Professional Documents
Culture Documents
Schwinger Et Al. - 1988 - Is Spin Coherence Like Humpty-Dumpty
Schwinger Et Al. - 1988 - Is Spin Coherence Like Humpty-Dumpty
Schwinger Et Al. - 1988 - Is Spin Coherence Like Humpty-Dumpty
© Springer-Verlag 1988
p2
by
H=~m---/~-B(r).
t
The equations of motion for the spatial variables of p (t): p + j"d t' 7 VB (r (t')) •rff (t') a t/(t')
0
the atom are
and
d r(t) = 1 p(t), t
dt m r(t) : r + t p + j" dt'(t_t') 7_ VB(r(t')).nt(t')at/(t ')
m o m
d
d~ p(t) = V [/~(t). B(r(t))] = y VB(r (t))-2S(t).
=r+--p(t)+ J dt'(-t') VB(r(t')).t/*(t') an(t').
Although we shall soon restrict attention to spin m o
it is annihilated by all the t/(t'). The result is the reduc- SG and SGI: leading approximation
tion of the differential equation to
SG
d (tp)
i~tl(t))=-Ta.B r+ t/(t)), The SG experiment is designed to measure the z-com-
ponent of an atomic magnetic moment through the
spatial deflection produced by a z-inhoraogeneity in
with the adjoint form the magnetic field. Thus the dominant components
of field and field gradient are B~ and ~ B~. In this
--t ~ (t/*(t)= -- (t/*(t) a . B t
circumstance the ~ equation of motion simplifies to
along with T
p~(T)--p~=2S~Ap~, Ap~= ~ dte(t),
o
or Az_TAp = T t --
B(r+-~P)~-B(t)+[(r--(r))+t
(p-(P))]'VB(t). is
Here we see the natural appearance of effectively time-
varying fields, in contrast with their introduction as z(T)-z--Tp~(T)=2S~ Az.
a dynamical approximation in Humpty-Dumpty I.
The q differential equation has now become (the pres- In the simple situation where F(t) is the constant F
ence of ) is understood) for all 0 < t < T,
1T 1T
i ~ tl(t)= {-- ~'a"B(t)-- , [(r-- (r ) ) /Jz ........
2m FT-- 2m Ap~= --Az.
+ t (p--(p))] "VB(t)'a} tl(t). SG criteria. It is well to recall here
the conditions
for a good SG experiment: that the splitting of the
138
3z Accordingly, we have
V,-~--
6pjm"
t/* (T) = r/* U(-- a~),
right hand side has the unique value of - t , whereas On recalling the relation
% on the left hand side takes the value + 1. All re-
maining dependence on spin appears in the factor Az=T--Ap~+jz,
m
<r/* ½(ax + iay) q> = <(S~(O) + iS,(O))>,
we recognize that the alternative form, referring to
referring only to the spin aspect of the initial state. the final state of the free particle, is
The outcome,
whereas the corresponding momentum wave func- and this means that the magnitude of C, which gives
tion, ~o (P~), and the second version, produce [p = p~] the length of the vector <2S(T)>, is the measure of
spin coherence 3
C= ; dptpo(p--Ap)*e21Va=¢o(p+Ap). In a good SG experiment,
-oo
SG:Az>>6z, Apz>>6p=
The wave functions ¢0 are those of a free particle,
at time zero. In the absence of the magnetic field - the combination of non-overlap of the wave func-
for a free particle - the z-momentum wave function, tions:
at time T, is just [p = p~] Or(z- A z)* Or(z + A z)~-O,
_ip~r with the rapid oscillation of the exponential factor:
Or(p)=e 2m ~/'o(P),
so that 6zApz~A~pp=>>1,
¢o(p--Ap)* @o(p+Ap) assures that
ApT
=@r(p- Ap)* e2ip m Or(p+ Ap). SG: IC[~0;
2 The symbol z in the wave function now stands for the eigenvalue 3 It is consistent that, as a diagonal matrix element of a unitary
of the operator z operator, C can never exceed unity in magnitude
140
spin coherence in the x - y plane is totally destroyed which provide the conditions
in a good SG measurement 4.
1
SGI, C~-I:IJzI~IAz[~ ~ ,
SGI criteria. What about the SGI ? Ideal maintenance opt
of spin coherence, C = 1, is possible only if 1
1 1 !(~ ~
Marginal detection. Perhaps one would be satisfied
Oo(Z)=(2rc)-4(rz) 2e 4 ~ # . with just enough coherence to be detectable? That
circumstance might be described by
The immediate outcome, I 1
SGI, C>~0:IAzI~@z, IAp~l~Tz,
[ 1 [[A z \ 2 [A pz\2]~
C=exp~-2tt~-z) +I~P~)l)' 1
IAz[ lapel'-' <1.
6zrp~
shows that a nearly perfect maintenance of spin co-
herence requires that Yet, a state close to that of minimum uncertainty
would still be required if the control involved is to
SGI, C~-I:IJzI~rz, [Ap~l~rpz, be on the microscopic level.
I,~zl IAp~l,~.6z6p=. Interference pattern. Incidentally, it is all very well
to speak of an interferometer, but where is the inter-
An alternative statement, for A z, combines
ference pattern? That comes into view if we recall
T that
Az=Ap~--+ J z
m
(2S~(T)) = C cos ~,
with T/m~rz/rpz to get
and recognize the possibility of a controlled shift in
IAzl¢rz, the Larmor phase angle • that can be produced by
[Azt tAPzI ¢rzrp~. having a stretch of homogeneous magnetic field be-
tween the two major stages of the apparatus - those
The implication, of splitting and recombination.
SGI, C ~- 1: [Az[ [A pd,~ 1, SG Larmor angle. It wilt be useful to have some idea
of the order of magnitude of
presents a sub-microscopic criterion for the near main-
tenance of spin coherence. It won't be easy. T
One can avoid reference to specific wave functions, oh=2 ~ dtyBz(t)
under the circumstance of nearly perfect coherence, o
by returning to the unitary operator form of C and
expanding the exponential. With the assumption of that is characteristic of a single SG magnet, for which
no initial correlation between z and Pz, the dominant
terms are IAp~lso = fr dt??-az Bz(t) > 6p~.
o
C -~ 1 - 2 [(3 z)2 (A p~)2 + (3 pz)2 (Az)2],
5 An attempted realization with the aid of slits m u s t approach the
4 Heisenberg recognized this long ago from a consideration equiva- diffraction limit, which, in view of the quite short wave length of
lent to the oscillation of the exponential factor the atom, requires rather extreme dimensions
141
The time-independent field and field inhomogeneity of which there are really only five, because of the
within the magnet define a length, field equation V. B = 0.
The field component By vanishes in the interior
of a magnet; it is characteristic of the fringing fields
encountered in entering and leaving a magnet. We
shall assume that the magnets can be so constructed
one that is assuredly macroscopic: that By vanishes everywhere along a free atom's aver-
age trajectory:
l~>>,Sz.
Then, on writing B (t) = 0, B,(t) O,
1
IA p~lsc ~-~ I~1 ,> 6p~, which are not independent statements, for
8 18 1
we learn that By= v= <p,>.
I~l>>Iz6p~>>6z,Sp~. So we are left with
Consequently,
I i,>1
must be quite a large number, involving as it does all of which are related to the z-components.
the product of two ratios between macroscopic and The t/equation now reads:
microscopic quantities.
d
i-d-~tl(t)
The next level
={-az[E(t)+(z +tpz) F(t)
It is time for a second look at the q equation of mo-
tion:
+((Y-<y))+t (Py-<Py)))l d E(t)]
i ~--~tl(t)={-~a'B(t)-7[(r-(r))
--axl--(x +t px) F(t)]
+t (p_ (p))]. VB(t).a}q(t).
-a,[(z +tp~) l d e(t)]} tl(t),
Coordinatesandfield components. Recall that the time which has used the properties ( x ) = O, (p~) = O.
dependences of B and VB are associated with the av-
erage motion of the free atom along a direction that
will be identified with the y-axis, the line with x = 0, ConservedS~
z = 0. The structure of the magnetic field is taken to
be invariant under reflection in the y-z
plane, as First, we set aside the ax and o-y terms, so that only
produced by x ~ - x . An immediate implication is a~ appears in the ~/equation and Sz continues to be
the vanishing of the following x = 0 quantities: conserved. The solution of the t/ equation that pre-
sents q(T) in linear relation to tt now contains two
B~,~x Br, ~x Bz, ~y B~,~z B* unitary operators. One, referring to z and Pz, is as
before. The other, apart from a numerical phase fac-
tor, is
which are actually only three in number, because of
the field equation g x B = 0. The remaining field quan-
tifies of interest are
8 0 8 0 0 py - <py))~,
Br, B~,~x Bx, -~y Br,~z B~,~fy B~=~z By,
142
which uses the significance of times zero and T to SG1 Larmor angle. This brings us to the question:
get What is the magnitude of q~ for a SGI, having in
mind that it is a large number for a single SG appara-
r d tus? It is in the nature of a SGI that individual SG
dt ~ E(t)= E( T)- E(O)=O, magnets with opposite field inhomogeneities are re-
0
quired. But this leaves open the choice of the orienta-
whereas tion of the magnetic field. In general, [~lsoi=>t~IsG,
except when the SG magnets are used in pairs, with
r r 1 oppositely oriented fields, to achieve
i dtt~--~E(t)= -- ~ dtE(t)= ---~
0 0
I~ISGI~ 1,
As a result, the expression for (S~(T)+iSy(T))
now has e -ie replaced by where one must remember that q) includes the vari-
able effect of the homogeneous field used to exhibit
the interference pattern.
If I~lsGi>>1, the degree of momentum control re-
quired ranges from absurd to incredibly difficult,
=/exp[-i~(l PY-(P')'I]\,~y).]j/ whereas I~lsG~ 1 demands no such unreasonable ef-
fort.
It is interesting to look again at the expectation
where one can recognize the first terms in the expan- value involving py, this time made explicit in terms
sion of of the initial y-wave function:
@ x p [ - i ~ (PY)]~py
J/ ([exP{2~bPY--(PY);]2\(p,)
JJ /
• : ~ . ; ~ 1. ~v,=2~Bz(t) 2S.
Py
143
The result is an accumulated displacement of y, rela- and A*(t) will differ only in the sign of i. One should
tive to that of a free particle 6, which depends on the notice that A(t) oscillates very rapidly in time, with
spin state: a period ~ 1/E, which represents the spill precession
in the x - y plane.
T 1 T
Ay= ~ dtbv,=--2 ~ dtyB~(t) S~ An iterative solution of the 0 equation, that begins
o Py o with t~(0)= i, must be carried to the second order,
because
(x)=0, (px)=0.
It is just these displacements that occur in the y-wave Then the two time integrals, ~(1/E) 2, combine with
functions, where, indeed, the displacement in the right the square of 6xF to give a loss of coherence, both
(left) hand wave function is that for S~ = - ½ ( + ½). for S~ and Sx, Sy, that is of the order
2
Non-conservedS~
If Sz is not conserved, spin coherence is inevitably ~E/ \ B~ / \tz]"
lost to some degree. It remains to find the order of
magnitude of the loss. For simplicity, let us include As the squared ratio of microscopic and macroscopic
just o-~ terms in the q differential equation, and omit, lengths, this is small, but it is not zero. A similar
as well, the effect of y-motion just discussed. Then treatment of the o-y terms associated with fringing
we are confronted with fields gives a measure of coherence loss
d rl(t)=[_azE(t)_az(Z+tmP~)F(t) ~(~@)2,