Skip to main content

Full text of "Step wise destruction of the pair correlations in micro clusters by a magnetic field"

See other formats


Step wise destruction of the pair correlations in micro clusters by a magnetic field 



N.K. Kuzmenko 
V.G. Khlopin Radium Institute, 194021, St. -Petersburg, Russia 
V.M. Mikhajlov 

Institute of Physics St .-Petersburg State University 198904, Russia 
and S. Frauendorf 

Department of Physics, University of Notre Dame, Notre Dame, IN 46556, USA and IKH, Research Center 

Rossendorf, Germany 



The response of nm-size spherical superconducting clusters 
to a magnetic field is studied for the canonical ensemble of 
electrons in a single degenerate shell. For temperatures close 
to zero, the discreteness of the electronic states causes a step 
like destruction of the pair correlations with increasing field 
strength, which shows up as peaks in the susceptibility and 
heat capacity. At higher temperatures the transition becomes 
smoothed out and extends to field strengths where the pair 
correlations are destroyed at zero temperature. 

The electron pair correlations in small systems where 
the single-particle spectrum is discrete and the mean level 
spacing is comparable with the pairing gap have recently 
been studied by means of electron transport through nm- 
scale Al clusters The pair correlations were found 
to sustain an external magnetic field of several Tesla, in 
contrast to much weaker critical field of H c = 99 Gauss 
of bulk Al. A step wise destruction of the pair correla- 
tions was suggested It is caused by the subsequent 
excitation of quasi particle levels, which gain energy due 
to the interaction of their spin with the external field. 
This mechanism is very different from the transition to 
the normal state caused by a magnetic field applied to a 
macroscopic superconductor. Hence it is expected that 
physical quantities, as the susceptibility \ an( i the spe- 
cific heat capacity C, which indicate the transition, be- 
have very differently in the micro cluster. The present 
letter addresses this question. 

For the micro clusters the energy to remove an elec- 
tron is much larger than the temperature T. The fixed 
number of the electrons on the cluster was demonstrated 
by the tunneling experiments Jl[. Hence, one must study 
the transition from the paired to the unpaired state in the 
frame of the canonical ensemble. The small number of 
particles taking part in superconductivity causes consid- 
erable fluctuations of the order parameter, which mod- 
ify the transition . Consequences of particle number 
conservation for the pair correlations in micro clusters 
have also been discussed recently in j4|-|^], where a more 
complete list of references to earlier work can be found. 

In order to elucidate the qualitative features we con- 
sider the highly idealized model of pair correlations be- 
tween electrons in a degenerate level, which permits to 
calculate the canonical partition function. A perfect Al- 
sphere of radius R — (l-5)»m confines N » 2-10 3 — 3-10 5 
free electrons. Its electron levels have good angular mo- 
mentum I. Taking the electron spin into account, each of 



these levels has a degeneracy of 2M, where M = 21 + 1. 
For a spherical oscillator potential, the average angular 
momentum at the Fermi surface is I k, iV 1 / 3 — 1.4, where 
N is the number of free electrons. The distance between 
these levels is Ae ~ 10 meV is much larger than the BCS 
gap parameter A, which is less than ImeV. Therefore, 
it is sufficient to consider the pair correlations within the 
last incompletely filled level. This single-shell model also 
applies to a hemisphere, because its spectrum consists 
of the spherical levels with odd I, and to clusters with a 
superconducting layer covering an insulating sphere (cf. 
H) or hemisphere. 
The single-shell model Hamiltonian 

H = Hpair — uj(L z + 2S Z ), (1) 

Hpair = —GA^ A, A^ = S ' Q^ajr", 

fc>0 

consists of the pairing interaction H pa i r , which acts be- 
tween the electrons in the last shell with the effective 
strength G, and the interaction with the magnetic field. 
We introduced the Larmour frequency Tiuj = hbB, the 
Bohr magneton /i^, the z-components of the total or- 
bital angular momentum and spin L z and S z . The label 
k = {A, a} denotes the z -projections of orbital momen- 
tum and spin of the electrons, respectively, and A + cre- 
ates an electron pair on states (k, k), related by the time 
reversal. 

The magnetic susceptibility and heat capacity of the 
electrons are 

Ml d 2 F (?» d 2 F{T,u) 
X h 2 V do; 2 ' ST 2 ' [ ' 

The free energy F derived from the Hamiltonian (|l|) gives 
only the paramagnetic part \p of the susceptibility, be- 
cause we left out the term quadratic in B. For the fields 
we are interested in (magnetic length is small as com- 
pared to the cluster size), the latter can be treated in 
first order perturbation theory, generating the diamag- 
netic part of the susceptibility 

= _ m/4<|2+j/2> ^ _ 4 i() _ 6 

where m is the electron mass is and V the volume of 
the cluster. It is nearly temperature and field indepen- 
dent [|| . The numerical estimate for Al assumes constant 



1 



electron density. For nm scale clusters xd ~ 10~ 3 . It is 
much smaller than xd ~ — 1 for macroscopic supercon- 
ductors, which show the Meissner effect. Since the mag- 
netic field penetrates the cluster it can sustain a very high 
field of B ~ Tesla. On the other hand, the xd is three 
orders of magnitude larger than the Landau diamagnetic 
susceptibility observed in normal bulk metals. 

The exact solutions to the pairing problem of parti- 
cles in a degenerate shell were found in nuclear physics 
fiof in terms of representations of the group SU2- The 
eigenvalues E v of H pair are 

E v = - — (N sh - v) (2M + 2 - N sh - v) . (4) 

where N s h is the number of particles in the shell. The 
seniority, which is the number of unpaired particles, is 
constrained by < v < N s h and v < M. The degenerate 
states {v, i} of given seniority v differ by their magnetic 
moments fJ,BTn u ,i, where i = {LASY*} takes all values 
of the total orbital (L) and total spin (S) momenta and 
their total z-projections (A, S) that are compatible with 
the Pauli principle for v electrons. In presence of a mag- 
netic field the states have the energy 

U„,i(u)) =E V — um Vt i, m Vj i = (A + 2£)„ )i , (5) 

and the canonical partition function becomes 

Z = ^exp(-0U v ,i) 
= exp(-/3^)[$„ - $„_ 2 (1 - *„.„)], (6) 
(3 = l/T, $„ = J2 exp(-pum v ,i). (7) 

i 

To evaluate the sums we take into account symmetry of 
the wave functions of the v unpaired electrons and reduce 
the sums ([?]) to products of sums over orbital projections 
of completely antisymmetric states (one column Young 
diagram, cf. with v/2 + £ and v/2 — X electrons. 

®u= Yl 2(l + 5 s . )- 1 |. iy/2+s $ iy/2 _ ff cosh(2/3 w E), (8) 

s=s mi „ 

s min - [i - (-n/4 

^ = 4, + (l-4.o)II sinh 4 ■ (9) 

fj,— i t 

The derivation of (JTJ - ^ will be published separately Q . 
The pair correlation energy is 

= -A»(T,o;)/a (10) 

Here we have defined the parameter A c , which measures 
the amount of pair correlations. Applying the mean field 



approximation and the grand canonical ensemble to our 
model, the thus introduced A c becomes the familiar BCS 
gap parameter A. Accordingly we also refer to A c as the 
"canonical gap". However, A c must be clearly distin- 
guished from A because it incorporates the correlations 
caused by the fluctuations of the order parameter A. For 
the case of a half filled shell and even N s h, the BCS gap 
is A(0) ee A(T = 0,w = 0) = GM/2. Ref. § found 
A(0) = 0.3-0.4 meV for Al-clusters with R = 5-10 nm, 
which sets the energy scale. 




0.3 0.4 0.5 0.6 
w/A(0) 

FIG. 1. Canonical gap A C (T, uj) (full lines) and BCS gap 
A(T,w) ( dotted lines) v.s. the Larmour frequency lu. 

Let us first consider the destruction of the pair cor- 
relations at T = 0. The lowest state of each seniority 
multiplct has the maximal magnetic moment 

m u = ?f{v{2M -v)-\[l- (-)"] + 4(1 - *„.„)}. 

(11) 

According to (Q) , @ and (|TT| ) the state of lowest energy 
changes from v to v + 2 at 



U v +2 — 



2 A(0) 
M 



M-v 
M - v - 1 



(1 - <S,.o) 



(12) 



At each such step m v increases according to (|1 1|) . The 
pair correlations are reduced because two electron states 
are blocked. At the last step leading to the maximum 
seniority v max all electron states are blocked. Hence the 
field B c corresponding to ui c = U! Vmax can be regarded as 
the critical one, which destroys the pairing completely at 
T = 0. For a half filled shell w c = 3A(0) /M for even 
electron number and 4A(0)/M for odd (v max = M — 1 
and M, respectively). 

Fig. |l| illustrates the step wise destruction of pairing 
by blocking for the half filled shell M = 11 (I = 5). This 
mechanism was discussed in jy, where the crossing of 
states with different seniority could be observed. It is a 
well established effect in nuclear physics, where the states 
of maximum angular momentum, are observed as "High- 
K isomers" |l3| . Fig. [j] also shows results for the mean 



2 



field (BCS) approximation (cf. p|,p^|) to the single shell 
model. The pair correlations are more rapidly destroyed. 
The quantum fluctuations of the order parameter stabi- 
lize the pairing. 



which is smaller than the coherence length, 





1.2r 




1.0- 




0.8- 




0.6- 


< 


< 


0.4- 




0.2- 













M=11 N.=10 

sh 










X B/B C =1 : 


B/B^=1J^^__ 




i . i 





0.0 



0.5 



1.5 



2.0 



1.0 

FIG. 2. Canonical gap A c (T,w) (full lines) and BCS gap 
A(T, uj) ( dotted lines) v.s. the temperature T 

We introduce T c as the temperature at which the mean 
field pair gap A(T C , uj — 0) takes the value zero when the 
magnetic field is absent. For the half filled shell T c = 
A(0)/2. Fig. [| shows the mean field gap A(T,w), It 
behaves as expected from macroscopic superconductors: 
The frequency where A = shifts towards smaller values 
with increasing T. Fig. |^ shows that the temperature 
where A = shifts from T c to lower values for uj > 0. 

However, fig. [I] also demonstrates that the canonical 
gap A c behaves differently. For T = 0.8T C there is a re- 
gion above uj c where there are still pair correlations. For 
T = 2T C this region extends to 2lo c . The pair correlations 
fall off very gradually with u>. Fig. || shows how these 
"temperature induced" pair correlations manifest them- 
selves with increasing T. For u> = there is a pronounced 
drop of of A c around T c , which signalizes the break down 
of the static pair field. Above this temperature there is 
a long tail of dynamic pairing. For uj > uj c the dynamic 
pair correlations only built up with increasing T. 

The temperature induced pairing can be understood in 
the following way: At T — 0, all electrons are unpaired 
when the state of maximum seniority becomes the ground 
state for w > uj cr it- At T > excited states with lower 
seniority enter the canonical ensemble, reintroducing the 
pair correlations. 

Here we have adopted the terminology of nuclear 
physics, calling "static" the mean field (BCS) part of the 
pair correlations and "dynamic" the quantal and statis- 
tical fluctuations of the mean field (or equivalently of 
the order parameter). The "pair vibrations" , which are 
oscillations of the pair field around zero 14 , are well 
established in nuclei. Fluctuation induced superconduc- 
tivity was discussed before |15|. The fluctuations play 
a particularly important role in the systems the size of 



1.2 
1.0 

0.8 

O 

<f 0.6 
<j" 

0.4 
0.2 



0.0 I— 
0.5 



0.6 




T7T.-0.1 
M=7 N sh =6 
T/T =0.02 



T/T c =0.2 
M=23, N sh =22 

T/T =0.1 



0.8 

co/co 



FIG. 3. Canonical gap A C (T, uj) v.s. uj. 



-1.0 

o. 





M=7, N sh =6 




T/T =0.02 


T/T =0.1 

M 


/ 





0.7 0.8 0.9 1.0 1.1 



T/T =0.2 [ 

\ 4 


\ M=23, N sh =22 
T/T=0.1 





0.5 0.6 0.7 0.8 0.9 1.0 1.1 

co/o 

FIG. 4. Susceptibility x(T, uj) v.s. uj. 

Fig. H shows a very small cluster (M = 7) with very 
pronounced steps. Already at T = 0.1T C the steps are 
noticeably washed out. In the single-shell model the step 
length is lo v — cj„_ 2 ~ A(0)/M. Accordingly, no indi- 
vidual steps are recognizable at T = 0.1T C for the large 
cluster (M = 23) shown. Yet there is some irregularity 
around uj — 0.7cj c , which is a residue of the discreteness 
of the electronic states. It is thermally averaged out for 
T = 0.2T C . Hence only for M < 50 i. e. N < 2 • 10 4 and 
T < 0.1T C the step wise change of A c is observable. 

The discreteness of the electronic levels has dramatic 
consequences for the the susceptibility at low tempera- 



3 



tures. As shown in the upper panel of fig. |I| \p nas pro- 
nounced peaks at the frequencies where the states with 
higher seniority and magnetic moment take over. The 
paramagnetic contribution is very sensitive to the tem- 
perature and to the fluctuations of the order parameter. 
Using the BCS mean field approximation we find much 
more narrow peaks, which are one to two orders of mag- 
nitude higher. For the larger cluster in the lower panel 
the individual steps are no longer resolved, resulting in 
a peak of \p near oj = 0.7lo c . Since for the considered 
temperatures it is unlikely to excite states with finite 
magnetic moment, \p is small at low to . It grows with 
uj because these states come down. It falls off at large 
u> when approaching the maximum magnetic moment of 
the electrons in the shell. The curve T = 0.1T C shows 
still a double peak structure, which is residue of the dis- 
creteness of the electron levels. 



the spin orbit coupling is small as in Al §. Most of 
the findings of the present paper are expected to hold 
qualitatively for these spin flips. 

In summary, at a temperature T < O.IT c an increasing 
external magnetic field causes the magnetic moment of 
small spherical superconducting clusters (R < 5nm) to 
grow in a step like manner. Each step reduces the pair 
correlations until they are destroyed. The steps mani- 
fest themselves as peaks in the magnetic susceptibility 
and the heat capacity. The steps are washed out at 
T > 0.2T C . For T ~ T c , reduced but substantial pah- 
correlations persist to a higher field strength than for 
T = 0. This phenomenon of the temperature-induced 
pairing in a strong magnetic field is only found for the 
canonical ensemble. 

Supported by the grant INTAS-93-151-EXT. 



2 



o 



M=23, N sll =22 
T/T =0.1 




FIG. 5. Heat capacity C(T, 



LO V.S. LO. 



The heat capacity is displayed in fig. Is has a dou- 
ble peak structure for the M = 7 cluster at T = 0.02T C . 
In this case C is very small because the spacing between 
the states of different v is much larger than T. Near a 
crossing the spacing becomes small and C goes up. The 
dip appears because at the crossing frequency the two 
states are degenerate. Then they do not contribute to 
C because their relative probability does not depend on 
T. For T = O.IT c the probability to excite states with 
different seniority has increased and C takes substantial 
values between the crossings. The dips due to the degen- 
eracy at the crossings remain. For the M — 23 cluster 
C shows only two wiggles, which are the residue of the 
discreteness of the electronic levels. 

The deviation of real clusters from sphericity will at- 
tenuate the orbital part of xp an d round the steps of 
A c already at T = 0. The back-bending phenomenon 
observed in deformed rotating nuclei ||l3fl is an exam- 
ple. How strongly the orbital angular momentum is sup- 
pressed needs to be addressed by a more sophisticated 
model than the present one. In any case, there will be 
steps caused by the reorientation of the electron spin, if 



[1 
[2 
[3 

[4] 
[5] 

[6] 
[7] 

[8 

[9 

[10 

[11 

[12 

[13 
[14 

[15] 



D.C.Ralf, C.T.Black, and M.Tinkham, Phys. Rev. Lett. 
78,4087(1997) and ealier works cited therein. 
B. Miihlschlegel, D. J. Scalapino and R. Denton, Phys. 
Rev. B 6, 1767 (1972) 

N. K. Kuzmenko, V. M. Mik hajlov, S. Frauend orf, J. 
Cluster Sci., 10, 195 (1999); |cond-mat 980701 1| v2 10 
Aug 1998. 

B. Janko, A.Smith, and VAmbegaokar, Phys. Rev. 
B50,1152(1994). 

F.Braun, J. von Delft, and M.Tinkham, Phys. Rev. Lett. 
79,921(1997). 

D. S. Golubev, A. D. Zaikin, Phys. Lett. A 195, 380 
(1994) 

A. Mastellone, G. Falci, and R. Fazio, Phys. Rev. Lett. 
80, 4542 (1998) 

F. Braun and J. v. Delft, Phys. Rev. Lett. 81, 4712 (1998) 
R. Neuendorf, M. Quinten, and U. Kreibig, J. Chem. 
Phys. 104, 6348 (1996) 

C. Esebag, J.L.Edigo, Nucl. Phys. A552,205(1993) and 
earlier work cited therein. 

M.Hammermesh, Group theory and its applicationM to 
physical problem, Addison- Wesley (1964) 
N. K. Kuzmenko, V. M. Mikhajlov, S. Frauendorf, Phys. 
Rev. B, in preparation 

Z. Szymanski, Fast Nuclear Rotation, Claredon, (1983) 
Bohr, A., and Mottelson, B., Nuclear Structure II, W.A. 
Benjamin, (1975) 

W. J. Skocpol and M. Tinkham, Rep. Prog. Phys. 38 
1049, (1975) 158,319(1967). 



4