The width of the first wide barrier for real systems can be taken equal to 8 instead of 20, required for perfect system
Algorithm for preparation of multilayer systems with high critical angle of total re ection a) b)
I. Carrona , V. Ignatovichb. Texas A & M University, College Station, TX, USA FLNP JINR, Dubna Moscow region, 141980, Russia
Abstract
The new development of theory of multilayer systems is presented. It shows precisely how to calculate thicknesses and number of layers to get re ectivity close to unity for a given, in principle, arbitrary critical angle. Application of the new approach to real systems is demonstrated.
«£®à¨â¬ ¤«ï ¯à¨£®â®¢«¥¨ï ¬®£®á«®©ëå á¨á⥬ á ¢ë᮪¨¬ ªà¨â¨ç¥áª¨¬ 㣫®¬ ¯®«®£® ®âà ¦¥¨ï . àà®, . £ ⮢¨ç.
®â æ¨ï
§« £ ¥âáï ®¢ë© «¨â¨ç¥áª¨© ¯®¤å®¤ ª «¨§ã ¬®£®á«®©ëå á¨á⥬. ¯®§¢®«ï¥â áâண® à ááç¨â âì ⮫é¨ë ¨ ç¨á«® á«®¥¢ ¥®¡å®¤¨¬ë¥ ¤«ï ¯®«ãç¥¨ï ¡«¨§ª®£® ª ¯®«®¬ã ®âà ¦¥¨ï ¤«ï § ¤ ®£®, ¢ ¯à¨æ¨¯¥ ¯à®¨§¢®«ì® ¢ë᮪®£®, ªà¨â¨ç¥áª®£® 㣫 . ¥¬®áâà¨àã¥âáï ¯à¨«®¦¥¨¥ ®¢®£® ¯®¤å®¤ ª ॠ«ìë¬ á¨á⥬ ¬.
1 Introduction We apply a not yet widely known physically transparent analytical method to calculation of thicknesses and number of layers in multilayer systems (MS) to achieve a high critical angle. Usually MS consist of many bilayers of two materials with dierent refraction indices, and the thickness ai of the bilayer varies with its number i according to theoretical prescriptions of [1]. In such a stuck all the bilayers have dierent thicknesses, and the change of neighboring layers is very small. We consider here a dierent construction: the MS consists of several periodic chains, and we show how to nd the period, number of periods for every chain, and number of chains to achieve the critical angle we wish. Applications of MS in experiments are discussed in many review papers (see, for instance [2, 3] and references there in), and we do not dwell on it too much. We only want to add some references [4] - [10], which were not mentioned in [3]. In [4, 5, 6, 7] the MS were used for polarization of neutrons by transmission [4] through them, by transportation along magnetized neutron guides [5, 7], and by splitting of unpolarized beam by a magnetized supermirror [6]. In [8] the pulsed beam was produced by re ection from a supermirror periodically magnetized in external eld. In [9] supermirrors were used in neutron guides to increase the transmitted ux. Some research on fabrication of supermirrors was presented in [10]. In our present paper we consider MS for polarization of neutrons. This purpose determines materials for bilayers. However our analytical method is not limited to this purpose but is applicable to all MS, even to those that contain more than two materials.
2 Our method First of all we should mention one dierence of our approach comparing to commonly used one. We consider re ection in terms of normal component k? of the incident neutron wave vector instead of the incidence angle. It is more convenient because re ection of a mirror at a given angle depends also on wave length, whereas in terms of the wave vector k it depends only on k? and properties of the mirror. In the following we even omit the index ?, and use simply k, because we deal only with specular re ection and for that the one dimension is sucient. To be more precise we consider a neutron propagating along x-axis normal to the supermirror, and calculate its re ection from the supermirror, which is a set of alternating layers of two materials. One of them is represented by a potential barrier of height ub and width lb, and another one is represented by a potential well of height uw and width lw . The potential barrier with lb ! 1 totally re ects neutrons with k < ub, and pub is called critical number kc. It is convenient to use 1=kc as a unit of length, then all the variables become dimensionless, the barrier becomes of height ub = 1, and the critical number is also unity. In the following we use a somewhat dierent normalization. We take for unity the dierence ub , uw , and for the unit length 1=pub , uw . We look for such MSs which give total re ection up to some Kc > kc = 1. In principle Kc can be arbitrary large, but practically it is possible achieve Kc not larger than 4. Our analytical method is based on an observation [11, 12] that every potential can be split by an in nitesimal gap into two separate ones, as shown in g. 1, and the re ection 2
1
1
2
x
2
Figure 1: Every potential can be split by an in nitesimal gap of width ! 0, into two ones, and the splitting does not change their re ection and transmission properties, because of total transmission of the gap. of the composite potential R is represented as the combination of re ections Ri and transmissions Ti of the separate barriers: R = R + T 1 ,RR R ; (1) where the denominator corresponds to multiple re ections inside the gap. For simplicity in (1) we did not take into account asymmetry of the potentials, we discuss it when needed. The expression (1) gives immediately the result [13] for a semiin nite periodic potential. If a single period of the potential is characterized by re ection and transmission amplitudes r and t respectively, then re ection amplitude of the whole potential (denoted R in [1]: eq-s (14-16) there) is 12
12
0
2
2 1
1
1
2
q q (1 + r)2 , t2 , (1 , r)2 , t2 q ; R= q
(1 + r) , t + (1 , r) , t 2
2
2
(2)
2
and the Bloch phase factor (denoted by in [1]: eq-s (12), (13) there) is q q (1 + t)2 , r2 , (1 , t)2 , r2 q exp(iqa) = q ;
(1 + t) , r + (1 , t) , r 2
2
2
(3)
2
where a is the period width, and q is the Bloch wave number. At Bragg re ection R = exp(i) and exp(iqa) = exp(,q0a) with real , and q0. (We neglect here imaginary part of the potential.) With the equations (2), (3) we can nd [11] re ection, RN , and transmission, TN , amplitudes of the periodic chain with nite number N of the periods: 1,R iqaN ) ; T = exp(iqaN ) (4) RN = R 1 1,,Rexp(2 N exp(2iqaN ) 1 , R exp(2iqaN ) : To see how do these formulas work we need to de ne the single period and its amplitudes r and t. A single period is a bilayer. It consists of a potential well and barrier. This period 2
2
2
u 6
lb
1
lw uw
a -
lb lw 2
lw 2
a -
Figure 2: A period containing a well and barrier, can be rearranged to symmetrical form. is nonsymmetrical, but we can make it symmetrical by shifting the barrier as shown in g. 2. This rearrangement is not principal, as we see later, but it facilitates our mathematics. For symmetrical period of width a = lw + lb we can immediately nd amplitudes r and t: 1 , exp(2ikblb) ; t = eik l eik l 1 , rwb r = eik l rwb 1 , (5) rwb exp(2ikblb) 1 , rwb exp(2ikblb) ; w w
q
2
w w
2
b b
2
where kw;b = k , uw;b, rwb = (kw , kb )=(kw + kb ) and potentials may contain imaginary part because of losses. Substitution of them into (2) and (4) gives the result shown in g. 3 and 4. In g. 3 we see the Bragg re ection with unit amplitude in the interval called width of the Darwin table. By decreasing lw and lb we can shift the interval toward larger k, and, if we can built a system of semiin nite potentials with dierent periods in such a way that intervals would overlap as intervals D in g. 5, we can considerably increase kc. However we can build periodic chains only with nite number of periods, so we must use jRN j of (4), which at Darwin table is smaller than unity because of exp(,2q0Na) in the nominator. This factor is small when N is large. If we tolerate re ection jRN j = 1 , with some small , we must have N = , 2lnaq0 : (6) 2
Figure 3: Re ection amplitude jR(k)j of a semiin nite periodic potential with period containing the potential well of depth uw = 0:5 and width lw = 1, and the barrier of the height 1 and width lb = 1.
Figure 4: Re ection amplitude jRN j of the periodic potential with N = 8 periods. The parameters of a single period are the same as in g. 3. So the strategy is very clear. We step by step cover the range of k, we needed, by overlapping intervals 0 < , and tuning parameters lw , lb, N nd maximal 0 to minimize the number of required chains, and therefore the total number of layers. To proceed further it is more convenient to transform (2,3) to the form q q q q cos + jrj , cos , jrj Re(r) + jrj2 , Re(r) , jrj2 q q =q ; R= q
cos + jrj + cos , jrj
(7)
Re(r) + jrj + Re(r) , jrj 2
2
q q q q 2 , sin + j t j , , sin , j t j Re( t ) + j t j , Re(t) , jtj2 iqa q q e =q =q ;
(8) Re(t) + jtj + Re(t) , jtj where is the phase and Re(r; t) are real parts of amplitudes r, t respectively. To derive (7,8) we use the relations valid for arbitrary potential [14]:
, sin + jtj + , sin , jtj
2
2
r = eijrj; t = ieijtj; r , t = e i: From (7) it follows that R is a unit complex number exp(i), when jrj > jRe(r)j. 2
2
2
(9)
2
jRj
2
6 D
D
0
1
1
D D D 2
3
4
0
k
0
1k k k k 1
2
3
4
-
k
Figure 5: A system of periodic potentials with overlapping Bragg peaks of widths Di gives total re ection in a range of k considerably wider than the common case 0 k 1.
3 Algorithm of calculations of lb, lw , N and 0. Substitution of (5) into (7) and (8) in the case k > ub gives 2
s
s
kb tan(kw lw =2) , kw cot(kblb=2) , kw tan(kw lw =2) + kb tan(kblb=2) s kb tan(kw lw =2) + kw tan(kb lb =2) R = s kw tan(kw lw ) , kb cot(kblb) kb tan(kw lw=2) , kw cot(kblb=2) + kw tan(kw lw =2) + kb tan(kb lb=2) k tan(k l ) , k cot(k l ) k tan(k l =2) + k tan(k l =2) w
or
w w
b
s
cos R = s cos cos cos
+ + +
+
and
b b
b
w w
s
+ rwb cos , , sin , rwb cos , s sin + rwb cos , + sin , r cos sin
+
+ +
,
wb
+
w
(10)
b b
+ rwb sin , , rwb sin , ; + rwb sin , , rwb sin ,
(11)
q q 2 2 2 2 cos , r cos , , sin2 + + rwb sin2 , + , wb iqa q e =q ;
(12) cos , rwb cos , + , sin + rwb sin , where = (kw lw kb lb)=2. It is easy to check that at the limit lb ! 0 we obtain R ! 0 and q ! kw , and at lw ! 0 we obtain R ! rwb , q ! kb . If k < ub instead of (11, 12) we obtain 2
+
2
2
2
+
2
2
2
q q cos2 , , exp(,2kb0 lb) cos2 + , i sin2 , , exp(,2kb0 lb) sin2 + q ; R=q 2 2
cos , , exp(,2kb0 lb) cos + i sin , , exp(,2kb0 lb) sin 2
2
+
(13)
+
v v u u 0 0 u u sin , exp( , k l ) sin , + b b t t cos , , exp(,kb lb ) cos + , sin , + exp(,kb0 lb) sin + cos , + exp(,kb0 lb) cos + iqa v e =v ; u 0 lb) sin + u 0 lb) cos + u u sin , exp( , k cos , exp( , k , , b b t +t
(14)
sin , + exp(,kb0 lb) sin cos , + exp(,kb0 lb) cos p p where = kw lw =2 , = arccos(kw = ub , uw ) and kb0 = ub , k . It is easy to check that in the limit lw ! 0 the periodic potential degenerates to potential step and we obtain R ! rwb = exp(,2i ), q = ik0. In the limit lb ! 0 barriers disappear, and we obtain empty space with R = 0 and q = kw . Now we consider k > ub. The Bragg re ections take place when expressions under two square roots in (11) have opposite signs. It happens when j cos j < rwbj cos , j, or j sin , j < rwb j sin j, i.e. for n=2 , n=2 + , where n is integer. The half width of the Bragg re ection = rwb j cos , j for odd n, and = rwbj sin , j for even n. To get this width maximal we must have , = m for odd n and , = m =2 for even n (m is also integer). From these considerations we obtain, that if we want to have the total re ection at some k = kv we must require at this point kb lb + kw lw = and kw lw , kb lb = 0, and immediately nd two parameters lb = =2kb, and lw = =2kw (as was correctly used in [1], eq. (7)). We can also require kb lb + kw lw = 2 and kw lw , kb lb = , and nd two other parameters lb = =2kb, and lw = 3=2kw . +
0
+
2
0
0
2
+
+
+
-
l
0
lb
1
lw lw=2 lw=2 Figure 6: The MS should contain also a barrier of large width l to provide total re ection almost up to k = 1 0
We cannot use the full width of the Darwin table, because the total re ection inside it is possible only for in nite number of periods. We need to nd such 0 and N which will give the maximal eective length = 0=N covered by a single period. For optimization we represent (12) in the form 1
v
u u (, )rwb , cos ( ) 1 , Q , qa e = 1 + Q (,2Q); where Q = t cos sin ( ) , sin (,)rwb : 2
2
0
2
2
+
(15)
+
2
2
In the case of small rwb we can expand near point kv where = n=2 and approximate (15) as q Q = rwb 1 , (x=x ) ; (16) where x = (k , kv )=kv and x 2rwb kw kb =nkv (kw + kb ). We see, that x , and therefore the width of the Darwin table is largest for n = 1. Thus it is the best to require = =2. +
+
0
0
2
2
2
2
2
2
0
+
Figure 7: Dependence of re ection coecient jRj of FeCo-Si MS on k. MS consists of a wide barrier of width 20 and 34 chains with dierent number of periods. Total number of bilayers is 1947. Critical kc for FeCo is equal to 1. Potentials do not include imaginary parts. 2
Now we need to nd the ends kve of the Darwin table around kv . They depend on what deviation from total re ection we tolerate. If we tolerate jRj = 1 , 2 , then 2
Q = , ln( )=4N:
(17)
Figure 8: Dependence of re ection coecient jRj of FeCo-Si MS on k. Parameters of MS are the same as for g. 7. Potentials include imaginary parts 2
From it we nd
!2
x 2
ln : = 1 , (18) x 4Nrwb To nd the optimal number N of periods we require the maximal eective width of k covered by a single period, i.e. we seek a maximum of (kve , kv )=N . This requirement gives (19) N = 2lnr ; 0
and
wb
p
p
p
(20) k = jkve , kv j = kv x 23 = k 3(rkwbk+w kkb ) 23 kv rwb; v w b where kw , kb and rwb are determined for k = kv . If we tolerate 2 = 1%, then N = 2:6=rwb . If we use another condition , = =2 we nd that x is approximately 2 times lower. So to use this condition is not pro table. p Now, the interval 0 = 2k = ( 3=)kv rwb around kv is closed, and we can make the step to a new kv0 = kv + 0 and nd a new periodic chain around the point kv0 . In practice we made steps 0 = 2k=1:1 to ensure the overlapping of the intervals, because every next width k is a little bit lower than the preceding one. In g. 7 the re ection coecient jRj is shown for MS consisting of 1947 bilayers with positions of the Bragg peaks chosen as prescribed above. The starting point was kv = 1:12. It was found empirically. We see that re ection coecient is perfectly equal to one. Above we did not take into account imaginary part of the potential, however formulas (2 | 5) and (10 | 12)| are valid for arbitrary potentials, so to take into account losses or gains (in the case of active media) we need only substitute in kw and kb complex potentials uw;b = u0w;b , iu00w;b, where minus sign means losses for u00 > 0. Of course, the number and widths of layers in periodic chains and the widths of the Bragg peaks are real numbers so for them we must use absolute magnitudes. The result of calculations for the real systems FeCo-TiZr, which is similar for FeCo-Si, is shown in g. 8. Here the number of bilayers is the same as in g. 7, and we see that 0
2
2
2
0
2
2
re ection coecient deviates from unity. It means that our requirement (17) with small is not necessary, because the makes us be tolerable to stronger deviation of the Bragg re ection from unity. So we can strongly decrease the number of periods in every chain. In g. 9 we show how re ection coecient presented in g. 7 changes, when number of bilayers is decreased to 271. We see that now it becomes alike to the one shown in g. 8. If we account for imaginary parts of potentials then for the real system FeCo-TiZr with 271 bilayers we obtain the re ection coecient shown in g. 10 In the case we are satis ed with smaller increase of the critical angle, we need even smaller number of layers. In g. 11 we show re ection coecient for FeCo-TiZr with only 46 bilayers. The parameters of these bilayers are shown in tabl. 1.
4 Re ection from the set of chains If we have two chains with re ection and transmission amplitudes RNi, TNi (i = 1; 2), then re ection R from two chains from the left (the chain 1 is to the right of the chain 2) is R = RN + 1 ,TNR RNR : (21) 21
2
21
2
1
2
N2 N1
Addition of the third chain to the left side gives the set with re ection amplitude (22) R = RN + 1 ,TNR R R : N Four chains will have re ection amplitude R and so on. It is a simple algorithm to calculate re ection from all the chains, and at the end we must add a single wide barrier as shown in g. 6, which provides total re ection for all k almost up to k = 1. Because of nite width l of the rst barrier, its re ectivity drops near k = 1. Indeed, the re ection from the barrier is q 0 p 1 , Q k w , ikb r = rwb 1 , r Q ; where rwb = k + ik0 ; kw = k , uw ; kb0 = 1 , k ; w wb b and Q = exp(,2l kb0 ). 2
321
3
3
21
3
21
4321
0
0
2
0
2
2
0
0
para- 1 2 3 4 5 6 7 8 9 meters kv 1.12 1.32 1.44 1.55 1.65 1.73 1.81 1.90 1.96 lb 3.11 1.84 1.5 1.331 1.20 1.11 1.04 0.98 0.93 lw 1.40 1.20 1.09 1.01 0.95 0.91 0.87 0.83 0.80 N 3 3 3 3 3 4 4 4 4
10
11
12
2.03 2.09 2.15 0.89 0.86 0.82 0.78 0.75 0.73 5 5 5
Table 1: Parameters of 12 periodic chains with re ection coecient shown in g. 8. kv is position of the Bragg peak, lw , lb and N are the widths of Si and FeCo layers and number of bilayers respectively for the chain with the Bragg re ection centered at kv . Total number of bilayers is 324. The unit of length is equal to 86.2 A.
Figure 9: Dependence of re ection coecient jRj of the same system as in g. 7 with number of periods in every periodic chain strongly decreased. Total number of bilayers is 271. 2
Suppose, we tolerate, when jrj = 1 , . Near the critical point k = 1 re ection coecient can be approximated as jrj = (1 , Q(1) ,+Q16)k k Q 1 , 4 kl ; w b 0
2
0
2
2
2
2
2
2 0
0
so, if we want pto have jrj to be everywhere in 0 < k < 1 larger than 1 , , we must choose l = 2= . In particular, for = 0:01 we must choose l = 20. In all the g-s. 7 | (11 we used this width of the rst barrier. However, if we take into account losses, this parameter is not too much critical. To show that we demonstrate g. 12, which is calculated 12 chains of FeCo-TiZr MS with parameters, shown in table 1, and for l = 8. Though re ection of smaller rst barrier can be a little bit less the losses in it are also less, so the result is nearly the same. So practically we have no gain, if increase the totally re ecting layer too much. 2
0
0
0
5 Asymmetry of the period Above we considered the case when periods of periodical chains are symmetrical, i.e. barrier of width lb is surrounded on both sides with wells of width lw =2, i.e. it is represented as a threelayer. In practice it is more simple to consider the period as a bilayer consisting of the well of width lw and the barrier of the width lb. Such a period is not symmetrical. Its re ection from the left rl is not equal to re ection from the right, rr , though transmissions from both sides is equal and is given by formula (5). The amplitudes rl and rr for the bilayer are 1 , exp(2ikblb) = eik l r; r = r 1 , exp(2ikblb) = e,ik l r; rl = e ik l rwb 1 , r wb rwb exp(2ikblb) 1 , rwb exp(2ikblb) (23) where r is re ection of the symmetrical period shown in (5). With nonsymmetrical period expression (2) should be modi ed. For instance, re ection from the semiin nite periodic 2
w w
2
w w
2
w w
Figure 10: Dependence of re ection coecient jRj of FeCo-TiZr MS on k. Parameters of MS are the same as for g. 8, but the number of bilayers is only 271. 2
potential beginning with the well we have Rl = q q s q(1 + pr r ) , t , q(1 , pr r ) , t (1 + r ) , t , (1 , r) , t r l r l rl q ik l q q q = e ; p p rr (1 + rr rl) , t + (1 , rr rl) , t (1 + r) , t + (1 , r) , t (24) or it is exp(ikw lw )R, where R is symmetrical amplitude given by (2). The re ection from semiin nite periodic potential beginning with the barrier will be exp(,ikw lw )R, i.e. asymmetry of r is inherited by R. Equation (3) for the Bloch phase factor does not change, because instead of r it contains rlrr , which is identical to r . Now it is easy to understand that re ection of a nite number periods RN for asymmetrical period will change in the same way as R, i.e. for re ection from the left and right we have RNl;r = exp(ikw lw )RN , where RN is for the symmetrical period. Now we need to see what happens when we stack two nonsymmetrical chains. For that we need to generalize expression (1) for nonsymmetrical potentials 1 and 2 shown in g. 1. This generalized expression is Rl = Rl + T 1 , RRl R ; (25) r l where indices l; r denote re ection from the left and right respectively. Taking into account this generalization we represent (21) in the form # " T exp( ik [ l , l ]) R w w w N (26) Rl = eik l 2 RN + 1 ,NR exp(ik [l , l ])R ; N w w w N where asymmetry is explicitly represented by the factor exp(ikw [lw , lw ]). It is easy to prove that, if the chain 1 at some k gives total re ection, i.e. RN = exp(i), then inclusion of chain 2 will not destroy this total re ection, i.e. Rl for these k is also a unit complex number: Rl = exp(i0). Indeed, taking into account relations (9), which are valid for RN and TN , we can transform (26) as follows ,ikw [lw , lw ] , i , i) ; (27) Rl = ,eik l 2 l 1 i i2 11,,jRjRN j jexp( N exp(ikw [lw , lw ] + i + i) 2
2
2
2
2
2
2
2
2
2
2
2
w w
2
2
2
2
2
2
12
2
w w
21
2 1
1
2
2
1
2
1
2
2
2
1
1
2
1
1
2
1
21
21
21
w [ w + w ]+
+2
2
1
2
1
2
2
2
2
Figure 11: Re ection of 46 bilayers of FeCo-TiZr system with account of losses. Parameters of layers are shown in table 1. where is the phase of the amplitude RN . Since the last factor is of the form exp(i ) the whole Rl is also of the form exp(i0) which corresponds to the total re ection. Of course, all these relations are precise only for real potentials. Imaginary part of the potentials gives a correction to them, and the smaller is imaginary part the smaller is the correction. 2
2
21
6 Similarity of all the MS systems All the MS can be represented as a system with barriers of height 1, and wells of height 0. Indeed, if in a real system barriers have the potential ub, and wells | the potential uw , then the potential step between well and barrier is ub , uw, and we can q normalize this dierence to unity, and take as a unit length the critical wave length = h = 2m(ub , uw . So, calculations for all the real systems is the same. The only dierence is that at the end we need to include re ection amplitude from the potential step from vacuum to the well. This potential step is now has normalized potential u~w = uw =(ub , uw). If re ection amplitude from MS without this correction is R, then after correction it will be r w )R ; where r = k , kw ; k = qk , u~ : r w + (11 , w w + r wR k + kw w We applied our method to real physical system, and considered only 34 chains, though it is not principal. With these chains we are able to triple the critical angle. To double it it is sucient to have only 12 chains. Their parameters are presented in Table 1. The rst row shows the points kv which are centers of the Bragg peaks. The rst number kv = 1:12 was chosen empirically. Next two rows show the width of the wells lw and barriers lb for those kv , And the last line shows the number of periods in every chain. We do not present here the numbers of periods that give the perfect re ection for real potentials and requires 324 bilayers. We show here the numbers of periods that give re ection g. 11 and 12 with 46 bilayers, and which is nearly the same as re ection for real system with 324 bilayers as can be seen from comparison with g. 8. 0
2 0
0
0
2
Figure 12: Re ection of 46 bilayers of FeCo-TiZr system with account of losses. Parameters of layers are shown in table 1. The width of the rst barrier is 8. Imaginary parts for potentials of real systems were normalized to dierence of real parts of ub ,uw. Thus for FeCo-Si system in which Si are wells with uw = 54:4,i6:2510, neV and FeCo are barriers with ub = 330:7 , i6:40 10, neV, the normalised potentials are ub = 1 , i2 10, and uw = 0 , i2:3 10, . In FeCo-TiZr system the normalized imaginary part of FeCo is 3 10, , and that of TiZr 1 10, . The main eect of losses comes from imaginary part of FeCo, so the results of calculations for real systems with Si and TiZr give nearly the same result. The width of the rst wide barrier for real systems can be taken equal to 8 instead of 20, required for perfect system with no losses. Accounting for losses shows that the increase of the rst barrier only increases losses in the whole range of energies and decreases the total re ection. 4
2
4
6
4
4
7 Conclusion We presented the method of calculation of a supermirror, having high critical angle of total re ection. We suppose that our method has some advantage, because it is analytical, and therefore more controllable. Change of parameters lb and lw from chain to chain is suciently large and therefore is less prone to errors related to technology of layers preparation. There is only few change of parameters comparing to common way, when the parameters change almost continuously, and lb, lw become lower than a monolayer. Such a small change of width is almost impossible to control. We want also to add that though our analytical method is very good for analysis, it is too slow for calculations. So, the calculation of re ection coecient, after all the parameters were de ned, was performed numerically with the matrix method. We have shown here how to prepare MS by increasing the range of total re ection step by step. However it is possible to proceed dierently. We can put one bilayer on a substrate and calculate its re ection. Then put another bilayer with parameters scanned in some intervals, choose parameters, which give the larger increase of the re ectivity. Then look for parameters of third bilayer and so on. If we do not restrict thickness of layers, we can get with 200 bilayers a good re ectivity as shown in g. 13 for some
Figure 13: Re ection of 399 layers for a model system with uw = ,1 model system with uw = ,1 even for interval k = 4kc . However in these bilayers some thicknesses are of the order 0:1 of interatomic distance. It is clear that it is impossible to achieve a good homogeneity for such thicknesses. We can restrict thicknesses to some values when scanning in the parameter space. It may give a multilayer system with smaller number of layers and with not perfect, but well tolerable re ectivity in a wide enough interval of k. Though this try and error method may give a tolerable re ection with smaller number of periods, the step by step method is promising for improving the technology for preparation of MSs. When we know exact thickness of a single monolayer we can prepare layers with better surfaces, and we can control the perfectness of the layers by comparing calculated and really obtained re ectivities in a wide range of energies. We considered here only re ection of neutrons from MSs. There is no problem to apply our method also to x-rays. However to do that we should nd a better way to optimize number of layers when we account for imaginary parts of the potentials. In the case of neutrons the number of layers can be easily found somewhat empirically. In the case of x-rays we can do the same, however because the imaginary parts for x-rays are considerably higher the analytical analysis must be more reliable. 1
References [1] [2] [3] [4] 1
J. B. Hayter and H. A. Mook. Appl. Cryst. (1989), 22, 35-41. O. Sharpf. Physica B (1991), 174, 514-527. P. Boni, D. Clemmens, M. S. Kumar, C. Papas. Physica B (1999), 267-268, 320-327. A. Ioe, P. Fischer, T. Krist, F. Mezei. J.Phys.Soc. Jap. (1996), 65, Suppl. A, 80-81.
If we normalize to unity the sum
ub
+
p , then increase in g. 13 is not 4, but 17 2 = 2 9.
juw j
=
:
[5] A. Schebetov a.o. J.Phys.Soc. Jap. (1996), 65, Suppl. A, 195-198. [6] S. Itoh, M. Takeda. J.Phys.Soc. Jap. (1996), 65, Suppl. A, 199-202. [7] M. Takeda, K. Kurahashi, , Y. Endoh, S. Itoh. J.Phys.Soc. Jap. (1996), 65, Suppl. A, 234-237. [8] T. Kawai a.o. J.Phys.Soc. Jap. (1996), 65, Suppl. A, 230-233. [9] K. Al Usta, K. N'guy, A. Menelle, F. Samuel. J.Phys.Soc. Jap. (1996), 65, Suppl. A, 249-251. [10] S. Basu a.o. J.Phys.Soc. Jap. (1996), 65, Suppl. A, 289-291. [11] Ignatovich V.K. Amer. J.Phys., 1989.- v.57, N.10, p.873-878 [12] F.Radu, V.K.Ignatovich. Physica B 267 (1999) 175-180 . [13] Ignatovich V.K. 1991.- v.175, N.1-3., Physica B., p.33- 38. [14] V.K.Ignatovich. Physics of Ultracold Neutrons (UCN). Oxford, Clarendon Press. 1990. [15] V.G.Syromyatnikov, A.F.Schebetov, Z.N.Soroko. Nucl. Instr.& Meth. in Phys. Res. A324 (1993) 401-403.