Solar Energy Materials and Solar Cells 186 (2018) 29–41

Contents lists available at ScienceDirect

Solar Energy Materials and Solar Cells

journal homepage: www.elsevier.com/locate/solmat

Review of accelerated ageing test modelling and its application to solar T

mirrors
⁎ ⁎
Coralie Avenela,b, Olivier Raccurta, , Jean-Luc Gardetteb, , Sandrine Theriasb
a
Univ. Grenoble Alpes, CEA LITEN, Thermal, Biomass and Hydrogen Department, Thermodynamic and Solar Systems Laboratory, F-38054 Grenoble, France; INES, F-
73375 Le Bourget du Lac, France
b
Université Clermont Auvergne – CNRS - SIGMA Clermont, ICCF F-63000 Clermont–Ferrand, France

A R T I C LE I N FO A B S T R A C T

Keywords: Solar mirrors for concentrated solar power (CSP) plants are expected to last at least 30 years. As this delay is far
Solar mirrors too long to obtain useful information regarding in-service degradation, accelerated approaches to weathering
Concentrated Solar Power (CSP) testing are performed by manufacturers and research laboratories in order to quickly assess the lifetime of
Ageing test commercial or new technologies. However, most published studies that have been performed in the CSP field are
Acceleration factor
based on phenomenological approaches. The characterization of the degradation, which mostly considers re-
Modelling
flectance loss, has rarely been linked to physical or chemical processes that are responsible for the degradation of
Lifetime prediction
properties. Furthermore, the general laws that can be established from these data to establish material behaviour
are empirical.
Ageing tests have been used for many years in other fields, particularly in the domain of polymeric materials.
The impacts on the material properties of stress factors such as temperature, irradiation and humidity have been
extensively studied, and models have been proposed for different kinds of materials, even though most are based
on empirical observations. One of the goals of this article is to determine how these models could be applied to
the weathering of solar mirrors, and as such, the goal of this paper is to provide a critical review of the various
models that are most used and accepted by the scientific community. All of these models include material-
dependent parameters, and the values that have been determined in these studies are reported here to list their
order of magnitude.

1. Introduction [14,15]. Issues involving durability then become a bottleneck in

bringing products to the market.
1.1. Solar mirrors for CSP Durability can be defined in many ways, but the most generally
accepted definition is that durability is the period of time during which
Concentrated solar power (CSP) plants are a carbon dioxide-free and a product in its service environment will survive before requiring re-
renewable way to produce energy in which sunlight is concentrated by placement or maintenance. The weathering resistance of mirrors ex-
mirrors on an absorber that heats a fluid. The different kinds of CSP posed to all of these stresses should then be evaluated to predict whe-
technologies and their historical developments have already been re- ther or not a technology is adapted to an implantation site. Mirrors can
viewed [1–9] and therefore are not presented here. Power plants only be classified as 3 types including glass, aluminium and polymer mirrors
use direct normal irradiation (DNI) and consequently must be installed [16,17]. Glass mirrors can be monolithic or laminated. All of these
in specific locations that meet several requirements such as high DNI, structures are shown in Fig. 2. Several plants using glass monolithic
large plane land, etc. Trieb et al. [10] determined potential locations for mirrors have been installed and exploited since the 1980s [9,18–21],
CSP plants on Earth, which are shown in Fig. 1. and little degradation has been observed on the mirrors. However,
Obviously, the stress factors within these locations are rather ag- mirror technologies have evolved since, and new mirrors in the market
gressive (light, temperature, humidity, etc.), and as a consequence, the are different from those installed in these plants [22]. Indeed, the
exposed materials are susceptible to rapid degradation of their prop- backing system of paints used to contain large amounts of lead to im-
erties [11–13]. It is known that plants are profitable only if the lifetime prove stability; nevertheless, recent environmental regulations require
of their components, and in particular that of mirrors, exceeds 30 years a drastic decrease in lead content, or even complete removal [23]. As a

⁎
Corresponding authors.
E-mail addresses: [email protected] (O. Raccurt), [email protected] (J.-L. Gardette).

https://doi.org/10.1016/j.solmat.2018.06.024
Received 9 April 2018; Received in revised form 7 June 2018; Accepted 12 June 2018
0927-0248/ © 2018 Elsevier B.V. All rights reserved.
C. Avenel et al. Solar Energy Materials and Solar Cells 186 (2018) 29–41

Fig. 1. Map of potential locations for CSP plants [10].

Fig. 2. Structures of (left to right) monolithic glass [11], laminated glass, aluminium [33] and polymer mirrors [34] with a magnitude order of layer thicknesses.

consequence but also in order to reduce costs, manufacturers have establishing models to predict the lifetime and the performance evo-
developed new coatings [24], new designs [25] and new technologies lution of the product under normal use [34].
(or have updated the old ones), aiming to make the devices lighter and The goal of any accelerated weathering test is to increase the rate of
easier to shape. These adaptations include aluminium [26–29], polymer degradation of material properties. Methodologies must be developed
[30–32] and glass laminated mirrors, and their efficiency and durability in order to assess the fate and the lifetime of new mirror technologies.
over the course of 30 years has not yet been proven. As stated above, this kind of methodology already exists in other do-
The back side of monolithic mirrors is protected by a paint system. mains, with general methods developed by statisticians [34–36] or
The binder of such paints is a polymer or a mix of polymers, holding methods applied in research fields such as those of polymers, paints and
together other elements such as inorganic fillers and pigments, which coatings [37–39] or microelectronic fields [40,41], even for other solar
provide desired protective properties such as anti-UV, oxygen and technologies [42,43]. All of these domains have a technological ma-
water barriers [33]. Laminated mirrors contain a polyvinyl butyral turity that is greater than that of CSP, and many of durability studies
(PVB) layer to affix the two glass parts together. Polymer mirrors, as have already been performed [44–46]. This could help in establishing
their name suggests, have a polymer top layer and often a polymer novel procedures that could be applied to the specific case of mirrors. In
substrate. Only aluminium mirrors do not contain polymers. In this particular, the experience gained in the domain of polymers is a sub-
regard, it seems relevant to base the study of mirror lifetime on the stantial asset, which has proven that an appropriate combination of
methodologies and models developed in the polymer field. testing conditions and set of methods assessing weather-induced
On the other hand, all solar technologies are confronted with more changes allows for a better understanding of property changes.
or less similar environmental conditions, and so it is also relevant to A general methodology for evaluating durability and lifetime can be
base this review on solar fields such as photovoltaic and solar thermal explained in the 4 steps described below [34,43]:
technologies. Define a performance requirement and a way to monitor the func-
tional properties that are associated with the performance.
Identify environmental conditions to define the main stress factors
1.2. Methodology for service life prediction
that are responsible for the loss of properties and measure their level in
a representative location. This allows establishing a hierarchy of the
Reliable methodologies are required to provide quantitative lifetime
stress factors that considers a probability of occurrence at the applica-
durability projections that support correct design decisions at the ear-
tion site. Meteorological data from sites all over the world are mon-
liest phases of product development. Relevant data should be obtained
itored, in particular in the framework of weather prediction and can be
in shortened times, and one could expect that this would allow for

30
C. Avenel et al. Solar Energy Materials and Solar Cells 186 (2018) 29–41

Many products are submitted to cyclic variations during utilization.

Environmental conditions in particular are cyclic, because of the nat-
ural alternation of day and night and of the seasons each year. This is
why it is sometimes considered that a test with cyclic conditions is
relevant to simulate such in-service conditions [34].
During a step-stress test, a constant stress is applied on a batch of
samples for a given time. If samples do not fail, the stress level is in-
creased to a higher constant level for a new period, and so on step-by-
step until all the samples fail. This method leads to failures in a rela-
tively short time compared to other methods. However, rapid failures
may be unrepresentative of in-service degradation. Furthermore,
modelling such a test is complex because it must account for the cu-
mulative effect of degradation. Step-stress interpretation needs a so-
Fig. 3. The specular reflectance in which θi is the incident angle and φ is the
phisticated computer programme. Thus, the mechanism must be the
acceptance angle [51]. same during all steps [34]. An example of such a model and its appli-
cation to electrical insulation cables was published by Nelson [48].
Progressive stress testing is similar to step-stress testing, except that
easily accessed. Nevertheless, these global data must be handled with
the stress is increased continuously. Consequently, the limits and
care because they correspond to the ambient atmosphere and not to the
complexity are of the same level [34].
material itself. As an example, black materials are heated by a strong
In some in-service conditions, the stress load can be random. An
irradiation and so their temperature is often higher than the ambient
example is building structures or planes that are submitted to wind
temperature [47]. This is why the microclimate of the samples of in-
variations. Such a test can be designed using in-service data, such as the
terest must be monitored in order to obtain a more realistic assessment
mean or median of the stress, its amplitude over a day or a year, etc. To
of the degradation on site. Furthermore, the microclimate should be
model the results, the test is usually regarded as a constant stress test,
determined with a real system during operating, rather than with a lone
using the mean value applied over the test period [34].
sample exposed on the site, because the system itself modifies the en-
vironment of the materials and components.
1.4. Methodology applied to solar mirrors
Laboratory ageing tests allow determination of the impact of each
factor on the product, and so determines which stresses are the most
To the best of our knowledge, the methodology described above has
aggressive. If a particular stress has no direct impact on the product,
not yet been applied to mirrors. Nevertheless, some steps can be taken
there is no real need to take it into account, except if the stress inter-
using data that are available in the literature. The advancement of each
feres with the other stresses, for example enhancing their effect. Every
step is presented below.
factor previously identified is then applied at a higher level than in use
conditions in order to accelerate degradation. Because accelerated
1.4.1. Step 1
testing is expected to provoke the degradation of the exposed materials
In the case of mirrors, the relevant functional property of interest is
in conditions that are relevant of the environmental use conditions, but
the specular reflectance, which is defined in Fig. 3. A first guideline
at higher rate, the real degradation mechanisms must be known in
[49] to monitor this property has been published in the literature, but
order to assess whether or not acceleration modifies the mechanisms.
the methods to properly perform the measurements are still under
This information can come from natural exposure of samples performed
discussion in the involved scientific community [50–52].
in parallel with accelerated tests, or from the experience gained from
existing systems. A long time is required for both methods.
1.4.2. Step 2
The final part aims at correlating accelerated ageing data with
Critical constraints applied to solar mirrors during power plant ex-
natural weathering data, taking into account the level of the various
ploitation have been identified using the experience gained from in-
stresses applied to the materials. The goal is to use mathematical
stalled plants and the analysis of stresses at the various sites. The main
models based on physical or chemical principles reported in the lit-
stress factors were identified, and obviously temperature, humidity,
erature to propose equations that allow for a better prediction of the
irradiation (both global and UV), rain, dew, wind, sand and dust come
long-term degradation of the properties. Even if mathematical model-
into play. In specific cases, other factors are also involved, such as
ling occurs in the end of the methodology, models should be specified
saline mist in coastal areas, chemical pollutants from nearby industrial
while planning the ageing tests. Indeed, all variables needed in the
sites, etc [12,53].
chosen model must be monitored during the test. If they are not, ageing
tests may lead to useless modelling data.
1.4.3. Step 3
Steps 2 and 3 constitute a risk analysis, which allows establishing a
Ageing tests that are currently used for mirror testing have been
hierarchy of the stress factors. This hierarchy includes the two fol-
recently reviewed [16,54–56] as the main studies performed on all
lowing parameters: sensitivity of the material to each stress factor and
kinds of solar mirrors [16]. These tests include thermal cycling, damp
the level of the stresses encountered on a typical location of the ap-
heat, humidity and condensation tests, irradiation tests with UV or
plication, which can be obtained from environmental data [43].
global solar spectrum, neutral and copper accelerated acetic acid salt
spray tests, acid rain and Kesternich tests. Nevertheless, there is cur-
1.3. Stress loading rently no consensus or proof that the current ageing tests are relevant.
Commonly used procedures come from other domains such as the
Once a stress has been chosen, it can be applied in several ways; photovoltaic [57] and paints [58–61] fields, but the performance cri-
constant, cyclic, step, progressive and random methods exist [34]. terion used in these domains cannot be easily transposed to the domain
Constant stress is the most common and simple of these methods. of CSP [55]. For example, one could erroneously conclude that the
Sample batches are run at a constant level of constraint, and several device is not degraded only because “no major visual damage” [62] is
levels are chosen judiciously. This method should allow for the lifetime observed, which is in many cases wrong. Thus, these studies are almost
prediction at low levels if the degradation varies linearly with the stress all only phenomenological and comparative, so none of the observed
level [34]. degradation has been linked to a physicochemical process. This is why

31
C. Avenel et al. Solar Energy Materials and Solar Cells 186 (2018) 29–41

standard ageing tests that account for the results obtained from ac- discussed. As varying the stress load over the testing time noticeably
celerated ageing tests of mirrors and their correlation with the results complicates the modelling work, this article reports only models with a
from natural exposure must be established. It is essential to assess that constant stress or a stress assimilated to a constant value. In this article,
the mechanism involved in accelerated tests and in natural exposure are we will also discuss the possibility and the relevance of using these
similar, otherwise any attempt to transfer the data from the lab to the relationships to obtain a drastic improvement of lifetime prediction of
ground would be meaningless. Several actors in the CSP field are trying mirrors for CSP plants.
to establish relevant standard procedures [55,63] in order to test mir- This kind of review article already exists in the other fields men-
rors with appropriate accelerated ageing tests and to obtain an esti- tioned above, such as in electronics [64] and photovoltaic (PV) tech-
mation of their lifetime, but even if many studies were done, the im- nologies [65]. Some studies review several stress factors [64,65],
pacts of stress factors on mirrors would still have to be ranked. whereas others are focused on one stress factor of interest [41,66].

1.4.4. Step 4 2. Mathematical models with one stress factor

Furthermore, once ageing tests are proven to reproduce the actual
degradation mechanism, it will be necessary to determine the accel- 2.1. Temperature
eration factor of accelerated ageing vs. natural ageing. The goal would
be to ensure that the performed test is sufficient to simulate 30 years or 2.1.1. Constant temperature
a mirror's lifetime, or even more. Temperature is a common factor that accelerates the degradation of
The global acceleration factor A of a test compared to an application a material because the kinetics of chemical processes vary with tem-
site is defined as the ratio between the time t s needed to reach a spe- perature. Most of the experimental data obtained with the temperature
cified extent of degradation and the time of accelerated test tt to reach stress factor are modelled with the Arrhenius relationship [34,64,65],
the same amount of degradation, defined by Eq. (1) [34,43]: shown in Eq. (2), assuming that a chemical degradation process is
ts controlled by a reaction rate constant k [67]:
A=
tt (1) −Ea
k = γ . exp ⎛ ⎞
⎝ R. T ⎠ (2)
Nevertheless, the time of degradation at site ts is seldom known
because the mirror technologies are built with the goal of lasting at where the pre-exponential factor γ is a constant called the frequency
least 30 years in harsh environmental conditions. It is for this reason factor, T is the absolute temperature in Kelvin, and R is either the
that a better understanding of the impact of the level of the relevant Boltzmann's constant or the universal gas constant, depending on units
stress factors on the degradation properties is necessary. Degradation in used for the activation energy Ea. The parameters Ea and γ are the
real conditions, which results from the application of low constraint characteristics of a defined product or material. The frequency factor γ
levels, is assumed to follow the same laws as those established for high is obtained from the collision theory (see the book The Theory of Rate
constraint levels, and an extrapolation of the time to reach a certain Processes [68] for a detailed explanation). The activation energy re-
extent of degradation can be made. presents the minimum amount of energy needed to pass an energy
barrier and achieve a single-step chemical process.
1.5. Purpose of this review However, most of the degradation processes that occur in real cases
are by far more complex and occur at different steps and with different
In this article, we report mathematical relationships that were es- activation energies. The difficulty then becomes to determine if the
tablished in other fields in order to model the effects of the three stress global process can be approximated by the Arrhenius law over the
factors that were previously identified. These laws were empirical or entire range of temperatures of interest. This is relevant if the reaction
sometimes based on rational approaches. To simplify the reading, rate is limited by one slower step, and so the calculated activation
symbols used in all equations were homogenised and summarised in energy represents the activation energy of this particular step. The re-
Table 1 with their usual unit. Since accelerated ageing tests should be sulting value is not an activation energy properly linked to one physical
performed first with only one stress factor at a time in order to identify or chemical process, and so it takes different names in the literature,
the impact of each stress factor independently, the first part of the such as global, apparent, experimental or quasi-activation energy
presented models is about the modelling of each identified stress factor. [64,68].
Then, in the second part, the question of how to mix several factors that The acceleration factor aT calculated from this law is given by Eq.
can be involved and how to model the coupling of these factors is (3) [43]:

Table 1
Meaning of symbols used in equations. Units of the parameters are an indication and can be adapted to the others in the equation used.
Symbol Unit Name Symbol Unit Name

A – global acceleration factor T K temperature

ai – partial acceleration factor of stress factor i Ts K temperature on site
B, C, D – constant Tt K temperature in test
ci mol.L−1 concentration in chemical compound i ts h time on site
ch chemical stress tt h time in test
Ea J.mol−1 thermic activation energy tf h time to failure
k h−1 kinetic constant tw h time of wetness
Is W.m-2 light irradiation on site w – liquid water
It W.m-2 light irradiation in test
I0 W.m-2 optimal intensity α – reaction order constant
m, n – material-dependent constant β % humidity activation energy
p – Schwarzschild's coefficient β’ %−1 humidity activation energy
R J.mol−1.K−1 perfect gas constant γ h−1 frequency factor
RD mol.h−1 rate of degradation (or reaction) θinc ° incidence angle
RHs % relative humidity on site θdec ° declination angle
RHt % relative humidity in test φ mrad acceptance aperture

32
C. Avenel et al. Solar Energy Materials and Solar Cells 186 (2018) 29–41

−E 1 1 to apply this relationship.

aT = exp ⎜⎛ a ⎛ − ⎞ ⎞⎟
⎜ ⎟

R T T (3) A semi-empirical method was proposed to calculate the activation

⎝ ⎝ t s ⎠⎠
energy, in the case of simple gas reaction [68,76–78]. Then, activation
The Arrhenius relationship has been applied to describe the fate of energies calculated using this method were compared to experimental
numerous materials and experiments such as polymers [34–36,38], determination for several gas reactions, and found to be in good
inorganic [69] and organic [70] PV devices, and even old glass solar agreement enough [78]. According to this work, activation energy for
mirrors technologies [13]. Nelson said “in certain applications, if the gas reactions ranges from about 0–480 kJ mol-1 [78]. A value of zero
Arrhenius life relationship does not adequately fit the data, the data are means that the kinetic of the process does not depend on temperature
suspect rather than the relationship” [34]. Although it works well for [68,78]. Even if the method works for simple molecules with few
many cases, some limits must be kept in mind: atoms, it can be complicated to apply it on complex reactions. More-
over, both calculation and experimental determinations contain hy-
• Activation energy assessment: the apparent activation energy cal- pothesis or uncontrolled parameters leading to uncertainties in the final
culated from data describes a global phenomenon, but it ideally values that should be kept in mind [68,76–78].
should be linked to a physical or chemical process to assess the Experimental activation energies were reported for thermoplastic
validity of the obtained value [64,68]. degradation mechanisms [79]: 10–30 kJ mol-1 for yellowing, 0–20 kJ
• One predominant degradation mechanism: if more than one reac- mol-1 for gloss loss depending on the kind and colour of resin, about
tion occurs competitively and if they have different activation en- 60 kJ mol-1 for photooxidation of polyethylene. Activation energy for
ergies, the Arrhenius model will fail to describe the degradation photooxidation of automotive coatings was assumed to be about
process [71]. 20–34 kJ mol-1 [47]. A review of activation energies for the degradation
• One degradation mechanism over the range of temperatures: the of different properties of several polymeric materials shown that it
Arrhenius relationship describes the reaction rate of a single de- varies from 28 to 210 kJ mol-1 [80].
gradation mechanism; consequently, all temperatures used to esti- For PV devices, assessments of activation energy range from 60 to
mate the apparent activation energy are assumed to lead to the same 100 kJ mol-1 [81,82]. Activation energies between 30 and 90 kJ mol-1
degradation mechanism. When the dominant process varies over the were associated to degradation of organic PV devices [70]. However,
temperature range of interest, as well as the related activation en- the activation energy is not always evaluated for the PV device of in-
ergy, several models must be used to describe the material beha- terest, and a range is often taken for further analysis, mostly between 0
viour: one for each interval of temperature, taking the correct ac- and 200 kJ mol-1 [83–86].
tivation energy [72–74]. As an example, it is known that in the case The thermic degradation of solar absorbers was assessed to have an
of polymer oxidation, non-Arrhenius behaviours can be observed, activation energy of 140–210 kJ mol-1 [87,88]. This value was asso-
which have different origins, presences of transitions within the ciated to an oxidation mechanism [88], in agreement with tungsten
range of investigated temperatures, existences of different processes oxidation activation energies of 142 and 153 kJ mol-1 found in another
with different activation energies, etc [74]. It is even said that the study [89].
kinetics of ageing polymers usually do not follow the Arrhenius law, The temperature can be considered as a full stress factor, but it is
except in some cases [75]. During data analysis, it is then necessary also a specific case. Indeed, temperature is always present, whatever
to verify that there is no variation in the apparent activation energy the level, and it is known to play a role in reaction kinetics.
with temperature, and therefore no variation of the degradation Consequently, a rise in temperature will contribute to an acceleration of
process, because the extrapolation of accelerated tests to outdoor any degradation initiated by other stresses, and this is why the models
climatic conditions is done over a wide range of temperatures for other stresses have frequently been established taking into account
[64,68]. the temperature. These models aimed at describing the effect of the
other stresses in the first place and the temperature is incorporated in
In a case in which the Arrhenius law fails to fit the experimental the mathematical relationship, but it is not considered as a second stress
data, other models have been developed. A thermodynamic model factor in its own right, such as humidity or irradiation would be. This is
based on quantum mechanics and involving free energy, heat and en- why they are not considered to be coupling models by their authors,
tropy of activation was developed by Glasstone, Laidler and Eyring and so in this review, they are classified as one stress factor models.
[68]. The simplest form is known as the Eyring Eq. (4). It is based on the
Arrhenius equation but contains one more term, FT, which is a function 2.1.2. Cycling temperature
of temperature. In general, the literature considers FT = Tm, with 0 ≤ m During outdoor use, the material is exposed to dramatic variations
≤ 1 [64–66]. in temperature. Indeed, the temperature varies according to day and
night alternation and seasonal alternation; thus, there are also small
−Ea
k = FT . γ . exp ⎛ ⎞ variations between years.
⎝ R. T ⎠ (4)
Regarding this fact, a constant temperature test does not seem re-
For the typical ranges of temperature that are adopted for ac- levant to simulate the outdoor behaviour of a material. Indeed, thermal
celerated tests, the factor m has a relatively small effect because it is cycling induces damages that are different than those resulting from
generally close to 0, so the relation given in Eq. (4) becomes the Ar- exposure to a constant temperature. The Coffin-Manson relationship (6)
rhenius equation. Furthermore, the values of both parameters are cal- is most commonly used to model the temperature cycling behaviour
culated from experimental data only, and so one value strongly depends [34,64,90],
on the assumed value of the other. Indeed, the effects of Ea and m on the C
reaction rate tend to compensate and the resulting acceleration factor is N=
(ΔT )n (6)
similar to that of the Arrhenius relationship [64]. This acceleration
factor is described by Eq. (5), where C is a material-dependent constant where N is the number of cycles, ΔT is the temperature range and C and
[43]: n are constants that are dependent on properties of the material and the
test used. The acceleration factor from this relation is described by Eq.
m
T 1 1 (7):
aT = ⎛ t ⎞ .exp ⎜⎛C ⎛ − ⎞ ⎞⎟
⎜ ⎟ ⎜ ⎟

T
⎝ ⎠
s ⎝ ⎝ Tt Ts ⎠⎠ (5) n
ΔT
aT = ⎛ t ⎞
⎜ ⎟

The same precautions as for the Arrhenius law must be kept in mind ΔT
⎝ s⎠ (7)

33
C. Avenel et al. Solar Energy Materials and Solar Cells 186 (2018) 29–41

An empirical extension of this relationship has been established to where p is the Schwarzschild's coefficient, also called the p-coeffi-
account for how the cycling frequency and the maximum temperature cient. When Schwarzschild established this law, he assumed that the p-
impact the ageing of the material. The Coffin-Manson relationship be- coefficient had a constant value of 0.86, which fitted reported experi-
comes Eq. (8), where freq is the cycling frequency, m is a property of the ments well. Nevertheless, it appeared in later experiments that the p-
material, Ea is a quasi-activation energy, and Tmax is the maximum coefficient depends on the studied material and that it could even vary
temperature reached during the cycle [64]. within a batch of the same material, and in some cases, it also changes
with the radiant flux range [46].
C 1 −Ea×11605 ⎞
N= × × exp ⎛⎜ ⎟ The reciprocity law should be considered as a specific case of the
(ΔT )n (freq)m ⎝ Tmax ⎠ (8) Schwarzschild's law with p = 1. Experiments made on the reciprocity or
All of these models contain parameters that depend on the tested Schwarzschild's laws have been reviewed by Martin et al., [46] who
material, and so the difficulty is to estimate them accurately. In parti- reported a list of all the values of p found in the literature. In the do-
cular, since the (apparent) activation energy is in the exponential main of photography, the study of photoconductive materials and of
factor, a little change in its value induces a large variation in the ac- photodegradation of materials, p ranges between 0.5 and 1, but there
celeration factor. are some cases where it may reach 1.2 [46].
The acceleration factor based on Schwarzschild's law can then be
2.2. Irradiation calculated by Eq. (11), assuming that p is constant over the entire range
of intensities [43]:
Degradation due to light irradiation is usually called photo- ts I
p

degradation. The impact of light on the properties of materials has been aI = = ⎛ t⎞

⎜ ⎟

tt ⎝ Is ⎠ (11)
studied in numerous fields [46,64], but it is mainly in the photographic
domain that methods to quantify the effect of the physical parameters Reciprocity and Schwarzschild's laws are the most used to describe
were first developed [91,92]. Models have subsequently been applied to the photoresponse of materials, but other models have been developed.
other fields such as medicine, biology or environmental ageing tests. Although most of them are seldom used, a few have gained acceptance
The basic law used in photodegradation is the reciprocity law, [46]. One is a graphical technique proposed by Halm which is based on
which says that the same dose produces the same effect, independent of Kron's catenary Eq. (12), where I0 and C are constants that vary ac-
the dose rate. In other words, sources of light of different intensities I cording to the material. This model describes the behaviour of de-
produce the same amount of degradation under different times of ex- gradation in the case of a reciprocity failure. The constant C governs the
posure t if the multiplication of irradiation by time results in a constant shape of the curve and can vary within the range of intensity, whereas
value. This means that the product of a photochemical reaction is only I0 corresponds to the optimal intensity, at which a minimum amount of
dependent on the total energy employed, which is the product of irra- exposure is required to produce a given degradation. It corresponds to
diation and time and is independent of the two factors separately. This the minimum of the curve on the graph [46,92,94].
is illustrated in Fig. 4, which shows different exposure regimes with C −C
varying intensity and time but with the same integrated area, indicating log (I × t ) = constant+log ⎡ ⎛I⎞ ⎛I⎞ ⎤
⎢ I0 + I0
⎜ ⎟ ⎜ ⎟
⎥
the same dose of irradiation. The photoresponse of a material is as- ⎣⎝ ⎠ ⎝ ⎠ ⎦ (12)
sumed to be the same in each case according to this law. The first ex-
The historical development and all of the experiments associated
periments to establish the reciprocity law were published in 1859 by
with this law are well reviewed in Chapter VI of the book written by
Bunsen and Roscoe [46,91,92].
Mees [92], so they are not described here. At low intensities, this
However, there are many cases that clearly show deviation from this
equation can be reduced to (13). This representation is shown in Fig. 5
law. In particular, the law often fails when very low or high intensities
[46,92].
are used [46]. Schwarzschild showed in 1900 that the law should in-
stead consider the product I ×t 0.86 [93]. log (I × t ) = constant −ClogI (13)
This empirical model was generalized later as the two equivalent
forms (9) and (10) [46]: The linear form of this equation is given by Eq. (14), and it is a
1
specific case of the Schwarzschild's equation with p = 1 ± C [92].
I × t p = constant (9)
1
(10) I × t1 ± C = constant (14)
Ip × t = constant
Other common graphical representations include log (I × t ) vs. log (t )
and log (I ) vs. log (t ) plots. Another solution is merely to represent
log (photoresponse ) vs. log (I ) . If a straight line is obtained,

Fig. 4. A selection of radiant flux VS exposure times for testing the law of re- Fig. 5. Halm's graphical method for presenting reciprocity data in which the
ciprocity, in which the integrated areas for each exposure regime are identical. logarithm of the dosage (log It) necessary to produce a fixed photoresponse is
When the reciprocity law is obeyed, the photoresponse is the same [48]. plotted against the logarithm of radiant intensity (log I) [48].

34
C. Avenel et al. Solar Energy Materials and Solar Cells 186 (2018) 29–41

Schwarzschild's law is obeyed and the slope provides the p-coefficient. If for the degradation due to the corrosion that includes the following four
it is not the case, Schwarzschild's law may still be obeyed but the p- steps [102]:
coefficient varies within the material or with the irradiation flux. These
representations are particularly used to link the photoresponse of a 1) permeation of water through the plastic encapsulant, usually mod-
material with the UV environmental exposure. To do this, they can be erately fast,
adapted to include temperature and humidity effects [46]. 2) diffusion of moisture through the passivation layer, usually low,
Even if the literature is abundant on the subject, especially in the 3) transport of contaminant ions by water, usually moderately fast,
photographic field, many of the published data have to be handled with 4) electrochemical corrosion of the metal, usually fast once initiated.
care. On one hand, before the 1970s, controlling the irradiant flux and
knowing its value could be difficult because almost no device was 2.3.2. Relative humidity vs water vapour pressure
available to monitor this parameter. On the other hand, measuring the Another difficulty to take into account with the moisture effect is
photoresponse of the exposed materials was complex and sometimes that it can be modelled using the partial vapour pressure of water,
consisted of subjective assessment, due to the lack of accurate mea- which describes the amount of water present, or the relative humidity
surement devices [46]. (RH) which is usually expressed in percent and is defined as [64,103]:
The reciprocity law and the reciprocity breakdown model only ac- Vapour Pressure
count for the total amount of light received by the material. However, RH = × 100
Saturation Vapour Pressure (15)
the wavelength distribution of the light to which the material is sub-
mitted also plays a part in degradation [95]. Polymers in particular are Identifying which parameter, between relative humidity and water
sensitive to specific wavelengths, because chemical bonds only absorb vapour pressure, is the most relevant has been discussed a lot in the
light with a wavelength below or equal to a specific value [96,97]. literature [41,66,103–105]. In another study from the microelectronic
Consequently, a large cumulative dose of light with a long wavelength field, Gunn et al. [104] designed several experiments with a goal to
can cause no damage to a polymer material, whereas a small dose of determine it. They kept vapour pressure constant while varying tem-
short wavelength light can cause significant damage [95]. This is why perature and so the relative humidity because saturation vapour pres-
the wavelength distribution of the laboratory sources of light used to sure is temperature dependent. The results showed that degradation is
perform accelerated ageing tests must be known precisely, especially faster when relative humidity increases, even if the temperature is
the short wavelength part. lower and the water vapour pressure is constant [104]. This agrees with
The temperature must be controlled carefully during experiments an earlier result from Striny and Schelling [105] for a dynamic random
because it has been shown that the degradation of polymer materials is access memory (DRAM) and an aluminium metallized N-type metal-
accelerated by temperature [95]. oxide semiconductor (NMOS) integrated circuit. They compared four
Bauer et al. developed several models [47,95,98–100] to predict the kinds of relationships to model the effect of humidity on these devices.
photooxidation of polymer materials. The authors worked in the auto- They demonstrated that the ones based on relative humidity and on an
motive domain and specifically studied the ageing of coatings exposed Eyring law, presented below by Eqs. (16) and (17), fitted the data better
to light. In this domain, South Florida is a reference site for natural than the one with vapour pressure. Furthermore, they represented these
ageing [95]. The photooxidation rates of polymers in different locations four models as a function of humidity and temperature to determine the
around the world were compared to the data obtained in the case of dependence on each factor in order to assess their reliability at low
exposures in South Florida, and on this basis, relative models of ageing humidity. This analysis confirmed the exclusion of the model using the
were extrapolated. The simplest model only uses the UV dose of the water vapour pressure because of physical non-sense at low humidity
exposure sites, whereas a second model includes the temperature of [105].
samples and a third adds the humidity effect [95]. The relative humidity has been accepted as the parameter that is
used in accelerated tests modelling, rather than the usual thermo-
2.3. Humidity dynamic parameter that is the vapour pressure, because its effect on
corrosion's progress seems to be independent of temperature [64–66].
2.3.1. Degradation mechanism due to humidity For this reason, only the models using relative humidity are presented
Humidity can participate in the degradation processes and at least in this review.
often accelerates these processes, especially corrosion mechanisms and
some chemical reactions that provoke degradation of the materials. 2.3.3. Eyring based models
Indeed, moisture can lead to corrosion if there is a path to the corrosion The two models developed by Striny and Schelling from the Eyring's
site, a contaminant and an electrical potential difference [64–66]. At law are shown by Eqs. (16) and (17), where tf is the time to failure and β
least two steps are involved in the corrosion process: diffusion of water and β’ are an activation energy term for humidity (others terms have
molecules to the corrosion site and the corrosion reaction itself. been described before and can be found in the Table 1) [105]:
In the case of microelectronic packaging, the Lawson experiments
E β ⎞
showed that diffusion was not the rate limiting step, but that it was the t f = C . exp ⎛ a +
corrosion reaction itself [101]. In this specific case, modelling only the ⎝ R. T RH ⎠ (16)
corrosion reaction can be accurate enough to represent the degradation E
process and allow for simplification of a model that can quickly become t f = C . exp ⎛ a − β′.RH ⎞
⎝ R. T ⎠ (17)
complex [66].
Nevertheless, another experiment was led by Gunn et al. [102] on When applied to the data reported in the article published by Striny
plastic encapsulated bipolar integrated circuits, and results showed that and Schelling [105], Eq. (16) leads to Ea = 67 kJ/mol and β = 528%,
the degradation process was limited by the diffusion process of water whereas Eq. (17) leads to Ea = 77 kJ/mol and β’ = 0.12%-1. Even if
through the passivation layer of the devices. Identical components were both fit the humidity test data, model (16) has a behaviour at low
passivated by quartz or by silicon nitride, and then they were aged in an humidity that is more relevant than model (17), so it has been selected
accelerated test with temperature and humidity. The activation en- by the authors. The accelerated factor so obtained is described by Eq.
ergies for the degradation, using the Arrhenius relationship only, were (18) and takes into account the temperature and relative humidity.
106 and 29 kJ/mol, respectively for the two passivation layers, and this E 1 1 1 1 ⎞⎞
difference was explained based on the different kinetics of water dif- aT, RH = exp ⎛⎜ a ⎛ − ⎞+β ⎛
⎜ ⎟ ⎜ − ⎟ ⎟

R
⎝ ⎝ Ts Tt ⎠ ⎝ RH s RHt ⎠⎠ (18)
fusion processes in these materials. The authors proposed a mechanism

35
C. Avenel et al. Solar Energy Materials and Solar Cells 186 (2018) 29–41

Another result of the work by Gunn et al. [104] is that degradation is model in general books on ageing tests methods and modelling [34,90],
different between high and middle levels of humidity, and such that the as well as in studies regarding the humidity effect [41,109].
linked processes are represented by different activation energies. This
highlights the possibility that the degradation of the devices may be 2.3.5. The T + HR correlation
different at in-service conditions, which commonly have a relative This empirical correlation was first developed for plastic packages
humidity that is even lower. Consequently, the extrapolation of lifetime for semiconductors devices to relate failure rate with temperature and
from the results obtained with high humidity tests must be handled relative humidity [110]. This study compared outdoor exposure in
with care, and degradation mechanisms must be studied closely and Panama to the method 106 of the MIL-STD-202 standard [111] for la-
compared to data obtained from natural ageing [104]. Keeping in mind boratory testing. It has been shown that 10 cycles are equivalent to at
these considerations, the authors applied model (16) that was estab- least 6 months of exposure in a tropical environment, which equates to
lished by Striny and Schelling [105] to their data. They found an ac- a laboratory to field acceleration factor of 18. Thus, field data were
tivation energy for the global process of approximately 29 kJ/mol and a obtained operating the device in a coastal area, so the authors noticed
β constant of 296% [104]. that the acceleration factor could be even greater compared to inland
Park et al. [106] also applied model (16) to the degradation of PV operation because of the device sensitivity to saline mist.
modules under humidity and temperature ageing tests, using form (19): Because this test method is performed by a cycle, it is difficult to
model. Consequently, the authors made the assumption that the de-
−Ea β ⎞
RD = C . exp ⎛ − gradation was similar in a test performed at 80 °C / 80% RH, which
⎝ R . T RH ⎠ (19)
seemed reasonable according to their experimental data and to the
They found Ea to be 47 kJ/mol and β to be 281.86%, which are of literature, but with an acceleration factor of only 7.5 compared to the
the same order of magnitude as the results of other previous studies. Panama site.
An acceleration factor to describe the humidity stress aH is given by They found that the degradation rate RD of the investigated plastic
Eq. (20), which is based on the Eyring-Flood relationship where β is a encapsulated device depends on the sum of the temperature T and the
material-dependent constant and RH is the relative humidity during test relative humidity RH of the test, according to Eq. (24) [110]:
(RHt ) or in service (RHs ) conditions [43]: RD = exp (C + n (T + RH )) (24)
1 1 ⎞⎞
aRH = exp ⎜⎛β ⎛ −
⎜ ⎟ ⎟
where C and n are material dependent constants.
⎝ ⎝ RHs RHt ⎠ ⎠ (20) Desombre studied the degradation mechanisms that occur on PV
modules and found the same empirical ageing correlation (24) that
This acceleration factor is the same type as that presented by Eq.
agreed well with the contact corrosion phenomenon [112].
(18) and developed by Striny and Schelling [105], but without the
Later, Cuddihy studied the electrical resistivity of polyvinyl butyral
temperature effect.
(PVB) during ageing tests at several temperature and humidity levels
[103]. PVB is often used as an encapsulant material for PV modules. He
2.3.4. The Peck model
applied the previous model to his data, changing the degradation rate
Another model has been established by Peck for epoxy packages of
for the electrical resistivity of PVB and using form (25):
microelectronic devices, which corresponds to Eq. (21) where n is a
material-dependent constant [107]: ln (u) = C − n (T + RH ) (25)

E where u is the electrical resistivity of PVB, C and n are constants con-

t f = (RH )n . exp ⎛ a ⎞
⎝ R. T ⎠ (21) taining an activation energy term, T is the temperature in degrees
Celsius and RH is the relative humidity in percent.
They found values of n between –2.5 and –3.0, with Ea between 74
Because it fitted the data well, the author tried to demonstrate that
and 78 kJ/mol. The acceleration factor is then given by Eq. (22), as-
this empirical observation had some theoretical validity. He assumed
suming that n and Ea values are constants over the range of temperature
that this behaviour was derived from the Arrhenius relationship to
and humidity. They found an acceleration factor of almost 20 for a test
model temperature and used the relative humidity parameter to ac-
at 130 °C / 85% RH compared to one at 85 °C / 85% RH, taking Ea
count for the water vapour pressure. Varying the temperature and
= 78 kJ/mol [107].
maintaining the same amount of humidity allows calculation of the
RH
n
E 1 1 global activation energy of the product. The tests were performed for
aRH = ⎛ s ⎞ . exp ⎜⎛ a ⎛ − ⎞ ⎞⎟
⎜ ⎟ ⎜ ⎟
30, 40 and 50 °C at 54%, 80% and 100% relative humidity. The plot of
⎝ RHt ⎠ ⎝ R ⎝ Ts Tt ⎠⎠ (22)
the electrical resistivity of the PVB against the relative humidity level
A more recent study [108] showed that photovoltaic modules aged for the 3 temperatures forms straight lines and allows extrapolation of
in a damp heat test (85 °C, 85% RH) obey to a similar law with a the electrical resistivity of the PVB for the 3 temperatures at a theore-
constant pre-exponential factor. They determined Ea to be 86 ± 11 kJ/ tical level of 0% RH. Then, the author used these values to calculate a
mol and n to be –2.2 ± 0.8. The authors also compared these values to global activation energy of the PVB corrosion of approximately 47 kJ/
data from the literature for microelectronic devices and found them to mol, accounting only for the temperature and not the humidity.
be consistent [108]. According to the previous model, the author plotted electrical re-
Another study was done by Park et al. [106] on PV modules, using sistivity as a function of T(°C)+RH(%) and obtained a good con-
form (23): vergence of data into a single straight line. Then, he tried to find
physical laws to explain this empirical observation. After some deri-
−Ea
RD = C . (RH )n . exp ⎛ ⎞ vations and assumptions which are not presented here because they are
⎝ R. T ⎠ (23)
detailed in the original publication, he obtained a complex model based
They found n to be 3.82, using Ea = 47 kJ/mol. It is apparent that n on theoretical relationships such as the Arrhenius equation and ther-
is positive in this case because the model considers the degradation rate modynamic concepts. He next simplified this model into the model
instead of the time to failure. Some mathematical changes easily show (25), which seems to prove that this empirical model has a fundamental
that the two forms lead to the same result, depending on where the basis. In the end, they found a global activation energy approximately
minus sign is placed. This model leads to the same acceleration factor as 50 kJ/mol for the PVB corrosion reaction, which was consistent with
before, described by relationship (22). their previous graphical determination of 47 kJ/mol [103].
This model is well accepted and is currently noted as the Peck's Even if the model works well for these experiments, this kind of

36
C. Avenel et al. Solar Energy Materials and Solar Cells 186 (2018) 29–41

relationship has also been studied by Striny and Schelling to model the 3. Coupling several stress factors
ageing of dynamic random access memory in plastic packages, and they
observed a poor correlation in their data [105]. Indeed, the model is The simplest hypothesis to couple several stress factors is to assume
based on empirical observations made for specific data, and its validity a linear interaction between them. So, relationship (28) has been pro-
for other materials must be carefully verified. To do that, an experi- posed, where the global acceleration factor A is the product of all
mental plan such as the one proposed by Cuddihy [103], with 3 tem- models established for each stress [43].
peratures and 3 humidity levels, can be performed. This kind of ex- A = aT × aRH × aI × ach × … (28)
perimental plan has also been used for photovoltaic modules [106], for
example. Thus, the final relationship has disharmonious dimensions However, sometimes a synergic effect can occur between several
(for example, degrees Celsius are added to percent) and that can be stress factors, and global degradation does not behave according to the
hazardous or not physically consistent. previous law. To take this into account, empirical relationships can be
established to model the degradation that occurs during a test using all
stress factors of interest. Relationship (28) can be a starting point in this
2.4. Others stress factors case, and some terms to model interactions will be added.
A model was developed by Jorgensen et al. [53] to link the loss of
While the three stresses presented above are considered to be the reflectance to the three main stress factors: temperature, light and hu-
main parameters, other constraints exist in the environment. They can midity. The effect of temperature is taken into account by the Eyring
vary according to specific locations, for example saline mist is a strong model and humidity is taken into account by an Eyring-based model.
stress factor in coastal areas but is almost non-existent inland. D
Δρ = C . IUV . T −ne− T e RH (B + T )
E
Pollutants, especially sulfur dioxide, are also specific stress factors be- (29)
cause of their strong concentrations near industries but weaker con- A similar model was used by Lee et al. [113] to model the loss of
centrations in wild areas such as deserts. specular reflectance, and it can be generalized to all kinds of de-
When a material is used in environmental conditions, it may be gradation with time.
exposed to liquid water such as dew and rain. This stress is also specific
Ea
because it is dependent on climatic conditions: the rate of rain is quite RD = C . IUV . e− R . T . e B . RH . t (30)
different in sand deserts than in tropical climates. Thus, the impact of
The first model proposed by Bauer [47,95,99] to predict the pho-
liquid water is different from the impact of humidity and necessitates its
tooxidation rate of automotive coatings at several locations (see above)
own model. A common acceleration factor to describe the liquid water
was completed a second time by the author in order to include the
stress a w is Eq. (26) [43], where tw is the time of wetness during test
temperature effect, and a third time to include the effect of the hu-
(t w, t ) or service (t w, s ) conditions:
midity. This last model is based on the empirical law (31):
tw , t RD, relative = 1 + 0.2 RH (31)
aw =
tw , s (26)
4. Assessment of outdoor parameters
This law is based on the assumption that degradation only occurs
during wet periods and stops when there is no more liquid water in
Once a model has been validated and the global activation energy
contact with the material.
has been determined, the acceleration factor can be calculated using the
The acceleration factor for a stress provoked by chemicals such as
temperature of the site of interest. Nevertheless, several ways to obtain
pollutants or saline mist is given by Eq. (27), where c i,t , cj,t , c i,s and cj,s
this value exist and can readily affect the lifetime prediction.
are the concentrations of a chemical compound i or j during the test or
The first and most common value used is the mean value of the
in service conditions, respectively, and αi and αj are reaction order
ambient temperature over the course of a year. Indeed, meteorological
constants [43]:
data at many sites over the world are monitored, in particular in the
αi αj framework of the weather prediction, and can easily be accessed.
ci, cj,
× ⎜⎛
t⎞
ach = ⎜⎛
t⎞
⎟ ⎟ ×… However, the mean temperature is not a relevant parameter, because
⎝ ci, s⎠ c
⎝ j, s⎠ (27) the degradation rates do not vary linearly with temperature.
Consequently, the degradation rate can be dramatically faster with a
The relationship can be adapted to the number of chemical com- high amplitude of temperatures than the one calculated with the mean
pounds that take part in the degradation process. temperature. This is the case for all of the processes that are thermally
Another way to calculate the acceleration factor of a test compared activated, such as chemical reactions and diffusion process. In such
to an application site for degradation due to corrosion is to use a re- cases, it is necessary to define a new average temperature to use for
ference metal that has a degradation similar to the studied material. service life prediction.
Data for reference materials are available in the literature, and so de- Tencer et al. [114] proposed to calculate an effective temperature
gradation rates during real ageing tests can be found for several loca- that would be defined as the temperature which would lead to the same
tions. Then, the time needed to reach the same extent of corrosion in degradation of the component if it was exposed to a constant tem-
the test is compared to the site according to the definition of the ac- perature, rather than varying temperatures. They assumed that the time
celeration factor (see Eq. (1)) [43]. of exposure could be divided into discrete equal intervals, because in
Modelling the atmospheric corrosion of a material, even a simple practice the measure is often done with an equal interval of time. Thus,
metal, is quite complex. The degradation mechanism involves several the temperature is continuous in a classical climate model, and they
stress factors in which saline mist is usually preponderant, but tem- assumed that the rate of degradation follows the Arrhenius equation. In
perature and humidity are also aggressive. The role of wind has also the end, the authors proposed the following Eq. (32):
been demonstrated, because it carries salt particles far away from the
sea [43]. −Ea
Teff =
t
To the best of our knowledge, no model has been properly devel-
⎣
1
Rln ⎡ t ∫0 exp ( ) dt ⎤⎦
−Ea
R.T (32)
oped for such stress factors, due to the high complexity and the speci-
ficity of phenomena. Because this equation depends on Ea, different materials and com-
ponent technologies will have different effective temperatures, and

37
C. Avenel et al. Solar Energy Materials and Solar Cells 186 (2018) 29–41

therefore different rates of degradation in similar conditions of ex- can be seen as the protection system formed by the plastic encapsulant
posure. Thus, the same material can have several effective temperatures and the passivation layer, whereas the silver layer would be the elec-
because it depends on the failure mechanism of interest, usually the tronic part. The functional property of electronic devices is the electric
fastest, which can vary according to the exposure conditions. conductivity, whereas the one of mirrors is the specular reflectance.
Pickett and Sargent [79] used an effective temperature defined as Both properties are governed by the central part of the structure: the
the irradiance weighted average temperature. Their hypothesis was that electronic part is surrounded by the encapsulant, whereas the silver
even if the temperature varies outdoors, there should be a constant layer is protected by glass or aluminium on one side and coatings on the
effective temperature which would result in the same amount of de- other side. As demonstrated in the two studies mentioned above
gradation after exposure to the same amount of sunlight. This hy- [101,102], the rate-limiting step depends on the material or on device.
pothesis is mathematically represented by Eq. (33). As a consequence, the conclusions regarding the limiting step and the
associated activation energy cannot be directly applied to the case of
Ea ⎛ 1 1 ⎞
∑ I Δt = ∑ Iexp ⎛⎜ ⎜
R ⎝ Teff
− ⎟⎞ ⎟ Δt
T⎠
mirrors, but the methodology to determine it can be used for mirrors
t t ⎝ ⎠ (33) and provides an order of magnitude for activation energies of these
processes.
The same method was then used by Dumbleton and Haillant [82]. Many studies [105,106] consider several models, usually among the
Although it is easier to know the air temperature of a site, the true 3 kinds presented above (Eyring, Peck and T + HR), and apply them to
temperature of interest is that of the exposed sample. One way to obtain their experimental data in order to determine which one is the best
that temperature is exposing a sample for a long period of time, at least adapted to their specific case. The basic way to identify the best model
a complete year, on the application site and monitoring its temperature would be to determine the degradation mechanism and the related
variations. Nevertheless, in most cases, such experiments cannot be physical law. In practice, this is a difficult task, which is why studies try
performed because of a lack of time and accessibility to real sites. to correlate experimental data as well as possible to existing models.
Another method used by Bauer et al. [95] includes estimating the Laminated mirrors contain a PVB layer, consequently its degradation
material temperature Tmat using the air temperature Tair, a constant C might be correctly modelled by the T + HR model that has previously
that is a function of the material colour and the declination angle θdec been used to model the degradation of the electrical resistivity of PVB
with relationship (34): [103].
Tmat = Tair + C . cos (θdec ) (34) While the relationships were discussed here for glass and polymer
solar mirrors, they could be applied to aluminium mirrors as well as to
other materials. The point is that for each new material, the degrada-
5. Discussion tion mechanism should be identified by material characterizations, and
the kinetic parameters should be determined experimentally and then
As discussed in the introduction, solar mirrors contain polymers, be correlated to this mechanism. Nevertheless in practice, the identi-
except for aluminium mirrors. Specific attention should be paid to fication of such mechanism is hard and necessitates a long period of
polymer parts, even if the specular reflectance seems to be stable. As an analysis with a lot of complementary characterization methods. The
example, works on monolithic mirrors [115–117] have shown that their complete mechanism is then rarely known, but some clues can be given
top paint colour changed during ageing tests under temperature before by similar degradations observed with some well-chosen analysis.
any loss of specular reflectance could be observed. Thus, this colour The minimum and maximum values of the four kinetic parameters
change indicates a modification in the paint structure, which has been Ea, β, n and p reported in this paper are summarised into the Table 2,
evidenced by infrared spectroscopy with an Attenuated Total Re- separately for each field of application.
flectance module (FTIR-ATR). A stress-step method has been applied to solar polymer mirrors
In the current testing procedures, solar mirrors are not often tested [113,120]. Workers used the Ultra-Accelerated Weathering System
at constant temperature [16,55,118] because of a poor representa- (UAWS) developed by NREL and ATLAS [113,121]. The acceleration
tiveness of temperature outdoor variations and because the community factor of the device is assumed to be 63 compared to the South Florida
is not yet aware of modelling methods. A cycling temperature test is example. They concluded that testing plans should be developed using
usually used instead, because it is assumed that it would better mimic this device and that statistical procedure adapted to these experiments
the natural day and night cycle [16,55,56,118]. The maximum tem- would be necessary, but complex to develop [113,120].
perature reached in the test must be chosen carefully, and should not It is well known that polymer degradation is sensitive to the initial
exceed or even be too close to the transition temperature of the poly- polymer composition, in particular to the presence of photooxidation
mers used in the device. catalysts, and thus to the conditions of polymerization. As an example,
As shown before, polymers are sensitive to irradiation, in particular see the work published by Gerlock et al. [97] This is of paramount
UV irradiation, and so the paint system of mirrors will be degraded by importance for the study of mirrors, because the back system of pro-
exposure to sunlight. When the binder is deteriorated, fillers and pig- tective paints has a composition that depends on the manufacturer and
ments are released and can no longer play their protective role [33].
Solar mirrors from different manufacturers have been aged by exposing
the front or back face to irradiation, and the composition of paints were Table 2
monitored by FTIR-ATR spectroscopy [119]. Results have shown that Kinetic parameters ranges from literature with the associated field, where Ea is
the loss of specular reflectance was faster with the back side exposed, in kJ/mol, β is in %, and n and p have no unit.
but only for mirrors whose paint was damaged by irradiation. For those Parameter min max field / application
with a few changes in paints, the loss of specular reflectance was similar
Ea 0 480 gas reactions
with the front and the back face exposed.
Ea 0 210 polymers
At least two steps are involved in the corrosion process: diffusion of Ea 0 200 PV
water molecules to the corrosion site and a corrosion reaction. In the Ea 140 210 solar absorbers
case of monolithic glass mirrors, the silver layer is protected by a β 290 530 microelectronic
system of paints that act as a coating. Water must then diffuse through β 282 – PV
n −3 − 2.5 microelectronic
this barrier or by the edges, which are not always sufficiently protected.
n −4 − 1.5 PV
A comparison can be made between the electronic devices with plastic p 0.5 1.1 photodegradation of materials
packages and monolithic mirrors: the paint system and the copper layer

38
C. Avenel et al. Solar Energy Materials and Solar Cells 186 (2018) 29–41

can even change based on a specific batch of samples. Indeed, as the and analysis (tech. rep.), Pacific Northwest Laboratory, 1979.
current market is still new, the production of mirrors is not stabilized [12] R. Girard, C. Delord, A. Disdier, O. Raccurt, Critical constraints responsible to solar
glass mirror degradation, Energy Procedia 69 (2015) 1519–1528.
yet and manufacturers adapt their production methods and materials. [13] A. Czanderna, K. Masterson, T.M. Thomas, Silver/glass mirrors for solar thermal
systems (tech. rep.), Solar Energy Research Institute (SERI), 1985.
6. Conclusions [14] International Energy Agency (IEA), Technol. Roadmap - CSP (2010).
[15] M. Mehos, C. Turchi, J. Vidal, M. Wagner, Z. Ma, C. Ho, W. Kolb, C. Andraka,
A. Kruizenga, Concentrating Solar Power Gen3 Demonstration Roadmap (tech.
The different relationships used in the literature to model the effect rep.), National Renewable Energy Laboratory and Sandia National Laboratories,
of the main environmental constraints on solar mirrors is reviewed, and 2017.
[16] A. García-Segura, A. Fernández-García, M. Ariza, F. Sutter, L. Valenzuela,
their application to the various technologies of solar mirrors is dis- Durability studies of solar reflectors: a review, Renew. Sustain. Energy Rev. 62
cussed. The impact of temperature is usually modelled by the Arrhenius (2016) 453–467.
law, sometimes by the Eyring law which contains one more term that [17] A. Czanderna, P. Schissel, Specularity and stability of silvered polymers, Sol.
Energy Mater. 14 (3) (1986) 341–356.
does not drastically change the results. The degradation mechanisms
[18] G.E. Cohen, D.W.K.G.J. Kolb, Final report on the operation and maintenance im-
must be determined and should not vary over the range of temperature provement program for concentrating solar power plants (Tech. Rep. SAND99-
including accelerated tests and service life. The effect of irradiation is 1290), Sandia National Laboratory, 1999.
usually modelled by the reciprocity law or is generalized using the [19] W. Fuqiang, C. Ziming, T. Jianyu, Y. Yuan, S. Yong, L. Linhua, Progress in con-
centrated solar power technology with parabolic trough collector system: a com-
Schwarzschild's law. Short wavelengths of light are more energetic and prehensive review, Renew. Sustain. Energy Rev. 79 (2017) 1314–1328.
can induce degradation which would not occur in outdoor conditions, [20] National Renewable Energy Laboratory (NREL), Concentrating solar power pro-
so the spectral distribution of lamps used in accelerated tests must be jects. 〈http://www.nrel.gov/csp/solarpaces/index.cfm〉.
[21] NREL, Solar two demonstrates clean power for the future, 2000.
monitored to ensure that they do not contain short wavelengths that are [22] L.A. Weinstein, J. Loomis, B. Bhatia, D.M. Bierman, E.N. Wang, G. Chen,
not present in the solar spectrum. Modelling humidity effects seems to Concentrating solar power, Chem. Rev. (2015).
be more complex, and three empirical models are still being used in the [23] Rapport du sommet mondial pour le développement durable, Organisation des
Nations Unies (ONU), 2002.
literature: Eyring-based model, Peck's model and T + HR correlation. [24] C. Atkinson, C.L. Sansom, H.J. Almond, C.P. Shaw, Coatings for concentrating
These models are more or less adapted to different cases, and they all solar systems - a review, Renew. Sustain. Energy Rev. 45 (2015) 113–122.
should be used in new studies in order to determine which has the best [25] E.M. Toygar, T. Bayram, O. Das, A. Demir, The design and development of solar
flat mirror (solarux) system, Renew. Sustain. Energy Rev. 54 (2016) 1278–1284.
fit with the experimental data. [26] R. Almanza, P. Hernandez, I. Martinez, M. Mazari, Development and mean life of
Accelerated tests may be designed for two distinct purposes: to try aluminum first-surface mirrors for solar energy applications, Sol. Energy Mater.
to be as representative as possible of outdoor conditions but with a Sol. Cells 93 (9) (2009) 1647–1651.
[27] R. Almanza, C. Jiefeng, G. Correa, M. Mazari, Further option for solar con-
faster degradation rate, or to determine the material-dependent con-
centrators: aluminum first surface mirrors, Sol. Energy 54 (5) (1995) 333–343.
stants used in modelling. The first rule to be followed is that the ob- [28] J. Sun, X. Li, W. Zhang, K. Yi, J. Shao, High-reflectivity mirrors by Al2O3, LaF3
served mechanism must be the same as the outdoor predominant me- and AlF3 for 193 nm application, Opt. Laser Technol. 56 (2014) 65–70.
chanism. Determine the whole mechanism needs a lot of in-depth [29] T. Fend, G. Jorgensen, H. Küster, Applicability of highly reflective aluminium coil
for solar concentrators, Sol. Energy 68 (4) (2000) 361–370.
characterization so in practice, most of studies only gives elements of [30] R.A. Assink, Abrasion resistant polymer reflectors for solar applications, Sol.
comparison between outdoor and indoor exposures. Energy Mater. 3 (1) (1980) 263–275.
[31] S. Akyol, S. Ahrens, F. Jahr, C. Rehberger, E. Lüpfert, Cost impact model for using
polymer film based lightweight mirror construction in CSP plant, in: Proceedings
Acknowledgement of SolarPACES, 2010.
[32] A. Zuzuarregui, B. Coto, J. Rodriguez, K.E. Gregorczyk, U. Ruiz de Gopegui,
The research leading to this article has received funding from J. Barriga, M. Knez, Highly reflective polymeric substrates functionalized utilizing
atomic layer deposition, Appl. Phys. Lett. 107 (6) (2015) 061602.
French national funding from Agence Nationale de la Recherche on the [33] D. Weldon, Failure Analysis of Paints and Coatings, Wiley, 2009.
program Investissements d’Avenir (no. ANR-11-EQPX-0014) and from [34] W. Nelson, Accelerated Testing: statistical models, tests plans, and data analyses,
the Auvergne region. John Wiley & Sons, Inc, 1990.
[35] W.Q. Meeker, L.A. Escobar, V. Chan, Using accelerated tests to predict service life
in highly-variable environments, Service Life Prediction: Methodology and
Declarations of interest Metrologies, (2000).
[36] R.L. Feller, Accelerated Aging - Photochemical and Thermal Aspects, The J. Paul
Getty Trust, 1994.
None
[37] I. Vaca-Trigo, W.Q. Meeker, A Statistical Model for Linking Field and Laboratory
Exposure Results for a Model Coating, Springer US, 2009, pp. 29–43.
References [38] K. Gillen, R. Clough, Time-temperature-dose rate superposition: a methodology for
extrapolating accelerated radiation aging data to low dose rate conditions, Polym.
Degrad. Stab. 24 (2) (1989) 137–168.
[1] G. Zhu, T. Wendelin, M.J. Wagner, C. Kutscher, History, current state, and future [39] J.W. Martin, S.C. Saunders, F.L. Floyd, J.P. Wineburg, Methodologies for
of linear Fresnel concentrating solar collectors, Sol. Energy 103 (2014) 639–652. Predicting the service lives of coating systems, U. S. Dep. Commer., Technol.
[2] V.S. Reddy, S. Kaushik, K. Ranjan, S. Tyagi, State-of-the-art of solar thermal power Adm., Natl. Inst. Stand. Technol. (NIST) (1994).
plants - a review, Renew. Sustain. Energy Rev. 27 (2013) 258–273. [40] H.-Z. Ding, Z. Wang, Time-temperature superposition method for predicting the
[3] A. Fernandez-Garcia, E. Zarza, L. Valenzuela, M. Perez, Parabolic-trough solar permanence of paper by extrapolating accelerated ageing data to ambient condi-
collectors and their applications, Renew. Sustain. Energy Rev. 14 (7) (2010) tions, Cellulose 14 (2007) 171–181.
1695–1721. [41] C.G. Shirley, THB reliability models and life prediction for intermittently-powered
[4] B. Gallego, The history of American CSP - A timeline, CSP Today (2012). non-hermetic components, in: Proceedings of the 32nd Annual International
[5] V. Jebasingh, G.J. Herbert, A review of solar parabolic trough collector, Renew. Reliability Physics Symposium, 1994, pp. 72–77.
Sustain. Energy Rev. 54 (2016) 1085–1091. [42] B. Carlsson, K. Moller, M. Köhl, M. Heck, S. Brunold, U. Frei, J.-C. Marechal,
[6] Concentrating solar power: its potential contribution to a sustainable energy fu- G. Jorgensen, The applicability of accelerated life testing for assessment of service
ture, policy report 16, European Academies Science Advisory Council, 2011. life of solar thermal components, Sol. Energy Mater. Sol. Cells 84 (14) (2004)
[7] M. Chaanaoui, S. Vaudreuil, T. Bounahmidi, Benchmark of concentrating solar 255–274.
power plants: historical, current and future technical and economic development, [43] M. Köhl, B. Carlsson, G. Jorgensen, A.W. Czanderna, Performance and Durability
Procedia Comput. Sci. 83 (2016) 782–789. Assessment: Optical Materials for Solar Thermal systems, Elsevier Science, 2004.
[8] X. Py, Y. Azoumah, R. Olives, Concentrated solar power: current technologies, [44] D.A. Cocuzzi, G.R. Pilcher, Ten-year exterior durability test results compared to
major innovative issues and applicability to west african countries, Renew. various accelerated weathering devices: joint study between ASTM international
Sustain. Energy Rev. 18 (2013) 306–315. and national coil coatings association, Progress. Org. Coat. (2012).
[9] D.A. Baharoon, H.A. Rahman, W.Z.W. Omar, S.O. Fadhl, Historical development of [45] J. White, A. Turnbull, Weathering of polymers: mechanisms of degradation and
concentrating solar power technologies to generate clean electricity efficiently: a stabilization, testing strategies and modelling, J. Mater. Sci. 29 (1994) 584–613.
review, Renew. Sustain. Energy Rev. 41 (2015) 996–1027. [46] J.W. Martin, J.W. Chin, T. Nguyen, Reciprocity law experiments in polymeric
[10] F. Trieb, C. Schillings, M. O’Sullivan, T. Pregger, C. Hoyer-Klick, Global potential photodegradation: a critical review, Progress. Org. Coat. 47 (2003) 292–311.
of concentrating solar power, in: SolarPACES Conference, 2009. [47] D.R. Bauer, Interpreting weathering acceleration factors for automotive coatings
[11] M. Thomas, M. Lind, C. Buckwalter, J. Daniel, J. Hartman, Heliostat mirror survey using exposure models, Polym. Degrad. Stab. 69 (3) (2000) 307–316.

39
C. Avenel et al. Solar Energy Materials and Solar Cells 186 (2018) 29–41

[48] W. Nelson, Accelerated life testing - step-stress models and data analyses, IEEE point, IEEE Trans. Electr. Insul. EI-15 (4) (1980) 331–334.
Trans. Reliab. vol. R-29, (2) (1980) 103–108. [81] B. Roeder, M. Köhl, J. Schlothauer, S. Hoffmann, Linking accelerated ageing tests
[49] S. Meyen, M. Montecchi, C. Kennedy, G. Zhu, M. Gray, J. Crawford, S. Hiemer, and outdoor testing, 2nd Atlas/NIST Workshop Photovolt. Mater. Durab. Agenda
W. Platzer, A. Heimsath, M. O’Neill, S. Ziegler, S. Brändle, A. Fernández, (2013).
Parameters and method to evaluate the solar reflectance properties of reflector [82] D. Dumbleton, O. Haillant, Environmental durability of PV modules - a model for
materials for concentrating solar power technology (tech. rep.), SolarPACES accelerated testing, in: Proceedings of the 37th IEEE Photovoltaic Specialists
workgroup, 2013. Conference, 2011, pp. 91–95.
[50] F. Sutter, P. Heller, S. Meyen, R. Pitz-Paal, C. Kennedy, A. Fernández-García, M. [83] P. Rajput, O. Sastry, G. Tiwari, Effect of irradiance, temperature exposure and an
Schmücker, A new method to characterize degradation of first surface aluminum Arrhenius approach to estimating weathering acceleration factor of Glass, EVA
reflectors, in: Proceedings of the SolarPACES Conference, 2010. and Tedlar in a composite climate of India, Sol. Energy 144 (2017) 267–277.
[51] J.S. Crawford, J. Stewart, J.A. Pérez-Ullivarri, A comparison of three portable [84] T.J. McMahon, Accelerated testing and failure of thin-film PV modules, Progress.
reflectometers for use in operations and maintenance of CSP plants, in: Photovolt.: Res. Appl. 12 (2004) 235–248.
Proceedings of the SolarPACES Conference, 2012. [85] B. Carlsson, K. Moller, M. Köhl, U. Frei, S. Brunold, Qualification test procedure for
[52] C.A. Arancibia-Bulnes, M.I. Pea-Cruz, A. Mutuberra, R. Diaz-Uribe, M. Sanchez- solar absorber surface durability, Sol. Energy Mater. Sol. Cells 61 (3) (2000)
Gonzalez, A survey of methods for the evaluation of reflective solar concentrator 255–275.
optics, Renew. Sustain. Energy Rev. 69 (2017) 673–684. [86] S. Kurtz, K. Whitfield, D. Miller, J. Joyce, J. Wohlgemuth, M. Kempe, N. Dhere, N.
[53] G. Jorgensen, H.-M. Kim, T. Wendelin, Durability studies of solar reflector mate- Bosco, T. Zgonena, Evaluation of high-temperature exposure of rack-mounted
rials exposed to environmental stresses, Durab. Test. Nonmet. Mater., ASTM STP photovoltaic modules, in: Proceedings of the 34th IEEE Photovoltaic Specialists
1294 (1996) 121–135. Conference (PVSC), 2009, pp. 002399–002404.
[54] F. Sallaberry, F. Sutter, H. Almond, C. Delord, A. Fernández-García, [87] O. Raccurt, A. Disdier, Accelerated ageing tests for durability study of solar ab-
T. Diamantino, Review of test methods for durability of reflectors (tech. rep.), sorber coatings on metallic substrate for solar thermal energy (STE) application,
Stage STE (2016). in: Energy Procedia, 2018.
[55] F. Sutter, A. Fernandez-Garcia, J. Wette, P. Heller, Comparison and evaluation of [88] O. Raccurt, F. Matino, A. Disdier, J. Braillon, A. Stollo, D. Bourdon, A. Maccari, In
accelerated aging tests for reflectors, Energy Procedia 49 (2014) 1718–1727. air durability study of solar selective coating for parabolic trough technology,
[56] A. Fernández-García, L. Martínez-Arcos, F. Sutter, J. Wette, F. Sallaberry, R. Erice, Energy Procedia 1850 (2017).
T. Diamantino, M.J. Carvalho, O. Raccurt, A.-C. Pescheux, G. Imbuluzqueta, [89] J. Wendel, Thermodynamics and kinetics of tungsten oxidation and tungsten oxide
M. Machado, Accelerated aging test of solar reflectors according to the new sublimation in the temperature interval 200–1100 °C (Master's thesis), Lund uni-
AENOR standard – results of a round robin test, Energy Procedia (2018). versity, 2014.
[57] “IEC 61215:2005: Crystalline silicon terrestrial photovoltaic (PV) modules - [90] D. Crowe, A. Feinberg, Design for Reliability (ch. 9 - Concepts in accelerated
Design qualification and type approval.” testing), CRC Press, 2001.
[58] ISO 11507:2007: Paints and varnishes - Exposure of coatings to artificial weath- [91] R. Bunsen, H. Roscoe, Photochemische Untersuchungen (I), Ann. der Phys. 184
ering - Exposure to fluorescent UV lamps and water. (10) (1859) 193–273.
[59] ISO 6270-2:2005: Paints and varnishes – Determination of resistance to humidity, [92] C.E.K. Mees, The Theory of the Photographic Process, The Macmillan Company,
2005. 1942.
[60] NF EN ISO 16474-2: Peintures et vernis - Méthodes d’exposition à des sources [93] K. Schwarzschild, On the deviations from the law of reciprocity for bromide of
lumineuses de laboratoire - Partie 1 : Lampes à arc au Xénon, 2014. silver gelatine, Astrophys. J. 11 (1900) 89–91.
[61] ISO 16474-3: Paints and varnishes – Methods of exposure to laboratory light [94] E. Kron, Über das schwärzungsgesetz photographischer trcokenplatten, Ann. der
sources - Part 3: Fluorescent UV lamps, 2014. Phys. 346 (9) (1913) 751–758.
[62] IEC 62108:2007: Concentrator photovoltaic (CPV) modules and assemblies - [95] D.R. Bauer, Global exposure models for automotive coating photo-oxidation,
Design qualification and type approval. Polym. Degrad. Stab. 69 (3) (2000) 297–306.
[63] STAGE-STE Project. 〈http://stage-ste.eu/〉. [96] J. Larché, P. Bussière, J. Gardette, Characterisation of accelerated ageing devices
[64] L.A. Escobar, W.Q. Meeker, A review of accelerated test models, Stat. Sci. 2006 21 for prediction of the service life of acrylic-melamine/urethane thermosets, Polym.
(4) (2006) 552–577. Degrad. Stab. 96 (8) (2011) 1530–1536.
[65] F. Dia, N. Mbengue, O.N. Sarr, M. Diagne, O.A. Niasse, A. Dieye, M. Niang, B. Ba, [97] J. Gerlock, D. Bauer, L. Briggs, J. Hudgens, Photostability of acrylic/melamine
C. Sene, Model associated with the study of the degradation based on the ac- enamels. effect of polymer composition and polymerization conditions on photo-
celerated test: a literature review, Open J. Appl. Sci. 6 (2016) 49–63. initiation rates, Progress. Org. Coat. 15 (3) (1987) 197–208.
[66] D.J. Klinger, Humidity acceleration factor for plastic packaged electronic devices, [98] D. Bauer, J. Gerlock, D. Mielewski, Photostabilization and photodegradation in
Qual. Reliab. Eng. Int. 7 (5) (1991) 365–370. organic coatings containing a hindered amine light stabilizer. Part VII. HALS ef-
[67] S. Arrhenius, Über die reaktionsgeschwindigkeit bei der inversion von rohrzucker fectiveness in acrylic melamine coatings having different free radical formation
durch säuren, Int. J. Res. Phys. Chem. Chem. Phys. 4 (1) (1889) 226–248. rates, Polym. Degrad. Stab. 36 (1) (1992) 9–15.
[68] S. Glasstone, K. Laidler, H. Eyring, The Theory of Rate Processes: The Kinetics of [99] D.R. Bauer, Predicting in-service weatherability of automotive coatings: a new
Chemical Reactions, Viscosity, Diffusion and Electrochemical Phenomena approach, J. Coat. Technol. 69 (864) (1997) 85–96.
(International chemical series), McGraw-Hill Book Company, Incorporated, 1941. [100] D.R. Bauer, J.W. Martin (Eds.), Service Life Prediction of Organic Coatings: A
[69] D.C. Miller, E. Annigoni, A. Ballion, J.G. Bokria, L.S. Bruckman, D.M. Burns, Systems Approach, 722 American Chemical Society, 1999.
L. Elliott, R.H. French, S. Fowler, X. Gu, C.C. Honeker, M.D. Kempe, H. Khonkar, [101] R.W. Lawson, A review of the status of plastic encapsulated semiconductor com-
M. Köhl, P.J. Krommenhoek, L.-E. Perret-Aebi, N.H. Phillips, K.P. Scott, F. Sculati- ponent reliability, Br. Telecom Technol. J. 2 (2) (1984) 95–111.
Meillaud, T. Shioda, S. Suga, S. Watanabe, J.H. Wohlgemuth, Extrapolating ac- [102] J.E. Gunn, S.K. Malik, P.M. Mazumdar, Highly accelerated temperature and hu-
celerated UV weathering data: perspective from PVQAT Task Group 5, NREL PV midity stress test technique (hast), 19th Int. Reliab. Phys. Symp (1981) 48–51.
Modul. Reliab. Workshop (2015). [103] E. Cuddihy, The Aging Correlation (RH + T): relative Humidity (%) +
[70] O. Haillant, D. Dumbleton, A. Zielnik, An arrhenius approach to estimating or- Temperature (°C) (tech. rep.), U.S. Department of Energy, 1986.
ganic photovoltaic module weathering acceleration factors, Sol. Energy. 95 (2011) [104] J.E. Gunn, R.E. Camenga, S.K. Malik, Rapid Assessment of the Humidity
1889–1895. Dependence of IC Failure Modes by Use of HAST, 21st Int. Reliab. Phys. Symp
[71] M. Celina, K. Gillen, R. Assink, Accelerated aging and lifetime prediction: review (1983) 66–72.
of non-arrhenius behaviour due to two competing processes, Polym. Degrad. Stab. [105] K. Striny, A. Schelling, Reliability evaluation of aluminum-metallized mos dy-
90 (3) (2005) 395–404. namic ram's in plastic packages in high humidity and temperature environments,
[72] K.T. Gillen, M. Celina, R.L. Clough, Limitations of the Arrhenius methodology, IEEE Trans. Compon., Hybrids, Manuf. Technol. 4 (4) (1981) 476–481.
Digit. Libr., Credit. UNT Libr. Gov. Doc. Dep. (1998). [106] N.C. Park, W.W. Oh, D.H. Kim, Effect of temperature and humidity on the de-
[73] M. Celina, Review of polymer oxidation and its relationship with materials per- gradation rate of multicrystalline silicon photovoltaic module, Int. J. Photo. 2013
formance and lifetime prediction, Polym. Degrad. Stab. 98 (12) (2013) (2013) 9.
2419–2429. [107] S. Peck, Comprehensive model for humidity testing correlation, IEEE / IRPS 215
[74] L. Achimsky, L. Audouin, J. Verdu, J. Rychly, L. Matisova-Rychla, On a transition (1986) 398–1222.
at 80 °C in polypropylene oxidation kinetics, Polym. Degrad. Stab. 58 (3) (1997) [108] G.M. Kimball, S. Yang, A. Saproo, Global acceleration factors for damp heat tests
283–289. of PV modules (in), 2016 IEEE 43rd Photovolt. Spec. Conf. (PVSC) (2016)
[75] L. Audouin, X. Colin, B. Fayolle, J. Verdu, Sur l’utilisation de la loi d’arrhenius 0101–0105.
dans le domaine du vieillissement des polymères, Mater. Tech. 95 (3) (2007) [109] G. Caswell, Temperature and Humidity Acceleration Factors on MLV Lifetime
167–177. (tech. rep.), DfR Solutions, 2015.
[76] H. Eyring, Steric hindrance and collision diameters, J. Am. Chem. Soc. 54 (8) [110] B. Reich, E.B. Hakim, Environmental factors governing field reliability of plastic
(1932) 3191–3203. transistors and integrated circuits (in), 10th Reliab. Phys. Symp (1972) 82–87.
[77] H. Eyring, M. Polanyi, On simple gas reactions, Z. für Phys. Chem. vol. Abteilung [111] MIL-STD-202G - Electronic & Electrical Component Testing Standards, 2002.
B, (12) (1931) 279–311. [112] A. Desombre, Methodology for a Reliability Study on Photovoltaic Modules,
[78] H. Eyring, The energy of activation for bimolecular reactions involving hydrogen Springer Netherlands, 1981, pp. 741–745.
and the halogens, according to the quantum mechanics, J. Am. Chem. Soc. 53 (7) [113] J. Lee, R. Elmore, C. Kennedy, M. Gray, W. Jones, Lifetime prediction for de-
(1931) 2537–2549. gradation of solar mirrors using step-stress accelerated testing, Workshop Accel.
[79] J.E. Pickett, J.R. Sargent, Sample temperatures during outdoor and laboratory Stress Test. Reliab., NREL (2011).
weathering exposures, Polym. Degrad. Stab. 94 (2) (2009) 189–195. [114] M. Tencer, J.S. Moss, T. Zapach, Arrhenius average temperature: the effective
[80] R.R. Dixon, Thermal aging predictions from an Arrhenius plot with only one data temperature for non-fatigue wearout and long term reliability in variable thermal

40
C. Avenel et al. Solar Energy Materials and Solar Cells 186 (2018) 29–41

conditions and climates, IEEE Trans. Compon. Packag. Technol. 27 (3) (2004) [118] F. Sutter, J. Wette, A. Fernández-García, S. Ziegler, R. Dasbach, Accelerated aging
602–607. testing of aluminum reflectors for concentrated solar power (tech. rep.), DLR,
[115] O. Raccurt, C. Delord, C. Bouquet, R. Couturier, Correlation between solar mirror CIEMAT, Alanod, Almeco (2016).
degradation and colorimetric measurement of protective back layer, Energy [119] C. Avenel, O. Raccurt, J.-.L. Gardette, S. Therias, Accelerated aging of solar mir-
Procedia 49 (2014) 1700–1707. rors: UV irradiation of mirrors back side. SolarPACES Conference, 2017.
[116] C. Delord, C. Bouquet, R. Couturier, O. Raccurt, Characterization of glass mirrors [120] J. Lee, R. Pan, Analyzing step-stress accelerated life testing data using generalized
durability for CSP, in: SolarPACES Conference, p. 10. linear models, IIE Trans. 42 (8) (2010) 589–598.
[117] C. Delord, C. Bouquet, R. Couturier, O. Raccurt, Accelerated weathering of solar [121] H. Hardcastle, G. Jorgensen, C. Bingham, Ultra-accelerated weathering system I:
mirror a new methodology based on the monitoring of protective back layer de- design and functional considerations, JCT Coat. 7 (8) (2010) 28–37.
gradation, 6th Eur. Weather. Symp. EWS (2013).

41