Phosphor thermometry for nuclear decommissioning and waste storage

Nuclear Engineering and Design 375 (2021) 111091

Contents lists available at ScienceDirect

Nuclear Engineering and Design
journal homepage: www.elsevier.com/locate/nucengdes

Phosphor thermometry for nuclear decommissioning and waste storage
Alberto Sposito a, Edward Heaps a, Gavin Sutton a, *, Graham Machin a, Robert Bernard b,
Sandra Clarke b
a
b

National Physical Laboratory, Teddington, Middlesex TW11 0LW, United Kingdom
Sellafield Ltd., Seascale, Cumbria CA20 1PG, United Kingdom

A R T I C L E I N F O

A B S T R A C T

Keywords:
Phosphor thermometry
Nuclear waste storage

Reliable surface thermometry of stored nuclear waste containers is essential for both health monitoring and
corrosion modelling. In this paper we investigate the feasibility of using phosphor thermometry for long-term
temperature monitoring of nuclear waste containers and storage racks. Two strands of research were conduct­
ed. Firstly, two thermographic phosphors (ruby and manganese-doped magnesium fluorogermanate [MFG]),
mixed with six different binders were coated onto stainless steel (316L) substrates and exposed for one month to
gamma radiation (γ-ray) and sodium hydroxide (NaOH) solution. Secondly, a hybrid fibre-optic phosphor
thermometer, capable of measuring these samples, was constructed and tested. The instrument can measure

temperatures using both the luminescence decay time and intensity ratio techniques. Following γ-ray and NaOH
exposure, no significant degradation in phosphor appearance or performance was observed. The MFG/silicone
binder combination gave the best results. The hybrid phosphor thermometer demonstrated for MFG/silicone
binder that:

• the intensity ratio technique gave the lowest measurement standard deviation (σ(T) < 0.1 ◦ C) but
suffered minor drift with thermal cycling.
• the decay time technique (650 nm emission) did not show any drift and was capable of a mea­
surement standard deviation of σ(T) ≤ 0.3 ◦ C.

1. Introduction
Reliable measurement of the surface temperature of stored nuclear
waste containers is crucial for health (integrity) monitoring and corro­
sion modelling. Currently temperature monitoring is performed with
infrared thermometers and spring-loaded thermocouples. However,
both approaches could have large uncertainties, due to a) unknown
surface emissivity and b) background radiation in the former case, c)
heat-flux and contact uncertainties in the latter:
1. Unknown surface emissivity: to convert a measured thermal radiance
to temperature requires accurate knowledge of the surface emissivity
– the emissive power of the surface relative to that of an ideal
blackbody (0 < ε < 1). Infrared thermometers are first calibrated

against a blackbody standard (with an ε very close to 1). However
once used on a practical surface such as a steel storage container, the
emissivity will be significantly <1 (typically around 0.3–0.5 for
brushed steel). The emissivity value used by the infrared thermom­
eter can be adjusted during initial measurements until the instrument
agrees with a calibrated contact sensor suitably placed in good
thermal contact with the surface. However, any changes in the

emissivity during storage (e.g. due to oxidation or corrosion) will
introduce temperature errors that are difficult to quantify, and
additionally the emissivity may also change with temperature and
measurement wavelength (i.e. a new calibration may be needed if
the infrared thermometer is changed). Even with a reasonable
knowledge of the surface emissivity, errors of 10 ◦ C or more are
common.

Abbreviations: MFG, manganese-doped magnesium fluorogermanate (Mn:Mg4FGeO6); NaOH, sodium hydroxide; γ-rays, gamma radiation; UV, ultraviolet; Al2O3,
aluminium oxide; SS316L, 316L stainless steel; pH, potential of hydrogen; NPL, national physical laboratory; PD, photodetector; FWHM, full width half maximum;
LED, light emitting diode; PC, personal Computer; NI, national instruments; BNC, Bayonet Neill–Concelman; dB, decibel; MSE, mean squared error.
* Corresponding author.
E-mail address: (G. Sutton).
/>Received 12 November 2020; Received in revised form 12 January 2021; Accepted 26 January 2021
Available online 20 February 2021
0029-5493/Crown Copyright © 2021 Published by Elsevier B.V. This is an open
( />
access

article

under

the

CC

BY-NC-ND

license

A. Sposito et al.

Nuclear Engineering and Design 375 (2021) 111091

2. Background radiation: for any surface with ε < 1, an infrared ther­
mometer will always collect an amount of thermal radiation that
originates from other hot objects nearby reflected into the mea­
surement path. This can be particularly challenging in the nuclear
storage environment where containers with different levels of ac­
tivity, and hence temperature are stored close together. Without a
detailed numerical model and knowledge of the environment around
the container being measured, it is extremely difficult if not impos­
sible to robustly determine the surface temperature. The magnitude
of errors from background radiation are difficult to assess but can
amount to between 5 ◦ C and 20 ◦ C for a container at 30 ◦ C sur­
rounded by other containers at 80 ◦ C for 0.3 < ε < 0.5, for example.
3. Heat-flux and contact uncertainties: for contact thermocouples that
are brought into contact with the surface to be measured, two factors
affect the accuracy of the measurement. Firstly, sufficient time must
elapse for the contact sensor and the surface to reach thermal equi­
librium. Secondly, the equilibrium temperature will depend on fac­
tors such as the contact pressure/angle and surface finish and the
thermal conductivity of both the sensor and the surface. These errors
can be minimized by waiting for thermal equilibrium to be reached
and by calibrating the contact sensor on the same material as the
container prior to use. In our experience, errors of up to 20 ◦ C are
possible for a container at 80 ◦ C when no mitigation strategies are
employed.

Fig. 1. Measurement principle: luminescence decay time thermometry. The
data shows the change in normalized decay time versus temperature for the
phosphor MFG. The decay time, τ, reduces monotonically with increasing
temperature (see insert).

excitation is switched off, the luminescence intensity decays exponen­
tially. In this example (for MFG), the emission is in a band around 660
nm, i.e. red light. The decay time, τ, is determined by fitting a single
exponential function I(t) = A +Be− t/τ to the decay signal. The decay time
decreases monotonically with increasing temperature, T, and through
suitable calibration the relationship between τ and T can be established
(see insert in Fig. 1). Following calibration, subsequent measurements of
τ allow the temperature to be determined.

Additionally, Infrared thermometers are unable to measure under­
water surfaces as the operating wavelengths are strongly absorbed by
water.
Phosphor thermometry does not suffer from these sources of uncer­
tainty and has the potential to provide a long-term, reliable approach to
surface thermometry of these containers. The technique works by
applying a phosphor coating to the surface, this is then interrogated
optically to determine the surface temperature. To our knowledge, the
technique has never been applied in an ionizing radiation or storage
pond environment, and research is required to qualify it for use in such
conditions. This work describes two interconnected pieces of research:
a) the performance of two thermographic phosphors, manganese-doped
magnesium fluorogermanate (Mn:Mg4FGeO6 or MFG) (Brübach et al.,
2008) and chromium-doped alumina (Cr:Al2O3 or ruby) (Atakan et al.,
2009), mixed with six chemical binders, exposed to gamma radiation

(γ-rays) and typical fuel pond solutions (sodium hydroxide, NaOH), and
b) the development of a hybrid fibre-optic phosphor thermometer
capable of measuring temperature using both the decay time and in­
tensity ratio phosphor thermometry techniques (Brübach et al., 2008,
2013). In the former, relevant to thermometry in fuel ponds, the tem­
perature range investigated was from 5 ◦ C to 65 ◦ C. In the latter, relevant
to thermometry of dry stored containers, the temperature range was
from 20 ◦ C to 220 ◦ C.

1.1.2. Luminescence intensity ratio thermometry
The principle of luminescence intensity ratio thermometry is shown
in Fig. 2. The phosphor is excited continuously with UV/blue light, and
the luminescence intensity measured simultaneously in two narrow
wavelength bands. In this example (for MFG), the emission peaks at 635
nm and 660 nm are used, with the former increasing and the latter
decreasing in intensity, respectively, with increasing temperature. The
intensity ratio, Φ = I(635nm)/I(660nm), versus temperature is then
established through suitable calibration (see inset in Fig. 2). Following
calibration, subsequent measurements of Φ allow the temperature to be
determined uniquely.

1.1. Phosphor thermometry – principle of measurement
There are numerous thermographic phosphors capable of measuring
temperature in the range from 5 ◦ C to 220 ◦ C, and many established
measurement techniques utilising either temporal or spectral properties
of the phosphor for thermometry. For the work described here with the
phosphors MFG and ruby, the luminescence decay time and intensity
ratio techniques are used. As a full exposition of phosphor thermometry
is beyond the scope of this work, we refer the reader to three compre­
hensive review articles for further information (Brübach et al., 2013;

Khalid and Kontis, 2008; Ald´
en et al., 2011), with details of our earlier
work given in (Sutton, 2019). Here we confine ourselves to the basic
principles of the two phosphor thermometry techniques.
Fig. 2. Measurement principle: luminescence intensity ratio thermometry.
The data shows the change in spectral intensity versus temperature for the
phosphor MFG. The intensity ratio, Φ, is seen to increase monotonically with
temperature (see insert).

1.1.1. Luminescence decay time thermometry
The principle of luminescence decay time thermometry is shown in
Fig. 1. The phosphor is excited with UV/blue light, and when the
2


A. Sposito et al.

Nuclear Engineering and Design 375 (2021) 111091

2. Methods

appearance.

2.1. Gamma radiation and NaOH exposure

2.1.4. Calibration of samples
Before γ-ray exposure or NaOH soaking, all samples were calibrated
in the temperature range from 5 ◦ C to 65 ◦ C against a traceably cali­
brated1 Type–N thermocouple in a stirred water bath, using a PICO TC08 temperature logger, and an existing fibre-optic phosphor thermom­
eter, with the decay time technique (McMillan et al., 2019). The ther­

mocouple temperature, T, and the phosphor decay time, τ, were
measured simultaneously. This temperature range covers that expected
for fuel racks within storage ponds.
The setup is illustrated in Fig. 3. The thermocouple was inserted into
a small hole in a 30 mm × 30 mm × 5 mm steel block, so that its tip was
located centrally within the block. The sample to be tested was placed on
top of this block such that the metal surfaces of the block and sample
were in good thermal contact. In order to be representative of the final
operating environment and to enhance temperature uniformity, the
sample and block were placed in a shallow stirred water bath. The probe
of the phosphor thermometer was placed 15 mm above the sample, so
that the area illuminated by it was approximately the size of the sample.
To determine the calibration curve of each sample, the temperature
of the water bath was varied in 5 ◦ C steps, with the temperature allowed
to stabilize before making measurements. The temperature of the water
bath was controlled by adjusting the set-point of the hot-plate for tem­
peratures above ambient temperature and by the addition of small
amounts of ice for below ambient temperatures. For each sample, a
calibration curve τ(T) was generated by taking an average of 60 datapoints from each stable temperature region of the continuous data.
The standard deviation was also recorded. After γ-ray exposure or NaOH
soaking, the samples were re-calibrated with the phosphor thermometer
and the differences from the initial calibration determined.

In this section we describe the selection of the phosphors and
binders, how the substrates were selected and coated, the pre-test cali­
bration of the phosphor samples and the test conditions both with the
gamma exposure and sodium hydroxide soak.
2.1.1. Phosphor selection
The two phosphors, MFG and ruby, were selected based on previous
experience with the former (Sutton, 2019) and the well-known radiation

resistance of the host material (Al2O3) of the latter (Wang et al., 2009).
Specifically, a) the authors have extensive experience of MFG from hightemperature phosphor thermometry research, where it has excellent
temperature sensitivity above 300 ◦ C; b) in our experience, MFG and
ruby both have strong luminescence signals in the temperature region of
interest for this work; c) we examined a number of other phosphors with
higher temperature sensitivity, including YAG:Dy, but found that the
luminescence was significantly less than MFG/ruby - this may in part be
due to the excitation scheme used (e.g. LED @ 420 nm); d) Al2O3 was
identified in an initial literature review to be resistant to radiation
damage (ruby being Cr doped Al2O3).
2.1.2. Binder selection
Six different chlorine-free binders (Aremco 529, 642, 642-A, 643–2,
830 and 880), were shortlisted based on their compatibility with the
material of typical nuclear waste containers, stainless steel 316L
(SS316L), and with typical fuel pond solutions. These were tested in
combination with the chosen phosphors. A list of the binders used for
coating trials is shown in Table 1, together with some specifications
taken from Aremco technical bulletins (Inc, 2020).
Of the binders selected, only one does not require heat curing after
coating (8 3 0). Also, it should be noted that only two binders have a low
viscosity (643–2 and 830), which means that only they can be easily airsprayed without thinning (with water, for example), whereas the others
have a high viscosity (greater than100 cP), meaning they cannot be
sprayed without prior thinning.

2.1.5. Gamma radiation exposure
Two gamma radiation exposure trials were performed using different
sets of samples:
1. Month-long, low-dose γ-ray exposure using a 60Co source of activity
of 13.3 GBq. The samples were fixed onto a cardboard mount at 0.69
m from the source, behind a 1 mm thick Perspex plate and irradiated

2.1.3. Preparation of substrates and phosphor coatings
Prior to coating, the substrates were roughened with sandpaper and
cleaned with solvent to improve coating adhesion. The phosphor pow­
ders and liquid binders were mixed in a ratio of 0.9:1 by weight (Sutton,
2019), agitated in an ultrasonic bath to remove any bubbles and
improve subsequent coating integrity, and then applied to the 25 mm ×
25 mm × 1 mm SS316L substrates. Unless otherwise stated, the coatings
were applied using either a paintbrush or an airbrush, depending on the
mixture viscosity. All samples were heat-cured up to 260 ◦ C (except 830
which required no heat-cure), in two or three steps, according to the
manufacturers’ specification. In all cases, the coating application was
completed once a uniform appearance was obtained, and subsequent
measurement of the coating thickness with a Sauter TC 1250–0.1
thickness gauge showed thicknesses of between 100 µm and 150 µm for
all samples. For the spray coatings specifically, between 3 and 5 passes
where needed at a stand-off of 10 cm, to give a uniform coating
Table 1
List and specifications of binders used for phosphor coating trials.
Binder

pH

Specific gravity
[g/cm3]

Viscosity
[cP]

Solid content by

weight [%]

Heat
cure?

529
642
642-A
643–2
830
880

N.A.
10.7
10.7
11.5
11.4
6.5

1.05
1.41
1.25
1.27
1.20
1.04

(150 – 250)
370
200
30

10
480

68
40
25
30
25
50

✓
✓
✓
✓
×
✓

Fig. 3. Phosphor calibration setup with hotplate and water bath: The Type–N
thermocouple (obscured by the phosphor probe) is inserted into the rear of the
steel block, below the sample.

1
That is, traceably calibrated to the International Temperature Scale of 1990
(ITS-90, the current temperature scale in use around the world) using national
reference standards.

3


A. Sposito et al.

Nuclear Engineering and Design 375 (2021) 111091

with a dose rate of 10 mSv/hr for one month (total dose: 7.2 Sv). This
dose rate is typical of that found on fuel pond storage racks.
2. Short-duration, high-dose γ-ray exposure using a Theratron machine.
The samples were placed on a plinth 800 mm from the source and
irradiated with a dose rate of 90.74 Sv/hr for 9 h and 40 min, giving a
total dose of 876 Sv. This dose was approximately equivalent to a 10
year exposure at 10 mSv/hr). This set-up is shown in Fig. 4.
2.1.6. NaOH soaking
A third set of samples were soaked for one month in a stirred beaker
of NaOH solution typical of that in the fuel pond environment, with a pH
of 11.25. For the duration of the test, the solution was maintained at a
temperature of approximately 40 ◦ C on a hotplate. Fig. 5 shows the
experimental setup.

Fig. 5. The experimental setup used for NaOH soaking of phosphorcoated samples.

2.2. Hybrid fibre-optic phosphor thermometer

was not possible to implement the intensity ratio technique for this
phosphor with this thermometer. However, it was still possible to
measure the temperature of a phosphor-coated sample in five different
ways (important for cross-checking of results):
MFG:

2.2.1. Instrument design
The instrument used for the gamma radiation and NaOH soak tests
described earlier was developed independently by NPL prior to the work

described here. However, it was not optimized for the temperature
ranges of interest here and was only capable of measuring phosphor
temperatures using the decay time technique. Hence a new and novel
hybrid fibre-optic phosphor thermometer was designed and built
capable of measuring temperature using both the luminescence decay
time and intensity ratio techniques. By measuring the temperature with
more than one technique, it may be ultimately possible to assess any
performance degradation due to gamma radiation or NaOH exposure
that would otherwise be missed. It also allows for the identification of
the most appropriate temperature measurement technique in terms of
sensitivity and stability.
A fibre-coupled violet light-emitting diode (LED) was chosen as the
excitation source, as its central wavelength (415 nm) falls within the
absorption bands of both selected phosphors: MFG, with absorption
peaks at ~280 nm and ~420 nm (Thorington, 1950), and ruby, with
absorption peaks at ~400 nm and at ~554 nm (Kumari and Khare,
2011). Two adjustable-gain silicon photodiodes were chosen as photo­
detectors (PDs), offering flexibility in terms of gain/bandwidth and
operational wavelength range (200 nm to 1100 nm). Additionally, a
longpass filter (to block any reflected excitation light) and bandpass
filters (to select a particular emission line) were used. By making use of
interchangeable filter sets, it was possible to measure phosphor tem­
peratures from both MFG and ruby.
The luminescence spectrum of MFG has broad emission peaks
centered at 635 nm and 660 nm (Thorington, 1950), whereas ruby emits
over a narrower wavelength range, with peaks at 690 nm and 694.3 nm
(Kumari and Khare, 2011). Due to the narrow emission range for ruby it

1) Intensity ratio: Φ(T) = I(635nm)/I(660nm)
2) Decay time: τ(T) = τ(635nm)

3) Decay time: τ(T) = τ(660nm)
Ruby:
1) Decay time: τ(T) = τ(690nm)
2) Decay time: τ(T) = τ(694.3nm)
To optimize the measurement wavelengths for the MFG/intensity
ratio configuration, preliminary measurements of the luminescence
spectra versus temperature over the range 20 ◦ C to 200 ◦ C were per­
formed, and, together with a numerical model, it was found that filters at
635 nm and 650 nm (with a FWHM of 10 nm) were most suitable.
In order to maximize delivery of the excitation light from the LED
onto the phosphor coatings and collection and transmission of the
emitted light from the latter to the PDs, large-core multi-mode fibreoptic components were chosen, with a core diameter of 1 mm and nu­
merical aperture of 0.22. These components include a bifurcated Y-fibre
bundle (for excitation and emission light delivery), a patch-cord (for
transmission of excitation light) and a 1 × 2 fibre coupler with 50:50
splitting ratio (to split the incoming phosphor emission signals to the
two PDs). A schematic of the instrument is shown in Fig. 6 and
comprises:
1) a personal computer (PC) with LabVIEW software.
2) a National Instruments (NI) data acquisition (DAQ) system.
3) an optoelectronic interrogator, with the aforementioned components
(LED, PDs, patch-cord and fibre coupler) and power supplies.
4) BNC cables for electrical connections.
5) an optical probe, i.e. the 2 m-long bifurcated Y-fibre bundle (1 × 2),
with stainless-steel monocoil sheathing and SMA connectors on all
ends.

2.2.2. Instrument calibration. The hybrid fibre-optic phosphor ther­
mometer was calibrated against a calibrated Type-N thermocouple with
both phosphors (MFG and ruby), using a variable temperature hotplate,

from 20 ◦ C to 220 ◦ C, in steps of 25 ◦ C. At each calibration point, the
temperature was stable to within ± 0.05 ◦ C for one minute, before
making the measurement and then increasing the hot plate temperature.
This calibration is separate from those carried out for the existing

Fig. 4. The Theratron machine used for short-duration, high-dose γ-ray expo­
sure, with the samples placed on a plinth below the γ-ray source.
4


A. Sposito et al.

Nuclear Engineering and Design 375 (2021) 111091

Fig. 6. Schematic of the hybrid fibre-optic phosphor thermometer.

phosphor thermometer used for the gamma and NaOH tests described
earlier. This temperature range covers that expected for dry stored
containers.
Two calibration targets were prepared (as described in Section 2.1.3)
by air-spraying a mixture of 643–2 binder and either MFG or ruby on
SS316 discs with 25 mm diameter, 5 mm thickness, a 1.5 mm diameter
hole on its side. This achieved a final coating thickness of between 100
µm and 150 µm. Fig. 7 shows a photo of the calibration targets with
uniform coatings, with the MFG and ruby samples appearing yellowish
and pink respectively.
The calibration targets were, in turn, placed at the center of the
hotplate, with the reference Type-N thermocouple secured into the
sample hole with thermal paste. The optical probe of the phosphor
thermometer was placed in suitable proximity to the coated surface of

the calibration target, with the distance between the two optimized to
maximize the phosphor emission signal. Optical alignment was not
changed between repeated tests, unless otherwise stated. In all cases, the
PDs were set with a gain of 60 dB, limiting their measurement band­
width to 9 kHz, and the LED was modulated with a 50% duty cycle
square wave, with 2.5 V amplitude, a period of 57 ms and 2000 samples
per period. A single measurement was performed by averaging 10

periods of the square wave (0.57 s). Data was continuously recorded and
post-processed to generate a calibration file. Standard deviations were
calculated over 60 measurements.
Specifically:
Decay time: 10 decay waveforms over 10 excitation square-wave
cycles are averaged first. Then a two stage fitting procedure is
performed:
1) An initial fit of the complete averaged time series is performed (the
duration of which will always be greater than 7 times the decay time)
– this generates first estimate τest .
2) The dataset is then truncated so that only the data for 1 × τest < t <
7 × τest remains, and a second fit is applied to give τtrue .
This procedure was found to minimise a small contribution from a
faster decay component (i.e. at the start of the decay) and remove a
small dependence of the decay time on the excitation intensity.
Intensity ratio: 10 exponential rise waveforms (i.e. when the LED is
‘on’) at both wavelengths are averaged over 10 excitation square-wave
cycles. Then the last 60 points of each signal is averaged for both
wavelengths and the intensity ratio determined. In all cases, the last 60
points occur when the emission signal is stable.
3. Results and discussion
In this section we describe the results of the tests of the phosphor

samples exposed to gamma radiation and NaOH solution, discuss the
implications of the findings, and identify which phosphor/binder is most
appropriate for the required measurement and environment.

Fig. 7. Calibration targets used for the hybrid fibre-coupled phosphor ther­
mometer: LHS – MFG/643-2, RHS – ruby/643-2.
5


A. Sposito et al.

Nuclear Engineering and Design 375 (2021) 111091

3.1. Gamma radiation and NaOH exposure
3.1.1. Selection of phosphor/binder combination
Twelve phosphor/binder combinations are possible (2 phosphors × 6
binders), with each combination applied to at least 3 SS316L substrates.
Although air-spraying is preferred, producing a more uniform coating,
most of the mixtures were applied with a paintbrush, due to the high
viscosity of most of the binders. Adhesion was tested by attempting to
scratch the coating with a diamond scribe drawn across the coating
surface applying consistent moderate hand pressure. A summary of the
coating trials is given in Table 2, with bubbling and cracking generally
associated with coatings applied with a paintbrush, whereas those that
were air-sprayed generally achieved a good finish.
Of the binders selected, 830 was immediately ruled out as it was
found to be impossible to achieve a good adhesion to the substrate: the
coating cracked and/or exfoliated while drying in air, as shown in Fig. 8.
Of the other five binders, 880 produced relatively good results with both
phosphors; the best results were obtained with the binders 529 and

643–2 (air-sprayed) with MFG and binder 643–2 with ruby.

Fig. 8. Binder 830 coating degradation: a) MFG, b) ruby. Both photos were
taken after the samples were air-dried for ~2 h at room temperature.

3.1.2. Sample calibration prior to testing
As mentioned in Section 2.1.4, each phosphor-coated sample was
initially calibrated by measuring the decay time τ at a number of fixed
temperatures T over the temperature range 5 ◦ C to 65 ◦ C, and the data
fitted with a 2nd-order polynomial of the form:

τi (T) = ai T 2 + bi T + ci
In all cases, the fit residuals were randomly distributed about zero
indicating that the behavior of the calibration data was well represented
by a second order polynomial over the temperature region considered. A
summary of the initial calibration for all samples is given in Figs. 9 and
10 for MFG and ruby phosphors respectively, showing a) the calibration
data, b) the fit residuals and c) the amplitude of the luminescence.
For the MFG phosphor samples, all calibrations yielded similar re­
sults, with little if any binder dependence. However, for ruby, the cali­
bration appeared to depend strongly on the binder used, with 529 and
especially 880 standing out from the others. This may be because these
two binders are silicone-based, whereas all the others are water-based.
Fig. 9 c) and 10 c) show that the luminescence signal amplitude for
MFG is approximately ten times stronger than that of ruby, which results
in a reduction in the statistical (Type A) uncertainty for the MFG mea­
surements compared to the ruby measurements. In principle, the in­
tensity of the excitation source could be increased to reduce this effect,
but this was not done for these measurements, in order to compare the
Table 2

Phosphor + binder combinations and results of coating trials.
Phosphor

Binder

Heat
cure?

Coating
method

Bubbles?

Cracks?

Adhesion?

MFG

529
642
642-A
643–2
830
880

Yes
Yes
Yes
Yes

Yes
No
Yes

Spray*
Brush
Brush
Brush
Spray
Brush
Spray*

No
Yes
Yes
No
Few
No
No

No
No
No
Yes
No
Yes
No

529
642

642-A
643–2
830
880

Yes
Yes
Yes
Yes
No
Yes

Brush
Brush
Brush
Brush
Brush
Brush
Spray*†

Yes
Yes
No
No
No
No
No

No
No

Yes
Yes
No
Yes
No

Good
Good
Good
Bad
Good
Bad
Not so
good
Good
Good
Good
Good
Good
Bad
Not so
good

Ruby

Fig. 9. Phosphor calibration – MFG: a) decay time, b) fit residuals and c) signal
amplitude, versus temperature.

performance of the two phosphors under similar conditions. It is also
evident that the signal amplitude for the 880 binder is lowest for both

phosphors. This is probably because this phosphor/binder mixture was
diluted with water, for thinning purposes. This produced a visually thin

*

= Coatings appear thinner than others, but otherwise ok.
= Phosphor/binder mixture was thick, so thinned with similar amount of
pure water to allow spraying.
†

6


A. Sposito et al.

Nuclear Engineering and Design 375 (2021) 111091

Fig. 11. Phosphor calibration: a) mean square error of the calibration fits and
b) mean signal amplitude versus binder.

3.1.3. Gamma exposure tests
During γ-ray exposure tests, the samples were monitored and pho­
tographed to look for evidence of scintillation, but none was observed.
Following γ-ray exposure, the samples showed no visual signs of
degradation. The lack of scintillation suggests that γ–rays are not
significantly absorbed by the selected phosphors or binders, giving
confidence that they are unlikely to be significantly damaged by expo­
sure to the γ-rays.
3.1.4. NaOH exposure tests
Some of the selected binders were immediately ruled out for further

analysis following the NaOH soak test due to coating degradation and
evidence of poor adhesion. An example of a degraded sample is shown in
Fig. 13, where the coating can be seen to have flaked off the substrate. A
summary of how each binder fared in the NaOH soak test is shown in
Table 3.

Fig. 10. Phosphor calibration – ruby: a) decay time, b) fit residuals and c)
signal amplitude, versus temperature.

looking coating.
In Fig. 11 a) the average mean squared error (MSE) of the polynomial
fits is presented, giving an indication of the quality of the fit, with MFG
showing a consistently lower MSE than ruby. Again, the 880 binder
stands out, in this case as having the largest MSE. Fig. 11 b) shows the
mean luminescence signal amplitude for all phosphor/binder combi­
nations (averaged over all temperatures): as mentioned earlier, the
signal amplitude for MFG is significantly larger than for ruby.
The relative temperature sensitivity of each phosphor can be deter­
mined from 1/τ(dτ/dT), where dτ/dT, the derivative of the calibration
curve τ(T) with respect to T. This is shown in Fig. 12. For both phos­
phors, the sensitivity is negative (i.e. τ decreases as T increases). For a
majority of the temperature range considered here, MFG has a higher
relative temperature sensitivity than ruby, as the magnitude of its de­
rivative is larger. It is also worth noting that the derivative approaches
◦
zero at T ≈ 0 C for ruby and at and around that temperature the phos­
phor cannot be used as a thermometer. Any uncertainty in τ for ruby
near that temperature translates into a much larger uncertainty in T.

3.1.5. Sample re-calibration after testing

Following γ-ray exposure and NaOH soak tests, each sample was
recalibrated to assess changes in the performance of the phosphor as a
surface thermometer. In all cases, and in a similar manner to the initial
calibration, the fit residuals remained randomly distributed about zero
indicating that the behavior of the calibration data was well represented
by a second order polynomial over the calibration temperature region
(5 ◦ C to 65 ◦ C). An example of the calibration data prior to and following
gamma radiation and NaOH exposure for binder 880 is shown in Fig. 14,
and a summary of all tests can be found in the Supplementary on-line
material (Supplements 1 and 2 for MFG and ruby respectively). In the
next section, the pre- and post- test calibration differences are examined
and discussed in more detail.
3.1.6. Calibration drift following gamma radiation and NaOH exposure
By using the calibration polynomial fit functions, it was possible to
7


A. Sposito et al.

Nuclear Engineering and Design 375 (2021) 111091

Fig. 12. Phosphor calibration – relative temperature sensitivity, 1/τ.(dτ/dT) versus temperature for all binders: a) MFG and b) ruby.

determine the differences in decay time at specific temperatures even
though the calibration temperatures varied from sample to sample. Once
the decay time differences Δτi were calculated, the equivalent temper­
ature differences were found using the small errors formula: ΔTi =
(Δτi )/(dτ0 /dT), where Δτi is the difference in decay time for sample i
from the initial calibration and dτ0 /dT is the derivative of the calibration
curve τ(T) with respect to T, also taken from the initial calibration.

The changes in phosphor calibration following the gamma radiation
and NaOH soak tests are given in Fig. 15, where dashed lines represent
the uncertainty in the difference measurement for each solid curve of the
same color. A comprehensive uncertainty budget for these measure­
ments can be found in the Supplementary on-line material (Supplement
3).
It should be noted that for some binders, results are not presented for
the NaOH soak. This is because those samples did not survive the
soaking intact, with significant loss of coating from the substrate.
Additionally, no month-long gamma exposure result is presented for the
ruby/529 combination. This is because preparation of the sample could
not be completed in time for the beginning of that test.
In general, the calibration drift is within 1 ◦ C for MFG and within
5 ◦ C for ruby. There are some exceptions to this, in particular MFG with
binder 643–2 after the month-long gamma exposure, where the
magnitude of the drift is up to 2 ◦ C. The general lack of any systematic
pattern in the results between different phosphor/binder combinations
suggests that the observed changes are likely to be caused by changes in
the measurement setup used for the calibrations than induced by γ-ray
exposure or NaOH soaking. The 1 ◦ C repeatability of MFG is consistent
with what would be expected from measurements made with the
phosphor thermometer.
The uncertainties associated with the differences between two

Fig. 13. Example of a coating degraded by the NaOH soak test: ruby phosphor
with 642 binder.
Table 3
Summary of binder conditions after NaOH soaking.
Binder

Condition

529
642
642-A
643–2
880

No damage
Coating came off
Coating came off
Coating is softer, but intact
No damage

Fig. 14. Calibration comparison for binder 880, mixed with: a) MFG, b) ruby.
8


A. Sposito et al.

Nuclear Engineering and Design 375 (2021) 111091

Fig. 15. Gamma and NaOH soak test results – calibration drift: all samples. Dashed lines represent the uncertainty in the difference measurement for each solid curve
of the same colour.

calibrations are ≤1 ◦ C for MFG, whereas those for ruby are generally
higher, between 2 ◦ C and 8 ◦ C. For both phosphors, uncertainties are
higher for samples with a thinner coating (i.e. 880 binder) and thus
lower emission intensity. Considering that the uncertainties are higher
than the observed differences in the calibrations following the tests it is

unlikely these are due to any test induced change to the coatings.

shorter wavelength increases, whereas that at the longer wavelength
decreases with temperature. The intensity ratio, Φ = V(635)/V(650), is
plotted versus temperature in the same figure: a clear monotonic trend
can be observed, allowing the generation of a unique calibration file. A
quadratic fit of experimental data generates residuals ≤ 1.5 ◦ C, but
higher-order polynomial fits produce smaller residuals (<0.1 ◦ C with
4th-order fit). In practice, the calibration data is stored in the software in
its entirety and spline interpolation used to convert any subsequent in­
tensity ratio measurement into phosphor temperature, leading to fit
residuals similar or superior to the 4th order polynomial method. By
making use of the initial calibration it is possible to determine changes in
the intensity ratio versus temperature relationship of subsequent mea­
surements in terms of temperature. This provides a means to calculate
temperature differences arising from the repeat calibrations separated
over time versus the thermocouple temperature (which is assumed sta­
ble for the duration of the results reported here).
(
)
In Fig. 17 a) the standard deviations, σ (Ttc ) and σ Tphos , of the
thermocouple and phosphor temperatures, respectively, are shown for
the initial calibration. While σ(Ttc ) is constant at approximately 0.01 ◦ C,
(
)
σ Tphos tends to increase with temperature. This is probably due to an
overall degradation of the signal-to-noise ratio with increasing tem­
perature, but it is still < 0.08 ◦ C across the whole temperature range.
Calibration residuals, i.e. the difference between the phosphor and
reference thermocouple temperature at each calibration point, were

insignificant being < 0.01 ◦ C. However, subsequent calibrations showed
that the repeatability was larger; this is presented in Fig. 17 b), which
shows that the temperature differences (ΔT) against the reference
thermocouple were as large as 0.6 ◦ C. It is worth noting that the optical
alignment of the phosphor thermometer was changed between the first
and subsequent calibrations which may have contributed to the
observed difference. In addition, three thermal cycles of the calibration
target were performed for other measurements between the second and
third intensity ratio calibrations and, therefore, it is possible that the
MFG coating underwent some annealing during the thermal cycling.

3.2. Hybrid fibre-coupled phosphor thermometer
3.2.1. Instrument calibration with MFG and the intensity ratio technique
Three calibrations were performed using the MFG calibration target
and the intensity ratio phosphor measurement technique. Fig. 16 shows
the amplitude of the luminescence signal at 635 nm and 650 nm versus
temperature from the initial calibration (similar results were observed in
the two other subsequent calibrations): the amplitude of the signal at the

3.2.2. Instrument calibration with MFG and the decay time technique
Three calibrations were performed using the MFG calibration target
and the decay time phosphor measurement technique. The lumines­
cence decay time versus temperature, measured at both wavelengths, is
shown in Fig. 18 for the initial calibration. In a similar manner to the
intensity ratio results, a clear monotonic trend is observed, and a suit­
able calibration file can be determined. A cubic fit to each dataset gave
residuals within ±0.3 ◦ C, but in this case higher-order polynomial fits

Fig. 16. Intensity ratio with MFG – initial calibration: signal amplitude and
Intensity Ratio versus temperature. Lines show 2nd-order polynomial fits. The

relative temperature sensitivity is also shown.
9


A. Sposito et al.

Nuclear Engineering and Design 375 (2021) 111091

Fig. 17. Intensity ratio with MFG: a) standard deviation (SD) of calibration thermocouple σ (Ttc ) and phosphor thermometer σ(Tphos ), and b) repeatability of the
phosphor calibration.

using the 650 nm decay time measurements for thermometry with MFG,
as additionally the luminescence amplitude is greater and the standard
deviation lower. It must be stressed that the observed difference at 635
nm is <1 ◦ C, which is around or more than an order of magnitude lower
than the achievable uncertainty of currently deployed temperature
measurement techniques such as infra-red thermometry or contact
thermocouples used under similar conditions (Temperature Measure­
ment et al., 1991; End-to-end uncertainty assessment of temperature
measurement in SNM packages, J.V. Pearce and D. Tucker, NPL
Restricted Report ENG (RES) 014, April, 2020).
3.2.3. Instrument calibration with ruby and decay time technique
Three calibrations were performed using the ruby calibration target
and the decay time phosphor measurement technique. Fig. 20 shows the
amplitude of the luminescence signal at 690 nm and 694.3 nm versus
temperature from the initial calibration, where both signals are seen to
decrease with increasing temperature (recall that for MFG, one signal
increased, while the other decreased). Although not used for ther­
mometry here, the intensity ratio Φ = V(690nm)/V(694.3nm) is also
plotted in Fig. 20, showing that ruby cannot be used in this mode of

operation due to the non-monotonic nature of Φ (e.g. when Φ = 0.75,
the temperature inferred from the luminescence could be either 150 ◦ C
or 220 ◦ C – see black dashed line in the figure). Additionally, the relative
temperature sensitivity of ruby is approximately 7 times less than that of
MFG, i.e. the intensity ratio for ruby changes by 9% relative to its mean
value, compared to 62% for MFG over the same temperature range.
The ruby luminescence decay time versus temperature, measured at
both wavelengths, is shown in Fig. 21 for the initial calibration. The
decay time of ruby at 20 ◦ C is approximately 1.3 ms lower than MFG, but
has a similar sensitivity to temperature, changing by 55%, compared to
47% for MFG, over the same temperature range (20 ◦ C – 200 ◦ C). A clear
monotonic trend is observed, allowing for the generation of a calibration
file. A cubic fit generates residuals within ± 3 ◦ C, with higher-order
polynomial fits producing smaller residuals (±0.5 ◦ C for a 5th-order
fit). It is worth noting that, unlike MFG (see Fig. 18), the decay time
calibration curves of ruby become flatter at both temperature extremes,
i.e. the temperature sensitivity (dτ/dT) tends to zero, meaning that the
phosphor temperature will have large uncertainties in these regions.
Fig. 22 a) shows the equivalent temperature standard deviation for
the two wavelengths decay time and thermocouple temperature values.
While σ (Ttc ) is constant at approximately 0.01 ◦ C, both σ(T690 ) and
σ (T694.3 ) change with temperature, increasing at the extremes of the
temperature range, as expected from the shape of the calibration curves.
Generally, the standard deviation of the phosphor temperatures is an
order of magnitude higher than the thermocouple temperatures, but the
same order of magnitude as decay time measurements with MFG

Fig. 18. Decay time with MFG – initial calibration: Decay time measurements
and decay time difference versus temperature. Lines show 3rd-order polynomial
fits. The relative temperature sensitivity is also shown.

did not reduce the magnitude of the residuals. The decay times
measured at each wavelength were similar, with minor differences
decreasing with increasing temperature, from 17 µs to 6 µs for temper­
atures from 20 ◦ C to 200 ◦ C, also shown in Fig. 18.
Fig. 19a) shows the equivalent temperature standard deviation for
the two decay times and thermocouple measurements. While σ(Ttc ) is
constant at approximately 0.01 ◦ C, σ(T635 ) and σ (T650 ) tend to decrease
and increase respectively with temperature. Generally, the phosphor
measurement standard deviations were approximately one order of
magnitude larger than those for the thermocouple. The difference be­
tween the phosphor and thermocouple temperatures, at calibration
points, were within ± 0.1 ◦ C. Fig. 19 b) shows the repeatability of the
decay time calibrations. While no significant change was observed for
the measurements at 650 nm, a negative offset was seen for measure­
ments at 635 nm, increasing (up to 0.8 ◦ C) with repeated thermal
cycling. The authors do not have an explanation for the different
behavior at different wavelengths. Considering this, we recommend
10


A. Sposito et al.

Nuclear Engineering and Design 375 (2021) 111091

Fig. 19. Decay time with MFG: a) standard deviation of calibration thermocouple and phosphor thermometer, and b) repeatability of the phosphor calibration.

Fig. 20. Intensity ratio with ruby – initial calibration: signal amplitude and
intensity ratio versus temperature. Lines show 3rd-order polynomial fits.

described earlier. Fig. 22 b) shows the repeatability of the decay time
calibrations. Above 60 ◦ C, differences were of the order of 0.7 ◦ C, and
similar in magnitude to those of the MFG decay time measurements.
However, repeatability issues were observed at lower temperatures (T <
50 ◦ C), where a negative shift of 2 ◦ C – 3 ◦ C was observed. The reason for
this is not clear and needs further investigation.

Fig. 21. Decay time with ruby – initial calibration: decay time measurements
and decay time difference versus temperature. Lines show 3rd-order polynomial
fits. The relative temperature sensitivity is also shown.

3.2.4. Uncertainty budget
The uncertainty budget for the hybrid phosphor thermometer in each
instrument configuration and for each phosphor is given in Table 4,
where all components have been combined in quadrature to give the
standard uncertainty (i.e. at k = 1). Despite the repeatability issues, two
of the configurations with MFG have standard uncertainties of<0.5 ◦ C. It
should be noted that, if the drift is eliminated, uncertainties as low as
0.25 ◦ C are possible for MFG with the intensity ratio approach (shown in
the last row of the table).
Comparing the different techniques and associated uncertainties in
Table 4, two measurement configurations using the MFG phosphor are
recommended:

2. Decay time (650 nm) gave good results, with calibration residuals
and measurement standard deviations of 0.1 ◦ C and 0.3 ◦ C respec­
tively, and no calibration drift was observed.
4. Conclusions
We have investigated MFG and ruby thermographic phosphors,
mixed with six different binders and applied to SS316L substrates, for

their suitability for temperature monitoring of nuclear waste storage
containers and racks. Neither month-long low-dose or short-term highdose γ-ray exposures caused any significant degradation in phosphor
appearance or performance. NaOH soak tests, carried out at 40 ◦ C for
one month, identified binders that were not compatible with the solu­
tion. Of the remaining samples, no change in appearance or performance
for either phosphor was observed.
In terms of thermometry, the phosphor MFG was demonstrated to be
superior to ruby due to its lower measurement uncertainty for a given
excitation strength and better sensitivity at lower temperatures. The 529

1. Intensity ratio gave the best results in terms of calibration residuals
and standard deviation, although a small drift (<1 ◦ C) was observed.
It is believed that this problem could be overcome with thermal
annealing of the phosphor coating before use. However, the source of
drift may be as a result of changes in the instrument configuration or
calibration set-up between tests and needs further investigation.
11


A. Sposito et al.

Nuclear Engineering and Design 375 (2021) 111091

Fig. 22. Decay time with ruby: a) standard deviation of calibration thermocouple and phosphor thermometer, and b) repeatability of the phosphor calibration.

thank Martin Kelly and Michael Homer of the NPL medical radiation
group, for providing access to the γ-ray sources.
The phosphor thermometer used for the γ-ray and NaOH tests was
developed under the European Metrology Research Programme (EMRP)
Project 14IND04 EMPRESS. This project has received funding from the

EMPIR programme co-financed by the Participating States and from the
European Union’s Horizon 2020 research and innovation programme.

Table 4
Uncertainty budget for fibre-optic hybrid phosphor thermometer.IR = intensity
ratio, DT = decay time, Rect. Dist. = rectangular distribution.
U(Phosphor/technique) (k = 1) [◦ C]
Component
Calibration Residuals
Measurement Type-A
Repeatability [Rect.
Dist.]
Repeatability [1σ]
Intensity dependence
Thermocouple
calibration
Temperature gradients
Total (quadraturesum)
Total (if drift is
removed)

MFG/
IR

MFG/DT (635
nm)

MFG/DT (650
nm)

Ruby/
DT

0.00
0.10
0.60

0.10
0.50
0.80

0.10
0.30
0.10

0.10
0.50
3.00

0.35
0.20
0.10

0.46
0.20
0.10

0.06
0.20
0.10

1.73
0.40
0.10

0.05
0.43

0.05
0.72

0.05
0.40

0.05
1.85

0.25

0.56

0.39

0.66

1

2

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.
org/10.1016/j.nucengdes.2021.111091.
References
Ald´en, M., et al., 2011. Thermographic phosphors for thermometry: A survey of
combustion applications. Prog. Energy Combust. Sci. 37 (4), 422–461.
Aremco Products Inc.; />Catalog_15.pdf. (accessed March 31, 2020).
Atakan, B., Eckert, C., Pflitsch, C., 2009. Light emitting diode excitation of Cr3+:Al2O3 as
thermographic phosphor: experiments and measurement strategy. Meas. Sci.
Technol. 20 (7), 075304. />Brübach, J., Feist, J.P., Dreizler, A., 2008. Characterization of manganese-activated
magnesium fluorogermanate with regards to thermographic phosphor thermometry.
Meas. Sci. Technol. 19 (2), 025602. />025602.
Brübach, J., Pflitsch, C., Dreizler, A., Atakan, B., 2013. On surface temperature
measurements with thermographic phosphors: A review. Progress Energy Combust.
Sci. 39, 37–60. ISSN 0360–1285.
Khalid, A.H., Kontis, K., 2008. Thermographic phosphors for high temperature
measurements: principles, current state of the art and recent applications. Sensors 8,
5673–5744. />Kumari, S., Khare, A., 2011. Epitaxial ruby thin film based photonic sensor for
temperature measurement. Rev. Sci. Instrum. 82, 066106 />1.3606443.
McMillan, J.L., Greenen, A., Bond, W., Hayes, M., Simpson, R., Sutton, G., Machin, G.,
Jowsey, J., Adamska, A., 2019. Thermometry of intermediate level waste containers
using phosphor thermometry and thermal imaging. Measurement 132, 207–212.
/>Michalski, L., Eckersdorf, K., McGhee, J., 1991. Temperature Measurement. John Wiley
& Sons. ISBN 0-471-92229-3.
Pearce, J.V., Tucker, D., 2020. End-to-end uncertainty assessment of temperature
measurement in SNM packages. NPL Restricted Report ENG (RES) 014.
Sutton, G., et al., 2019. Imaging phosphor thermometry from T = 20 ◦ C to 450 ◦ C using
the time-domain intensity ratio technique. Meas. Sci. Technol. 30, 044002 https://
doi.org/10.1088/1361-6501/ab04ea.
Thorington, Luke, 1950. Temperature dependence of the emission of an improved
manganese-activated magnesium germanate phosphor. J. Opt. Soc. Am. 40 (9), 579.

/>Wang, G.G., et al., 2009. Radiation resistance of synthetic sapphire crystal irradiated by
low-energy neutron flux. Cryst. Res. Technol. 44 (9), 995–1000. />10.1002/crat.v44:910.1002/crat.200900243.

and 880 silicone-based binders were identified as the most suitable due
to their resilience to NaOH solution.
We also developed a fibre-coupled phosphor thermometer and
determined that both intensity ratio and decay time temperature mea­
surement modes with low uncertainty (<0.5 ◦ C) were possible for the
samples tested here. The intensity ratio technique for MFG produced the
lowest measurement standard deviation (<0.1 ◦ C) but showed a small
drift with repeated thermal cycling. The decay time technique with MFG
measuring at 650 nm did not show any drift but had a larger measure­
ment standard deviation (0.3 ◦ C). Further investigation of the drift
behaviour is planned. Of the phosphor/binder combinations studied
here MFG/529, using either measurement technique, is the most suit­
able for the nuclear waste storage environment, for storage in air or
ponds.
Declaration of Competing Interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.
Acknowledgments
This work was funded by Sellafield Ltd. The authors would like to

12