Superconductor Science and Technology

Supercond. Sci. Technol. 35 (2022) 043003 (14pp) https://doi.org/10.1088/1361-6668/ac5163

Topical Review

The T-A formulation: an efficient

approach to model the macroscopic
electromagnetic behaviour of HTS
coated conductor applications
Felix Huber1, Wenjuan Song2, Min Zhang1,∗ and Francesco Grilli3,∗
1
Electronic and Electrical Engineering, University of Strathclyde, Glasgow G11XW, United Kingdom
2
Electronic and Electrical Engineering, University of Bath, Bath BA2 7AY, United Kingdom
3
Institute for Technical Physics, Karlsruhe Institute for Technology, 76131 Karlsruhe, Germany

E-mail: [email protected] and [email protected]

Received 11 September 2021, revised 17 December 2021

Accepted for publication 2 February 2022
Published 1 March 2022

Abstract
In recent years, the T-A formulation has emerged as an efficient approach for modelling the
electromagnetic behaviour of high-temperature superconductor (HTS) tapes in the form of
coated conductors (CCs). HTS CCs are characterized by an extremely large width-to-thickness
ratio of the superconducting layer, normally up to 1000 ∼ 6000, which in general leads to a very
large number of degrees of freedom. The T-A formulation considers the superconducting layer
to be infinitely thin. The magnetic vector potential A is used to calculate the magnetic field
distribution in all simulated domains. The current vector potential T is used to calculate the
current density in the superconducting layer, which is a material simulated with a highly
nonlinear power-law resistivity. This article presents a review of the T-A formulation. First, the
governing equations are described in detail for different cases (2D and 3D, cartesian and
cylindrical coordinates). Then, the literature on the implementation of T-A formulation for
simulating applications ranging from simple tape assemblies to high field magnets is reviewed.
Advantages and disadvantages of this approach are also discussed.
Keywords: HTS modelling, T-A formulation, large-scale application, electromagnetic behaviour

(Some figures may appear in colour only in the online journal)

1. Introduction several applications, some of which have already reached a

pre-commercial stage. For the design of HTS applications,
The availability of high temperature superconductors (HTS) numerical modelling plays a key role, especially when the
at lengths over a kilometre has enabled the development of electromagnetic behaviour needs to be investigated. Here, the
electromagnetic characterisation of superconductors in gen-
∗
eral and HTS in particular is quite peculiar, due to the aniso-
Authors to whom any correspondence should be addressed.
tropic critical current density in HTS materials. In addition,
the numerical methods adopted for conventional materials are
Original content from this work may be used under the terms
often not applicable.
of the Creative Commons Attribution 4.0 licence. Any fur-
ther distribution of this work must maintain attribution to the author(s) and the Over the years, different approaches to model the macro-
title of the work, journal citation and DOI. scopic electromagnetic behaviour of HTS have been proposed

1361-6668/22/043003+14$33.00 Printed in the UK 1 © 2022 The Author(s). Published by IOP Publishing Ltd
Supercond. Sci. Technol. 35 (2022) 043003 Topical Review

[1]. Among them, the H formulation became the most widely The resistivity in type II superconductors is modelled by the
used method in the applied superconductivity community E-J power law, coupling the local electric field with the local
[2, 3], although other methods, such as the minimum electro- current density within the superconducting domain only
magnetic entropy production method [4], proved to be more n(B)−1
efficient from the computational point of view [5]. The main J J
E = ρHTS · J = Ec . (3)
reason for the success of the H formulation is its ease of Jc (B) Jc (B)
implementation in the widespread commercial software pack- Here, ρHTS denotes the resistivity of the HTS material, Jc is the
age COMSOL Multiphysics. This enabled researchers who are critical electric field at 10−4 V m−1 , Jc is the critical current
not necessarily experts in the field to use numerical model- density and n is the superconducting n-value. Both Jc and n
ling and made it easy for research groups as well as the whole can depend on the magnetic field.
community to share finite element method (FEM) models and Applying Maxwell-Faraday’s law to the HTS domain gives
transfer knowledge.
One of the limitations of the H formulation is its compu- ∂B
tational speed in cases that contain large numbers of super- ∇×E = − . (4)
∂t
conducting tapes (or coil turns), especially in the case of HTS
coated conductors: those superconductors are characterized by In the whole domain Maxwell-Ampere’s law is applied to
an extremely large width-to-thickness ratio of the supercon- solve the traditional A formulation
ducting layer, which leads to a considerable increase of the ∇ × ∇ × A = µ0 µr J, (5)
total number of the degrees of freedom (DOF) of the problem.
In recent years, the T-A formulation has proved to be a com- where µ0 and µr are the magnetic and relative permeability of
petitive alternative for modelling systems with a large number the free space, respectively.
of HTS coated conductors, where the superconducting layer is The current density is coupled with the A formulation by
regarded as an infinitely thin sheet and the variation of the elec- an applied external surface current density Je
tromagnetic quantities across its thickness can be neglected
Je = J · δ, (6)
[6, 7].
This article provides a review of the works based on the T-A where δ denotes the thickness of the HTS tape. Je is imposed
formulation and is organized as follows. Section 2 describes in the tape in A m−1 to accommodate for the thin strip approx-
the governing equations of the formulation in 2D and 3D, imation, which is neglecting the thickness of the HTS layer.
including the extension to the homogenisation and multi- The transport current flowing in the superconductor is
scale approaches, used to handle cases with large numbers applied at the tape terminals. It can be represented by an
of HTS tapes. The validation of the methodology as well as integral of the current density over the cross-section of the
an efficiency analysis is reported in section 3. Section 4 is conductor
dedicated to applications: it starts with basic tape and coil ¨ ¨ ˛
modelling, then addresses several power applications, and it I= J dS = ∇ × T dS = T ds. (7)
finally concludes with high field magnets. Section 5 contains
S S ∂S
a discussion on the T-A formulation, highlighting its advant-
ages and disadvantages. Eventually, section 6 draws the main Here, S is the cross section of the conductor and ∂S represents
conclusions. the boundary edges of the cross section.

2.2. Formulation in 2D (infinitely long strip, axial-symmetric)

2. T-A formulation
Due to the high aspect ratio of HTS coated conductors, model-
The T-A formulation was first introduced by Zhang et al [6] ling the whole surface requires long simulation time. The T-A
and Liang et al [7] in 2016. The concept of this numerical formulation makes unique use of this issue, by applying the
model is to solve the current vector potential T only in the thin strip approximation to the tape. This geometrical simpli-
superconducting domain, while the magnetic vector potential fication reduces the surface area of the HTS tape to a thin sheet
A is solved for the whole domain. The two state variables T (figure 1).
and A are used to calculate the current density J and the mag- Therefore, the current is restricted to flow within the super-
netic flux density B, respectively. The numerical model is fully conducting sheet, while T is reduced to the component per-
coupled by solving for both J and B at the same time and con- pendicular to the tape. The current vector potential can then
 T
tinuously exchanging these variables. be defined as T · n, where n = nx ny nz and n =
[nr nφ nz ]T are the normal vectors perpendicular to the wide
face of the superconducting tape in cartesian and cylindrical
2.1. General formulation
coordinates, respectively. In 2D the surface element of the
The governing equations of the general form are, HTS tape is reduced to a 1D line element and equation (1)
can be written as
J = ∇×T (1)
∂(T · ny ) ∂ (T · nx )
Jz = − (8)
B = ∇ × A. (2) ∂x ∂y

2
Supercond. Sci. Technol. 35 (2022) 043003 Topical Review

 
∂ (T · nz ) ∂ (T · ny )
−
   ∂y ∂z 
Jx  
 ∂ (T · nx ) ∂ (T · nz ) 
 Jy  =  − . (13)
 
Jz  ∂z ∂x 
 ∂ (T · ny ) ∂ (T · nx ) 
−
∂x ∂y

The implementation of Faraday’s law in the superconduct-

ing domain can now be written as
   
∂Ez ∂Ey ∂Bx
 −   
 ∂y ∂z   ∂t 
Figure 1. T is computed only in the superconducting domain (black
 ∂Ex ∂Ez   ∂By 
lines), while A is computed everywhere. The thickness of the HTS  − ·n+  · n = 0. (14)
 ∂z ∂x   
tape is collapsed into a superconducting layer, represented by a 1D    ∂t 
line element. The current is enforced by boundary conditions at the  ∂Ey ∂Ex   ∂Bz 
edges T1 and T2 of the tape. −
∂x ∂y ∂t

in cartesian coordinates, or for axial-symmetric problems in The boundary condition for the transport current is applied
cylindrical coordinates as in the same way as for a 2D problem in equation (12), only
instead of defining it at the terminal points, it is applied to the
∂(T · nr ) edge of the superconducting sheet. In 3D there is an additional
Jφ = . (9) boundary condition to consider. When symmetry is exploited
∂z
the conductor cross section is defined with a zero-flux condi-
Depending on the orientation of the tape, T · ny or T · nx in tion in the T formulation to contribute with the magnetic insu-
equation (8) can further be neglected. Equation (9) expresses lation boundary condition of the A formulation.
a general case for a tape wound with an ‘easy bend’ where
the current flows in the φ direction, which reduces the cur-
rent vector potential to T · nr . Furthermore, when implement- 2.4. Multi-scale and homogenisation method
ing Faraday’s law, equation (4) can be reduced to Based on the general form of the T-A formulation, two sim-
 ∂Ez  ∂B  plification methods can be implemented in order to reduce
x
the amount of computational resources needed to model large
 ∂y   ∂t 
 ·n+ ·n = 0 (10) numbers of tapes in 2D and 3D. Although simplification
∂Ez ∂By
− assumptions are made, the accuracy for hysteresis losses, cur-
∂x ∂t rent distribution and magnetic field distribution is maintained
    [8]. Both these methods have been validated in 2D by [8, 9]
∂Eφ ∂Br
− ·n+ ·n = 0 (11) and in 3D by [10, 11].
∂z ∂t

for cartesian and cylindrical coordinates respectively. Again, 2.4.1. Multi-scale method. In the multi-scale method, the
only the x or y component of the magnetic flux density in current vector potential is defined for each analysed tape
cartesian coordinates needs to be considered, depending on the individually.
orientation of the tape. The current density distribution is obtained by computing T
A transport current, as defined in equation (7), can then be in the analysed tapes, while the non-analysed tapes in between
applied to the edges of the tape using are approximated by linear interpolation, shown in figure 2.
The magnetic flux density is computed via the A formulation,
I = (T1 − T2 ) δ (12) which is defined in the whole space. In order to couple the T
and A formulation, the current density is imposed into the A
where T1 and T2 are the current vector potentials at the respect- formulation using equation (6), where J is multiplied by the
ive edge points and δ is the thickness of the superconducting thickness of the tape, to obtain the surface current density.
tape.

2.4.2. Homogenisation method. The homogenisation

2.3. Formulation in 3D
method assumes that a stack of tapes can be considered as
In a 3D geometry, the superconductor is no longer represented an anisotropic bulk, which will not compromise the electro-
as a line, instead the conductor will be modelled as a thin sheet dynamic behaviour of the coil [8]. In the T-A formulation, a
[6]. This surface element also does not consider the thickness tape is considered as a thin sheet with the thickness δ, where
of the HTS. Using the same set of equations for a 3D geometry, we now define a unit cell around it with the thickness Λ. The
the current density vector is defined as unit cell assumes the thickness of the additional layers in a

3
Supercond. Sci. Technol. 35 (2022) 043003 Topical Review

Table 1. T-A full, multi-scale and homogenous model comparison

[8].

T-A model Losses (J m−1 ) Error (%) ct (h)

full 0.8832 — 9 h 04 min

multi-scale 0.8807 0.28 1 h 13 min
homogeneous 0.9025 2.18 0 h 37 min

 
√ x
 x 2 +y 2

n=ρ=

√ 2y 2 .
 (17)
Figure 2. T-A multi-scale approach in 2D. In this example, there are x +y
three analysed tapes (black) whereby J is obtained by computing T. 0
J in the remaining non-analysed tapes (green) is approximated by
linear interpolation. The surface current Je is imposed in the A
formulation by means of a boundary condition. See also [8].
Here, i defines the normal vector in a cartesian coordinate sys-
tem and ρ is defined in a cylindrical frame.
The boundary conditions for the transport current are
applied in the same way as in the 3D tape model, with Dirichlet
boundary conditions at the top and bottom of the homogen-
ized bulk. For the additional boundaries that come from the
homogenisation technique, a Neumann boundary condition is
defined as,

∂ (Tx nx + Ty ny )
= 0. (18)
∂n
In a 2D homogenisation problem, equation (18) can be reduced
to only the scalar component of T.
Figure 3. T-A homogenisation in 2D. The superconducting sheets
In the case of a closed coil, symmetry conditions can be
are replaced by a homogeneous bulk. The current density inside the useful to simplify the problem. When considering a quarter
bulk is computed by boundary conditions applied to the edges T 1 model, the symmetry planes use a zero flux condition for the
and T 2 of the bulk and the scaled current density Je is imposed in superconducting domain. In the A formulation the symmetry
the A formulation. plane is considered a perfect magnetic conductor, mirroring
the current of the superconductor.
HTS tape such as copper, hastelloy, buffer layer, insulation
material, but the materials themselves are not modelled sep- 3. Validation of modelling strategies
arately. Therefore, the actual size of the bulk is given by the
stack of tapes with the unit cells surrounding them. The tapes The validation of the T-A formulation, especially of the
are then transformed into a single block, which is shown in advanced modelling strategies (multi-scale, homogenisation),
figure 3. was carried out by comparisons with experiments and other
The scaled external current density Je in the homogenized numerical models, mainly the H formulation. This section is
bulk is imposed in the A formulation using, dedicated to show a quantitative evaluation of the performance
of the T-A formulation compared to the H formulation.
δ In [8], the authors proposed the T-A multi-scale and homo-
Je = J · . (15) geneous approach as an efficient alternative for large-scale
Λ
HTS applications.Their simulations compared the accuracy
When expanding the homogenisation method to 3D geo- and computational time (ct) of the different models for a coil
metries, software like COMSOL only features the normal vec- configuration of 200 tapes for each of the 10 pancakes. Making
tor on boundary elements. Therefore, it is necessary to define use of symmetry conditions reduced the total number of tapes
the normal vector n inside the bulk. to 500. The results of this comparison are shown in table 1.
Considering a simple 3D geometry that is commonly stud- Furthermore, an extensive comparison with the H formula-
ied like in the case of a racetrack coil, n will be defined for the tion was conducted to validate these approaches with different
straight and circular part respectively orders of elements for the T and A formulation and the results
are shown in table 2.
  Further investigations about the validity of the T-A formula-
1 tion in 2D was conducted in [12], where different orders for the
n = i =  0 . (16) dependent variables were evaluated and compared with the H
0 formulation. The authors concluded that the T-A formulation

4
Supercond. Sci. Technol. 35 (2022) 043003 Topical Review

Table 2. H and T-A full models comparison [8].

Model Order Losses (W m−1 ) Error (%) ct (h)

H full 127.23 — 31 h 32 min

T-A full T 1◦ , A 1◦ 126.02 0.96 02 h 46 min
T 2◦ , A 2◦ 126.38 0.67 09 h 01 min
T 1◦ , A 2◦ 128.05 0.64 03 h 14 min

Table 3. Comparison of different modelling approaches with the H

and T-A formulation [9].
Losses
(W m−1 ) ct (h) DOF

Reference 127.24 31 h 32 min 563 893

Model Error (%) ct (%) DOF

H full 1.62 55.81 359 408

H homogenous 1.28 1.94 11 838
H densified 6.67 29.03 120 424
T-A full 0.64 10.25 548 624
T-A 0.71 0.78 20 612
homogenous
T-A densified −2.62 3.22 103 638

is the preferred choice for large-scale applications, due to the

faster computation time and high accuracy.
Another study regarding the element order of the depend-
ent variables of the T-A formulation was carried out by the
authors in [13]. Simulations using T 1◦ , A 1◦ and T 1◦ , A 2◦
were compared with the H formulation using first order curl
elements. Although all models showed good results, the T-A Figure 4. Bubble chart presenting the trade-off between accuracy
formulation was advantageous for complex problems like the and computation time. The bubbles closer to the origin are those of
simulated 14-strand Roebel cable. The results of this paper are the models that better fulfil the compromise between accuracy and
computation time. The areas of the bubbles are proportional to the
reviewed in more detail in section 4.2.1.1.
number of DOF. Reproduced from [9]. © IOP Publishing Ltd
In their latest publication to date [9], the authors expanded CC BY 3.0.
on their previously conducted study in [8] and provided an
even more detailed comparison of the different T-A and H for-
mulation models, including densified modelling strategies as In [11], the implementation of the T-A homogenisation
well as iterative and simultaneous approaches. Also included method was achieved for the first time in 3D. The valida-
was a comparison of models with regards to the degrees of tion of the method was done by experimental analysis of an
freedom (DOF), which have a direct impact on the simulation 86-turn double pancake racetrack coil. For applications such
time. as a racetrack coil where a straight as well as a curved seg-
A compact version of these results is presented in table 3 ment of the coil needs to be modelled, 2D simplifications are
where ct denotes the normalized computation time. It can be often not sufficient to accurately simulate the electromagnetic
seen that especially the homogenous approach for both H behaviour. The authors concluded that the AC loss of 2D mod-
and T-A formulation greatly reduced the number of DOF and els could show a larger discrepancy if the number of turns
subsequently reduced the computational time tremendously. was increased. Furthermore, the 3D T-A homogenized model
In figure 4, the authors compared the normalized computa- improves in its efficiency as the number of turns increases,
tion time with the absolute value of the losses error shown in compared to other FEM models such as the H formulation.
table 3. Again, the results confirmed the improved computa- Therefore, this method could be valuable for the numerical
tional speed with similar accuracy for the T-A formulation. analysis of large-scale HTS problems, which cannot be rep-
It was shown that for the T-A formulation (striped bubbles), resented accurately in 2D.
the homogeneous and the simultaneous multi-scale were In [14], the authors continued their investigation of the
within the first square of the grid. It was concluded that these homogenisation method in 3D using the T-A formulation.
methods have an error of less than 1% and less than 10% of The 3D approach was compared with 2D infinitely long and
the normalized computation time. axial-symmetric models to show the difference in the current

5
Supercond. Sci. Technol. 35 (2022) 043003 Topical Review

distribution as well as the losses. Although a 2D estimation of currents. It was concluded that in reducing both quantities,
a racetrack coil is often sufficient, it was shown that there are the two problems were negated, which is contradictory to the
differences in the penetration of the HTS coil, due to the aniso- use of HTS, since high Jc and high n-values are desired. Other
tropy of the critical current. In comparison with the homogen- ways of mitigating the SCIF and field drift are the striation of
ised H formulation the higher penetration in the arc segment HTS tapes and the current sweep reversal, which are shown
of the racetrack coil was validated since the influence of the to be effective methods. As these methods depend on the coil
self-field in this region is higher than in the straight part. The configuration, they need to be evaluated for each magnet and
authors had to concede that the H formulation was faster than operating cycle.
the T-A formulation, unlike in 2D. Nevertheless, they state that The screening current induced strain gradients of a small
the mode is extremely necessary and will enable the design and REBCO pancake coil were investigated experimentally and
optimisation of HTS applications in 3D in the future. analytically in [17]. Two coils with a diameter of 150 mm and
monofilament and another with 3-striate/4-filament REBCO
tapes were tested. A 5 T room temperature magnet was used
4. Applications to excite a strong screening current and further apply a non-
uniform Lorentz force to each coil in the experimental setup.
In this section the research conducted using the T-A formula- The simulations were conducted using the axial-symmetric 2D
tion is reviewed. In power applications, the T-A formulation T-A formulation. At low strain gradients, the experimental res-
is of particular interest for the simulation of HTS cables and ults agreed with the numerical solution. The striated multifila-
large-scale magnet applications making use of the advantages ment conductor provided a solution to the mitigation of screen
from advanced modelling strategies such as the homogenisa- current induced strain gradient.
tion method to model large numbers of HTS turns. An extension of the T-A formulation model to thick super-
conductors was proposed in [18], which was used to simulate
a racetrack coil with 52 turns. The turns were considered to be
4.1. Basic tape and coil modelling
either in electrical contact (coupled) or electrically insulated
The critical current of superconducting coils was investigated (uncoupled). Three different models, the T-A and H formu-
in [15], as this is crucial for the limitation of the performance lation as well as the MEMEP method were used to compare
of the coil. The limiting factors such as temperature, the mag- the losses of this stator coil for a superconducting motor. The
nitude and orientation of the magnetic field inside the super- results showed a good agreement between the methods, only
conductor were evaluated using H and T-A formulation mod- at low currents for the coupled T-A formulation there was a
els. Also, static simulations with the P-model and the modified slight discrepancy. This was because in the T-A formulation
load-line method were used to estimate the critical current of the coupled tapes are simulated as bulks rather than individual
coils when time approaches infinity. In order to model a super- tapes when using the H formulation and MEMEP method. The
conducting double pancake racetrack coil with more than 200 losses of the coupled case were twice as high compared to the
turns of a 4 mm-wide tape, a 2D infinitely long conductor was uncoupled case, due to the non-uniform current distribution in
considered wherein the circular part was neglected. This was the tapes. It was shown that the T-A formulation was advant-
justified by the authors since the critical current of a racetrack ageous especially for 2D problems that can be approximated
coil is determined by the average electric field of the circular as infinitely long or have an axial symmetry.
and the straight part. When the straight part was longer than In [19], the authors implemented the T-A formulation in
the circular part, comparisons of 2D T-A and 3D H formula- GetDP. The authors proposed a new derivation of the T-A for-
tion models showed a good agreement. Their conclusion listed mulation, wherein global constraints were imposed on the cur-
the T-A formulation as reliable for the estimation of the critical rent or the voltage for the individual tapes. It was concluded
current density, although it overestimated the Jc by 19%. This that the T-A formulation is an efficient formulation to model
was assumed to be due to the varying n-values, length uniform- the electromagnetic behaviour of superconducting tapes.
ity of the critical current and angular dependence, manufactur-
ing process as well as the different cooling efficiency between 4.2. Power applications
the short sample, which was used to verify the findings exper-
imentally, and the whole coil. This section reviews the works implementing different forms
In [16], the effects of screening current, caused by the pen- of the T-A formulation discussed in section 2 for calculating
etration of the magnetic field into the HTS tapes was invest- AC losses in various applications and scenarios.
igated. Two major problems, namely the screening current
induced field (SCIF) and the field drift were identified for 4.2.1. High current cables. The T-A formulations have been
wound HTS coils. The effects of Jc and the n-value were stud- implemented to compute the AC losses in various types of high
ied for both problems. Using the T-A formulation and making current cables which are comprised of HTS tapes.
use of the homogenisation method allowed the authors to run
multiple simulations, which would have taken months to simu- 4.2.1.1. Roebel cables. In [20], the authors studied the rem-
late with the conventional H formulation. The results showed nant field characteristics of REBCO based 10-strand Roebel
that a higher Jc causes a higher screening current, while an cables after removal of the external magnetic field, numer-
increase in n-value hinders the relaxation of the screening ically and experimentally. A 3D FEM model using T-A

6
Supercond. Sci. Technol. 35 (2022) 043003 Topical Review

formulation was built and validated with the measured results CORC® cable facing a hot spot could be significantly reduced
mapped out using a Hall probe. They found that the remnant by minimizing the TCR. With a low TCR, such as Rt < 5 µΩ,
field was dependent on the maximum external magnetic fields the current redistribution mainly depended on the critical cur-
and different height with respect to the tape wide surface. rent of the tape without a hot spot. With high terminal resist-
In [13], the authors carried out a performance assessment of ances (Rt > 5 µΩ), the current redistribution mainly depended
T-A formulation against H formulation on AC loss calculation on the TCR.
of superconductor, in terms of the number of DOF, computa- In [23], the authors focused on the magnetization loss ana-
tion time and accuracy. A combination of element order for T lysis of CORC® cables by an experimentally validated 3D
formulation and A formulation were used, varying from linear model using T-A formulation. In particular, the authors stud-
to cubic, while first-order curl element and second-order curl ied how the winding direction, multilayer structure and utiliz-
elements were considered for H formulation model. Different ation of striation tapes could affect the magnetization loss of
relative tolerances were investigated as well, such as 3 × 10−4 , CORC® cables. They found that an increase of the number of
1 × 10−4 and 1 × 10−5 . Two benchmark models were chosen: cable layers could decrease cable losses due to the field shield-
a single REBCO straight tape modelled both in 2D and 3D, and ing effects. For CORC® cables with multiple layers, the mag-
a 14-strand Roebel cable modelled in 3D. It was found that netization loss was significantly affected by the winding direc-
in the model of 2D single tape with coarse mesh, the solving tion and the packing arrangement of air gaps in each layer. In
time, and the number of DOF were comparable in the T-A and addition, tape striation could remarkably reduce the magnet-
H models set with equivalent element orders. When the mesh ization loss only when the CORC® cable was exposed to high
was finer, T-A model with linear Lagrange element for T for- magnetic fields.
mulation and quadratic element for A formulation was much In [10], the authors proposed an accelerated 3D T-A for-
faster, as compared to H formulation model with curl linear mulation model to study the AC losses of a CORC® -like con-
element, although the number of DOF was larger. The 3D T- ductor based on quasi-isotropic conductor. The acceleration
A model where linear Lagrange element and linear element of the T-A model was achieved by reducing the solution area
were set for T and A formulation achieved better accuracy of employed with T formulation. In other words, the J distri-
AC losses. However, it took longer time compared with H for- butions of some selected superconductors were solved by T
mulation model with curl linear element. In the case of the 14- formulation and J distributions of other superconductors were
strand Roebel cable, the 3D T-A model with linear Lagrange approximated by linear interpolation. It was concluded that the
and quadratic element set for T and A variables was advant- pitch length of the CORC® -like cable could significantly affect
ageous in terms of time per thousand DOF but required much the current distribution among superconductors and, hence,
greater computation memory, as compared to curl linear ele- their AC loss. An optimal pitch length was found with min-
ment H formulation model. imized AC losses.
In [24], the quench of a single-layer REBCO-based
4.2.1.2. CORC® cables. In [21], the authors studied the CORC® cable wound by three tapes with non-uniform ter-
influence of harmonic current on transport AC losses of a minal contact resistance Rt was investigated. A 3D multi-
10 MW/3 kV 2 kA tri-axial CORC® HTS cable aimed for physics model was built, including three modules coupled with
hybrid-electric aircraft, by means of a 2D model based on T-A each other, a T-A formulation model, a heat transfer model and
formulation. It was observed that under three-phase balanced an equivalent circuit model. A heat pulse was applied to sim-
load and sinusoidal current, the loss in the outer phase could ulate and initiate a local hotspot on one tape. Current redis-
be twice that of the inner phase due to the magnetic field dis- tribution, voltage, and temperature of each tape during the
tribution; a small harmonic current (less than 10% of the rated hotspot induced quench process were reported and compared,
current) could result in a significant increase in AC losses, up when the heat pulse was imposed on the tape with different
to 40%. Rt . It was concluded that for a single-layer CORC® cable with
Although the 2D model could simulate the electromagnetic non-uniform terminal contact resistance, when a hotspot was
behaviour and predict the AC losses of CORC® cable within induced on the tape with middle value of Rt , the CORC® cable
an acceptable accuracy, the real structure of the CORC® cable was most stable due to the highest minimum quench energy.
was not totally included. Then, a 3D model of the CORC® In [25], the author showed her study on advanced 3D and
cable using T-A formulation was proposed and developed to 2D modelling of HTS CORC® cable with T-A formulation
enhance on this. implemented for the electrical propulsion system. The focus of
In [22], to investigate the physical performance of a the thesis was to understand how the cable structure affects the
REBCO based CORC® cable when a hot spot was induced on electromagnetic and quench behaviour, along with how these
one tape, the authors developed a 3D multi-physics electro- effects could be minimized. The author developed and applied
magnetic-thermal model by coupling four modules, including T-A formulation to 3D modelling of CORC® cables. T-A for-
a T formulation model, an A formulation model, a heat trans- mulation demonstrates advantageous in simulating the real
fer model and an equivalent circuit model. They analysed and structure of CORC® cables to achieve acceptable accuracy by
discussed the current redistribution among all tapes (figure 5), relatively fast speed, as compared to simple 2D model which
voltage drop, and temperature variation of the tapes caused could not fully represent the complex electromagnetic field
by the hot spot, under different amplitude of terminal contact coupling due to the real structure, and 3D models using con-
resistance (TCR). It was found that the quench risks of the ventional H formulation which could have taken up to months.

7
Supercond. Sci. Technol. 35 (2022) 043003 Topical Review

Figure 5. Distribution of the current density in the superconducting layer, metallic layers, and temperature during a hot-spot-induced
quench in a CORC® cable. The heat quench was 183 mJ and terminal contact resistance was 100 µΩ. Figure reproduced with permission
from [22]. © IOP Publishing Ltd All rights reserved.

The 3D model of CORC® cable using T-A formulation allows at any transport current, while AC loss in double coaxial cable
authors to investigate more critical issues of CORC® cable, was 20% lower than that of CORC® cable when exposed to
including to calculate accurate AC losses, to investigate hot background magnetic field.
spot-induced quench along the cable, to explore current shar- In [28], to achieve high current capability of supercon-
ing in different layers of the CORC cable mainly caused by ductor, the authors proposed a quasi-isotropic straight super-
terminal contact resistance, etc. conducting strand, and characterized a 35 cm short sample by
In [26], the authors studied the CORC® cable loss using a means of experiment and numerical calculation. The quasi-
multi-physics electromagnetic-thermal coupled model based isotropic strand was comprised of four units, each unit of
on T-A formulation. The model was validated by comparing which was stacked alternatively by 7 copper tapes and 7 HTS
the transport AC losses of a single tape solely calculated by H tapes with the width of 2 mm. Transport AC losses in the pro-
formulation, 2D T-A formulation and 3D T-A formulation. The posed strand were measured under different frequencies ran-
authors proposed an improved 3D T-A model with the func- ging from 30 Hz to 200 Hz. Magnetic flux density distribu-
tionality of analysing current redistribution between supercon- tion and AC loss in each HTS tape were also discussed, at
ducting layer and other metal layers during over current and peak current of 800 A, 50 Hz. The AC loss results calculated
fault regime, by post-processing the electromotive force E, from a 3D T-A model was consistent with the measured ones
although the superconductor was regarded as a thin shell in at frequencies below 75 Hz. At frequencies of 30 Hz, 50 Hz
the model. and 75 Hz, the calculated AC loss dependence on normalized
current Ipeak / Ic keeps consistent with the experimental results,
but much higher than the Norris strip model, due to two main
4.2.1.3. Stacked-tape cables. In [27], 3D electromagnetic reasons. One reason is the contribution of eddy current loss
numerical models were built based on T-A formulation, to in the copper tapes, and the other reason is that each HTS
study the AC loss characterization of CORC® cable, twisted- tape in the quasi-isotropic strand is affected by the magnetic
stacked-tape cable (TSTC), and double coaxial cable, under fields from other tapes, whereas Norris model assumes a uni-
different transport current and external magnetic field. The form current transporting in self-field. AC losses in the quasi-
effect of pitch length in double coaxial cable inner conducting isotropic stack shows frequency dependency which indic-
layer was also discussed. The results showed that AC loss in ates that the eddy current loss in the copper tapes cannot be
TSTC cable increased an average of 40% over the other cables, neglected.

8
Supercond. Sci. Technol. 35 (2022) 043003 Topical Review

4.2.1.4. Cables with twisted structures. In [29], the authors the magnetic flux density and normalized current density.
proposed a finite element 2D model of a 5-slots twisted- It was found that minimal fill factor in the superconducting
stacked-tape slotted-core HTS Cable-in-conduit conductor bias coil was not suitable for minimizing its critical current
(CICC) using T-A formulation, to investigate the current dis- density.
tribution among tapes in HTS stack(s). The numerical model
was validated by the agreement with experimentally measured 4.2.3. Electrical machines. In [32], the authors aimed to
V–I curves of the CICC. To study the effect of terminal con- establish T-A formulation models to simulate superconducting
tact resistance Rterm on the current distribution between tapes electrical machines and AC losses in superconducting wind-
within stacks, two cases were considered: Rterm was identical ings. Firstly, a 2D time dependent model based on T-A for-
for each tape-copper joint, and Rterm showed a statistical dis- mulation was developed to evaluate the AC losses of a spe-
tribution, under three experiments: #1 stack of 20 tapes, #2 cific synchronous motor. The model was further validated by
two adjacent stacks, and #3 a bent stack with bending radius comparing the current density distribution and AC losses with
of 0.15 m. It was found that the variation of Rterm could signi- the previously validated results calculated using the minimum
ficantly change the initial current distribution within the tapes electromagnetic entropy production method (MEMEP). The
of stacks. What’s more, V–I curves would not change drastic- AC losses computed by the two methods had a difference
ally if Rterm was smaller than 500 nΩ. I c degradation and non- around 5.6%, which indicated a good agreement. After this,
uniform current distribution of the tapes were recorded in the the authors added the J c anisotropy into the model, by incor-
bending stack, which could further lead to thermal instability porating the perpendicular-field-dependent I c data into the
during operation. motor model. The AC losses in the stator winding at 65 K and
In [30], AC losses in a twisted quasi-isotropic supercon- 77 K calculated using the proposed T-A model and MEMEP
ducting conductor stacked by HTS tapes and copper tapes model showed a difference of 1.5% and 5.6% only, which
alternatively were investigated by 3D simulations based on again proved the accuracy of the proposed model.
T-A formulation and experiments. A twisting apparatus was In [33], the authors studied the electromagnetic behaviour
developed to accurately adjust and fix the twist angles of the and AC loss estimation of a 10 MW synchronous generator,
quasi-isotropic conductor. Critical current and AC loss beha- in which the rotor was comprised of permanent magnets, and
viour under different twist pitch length were measured at 77 K the stator was comprised of superconducting coils, using T-A
in liquid nitrogen bath. The measured critical current of the formulation. To obtain current distribution in stator windings
twisted conductor tended to degrade with the decrease of pitch for feeding the T-A formulation model, a resistive model was
length, and a 5% critical current degradation was found when implemented firstly by connecting a load to the generator;
the pitch length was 240 mm. The AC loss of the twisted quasi- once the induced current distribution was captured, the resist-
isotropic conductor was reduced to 89% and 79% of that in the ive model was disconnected, and the superconducting model
non-twisted conductor at low frequencies, when the twisted started to compute. Three generator designs (figure 6) were
angle was 180◦ and 300◦ , due to the uniform current distribu- studied and compared: the first one was based on distributed
tion among the HTS tapes. coils wound with 4 mm-wide tape located in the stator with
iron teeth; the second design, similar to the first one, had no
magnetic teeth but air-gap windings; the third design was sim-
4.2.2. Fault current limiters. In [31], the authors focused ilar to the first one, but 2 mm-wide tapes were used instead
on the superconducting DC bias coil of saturated iron of 4 mm.
core superconducting fault current limiter (SIC-SFCL). A Conclusions were drawn that the slot-less design yielded
multi-objective optimization approach using the Nelder-Mead to uniform current distribution among coils, and thus, losses
algorithm for an optimal coil geometry was proposed and were dissipated more uniformly. A roughly 44% AC loss
presented, with three objective functions: to maximize the crit- reduction was observed in the third design, as compared to
ical current density in the HTS tape, minimize the current flow- the first design. A sensitivity study was also carried out over
ing the DC biased superconducting coil, and minimize the coil the frequencies and cryogenic temperatures. A linear increase
price. The SIC-SFCL was simulated by a 2D T-A formula- of losses in kW was observed when the frequency increased
tion model in COMSOL, which was validated by measure- at given temperature. Much higher losses were produced in
ments of a non-superconducting saturated iron core prototype. the first and second design when the temperature was above
The finite element model was coupled with an electrical cir- 68 K, whereas the third design showed steady loss perform-
cuit by electrical lumped parameters, representing the DC bias ance against temperature variation.
superconducting coil and AC copper coils. In terms of the coil In [34], the authors studied the AC losses in an HTS gen-
geometry optimization, the authors showed the normalized erator with superconducting coils as rotor, by means of T-A
critical current density by varying the stack and layers of the formulation. The model was validated by comparing the res-
biased coil, and the algorithm indicates that fewer layers and ults calculated using A-H formulation on the aspects of mag-
more stack number was preferable. The authors also invest- netic flux density distribution of the whole generator and HTS
igated the fault analysis with the optimized biased coil geo- coils area, as well as instantaneous AC loss of the studied field
metry in SIC-SFCL, in the system where steady current was winding. Using the validated model, the effects of the ferro-
25 Arms and the peak short circuit current was 1.6 kArms . A-V magnetic stator teeth and the shielding layer on AC losses
and T-A formulations were both used and compared to observe were further investigated, by three cases. Conclusions were

9
Supercond. Sci. Technol. 35 (2022) 043003 Topical Review

Figure 6. Magnetic flux density norm in tesla for one pole pair of the three generator designs considered in [33], at the beginning of the
second cycle of simulation. Figure courtesy of Carlos Vargas-Llanos (Karlsruhe Institute of Technology). See [33] for more details.

drawn that the adoption of ferromagnetic stator teeth could sig- permanent magnet rotating past a stationary HTS coated con-
nificantly cause sixth-order harmonic magnetic field, hence, ductor wire in an open-circuit configuration. The benchmark
increased the AC losses in field winding. Addition of a copper was calculated using multiple methods, including H formula-
shielding layer at the rotor side could reduce the AC losses in tion, coupled H-A and T-A formulations, the Minimum Elec-
rotor winding in the steady state and also load variation pro- tromagnetic Entropy Production (MEMEP), Integral Equation
cess, but extra attention should be paid on eddy current loss (IE) and Volume integral equation-based equivalent circuit
generated in the shielding layer itself. (VIE). Excellent qualitative and quantitative agreement was
In [35], the electromagnetic and mechanical properties of a observed between all models for the open circuit equivalent
fully superconducting generator were investigated, which was instantaneous voltage and the cumulative time-averaged equi-
composed of field windings and armature windings made from valent voltage, together with the current density and electric
YBCO superconducting racetrack coils. The moving mesh field distributions within the HTS wire at specific positions
method was incorporated into the 2D T-A formulation model during the magnet transit. A critical analysis and comparison
to simulate the rotating structure of the generator. Virtual tapes with all modelling approaches was presented, regarding key
were introduced to compute the induced current in armature metrics: number of mesh elements in the HTS wire and the
winding, which was further fed back to the T formulation. The entire model, number of degree of freedoms, tolerance settings
model was validated by comparing the calculated results from and the approximate time taken per cycle for each model. The
H-A model on the aspects of radical gap field, armature cur- benchmark provided researchers with a suitable framework to
rent, magnetic flux distribution and total AC losses in the gen- validate, compare and optimize their own approaches for sim-
erator windings. In addition, the stress distribution in the rotor ulating the HTS dynamo.
half coil was shown and analysed. It was concluded that the In [37], design, fabrication, and testing of a non-insulation
AC losses in the armature winding was increased significantly double-pancake racetrack coil for an HTS synchronous motor,
when field currents increased. The body force and stress were which was energized by HTS flux pump, were presented. 2D
concentrated on the boundary of the field winding and arma- T-A formulation model was built to simulate the electromag-
ture winding, and the field winding suffered greater stress than netic distribution of the double pancake coil. The authors pro-
the armature winding due to stronger magnetic field. posed a mechanical way to fix the HTS coil instead of epoxy
In [36], the authors studied the electromagnetic and thermal resin impregnation, due to the concern on different expansion
behaviours of a synchronous generator comprised of HTS and contraction ratios of HTS tape and epoxy resin that might
armature winding by proposing a numerical model which was cause coil damage. Critical current test for the HTS coil was
coupled with three sub-models, including a field model, an presented at 77 K. A charging test with rotating flux pump was
equivalent circuit model and a thermal model. AC loss of the also carried out. The charging test results showed the pump
armature winding under rated operation was calculated. Tran- current got saturated once the output DC voltage was con-
sient electromagnetic-thermal behaviour of the HTS coil under sumed by internal resistance.
a three-phase short circuit fault at the no-load condition was In [38], a superconducting linear induction pump was ana-
analysed. It was observed that the critical current of each turn lysed, using the T-A formulation to evaluate the AC loss char-
of the armature winding was not identical under the fault due acteristics under self-field and an external alternating magnetic
to different ambient magnetic field, and this critical current field. The study showed that the AC loss, when a three-phase
deviation of turns led to different current distribution among alternating current was applied, featured a symmetric distribu-
layers and various temperature rise. Thermal analysis showed tion due to the self-field. The outer coils of the pump gener-
that a thicker copper stabilizer improves the thermal stability ated less AC loss than the ones in the center. When an external
of HTS coil upon the fault. alternating magnetic field was applied, the coils at the edges
In [5], the authors proposed HTS dynamo as a new bench- had the highest losses, due to the diamagnetic properties of
mark problem for the HTS modelling community, consisting a superconductors.

10
Supercond. Sci. Technol. 35 (2022) 043003 Topical Review

Figure 7. (a) Magnetic flux density magnitude and (b) normalized current density at 1 h, the first peak of the ramping cycle. The upper
pancakes are fully penetrated by screening currents. Figure reproduced with permission from [42]. © 2020, IEEE. Reprinted, with
permission, from [42].

4.3. High field magnets formulation, this design was analysed numerically and con-
firmed by experiments. It was demonstrated that the SCF in
Large-scale HTS magnets usually consist of stacked HTS pan- the NS wire was reduced, and it also met the critical current
cake coils. The coils are made of hundreds of turns of HTS requirements for applications such as NMR, MRI, and HEP.
tapes, which due to the high aspect ratio of HTS tapes, will In [42] the screening currents and corresponding hysteresis
result in a large number of elements for the numerical model. losses in the 32 T all-superconducting magnet of the National
Here, the T-A formulation can help to reduce the complexity of High Magnetic Field Laboratory (NHMFL) were investigated.
the simulation by making use of geometrical simplifications. This magnet was tested in 2017 and consisted of HTS insert
Applying the thin strip approximation combined with a multi- and LTS outsert magnets, which produced a total magnetic
scale or homogenisation method to simulate a stack of tapes, field of over 30 T. The challenge was to model the 20 000 turns,
can significantly reduce the number of elements of the geo- which was tackled by using the homogeneous T-A formulation
metry, while keeping the electromagnetic integrity of the HTS (figure 7). A comparison with other numerical models (homo-
coil. It is essential to numerically investigate such large-scale geneous and multi-scale H formulation) showed that it was
systems before they are built in order to evaluate the perform- possible to significantly reduce the computational load of the
ance and potential problems of the magnet. simulation. Therefore, it was possible to model a full-size HTS
In his viewpoint, Ainslie [39] identified the screening insert and include the background field of the LTS outsert mag-
current-induced magnetic field and its associated stress/strain net. The whole simulation was achieved within hours using a
as one of the major challenges for the design of high-field mag- desktop computer.
nets. This phenomenon was investigated numerically, using The T-A formulation was used in [43] for the simulation of
the T-A formulation by several researchers. 40 REBCO pancakes with 12 000 total turns, which produced
The influence of shielding/screening current effect in high a field of 15 T. Due to screen current induced field (SICF),
temperature superconductors was investigated in [40]. The it was shown that the field in the centre of the magnet was
experimental setup consisted of an LTS background magnet reduced significantly. The investigation of the mechanical sta-
providing up to 9 T parallel field and an HTS insert coil. The bility and safety of the high field magnet showed the risk of
measured data for the hoop strain of a 10 mm REBCO tape high residual fields at the top and bottom of the magnet coils.
test sample was compared with numerical results using the T-A Since the top and bottom coils had a higher radial field than
formulation. the centre coils, the induced screening currents and magnet-
The issue of distorted magnetic fields produced by an ization loss were increased. Furthermore, during ramp cycles,
induced persistent screening current in the tape under time- the magnetization loss was increased due to the screening cur-
varying conditions was studied in [41]. A novel REBCO con- rents generated by the superconductor. By using an equivalent
ductor design, consisting of narrow-stacked (NS) wire, bund- turn model with the T-A formulation, the number of turns that
ling 1 mm wide tapes was used for this study. This design needed to be modelled was reduced, while keeping the same
makes use of the fundamental property that a smaller width electromagnetic performance. This approach was validated by
features a smaller screen-current field (SCF). Using the T-A simulations using the H formulation.

11
Supercond. Sci. Technol. 35 (2022) 043003 Topical Review

In [44] a HTS synchronous driving system for an EDS formulation is easier to use, especially for researchers who are
(electrodynamic suspension) train was studied. The authors not experts in the software.
computed the propulsion force under different operating con- This was shown for a wide range of different power applic-
ditions analytically and numerically using the T-A formulation ations from high power cables, fault current limiter, elec-
for a REBCO magnet. It was shown that due to the high excit- trical machines to high field magnets. Especially in large-scale
ation speed of transport current, the critical current of the coil applications, where the number of HTS tapes is large, the
tends to be underestimated. T-A formulation has the efficiency to solve these simulations
The T-A formulation was used in [38], to validate the within reasonable time.
developed self-consistent model, in order to estimate the crit- Expanding on the conventional T-A formulation for tapes
ical current of a racetrack REBCO magnet. The evaluated rel- and making use of simplifications such as the multi-scale or
ative error between the models was 0.48% for the whole coil the homogenisation method can further improve the com-
and an identical magnetic field distribution. putation time. These methods are validated for 2D and 3D
A persistent-current superconducting magnet system with using experiments as well as numerical comparison with the
solid nitrogen cooling for magnetic levitation trains was pro- H formulation.
posed in [45]. The superconducting magnets need to cover the There are limitations to the T-A formulation and the use
vibratory motion range when Maglevs are operated at high should be evaluated for each problem individually. Practically
speeds >600 km h−1 . The field was designed to be >0.8 T all T-A modelling strategies are limited to cases where the thin
using no-insulation 2G HTS wires. These wires feature high sheet approximation of the superconducting layer is meaning-
in-field current, enhanced self-protective stability and volu- ful [9]. This means that although this modelling technique has
metric compactness. The T-A formulation was used to model a a great potential to simulate HTS application simply and fast,
300 turn double pancake coil and evaluate its electromagnetic there are applications where it is still required to use the H
performance and demonstrate a 2G HTS magnet system for formulation, such as wires with different geometries (MgB2 ).
maglev applications.
6. Conclusion

5. Discussion In this paper the modelling of high temperature superconduct-

ors using the T-A formulation is reviewed. In recent years this
The T-A formulation gained popularity in the last years due to formulation gained popularity due to its simple implementa-
two main factors. First, the T-A formulation is a more efficient tion and efficient computation speed. By adapting the multi-
numerical method than the widely used H formulation, due to scale and homogenisation method to the T-A formulation, it
the simultaneous computation of two state variables and the is now possible to model large-scale applications with large
continuous exchange of the current density J and magnetic numbers of turns in excess of the capabilities of other formu-
flux density B. The other is the convenient implementation lations, such as the H formulation. This makes the T-A formu-
and user friendliness in FEM software like COMSOL, which lation even more promising for future modelling of HTS.
is mainly used to run simulations for HTS applications. As With these improvements, the T-A formulation was used
pointed out by the authors in [9], setting up a model using the to simulate the superconducting behaviour in a multitude of
H formulation, can be quite time consuming since the trans- applications: high field cables such as Roebel cables, CORC®
port current is imposed via integral constraints, which need to cables, stacked cables and cables with a twisted structure; fault
be defined for each tape individually. In comparison, the T-A current limiters, electrical machines, and high field magnets.
formulation allows imposing the transport current via Dirich- For these applications the evaluation of AC loss as well as
let boundary conditions at the edges of the tape, which makes screening currents are of particular interest to researchers.
it very easy to build. In the case of simultaneous multi-scale The main implementation of the T-A formulation is done
models, the transport current just needs to be applied to the in COMSOL, a multi-physics software where the electromag-
analysed tapes. Although this does not simplify the model netics can be coupled with a thermal model. Another software
compared to the full T-A model, it is easier to implement that is used to study HTS tapes using the T-A formulation is
than the same approach using the H formulation. The easiest GetDP.
models to build are the homogeneous models, as the simpli- The T-A formulation has shown to be a reliable alternative
fication of a stack of tapes into a homogenised bulk reduces to the H formulation. It is especially desired for large-scale
the amount of boundary conditions especially for large num- applications since the computational speed can be increased
bers of tapes. In the T-A formulation, the transport current is tremendously.
still imposed via Dirichlet boundary conditions at the edges, Here we reviewed the basic concept and implementation of
but especially for the H formulation, the integral constraint the T-A formulation as well as the investigation and application
only needs to be defined for each bulk segment, rather than to real problems. Although the T-A formulation is suitable and
each individual tape. Nevertheless, in order to prevent current efficient for most problems, there are some cases where it is
sharing, air gaps need to be implemented between the bulk still necessary to use alternative formulations (H, T-Ω, H-A,
segments when using the H formulation, which complicates A-V, H-φ, MEMEP, P-model, etc) and therefore, the choice of
the geometrical design. Overall, it can be said that the T-A the numerical model needs to be done on a case-by-case basis.

12
Supercond. Sci. Technol. 35 (2022) 043003 Topical Review

Future works will extend the use of the T-A formulation temperature superconductor systems based on H and T-A
in 3D to geometries that cannot be accurately represented in formulations Supercond. Sci. Technol. 34 044002
2D. This is especially important for twisted cable and coil con- [10] Wang Y et al 2020 AC losses of a like-corc conductor using
accelerated 3D T-A model Physica C 579 1353770
figurations, where the homogenous T-A formulation will be a [11] Huang X, Huang Z, Xu X, Li W and Jin Z 2019 Effective 3D
useful tool to enable numerical investigations. Furthermore, FEM for large-scale high temperature superconducting
coupling of the electromagnetic model with other physical racetrack coil Prog. Supercond. Cryog. 21 32–37
models i.e. heat transfer or mechanical models are of interest [12] Mu S et al 2021 Calculation and comparison of HTS
when investigating HTS applications. electromagnetic characteristics with different models
Physica C 589 1353930
[13] Yan Y, Qu T and Grilli F 2021 Numerical modeling of AC loss
in HTS coated conductors and roebel cable using T-A
Data availability statement
formulation and comparison with H formulation IEEE
Access 9 49649
All data that support the findings of this study are included [14] Xu X et al 2021 3D finite element modelling on racetrack coils
within the article (and any supplementary files). using the homogeneous T-A formulation Cryogenics
119 103366
[15] Liu Y et al 2019 Comparison of 2D simulation models to
estimate the critical current of a coated superconducting
Acknowledgment
coil Supercond. Sci. Technol. 32 014001
[16] Berrospe-Juarez E, Trillaud F, Zermeño V M R and Grilli F
This work was supported by the COST Action CA19108 2020 Screening current-induced field and field drift study in
‘High-Temperature SuperConductivity for AcceLerating the HTS coils using T-A homogenous model J. Phys.: Conf.
Energy Transition’ (Hi-SCALE). Ser. 1559 012128
[17] Li Y et al 2020 Screening-current-induced strain gradient on
REBCO conductor: an experimental and analytical study
ORCID iDs with small coils wound with monofilament and striated
multifilament REBCO tapes IEEE Trans. Appl. Supercond.
30 3–7
Wenjuan Song  https://orcid.org/0000-0001-8003-7038 [18] Grilli F et al 2021 Electromagnetic modeling of
Min Zhang  https://orcid.org/0000-0003-4296-7730 superconductors with commercial software: possibilities
Francesco Grilli  https://orcid.org/0000-0003-0108-7235 with two vector potential-based formulations IEEE Trans.
Appl. Supercond. 31 1–9
[19] Dular J, Harutyunyan M, Bortot L, Schops S, Vanderheyden B
References and Geuzaine C 2021 On the stability of mixed
finite-element formulations for high-temperature
[1] Grilli F, Pardo E, Stenvall A, Nguyen D N, Yuan W and superconductors IEEE Trans. Appl. Supercond. 31 1–12
Gomory F 2014 Computation of losses in HTS under the [20] Yan Y et al 2019 Experimental and numerical study on the
action of varying magnetic fields and currents IEEE Trans. magnetization process of roebel cable segments IEEE
Appl. Supercond. 24 78–110 Trans. Appl. Supercond. 29 1–5
[2] Shen B, Grilli F and Coombs T 2020 Review of the AC loss [21] Zhu Z, Wang Y, Venuturumilli S, Sheng J, Zhang M and
computation for HTS using H formulation Supercond. Sci. Yuan W 2018 Influence of harmonic current on
Technol. 33 033002 magnetization loss of a triaxial CORC REBCO cable for
[3] Shen B, Grilli F and Coombs T 2020 Overview of hybrid electric aircraft IEEE Trans. Appl. Supercond.
H-formulation: a versatile tool for modeling 28 4–8
electromagnetics in high-temperature superconductor [22] Wang Y, Zheng J, Zhu Z, Zhang M and Yuan W 2019 Quench
applications IEEE Access 8 100403–14 behavior of high-temperature superconductor
[4] Pardo E, Šouc J and Frolek L 2015 Electromagnetic modelling (RE)Ba2Cu3Ox CORC cable J. Phys. D: Appl. Phys. 52
of superconductors with a smooth current-voltage relation: 345303
variational principle and coils from a few turns to large [23] Wang Y, Zhang M, Grilli F, Zhu Z and Yuan W 2019 Study of
magnets Supercond. Sci. Technol. 28 044003 the magnetization loss of CORC® cables using a 3D T-A
[5] Ainslie M, Grilli F, Qu L and Pardo E 2020 A new benchmark formulation Supercond. Sci. Technol. 32 1–21
problem for electromagnetic modelling of superconductors [24] Zhu Z, Wang Y, Xing D, Pei X, Zhang M and Yuan W 2019
: the high- T c superconducting dynamo Supercond. Sci. Quench of a single-layer ReBCO CORC cable with
Technol. 33 105009 non-uniform terminal contact resistance IEEE Trans. Appl.
[6] Zhang H, Zhang M and Yuan W 2017 An efficient 3D finite Supercond. 29 2–6
element method model based on the T-A formulation for [25] Zhu Z 2020 Advanced 3D and 2D Modelling of HTS CORC
superconducting coated conductors Supercond. Sci. Cable Based on the T-A Formulation for the Propulsion
Technol. 30 024005 System of Hybrid-Electric Aircraft (Bath, UK: University of
[7] Liang F et al 2017 A finite element model for simulating Bath)
second generation high temperature superconducting [26] Li X et al 2021 Calculation of CORC cable loss using a
coils/stacks with large number of turns J. Appl. Phys. coupled electromagnetic-thermal T-A formulation model
122 043903 IEEE Trans. Appl. Supercond. 8223 1
[8] Berrospe-Juarez E, Zermeño V M R, Trillaud F and Grilli F [27] Fu S et al 2018 Numerical study on AC loss properties of HTS
2019 Real-time simulation of large-scale HTS systems: cable consisting of YBCO coated conductor for HTS power
multi-scale and homogeneous models using the T-A devices IEEE Trans. Appl. Supercond. 28 4802005
formulation Supercond. Sci. Technol. 32 065003 [28] Pi W et al 2020 Numerical simulation and experiment of AC
[9] Berrospe Juarez E, Trillaud F, Zermeno V and Grilli F 2021 loss in quasi-isotropic superconducting strand IEEE Trans.
Advanced electromagnetic modeling of large-scale high Appl. Supercond. 30 1–5

13
Supercond. Sci. Technol. 35 (2022) 043003 Topical Review

[29] De Marzi G, Celentano G, Augieri A, Marchetti M and [37] Gao Y et al 2020 Design, fabrication, and testing of a YBCO
Vannozzi A 2021 Experimental and numerical studies on racetrack coil for an HTS synchronous motor with HTS flux
current distribution in stacks of HTS tapes for pump IEEE Trans. Appl. Supercond. 30 1–5
cable-in-conduit-conductors Supercond. Sci. Technol. [38] Gong T et al 2020 An improved self-consistent model and its
34 035016 application to estimate the critical current of REBCO
[30] Pi W et al 2020 Transport AC losses analysis of twisted magnet IEEE Trans. Appl. Supercond. 30 1–6
quasi-isotropic conductor Physica C 576 1353707 [39] Ainslie M D 2020 Guiding future high-field magnet design by
[31] Dos Santos G, Sass F, Sotelo G G, Fajoni F, Baldan C A and better predicting and mitigating screening current-induced
Ruppert E 2021 Multi-objective optimization for the stress and strain Supercond. Sci. Technol. 33 3–6
superconducting bias coil of a saturated iron core fault [40] Yan Y, Xin C, Guan M, Liu H, Tan Y and Qu T 2020
current limiter using the T-A formulation Supercond. Sci. Screening current effect on the stress and strain distribution
Technol. 34 025012 in REBCO high-field magnets: experimental verification
[32] Benkel T et al 2020 T-A-formulation to model electrical and numerical analysis Supercond. Sci. Technol. 33 05LT02
machines with HTS coated conductor coils IEEE Trans. [41] Wang M et al 2019 A novel REBCO conductor design to
Appl. Supercond. 30 1–7 reduce screening-current field in REBCO magnets Phys.
[33] Vargas-Llanos C R, Lengsfeld S and Grilli F 2020 T-A Scr. 94 0–7
formulation for the design and AC loss calculation of a [42] Berrospe-Juarez E, Trillaud F, Zermeno V M R, Grilli F,
superconducting generator for a 10 MW wind turbine IEEE Weijers H W and Bird M D 2020 Screening currents and
Access 8 208767–78 hysteresis losses in the REBCO insert of the 32 T
[34] Huang X, Huang Z, Xu X, Wang L, Li W and Jin Z 2020 A all-superconducting magnet using T-A homogenous model
fully coupled numerical method for coated conductor HTS IEEE Trans. Appl. Supercond. 30 1–5
coils in HTS generators IEEE Trans. Appl. Supercond. [43] Wang Y, Bai H, Li J, Zhang M and Yuan W 2020
30 11–16 Electromagnetic modelling using T-A formulation for
[35] Yang Y, Yong H, Zhang X and Zhou Y 2020 Numerical high-temperature superconductor (RE)Ba2Cu3Oxhigh field
simulation of superconducting generator based on the T-A magnets High Voltage 5 218–26
formulation IEEE Trans. Appl. Supercond. [44] Gong T et al 2019 Calculation and optimization of propulsion
30 1–11 force of a real-scale REBCO magnet for EDS train IEEE
[36] Huang X, Huang Z, Xu X, Li W and Jin Z 2020 An Trans. Appl. Supercond. 29 1–6
electromagnetic-thermal coupling numerical study of the [45] Dong F, Huang Z, Hao L, Xu X, Jin Z and Shao N 2019 An
synchronous generator with second-generation on-board 2G HTS magnets system with cooling-power-free
high-temperature superconducting armatures Appl. Sci. and persistent-current operation for ultrahigh speed
10 5228 superconducting maglevs Sci. Rep. 9 1–12

14