Numerical quality factor statistics for superconducting rf cavities with spatially inhomogeneous multilayer coatings modeled by Gaussian random fields

arXiv:2605.05505 · physics.acc-ph · Submitted 2026-05-06 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: We're looking at the paper: "Numerical quality factor statistics for superconducting rf cavities with spatially inhomogeneous multilayer coatings modeled by Gaussian random fields".

Mira: Here's the gist of the paper. Numerical quality factor statistics for SRF cavities with spatially inhomogeneous multilayer coatings modeled by Gaussian random fields is a numerical study quantifying how coating thickness variations affect the quality factor of a single cell TESLA cavity, which is important because extending coating processes to non-flat geometries introduces macroscopic inhomogeneities that are difficult to assess experimentally Modeling Inhomogeneities The authors model the coating thickness variations using Gaussian random fields parametrized by a length scale, generated by solving a stochastic partial differential equation.

Kai: This resulting field is incorporated into the boundary condition of the cavity eigenvalue problem, from which quantities of interest—such as resonant frequency and quality factor—are computed.

Mira: The procedure is repeated for eight length scales, with 2048 samples per length scale, where the resulting quality factors are recorded.

Kai: Statistical Findings The results show that the quality factors follow a normal distribution.

Title and authors: Kai: Now that we’ve covered the setup and the statistical results, let's go over what the paper actually found in "Numerical quality factor statistics for superconducting rf cavities with spatially inhomogeneous multilayer coatings modeled by Gaussian random fields."

Mira: Basically, they use a stochastic partial differential equation to create a Gaussian random field parameterized by length scale l to describe coating thickness variations. This field then feeds into the boundary condition of the cavity eigenvalue problem, and from that they compute quality factors.

Lev: So it’s a simulation of how those thickness variations impact the resonant frequency and quality factor when you feed them into Maxwell's equations via that specific surface impedance model.

Kai: They are looking for quantities like the quality factor, which is quantified by taking the square of the real part of the complex angular frequency and multiplying by its imaginary part.

Mira: The main thing they established is that those quality factors follow a normal distribution, but again, they found that while mean values stay largely unchanged.

Lev: So if we can't find a systematic shift in the average quality factor due to these random thickness variations, it simplifies our design assumptions significantly.

Kai: They also found that even in extreme cases, depending on the length scale l, the quality factor can differ from the uniform case by two percent to six percent.

Mira: That variability is a measurable consequence of those inhomogeneities that needs to be accounted for in modeling device reliability.

Lev: So it boils down to quantifying how much variance we get from geometry uncertainty when designing superconducting cavities.

The paper's summary: Kai: Moving on, the paper suggests some ways they can improve their simulation scheme going forward. They suggest introducing rough surface contributions directly into the surface impedance model or moving away from the Leontovich boundary conditions to consider higher order SIBCs.

Mira: They also point out that their results might depend on deposition method, such as sputtering versus atomic layer deposition. And they mention that they aren't even looking at micro-scale deviations like those you might measure with atomic force microscopy right now.

Lev: From an error correction standpoint, we need to see how much overhead the computational cost lands when we introduce these more complex boundary conditions, because those higher order SIBCs are going to be computationally expensive.

Kai: That’s a fair point. The paper itself is quite clear about what it doesn't cover; for example, they don't consider micro-scale deviations from atomic force microscopy at this stage.

Mira: They also note that obtaining the precise distribution from experimental samples isn't realistic right now, so they have to rely on numerical simulations of the coating process itself.

Lev: So the suggested improvements are mostly about making the simulation more robust, rather than finding a new physical phenomenon entirely at this stage.

The paper's improvements: Kai: To wrap up this discussion on "Numerical quality factor statistics for superconducting rf cavities with spatially inhomogeneous multilayer coatings modeled by Gaussian random fields." The core message is that mean values were mostly indifferent with only weakly significant exceptions, but the standard deviations showed a clear increase with length scale l.

Mira: This means we can expect increased variance when we design these cavities because of the geometry uncertainties introduced by coating processes.

Lev: For our hardware, this translates directly to needing more robust designs that handle that known spread in performance statistics.

Kai: So this paper provides a way to quantify the impact of coating inhomogeneities on quality factor variance through simulation with Gaussian random fields.

Mira: It’s a useful tool for predicting how manufacturing imperfections affect the statistical properties of our superconducting components.

Lev: We're ready to look at what those increased standard deviations mean for real-world error correction overhead when we start designing systems that need to be more tolerant of these variations.

Conclusion: Kai: So we looked at "Numerical quality factor statistics for superconducting rf cavities with spatially inhomogeneous multilayer coatings modeled by Gaussian random fields." It’s about how those coating thickness variations, modeled by a Gaussian random field, mess with the quality factor of a single cell TESLA cavity.

Mira: Exactly. They used this field to change the boundary condition of the eigenvalue problem to calculate things like resonant frequency and that quality factor itself.

Lev: It’s a simulation looking at how these macroscopic inhomogeneities affect the physics when you're dealing with superconducting radio frequency cavities.

Kai: The main finding is that while the mean quality factors stayed pretty stable, they found a normal distribution for those factors across different length scales.

Mira: That’s interesting because it suggests that the average performance isn't drastically shifted by these thickness changes, which is good for baseline design assumptions.

Lev: But they also pointed out that the spread of those quality factors—the standard deviation—grows as the length scale l increases, and that you can statistically tell if those spreads are different.

Kai: In extreme cases, depending on the length scale l, they saw the quality factor could swing by two percent to six percent compared to a perfectly uniform coating.

Mira: So it shows that even small variations in the coating thickness create noticeable statistical uncertainty in how good those cavities are.

Lev: From an error correction standpoint, that increased spread means we’d need much more robust error correction schemes if we rely on these geometries at scale.

Kai: This paper really lays out the problem of modeling macroscopic inhomogeneities when extending coating processes to non-flat shapes.

Mira: The authors are also cautious, noting that the Gaussian random field they used might not perfectly represent real-world deposition methods like sputtering versus atomic layer deposition.

Lev: They flag that they aren't even looking at micro-scale deviations from things like atomic force microscopy right now, so this is a macro scale study.

Kai: Overall, the simulation scheme is versatile, and they suggest ways to improve it by adding rough surface contributions or moving to higher order boundary conditions.

Mira: They also mention that changing the underlying surface impedance model itself, maybe replacing the two-fluid model with something like BCS theory or Dynes superconductor model.

Lev: Those modifications would certainly add computational overhead, but they might give us a more accurate physical description of the damping mechanism.

Kai: So this study of "Numerical quality factor statistics for superconducting rf cavities with spatially inhomogeneous multilayer coatings modeled by Gaussian random fields" gives us concrete numbers on how geometry uncertainty affects performance variance.

Mira: It’s a solid piece of work for understanding the statistical landscape when we move beyond perfectly flat surfaces in our cavity designs.

Lev: Next up, we'll be looking at some papers on resistive wall modes and how those boundary conditions play out in full magnetohydrodynamics codes.

Aaron Gobeyn, Wolfgang Ackermann, Herbert De Gersem

Institute of Accelerator Science and Electromagnetic Fields (TEMF), TU Darmstadt

physics.acc-ph

Submitted: 2026-05-06

Updated: 2026-11-09

Journal ref: Phys. Rev. Accel. Beams 29, 103802 (2026)

DOI: 10.1103/hjqj-v5yc

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 79/100

The gist: Numerical quality factor statistics for SRF cavities with spatially inhomogeneous multilayer coatings modeled by Gaussian random fields is a numerical study quantifying how coating thickness…

Key concepts

Gaussian Random Fields
These mathematical models are used to simulate coating thickness variations across a surface. They are generated using a stochastic partial differential equation, allowing researchers to mimic the complex, non-uniform thickness found when extending coating processes beyond flat samples.
Quality Factor (Q)
The quality factor measures how well an SRF cavity stores energy and how much power it loses over time. In this study, Q is calculated by solving a complex eigenvalue problem derived from Maxwell's equations, which are influenced by the coating thickness variations.
Length Scale (l)
This parameter defines the characteristic size or spatial extent of the coating thickness variations being modeled. By testing eight different length scales, the study determines how sensitive the cavity's quality factor is to changes in this spatial scale.

Terminology

Summary

Numerical quality factor statistics for SRF cavities with spatially inhomogeneous multilayer coatings modeled by Gaussian random fields is a numerical study quantifying how coating thickness variations affect the quality factor of a single cell TESLA cavity, which is important because extending coating processes to non-flat geometries introduces macroscopic inhomogeneities that are difficult to assess experimentally

Modeling Inhomogeneities

The authors model the coating thickness variations using Gaussian random fields parametrized by a length scale, generated by solving a stochastic partial differential equation. This resulting field is incorporated into the boundary condition of the cavity eigenvalue problem, from which quantities of interest—such as resonant frequency and quality factor—are computed. p. 0

Where the paper says it

We model these variations using Gaussian random fields parametrized by a length scale, and generated by solving a stochastic partial differential equation. The resulting field is incorporated into the boundary condition of the cavity eigenvalue problem, from which quantities of interest—such as resonant frequency and quality factor—are computed.

The procedure is repeated for eight length scales, with 2048 samples per length scale, where the resulting quality factors are recorded. p. 0
Where the paper says it

The resulting field is incorporated into the boundary condition of the cavity eigenvalue problem, from which quantities of interest—such as resonant frequency and quality factor—are computed. This procedure is repeated for eight length scales, with 2048 samples per length scale, where the resulting quality factors are recorded.

Statistical Findings

The results show that the quality factors follow a normal distribution. The standard deviation increases with the length scale and can be statistically distinguished. p. 0

Where the paper says it

Our results show that the quality factors follow a normal distribution. The standard deviation increases with the length scale and can be statistically distinguished.

p. 0
Where the paper says it

Our results show that the quality factors follow a normal distribution. The standard deviation increases with the length scale and can be statistically distinguished.

In contrast, the mean values remain largely unchanged, with only a few significant differences. p. 0
Where the paper says it

The standard deviation increases with the length scale and can be statistically distinguished. In contrast, the mean values remain largely unchanged, with only a few significant differences.

p. 0
Where the paper says it

The standard deviation increases with the length scale and can be statistically distinguished. In contrast, the mean values remain largely unchanged, with only a few significant differences.

In extreme cases, depending on the length scale, the quality factor may differ from the uniform case by 2 % to 6 %. p. 0
Where the paper says it

In contrast, the mean values remain largely unchanged, with only a few significant differences. In extreme cases, depending on the length scale, the quality factor may differ from the uniform case by 2 % to 6 %.

p. 0
Where the paper says it

In contrast, the mean values remain largely unchanged, with only a few significant differences. In extreme cases, depending on the length scale, the quality factor may differ from the uniform case by 2 % to 6 %.

Simulation Methodology

The quality factor is obtained using the finite element (FE) method to numerically solve a specific eigenvalue problem involving Maxwell’s equations and the Leontovich surface impedance boundary condition. The surface impedance model is based on an SIS multilayer structure where coating thickness is inhomogeneous over the boundary, viewed as a semi-infinite space with varying thickness parameter dS. p. 1

Where the paper says it

The material assigned to the boundary 𝜕Ω is a SIS multilayer, where the coating thickness is inhomogeneous over the boundary. However, due to the order of magnitude difference in length scales, ∼1 mm to 10 mm for the size of mesh cells on the boundary and ∼10 nm to 100 nm for the layer thicknesses, we view each boundary cell as a semi-infinite space as illustrated in Fig.

The position-dependent coating thickness is defined by dS (r) = dˆ S + ΔdS (r) (41). p. 3
Where the paper says it

𝒈 is sampled from a zero-mean multivariate normal distribution with the scalar mass matrix as covariance matrix. The position-dependent coating thickness is subsequently defined by 𝑑𝑆 (𝒓) = 𝑑ˆ 𝑆 + Δ𝑑𝑆 (𝒓) (41) where 𝑑ˆ 𝑆 is a global average coating thickness, and Δ𝑑𝑆 (𝒓) is a rescaling of a sample 𝑠(𝒓)∼GP (0, 𝑐𝑠 (𝒓, 𝒓 ′ )) to a desired interval.

Analysis of Results

The distribution of quality factors does not necessarily have to be normal, but the authors observed that for each length scale, the quality factors were normally distributed, i.e., Q ∼ N (μQ, σQ). The spread of the quality factors for each l is illustrated using boxplots. p. 5

Where the paper says it

𝑄 ∼ N (𝜇𝑄, 𝜎𝑄). 5, the spread of the quality factors for each 𝑙 is illustrated using boxplots.

The standard deviations increase with increasing l, and the quality factor of a homogeneous coating lies significantly above the confidence intervals of all population means. p. 7
Where the paper says it

We found that the standard deviations increase with length scale. Furthermore, the quality factor of a homogeneous coating was found to lie above the confidence interval of all population means.

Extreme observations indicate that, depending on the length scale, the quality factor may increase or decrease by approximately 2 % to 6 %. p. 7
Where the paper says it

Furthermore, the quality factor of a homogeneous coating was found to lie above the confidence interval of all population means. Finally, extreme observations indicate that, depending on the length scale, the quality factor may increase or decrease by approximately 2 % to 6 %.

p. 7
Where the paper says it

Furthermore, the quality factor of a homogeneous coating was found to lie above the confidence interval of all population means. Finally, extreme observations indicate that, depending on the length scale, the quality factor may increase or decrease by approximately 2 % to 6 %.

Future Directions

The simulation scheme can be modified by introducing rough surface contributions to the surface impedance model or by moving away from Leontovich boundary conditions to consider higher order SIBCs. The authors note that the representative distributions may differ depending on the deposition method, such as sputtering based approaches versus atomic layer deposition. p. 7

Where the paper says it

Furthermore, one can expect the representative distributions to be quite different between e. sputtering based approaches such as co-sputtering [12, 13] and DC magnetron sputtering [14], or atomic layer deposition [10].

They also mention that micro-scale deviations, like those from atomic force microscopy, are not considered in this work. p. 7
Where the paper says it

There are, however, also micro-scale deviations, see e. the atomic force microscopy in Ref. p. 7

Where the paper says it

There are, however, also micro-scale deviations, see e. the atomic force microscopy in Ref.

Conclusion

The study concluded that the population means were mostly indifferent, with exceptions being only weakly significant. Pronounced differences were observed in the standard deviations, indicating that increasing l primarily results in an increase in σQ. p. 7

Where the paper says it

Our analysis revealed that the population means were mostly indifferent, with exceptions being only weakly significant, however, pronounced differences were observed in the standard deviations. We found that the standard deviations increase with length scale.

The quality factor of a homogeneous coating was found to lie above the confidence interval of all population means. p. 7
Where the paper says it

We found that the standard deviations increase with length scale. Furthermore, the quality factor of a homogeneous coating was found to lie above the confidence interval of all population means.

Extreme observations indicate that, depending on the length scale, the quality factor may increase or decrease by approximately 2 % to 6 %. The simulation scheme presented in Fig. p. 7
Where the paper says it

Depending on the length scale, extreme observations show a 2 % to 6 % increase or decrease in the quality factor. The simulation scheme presented in Fig.

3 can be modified and expanded upon in a number of ways. p. 7
Where the paper says it

The simulation scheme presented in Fig. 3 can be modified and expanded upon in a number of ways, however, the general structure remains.

The authors used Gaussian random fields to emulate the distribution from a coating process, however, depending on the deposition method, certain random fields may not be representative. p. 7
Where the paper says it

3 can be modified and expanded upon in a number of ways, however, the general structure remains. We used Gaussian random fields to emulate the distribution from a coating process, however, depending on the deposition method, certain random fields may not be representative.

As of time of writing, obtaining the precise distribution from experimental samples is not realistic, hence one would have to resort to numerical simulations of the coating process itself. p. 7
Where the paper says it

sputtering based approaches such as co-sputtering [12, 13] and DC magnetron sputtering [14], or atomic layer deposition [10]. As of time of writing, obtaining the precise distribution from experimental samples is not realistic, hence one would have to resort to numerical simulations of the coating process itself.

The current scheme considers macro scale deviations in coating thickness due to difficulties depositing on non-flat surfaces. p. 7
Where the paper says it

As of time of writing, obtaining the precise distribution from experimental samples is not realistic, hence one would have to resort to numerical simulations of the coating process itself. The current scheme considers macro scale deviations in coating thickness due to difficulties depositing on non-flat surfaces.

There are, however, also micro-scale deviations, see e. p. 7
Where the paper says it

The current scheme considers macro scale deviations in coating thickness due to difficulties depositing on non-flat surfaces. There are, however, also micro-scale deviations, see e.

g. the atomic force microscopy in Ref. [10]. One approach could be to introduce rough surface contributions to the surface impedance model [44], another is to move away from the Leontovich boundary conditions and consider higher order SIBCs [16]. p. 7
Where the paper says it

These contributions are not considered here. One approach could be to introduce rough surface contributions to the surface impedance model [44], another is to move away from the Leontovich boundary conditions and consider higher order SIBCs [16].

Aside from that, the surface impedance model can be improved in other ways as well, for instance by replacing the two-fluid model by more advanced formulations such as BCS theory [45] or Dynes superconductor model [46]. p. 7
Where the paper says it

One approach could be to introduce rough surface contributions to the surface impedance model [44], another is to move away from the Leontovich boundary conditions and consider higher order SIBCs [16]. Aside from that, the surface impedance model can be improved in other ways as well, for instance by replacing the two-fluid model by more advanced formulations such as BCS theory [45] or Dynes superconductor model [46].

The surface impedance model can also be replaced for other reason, such as studying thin-layer structures that are not SIS, for example SS-bilayers. p. 7
Where the paper says it

Aside from that, the surface impedance model can be improved in other ways as well, for instance by replacing the two-fluid model by more advanced formulations such as BCS theory [45] or Dynes superconductor model [46]. The surface impedance model can also be replaced for other reason, such as studying thin-layer structures that are not SIS, for example SS-bilayers.

REFERENCE

arXiv:2605.05504v2 physics. p. 0

Where the paper says it

In extreme cases, depending on the length scale, the quality factor may differ from the uniform case by 2 % to 6 %. INTRODUCTION Bulk niobium (Nb) has long been the material of choice for superconducting radio-frequency (SRF) cavities in accelerator physics applications, and has therefore been extensively studied to approach its theoretical capabilities.

acc-ph 11 May 2026

--- Page 1 ---

Numerical quality factor statistics for SRF cavities with spatially inhomogeneous multilayer coatings modeled by Gaussian random fields Aaron Gobeyn, Wolfgang Ackermann, and Herbert De Gersem Institute of Accelerator Science and Electromagnetic Fields (TEMF), TU Darmstadt (Dated: May 12, 2026) Bulk niobium has long been the material of choice for superconducting radio-frequency applications. An alternative approach is the superconductor-insulator-superconductor multilayer structure, which enables the use of brittle high-Tc materials such as NbTiN. p. 0

Where the paper says it

Numerical quality factor statistics for SRF cavities with spatially inhomogeneous multilayer coatings modeled by Gaussian random fields Aaron Gobeyn,∗ Wolfgang Ackermann, and Herbert De Gersem Institute of Accelerator Science and Electromagnetic Fields (TEMF), TU Darmstadt (Dated: May 12, 2026) Bulk niobium has long been the material of choice for superconducting radio-frequency applications. An alternative approach is the superconductor-insulator-superconductor multilayer structure, which enables the use of brittle high-𝑇𝑐 materials such as NbTiN.

At present, SIS coatings are limited to flat samples, with the single-cell TESLA cavity representing a key milestone. p. 0
Where the paper says it

An alternative approach is the superconductor-insulator-superconductor multilayer structure, which enables the use of brittle high-𝑇𝑐 materials such as NbTiN. At present, SIS coatings are limited to flat samples, with the single-cell TESLA cavity representing a key milestone.

Extending coating processes to non-flat geometries is expected to introduce macroscopic inhomogeneities in coating thickness. p. 0
Where the paper says it

At present, SIS coatings are limited to flat samples, with the single-cell TESLA cavity representing a key milestone. Extending coating processes to non-flat geometries is expected to introduce macroscopic inhomogeneities in coating thickness.

We model these variations using Gaussian random fields parametrized by a length scale, and generated by solving a stochastic partial differential equation. p. 0
Where the paper says it

Extending coating processes to non-flat geometries is expected to introduce macroscopic inhomogeneities in coating thickness. We model these variations using Gaussian random fields parametrized by a length scale, and generated by solving a stochastic partial differential equation.

The resulting field is incorporated into the boundary condition of the cavity eigenvalue problem, from which quantities of interest—such as resonant frequency and quality factor—are computed. p. 0
Where the paper says it

We model these variations using Gaussian random fields parametrized by a length scale, and generated by solving a stochastic partial differential equation. The resulting field is incorporated into the boundary condition of the cavity eigenvalue problem, from which quantities of interest—such as resonant frequency and quality factor—are computed.

This procedure is repeated for eight length scales, with 2048 samples per length scale, where the resulting quality factors are recorded. p. 0
Where the paper says it

The resulting field is incorporated into the boundary condition of the cavity eigenvalue problem, from which quantities of interest—such as resonant frequency and quality factor—are computed. This procedure is repeated for eight length scales, with 2048 samples per length scale, where the resulting quality factors are recorded.

Our results show that the quality factors follow a normal distribution. p. 0
Where the paper says it

This procedure is repeated for eight length scales, with 2048 samples per length scale, where the resulting quality factors are recorded. Our results show that the quality factors follow a normal distribution.

The standard deviation increases with the length scale and can be statistically distinguished. In contrast, the mean values remain largely unchanged, with only a few significant differences. In extreme cases, depending on the length scale, the quality factor may differ from the uniform case by 2 % to 6 %.

--- Page 2 ---

A. Eigenproblem formulation In order to obtain the quality factor of the cavity, we use the finite element (FE) method. p. 0

Where the paper says it

Different length scales of the variations are considered and compared statistically. Eigenproblem formulation In order to obtain the quality factor of the cavity, we use the finite element (FE) method.

In particular, we numerically solve arXiv:2605. p. 0
Where the paper says it

Eigenproblem formulation In order to obtain the quality factor of the cavity, we use the finite element (FE) method. In particular, we numerically solve arXiv:2605.

05505v2 [physics. p. 0
Where the paper says it

In particular, we numerically solve arXiv:2605. 05505v2 [physics.

acc-ph] 11 May 2026 the eigenvalue problem: ∇ × (∇ × E) = ω c2 E, on Ω (1) n × E = Z(r, ω) [n × (n × H)], on ∂Ω (2) where E and H denote the electric and magnetic field strengths respectively, c is the speed of light, n is the normal to the surface ∂Ω and Z(r, ω) is the frequency-dependent and position-dependent surface impedance. p. 1
Where the paper says it

2 the eigenvalue problem: ∇ × (∇ × 𝑬) = 𝜔 𝑐 2 𝑬, on Ω (1) 𝒏 × 𝑬 = 𝑍(𝒓, 𝜔) [𝒏 × (𝒏 × 𝑯)] , on 𝜕Ω (2) where 𝑬 and 𝑯 denote the electric and magnetic field strengths respectively, 𝜔 ∈ C is the angular frequency, 𝑐 is the speed of light, 𝒏 is the normal to the surface 𝜕Ω and 𝑍(𝒓, 𝜔) is the frequency-dependent and position-dependent surface impedance. The former, Eq. p. 1

Where the paper says it

2 the eigenvalue problem: ∇ × (∇ × 𝑬) = 𝜔 𝑐 2 𝑬, on Ω (1) 𝒏 × 𝑬 = 𝑍(𝒓, 𝜔) [𝒏 × (𝒏 × 𝑯)] , on 𝜕Ω (2) where 𝑬 and 𝑯 denote the electric and magnetic field strengths respectively, 𝜔 ∈ C is the angular frequency, 𝑐 is the speed of light, 𝒏 is the normal to the surface 𝜕Ω and 𝑍(𝒓, 𝜔) is the frequency-dependent and position-dependent surface impedance. The former, Eq.

The former, Eq. (1), is obtained from Maxwell’s equations in frequency domain, without sources, and with homogeneous, linear and isotropic constitutive relations. p. 1
Where the paper says it

The former, Eq. (1), is obtained from Maxwell’s equations in frequency domain, without sources, and with homogeneous, linear and isotropic constitutive relations.

The latter, Eq. p. 1
Where the paper says it

(1), is obtained from Maxwell’s equations in frequency domain, without sources, and with homogeneous, linear and isotropic constitutive relations. The latter, Eq.

(2), is the Leontovich surface impedance boundary condition (SIBC) [16]. p. 1
Where the paper says it

The latter, Eq. (2), is the Leontovich surface impedance boundary condition (SIBC) [16].

The boundary conditions can be specified fully in terms of the electric field through Faraday’s law H = (j/ωμ0)∇ × E. p. 1
Where the paper says it

(2), is the Leontovich surface impedance boundary condition (SIBC) [16]. The boundary conditions can be specified fully in terms of the electric field through Faraday’s law 𝑯 = ( 𝑗/𝜔𝜇)∇ × 𝑬.

To write down the weak formulation, we define the space Ht(curl; Ω) = u ∈ H(curl; Ω) γt(u) ∈ L2 (∂Ω) 3 (3) where the tangential trace operator γt is defined by γt: H(curl; Ω) → H−1/2 (div∂Ω; ∂Ω): u ↦→ n × u ∂Ω (4). p. 1
Where the paper says it

The boundary conditions can be specified fully in terms of the electric field through Faraday’s law 𝑯 = ( 𝑗/𝜔𝜇)∇ × 𝑬. To write down the weak formulation, we define the space 𝐻𝑡(curl; Ω) = 𝒖 ∈ 𝐻(curl; Ω) | 𝛾𝑡(𝒖) ∈ 𝐿 2 (𝜕Ω) 3 (3) where the tangential trace operator 𝛾𝑡 is defined by 𝛾𝑡 : 𝐻(curl; Ω) → 𝐻 −1/2 (div𝜕Ω; 𝜕Ω) : 𝒖 ↦→ 𝒏 × 𝒖 |𝜕Ω (4) We further define the inner product ⟨.

We further define the inner product Ω: L2(Ω) 3 × L2(Ω) 3: (u, v) ↦→ ∫ Ω dμ v†u (5) where dμ is the appropriate Lebesgue measure and v† denotes the Hermitian conjugate. p. 1
Where the paper says it

To write down the weak formulation, we define the space 𝐻𝑡(curl; Ω) = 𝒖 ∈ 𝐻(curl; Ω) | 𝛾𝑡(𝒖) ∈ 𝐿 2 (𝜕Ω) 3 (3) where the tangential trace operator 𝛾𝑡 is defined by 𝛾𝑡 : 𝐻(curl; Ω) → 𝐻 −1/2 (div𝜕Ω; 𝜕Ω) : 𝒖 ↦→ 𝒏 × 𝒖 |𝜕Ω (4) We further define the inner product ⟨. ⟩Ω : 𝐿 2 (Ω) 3 × 𝐿 2 (Ω) 3 : (𝒖, 𝒗) ↦→ ∫ Ω d𝜇 𝒗 †𝒖 (5) where d𝜇 is the appropriate Lebesgue measure and 𝒗 † denotes the Hermitian conjugate.

Finally, we assume that, for all ω ∈ C, the function Z−1(r, ω) ∈ L∞(∂Ω), such we may define the bounded operators Mω Z−1: L2(∂Ω) 3 → L2(∂Ω) 3: u ↦→ Z−1u (6) whose action is effectively multiplication by Z−1. p. 1
Where the paper says it

⟩Ω : 𝐿 2 (Ω) 3 × 𝐿 2 (Ω) 3 : (𝒖, 𝒗) ↦→ ∫ Ω d𝜇 𝒗 †𝒖 (5) where d𝜇 is the appropriate Lebesgue measure and 𝒗 † denotes the Hermitian conjugate. Finally, we assume that, for all 𝜔 ∈ C, the function 𝑍 −1 (𝒓, 𝜔) ∈ 𝐿 ∞(𝜕Ω), such we may define the bounded operators 𝑀𝜔 𝑍−1 : 𝐿 2 (𝜕Ω) 3 → 𝐿 2 (𝜕Ω) 3 : 𝒖 ↦→ 𝑍 −1𝒖 (6) whose action is effectively multiplication by 𝑍 −1 .

The weak formulation then reads: find u ∈ Ht(curl; Ω) and ω ∈ C such that ∀v ∈ Ht(curl; Ω) [17, 18]: Ω + j Z0 ω c−1 ∂Ω − ω c2 Ω = 0 (7) where Z0 is the impedance of free space. p. 1
Where the paper says it

Finally, we assume that, for all 𝜔 ∈ C, the function 𝑍 −1 (𝒓, 𝜔) ∈ 𝐿 ∞(𝜕Ω), such we may define the bounded operators 𝑀𝜔 𝑍−1 : 𝐿 2 (𝜕Ω) 3 → 𝐿 2 (𝜕Ω) 3 : 𝒖 ↦→ 𝑍 −1𝒖 (6) whose action is effectively multiplication by 𝑍 −1 . The weak formulation then reads: find 𝒖 ∈ 𝐻𝑡(curl; Ω) and 𝜔 ∈ C such that ∀𝒗 ∈ 𝐻𝑡(curl; Ω) [17, 18]: ⟨∇ × 𝒖, ∇ × 𝒗⟩Ω + 𝑗 𝑍0 𝜔 𝑐 ⟨𝑀𝜔 𝑍−1𝛾𝑡(𝒖), 𝛾𝑡(𝒗)⟩𝜕Ω − 𝜔 𝑐 2 ⟨𝒖, 𝒗⟩Ω = 0 (7) where 𝑍0 is the impedance of free space.

The domain Ω is discretized into tetrahedral elements, and the Ht(curl; Ω) space is approximated via Ned´ elec basis functions of the first kind ∈ Ht(curl; Ω) = ∑ i=1 to N w i(r) (8) which reduces the continuous problem Eq. p. 1
Where the paper says it

The weak formulation then reads: find 𝒖 ∈ 𝐻𝑡(curl; Ω) and 𝜔 ∈ C such that ∀𝒗 ∈ 𝐻𝑡(curl; Ω) [17, 18]: ⟨∇ × 𝒖, ∇ × 𝒗⟩Ω + 𝑗 𝑍0 𝜔 𝑐 ⟨𝑀𝜔 𝑍−1𝛾𝑡(𝒖), 𝛾𝑡(𝒗)⟩𝜕Ω − 𝜔 𝑐 2 ⟨𝒖, 𝒗⟩Ω = 0 (7) where 𝑍0 is the impedance of free space. The domain Ω is discretized into tetrahedral elements, and the 𝐻𝑡(curl; Ω) space is approximated via Ned´ elec basis functions of the first ´ kind {𝒘𝑖 }𝑖=1,.

(7) to the discrete eigenvalue problem K + j Z0 ω c B(ω) − ω c2 M x = 0 (9) where the components of the stiffness matrix K, boundary matrix B(ω) and mass matrix M are given by Ki j = Ω, (10) Bi j(ω) = ∂Ω, (11) Mi j = Ω (12) for i, j = 1. p. 1
Where the paper says it

𝒖(𝒓, 𝜔) = ∑︁ 𝑁 𝑖=1 𝑥𝑖(𝜔)𝒘𝑖(𝒓) (8) which reduces the continuous problem Eq. (7) to the discrete eigenvalue problem 𝑲 + 𝑗 𝑍0 𝜔 𝑐 𝑩(𝜔) − 𝜔 𝑐 2 𝑴 𝒙 = 0 (9) where the components of the stiffness matrix 𝑲, boundary matrix 𝑩(𝜔) and mass matrix 𝑴 are given by 𝐾𝑖 𝑗 = ⟨∇ × 𝒘𝑖 , ∇ × 𝒘𝑗⟩Ω, (10) 𝐵𝑖 𝑗(𝜔) = ⟨𝑀𝜔 𝑍−1𝛾𝑡(𝒘𝑖), 𝛾𝑡(𝒘𝑗)⟩𝜕Ω, (11) 𝑀𝑖 𝑗 = ⟨𝒘𝑖 , 𝒘𝑗⟩Ω (12) for 𝑖, 𝑗 = 1, .

.., N. The resonance frequency of an eigenmode f and the quality factor Q, which quantifies the damping, can be obtained from the eigenvalue ω via f = 1/2π Re(ω), Q = Re(ω)2 Im(ω) (13)

--- Page 3 ---

B. Surface impedance model Before the eigenvalue problem Eq. p. 1

Where the paper says it

6 Increment 𝑖 ← 𝑖 + 1 7 until |𝜔𝑖 − 𝜔𝑖−1 |/|𝜔𝑖 | < 𝜀 B. Surface impedance model Before the eigenvalue problem Eq.

(9) can be solved, a surface impedance model, i. p. 1
Where the paper says it

Surface impedance model Before the eigenvalue problem Eq. (9) can be solved, a surface impedance model, i.

e. description of Z(r, ω), must be specified. p. 1
Where the paper says it

(9) can be solved, a surface impedance model, i. description of 𝑍(𝒓, 𝜔), must be specified.

The material assigned to the boundary ∂Ω is a SIS multilayer, where the coating thickness is inhomogeneous over the boundary. p. 1
Where the paper says it

description of 𝑍(𝒓, 𝜔), must be specified. The material assigned to the boundary 𝜕Ω is a SIS multilayer, where the coating thickness is inhomogeneous over the boundary.

However, due to the order of magnitude difference in length scales, ∼1 mm to 10 mm for the size of mesh cells on the boundary and ∼10 nm to 100 nm for the layer thicknesses, we view each boundary cell as a semi-infinite space as illustrated in Fig. p. 1
Where the paper says it

The material assigned to the boundary 𝜕Ω is a SIS multilayer, where the coating thickness is inhomogeneous over the boundary. However, due to the order of magnitude difference in length scales, ∼1 mm to 10 mm for the size of mesh cells on the boundary and ∼10 nm to 100 nm for the layer thicknesses, we view each boundary cell as a semi-infinite space as illustrated in Fig.

1, with varying thickness parameter dS. p. 1
Where the paper says it

However, due to the order of magnitude difference in length scales, ∼1 mm to 10 mm for the size of mesh cells on the boundary and ∼10 nm to 100 nm for the layer thicknesses, we view each boundary cell as a semi-infinite space as illustrated in Fig. 1, with varying thickness parameter 𝑑𝑆.

Hence we need to determine Z(r, ω) for such a semiinfinite space, at a specific boundary point r ∈ ∂Ω. p. 1
Where the paper says it

1, with varying thickness parameter 𝑑𝑆. Hence we need to determine 𝑍(𝒓, 𝜔) for such a semiinfinite space, at a specific boundary point 𝒓 ∈ 𝜕Ω.

The model by Kubo et al. p. 2
Where the paper says it

The layers are planar and parallel to the 𝑦𝑧-plane and perpendicular to the 𝑥-axis. model by Kubo et al.

[8] provides the magnetic fields under reasonable assumptions and simplifications for precisely this setup. p. 2
Where the paper says it

model by Kubo et al. [8, 20, 21] provides the magnetic fields under reasonable assumptions and simplifications for precisely this setup.

The electric and magnetic fields are of the form E = E(x) exp(-jωt)ey and B = B(x) exp(-jωt)ez. p. 2
Where the paper says it

[8, 20, 21] provides the magnetic fields under reasonable assumptions and simplifications for precisely this setup. The electric and magnetic fields are of the form 𝑬 = 𝐸(𝑥) exp(−𝑗𝜔𝑡)𝒆𝑦 and 𝑩 = 𝐵(𝑥) exp(−𝑗𝜔𝑡)𝒆𝑧 .

Solving Maxwell’s equations and the London equations in each layer and imposing continuity in the fields at layer boundaries, one finds that the amplitudes of the magnetic fields in each layer are given by BI(x) = γB0 cosh dS − x/λ1 + r sinh dS − x/λ1, BII(x) = γB0, BIII(x) = γB0 exp - x - dS - dI / λ2 (14). p. 2
Where the paper says it

The electric and magnetic fields are of the form 𝑬 = 𝐸(𝑥) exp(−𝑗𝜔𝑡)𝒆𝑦 and 𝑩 = 𝐵(𝑥) exp(−𝑗𝜔𝑡)𝒆𝑧 . Solving Maxwell’s equations and the London equations in each layer and imposing continuity in the fields at layer boundaries, one finds that the amplitudes of the magnetic fields in each layer are given by 𝐵I(𝑥) = 𝛾𝐵0 cosh 𝑑𝑆 − 𝑥 𝜆1 + 𝑟 sinh 𝑑𝑆 − 𝑥 𝜆1 , 𝐵II(𝑥) = 𝛾𝐵0, 𝐵III(𝑥) = 𝛾𝐵0 exp − 𝑥 − 𝑑𝑆 − 𝑑𝐼 𝜆2 (14) The magnetic field value 𝐵0 is the value at the vacuum and coating layer interface, and the constants 𝛾 and 𝑟 are defined by 𝛾 −1 = cosh 𝑑𝑆 𝜆1 + 𝑟 sinh 𝑑𝑆 𝜆1 , 𝑟 = 𝜆2 + 𝑑𝐼 𝜆1 .

The magnetic field value B0 is the value at the vacuum and coating layer interface, and the constants γ and r are defined by γ −1 = cosh dS / λ1 + r sinh dS / λ1, r = λ2 + dI/λ1 (15). p. 2
Where the paper says it

The electric and magnetic fields are of the form 𝑬 = 𝐸(𝑥) exp(−𝑗𝜔𝑡)𝒆𝑦 and 𝑩 = 𝐵(𝑥) exp(−𝑗𝜔𝑡)𝒆𝑧 . Solving Maxwell’s equations and the London equations in each layer and imposing continuity in the fields at layer boundaries, one finds that the amplitudes of the magnetic fields in each layer are given by 𝐵I(𝑥) = 𝛾𝐵0 cosh 𝑑𝑆 − 𝑥 𝜆1 + 𝑟 sinh 𝑑𝑆 − 𝑥 𝜆1 , 𝐵II(𝑥) = 𝛾𝐵0, 𝐵III(𝑥) = 𝛾𝐵0 exp − 𝑥 − 𝑑𝑆 − 𝑑𝐼 𝜆2 (14) The magnetic field value 𝐵0 is the value at the vacuum and coating layer interface, and the constants 𝛾 and 𝑟 are defined by 𝛾 −1 = cosh 𝑑𝑆 𝜆1 + 𝑟 sinh 𝑑𝑆 𝜆1 , 𝑟 = 𝜆2 + 𝑑𝐼 𝜆1 .

The functions hold in a specific domain, namely BI(x) is valid for 0 ≤ x ≤ dS, BII(x) is valid for dS ≤ x ≤ dS +dI and BIII(x) holds when x ≥ dS +dI (14). p. 2
Where the paper says it

Solving Maxwell’s equations and the London equations in each layer and imposing continuity in the fields at layer boundaries, one finds that the amplitudes of the magnetic fields in each layer are given by 𝐵I(𝑥) = 𝛾𝐵0 cosh 𝑑𝑆 − 𝑥 𝜆1 + 𝑟 sinh 𝑑𝑆 − 𝑥 𝜆1 , 𝐵II(𝑥) = 𝛾𝐵0, 𝐵III(𝑥) = 𝛾𝐵0 exp − 𝑥 − 𝑑𝑆 − 𝑑𝐼 𝜆2 (14) The magnetic field value 𝐵0 is the value at the vacuum and coating layer interface, and the constants 𝛾 and 𝑟 are defined by 𝛾 −1 = cosh 𝑑𝑆 𝜆1 + 𝑟 sinh 𝑑𝑆 𝜆1 , 𝑟 = 𝜆2 + 𝑑𝐼 𝜆1 . (15) The functions hold in a specific domain, namely 𝐵I(𝑥) is valid for 0 ≤ 𝑥 ≤ 𝑑𝑆, 𝐵II(𝑥) is valid for 𝑑𝑆 ≤ 𝑥 ≤ 𝑑𝑆 +𝑑𝐼 and 𝐵III(𝑥) holds when 𝑥 ≥ 𝑑𝑆 +𝑑𝐼 .

We then obtain the surface impedance through the complex Poynting theorem in frequency domain [22] 1/2 ∫ Ω dV (E · J∗) + 2 jω ∫ Ω dV (we − wm) +∮ ∂Ω dS (n · S) = 0 (17). p. 2
Where the paper says it

(15) The functions hold in a specific domain, namely 𝐵I(𝑥) is valid for 0 ≤ 𝑥 ≤ 𝑑𝑆, 𝐵II(𝑥) is valid for 𝑑𝑆 ≤ 𝑥 ≤ 𝑑𝑆 +𝑑𝐼 and 𝐵III(𝑥) holds when 𝑥 ≥ 𝑑𝑆 +𝑑𝐼 . We then obtain the surface impedance through the complex Poynting theorem in frequency domain [22] 1 2 ∫ Ω d𝑉 (𝑬 · 𝑱 ∗ ) + 2 𝑗𝜔 ∫ Ω d𝑉 (𝑤𝑒 − 𝑤𝑚) + ∮ 𝜕Ω d𝑆 (𝒏 · 𝑺) = 0 (16) where Ω is an arbitrary volume, 𝜕Ω its boundary, 𝒏 the normal vector to that boundary and 𝑤𝑒 = 1 4 𝑬 · 𝑫 ∗ , 𝑤𝑚 = 1 4 𝑩 · 𝑯 ∗ , 𝑺 = 1 2 𝑬 × 𝑯 ∗ .

Here, we consider the x ≥ 0 semi-infinite space as the volume Ω. p. 2
Where the paper says it

(17) Here, 𝑤𝑒 and 𝑤𝑚 are the energy densities represented by the electric and magnetic components respectively, and 𝑺 is the Poynting vector. We consider the 𝑥 ≥ 0 semi-infinite space as the volume Ω.

Since the fields must decay to zero at infinity, this means only the x = 0 plane contributes to the surface integral. p. 2
Where the paper says it

We consider the 𝑥 ≥ 0 semi-infinite space as the volume Ω. Since the fields must decay to zero at infinity, this means only the 𝑥 = 0 plane contributes to the surface integral.

Using that the Leontovich boundary condition Eq. p. 2
Where the paper says it

Since the fields must decay to zero at infinity, this means only the 𝑥 = 0 plane contributes to the surface integral. Using the Leontovich boundary condition Eq.

(2), one also finds that n · S x=0= -1/2 Z(ω)H02 (18). p. 2
Where the paper says it

The electric and magnetic fields are of the form 𝑬 = 𝐸(𝑥) exp(−𝑗𝜔𝑡)𝒆𝑦 and 𝑩 = 𝐵(𝑥) exp(−𝑗𝜔𝑡)𝒆𝑧 . Solving Maxwell’s equations and the London equations in each layer and imposing continuity in the fields at layer boundaries, one finds that the amplitudes of the magnetic fields in each layer are given by 𝐵I(𝑥) = 𝛾𝐵0 cosh 𝑑𝑆 − 𝑥 𝜆1 + 𝑟 sinh 𝑑𝑆 − 𝑥 𝜆1 , 𝐵II(𝑥) = 𝛾𝐵0, 𝐵III(𝑥) = 𝛾𝐵0 exp − 𝑥 − 𝑑𝑆 − 𝑑𝐼 𝜆2 (14) The magnetic field value 𝐵0 is the value at the vacuum and coating layer interface, and the constants 𝛾 and 𝑟 are defined by 𝛾 −1 = cosh 𝑑𝑆 𝜆1 + 𝑟 sinh 𝑑𝑆 𝜆1 , 𝑟 = 𝜆2 + 𝑑𝐼 𝜆1 .

Using that the fields depends only on the x-direction and rearranging for Z(ω) one finds that Z(ω) = 1/H02 ∫ R+ d x (E(x) · J(x)) + 4 j ω ∫ R+ d x (we (x) − wm(x)) (19). p. 2
Where the paper says it

Using the Leontovich boundary condition Eq. (2), one also finds that 𝒏 · 𝑺 |𝑥=0= − 1 2 𝑍(𝜔)|𝐻0| 2 (18) Using that the fields depends only on the 𝑥-direction and rearranging for 𝑍(𝜔) one finds that 𝑍(𝜔) = 1 |𝐻0| 2 ∫ R+ d𝑥 (𝑬(𝑥) · 𝑱(𝑥) ∗ ) + 4 𝑗𝜔 ∫ R+ d𝑥 (𝑤𝑒 (𝑥) − 𝑤𝑚(𝑥)) (19) The electric field and current density are connected through the modification of Ohm’s law by the two-fluid model [23] 𝜕𝑱 𝜕𝑡 = 𝜎𝑛 𝜕𝑬 𝜕𝑡 + 1 𝜇0𝜆 2 𝑬 (20) where 𝜎𝑛 is the normal conductivity.

The electric field and current density are connected through the modification of Ohm’s law by the two-fluid model [23] ∂J/∂t = σn ∂E/∂t + 1/μ0λ2 E (20) where σn is the normal conductivity. p. 2
Where the paper says it

Using the Leontovich boundary condition Eq. (2), one also finds that 𝒏 · 𝑺 |𝑥=0= − 1 2 𝑍(𝜔)|𝐻0| 2 (18) Using that the fields depends only on the 𝑥-direction and rearranging for 𝑍(𝜔) one finds that 𝑍(𝜔) = 1 |𝐻0| 2 ∫ R+ d𝑥 (𝑬(𝑥) · 𝑱(𝑥) ∗ ) + 4 𝑗𝜔 ∫ R+ d𝑥 (𝑤𝑒 (𝑥) − 𝑤𝑚(𝑥)) (19) The electric field and current density are connected through the modification of Ohm’s law by the two-fluid model [23] 𝜕𝑱 𝜕𝑡 = 𝜎𝑛 𝜕𝑬 𝜕𝑡 + 1 𝜇0𝜆 2 𝑬 (20) where 𝜎𝑛 is the normal conductivity.

In frequency domain this yields, J = σ(ω)E, σ(ω) = σn + j μ0 ω λ2:= σ' + jσ'' (21).

--- Page 4 ---

C. Gaussian random field coating With the previous computation, we determined the surface impedance Z(ω) for a particular boundary element. p. 3

Where the paper says it

(31) and (32) also include the reactance, which has not previously been mentioned for this model, but is required for the eigenvalue problem. Gaussian random field coating With the previous computation, we determined the surface impedance 𝑍(𝜔) for a particular boundary element.

To extend this to a position-dependent surface impedance Z(r, ω), we make the coating thickness position-dependent, i. p. 3
Where the paper says it

Gaussian random field coating With the previous computation, we determined the surface impedance 𝑍(𝜔) for a particular boundary element. To extend this to a position-dependent surface impedance 𝑍(𝒓, 𝜔), we make the coating thickness position-dependent, i.

e. dS = dS (r) for r ∈ ∂Ω. For this purpose, we use Gaussian random fields. p. 3
Where the paper says it

𝑑𝑆 = 𝑑𝑆 (𝒓) for 𝒓 ∈ 𝜕Ω. For this purpose, we use Gaussian random fields.

It is known that for zero-mean Gaussian random fields with Matern covariance kernel, a sample can be generated by solving a stochastic partial differential equation (SPDE) [24]. p. 3
Where the paper says it

For this purpose, we use Gaussian random fields. It is known that for zero-mean Gaussian random fields with Matern covariance kernel, a sample can be generated by solving ´ a stochastic partial differential equation (SPDE) [24].

Here, we will consider the simplest non-trivial Matern kernel, cs (r, r′) = σ squared exp(- r - r' / l), (33) where σ is the standard deviation and l the length scale. A sample s(r)∼GP (0, cs(r, r')) can then be generated by solving the SPDE 1/l2 −∇2 s(r) = √8πσ 2/l g(r):= αg(r) on ∂Ω (34) where g(r∼GP (0, δ(r - r')) is a Gaussian white noise field and ∇2 is understood as the Laplace-Beltrami operator. Using the FE method, we can solve for s(r) on complex geometries numerically. p. 3
Where the paper says it

A sample 𝑠(𝒓)∼GP (0, 𝑐𝑠 (𝒓, 𝒓 ′ )) can then be generated by solving the SPDE 1 𝑙 2 − ∇2 𝑠(𝒓) = √︂ 8𝜋𝜎2 𝑙 𝑔(𝒓) := 𝛼𝑔(𝒓) on 𝜕Ω (34) where 𝑔(𝒓)∼GP (0, 𝛿(𝒓 − 𝒓 ′ )) is a Gaussian white noise field and ∇ 2 is understood as the Laplace-Beltrami operator. Using the FE method, we can solve for 𝑠(𝒓) on complex geometries numerically.

As before, we define the inner product Ω: L2(Ω) × L2(Ω): (u, v) ↦→ ∫ Ω dμ u v. p. 1
Where the paper says it

The boundary conditions can be specified fully in terms of the electric field through Faraday’s law 𝑯 = ( 𝑗/𝜔𝜇)∇ × 𝑬. To write down the weak formulation, we define the space 𝐻𝑡(curl; Ω) = 𝒖 ∈ 𝐻(curl; Ω) | 𝛾𝑡(𝒖) ∈ 𝐿 2 (𝜕Ω) 3 (3) where the tangential trace operator 𝛾𝑡 is defined by 𝛾𝑡 : 𝐻(curl; Ω) → 𝐻 −1/2 (div𝜕Ω; 𝜕Ω) : 𝒖 ↦→ 𝒏 × 𝒖 |𝜕Ω (4) We further define the inner product ⟨.

¯ (35). Then the weak formulation reads: for g(r∼GP (0, δ(r - r')), find u ∈ H1(∂Ω) such that ∀v ∈ H1(∂Ω) 1/l2 ∂Ω + (∇u, ∇v)∂Ω − α(g, v) ∂Ω = 0 (36). Note that no boundary condition was required. p. 3
Where the paper says it

)Ω : 𝐿 2 (Ω) × 𝐿 2 (Ω) : (𝑢, 𝑣) ↦→ ∫ Ω d𝜇 𝑢𝑣. ¯ (35) Then the weak formulation reads: for 𝑔(𝒓)∼GP (0, 𝛿(𝒓 − 𝒓 ′ )), find 𝑢 ∈ 𝐻 1 (𝜕Ω) such that ∀𝑣 ∈ 𝐻 1 (𝜕Ω) 1 𝑙 2 (𝑢, 𝑣)𝜕Ω + (∇𝑢, ∇𝑣)𝜕Ω − 𝛼(𝑔, 𝑣)𝜕Ω = 0 (36) Note that no boundary condition was required.

Since Ω ⊆ R cubed is a Lipschitz domain, ∂Ω is closed and ∂(∂Ω) = ∅, hence integrals over ∂(∂Ω) appearing due to Stokes’ theorem vanish without imposing a boundary condition. p. 3
Where the paper says it

¯ (35) Then the weak formulation reads: for 𝑔(𝒓)∼GP (0, 𝛿(𝒓 − 𝒓 ′ )), find 𝑢 ∈ 𝐻 1 (𝜕Ω) such that ∀𝑣 ∈ 𝐻 1 (𝜕Ω) 1 𝑙 2 (𝑢, 𝑣)𝜕Ω + (∇𝑢, ∇𝑣)𝜕Ω − 𝛼(𝑔, 𝑣)𝜕Ω = 0 (36) Note that no boundary condition was required. Since Ω ⊆ R 3 is a Lipschitz domain, 𝜕Ω is closed and 𝜕(𝜕Ω) = ∅, hence integrals over 𝜕(𝜕Ω) appearing due to Stokes’ theorem vanish without imposing a boundary condition.

The space H1(∂Ω) is approximated via Lagrange basis functions u = sum i=1 to Ns u i φ i (37) where Ns denotes the scalar DOF. p. 3
Where the paper says it

The space 𝐻 1 (𝜕Ω) is approximated via Lagrange basis functions {𝜙𝑖(𝒓)}𝑖=1,. ,𝑁𝑠 , so the test and trial functions are expanded as 𝑢(𝒓) = ∑︁ 𝑁𝑠 𝑖=1 𝑢𝑖𝜙𝑖(𝒓) (37) where 𝑁𝑠 denotes the scalar DOF.

This leads to the discrete linear system 1/l2 M S + K S u = αg, (38) where the components of the scalar mass matrix MS and scalar stiffness matrix KS are given by: MS,ij = (φ i, φ j)∂Ω, (39) and KS,ij = ∂Ω (40). p. 3
Where the paper says it

,𝑁𝑠 , so the test and trial functions are expanded as 𝑢(𝒓) = ∑︁ 𝑁𝑠 𝑖=1 𝑢𝑖𝜙𝑖(𝒓) (37) where 𝑁𝑠 denotes the scalar DOF. This leads to the discrete linear system 1 𝑙 2 𝑴𝑆 + 𝑲𝑆 𝒖 = 𝛼𝒈, (38) where the components of the scalar mass matrix 𝑴𝑆 and scalar stiffness matrix 𝑲𝑆 are given by: 𝑀𝑆,𝑖 𝑗 = (𝜙𝑖 , 𝜙𝑗)𝜕Ω, (39) 𝐾𝑆,𝑖 𝑗 = (∇𝜙𝑖 , ∇𝜙𝑗)𝜕Ω (40) and the vectors 𝒖 and 𝒈 are the expansion coefficients in the FE basis.

One can show that g∼N (0, M S), i. p. 3
Where the paper says it

(28) Naturally, Eq. (28) can be split into three components where the integration bounds are separated according to the domains of 𝐵I(𝑥), 𝐵II(𝑥) and 𝐵III(𝑥), which we denote as 𝑍(𝜔) = 𝑍I(𝜔) + 𝑍II(𝜔) + 𝑍III(𝜔) (29) One can show that for the semi-infinite superconductor where 𝐵(𝑥) = 𝐵0 exp(−𝑥/𝜆), Eq.

e. g is sampled from a zero-mean multivariate normal distribution with the scalar mass matrix as covariance matrix. p. 3
Where the paper says it

One can show that 𝒈∼N (0, 𝑀𝑆), i. 𝒈 is sampled from a zero-mean multivariate normal distribution with the scalar mass matrix as covariance matrix.

The position-dependent coating thickness is subsequently defined by dS(r) = dˆ S + ΔdS (r) (41) where dˆ S is a global average coating thickness, and ΔdS (r) is a rescaling of a sample s(r∼GP (0, cs(r, r')) to a desired interval. Note that the rescaling effectively makes the standard deviation σ irrelevant. p. 3
Where the paper says it

The position-dependent coating thickness is subsequently defined by 𝑑𝑆 (𝒓) = 𝑑ˆ 𝑆 + Δ𝑑𝑆 (𝒓) (41) where 𝑑ˆ 𝑆 is a global average coating thickness, and Δ𝑑𝑆 (𝒓) is a rescaling of a sample 𝑠(𝒓)∼GP (0, 𝑐𝑠 (𝒓, 𝒓 ′ )) to a desired interval. Note that the rescaling effectively makes the standard deviation 𝜎 irrelevant.

--- Page 5 ---

FIG. 1. SIS multilayer structure consisting of a superconducting coating layer of thickness dS with London penetration depth λ1 and normal conductance σ1, followed by an insulator layer of thickness dI whose dielectric properties are considered negligible, all layered on top of a superconducting substrate with material parameters λ2 and σ2 which is considered effectively infinitely thick. p. 2

Where the paper says it

3 Vacuum S 𝜆1, 𝜎1 𝑑𝑆 I 𝑑𝐼 S 𝜆2, 𝜎2 +∞ 𝑥 𝑦 𝑧 FIG. SIS multilayer structure consisting of a superconducting coating layer of thickness 𝑑𝑆 with London penetration depth 𝜆1 and normal conductance 𝜎1, followed by an insulator layer of thickness 𝑑𝐼 whose dielectric properties are considered negligible, all layered on top of a superconducting substrate with material parameters 𝜆2 and 𝜎2 which is considered effectively infinitely thick.

The layers are planar and parallel to the yz-plane and perpendicular to the x-axis. p. 2
Where the paper says it

SIS multilayer structure consisting of a superconducting coating layer of thickness 𝑑𝑆 with London penetration depth 𝜆1 and normal conductance 𝜎1, followed by an insulator layer of thickness 𝑑𝐼 whose dielectric properties are considered negligible, all layered on top of a superconducting substrate with material parameters 𝜆2 and 𝜎2 which is considered effectively infinitely thick. The layers are planar and parallel to the 𝑦𝑧-plane and perpendicular to the 𝑥-axis.

--- Page 6 ---

FIG. 2. Singe cell TESLA cavity with Gaussian random field on the boundary for different length scales The top plot displays the result for a length scale of l = 20 mm and the bottom plot for a length scale of l = 70 mm The field is rescaled to the interval [-26 nm, 26 nm] and represents the deviation from the global mean of the superconducting coating layer dS (r)

--- Page 7 ---

FIG. 3. Diagrammatic representation of the simulation scheme We solve two FE problems, one on the boundary and one on the entire domain The former yields a sample of the coating thickness distribution dS (r), parameterized by the length scale l. p. 4

Where the paper says it

Diagrammatic representation of the simulation scheme. We solve two FE problems, one on the boundary and one on the entire domain.

This in turn defines Z(r, ω), which is present in the boundary condition Eq. p. 4
Where the paper says it

(38), which yields a sample of the coating thickness distribution 𝑑𝑆 (𝒓), parameterized by the length scale 𝑙. This in turn defines 𝑍(𝒓, 𝜔), which is present in the boundary condition Eq.

(2) and hence a component of the eigenvalue problem Eq. p. 4
Where the paper says it

This in turn defines 𝑍(𝒓, 𝜔), which is present in the boundary condition Eq. (2) and hence a component of the eigenvalue problem Eq.

(9). Solving this problem yields an eigenvalue ω from which the quality factor Q can be extracted through Eq. p. 4
Where the paper says it

(2) and hence a component of the eigenvalue problem Eq. Solving this problem yields an eigenvalue 𝜔 from which the quality factor 𝑄 can be extracted through Eq.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Statistical Differentiation of Mean Quality Factors: The simulation scheme can be adapted to specifically test for statistical differences in mean quality factors across various length scales, as the null hypothesis of equal means can be rejected with p = 4.8 x 10-5, indicating that not all means are equal.

  2. Quantification of Distribution Spread: The system can be used to precisely quantify the increase in the spread of quality factors by length scale, noting that the spread becomes larger and the standard deviations increase with length scale.

  3. Identification of Extreme Deviations: The improved AI system can identify extreme cases where depending on the length scale, the quality factor may differ from the uniform case by 2 % to 6 %, providing actionable insights into coating thickness variability impact.

  4. Statistical Comparison of Length Scales: The simulation process allows for rigorous statistical comparison between different length scales by applying tests like Levene's test and Welch's ANOVA, as these methods were used to reject the null hypothesis of equal means and determine if variances are statistically distinct.

  5. Predictive Modeling for Coating Processes: By linking the modeled Gaussian random fields to deposition methods, the system can be used to predict which coating process distributions are most representative of real-world results, addressing the limitation that certain random fields may not be representative depending on the method.

Abstract

Bulk niobium has long been the material of choice for superconducting radio-frequency applications. An alternative approach is the superconductor-insulator-superconductor multilayer structure, which enables the use of brittle high- T c materials such as NbTiN. At present, SIS coatings are limited to flat samples, with the single-cell TESLA cavity representing a key milestone. Extending coating processes to non-flat geometries is expected to introduce macroscopic inhomogeneities in coating thickness. We model these variations using Gaussian random fields parametrized by a length scale, and generated by solving a stochastic partial differential equation. The resulting field is incorporated into the boundary condition of the cavity eigenvalue problem, from which quantities of interest -- such as resonant frequency and quality factor -- are computed. This procedure is repeated for eight length scales, with samples per length scale, where the resulting quality factors are recorded. Our results show that the quality factors follow a normal distribution. The standard deviation increases with the length scale and can be statistically distinguished. In contrast, the mean values remain largely unchanged, with only a few significant differences. In extreme cases, depending on the length scale, the quality factor may differ from the uniform case by 6 percent.

Sources

Related papers