (highlands), starting yttria-stabilized zirconia (YSZ) powder, and YSZ/LHS-1

mixture (50:50 molar ratio) before and after heat treatment. (B) XRD spectra of JSC-1A (maria), starting YSZ powder, and YSZ/JSC-1A mixture

(50:50 molar ratio) before and after heat treatment. Spectra show presence of anorthite (

■

), tetragonal YSZ (

●

), and cubic YSZ (

▲

FIGURE 2

Backscattered electron (BSE) micrograph of (A) heat-treated, 50:50 molar ratio yttria-stabilized zirconia (YSZ)/LHS-1

(highlands) and (B) heat-treated, 50:50 molar ratio YSZ/JSC-1A (maria).

indicating that the regolith simulant had melted in both

samples, and remained amorphous through solidification

due to rapid quenching. In both samples, the original

YSZ peaks (mixture of tetragonal and cubic YSZ) are still

present, with no additional crystalline phases observed.

TheseXRDscansindicatethatYSZdidnotformadditional

crystallinereactionproductswithexposuretoeitherLHS-1

or JSC-1A.

SEM micrographs of the 50:50 molar ratio heat-treated

YSZ/LHS-1 and YSZ/JSC-1A samples, obtained using a

BSE detector, are shown in Figure

2. There are two distinct

phasesinbothsamples—abrighterYSZphaseandadarker

simulantglassmatrix.Theseresultsareconsistentwiththe

XRD spectra of the heat-treated samples in that YSZ is the

only observed crystalline phase. Since no reaction phase

wasdetectedbyXRDandSEM,dissolution/reprecipitation

is predicted to be the dominant degradation mechanism in

both YSZ/simulant samples. The YSZ particles in the JSC-

1A system are larger than those in the LHS-1 system. This

difference in particle size is attributed to the ball milling

used to mix these powders. The LHS-1 system contained

a higher content of harder minerals such as anorthite and

may cause a faster rate of YSZ particle size reduction com-

pared to JSC-1A. EPMA compositional analysis (Table

confirms that both sets of YSZ/LHS-1 and YSZ/JSC-1A

powder samples were at equilibrium. Thus, the difference

YU et al.

TABLE 2

Electron-probe microanalysis concentrations of Zr and Y in glassy matrix and yttria-stabilized zirconia (YSZ) particles of

heat-treated YSZ/simulant samples.

Phase

Molar ratio

(simulant:YSZ)

LHS-1 (highlands)

JSC-1A (maria)

Zr at%

Y at%

Zr at%

Y at%

Glassy matrix

25:75

1.09 (0.06)

1.50 (0.04)

1.27 (0.16)

1.78 (0.03)

50:50

1.38 (0.18)

1.27 (0.02)

1.65 (0.44)

1.38 (0.04)

YSZ particles

25:75

29.36 (0.25)

4.09 (0.35)

29.24 (0.09)

3.86 (0.02)

50:50

29.72 (0.15)

3.53 (0.18)

29.34 (0.20)

3.39 (0.33)

Nominal composition

30.00

4.00

30.00

4.00

Note

: Standard deviations are presented in parentheses.

inYSZparticlesizesisnotbelievedtobesignificantforthis

analysis.

The dissolution should result in a measurable concen-

tration of Zr and Y within the glassy matrix. Additionally,

metal ions, such as Ca, Mg, Fe, and Ti, in the molten

regolith simulant are soluble in YSZ and are likely to dif-

fuse into the YSZ particles at 1600

◦

C. Low concentrations

(

0.5 at%) of these cations are detected in the YSZ parti-

cles. EPMA compositional analysis of Zr and Y in both the

glassy matrix and YSZ particles is provided in Table

2.The

nominal composition of the starting YSZ powder is also

provided in Table

2. Full EPMA compositions of both the

glassy matrix and YSZ particles are provided in Supporting

Information S1.

The measured concentrations of Zr and Y in the glassy

matrix are in close agreement between the 25:75 and 50:50

molar ratio samples for both regolith simulants, indicat-

ing that the samples have reached solubility equilibrium.

Based on the FactSage results, it is determined that the

system would not produce any high-temperature reaction

phases, which is consistent with the XRD (Figure

1)and

SEM (Figure

2) results. Additionally, FactSage predictions

include a 1.96 and 1.95 at% Zr solubility for LHS-1 and JSC-

1A, respectively, which is consistent with the measured

solubilities of Zr. Based on combined experimental results,

the two regolith simulants have similar solubilities for Zr

and Y, with molten JSC-1A having marginally higher solu-

bilities than molten LHS-1. The solubilities of Zr in Table

are consistent with those reported in literature for pure

ZrO

(no Y

) in aluminosilicate glasses at similar tem-

peratures, which range from 1.3 to 3.3 at%, depending on

the composition of the glass.

29,30

The measured Y concentrations in the glassy matrix for

each sample are also similar to the Zr concentrations even

though the starting YSZ material is only 8.2 mol% Y

Since YSZ dissolution should result in a concentration of

Zr and Y that maintains the same ratio as the starting

material, it is clear that YSZ dissolution is not the only

degradation mechanism; Y must also be diffusing out of

the YSZ. The measured Y concentration in the YSZ par-

ticles decreased with increasing regolith simulant, which

supports the idea of Y diffusion out of the starting YSZ

material. As more molten regolith is available for Y to dif-

fuse into, more Y is depleted from the YSZ particles until

the Y solubility limit is reached.

Guo et al. investigated the effect of varying optical

basicity of an oxy-fluoride flux to understand Y depletion

from YSZ. Their results indicate that Y depletion will

occur when the optical basicity (

) of the flux is lower

than that of Y

(

0.72).

The optical basicities of

LHS-1 and JSC-1A were calculated based on the compo-

sitions presented in Table

1, resulting in

LHS-1

0.57 and

JSC-1A

0.60. Thus, both regolith simulants are lower

(more acidic) in optical basicity than Y

; therefore, Y

depletion from YSZ is expected to occur in the presence of

both molten regolith simulants. The observed Y depletion

behavior in an oxide melt is consistent with previously

reported results in an oxy-fluoride melt.

The Y depletion from YSZ in turn increases the Y

content in both molten regolith simulants, resulting in

comparable solubilities for Zr and Y, as shown in Table

The similarity of Zr and Y solubilities observed is consis-

tent with the observations of Krämer et al. at 1300

◦

However, this combined YSZ dissolution and Y depletion

from YSZ, in both cases, appears to occur at temperatures

higher than the liquidus temperature of the glass. In con-

trast, Zhao et al. investigated YSZ degradation at 1250

◦

just above the liquidus temperature of the aluminosilicate

glass used in their system and observed the formation of Y-

enriched YSZ instead of Y depletion into the glass phase.

This difference is attributed to the temperature at which

these experiments were performed.

3.2

Crucible experiments

3.2.1

Bulk regolith behavior

Optical and scanning electron microscopy was performed

on the cross-sectioned YSZ crucible samples and is pre-

sented in Figure

3. The solidified LHS-1 in the YSZ crucible

(Figure

3A) contains minimal microstructural features,

YU et al.

FIGURE 3

Optical and backscattered electron (BSE) images of the cross-sectioned heat-treated (A) yttria-stabilized zirconia

(YSZ)/LHS-1 (highlands) crucible and (B) YSZ/JSC-1A (maria) crucible at 1600

◦

Cfor3h.

with most of the solidified LHS-1 appearing dark and

glassy. YSZ dendrites grow from the bottom YSZ/LHS-1

interface. Other than the dendrite formation, optical

and electron microscopy did not reveal any additional

YSZ particles within the bulk of the solidified LHS-1. In

contrast to LHS-1, the solidified JSC-1A exhibited more

microstructural features within the bottom half of the

sample. The region closest to the YSZ/JSC-1A interface

contains larger and coarser dendrites compared to the

YSZ/LHS-1 interface. The dendrite size decreases with

increasing distance from the YSZ/JSC-1A interface. An

optical micrograph is inset in Figure

3B to highlight the

dendritic features spanning roughly the bottom half of the

sample. The top half of this sample is similar to the LHS-1

crucible in that there are no observed YSZ dendrites in

the solidified JSC-1A. EDS analysis of different regions in

the solidified regolith indicates that the Zr and Y contents

are not homogenous and are generally higher near the

YSZ interface, suggesting that equilibrium has not been

reached in these crucible experiments, unlike the YSZ

powder experiments described in Section

3.1.

YSZ dendrite formation in both regolith simulants was

hypothesized to be caused by the slow cooling of the cru-

cibles from 1600

◦

C to ambient temperature compared to

thatofthequenchedpowdersamplesdescribedpreviously.

YSZ dissolves into the molten regolith simulants at 1600

◦

according to the solubility limits observed in the powder

experiments. As the temperature of the sample is slowly

decreased, the solubility limit of Zr and Y decreases until

YSZ nucleated in the molten regolith simulant, forming

dendrites in the process. The larger number of dendrites

observed in the JSC-1A sample suggests that more Zr and

Y have dissolved into JSC-1A as compared to LHS-1. There

is also higher localized corrosion at the triple boundary

between the YSZ crucible, molten regolith simulant, and

argon. This type of corrosion is accelerated by Marangoni

convection at this interface, which has been observed in

other experiments with molten lunar regolith simulants.

The Marangoni-mediated corrosion in the JSC-1A sample

is also more severe than that in the LHS-1 sample. Both

results suggest that the dissolution rate of YSZ in JSC-1A is

higher than the dissolution rate of YSZ in LHS-1.

The powder experiments (Table

2) revealed comparable

solubility limits of Zr and Y in molten JSC-1A and LHS-1

at equilibrium, but a higher dissolution rate of YSZ in JSC-

1A was observed in the larger-scale crucible experiments.

This behavior is attributed to the difference in viscosities

between JSC-1A and LHS-1. From the FactSage viscosity

model, approximately an order-of-magnitude difference

between the two molten regolith simulants (without dis-

solved Zr or Y) at 1600

◦

Ciscomputed(

JSC-1A

:0.7Pa-s,

LHS-1

: 6.7 Pa-s). This is likely due to the higher content

YU et al.

of silica network modifiers such as FeO and MgO in JSC-

1A compared to LHS-1.

Humbert et al. modeled regolith

thermophysical properties for measured compositions of

lunarregolithat1850

◦

C and predicted a similar order-of-

magnitude increase in viscosity for the highlands (LHS-1)

compared to the maria (JSC-1A).

Kato and Araki investigated the effect of glass compo-

sition on the dissolution behavior of zirconia at 1500

◦

andfoundsimilarresults;theoveralldissolution/corrosion

loss of zirconia was higher with decreasing molten glass

viscosity.

Additionally, they discovered that the dis-

solution rate of zirconia is controlled by the rate of

dissolved Zr transport into the bulk molten glass. The

Noyes‒Whitney‒Nernst equation can be used to describe

dissolution dominated by solute (Zr) diffusion away from

the dissolving surface:

푗=

퐷

훿

∗

(

퐶

푖

−퐶

∞

)

(1)

where

is the dissolution flux,

is the diffusivity of the

dissolved species,

is the concentration of the dissolved

species at the interface,

∞

is the concentration of the

dissolved species in the bulk molten glass, and

is the

diffusion layer thickness, which is defined as:

훿

∗

퐶

푖

−퐶

∞

(

푑퐶∕푑푦

)

푖

where (

)

is the concentration gradient at the inter-

face and

is the position along the direction normal to

the interface.

For a given temperature and solute, the

Stokes‒Einstein equation, relating viscosity to diffusivity,

can be reduced to:

퐷휂 = 푘

(2)

where

is the viscosity of the glass and

is a constant that

varies with temperature and size of the diffusing species.

By combining the previous equations, the following rela-

tion is obtained:

푗=

푘

훿

∗

휂

(

퐶

푖

−퐶

∞

)

(3)

Based onthepowder experiments,

(solubility concen-

tration of Zr) is similar for JSC-1A and LHS-1, while

∞

is zero for both regolith simulants. This implies that

also similar between the two regolith simulants at the start

of dissolution. The constant

in the context of these cru-

cible experiments is the same for JSC-1A and LHS-1 since

the diffusing species and temperature are identical. Thus,

changes in the dissolution flux

are dominated by the vis-

cosity term and

is inversely proportional to the viscosity

of the melt. For a glass-forming melt, the Stokes‒Einstein

relation can become inaccurate due to the coexistence of

various silicon‒oxygen complexes in the glass, and a more

suitable model can be expressed as:

퐷

푚

휂=푘

(4)

where

is a positive coefficient and is the ratio of the acti-

vation energies of viscous flow and diffusion.

However,

the same overall trend applies in that the dissolution flux

will be dependent on the inverse of

raised to a positive

exponent.

As a result of this analysis and given the lower viscos-

ity of JSC-1A compared to LHS-1, the dissolution rate of

YSZ in JSC-1A is expected to be higher than that of YSZ

in LHS-1. The greater volume of recrystallized dendrites

observed in the JSC-1A sample indicates that more YSZ

is dissolved into JSC-1A compared to the LHS-1 sample,

which is consistent with the expected dissolution behavior

of the molten regolith simulants based on their viscosi-

ties. The composition of the lunar regolith simulant, and

resulting viscosity, has a significant effect on the overall

corrosion of the YSZ crucible. This suggests that lunar

maria regolith, which has a higher concentration of net-

work modifiers, will degrade crucibles faster than lunar

highlands regolith.

3.2.2

YSZ/regolith interfacial behavior

To better understand the microstructural and composi-

tional changes that occur due to interactions between the

YSZ and molten regolith simulants, further analysis of

the YSZ/simulant interface was required. Combined SEM

images, EDS line scans, and EBSD maps were collected at

the bottom YSZ/simulant interface, over a width of 200

μ

and extending approximately 650

μ

m into the YSZ. These

results are presented in Figure

4. Both samples contain

three distinct regions based on the changes in Y content

in EDS line scans and are demarcated by horizontal green

lines. These three areas will be referred to as the repre-

cipitation region (top), surface region (middle), and bulk

region (bottom) in the upcoming discussions.

The reprecipitation region (top) exhibits decreased Y

content with respect to the starting YSZ crucible mate-

rial (5.2 at% Y) and is approximately 100‒150

μ

m thick for

both regolith simulants. EBSD phase mapping indicates

that this region contains mostly monoclinic YSZ as well

as pockets of amorphous, solidified regolith simulant. A

dissolution‒reprecipitation mechanism is responsible for

producing the monoclinic YSZ phase. As the regolith sim-

ulants melt, YSZ begins to dissolve from the crucible wall

until the local molten regolith environment is saturated

with Zr. At this point, Y-depleted (

∼

1 at% Y) YSZ particles

YU et al.

FIGURE 4

Combined energy dispersive spectroscopy (EDS)/scanning electron microscopy (SEM)/electron backscatter diffraction

(EBSD) images of (A) yttria-stabilized zirconia (YSZ)/LHS-1 (highlands) interface (crucible bottom) and (B) YSZ/JSC-1A (maria) interface

(crucible bottom).

YU et al.

begin to reprecipitate into smaller grain sizes compared

to those in the original YSZ crucible. The effect on cool-

ing is exacerbated as the solubility of Zr decreases with

temperature,andahigherdegreeofundercoolingisexperi-

enced.Thegrainsindirectcontactwiththemoltenregolith

simulants demonstrate cellular growth that eventually

transitions to dendrites, as observed in Figure

3. These

Y-depleted particles are tetragonal at high temperature,

but do not contain sufficient Y to stabilize the tetrago-

nal phase and ultimately transform into the monoclinic

phase on cooling. This effect has been observed in YSZ

interactions with molten silicates at temperatures lower

than 1600

◦

30,38

The tetragonal-to-monoclinic transfor-

mation causes significant stress on the crucible walls and

likely gives rise to the crucible cracks that are observed in

Figure

3. Based on the measured Y content in the mono-

clinic region, the tetragonal-to-monoclinic transformation

is expected to occur around 1100

◦

C, while the liquidus

temperature of the regolith simulants is predicted to be

between 1450

◦

C and 1550

◦

18,39

Thus, it is concluded

that the cracks in the YSZ crucible form after the molten

regolith simulant has solidified, which is supported by the

lack of molten regolith penetration along the majority of

cracks in the YSZ crucible wall.

The surface region (middle) is the interior of the YSZ

crucible wall after the initial dissolution of YSZ occurs and

the molten regolith near the interface is locally saturated

with Zr and Y. For both LHS-1 and JSC-1A, this region

exhibits an increase in Y content (from top to bottom) that

eventuallyreturnstothenominalYcontentoftheYSZcru-

cible. EDS line scans elucidate the diffusion profile of Y

from the bulk YSZ into the molten regolith simulant. This

is consistent with the previous EPMA powder results that

also exhibited Y depletion in the YSZ particles (Table

2). In

additiontothegreateramountofYSZdissolutionobserved

withJSC-1AthanwithLHS-1,moreYdepletionisobserved

in the YSZ exposed to JSC-1A. The lower viscosity of JSC-

1A likely allowed for faster transport of Y out of YSZ and

into the bulk of the molten JSC-1A, resulting in a more

extensive Y depletion region compared to that in the LHS-1

sample. In contrast to the reprecipitation region, the thick-

ness of the surface region varies between the two regolith

simulants and is approximately 100

μ

m for LHS-1 and 380

μ

m for JSC-1A. EBSD phase mapping of the surface region

reveals that this region is largely a mixture of cubic and

tetragonal YSZ, with no monoclinic YSZ content. Addi-

tionally,theYSZgrainsinthesurfaceregionarelargerthan

those in the bulk region below it, suggesting that interac-

tions between YSZ and both molten regolith simulants are

accelerating grain coarsening, which is more pronounced

in the case of JSC-1A.

The bulk region (bottom) is representative of the

remaining YSZ crucible material. The Y level is consistent

with the nominal Y content of the starting YSZ crucible

material and does not vary in the EDS line scan. Similar

to the surface region, EBSD mapping of the bulk region

reveals that it is also a mixture of cubic and tetragonal

YSZ, with no monoclinic YSZ content. EBSD scans on

the exterior crucible wall (

∼

2.5 mm below the regions in

Figure

4) for each crucible (not shown here) are consistent

with the bulk region of each crucible, allowing the desig-

nated bulk region to be used as a baseline for grains that

are far from the YSZ/simulant interface and unaffected by

interactions with the molten regolith simulant. A compar-

ison of YSZ grain sizes in the bulk region and the exterior

crucible for both LHS-1 and JSC-1A is provided in Sup-

porting Information S2. There is a difference in the bulk

region grain sizes between JSC-1A and LHS-1, but this dif-

ference is attributed to manufacturing variations during

YSZ crucible production.

Using the bulk region as a baseline, it is clear that

YSZ grain coarsening had occurred in the surface region

for both regolith simulants. Grain segmentation and size

analysis were performed on the grains within surface and

bulk regions of both samples. Grains intersecting the map

boundaries were not included to avoid skewing the data

based on arbitrary boundary locations. An area-weighted

histogram of the equivalent grain diameters is presented

in Figure

Grain coarsening occurs in both samples as shown in

Figure

5. For both regolith simulants, there is a decrease in

the fractional area of smaller grains and an increase in the

fractional area of larger grains in the surface region com-

pared to the bulk region. In the YSZ/LHS-1 bulk region,

grains larger than 30

μ

mmakeup

∼

10% of the area, while

in the surface region, grains larger than 30

μ

mmakeup

∼

55% of the area. Additionally, there is a clear decrease in

the fractional area of all grain sizes smaller than 30

μ

in the surface region compared to the bulk region. In the

YSZ/JSC-1A bulk region, there are no grains larger than 25

μ

m (0% area), but in the surface region, grains larger than

μ

mmakeup

∼

50% of the area. Similar to the YSZ/LHS-

1 crucible, there is a decrease in the fractional area of all

grain sizes smaller than 15

μ

m in the surface region com-

pared to the bulk region. These results are consistent with

the visibly larger grains in the surface regions of Figure

and indicate that grain coarsening was accelerated in the

surface region of both YSZ crucibles. However, in com-

parison to LHS-1, JSC-1A causes grain coarsening to occur

more quickly.

Since the bulk region of YSZ grains underwent the same

thermal treatment as the surface region, it is imperative

to consider mechanisms that facilitate grain coarsening

locally at the YSZ/simulant interface. Large pockets of

regolith simulant in the surface region of either crucible

are not detected in EDS scans. This suggests that the

YU et al.

FIGURE 5

Area-weighted histogram of grain diameters (5

μ

m bins) for (A) LHS-1 (highlands) and (B) JSC-1A (maria).

regolith simulant penetrates the YSZ crucible wall along

theYSZgrainboundaries.Thismechanismofgrainbound-

ary dissolution and penetration has been observed in

YSZ interactions with molten silicates.

38,40,41

The molten

regolithsimulantincreasesthediffusivityofZrandYalong

grain boundaries, accelerating the rate at which grains can

coarsencomparedtothebulkoftheYSZ.Furthermore,the

YSZ/JSC-1A sample exhibits a higher degree of coarsen-

ing, almost certainly due to its lower viscosity at 1600

◦

The diffusivities of Zr and Y are expected to be higher

in JSC-1A (Section

3.2.1), so YSZ grain coarsening should

also occur faster than in LHS-1 as a result. Additionally,

the lower viscosity of JSC-1A results in a higher degree

of penetration into the YSZ material, allowing coarsening

to occur deeper into the YSZ material compared to LHS-1

material.

The grain coarsening behavior in both samples has

important implications for the long-term degradation

behavior of YSZ in contact with molten regolith simu-

lants. If the molten regolith simulant is dissolving and

penetrating along grain boundaries, then grain coarsening

should slow the rate of penetration by reducing the num-

ber of grain boundaries at the YSZ/simulant interface.

Kowalski et al. investigated the effect of YSZ grain size on

molten silicate degradation and found that larger grain

sizes can significantly slow the rate of penetration into

YSZ; they attributed this behavior to a change in grain

boundary density, width, and shape associated with the

change in YSZ grain size.

Additionally, the authors

concluded that for a finite amount of molten silicate

and given enough time, the depth of penetration will be

equal regardless of the grain size and grain boundary

morphology. In containment applications, the amount of

molten lunar regolith will be much larger compared to

that of the YSZ crucible walls, so the molten lunar regolith

will eventually penetrate the YSZ crucible wall and cause

crucible failure. However, the results of Kowalski et al.

suggested that it is possible that YSZ grain coarsening can

slow the degradation rate of YSZ and extend the usable

life of a YSZ crucible for containing molten lunar regolith.

CONCLUSION AND IMPLICATIONS

High-temperature YSZ degradation by molten lunar

regolith simulants was evaluated by powder and crucible

experiments. Two different lunar regolith simulants were

tested,representingthehighlands(LHS-1)andmaria(JSC-

1A) regions of the lunar surface. Based on the YSZ powder

experiments, both molten lunar regolith simulants had

a comparable equilibrium solubility for Zr and Y, with

JSC-1A having marginally higher solubility for Zr and Y.

Two degradation mechanisms were observed in the pow-

der experiments—YSZ dissolution and Y depletion in the

YSZ. Based on YSZ crucible experiments, greater amounts

ofrecrystallizedYSZwereobservedinthesolidifiedJSC-1A

regolith simulant than in LHS-1, indicating that more YSZ

had dissolved. However, this increase in YSZ dissolution

wasnotfullyaccountedforbythemarginaldifferenceinZr

and Y solubilities between LHS-1 and JSC-1A. Instead, the

higher degradation observed with JSC-1A stemmed from

its lower viscosity compared to LHS-1, which was credited

to higher FeO and MgO contents that disrupted the sil-

ica network at high temperatures. In addition to causing

a higher YSZ dissolution rate, the lower viscosity of the

JSC-1AallowedittopenetratefurtherintotheYSZcrucible

YU et al.

wall, enabling significant grain coarsening deeper into the

sample.

Overall, YSZ crucibles successfully contained both

molten lunar regolith simulants at 1600

◦

Cfor3h,but

ultimately cracked on cooldown. This cracking behavior

is attributed to the formation of Y-depleted YSZ at the

YSZ/simulant interface which undergoes a phase transfor-

mationfromtetragonaltomonoclinicphases.Basedonthe

crucible experiments, YSZ may be well suited for contain-

ment or contact with molten lunar highlands regolith for

longer times than explored in this work since the LHS-1

regolith simulant suffered less degradation than JSC-1A.

However, theformation of Y-depleted YSZ and subsequent

transformation-induced cracking may limit the use of YSZ

in single-use applications to support oxygen extraction

ISRU processes. The ultimate goal of lunar ISRU for oxy-

gen extraction is to scale up to continuous oxygen produc-

tion, in which case, YSZ may be a promising containment

material for high-temperature processes involving molten

lunar regolith, such as MRE and carbothermal reduction.

ACKNOWLEDGMENTS

This work was supported by the NASA Space Tech-

nology Graduate Research Opportunity (grant no.

80NSSC22K1184) for K. Yu, K.T. Faber, and L. Reidy

and NASA Aeronautics Research Mission Directorate’s

Transformational Tools and Technologies Project for J.

Stokes and B. Harder. The authors gratefully acknowledge

Dr. Chi Ma of the Caltech Geology and Planetary Sciences

Analytical Facility for his assistance with EDS, EBSD,

and EPMA analysis and Dr. Richard Rogers of the NASA

Glenn Research Center X-ray Characterization Laboratory

for his assistance with XRD data collection and analysis.

The authors would like to thank Dr. Gustavo Costa and Dr.

Nate Jacobson for their helpful discussions and assistance

with FactSage modeling. The authors would also like

to thank the NASA Jet Propulsion Laboratory Thermal

Energy Conversion Group for the use of their microscopy

facilities.

ORCID

KevinYu

https://orcid.org/0000-0003-3130-4309

JamesaStokes

https://orcid.org/0000-0002-0675-9988

BryanHarder

https://orcid.org/0000-0002-5304-5267

KatherineT.Faber

https://orcid.org/0000-0001-6585-

2536

REFERENCES

1. Cassel AM. Forward to the moon: NASA’s strategic plan

for human exploration. 2019.

https://ntrs.nasa.gov/citations/

20190032000

2. Crawford IA. Lunar resources: a review. Prog Phys Geog.

2015;39(2):137–67.

https://doi.org/10.1177/0309133314567585

3. Kleinhenz J E. In-situ resource utilization options for the moon

and mars. SEG technical program expanded abstracts 2020.

Virtual: Society of Exploration Geophysicists; 2020. p. 3527–29.

https://doi.org/10.1190/segam2020-3426725.1

4. Heiken G, Vaniman D, French BM, editors. Lunar sourcebook:

a user’s guide to the moon. Cambridge (England); New York:

Cambridge University Press; 1991.

5. Guerrero-Gonzalez FJ, Zabel P. System analysis of an ISRU

production plant: extraction of metals and oxygen from lunar

regolith. Acta Astronaut. 2023;203:187–201.

https://doi.org/10.

1016/j.actaastro.2022.11.050

6. Gustafson RJ, White BC, Fidler MJ. Oxygen production via car-

bothermal reduction of lunar regolith. SAE Int J Aerospace.

2009;4(1):311–16.

https://doi.org/10.4271/2009-01-2442

7. Troisi I, Lunghi P, Lavagna M. Oxygen extraction from lunar dry

regolith: thermodynamic numerical characterization of the car-

bothermal reduction. Acta Astronaut. 2022;199:113–24.

https://

doi.org/10.1016/j.actaastro.2022.07.021

8. Paz A. Carbothermal reduction demonstration (CaRD). 2023.

https://ntrs.nasa.gov/citations/20230003977

9. Hegde U, Balasubramaniam R, Gokoglu S. Development and

validation of a model for hydrogen reduction of JSC-1A. 47th

AIAA aerospace sciences meeting including the new horizons

forum and aerospace exposition. Orlando, FL: American Insti-

tute of Aeronautics and Astronautics; 2009.

https://doi.org/10.

2514/6.2009-1389

10. Lu Y, Mantha D, Reddy RG. Thermodynamic analysis on

lunar soil reduced by hydrogen. Metall Mater Trans B. 2010,

41(6):1321–27.

https://doi.org/10.1007/s11663-010-9411-3

11. Sargeant HM, Barber SJ, Anand M, Abernethy FAJ, Sheridan

S, Wright IP, et al. Hydrogen reduction of lunar samples in a

staticsystemforawaterproductiondemonstrationonthemoon.

Planet Space Sci. 2021;205:105287.

https://doi.org/10.1016/j.pss.

2021.105287

12. LomaxBA,ContiM,KhanN,BennettNS,GaninAY,SymesMD.

Provingtheviabilityofanelectrochemicalprocessforthesimul-

taneous extraction of oxygen and production of metal alloys

from lunar regolith. Planet Space Sci. 2020;180:104748.

https://

doi.org/10.1016/j.pss.2019.104748

13. Meurisse A, Lomax B, Selmeci Á, Conti M, Lindner R, Makaya

A, et al. Lower temperature electrochemical reduction of

lunar regolith simulants in molten salts. Planet Space Sci.

2022;211:105408.

https://doi.org/10.1016/j.pss.2021.105408

14. Ellery A, Mellor I, Wanjara P, Conti M. Metalysis Fray Farthing

Chen process as a strategic lunar in situ resource utilization

technology. New Space. 2022;10(2):224–38.

https://doi.org/10.

1089/space.2021.0047

15. Sirk AH, Sadoway DR, Sibille L. Direct electrolysis of molten

lunar regolith for the production of oxygen and metals on the

moon. ECS Trans. 2010;28(6):367–73.

https://doi.org/10.1149/1.

3367929

16. Grossman KD, Sibille L, Bell E, Petersen E, Toro Medina

JA, Williams H, et al. GaLORE (gaseous lunar oxygen from

regolith electrolysis): technology advances for a cold-walled

molten regolith electrolysis reactor. ASCEND 2021. Las Vegas,

Nevada and Virtual: American Institute of Aeronautics and

Astronautics; 2021.

https://doi.org/10.2514/6.2021-4019

17. Runyon K, Berdis J, Summers R, Steven A, Michael N,

Karl H, et al. A system-level approach to extracting oxygen

from lunar regolith via molten regolith electrolysis. 2022.

YU et al.

https://lsic.jhuapl.edu/Resources/files/reference-materials/

MRE_SIS%20report%20for%20IAC%20formated%20iac-2022-

manuscript-template.pdf

18. Humbert MS, Brooks GA, Duffy AR, Hargrave C, Rhamdhani

MA. Thermophysical property evolution during molten regolith

electrolysis. Planet Space Sci. 2022;219:105527.

https://doi.org/

10.1016/j.pss.2022.105527

19. Schreiner SS, Hoffman JA, Sanders GB, Lee KA. Integrated

modeling and optimization of lunar in-situ resource utilization

systems. 2015 IEEE aerospace conference. Big Sky, MT: IEEE;

2015.p.1–11.

https://doi.org/10.1109/AERO.2015.7119077

20. Morita K, Wiraseranee C, Shuto H, Nakamur S, Iwasawa K,

Okabe TH, et al. Dissolution behaviour of platinum group met-

als into molten slags. Min Proc Ext Met. 2014;123(1):29–34.

https://doi.org/10.1179/0371955313Z.00000000070

21. Kim H, Paramore J, Allanore A, Sadoway DR. Electrolysis of

molten iron oxide with an iridium anode: the role of electrolyte

basicity.JElectrochemSoc.2011;158(10):E101.

https://doi.org/10.

1149/1.3623446

22. Standish E, Stefanescu D, Curreri P. Ceramics for molten mate-

rials containment, transfer and handling on the lunar surface.

47th AIAA aerospace sciences meeting including the new hori-

zons forum and aerospace exposition. Orlando, FL: American

Institute of Aeronauticsand Astronautics;2009.

https://doi.org/

10.2514/6.2009-1657

23. Rosenberg SD, Musbah O, Rice EE. Carbothermal reduction of

lunar materials for oxygen production on the moon: reduction

of lunar simulants with carbon. Lunar Planet Sci. 1996;27:1103.

24. Exolith Lab. LHS-1 lunar highlands simulant fact sheet (002-

01-001-0621). 2021.

https://cdn.shopify.com/s/files/1/0398/9268/

0862/files/lhs-1-spec-sheet-Dec2022.pptx.pdf?v=1688051876

25. McKay DS, Carter JL, Boles WW, Allen CC, Allton JH. JSC-1: a

new lunar soil simulant. Engineering, construction, and opera-

tions in space IV. Albuquerque, NM: American Society of Civil

Engineers; 1994. p. 857–66.

26. Bachmann F, Hielscher R, Schaeben H. Grain detection from

2d and 3d EBSD data—specification of the MTEX algorithm.

Ultramicroscopy. 2011;111(12):1720–33.

https://doi.org/10.1016/j.

ultramic.2011.08.002

27. Bale CW, Bélisle E, Chartrand P, Decterov SA, Eriksson G,

Gheribi AE, et al. FactSage thermochemical software and

databases, 2010–2016. Calphad. 2016;54:35–53.

https://doi.org/

10.1016/j.calphad.2016.05.002

28. Henke BL, Gullikson EM, Davis JC. X-ray interactions: pho-

toabsorption, scattering, transmission, and reflection at

50–

30,000 eV,

1–92. Atom Data Nucl Data. 1993;54(2):181–342.

https://doi.org/10.1006/adnd.1993.1013

29. Kato K, Araki N. The corrosion of zircon and zirconia refracto-

ries by molten glasses. J Non-Cryst Solids. 1986;80(1–3):681–87.

https://doi.org/10.1016/0022-3093(86)90462-X

30. Ficheux M, Burov E, Aquilanti G, Trcera N, Montouillout V,

Cormier L. Structural evolution of high zirconia aluminosilicate

glasses. J Non-Cryst Solids. 2020;539:120050.

https://doi.org/10.

1016/j.jnoncrysol.2020.120050

31. Guo J, Villalon T, Pal U, Basu S. Effect of optical basicity on the

stability of yttria-stabilized zirconia in contact with molten oxy-

fluoride flux. J Am Ceram Soc. 2018;101(8):3605–16.

https://doi.

org/10.1111/j

ace.15496

32. Krämer S, Yang J, Levi CG, Johnson CA. Thermochem-

ical interaction of thermal barrier coatings with molten

CaO–MgO–Al

–SiO

(CMAS) deposits. J Am Ceram

Soc. 2006;89(10):3167–75.

https://doi.org/10.1111/j.1551-

2916.2006.01209.x

33. ZhaoH,LeviCG,WadleyHNG.Moltensilicateinteractionswith

thermalbarriercoatings.SurfCoatTech.2014;251:74–86.

https://

doi.org/10.1016/j.surfcoat.2014.04.007

34. Dominguez JA, Whitlow J. Marangoni effect and its potential

utilization in supporting lunar habitats and other extraterres-

trial endeavors. Adv Space Res. 2022;69(5):2259–67.

https://doi.

org/10.1016/j.asr.2021.12.023

35. Webster R I, Opila E J. Viscosity of CaO–MgO–Al

–SiO

(CMAS) melts: experimental measurements and comparison to

model calculations. J Non-Cryst Solids. 2022;584:121508.

https://

doi.org/10.1016/j.jnoncrysol.2022.121508

36. Cooper AR, Kingery WD. Dissolution in ceramic systems: I,

molecular diffusion, natural convection, and forced convection

studies of sapphire dissolution in calcium aluminum silicate.

J Am Ceram Soc. 1964;47(1):37–43.

https://doi.org/10.1111/j.1151-

2916.1964.tb14638.x

37. Guloyan YA, Pustyl’nikov OM. Physical–chemical interactions

during contact between molten glass and refractories. Glass

Ceram. 2011;67(9–10):297–303.

https://doi.org/10.1007/s10717-

011-9283-2

38. Holgate CS, Seward GGE, Ericks AR, Poerschke DL, Levi CG.

Dissolution and diffusion kinetics of yttria-stabilized zirco-

nia into molten silicates. J Eur Ceram Soc. 2021;41(3):1984–94.

https://doi.org/10.1016/j.jeurceramsoc.2020.10.056

39. Trice RW, Su YJ, Mawdsley JR, Faber KT, De Arellano-López

AR, Wang H, et al. Effect of heat treatment on phase stabil-

ity, microstructure, and thermal conductivity of plasma-sprayed

YSZ. J Mater Sci. 2002;37(11):2359–65.

https://doi.org/10.1023/

A:1015310509520

40. Wu D, Shan X, Cai H, Yang F, Zhao X. Grain boundary cor-

rosion mechanism of YSZ thermal barrier oxides under CMAS

attack. Corros Sci. 2022;209:110803.

https://doi.org/10.1016/j.

corsci.2022.110803

41. Morelli S, Testa V, Bolelli G, Ligabue O, Molinari E, Antolotti

N, et al. CMAS corrosion of YSZ thermal barrier coat-

ings obtained by different thermal spray processes. J Eur

Ceram Soc. 2020;40(12):4084–100.

https://doi.org/10.1016/

j.jeurceramsoc.2020.04.058

42. Kowalski B, Stokes J, Albert PE, Lauer PE, Wolfe DE. Effect

of test parameters in the exposure of Ca–Mg–Al–silicate on

yttria stabilized zirconia. J Am Ceram Soc. 2023;106(6):3911–21.

https://doi.org/10.1111/j

ace.19025

SUPPORTING INFORMATION

Additional supporting information can be found online

in the Supporting Information section at the end of this

article.

How to cite this article:

Yu K, Stokes J, Harder B,

Reidy L, Faber KT. Thermochemical interactions

between yttria-stabilized zirconia and molten lunar

regolith simulants. J Am Ceram Soc. 2024;1–12.

https://doi.org/10.1111/jace.19821