Thermal stability of SiO2-B2O3-Al2O3-Na2O-CaO glasses with high Nd2O3 and MoO3 concentrations

Journal of Alloys and Compounds 671 (2016) 84e99

Contents lists available at ScienceDirect

Journal of Alloys and Compounds

journal homepage: h ttp://www.elsevier.com/locate/jalcom

Thermal stability of SiO₂eB₂O₃eAl₂O₃eNa₂OeCaO glasses with high

Nd₂O₃and MoO₃concentrations

,

b

,

*

a

Nolwenn Chouard ^{a 1}, Daniel Caurant ^a, Odile Majerus , Nadia Guezi-Hasni ,

ꢀ

Jean-Luc Dussossoy ^c, Rita Baddour-Hadjean ^d, Jean-Pierre Pereira-Ramos ^d

^aChimie ParisTech, PSL Research University, CNRS, Institut de Recherche de Chimie Paris (IRCP), 75005, Paris, France

Laboratoire d'Etude et Developpement des Matrices de Conditionnement, CEA/DEN/DTCD/SECM, Marcoule, 30207, Bagnols-sur-Ceze, France

Laboratoire des Materiaux et Procedes Actifs, CEA/DEN/DTCD/SECM/LMPA, Marcoule, 30207, Bagnols-sur-Ceze, France

b

ꢀ

ꢁ

c

ꢀ

ꢀ ꢀ

ꢁ

d

ꢀ

Groupe Electrochimie et Spectroscopie des Materiaux (UMR CNRS 7182), Institut de Chimie et des Materiaux Paris-Est, 94320, Thiais, France

a r t i c l e i n f o

a b s t r a c t

Article history:

The incorporation of high MoO₃amounts in borosilicate glasses developed for the immobilization of

radioactive waste may lead to the crystallization of Mo-rich phases that may induce a decrease of the

long term performances of glasses. It is thus essential to understand their crystallization mechanisms

and the possible effect of other abundant ﬁssion products present in the wastes (such as rare earths) in

order to control or to avoid their formation during glass preparation. This paper presents a study, per-

formed by X-ray diffraction, scanning electron microscopy, Raman and optical absorption spectroscopies

of the stability as a function of the thermal treatment temperature T_Cof a simpliﬁed Mo-rich nuclear

waste glass. The impact of the addition of a high amount of Nd₂O₃on the thermal stability of this glass is

studied. For comparison, the thermal stability of a Nd₂O₃-rich glass without Mo is also presented. The

crystallization range of all phases formed in these glasses (CaMoO₄, Na₂MoO₄, Ca₂Nd₈(SiO₄)₆O₂(apatite))

and the evolution of their structure and microstructure as a function of T_Care presented. The intro-

duction of Nd₂O₃in the MoO₃-rich glass inhibits the crystallization of molybdates (increase of Mo sol-

ubility), as long as apatite does not form which suggests that [MoO₄]^2-entities and Nd^3þcations are close

to each other in the glass structure. Besides, when a high density of apatite crystals form, for instance

from glass surface, small Mo-rich partially crystallized globular heterogeneities are observed between

these crystals that exacerbate the nucleation of new apatite crystals (nucleating effect).

Received 9 July 2015

Received in revised form

21 January 2016

Accepted 6 February 2016

Available online 11 February 2016

Keywords:

Borosilicate glass

Molybdenum

Neodymium

Crystallization

Powellite

Apatite

Raman spectroscopy

Optical absorption

1. Introduction

water, high self-irradiation resistance). Research is in progress in

nuclear industry on new glass compositions to immobilize higher

Vitriﬁcation is the most widely used technology for the immo-

bilization of high level radioactive wastes throughout the world

[1e4]. The aim is to dissolve the radionuclides as homogeneously as

possible at the atomic scale in the aperiodic structure of a glassy

matrix for more than 10000 years [5]. After the reprocessing of

spent nuclear fuel, the radioactive waste solutions are generally

conﬁned in borosilicate glasses that present very good character-

istics (high glass transition temperature T_g, low crystallization

tendency during melt cooling, high leaching resistance against

concentrations of nuclear waste (ﬁssion products þ actinides) than

today while keeping good physicochemical properties and long

term performances [6e11]. This will enable to decrease the number

of glass canisters and to reduce the cost of disposal.

Molybdenum is abundant in commercial HLW solutions (de-

fense HLW does not contain much Mo) [12] and its incorporation in

glassy waste forms is a big challenge as MoO₃presents a limited

solubility in silicate and borosilicate glasses [13e17]. Indeed, when

these glasses are prepared under oxidizing or neutral atmosphere,

the main and most stable oxidation state of molybdenum is þVI

[15,18,19]. According to Mo EXAFS and XANES studies [13,18,20],

Mo^6þcations are present in silicate and borosilicate glasses as

molybdate tetrahedral entities [MoO₄]^2-(oxoanions) that are not

directly connected to the silicate network (absence of direct

MoeOeSi bondings). It has been proposed that these oxoanions are

* Corresponding author.

E-mail address: daniel.caurant@chimie-paristech.fr (D. Caurant).

Present address: BU Recyclage DIRP/RDP, Tour Areva, 1 place Jean Millier, 92084

1

ꢀ

PariseLa Defense, France.

http://dx.doi.org/10.1016/j.jallcom.2016.02.063

N. Chouard et al. / Journal of Alloys and Compounds 671 (2016) 84e99

85

Table 1

located in depolymerized regions of the glassy network (i.e. in re-

gions rich in non-bridging oxygen atoms (NBOs)) where their

negative charge is compensated by the alkali and alkaline-earth

cations that are also located in these regions [13,21]. As a conse-

quence, the presence of high contents of both molybdenum and

alkali and alkaline-earth cations in the same depolymerized re-

gions of the glass structure where these species are rather mobile -

because they are not strongly connected to the silicate network -

would favor the separation and the crystallization of alkali and

alkaline-earth molybdate phases such as Na₂MoO₄and CaMoO₄

(scheelite structure) during melt cooling which would explain the

poor molybdenum solubility in these glasses [3,16,21e23]. Indeed,

the structure molybdate phases such as CaMoO₄and Na₂MoO₄

consists in the same kinds of isolated [MoO₄]^2-entities charge

compensated by Na^þor Ca^2þcations [21]. The crystallization of

Na_0.5Nd_0.5MoO₄(scheelite structure) during the cooling of

SiO₂eB₂O₃eNa₂OeZnO-Nd₂O₃eMoO₃glasses containing or not

CaO has also been reported by Kashchieva et al. [24]. The same team

also studied glass formation and crystallization of Mo- and rare

earths-rich phases in simple B₂O₃eMoO₃-RE₂O₃systems (RE: La,

Nd) [25e28]. Using a thermodynamic approach with the Calphad

Nominal composition (in mol% (a) and wt% (b)) and glass transition temperature (T_g)

determined by DTA [34] for Mo, Nd and MoNd glasses. The true glass compositions

(c) as determined by chemical analysis (ICP-AES) are also given for comparison (in

wt%).

Glass

SiO₂

B₂O₃Al₂O₃Na₂O CaO MoO₃Nd₂O₃T_g(^ꢀC)

Mo (mol%)^a

Mo (wt%)^b

Mo (wt%)^c

Nd (mol%)^a

Nd (wt%)^b

Nd (wt%)^c

57.95 10.45 5.11

53.94 11.27 8.07

53.72 10.98 8.47

56.78 10.24 5.00

46.69

46.64

16.49 8.40 1.61

15.84 7.30 3.59

15.57 7.46 3.55

16.16 8.23 0.00

13.71 6.32 0.00

13.57 6.17 0.00

15.91 8.10 1.55

13.30 6.13 3.01

13.56 5.93 2.89

0.00 556

0.00

3.59 584

16.54

16.83

3.54 580

16.05

15.79

9.75 6.98

9.66 6.65

MoNd (mol%)^a55.90 10.08 4.93

MoNd (wt%)^b

MoNd (wt%)^c

45.29

44.93

9.46 6.77

8.60 6.29

studied and compared to that of the MoNd glass. In a previous

paper [37], we presented the impact of the addition of increasing

Nd₂O₃contents (0e16 wt%) on the structure and the crystallization

tendency of the melt at constant cooling rate (1 ^ꢀC/min) to simulate

natural cooling in metallic canisters after melt casting. Moreover, in

another paper we compared the crystallization tendency during

cooling from the melt or prolonged isothermal heating (750 ^ꢀC)

from the glassy state of Mo-, Nd- and Mo þ Nd-rich glasses with or

without ruthenium oxide, RuO₂representing in this study the

platinoids present in waste solutions [33,34]. In the present paper,

in order to study more precisely the tendency of the supercooled

melts to crystallize as a function of the temperature, isochronal

heat treatments above the glass transformation temperature T_g

with a nucleation stage at a ﬁxed temperature (T_N) followed by a

growth stage during 6 h at increasing temperature (T_C) were per-

formed. The quenched glasses and heat treated glass samples were

then characterized by X-ray diffraction (XRD), scanning electron

microscopy (SEM), transmission electron microscopy (TEM),

Raman and optical absorption spectroscopies.

ꢀ

method, Gosse et al. [29] recently predict the separation of a MoO₃

and Na₂O rich-phase (that would then crystallize to form Na₂MoO₄)

in the SiO₂eNa₂OeMoO₃system in accordance with experimental

results [30]. As alkali molybdate crystalline phases such as

Na₂MoO₄exhibits very poor chemical durability against water and

that radioactive cesium can partially replace sodium in their

structure, the crystallization of such phases must be avoided in

nuclear waste glasses. In previous studies, we showed that it was

possible to control the formation of Na₂MoO₄in favor of CaMoO₄in

Mo-rich borosilicate glasses for instance by increasing B₂O₃or CaO

concentrations [21e23]. Nevertheless, even if the chemical dura-

bility of alkaline earth molybdate phases such as CaMoO₄is rela-

tively good, their potential crystallization during melt cooling or

glass heating must be controlled because such phases may incor-

porate actinides (Np, Am, Cm present in waste) in their structure by

partial substitution of alkaline-earth cations [31] that may then

2. Experimental procedure

induce local swelling under the effect of a-self-irradiation followed

by strains and glass cracking during waste forms disposal. Indeed,

recent external Ar irradiation studies performed on RE-doped

CaMoO₄single crystals (RE ¼ rare earths being used as actinide

2.1. Samples preparation and thermal treatment

A Mo-rich glass composition (Mo glass) was chosen in the

SiO₂eB₂O₃eAl₂O₃eNa₂OeCaOeMoO₃system and derives from

that of a more complex glass envisaged to immobilize highly

concentrated waste. From Mo glass, another glass was prepared by

adding Nd₂O₃(MoNd glass). The nominal composition of these two

glasses is given in Table 1. For comparison, a Nd-rich glass without

molybdenum (Nd glass) was also prepared keeping constant the

molar ratios of other oxides (Table 1). The composition of Nd glass

derived from that of MoNd glass by removing molybdenum oxide

from its composition. The true composition of the three glasses was

analysed by inductively coupled plasma-atomic emission spec-

troscopy method (ICP-AES) at the Service Central d'Analyses (Ver-

naison, France) and the results obtained showed good agreement

between experimental and nominal glass compositions (Table 1).

Neodymium oxide was chosen here to simulate all the rare earths

and actinides occurring in high-level waste solutions because

neodymium is the most abundant rare earth in these waste solu-

tions [12] and it is considered as a good surrogate for all trivalent

rare earths and actinides [36].

surrogates) to simulate the effects of

a-decays showed that

powellite may swell by more than 5% at saturation but this phase

has a high radiation resistance because amorphization was not

reached during these experiments [32].

The present work joins within the continuity of previous studies

concerning the crystallization of Mo-rich phases in simpliﬁed nu-

clear borosilicate glasses [10,21e23,33e34]. Particularly, we pre-

sent a study on the crystallization of molydbate phases as a

function of temperature (650e1100 ^ꢀC), in a simpliﬁed nuclear

waste glass composition (Mo glass) belonging to the

SiO₂eB₂O₃eAl₂O₃eNa₂OeCaOeMoO₃system with 1.61 mol%

(~3.6 wt%) MoO₃(Table 1). Besides, as RE are also abundant ﬁssion

products in high level radioactive waste [12], which may crystallize

during melt cooling to form RE-rich silicate phases such as apatite

(Ca₂RE₈(SiO₄)₆O₂) [9,35,36], the complexity of the previous glass

composition (Mo glass) was increased by adding a high amount of

Nd₂O₃(3.54 mol% i.e. ~ 16 wt%, Table 1) in order to become closer to

a real nuclear waste glass with high waste loading (RE ¼ Nd is the

most abundant RE and simulates here all the RE and actinides

present in wastes [3,36]). The thermal stability of this glass con-

taining both Mo and Nd (MoNd glass) was studied and compared to

that of the Mo glass. To complete this work, the crystallization as a

function of temperature of a Nd₂O₃-rich glass (Nd glass) containing

3.6 mol% Nd₂O₃(~16.5 wt%, Table 1) without molybdenum was also

Mo, MoNd and Nd glasses (50 g batch) were prepared using

appropriate quantities of reagent grade SiO₂, H₃BO₃, Al₂O₃, Na₂CO₃,

CaCO₃, MoO₃and Nd₂O₃powders and melted in air in an electrical

furnace at 1300 ^ꢀC for 3 h in PteAu (5 wt% Au) crucibles, and then

cast. The glass obtained was then ground and melted again at

1300 ^ꢀC for 2 h to ensure homogeneity before casting on a steel

86

N. Chouard et al. / Journal of Alloys and Compounds 671 (2016) 84e99

plate at room temperature. The MoNd and Nd quenched glasses

were transparent, macroscopically and microscopically homoge-

neous. The Mo quenched glass was lightly opalescent because of

the existence of a phase separation phenomenon as detected by

scanning electron and transmission microscopies. Indeed, Mo-rich

globular particles (~50 nm size) homogeneously dispersed in the

Mo sample were clearly observed by these techniques (images not

shown) and are characteristic of a liquideliquid phase separation

occurring in the melt or in the supercooled melt. For Mo glass,

phase separation (at least on the optical scale) can be avoided by

rapidly quenching the melt between two metallic plates after

casting. However, despite this little opalescence for the Mo glass, all

quenched glasses were amorphous according to XRD and Raman

spectroscopy [34].

(a)

Ca₂Nd₈(SiO₄)₆O₂

CaMoO₄

After quenching, a two-step thermal treatment of nucleation

and growth was performed on bulk samples for the three glass

compositions in order to induce crystallization. The nucleation step

was performed at T_N¼ T_gþ 20 ^ꢀC for 2 h (T_gis the glass transition

temperature determined by differential thermal analysis (DTA)

[34], Table 1) and followed by a growth step for 6 h at T_Cbetween

650 and 1100 ^ꢀC. The subscripts N and C for the heating tempera-

tures T_Nand T_Crespectively refer to the Nucleation and Crystalli-

zation steps. For these thermal treatments, pieces of glass were put

in PteAu crucibles and introduced successively in two preheated

furnaces at the corresponding temperatures (T_Nand T_C) and then

quenched at room temperature.

For comparison with the partially crystallized samples obtained

above, two ceramics (used in this paper as references for XRD and

Raman spectroscopy) have also been synthesized: a powellite

CaMoO₄ceramic and an apatite Ca₂Nd₈(SiO₄)₆O₂ceramic. The

powellite ceramic was synthesized by solid state reaction with a

ﬁrst heat treatment during 6 h at 800 ^ꢀC (heating and cooling at

20

25

30

35

2 theta (°)

40

45

50

(b)

Apatite

CaMoO₄

Na₂MoO₄

5

^ꢀC/min) of a mixture of CaCO₃þ MoO₃in stoichiometric pro-

700

800

900

1000

1100

1200

portions followed by a sintering during 3 h at 1150 ^ꢀC (heating and

cooling at 5 ^ꢀC/min). The apatite ceramic was synthesized by solid

state reaction with a ﬁrst heat treatment during 24 h at 900 ^ꢀC

Raman shift (cm^-1)

Fig. 1. (a) XRD patterns of the powellite (CaMoO₄) and apatite (Ca₂Nd₈(SiO₄)₆O₂)

reference samples. The purity of the phases was veriﬁed by comparison with the ﬁles

provided by the International Centre for Diffraction Data (ICDD): CaMoO₄powellite

(00-029-0351) and Ca₂Nd₈(SiO₄)₆O₂apatite (04-007-5968). (b) Raman spectra of

CaMoO₄, Na₂MoO₄and Ca₂Nd₈(SiO₄)₆O₂reference samples.

(heating and cooling at

5

^ꢀC/min) of

a

mixture of

SiO₂þ CaCO₃þ Nd₂O₃in stoichiometric proportions followed by a

sintering during 100 h at 1500 ^ꢀC (heating and cooling at 5 ^ꢀC/min).

According to XRD measurements, pure powellite and apatite pha-

ses were obtained (Fig. 1a). Besides, a Na₂MoO₄crystalline com-

mercial powder (Aldrich, AD2690) was also used as reference (a-

200 kV. In this last case the samples were prepared by ﬁnely

crushing glasses in an agate mortar and ultrasonically stirring them

in butanol. The smallest particles were then collected in the su-

pernatant liquid using a holey carbone-coated copper grid.

form, stable at room temperature and until 440e460 ^ꢀC [38,39]).

2.2. Characterization methods

Raman spectroscopy was conducted at room temperature on a

Horiba Jobin-Yvon HR800 micro-Raman spectrometer equipped

with a CCD detector. The 488 nm line of a coherent Ar^þlaser was

used as the exciting source. Experiments were carried out with a

XRD patterns of the ceramics, quenched and heat treated glass

samples were recorded at room temperature on an X'Pert PRO

PANalytical instrument using Cu-K

Bragg-Brentano ( e2 ) geometry. For this analysis, the samples

were crushed into powders with particle size <80 m. XRD patterns

were recorded for 2

ranging from 10^ꢀto 80^ꢀwith a step of

0.026^ꢀ(2

) and a time acquisition per step of 79s. Before performing

a

radiation (

l

¼ 0.15406 nm) in

q

confocal ꢁ100 objective and a confocal hole of 300

were recorded between 700 and 1200 cm^ꢂ1using a 1800 grooves

per mm grating and a 100 m slit, and with a total acquisition time

mm. Spectra

m

q

m

q

of 120s (60s acquisition time and 2 successive acquisitions to

reduce spurious noise). All the spectra shown in this paper were

corrected for temperature and frequency dependency of the scat-

tering intensity using a correction factor of the form proposed by

Long [40]. A polynomial baseline was ﬁtted directly to the corrected

Raman spectra which were then normalized. For comparison with

the heat treated glass samples, the Raman spectra and XRD pat-

terns of the crystalline references Ca₂Nd₈(SiO₄)₆O₂, CaMoO₄and

Na₂MoO₄were also recorded (Fig. 1).

the thermal treatment of nucleation at T_N(T_gþ 20 ^ꢀC), T_gwas

determined by DTA for all quenched glasses with a Netzsch Ger-

atebau STA 409 apparatus with heating at 10 ^ꢀC/min (Table 1).

Field-emission gun scanning electron microscopy (FEG-SEM) ob-

servations were performed on quenched and heat treated glass

samples using a Zeiss supra 55VP microscope with a tungsten wire

of 200 nm operating at 25 kV with a working distance of 5 mm.

Before SEM observation, the samples were cut, embedded in resin,

polished with diamond paste at 1 mm (Fig. 2) and coated with a thin

Optical absorption spectra were recorded in transmission at 10 K

with a double-beam CARY-6000i spectrometer to follow the evo-

lution of Nd^3þions environment with T_C(Nd glass). Translucent

carbon layer to eliminate electrostatic effects. Transmission elec-

tron microscopy (TEM) observations were carried out on quenched

glasses with a JEOL JEM 2010F electron microscope operating at

N. Chouard et al. / Journal of Alloys and Compounds 671 (2016) 84e99

87

Fig. 2. Pictures of the heat treated Mo (T_C¼ 700 ^ꢀC (a), 750 ^ꢀC (b)), Nd (T_C¼ 700 ^ꢀC (c),

800 ^ꢀC (d)) and MoNd (T_C¼ 700 ^ꢀC (e), 750 ^ꢀC (f)) glasses after cutting, embedding in

resin and polishing (samples size ~ 1 cm).

pellets were prepared from 25 mg of ﬁnely ground material mixed

with KBr (125 mg), and spectra were collected in the 420e440 nm

(22727e23810 cm^ꢂ1) region using a spectral band width of 0.6 nm,

data interval of 0.2 nm, and 1 s acquisition time. This wavelength

region corresponds to the ⁴I_9/2/ ²P_1/2transition of the Nd^3þion. At

4

10 K, only the lowest Stark level of the I_9/2ground state of Nd^3þ

2

ions is populated. As the P_1/2state is not split by the crystal ﬁeld

around Nd^3þions, each kind of neodymium environment is char-

acterized by a single ⁴I_9/2/ ²P_1/2absorption band whose position

depends on the degree of NdeO bonds covalency (nephelauxetic

effect) [41]. For the MoNd sample, the evolution with T_Cof the ⁴I_9/

/

⁴G_5/2,²G_7/2transition occurring in the 550e610 nm

2

(16400e18200 cm^ꢂ1) region was also to detect the possible pres-

ence of Nd^3þions in the molybdate crystals. This last transition is

known as hypersensitive transition as it is very sensitive to changes

in the environment of Nd^3þions such as the symmetry of the

crystal ﬁeld [41].

Fig. 3. (a) XRD patterns of the heat treated Mo glass at T_Cvarying from 650 to 1100 ^ꢀC.

All the diffraction peaks can be attributed to powellite (Fig. 1a). (b) XRD patterns of the

heat treated Nd glass at T_Cvarying from 650 to 1100 ^ꢀC. All the diffraction peaks can be

attributed to apatite crystals (Fig. 1a) except several small peaks for T_C¼ 800 ^ꢀC

(indicated by asterisks) that were not identiﬁed. (c) XRD patterns of the heat treated

MoNd glass at T_Cvarying from 650 to 1100 ^ꢀC. All the diffraction peaks detected on the

patterns are attributed to apatite crystals (Fig. 1a).

3. Results

3.1. XRD analysis

3.1.1. Crystallization of the Mo-rich glass

in all the samples heat treated between 650 and 900 ^ꢀC. No

Na₂MoO₄crystals or other crystalline phases were detected by this

method. For the sample heat treated at 950 ^ꢀC, it is still possible to

The XRD patterns of the Mo glass heat treated between 650 and

1100 ^ꢀC and then quenched to room temperature are presented in

Fig. 3a. According to the XRD pattern of the powellite ceramic

(Fig. 1a), the presence of CaMoO₄crystals is clearly put in evidence

88

N. Chouard et al. / Journal of Alloys and Compounds 671 (2016) 84e99

9.535

(a)

9.530

9.525

9.520

9.515

9.510

Apatite

Nd

MoNd

800

900

1000

1100

1500

Temperature (°C)

7.025

7.020

7.015

7.010

7.005

Fig. 4. Evolution of the intensity (integration) of the XRD line located at 2

q

¼ 47.1^ꢀof

(b)

the powellite (CaMoO₄) phase formed in the heat treated Mo and MoNd glasses as a

function of T_C. This diffraction line was selected because in the case of both apatite and

powellite crystallization, it remains isolated and not hidden by one of the diffraction

lines of apatite (as it would have been the case for instance for the most intense XRD

line of powellite, located at 2

q

¼ 28.8^ꢀ(Figs. 1 and 3)).

Apatite

NdM

MoNdM

5.232

5.230

5.228

5.226

5.224

5.222

(a)

800

900

1000

1100

1500

CaMoO

4

Temperature (°C)

Fig. 6. Unit cell parameters a ¼ b and c, determined from the XRD patterns (Fig. 3b), of

the apatite crystals (space group P6₃/m) formed in the heat treated Nd glass as a

function of T_C. (a) a ¼ b parameter, (b) c parameter. For comparison, the lattice pa-

rameters of the Ca₂Nd₈(SiO₄)₆O₂ceramic sample synthesized at 1500 ^ꢀC are also re-

ported on ﬁgure (a ¼ b ¼ 9.529 Å, c ¼ 7.022 Å).

observe the most intense diffraction peak of the powellite phase

but its intensity is very weak, which means that powellite crys-

tallization rate is very low at this temperature. Then, beyond

1000 ^ꢀC, no more crystallization is observed according to XRD

measurements (Fig. 4). Thus, in Mo glass, powellite crystals are not

stable above 950 ^ꢀC. This result is in good agreement with the high

temperature in situ Raman spectroscopy study of Magnin et al.

[42,43] performed between 1100 and 25 ^ꢀC by cooling from the

melt a soda lime borosilicate composition with 2 mol% MoO₃.

Indeed these authors showed, by comparing their Raman spectra

with those of pure crystalline CaMoO₄and Na₂MoO₄at the same

temperatures, that the crystallization of CaMoO₄in their system

(close to the one studied in the present paper) only occurred below

950 ^ꢀC whereas Na₂MoO₄crystallization was detected near 650 ^ꢀC.

In our work, quenching from T > 950 ^ꢀC probably enabled to avoid

the crystallization of CaMoO₄during cooling.

The a and c lattice parameters of the powellite crystals (scheelite

structure, I4_1/aspace group) [44] formed in the Mo glass have been

determined by full pattern proﬁle ﬁtting for the samples heat

treated between 700 and 900 ^ꢀC. These parameters do not evolve

signiﬁcantly with T_C(Fig. 5a,b) which means that by increasing T_C

there is no effect of thermally activated rearrangement within the

structure of powellite and that the composition of the crystals

probably does not change. The a parameter of the crystals formed in

Mo glass is consistent with the value obtained for the powellite

700 750 800 850 900 1100

Temperature (°C)

1200

11.48

11.47

11.46

11.45

11.44

11.43

11.42

(b)

CaMoO

4

700 750 800 850 900 1100

Temperature (°C)

1200

Fig. 5. Unit cell parameters a ¼ b and c, determined from the XRD patterns (Fig. 3a), of

the powellite crystals (I4_1/aspace group) formed in the heat treated Mo glass as a

function of T_C. (a) a ¼ b parameter, (b) c parameter. For comparison, the lattice pa-

rameters of the powellite ceramic sample synthesized at 1150 ^ꢀC (B) are also reported

on ﬁgure (a ¼ b ¼ 5.2267 Å, c ¼ 11.4363 Å).

N. Chouard et al. / Journal of Alloys and Compounds 671 (2016) 84e99

89

Fig. 7. Back-scattered SEM images of the heat treated Mo glass at T_Cvarying from 650 to 1100 ^ꢀC. (a) 650 ^ꢀC, (b) 700 ^ꢀC, (c) 750 ^ꢀC, (d) 850 ^ꢀC, (e) 900 ^ꢀC, (f) 950 ^ꢀC, (g) 1000 ^ꢀC, (h)

1100 ^ꢀC. In several globular heterogeneities (see (f) for instance) white spots can be observed that have been identiﬁed by EDX as gold particles originating from the corrosion of the

PteAu crucible by the melt. By comparison, melting performed in pure Pt crucibles followed by glass heat treatment has shown that the gold particles observed here have not any

impact on phase separation and molybdate crystallization.

Fig. 8. Back-scattered SEM images of the heat treated Nd glass at T_Cvarying from 650 to 1100 ^ꢀC. (a) 650 ^ꢀC, (b) 750 ^ꢀC, (c) 800 ^ꢀC, (d) 850 ^ꢀC, (e) 900 ^ꢀC, (f) 950 ^ꢀC, (g) 1050 ^ꢀC, (h)

1100 ^ꢀC.

ceramic synthesized for this study (a ¼ 5.227 Å, Fig. 5a) and with

values reported in literature (a ¼ 5.2261 Å) [44]. However, the c

parameter of these crystals is quite different (~0.03 Å higher,

Fig. 5b) from that of the synthesized powellite ceramic

(c ¼ 11.436 Å) and from values reported in literature (11.4329 Å)

[44]. The origin of this difference concerning the c parameter is

unclear but several explanations may be proposed. The composi-

tion of the powellite crystals grown in the supercooled borosilicate

melt may deviate from the CaMoO₄stoichiometry of the ceramic,

but this would be in contradiction with the lack of important dif-

ference for the a parameter (Fig. 5a). Nevertheless, because of their

(EDX) or electron probe micro-analysis (EPMA). A more plausible

explanation may arise from the fact that powellite is known to

exhibit an important anisotropic behavior of its average lattice

thermal expansion coefﬁcient

a along a- and c-axis between 25 and

1000 ^ꢀC: a_a¼ 13.5.10^ꢂ6^ꢀC and a_b¼ 22.8.10^ꢂ6^ꢀC [44]. Thus, the

powellite crystals formed in the supercooled melt and quenched to

room temperature may have retained a slight length excess along c-

axis after samples quenching which could explain the lattice pa-

rameters difference observed at room temperature (Fig. 5b).

3.1.2. Crystallization of the Nd-rich glass

too small size (<1

mm, see Fig. 7 in Section 3.2.1), it was not possible

The XRD patterns of the Nd glass heat treated between T_C¼ 650

to determine their composition by energy dispersive X-ray analysis

and 1100 ^ꢀC and then quenched to room temperature are presented

90

N. Chouard et al. / Journal of Alloys and Compounds 671 (2016) 84e99

in Fig. 3b. According to the XRD pattern of the apatite ceramic

(Fig. 1a), the presence of apatite crystals is clearly put in evidence in

all the samples heat treated between 800 and 1100 ^ꢀC. For

T_C¼ 750 ^ꢀC only very weak intensity lines due to apatite crystal-

lization from the surface are detected (see Fig. 8b in Section 3.2.2).

From Fig. 3b it also appears that the intensity of the pattern is

maximal at about 850 ^ꢀC which indicates that after increasing until

850 ^ꢀC, the amount of apatite crystals formed decreases with T_C.

Moreover, several low intensity peaks, only detected for T_C¼ 800 ^ꢀC

(indicated by asterisks), were not identiﬁed. By comparison with

the results concerning the Mo glass presented above (Fig. 3a), it

appears that the crystallization of the Nd glass occurs at higher T_C,

Nd glass is thus more stable than Mo glass. This is consistent with

Mo glass (Fig. 3a), this puts in evidence that adding neodymium

oxide to the glass composition inhibits the crystallization of

molybdate phases. This observation is consistent with the results of

our previous works [33,34,37] and with the SEM observations on

the quenched Mo and MoNd glasses presented in this paper, where

it appeared that Nd₂O₃addition suppressed the formation of the

globular Mo-rich phase detected in Mo glass. By comparison with

the results concerning Mo glass presented above (Fig. 3a), it ap-

pears that the crystallization of MoNd glass occurs at higher T_C,

MoNd glass is thus more stable than Mo glass. This is consistent

with our previous DTA results [34] that showed that the difference

DT between the maximum of the exothermic peak T_passociated

with the crystallization of apatite or powellite and T_gwas signiﬁ-

our previous DTA results [34] that showed that the difference

between the maximum of the exothermic peak T_passociated with

the crystallization of apatite or powellite and T_g(

D

T

cantly higher for MoNd glass (360 ^ꢀC) than for Mo glass (192 ^ꢀC).

DT ¼ T_pe T_g) was

3.2. SEM analysis

signiﬁcantly higher for Nd glass (398 ^ꢀC) than for Mo glass (192 ^ꢀC).

This lower stability of Mo glass can be explained structurally by the

higher mobility of [MoO^-₄] entities that, unlike Nd^3þions, are not

directly connected to NBOs. This difference would be at the origin of

the highest phase separation tendency and preferential crystalli-

zation of powellite in the bulk of Mo glass (see Fig. 7 in Section

3.2.1).

3.2.1. Crystallization of the Mo-rich glass

SEM shows that powellite crystallization (facetted crystals with

pyramidal shape) is uniform within the bulk of the Mo glass heat

treated between 650 and 900 ^ꢀC (Fig. 7aee). The crystal size is very

small for all the samples (<1 mm) but increases with T_C. Besides, it

appears that the number of crystals per unit volume decreases

when the temperature increases. These two observations are

consistent with an Oswald ripening. When T_C> 950 ^ꢀC, despite the

fact that no crystalline phase is detected by XRD, the glass is not

homogeneous. Indeed, Mo-rich globular heterogeneities (as ana-

The a and c lattice parameters of the apatite crystals (hexagonal

structure, space group P6₃/m) formed in the Nd glass have been

determined by full pattern proﬁle ﬁtting for the samples heat

treated between 850 and 1100 ^ꢀC. It appears that a and c increase

signiﬁcantly with T_C(Fig. 6a,b) and become closer to the lattice

parameters of the apatite ceramic prepared at 1500 ^ꢀC (a ¼ 9.529 Å,

c ¼ 7.022 Å). Different explanations can be proposed to explain the

variation with T_Cof the apatite crystals lattice parameters (Fig. 6).

For instance, the composition of the apatite crystals may deviate

from the stoichiometric Ca₂Nd₈(SiO₄)₆O₂composition. Indeed, in a

previous study performed on Nd-rich partially crystallized glasses

with composition close to that of the Nd glass, we showed that the

lysed qualitatively by EDX) with a size of about 1

mm can be

observed for the heat treated Mo glass (Fig. 7geh). Similar globular

heterogeneities were also observed in the sample heat treated at

950 ^ꢀC (Fig. 7f) for which only weak CaMoO₄crystallization was put

in evidence by XRD (Fig. 3a). This globular morphology, associated

with the absence of diffraction peaks on the XRD patterns between

950 and 1100 ^ꢀC, indicates that a phase separation phenomenon

occurs at these temperatures and that the Mo-rich heterogeneities

are amorphous. This is consistent with the study performed by

Magnin et al. [42,43] on a Mo-rich soda lime borosilicate compo-

sition that showed that liquideliquid phase separation of Mo-rich

droplets occurred in the melt below 1100 ^ꢀC. For the Mo glass

sample heat treated at 950 ^ꢀC, it seems that the very small

diffraction peaks observed on the XRD pattern (Fig. 3a) correspond

to crystals that did not appear during the thermal treatment of

growth but only after, when the heat treated sample was quenched

from 950 ^ꢀC to room temperature. In this case, a small amount of

powellite crystals was probably formed inside the separated Mo-

rich globular heterogeneities and the crystallization mechanism is

consequently different from the one observed in the Mo glass heat

treated between 650 and 900 ^ꢀC. Indeed, the geometrical

morphology of the CaMoO₄crystals formed in this temperature

range (Fig. 7aee) indicates that they have grown within the bulk of

the melt and not within globular heterogeneities. Nevertheless, the

very small and homogeneously dispersed globular Mo-rich het-

erogeneities (~50 nm) present in the quenched Mo glass and

originating from a small scale phase separation in the melt during

quenching, probably act as nucleation sites in the bulk for the

crystallization of the facetted CaMoO₄particles during thermal

treatment below 950 ^ꢀC. During growth, these crystals probably

consume the molybdenum initially present in the small globular

Mo-rich heterogeneities, so that after 6 h heating, there are no

traces of the initial small scale phase separation (Fig. 7bee).

Macroscopic observations show that the heat treated Mo sam-

ples are totally opaque when T_C< 950 ^ꢀC (the samples corre-

sponding toT_C¼ 700 and 750 ^ꢀC are shown in Fig. 2a,b) and become

slightly translucent when T_Cꢃ 950 ^ꢀC. According to the SEM results,

the presence of a high density of crystalline particles dispersed

lattice parameters of the apatite crystals (Ca_2þxNd_8-x(SiO₄)₆O_2e0.5x

)

decrease with the non-stoichiometry [45]. Thus, in the present

study, this would mean that at low heating temperature

(T_C¼ 850 ^ꢀC) the non-stoichiometry x of the apatite crystals would

be important (x ~ 0.4 at 850 ^ꢀC according to the data given by

Quintas et al. [45]) and would then decrease with T_cto reach a value

close to x ¼ 0 (stoichiometric apatite ceramic composition, Fig. 6).

Nevertheless, it is important to note that in Ref. [45] the decrease of

x was observed by increasing the CaO/Na₂O concentration ratio

while keeping the same thermal treatment (slow cooling from the

melt or nucleation þ growth). As in the present case the compo-

sition of the glass remained the same for all the heat treatment

temperatures, it is possible that the increase of lattice parameters

with T_Chas a different origin. It is probably due to the evolution

with T_Cof the ordering of Ca^2þand Nd^3þions in the 6 h and 4f sites

of the apatite structure as discussed below from the optical results

of the partially crystallized samples.

3.1.3. Crystallization of the Mo and Nd-rich glass

For the heat treated MoNd glass, the XRD patterns presented in

Fig. 3c bring to light the crystallization of apatite Ca₂Nd₈(SiO₄)₆O₂

apparently as unique phase. Apatite crystallization is always

observed beyond T_C¼ 750 ^ꢀC, with a maximum crystallization rate

in the range T_C¼ 850e900 ^ꢀC (Fig. 3c). An evolution with T_Cof the

XRD patterns and apatite crystals lattice parameters very similar to

that of Nd glass was observed (Figs. 3 and 6). Whatever T_C, no

molybdate crystalline phase was detected by XRD (Fig. 3c). The

effect of neodymium oxide addition on powellite crystallization

clearly appears in Fig. 4. As no molybdate crystallization is detected

by XRD in the heat treated MoNd glass contrary to the heat treated

N. Chouard et al. / Journal of Alloys and Compounds 671 (2016) 84e99

91

Fig. 9. Back-scattered SEM images of the heat treated MoNd glass at T_Cvarying from 650 to 1100 ^ꢀC: (a) 650 ^ꢀC, (b) 700 ^ꢀC, (c and i) 750 ^ꢀC, (d) 850 ^ꢀC, (e) 900 ^ꢀC, (f) 950 ^ꢀC, (g)

1000 ^ꢀC, (h) 1100 ^ꢀC, (j) 800 ^ꢀC. A: apatite Ca₂Nd₈(SiO₄)₆O₂; M: Mo-rich globular heterogeneities.

everywhere in the bulk with size higher than 100 nm in the glass

heated below 950 ^ꢀC (Fig. 7) enables to explain why these samples

appear totally opaque (Fig. 2a,b) whereas the samples heated at

T_Cꢃ 950 ^ꢀC are slightly opalescent due to the lower density of

particles (Fig. 7).

3.2.3. Crystallization of the Mo- and Nd-rich glass

The SEM images of the heat treated MoNd glass bring additional

information (Fig. 9). Firstly, they conﬁrm the XRD results as it is

possible to observe the dendritic-shape and then hexagonal-shape

apatite crystals for T_Cranging from 750 to 1100 ^ꢀC. The evolution

with T_Cof the shape and distribution (surface/bulk) of apatite

crystals is globally similar to that of the Nd sample (Fig. 8).

Nevertheless, a careful examination of the apatite crystals growing

from the surface shows signiﬁcant differences between the Nd and

MoNd samples. Indeed, the thickness of the crystallized layer

signiﬁcantly increases when MoO₃is added to the glass composi-

3.2.2. Crystallization of the Nd-rich glass

SEM images show that the microstructure and the distribution

of the apatite crystals formed in the Nd glass strongly depend on T_C

(Fig. 8). At T_C¼ 650 ^ꢀC, in accordance with XRD (Fig. 3c) no crystals

are detected neither in the bulk nor on the surface of the sample

(Fig. 8a). For T_C¼ 750 ^ꢀC (Fig. 8b), apatite crystallization (hetero-

geneous) is almost only detected from the glass surface (layer

tion, for instance it grows from 30 to 200

m

m for T_C¼ 750 ^ꢀC

(Figs. 8b and 9i) and from 170 to 700

m

m for T_C¼ 800 ^ꢀC (Figs. 8c

and 9j). The layer microstructure is also different between the Nd

and MoNd samples. Whereas the crystals constituting the layer of

the heat treated Nd glass almost all originate from the surface

(Fig. 8), the crystallized layer of the heat treated MoNd glass is

made of a succession of apatite crystals that have nucleated at

different distances from the surface (Fig. 9iej). Besides, by

comparing the SEM images obtained for the heat treated Mo and

MoNd glasses, the strong effect of neodymium oxide on the crys-

tallization of molybdate phases put in evidence by XRD can be

described more precisely. Indeed, at 650 and 700 ^ꢀC, no crystals are

observed in the bulk of the heat treated MoNd glass (Fig. 9aeb),

whereas powellite clearly appeared in the bulk of the heat treated

Mo glass for the same T_C(Fig. 7aeb). Macroscopic observation

conﬁrms that in this temperature range the MoNd samples - unlike

the heat treated Mo samples (Fig. 2a) e remain transparent in the

thickness ~ 30

m

m). For 800 ^ꢀC ꢄ T_Cꢄ 950 ^ꢀC (Fig. 8def), the

amount of crystals has strongly increased in accordance with XRD

(Fig. 3c) and apatite crystallization is detected both from the surface

and in the bulk. As observed by XRD, the maximal density of

crystals occurs at T_C¼ 850 ^ꢀC. Below 900 ^ꢀC the crystals exhibit a

dendritic shape that can be explained by the high viscosity of the

supercooled melt in this temperature range. Above T_C¼ 950 ^ꢀC

(Fig. 8g,h), the density of crystals has signiﬁcantly decreased.

Crystals are still formed in the bulk but their microstructure is no

more dendritic due to the decrease of the viscosity of the super-

cooled melt in this temperature range. All these SEM results are in

agreement with macroscopic observations showing that the heat

treated Nd samples remain transparent in the bulk when

T_Cꢄ 750 ^ꢀC but become opaque when T_Cꢃ 800 ^ꢀC (the samples

corresponding to T_C¼ 700 and 800 ^ꢀC are shown in Fig. 2c,d).

92

N. Chouard et al. / Journal of Alloys and Compounds 671 (2016) 84e99

residual glass has been neodymium depleted (as shown in Ref. [37]

by electron microprobe analysis of Nd close to the apatite crystals

formed during slow cooling from the melt in MoNd glass). In the

next section, it is shown by Raman spectroscopy that these het-

erogeneities are not totally amorphous. The presence of such Mo-

rich particles in the neighborhood of apatite crystals was already

observed after slow cooling from the melt in a previous paper [34].

3.3. Raman spectroscopy

3.3.1. Crystallization of the Mo-rich glass

Raman spectra of the Mo glass heat treated between 650 and

1100 ^ꢀC are presented in Fig. 10a. They are consistent with the re-

sults obtained by XRD and SEM (Figs. 3a and 7). Indeed, for the Mo

glass heat treated between 650 and 900 ^ꢀC, the symmetric

stretching vibration bands of the [MoO₄]^2-tetrahedral entities in

CaMoO₄crystals clearly appear with very high intensities. These

bands can be associated with this crystalline phase thanks to the

Raman spectrum of the pure powellite ceramic (Fig. 1b). Besides,

these vibration bands are so intense that it is very difﬁcult to

observe the vibration bands of [MoO₄]^2-tetrahedral entities and

[SiO₄]^4-(Q_n) units that are present in the residual glass. It is also

interesting to note that the vibration bands of [MoO₄]^2-tetrahedral

entities in Na₂MoO₄crystals are not observed for T_Cvarying from

650 to 900 ^ꢀC (the spectrum of the Na₂MoO₄reference is given

Fig. 1b), which conﬁrms that the only molybdate phase that crys-

tallizes in the heat treated Mo glass is powellite. Even for a pro-

longed heat treatment (30 h) at T_C¼ 750 ^ꢀC followed by quenching

to room temperature, we showed in Ref. [34] by Raman spectros-

copy that CaMoO₄was the only crystalline molybdate phase crys-

tallizing in Mo glass whereas both Na₂MoO₄and CaMoO₄crystals

were detected after controlled cooling from the melt. This differ-

ence of crystallization tendency of Na₂MoO₄during melt cooling

and glass heating was explained from structural considerations by

taking into account the temperature dependence of the amount of

Na^þions available for the charge compensation of molybdate en-

tities [34]. Moreover, the fact that sodium molybdate crystals were

not formed in our samples (all heat treated at T_Cꢃ 650 ^ꢀC and then

quenched to room temperature) is in agreement both with the low

melting point of Na₂MoO₄(~690 ^ꢀC) [38,39] and with the in-situ

high temperature Raman spectroscopy study performed by Mag-

nin et al. [42,43] on a Mo-rich soda lime borosilicate glass that

showed that Na₂MoO₄crystallization only occurred below around

650 ^ꢀC during melt cooling. Then, for T_Cvarying from 950 to

1000 ^ꢀC, no more vibration bands associated with the crystalliza-

tion of molybdate phases are observed and the vibration bands of

[MoO₄]^2-tetrahedral entities and [SiO₄]^4-(Q_n) units in the glass are

then clearly visible. The fact that no vibration bands associated with

powellite crystals are detected for T_C¼ 950 ^ꢀC in Fig. 10a whereas

low intensity lines due to powellite are detected on the XRD pattern

(Fig. 3a) is unclear but could be due to a sampling problem during

Raman or XRD characterization. Anyway, we can conclude from

these results that the Mo-rich globular heterogeneities observed by

SEM when T_C¼ 950e1000 ^ꢀC (Fig. 7feg) are essentially amorphous.

However, it is important to note the slight crystallization observed

for the Mo glass heat treated at 1100 ^ꢀC. Indeed, for this sample, the

Raman spectrum shows the presence of vibration bands associated

with the crystallization of a small amount of both CaMoO₄and

Na₂MoO₄phases (Fig.10a). As we explained above, these crystalline

phases are no more stable at this temperature according to the in

situ high temperature Raman study of Magnin et al. [42,43]. As a

consequence, it seems that these crystals were not formed during

the thermal treatment at T_C, but only after, when the heat treated

sample has been quenched from 1100 ^ꢀC to room temperature. In

this case, a small quantity of CaMoO₄and Na₂MoO₄crystals was

Fig. 10. Evolution of the normalized Raman spectra (in the region 700-1200 cm^ꢂ1) of

the heat treated Mo (a) and MoNd (b) glasses at T_Cvarying from 650 to 1100 ^ꢀC. The

spectra have been corrected for temperature and frequency dependent scattering in-

tensity. The black circles (C) correspond to the vibration bands of [MoO₄]^2-tetrahedral

entities in the powellite phase, the white square ( ) corresponds to a vibration band of

[MoO₄]^2-tetrahedral entities in the Na₂MoO₄crystalline phase and the star (*) cor-

responds to the symmetric stretching vibration of [MoO₄]^2-tetrahedral entities in the

glassy phase. The broad band located around 1070 cm^ꢂ1is due to the symmetric

stretching vibration of the [SiO₄]^4-(Q_n) tetrahedral entities in the glass (the dotted line

is a guide for the eyes), it is difﬁcult to detect this band for the Mo glass heat treated

between 650 and 900 ^ꢀC because of the presence of the intense bands due to the

vibration of [MoO₄]^2-tetrahedral entities in the powellite CaMoO₄phase. The white

circles (B) correspond to the stretching vibrations of the [SiO₄]^4-Q₀units in the

apatite Ca₂Nd₈(SiO₄)₆O₂phase. In accordance with XRD, no crystalline phases were

detected by Raman spectroscopy (spectra not shown) for the quenched Mo and MoNd

glasses [34].

▫

bulk (Fig. 2e). Thus, in agreement with the XRD results (Fig. 3a and

c), it is clearly shown that Nd₂O₃addition inhibits the crystalliza-

tion of the powellite phase in the bulk when T_C< 750 ^ꢀC. In this

temperature range only a very thin partially crystallized layer

(10e15 mm) - probably constituted of apatite crystals - is formed

from the surface (not shown in Fig. 9). When T_Cꢃ 750 ^ꢀC, apatite

crystallizes both from the surface and the bulk of MoNd glass

(Figs. 2f and 9cej) as in Nd glass (Fig. 8). Nevertheless, contrary to

Nd glass, it is here possible to observe small (<1

mm) globular

heterogeneities (Mo-rich, as analysed qualitatively by EDX) be-

tween and close to the apatite crystals (Fig. 9ceg). According to the

study of the heat treated Mo glass presented above, these globular

heterogeneities probably originate from a phase separation phe-

nomenon in the neighborhood of the apatite crystals where the

N. Chouard et al. / Journal of Alloys and Compounds 671 (2016) 84e99

93

formed inside the separated globular particles during cooling of the

melt probably because the time that the sample reaches room

temperature was longer than at 950 and 1000 ^ꢀC. Nevertheless, the

intensity of these vibration bands remains very low compared to

the intensity of the vibration bands associated with powellite

crystals in the Mo glass heat treated between 650 and 900 ^ꢀC. This

indicates that the crystallization detected in the Mo glass heat

treated at 1100 ^ꢀC is very weak and explains why CaMoO₄and

Na₂MoO₄were not observed by XRD. Thus, this little crystallization

artefact occurring during sample cooling will be not considered for

the explanation of crystallization mechanisms of molybdate phases

versus temperature.

3.3.2. Crystallization of the Mo and Nd-rich glass

The Raman spectra of the heat treated MoNd glass are presented

in Fig. 10b. For the MoNd glass heat treated at 650 and 700 ^ꢀC, the

only vibration bands observed are those attributed to the sym-

metric stretching vibration of [MoO₄]^2-and [SiO₄]^4-(Q_n) entities in

the glass which is consistent with the XRD and SEM results (Figs. 3c

and 9). At 750 ^ꢀC, a new band is observed at about 860 cm^ꢂ1, which

is associated with the beginning of the crystallization of the apatite

Ca₂Nd₈(SiO₄)₆O₂phase (Fig. 1b) [33,37]. In accordance with XRD

(Fig. 3c), the intensity of this band decreases for high T_C(1100 ^ꢀC)

due to the decreasing amount of apatite crystals formed in the

glass. It is also interesting to note that the vibration band at about

912 cm^ꢂ1of the [MoO₄]^2-tetrahedral entities in the glass signiﬁ-

cantly shifts toward lower frequencies (~900 cm^ꢂ1) as soon as

apatite crystallizes. This phenomenon, linked to the evolution of

the concentration of neodymium - and probably also of calcium - in

the residual glass depleted in Nd₂O₃and CaO because of apatite

crystallization, has already been observed and explained in detail

elsewhere [37]. It is linked to the evolution of the average ﬁeld

strength of the charge compensator cations (Na^þ, Ca^2þ, Nd^3þ) in the

neighborhood of [MoO₄]^2-entities in the glass. Consequently, the

shift observed for the vibration band of the [MoO₄]^2-entities in the

glass for the heat treated MoNd glass depends on the amount of

apatite formed. Moreover, it has also been observed that this shift

depends on the glassy regions of the sample that are probed by the

exciting laser beam during Raman experiments: the closer to the

apatite crystals, the more depleted in neodymium and calcium is

the residual glass and the lower is the frequency of the vibration

band of [MoO₄]^2-entities. From T_C¼ 800e950 ^ꢀC, new vibration

bands associated with the crystallization of CaMoO₄and Na₂MoO₄

are detected (compare Fig. 10b with Fig. 1b), whereas these crys-

talline phases were not observed by XRD (Fig. 3c). As the SEM

images show the presence of Mo-rich globular heterogeneities

between the apatite crystals (Fig. 9), it seems that these particles

are not totally amorphous but contain molybdate crystals, at least

when T_C> 750 ^ꢀC. However, the amount of crystalline molybdates

is probably too low to be detected by XRD. Finally, beyond 950 ^ꢀC,

no crystallized molybdate phases can be detected on the Raman

spectra which can be explained by the fact that such phases are no

more stable in this T_Crange as indicated above [42,43]. Thus, mo-

lybdates would not crystallize within the Mo-rich globular het-

erogeneities which would remain amorphous after quenching

when T_C> 950 ^ꢀC. Indeed, for these high T_C, Raman spectra only

show the vibration bands associated with the apatite crystalline

phase, and with the [MoO₄]^2-and [SiO₄]^4-(Q_n) entities in the re-

sidual glass.

Fig. 11. (a) Optical absorption spectra of Nd^3þions (⁴I_9/2

/

²P_1/2transition) recorded

at T ¼ 10 K of the quenched and heat treated (T_C¼ 650e1100 ^ꢀC) Nd glass. The

spectrum of apatite Ca₂Nd₈(SiO₄)₆O₂ceramic is given at the top. For the ceramic, the

two intense bands correspond to Nd^3þions located in the 6 h and 4f sites. The small

band indicated by an asterisk (*) on the ceramic spectrum is probably due to Nd^3þions

also located in the 4f site of apatite crystals but that are slightly different to that of the

most intense band (this small band was not considered for the spectra simulation used

in Fig. 12). (b) Optical absorption spectra of Nd^3þions (⁴I_9/2

/

⁴G_5/2,²G_7/2hypersen-

sitive transition) recorded at T ¼ 10 K of the heat treated (T_C¼ 650e950 ^ꢀC) MoNd

glass. The spectrum of apatite Ca₂Nd₈(SiO₄)₆O₂ceramic is shown at the top for com-

parison. G: bands due to Nd^3þions located in the glass. A: bands due to Nd^3þions

located in the 6 h and 4f sites of apatite crystals. P: bands due to Nd^3þions in powellite

crystals [22,37].

by optical absorption spectroscopy on the quenched and heat

treated Nd glass to follow the incorporation of Nd^3þions in apatite

crystals. The evolution with T_Cof the spectra corresponding to the

⁴I_9/2

/

²P_1/2transition is shown in Fig. 11a. For comparison, the

optical absorption spectrum of the apatite Ca₂Nd₈(SiO₄)₆O₂ceramic

is also given. Between T_C¼ 650 and 750 ^ꢀC, only one broadband

(~100 cm^ꢂ1width) with maximum at 23217 cm^ꢂ1is detected. This

band is similar to that of neodymium in the quenched Nd glass

which shows (bottom spectrum in Fig. 11a) that, for this T_Crange,

the great majority of Nd^3þions remained in the glassy phase (for

T_C¼ 750 ^ꢀC only a very small amount of apatite crystals formed on

glass surface was detected by XRD and SEM (Figs. 3b and 8b)). From

800 ^ꢀC, two new narrower bands are detected on the spectra at

about 23143 and 23245 cm^ꢂ1. By comparison with the spectrum of

the apatite Ca₂Nd₈(SiO₄)₆O₂ceramic (top spectrum in Fig. 11a),

these two new contributions can be attributed to the absorption of

3.4. Optical absorption

3.4.1. Crystallization of the Nd-rich glass

As strong structural and microstructural evolutions with T_Chave

been reported in the previous sections, a study has been conducted

94

N. Chouard et al. / Journal of Alloys and Compounds 671 (2016) 84e99

Fig. 12. Evolution with T_Cof the position of the optical absorption bands (⁴I_9/2

/

²P_1/2transition) of Nd^3þions in the heat treated Nd glass: (a) 4f site, (b) 6 h site, (c) residual glass.

The position of the bands associated with Nd in the 6 h and 4f sites of the apatite Ca₂Nd₈(SiO₄)₆O₂ceramic (AC) is also shown (B).

Nd^3þions in the 6 h and 4f sites respectively. All spectra were

simulated with three Gaussian bands: one for Nd^3þions in the

residual glass and one for each of the sites (6 h, 4f) occupied by

Nd^3þions in apatite crystals. The evolution of the position of these

three bands with T_Cis shown in Fig. 12. Firstly, concerning the

evolution of the position of the band associated with Nd^3þions in

the glassy phase (Fig.12c), for T_C< 800 ^ꢀC no signiﬁcant evolution is

observed, but as soon as apatite crystallized in the bulk, a signiﬁ-

cant shift towards low energies is put in evidence (the shift

maximum (17.7 cm^ꢂ1) is observed for T_C¼ 850 ^ꢀC). When

T_Cꢃ 900 ^ꢀC, a shift in the opposite direction, towards higher en-

ergies, is observed to reach a value (23215 cm^ꢂ1) for T_C¼ 1100 ^ꢀC

only slightly lower than that of the quenched Nd glass. Concerning

now the evolution with T_Cof the bands associated with Nd^3þions

in the apatite crystals (4f and 6 h sites), a signiﬁcant shift with T_Cof

their position (Fig. 12a,b) and an evolution of their relative in-

tensities (increase of the intensity of the 6 h band, Fig. 11a) towards

that of the Ca₂Nd₈(SiO₄)₆O₂ceramic are put in evidence.

4

The variation with T_Cof the position of the I_9/2

/

²P_1/2tran-

sition associated with the neodymium ions remaining in the glassy

phase (Fig. 12c) indicates that the environment of these ions is

signiﬁcantly modiﬁed after crystallization. This is related to a

variation of the average NdeO bond covalence linked to an evolu-

tion of the average NdeO bond length in the glass surrounding the

crystals (nephelauxetic effect) [41,46]. For instance, a band shift

towards lower energies indicates an increase of the NdeO bond

N. Chouard et al. / Journal of Alloys and Compounds 671 (2016) 84e99

95

Fig. 13. Schematic representation proposed for the structure of peralkaline aluminoborosilicate glasses containing sodium, calcium and neodymium according to previous works

[3,47]. In peralkaline glass compositions, as it is the case of the Nd glass of the present study, the amount of charge compensator cations (here Na^þand Ca^2þ) available is higher than

the amount of AlO^-₄entities. In this ﬁgure are shown: bridging oxygen atoms (BOs), non-bridging oxygen atoms (NBOs), SiO₄tetrahedra with or without NBOs, AlO₄^-and BO^-₄

tetrahedra charge compensated by Na^þor Ca^2þcations, BO₃triangles and Nd^3þcations with their nearest neighbor oxygen atoms. According to both molecular dynamics simulation

results [51e53] and bond valence-bond length considerations [3] reported on RE-bearing silicate and aluminosilicate glasses, the nearest neighbor oxygen atoms of neodymium

would be mainly NBOs. Nd^3þions would be preferentially located in Na^þ, Ca^2þand NBO-rich regions of the network (depolymerized regions indicated by DR) separated by BO-rich

regions (polymerized regions indicated by PR). In DR, Na^þor Ca^2þcations would both contribute to the formation of NBOs and to compensate the negative charge excess of the

NBOs present in the neighborhood of Nd^3þions. The dotted lines separate DR and PR regions.

covalence and a decrease of the NdeO distance [46]. Thus, the band

shift towards lower energies for 750 ^ꢀC < T_C< 900 ^ꢀC indicates that

the average NdeO distance in the residual glass has decreased in

comparison with the average NdeO distance in the quenched glass

and in the samples heat treated at T_Cꢄ 750 ^ꢀC. Then, for

900 ^ꢀC < T_C< 1100 ^ꢀC, the average NdeO distance increases again to

ﬁnally reach an average NdeO distance close to that of the

quenched glass. This evolution of the NdeO distance in the glassy

phase with T_Ccan be explained by the variation of the total amount

of apatite crystals formed (Figs. 3b and 8) and by the consecutive

evolution of the relative proportions of Na^þand Ca^2þcations

available in the glassy phase for local charge compensation in the

neighborhood of the non-bridging oxygen atoms (NBOs) sur-

rounding Nd^3þions (see the structural scheme shown Fig. 13 pro-

posing that Nd^3þions are located in the (Na^þ,Ca^2þ)-rich

depolymerized regions of the glass structure) [3,47]. Indeed, the

crystallization of apatite, that contains two calcium atoms by for-

mula unit, induces a depletion of Ca^2þions - and thus a relative

enrichment in Na^þcations - in the residual glass. Thus, using bond

valence - bond length considerations [46], it can be shown that the

Na^þ/Ca^2þconcentration ratio increase in the surrounding of Nd^3þ

ions induces a decrease of the average NdeO distance. The opposite

tendency (NdeO increase) observed when T_Cꢃ 900 ^ꢀC, can be

explained by the progressive decrease with T_Cof the amount of

apatite crystals - and thus by the decrease of the Na^þ/Ca^2þcon-

centration ratio in the residual glass - as shown by XRD and SEM

(Figs. 3b and 8).

between the 6 h and 4f sites from a random occupancy of Nd^3þions

between the 6 h and 4f sites for the lowest temperature to an oc-

cupancy closer to that existing in the Ca₂Nd₈(SiO₄)₆O₂ceramic

apatite prepared at 1500 ^ꢀC (preferential occupancy of 6 h sites by

TR^3þions [48,49]) for the highest temperature. A similar evolution

of ions occupancy was already observed and discussed in Ref. [50]

about the effect of the heating temperature on the structure of the

apatite crystals formed in a Nd-rich soda-lime aluminoborosilicate

glass. The lower thermal energy available at low T_Cfor cations

diffusion is probably not sufﬁcient to enable cationic organization

inside apatite crystals during heat treatment to minimize their free

energy which would explained the evolution observed.

3.4.2. Crystallization of the Mo and Nd-rich glass

4

The evolution with T_Cof the spectra corresponding to the I_9/

/

²P_1/2transition of Nd^3þin MoNd glass (not shown) is very

2

similar to that of the heat treated Nd glass (Fig. 11a). The bands

associated with Nd^3þions in the 6 h and 4f sites of the apatite

crystals are detected above 700 ^ꢀC and present a similar evolution

with T_Ctowards the spectra of the Ca₂Nd₈(SiO₄)₆O₂ceramic. This

conﬁrms the XRD results presented above that have shown that the

presence of MoO₃does not strongly disturb apatite crystallization.

Nevertheless, following only this transition, the presence of a

fraction of Nd^3þions in the molybdate crystals (detected by Raman

spectroscopy in the MoNd heat treated glass, Fig. 10b) cannot be

excluded. Indeed, in a previous work [37], we showed that the

4

/

²P_1/2band associated with Nd^3þin the 6 h

4

position of the I_9/2

The variation with T_Cof the intensity and position of the I_9/

²P_1/2transitions associated with Nd^3þions in the 4f and 6 h

site of apatite crystals occurs at the same position as the band

/

2

associated with Nd^3þin powellite crystals. However, the study of

sites of apatite crystals (Figs. 11a and 12a,b) can be explained by a

strong evolution with T_Cof the distribution of Ca^2þand Nd^3þions

4

the evolution with T_Cof the I_9/2

/

⁴G_5/2,²G_7/2transition clearly

demonstrates that a small fraction of Nd^3þions is detected in the

96

N. Chouard et al. / Journal of Alloys and Compounds 671 (2016) 84e99

Fig. 14. Schematic representation of phase separation and crystallization phenomena of molybdates in a soda-lime aluminoborosilicate glass as a function of both Nd₂O₃addition

and T_C, according to this work. (a) crystallization of Mo glass at low T_C(<750 ^ꢀC) / formation of facetted powellite crystals throughout the sample; (b) effect of the addition of

Nd₂O₃to Mo glass (MoNd glass) at low T_C(<750 ^ꢀC) / suppression of molybdate crystallization; (c) effect of increasing T_C(ꢃ750 ^ꢀC) / crystallization of apatite from the surface

and in the bulk; (d) phase separation and crystallization (CaMoO₄þ Na₂MoO₄) at T_Cof a globular Mo-rich phase (in yellow) in the neighborhood of the apatite crystals formed at

T_Cꢃ 750 ^ꢀC (when T_Cꢃ 1000 ^ꢀC, molybdate crystallization is not observed, Fig. 10b). P: powellite facetted crystals, RG: residual glass remaining between the crystals, HG: ho-

mogeneous glass without any phase separation and crystallization, A: Apatite Ca₂Nd₈(SiO₄)₆O₂crystals (pink). The evolution of the color of the residual glass shown in the ﬁgure

between (b) and (c) indicates that this one is depleted in Nd after apatite crystallization. (For interpretation of the references to colour in this ﬁgure legend, the reader is referred to

the web version of this article.)

powellite crystals when 750 ꢄ T_Cꢄ 900 ^ꢀC (Fig. 11b) which is close

to the T_Crange where powellite is detected by Raman spectroscopy

(Fig. 10b).

with our previous works [34,37]. We proposed that it is the close

relationship existing between Nd^3þcations and [MoO₄]^2-tetrahe-

dral entities in the glassy network [37] that is at the origin of the

increase of molybdenum solubility and of the disappearance of the

phase separation and crystallization of powellite when neodymium

is added to the glass composition. However, in accordance with the

results concerning the glass composition with only Nd₂O₃(Nd

glass, Figs. 3b and 8), when T_Cꢃ 750 ^ꢀC apatite crystallizes ﬁrstly

only from the surface of the MoNd glass and then also in the bulk

(Figs. 3c and 9). Consequently, the Mo-rich residual glass remaining

between and close to the apatite crystals is depleted in neodymium

(Fig. 14c). As a matter of fact, the composition of this residual glass

becomes closer to that of the Mo glass that does not contain Nd₂O₃

and for which Mo-rich phase separation or powellite crystallization

may occur during heat treatment (Fig. 7). In these conditions, the

solubility of molybdenum is affected which can explain the for-

mation of Mo-rich globular particles containing CaMoO₄and

Na₂MoO₄crystals between apatite crystals (Fig. 14d). The fact that

the formation of Na₂MoO₄is almost never observed in the heat

treated Mo glass (Fig. 10a) whereas it is detected along with

CaMoO₄in the heat treated MoNd glass (Fig. 10b) can be explained

by the increase of the Na/Ca ratio in the residual glass close to the

apatite crystals (Ca depletion).

Thus, this study of thermal stability versus temperature con-

ﬁrms that neodymium addition to a Mo-rich glass inhibits the

separation and crystallization of molybdate phases as long as

apatite crystallization is avoided (T_C< 750 ^ꢀC). Otherwise, the

separation of a Mo-rich phase followed by molybdates crystalliza-

tion appears again in the Nd-depleted regions of the residual glass

(i.e. close to apatite crystals). In particular, it is important to avoid

the crystallization of Na₂MoO₄that could incorporate radioactive

Cs and is poorly durable, and consequently prohibited for the nu-

clear application envisaged here. As during the interim storage and

geological disposal of nuclear waste glasses their temperature is

expected to always remain about 100 ^ꢀC lower than Tg (Tg ~ 580 ^ꢀC

for the simpliﬁed nuclear waste MoNd glass studied in this work),

no apatite crystallization is expected during storage (T < 750 ^ꢀC,

Fig. 10b) and all the RE present could help the incorporation of

Mo^6þcations in the glass structure, thus preventing molybdates

separation and crystallization. Nevertheless, it is important to un-

derline that during nuclear glass production the melt (at

1150e1300 ^ꢀC) is not rapidly quenched below 750 ^ꢀC but it is

4. Discussion

4.1. Effect of neodymium addition on the crystallization of

molybdate phases

According to the results presented above, the comparison be-

tween the quenched and heat treated Mo and MoNd glasses allows

putting in evidence compositional effects on the phase separation

and crystallization tendency of molybdate phases in these glasses:

- According to SEM, Nd₂O₃addition to the glass composition in-

hibits the phase separation of molybdates in the quenched

glasses.

- According to XRD measurements (Figs. 3b and 4), Nd₂O₃addi-

tion also inhibits the crystallization of molybdates in the heat

treated glass samples when T_C< 750 ^ꢀC. This observation is

conﬁrmed by SEM for T_C¼ 650 and 700 ^ꢀC (compare Fig. 7aeb to

Fig. 9aeb). However, for the MoNd glass heated at T_Cꢃ 750 ^ꢀC,

Mo-rich globular heterogeneities are observed between the

apatite crystals formed in the bulk or from the surface

(Fig. 9ceh). Because of their characteristic morphology and their

location only in the neighborhood of the apatite crystals, it

seems that these Mo-rich particles come from a phase separa-

tion phenomenon in the Nd-depleted supercooled melt close to

the apatite crystals formed at T_C. According to Raman spectra

(Fig. 10b), these globular particles contain CaMoO₄and

Na₂MoO₄crystals, but in a concentration too low to be detected

by XRD (Fig. 3c).

All these observations can be summarized by a simple scheme

(Fig. 14) that can be described as follows. When glass composition

does not contain Nd₂O₃(Mo glass), a Mo-rich phase is formed

throughout the sample during the thermal treatment, either as

facetted powellite crystals (T_C< 950 ^ꢀC, Fig. 14a) or as globular

amorphous particles (T_Cꢃ 950 ^ꢀC, Fig. 7). When Nd₂O₃is added to

the glass composition (MoNd glass), the crystallization of the

powellite phase disappears for thermal treatments at T_C< 750 ^ꢀC

(Figs. 9aeb and 14b), which means that neodymium addition im-

proves the solubility of molybdenum in the glass in accordance

poured in

a metallic canister and undergo natural cooling

N. Chouard et al. / Journal of Alloys and Compounds 671 (2016) 84e99

97

Fig. 15. Schematic representation of crystallization and phase separation phenomena close to the surface of Nd (top) and MoNd (bottom) glasses as a function of time at a given

temperature T_C. (a) quenched Nd glass; (b,c) apatite crystallization from the surface of the Nd glass with increasing the duration of heat treatment; (d) quenched MoNd glass; (e)

apatite crystallization from the surface of the MoNd glass followed by the separation of Mo-rich particles between apatite crystals during heat treatment; (f) nucleation and growth

of new apatite crystals from the region containing Mo-rich particles with increasing the duration of heat treatment of MoNd glass. To simplify the ﬁgure, apatite crystals nucleating

directly in the bulk (observed in Figs. 8 and 9) have not been represented. G: glass; A: Apatite Ca₂Nd₈(SiO₄)₆O₂crystals (in pink); M; Mo-rich globular partially crystallized particles

(in yellow). (For interpretation of the references to colour in this ﬁgure legend, the reader is referred to the web version of this article.)

(~1 ^ꢀC min^ꢂ1). During this cooling (several hours), partial crystal-

lization of the melt could occur. In a previous paper performed on

similar glass compositions containing both MoO₃and Nd₂O₃[37]

we showed that if the amount of Nd₂O₃remains lower than

14 wt% (i.e. 3.0 mol%) and higher than 8 wt% (i.e. 1.65 mol%) no

crystallization (apatite or molybdates) occurs during controlled

cooling of the melt from 1300 ^ꢀC to room temperature at

apatite crystals formed during heat treatment. Nevertheless, SEM

observations have put in evidence that the thickness of the crys-

tallized layer (mainly constituted of apatite crystals) and growing

from the surface was signiﬁcantly higher when MoO₃was present

in glass composition (Figs. 8 and 9). Important microstructural

differences were also observed by SEM between the crystallized

layers formed in Nd and MoNd glasses. Indeed, Fig. 9iej shows that,

during the heat treatment of MoNd glass, numerous new apatite

crystals nucleate and grow between the apatite crystals that are

growing from the surface. Such a phenomenon, which was not

observed for Nd glass (Fig. 8), may explain the higher thickness of

the crystallized layer in MoNd glass. As numerous Mo-rich globular

heterogeneities appear between apatite crystals during their

growth in the MoNd glass (Fig. 9c), we propose in accordance with

a previous paper [34] that these Mo-rich particles are at the origin

of the nucleation of the new apatite crystals in the crystallized layer

and act as nucleating agent. Nevertheless, the fact that these apatite

crystals nucleate directly or not from the surface of these Mo-rich

particles remains to be checked by direct observation by TEM for

instance. A scheme is proposed in Fig. 15 that compares the pro-

cesses of crystallization from the surface in Nd and MoNd glasses,

with the Mo-rich globular particles acting as nucleating agent for

apatite and leading to an increase of the density of apatite crystals

in the crystallized layer and of its thickness. The same processes

also occur in the neighborhood of the apatite crystals formed in the

bulk of glass samples (Fig. 9). The following sequence explains the

formation of the complex crystallized layer in MoNd glass

(Fig. 15def): (1) heterogeneous crystallization of apatite from glass

surface and growth towards the bulk; (2) neodymium depletion in

the residual glass between the apatite crystals in the layer; (3)

decrease of molybdenum solubility in this residual glass; (4) sep-

aration of Mo-rich globular particles followed by molybdates

crystallization within them; (5) nucleation of new apatite crystals

from the Mo-rich particles (nucleating agent) and growth toward

1

^ꢀC min^ꢂ1. However, for the MoNd glass studied in the present

paper which is slightly above the upper limit - because it contains

16 wt% Nd₂O₃(Table 1) - we showed in Ref. [37] that during cooling

at 1 ^ꢀC min^ꢂ1the melt partially crystallized in apatite followed by

phase separation and crystallization of Mo-rich particles. Therefore,

to beneﬁt from the positive effect of the presence of rare earths on

the solubility of molybdenum both during nuclear glass synthesis

and storage, it is very important to introduce a controlled amount

of rare earths in the composition (enough to prevent molybdates

crystallization but not too much to remain below rare earths sol-

ubility limit in the system).

4.2. Effect of molybdenum addition on the crystallization of the

apatite phase

XRD results have shown that the evolution of the patterns

(Fig. 3b and c) and lattice parameters (Fig. 6) were very similar for

the Nd-rich glasses without (Nd glass) or with molybdenum (MoNd

glass). Independently of the presence of MoO₃, apatite crystalliza-

tion is not detected below T_C¼ 750 ^ꢀC. Moreover, the optical ab-

sorption results have shown that the presence of MoO₃in the glass

composition had no impact on the evolution of the environment of

Nd^3þions in the apatite crystals formed during heat treatment

(same evolution of spectra with T_Cfor Nd and MoNd glasses). All

these results demonstrate that the addition of molybdenum to the

Nd glass has no effect on the temperature above which apatite

crystallization is detected (T_Cꢃ 750 ^ꢀC) and on the structure of the

98

N. Chouard et al. / Journal of Alloys and Compounds 671 (2016) 84e99

the bulk. In comparison, in Nd glass, the apatite crystals forming the

layer have mainly grown from the surface (Fig. 15aec).

glass remaining between these crystals is depleted in Nd. In the

regions of the samples where these Mo-rich particles are observed

numerous new apatite crystals nucleate and grow between the

previous apatite crystals. In comparison with the glass with only Nd

the microstructure is thus signiﬁcantly modiﬁed: increase of the

thickness of the apatite layer growing from the surface and higher

density of apatite crystals (Fig. 15f). It is proposed that the Mo-rich

particles are at the origin of the nucleation of these new apatite

crystals and act as nucleating agent. Nevertheless, in comparison

with the glass with only Nd, the presence of MoO₃in the compo-

sition has not effect on the structure of the apatite crystals formed

during heat treatments and on the environment of Nd^3þions in

these crystals.

All the results presented in this work have thus shown that the

simultaneous presence of Nd₂O₃and MoO₃in the simpliﬁed soda-

lime aluminoborosilicate nuclear glass increases the glass stability

if the heating temperature remains lower than 750 ^ꢀC which is

signiﬁcantly higher than the glass transformation temperature

(T_g~ 580 ^ꢀC). This enables to avoid the crystallization of Na₂MoO₄

that is prohibited for the nuclear application envisaged here.

Nevertheless, to beneﬁt from the positive effect of the presence of

rare earths on the solubility of molybdenum also during nuclear

glass synthesis and more particularly to avoid partial crystallization

of apatite and molybdates during natural cooling of the melt in

canisters (~1 ^ꢀC min^ꢂ1), it is very important to introduce a

controlled amount of rare earths in the composition (enough to

prevent molybdates crystallization but not too much to remain

below rare earths solubility limit in the system) [37].

5. Conclusions

Mo and Nd are abundant ﬁssion products in nuclear wastes. The

wish to increase in the future the amount of waste in nuclear

glasses may be problematic because of the risk of separation and

crystallization of Mo-rich and Nd-rich phases due to the high ﬁeld

strength of Mo^6þand Nd^3þcations. In order to evaluate indepen-

dently the effect of the introduction of high amounts of MoO₃or

Nd₂O₃on the thermal stability of a simpliﬁed nuclear glass and

then to study the consequence of the simultaneous presence of

these two oxides in this glass, three soda-lime aluminoborosilicate

glasses containing only MoO₃or Nd₂O₃or both MoO₃and Nd₂O₃

have been prepared and heat treated above T_gbetween 650 and

1100 ^ꢀC. The effect of the heating temperature on the phase sepa-

ration and crystallization phenomena has been investigated for

these glasses and comparison has been made with crystalline

reference Mo-rich or Nd-rich phases.

In the glass with only Mo, CaMoO₄(powellite) crystallization

(facetted crystals) has been detected between 650 and 900 ^ꢀC and is

uniform throughout the bulk of the samples (Fig. 14a). The evolu-

tion with heating temperature of the structure of CaMoO₄crystals

has been followed. In this temperature range, the small Mo-rich

phase separated amorphous globular particles (~50 nm size) that

are present in the quenched glass act as heterogeneous nucleation

sites for the crystallization of CaMoO₄(but not Na₂MoO₄as shown

by XRD and Raman spectroscopy). Above 950 ^ꢀC, CaMoO₄is no

more stable and does not crystallize anymore, but Mo-rich amor-

Acknowledgments

phous globular particles (~1

ration phenomenon (Fig. 7).

mm size) are formed by a phase sepa-

ꢀ

The authors would like to thank the Chaire ParisTech “Ingenierie

In the glass with only Nd, no phase separation or crystallization

has been observed below 750 ^ꢀC, this glass is thus more stable than

the glass with only Mo. Above this temperature, Ca₂Nd₈(SiO₄)₆O₂

(apatite) crystals are ﬁrstly detected only close to the glass surface

(heterogeneous nucleation, Fig. 15b) and then, at higher tempera-

ture, also in the bulk (Fig. 8). The evolution with heating temper-

ature of the structure of apatite crystals, that has been followed by

XRD and optical absorption spectroscopy, has shown an increase of

their lattice parameters that may be explained by the evolution

with temperature of the ordering of Ca^2þand Nd^3þions in the 6 h

and 4f sites of the apatite structure. The evolution with the heating

temperature of the environment of the Nd^3þions remaining in the

residual glass has been also followed by optical absorption spec-

troscopy and correlated with the evolution of the amount of apatite

crystals present in the glass.

In the glass with both Mo and Nd, it has been shown that, as long

as Nd^3þions remain solubilized in the glass structure (i.e. as long as

the apatite phase does not crystallize in the glass composition

studied here, which is the case when the heating temperature re-

mains lower than 750 ^ꢀC), the presence of Nd₂O₃inhibits the Mo-

rich phase separation and the crystallization of powellite

(Fig. 14aeb) in comparison with what is observed for the glass with

only Mo. Consequently, neodymium addition improves the solu-

bility of molybdenum in the glass in agreement with our previous

studies [34,37] and increases its thermal stability. It is proposed

that this beneﬁcial effect is due to a close relationship existing

between Nd^3þions and [MoO₄]^2-tetrahedral entities in the regions

rich in non-bridging oxygen atoms of the glassy network. However,

as soon as the apatite phase is formed either from the surface or in

the bulk of the glass (i.e. when the heating temperature is higher

than 750 ^ꢀC), the separation of globular Mo-rich particles con-

taining CaMoO₄and Na₂MoO₄crystals is observed close and be-

tween the apatite crystals (Figs. 14d and 15e) because the residual

ꢀ

Nucleaire” funded by Areva for ﬁnancial support. The authors

would also like to acknowledge O. Boudouma of the ISTEP from the

University of Paris VI for his grateful help for the ﬁeld-emission gun

scanning electron microscopy. P. Vermaut of the Institut de

Recherche de Chimie Paris (IRCP) is also gratefully acknowledged

for the characterizations by transmission electron microscopy.

References

[1] I.W. Donald, Waste Immobilization in Glass and Ceramic Based Hosts.

Radioactive, Toxic and Hazardous Wastes, Wiley, Chichester, 2010.

[2] M.I. Ojovan, W.E. Lee, New Developments in Glassy Nuclear Wasteforms, Nova

Science Publishers, New York, 2007.

ꢀ

[3] D. Caurant, P. Loiseau, O. Majerus, V. Aubin-Chevaldonnet, I. Bardez,

A. Quintas, Glasses, Glass-ceramics and Ceramics for Immobilization of Highly

Radioactive Nuclear Wastes, Nova Science Publishers, New York, 2009.

[4] C.M. Jantzen, Development of glass matrices for high level radioactive waste,

in: M.I. Ojovan (Ed.), Handbook of Advanced Radioactive Waste Conditioning

Technologies, Woodhead Publishing Series in Energy, Cambridge (, 2011, pp.

230e292.

[5] C.M. Jantzen, K.G. Brown, J.B. Pickett, Durable glass for thousands of years, Int.

J. Appl. Glass Sci. 1 (2010) 38e62.

[6] A. Kidari, J.-L. Dussossoy, E. Brackx, D. Caurant, M. Magnin, I. Bardez-Giboire,

Lanthanum and neodymium solubility in simpli ﬁ ed SiO₂-B₂O₃-Na₂O-Al₂O₃-

CaO high level waste glass, J. Am. Ceram. Soc. 95 (2012) 2537e2544.

[7] J.-L. Dussossoy, S. Schuller, T. Advocat, The Formulation of New High Level

Waste Glasses with an Increased Waste Loading Charges, Extended abstracts

of the International Congress on Glass (CD-ROM), Strasbourg, France, 2007.

[8] J.-M. Gras, R. Do Quang, H. Masson, T. Lieven, C. Ferry, C. Poinssot, M. Debes, J.-

M. Delbecq, Perspectives on the closed fuel cycle e Implications for high-level

waste matrices, J. Nucl. Mater. 362 (2007) 383e394.

[9] I. Bardez, D. Caurant, J.L. Dussossoy, P. Loiseau, C. Gervais, F. Ribot,

D.R. Neuville, N. Baf ﬁ er, C. Fillet, Development and characterization of rare-

earth rich glass matrices envisaged for the immobilization of concentrated

nuclear waste solutions, Nucl. Sci. Eng. 153 (2006) 272e284.

ꢀ

[10] M. Magnin, S. Schuller, C. Mercier, J. Trebosc, D. Caurant, O. Majerus, F. Angeli,

T. Charpentier, Modi ﬁ cation of molybdenum structural environment in bo-

rosilicate glasses with increasing content of boron and calcium oxide by ⁹⁵Mo

MAS NMR, J. Am. Ceram. Soc. 94 (2011) 4274e4282.

[11] E. Gasnier, I. Bardez-Giboire, V. Montouillout, N. Pellerin, M. Allix, N. Massoni,

N. Chouard et al. / Journal of Alloys and Compounds 671 (2016) 84e99

99

ꢀ

S. Ory, M. Cabie, S. Poissonnet, D. Massiot, Homogeneity of peraluminous

SiO2eB2O3eAl2O3eNa2OeCaOeNd2O3 glasses: effect of neodymium con-

tent, J. Non Cryst. Solids 405 (2014) 55e62.

[34] N. Chouard, D. Caurant, O. Majerus, D. Thibault, J.-L. Dussossoy, S. Klimin,

D. Pytalev, R. Baddour-Hadjean, J.-P. Pereira-Ramos, Effect of MoO₃, Nd₂O₃

and RuO₂on the crystallization of soda-lime aluminoborosilicate glasses,

J. Mater. Sci. 50 (2015) 219e241.

ꢁ ꢀ

[12] R. Guillaumont, Elements Chimiques

a Considerer dans l'aval du Cycle

ꢀ

Nucleaire, C. R. Chim. 7 (2004) 1129e1134.

[35] A. Quintas, D. Caurant, O. Majerus, J.L. Dussossoy, T. Charpentier, Effect of

[13] G. Calas, M. Le Grand, L. Galoisy, D. Ghaleb, Structural role of molybdenum in

nuclear glasses: an EXAFS study, J. Nucl. Mater. 322 (2003) 15e20.

[14] O. Pinet, J.L. Dussossoy, C. David, C. Fillet, Glass matrices for immobilizing

nuclear waste containing molybdenum and phosphorus, J. Nucl. Mater. 377

(2008) 307e312.

changing the rare earth cation type on the structure and crystallization

behavior of an aluminoborosilicate glass, Phys. Chem. Glas. Eur. J. Glass Sci.

Technol. B 49 (2008) 192e197.

ꢀ

[36] D. Caurant, O. Majerus, P. Loiseau, I. Bardez, N. Baf ﬁ er, J.L. Dussossoy, Crys-

tallization of neodymium-rich phases in silicate glasses developed for nuclear

[15] R.J. Short, R.J. Hand, N.C. Hyatt, Molybdenum in nuclear waste glasses.

incorporation and redox state, Mat. Res. Soc. Symp. Proc. 757 (2003) 141e146.

[16] S. Kroeker, C.S. Higman, V.K. Michaelis, N.B. Svenda, S. Schuller, Precipitation

of mixed-alkali molybdate during HLW vitri ﬁ cation, Mat. Res. Soc. Symp. Proc.

1265 (2010) 95e101.

waste immobilization, J. Nucl. Mater. 354 (2006) 143e162.

[37] N. Chouard, D. Caurant, O. Majerus, J.-L. Dussossoy, A. Ledieu, S. Peuget,

ꢀ

R. Baddour-Hadjean, J.-P. Pereira-Ramos, Effect of neodymium oxide on the

solubility of MoO₃in an aluminoborosilicate glass, J. Non Cryst. Solids 357

(2011) 2752e2762.

[17] J.E.C. Wren, B.J. Greer, V.K. Michaelis, C.S. Higman, S. Kroeker, Multinuclear

magnetic resonance investigation of crystalline alkali molybdates, Inorg.

Chem. 54 (2015) 9853e9861.

[38] Y.G. Petrosyan, E.V. Tkachenko, V.M. Zhukovskii, Phase diagrams of the sys-

tems Na₂MoO₄-MMoO₄(M

1381e1383.

¼

Ca, Sr or Ba), Inorg. Mater. 11 (1975)

€

[18] R.J. Short, R.J. Hand, N.C. Hyatt, G. Mobus, Environment and oxidation state of

[39] C.W.F.T. Pistorius, Phase diagrams of sodium tungstate and sodium molybdate

to 45 kbar, J. Chem. Phys. 44 (1966) 4532e4537.

molybdenum in simulated high level nuclear waste glass compositions,

J. Nucl. Mater. 340 (2005) 179e186.

[19] R.J. Short, R.J. Hand, N.C. Hyatt, An investigation into the oxidation state of

molybdenum in simpli ﬁ ed high level waste glass compositions, Mat. Res. Soc.

Symp. Proc. 807 (2004) 169e174.

[40] D.A. Long, Raman Spectroscopy, McGraw-Hill, New-York, 1977.

[41] K. Gatterer, G. Pucker, W. Jantscher, H.P. Fritzer, S. Arafa, Suitability of Nd(III)

absorption spectroscopy to probe the structure of glasses from the ternary

system Na₂OeB₂O₃eSiO₂, J. Non Cryst. Solids 231 (1998) 189e199.

ꢀ

[20] N. Sawaguchi, T. Yokokawa, K. Kawamura, Mo K-Edge XAFS in Na₂O-K₂O-SiO₂

glasses, Phys. Chem. Glas. 37 (1996) 13e18.

[42] M. Magnin, Etude des Processus de Demixtion et de Cristallisation au Sein de

ꢀ

ꢁ

Liquides Fondus Borosilicates Riches en Oxyde de Molybdene, University P.

and M. Curie, Paris, France, 2009. PhD Thesis.

ꢀ

[21] D. Caurant, O. Majerus, E. Fadel, A. Quintas, C. Gervais, T. Charpentier,

ꢀ

D. Neuville, Structural investigations of borosilicate glasses containing MoO₃

[43] M. Magnin, S. Schuller, D. Caurant, O. Majerus, D. de Ligny, T. Advocat, Phase

by MAS NMR and Raman spectroscopies, J. Nucl. Mater. 396 (2010) 94e101.

[22] D. Caurant, O. Majerus, E. Fadel, M. Lenoir, C. Gervais, O. Pinet, Effect of mo-

separation and crystallization in soda-lime borosilicate glass enriched in

MoO₃studied by in situ Raman spectroscopy at high temperature, in: Pro-

ceedings of the Conference Atalante 2008: Nuclear Fuel Cycles for a Sustain-

able Future, 19e22 May 2008. Montpellier, France, http://hal.archives-

ouvertes.fr/hal-00994577.

ꢀ

lybdenum on the structure and on the crystallization of SiO₂-Na₂O-CaO-B₂O₃

glasses, J. Am. Ceram. Soc. 90 (2007) 774e783.

ꢀ

[23] M. Magnin, S. Schuller, D. Caurant, O. Majerus, D. de Ligny, C. Mercier, Effect of

compositional changes on the structure and crystallization tendency of a

borosilicate glass containing MoO₃, Ceram. Trans. 207 (2009) 59e68.

[24] E. Kashchieva, I. Petrov, L. Aleksandrov, R. Iordanova, Y. Dimitriev, Micro-

structure of multicomponent borosilicate glasses containing MoO₃, Phys.

Chem. Glas. Eur. J. Glass. Technol. B 53 (2012) 264e270.

[44] S.N. Achary, S.J. Patwe, M.D. Mathews, A.K. Tyagi, High-pressure crystal

chemistry of sheelite-type tungstates and molybdates, J. Phys. Chem. Solids 67

(2006) 774e781.

ꢀ

[45] A. Quintas, O. Majerus, D. Caurant, J.-L. Dussossoy, P. Vermaut, Crystallization

of a rare earth-rich aluminoborosilicate glass with varying CaO/Na₂O ratio,

[25] L. Aleksandrov, R. Iordanova, Y. Dimitriev, Glass formation in the MoO₃-

Nd₂O₃-La₂O₃-B₂O₃system, J. Non Cryst. Solids 355 (2009) 2023e2026.

[26] L. Aleksandrov, R. Iordanova, Y. Dimitriev, Glass formation in the MoO₃-La₂O₃-

B₂O₃system, Phys. Chem. Glas. Eur. J. Glass Sci. Technol. B 48 (2007) 242e245.

[27] L. Aleksandrov, T. Komatsu, R. Iordanova, Y. Dimitriev, Study of molybdenum

coordination state and crystallization behavior in MoO₃eLa₂O₃eB₂O₃glasses

by Raman spectroscopy, J. Phys. Chem. Solids 72 (2011) 263e268.

J. Am. Ceram. Soc. 90 (2007) 712e719.

[46] A. Quintas, O. Majerus, M. Lenoir, D. Caurant, K. Klementiev, A. Webb, J.-

ꢀ

L. Dussossoy, In ﬂ uence of the nature of alkali and alkaline-earth cations on the

neodymium environment in a rare-earth rich aluminoborosilicate glass, J. Non

Cryst. Solids 354 (2008) 98e104.

[47] D. Caurant, Spectroscopic investigations on glasses, glass-ceramics and ce-

ramics developed for nuclear waste immobilization, Opt. Spectrosc. 116

(2014) 667e676.

[48] L.W. Schroeder, M. Mathew, Cation ordering in Ca₂La₈(SiO₄)₆O₂, J. Solid State

Chem. 26 (1978) 383e387.

[49] J.A. Fahey, W.J. Weber, F.J. Rotella, An X-ray and neutron powder diffraction

study of the Ca_2þxNd_8-x(SiO₄)₆O_2-0.5xsystem, J. Solid State Chem. 60 (1985)

145e158.

[50] D. de Ligny, G. Panczer, D. Caurant, D.R. Neuville, Contribution of neodymium

optical spectroscopy to the crystal growth study of a silicate apatite in a glassy

matrix, Opt. Mater. 30 (2008) 1694e1698.

[28] L. Aleksandrov, T. Komatsu, K. Shinozaki, T. Honma, R. Iordanova, Structure of

MoO₃-WO₃-La₂O₃-B₂O₃glasses and crystallization of LaMo_1-xW_xBO₆solid

solutions, J. Non Cryst. Solids 429 (2015) 171e177.

ꢀ

[29] S. Gosse, C. Gueneau, S. Bordier, S. Schuller, A. Laplace, J. Rogez,

A thermodynamic approach to predict the metallic and oxide phases pre-

cipitations in nuclear waste glass melts, Procedia Mater. Sci. 7 (2014) 79e86.

[30] M. Stemprok, Geological signi ﬁ cance of immiscibility in fused silicate systems

containing tungsten and molybdenum, Int. Geol. Rev. 17 (1975) 1306e1316.

€

[31] D. Bosbach, T. Rabung, F. Brandt, T. Fanghanel, Trivalent actinide coprecipi-

tation with powellite (CaMoO₄): secondary solid solution formation during

HLW borosilicate-glass dissolution, Radiochim. Acta 92 (2004) 639e643.

[32] X. Wang, G. Panczer, D. de Ligny, V. Motto-Ros, J. Yu, J.L. Dussossoy, S. Peuget,

[51] B. Park, H. Li, L.R. Corrales, Molecular dynamics simulation of

La₂O₃eNa₂OeSiO₂glasses. I. The structural role of La^3þcations, J. Non Cryst.

Solids 297 (2002) 220e238.

[52] A. Bonamartini Corradi, V. Cannillo, M. Montorsi, C. Siligardi, A.N. Cormack,

Structural characterization of neodymium containing glasses by molecular

dynamics simulation, J. Non Cryst. Solids 351 (2005) 1185e1191.

[53] A. Bonamartini Corradi, V. Cannillo, M. Monia, C. Siligardi, Local and medium

range structure of erbium containing glasses: a molecular dynamics study,

J. Non Cryst. Solids 354 (2008) 173e180.

ꢀꢀ

ꢀ

I. Jozwik-Biala, N. Bererd, J. Jagielski, Irradiated rare-earth-doped powellite

single crystal probed by confocal raman mapping and transmission electron

microscopy, J. Raman Spectrosc. 45 (2014) 383e391.

ꢀ

[33] N. Chouard, D. Caurant, O. Majerus, J.-L. Dussossoy, A. Ledieu, S. Klimin,

D. Pytalev, Effect of molybdenum and ruthenium on the crystallization ten-

dency of a new nuclear glass containing high rare-earth concentration, Mat.

Res. Soc. Symp. Proc. 1265 (2010) 115e120.

Article Doi

Thermal stability of SiO2-B2O3-Al2O3-Na2O-CaO glasses with high Nd2O3 and MoO3 concentrations

DOI: 10.1016/j.jallcom.2016.02.063

Source and publish data:

Authors:

Article abstract of DOI:10.1016/j.jallcom.2016.02.063

Full text of DOI:10.1016/j.jallcom.2016.02.063

Products guided by the article

R&D Labs maybe for 7789-82-4

Relevant to this article

Hot Product