DIFFUSION OF A PLASIA AC£OS A MAOMNST C FIELD a~IaD STsPH FIN B., NOasachusetta Isitute Of Teehnuosq (1959) SU~UTTED IN PARTIAL FULFILISW OF THE RSQUIRMWS F THE . ss. INST MgROF MASTR OF LIBRAREY SCIg c LINDGREN at the ASSLOCUSETTS INSTITUTE OF TECHNOLOXI lOt .oJa Aut96 1 w en4dAgIphsics, Hy 29 1961 of Gm0 CerLtied by . . n .1 Accepted by. . ... * Thesis Supervisor - * * * * * * * Chairma, Departmental Committee on raduste Students Abstract An invetigation of the problem of plasma diffusion through a magnetic field is made through the macroscopic fluid equations. The problem is set up in a simple one-dimensional geometry of a semi-infinite half-space filled with plasma facing a semi.infinite half-*pace containing a ~unir magnetic field, and the equations are used to describe their subsequent behavior. An attempt is mede to solve the equations analyti- cally, but their final solution is obtained numerically. The solutions are also used to determine temperature and entropy increases in the plasma. Absi'omt I* Intdmitoz .U,*OsawaI MI. Dismusio 2 Ths Infinite On.. Dimmrnanu IT,* AmIA17tma Intogwation To rmwio* Iwto sz can 7 17 20 Thare hasbson very reomtly a greaty increased interest in sientUM and engineering applications of tonised gases and plasmas this has dwsnded ftom known theory a quantitative idea of how pL ase behave under given enditions. An example of these is Vt. problem of plasm oontainment by a sagnetic fleld; or, stating it more generally, the inter aioGn of a plasma with a magneti k tild Mhis plasmnfield interaetion has a B generated by external sources. whrth is the subject of this discussion, ofaspeets that complicate its analysis; tbe complications resulting in soulinear differential equations. As a mcnsequence, the inmatigation of this interaction is limited to a simple example; this is to make the equations consernd tractable enough to yield quantitative data with a reasonao amount of effort. To illustrate, the processes considered can be crndely deseribed as 'Naaw*mations of plass through 'strong' magnetie fields, in which the electromagnetic and thermo properties of the plasse do not change appreciably. 0ao What is meant by mations and "strong' fields will be made clearer further on. In addition to this, ti. form. ni c geometry of the proble is reduced to its simplest Doing this naturally limits the applicability of the resultine data;, st*il it produces a good approximation of ti- saoan in a variety of cases. plasmanef d inter- Te formal stateant of tats prebla most amenablo to qualtative investgati n is in terms of the no*alUd tion. seroeeeapt esentialy these are the equationa that re equations of ltfrom averaging this averaging being done by integration thie mtions of many particle y Mseas of this integration of thn bitUman eqat±*t in velocity spaceo. as average iSid velocity %#and pitser(2) defines a ou~r t dnsityn derives an equation of fluid ation and a generalised tfr In the case to be eomposed oa oansidered bore the fluid to asamed to be of sangly chargi d ions of identical to generalse to ta ation, but itSt mber o asses and of elecbtrone types of ons does not coplicate the sito s out that this does not lead mstin. A mnaoasd" gas as *eeea simply to -* lalations done later. eqaol to 5/3 in of Ohs's law. say additional Inftr- the gas oonetant f 1Fr this caSen then, the equation of motion take the fram thare B is the agnetie field strength, p is the seaar pressure, and p is the mass density of the iosn, the addition of a Xj I This is asi ly itler's equation with f raOe. There are certain assuamptions aready Implicit in this equations mass of the eleotna the aW the gravitational fOrce are both negleoted, and the amber densites of the electrs and iLns are taken to be exatly equal at all points in t3e fluid. B ssee is never an eleotef pace charge. field dua to of this last assumption, there Also, viscosity eaffocts are neglaote llowig what is actually the stress tensor to be tsan as a sUcar pressure. Aaoounti dae to external ma4te"g ditferent for direott al pressure differenes c fields by wdag a stress tensor with at least nournese diaoaal terms might seow actualy anessary for this tip. of problm differen howavw, it rill S rn out tat t this ossawt e in the saculations Also for this case5 the "J where 1 IsoSti = ensrased Oha's ]A E + reduces to the for kxB eectrical reisatvity and e i twe electric field. assuaption made here is that the resistivity is a constant scalarl again, rwiti it fields & e not affer On once as a diagonal tensor to acooOunt for external magnetic t the calculation. It must be admitted that ae. g~eMg its dependence on the current, tSperature, and fluid density considerably limits the -usefless of tU solution; nevertheless, this is the first assrupton that will have to be made strioe1y 9br simV-pi ficaton. Anther eassumptan is that a nall compared to the othersl terMn is neggibly here the limitation of salowaser is first applied. The two equations above resulted from the "plasma properties of the fluid in this proble . The reumining equations to be introced are of a sore general nature. Flrst i p + there are the continuity equation (P " ) = o and the equation of .ats = PRT here R is the gas eonstanat a T ir he Sapertre. In addition, thare are the Mawsell equations and J V-x8 = The arrnts due to polariation and the tia rate of change of the electric field have been mngleotd as being smll oapared to the magnetisation curret watch is incm pawebility ded in the V ierm in this last equation. 8 The o is taken equal to that of free space. ?lnUy there is the energy equation j + >t LN( f - owlm heat mapacity and Y is the ratio of spoei io where , it the oWnsta-to heats, This equation implios two vry aportast assumptionst She or 0 sour a t energy is ohbie heating ( viscosity having alread been n glected )g and the ther, i*o, one, that hat there is no heat transfer in the fluid, all proeeose are a4abatia. This lat asaption is valid when heat s an rtci ditaeio to be demwibed. than the preoe ear mw tabulating the equatims obtaind Se( t) = j x7 -v xg + '14 t (p) 9t + , = pRT = 0 j3 (Vt V xE ,AAO Cv -V it is seen that tha*, soy als ads seven equations , pT ntaining sevn viaria e. It be ssn that tin #quations re nnlin ar, wvt4ch Iamdlately pro- anW strn4ahtgbrward analytical integration of them. with theria Jbenfw tio Thaus proceding reqUires some additonal sipificateions. continuing wita thi 9, it is astractive to rougly sketch seas pImpertis oof pleas irn etrong magneti* fteids predued by constant exernal earrents. ere"strongfmaget i. fields" means moh larger than sq field met uap 1 order, gas. ne mfor ti curaents in the plea. e line tend to be 'fse" tfieda that are to tie fLrst in a highly oudtacting If a motin is imparted to the gas, the foree lime are wdragged along* wih it. resistivit field lias. To the next ordsr, trrent dissipation due to the finite of she gas allows a relatve aetiet To see th smea between the gas and tbe clearly, one can lmagine a claster of s* ternally produced parallel magunee tield lines in a plasma with a finite resittvity. in it Beause tbe pleasm will have to decay if Iag he fields. it is resistiv*e is not regenerated by externally intaesi4- Tias the number of flux limes passing through the area will have to be reduced, and if their total number is oonstant, some of them mast appear outside of the area. ard what the whole proeses 'orks throug the crrant around an area the plans Therefore, the cluster spreads out like' is a diffutsion of the field line WV~* O2flflI S&" ,p s1 .xn do. x iwyd tam WqR 0#UUf*J fruv140z-T WTfll*~ pu*Vs s~ou q.'t *%wzr 4n p~jo 0 em C17 f0U Wtiwr 01o% polf tft") **yp*4 WIWood CWn'~d P WqSVMWv MonVad np wte sn am~ 3uoy*^eUT A0w n ppgq~ oqj **pnc r&"vbu matvPuwunx. o * i tv 06 Awwwr cii % -'' znraw ej;#.r Mq uy pwpa;4 OffasS wOW'~ 'eanclw0* tzoyvi rawqo"luona AmO saq 0% PS*ATT IOU wr umoid "q& 'O*pvOg ay~p;.d a"t 2Iwu £^'pp o% .mqd au \vli £*TV#wn~ n MAn 01M ST'R So fldowo F,, pO U*OTfl u # snpstT V; n~ OTuIadS V ERt VpfoOo #flbmdI ps NfliD ^dT*msp :0vv " OuMoMua ,x pufl iri TMnft 0%tOt.d am n4 %'Ind fl0 30enw Sit pwrW U anThIT -~,VWM-, Ot'l"01MTO - covertfg the x * 0 srfte of the pl as. To seek a solation for the behavior of thisyastm for timest one £irst eonsiders the equation of nation ( ) S+ Jx 0, - Sice the plasm is always uniform in any plane perpendiclar to the x-eais thUe V ater redee to /a x . Also, if the l-field is always onaslderd in the s-dLrectiou, the not currents in the plasm always £Low in the yedirootion. Then the equation of motion becomes + r x =x New it may be men why it was not neessay to regard the resistivity and the pressure as tensors JEven if the tensor form had been retained, only one term in each aould have been used. At this point the mriterin of notion of the plasma. This smea eslowness" is again applied to the that the teras on the left are negligibLy small, or that the ters on the right nearly cancel. J ax Sinee for this configuration the equation -A o Vx x Thbs becae s oaly hao a twntional dependoenc on the xn oPrdnatetela equatSon of motion then becomeM and this may be integrated to give = + Tbe r/2 oy be oonsidered asa ua pressure at all pott preseures vary. in - tite pressure0 , and the total pace remains constant as the manetic and plasm Sitqly from this equation it plase leaks alowly into ti poetAngl CONSTANT is possible to see that the magnetic field while the magnetic field corre-* leaks into the plam. Neither the pesma nor the field has a sharp boundary at t > 0; and the current setting up the magnetic field boomes distributed through the plasma rabter than being eonoertrated in a shtra . A plot of plame density and field strength might the to look l It. ( fIguer 2. ). Fiare 2, be expected 10. The next series of stops in seeking a solution to the equations consiste of redicing than as far as possible. The calculations are first siepli- fled b- tusing the facts that the magnetic fields are only in the s-directon and the curaents and therefore electric fields are only in the y-direction ( there are no external fields )g and that all quantities depend funotioally on the xdirecbtone 1 Oha's law then becomes = E + <B and differentiating it with respect to x yields +t Jle ± 'b _- The cont tuilt equation becomes ~e 4 )&± JN + and the other Nswell equation becomes &Sabstattng for from the NaxImal / ivrom the continuity equation and frr equation into the differentAated for 1; -n + - e ~td X of Ohm's las yields aO labn ubeibStUgfor 3 ron the first PanLl equatin yields xl) A. AO- 4a 3 u 7 ata whish v$l be called equenoc 1). Sw the proe.as ea of re*otcen is carried tbroug starting from the ecaton At AxpT s t jp sd substituting 5/a Maupliss 4u &tet tutingp yoidtsaeite . W ntsis the ) 2) or Y" txlandj Y ? fro te iedn et duera - a o ta aiig Ge~, ex GC) + ±t by -' asrso fOr P in term of b3 deri'ved from the eqUmo ubettted, yielding :(z+ ; t.s$ go G_ ±v±>= P (273 of _Y Y- whieb will be called equation 2). Staple subsiaticm produces no further redation after this point, U4. tvenaforNUa *ad tb* pmp.. most be oontiwed b~y thwt1wie owdmsI lot = thu. 3T3 e @a La' -+ t )T DOI ~tu man of .. ti0~,u Tm ad s"bauutdzg +~ that + j %"tpxaly in .qattons 1) cmd 2) yieldo " of 13. - /-I "- #AT using the 0seasionless variables a aMs tmornntm Uen has served its variaMel niv axl intent by eminatng but it bae also had other effeots in changing the is t alttf I'tabr equeeeod* /is axis. The ompared to the real x- xis when the proporton- *mal. Also X is neaer less than ero Iease the defining itegral is neaver angatiwr tive vales on the as a What this mnans rloprrspod to sial ee is that large asep* poetive values on the %Z is$, wt4le large positie valaes on the saxis arrespond to large po81si tive va~es on the Xmaos. tf ions c uartherrwe it and A are determined as of 1, tXe invesfm trasortbao 0 des not loate them 4wth reference to the original saeordate. reason for this is that the iesttrnnsfmatioaea The offetod by nying 14 is also true tha t tAt it + Thus al _ the inverse tranaformation can do is replot the functions in At least, however their proper sca~l. this trasxormation has reduced the problem to two equations in two unloowas, and the remininnt disoussion is involved with trying to solve th". For further oamvninoe, the two equations will be rartten in tere of one variable not to be conAzsed with the scasordinate, so thst and _ A - . f _ ~ + _ Equation 1) then becomes 4 - -I t( ta7 t-tTy L r 3 15* ad eVanding the tmr _ /2-I on tho left yields A + I)- "%C) - I ) 4---- ~ X4- A~ 4L~+ Tkas equation 1) beuwMes Orepdingly#r equation 2) beecues ".f-. I .- %( Af V4-t ) 4) hfI" If- YX =A or + Thus, writg primss for { Oj / - o + ± Q-~± ~/ # / x it may imedlately be shown that these equations lead to a process Taking the limiting case that at least resembles ordinary diffusion. of an incompressible plasma i.e. g -- 1, g' -- 0, and ( -- o0 causes equation 2) to become indeteraminate and equation 1) to become Zif/ = -j Rearrangig this as I = (LN ') / _ and integrating it yields _ / and integrating it again yielde Thus f in this case is of the form of the error function. What this ma .ais that the plasa sits perfectly still, being incompreaible and infa-kiE and the magnetic field diffuses into it. This is exactly the process undergone by a magnetic field diffusing into a motionless metal condActor. 17, IV. totQ fSlb Iatgration Ieomse eqaations 1) and 2) are nonlirar second order differential equations, they cannot be straightforwardly integrated over all values of s. What can be done, however, is to look at the forms of theso equations as f and reasonaig. g uasyttial approach values determid from physiqoa rSmmabriag that tlhe functions are n in terns of the transfrrmed ooordinate X, it may be seen tha att large as one and f approaches ro. ono and g appmroahes sa,. g a proach Oonweieyp, vWry near the origin tf approaches Ths sratements may be made in light of the test that, at large distances from She origin of the real coordinate the magnetic field and the plama shoald each be relatively undisturbed if only a finite tim has el sed. With this muach knowledge in mind about the behavior of the variables, an attiapt may be made to at least partially understand the equations. Taking equatton 1) Integrating SS/'3 I _ )-~1 i- 9. 1 1 18. Integratng again + f - Both f' and g' go to sero at large sg i +;& A' thus it is dffticult to discuss the behavior of the rightsnd side of the exponent at this limit. heless, Never- it might be expeted to go to sro, since the resmining terms give an expression for f similar to that in the case of the incompressible fluid. This is what might be expected at large s where g is nearly un- disturbed. Ixpressions for the term gf' may be derived from both equations. Equation 1) ay be writt= whish integrates to Also from equation 2) SC) 19. Solving tais quadratic both of the mqa be integrated to give expressions for f. Unfortunate ly nes of the above yield azy nw informotion about the f netional form of either f or g. It tarned out that none of these methods ere roally helpfu in un- ooupling the equations to obtain an expression for either f or go They either led to omplicated integral fors or approximations too crude to be tuserl. It is possible that methods for treating nonlinear differ- ential equatlons that at least lead to further simplifloations in this asoe ight exist. Such methods are dia Lplian()j howvear, their spli sed in texts like Ieo(1 ) and ton is suffciwntl difficult that no attmpt was made to use them on this problem. Aowther possible method of investigation is to substituto various woll-known fmctions for the variables and see if they balance the equations at asystotic values. A great deal of time and effort was spent trying this, and it proved unequivoeally fritless. functions, though they The actual ight be wllbehaved, are of aufficient complex* ity that no Olosed funoLon reembles either of thesm This is llus- tnated by aetag Uhaot, in the greatly sres*rotive case of an incompressible plasma, f still had the fora of an open integral. 20. wIaieJ l Interton m y In order to obtain specific data on the plaia-field interaction, it was necessary to resort to numerical integration. The prooedre used is described in KEplan, and simply consists of selecting values of the functions and their divatives at a points, casulating their values at a point close and iterating. i ty V Thus the method is a nrmerical first integrationg and since appears to the second order In tie equations, it is necessary to intoe grate t4See to obtain values for it. S 1( / / The equations are writtn in the fore / / / -=- wher the first equation is equation 1) and the tird is equation 2). In doing he oaJclation, values for f and g and their derivatives were eflected at large positive a and the iteration earied toward the origin. These initial vaLues wre adjusted until f approaohed one and g aproaoed e at the origin and both held their asmtotic values at large a. It would seem more acurate to start at the origin and look for asyatotic values of the funotions by iterating away from the origin, but it was 21. ezrtrtely dtffioult to oestIate the first derivatives at this point. The results of the numerical integration are plotted in ( Figure . ). The variables are left in the dimensionless form of f and g and are plotted as functions of the transformed coordinates. These transformed oordLnates were then reoast in the form of the real coordLnates by means of the inverse transformations xn = Ax This was also done numerically. T The variables f and g are thus also plotted as funtions of x and t in ( Figure 4. ). viously, the origin of the real x-axis cannot sero-point in ( Fgur 4. ) is arbitrary. the napeti As was mentioned pre- eo redetermined and the These figwrs thas detAerine field strength and the plama density in both space and time as the diffusion process takes place. y also calculating temperature and entropy changes in the gas# its total behavior ma be considered ade- quately defined. The temperature of the gas at each point was calculated from the x-exis ourves of f and g as follows# sinco them -r 13 T-~--PC R 22. Chmaging to dlmsienlAeas variables (I- ~7 -- o'k "T=~cop i ) 5. ) with wich in plotted in ( Figure 'L Since tk prOoess was asummed adiabatic, th speific entropy was eJ.ualated from the xmaxis crves of g and f by using the expression 5 - S o 0. C-I e -*1- - Cr 'j ' Changing to daillmelse variables L- - 5o +. ±t CV 'h _____ <t, N _ +,+r r + CV N ( I-Z Or, subtratng *,and dividing by c ~ - ~wta)9/ Then the increase in entropy as a t- "(I, (fL tanction of x is giive by vhich was oal"ulatod and plotted in ( Figure 6. ) using Y * 51/3. The total increase in entropy as a funtion of time was obtained by graphicagly 23. integrating the Mr"n in ( Pigre 6, ), nd wa s found to be All of the caloultdons for these graph aocuracy. Because of tie, are only to s-ldesrule and because of the errar involved in s*t- ating the boundary oonditions for the f arn g carves in ( Figu 3. ), the curve shapes and the derivod quantities nay only be oansidered to be carrect to soewhat bettter tan an order of magnitude. acuraq would requte the services of a computer. Gratr I I I -I m-i 4117±f4I" _ -l 1 11 -1 ++H+ H' j - __117 I-~4 _ _ -i"T<- _ T Hf_ ---- -T I,4J T _ _ _ _ H 4, L-+ F-TFFITF:T *11 + -tV - rT. ITM i FF 1 ~ fill -F _ -h - Ii t II III p Jill fill + 4: - LFr _T II TFFr I I_ TF V' JL T- E_7 7---I H4~ H +4T-rj -- n I III L U1 U - iHH 1+ 1 ++ -M :4 ir-::Ft _H_ 4I 4774 FWLH I AI 1- r4 -T + F -7-Fl- r A I [ FFI + 7+ + I T 4F T 0 25. 1±] s .J iii-', KTiRH K17 1, HE- K -4114 '1K TEFTi 27. 1. E. L. Inoe, Sutiono OrdiaMy Ifrential Dvemr Puthications Ino. Nw York (1956). 2. Spitzer , gLnom Q FUll ~, ! Gases, IntersoaJnce lonised oJr., Publishers, Inc., New York (1956). 3. H. W. Liopmann and A. Roshkos Silents of Ga"nmsa, John Wiley & sons, Inc., Now York (195). 4, Wjilred rdinary Diffterential quatons . lan, dAddison-Wesley Publishing Company, Inc., Reading, Mass. (1958). 5. 6. s. Chandr'asekbar J, O. Linhart, Plans P? Chicago (1960). P Pbi,, Amsterda (1960). icgs Uni rity of Chicagp Press, NPlano th olland oor Publishing Company,