Engineering Perspective of the CFQS Quasi-Axisymmetric Stellarator and its Upgrade Status Toward the 1 Tesla Experiment

Abstract

The National Institute for Fusion Science in Japan and Southwest Jiatong University in China are proceeding with the CFQS quasi-axisymmetric stellarator project since 2017. CFQS principal physical properties are as follows: toroidal periodic number N_p = 2, aspect ratio A_p = 4, and major radius R₀ = 1 m. Sixteen modular coils are employed to produce the quasi-axisymmetric magnetic field configuration. Its upgrade from the 0.1 T operation as the CFQS-TEST, hereafter called CFQS-T to the 1 T operation as the CFQS is ongoing. In this report, the finalized engineering design, R&D, established essential component manufacturing, and assembly of the CFQS-T are described. Subsequently, the upgrade status toward the CFQS is given, focusing on the further improved design of a support structure that can withstand the complex and enormous electromagnetic forces acting on the modular coils at the 1 T operation. Iterative design and validation using the finite element method revealed a rational support structure with low-cost, good port accessibility, and manufacturability even under a packed configuration of the CFQS main body due to its low-aspect ratio. The engineering knowledge obtained through the CFQS project, particularly regarding the improved support structure design, would contribute to developing next-generation advanced stellarators.

1.  Introduction

A Quasi-axisymmetric (QA) stellarator has been attracting significant attention across the world as one of a number of optimized stellarators, providing both good plasma confinement properties comparable to those of a tokamak, and steady-state operation capability. Its concept was advocated in the mid-1990s [1, 2]. However, no QA stellarator had been realized until the 2010s, although several device projects striving to provide a proof-of-principle demonstration of a QA stellarator configuration were planned from Japan [3, 4], USA [5, 6], and France [7, 8]. In the 2020s, two QA device projects, known as the MUSE and the CFQS have been greatly progressed. The MUSE, having a major radius of R₀ = 0.3 m, two field periods, on-axis magnetic field strength B_t = 0.15 T at the Princeton Plasma Physics Laboratory (PPPL) was completed in 2023. It provides the QA magnetic field configuration by 16 planar, circular toroidal field coils and approximately 10,000 tiny permanent magnets embedded in three-dimensional (3-D) printed holders [9]. The MUSE’s magnetic field topology and error fields have been measured using an electron beam and a fluorescent rod [10].

The CFQS quasi-axisymmetric stellarator project, an international joint research project between the National Institute for Fusion Science (NIFS) in Japan and Southwest Jiatong University (SWJTU) in China, was launched in 2017 [11]. The principal parameters of the CFQS are as follows: Toroidal periodic number N_p = 2, aspect ratio of plasma A_p = 4, major radius of R₀ = 1 m, average plasma minor radius a_p = 0.25 m, and on-axis magnetic field strength B_t = 1 T. The magnetic field configuration is based on CHS-qa, which was designed by NIFS in the 2000s [3, 4]. Previous papers provide the principal physical properties of the CFQS, such as the shape of the magnetic surface and rotational transform etc. [12, 13]. We have been successful in the completion of the device as the CFQS-TEST, hereafter called CFQS-T, capable of 0.1 T long pulse operation with a simplified magnetic field coil support structure, in July, 2024. The measurement of the magnetic field topology in the CFQS-T has been conducted with a mapping system consisting of an electron gun, a 400 mm² fluorescent mesh and a high-sensitivity camera [14, 15]. Subsequently, at the end of August, 2024, the CFQS-T successfully generated the first plasma driven by 2.45 GHz electron cyclotron resonance heating (ECRH). Following the initial experiment series, which ran until May 2025 and yielded several notable achievements, e.g., confirmation of good nested magnetic flux surfaces, and observations of zonal flow-like oscillation and suprathermal electron-driven MHD instabilities, we initiated an upgrade toward the 1 T short pulse operation as the CFQS, an original name of the device with numerous additional support structures, reinforcing the modular coils (MCs) against complex and massive electromagnetic (EM) forces.

The CFQS-T/CFQS is equipped with 16 three-dimensionally twisted MCs of four different shape types to realize a QA magnetic field configuration. Regarding the vacuum vessel (VV), complex geometry along with a plasma shape is adopted as well. Furthermore, the low-aspect ratio makes the main body packed, resulting in a minimum gap of only 13 mm between components. Thus, engineering design and validation with the finite element method (FEM) analysis, as well as the R&D to elucidate the manufacturability of essential components, rigorously carried out [16–22]. However, despite being designed and modified several times so far [16, 19, 20], the new EM force support structure for MCs at the 1 T operation still has engineering concerns and room for improvement in terms of cost, manufacturability, and port accessibility.

This paper gives the finalized engineering design and R&D for the CFQS-T main body, and further improved design of the MC support structure for the 1 T operation as the CFQS first, followed by the established manufacturing procedure of crucial components and assembly of the CFQS-T main body, in more detail than the previous report [14]. In addition, the upgrade status for the 1 T operation of the CFQS is reported. We plan to initiate the first plasma of the CFQS in January, 2027.

We have overcome some technical roadblocks and obtained much invaluable engineering knowledge throughout this construction project. In particular, establishing engineering designs and precise manufacturing methods for complex MCs and the VV, together with a reasonable support structure design, will greatly contribute to the development of a next-generation advanced stellarator [23].

2.  Engineering Design of the CFQS-T/CFQS

2.1  Magnetic field coil configuration

Figure 1 depicts the CFQS-T/CFQS magnetic field coil configuration. It consists of four types, 16 MCs for providing the QA magnetic field configuration, two pairs of poloidal field coils (PFCs) for magnetic axis position control, and 12 toroidal field coils (TFCs) of three different types for rotational transform control. The coils are electrically connected in series for each type, and the current in each coil type is independently controlled to enable the experiment with various magnetic field configurations, e.g., an island bundle divertor configuration [24]. The maximum flat-top duration of MCs for the 0.1 and 1 T operations are 269 and 0.6 s respectively, limited by coil temperature rise. The discharge interval of 600 s is determined by the MC and PFC cooling-down time.

Fig. 1.  CFQS-T/CFQS magnetic field coil configuration. It consists of four types of modular coils (MCs), two pairs of poloidal field coils (PFCs), and three types of toroidal field coils (TFCs).

2.2  Modular coil winding pack

Figure 2 illustrates the design of an MC winding pack. It is composed of water-cooled oxygen-free copper hollow conductors with 72 turns in total. The conductor has an 8.5 × 8.5 mm cross-section, with a hollow diameter of 4 mm. The hollow conductor is insulated by glass-cloth tapes and polyimide ones, half-overlapped. The appropriate width of tapes was investigated to maintain the half-overlapping, even under a complex winding route. The coil-current density for the 1 T operation is 73.8 A mm⁻², which is severe in terms of coil temperature rise, to make the MCs as compact as possible for avoiding interference with other components. The winding method of an MC is a double pancake with three simultaneous windings, which shortens the coil winding time and suppresses the water pressure loss up to 1 MPa. The gross length of the hollow conductors for one MC is approximately 300 m. Three conductors are jointed at the outboard of the MC by copper blocks, as shown in Fig. 2, so that only one coil power supply feeds it. Although fundamental geometry is defined by the center trajectory of an MC, a so-called filament coil calculated with the NESCOIL code [25], the 3-D MC structure is adjusted in terms of manufacturability, the required minimum gap between MCs or an MC and other components, as well as assemblability with an MC support structure [21, 26, 27].

Fig. 2.  Design of a modular coil winding pack. It is a double pancake coil with three simultaneous windings. Three conductors are jointed at the outboard of a modular coil by copper blocks.

2.3  Entire modular coil support structure

Figure 3 summarizes the EM force on each type of MC at the 1 T operation obtained by FEM analysis software ANSYS/Maxwell [19]. The EM force on the inboard is larger than that on the outboard of an MC. The EM force is primarily classified into three directions: vertical, centripetal and toroidal. However, because each MC type has a unique 3-D asymmetric structure, the dominant direction of the EM force differs, as shown in Fig. 3 in red. An appropriate support structure corresponding to the complex EM force on each MC had to be designed.

Fig. 3.  Summary of estimated electromagnetic force on modular coils.

Generally, magnetic field coils for a fusion device are mechanically connected to cancel out a massive EM force. Figure 4 represents two candidates for the CFQS MC support structure for the 1 T operation. In the initial phase, we tried to design the support structure based on a shear-panel/shell type. This structure combines all MCs by using shear-panels or shells, as with the toroidal field coils of a tokamak [28, 29]. It has excellent reliability for coil support because the connection area between the MCs is even wider than that of the cage-type shown in Fig. 4. The Large Helical Device (LHD) at NIFS [30], the National Compact Stellarator eXperiment (NCSX) at PPPL [5], and the Wendelstein 7-AS (W7-AS) at the Max Planck Institute for Plasma Physics (IPP) [31], employ the shear-panel/shell type-based support structure for their three-dimensionally twisted coils. However, we needed to implement a vacuum vessel with numerous large ports, including two large rectangular ones for neutral beam injection (NBI) heating [32], and for Thomson scattering diagnostics. Thus, we decided on the challenge of designing a cage-type MC support structure, prioritizing ease of experimentation over engineering reliability. This design can be quickly constructed at a low-cost and give more space for ports than that of the shear-panel/shell type. Furthermore, it enables us to adjust the MC alignment during the main body assembly. The cage-type MC support structure uses a central support and some linked beams instead of a shear-panel or a shell. The central support absorbs a large centripetal EM force. It was designed to be as small and simple as possible to make sufficient space for diagnostic installation at the inner ports. A pair of pillars are mounted on the central support to stabilize it during assembly work. The Helically Symmetric eXperiment (HSX) at the University of Wisconsin-Madison, one of the optimized stellarators with 48 normal modular twisted coils also has this type of support [33]. The role of the linked beams is to cancel out any toroidal EM force on the MCs. This support structure: linked beams and the central support for the 1 T operation are under fabrication and assembly with MCs.

Fig. 4.  Two candidates for the CFQS entire modular coil support structure for the 1 T operation. The cage-type support structure has been selected for the CFQS. It employs a central support and some linked beams instead of a shear-panel/shell type support structure.

2.4  Modular coil case

We reviewed and modified the design of the MC coil cases several times, as shown in Fig. 5 to overcome some engineering hurdles and operational requirements. It would be ideal to adopt a continuous metal case totally covering an entire coil body which would surely prevent deformation. In the past, we designed continuous coil cases for MCs, as shown in Fig. 5(a) and assessed their robustness against the EM force [19]. The Wendelstein 7-X (W7-X) at IPP uses this type of coil case [34]. However, several concerns about adopting continuous coil cases for the CFQS MCs came up from an engineering perspective. The most serious concern was fabrication cost. The case would be fabricated by pressing a lot of thin metal plates at most 13 mm of thickness and welded together. Therefore, a variety of dies for press work are required. In addition, concerns arose about whether the MCs could be inserted into the continuous case, forming a three-dimensionally twisted geometry and whether the MCs might be damaged while welding the thin pressed plates together. In order to address such engineering concerns, we made a challenging decision to select a partial-clamping method, more affordable and easier to assemble, as shown in Fig. 5(b). This method is likely to be less reliable than the continuous coil case because the stiffness of the MC winding pack itself must be expected.

Fig. 5.  The evolution of the modular coil case design. It ultimately became a reasonable structure in terms of cost, manufacturability, and fabrication time. The designs (b) and (c) are the final versions for the CFQS and CFQS-T, respectively.

The CFQS-T has been completed with a further simplified and low-cost MC support structure depicted in Fig. 5(c) to quickly realize a platform for a proof-of-principle demonstration of a QA configuration. The simplified partial-clamping structure only composed of coil clamps and coil legs for fixing MC winding packs was adopted. To ease temporary disassembly of the MC support structure for the upgrade, the connections between the coil clamps and legs, and between the clamps and an MC winding pack, were secured with bolts. The details of support structure are available in previous paper [21].

Figure 6 gives all types of upgraded MCs reinforced with U-shaped partial clamps for the 1 T operation. There are three types of clamps on MCs. Clamp type A, a complex and long structure is assembled by welding several pieces fabricated with a five-axis computerized numerical control (CNC) machine. Clamp type B, a relatively simple structure, can be fabricated by welding flat plates by a conventional milling machine. Clamp type C, a complex but shorter structure than type A, is fabricated by 3-D machining. The thickness of the clamps is set to a minimum of 20 mm, as shown in Fig. 7, to ensure sufficient stiffness and to accommodate fabrication constraints. If the clamps are designed to be thinner than 20 mm, significant deformation during the machining process might occur. The gap between an MC winding pack and a coil clamp is filled with spacers composed of fiber-glass cloth fixed with resin.

Fig. 6.  Design of U-shaped partial clamps for modular coils for the 1 T operation. There are three types of clamp. Clamp type A is the most complex and longest. Meanwhile, Clamp type B has a relatively simple structure. It can be fabricated with a conventional milling machine. Clamp types A and C must be fabricated with a five-axis computerized numerical control machine due to their complex shape.

Fig. 7.  Cross-sectional shape of the U-shaped partial clamp for modular coils. All types of clamp were designed to be 20 to 30 mm thick. The gap between coil winding pack and clamp is filled with spacers composed of fiber-glass cloth fixed with resin.

2.5  Validation of modular coil support

To validate the stiffness of the MC support structure for the 1 T operation, the distribution of deformation and Von-Mises stress on the MCs induced by the EM force were estimated with the ANSYS Maxwell and Mechanical. This FEM analysis software is widely used to evaluate the designs of various support structure for fusion devices [35, 36]. As a result, we have confirmed that the design meets acceptance criteria. Figure 8 depicts the model of the entire MC support structure including top /bottom flames and main body legs receiving the total gravity of the main body and the vertical EM force of the MCs. The bottom surfaces of the 12 main body legs were set as fixed boundaries. The assumed Young’s modulus of components are listed below;

• 1 GPa for the spacer indicated in Fig. 7,

• 110 GPa for an MC winding pack, and

• 193 GPa for an MC support structure made of SUS.

As for the spacer, the Young’s modulus would be much smaller than the presumed actual value of approximately 50 GPa, to make plenty of margin. An MC winding pack, comprising a total of 72 conductor turns including insulation tapes and resin was replaced by a single solid composite. Gravity and temperature effects, which would not significantly contribute to Von-Mises stress were neglected. The contact condition between components was set as all-bonded (AB), which assumes that all contact surfaces were completely bonded.

Fig. 8.  Model of the entire MC support structure including top /bottom flames and main body legs for finite element method analysis. The bottom surfaces of 12 main body legs were set as fixed boundary.

Figure 9 shows contour plots displaying the deformation distribution of the MC winding packs and their support structure analysed under the AB contact condition. The maximum deformation is less than 1 mm. It is clarified that this degree of deformation, less than 5 mm does not significantly lead to degradation of magnetic field configuration or related physical properties [37–39]. The Von-Mises stress distribution on the MC winding packs and their support structure under the AB contact condition is shown in Fig. 10. The maximum stress on the MC winding pack and support structure is 46 and 105 MPa, respectively, which meets acceptance criteria. However, it is presumed that although the contact condition between a spacer and an MC clamp is almost AB at the initial state of the CFQS operation, it gradually degrades throughout the repetitive pulse energization and ultimately reaches electrical breakdown due to an occurrence of cracks on the MC insulation surface. Thus, we tried to estimate the deformation and Von-Mises stress with a frictionless (FL) contact condition which simulates a highly deteriorated bonding status between the spacer and the MC clamp. In ANSYS, the FL contact condition is defined such that the contact surfaces between components can separate or slide without friction. Table 1 summarizes maximum deformation and Von-Mises stress on an MC winding pack and MC support structure under two different contact conditions. Although the maximum Von-Mises stress on the MC support structure slightly exceeds the acceptance criterion, average values between AB and FL are presumed to be better for judgement because the FL contact condition is far from a realistic status. Repetitive 1 T pulse energizations may also cause fatigue fractures in MC clamps or MC hollow conductors. However, such metal components are more robust than the insulation material used for MCs. Thus, occurrence of cracks on the MC insulation surface are more likely to occur than metal fatigue. Fatigue fractures, including metal fatigue and cracking on the MC typically occur after receiving more than a million cycles of fluctuating loads. This number is even greater than the expected total pulse shot numbers for the CFQS 1 T operation. Anyhow, a regular visual inspection on the MC support structure must be done to find any deterioration at an initial phase.

Fig. 9.  Contour plots of the deformation distribution of modular coil winding packs and their support structure under all-bonded contact condition. Estimated maximum deformation is less than 1 mm.

Fig. 10.  Contour plots of the Von-Mises stress distribution of modular coil winding packs and their support structure under all-bonded contact conditions. Estimated stresses are in the acceptable range.

Table 1. Summary of estimated maximum deformation and Von-Mises stress under two different contact conditions.

Analysis target		Maximum value		Average	Acceptance criterion
		All-bonded (AB)	Frictionless (FL)
Deformation (mm)		0.83	1.44	1.14	5
Von-Mises stress (MPa)	Coil winding pack	46	68	57	70
	Support structure	105	146	126	137

2.6  Auxiliary coils and the vacuum vessel

PFCs are composed of water-cooled oxygen-free copper hollow conductors, the same as those used in MCs. The winding method is solenoid winding with one conductor for the inner vertical (IV) coils, and double pancakes with two simultaneous windings for the outer vertical (OV) coils. The number of turns is 64 with four layers for the IV and 32 with four layers for the OV. Coil current density for the 1 T operation is set at 106.2 A mm⁻² which is more severe than that of the MCs. It leads to a shorter flat-top duration of 0.3 s than that of the MCs. PFCs are designed to be installed on the top/bottom frames of the main body of the CFQS-T/CFQS with numerous coil clamps.

TFCs, having three-dimensionally twisted geometry coils, like an MC are designed to be directly wound on the VV with numerous coil clamps to allow them to have a simple support structure. Although a water-cooled copper conductor like that used for MCs and PFCs was a desirable material to realize long duration with short intervals, the enlargement of the cross-sectional area of the TFCs inevitably increases the risk of interference with the MCs. Instead of using a hollow conductor, we selected two candidates and carried out a winding test using a life-size wooden mock-up VV section to confirm the manufacturability, as shown in Fig. 11. The method is solenoid winding with one conductor. The test revealed that TFC winding with a solid conductor, slimmer than the other candidate, resulted in an unacceptable gap of about 5 mm between the conductor and the VV surface, due to the rigidity of the conductor and the large torsion of TFCs. Furthermore, hammering, which was likely to lead to damage to the conductor insulation was unavoidable in order to adjust the conductor winding route. Meanwhile, winding using the other candidate: a flexible, air-cooled, circular cable, known as a crosslinked polyethylene insulated vinyl sheath (CV) cable was much easier than winding with a solid conductor. It did not require hammering. Based on the test results above, the CV cable was selected as conductor for the TFCs.

Fig. 11.  Toroidal field coil winding test with a wooden mock-up vacuum vessel section and two candidates for coil conductors. Flexible cable, which allows us to wind easily and quickly was selected. The cable is fixed by numerous clamps installed on the surface of the vacuum vessel.

The VV consists of four sections in the toroidal direction: two Type-A and two Type-B, as shown in Fig. 12. It has a three-dimensionally twisted geometry with 46 ports and 12 TFCs. The thickness of the VV wall is set at 6 mm to make enough inboard space to install some diagnostics by a worker, realized by welding several SUS316L thin plates. Two 340 × 580 mm large rectangular ports are integrated on the VV type-A for the NBI heating system and the Thomson scattering diagnostic. A pair of split flanges are attached on both edges of the VV section to suppress thermal deformation during the welding process with other VV sections. Sheathed heaters for baking are divided into 14 systems around the entire VV surface. These systems are individually controlled by power supplies to produce a homogeneous VV temperature distribution at 130 degrees Celsius. The wiring route of baking heaters was carefully designed with a 3-D printed VV, shown in Fig. 13. The VV was stabilized by eight leaf-spring type legs, that can deform and absorb the VV’s thermal expansion during baking. The ANSYS Mechanical and Maxwell helped in estimating the maximum stress and deformation of the VV under vacuum conditions, baking, and coil excitation/demagnetization, inducing the EM force derived from the eddy current [17, 18]. It was ensured that these factors were within an acceptable range.

Fig. 12.  Design of the vacuum vessel (VV) consisting of four sections in the toroidal direction. Each VV section has split flanges, toroidal field coils, leaf-spring type legs and sheathed baking heaters. Two rectangular ports are installed as well for a neutral beam injection heating system and a Thomson scattering diagnostic.

Fig. 13.  Three-dimensional printed vacuum vessel for consideration of baking heater wiring.

3.  Manufacturing of Key Components

Figure 14 shows the established major manufacturing procedure for MC winding packs with a simplified MC support structure for the CFQS-T. Manufactured MC winding packs will be used for the CFQS with a reinforced support structure. As a first manufacturing step, for each type of MC, one complex winding mould was fabricated by machining a large metal block with a five-axis CNC machine. Although the fabrication method took a few months to finish each mould, maximum dimensional deviation was accomplished to be within a few tenths of a millimeter. The MC winding was carried out with the mould on a turntable, as shown in procedure No. 2 in Fig. 14. The moulds were employed not only for coil winding, but also for the Vacuum Pressure Impregnation (VPI) casing, which prevented an MC winding pack from deforming.

Fig. 14.  Established major manufacturing procedure for modular coil (MC) winding packs with simplified coil support structure for the CFQS-T. Vacuum Pressure Impregnation has been performed in two steps at procedures No. 3 and 5 to suppress spring-back effect during ground insulation taping.

The distinctive feature of the MC manufacturing is the adoption of a two-step VPI process to maintain the shape of an MC winding pack during ground insulation taping, as shown in procedures No. 3 and 5 in Fig. 14. Although finishing an MC winding pack with only one VPI was possible, an unacceptable spring-back effect occurred when it was dismounted from a mould for ground insulation taping without VPI. This issue was clarified through the fabrication of a mock-up MC to verify manufacturability.

To perform VPI a couple of times, a five-axis CNC machine was utilized again to cut 3 mm inside the mould so that it matched the dimensions of a ground insulation taped MC winding pack, which increased in thickness by 3 mm, as shown in procedure No. 4 in Fig. 14.

All MCs were manufactured and passed quality inspections. They had less than 2 mm of maximum dimensional deviation, which met the acceptance criterion of 5 mm determined to avoid interference between an MC and any components [21]. This order of dimensional deviation would not significantly influence the magnetic flux surface [37–39]. Meanwhile, PFCs have been successfully manufactured with only one VPI due to their simple geometry, and they passed all inspections.

The manufacturing process of the VV section is shown in Fig. 15. First, SUS316L plates of 6 mm thickness were formed by hot-press processing to suppress spring-back deformation. Then, the plates were welded together on a mould and shaped by hammering as shown in procedure No. 2 in Fig. 15. Four and seven pressed plates were used to form the VV types-A and B, respectively. Because of the highly twisted structure of the VV section, the mould needed to be dismantlable when it was pulled out from a VV section after plate welding and port drilling. TFCs were wound on the surface of each VV section. The installation of TFC clamp bases to fix CV cables were aligned with a paper pattern, as shown in procedure No. 3 in Fig. 15. Baking heaters were installed on the surface of the VV sections as well, by means of point welding. A 3-D measurement of the VV sections was taken using a laser tracker. It revealed that the maximum deviation of the VV Type-A was 4.7 mm, within the acceptable range of 10 mm, while that of Type-B was 20.2 mm. This significant deviation necessitated a design modification of the split flanges of the VV type B, in accordance with the dimension of the as-manufactured VV section. Moreover, it resulted in difficulties in assembling the CFQS-T main body. However, fortunately, the gap between the MC and the maximum deviation part of the VV Type-B was wider than 20.2 mm according to an investigation with a 3D-CAD model. This significant deviation was presumed to come from a greater number of SUS316L plates and a larger surface area compared to the VV Type-A.

Fig. 15.  Established major manufacturing procedure for vacuum vessel (VV) sections. Although the maximum dimensional deviation of the VV section Type-B was 20.2 mm, larger than expected deformation, fortunately, the gap between the MC and the maximum deviation part of the VV Type-B was wider than 20.2 mm according to an investigation with a 3D-CAD model.

4.  Assembly of the CFQS-T Main Body

The primary assembly process of the CFQS-T main body is shown in Fig. 16. Four VV sections with MCs were carefully installed and welded together via split flanges one by one. Although the estimated minimum gap between MCs and VV was only 15 mm in the case of the CFQS-T, the main body was assembled in the factory by a few fully skilled mechanics without any specialized assembly tools. The simplified MC support structure allowed the assembly to be completed in a couple of weeks of actual work. After the assembly of all components, the MC installation position was measured by a laser tracker. It revealed that the maximum installation error of the MC was achieved to be less than 3 mm [14]. We analysed the effects of MC misalignment on the magnetic configuration properties, such as the shape of magnetic flux surfaces, rotational transform, and neoclassical transport properties, and changes in them were not significant if the error level was less than 5 mm [37–39]. The commissioning of the CFQS-T including coil energization tests has been successfully carried out. It has indicated that the CFQS-T could provide QA magnetic field configuration at 0.1 T with good vacuum conditions of 1.5 × 10⁻⁵ Pa [14] and a designed discharge interval of 600 s.

Fig. 16.  Assembly process of the CFQS-T main body. Four vacuum vessel sections with modular coils (MCs) were installed one by one. No specialized tool was required to complete the assembly. The maximum installation error of the MCs was achieved at less than 3 mm which will not influence magnetic field configuration significantly.

5.  Upgrade Status Toward the 1 T Operation

The upgrade activity toward the 1 T operation as the CFQS launched in June, 2025. Sixteen MCs are now being reinforced by the new MC support structure, as shown in Fig. 5(b). The assembly and welding tooling for all MC types were designed and manufactured. Subsequently, one set of clamps (coil cases) has so far been attached to an MC4, as shown in Fig. 17. Each MC4 clamp is fabricated by welding together four pieces, 3-D machined. Fabrication of other parts of the MC support structure, such as the central support and linked beams is also in progress. In addition to reinforcing the MC support structure, an upgraded coil power supply system employing a huge supercapacitor bank, a plasma heating system consisting of an NBI system and an ECRH system [40] and some diagnostics [41] will be installed. In parallel, a new laboratory building for the CFQS operation is being constructed in Chengdu, China. We are striving to complete the CFQS main body by the end of March, 2026 and generate the first plasma in January, 2027.

Fig. 17.  MC4 reinforced with U-shaped partial clamps for the 1 T operation. One set of partial clamps has so far been attached to an MC4.

6.  Summary

The engineering design of the CFQS-T main body with a low-aspect ratio was finalized based on the results of R&D. The design was validated with FEM analysis software ANSYS. Essential components, such as MCs and the VV, which have complex geometry have been successfully manufactured using several ingenuities. Subsequently, we quickly completed assembly of the CFQS-T main body with a simplified MC support structure in 2024. The dimension measurement by a laser tracker and commissioning indicated that the CFQS-T had a designated operational function with good accuracy of MC alignment, which did not influence magnetic field configuration significantly. The design of the MC support structure, which allows the MCs to withstand strong and complex EM forces, was further improved through iterative design and validation with ANSYS. The cage-type support structure and the partial-clamping coil casing were adopted to realize the good balance between strength reliability, and experimental degrees of freedom. So far, one set of MCs with the new support structure has been finished. The CFQS main body will be installed in the new laboratory building in the Tianfu area, Chengdu and the first plasma will be initiated in January, 2027. The engineering experience and knowledge gained from the CFQS project will contribute to developing a next-generation advanced stellarator at NIFS.

 Acknowledgments

The authors would like to give substantial thanks for the strong support of the NIFS-SWJTU joint project (NSJP) for the CFQS from former Director General Prof. Y. Takeiri, Z. Yoshida, and current Director General Prof. H. Yamada of NIFS. The authors also greatly appreciate the tremendous contribution to the CFQS project by Dr. S. Kinoshita. The authors are also grateful for the support to the NSJP for CFQS from other related staff in the CFQS team made up of people from NIFS, SWJTU, and Hefei KEYE. This research is supported by NIFS international collaborations with overseas laboratories (UFEX105), NIFS promotion of magnetic confinement research using helical devices in Asia (URSX401, UFEX108), the NINS program of Promoting Research by Networking among Institutions (grant number 01412302), the Japan-China Scientific Cooperation Program between the JSPS and NSFC (grant number 120237401), the “PLADyS”, JSPS Core-to-Core Program, A. Advanced Research Networks, the NIFS general collaboration project, NIFS18KBAP041, NIFS20KBAP067, NIFS20KBAE001, NIFS22KIPH009, NIFS22KIPH011, NIFS22KIEE001, and NIFS24KIEE002.

References

• [1]   J.  Nührenberg et al., Theory of Fusion Plasmas (Proc. Joint Varenna-Lausanne Int. Workshop Varenna, 1994), (Editrice Compositori, Bologna, 1994) p. 3.
• [2]   P.R.  Garabedian, Phys. Plasmas 3, 2483 (1996).
• [3]   S.  Okamura et al., Nucl. Fusion 41, 1865 (2001).
• [4]   K.  Matsuoka et al., Fusion Sci. Technol. 46, 378 (2004).
• [5]   B.E.  Nelson et al., Fusion Eng. Des. 66–68, 169 (2003).
• [6]   G.H.  Neilson et al., Fusion Sci. Technol. 56, 485 (2009).
• [7]   M.  Drevlak et al., Contrib. Plasma Phys. 53, 459 (2013).
• [8]   F.  Brochard et al., 41st EPS Conference on Plasma Physics (Berlin, June 23–27, 2014) P2.081.
• [9]   T.M.  Qian et al., J. Plasma Phys. 89, 955890502 (2023).
• [10]   X.  Chu et al., Plasma Phys. Control. Fusion 67, 125016 (2025).
• [11]   M.  Isobe et al., Plasma Fusion Res. 14, 3402074 (2019).
• [12]   H.  Liu et al., Plasma Fusion Res. 13, 3405067 (2018).
• [13]   A.  Shimizu et al., Plasma Fusion Res. 13, 3403123 (2018).
• [14]   J.  Cheng et al., Plasma Phys. Control. Fusion 67, 105011 (2025).
• [15]   M.  Shoji et al., Plasma Fusion Res. 18, 2405026 (2023).
• [16]   S.  Kinoshita et al., Plasma Fusion Res. 14, 3405097 (2019).
• [17]   S.  Nakagawa et al., Plasma Fusion Res. 15, 2405066 (2020).
• [18]   T.  Murase et al., Fusion Eng. Des. 161, 111869 (2020).
• [19]   G.  Xiong et al., Fusion Eng. Des. 160, 112021 (2020).
• [20]   A.  Shimizu et al., Nucl. Fusion 62, 016010 (2022).
• [21]   H.  Tanoue et al., Fusion Eng. Des. 212, 114853 (2025).
• [22]  CFQS TEAM, Research Report NIFS-PROC Series NIFS-PROC-126, (2024).
• [23]   H.  Yamaguchi et al., 2023 7th Asia-Pacific Conference on Plasma Physics (Japan, Nov. 12–17, 2023) MFP–57.
• [24]   S.  Okamura et al., J. Plasma Phys. 86, 815860402 (2020).
• [25]   M.  Dravlak, Fusion Technol. 33, 106 (1998).
• [26]   H.  Liu et al., Nucl. Fusion 61, 016014 (2021).
• [27]   Y.  Li et al., Plasma Phys. Control. Fusion 62, 125004 (2020).
• [28]   L.  Sonnerup et al., Nucl. Eng. Des. Fusion 3, 233 (1986).
• [29]   H.  Tamai et al., Plasma Sci. Technol. 6, 2141 (2004).
• [30]   A.  Iiyoshi et al., Nucl. Fusion 39, 9Y (1999).
• [31]   R.  Mathis et al., Fusion Eng. Des. 11, 399 (1990).
• [32]   K.  Ogawa et al., Plasma Fusion Res. 14, 3402067 (2019).
• [33]   F.  Simon et al., Fusion Technol. 27, 273 (1995).
• [34]   M.  Wanner et al., Nucl. Fusion 43, 416 (2003).
• [35]   L.  Myatt et al., Fusion Sci. Technol. 47, 916 (2005).
• [36]   V.  Bykov et al., Fusion Eng. Des. 82, 1538 (2007).
• [37]   A.  Shimizu et al., Plasma Fusion Res. 14, 3403151 (2019).
• [38]   G.  Xiong et al., Plasma Phys. Control. Fusion 65, 035020 (2023).
• [39]   X.  Zhang et al., Fusion Eng. Des. 211, 114820 (2025).
• [40]   Y.  Yoshimura et al., Plasma Fusion Res. 17, 2402039 (2022).
• [41]   A.  Shimizu et al., J. Instrum. 17, C06004 (2022).