Beruflich Dokumente
Kultur Dokumente
for a near-
eld optical system such as a near-eld optical microscope
and spectroscope be
H =
H
0
V; (1)
H
0
=
H
A
H
B
H
bath
; (2)
where the Hamiltonians H
A
and H
B
describe the states of the
probe tip and sample, respectively, as isolated quantum-
mechanical systems. The Hamiltonian H
bath
represents the
states of light and innite matter (substrate or bre probe),
while V
and H
0
are
assumed in the following discussion, and the eigenvalues
and eigenstates of H
are written as E
l
and [W
l
), respectively,
i.e.
H[W
l
) = (
H
0
V)[W
l
) = E
l
[W
l
): (3)
Dening the static projection operators P and Q=1P in
the usual manner as
[W
(1)
l
) = P[W
l
); [W
(2)
l
) = Q[W
l
); (4)
P
2
= P; PQ = QP = 0; (5)
we divide the eigenstates [W
l
) into two groups, [W
( 1)
l
) in P-
space and [W
( 2)
l
) in Q-space. It is possible to formally express
Q[W
( 2)
l
) by P[W
( 1)
l
) as
Q[W
(2)
l
) = J(E
l
H
0
)
1
Q
VP[W
(1)
l
); (6)
J =
_
1 (E
l
H
0
)
1
Q
1
; (7)
and the eigenstates [W
l
) for the total Hamiltonian can then
be expressed in terms of the eigenstates in P-space as
[W
l
) = (P Q)[W
l
) = JP[W
(1)
l
): (8)
Since both of the states satisfy the normalization condition,
we can rewrite this as
[W
l
) = JP(PJ
JP)
1=2
[W
(1)
l
): (9)
We thus obtain an effective operator O
eff
in the P-space,
which is equivalent to an arbitrary operator O
in the full
space, corresponding to physical observables (Kobayashi &
Kohmura, 1984):
W
m
[
O[W
l
) = W
(1)
m
[
O
eff
[W
(1)
l
); (10)
O
eff
= (PJ
JP)
1=2
(PJ
OJP) (PJ
JP)
1=2
: (11)
From the above Eqs (1011), it follows that relevant
quantities such as the probe tipsample interaction and
transition dipole moments can be exactly expressed by using
a small number of bases in P-space after renormalizing the
effects from Q-space. This is one of the advantages of the
method, and is desirable from the computational viewpoint.
3. Effective sampleprobe interaction and virtual photon
model
On the basis of the above formalism, we now discuss the
virtual photon model (Hori, 1993; Kobayashi & Ohtsu,
1998; Ohtsu & Hori, 1998) a unique and intuitive
model in particular, its empirical assumption regarding
250 K. KOBAYASHI AND M. OHTSU
q 1999 The Royal Microscopical Society, Journal of Microscopy, 194, 249254
the probesample interaction in a fundamental process. The
model assumes virtual photons with nonzero mass (m
eff
)
tunnelling between nanometric objects, similar to electron
tunnelling in scanning tunnelling microscopy (STM). This
means that the Yukawa-type interaction is responsible for
the interaction between a nanometric probe tip and sample.
In other words, the Yukawa potential (e
m
eff
t
/t) is assumed
to be generated near the nanometric object (source) as a
result of the interaction between incident light and the
object, for both of the illumination and collection modes in
near-eld optical microscopy and spectroscopy. Owing to the
assumption, it can describe optical near elds localized on
the material surface without solving the Maxwell equation,
and a novel scheme for manipulating atoms by using the
optical near eld was successfully proposed and analysed
(Hori, 1993; Ohtsu et al., 1993).
Let us now adopt more specic Hamiltonians as
components of H
0
in order to derive an explicit form of
the effective sampleprobe interaction. First we choose a
Hamiltonian for H
a
(a=A, B),
H
a
=
n
E
n
(a) e
na
c
na;
(12)
E
n
(a) =
"
2
k
2
a
2m
c
=
"
2
2m
e
np
a
_ _
2
or
"
2
2m
e
np
b
_ _
2
; (13)
where the creation and annihilation operators are denoted
as c
na
and c
na
, and the relevant energy levels are designated
as E
n
(a), and electric dipole transitions between those levels
are allowed. Note that this Hamiltonian corresponds to one
in which electrons with effective mass m
e
and wave number
k
a
are conned in an innite well potential of size a(b). The
ground and excited states will be designated as [A), [A*),
[B), [B*), for later use.
Next, let us describe H
bath
in terms of excitons (B
, B
),
photons (a
, a
~
j
D
~ (r ~
j
) in a multipolar
QED Hamiltonian (Craig & Thirunamachandran, 1984),
where m
~and D
~(r ~
j
) are the dipole operator of matter and the
displacement eld operator at site j, respectively. As is well
known (Hopeld, 1958; Knoester & Mukamel, 1989;
Juzeliunas & Andrews, 1994), this Hamiltonian can be
diagonalized in terms of exciton-polariton modes (energy
"q
k~n
, creation/annihilation operator y
k~v
, y
k~n
as
H
bath
=
3
i=1
"Q
ji
B
ji
~
k;
~
G
2
l=1
"q
~
k
~
C
a
~
k
~
G;l
a
~
k
~
G;l
~ m
j
~
D(~r
j
) (14)
=
~
k;n
"q
~
kn
~
kn
y
~
kn
: (15)
Here the indices i and l label the three Cartesian
axes and the transverse polarization of light, respectively.
The k~ summation extends over the rst Brillouin zone,
and G~ runs over the reciprocal lattice. The summation of u
runs over all branches of exciton-polaritons. We will use the
ground state of H
bath
, [0) as one component of the P-space.
It follows from Eq. (15) that photons and innite matter
may be a collection of harmonic oscillators. In this
representation, a near-eld optical probe tip interacts with
a nanometric sample via exciton-polariton mediation as
q 1999 The Royal Microscopical Society, Journal of Microscopy, 194, 249254
Fig. 1. Schematic illustration of how to compose a total Hamiltonian
H
A
, H
B
,
and H
bath
. (a) The states of the probe tip, the sample, and the
bath system, as each isolated system, are described by the Hamilto-
nian H
A
, H
B
, and H
bath
, respectively. Depending on the sample, H
B
may contain the centre of mass motion as well as internal degrees
of freedom. Atom deection system by the optical near eld
described by the total Hamiltonian H
as an example. Incident
and deected atoms have the momenta "k~and "k~
/
, respectively.
QUANTUM THEORETI CAL APPROACH TO NEAR- FI ELD OPTI CAL SYSTEM 251
follows:
V =
B
a=A
~ m
a
~
D(~r
a
)
=
B
a=A
~
kn
i"
V
0
_
2p
"
_
(K
/
~
kn
(a)
y
~
kn
K
/+
~
kn
(a)
~
kn
)(
a
b
a
);
(16)
where K
/
k~n
(a) and K
/
*
k~n
(a) are the coupling coefcients
between the probe (sample) and exciton-polaritons, and b
a
and b
a
are the creation and annihilation operators leading
to electric dipole transitions of the sampleprobe system,
relating to Eq. (12). It is worth reminding ourselves that
there is no direct interaction between sample and probe tip
in this Hamiltonian.
As explained in the previous section, one can exactly
express the relevant interaction by using a small number of
bases in the P-space, after renormalizing the effects from the
Q-space. We then choose a combination of the ve states,
[A), [A*), [B), [B*) and [0), as the P-space bases illustrated in
Fig. 2, and set [W~
(1)
l
) = [A) [B*) [0), [W
(1)
m
) = [A*) [B) [0), in
order to evaluate the effective sampleprobe interaction V
eff
(AB) from Eqs (10) and (11):
V
eff
(AB) = W
(1)
m
[
V
eff
[W
(1)
l
_
= W
(1)
m
[(PJ
JP)
1=2
(PJ
VJP) (PJ
JP)
1=2
[W
(1)
l
_ _
(17)
Noting that the operator J satises the equation
[ J;
H
0
]P = (E
H
0
)JP J(E
H
0
)P
=
VJP JP
VJP; (18)
let us expand J as J =PS
n=1
g
(n)
J
(n)
, and obtain pertur-
bative solutions of J. We can then determine V
eff
(AB)
perturbatively in the order of V
because PV
P=0,
the nonzero contributions to V
eff
(AB) come from the second
order of V
as
V
eff
(AB) = A
+
[B[0[
_
P
VQ(E
0
p
E
0
Q
)
1
VP
P
V(E
0
p
E
0
Q
)
1
Q
VP
[A)[B
+
)[0) (19)
=
4p
(2p)
3
n
_
d
3
k
K
/
~
kn
(A)K
/+
~
kn
(B)
q
~
kn
Q
0
(B)
K
/
~
kn
(B)K
/+
~
kn
(A)
q
~
kn
Q
0
(A)
_ _
:
(20)
Here the summation of k~ is replaced by integrating k~, and
H
0
in the denominators with respect to the corresponding
eigenvalues. The excitation energies, "Q
0
(A) and "Q
0
(B),
are
"Q
0
(A) =
3"
2
2m
e
p
a
_ _
2
; "Q
0
(B) =
3"
2
2m
e
p
b
_ _
2
; (21)
from Eq. (13). We can similarly show how W
( 1)
l
[V
eff
[W
(1)
m
)
contribute to the effective sampleprobe interaction
V
eff
(BA).
On the basis of the dispersion relation of an exciton-
polariton (effective mass: m
p
, one of the solutions of the
dispersion relation: q
k
) obtained from Eq. (15), we make the
following assumptions, which would be appropriate for
actual near-eld experiments:
(i) q
~
kn
= q
k
"
2m
p
k
2
,
"
2m
p
k
2
; (q
k
< Q
0
);
(ii) K
/
~
kn
,const exp (i
~
k~r):
Dening the ratio g =(3m
p
/m
e
)
1/2
for exciton-polaritons
and electrons in the probe tip (sample), we apply these
assumptions to Eq. (20), and obtain the explicit function
form of
V
eff
(AB) V
eff
(BA) ~
exp[(igpt)=b]
t
exp[(gpt)=a]
t
exp[(igpt)=a]
t
exp[(gpt)=b]
t
;
t = [~t
A
~t
B
[: (22)
This gives the effective sampleprobe interaction derived
from microscopic consideration, and the second and fourth
terms are the Yukawa function, which was empirically
assumed in the virtual photon model. The formula also
shows that optical near eld is effectively localized near the
sampleprobe system, and that the decay length is
proportional to the inverse of the probe (sample) size and
consistent with the experimental results (Saiki et al., 1996).
In addition, we can infer the so-called resonance condition
that a similar size gives maximum coupling between sample
and probe tip, as can be determined by a simple estimation
based on the static dipole model.
Fig. 2. P- and Q-space spanned by bases each of which corresponds
to an eigenstate of H
A
, H
B
, and H
bath
. The P-space is depicted by
hatching.
252 K. KOBAYASHI AND M. OHTSU
q 1999 The Royal Microscopical Society, Journal of Microscopy, 194, 249254
4. Practical implications of the virtual photon model and
application of our formulation
Since the empirical assumption of the virtual photon model
was justied on the basis of microscopic theory in the
previous section, we can take advantage of the character-
istics of the model to develop an application of our
formulation. We have, so far, emphasized the quantum-
mechanical aspects of the optical near-eld systems. The
model also predicts the fundamental performance of near-
eld optical microscopy much faster than the conventional
approach in computation time, that is, the intensity
distribution in the near-eld region of a subwavelength
metallic circular aperture irradiated by the propagating
light. The resolution of the microscope was also clearly
explained by the model. In addition, we can easily handle
more realistic probesample systems, and discuss the
dependence of detected signals on the probe shape, or on
the taper angle v (see Fig. 3). Let the probe be a set of
dielectric spheres of different diameters that approximates a
conical shape. From Eq. (22), the Yukawa potential is
generated between two arbitrary points on the surfaces of
the probe sphere p
i
and sample sphere s
j
. Thus pickup signal
I can be written as
I ~
probe
i
sample
j
_
dj
dr
1
_
pi
dt
2
V
eff
(t
12
)
2
; t
12
= [~t
1
~t
2
[:
(23)
If the so-called resonance condition is satised, the pickup
signal becomes proportional to exp[const/sin(v/2)].
Turning to the quantum-mechanical system again, we
discuss the possibility of applying the formulation to atom
deection and manipulation with a sharpened bre probe.
These techniques are unique and will be essential for
carrying an atom to a desired point on a substrate with high
spatial accuracy far beyond the diffraction limit. As shown
in Fig. 1, suppose that an incident atom with momentum
"k
~
is deected into momentum "k
~
/
by an optical near eld
V
eff
. Then the differential cross-section or angular distribu-
tion of the deected atom is given by
dj
dQ
~
M
2p"
2
~
k
/
_
V
eff
~
k
2
(24)
where M is the mass of the atom. This kind of evaluation
will be required in the design of atom guidance and
manipulation systems using the optical near eld, as well as
in the discussion of basic experimental results.
One implication related to cavity QED problems is that the
transition dipole moment of an atom or a molecule m~
0
is
changed by the inuence of the probe tip and optical near
eld, that is, the difference between the isolated and coupled
states. Perturbative evaluation of Eqs (10) and (11) with
O
=m