aa r X i v : . [ a s t r o - ph ] M a y Astron. Nachr. / AN , No. 6, 499 – 506 (2007) /
DOI
Reduction of boundary effects in the spiral MRI experiment PROMISE
J. Szklarski ⋆ Astrophysikalisches Institut Potsdam, An der Sternwarte 16, D-14482 Potsdam, GermanyReceived 2007 Apr 18, accepted 2007 Apr 22Published online 2007 Jun 18
Key words methods: numerical – magnetic fields – magnetohydrodynamics (MHD)Magnetorotational instability (MRI) is one of the most important and most common instabilities in astrophysics. It iswidely accepted that it serves as a source of turbulent viscosity in accretion disks – the most energy efficient objects in theUniverse. However it is very difficult to bring this process down on earth and model it in a laboratory experiment. Severaldifferent approaches have been proposed, one of the most recent is PROMISE (Potsdam-ROssendorf MagnetorotationalInStability Experiment). It consists of a flow of a liquid metal between two rotating cylinders under applied current-freespiral magnetic field. The cylinders must be covered with plates which introduce additional end-effects which alter theflow and make it more difficult to clearly distinguish between MRI stable and unstable state. In this paper we proposesimple and inexpensive improvement to the PROMISE experiment which would reduce those undesirable effects. c (cid:13) Velikhov (1959) showed that, for ideal magnetohydrody-namics, an axial magnetic field applied to a flow of a liquidmetal between two concentric, differentially rotating cylin-ders (Taylor-Couette flow) can lower the critical rotationratio or even destabilize the flow, although it is hydrody-namically stable (when the Prandtl number is large enough).That type of instability is called magnetorotational instabil-ity, and not long ago Balbus & Hawley (1991) demonstratedthat it plays an important role in astrophysics. MRI serves asan essential mechanism for transporting angular momentumin a wide range of astrophysical objects, stellar interiors, jetsand in particular it is crucial for process of accretion whereit provides necessary amount of turbulent viscosity. Recentexperiments by Ji et al. (2006) suggested that purely hydro-dynamical quasi-Keplerian flows, i.e. Taylor-Couette flowswhich resemble Keplerian disks, do not provide viscosityrequired to transport the angular momentum effectively, andtherefore this instability is ruled out as a source of viscousturbulence in the disks.One of the most convenient laboratory models for MRIis still the magnetohydrodynamical, cylindrical Taylor-Couette flow with an imposed external magnetic field alongthe axis (R¨udiger & Zhang 2001; Ji, Goodman & Kageyama2001). For magnetic Prandtl numbers Pm ≈ ⋆ Corresponding author: [email protected] (due to low conductivity) so that for a purely axial field thecritical Reynolds number is of order O (10 ) , and thereforevast rotation rates are necessary for MRI to grow.Recently it has been shown by Hollerbach & R¨udiger(2005) and R¨udiger et al. (2005) that a current-free externalazimuthal magnetic field in addition to the axial one can re-duce the critical Reynolds number to O (10 ) which makesit much easier to design an MRI experiment. Moreover dueto symmetry-breaking there exists a drift of Taylor vorticesassociated with the configuration of the applied magneticfield. The frequency of the traveling wave that can be mea-sured and compared with theory is an important feature ofthis type of instability.The idea of an additional toroidal field was success-fully implemented in the PROMISE experiment by Stefaniet al. (2006) where modes corresponding to so called spi-ral (or helical) MRI were observed for the first time (seealso R¨udiger et al. 2006; Stefani et al. 2007). Results of thisexperiment also show that in the basic stable state, withoutany toroidal field, there exists a nonzero axial velocity fieldwhich arises due to presence of the rigid endplates enclos-ing the cylinders. These plates, undoubtedly present in anyreal experiment, are responsible for additional effects whichdo not take place for an idealized infinitely long container.The boundary layer which exists in the vicinity of theendplates consists of an Ekman layer which is the resultof the rotation of a rigid surface, and a Hartmann layerwhich develops when a conducting fluid is used and an ex-ternal axial magnetic field is applied (see e.g. Ekman 1905;Roberts 1967). Consequently, the global properties of theflow change when compared to the idealized case of in-finitely long cylinders: a secondary flow, i.e. two large Ek-man vortices appear, and the Hartmann current is drawn intothe bulk of the fluid. All these effects depend on the mechan- c (cid:13)
00 J. Szklarski: Reduction of boundary effects in the spiral MRI experiment PROMISE ical and magnetic properties of the lids. In the PROMISEexperiment one of the lids is made of copper and is attachedto the outer cylinder, the other one is a stationary plexiglassplate.In this work we review simple improvements which canreduce undesired effects induced by the lids and providetherefore the possibility to distinguish more clearly betweenstable and unstable states of MRI.
We consider two concentric cylinders with radii R in = 4 cm, R out = 8 cm and height H = 40 cm which ro-tate with angular velocities Ω in , Ω out . The rotation ra-tio ˆ µ = Ω out / Ω in is chosen in such a way that the flowis hydrodynamically stable, i.e. Rayleigh stability criterion ∂ r ( r Ω) > , r being the distance from the axis of rota-tion, is fulfilled. Through this paper we use ˆ µ = 0 . . Forinfinite cylinders the basic rotational profile for the flow isthe Couette solution Ω ( r ) = a + br , (1)where a, b are constants dependent on radii and rotationspeeds a = Ω in ˆ µ − ˆ η − ˆ η , b = 1 − ˆ µ − ˆ η R Ω in . (2)The external magnetic field (steady, current free) has theform of B = B (cid:18) βR in r ˆ e φ + ˆ e z (cid:19) , (3)and its strength is measured by the Hartmann number Ha = B s R in ( R out − R in ) µ ρνη . (4)The magnetic properties of the conducting fluid are de-scribed by the magnetic Prandtl number which is the ratioof the kinematic viscosity ν to the magnetic diffusivity η , Pm = ν/η , µ denotes the magnetic permeability, and ρ denotes the density. The Reynolds number Re is defined as Re = ν − Ω in p R in ( R out − R in ) . The liquid used in thePROMISE experiment is the euticetic alloy GaInSn giving Pm = 1 . × − . Thus it is reasonable to solve the MHDequations (dimensionless) in the small Prandtl number limit(Youd & Barenghi 2006; Roberts 1967; Zikanov & Thess1998), ∂ t u + ( u · ∇ ) u = −∇ p + ∇ u + Ha rot b × B B , (5) ∇ b = − rot( u × B /B ) , (6)and div u = 0 , div b = 0 , where u is the velocity field and b is the perturbed magnetic field.We simulate the above nonlinear equations for a 2 Daxisymmetric flow in cylindrical coordinates ( r, φ, z ) . Fordetails on the numerical method and the boundary condi-tions see Youd & Barenghi (2006) and Szklarski & R¨udiger Fig. 1
Profiles of u z ( z, t ) at r = R in + 0 . D for periodiccylinders just above the critical characteristic values: Re =1000 , Ha = 9 . , β = 4 . The critical Reynolds number inthis case is Re c = 842 .(2006). The cylinders are assumed to be perfectly conduct-ing, and the endplates are either conducting or insulating.For the latter the pseudo-vacuum approximation is used . From the point of view of the MRI experiment we are inter-ested in obtaining a stable, uniform rotation profile whichfor subcritical characteristic parameters is as close as possi-ble to the idealized basic state Ω . On the other hand weexpect a clear pattern of traveling vortices for supercriti-cal conditions. For infinite cylinders with an external axialmagnetic field and liquids with Pm of our interest the ba-sic Couette profile Ω is not altered until a critical Reynoldsnumber of order O (10 – ) is reached (corresponding toa rotation frequency f ≈ Hz). For an instability due tothe additional toroidal field with β = 4 we expect Re to beof order O (10 ) (implying f ≈ . Hz), Ha of order O (10) ,and therefore we search for conditions for which the flow isclosest to the Ω profile for these parameters (for details onthe critical values see Hollerbach & R¨udiger 2005; R¨udigeret al. 2005).Figure 1 displays values of the velocity component u z measured along the z axis at r = R in + 0 . D for super-critical values of rotation and magnetic fields. R in = 4 cm, R out = 8 cm, and the physical properties of gallium forthe viscosity and the magnetic diffusivity are used in orderto obtain values in physical scales comparable with thoseof the PROMISE experiment. The results in this figure arefor cylinders with periodic boundary conditions so that theprofiles are not constrained by end effects and are directlycomparable with results from linear theory for infinite cylin-ders. We notice clear traces of the drifting Taylor vortices. However, even for copper the assumption for perfect conductors maybe not very realistic. c (cid:13) stron. Nachr. / AN (2007) 501 All undesirable effects induced by the endplates arise as aconsequence of vertical shears near the boundaries. Thus weattempt to reduce the shears by using appropriate boundaryconditions. Some experiments (e.g. Ji et al. 2004; Noguchiet al. 2002) must deal with vast rotation rates since a toroidalfield is not applied (i.e. β = 0 ), and the rigidly rotat-ing boundaries dominate the whole flow (see Hollerbach &Fournier 2004). In this case it is necessary to split the end-plates into many independently rotating rings (Kageyamaet al. 2004; Burin et al. 2006). When the rotation rates arerelatively slow, so that the corresponding Reynolds numberis of order O (10 ) , the desired result can be achieved ei-ther by allowing the endplates to rotate rigidly and indepen-dently (see e.g. Abshagen et al. 2004) or by splitting theminto two rings which are attached to both cylinders. Fromthe technical point of view the latter configuration is easierto implement and can be considered as a possible extensionto the next spiral MRI experiment.Firstly we consider a criterion according to which wesay that the boundary conditions are more suitable. In thebasic state for subcritical parameters and for the case of pe-riodic cylinders, the rotational profile of the fluid is Ω ( r ) and is independent of z , and the magnetic perturbations b are zero everywhere. Introducing endplates leads to the de-velopment of z and r gradients in velocity, especially closeto the vertical boundaries where Ω( r ) from the bulk of thefluid must match the imposed conditions at z = 0 , z = H .Consequently two Ekman vortices, new currents and mag-netic fields are generated (we assume the lids to be insulat-ing unless explicitly stated otherwise). Any deviation from Ω will result in generating an azimuthal component of themagnetic field, b φ , which enters the momentum equation(and in our 2D axisymmetric formulation it is the only termwhich gives rise to the Lorentz force). Vertical profiles of b φ in the middle, i.e. for r = D/ , D = R out − R in , are shownin Fig. 2 for two different boundary conditions. If the end-plates rotate rigidly with the outer cylinder, Ω end = Ω out ,the Ekman circulation at the bottom lid has a clockwise di-rection, if they rotate with the inner cylinder, Ω end = Ω in ,counter-clockwise – all the gradients have opposite sign. Weconclude that, not surprisingly, there exists a condition with Ω out < Ω endmin < Ω in for which the shears are minimized,and the generated magnetic field as well.We are interested in obtaining a rotational profile forwhich the energy in b φ is minimized, E b = Z Z b φ ( r, z ) d r d z, (7)where the integration is done over the total volume. As ameasure of the deviation from Ω one could also consider,for example, the kinetic energy of the flow, but our aim is toobtain good rotational profiles also for Ha of order 10 (and β = 0 ). This is not necessarily a good approach since theaxial field can inhibit the flow velocity while the rotationalprofile will still be significantly different from Ω . Fig. 2
Vertical profiles for induced b φ in the middle ofthe gap, r = D/ for β = 0 , Re = 100 , Ha = 1 andpseudo-vacuum boundary conditions. —— Ω end = Ω out ,– – – Ω end = Ω in . Fig. 3
The magnetic energy E b as a function of an-gular velocity of independently rotating lid Ω end for β = 0 , Re = 100 . ——
Ha = 1 , – – –
Ha = 10 .Figure 3 shows how E b depends on the rotation rates ofthe rigid endplates, Ω in < Ω end < Ω out . We notice that aminimum occurs for Ω endmin ≈ . in − Ω out ) + Ω out which is even three orders of magnitude smaller than for Ω end = Ω in .When considering endplates divided into rings, we as-sume that the ring attached to the inner cylinder has a width w , the other one attached to the outer cylinder has w = D − w . Because it is not obvious which value for w shouldbe chosen we search for the optimal w , i.e. for which E b has a minimum, by performing simulations for several dif-ferent values (Wendt 1933, for example, used w /D = 0 . ).From Fig. 4 we see that the energy of the induced b φ has aminimum for w /D ≈ . which is roughly independent c (cid:13)
02 J. Szklarski: Reduction of boundary effects in the spiral MRI experiment PROMISE
Fig. 4
The magnetic energy E b as a function of radius ofthe inner ring for β = 0 , Re = 100 . ——
Ha = 1 ,– – –
Ha = 10 .of the applied axial magnetic field. It has also been checkedthat the minimum holds for larger Reynolds numbers (forthe Fig. 3 as well). Again we notice the improvement of E b of two to three orders of magnitude when compared toone end-ring attached either to the inner ( w = D ) or outer( w = 0 ) cylinder. The minimum value is very similar tothat for independently rotating endplates.A qualitative view of the resulting rotational profilesgives Fig. 5 which displays deviations of ¯Ω( r ) – the an-gular velocity averaged in the z domain – from Ω ( r ) fordifferent rotational properties of the endplates and for var-ied Re and magnetic fields. The case with independentlyrotating endplates refers to boundary conditions where bothlids rotate with angular velocity Ω end corresponding to theminimum value of E b . For comparison we also present thecase for two rings attached to the cylinders with equal width w = w = D/ .We see that applying independently rotating or split end-plates produce significantly more suitable profiles – flatterand closer to 1. We also notice that using w = 0 . D givessomewhat better results than w = 0 . D , especially for r > R in + D/ where the former profile is almost flat. Figure 6 shows values of the velocity component u z , simi-larly like Fig. 1 but for finite cylinders. The velocity field u z in the basic state, i.e. β = 0 for which there is no instability,is u z = 0 everywhere when considering infinite or periodiccylinders for Re ≈ , Ha ≈ . For the enclosed cylin-der this is not the case. In Fig. 6a (left) we present resultsfor symmetrically, rigidly rotating (with Ω out ), insulatingendplates. We notice that u z is quite large and, more im-portantly, time dependent (this is even more evident for u z closer to the inner cylinder). The right panel in this figure Fig. 5
Deviations of the averaged ¯Ω( r ) from the ba-sic state Ω ( r ) for different vertical boundary conditions;rigidly rotating endplates (both with Ω end ): —— Ω end =Ω out , · · · Ω end = Ω in , – – – Ω end = Ω endmin ; dividedinto two rings: · – · – w = 0 . , · · · — · · · w = 0 . .(a) Re = 1000 , Ha = 0 , (b)
Re = 1000 , Ha = 10 .displays the same flow with the toroidal field applied, u z ( z ) is averaged in time and subtracted in order to filter out thebackground. We clearly see the instability and structure oftraveling vortices, the frequency of this motion agrees withthe predictions based on the linear analysis (see e.g. R¨udigeret al. 2005). c (cid:13) stron. Nachr. / AN (2007) 503 Fig. 6
The axial velocity u z ( z, t ) at r = R in + 0 . D as a function of time t and z . Left : basic state β = 0 . Right : β = 6 , the averaged u z ( t ) is subtracted in order to eliminate the background from the velocity field, except in (d). (a): bothendplates rotate rigidly with Ω end = Ω out . (b): both endplates are divided into rings attached to the cylinders; the inner ringhas the width . D . (c), (d): the bottom endplate is stationary Ω bot = 0 , the upper rotates with Ω top = Ω out . (a), (b), (c):insulating endplates, Re = 1775 ( Ω in = 0 . Hz),
Ha = 9 . . (d): perfectly conducting endplates, Re = 1000 , Ha = 10 .The traveling wave frequency for (a), (b), (c) is respectively f / Ω in = 0 . , . , . whereas the linear stabilityanalysis yields f / Ω in = 0 . . c (cid:13)
04 J. Szklarski: Reduction of boundary effects in the spiral MRI experiment PROMISE
As we have shown above, one can obtain a much bet-ter basic state for the finite cylinders by dividing endplatesinto two rings. The results for such conditions are presentedin Fig. 6b. We notice that the background state quickly be-comes entirely steady. Naturally the Ekman pumping mech-anism is still present in this case, and traces of two Ekmanvortices can be seen. The flow, however, is laminar. For β = 6 the pattern of the traveling vortices is clearly moreregular (cf. Fig. 6b, right).When one considers two endplates with different rota-tional properties, additional velocity and current gradientsin the vertical direction arise and disturb further the flow. Itis clearly seen in Fig. 6c that disturbances exist in the casewhere the upper endplate rotates with Ω top = Ω out , and thebottom one is fixed, Ω bot = 0 . The background flow for β = 0 is highly irregular and time-dependent, especially inthe middle part of the cylinder, the circulation close to theendplates is roughly steady. Nonetheless, the external B φ produces, again, a clear periodic motion with a frequencycorresponding to that of the helical MRI.Using conducting boundaries instead of insulating onesleads to an increase of the Ekman circulation and the Hart-mann current, the latter being drawn from the plates. Thiscurrent is significantly stronger than the current generatedin the Ekman-Hartmann layer, and therefore we expect thatan experiment with conducting plates would undergo addi-tional problems due to magnetic forces acting on the fluid.Let us consider a perfectly conducting endplates with asym-metric rotation (again at the top Ω top = Ω out and at thebottom Ω bot = 0 ), then there exists an important gradientin the radial current which, acting in concert with the axialmagnetic field, is strong enough to “drag” vortices in the di-rection of decreasing field strength. This situation is shownin Fig. 6d where we see a periodic vertical motion even if β = 0 . Moreover, if we introduce a toroidal field with ap-propriate sign (i.e. positive in this case) it will act againstthe force due to the current gradient and can reduce the pe-riodic vertical motions in the flow (Fig. 6d, right panel). If B φ would have a different sign both effects would interactresulting in a highly irregular time-dependent behavior.We notice that in the real PROMISE experiment the bot-tom endplate rotating with Ω out (which, after taking intoaccount the directions of rotation and the applied magneticfield, corresponds to the top endplate in our simulations)was made of copper, and the stationary top endplate (bot-tom in the simulations) was made of plexiglass. Thereforean additional asymmetry in the magnetic boundary condi-tions was present. Although copper is a good conductor itshould not be directly compared with perfectly conductingboundaries used in the simulations since the latter representstronger assumptions and induce stronger currents. How-ever it is clear that using insulating material on both endswould prevent additional currents from disturbing the flow. Fig. 7
Profiles of u z ( z, t ) for Re = 1775 , Ha = 9 . , β =2 and rigidly rotating ends with Ω end = Ω out . The criti-cal β c for the corresponding Re , Ha in the limit of infi-nite cylinders is β c = 2 . , and one would expect that thetraveling wave decays. This is not the case for the bound-ary conditions shown here where a clear periodic motion isvisible. Its frequency f / Ω in = 0 . agrees with the pre-diction of a linear analysis for marginal stability in the limitof infinite cylinders, yielding f / Ω in = 0 . . However,the latter approach yields negative growth rate (exact num-bers for frequencies and wavenumbers for the linear resultspresented in this paper were provided by R.Hollerbach whoused them to generate figures in R¨udiger et al. (2006). Noting that the background state for sufficiently fast ro-tation and rigidly rotating endplates Ω end = Ω out is notsteady, it is interesting to investigate what happens when aspiral magnetic field with strength below the critical value isapplied. One could expect that a viscid process (like the Ek-man pumping) excites fluctuations which could then be am-plified and, due to geometry of the applied magnetic field,drifting.Figure 7 shows that for endplates causing strong dis-turbations the traveling wave can indeed be observed evenfor subcritical characteristic values. This is also somewhatin agreement with the experiment – traces of moving vor-tices were observed for states which are stable in the limitof infinite cylinders. We see that the amplitudes of the ver-tical component of the velocity u z are almost unchangedwhen compared to the background state (Fig. 6a, left). Al-though the pattern of the vortices is not very regular, thereexists a clear frequency peak for the vertical traveling wave.The frequency and the drift direction (which is reversed bya sign change of, for example, β ) corresponds to results ofthe linear analysis for infinite cylinders. This leads to theconclusion that, although excitations do not grow due to he-lical MRI, still the same mechanism is responsible for thedrift.If two rings are used and the basic state is steady the sit-uation changes since the additional excitations due to theendplates are minimized. Surprisingly, it is possible thateven for supercritical parameters the traveling wave, al- c (cid:13) stron. Nachr. / AN (2007) 505 Fig. 8
The velocity u z ( z, t ) for Re = 1775 , Ha =9 . , β = 5 and endplates divided into two rings with w = 0 . . Although in the limit of infinite cylinders the flowis unstable, we clearly see that the disturbances (which de-veloped after a sudden switch on of the magnetic field) de-cay. The frequency of the decaying wave is f / Ω in = 0 . which again is in agreement with f from the result of thelinear analysis, f / Ω in = 0 . .though excited for a moment, decays (see Fig. 8). It is stillpossible to get sustained instability by increasing, for exam-ple, β (see Fig. 6b).The reason for this damping might be height of thecylinders which does not match an integer value of the ver-tical wavenumber k . For Re = 1775 , β = 5 , Ha = 9 . thecorresponding wavelength is λ = 2 πD/ . and doesnot suit the assumed aspect ratio Γ =
H/D = 10 . If theheight was changed to
Γ = 4 λ = 11 . the observed decayof the traveling wave was significantly slower, so slow thatafter the sudden switching on of the external azimuthal mag-netic field the wave could be observed with the PROMISEfacility still several hours later. Bearing in mind that wave-lengths for given Reynolds numbers are longer with de-creasing beta (for β = 3 , k = 1 . D − , β = 1 , k =0 . D − ), the constant height of the cylinders ( Γ = 10 inthe experiment) can be an issue when looking for criticalnumbers. It should be noted that due to the boundary lay-ers the effective region where the traveling wave can existsfor configuration with two rings is smaller than Γ by ap-proximately . D (distances up to about . D from theendplates are influenced by their presence).Although endplates clearly can serve as the source ofviscous excitations and the axially traveling wave developsalso for subcritical parameters, we shall notice that there areno periodic motions in the background state. In this sensethe “imperfect” background state serves as a catalyst for thehelical MRI instability. When the endplates are divided intorings the resulting hydrodynamic flow is laminar, and onlyafter the magnetic field is applied the periodic fluctuationsoccur, and, moreover, their frequency corresponds exactlyto that predicted from the linear analysis for infinite cylin-ders. Fig. 9
The velocity u z ( z, t ) for Re = 1775 , Ha =9 . , β = 6 and endplates rotating differentially so that Ω end = Ω .Liu et al. (2006) suggested that the observed fluctuationscan have theirs origin in the underlying hydrodynamical un-steady flow as reported, for example, by Kageyama et al.(2004). In the latter work the purely hydrodynamic flow for Re ≈ with rigidly rotating ends Ω end = Ω out andshort aspect ratio Γ = 1 was already unsteady. We confirmthese results with the method used here. However if longercylinders are used, like
Γ = 10 , the flow becomes steadyfor
Re = 1000 , and only after imposing strong enough mag-netic fields (say
Ha = 12 , β = 6 ) a traveling wave developswith a frequency that matches calculations from the linearanalysis.
We have also performed simulations for differentially ro-tating plates with ideal Couette profile for periodic cylin-ders so that Ω end ( r ) = Ω ( r ) . In another recent work (Liu,Goodman & Ji 2007) it has been shown that for parameterscorresponding to Re = 1775 , Ha ≈ , β ≈ the travel-ing wave decays for such boundary conditions. We confirmthis result with our method, although our treatment of themagnetic boundaries is simplified.The explanation for this fact might be again the inappro-priate height of the cylinders which is far from an integervalue of the expected vertical wavelength. For these param-eters λ = 3 . D according to the linear theory so that lessthan three wavelengths can fit in the container. On the otherhand, if β = 6 is used, λ = 2 . D and then Γ = 10 al-most exactly corresponds to λ . From Fig. 9 we see indeedthat in this case persistent fluctuations exist with a frequencycorresponding to the helical MRI instability. We have alsomade calculations for β = 4 with longer cylinders so that H = 4 λD = 13 . D and Γ = 5 λD = 17 . D . In eachcase a sustained traveling wave has been observed. It shouldbe mentioned that the vortices do not develop very close tothe upper boundary so that it is convenient to take longercylinders. c (cid:13)
06 J. Szklarski: Reduction of boundary effects in the spiral MRI experiment PROMISE
It is easier to perform an experiment showing spiral MRIbecause a much slower rotation of the cylinders is requiredfor the instability to set in compared with MRI with an axialmagnetic field only. Moreover, there exists additional quan-tity, i.e. drift frequency, which is easy to measure and canserve as an important indicator for the associated phenom-ena. It is claimed that in the PROMISE experiment frequen-cies and amplitudes corresponding to the spiral MRI wereobserved, and the results agreed with theoretical calcula-tions of both linear and nonlinear 2D simulations (see Ste-fani 2007) for a review). However it is still possible to im-prove the experiment so that the basic state is a completelysteady flow.In this paper we have presented a relatively simple andinexpensive modification which is suitable for such an im-provement. Firstly, the endplates should be both made ofinsulating material and both should rotate in the same wayso that the system is symmetric in the z direction. Secondly,it is convenient to divide the lids into two rings which canbe attached to the cylinders so that no separate driving isneeded. The optimal, width of the inner ring, in the sense ofminimizing the induced azimuthal magnetic field, is 1.6 cmfor the current experimental setup.Our calculations also show that spiral MRI modes canbe driven by endplate effects even for subcritical charac-teristic values (see Fig. 7). On the other side when provid-ing a steady background flow by applying rings one has topay more attention to the height of the cylinders and to takeinto account the vertical wavelengths of the traveling wavewhich depend on the magnetic configuration. For the cur-rent aspect ratio Γ = 10 and
Re = 1775 it is reasonableto consider
Ha = 9 . , β = 6 which almost exactly corre-sponds to Γ = 4 λ . References
Abshagen, J., Cliffe, K. A., Langenberg, J., Mullin, T., Pfister, G.,Tavener, S.J.: 2004, ThCFD 18, 129Balbus, S.A., Hawley, J.F.: 1991, ApJ376, 214Barenghi, C.F., Youd, A.J., Willis, A.P.: 2004, in: R. Rosner,G. R¨udiger, A. Bonanno (eds.),
MHD Couette Flows: Experi-ments and Models , AIPC 733, p. 83Burin, M.J., Ji, H., Schartman, E., Cutler, R., Heitzenroeder, P.,Liu, W., Morris, L., Raftopolous, S.: 2006, ExFl 40, 962Ekman, V.: 1905, Ark. Mat. Astr. Fys 2, 1 Hollerbach, R. , Fournier, A.: 2004, in: R. Rosner, G. R¨udiger,A. Bonanno (eds.),
MHD Couette Flows: Experiments andModels , AIPC 733, p. 114Hollerbach, R., R¨udiger, G.: 2005, PhRvL 95, 124501Ji, H., Goodman, J., Kageyama, A.: 2001, MNRAS325, L1Ji, H., Goodman, J., Kageyama, A., Burin, M., Schartman, E., Liu,W. 2004, in: R. Rosner, G. R¨udiger, A. Bonanno (eds.),
MHDCouette Flows: Experiments and Models , AIPC 733, p. 21Ji, H., Burin, M., Schartman, E., Goodman, J.: 2006, Nature 444,343Kageyama, A., Ji, H., Goodman, J., Chen, F., Shoshan, E.: 2004,physics/0405123Liu, W., Goodman, J., Ji, H.: 2007, astro-ph/0703525Liu, W., Goodman, J., Herron, I., Ji, H.: 2006, PhRvE 74, 056302Noguchi, K., Pariev, V.I., Colgate, S.A., Beckley, H.F., Nordhaus,J.: 2002, ApJ575, 1151Roberts, P. H.: 1967,
An Introduction to Magnetohydrodynamics ,Longmans, Greeen and Co. Ltd.R¨udiger, G., Zhang, Y.: 2001, A&A 378, 302R¨udiger, G., Schultz, M., Shalybkov, D.A.: 2003, PhRvE 67,046312R¨udiger, G., Hollerbach, R., Schultz, M., Shalybkov, D.A.: 2005,AN 326, 409R¨udiger, G., Hollerbach, R., Stefani, F., Gundrum, T., Gerbeth, G.,Rosner, R.: 2006, ApJL, 649, L145Stefani, F., Gundrum, T., Gerbeth, G., R¨udiger, G., Schultz, M.,Szklarski, J., Hollerbach, R.: 2006, PhRvL 97, 184502Stefani, F., Gundrum, T., Gerbeth, G., R¨udiger, G., Szklarski, J.,Hollerbach, R.: 2007, astro-ph/0701030Szklarski, J., R¨udiger, G.: 2006, AN 327, 844Velikhov, E.P.: 1959, SJETP 36, 995Wendt, G.: 1933, AnP 409, 445Youd, A.J., Barenghi, C.F.: 2006, JFM 550, 27Zikanov, O., Thess, A.: 1998, JFM 358, 299 c (cid:13)2007 WILEY-VCH Verlag GmbH&Co. KGaA, Weinheim