MAGNETICALLY LEVITATED LINEAR STAGE FOR IN-VACUUMTRANSPORTATION TASKS

Lei Zhou, David L. TrumperDepartment of Mechanical Engineering

Massachusetts Institute of Technology, Cambridge, MA, USA

INTRODUCTIONThe reticle handling task in photo lithography machines isbecoming increasingly challenging in the extreme ultra-violet (EUV) era [1]. In the next generation lithographymachines, the reticle needs to be transported througha distance of 2 m between the storage position and thescanning position in high vacuum, and the transportationmechanism needs to satisfy the ultra-tight contamina-tion requirements. In addition, due to the increased sys-tem complexity, the allowed height for the reticle trans-portation mechanism in the machines is constraint within100 mm. To fulfill these requirements, the use of the tra-ditional robot manipulator is challenging.

In this work, we present a new concept of in-vacuumtransportation mechanism using a magnetically levitatedlinear stage (MLLS). Compared with robot manipulators,linear stages usually require less volume for long dis-tance transportation tasks, and thus can satisfy the tightvolume constraint with less challenge. In addition, weintroduce magnetic levitation to the linear stage to elim-inate particle generation of mechanical bearings. To re-duce the system’s complexity, we present a linear bear-ingless slice motor design [2], where the stage’s mag-netic levitation in several degrees of freedom (DOFs) areachieved passively. This design helps reduce the numberof actuators and sensors in the system.

Linear hysteresis motor is selected as the driving princi-ple for the MLLS. Hysteresis motor has the advantagesof simple structure and high thermal and mechanical ro-bustness [3]. In addition, the secondary of a hysteresismotor can be made from permanent-magnet-free solidmetal material. This is desirable for in-vacuum opera-tion, since permanent magnets (PM) can out-gas in highvacuum when it is not encapsulated. Hysteresis linearmotor also allows using the same secondary material formagnetic and structural purposes, thereby beneficial fora compact design. In the past, most of the developedhysteresis motors are in rotary form and operate in open-loop. Our past work [4] studied the position control forrotary hysteresis motors. To our knowledge, this paperpresents the first experimental study of linear hysteresismotor.

Air-gapsensorPCB

PermanentmagnetStatorbackiron

Mainstator

SpacerYawcontrolstator

Stagebackiron HysteresissecondaryEncoderarray

x

y

z

100mm

(a) Cross-section view of CAD model.

(b) Photograph of linear stage.

FIGURE 1. Cross-section of the magnetically levitatedlinear motor.

HARDWARE OVERVIEWFig. 1 shows the CAD model and the photograph of themagnetically levitated linear stage system, which com-prising two stators, one moving stage, and a sensing sys-tem. The coordinate system is shown in Fig. 1a. Herethe moving stage is driven along the y-axis. The mag-netic levitation of the moving stage is active in x- andθz-directions, and is passive in z-, θx-, and θy-directions.In our design, the moving stage is vacuum compatible.When the system is integrated in the lithography ma-chine, a channel with thin walls (not shown in figures)needs to be configured along the motion range of themoving stage, and separates the moving stage and thestators. The moving stage is levitated and transports thereticle inside the channel in clean vacuum, while the sta-tors are configured outside the channel, and are in a rela-

Motor Stator: 3-phase windings

Yaw control stators:5-phase windings

FIGURE 2. Photograph of front view of the stator.

Stage base:6061 Aluminum

Hysteresis motor secondary:D2 tool steel

Bias flux collector: 1018 low carbon steel

Back-iron:1018 Low carbon steel

FIGURE 3. Photograph of the moving stage in the mag-netically levitated linear stage.

tively dirty vacuum environment. The air-gap of the linearmotors is 1.5 mm.

Fig. 2 and Fig. 1 show the photograph of the front andside views of the stator, respectively. The stator assem-bly mainly consists of a motor stator, two yaw controlstators, and a flux-biasing structure comprising two rowsof PMs and one stator back-iron. The stators are fixedon the base via three angle plates. Lumped windingsare selected for the stators due to volume constraints.The windings on the motor stators are 3-phase, while thewindings in the yaw control stators are 5-phase. Thereare 11 currents independently controlled in the system,including two sets of 3-phase windings in the left andright motor stators, and one set of 5-phase windings inthe yaw control stators. The current controlled switchingtype power amplifiers AMC B30A40 is being used to drivecurrents in the windings. The DC bus voltage is 300V.

Fig. 3 shows a photograph of the moving stage, whichcomprising an aluminum stage base, two stage back-irons, two hysteresis motor secondaries, and four biasflux collectors. The hysteresis motor secondaries aremade of D2 tool steel. The D2 tool steel has relativelylarge magnetic hysteresis, which allows it being used forhysteresis motor secondary. It also has large permeabil-ity, which is advantageous for reluctance force generationfor magnetic suspension purpose. The total mass of thestage is 4.9 kg.

Two kinds of displacement sensors are used to measurethe stage’s motion in x-, y-, and θz-DOFs. To measurethe x- and θz-directional displacements of the stage atdifferent y-positions, 16 optical displacement sensors are

N S

NS

NS

N S

xy

z

FIGURE 4. Magnetic fluxes in the MLLS system. Blacklines: PM fluxes for passive magnetic suspension in z,θx, and θy-DOFs; red lines: motor fluxes for x-directionalsuspension force and y-directional thrust force, blue line:yaw control fluxes for θz-directional suspension torque.

𝐴 − δ𝐴 𝐴 + δ𝐴

𝐹&

𝐹' a

Permanent magnet bias flux

Yaw control flux

T"

c

b d

(a) Suspension force generationin x-direction.

(b) Torque generation about z-axis.

FIGURE 5. Top view of air-gap magnetic flux distributionand controlling force/torque generation.

arranged along the stator on the printed circuit boards(PCBs) mounted on the front surface of the stators. Thesignals from the sensors engaged with the stage areused for feedback control to stabilize the stage’s mag-netic suspension. In addition, two rows of magneticencoders are arranged along the motion direction formeasuring the y-directional displacement of the stage.Two magnetic encoder scale are attached on the movingstage, and in each row there are 4 encoder read headsconfigured along the moving direction of the stage. Withthis sensing system design, there is no cable attached onthe moving stage. Note that the magnetic encoders arenot vacuum compatible. When the system is operating invacuum, different displacement sensors are needed, forexample laser interferometers.

OPERATING PRINCIPLE

Suspension Force GenerationFig. 4 shows a cross-section diagram of the MLLS andthe magnetic flux distributions in the system, and Fig. 5shows the top view of the air gap magnetic fluxes. Thereare three kinds of magnetic fluxes in the system. Theblack lines in Fig. 4 and Fig. 5b show the PM bias mag-netic fluxes, which are used to generate passive mag-netic suspension force/torque in z-, θx-, and θy-DOFs.When the stage is displaced in these directions, the

𝐶"(𝑠)𝑥'() 𝑒"+-

𝐼,-./

𝑢"

+

++

-

𝐼2

𝑥 and 𝜃4estimator

𝑥5+

-

𝑥26

𝑥76

+

-

𝑥28

𝑥78

DAC

x suspensioncontroller

ADC

PowerAmps

PowerAmps

𝑥6

𝑥8

LeftMotorStator

RightMotorStator

Real-time Controller

DIO

𝑦2

𝑦7

Left Encoder+

+12 Right Encoder

𝑦'() 𝑒<+

-y positioncontroller

𝐼7

𝐶<(𝑠)𝜙

Mag 3-phasedistributorPhase

Mag 3-phasedistributorPhase

𝑦

𝐶>4(𝑠)𝜃4'()𝑒>4+

-𝜃4 suspensioncontroller

PowerAmps

Yaw ControlStators𝑢>4

Phase 5-phasedistributorMag

𝜃?4

+

+@ABC

@ABD

DAC

FIGURE 6. Block diagram of the control system for the magnetically levitated linear stage.

PM fluxes in the air gaps provide restoring forces, andtherefore stabilize the magnetic suspension passively.Note that the PM bias flux also generates destabilizingforce/torque in x- and θz-directions, and feedback controlis required to stabilize the magnetic suspension in theseDOFs. Rotary motors with this magnetic suspension de-sign are often referred as slice bearingless motors, andtypical applications of bearingless slice motors includepumps [5], high-speed fans [6], centrifuges [7], etc. Toour knowledge, our paper presents the first linear slicebearingless motor design.

The red lines in Fig. 4 and Fig. 5a represents the motorfluxes, which are generated by the windings in the mo-tor stators. The common mode of the left and right mo-tor fluxes is used to generate y-directional thrust force tothe stage by interacting with the hysteresis motor secon-daries. The differential between the left and right motorfluxes generates x-directional reluctance force, which isused to control the x-directional magnetic suspension.

The blue lines in Fig. 4 and Fig. 5b show the yaw sus-pension control fluxes, which are generated by the 5-phase windings in the yaw control stators. In the top andbottom air gaps, the yaw suspension control flux is dis-tributed sinusoidally, and is synchronous to the movingstage. This flux steers the PM bias flux to generate θz-directional suspension torque. For example, in Fig. 5b,the yaw control flux attenuates the PM bias flux in areasb and c, while intensify the flux in areas a and d, thereforethe reluctance forces in the air-gaps generates a torqueabout z-axis as is shown by the arrow.

Thrust Force GenerationThe proposed MLLS uses linear hysteresis motors for thethrust force generation. When the motor windings are ex-

!2

!4

0−!4−!2−! !

Hysteresis Thrust Force

Reluctance Thrust Force

Total Thrust Force

Phase Angle (rad)

FIGURE 7. Thrust force and phase angle relationships inhysteresis-reluctance hybrid motors.

cited, the induced magnetization in the secondaries lagsbehind the external field due to the magnetic hysteresisof the secondary material, and therefore generates thrustforces. The hysteresis motor secondaries are often pre-magnetized using a large current amplitude to improvethe force generation capability of the motor. In addition,reluctance thrust forces are existing in linear hysteresismotors at the edges of the motor secondaries. As aresult, linear hysteresis motors’ characteristics resemblethat of hysteresis-reluctance hybrid motors [8].

The linear hysteresis motor in our MLLS system oper-ates in synchronous mode. There are two reasons forthis selection. First, the reluctance force in the linearhysteresis motor is oscillatory when the motor is asyn-chronous, which introduces undesirable vibration to thesystem. Second, synchronous operation of the motoreliminates secondary hysteresis and eddy current losses.This is highly desirable for the reticle transportation sys-tem because cooling the moving stage is extremely chal-lenging since the stage is in vacuum.

TABLE 1. Magnetic levitation stiffnesses and resonancefrequencies in the passive levitated DOFs

DOF Resonance Frequency Passive Stiffnessz (Vertical) 11 Hz 37 N/mmθx (Pitch) 9 Hz 54 Nm/radθy (Roll) 9 Hz 163 Nm/rad

Fig. 7 shows the thrust force and phase relationship oflinear hysteresis-reluctance hybrid motor in synchronousoperation. Here, the horizontal axis shows the phase dif-ference between the stator excitations and the positionof the stage. It can be seen that the hysteresis thrustforce demonstrates its maximum values at phase angleof ±π/2. With the reluctance thrust force added, the peakof the total thrust force shifts toward the center.

CONTROL DESIGNFig. 6 shows a block diagram of the control system forthe magnetically levitated linear stage. The x- and θz-directional displacements of the moving stage are esti-mated from the encoder and air-gap sensor measure-ments, and are fedback for suspension control. The x-DOF magnetic suspension control signal ux is the differ-ential magnitude of the left and right motor current am-plitudes, and a bias current Ibias added to the motor sta-tor currents for thrust force generation and maintainingmagnetic suspension. the θz-DOF suspension controleffort signal is used to determine the yaw control sta-tor current amplitude. The two stators operates in syn-chronous mode. The position control loop for the linearstage is closed with the encoder signals being used forfeedback, and the control effort signal is used to deter-mine the phase between the motor stator excitation andthe position of the linear stage. The signals from the D/Aconverters of the real-time controllers are send to poweramplifiers as current commands.

EXPERIMENTS

Suspension TestsTable 1 shows the magnetic suspension stiffness in thepassive levitated DOFs, which is estimated through mea-suring the natural frequency of these modes when thestage is levitated. The stage sags below the equilibriumposition in the vertical direction by 1.45 mm due to itsweight.

Fig. 8 shows the measured plant and loop frequency re-sponses of the x- and Rz-directional magnetic suspen-sion of the moving stage under a bias current of 2.5 A.In the plant frequency responses, the magnitude curvesshow a notch and peak around 9 Hz, which correspondsto a pair of complex zeros and poles in the x- and θz-directional magnetic levitation plants. To our understand-

100 101 10210-310-210-1100101102

Ampl

itude

(abs

)

PlantLoop

100 101 102

Frequency (Hz)

-360

-270

-180

-90

Phas

e (d

eg)

(a) x-directional levitation

100 101 10210-410-310-210-1100101102

Ampl

itude

(abs

)

PlantLoop

100 101 102

Frequency (Hz)

-450-360-270-180

-90Ph

ase

(deg

)

(b) θz-directional levitation.

FIGURE 8. Measured plant and loop Bode plots of the x-and θz-directional magnetic suspension systems.

ing, this is due to the 2nd-order θy-directional mode isadded to the measured modes. Data show the cross-over frequencies of the x- and θz-directional magneticlevitation are 70 Hz and 60 Hz, respectively, and thephase margin of both loops are around 20 degrees.

Linear Motor TestsThe thrust force and phase relationship for the linearstage is measured under different bias current ampli-tudes, as shown in Fig. 9. In this measurement, thestage is levitated, and the motor secondaries are pre-magnetized with a zero air-gap and 5A current amplitude.It can be seen that the phase angles of the peak thrustforce are between ±π/4 and ±π/2. This is a result ofthe combination of hysteresis force and reluctance force.The maximum thrust force that the stage has demon-strated is 3.2 N under 3A bias current amplitude, whichcorresponds to an acceleration of 650 mm/s2.

We conducted initial test for the closed-loop position con-trol of linear stage using the method shown in Fig. 6.

- -3 /4 - /2 - /4 0 pi/4 /2 3 /4

Phase (rad)

-4

-2

0

2

4

Thr

ust F

orce

(N

)

Ibias

=2A

Ibias

=2.5A

Ibias

=3A

FIGURE 9. Measured thrust force and phase relationshipof linear stage under different bias current amplitudes.

0 0.5 1 1.5 2 2.5 3 3.5

Time (s)

-0.5

0

0.5

1

1.5

Dis

plac

emen

t (m

m)

FIGURE 10. Measured position step response of theMLLS.

Fig. 10 shows the measured position step response.Data show that the bandwidth of the position control loopis about 5 Hz. It can be observed that there is a 9 Hz os-cillation in the steady-state position signal. This is dueto the θx-directional (pitch) mode is coupled to the y-directional position control loop. We expect that this os-cillatory signal in position can be eliminated if the cross-over frequency of the position control loop is above thepitch mode natural frequency. This test shows the avail-ability for the position control for linear hysteresis motors.

CONCLUSIONS AND FUTURE WORKThis work presents a MLLS driven by linear hysteresismotors. Passive magnetic suspension is utilized to re-duce actuators and sensors. The prototype system canmagnetically levitate the stage and drive its linear motionwith a maximum acceleration of 650 mm/s2.

In the position control tests for the linear stage, we havedemonstrated the feasibility of the linear position control,however the bandwidth of the control is relatively low inthe presented tests. We are currently working on improv-ing the position control bandwidth of the linear stage. Wealso plan to test different motor secondary materials withlarger magnetic hysteresis to improve the motor’s thrustforce capability.

ACKNOWLEDGEMENTWe thank ASML for supporting this work. We thank Na-tional Instruments Inc. for contributing the real-time con-troller used in this project.

REFERENCES[1] Peeters R, Lok S, Mallman J, van Noordenburg M,

Harned N, Kuerz P, et al. EUV lithography: NXEplatform performance overview. In: Extreme Ultra-violet (EUV) Lithography V. vol. 9048. InternationalSociety for Optics and Photonics; 2014. p. 90481J.

[2] Schoeb R, Barletta N. Principle and application of abearingless slice motor. JSME International JournalSeries C Mechanical Systems, Machine Elementsand Manufacturing. 1997;40(4):593–598.

[3] Teare BR. Theory of hysteresis-motor torque. Elec-trical Engineering. 1940;59(12):907–912.

[4] Zhou L, Gruber W, Trumper DL. Position Controlfor Hysteresis Motors: Transient-time Model andField-oriented Control. IEEE Trans on Ind Appl.2018;54(4):3197–3207.

[5] Nussbaumer T, Karutz P, Zurcher F, Kolar JW. Mag-netically levitated slice motorsAn overview. IEEETrans on Ind Appl. 2011;47(2):754–766.

[6] Sugimoto H, Shimura I, Chiba A. Design of SPMand IPM rotors in novel one-axis actively positionedsingle-drive bearingless motor. In: Energy Conver-sion Congress and Exposition (ECCE), 2014 IEEE.IEEE; 2014. p. 5858–5863.

[7] Asama J, Kanehara D, Oiwa T, Chiba A. Devel-opment of a compact centrifugal pump with a two-axis actively positioned consequent-pole bearing-less motor. IEEE Transactions on Industry Applica-tions. 2014;50(1):288–295.

[8] Rahman MA, Osheiba AM. Steady-state perfor-mance analysis of polyphase hysteresis-reluctancemotors. IEEE Transactions on Industry applications.1985;(3):659–663.