Uncertainty in mechanical deformation of a Fabry-Perot cavity due to pressure: towards best mechanical configuration


ACTA IMEKO 
ISSN: 2221-870X 
September 2022, Volume 11, Number 3, 1 - 8 

 

ACTA IMEKO | www.imeko.org September 2022 | Volume 11 | Number 3 | 1 

Uncertainty in mechanical deformation of a Fabry-Perot 
cavity due to pressure: towards best mechanical 
configuration 

S. Moltó1, M. A. Sáenz-Nuño2, E. Bernabeu3, M. N. Medina1 

1 Centro Español de Metrología, Calle Alfar 2, 28760 Tres Cantos, España 
2 Instituto de Investigación Tecnológica, Escuela Técnica Superior de Ingeniería-ICAI, Universidad Pontificia Comillas, 28015 Madrid, España 
3 Universidad Complutense de Madrid, España 

 

 

Section: RESEARCH PAPER  

Keywords: Fabry-Perot; pressure measurement; refractrometry; quantum pascal  

Citation: S. Moltó, M. A. Sáenz-Nuño, E. Bernabeu, M. N. Medina, Uncertainty in mechanical deformation of a Fabry-Perot cavity due to pressure: towards 
best mechanical configuration, Acta IMEKO, vol. 11, no. 3, article 15, September 2022, identifier: IMEKO-ACTA-11 (2022)-03-15 

Section Editor: Francesco Lamonaca, University of Calabria, Italy 

Received October 27, 2021; In final form August 30, 2022; Published September 2022 

Copyright: This is an open-access article distributed under the terms of the Creative Commons Attribution 3.0 License, which permits unrestricted use, 
distribution, and reproduction in any medium, provided the original author and source are credited. 

Corresponding author: S. Moltó, e-mail: smolto@cem.es  

 

1. INTRODUCTION 

The 18SIB04 QuantumPascal EMPIR project deals with the 
development of a quantum-based pressure standard in order to 
improve the uncertainty and traceability that the current methods 
offer. This project is based on the application of a Fabry-Perot 
interferometer to measure in a pressure range of 1 Pa to 100 kPa. 
However, some parameters must be carefully handled, such as 
the deformation, the temperature and the gas properties.  

The cavity deformation due to the gas pressure was simulated 
using a finite element method (FEM) software. Therefore, a 
study of the uncertainty obtained in these simulations and its 
propagation in the evaluation of pressure was needed.  

A previous study of the simulated cavity deformation due to 
pressure was carried out in [1], and its results were used herein 
to show the difference obtained due to the use of different 
software. Finally, a new cylindrical design of the cavity was 
proposed and compared to the design proposed in [1]. 

2. THEORY 

The relative deformation in the length with the gas pressure 
of the Fabry-Perot cavity can be described with a linear 

dependency as in equation (1). Where 𝐿0 is the without 
deformation cavity length, Δ𝐿 is the change in cavity length (the 
deformation), 𝑃 is the pressure and 𝜅 is the pressure-normalized 
deformation. 

Δ𝐿

𝐿0
= 𝜅𝑃 . (1) 

𝐿0 is calculated as the distance between the centre of the 
reflective face of the mirrors, and Δ𝐿 as the change in this 
distance. 

Table 1 summarizes the data used in this work, so is it possible 
to calculate the pressure with equations below with the values of 
the table. 

ABSTRACT 
In this paper, a study of the deformation of a refractometer use to achieve a quantum realization of the Pascal is made. First, the 
propagation of the uncertainty in the measure of pressure due to mechanical deformation was made. Then deformation simulations 
were made with a cavity designed by the CNAM and whose results are reported in the 18SIB04 Quantumpascal EMPIR project. This step 
aims to corroborate the methodology used in the simulations. The pressure-normalized relative deformation of this design obtained in 
18SIB04 is (-6.390 ± 0.015) × 10-12 Pa-1, the result obtained with our method is (-6.384 ± 0.032) × 10-12 Pa-1, which differs 0.001 times 
from the value obtained in 18SIB04. Finally, a cylindrical cavity design is presented and simulated obtaining a pressure-normalized 
relative deformation of (-5.758584 ± 0.00000047) × 10-12 Pa-1, which deforms 0.1 times less.  

mailto:smolto@cem.es


 

ACTA IMEKO | www.imeko.org September 2022 | Volume 11 | Number 3 | 2 

The gas parameters can be calculated from the refractivity, 

whose change is related to the change of the laser frequency Δ𝜈, 
the change in the number of modes (Δ𝑞) and the deformation 
coefficient (𝜖) using equation (2), [3]. This equation can be used 
with absolute pressures up to 10 kPa. 

Δ(𝑛 − 1) = (𝑛 − 1) =

Δ𝑞
𝑞0
+ (1 + 2 𝜉 + 𝜂)

Δ𝜈
𝜈0

1 + 𝜖 + (1 + 2 𝜉 + 𝜂)
Δ𝜈
𝜈0

 . (2) 

The deformation coefficient (𝜖) [3] depends on the pressure-
normalized deformation (𝜅) as it is shown in equation (3). The 
change in number of modes is an experimental measurement but 

it can be simulated using equation (4), where 𝜆0 is the wavelength 
in vacuum and 𝜆 is the wavelength affected by the change in 
refractive index (𝜆 = 𝜆0/𝑛). Also, Δ𝜈 is the laser change in 
frequency which maximum value is 𝜈𝐹𝑆𝑅 when mode jumping is 
used. 

ϵ = 
𝜅 𝑃

(𝑛 − 1)
= 𝜅

2 𝑘B 𝑇 𝑁A
3 𝐴r

 (3) 

Δ𝑞 = 𝑞 −𝑞0 =
𝐿

𝜆
2

−
𝐿0
𝜆0
2

= 2𝐿0 [
𝜆0 + 𝜆0 𝜅 𝑃 −𝜆

𝜆 𝜆0
] . (4) 

The molar density is related with the refractivity by the 
equation (5) [3] and the molar density is related with the pressure 

by the equation (6) [3]. Where 𝐵𝜌 is the first virial coefficient and 

𝐵𝑃(𝑇) is the first pressure virial coefficient which is calculated 
with [−4.571 +  0.1974(𝑇(𝐾)− 300)− 5.137 ×
10−4(𝑇(𝐾)− 300)2] ×10−6 m3/mol. 

𝜌 =
2

3 𝐴r
(𝑛 − 1)[1+ 𝐵�̃�(𝑛 − 1)] (5) 

𝑃 = 𝑘B 𝑇 𝑁A 𝜌[1 + 𝐵𝑃(𝑇)𝜌] . (6) 

3. UNCERTAINTY PROPAGATION 

In order to calculate the uncertainty of pressure due to the 

uncertainty in 𝜅 is necessary to obtain an expression of pressure 
depending on 𝜅 directly. From equations (5), (6) and the 
expression of 𝐵𝑃(𝑇) is possible to calculate that relation 
(equation (7)), where the value of 𝐶0 is shown in equation (8) 

𝑃(Δ𝜈,Δ𝑞,𝜅,𝑇) = 

𝐶0 𝑇 Δ𝑞

(𝐶0 𝑇 𝜅 + 1)𝑞0
(

𝐵𝜌 Δ𝑞

(C0 T κ+ 1)𝑞0
+1) 

×(1 +
2 Δ𝑞(

𝐵𝜌 Δ𝑞

(𝐶0 𝑇 𝜅 + 1)𝑞0
+1)

3 𝐴r(𝐶0 𝑇𝜅 + 1)𝑞0
𝐵𝑃(𝑇)) 

(7) 

𝐶0 =
2 𝑁A 𝑘B
3 𝐴𝑟

 . (8) 

Considering that Δ𝑞 =
2𝐿0(1+𝜅𝑃Nominal)

𝜆/2
− 𝑞0 and Δ𝜈 = 0 an 

expression of pressure as 𝑃 = 𝑃(𝜅,𝑃Nominal,𝑇) is obtained 
(Equation (9)) 

𝑃(𝜅,𝑃Nominal,𝑇) = 

𝐶0𝑇(
2(𝐿0 𝑃Nominal 𝜅 + 𝐿0)

𝜆
− 𝑞0)

(𝐶0 𝑇 𝜅 + 1)𝑞0
× 

(
𝐵𝜌 (

2(𝐿0 𝑃Nominal 𝜅 +𝐿0)
𝜆

− 𝑞0)

(𝐶0 𝑇 𝜅 + 1)𝑞0
+ 1) × 

(9) 

Table 1. Coefficients and constants used in this work 

Parameter Designation Value 

𝐿0 Cavity length in vacuum 0.1 m 

𝑛 Refractive index 𝑛 ≥ 1 

𝐴r Molecular polarizability 4.44613930 × 10
-6 m3/mol [2] 

𝐵r First refractive virial coefficient 0.812 × 10
-12 m6/mol2 [2] 

𝐵�̃� First density virial coefficient −[1+4(𝐵𝑟/𝐴𝑟
2)]/6 [2] 

𝐵𝑃(𝑇) First pressure virial coefficient [−4.571+0.1974(𝑇(𝐾)−300)−5.137×10
−4(𝑇(𝐾)−300)2]×10−6 m3/mol [2] 

𝜉 Mirror dispersion coefficient 11 × 10-6 [2] 

𝜂 Gas dispersion coefficient 1 × 10-6 [2] 

𝜆0 Wavelength in vacuum 632.8 nm 

𝜆 Wavelength affected by pressure 𝜆 = 𝜆0/𝑛 

𝜈0 Frequency in vacuum 𝑐/𝜆0 

𝑞0 Number of length modes in vacuum 2𝐿0/𝜆0 

Δ𝜈 Laser change in frequency Assumed as 0 

Δ𝑞 Change in number of modes Estimate in Equation (4) 

𝜖 Deformation coefficient Equation (3) [3] 

𝑇 Temperature 273 K 

𝑃 Nominal Pressure From 1Pa to 10 kPa 

𝑁𝐴 Avogadro number Used value of [3] 

𝑘𝐵 Boltzmann constant Used value of [3] 



 

ACTA IMEKO | www.imeko.org September 2022 | Volume 11 | Number 3 | 3 

(

 
 
 
 
2𝐵𝑃 (

2(𝐿0 𝑃Nominal 𝜅 +𝐿0)
𝜆

+ 𝑞0)

3 𝐴r(𝐶0 𝑇 𝜅 +1)𝑞0
×

(

2(𝐿0 𝑃Nominal 𝜅 +𝐿0)
𝜆

− 𝑞0

(𝐶0 𝑇 𝜅 + 1)𝑞0
+1) + 1

)

 
 
 
 

 . 

In this work, it is assumed that the only source of uncertainty 

is the pressure-normalized deformation (𝜅), as the aim is to 
analyse how this particular magnitude affects to the measurement 
of pressure. From equation (7) and applying the rule of the 
propagation of uncertainty the equation (10) is obtained, which 

estimates the relative uncertainty in pressure with respect 𝜅. 
Where 𝐶𝜅 is the sensitivity coefficient, which is calculated with 
equation (11). As the pressure of equation (9), depends on the 
nominal pressure, the relative uncertainty depends on that 
parameter 

𝑤(𝑃) =
𝑢(𝑃)

𝑃
≈ |𝐶𝜅

𝑢(𝜅)

𝜅
| (10) 

𝐶𝜅 =
𝜕𝑃(𝜅,𝑃Nomina,𝑇)

𝜕𝜅
 . (11) 

4. CNAM CAVITY DESCRIPTION 

The cavity modelled for the simulation was built by the 
CNAM ([5]). In Figure 1, it is shown a 3D model of the Fabry-
Perot cavity. The spacer is made of Zerodur and the mirrors are 
made of fused silica. The dimensions of the Zerodur spacer are 
specified in Figure 2. The mirrors were considered to be flat.  
The properties of the materials used are collected in Table 2. In 
order to improve the computational time, only an octave of the 
cavity was simulated using the symmetry restrictions available in 
the software. This selection is represented in Figure 3 where the 
simulated section of the Fabry-Perot cavity is dyed with red.  

With the purpose of simulating the cavity inside a vacuum 
chamber, the same pressure load was introduced to every face of 
the Fabry-Perot cavity (same inner and outer pressure). The only 

restriction applied to the model was that the spacer and the 
mirrors would not rotate. 

5. SIMULATION RESULTS 

The Solid Edge 2021 results are shown with a colormap of 
the deformation in mm in Figure 4. It was shown an 
inhomogeneous deformation, but it was corrected using the 
symmetries of the design. Simulating a part of the model 
incorporates these symmetries without including extra 
conditions to the model. 

5.1. Cavity Relative Deformation Dependency with the Number of 
Mesh Elements. 

The relative deformation of the cavity calculated in the 
simulation depends on the number of mesh elements used, as it 

is shown in Figure 5. The convergence of 𝜅 with the number of 
mesh elements was not achieved so two curve fits were made. 

One to the function 𝑦1(𝑥) = 𝑎 𝑥
−4 + 𝑏 𝑥−2 + 𝑐 𝑥−1 + 𝑑 and 

another to 𝑦2(𝑥) = 𝑎 (log(𝑏 𝑥))
−1 + 𝑑, where 𝑥 is the number 

of mesh elements and 𝑦𝑖 is the value of 𝜅. These functions where 
selected to be constant when the number of mesh elements from 

Table 2. Materials properties used in the simulations. 

Parameter Spacer material 
(Zerodur) 

Mirror material 
(fused silica) 

Young modulus (GPa) 90.3 73 
Poisson ratio 0.24 0.155 
Density (kg/m3) 2530 2195 

 
 

 

Figure 1. FP cavity model. 

 

Figure 2. Dimensions of the Zerodur spacer in mm, all drillings diameters are 
10 mm. 

 

 

Figure 3. CNAM FP cavity model with the simulated section dye in red. 



 

ACTA IMEKO | www.imeko.org September 2022 | Volume 11 | Number 3 | 4 

among group of functions that tends to infinite as it is shown in 
equation (12): 

lim
𝑥→∞

(𝑎 𝑥−4 + 𝑏 𝑥−2 + 𝑐 𝑥−1 + 𝑑) = 𝑑 =  κ 

lim
𝑥→∞

(
𝑎

log (𝑏 𝑥)
+ 𝑑) = 𝑑 =  κ . 

(12) 

The values obtained in the fitting are collected in Table 3. 

Using equation (12) it is obtained the value of 𝜅 for each curve 
fit. The value for 𝑦1 is 𝜅1 = (-6.38378 × 10

-12 ± 3.2 10-16) Pa-1 

and for 𝑦2 is 𝜅2 = (-6.3868 × 10
-12 ± 3.2 10-15) Pa-1. The 

uncertainty is calculated with the variance and covariance matrix 
obtained with function curve_fit() from the package optimize of 

Scipy [3], that uses the equation cov(𝑥,𝑦) = 𝐶𝑥𝑦 =
1

𝑁
∑ (𝑥𝑖 −𝑖

�̅�)(𝑦𝑖 −�̅�) to estimate it. It is shown that the first curve gives 
less uncertainty than the second. 

With the relative deformation of the cavity over the pressure 
obtained in Table 3, the values of Table 1 and the equation (9) is 
possible to calculate the pressure that the Fabry-Perot cavity will 
have for a given nominal pressure, obtaining in Figure 6 the 
difference between the calculated and nominal values. It is shown 

that near 0 Pa there is an asymptotic behaviour, but if the 
nominal pressure increases a linear dependency is presented. 

As it was said before, the relative uncertainty of pressure 
depends on the nominal pressure (equation (10) and (11)). Figure 
7 represents the evolution of pressure uncertainty with the 
nominal pressure. It is shown a linear dependency and the 

uncertainty obtained by using 𝜅2 increases faster than the one 
obtained with 𝜅1. Also, the relative uncertainty in pressure 
calculated in a range of pressure under 10 kPa is under 9 × 10-12. 

5.2. Comparison with Results from [1]  

A comparison between the results of simulations made with 
different software and institution are needed to verify that the 
procedure of simulation of every participant in the 18SIB04 
QuantumPascal EMPIR project is correct. 

In [1] the results of simulating the deformation of the Fabry-
Perot cavity are shown, and they are collected in Table 4. The 
result simulated in Solid Edge software by the CEM is less than 
10-3 time bigger than the average of [1]. The uncertainty in CEM’s 
simulation was calculated as a rectangular distribution centred in 
-6.3846 × 10-12 Pa-1 and with a width equal to two times the 

 

Figure 4. Cavity deformation for a pressure of 80 kPa.  

 

Figure 5. Pressure-normalized deformation versus number of mesh 
elements. The data was fitted to two curves 

 

Figure 6. Relative difference between calculated pressure and nominal 
pressure ((PNominal-P)/PNominal), where 𝜿𝟏 is the value of pressure-normalized 
deformation calculated with 𝒚𝟏and 𝜿𝟐 is the value of pressure-normalized 
deformation calculated with 𝒚𝟐. 

 

Figure 7. Relative uncertainty of calculated pressure with 1 = -6.38378 Pa-1 

and 2 = 6.3868 Pa-1 versus the nominal pressure. 

Table 3. Values of the mean square error (MSE) of the fit, the mean absolute 

error (MAE) of the fit and  in Pa-1 for each fit. 

𝑦1(𝑥) = 𝑎𝑥
−4 +𝑏𝑥−2 +𝑐𝑥−1 +𝑑 

MSE 
1.41 × 10-23 

MAE 
2.83 × 10-12 

 

𝜅 
-6.38378 × 10-12 

u(𝜅) 
3.2 × 10-16 

u(𝜅)/𝜅 
4.9 × 10-5 

𝑦2(𝑥) = 𝑎(log(𝑏𝑥))
−1 +𝑑 

MSE 
4.78 × 10-31 

MAE 
4.85 × 10-16 

 

𝜅 
6.3868 × 10-12 

u(𝜅) 
1.7 × 10-15 

u(𝜅)/ 𝜅 
2.7 × 10-4 



 

ACTA IMEKO | www.imeko.org September 2022 | Volume 11 | Number 3 | 5 

difference of CEM’s value and the average value of [1] 
(2.24 × 10-14 Pa-1). 

Using the values of Table 4, Table 1 and equation (7), the 

value of pressure for the average value 𝜅 and the value calculated 
with Solid Edge are obtained. In Figure 8 the relative difference 
of each pressure with respect the nominal pressure is 
represented. It is shown the same behaviour as in Figure 6. Also, 
the values are so similar that the effect on the pressure cannot be 
distinguished with this figure for pressures over 10 Pa; hence, in 
Figure 9, the relative difference of the pressure calculated with 

the 𝜅 obtained by [1] and the pressure calculated with 𝜅 obtained 
with Solid Edge is represented. 

In Figure 9, the same asymptotic behaviour is shown. The 
maximum relative difference obtained is near 0 Pa, where the 
asymptote is located, is around 3.5 × 10-8. After the elbow value 
of the function the relative difference in pressure is less than 2 × 
10-10. 

As the relative uncertainty of pressure depends on the nominal 
pressure (Equation (10) and (11)), Figure 10 depicts the evolution 
of pressure uncertainty with the nominal pressure. It is shown a 
lineal dependency and the uncertainty of the pressure obtained 
with the CEM’s value increase faster. Also, the relative 
uncertainty calculated for the range of pressure where equation 
(2) is acceptable is under 3 × 10-10. 

6. CEM’S CYLINDRICAL CAVITY 

With the purpose of solving the symmetry rupture that 
supposes using the spacer as an inner vacuum chamber and gases 
chamber, a new cylindrical design was proposed. 

6.1. Cavity Description 

The designed cavity is represented in Figure 11, where the 
simulated section is coloured in red. Although this time the 
symmetry is only a quarter of the model and not an octave as 
before, an octave was taken after checking that the results were 
the same as the full model. This way, a similar number of mesh 
elements are analysed for both models. The spacer and mirrors 
of this cavity are made of Schott’s Zerodur Class 0, whose 
properties are shown in Table 2. The Zerodur Class 0 mirrors 
provide a reflectance over 97 % for a wavelength of 633 nm, 
which is the wavelength that will be used in the pattern. 

The dimensions of the Zerodur spacer are specified in Figure 
12. It is shown that the geometry of this spacer is simpler than 
the presented in Figure 2, this will deal with faster calculations. 

The inner diameter of the Zerodur spacer was selected to be 
20 mm, so as to be compatible with the optic requirements of 
numeric aperture of 0.2. Also, this diameter provides a more 

Table 4. Summary of the simulated pressure normalized cavity deformation 
using the finest meshing by each partner and software. 

Partner - Software 
Used 

Pressure-normalized 
deformation (𝜿) 

𝚫𝑳/𝑳/𝑷 in 10-12 Pa-1 

Number of mesh 
elements 

A-Comsol [1] -6.3900 1 735 324 

B-Comsol [1] -6.3899 190 000 

B-Ansys [1] -6.3908 2 948 108 

C-Ansys [1] -6.3900 21 061 

D-Comsol [1] -6.3902 3 168 777 

Average value [1] -6.3902 -- 

Standard error of the 
mean (𝜎𝑥) [1] 

0.00015 -- 

CEM-Solid Edge -6.3846 2197477 

Uncertainty 0.0032 -- 

 

Figure 8. Relative difference between calculated pressure and nominal 
pressure ((PNominal-P)/PNominal). 

 

Figure 9. Relative difference between calculated pressure and nominal 
pressure ((P-PCEM)/P), where P is the nominal pressure and PCEM is the 
calculated pressure. In the first figure the uncertainty is not appreciable, but 
in the enlargement below the uncertainty can be seen. 

 

Figure 10. Relative uncertainty of calculated pressure with  obtained by [1] 

and  calculated by the CEM with Solid Edge software versus the nominal 
pressure. 𝑹𝟐 is the linear correlation coefficient. 

 

Figure 11. CEM FP cavity model with the simulated section dyed in red. 



 

ACTA IMEKO | www.imeko.org September 2022 | Volume 11 | Number 3 | 6 

laminar regimen in rapid pressure variations than with smaller 
diameters. The Zerodur spacer thickness was designed to be 9 
mm, but the mirror thickness was 11 mm, which corresponds to 

a relative factor 
𝑀𝑖𝑟𝑟𝑜𝑟 𝑇ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠

𝑆𝑝𝑎𝑐𝑒𝑟 𝑇ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠
≈ 1.22, similar to the Bessel’s 

function first zero divided by π (
𝐽1(𝑧)

𝜋
= 1.216). The Bessel 

function is associate to the circular revolution of 2 π of the 
cylindrical symmetry. 

As made with the previous cavity, in order to simulate that the 
cavity is into a vacuum chamber, the same pressure load was 
introduced to every face of the Fabry-Perot cavity. The only 
restriction applied to the model was that the spacer and the 
mirrors would not rotate. 

6.2. Results of the simulations 

Figure 13 presents the results of the deformation simulation 
for a pressure of 80 kPa. As shown in Figure 4, there was an 
inhomogeneous deformation, but it is corrected by using the 
partial simulation of the model as the partial only simulates one 
mirror assuming the symmetry. 

The simulation data were fitted to three different equations. 
There were selected to be constant when the number of mesh 
elements tends to infinite. The selected curves are shown in 

equation (13), the quadratic equation was not added in equation 
(12) because the software could not fit the data to that curve. The 

results of fitting are shown in Figure 14, where the curves 𝑦1(𝑥) 
and 𝑦2(𝑥) cannot be distinguish. 

𝑦1(𝑥) = 𝑎 𝑥
−4 + 𝑏 𝑥−2 + 𝑐 𝑥−1 + 𝑑 

𝑦2(𝑥) = 𝑎 𝑥
−2 + 𝑏 𝑥−1 + 𝑑 

𝑦3(𝑥) = 𝑎 (log(𝑏 𝑥))
−1 + 𝑑 . 

(13) 

The parameters obtained in the fitting are shown in Table 5. 

The curve 𝑦1(𝑥) shows the biggest mean square error and mean 
absolute error o of the fitting. The relative uncertainties of curves 

𝑦2(𝑥) and 𝑦3(𝑥) are quite similar, but the MSE and the MAE 
obtained by fitting to the curve 𝑦3(𝑥) are several orders of 
magnitude smaller. Also, the best relative uncertainty obtained is 
not reached with the fitting, but it is achieve using the result of 
the finest mesh, the simulation with the highest number of mesh 
elements. 

Using the relative deformation of the cavity over pressure 

obtained in Table 5 for 𝑦2(𝑥) and the finest mesh, the values of 
Table 1 and the Equation (9) is possible to calculate the pressure 
that the Fabry-Perot cavity will have for a given nominal 
pressure. The difference between the calculated and nominal 

 

Figure 12. Dimensions of the Zerodur spacer in mm. 

 

Figure 13. Cylindrical cavity deformation for a pressure of 80 kPa. 

 

Figure 14. Pressure-normalized deformation versus the number of mesh 
elements. The data were fitted to three equations. 

Table 5. Values of the mean square error (MSE) of the fit, the mean absolute 

error (MAE) of the fit and  in Pa-1 for each fit and the value of  for the finest 
mesh. 

𝑦1(𝑥) = 𝑎 𝑥
−4 +𝑏 𝑥−2 +𝑐 𝑥−1 +𝑑 

MSE 
7.36 × 10-18  

MAE 
1.97 × 10-9 

 

𝜅  
-5.758610 × 10-12 

u(𝜅) 
1.9 

u(𝜅)/ 𝜅 
3.3 × 10-6 

𝑦2(𝑥) = 𝑎 𝑥
−2 +𝑏 𝑥−1 +𝑑 

MSE 
9.17 × 10-30  

MAE 
2.51 × 10-15 

 

𝜅  
-5.758573 × 10-12 

u(𝜅) 
1.1 × 10-17 

u(𝜅)/ 𝜅 
1.9 × 10-6 

𝑦3(𝑥) = 𝑎 (log(𝑏 𝑥))
−1 +𝑑 

MSE 
3.19 × 10-34  

MAE 
1.24 × 10-17 

 

𝜅  
-5.758600 × 10-12 

u(𝜅) 
1.5 × 10-17 

u(𝜅)/ 𝜅 
2.6 × 10-6 

𝜅 of the finest mesh (2602131elements) 

𝜅 
-5.7585840x10-12 

u(𝜅) 
4.7 × 10-18 

u(𝜅)/𝜅 
8.2 × 10-7 



 

ACTA IMEKO | www.imeko.org September 2022 | Volume 11 | Number 3 | 7 

values are shown in Figure 15. The asymptotic behaviour is 
shown again, and the difference of the relative difference 
obtained with these values is so small that over pressures of 
100 Pa cannot be distinguish (under 7 × 10-10). The values of 
relative difference between the nominal pressure and the 
calculated values of pressure are similar to the obtained in Figure 
6. 

As it was discussed before the relative uncertainty of pressure 
depends on the nominal pressure (Equation (10) and (11)). 
Figure 16 represents the evolution of pressure uncertainty with 
the nominal pressure. A linear dependency is shown again and 
the increase of the uncertainty using the value obtained with the 

curve 𝑦2(𝑥) is bigger than the obtained with the finest mesh. 
Furthermore, this simulation gives the best relative uncertainty, 
under 6.5 × 10-14 for both cases. 

Table 6 shows a comparison between the best values of 
pressure-normalized deformation obtained for each model, 
including the results obtained by the partners. It is shown that 
the cylindrical model presents less pressure-normalized 
deformation and uncertainty.  

7. FINAL DISCUSSION 

The results of the simulations of the Fabry-Perot cavity 
deformation proposed by CNAM depends on the number of 
mesh elements introduced in the FEM software to simulate it. 
Also, different software can have different results. Because of 
this issue, a study of the evolution of the relative deformation 
over pressure of the Fabry-Perot cavity was made, fitting the 
curve to two equations in order to achieve the value when the 
number of mesh elements tends to infinity obtaining a relative 

error in 𝜅 between 5 × 10-5 and 3 × 10-4 depending on the curve 
used.  

Moreover, the propagation of uncertainty of 𝜅 to the pressure 
was studied for the values obtained with this method, showing 

that the formula which fits the best is 𝑦1(𝑥) =
𝑎

𝑥4
+

𝑏

𝑥2
+
𝑐

𝑥
+ 𝑑 

whose parameters are collected in Table 5.It was shown, that the 
relative difference with the nominal value of pressure is under 
5 × 10-7 times the value of the nominal pressure and an 
asymptotic behaviour was shown in Figure 6. 

Furthermore, the difference between values of pressure-
normalized deformation obtained by other laboratories were 
compared with the values calculated using Solid Edge software. 
We conclude that the relative difference between the nominal 
pressure and the calculated pressure is under 5 × 10-8 times the 
value of the nominal pressure for 10 kPa. Also, the difference 
between the values obtained by the CEM and the pressure 

calculated using the 𝜅 of [1] is under 4 × 10-8 times the value of 
the pressure calculated with the 𝜅 of [1]. For Figure 8 and Figure 
9, the asymptotic behaviour shown before was followed. 

Moreover, the pressure relative uncertainty calculated for 
both cases is under 3 × 10-10 times the nominal pressure, which 

means that the contribution of the uncertainty of pressure-

normalized deformation (𝜅), for a range of pressure between 
1 Pa and 10 kPa, to the relative pressure uncertainty is not 
decisive for the measure. 

Finally, the CEM’s Fabry-Perot cavity provides smaller 
pressure-normalized deformation (-5.7548684 × 10-12 Pa-1) with 
an uncertainty of 4.7 × 10-18 Pa-1, which is 0.1 times smaller than 
the other cavity. The Fabry-Pérot cylindrical cavity provides a 
relative uncertainty under 6.5 × 10-14 for 10 kPa, one order of 
magnitude less than the best results using the other geometry. It 
would be interesting making a study of the deformation of the 
mirrors for different Fabry-Pérot cavity configurations.  

8. CONCLUSION 

In conclusion, CEM’s results using finite elements method 
through the software Solid Edge are similar to the results 
presented in [1] (with a discrepancy of 0.001 times the values 
obtained by [1]).  

Also, the CEM’s cylindrical has a pressure-normalized relative 
deformation 0.1 times lower than the model analysed in [1]. 
Moreover, the model presents more advantages as it only uses 
one material and, with the selected dimensions, there is a more 
laminar regimen in rapid pressure variations. 

In future projects the uncertainties of the constants and values 
of Table 1 will be introduced in order to have a value of the total 
uncertainty of the measurement. Also, due to the capability of 
the cylindrical design to have an inner pressure different to the 
outer pressure, an analysis of the FP cavity deformation where 
vacuum is applied inside the cavity and atmospheric pressure is 
applied outside the cavity will be studied. 

Table 6. Comparison between the results of the first model analysed by the 
partners (Model 1 partner) and analysed by CEM (Model 1 CEM) and the 
cylindrical model (Model 2). 

Model 
Pressure-normalized 

deformation (𝜅) 
(Δ𝐿/𝐿)/𝑃 in 10-12 Pa-1 

Uncertainty 
in 10-12 Pa-1 

Model 1 partner -6.3902 0.00015 

Model 1 CEM -6.3842 0.032 

Model 2 -5.758584 0.00000047 

 

Figure 15. Relative difference between calculated pressure and nominal 
pressure (PNominal - P)/PNominal.  

 

Figure 16. Relative uncertainty of calculated pressure with 1 and 2 versus 
the nominal pressure.  



 

ACTA IMEKO | www.imeko.org September 2022 | Volume 11 | Number 3 | 8 

REFERENCES 

[1] J. Zakrisson, I. Silander, C. Forssén, Z. Silvestri, D. Mari, S. 
Pasqualin, A. Kussicke, P. Asbahr, T. Rubin, O. Axner, Simulation 
of pressure-induced cavity deformation - the 18SIB04 
Quantumpascal EMPIR project, ACTA IMEKO 9(5) (2020), pp. 
281-286. 
DOI: 10.21014/acta_imeko.v9i5.985 

[2] I. Silander, T. Hausmaninger, M. Zelan, O. Axner, Gas modulation 
refractometry for high-precision assessment of pressure under 
non-temperature-stabilized conditions, Journal of Vacuum Science 
& Technology 36(3) (2018), art. 03E105, pp. 1-8. 
DOI: 10.1116/1.5022244 

[3] O. Axner, I. Silander, T. Hausmaninger, M. Zelan, Drift-free 
Fabry-Perot-cavity-based optical refractometry-Accurate 

expressions for assessments of gas refractivity and density, Jan. 
2018. 
DOI: 10.48550/ARXIV.1704.01187 

[4] Pauli Virtanen (+ 34 authors), SciPy 1.0: Fundamental algorithms 
for scientific computing in Python, Nature Methods 17(3) (2020), 
pp. 261-272. 
DOI: 10.1038/s41592-019-0686-2 

[5] Z. Silvestri, F. Boineau, P. Otal, J. Wallerand, Helium-Based 
refractometry for pressure measurements in the range 1-100 kPa, 
in 2018 Conference on Precision Electromagnetic Measurements 
(CPEM 2018), July 08-13, 2018, Paris, France.  
DOI: 10.1109/CPEM.2018.8501259 

 

 

http://dx.doi.org/10.21014/acta_imeko.v9i5.985
https://doi.org/10.1116/1.5022244
https://doi.org/10.48550/arXiv.1704.01187
https://doi.org/10.1038/s41592-019-0686-2
https://doi.org/10.1109/CPEM.2018.8501259