Critical-state and cylinder magnetization geometries of type-11 superconductors in rectangular slab T. H. Johansen and H. Bratsberg Department of Physics, University of Oslo, P.0. Box 1048, Blindern, 0316 Oslo 3, Norway (Received24 October 1994; acceptedfor publication 20 December 1994) A scheme is described for analytical calculation of critical-state magnetization 11/1of superconductorsin the geometry of long rectangularslabs and cylindrical specimensin a parallel magnetic field. The simplicity of the general scheme is demonstratedby deriving compact expressionsfor the ascending and descending field branches of M in the exponential model Jc=jcO exp(-BIB,) and in the Rim, Hempstead,and Strnad model [Phys. Rev. 129, 528 (1963)], Jc=jcol( 1 + BIBo). The analysesfocus on the vertical width AM of large field magnetization hysteresis loops. While Bean’s result [Phys. Rev. Lett. 8, 250 (1962)-j,J,mAM, today is used extensively to infer the critical current, it is well known that the method lacks consistencywhen a field dependenceis seenin AM. For the two models it is shown explicitly that in the expansionof the functional relation AM( J,), Bean’sresult correspondsto the lowest-orderterm. Also to the next order in the functional expansionwe find a unifying form of expressingthe model behaviors.This term containsthe secondderivative of J:(B) with a prefactor that dependson the samplegeometry. A model-independentproof for the two first terms in the expansionof AM(J,) is also given, which allows the significanceof size and shape,i.e., thicknessand aspectratio, to be discussedon a general basis. New methods to extract J, from AM data, one of them without having to invoke specific critical-state models, are indicated. 0 1995 American Institute of Physics. I. INTRODUCTION Whenever flux pinning centers are present in a type-II superconductorthe distribution of penetrated flux will be nonuniform and dependon the past magnetic history of the specimen. To describe this behavior, Bean’ introduced in 1962 the critical-state model, which in recent years also has been applied extensively to studies of high-temperaturesuperconductors.In this model one assumesthat the local current density, which by Ampere’s law is associatedwith the nonuniform flux distribution, has a certain magnitude J,. This critical current, which originally was taken as independent of the flux density, has in later refinementsbeen given various dependenceson the local induction magnitudeB. Up to now the most frequently used B-dependentcritical-state functions are due to Kim, Hempstead,and Strnad,2 J,= JcO 1 +BIBo’ and Fietz et a1.,3 J,= j,ge-B’Bn. Here jc, and B, are positive parameters.See,e.g., Ref. 4 for a review of J,(B) functions introducedin the literature. Even more model functions are discussedin Refs. 5-7. Measurementsof the irreversible behavior in magnetization has become a standardmethod to determineJ, and its dependenceon B in a superconductingsample.On the other hand, one has today no meansto derive J,(B) directly from a given set of data for the vertical width AA4 of the magne tization hysteresisloop. This is due to the fact that A.M is not a function of J, , but rather a functional. For a given magnetized state the magnetization dependson an integral of an J. Appl. Phys. 77 (8), 15 April 1995 a priori unknown function J,(B). Thus, in order to analyze data one relies on the applicability of a chosenmodel function which also should be integrable. Most commonly one uses the simple result of Bean’s model, namely, that J, is proportional to AM, or J,===AM/d, (3) where d is a length characteristic of the sample size and geometry.A measuredfield dependenceof AM is then used together with Eq. (3) to get the B dependenceof J,. This method is not self-consistent,and errors are inevitable. The inherent problem has been discussedby several authors,4.8 and it has been shown that Eq. (3) gives misleading results for small fields if the sampleactually behavesin accord with Eqs. (1) and (2). A second method is to choose a particular function J,(B), from which one computesAM and fits model parameters in a consistent manner.6V7 For this far more attractive approachto be versatile, one needsthe solution of different models in various geometriesat hand. Based on the Kim and co-workers model, Chen and Goldfarb4 derived analytically the magnetizationcurves of an infinitely long sample of any rectangular cross section placed in a parallel field. They also showed that the results are applicable to other simple shapessuch as the circular cylinder. In their treatment of A&f versus applied field H they found that for this model Eq. (3) is modified according to AM(H) 7=J.‘H)[ l+s( ~j*++ where 2w is the samplethickness,and s is a numberdepending on the geometry.Chen and Goldfarb state that s = l/4 for 0021-8979/95/77(8)/3945/8/$6.00 D 1995 American Institute of Physics 3945 / / ILH In this work we consideronly systems with isotropic propertiesin the xy plane.As arguedin Ref. 4, a valid assumptionfor this caseis that the fiux penetratesequallyfrom all sides.This implies that the current flows in a patternof concentricrectangularloops in the xy plane, where each loop is equidistantfrom the externalboundary.By definition, the magnetizationis evaluatedby integratingthe magnetic momentof theseloops divided by the total samplevolume. This is expressedby W -X a! M=--T / an infinite slab and ~=I/20 for a cylinder, and conclude thereforethat the error in using the simpler Eq. (3) is much smallerfor the cylindrical case. Applying the same schemeof calculation,Chen, Sanchez,and M u iiozglater derivedexpressionsfor the magnetization using the exponentialm o d e l. That article, however, containsno analytical discussionof AM. The long expressionscontainingnumerousnesteddefinitionsof terms make suchfurther analysisdifficult. In the following sectionwe describea different scheme of calculatingthe critical-statemagnetizationfor the general slab and cylinder geometries.The schemereadily leadsto compactformulas for the ascendingand descendingfield branchesof the hysteresisloop. This is demonstrated in Sets. III and IV, where the two m o d e ls, Eqs. (1) and (2), are treated. In the discussionof the results we focus on the analysisof AM, which is the central quantity for comparisonswith experiment.For the Kim andco-workersm o d e lwe confnm Eq. (4), althoughwith a different value for s in the cylinder geometry.W e also derive a similar relation for the exponentialm o d e l. A m a in result of this article is to show that the secondterm in the expansioncan be expressedin a unified way, i.e., written on a m o d e l-independent form. In Sec.V this result is generalized,and we derive the two first terms in the expansionof the functional AM(J,), valid for any critical-statefunction J,(B) proposedin the literature. I ! i I II. MAGNETIZATION Consideran infinitely long superconductor of rectangular crosssection.Let 2w and2w/a! denotethe short andlong side of the rectangle,respectively.The parameterQ , which is the aspectratio of the cross section, can have any value within the range O=%Y=G~, where rw=Ocorrespondsto the infinite slab geometry.W e use a coordinatesystem aligned with the sampleas shown in F ig. 1. The appliedmagnetic field H is directedalong the z axis, and we defineB, from H=(B,/&. 3946 J. Appl. Phys., Vol. 77, No. 8, 15 April 1995 (5) 4w- I 0 Ab)j,bW. Herej,(x) denotesthe currentdensitycrossingthe x axis at a given positive x, and A(x) is the area enclosedby the currentloop, Ajx)=4x FIG. 1. Cross section of an infinitely long type-II superconductor placed in a parallel magnetic field. In the critical state the supercurrent flows in concentric rectangular loops, each loop equidistant from the external boundary. w i-,+x (- i . Note that in Eq. (5) the sign of j, with respectto the y axis correctly gives the sign of M . W e want to determineM as function of appliedfield whencycled througha m a jor hysteresisloop. As usualwhen discussingcritical-statem o d e ls,we neglectthe lower critical field and surface effects. F igure 2 illustrates then the sequenceof different situationsthat occur during sucha cycle. Except for an interval immediatelyafter the field sweepdirection is reversed,which we o m it in the presentdiscussion, the flux density profile will always be such that x(B) is a single-valuedfunction whereverjY#O; thus, B can be used insteadof x as integrationvariable.From Ampere’slaw, j,dx= - dBlpo , (7) andEqs. (5) and (6), the magnetizationcanthereforebe written as X[l+a(X-l)]dB. p&f=- I BCl BtIl Here X=x/w, and B, is the induction at the m idpoint. The profile function X(B) is obtainedby integrationof Eq. (7), which gives (9) The current densityj,(B) has the m a g n itudeof the actual critical-statefunction J,(B), and the sign is the oppositeof that of the slopedBldx. One seesthat in order to calculateM analyticallyin the presentgeometrya requirementon the critical-statefunction is that its inverseis twice integrable.The secondintegration involvesthe squareof the first obtainedfunction, which representsthe inductionprofile. A m a jority of the critical-state functions proposedin the literature satisfiesthis condition, and in fact leadsto integralsthat are easily solved. The aboveschemeof calculationis differentfrom that of Chen et aL4,’ for the same geometry.The connectionbetweenthemis seenby integratingEq. (8) by parts.This gives the alternativeformulas T. H. Johansen and H. Bratsberg as a dimensionlessparameter.Moreover, throughoutthis article, wheneverinduction symbols are typed in lower caseit indicates that the quantity is normalized by Bo; we define ba= B,IBo, etc. Ill. EXPONENTIAL A. Increasing MODEL field Assumethat initially the superconductorhas beencooled below T, in zero field. As the field is steadily increasedthe penetratedflux will, in the part of the sample used in the calculations,have a density profile with dBldx>O [see Fig. 2(a)]. Thus, jY= -J, where in this model J,(e) is given by Eq. (2). From EZq.(9) we obtain the profile function Increasing x+; Ba B”eB’~~odB’/Bo = 1 - $a - ,&y/k. s6 03) ln the first stageof magnetizationthe flux entersthe sample only in an outer layer of thickness S. This penetrationdepth is given by b) s=l-X(b=O)=(eb+l)lk. (141 The initial stagelasts until b, reachesa value b, corresponding to complete penetration,or 6= 1. It follows that Decreasing BS b,=ln(l+k). (15) 1. O<b,sb6, The virgin branch of the magnetizationcurve is found from Eq. (8) setting the lower limit of integration equal to zero. With X given by Eq. (13) one obtains 1+a -[I-(1--b,)ebr]+$$l-4e”a k WOM FIG. 2. Schematic picture of the local flux density distribution in the sample along the positive x axis (see top) at various stages of a magnetization hysteresis loop. -=-b,+ Bo +(3-2b,)e”ba]. (16) By expandingthe expressionin powers of 6, one finds poM= -B,+ 1 I B(X)(2aX+ 1 -a)dX, 0 (10) POM w--o-=.-b + ‘+a b;+... a 2k Bo , (17) which shows that the initial slope of ,@4 vs B, equals - 1. 1 -a)dX. ill) One can benefit substantially by properly choosing the schemeof integration accordingto the aim of the analysis. As pointed out in Refs. 4 and 9 also geometriesother than rectangularcan be directly included in such analyses.In particular, the case a= 1, in this article, applies to both squareand circular cross sections,with the cylinder diameter being equal to 2w. In the following two sectionswe apply Eqs. (8) and (9) to derive M(B,) for the two important critical-state models, Eqs. (1) and (2). In both caseswe define 2. b,ab,, When the applied field exceedsthe full penetrationvalue the midplane induction b,n, becomesnonzeroand dependent upon b,. We find b, as the b satisfying X=0 in Eq. (13), i.e., b,=ln(eba--k). To obtain M the integrationbecomesthe sameas that in Sec. III A 1 and with the nonzerob, we obtain a k = ,uunjcow/Bo, J. Appt. Phys., Vol. 77, No. 8, 15 April 1995 (12) (18) o! - 1- 2- + k eba. T. H. Johansen and H. Bratsberg 3947 B. Decreasing field The situations under discussion are illustrated in Fig. 2(b). When the appliedfield has beenloweredfrom its maximum valueby a sufficient amountthe induction gradienthas changedsign throughoutthe sample.In these situationswe havej,=+J,. From Eq. (9) one sees immediately that changing the sign of j, leadsto a new profile function given by Eq. (13) with the substitutionk-+ -k, i.e., x= 1 -(eb-ebqk. (20) Again, settingX=0 we find now I b,=ln(eba+k). (21) Since k-+-k in both X and b, , the same substitutionapplies for the entire expressionfor M in Eq. (8). Thus, from Eq. (19) we immediately obtain (22) 3 I -2 I -1 I I ia I I I 1 I 2 I I 3 FIG. 3. Magnetization curves for the exponential model. The branches for ascending (lower) and descending (upper) field are calculated for (~=l, 0.5, and 0 (infinite slab). The virgin curves merge with the major loops at the point of full penetration, b,=ln(l+k)=ln(2). POM 1-ff -= --k--[1-(1-b,)cbm]-s[l-4ebm+(3 Bo -2b,)e2bm]- 1+a - k [l-(l+b,)evba] 2. -b,,G b,GO In this interval the induction becomesnegative in an outer regionX>Xa [see Fig. 2(b), curve (IV)]. Since J, depends only upon the absolute value of the induction two separatefunctions describethe profile. Note first that X0 satisfies X0=1-6(jb,])=1+(1-e-b~)lk. (23) In the inner region, 0 =GXGXa, integrationof Eq. (7) yields 1 x(i) = __ -=-(ebm-eb)lk. (24) POW The midplaneinduction is found using that X(‘)( b = 0) =X0, which gives b,=ln(2+k-e-‘a). (25) In the outer region, XaGXGl , Eq. (9) can againbe applied. Since B<O we set j,,(B) =J,( - B) and obtain (26) The magnetizationintegral, Eq. (8),:now becomes I poM=- ’ X(‘)[ 1 + a(X(‘)-- l)]dB*, I 0 and we find 3948 J. Appl. Phys., Vol. 77, No. 8, 15 April 1995 (27) (28) where b, is given by Eq. (25). 3. b&-b, It is evident that in this range of b,, illustrated in Fig. 2(b) curve (V), the physical situation is the same as for baa b, , except now the direction of the supercurrentis reversed.Indeed, the entire magnetizationloop possessesthe symmetry M(-b,)=-M(b,). In the specialcasesof cr=O (infinite slab) and CY=1 (cylinder) all our expressionsfor M coincide with the results given in Ref. 8. Fire 3 shows the ascending and descendingfield branchesof the magnetizationloops obtainedfor the aspect ratios a=O, l/2, and 1. The plot covers the range -4b,~b,~4b,, and we have set k=l. ‘By this choice one can seethat our curvesagreewith Fig. 3(b) in Ref. 9 (Ref. 9 usesthe full penetrationfield and not B, as normalizingparameter). C. Width of hysteresis *Cl X(O’[ 1 -t a(X@)- l)]dB, -$[1-4e-b~+(3+2b,)e-2b~]-b,, loop In Fig. 4(a) we show how the vertical width, p&MIB,, betweenthe branchesin Fig. 3 is reducedas the appliedfield increases.When plotting the samedata with a logarithmicy axis, as in Fig. 4(b), it becomesevident that the curvesform a set of nearly perfect straight lines over the entire range b,>b,. This suggests that in this region AM to an T. H. Johansen and H. Bratsberg which indeedcan be expressedas AM=( 1-~)wJ,(B,)[ l+&E(ke-ba)2+*** 1 1’ (3 1) It is clear that as 6, exceedsb, even the lowest-ordercorrectionterm very rapidly becomesnegligible.In fact, already at b,= b,=ln(l fk), whereone can show that the total correction is maximum,the term has the value m" 0.4 2 4 k* 1 5-30~ 10 3-a (l+k)” 9 -~ 0.2 0 2 I 3 ba (a) W ith k= 1 this amountsto l/24 for (Y=O,and even smaller for a=1 where it becomesl/40. Note that since‘k2/( 1 + k)*<l the correction to the leading behavior AM(B,)~J,(B,) can never be large at b,=bp, and it always decaysexponentiallyfast as b, increases. Becauseof the rapid convergencetoward a straight-line behaviorit shouldin practicebe possibleto judge critically from a log plot of experimentalAM datawhetherthe exponential m o d e lis a properchoice.Providedthat the measurements include a range abovethe full penetrationfield, and that this m o d e l is a reasonableone, the high field part of a log plot is fitted well by a straight line. The parametersjCo and B. can then be found from the line’s slopeand intersection with the vertical axis accordingto ln(AM)=ln[(l-cu/3)wj,a]-B,IBs. (32) If one finds a significantsystematicdeviationfrom such behavior,other critical-statem o d e lsshouldbe considered. IV. KIM AND CO-WORKERS MODEL -0 (b) 2 1 3 A. Increasing field In this m o d e l J,(B) is given by Eq. (1). The profile function, the analogof Eq. (13), now becomes ba FIG. 4. (a) Magnitude of the hysteresis in magnetization for the exponential model. The graphs represent the vertical separation of the magnetization branches in Fig. 3. (b) Logarithmic plot of the curves in (a). The straight lines appearing as distinct near b,=b, are the high-field asymptotes. Note the rapid convergence. extremelygood approximationis proportionalto the exponential function J,(B,), with a proportionality factor depending on cr. To verify this analytically we use the expressionsfor M found in Sets. III A 2 and III B 1, which gives X= 1 -[(b,+ l)*-(b+ 1j2]/2k. The full penetrationfield is found by setting X( b = 0) =O, which gives b,= Ji?k- 1. (34) 1. Ocb,cb,, By integrationof Eq. (8) one gets for the virgin magnetization POM -=-bo+F( Bo 1+cY lf ; b, ) b,2 (35) By expandingin the quantity ke -ba we find (ke-ba)3+*+* J. Appl. Phys., Vol. 77, No. 8, 15 April 1995 (33) , (30) One sees directly that the initial slope of m equals- 1. T. H. Johansen and H. Bratsberg vs B, 3949 2. b,ab6, The Eq. (8) now gives 0.4 1+CY POM I-a -=--b,-l+~(,b0+1)3Bo 3k ib,+ 1J3 0.2 2ff -IskZ~tb,+1)5--ib,+1)sl, im a? where the midplaneinduction is b,= &T,+ 1)2-2k- 0.0 2 1. (3jj I" _ -0.2 B. Decreasing field 1. baa0 As in the treatmentof the exponentialmodel we can for this region make use of the substitutionk~-- k, and from the resultsof Sec. IV A 2, we immediatelyget 1+cY POM ~=-h,-l-~(b~flj3+3k(b,+lj~ Bo -04 I -3 -2 -1 I 4 IO I 1 2 I I 3 1-a 2a “-y-y-p [(b,+- ljSU(b,f 1F1, (38) FIG. 5. Magnetization curves for the Kim and co-workers model. As in Fig. 3, the branches are calculated for LY=1, 05, and 0 (infinite slab). Here the virgin curves merge with the major loops at the point, b, = i/i?% -1=\/5I. where b,= \I(&+ 1)2+2k- 1. (39) 2. -bps b,<O Following the procedureof Sec. III B 2 we find that this time the midplaneinduction dependson 6, accordingto b,= J2+2k--(b,- 1)2- 1. The magnetizationbecomes I-a POM -zz 6k Bo (3+2b,)b;-!- 1fCY - 6k 60kS cy (8b;+25bm+20)b; (3-2b,)bz- & valid for b,&b, . Expandingthis time in the quantity l/(1 + b,) we tind a high degreeof term cancellation,and obtain 5-3ff 20 (8bz-25b,+20)b; k3 (l+bJ5 +-.. ’ i43) -b,. (41) The remainingpart of the magnetizationcurve againfollows from the condition M( - b,) = - M(b,). For both models we have verified analytically that the expressionscombine to give a continuousascendingand descendingbranch in the magnetizationcurve. C. Width of hysteresis loop The magnetizationfor ascendingand descendingapplied field is shownin Fig. 5 for cu=O,0.5 and 1, againusing k=l. The correspondinghysteresiswidth AM is plotted in Fig. 6(a) as a function of applied field. Also for the Kim and co-workersmodel it is possibleto produceessentiallylinear graphsabove the full penetrationfield; this is now obtained by plotting l/AM as shown in Fig. 6(b). To derive the analog of the expansionEq. (31) we subtract Eqs. (36) and (38) using the respectiveexpressionsfor 6,. The result is 3950 J. Appl. Phys., Vol. 77, No. 8, 15 April 1995 It follows that AM can be written as AM--(~-T)wJ,(B,)~ l+$~(l~~,)~+...). (44) This result can be comparedwith the previously published Eq. (4). One seesthat both the first and secondterm in the expansionshave the samefield dependence,and we identify 3 5-3cY s=syq-. For the infinite slab case (cx=O)we get s-1/4 in agreement with Chen and Goldfarb.” For cylinder geometry (a=l), however, the above relation gives s=3/20, three times the value given in Ref. 4. Thus, the benefit of using cylindrical samplesin a Bean model type of analysisis not as dramatic as previously thought. Like in the exponentialmodel, we find that even the lowest-ordercorrectionis very small. At b,= 6, it equals T. H. Johansen and H. Bratsberg PIG. 7. Two flux density distribution consistent with a given applied field. BI(X) is the induction profile when B, is reached from above (descending field), while Bt(X) is the profile when coming from the lower side (ascending field). 0.0 1 I 1 0 I 1 I I 2 I AM&)=( 1-+,(B,)( 3 d2 XdB” J;(&)+*** ba (a) l+;it,Ew2,; a 1 (45) > valid for B,>B,. We now show that this unified form in fact holds for any critical-state modelsfor the presentgeometry.Let BT(X) and B I (X) denotethe internal induction profile in the ascending and descendingfield situationsillustrated in Fig. 7. The differencein magnetizationbetweenthe two statesequals h 4 m” ,u~AM= ‘[LQ(X) --BT(X)](2cuX+ 1 - a)dX, J0 (461 where we here have used the alternative magnetizationformula, Eq. (IO). Expanding the profiles in a Taylor series about X= 1 gives I I II 0 1 1 I 2 I I 3 PIG. 6. (a) Magnitude of the hysteresis in magnetization for the Kim and co-workers model. The graphs represent the vertical separation of the curves in Fig. 5. (b) The field dependence of I/A&f. Plotted this way the high-field asymptotes appear as straight lines near b,=b, also for this model. Like in Fig. 4(b) the convergence is strong. 3 5-3a 20 3-cu +$?;‘(l)(x-l)3+... , (47) bs (4 k2 (1-t2k)Z’ which amounts from l/36 (a=O) to l/60 (a=l) for k=l. Again the rapid convergenceof the expansionjustifies the use of a simple graphical test of applicability, similar to the one suggestedfor the exponential model. This time l/AM should be plotted against the applied field, and again if the high-field behavior appearslinear, the model parametersj, and B, are determinedby the fit of a straight line. V. DISCUSSION By inspectionwe seethat in both Eqs. (31) and (44) the expansionfor AA4 can be written in terms of the critical-state function J,(B) in the following way: J. Appl. Phys., Vol. 77, No. 8, 15 April 1995 and a similar seriesfor BL(X). From Ampere’s law one has B;(l) = powJ,(l) and B;(l) = -,~~wJ,(l), where in this context J,(l) meansJ,(X= 1). It is readily seenthat the secondderivatives of BT and BL becomeequal, also in sign, and the terms cancel in Eq. (46). The third-order terms again have oppositesign. Thus, one obtains B~(X)-B~(X)=-2~~w~,(.l)(x-1)-~~w~~(1> X(X- 1)3/3f*** . (48) Equation (46) then integratesto give A,=( 1-;)w,(l,( l+$-,~~+.+ (49) This expansiongeneralizesa similar result obtainedby Fietz and Webb.” It relates the magnetizationhysteresiswidth to the local current distribution near the surfaceof the sample. The secondterm in this expansiondependson the curvature of the distribution profile. Note now that Ampere’s law yields T. H. Johansen and H. Bratsberg 3951 JkYw=(Pow)2J, d -& Applied to the point where X=1, B=B, tains one thereforeob- M=(l-a)M,fcU&,, Togetherwith Eq. (49) this provesthe generalvalidity of Eq. (45). Since the Taylor expansionis a good approximation only as long as the induction profile is an analyticalfunction, the above derivation is restricted to B,>B, due to the nonanalyticityin the B dependenceof J, . A nice feature of Eq. (45) is that the sample’sgeometry and magneticproperties,representedby the function J,(B), enter as separatefactors. This allows us to discussthe significanceof shapeand size on a generalbasis. First, we note that the leading term representsthe Bean model result derived in Ref. 4. The secondterm, which obviously vanishesin the Bean model, has an a-dependentcoefficient which monotoneouslyincreasesfrom 1 to 513as LY is varied from 1 to 0. Thus, the geometricalaspectratio of the sampledeterminesonly weakly the contribution of the secondterm. Nevertheless,the deviation from the leading order behavior will always be minimum for a sample of squareor circular cross section. For the two models treated in the previous sections,the graphs in Figs. 4(b) and 6(b) illustrate this point. It is clear that the relative weight of the secondterm in Eq. (45) can be altered much more efficiently through the quadraticfactor w2. Physically,this is due to the fact that the secondterm representsdeviationsfrom a linear flux distribution. Such nonlinearitiessoon becomeless pronouncedthe thinner the sample. We saw in the previous sectionsthat in both modelsthe size of the hysteresisloops is largestfor the infinite slab case, a=O. Although this observationis consistentwith Eq. (45), a generalargumentis most easily formulatedfrom Eq. (8). If M, denotesthe magnetizationof an infinite slab of a given thickness,the magnetizationof an equally thick finite slab can be written Ba p&f = poM, + CY X( 1 -X)dB. I Bl?t (52) Along the ascendingfield branch one has B,< B, , and the integralis positive. For the descendingbranchB,> B, , and the integral becomesnegative.Thus, when a is increased from zero both branchesof M are shifted away from M, toward smaller magnitudes,i.e., the loop shrinks monotonically with larger a’. 3952 J. Appl. Phys., Vol. 77, No. 8, 15 April 1995 Another point can be seen directly from Eq. (8). Together with the fact that a=1 includes the cylinder case where diameter equals slab thickness,4it follows that the magnetizationof a rectangularslab can be expressedas (53) where M,, is the magnetizationof the cylinder. Thus, the generalslab has a magnetizationgiven by a weightedsum of the results for two simpler geometricalcases. Returningto the result, Eq. (45), we seethat the relation between.J,(B,) and AM(B,) also can be written as follows: 1 5-3a l-40 3-a J,(B,) = w2& (54) AM@,) %w(l-a/3) 1 I-3a/5 1-z(1-cr13)3 i-4 (55) This describeshow to go beyond the Bean model approximation when extractingJ, from AM datawithout invoking a specific model. The very simple critical-state functions J,(B) that one uses today are, with few exceptions,motivated on a phenomenologicalbasis.For future developments of microscopically derived models, the possibility shown here, to extract on a more generalbasis the function J,(B) from experimentaldata, seemsto be important. ACKNOWLEDGMENTS The authors wish to thank Professor Y. Galperin for stimulating discussions.The financial supportfrom the Research Council of Norway and the Nansen Foundationis gratefully acknowledged. ‘C. P. Bean, Phys. Rev. Lett. 8, 250 (1962). “Y.-B. Kim, C. F. Hempstead, and A. R. Stmad, Phys. Rev. 129, 528 i1963). s W. A. Fietz, M. R Beasley, J. Silcox, and W. W. Webb, Phys. Rev. 136, A335 (1964). 4D.-X. Chen and R. B. Goldfarb, J. Appl. Phys. 66, 2489 (1989). ‘M. Xu, D. Shi, and R. Fox, Phys. Rev. B 42, 10773 (1990). 6H. Ishii, T. Hara, S. Hiiano, A. M. Figueredo, and M. J. Cima, Physica C 225, 91 (1994). 7T. H. Johansen and H. Bratsberg, J. Appl. Phys. 76, 8001 (1994). 8P Chaddah, K. V. Bhagwat, and G. Ravikumar, Physica C 159, 570 (1989). 9D.-X. Chen, A. Sanchez, and J. S. Muiioz, J. Appl. Phys. 67,343O (1990). “W. A. Fietz and W. W. Webb, Phys. Rev. 178, 657 (1969). T. H. Johansen and H. Bratsberg