5.18

Waste Glass

E. Vernaz and S. Gin
Commissariat a` l’Energie Atomique et aux Energies Alternatives, Bagnols sur Ce`ze, France

C. Veyer
Veyer-Consultant, St Waast la Valle´e, France

ß 2012 Elsevier Ltd. All rights reserved.

5.18.1

Introduction

452

5.18.2
5.18.2.1
5.18.2.2
5.18.2.3
5.18.2.4
5.18.2.5
5.18.2.6
5.18.3
5.18.3.1
5.18.3.1.1
5.18.3.1.2
5.18.3.1.3
5.18.3.2

5.18.3.2.1
5.18.3.2.2
5.18.3.2.3
5.18.3.2.4
5.18.3.3
5.18.3.3.1
5.18.3.3.2
5.18.3.3.3
5.18.3.3.4
5.18.4
5.18.4.1
5.18.4.2
5.18.4.2.1
5.18.4.2.2
5.18.4.2.3
5.18.4.3
5.18.4.3.1
5.18.4.3.2
5.18.4.3.3
5.18.4.3.4
5.18.4.3.5
5.18.4.3.6
5.18.4.3.7
5.18.4.3.8
5.18.4.3.9
5.18.4.4
5.18.5
5.18.5.1
5.18.5.1.1
5.18.5.1.2


Glass and Vitreous State
Phenomenological Approach of Glass
Glass Transition Temperature
Glass Structure at Atomic Scale
Polymerization and Depolymerization of the Glass Network
Structure of R7T7-Type Containment Glass
Crystallization Mechanisms
Waste Glass Definition and Characterization
The Waste Streams to Vitrify
Nature and composition of HLW solutions
Other kinds of nuclear waste
Hazardous waste
Glass Formulation
Establishment of glass formation diagrams
Optimization of glass formulation
Validation of the reference formulation
Sensitivity to chemical composition
Glass Characteristics of Interest
Microstructural homogeneity
Physical properties
Thermal stability and crystallization potential
Chemical durability
Long-Term Behavior of Nuclear Waste Glasses
Glass Crystallization and Long-Term Thermal Stability
Glass Resistance to Self-Irradiation
Investigations of glasses doped with a short half-life actinide
Atomistic modeling of glass self-irradiation
External irradiation of glasses
Nuclear Glass Alteration by Water

Basic mechanisms of glass alteration
Initial rate of glass dissolution
Alteration rate in saturated conditions and final rate of glass dissolution
Essential role of the ‘passivating reactive interphase’
Influence of glass composition
Influence of groundwater and environmental materials
Influence of glass fracturing
Modeling glass long-term behavior
Natural an archaeological analogues
Conclusions on Glass Long-Term Behavior
Vitrification Processes
Existing Processes for Radioactive Waste Vitrification
The French two-step continuous vitrification process
Liquid-fed ceramic melters

453
453
454
454
456
456
456
457
457
457
458
458
459
459
461

461
461
462
462
463
464
464
465
466
467
467
468
469
470
470
470
470
470
471
471
471
472
473
473
474
475
475
477
451



452

Waste Glass

5.18.5.2
5.18.5.2.1
5.18.5.2.2
5.18.6
References

Emerging Processes for Radioactive Waste Vitrification
Cold-crucible induction melters
Incineration–vitrification processes
Conclusions and Outlook on Waste Glasses

Abbreviations
AVM

Atelier de Vitrification de Marcoule
(the first industrial vitrification plant in
France)
BO
Bridging oxygen
CEA
Commissariat a` l’Energie Atomique et
aux e´nergies alternative (The French
Atomic and Alternative Energy
Commission)
EXAFS

Extended X-ray absorption fine
structure
FP
Fission product
GCR
Gas-cooled reactor
GRAAL
Glass reactivity with allowance for the
(model)
alteration layer
HLW
High-level waste
ICP-AES
Inductively coupled plasma – atomic
emission spectroscopy
ICP-MS
Inductively coupled plasma – mass
spectroscopy
LWR
Light water reactor
MAs
Minor actinides
MOX (fuel) Mixed oxide (fuel)
NBO
Nonbridging oxygen
NMR
Nuclear magnetic resonance

PUREX
SEM

SIMS
STEM
TEM
UP1

479
479
480
481
482

Plutonium and uranium refining by
extraction
Scanning electron microscope
Secondary ion mass spectrometry
Scanning transmission electron
microscope
Transmission electron microscope
The first French reprocessing plant
located in Marcoule

5.18.1 Introduction
Fission products (FPs) and minor actinides (MAs)
produced during fuel irradiation in a nuclear reactor
represent only about 5% of the weight of used nuclear
fuel, but about 98% of its radioactivity. When the fuel
is reprocessed, these FPs and MAs end up in concentrated solutions (called high-level waste (HLW) solutions) that are stored in tanks fitted with stirring
systems and cooling facilities to evacuate the heat
resulting from radioactive decay. Figure 1 presents
an example of such a storage tank, made of stainless

steel, in a commercial reprocessing plant.

Figure 1 Example of a high-level waste storage tank for concentrated fission product solutions.


Waste Glass

Such a storage principle can be safe for several
decades. However, it requires active monitoring and
maintenance, and cannot reasonably be extended for
the durations required for complete decay of the activity (thousands of years). As early as the mid-1950s, the
major western countries started designing plans for
their nuclear waste, and work on FP immobilization
was initiated at Oak Ridge (USA), Harwell (UK),
Chalk River (Canada), and Saclay (France).
Several materials were considered at first, with a
rapid convergence on glass or glass-ceramics compositions. For instance, the first attempts at the
Commissariat a` l’Energie Atomique et aux e´nergies
alternative (CEA) in 1957 targeted crystals of
mica-phlogopite (M2Mg6(AlSi3)2O20F4, M being
an alkali or an alkaline earth metal), but this was
soon abandoned because of the impossibility of
incorporating all the elements of the concentrated
solution within one specific mineral. During these
first tests, a glassy component was frequently
observed at the bottom of the crucible, with often
better durability than that of the targeted mineral.
As, at the same time, some favorable results had
been obtained at Chalk River, where a confining
glass had been obtained by melting natural aluminosilicates impregnated with FP solutions at

1350  C, glass was selected for further investigations in France.1
A new application for glass was born: glass for the
containment of radioactivity.
At least, the idea was born, but a long path
remained to be covered from the idea to industrial
deployment, to optimize glass compositions adapted
for each type of FP solution, and to develop processes
operable in highly radioactive environments.
Similar exploratory work was performed in the
other countries, which ended up 20 years later in
the quasi-unanimous selection of borosilicate glass
as the preferable matrix.2

5.18.2 Glass and Vitreous State
Glass is one of the oldest materials known to man.
During prehistoric times, man used natural glasses (volcanic) to make knives or arrowheads. The first glass
actually melted by man could date back to 4500 BC.
Although, in common language, the term ‘glass’
often refers to a fragile and transparent material, the
scientific approach regarding vitreous state is both
much wider (for instance, nuclear glasses are not
transparent) and more difficult to define. This

453

chapter aims at providing the basis to understand
vitreous state, by considering the aspects of the formation of a glassy structure, the major glass properties (viscosity, durability, and thermal stability), and
the fine atomic scale structure of glass.3
5.18.2.1 Phenomenological Approach
of Glass

Why are most of natural rocks crystalline, while only
a small number of them display amorphous structures
(absence of diffraction peaks as evidenced by XRD)?
Most mineral compounds, when in the molten state,
form liquids with a low viscosity (some centipoises).
(The Poise (P) is a viscosity unit commonly used in the
glass industry; 1 P ¼ 0.1 Pa sÀ1, 1 cP ¼ 10À3 Pa sÀ1.)
Upon cooling, these liquids easily crystallize when
they reach their melting temperature. Some of these
liquids, however, are very viscous in the range of their
melting temperature (typically 105–107 P). Such
liquids, if they are kept below their liquidus temperature (in this case, they are supercooled liquids), will tend
to crystallize very slowly. If the cooling rate is faster
than the crystallization rate, crystallization will not
occur. During cooling, the viscosity of the supercooled
liquid increases progressively until the material rigidifies: the liquid ‘vitrifies’ or transitions from supercooled
liquids to the ‘vitreous state.’ A phenomenological definition of ‘glass’ could then be ‘glass is a rigidified
supercooled liquid.’
This definition is nevertheless too restrictive,
because a glass can be obtained by other routes,
(sol–gel for instance). Several alternative approaches
can be proposed:
 Structural approach: absence of order in the distribution of elementary structural units at scales
larger than 10–30 A˚,
 Thermodynamic approach: glass is in a metastable
state. It is nevertheless not unstable because the
energy gap that must be overcome to bring it to its
more stable crystallized state is generally significant due to the high viscosity,
 Physical approach: glass is a nonporous, impermeable, isotropic, noncleavable, elastic solid with a
fragile rupture behavior (absence of plastic deformation before failure),

 Kinetic approach: glass is a material which transitions
continuously and reversibly from liquid to solid
state with temperature (Figure 2).
Figure 2 provides an example of a typical evolution
of viscosity between 250 and 1500  C. One can


454

Waste Glass

h = 10x P
A

Solid glass (elastic)

B
x = 15
Tg

Blowing
(hollow glass)

x = 10

C
Working
range

x=5


Transformation domain
(plastic glass)

Devitrification zone

D
E

Devitrification
domain
250

500

Liquid glass (viscous)

750

1000

1250

1500 ؇C

Figure 2 Typical evolution of glass viscosity with temperature.

uid

Volume

d

liq

ole

co
er

p

Su

A

id
iqu

L

B
E

s

Glas

l

Crysta


C

D

Temperature
Tg

Tf

Figure 3 Evolution of the specific volume V of a glass or a
crystal during cooling.

distinguish an elastic solid domain below 500  C,
a plastic domain between 500 and 1000  C, where
the glass can be worked (blown, made to fibers,
moulded, etc.), and a liquid domain above 1000  C.
5.18.2.2

Glass Transition Temperature

Figure 3 compares the evolutions of the specific
volumes (inverse of densities) for a glass and a crystal
with temperature. During cooling of a liquid, a sharp
step is observed in the evolution of density when the
material crystallizes at melting temperature. If the
material does not crystallize, the specific volume
continues to decrease smoothly below this melting
temperature until a change of slope is observed at a


temperature Tg. This temperature is called the glass
transition temperature. At this temperature, the material
transitions from a supercooled liquid to a solid whose
expansion coefficient (the slope of the curve) is
roughly one-third of that of the liquid.
Similar glass transition patterns are observed for
other thermodynamic parameters such as specific
heat. The glass transition temperature corresponds
to a glass viscosity of 1013 P.
One can then propose another definition: ‘‘glass is
an amorphous solid that displays the glass transition
behavior.’’
5.18.2.3

Glass Structure at Atomic Scale

Numerous compounds can be stabilized in the
vitreous structure: oxide (silicates, borates, phosphates, etc.), chalcogenide (sulfides, selenides,
tellurides, etc.), ionic compounds (BeF2, ZrF4–BaF2,
AlF3, ZnCl2, etc.), specific metallic alloys subjected to
overhardening (Pd4Si, FeB, ZrCo, CaMg, etc.), and
also organic compounds such as glycerine, polyethylene, or glucose. For instance, the shiny caramel of a
tiered cake is obtained by cooling molten sugar fast
enough to obtain a vitreous state; candy floss is an
organic glass fiber whose Tg is around 55  C!4
The most common glass compositions are silicabased oxide glasses. Even if pure silica can vitrify by
itself (application to optic fibers), all-days glasses
are enriched with other oxide-based components.
They include, for instance, window glass (alumina–
sodium–calcium–silicate), Pyrex® (a high silica



Waste Glass

borosilicate), crystal-glass (silicate glass with a high
content of lead oxide), optical glass (such as ‘flint’
glass with high barium oxide), nuclear glass (alumina
borosilicate), and so on.
The constitutive oxides of an oxide glass can be
categorized into three families:
 Glass network forming oxides (network formers): these
oxides are able to form glass by themselves; SiO2,
B2O3, GeO2, and P2O5 are the most common.
In silica glass, the basic structural units are silica
tetrahedra [SiO4] sharing corners. Their linking
creates a continuous network with some degree of
disorder (groupings in cycles of five or six tetrahedra for instance), while, in a crystalline form such as
quartz, the tetrahedra are perfectly ordered.
 Network modifiers: these oxides cannot give a glass
by themselves. When they are coupled with network formers they are inserted into the vitreous
structure and modify the properties of the material. Typically, these oxides are alkali or alkaline
earth oxides (Na2O, Li2O, Cs2O, CaO, BaO, etc.).
As an example, the introduction of sodium oxide in
a silicate network induces the breakdown of
strongly covalent Si–O bonds and their replacement by Si–OÀ. . .Naþ bonds with a more ionic
character.
O
|

O

|

O

O
|

Na+
−

O − Si − O − Si − O + Na2O → O − Si − O
|
|
|
O
O
O

Na+

|
−
O − Si − O
|
O

The oxygen atoms bonding two silicon atoms are said
to be ‘bridging oxygen’ (BO) atoms. The introduction
of two sodium atoms in the structure creates two
‘nonbridging oxygen’ (NBO) atoms. The same

mechanisms may be observed with CaO, but here
only one Ca2þ is needed to compensate the negative
charge of the two NBO atoms.
The presence of NBO atoms weakens the vitreous
structure and allows decreasing melting temperature
(the glass becomes less refractory) and viscosity (the
glass can be poured more easily). This explains why
alkali oxides are often used as fluxes. NBO atoms also
loosen the network, and then help incorporating
more elements in the structure. A counterpart is a
decrease in chemical durability.
Finally, it should be recalled that the amount of
modifiers must remain limited if one wants to obtain
a glass from a molten liquid. If the number of NBO
atoms is too high, the liquid becomes very fluid and
tends to crystallize easily upon cooling.

455

 Intermediate oxides: these oxides cannot give a glass
by themselves. However, when mixed with network modifiers, they behave like network formers
within the vitreous structure. These oxides are
typically Al2O3, Fe2O3, ZnO, ZrO2, PbO, TiO2,
and so on.
Alumina is specifically important for the glass
industry, as it improves chemical durability (for
container glass, particularly). In order to behave
like a network former and to form a tetrahedron
similar to that of silica, the [AlO4/2]À ion needs a
positive charge to maintain local electroneutrality.

This will be achieved by fixing an alkali ion (or one
alkaline earth ion for two [AlO4/2]À).This results
in the following structure:
O
Na+
|
O − Si − O−
|
Na+
O

O
|
O− − Si − O + Al2O3 → 2 O −
|
O

O
|

O
|

|
O

|
O

Na

Si − O − Al − O

Introducing alumina into an alkali silicate glass
thus ‘hardens’ the glass, as NBO atoms are eliminated according to the stoichiometry (1Naþ for
1Al or 1Ca2þ for two Al). This will allow keeping
a structure slightly looser than that of a pure silica
glass (as the Al–O bonds are slightly weaker than
the Si–O bonds), while leaving only a small number of NBO atoms. The introduction of calcium
and some alumina into the formulation allowed the
transition from the poorly durable alkali glass used
for medieval stained glass to the window glass
used at present.
 The specific role of boron: the behavior of boron oxide
in the presence of alkalis differs from the general
pattern described for silica. Adding alkalis to pure
B2O3 or to a mixture of SiO2 and B2O3 leads at first
to the formation of BO atoms, as a result of the
formation of BO4 tetrahedra from the BO3 triangles
initially present in the B2O3 glass (in this case, the
alkali cation is used to compensate the charge of the
[BO4/2]À structural unit). (The elementary structural units of glass are often written [SiO4], [BO4],
[AlO4] to indicate the tetrahedral environment of
Si, B, or Al; however, for more detailed structural
descriptions, the more rigorous notation [SiO4/2],
[BO4/2]À, [AlO4/2]À, is preferred. In addition to
indicating that the base atom is surrounded by four
oxygen atoms, this notation stresses the fact that the
oxygen atoms are shared between two tetrahedra and
that, for B and Al, a negative charge exists, which
will need compensation by a neighboring cation.)



456

Waste Glass

It is only for higher alkali contents that NBO atoms
are formed, in the environment of boron or silicon
atoms. Consequently, the addition of alkalis to a
borate or borosilicate glass initially induces an
increase in viscosity. It is only for higher additions
that viscosity starts to decrease. This deviation in
the behavior of boron when compared to silica is
often termed ‘boron anomaly.’
Boron, in right amount, plays a key role in nuclear
borosilicates for several reasons: (1) it helps
decrease the glass melting temperature without
drop of the durability as [BO4] units decrease the
number of NBO by fixing an alkali ion; (2) it helps
digest a number of chemical elements that would
be only sparingly soluble in pure silica; (3) it prevents glass crystallization; (4) it contributes to the
long-term good glass behavior by allowing the formation of a very fine gel structure (without boron)
and by decreasing the final pH.
5.18.2.4 Polymerization and
Depolymerization of the Glass Network
On the basis of the above classification, it is observed
that the progressive addition of network modifiers to
silica (SiO2) leads to network depolymerization by
the formation of NBO atoms. At a certain limit, the
liquid is strongly depolymerized, its viscosity

becomes very low, and it is not possible to obtain a
glass by quenching any more. On the contrary, adding
boron or intermediate oxides favors network polymerization by fixing alkalis or alkaline earths as
charge compensators.
The very good performances of nuclear borosilicate (moderate melting temperature coupled with
good durability) come from an adequate balance
between boron and intermediate elements (Al, Fe,
Zr, etc.) on one side and the alkali elements on the
other side, allowing a high polymerization rate as
most of the alkalis are found as charge compensators
rather than forming NBOs.
5.18.2.5 Structure of R7T7-Type
Containment Glass
The atomic structure of nuclear waste glasses has
been studied using spectroscopic techniques such as
nuclear magnetic resonance (NMR) or extended
X-ray absorption fine structure (EXAFS), at first on
simplified glass compositions (SiO2–B2O3–Na2O–
Al2O3), and then on compositions with increasing
complexity.

These studies show that all the intermediate
cations in the R7T7 glass composition are in positions of network formers. Their substitution on Si4þ
sites induces charge deficits which are compensated
by a network modifier cation in order to ensure
charge neutrality. The sum of negative charges cre2À
À
À
ated by the (AlOÀ
4/2), (ZrO6/2), (FeO4/2), and (BO4/2)

groups in R7T7 amounts to 5.19 mol per 100 mol of
elements (computed from elemental mol%). On the
other hand, the sum of positive charges created by
alkalis and alkaline earths is 9.44 mol%. This results
in the following situation:
 All the intermediate cations behave as network
formers,
 A somewhat limited number of network modifiers
remain available to create NBOs.
These considerations show that the network of this
glass composition is homogeneous and very well
copolymerized, owing to the good incorporation of
intermediate cations in the network. The network is
most probably composed of numerous mixed Si–O–
M bonds (with M ¼ Al, B, Zr, Zn, and even some rare
earths). This strong polymerization, as well as all the
structural data, is confirmed by molecular dynamics
modeling.
5.18.2.6

Crystallization Mechanisms

Crystallization in a supercooled liquid or a glass
occurs via a nucleation-growth mechanism which
starts by the formation of small ordered germs. The
germination can be enhanced by the presence of
rough interfaces in the melt.
Figure 4 gives an example of nucleation and
growth curves for a glass.
A

r
b
i
t
r
a
r
y
u
n
i
t
s

(n), (g)

(n) is the
nucleation rate in
number of seeds
per time unit

Tg 600 ЊC

700 ЊC

(g) is the seeds
growth rate in
microns per time
unit


800 ЊC

Temperature

Figure 4 Nucleation and growth of a crystalline phase
within a glass.


Waste Glass

Several crystalline phases can be formed during
cooling of a supercooled liquid, according to its composition. The chemical composition of the formed
crystals can be very different from those of the liquid
or of the glass. For a given phase, graph (n) plots
the number of supercritical germs per unit volume
and time as a function of temperature. Graph (g) plots
the growth rate for this phase as a function of
temperature.
If those two curves do not overlap, the given
crystal cannot form spontaneously during cooling.
In this case, when the glass is in the growth zone,
there are no germs liable to grow, and when it is in the
nucleation zone, the temperature is too low to allow
the formed nuclei to grow. Crystallization will then
be possible only after a two-stage heat treatment, one
stage for nucleation and the second for growth. This
is a favorable situation for the production of glassceramics with homogeneous and well controlled
crystallization.5,6 On the other hand, if the two curves
overlap, a risk of uncontrolled crystallization appears
in the overlap region, where germs can be formed and

can grow. The relative position of the nucleation and
growth curves in the temperature field will then be an
important parameter determining the sensitivity of
the glass to crystallization; it must be studied for each
of the phases liable to be formed in a given glass
composition.

5.18.3 Waste Glass Definition and
Characterization
Vitrification is not a simple encapsulation process
(as immobilization in bitumen for instance) but consists of making a new material in which the waste
components are contained at the atomic scale within
the matrix and can only be released by destruction of
the network bonds.
One major requirement is that the selected matrix
should be able to incorporate all of the waste stream
components in its structure. By using the flexibility
brought about by the disordered and relatively loose
structure of a glass, it is possible to design glass
compositions able to integrate a very wide range of
elements within their structure, and which are tolerant to compositional variations in the waste stream.
This approach constitutes waste vitrification, where
waste components are usually mixed with suitable
additives and molten to give a glass wasteform.
A recent development of vitrification has been the
design of glass-ceramics that combine the flexibility

457

of glass formulation to digest most of the waste components with the possibility of targeting well defined

crystalline phase(s) for specific waste components
that may not be soluble in large amounts in glass, as
molybdenum for instance. Vitrification usually
involves a small number of processing steps, with a
robust design, compatible with operation in a highly
radioactive or hazardous environment.

5.18.3.1

The Waste Streams to Vitrify

A large variety of nuclear waste compositions have
been considered for vitrification since the first
attempts in the late 1950s, including not only highlevel, but also intermediate or low-level effluents.
More recently, this approach has been extended
to other types of inorganic hazardous waste for
which other types of immobilization were not considered suitable.
5.18.3.1.1 Nature and composition of HLW
solutions

FPs and MAs produced in the reactor by fission or
neutron capture display a wide range of atomic numbers. One can find for instance alkalis (rubidium and
cesium), alkaline earths (strontium and barium), a
wide range of transition elements (zirconium, molybdenum, etc.), noble metals (ruthenium, rhodium,
and palladium), chalcogenides (Se and Te), nonmetals (As, Sb, etc.), lanthanides, and actinides. The
amount and composition spectrum of the FPs and
MAs varies with fuel initial composition, enrichment,
and burnup.
The separation process used for light water reactor (LWR) oxide fuel at the commercial reprocessing
plants of La Hague (France), Rokkasho Mura ( Japan),

or Sellafield (UK) is a hydrometallurgical process
based on plutonium and uranium refining by extraction (PUREX), where, after nitric dissolution of the
fuel and a series of solvent extraction steps used
for U and Pu recycling, most FPs and MAs from
the fuel end up in a concentrated nitric solution
(HLW solution), which constitutes the major target
feed for vitrification. In addition to the isotopes
extracted from the fuel, the HLW stream holds some
chemicals added during reprocessing. The current
PUREX-based process for LWR oxide fuel has been
designed to minimize these additions, by using essentially chemicals that will not add to the waste load.
However it is not possible to avoid the addition of
some selected chemicals, mainly sodium, during


458

Waste Glass

ancillary operations such as solvent purification or
equipment cleaning operations. Some impurities
resulting from a slight corrosion of the piping (namely
Fe, Cr, and Ni) or solvent degradation (phosphate)
also end up in the HLW solution.
In addition to commercial LWR fuel, other fuels
have been reprocessed worldwide in the past, and
other processes have been used or are considered for
separating uranium and/or plutonium from the fuel.
Fuel alloy metals (Al and Mo) or dissolved cladding
material (Al, Mg, and stainless steel) may follow the

HLW stream. The chemicals used to perform the
dissolutions or separations can also be very diverse
(mercury, fluorine, ferrous sulfamate, bismuth phosphate, etc.). In several instances, the HLW solution
(which is, initially, acidic) has been neutralized by
adding massive amounts of caustic to prevent corrosion of the tanks, thereby precipitating most of the
FPs and MAs as hydroxides. This is for instance the
case in the United States, at Hanford or Savannah
River; The HLW solutions are thus quite complex
and not unique, with a large number of constituents.
Typically, in France, concentrated HLW solutions
are nitric solutions (1–2 N) with high bg activities
(several tens of TBq per liter) including suspended
solids such as colloids (zirconium phosphate and
cesium phosphomolybdate) and some metallic fines
(insoluble residue, cladding fines generated during
the fuel shearing operation). Table 1 gives examples
of solutions derived from reprocessing various types
of fuel from past or present reactors in France.
5.18.3.1.2 Other kinds of nuclear waste

In addition to HLW, vitrification is increasingly considered for waste of lower radioactivity content,
although other, less costly, immobilization methods,
such as grouting, are more common. Despite the
cost, vitrification has the advantage of providing
durable matrices, with significant volume reduction.
For liquid effluents, a vitrification plant is for
instance being designed and built at the Hanford
site, USA, as part of the Waste Treatment Plant
(WTP) to immobilize the low activity fraction of the
waste from 177 underground tanks. This low activity

fraction consists mainly of concentrated sodium
nitrate and sodium hydroxide solutions, with some
aluminum, chrome, sulfur, and some minor metals.
In Russia, at the Radon facility near Moscow, borated
low activity power plant effluents are also vitrified.
Vitrification is also applied to low activity solid
combustible waste, in conjunction with incineration:
the combustible fraction is incinerated and the

Table 1
France

Example of HLW compositions to vitrify in

Reactor type

PWR

Gas-cooled, graphite
moderated, natural
uranium reactor

Fuel type

UO2

SiCrAl

UMo–MoSnAl


Burnup
(MWd tÀ1)
HLW solution
(l tUÀ1)
Oxide contents
(g lÀ1)
Free acidity (N)
Oxide
composition
(g lÀ1)
Fe
Al
Cr
Ni
Na
Mg
Zn
Mo
Sn
P
F
PF oxides
(g lÀ1)
Actinide oxides

33 000

4000–6000

1500–3500


660

100–120

75–100

90–100

240

0.95

1.0

0.8

13
0–2
2.3
1.9
20
–
0–1
–
–
1.26
–
52.23


7–14
19–38
0–1.5
0–1.3
7–11
3–7
–
–
–
1–2
2–8
45.1

3.14
2.76
0.65
0.31
14
1.6
–
163.2
0.67
41.16
–
11.46

3.83

3.65


3.62

inorganic fraction, under the form of ashes, is vitrified. This is for instance the case at Ulchin, S. Korea,
where a facility based on French technology has been
recently commissioned to simultaneously incinerate
and vitrify dry active waste and resins from nuclear
power plants. At the Zwilag waste management facility, in Switzerland, a plasma-torch powered facility
has been built to incinerate, melt, and vitrify low
activity waste conditioned in drums.
In the end, vitrification has also sometimes been
considered for actinide-bearing waste or residues.
In the United States, for instance, a significant program has been launched in the 1990s in the frame
of the START-II program, to study the possibility of
vitrifying plutonium-rich residues from the weapon
industry that could not be easily made into mixed
oxide (MOX) fuel.
5.18.3.1.3 Hazardous waste

Vitrification, alone or combined with incineration,
provides a convenient method for immobilizing or
rendering hazardous waste innocuous. For instance,


Waste Glass

in the United States, vitrification has been considered
as Best Demonstrated Available Technology (BDAT)
by the US-EPA (US Environmental Protection
Agency) for waste containing toxic metals such as
arsenic. In situ vitrification of soils has been applied

in some contaminated sites in the United States.
Owing to the potential for immobilizing inorganic, nonvolatile, toxic metals into a stable and
durable matrix, vitrification is used in several facilities worldwide, including France and Japan, to
immobilize the slag and ashes from municipal waste
incineration or waste-to-energy conversion facilities.
The product is completely inert and can be disposed
of in a conventional disposal facility. Efforts are under
way in several institutions to improve the product for
reuse, mostly in the building industry (as road base or
tiles for instance). Another significant application of
vitrification is the destruction of asbestos: once vitrified, asbestos becomes a compact, harmless substance.
Such a facility for asbestos vitrification has been
operating in southwestern France since 2003.
5.18.3.2

Glass Formulation

Glass formulation is aimed at defining a compositional domain within which the matrix will display a
number of required characteristics related to technological feasibility, durability, containment properties,
and all the properties required for the intended use or
destination of the product. For HLW glass, for instance,
the product will have to ensure long-term safety in the
proposed geological disposal environment.
Glass formulation then consists in reaching
the best compromise between a large number of

constraints: glass formation domain (solubility of the
various waste components, waste loading, and homogeneous glass), technological feasibility (melting
temperature limits, viscosity allowing to pour the
product, minimum volatilization, and minimum

corrosion of the melter), as well as stability and containment properties (thermal stability, resistance
to self-irradiation, chemical durability, mechanical
properties, etc.) (Figure 5).
For radioactive waste immobilization, borosilicate
systems represent the best compromise between the
various constraints, and they have been selected as
the reference matrix compositions for HLW in most
countries (France, USA, UK, Japan, Germany,
Belgium, etc.). Other matrices for HLW include
phosphate glasses in Russia.
Glass formulation involves several successive (and
often iterative) steps illustrated below by the methodology used at the French CEA for HLW borosilicate glass formulation: in this instance, it was
necessary to design a matrix with very good containment and long-term properties, to immobilize a highly
radioactive, heat-generating, waste stream, whose composition was expected to be very stable throughout the
years. In other countries, or for other waste streams,
the practical organization of formulation studies may
differ, but these steps are always necessary to ensure
consistent and reliable waste form properties.
5.18.3.2.1 Establishment of glass formation
diagrams

The glass-forming domain is determined by establishing quaternary phase diagrams with the expected
four major components of the glass, to identify the

Ability to accommodate the waste
Solubility (Cr, Ru, Rh, Pd, Ce, Pu, SO4, Cl)
Phase separation (Mo, SO4, Cl, P)
Devitrification (Mo, P, F, Mg, ...)
Maximize the waste loading


Process/technology
Ease of processing
Melting temperature
Viscosity, reactivity, residence time,
electrical condition, thermal condition
Additives needed

Figure 5 Waste glass formulation is a compromise.

459

Glass performance
Properties for storage/disposal
Thermal stability
Chemical durability
Resistance to self-irradiation
Mechanical properties


460

Waste Glass

glass-forming regions. For waste solutions dominated
by FP oxides, the four major components can be
summarized as silica SiO2, boron oxide B2O3, sodium
oxide Na2O, and FP oxides (considered collectively,
with a chemical spectrum corresponding to the
expected spectrum for the solution). This leads to
‘SON’-type glass formulations such as those selected

for the LWR FP solutions at La Hague. When the
waste solution is dominated by aluminum, the quaternary diagram considers SiO2, Al2O3, B2O3, and
Na2O. This leads to ‘SAN’-type glass formulations,
such as those selected for the gas-cooled reactor
(GCR) FP solutions processed at the Atelier de
Vitrification de Marcoule (AVM) facility at Marcoule
(Figure 6).
The glass-forming domain is established by melting a large number of glass compositions in small
laboratory crucibles and performing visual or

microscopic observations. As can be seen in Figure 7,
boron helps to dissolve the entire FP spectrum into
the glass, prevent crystallization, and lower viscosity.
The chosen boron content is however limited to the
minimum needed to have a sufficient domain of
homogeneous glass while keeping the best durability
(for instance 18 wt% B2O3 is sufficient in the quaternary systems shown on Figure 7).
Once this glass-forming region has been determined, additional constraints are used to further
limit this domain, for instance, limitation of the melting temperature, which will result in limits on the
refractory elements such as silica or alumina, limitation of waste loading by thermal considerations (for
HLW), and limitation of network modifiers to keep
an acceptable chemical durability. Each of these constraints establishes one side of the diamond in which
the glass composition will be chosen.
B2O3

B2O3

24% B2O3
24% B2O3
10% B2O3

10% B2O3
SiO2

Na2O

SiO2

Na2O

Ox PF

Al2O3

Figure 6 Basic quaternary glasses for gas-cooled reactor on left (SAN-type glasses) and light water reactor on right
(SON-type glasses).

SiO2
SiO2
730

Ox.
PF +

110
40

315
75

80


310
230

Ox.

115

Na2O
10

PF +
Act.

Act.

24% B2O3

18% B2O3
Nonglassy state
Heterogeneous glasses
Homogeneous glasses
Viscosity (P [= dPa.s] )

Figure 7 Effect of boron oxide content on the vitrification domain in quaternary glasses. Domains of glass formation
into the quaternary diagram SiO2–Na2O–OxPF–B2O3 for two different B2O3 contents.


Waste Glass


5.18.3.2.2 Optimization of glass formulation

Optimization consists in adding or substituting various additives (such as Li2O, CaO, ZnO, ZrO2, etc.)
to improve the matrix properties. Among all properties of interest, the waste load (considering the final
glass homogeneity), the viscosity of the molten glass,
and the chemical durability have to be specially
mentioned. Depending on the choice for industrial
process, other properties may be targeted in addition.
As an example, small amounts of lithium may be
substituted for sodium to decrease melt viscosity,
without affecting the chemical durability. Aluminum
and calcium oxide may be substituted for a small
amount of silica to improve durability.
It has to be pointed out that those optimizations
result from sharp compromises and high-level expertises in mixing the appropriate oxides in a synergic
way (for instance too high aluminum content would
enhance short term durability but would decrease
long term durability; it would also increase the risk
of crystallization of undesirable phases; controlled
etc.).
At the end of this phase, a reference composition is
proposed for a given effluent composition and a given
melting technology. Dozens of compositions are published in the literature, for a very wide range of waste
compositions, in many countries. Table 2 gives the
composition of the French reference glass for the
HLW solutions derived form processing 33 GWdtÀ1
LWR fuel at La Hague: the so-called ‘R7T7’ composition, named after the vitrification facilities R7 and
T7 at La Hague plant.
5.18.3.2.3 Validation of the reference
formulation


Once a reference composition has been defined at
lab-scale for a specific simulated waste stream, it
needs to be validated for actual industrial application.
(During the development stage of reference matrix,
as the main consideration is on the basis of chemical
incorporation, radioisotopes issued from FPs and
MAs are replaced by their inactive representatives.)
To this effect, several additional programs are
performed:
 complete characterization of the reference composition and determination of all its important
properties,
 validation of technological feasibility in industrial
conditions by performing long-duration demonstrations on an inactive industrial-scale pilot
facility, and confirmation of the maintenance of

461

Table 2
Chemical composition range of the French
R7T7 glass
Chemical composition range of R7T7 glasses produced in
the AREVA – La Hague plant workshops
Oxides

SiO2
B2O3
Al2O3
Na2O
CaO

Fe2O3
NiO
Cr2O3
P2O5
Li2O
ZnO
Oxides
(FP þ Zr þ actinides)
fines suspension
Actinide oxides
SiO2 þ B2O3 þ Al2O3

Specified
interval for
the industry
(wt%)
Min

Max

42.4
12.4
3.6
8.1
3.5

51.7
16.5
6.6
11.0

4.8
<4.5
<0.5
<0.6
<1.0
2.4
2.8
18.5

1.6
2.2
7.5

>60

Average
composition of
industrial
glasses (wt%)

45.6
14.1
4.7
9.9
4.0
1.1
0.1
0.1
0.2
2.0

2.5
17.0

0.6
64.4

the characteristics defined at lab-scale on the
product made during scale one demonstrations,
and
 validation of the representativeness of the inactive
glasses made in the laboratory by making a glass of
the same composition with actual radioactive waste
in a hot laboratory and determining its characteristics. The fabrication of active samples is also
necessary to study the effect of self-irradiation
and the specific behavior of some radioactive
elements.
5.18.3.2.4 Sensitivity to chemical composition

In industrial situations, one must expect day-to-day
variations in the composition of the solution to
be vitrified, and also process upsets or variability
which may affect the product. It is then necessary
to determine what is the flexibility of the formulation towards these variations, in order to provide
a wide enough operational domain, while keeping
acceptable glass properties. For this, a sensitivity
study is launched, to study systematically the
effect of selected composition variations on glass
properties, and establish acceptable limits for these
components.



462

Waste Glass

For instance, in order to be pourable, the glass
must display a viscosity lower than about 100 P at
the melting temperature. This will limit the possible
increase in refractory elements such as SiO2 or Al2O3
or the decrease in fluxing components such as
Na2O or Li2O. On the other hand, in order to keep
an acceptable chemical durability, the increase in
alkali oxides will be limited.
In such a multidimensional space, with more than
ten constituents of interest in a formulation, sensitivity studies can become very cumbersome and involve
a large number of crucible melts and the associated
characterization measurements. In order to minimize the number of experiments while obtaining a
reliably acceptable composition domain, statistical
tools, property–composition models, and experiment
design are implemented. These allow identifying
acceptability limits and modeling matrix properties
within these limits.
For illustration, Table 2 displays the range of
acceptable composition for the R7T7 glass produced
at La Hague. One can see that the acceptable variations may be small for some components, and much
larger for other components.

5.18.3.3

(Figure 8). The black color, analogous to the color of

some volcanic glasses, is the result of the wide diversity of chemical elements included in the glass and,
more particularly, transition metals and rare earths,
which absorb light over a wide range of wavelengths.
By visual observation, some small bubbles and
some cracking associated with stress relaxation during
cooling can be observed on industrial-size blocks. Conservative durability modeling accounts for the impact
of this physical heterogeneity (see Section 5.18.4.3.7).
Under the microscope (optical or scanning electron microscope, SEM), a small volume of chemical
heterogeneities can be observed, essentially noble
metal inclusions (Pd, Rh, or Pd–Te alloy), ruthenium
oxide, nickel, zinc, and iron chromites (Figure 9).
This type of inclusion represents a small volume
of the matrix, and has been shown to have no effect
on the long-term behavior of R7T7 glass. For other
waste glass compositions, it is necessary to make
sure that the precipitated phases do not alter the
matrix properties, such as chemical durability. For
instance, for Fe- and Cr-rich glasses, spinels have
been observed, which do not alter glass chemical
durability. On the other hand, for Al-rich glasses,
it is necessary to avoid the formation of nepheline

Glass Characteristics of Interest

Glass characterization is the determination of all the
physical and chemical properties of the product.
Generally, and as observed in many laboratories, the
products made with simulated waste by substituting
the inactive counterparts for the radioactive components of the waste display properties that are very
representative of those of the actual radioactive product. Glass characterization is thus generally performed on a simulated, nonradioactive glass for

convenience, and validated in the end on radioactive
glasses (see Section 5.18.3.2).

Figure 8 Obsidian glass and nuclear waste glass (cannot
be distinguished with the naked eye).

20 μm

5.18.3.3.1 Microstructural homogeneity

As it is generally more difficult to characterize and
demonstrate a favorable long-term behavior for a
chemically heterogeneous material, the objective is
usually to produce a homogenous glass, or a glass that
contains, after cooling of industrial-size blocks, only a
small amount of harmless precipitates or insoluble
phases. Nevertheless, the notion of homogeneity is
relative and depends on the resolution of the measuring instrument used.
Upon visual observation, HLW glasses such as the
R7T7 composition are homogenous, black, and shiny

Palladiumtellure

Ruthénium
oxide
Chromites

Figure 9 Scanning electron microscope observation on a
zone of platinoI¨de concentration into the glass.



Waste Glass

(aluminosilicate) crystals on cooling, as these tend
to be detrimental to resistance to aqueous leaching.
In the end, the presence of precipitates or insoluble
alloys in the molten glass may promote settling in the
melter and have detrimental effects on the process
(pouring difficulties and electrical short-circuiting
in some types of melters). For the La Hague process,
stirring allows processing melts with significant
amounts of insoluble noble metals.
Another form of heterogeneity is phase separation, whereby the glass separates (during melting or
during cooling) into two or more different liquid or
glassy phases. This type of phenomenon has to be
carefully controlled to avoid the formation of a phase
that is ‘weaker’ than the other, inducing some loss of
general durability.
Nevertheless, in some specific instances, it may be
acceptable to formulate microcrystalline or phaseseparated glasses, if it can be demonstrated that this
does not alter chemical durability. Such an approach
has been retained in France to formulate glasses to
immobilize molybdenum-rich solutions resulting
from the processing of U–Mo fuel.
Lower resolutions of observations (such as the one
reachable by transmission electron microscope,
TEM, as well as structural techniques) are not relevant to link the heterogeneities at that scale to a
significant effect on durability, but are of great use
for understanding.
5.18.3.3.2 Physical properties


The physical properties of nuclear waste glasses are
usually quite comparable to those of classical industrial glasses,7 as illustrated on Table 3. Most of them
have to range in between a minimum and a maximum
value to suit industrial processes.
The density of waste glasses is slightly higher than
those for the industrial glasses, owing to the presence of heavy metals. The viscosity at 1100  C is
much lower. This is because nuclear glasses are

Table 3

463

formulated to be poured at 1100  C while the melting temperature of industrial glasses is significantly
higher. It should also be noticed that the presence of
noble metals or other heterogeneity in the nuclear
glass can significantly modify its rheological behavior. Glass transition temperature occurs in similar
ranges, although it is slightly lower for nuclear
waste glass. The thermal expansion coefficient of nuclear
waste glasses is similar to that of window glass,
and significantly higher than that of Pyrex: Pyrex
is formulated specifically to resist thermal shocks.
Thermal conductivity is similar for all three types
of glass. It should be noted that this low value is
significant in nuclear waste glasses which hold
heat-generating FPs: temperature at the centre of
the glass block will need to be controlled. Young’s
modulus, which characterizes rigidity, is slightly
higher for nuclear waste glass, while fracture toughness is similar.
Electrical resistivity, not included in the table is

another significant parameter if the glass is heated
by direct electromagnetic induction or within electrodes (in the case of Joule Melter technology, this
property can be neglected). Electrical resistivity
decreases when temperature increases and when
the alkali content of the glass increases. Electrical
resistivity depends mainly on ionic diffusion in the
material. It decreases from around 1.5 Â 104 at
500  C to a few O cm at 1200  C. Other parameters
not included are the thermal conductivity and the
redox behavior. Thermal conductivity impacts
directly the energetic efficiency in the melter. If
too high, the thermal losses out of the melter
make the process energetically (and economically)
inefficient; if too low, the energy transmitted to the
glass is nonhomogenous. Regarding redox behavior,
the presence of multivalent elements (Ce oxide for
instance) in the glass composition may induce
foaming at high temperature due to their reduction.
The monitoring of oxygen partial pressure in the

Comparison of physical properties of selected nuclear and industrial glass compositions

À3

Density (kg m )
Viscosity at 1100  C (P)
Tg ( C)
Expansion coefficient (10À6 KÀ1)
Thermal conductivity (W mÀ1 KÀ1)
Young’s modulus (1010 Pa)

Fracture toughness KIc (106 Pa m1/2)

Typical HLW glass compositions

Pyrex glass

Window glass

2.50–2.75
50–150
510
8.3–9.9
1.0
8.4–8.6
0.75–0.95

2.28
80 000
565
3.2
1.09
6.1
0.85

2.46
4000
527–547
9.3
1.05
7.3

0.70–0.80


464

Waste Glass

molten glass, and if necessary the use of redox
buffer (FeII/FeIII) as well as physical means can
overcome this inconvenience.
5.18.3.3.3 Thermal stability and
crystallization potential

Devitrification is the process by which the glass loses
part or all of its glassy nature through crystallization.
It depends on the composition of the glass and its
thermal history.
For instance, increasing the levels of FPs, noble
metals, molybdenum, phosphorus, chrome, nickel,
iron, or magnesium can favor crystallization in a
nuclear waste glass. Furthermore, the time needed
to reach the glass transition temperature (Tg) from
melting temperature is also important (this depends
mainly on glass thermal conductivity and specific
heat, canister geometry, and process parameters such
as pouring rate). However, it is considered that once
Tg is reached in cooling conditions, the devitrification
process is kinetically frozen (cf. Section 5.18.4.1).
Devitrification studies are on the basis of subjecting glass samples to short-duration heat treatments
(around 15 h) at stabilized temperatures and observing the heat-treated samples under the microscope to

detect, observe, and quantify the crystals formed.
Several indicators are determined:
 starting crystallization temperature: the temperature
below which crystalline phases can be observed in
the bulk of the sample after about 10 h of isothermal heat treatment,
 crystallization temperature range: range in which these
crystals are observed,
 maximal crystal growth rate (generally expressed
in mm mnÀ1),
 crystallization potential: characterizes the ability of a
composition to devitrify. It is the maximum percentage of crystals that can form after a heat treatment. This can vary from 0% (pure borate glass) to
100% (lithium disilicate glass). It is about 4% for
R7T7 glass.
XRD and X-ray microprobe are used to identify the
crystalline phases formed within a glass while image
analysis and quantitative XRD are used to evaluate
the percentage of phases formed.
The amount of crystals that form in an actual large
size industrial glass block is different from the maximal values found at lab-scale: indeed, the thermal
profile in the glass canister involves a continuous
decrease of temperature, and is different in the various parts of the canister (close to the canister wall,

cooling is faster than in the centre of the glass block).
Several approaches can be used to bracket an estimation of the amount of crystals in the industrial
glass block. In France, for instance, the maximum
possible amount of crystallization is determined on
laboratory samples, by subjecting them to a heat
treatment designed to promote crystallization (5 h
at 610  C – nucleation temperature – and 100 h at
780  C – maximum growth temperature – this cannot

happen in a real glass block) and it is postulated that
the amount of crystals in the glass block cannot be
higher than the fraction determined in this way
(which is in fact quite small for the R7T7 glass).
In the United States, where the glass blocks are
quite large, crystallization studies are performed by
establishing systematic TTT (time–temperature–
transformation) diagrams and by considering the
cooling profile at the centre of the canister (CCC,
canister centerline cooling curve), which is the slowest cooling part of the canister.
Whatever be the approach, the important aspect is
the fact that crystallization must not be detrimental
to glass durability. It is thus necessary to avoid
depleting the glass matrix of elements that are favorable to durability, such as silica or alumina.
5.18.3.3.4 Chemical durability

Water is the major cause of glass alteration and radionuclide (or hazardous metal) dispersion during the
life of the product, and more particularly during
geologic disposal. The resistance of glass to aqueous
alteration is generally called chemical durability. It is
the essential property required for a containment
matrix. It is also essential for classical industrial glass
compositions (container glass, window glass, etc.) and
as such is also considered during their formulation.
The process by which glass constituents are
washed into water is called leaching, and it is the
combination of a variety of mechanisms described
in Section 5.18.4.
Chemical durability is assessed by various ‘leach
tests’ during which glass samples are contacted with

aqueous solutions, under a large range of experimental conditions. The leaching behavior of glasses varies
according to glass composition, test conditions, and
time. There is no unique evaluation of chemical
durability, and the performance of different glass
compositions can only be compared under the conditions of a given test. In order to fully evaluate the
behavior and performance of a given glass, it is advisable to understand the various mechanisms that are
involved during its interaction with water and the


Waste Glass

environment, and to design testing to obtain the
parameters that allow modeling this performance.
Several types of leach tests can be listed:
 Static tests, during which glass samples, either monolithic or crushed to powders, are exposed to a solution and left standing there for the duration of the
test. During these tests, the components dissolved
from the glass progressively accumulate into the
solution and are left to interact between one another
and with the sample. Two of these tests have been
normalized in the United States: the so-called
‘MCC-1’ test on monolithic samples (also known as
ASTM-C1220) and the ‘PCT’ test on crushed glass
samples (also known as ASTM-C1285). Both these
tests are available with several variants (temperature,
volume of solution, type of solution, etc.).
 Flow-through tests, during which the sample is
exposed to a continuous flow of fresh leachant to
prevent the accumulation of reaction products into
the solution, thereby testing the ‘initial’ alteration.
Among those tests, one can list the ‘Soxhlet’ test

(also known as ISO 16797:2004), by which a monolithic sample is exposed to a continuous flow of
condensed water at 100  C recirculated in a distillation apparatus, or the ‘flow-through’ test (also
known as ASTM C1662-07) by which the sample
is inserted in a column or a container and subjected
to a continuous flow of fresh solution.
 ‘Integral tests’ or ‘service conditions tests’ or ‘tests
including environmental materials’ are tests
designed to account for the overall service environment expected during the life of the product.
Glass alteration may be measured in three different ways:
 Elemental analysis (ICP-MS, ICP-AES, etc.) of the
leachate (solution after contacting the sample), in
order to obtain information on the leaching kinetics for each element, as a function of time.
 Weighing of the sample, to determine the overall
weight loss.
 Analysis of the alteration rind on the sample (SEM,
secondary ion mass spectrometry, SIMS, scanning
transmission electron microscope, STEM, etc.) in
order to study the thickness of the altered layer and
to understand the fate of the various elements
released from the glass and retained in the alteration layer.
The quantitative expression describing the rate of
release of an element in water is the ‘leach rate,’
usually expressed in g m2 dÀ1, and normalized with

465

respect to the mass fraction of the given element in
the glass.
LRðiÞ ¼ mi =xi St
where mi is the weight (or activity) of element i

released into solution, xi is the mass fraction (or
specific activity) of this element in the glass, S is the
area of the glass surface exposed to the leachant, and
t is time.
During a static test the ‘normalized mass loss’ is
often used as an indicator:
NLðiÞ ¼ mi =xi S;

expressed in g mÀ2

The evolution of NL(i ) ¼ f (t) describes the overall
kinetics of the process for the given experimental
conditions. This curve is the basis for all the glass
behavior studies.
If an element (i) is a good alteration tracer, that is,
congruently released with glass dissolution and not
trapped back into the alteration products, then the
equivalent thickness of altered glass Eth can be calculated by dividing the normalized mass loss by the
glass density r:
Eth ðiÞ ¼ NLðiÞ=r
Boron is most often a good tracer of glass alteration.
Sodium, lithium, and molybdenum are good tracers
in dilute media, but they may be integrated in alteration products in more concentrated media.
More information on glass alteration mechanisms
is provided in Section 5.18.4.
Exhaustive leaching characterizations are performed on inactive but chemically representative
samples; some tests are performed on fully active
material to check the impact of radioactive environment on leaching behavior.

5.18.4 Long-Term Behavior of

Nuclear Waste Glasses
The main phenomena that could alter glass containment properties over the long term are heat (for
HLW only), radiation damage, and alteration by
water. Their occurrence is not expected at the same
time scale (Figure 10). For instance, the risk of crystallization is principally limited to the thermal phase,
that is, in interim storage over the few decades during
which the maximum glass temperature will decay
from about 400 to <100  C.
For thousands of years, the glass matrix is
expected to remain dry and the major potential


466

Waste Glass

glass alteration mechanism is self-radiation damage.
Can it change the glass containment properties when,
after a few thousands of years, canister and over-pack
breach, and glass alteration by water start? Many
studies on glass self-radiation damage address this
question.
Eventually, in the very long term, the rate of radio
nuclides released into the near field will be controlled by the rate of glass alteration by water. Worldwide, for over 30 years, large research efforts have
been conducted to understand all the mechanisms of
glass alteration by water and to develop comprehensive models, and to adapt them for the evaluation of
repository performance. (A repository is the final
destination of HLW glass, a disposal site selected
and engineered to definitively isolate the radioactivity contained within the waste from the biosphere
and man.)

5.18.4.1 Glass Crystallization and
Long-Term Thermal Stability
Thermal stability constitutes one of the prime criteria for glass selection. It underlies the preservation
of a homogeneous glass over time. Theoretically,
glass could naturally evolve to a crystalline state
that is thermodynamically more stable. But such
thermally induced transformation gets dramatically
slow (or even stops), when the glass is maintained at
temperatures lower than the glass transition temperature (Tg). Predicting thermal stability at low temperature and in the long term therefore requires
experiments performed in supercooled liquids, as
well as modeling.
For nuclear glasses, the main work in this field was
performed by Orlhac,8 which helped confirm the
thermal stability of R7T7 glass in the very long term.
Devitrification experiments conducted on this
glass9 made it possible to identify three major

100

crystalline phases (CaMoO4, CeO2, and ZnCr2O4)
and two minor phases (albite NaAlSi3O8 and silicophosphate) between 630 and 1200  C. Yet, their
crystallization remains limited (a maximum 4.24 wt%),
as these phases consist of minor glass constituents. Even
after a heat treatment designed for a maximum crystallization (100 h at 780  C), no change can be observed in
the main properties of the nuclear glass (chemical
durability and mechanical properties).
Plotting the nucleation and growth curves of these
phases highlighted several essential points:
 Nucleation sharply emerges during the first few
hours of the treatment, and then stops beyond this

period of time. Nucleation is heterogeneous,
inducing crystallization on the preexisting active
sites. Moreover, nucleation curves are strongly
amplified and shift to lower temperatures in the
presence of insoluble noble metal particles;
 Seed crystal growth is very low, and, after a few
dozen hours, displays a saturation phenomenon;
 Strong nucleation coupled with slow growth globally leads to a material which can hardly be devitrified (Figure 11).
The stability of high-level R7T7-type nuclear glass
at low temperature and in the long term was then
investigated by modeling. The mathematical model
selected is on the basis of the KJMA theory (KJMA,
for Kolgomorov, Johnson, Mehl, and Avrami), and
describes the transformation kinetics as a function
of time and temperature.10
Atomic diffusion is the main factor which limits
crystallization, as demonstrated by measuring the
diffusion activation energy. Viscosity is therefore the
key parameter which determines the nucleationgrowth kinetics in glass: it is this very parameter
which conditions diffusive atomic transport in the
silicate-based liquid. Consistently, nucleation-growth
kinetic processes can be determined by means of

1000

10 000
Time (years)

Figure 10 The sequence of alteration of a vitrified waste package.



Waste Glass

467

Maximum
theoretical
crystallized
fraction (wt%)
0.6%

ZnCr2O4 zincochromite

1%

2.44%

CeO2 cerianite

CaMoO4 powellite

Total 4.24%
NaAISi3O8 albite
Silicophosphate

600

700

800


900

1000

1100
1200
Temperature (ЊC)

Figure 11 Temperature range for the nucleation and growth of the main crystalline phases likely to be formed after a
devitrifying thermal treatment of a borosilicate nuclear glass, which confines high-level effluents arising from Uranium Oxide
(UOX) fuel treatment. From CEA DEN Monographs ‘‘Nuclear Waste Conditioning’’. Etienne Vernaz, E´ditions du Moniteur:
Paris, 2009; ISBN 978-2-281-11380-8.

independent viscosity measurements in a broad
range of temperatures.
The model validation was achieved under isothermal conditions on a simplified barium disilicate glass,
known for its homogeneous, fast crystallization.
Applying this model to the R7T7 glass shows that
periods of several millions of years are required for
the three main phases to be completely crystallized at
any temperature below 600  C. Clearly, if during the
slow cooling of large industrial block there is no,
or minute, crystallization in the temperature range
900–600  C, then there will be neither any other
crystallization, nor crystal growth in the long term,
for kinetics reasons. These results confirm the thermal stability of actual HLW confining glasses.
5.18.4.2 Glass Resistance to
Self-Irradiation
The main sources of irradiation in nuclear glasses are

a-decays from actinides, b-decays from FPs, and g
transitions accompanying a- and b-decays.11
Alpha disintegrations are characterized by the production of a heavy recoil nucleus and the emission of a
light a-particle, yielding a helium atom at the end of
the path. Recoil nuclei (RN), shedding large amounts

of energy over a short distance result in atom displacement cascades, thus breaking a large number
of chemical bonds. Alpha decays are thus the main
cause of atomic displacement. On the other hand,
b disintegrations produced by FPs and g transitions
lead mainly to electronic interactions (electron excitation and ionization) with the glass network atoms
but not to atomic displacements.
The effects of self-irradiation have been studied
by investigation of glasses doped with short half-life
actinides, by atomistic modeling, and by external
irradiations.
5.18.4.2.1 Investigations of glasses doped
with a short half-life actinide

This investigation method is the most representative
of nuclear glasses aging: isotropic irradiation is produced in the whole of the glass volume preserving
the electrical neutrality (unlike external irradiations),
a particle and the recoil nucleus are produced allowing electronic interactions and atomic displacements,
and eventually helium builds up in the glass, exactly
as it will in the real case.
The effects of a disintegrations were mainly
investigated through studying 244Cm-doped-glass
glasses that can integrate within a few years doses



468

Waste Glass

equivalent to those to be delivered to the nuclear
glass during thousands of years under disposal
conditions.12
Figure 12 shows the ‘DHA’ Atalante laboratory.
It is an example of a shielded line at Marcoule
(France) devoted to HLW studies, where actinidedoped glass samples are fabricated to carry out
‘accelerated’ studies of the effects of self-irradiation.
The inset in the figure shows a 238Pu-doped glass
block manufactured in the Vulcain laboratory in 1975.
Results produced all over the world, in France,13
UK,14 US,15 or Japan16 are quite consistent. Because
of the effect of a-decay, the glass density decreases

slightly (Figure 13) and its mechanical properties
appreciably improve, especially fracture toughness
that characterizes glass resistance to cracking. The
variations in these properties reach a saturation level
and stabilize beyond 2Â1018 a gÀ1.
Up to a dose of 1019 a gÀ1, no helium accumulation (He bubble) is observed. No significant change
in glass durability is observed either. Furthermore,
there is no dose rate effect, as variations can be
reproduced among the various glasses under study
that exhibit quite different integration rates, spreading over four orders of magnitude.
5.18.4.2.2 Atomistic modeling of glass
self-irradiation


The second approach to understand radiation
damages focuses on atomistic modeling. In particular,
molecular dynamics can provide insight into the
ballistic effects induced by the deceleration of RN
emitted at the end of a decay. Numerous studies
conducted on simplified glasses representative of
the basic matrix nuclear glass (SiO2, B2O3, Al2O3,
Na2O, and ZrO2) demonstrated the remarkable
capacity of this type of glass to restore its structure
following the passage of a recoil nucleus.
The following conclusions could be drawn from
the whole of the calculations performed in relation to
individual cascades in glasses:

Figure 12 The Atalante laboratory for high-level waste
‘DHA.’

a-decay dose (α g−1)
1018

2.1018

3.1018

4.1018

5.1018

Density variation (%)


0.1
0

244Cm-doped

glass (0.04%)

244Cm-doped

glass (0.4%)

−0.1

244Cm-doped

glass (1.2%)

244Cm-doped

glass (3.25%)

Fitting by an exponential law

−0.2
−0.3
−0.4
−0.5
−0.6
−0.7


Figure 13 Evolution of the density of curium-doped glasses with the a-decay dose. From CEA DEN Monographs ‘‘Nuclear
Waste Conditioning’’. Etienne Vernaz, E´ditions du Moniteur: Paris, 2009; ISBN 978-2-281-11380-8.


Waste Glass

Figure 14 Evolution of displacement cascade from the
initial glass to the reconstruction of the glass structure after
the ballistic phase. From CEA DEN Monographs ‘‘Nuclear
Waste Conditioning’’. Etienne Vernaz, E´ditions du Moniteur:
Paris, 2009; ISBN 978-2-281-11380-8.

 displacement cascades take place in two separate
steps (Figure 14):
– The ballistic stage during which collisions
between atoms take place as a whole. This phase
coincides with the strong heating of the matrix
and a depolymerization of the structure by interatomic bond breaking. In parallel, a decrease in
atom density can be observed within the cascades;
– The relaxation stage during which glass structure reconstruction takes place. The glass
structure then experiences significant reconstruction to a state close to its initial state, but
still with a slight structural depolymerization
and a slight swelling on the whole.
In Figure 14, four stages of the evolution of displacement cascade can be seen. In the top left corner, the
initial glass containing the uranium atom (light blue
atom) to be accelerated with 800 eV energy (t ¼ 0 ps).
In the top right corner, the start of the ballistic phase
induced by the uranium projectile (t ¼ 0.013 ps).
In the bottom left corner, the final step of the ballistic
phase when the maximum number of broken bonds

can be observed (t ¼ 0.038 ps). In the bottom right
corner, reconstruction of the glass structure after
the ballistic phase (t ¼ 0.25 ps).
A model of cumulative local quenching was built
from these data in order to help explain the origin of

469

the small evolution observed under irradiation, as
well as the origin of their stabilization under high
doses.17
As displacement cascades accumulate, the glass
structure is fully destabilized by nuclear interactions.
Then the material can be quickly reconstructed without any external energy and its structure is close to
that of glass frozen at high temperature, which results
in the observed evolutions.
When the whole of the glass volume has been
damaged once by the displacement cascades, any
new a disintegration produced will again temporarily destabilize the structure, but the latter will
be rebuilt in the same way as after the first damage.
So the glass no longer undergoes significant change,
which could explain why its properties are stabilized beyond a given dose. It is worthwhile mentioning that the saturation dose experimentally
observed in relation to macroscopic property evolutions (as seen on Figure 13) coincides with that
required for full glass damaging by displacement
cascades, which corroborates with the proposed
model.
As a conclusion, the insignificant evolution of
nuclear glasses under a self-irradiation with respect
to crystallized minerals could be explained by the self
repairing properties of the glass structure.

5.18.4.2.3 External irradiation of glasses

This complementary type of investigation is based
upon the use of nonradioactive glasses in which irradiation stress is simulated by external irradiation
techniques (neutrons, heavy ions, electrons, and g).
The major disadvantages of this experimentation
type consist of the upsetting effects of injected high
dose rates in low irradiated volumes. Today, accurate
knowledge of these offsets allows relevant effects to
be sorted out from experimental artifacts and the
results obtained are quite comparable to those
obtained with other investigation methods.
This type of study was undertaken as early as the
1980s to evaluate the effects of b disintegrations.
Glasses with chemical compositions representative
of the industrially produced nuclear glasses were
thus irradiated by electrons during 1 year up to
doses equivalent to those received in about
1000 years of disposal, that is, 70% of the total bg
dose received in 1 My. These irradiations have not
entailed detectable modifications of the macroscopic
properties (density and mechanical properties).
In addition, glass still displays a homogeneous
microstructure after irradiation.


470

Waste Glass


5.18.4.3

Nuclear Glass Alteration by Water

The mechanisms which control nuclear glass leaching kinetics have been investigated worldwide for
more than three decades. This large accumulated
knowledge allows building computational models
likely to be used for performance assessment of a
geological repository. These models have to be applicable to all the vitrified waste packages industrially
produced, taking into account many different environmental conditions.

With an activation energy of about 75 kJ molÀ1, r0
ranges over seven orders of magnitude between the
temperatures of 4 and 300  C. This is why natural
glasses exhibit very low alteration at room temperature even after millions of years but high alteration
under hydrothermal condition (‘hyaloclastites’). This
value explains also the inherent difficulty of measuring r0 at room temperature, (alteration of about 1 nm
per day hindered by interdiffusion) while a ‘Soxhlet’
measurement operating at 100  C will allow a prompt
measurement, representative of the sample in its mass.

5.18.4.3.1 Basic mechanisms of glass
alteration

5.18.4.3.3 Alteration rate in saturated
conditions and final rate of glass dissolution

In contact with water, the main alteration mechanisms of borosilicate glass are the following18,19:

In a closed system or under conditions in which the

water renewal rate is very slow, such as in the case of a
geological repository, apparent silica saturation is
observed in the leachate and a strong ‘rate drop’ is
systematically evidenced for borosilicate glass.
(‘Leachate’ refers to the leaching solution after contact
with the glass sample.) This rate drop has been related
both to affinity effects (a decrease in the hydrolysis rate
coupled with an increase in the concentrations in solution), and to the formation of an alteration gel standing
as a diffusive barrier between glass and solution.20
Once these ‘saturation conditions’ are established, a
steady rate of glass alteration is generally observed.
This ‘residual rate’ seems to be related to the phenomenon of gel dissolution and to the rate of secondary
phase precipitation. For most nuclear borosilicate glass
composition this final rate is very slow (about 5 nm
yearÀ1 at 50  C for R7T7-type glass).
The origin of such a small residual rate of glass
alteration observed in saturation conditions seems to
be related to the persistence of a weak dissolution
rate of the gel probably itself resulting from the slow
precipitation of secondary phase (mainly phyllosilicates if pH does not exceed 10).15
In some specific cases ‘alteration resumption’ can
be observed, once a residual rate regime has been
established. This so-called ‘renewed alteration’ is
observed only for very alkaline conditions (equilibrium pH higher than 10) and specific glass composition (with high Al/Si ratio). This can be related to a
high precipitation rate of some specific secondary
phases (zeolites).21

 Exchange and hydrolysis reactions involving the
mobile glass constituents (alkalis, boron, etc.) rapidly occur during the initial stages.
 Slower hydrolysis, especially of silicon, drives the

initial glass dissolution rate.
 The in situ condensation of many hydrolyzed species (Si, Zr, Al, Ca, RE, etc.) results in the creation
of an amorphous gel layer at the glass/solution
interface regardless of the alteration conditions.
This layer is more or less reorganized by hydrolysis and condensation mechanisms according to the
environmental conditions.
 This amorphous layer can soon constitute a barrier
against the transport of water toward the glass and
of solvated glass ions into solution. The existence
of this transport-inhibiting effect rapidly causes
this layer to control glass alteration when water
renewal becomes very low.
 Some glass constituents released from the glass
during process can precipitate as crystallized secondary phases. The precipitation of these crystallized phases within or on top of this amorphous
layer or in solution, can sustain glass alteration by
consuming the elements that form the protective
barrier.
5.18.4.3.2 Initial rate of glass dissolution

The initial dissolution rate r0 is the hydrolysis rate
obtained in pure water when no diffusion barrier
slows down the kinetics of alteration. It is an intrinsic
property of the material that characterizes its chemical durability in water. This rate depends mainly on
the glass composition, the temperature, and the pH.
It is the maximum rate of glass alteration for a given
temperature and pH.

5.18.4.3.4 Essential role of the ‘passivating
reactive interphase’


Historically, this apparent saturation state described
above was expressed in equations as if equilibrium
with the fresh glass could be achieved. Today it is


Waste Glass

considered that a saturation state can only be
achieved with respect to a hydrated layer. As saturation is approached in solution, the rate of condensation of many gel forming elements (Si, Al, Zr, Ca, etc.)
increases, allowing the formation of a thin amorphous
layer. Frugier et al.19 proposed the term ‘passivating
reactive interphase’ (PRI) to take into account the
fact that not all of the gel layer becomes passivating
but only a thin inner layer in which a high condensation rate has led to closed porosity.22
Diffusion coefficients in the PRI are consistent
with diffusion in solids with values of the order of
magnitude of 10À20 m2 sÀ1. Furthermore, Monte
Carlo modeling of the gel layer formation by hydrolysis and condensation mechanisms allows describing
the conditions for which porosity closure is reached,
in good agreement with experimental data.23
It will be noticed that phosphate glass that can in
some cases24 display quite low initial alteration rates is
not expected to form any PRI, and therefore, no large
rate drop can be expected in saturation condition.
5.18.4.3.5 Influence of glass composition

Within a given domain, glass dissolution kinetics
strongly depends on its composition.25 Some nonlinear effects have been evidenced on the basis of semiempirical statistical methods26 or, in a few cases, fully
explained using experimental approaches and Monte
Carlo numerical model. A given element usually

modifies specifically each kinetic regime. As an example, Ca has no effect on the forward rate but strongly
favors the rate drop as it is incorporated into the PRI
in contrast to Mg that is incorporated in secondary
clay minerals acting as a silicon sink and promoting
higher dissolution rate. From an operational point of
view, rates of actual glass or rates corresponding to the
best or the worst glass within a given domain are
calculated from an empirical equation built from a
limited number of tested glasses. Considering the
R7T7 domain, forward rate ranges from 1.6 to 4.1 g
mÀ2 dÀ1 at 100  C and residual rate from 2.7Â10À5 to
3.2Â10À4 g mÀ2 dÀ1 at 90  C in initially pure water.
5.18.4.3.6 Influence of groundwater and
environmental materials

In a geological repository, water chemistry initially in
equilibrium with the host rock will be disturbed by
engineered materials (stainless steel canister, overpack, liner, concrete, etc.) and also by the heat produced by the glass canister itself. As a consequence, it
is expected that glass dissolves in an open medium
in which temperature, water flow, and composition

471

evolve with time and space, at least during the first
ten thousand years (this time depends on the host
rock and the disposal design). Chemical elements
brought by water, like Ca, Mg, organic matter, etc.,
may influence glass dissolution mechanisms, for
example, by promoting the PRI condensation,27
increasing the hydrolysis rate of the silicate network,28 or the allowing the precipitation of secondary

crystalline phases.29
In a fractured rock environment, such as a granitic
disposal site, the water renewal rate will be the main
environmental parameter. In a clay environment,
with no flow rate (or too low compared to diffusion),
the predominant effects of environmental materials
will be silica sorption onto likely oxide and hydroxide
minerals, and low precipitation of silicate minerals
that act as silica sinks.30 In any environment, these
phenomena can also be expected with iron corrosion
products. This kind of reaction will maintain a high
glass dissolution rate until the close environment of
the glass is saturated. Beyond this transient regime
that can be investigated by reactive-transport codes,
the final rate regime will control long-term glass
dissolution. Predicting such final rate is a challenge
that requires specific integrated mock-ups, in situ
tests, simulation by reactive-transport codes, and validation by natural or archaeological analogues.
Finally, in salt rock, lower rates than in pure water
are expected, especially if the water is weakly
renewed. Numerous studies have investigated effects
of ionic strength and chemical composition of the
brine.31–34 Because most of these studies are quite
ancient, and also because the prediction of the
water availability near the glass and the migration
of soluble species in salt is more complicated than in
other host rocks, their conclusions, namely in terms
of performance assessment should be revisited as
mechanistic knowledge has been improved.
5.18.4.3.7 Influence of glass fracturing


As glass is fractured after cooling, the reactive surface
is greater than the geometrical one. Neither the surface of an actual glass block nor the evolution of the
reactive surface in geological disposal condition has
been precisely determined or estimated. Several
experimental techniques have been employed to
investigate glass blocks cracking networks and determine their impact on glass lifetime.35 Considering an
inactive R7T7 glass block, the largest cracks are
estimated to increase the geometric surface by a
factor around 5 and smallest ones by a factor around
40. Up to now, all countries have calculated glass


472

Waste Glass

package lifetimes considering a constant cracking
factor. Mechanistic modeling under development
will help address this issue in the near future. However, most studies on natural or archaeological analogues have shown that inner cracks have generally
a minor contribution to the overall glass alteration
(cf. Section 5.18.4.3.9).
5.18.4.3.8 Modeling glass long-term behavior

Predicting long-term behavior of glass requires a
multiscale approach as space and time scales related
to key phenomena are too large to be simulated by a
single mechanistic model. As a consequence, discrete
modeling approaches have been developed from
ab initio calculation at atomistic level36 up to performance assessment model at macroscopic level (also

called operational models).37 In between, Monte
Carlo model allows bridging the gap between atomistic level and measured dissolution kinetics.38,39 One
key point is the glass dissolution rate law.
Many mechanistic models based on rate equations
have been proposed.40–42 The most advanced one is
probably the ‘glass reactivity with allowance for the
alteration layer, GRAAL,’ model proposed by Frugier
et al.19 In this model, the glass-related parameters are
the solubility limit of the PRI, the water, and solvated
ions interdiffusion coefficient in this interphase, the
PRI dissolution rate. The other model parameters are
relative to secondary phases likely to precipitate,
depending on the chemical elements supplied by
the glass or by the surrounding medium: phase solubility limits and precipitation kinetics. For R7T7

glass, a very good agreement is observed between
simulation and experimental data for a very large
set of experimental conditions.43
The operational model that is proposed to assess
the R7T7 glass long-term behavior in the proposed
setting for the French geological repository, is the socalled ‘r0!rf ’ model.44 In this model, the rate of glass
alteration is supposed to keep its initial r0 value until
all conditions are obtained to get the final rate (i.e.,
full saturation of the media, with all silica sorption
sites saturated). Then the final rate rf is applied.
For R7T7 glass, the model parameters (alteration
rates r0 and rr, and glass surface area accessible
toward water) were determined as a function of temperature, pH, and glass composition throughout the
whole R7T7 composition range. Uncertainties on
the parameters values were also determined. The

model can be used to calculate the glass block lifetime depending on the time–temperature profile,
the pH of the medium, the date of water ingress in
contact with the glass, and the quantity of accessible
silicon sorption sites on the metal canister alteration
products.
A typical calculated glass lifetime plot is given on
Figure 15. Two assumptions concerning the quantity
of unsaturated sorption sites during the initial rate
phase are proposed and water ingress in contact with
the glass after 4000 years45 is assumed. Should the
environment be saturated, the total package lifetime
of a R7T7 glass will exceed 300 000 years, even in a
pessimistic scenario where a large amount of iron
corrosion product acts as a silica sink.

60

100.0%

50
10.0%

Numerical uncertainties (1s)

40

30
1.0%

20


Temperature (ЊC)

Total fraction of altered glass (%)

Mass of corrosion products = 2730 kg
Mass of corrosion products = 100 kg

Two environmental conditions

10

0.1%
1000

10 000

100 000
Time (years)

1000 000

0
10 000 000

Figure 15 Operational modeling; example of calculation of R7T7 glass lifetime in two different environmental conditions.


Waste Glass


5.18.4.3.9 Natural an archaeological
analogues

Validating predictive models is one of the major
difficulties of investigating the long-term behavior of
containment materials because the relevant time
scales largely exceed what is accessible to laboratory
experimentation. Whenever possible, therefore, natural or archaeological analogues are examined for this
purpose. They enable to check that no long-term
mechanism is forgotten. They give us some very
long-term integrated experiments, against which predictive models can be qualitatively validated.
For instance, archaeological glass blocks from
a shipwreck discovered near the French Mediterranean
island of Les Embiez have been examined because
of their morphological analogy with nuclear glasses
and their known, stable environment. Like nuclear
glasses, these blocks were fractured after production;
they were then leached for 1800 years in seawater.46
In that specific case, a quantitative agreement has
been achieved between geochemical simulation and
measurements on the archaeological artifact showing
that bridging the gap between short-term laboratory
data and long-term natural system is possible via a
rigorous methodology47 (Figure 16).
The same methodology could be applied to much
older basaltic glasses for which the environment can
be characterized. These glasses not only exhibit the
same alteration mechanisms and kinetics as nuclear
glasses in laboratory experiments, but their alteration


473

products also reveal strong similarities, especially
between the palagonite on basaltic glasses and the
gel on nuclear containment glasses, which can constitute a diffusion barrier. These studies can contribute to a finer definition of the chemical model of
nuclear glasses and to the long-term validation of
the gel protective properties.48

5.18.4.4 Conclusions on Glass Long-Term
Behavior
For more than 30 years, a very significant research
effort on nuclear glass alteration mechanisms has
been carried out worldwide, and a large database
has been produced. Academic researches on longterm crystallization, radiation damage, and alteration
by water, enable nowadays a good mechanistic understanding of the key phenomena that can alter nuclear
waste glass properties in the long term.
A sound methodology was established to use the
best of academic knowledge of alteration mechanisms for performance assessment of glass package in
complex environments. This methodology includes
the following:
 Assessing the evolution of the boundary conditions
including normal and incidental scenario of
evolution.
 Understanding elementary alteration processes at
a mechanistic level.

100
Embiez glass
80


Total

% of altered glass

Sext. (r0)
Sint. (D)
60

Total (measured)
Sext. (measured)

40

Sint. (measured)

20

0
102

103

104

105

Time (years)
Figure 16 Predicted percentage of alteration for Embiez glass (curves) and measured alteration of both kinds of
surfaces (stars). Geochemical modeling has been achieved using the Hytec code including the glass reactivity with
allowance for the alteration layer model for glass dissolution, water diffusion within smallest cracks, advection within the

largest ones, and specific boundary conditions related to 56-m deep seawater.


474

Waste Glass

 Assessing the couplings between the different
mechanisms which simultaneously occur within a
given scenario. Such couplings may indeed modify
significantly the global evolution.
Finally, the models describing the different processes
have to be integrated in a global predictive model
which often requires to be simplified by selecting the
most significant processes and parameters. Operational models also include a conservative approach
to overcome the lack of knowledge and wrap the
general trend.
The use of this kind of operational model demonstrates that waste glass lifetime can be over millions of
years if the glass composition is optimized and disposal conditions are appropriate.
Furthermore, through this long-term research on
waste glass a new ‘science of long-term behavior’ has
been developed. This science and methodology is
now applied to numerous other matrixes (cement,
bitumen, spent fuel, etc.).

5.18.5 Vitrification Processes
As nuclear energy is very concentrated, the overall
volume of nuclear waste is small. This is especially
true for HLW that will be concentrated into a small
volume of glass. (For example, the amount of HLW

glass produced each year in France, related to the
reprocessing of the spent fuel of about 50 reactors, is
in the range of 100 m3.) Consequently, the scale of
radioactive waste vitrification facilities is usually
much smaller than that found in traditional glassmaking. In addition, and especially when processing
HLW, the very high levels of radiation preclude
direct contact with the equipment. Any waste resulting from exchange of failed equipment, for instance,
becomes radioactive waste and must be managed as
such. Consequently, HLW vitrification facilities must
be designed to be remotely operated, and to minimize
maintenance as well as secondary waste generation.
Off-gas treatment systems must be very efficient to
remove any volatilized or entrained radionuclide. As
most of waste streams are nitrate-rich, NOx fumes
are produced and must be abated. The whole vitrification process must be contained efficiently in order to
prevent release of radionuclides to the environment.
Another significant difference between traditional
glassmaking and waste vitrification is that, most often,
the waste is in a liquid form while, in glassmaking, the
batch materials are dry solids. For waste vitrification

it is then necessary to evaporate the liquid and calcine the salts prior to reacting them with the glassformers. This operation requires large amounts of
additional energy provided directly in the melter or
in a specific pretreatment step.
In the end, the glass product must be disposed of,
usually in metallic canisters. For that purpose, most
of the time, the glass product must be poured into
these canisters. This requires, first, that glass viscosity
be around 100 P or lower at the time of pouring and,
second, that the vitrification equipment be designed

with a pouring function.
According to the nature of the waste to be vitrified, and the context, a number of processes have
been studied, among which several have been
deployed industrially.
The first attempts were extrapolations of the crucible work performed in the laboratories. The process
was performed batch wise in a single crucible, where
all the operations of evaporation, calcination, vitrification, and evacuation of the product were performed successively. The melting crucible could be
the canister itself (the process was then a ‘lost-crucible’ process) or a melter from which the glass product
was poured into the canister. The first French industrial facility, PIVER (Figure 17), for instance, was of
this type. The metallic melter was heated from the
outside by a stack of inductors.
The facility was used to process actual high-level
radioactive waste into 100 kg glass blocks. Similar
facilities, operated with lost crucibles or not, were
designed or built in various other countries (UK,
Italy, etc.). Very soon, however, it was concluded
that batch processes did not allow throughputs compatible with commercial operation. The PIVER
throughput, for instance, was around 5 kg hÀ1 of glass.
Most countries, then, decided to abandon batch
processes and design continuous vitrification processes, with two major options for feeding the waste:
 One-step processes, where liquid waste is fed
directly to the melter and all the steps of evaporation, calcinations, and vitrification are performed
in it. This is the case for instance of the Defense
Waste Processing Facility at Savannah River, USA.
 Two-step processes where liquid waste is first fed
to a calciner before entering the melter. This is the
case for instance in France, at the AVM facility at
Marcoule or at the R7 and T7 facilities at La
Hague.
In the following sections we will describe the major

existing facilities and the emerging new processes


Waste Glass

HLW solutions

475

Glass frit

Inductors

Feeding
evaporation

Calcination

Melting
refining

Pouring

Figure 17 The PIVER process in France (1969–1980).

being designed to further improve the capabilities
and efficiency of these processes.
5.18.5.1 Existing Processes for
Radioactive Waste Vitrification


FP solution
Recycling

Additive

Calciner

Gas

5.18.5.1.1 The French two-step continuous
vitrification process

Following the PIVER experience, the French CEA
started to develop a two-step process, in order to separate the functions of evaporation-calcination and vitrification, as illustrated in Figure 18. This allows
keeping a melter of relatively small size, as most of
the energy is provided at the level of the calciner.
Another major decision was to select a vitrification
method by which power is supplied to the glass from
the outside, without direct contact of the glass with the
power source. A metallic melter heated by induction
provided by an external stack of inductors was selected,
following the good results obtained with PIVER. This
disposition allows protecting the power source from
contamination. On the other hand, the size of the
melter is limited by the ability to transmit heat to the
core of the molten bath.
The first industrial facility for vitrifying HLW in
France was the AVM which was the first industrial
vitrification facility in the world, commissioned in
1978. This facility has vitrified the HLW solutions

from the UP1 reprocessing plant and is now used to
vitrify the effluents resulting from the decommissioning and decontamination of the same UP1 plant.
This mission is nearing completion, and the AVM
facility is now facing decommissioning after more
than 30 years of successful operation. The experience

Glass
frit

Scrubber

Melter

Canister

Figure 18 The French two-step continuous vitrification
process.

gained from the operation of AVM has been later
incorporated into the design of the larger facilities
R7 and T7 at La Hague, with three vitrification lines
each, which started operation in 1989 and 1992
respectively. The same technology has been selected
for the WVP (Waste Vitrification Plant) at Sellafield
in the United Kingdom.
In the French continuous process used at
La Hague, the concentrated HLW solutions are
received and stored in cooled and stirred tanks. After
sampling and analysis, they are fed at a metered rate to