Content uploaded by Christoph Lienau
Author content
All content in this area was uploaded by Christoph Lienau
Content may be subject to copyright.
Coherent Exciton–Surface-Plasmon-Polariton Interaction
in Hybrid Metal-Semiconductor Nanostructures
P. Vasa,
1,2
R. Pomraenke,
1
S. Schwieger,
2
Yu. I. Mazur,
3
Vas. Kunets,
3
P. Srinivasan,
4
E. Johnson,
4
J. E. Kihm,
5
D. S. Kim,
5
E. Runge,
2
G. Salamo,
3
and C. Lienau
1,
*
1
Institut fu
¨r Physik, Carl von Ossietzky Universita
¨t, D-26111 Oldenburg, Germany
2
Technische Universita
¨t Ilmenau, Fachgebiet Theoretische Physik I, Postfach 100565, D-98684 Ilmenau, Germany
3
Department of Physics, University of Arkansas, Fayetteville, Arkansas 72701, USA
4
The Center for Optoelectronics and Optical Communications, University of North Carolina at Charlotte,
Charlotte, North Carolina 28223, USA
5
School of Physics, Seoul National University, Seoul 151-742, Korea
(Received 18 December 2007; revised manuscript received 1 August 2008; published 8 September 2008)
We report measurements of a coherent coupling between surface plasmon polaritons (SPP) and
quantum well excitons in a hybrid metal-semiconductor nanostructure. The hybrid structure is designed
to optimize the radiative exciton-SPP interaction which is probed by low-temperature, angle-resolved, far-
field reflectivity spectroscopy. As a result of the coupling, a significant shift of 7 meV and an increase in
broadening by 4 meV of the quantum well exciton resonance are observed. The experiments are
corroborated by a phenomenological coupled-oscillator model predicting coupling strengths as large as
50 meV in structures with optimized detunings between the coupled exciton and SPP resonances. Such a
strong interaction can, e.g., be used to enhance the luminescence yield of semiconductor quantum
structures or to amplify SPP waves.
DOI: 10.1103/PhysRevLett.101.116801 PACS numbers: 73.20.Mf, 71.35.Gg, 78.67.De
The radiative coupling between an emitter and its elec-
tromagnetic environment gives rise to fundamental
quantum-optical phenomena like vacuum Rabi or normal
mode splittings [1–3] and the Purcell effect [4]. Modifica-
tions of the spontaneous emission due to radiative coupling
have been demonstrated for various quantum systems rang-
ing from single atoms [1] to semiconductor quantum dots
[2] and quantum wells (QW) [3,5] coupled to a micro-
cavity. Many groups have investigated similar interactions
in close proximity of metallic structures [6,7]. Various ex-
perimental [8–10] and theoretical [6,11] studies have
shown an enhancement in the luminescence yield as well
as a reduction of the excitonic lifetime of a QW when
placed close to a metallic nanostructure due to its coupling
to surface plasmon polaritons (SPPs). Other reports, how-
ever, show a suppression of luminescence in the proximity
of a metal surface [12,13]. The discussion of these results
has partly remained controversial because present experi-
ments give limited insight into the microscopic nature of
the coupling. Hence the interaction between a metallic
nanostructure and a quantum emitter is not fully under-
stood. One of the most direct ways to probe this interaction
is to study the normal mode splitting between QW excitons
and SPP.
In this Letter, we investigate the linear optical properties
of a hybrid nanostructure consisting of a GaAs QW placed
in the vicinity of a metallic nanoslit array [14,15]. We
report the first direct observation of the normal mode
splitting between QW excitons and SPPs in the nanoslit
array. As a result of this coupling, low-temperature angle-
resolved far-field reflectivity spectra reveal a significant
shift in the exciton resonance position together with an
increase in the radiative exciton damping.
We investigate a multilayer, metal-semiconductor hy-
brid structure consisting of a gold nanoslit grating depos-
ited on a GaAs=AlGaAs QW heterostructure. The sample
consists of a 10 nm wide GaAs QW layer embedded in
Al0:3Ga0:7As barriers grown by molecular beam epitaxy on
a GaAs substrate. On the top side, the barrier has a height
of only 20 nm and is capped by a 3 nm GaAs buffer to
ensure that the QW is close to the surface while maintain-
ing sufficient optical quality. A h¼80 nm thick gold film
is then deposited on the top surface of the semiconductor.
Then an array of d¼140 nm wide slits with a grating
period of a0¼500 nm is created by electron beam lithog-
raphy. A schematic of the sample cross section and a
scanning electron microscope image of the top surface
are shown in Figs. 1(a) and 1(b), respectively. Such a
nanoslit grating is chosen (i) because these arrays provide
a particularly efficient coupling between SPP and far-field
radiation and (ii) because its linear optical properties are
now reasonably well understood [14,15].
In our experiments such a nanoslit array is illuminated
with p-polarized light, with its electric field vector perpen-
dicular to the slit axis. In this case, SPPs are excited at both
air-metal (AM) and semiconductor-metal (SM) interfaces
by transferring momentum n2
a0,n2Z, to the incident
photons. Angle-dependent reflectivity spectra of such
nanoslit arrays reveal the excitation of different AM and
SM resonances as illustrated in the simulations shown in
Fig. 1(c). The hybrid structure parameters are chosen
similar to our sample and the spectra have been calculated
PRL 101, 116801 (2008) PHYSICAL REVIEW LETTERS week ending
12 SEPTEMBER 2008
0031-9007=08=101(11)=116801(4) 116801-1 Ó2008 The American Physical Society
using a full vectorial diffraction model [15,16] by assum-
ing a spatially homogeneous refractive index ns¼3:66 of
the semiconductor layers. The spectra indicate that at
excitation energies of 1:5eV, i.e., around the exciton
resonance of a 10 nm GaAs QW and at an incidence angle
¼40, three different SPP modes, AM½1,SM½þ2,
and SM½3, are almost in resonance leading to an effi-
cient coupling among these modes via the light transmitted
through the nanoslits.
In Fig. 1(d), the spatial distribution of the electric field
component Experpendicular to the slit axis is shown for
illumination by a p-polarized monochromatic plane wave
with Eex ¼1:517 eV, incident at ¼38. This is close to
the crossing of the SM½3,SM½þ2, and AM½1SPP
resonances. The calculations indeed reveal the formation
of coupled SPP modes with spatially localized field inten-
sities at both the AM and SM interfaces. Inside the semi-
conductor, the SPP fields are characterized by a short
modulation period along xof a0=j2nj100 nm due to
the dominant contribution from the n¼þ2and 3dif-
fraction orders. The SPP field is mostly evanescent with a
decay length of only 50 nm, due to the large nsof the
semiconductor. When placed within this decay length, the
transition dipole moment of the QW exciton couples to this
strong evanescent field.
Both the SPP field at the SM interface and the electric
field emitted by the QW are coupled via the nanoslits to
evanescent SPP fields at the AM interface. Therefore they
are rescattered into propagating radiation [14,15] and can
be detected in the far field in a reflection geometry. It is,
however, not necessarily needed for observing exciton-SPP
couplings. The parameters of our sample are chosen so that
the heavy hole (HH) and light hole (LH) exciton reso-
nances of the QW at a temperature of T¼10 K occurring
at 1.546 and 1.554 eV, respectively, match well with the
coupled AM-SM resonance. A strong exciton-SPP interac-
tion is expected only in a narrow range of incidence angles,
when the almost dispersionless QW exciton and the highly
angle-dependent SPP are brought into resonance.
To study this interaction, the sample is illuminated in the
energy range between 1.503 and 1.569 eV with a weakly
focused p-polarized beam from a tunable continuous wave
Ti:sapphire laser. Angle-resolved, linear reflectivity spec-
tra are recorded for ¼26–44at T¼10 K with an
angular resolution of 0.2and a spectral resolution of
0.2 meV [Fig. 2(a)]. The spectra are almost entirely domi-
nated by the strong reflectivity peak of the SPP resonance
resulting from the coupling among the AM½1,SM½þ2,
and SM½3modes. This resonance is highly dispersive
and is essentially unaffected by the presence of the QW.
Very faint features from the QW and GaAs substrate (at
1.512 eV) resonances are also visible. As expected the QW
signature is weak, since most of the incident light is simply
reflected off the metal grating.
We expect that the weak QW signal has little effect on
the strong SPP peaks in the reflectivity spectra. Therefore
more detailed information on the QW reflectivity can be
obtained by subtracting the contribution from the SPP
resonance from the data. Recent studies on metallic nano-
arrays have shown that SPP line shape is well described by
a Fano-like line shape [14,17] given by Rð!; Þ¼
jrð!; Þj2with
rð!; Þ¼aþb
!!SPPðÞþiðþÞ:(1)
Here, aand bare frequency-independent complex ampli-
tudes and !SPPðÞis the SPP resonance frequency. The
nonradiative SPP damping rate is given by , whereas is
the dominant radiative damping rate [14,18].
In agreement with previous work, this Fano model
satisfactorily describes the experimental line shapes
[Fig. 2(a)]. The parameters of the model, i.e., the disper-
sion !SPPðÞof the coupled SPP resonance and the line-
width þ, agree well with our model calculations [4].
The SPP dispersion is mainly given by that of the AM½1
resonance.
Difference spectra obtained by subtracting these simu-
lations from the original data are shown in Fig. 2(c).We
observe a small remaining signature of the strongly dis-
persive SPP, a distinct LH exciton resonance at around
1.554 eV, a HH exciton resonance at 1.546 eV, and the
FIG. 1 (color online). (a) Schematic of the metal-
semiconductor hybrid structure consisting of a gold nanoslit
grating deposited on a GaAs QW. (b) Scanning electron micro-
scope image of the gold grating with a0¼500 nm,d¼
140 nm, and h¼80 nm. (c) Calculated angle-resolved far-field
reflectivity spectra of this structure. Dispersion relations for
different AM and SM SPP resonances are indicated as dash-
dotted lines. (d) Calculated spatial distribution of the normal
electric field Exfor Eex ¼1:517 eV and ¼38, near the
crossings of the SM½þ2,SM½3, and AM½1SPP reso-
nances.
PRL 101, 116801 (2008) PHYSICAL REVIEW LETTERS week ending
12 SEPTEMBER 2008
116801-2
excitonic excitation of the GaAs substrate at 1.512 eV. The
substrate feature is split into two closely spaced narrow
resonances reflecting a minor strain-induced lifting of the
degeneracy of LH and HH excitons. Their resonance en-
ergies are mainly unchanged when varying . In contrast,
for the QW excitonic transitions we observe a clear bend-
ing of both QW LH and HH resonances together with a
slight increase in broadening near the crossing with the
SPP resonance. The changes in line position and width are
a clear signature of exciton-SPP coupling in hybrid metal-
semiconductor nanostructures.
To quantitatively explain the SPP-QW interaction, a
phenomenological model is used, in which the AM½1,
SM½þ2, and exciton resonances are treated as three
coupled Lorentzian oscillators. Such a classical coupling
model is commonly used to study normal mode splittings
in various systems because the dispersion relations of the
coupled modes are mainly affected by the quantum prop-
erties of the emitters [19,20].
As a first approximation, the SPP couplings of HH and
LH excitons are treated separately because of their suffi-
ciently different energies. For each excitonic resonance the
reflectivity spectrum is modeled by a function of the form
of Eq. (1), but now with three Lorentzian oscillators in-
stead of one. The uncoupled SPP and exciton resonances
are represented by a diagonal matrix Huc where the com-
plex entries Vii are !iii, with i¼1corresponding to
AM½1,i¼2to SM½þ2, and i¼3to the QW HH, LH
or substrate exciton with known dispersion relations. The
total matrix for the coupled system is given as
Huc þHc¼
V11 00
0V22 0
00V33
0
@1
Aþ
0V12 0
V21 0V23
0V32 0
0
@1
A:
(2)
In Hc, the parameters Vij are complex coupling ele-
ments. V12 ¼V21 denotes the slit-induced coupling be-
tween AM½1and SM½þ2SPPs [14,15]. The
parameters V23 ¼V32 denote the coupling between the
exciton and the SM½þ2SPP. This reflects the coupling ~
~
ESPP between the exciton dipole moment ~ and the local
SPP field. Under our weak excitation conditions ~
ESPP is
dominated by the vacuum SPP field, as in previous experi-
ments [2,3]. In our structure there is no direct coupling
between air side plasmons and the QW [V13 ¼V31 ¼0].
Eigenvalues of this coupled matrix are used to calculate
Rð!; Þ. The calculated spectra compare well with the
experimental data [Fig. 2(b)]. To compare with Fig. 2(c),
the SPP contribution is subtracted from the simulated
Rð!; Þand the resulting QW dispersion is illustrated in
Fig. 2(d). Satisfying agreement with experiment is ob-
tained for the following coupling parameters: V12 ¼ð90 þ
30iÞmeV,VLH
23 ¼ð40 þ5iÞmeV,VHH
23 ¼ð50 6iÞmeV,
and Vsub
23 ¼ð18 2iÞmeV [Fig. 3(a)]. Because of the
considerable field enhancement in the metallic slit struc-
tures, these values for the coupling strengths are larger than
in many dielectric microcavities.
The model correctly predicts various salient features of
our measurements. Most importantly, it reproduces the
strongly asymmetric shift of the QW resonances seen in
Fig. 3(b). This shows that the QW exciton is coupled to at
least two interacting SPP modes. For small angles (<
36), these two modes interact constructively, giving rise to
a large spectral shift. On the high-angle side, destructive
interference diminishes the exciton-SPP interaction. The
model also explains why the observed shift of 6–8 meV is
much smaller than the predicted coupling strengths of
50 meV. In our structure the crossing between the AM
and SM modes is not perfectly matched to the exciton-SPP
resonance. This off-resonance condition reduces the effec-
tive coupling strength significantly. In Fig. 3(b), the shift of
the LH resonance is less pronounced than the HH shift.
This obviously reflects the larger in-plane component of
the HH dipole moment ~kand shows that the coupling is
induced by the in-plane components of the SPP field. The
very small shift of the substrate resonance evidently re-
flects its large spatial separation from the interface. This
FIG. 2 (color online). (a) Angle-resolved reflectivity spectra
(T¼10 K) of the hybrid structure. The dominant feature is the
resonance resulting from the coupling of the AM½1,SM½þ2,
and SM½3SPP modes. (b) Spectrum at ¼38(open circles)
together with simulations based on the Fano (blue solid line) and
the coupled-oscillator model (red dashed line). (c) Reflectivity
spectra obtained after subtracting the SPP contribution from (a).
The spectra reveal a clear shift of the HH and LH QW reso-
nances. (d) Spectra obtained from the coupled-oscillator model.
Red dash-dotted lines indicate the SPP and QW HH and LH
dispersions, respectively.
PRL 101, 116801 (2008) PHYSICAL REVIEW LETTERS week ending
12 SEPTEMBER 2008
116801-3
results in a reduced spatial overlap with evanescent SPP
fields [Fig. 2(d)] which cause the exciton-SPP coupling.
The results in Fig. 3(c) show a rather surprising angle-
dependent change in the linewidth of the exciton reso-
nances. The HH resonance width changes from a nearly
constant width of 6 meV to about 10 meV when brought
into resonance with the SPP mode. This means that the on-
resonance linewidth is not purely given by the predomi-
nantly inhomogeneous broadening of the QW exciton line,
but that the coupling to SPP induces an additional, and
surprisingly strong, radiative damping of the coherent ex-
citon polarization.
This is of interest because it means that the radiative
damping time of a quantum emitter can be decreased to
values of few ps or even below by coupling it to SPP in
metal nanoslit arrays, provided that (i) its emission energy
is in resonance with the SPP excitation and (ii) its dipole
moment is appropriately aligned to strongly couple to the
SPP modes. In such arrays SPP excitations are predomi-
nantly radiatively damped [in Eq. (1)] [14,18].
Therefore the luminescence efficiency of an emitter with
a low quantum yield (e.g., carbon nanotubes) can be en-
hanced by coupling its emission via the SPP excitations
into the far field. In contrast, metal structures with domi-
nant nonradiative damping (<)[12,13] will reduce the
luminescence yield. Of course for emitters with an inher-
ently large quantum yield, such as excitons in the GaAs
QW studied here, only minor enhancements of the lumi-
nescence yield are possible.
In conclusion, we have investigated for the first time the
coherent interaction between quantum well excitons and
surface plasmon polaritons in a novel hybrid metal-
semiconductor nanostructure. This coupling gives rise to
spectral shifts of the exciton resonance and to a surprising
increase in the radiative exciton damping. Our results
present a quantitative measure of the coupling strength
and show that couplings as large as 50 meV can be reached
in samples with optimized geometries. Because of the large
optical nonlinearities of QW excitons, such a strong
exciton-SPP coupling [21] is of considerable interest for
various future applications. This strong coupling may be
beneficial for enhancing the quality factor of metallic
nanoresonators by SPP amplification [22,23], or it may
help to build SPP lasers [22] and to transfer quantum
information over mesoscopic distances.
We gratefully acknowledge helpful discussions with
Q. H. Park (Seoul) and C. Ropers (Berlin). We thank
Alexander von Humboldt Foundation (P. V.) and DFG
(SFB 296) for financial support.
*christoph.lienau@uni-oldenburg.de
[1] T. J. Thompson, G. Rempe, and H. J. Kimble, Phys. Rev.
Lett. 68, 1132 (1992).
[2] T. Yoshie et al., Nature (London) 432, 200 (2004).
[3] C. Weisbuch, M. Nishioka, A. Ishikawa, and Y. Arakawa,
Phys. Rev. Lett. 69, 3314 (1992).
[4] E. M. Purcell, Phys. Rev. 69, 681 (1946).
[5] J. M. Gerard et al., Phys. Rev. Lett. 81, 1110 (1998).
[6] K. X. Drexhage et al.,Progress in Optics (North Holland,
Amsterdam, 1974), Vol. 12, p. 165.
[7] I. Pockrand, A. Brillante, and D. Moebius, J. Chem. Phys.
77, 6289 (1982).
[8] N. E. Hecker, R. A. Hoepfel, and N. Sawaki, Physica
(Amsterdam) 2E, 98 (1998).
[9] K. Okamoto et al., Nature Mater. 3, 601 (2004).
[10] Y. Fedutik et al., Phys. Rev. Lett. 99, 136802 (2007).
[11] W. Zhang, A. O. Govorov, and G. W. Bryant, Phys. Rev.
Lett. 97, 146804 (2006).
[12] I. Gontijo et al., Phys. Rev. B 60, 11 564 (1999).
[13] A. Neogi et al., Phys. Rev. B 66, 153305 (2002).
[14] C. Ropers et al., Phys. Rev. Lett. 94, 113901 (2005).
[15] K. G. Lee and Q. H. Park, Phys. Rev. Lett. 95, 103902
(2005).
[16] H. Lochbihler, Phys. Rev. B 50, 4795 (1994).
[17] C. Genet, M. P. van Exter, and J. P. Woerdman, Opt.
Commun. 225, 331 (2003).
[18] D. S. Kim et al., Phys. Rev. Lett. 91, 143901 (2003).
[19] R. Houdre
´et al., Phys. Rev. Lett. 73, 2043 (1994).
[20] D. G. Lidzey et al., Nature (London) 395, 53 (1998).
[21] J. Dintinger et al., Phys. Rev. B 71, 035424 (2005).
[22] D. J. Bergman and M. I. Stockman, Phys. Rev. Lett. 90,
027402 (2003).
[23] J. Seidel, S. Grafstroem, and L. Eng, Phys. Rev. Lett. 94,
177401 (2005).
FIG. 3 (color online). (a) Experimental (black solid lines) and
simulated (blue dashed lines) reflectivity spectra at ¼26and
34, respectively. The red dash-dotted lines are guides to the eye.
(b),(c) Experimental (open circles) and simulated (lines) energy
shifts (b) and line widths (c) of the LH (blue), HH (black), and
substrate (red) excitonic resonances as a function of angle . The
LH and substrate transitions are vertically shifted for clarity. In
(c), the green dashed lines mark the new reference positions.
PRL 101, 116801 (2008) PHYSICAL REVIEW LETTERS week ending
12 SEPTEMBER 2008
116801-4