This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication. The final version of record is available at http://dx.doi.org/10.1109/TASC.2017.2766132 1 Strain Distribution in REBCO Coated Conductors Bent with the Constant-Perimeter Geometry Xiaorong Wang, Diego Arbelaez, Shlomo Caspi, Soren O. Prestemon, GianLuca Sabbi, Tengming Shen Abstract—Cable and magnet applications require bending REBa2 Cu3 O7−δ (REBCO, RE = Rare Earth) tapes around a former to carry high current or generate specific magnetic fields. With a high aspect ratio, REBCO tapes favor the bending along their broad surfaces (easy way) than their thin edges (hard way). The easy-way bending forms can be effectively determined by the constant-perimeter method that was developed in the 1970s to fabricate accelerator magnets with flat thin conductors. The method, however, does not consider the strain distribution in the REBCO layer that can result from bending. Therefore, the REBCO layer can be over-strained and damaged even if it is bent in an easy way as determined by the constant-perimeter method. To address this issue, we developed a numerical approach to determine the strain in the REBCO layer using the local curvatures of the tape neutral plane. Two orthogonal strain components are determined: the axial component along the tape length and the transverse component along the tape width. These two components can be used to determine the conductor critical current after bending. The approach is demonstrated with four examples relevant for applications: a helical form for cables, forms for canted cos θ dipole and quadrupole magnets, and a form for the coil end design. The approach allows to optimize the design of REBCO cables and magnets based on the constantperimeter geometry and to reduce the strain-induced critical current degradation. Index Terms—REBCO Coated conductors, rectifying developable surface, bending strain I. I NTRODUCTION ENDING REBa2 Cu3 O7−δ (REBCO, RE = Rare Earth) coated conductors around a former is generally required to make REBCO cables and magnets. The flat thin tape geometry, however, makes the bending along the thin edge of the REBCO tapes more difficult than along their broad surface. Therefore bending REBCO tapes is more constrained than bending round wires [1]. In addition, knowing the shape of the bent REBCO tapes allows to properly support the tapes against the electromagnetic force. In the 1970s, the flat thin Nb3 Sn tapes faced similar issues for the development of beam steering magnets and an effective solution called the constantperimeter method was developed [2–5]. The constant-perimeter method models the tape conductor as a developable surface that can be unfolded to a plane without stretching or tearing [4]. Thus, it approximates an easy-way bend of the tape conductor that can facilitate the cable and magnet applications. Besides the most straightforward example of single pancake coils, the method has B The authors are with Lawrence Berkeley National Laboratory, Berkeley, CA 94720, U.S.A. E-mail: xrwang@lbl.gov This work was supported by the Director, Office of Science, Office of High Energy Physics and Office of Fusion Energy Sciences, of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231. been successfully applied to the design of accelerator magnets based on the high-temperature superconducting tapes, such as a quadrupole magnet wound with Bi-2223 tapes [6, 7] and cos θ dipole magnets made of REBCO tapes [8–10]. A coil end design that is useful for saddle, block and cos θ coils using thin tape conductors was also proposed in [11]. The constant-perimeter method was extended to model the bending of Rutherford cables [12]. The method was also implemented in ROXIE, a software package that is widely used for accelerator magnet design [13]. In addition to accelerator magnets, the method was also used for the design of fusion magnets [14]. Although the constant-perimeter method enables an easy way bend of REBCO tapes, it does not address the strain distribution in the REBCO layer because the tape thickness is not considered. For most commercial conductors available today, the REBCO layer will strain during bending because it is off the neutral plane of the tape. Excessive strain can permanently degrade the conductor critical current (Ic ) by damaging the brittle REBCO layer. Therefore, understanding the strain distribution in the REBCO layer is critical to minimize the Ic degradation for REBCO cables and magnets [15, 16]. Few studies have been reported on the strain distribution in the REBCO layer when the tape is bent in the constantperimeter geometry. van der Laan studied the strain along the REBCO tape length when the tapes are helically wound around a former [17]. A similar study was reported by Zhu et al. based on a different analysis method [18]. Amemiya et al. showed that significant edge-wise strain appears if the REBCO tape in the winding deviates from the constantperimeter geometry [9]. Although these studies clarify the strain distribution for specific cases, strain distribution for general bending cases based on the constant-perimeter geometry remains to be addressed. In this paper, we present a curvature-based approach to determine the local strain distribution of the REBCO layer bent with the constant-perimeter geometry. Four cases relevant for cable and magnet applications are analyzed: 1) a helical winding, 2) a canted cos θ (CCT) dipole winding, 3) a CCT quadrupole winding, and 4) a coil end design proposed in [11]. The amplitude and distribution of two orthogonal strain components, one along the conductor length and the other along the conductor width, are determined. We show how the strain depends on the cable and magnet design parameters. With the presented method, one can optimize the design for REBCO cables and magnets to minimize the strain-induced Ic degradation. Section II briefly reviews the constant-perimeter method and Copyright (c) 2017 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org. This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication. The final version of record is available at http://dx.doi.org/10.1109/TASC.2017.2766132 2 how the shape of a bent tape is determined. In section III, we present the method to determine the local strain distribution based on the curvature of the neutral plane. The strain distributions for several practical cases are analyzed in section IV as examples. Finally, we discuss the implications of the results and limitation of the presented method. II. A PPROXIMATE REBCO COATED CONDUCTORS WITH RECTIFYING DEVELOPABLE SURFACES We first review the constant-perimeter method based on the concept of rectifying developable surface [4]. Our goal is to determine the developable surface that approximates the shape of the bent tape. The tape is neither extendable nor stretchable, similar to a piece of paper. This property implies the isometric bending that preserves the length of every arc and the angle between every pair of intersecting curves on the surface. Only a developable surface can be isometrically mapped to a plane [19]. Fig. 1 illustrates a tape bending around a cylinder as an example of the rectifying developable surface. After bending, the center line of the tape becomes a space curve, X. For each point along X, a trihedron of orthogonal unit vectors can be established: the tangent vector v1 , the principal normal vector v2 and the binormal vector v3 [19]. Fig. 1. (a) A REBCO tape on a plane before bending. The axial and transverse directions are defined. (b) The tape around a cylinder as a rectifying developable surface. The white dashed line is the center line of the tape and the base curve X. Along X, a moving trihedron of unit vectors can be established: the tangent vector v1 , normal vector v2 (pointing toward the axis of the cylinder), and binormal vector v3 . r is the unit Darboux vector in the v1 -v3 plane. The transport current flows along the axial direction. The 3D interactive version of the figure can be viewed in the attached pdf file with Acrobat Reader R . The rectifying developable surface resulting from the space curve X is given by the vector Y = X + v r, (1) where the vector r and scalar v are to be determined based on X, the base curve for the surface Y. The unit vector r is given by the tangent and binormal vectors of X (Fig. 1), i.e., τ v1 + κv3 , r= √ τ 2 + κ2 (2) where τ is the torsion and κ is the curvature along the base curve X. The torsion and curvature can be determined once the base curve is given (Appendix A). The vector τ v1 + κv3 in (2) is also called the Darboux vector [19]. Equation (2) shows that the developable surface is given by a generator vr that is parallel to the Darboux vector along the base curve X. Next, we determine the scalar v which is related to the tape width. We use the property of isometric transformation (length and angle are preserved) [20]. The unit Darboux vector r is in the plane of v1 and v3 because τ and κ are scalars. Since v1 and v3 are orthogonal unit vectors, the angle β (Fig. 1) between r and v1 is given by κ (3) β = tan−1 . τ From the isometric bending, we have g , (4) v= sin β where g ranges between ±w and 2w is the conductor width (Fig. 1). The generator vr is now determined. With the base curve X, we can now determine the rectifying developable surface Y corresponding to a rectangular tape with zero thickness according to (1). III. D ETERMINE LOCAL ELASTIC STRAIN IN REBCO LAYER BASED ON CURVATURE The REBCO layer experiences strain when it is off the neutral plane of the bent tape (Fig. 2). We treat the conductor as a thin bending beam to estimate the local strain distribution in the REBCO layer [15, 21, 22]. Elastic strains are considered such that the deformations of the conductor and changes of tape geometry are neglected. The Poisson effect is also neglected. Fig. 2. Sketch of the cross section for a typical commercial REBCO tape (not to scale). The width of the tape is on the order of 1 mm. The thickness of the tape is on the order of 10 µm. The tape has Cu stabilizers of the same thickness on the top and bottom. The neutral plane of the tape lies at the center of the substrate. The distance between the REBCO layer and the neutral axis, d, is half of the substrate thickness. The strain in the REBCO layer is determined by the local curvature of the neutral plane κ and the distance between the REBCO layer and the neutral plane d, i.e., = dκ = d , ρ (5) where ρ is the radius of the curvature (ρ = κ−1 ). For a given conductor architecture, d is assumed to be constant during bending. The rectifying developable surface determined by the base curve X (section II) is the neutral plane because by definition, the developable surface preserves the length of any two points on the surface and hence experiences zero strain. Copyright (c) 2017 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org. This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication. The final version of record is available at http://dx.doi.org/10.1109/TASC.2017.2766132 3 We focus on two orthogonal strain components on the REBCO layer. The first is an axial component along the conductor length and parallel to the direction of transport current (Fig. 1(a)). Its impact on the conductor critical current, in particular for the tensile strain, has been extensively studied. The irreversible tensile strain limit, beyond which the conductor Ic permanently degrades, ranges from 0.4% to 0.7% for recent commercial conductors with Hastelloy substrates [16, 22–24]. For the compressive axial strain, a large intrinsic and reversible effect on conductor Ic has been observed [25]. An axial compressive strain between 0.5% and 1.0% can reduce the conductor Ic for 20% [23, 26, 27]. The second component is the transverse strain along the conductor width. An example of the transverse bend is given by the cross sectional picture of the Conductor on Round Core (CORC R ) cable (Fig. 8 in [26]). The transverse strain and its impact on conductor Ic is less known compared to the axial strain. To determine these two strain components, we first determine the curvatures of the axial and transverse lines on the neutral plane of the tape. The curvature κ of a curve on a surface can be determined by the normal curvature κn and geodesic curvature κg according to κ2 = κ2n + κ2g [19]. The axial and transverse straight lines on a plane tape become the geodesics on the developable surfaces after bending. By definition, the geodesics have zero κg . Therefore, the curvatures for the axial and transverse lines on the neutral plane of a bent tape are given by the their normal curvatures, i.e., κ = κn . The normal curvature κn can be determined analytically based on the principal curvatures κ1 and κ2 according to Euler’s Theorem [19]. Developable surfaces feature κ1 = 0 and κ2 can be determined by the mean curvature based on the first and second fundamental forms [19]. The analytic approach is convenient for special cases, e.g., a circular helix, but becomes less straightforward for more general space curves. Here we present a numerical approach to determine the normal curvature κn . The axial and transverse lines on the neutral plane are represented by discrete space curves composed of vertexes. The κn of the discrete space curve can be determined by the vertex osculating circles passing three consecutive vertexes along the discrete space curve. We choose a uniform mesh for the plane tape and determine the corresponding coordinates for each vertex after bending. The local curvature at each vertex can then be determined and gives the distribution of the strains according to (5). More details are explained in Appendix B. IV. C ASE STUDIES RELEVANT FOR CABLE AND MAGNET APPLICATIONS Using the presented approach, we analyze the strain distribution in REBCO layer for several cases that are relevant for cable and magnet applications. These cases are arranged with increasing complexity of the strain distribution. For all cases, the REBCO layer faces the former and is in compression after bending. A. Helically wound REBCO tapes The configuration of a helical tape is used for CORC R [17, 26, 28, 29] and other round cables [30, 31]. The base curve for this configuration is a circular helix X = (r cos t, r sin t, p t), (6) where the radius r is the sum of the former radius and half of the tape thickness; t is the parameter in radian and 2πp is the pitch length. Fig. 3 shows the resulting rectifying developable surface. Fig. 3. A tape bent in a helical form around a cylindrical former. It is generated from a circular helix given by (6). The circular helix features a constant curvature according to (20), i.e., r κ= 2 . (7) r + p2 All the axial lines are in parallel with the base curve and they have the same constant curvature as the base curve. If α is the winding angle between the base curve and rotation axis of the cylindrical former (Fig. 3), then p = r cot α and (7) becomes κaxial = sin2 α . r (8) Similarly, all the transverse lines have the same curvature κtransverse = cos2 α . r (9) The axial and transverse strain on the REBCO layer are then given by d sin2 α , r d cos2 α transverse = . r axial = (10a) (10b) The strain distribution calculated using the numerical method presented in section III is consistent with (10). Fig. 4 shows the axial strain in the REBCO layer as a function of the substrate thickness for a 1 mm diameter former. For a fixed winding angle, the axial strain decreases with thinner substrates (smaller d). For a given substrate thickness, smaller winding angle (longer pitch) reduces the axial strain but increases the transverse strain. To limit the compressive axial strain within 1% for a 1 mm diameter former, the substrate should be thinner than 20 µm for α = 45◦ . Alternatively, one could wind tapes with a substrate thicker than 20 µm with α < 30◦ . Copyright (c) 2017 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org. This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication. The final version of record is available at http://dx.doi.org/10.1109/TASC.2017.2766132 4 α=30˚ -0.75 45˚ -1.00 60˚ -1.25 90˚ -1.50 0 10 20 30 40 Substrate thickness (micron) 50 Fig. 4. The axial strain in the REBCO layer as a function of substrate thickness and winding angles for a 1 mm diameter former. According to (10), the lines for α = 30◦ , 45◦ and 60◦ also give the transverse strains for α = 60◦ , 45◦ and 30◦ , respectively. B. A canted cos θ dipole winding Canted cos θ (CCT) winding configuration is a concept of generating multipole magnetic fields using helical current paths for beam steering magnets. Originally proposed by Meyer and Flasck in 1969 [32], the concept has recently gained interest for various accelerator magnet applications [33–38]. The base curve for a CCT dipole winding can be given by r sin t X = r cos t, r sin t, + pt , (11) tan α where r is the radius of the conductor base curve; α is the tilt angle between 0 and π/2. The pitch length is given by 2w + wrib , (12) sin α where 2w is the tape width and wrib is the rib width at t = 0. More details, e.g., on the ribs, can be found in [36]. Fig. 5 shows a tape forming one turn of CCT dipole winding. 2πp = Fig. 5. A tape forms a CCT dipole winding. It is generated from the base curve given by (11). The peak axial strain occurs at the poles at t = 0.5π and 1.5π. The attached pdf file gives the 3D interactive version of the figure that can be viewed with Acrobat Reader R . The winding scheme is similar to a pancake coil except that the base curve approaches an ellipse rather than a circle. Therefore, one expects that the CCT dipole winding leads to mostly axial strain with minimum transverse strain in the REBCO layer. Fig. 6 shows the calculated strain distribution in the REBCO layer using the method presented in section III. We consider (a) 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.03 -0.04 -0.05 -0.06 -0.07 -0.08 -0.09 -0.10 -0.11 -0.12 -0.13 (b) 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 -0.4 -0.5 0 0.5π Tape length in terms of parameter t π 0.0e+00 -1.0e-03 -2.0e-03 -3.0e-03 -4.0e-03 -5.0e-03 -6.0e-03 -7.0e-03 -8.0e-03 -9.0e-03 -1.0e-02 0 0.5π Tape length in terms of parameter t Transverse strain (%) -0.50 Normalized tape width Axial strain (%) -0.25 Normalized tape width 1.0 mm diameter former Axial strain (%) the following design parameters: r = 32 mm, α = 40◦ , p = 1.05 mm, 2w = 4 mm and d = 25 µm (a 50 µm thick substrate). Due to the symmetry, half turn of the tape is shown with t ranging from 0 to π. 0.00 π Fig. 6. Strain distribution in the REBCO layer for a CCT dipole turn: (a) axial, (b) transverse. r = 32 mm, α = 40◦ , p = 1.05 mm, 2w = 4 mm, d = 25 µm. The axial strain dominates, as expected. The peak axial strains occur at the poles of the winding (t = π/2 and 3π/2) with the smallest bending radii (Fig. 5). Negligible transverse strain occurs along the base curve of the tape due to the nonzero pitch and the resulting torsion of the base curve (Fig. 6). Longer pitches (larger p) will increase the torsion and hence the transverse strain. For instance, doubling the p will increase the magnitude of the peak transverse strain from 0.01% to 0.03% which is still negligible. Longer pitches also reduce the effective current density for field generation, undesirable for high-field CCT accelerator magnet applications. So we will focus on shorter pitches (p . 2w) for the dipole case here and the quadrupole case in the next section. The peak axial strain can be approximated analytically when p approaches zero to provide quick feedback for the CCT dipole design based on the REBCO tapes. With zero p, the base curve becomes an ellipse with no torsion. The ellipse has a major axis of 2r/ sin α and a minor axis of 2r. Therefore, the curvature along the ellipse satisfies 1 sin2 α ≤ κellipse ≤ (13) r r sin α according to (21). Thus, the magnitude of the maximum axial strain for a CCT dipole winding can be approximated as [33] axial,max = |d| . r sin α Copyright (c) 2017 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org. (14) This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication. The final version of record is available at http://dx.doi.org/10.1109/TASC.2017.2766132 5 1.00 30 µm 10 µm 5 TF 0.85 0.80 0.75 10 15 20 25 30 35 40 450.70 α (degree) Fig. 7. The peak axial strain in the REBCO layer as a function of tilt angle and conductor substrate thickness (2d) for a CCT dipole turn. r = 32 mm. The secondary y-axis gives the transfer function (TF) in dashed line for the dipole field normalized to that of α = 0. Fig. 7 highlights the direct correlation between the peak axial strain and magnet performance that is enabled by the presented approach. For instance, even with a 50 µm thick substrate, a high dipole transfer function of 0.95 can be reached with the peak axial compressive strain around 0.3% in the REBCO layer for r = 32 mm. This strain level is lower than the 0.5% to 1% level where the conductor Ic can reduce by 20% [23, 26, 27]. Larger magnet apertures will further reduce the axial strain for the same α as indicated by (14). However, thinner tapes are still preferred because they can not only enable smaller α but also increase the engineering current density of the winding pack, both benefiting the field generation. C. A canted cos θ quadrupole winding The base curve for a CCT quadrupole winding [37, 39–41] can be given by r sin 2t X = r cos t, r sin t, + pt , (15) 2 tan α where the variables are defined as in the CCT dipole case. Fig. 8 shows a tape forming one turn of CCT quadrupole winding with four poles in a constant-perimeter geometry. Note that the REBCO layer is in compression for the whole winding. Fig. 9 shows the calculated strain distribution in the REBCO layer with the following design parameters: α = 40◦ , p = 1.05 mm, 2w = 4 mm and d = 25 µm (a 50 µm thick substrate). We choose r = 32 mm for the direct comparison with the dipole case. As expected, the peak axial strain appears at four poles with the smallest bending radius (t = 0.25π, 0.75π, 1.25π, 1.75π). Fig. 8. A tape forms a CCT quadrupole winding. It is generated from the base curve given in (15). The four poles are located at t = 0.25π, 0.75π, 1.25π and 1.75π. The 3D interactive version of the figure can be viewed in the attached pdf file with Acrobat Reader R . (a) 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 -0.4 -0.5 (b) 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.15 -0.02 -0.04 -0.06 -0.08 -0.10 -0.12 -0.14 -0.16 -0.18 -0.20 -0.22 -0.21 -0.20 -0.20 -0.21 -0.15 0 0.25π 0.5π 0.75π Tape length in terms of parameter t Axial strain (%) 50 µm π 0.00 -0.10 -0.20 -0.30 -0.40 -0.50 0 0.5π Tape length in terms of parameter t Transverse strain (%) 0.90 Normalized tape width 0.95 Normalized tape width 0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 -0.9 -1.0 0 Normalized dipole TF Peak axial strain (%) Based on (14), Fig. 7 plots the peak axial strain as a function of the winding tilt angle α for a CCT dipole coil with r = 32 mm and three substrate thicknesses: 50, 30 and 10 µm. Also shown is the dipole transfer function as a function of α normalized to the value at α = 0 [37]. The strain value with α = 40◦ and d = 25 µm from (14) is consistent with Fig. 6(a) (within 7%). -0.60 π Fig. 9. Strain distribution in the REBCO layer for a CCT quadrupole turn: (a) axial. Equi-strain lines of −0.15%, −0.20% and −0.21% are plotted, (b) transverse. r = 32 mm, α = 40◦ , p = 1.05 mm, 2w = 4 mm and d = 25 µm. In addition, the transverse strain appears when the tape transits between the poles, with the peak magnitude up to 0.6% appearing at the tape edges. The maximum strain alternates between the axial and transverse components (Fig. 9). The magnitude of the peak axial strain can still be correlated with the magnet design parameters and conductor substrate thickness even though the distribution of the axial strain across the conductor width is less uniform (Fig. 9(a)) compared to the dipole case. With a zero pitch length, the maximum axial Copyright (c) 2017 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org. This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication. The final version of record is available at http://dx.doi.org/10.1109/TASC.2017.2766132 6 strain can be approximated by |d| p 1 + 3 cos2 α (16) r sin α according to (22). Based on (16), we expect the ratio of the peak axial strain between the quadrupole and dipole windings to be 1.66 for α = 40◦ , consistent with Figs. 9(a) and 6(a). Comparing (16) and (14), we see that, given the same r, d and α, the peak axial strain in a quadrupole winding given by (15) will be higher than but at most double that in a dipole winding given by (11). Fig. 10 plots the peak axial strain as a function of the tilt angle α and the substrate thickness for CCT quadrupole windings according to (16). Three substrate thicknesses are considered: 50, 30 and 10 µm. Also shown is the normalized quadrupole gradient as a function of the tilt angle. The strain value from (16) is consistent with the peak axial strain in Fig. 9(a) (within 8%) for α = 40◦ and d = 25 µm. D. Saclay coil end To use the flat thin conductors to form the coil ends for cos θ or block magnet designs, Rochepault et al. at CEA Saclay presented a base curve to determine the bending of the conductors consistent with the constant-perimeter method [11]. The base curve is given by π π X = r cos + φ0 cos t , r sin + φ0 cos t , δ sin t , 2 2 (17) where r is the radius of the conductor center line; φ0 is the initial radial angle of the tape with respect to the vertical plane; δ is the length of the coil end; and t is the parameter. We consider two starting positions for the tape: a horizontal and a radial position [11]. Fig. 11 shows an example coil end with a 12 mm wide tape bending around a 50 mm diameter former. The substrate thickness is 50 µm. Here φ0 = π/2 such that the tape starts and ends on the same horizontal plane. Fig. 12 shows the corresponding axial and transverse strain distribution in the REBCO layer. Half turn of the tape is shown with t ranging from 0 to π/2 due to the symmetry. The equistrain lines illustrate the pattern of the axial strain distribution in the REBCO layer. Similar to the CCT quadrupole case, both axial and transverse bending modes appear in the Saclay coil end winding scheme. As opposed to the CCT winding (b) 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.02 -0.08 -0.04 -0.12 -0.06 -0.08 -0.08 -0.10 -0.12 -0.14 -0.15 0 0.25π 0.5π Tape length in terms of parameter t Axial strain (%) Normalized tape width (a) 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.16 -0.18 0.00 -0.05 -0.10 -0.15 -0.20 0 0.25π 0.5π Tape length in terms of parameter t Transverse strain (%) Fig. 10. The peak axial strain in the REBCO layer as a function of tilt angle and conductor substrate thickness (2d) for a CCT quadrupole turn. r = 32 mm. The secondary y-axis gives the transfer function for the quadrupole gradient normalized to that of α = 0. Fig. 11. A tape forms a coil end on a 50 mm aperture mandrel based on [11]. The bending form is generated from the base curve given by (17). The tape is 12 mm wide. r = 31 mm (the radius of the aperture plus half width of the tape), φ0 = π/2, δ = 50 mm, d = 25 µm. The 3D interactive version of the figure can be viewed in the attached pdf file with Acrobat Reader R . Normalized tape width 0.0 1.00 -0.1 10 µm 0.95 -0.2 -0.3 0.90 30 µm -0.4 0.85 -0.5 TF -0.6 50 µm 0.80 -0.7 -0.8 0.75 -0.9 -1.0 0 5 10 15 20 25 30 35 40 450.70 α (degree) Normalized quadrupole TF Peak axial strain (%) axial,max = -0.25 Fig. 12. Strain distribution in the REBCO layer for a Saclay coil end turn bent on a 50 mm aperture mandrel with the horizontal start position: (a) axial, (b) transverse. The tape is 12 mm wide. r = 31 mm, φ0 = π/2, δ = 50 mm, d = 25 µm. cases, the peak axial strain at t = π/2 is no longer uniform across the tape width because of the non-negligible torsion of the base curve at t = π/2. The peak axial strain appears at the center of the outer edge of the tape (t = π/2, normalized width = −0.5). The axial strain also increases at both ends of the inner edge (t = 0 and π/2, normalized width = 0.5). This is related to the boundary conditions for the Saclay coil end design with φ0 = π/2 such that they match the conductors Copyright (c) 2017 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org. This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication. The final version of record is available at http://dx.doi.org/10.1109/TASC.2017.2766132 7 0.00 -0.10 -0.2 -0.15 -0.20 -0.35 Fig. 13. The same coil end design as in Fig. 11 except for φ0 = π/4. The 3D interactive version of the figure can be viewed in the attached pdf file with Acrobat Reader R . Tilting the tape from the horizontal to a radial starting position leads to tighter bending at t = π/2 and less torsion of the tape. As a result, the magnitude of the peak axial strain increases from −0.18% (Fig. 12(a)) to −0.40% (Fig. 14(a)). The magnitude of the peak transverse strain decreases from −0.25% to −0.14% compared to the horizontal starting position. A third initial position as discussed in [11] is that the tape tilts by an additional angle with respect to the radial positioning angle φ0 toward π/2 − φ0 , i.e., from the radial to the horizontal position. In this case, the magnitude of the peak axial strain will decrease while the peak transverse strain will increase, as demonstrated here by Figs. 12 and 14. The detailed strain distribution can be determined by the approach presented here. The design parameter δ affects mostly the axial strain at t = π/2 with negligible impact on the transverse strain: longer coil end (larger δ) makes a tighter bend at t = π/2 and hence a higher axial strain. Using the case shown in Fig. 14 as an example, the magnitude of the peak axial strain increases 20% from 0.40% to 0.60% when δ increases 20% from 50 mm to 75 mm. One can again perform the detailed design optimization on the strain distribution in REBCO layer with the presented approach. V. D ISCUSSION The REBCO tapes can be bent into various easy-way bending forms based on the constant-perimeter method. We used this method to describe the helical winding form that is relevant to the round REBCO wires. Two bending forms were also demonstrated for the canted cos θ accelerator magnet concept. Although examples of bending single tapes are shown -0.25 -0.30 -0.35 0 0.25π 0.5π Tape length in terms of parameter t -0.40 0.00 -0.02 -0.04 -0.06 -0.08 -0.10 Transverse strain (%) (b) 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.1 Axial strain (%) -0.05 Normalized tape width (a) 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3 -0.4 -0.5 Normalized tape width from the magnet straight section. This boundary condition also results in zero transverse strain at the flat ends of the tape. The transverse strain starts increasing in the first and last quarter of the tape where the transition toward the axial folding completes, a pattern similar to the CCT quadrupole case. Fig. 13 shows the second case where the tape starting at π/4 radial position as opposed to the horizontal position to study the impact on the strain distribution. -0.12 0 0.25π 0.5π Tape length in terms of parameter t -0.14 Fig. 14. Strain distribution in the REBCO layer for a Saclay coil end turn bent on a 50 mm aperture mandrel at the 45◦ radial position: (a) axial, (b) transverse. The tape is 12 mm wide. r = 31 mm, φ0 = π/4, δ = 50 mm, d = 25 µm. here, the method is applicable to a stack of tapes and Roebel cables [42]. The constant-perimeter method describes an easy-way bend for the REBCO tapes; but it does not consider the strain distribution in the REBCO layer. We attempted to address this issue with a numerical approach to determine the distribution of the axial and transverse strains in the REBCO layer based on the local curvature of the bent tape. The approach is consistent with earlier analysis on the axial strain for the helical tape case [17, 18] (Appendix C). Our approach can be applied to more general bending forms given the local curvature of the neutral plane. The most effective way to reduce the bending strain in the REBCO layer is to place it on the neutral plane of the tape as demonstrated previously [43, 44]. For helical tapes around a former, the conductor Ic degraded less when the REBCO layer was closer to the tape neutral plane through additional Cu lamination [17] or customized conductor architecture [31]. For a conductor layout shown in Fig. 2, a thinner substrate places the REBCO layer closer to the neutral plane and reduces the minimum bending diameter for round REBCO wires [28, 29]. For conductors with the REBCO layer that is off the tape neutral plane, the winding strain also depends on the design parameters of REBCO cables and magnets, e.g., the former radius and winding angle with respect to the former rotation axis. Formers with small radii and/or small winding angles are desired as they can lead to compact round REBCO wires and compact magnets with higher fields or gradients. However, Copyright (c) 2017 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org. This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication. The final version of record is available at http://dx.doi.org/10.1109/TASC.2017.2766132 8 they also result in higher strains in the REBCO layer (Figs. 7 and 10 in section IV). This design optimization issue can be addressed with our method that correlates cable/magnet designs and strain distribution in the REBCO layer. The presented approach can help ensure proper strain margin to avoid conductor degradation during cable fabrication and magnet winding. Given the strain dependence of Ic , the presented approach allows to determine the expected Ic of the REBCO tape after bending [45–48]. Considering the Ic dependence on both strain and magnetic field, one can determine the expected cable and magnet performance in order to assess if the fabrication process has introduced any conductor degradation [49]. We assumed that the conductor Ic depends primarily on the axial strain component in the REBCO layer. If the assumption holds, small winding angles (longer pitches) are favored for a helical tape configuration as they reduces the axial strain, as shown by (10). This assumption is supported by the experimental results on helically wound tapes in tension [18] and compression [50]. The transverse strains are expected in the bending forms relevant for cable and magnet applications but their impact on the conductor Ic is largely unknown. For REBCO tapes fabricated with the metal-organic chemical-vapor deposition route, both axial and transverse strains lead to the highest strain sensitivity of the Ic through the impact on the critical temperature of REBCO [51–53]. The transverse and axial strain in the helical configuration can be varied with the winding angle (section IV-A). Thus, comparing the Ic of tapes wound with different angles may help to clarify the impact of each strain component on the Ic of REBCO layer. Although several practical cable and magnet schemes were presented here, it can be less trivial to obtain the explicit base curve as required by the constant-perimeter method for some winding cases, e.g., the continuous ramp between the two layers of pancake or racetrack coils where the strain in the tapes can limit the critical current [15, 16, 54]. This can be a challenge to apply the constant-perimeter method. With the approach presented here, however, it is now possible to compare different base curves and determine the optimal one that minimizes the bending strain in the REBCO tapes. The approach also gives the shape of the bent tape which can help the mandrel design to provide proper mechanical support to the tape. VI. C ONCLUSION We present a curvature-based approach to determine the local axial and transverse strain in the REBCO layer after being bent with the constant-perimeter geometry. The presented approach allows to determine the strain distribution in the REBCO layer during the cable and coil winding process against the key design parameters. It offers an effective and simple tool for REBCO cable and magnet designers. Four cases relevant for cable and magnet applications were given as an example for the application of the presented method: the helical winding, the CCT dipole and quadrupole configuration, and the Saclay coil end design. Based on the presented method, one can determine the expected conductor Ic after bending. The transverse strain is not negligible for some cable and magnet winding schemes and its impact on conductor Ic needs to be clarified. The presented method can be further improved by considering the Poisson effect as a next step through, e.g., finite-element analysis. ACKNOWLEDGMENTS We thank the following colleagues for the useful discussion: D. R. Dietderich and S. A. Gourlay of LBNL, D. van der Laan and J. Weiss of Advanced Conductor Technologies LLC, and E. Rochepault of CERN. We thank the referees for suggesting many improvements. A PPENDIX A C URVATURE AND TORSION OF A SPACE CURVE WITH ARBITRARY SPEED Suppose X(t) = (x1 (t), x2 (t), x3 (t)) is a regular space curve with the parameter t and arbitrary speed. The unit tangent vector v1 is defined as v1 = X0 , |X0 | (18) where X0 = dX(t)/dt. The unit binormal vector v3 is given by X0 × X00 v3 = . (19) |X0 × X00 | The unit normal vector v2 is given by v2 = v3 × v1 [55]. The curvature is given by κ= |X0 × X00 | . |X0 |3 (20) When p = 0 in the base curves for CCT dipole (11) or quadrupole (15), we have √ a2 + 1 κdipole (t) = , (21) r(a2 cos2 t + 1)3/2 q 3 a2 sin2 (2 t) + a2 + 1 , (22) κquadrupole (t) = 3/2 r (a2 cos2 (2 t) + 1) where a = cot α. The torsion is given by τ= |X0 X00 X000 | (X0 × X00 ) · (X0 × X00 ) . (23) More details can be found in [19]. A PPENDIX B L OCAL CURVATURE BASED ON VERTEX OSCULATING CIRCLES To determine the local curvature, we first discretize the axial and transverse lines on the neutral plane of the flat tape (Fig. 15(a)). Except for the vertexes on the tape boundaries, we can define a vertex osculating circle for each vertex along the axial and transverse lines [56]. For example, the vertex osculating circle at vertex γk is the circle through γk and its two neighbors γk−1 and γk+1 (Fig. 15(b)). Copyright (c) 2017 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org. This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication. The final version of record is available at http://dx.doi.org/10.1109/TASC.2017.2766132 9 The axial strain in REBCO layer per pitch length is then lREBCO − lneutral axis lneutral axis s (r − d)2 + p2 = −1 r 2 + p2 s d2 − 2rd −1 = 1+ 2 r + p2 = (27) 2 Since | dr2−2rd +p2 | < 1, (27) can be expanded with Taylor series as s 1+ d2 − 2rd 1 d2 − 2rd rd −1 ≈ 1+ −1 ≈ − 2 , (28) r 2 + p2 2 r 2 + p2 r + p2 2 by neglecting the term 2(r2d+p2 ) . Equation (28) is equivalent to (10a). R EFERENCES Fig. 15. (a) The mesh along the axial and transverse lines of the flat tape before bending. The black dots are the vertexes. (b) The local curvature of the bent tape is determined by the osculating circles passing three consecutive vertexes. (c) The osculating circle at vertex γk . After bending, the axial and transverse lines become space curves. The coordinate of each vertex can be determined based on (1). The curvature at vertex γk of the discrete space curve can be determined based on the sine rule, i.e., κk = 2 sin ϕk , ||γk+1 − γk−1 || (24) where ϕk is the turning angle at vertex γk (Fig. 15(c)). Define Tk−1 as the unit vector given by γk−1 and γk and Tk as the unit vector given by γk and γk+1 (Fig. 15(c)), then ϕk = cos−1 (Tk · Tk−1 ). A PPENDIX C A XIAL STRAIN AS DETERMINED BY THE LENGTH CHANGE Here we show for the helical winding, the axial strain defined as the length change over a pitch length [17] is equivalent to the local strain. The length of the neutral axis is lneutral axis = 2π p r 2 + p2 . (25) The length of the REBCO layer in compression is lREBCO = 2π p (r − d)2 + p2 . (26) [1] G. A. Levin and P. N. Barnes, “The integration of YBaCu3 O6+X coated conductors into magnets and rotating machinery,” AIP Conference Proceedings, vol. 824, no. 1, pp. 433–439, 2006. [2] B. Colyer, “The geometry of constant perimeter dipole windings,” Rutherford Laboratory, Tech. Rep. RL-73143, 1973. [3] H. I. Rosten, “The constant perimeter end,” Rutherford Laboratory, Tech. Rep. RL-73-096, 1973. [4] T. C. Randle, “A few mathematical notes on constant perimeter layers in coil windings,” Rutherford Laboratory, Tech. Rep., December 1974, RL-74-165, and references therein. [5] M. N. Wilson, R. B. Hopes, and R. L. Roberts, “A pressure impregnated superconducting dipole,” in Proceedings, 5th International Conference on Magnet Technology (MT-5): Rome, Italy, April 21-25, 1975, 1975, pp. 511– 518. [6] G. Dugan, “Developable constant perimeter surfaces: application to the end design of a tape-wound quadrupole saddle coil,” Cornell University, Tech. Rep., 1998, cBN 98-1. [7] M. Manlief, H. Picard, G. Snitchler, A. Ghosh, M. Harrison, W. Sampson, P. Wanderer, G. Dugan, J. Rogers, A. Temnykh, J. Brandt, P. Limon, and R. Sood, “The performance of Bi-Sr-Ca-Cu-O superconducting quadrupole coils,” IEEE Trans. Appl. Supercond., vol. 9, no. 2, pp. 293–296, 1999. [8] K. Takahashi, N. Amemiya, T. Nakamura, Y. Mori, T. Ogitsu, M. Yoshimoto, I. Watanabe, and T. Yoshiyuki, “Magnetic field design of coil-dominated magnets wound with coated conductors,” IEEE Trans. Appl. Supercond., vol. 22, no. 3, p. 4901705, 2012. [9] N. Amemiya, H. Miyahara, T. Ogitsu, and T. Kurusu, “Design of a cosine-theta dipole magnet wound with coated conductors considering their deformation at coil ends during winding process,” Physics Procedia, vol. 67, Copyright (c) 2017 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org. This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication. The final version of record is available at http://dx.doi.org/10.1109/TASC.2017.2766132 10 pp. 776–780, 2015, Proceedings of the 25th International Cryogenic Engineering Conference and International Cryogenic Materials Conference 2014. [10] K. Koyanagi, S. Takayama, H. Miyazaki, T. Tosaka, K. Tasaki, T. Kurusu, and Y. Ishii, “Development of saddle-shaped coils for accelerator magnets wound with YBCO-coated conductors,” IEEE Trans. Appl. Supercond., vol. 25, no. 3, p. 4003104, June 2015. [11] E. Rochepault, G. Aubert, and P. Vedrine, “Threedimensional magnetic optimization of accelerator magnets using an analytic strip model,” Journal of Applied Physics, vol. 116, no. 2, p. 023910, 2014. [12] J. M. Cook, “Strain energy minimization in SSC magnet winding,” IEEE Transactions on Magnetics, vol. 27, no. 2, pp. 1976–1980, March 1991. [13] S. Russenschuck, Field Computation for Accelerator Magnets: Analytical and Numerical Methods for Electromagnetic Design and Optimization. Weinheim: John Wiley & Sons, 2010, ch. 19. [14] P. L. Walstrom, “Twisted coil geometry in plasma confinement devices,” Journal of Fusion Energy, vol. 6, no. 3, pp. 265–273, 1987. [15] W. D. Markiewicz and C. A. Swenson, “Winding strain analysis for YBCO coated conductors,” Superconductor Science and Technology, vol. 23, no. 4, p. 045017, 2010. [16] C. Barth, P. Komorowski, R. Tediosi, R. Herzog, R. Schneider, and C. Senatore, “A size-constrained 3-T REBCO insert coil for a 21-T LTS magnet: Mechanical investigations, conductor selection, coil design, and first coil tests,” IEEE Trans. Appl. Supercond., vol. 26, no. 4, p. 4600609, 2016. [17] D. C. van der Laan, “YBa2 Cu3 O7−δ coated conductor cabling for low ac-loss and high-field magnet applications,” Superconductor Science and Technology, vol. 22, no. 6, p. 065013, 2009. [18] Y. Zhu, P. Yi, M. Sun, X. Yang, Y. Zhang, and Y. Zhao, “The study of critical current for YBCO tape in distorted bending mode,” Journal of Superconductivity and Novel Magnetism, vol. 28, no. 12, pp. 3519–3523, 2015. [19] E. Kreyszig, Differential geometry. New York: Dover Publications, Inc., 1991. [20] P. Bo and W. Wang, “Geodesic-controlled developable surfaces for modeling paper bending,” Computer Graphics Forum, vol. 26, no. 3, pp. 365–374, 2007. [21] D. C. van der Laan, J. W. Ekin, J. F. Douglas, C. C. Clickner, T. C. Stauffer, and L. F. Goodrich, “Effect of strain, magnetic field and field angle on the critical current density of YBa2 Cu3 O7−δ coated conductors,” Superconductor Science and Technology, vol. 23, no. 7, p. 072001, 2010. [22] S. Otten, A. Kario, A. Kling, and W. Goldacker, “Bending properties of different REBCO coated conductor tapes and Roebel cables at T =77 K,” Superconductor Science and Technology, vol. 29, no. 12, p. 125003, 2016. [23] C. Zhou, K. A. Yagotintsev, P. Gao, T. J. Haugan, D. C. van der Laan, and A. Nijhuis, “Critical current of various REBCO tapes under uniaxial strain,” IEEE Trans. Appl. Supercond., vol. 26, no. 4, p. 8401304, 2016. [24] H.-S. Shin and M. J. Dedicatoria, “Intrinsic strain effect on critical current in Cu-stabilized GdBCO coated conductor tapes with different substrates,” Superconductor Science and Technology, vol. 26, no. 5, p. 055005, 2013. [25] D. C. van der Laan and J. W. Ekin, “Large intrinsic effect of axial strain on the critical current of high-temperature superconductors for electric power applications,” Applied Physics Letter, vol. 90, no. 5, p. 052506, 2007. [26] D. C. van der Laan, X. F. Lu, and L. F. Goodrich, “Compact GdBa2 Cu3 O7−δ coated conductor cables for electric power transmission and magnet applications,” Superconductor Science and Technology, vol. 24, no. 4, p. 042001, 2011. [27] M. Sugano, K. Shikimachi, N. Hirano, and S. Nagaya, “The reversible strain effect on critical current over a wide range of temperatures and magnetic fields for YBCO coated conductors,” Superconductor Science and Technology, vol. 23, no. 8, p. 085013, 2010. [28] D. C. van der Laan, J. D. Weiss, P. Noyes, U. P. Trociewitz, A. Godeke, D. Abraimov, and D. C. Larbalestier, “Record current density of 344 A mm−2 at 4.2 K and 17 T in corc accelerator magnet cables,” Superconductor Science and Technology, vol. 29, no. 5, p. 055009, 2016. [29] J. D. Weiss, T. Mulder, H. J. ten Kate, and D. C. van der Laan, “Introduction of CORC R wires: highly flexible, round high-temperature superconducting wires for magnet and power transmission applications,” Superconductor Science and Technology, vol. 30, no. 1, p. 014002, 2017. [30] W. Luo, S. Kar, A. Xu, X. F. Li, A. Ben Yahia, and V. Selvamanickam, “Fabrication and electromagnetic characterization of ultra-small diameter REBCO wires,” IEEE Trans. Appl. Supercond., vol. 27, no. 4, p. 6602705, June 2017. [31] S. Kar, W. Luo, and V. Selvamanickam, “Ultra-small diameter round REBCO wire with robust mechanical properties,” IEEE Trans. Appl. Supercond., vol. 27, no. 4, p. 6603204, June 2017. [32] D. Meyer and R. Flasck, “A new configuration for a dipole magnet for use in high energy physics applications,” Nuclear Instruments and Methods, vol. 80, no. 2, pp. 339–341, 1970. [33] C. L. Goodzeit, M. J. Ball, and R. B. Meinke, “The double-helix dipole — a novel approach to accelerator magnet design,” IEEE Trans. Appl. Supercond., vol. 13, no. 2, pp. 1365–1368, June 2003. [34] A. Devred, B. Baudouy, D. E. Baynham, T. Boutboul, S. Canfer, M. Chorowski, P. Fabbricatore, S. Farinon, H. Félice, P. Fessia, J. Fydrych, V. Granata, M. Greco, J. Greenhalgh, D. Leroy, P. Loverige, M. Matkowski, G. Michalski, F. Michel, L. R. Oberli, A. den Ouden, D. Pedrini, S. Pietrowicz, J. Polinski, V. Previtali, L. Quettier, D. Richter, J. M. Rifflet, J. Rochford, F. Rondeaux, S. Sanz, C. Scheuerlein, N. Schwerg, S. Sgobba, M. Sorbi, F. Toral-Fernandez, R. van Weelderen, P. Védrine, and G. Volpini, “Overview and status of the Next European Dipole Joint Research Activity,” Superconductor Science and Technology, vol. 19, Copyright (c) 2017 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org. This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication. The final version of record is available at http://dx.doi.org/10.1109/TASC.2017.2766132 11 no. 3, p. S67, 2006. [35] H. Witte, T. Yokoi, S. L. Sheehy, K. Peach, S. Pattalwar, T. Jones, J. Strachan, and N. Bliss, “The advantages and challenges of helical coils for small accelerators —a case study,” IEEE Transactions on Applied Superconductivity, vol. 22, no. 2, p. 4100110, April 2012. [36] S. Caspi, F. Borgnolutti, L. Brouwer, D. Cheng, D. R. Dietderich, H. Felice, A. Godeke, R. Hafalia, M. Martchevskii, S. Prestemon, E. Rochepault, C. Swenson, and X. Wang, “Canted-cosine-theta magnet (CCT) — a concept for high field accelerator magnets,” IEEE Transactions on Applied Superconductivity, vol. 24, no. 3, p. 4001804, June 2014, and references therein. [37] L. N. Brouwer, “Canted-cosine-theta superconducting accelerator magnets for high energy physics and ion beam cancer therapy,” Ph.D. dissertation, University of California, Berkeley, 2015. [38] L. Quéval and R. Gottkehaskamp, “Analytical field calculation of modulated double helical coils,” IEEE Trans. Appl. Supercond., vol. 25, no. 6, p. 4901307, 2015. [39] R. B. Meinke, C. L. Goodzeit, and M. J. Ball, “Modulated double-helix quadrupole magnets,” IEEE Trans. Appl. Supercond., vol. 13, no. 2, pp. 1369–1372, June 2003. [40] S. Caspi, F. Trillaud, A. Godeke, D. R. Dietderich, P. Ferracin, G. Sabbi, C. Giloux, J. G. Perez, and M. Karppinen, “Test of a NbTi superconducting quadrupole magnet based on alternating helical windings,” IEEE Trans. Appl. Supercond., vol. 19, no. 3, pp. 1195–1198, June 2009. [41] F. Bosi, P. Fabbricatore, S. Farinon, U. Gambardella, R. Musenich, R. Marabotto, and E. Paoloni, “Compact superconducting high gradient quadrupole magnets for the interaction regions of high luminosity colliders,” IEEE Trans. Appl. Supercond., vol. 23, no. 3, p. 4001004, June 2013. [42] C. Lorin, M. Durante, P. Fazilleau, G. Kirby, and L. Rossi, “Development of a Roebel-cable-based cos ϑ dipole: Design and windability of magnet ends,” IEEE Trans. Appl. Supercond., vol. 26, no. 3, p. 4003105, April 2016. [43] D. Verebelyi, E. Harley, J. Scudiere, A. Otto, U. Schoop, C. Thieme, M. Rupich, and A. Malozemoff, “Practical neutral-axis conductor geometries for coated conductor composite wire,” Superconductor Science and Technology, vol. 16, pp. 1158–1161, 2003. [44] M. Sugano, S. Machiya, M. Sato, T. Koganezawa, K. Shikimachi, N. Hirano, and S. Nagaya, “Bending strain analysis considering a shift of the neutral axis for YBCO coated conductors with and without a Cu stabilizing layer,” Superconductor Science and Technology, vol. 24, no. 7, p. 075019, 2011. [45] D. C. van der Laan and J. W. Ekin, “Dependence of the critical current of yba2 cu3 o7−δ coated conductors on inplane bending,” Superconductor Science and Technology, vol. 21, no. 11, p. 115002, 2008. [46] M. Takayasu, L. Chiesa, L. Bromberg, and J. V. Minervini, “HTS twisted stacked-tape cable conductor,” Su- perconductor Science and Technology, vol. 25, no. 1, p. 014011, 2012. [47] N. C. Allen, L. Chiesa, and M. Takayasu, “Structural modeling of HTS tapes and cables,” Cryogenics, vol. 80, pp. 405–418, 2016. [48] N. Bykovsky, D. Uglietti, R. Wesche, and P. Bruzzone, “Design optimization of round strands made by twisted stacks of HTS tapes,” IEEE Trans. Appl. Supercond., vol. 26, no. 2, p. 4201207, 2016. [49] X. Wang, D. R. Dietderich, A. Godeke, S. A. Gourlay, M. Marchevsky, S. O. Prestemon, and G. L. Sabbi, “Performance correlation between YBa2 Cu3 O7−δ coils and short samples for coil technology development,” Superconductor Science and Technology, vol. 29, no. 6, p. 065007, 2016. [50] M. Vojenčiak, A. Kario, B. Ringsdorf, R. Nast, D. C. van der Laan, J. Scheiter, A. Jung, B. Runtsch, F. Gömöry, and W. Goldacker, “Magnetization ac loss reduction in HTS CORC R cables made of striated coated conductors,” Superconductor Science and Technology, vol. 28, no. 10, p. 104006, 2015. [51] D. C. van der Laan, D. Abraimov, A. A. Polyanskii, D. C. Larbalestier, J. F. Douglas, R. Semerad, and M. Bauer, “Anisotropic in-plane reversible strain effect in Y0.5 Gd0.5 Ba2 Cu3 O7−δ coated conductors,” Superconductor Science and Technology, vol. 24, no. 11, p. 115010, 2011. [52] M. Sugano, S. Machiya, H. Oguro, M. Sato, T. Koganezawa, T. Watanabe, K. Shikimachi, N. Hirano, S. Nagaya, T. Izumi, and T. Saitoh, “The effect of the 2D internal strain state on the critical current in GdBCO coated conductors,” Superconductor Science and Technology, vol. 25, no. 5, p. 054014, 2012. [53] S. Awaji, T. Suzuki, H. Oguro, K. Watanabe, and K. Matsumoto, “Strain-controlled critical temperature in REBa2 Cu3 Oy -coated conductors,” Scientific Reports, vol. 5, p. 11156, 2015. [54] X. Wang, S. Caspi, D. W. Cheng, D. R. Dietderich, H. Felice, P. Ferracin, A. Godeke, J. M. Joseph, J. Lizarazo, S. O. Prestemon, and G. Sabbi, “Measurements on subscale Y-Ba-Cu-O racetrack coils at 77 K and self-field,” IEEE Trans. Appl. Supercond., vol. 20, pp. 368–372, 2010. [55] A. Gray, E. Abbena, and S. Salamon, Modern Differential Geometry of Curves and Surfaces with Mathematica, 3rd ed. Chapman & Hall/CRC, 2006, ch. 7. [56] A. I. Bobenko, “Geometry II: Discrete differential geometry,” December 2015, http://page.math.tuberlin.de/ bobenko/Lehre/Skripte/DDG Lectures.pdf. Copyright (c) 2017 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.