Simultaneous Inhibition and
Redistribution of Spontaneous
Light Emission in Photonic Crystals
Masayuki Fujita, Shigeki Takahashi, Yoshinori Tanaka,
Takashi Asano, Susumu Noda*
Inhibiting spontaneous light emission and redistributing the energy into useful
forms are desirable objectives for advances in various fields, including photonics,
illuminations, displays, solar cells, and even quantum-information systems. We
demonstrate both the ‘‘inhibition’’ and ‘‘redistribution’’ of spontaneous light
index is changed two-dimensionally. The overall spontaneous emission rate is
Simultaneously, the light energy is redistributed from the 2D plane to the
direction normal to the photonic crystal.
Spontaneous light emission is a fundamental
factor (or bottleneck) limiting the performance
of devices in various fields, including photon-
ics (1), illuminations (2), displays (3), solar
cells (4), and quantum-information systems
(5). For example, in light-emitting diodes,
spontaneous emission that is not extracted
from the device contributes to losses. Similar-
ly, in lasers, spontaneous emission that does
not couple to the lasing mode results in both
losses and noise. Therefore, inhibiting un-
desirable spontaneous light emission and
redistributing (6) the energy into useful forms
will allow advances in these fields. Three-
dimensional (3D) photonic crystals (7–13),
which have a 3D periodic refractive-index
distribution that eliminates the optical modes
in all 3D directions via the photonic bandgap
(PBG) effect, have been used to demonstrate
the inhibition of spontaneous light emission
(11, 13). However, redistribution of the energy
has yet to be demonstrated.
Here, we investigate both the Binhibition[
and Bredistribution[ of spontaneous light
emission using 2D photonic crystals (14–19).
These crystals are predicted to provide a
mechanism for controlling spontaneous emis-
sion, which reflects their 2D nature (14): The
overall spontaneous emission rate is expected
to decrease as a result of the inhibition of
optical modes in all 2D directions by the 2D
PBG effect, whereas the emission efficiency
the 2D PBG effect does not appear) should
increase via redistribution of the saved energy.
Although an experimental trial has been
reported recently (20), it is not clear whether
the spontaneous emission is inhibited by the
2D PBG effect.
Spontaneous emission originates from fluc-
tuations in the vacuum field. Under a weak
light–matter coupling regime, the rate of spon-
taneous emission (Rspon) is given by Fermi_s
golden rule (21) and is determined by the
number of optical modes. In order to control
spontaneous light emission, the optical modes
must therefore be manipulated. Figure 1 shows
the 2D photonic crystal used, in which a
triangular-lattice 2D photonic crystal is formed
in a semiconductor (GaInAsP) slab. The struc-
ture incorporates a single quantum well (QW)
as the light-emitting material, emitting light
with a transverse-electric (TE) polarization in
which the dipole moment is orientated parallel
to the slab plane (22). The optical modes in
the thin slab can be categorized into two
modes: Bslab modes,[ which are confined to
Department of Electronic Science and Engineering,
Kyoto University, Katsura, Nishikyo-Ku, Kyoto 615-
*To whom correspondence should be addressed.
Fig. 1. The semiconductor (GaInAsP) 2D
photonic-crystal slab. A scanning electron
micrograph is shown. The 2D photonic-crystal
slab has a triangular lattice structure with an
air-hole radius r of 0.29a (lattice constant a 0
300 to 500 nm) and a thickness of 245 nm. A
5-nm-wide single QW is inserted at the center
of the slab to form the light-emitting layer.
Fig. 2. Theoretical analyses of the
effects of the 2D photonic-crystal
inhibition and energy redistri-
bution. A flat semiconductor
(InP) substrate is used as a refer-
ence. The refractive index and
thickness of the semiconductor
slab are 3.27 and 0.6a, respec-
tively. The refractive index of the
substrate and air are 3.13 and 1.0,
respectively. (A) Calculated spon-
taneous emission rate as a func-
tion of frequency, normalized by
the result calculated for the refer-
ence structure. (B) Emission effi-
ciency in the vertical direction for
various frequencies, normalized
by the results calculated for the
reference. (Top) Schematic views
illustrate how emission is con-
trolled by the 2D PBG effect. The
arrow denotes the corresponding
frequency region. (C) Photonic-
band diagram calculated by 3D
FDTD using periodic boundary
conditions. The solid lines show
the dispersion relation of the 2D
slab modes. The frequency region
between 0.267 and 0.330 cor-
responds to the 2D PBG region.
‘‘Leaky’’ refers to the region in
which the TIR condition is not
satisfied and light is emitted
outside the crystal surface.
R E P O R T S
27 MAY 2005 VOL 308SCIENCE www.sciencemag.org
the 2D plane by satisfying the total internal
reflection (TIR) condition for the vertical di-
rection, and Bvertical modes,[ which do not
satisfy the TIR condition and are emitted out-
side the slab. Excited carriers that are con-
fined in the QW generate spontaneous
emission coupled to both slab and vertical
modes, and the spontaneous emission rate
(Rspon) can be expressed by
Rspon0 Rslabþ Rvertical
Here, Rslaband Rverticaldenote the spontaneous
emission rate for slab modes and vertical
modes, respectively. For a semiconductor slab
with a large refractive index that is sur-
rounded by low–refractive index air cladding
(Fig. 1), light is strongly confined within the
slab, and the condition Rslabd Rverticalis
satisfied. Therefore, spontaneous emission
from the QW is mostly coupled to the slab
modes. However, when a 2D photonic-crystal
structure is incorporated in the slab, Rslabis
expected to be strongly reduced, whereas
basically no modification of Rverticaloccurs.
As a result, a marked reduction in Rsponis
expected through the reduction of Rslab. The
2D PBG is responsible for inhibiting emission
into the slab modes so that photons emitted
from the QW can only couple into the vertical
modes. Therefore, the efficiency of spontane-
ous emission in the vertical direction, which
can be detected, will increase. These two
effects—the reduction of Rspondue to the
inhibition of spontaneous emission into 2D
slab modes within the PBG and the increased
emission efficiency of the vertical modes—
will occur simultaneously. This implies that
the 2D photonic crystal effectively redistrib-
utes the light energy from the 2D plane to the
We quantitatively investigated the sponta-
neous emission modification and energy re-
distribution within the 2D photonic-crystal
structure. For these purposes, we calculated
efficiency intheverticaldirection usingthe 3D
finite-difference time-domain (FDTD) meth-
od (23, 24) and normalized the results using
values calculated for an equivalent structure
without a photonic crystal. The details of the
FDTD calculation are described in the sup-
porting text (25). Figure 2, A and B, show the
results, and Fig. 2C illustrates the corre-
sponding photonic band diagram for the 2D
spontaneous emission rate is reduced by more
thana factorof 15 withinthe PBGregion(blue
region) compared with that outside the PBG
region. Simultaneously, the emission efficien-
cy for the direction normal to the crystal (Fig.
2B) increases by more than a factor of 15
within the PBG region compared with that
outside the PBG region. Schematic views of
the cross-sectional electric field on such
situations are shown also in Fig. 2, where a
dipole oscillator is placed at one point on the
slab and is continuously excited so that the
total power emitted in all directions becomes
constant for each case. When the emission
wavelength is outside the PBG, emission
normal to the crystal is weak because Rslabd
Rverticaland the energy is distributed mostly in
the 2D slab modes. However, when the
wavelength is within the PBG, emission
normal to the crystal is increased notably,
despite the small value of Rvertical, because
emission into 2D slab modes is inhibited
(RslabÈ 0) and the energy is redistributed
into vertical emission modes.
Our samples were fabricated using a com-
bination of epitaxial growth, electron-beam
lithography, plasma etching, and chemical etch-
ing (26). A series of samples were prepared on
the same wafer, with the lattice constants a
ranging from 300 to 500 nm in 10-nm intervals,
in order to investigate a wide normalized-
frequency range. The samples were optically
pumped using a Ti-Al2O3laser emitting at
980 nm with a pulse width of 2 ps and a
repetition frequency of 2 MHz. The PBG
region was at first estimated for each sample
under a sufficiently strong excitation condition
(10 W/cm2on average) (27). Then,theaverage
laser power was reduced to 0.5 W/cm2so that
only recombination processes in the QW
involving emission of TE polarized light
could occur, in order to satisfy the assumption
made in the calculation. If the optical
absorption coefficient of the slab is assumed
to be 2 ? 104cm–1, the excited-carrier density
is estimated as È8 ? 1017cm–3. Under the
weak excitation condition, time-integrated
emission spectra (Fig. 3A), which indicate the
emission efficiency in the vertical direction,
were measured at 4 K using a multichannel
GaInAs detector system. Then, time-resolved
emission measurements (Fig. 3B) were made
single-photon-counting system (28) incorpo-
rating a photomultiplier detector, where we
collected spectra across the whole emission
wavelength range, in order to measure the
overall emission rate.
When the emission wavelength is within
the PBG region (Fig. 3, A and B) (a 0 390 to
480 nm), the spontaneous emission lifetime
(which corresponds to 1/Rspon) increases
compared with that observed in structures not
incorporating photonic crystals. Simultaneous-
ly, the emission efficiency increases in the
direction normal to the crystal compared with
that in structures not incorporating photonic
crystals. In contrast, when the emission wave-
Fig. 3. Experimental results. (A)
Time-integrated emission spectra
for samples with a range of lat-
tice constants between 350 and
500 nm. The blue cross-hatching
denotes the PBG region. (B) Time-
resolved photoluminescence mea-
surements for various samples.
When the spontaneous-emission
spectrum lies within the PBG region,
the emission lifetime increases by a
factor of 5 relative to that observed
when the spectrum lies outside the
PBG region (and compared with
that of the sample without a
photonic-crystal structure). A cor-
responding increase in the light-
emission efficiency in the vertical
direction, where the PBG effect
does not occur, is clearly observed
in the PBG region, as seen in (A).
R E P O R T S
www.sciencemag.org SCIENCEVOL 308 27 MAY 2005
length lies outside the PBG (Fig. 3, A and B) Download full-text
(a 0 350 and 500 nm), the spontaneous
emission lifetime and the emission efficiency
in the normal direction are both comparable to
those observed in structures without photonic
crystals. Figure 4 summarizes these results
over a range of normalized frequencies (open
circles) with the emission lifetimes converted
to corresponding emission rates. The 2D PBG
effect is clearly seen in Fig. 4A. The overall
emission rate decreases by a factor of 5 within
the PBG region compared with those outside
the PBG region, and the increase in emission
efficiency in the direction normal to the crystal
is clearly seen in Fig. 4B. These results are in
good agreement with the theoretical analyses
shown in Fig. 2. Therefore, simultaneous
inhibition and redistribution of spontaneous
emission are successfully demonstrated by this
We also fitted the experimental results
more quantitatively by considering the energy
loss due to nonradiative recombination of
excited carriers in the QW layer. Surface
recombination is the most likely mechanism
for nonradiative relaxation, and the rate (Rnon)
is proportional to the ratio of the exposed QW
surface area (Sw) to its overall volume (Vw)
(22), which is expressed as
Here, vsis the surface-recombination velocity
and r is the air-hole radius. The total re-
combination rate (RT) is therefore expressed as
RT0 Rsponþ Rnon
The experimentally observed rate should be
proportional to RT. Similarly, the emission
efficiency for the direction normal to the
photonic crystal can be modified in light of
this nonradiative recombination process. Un-
der these conditions, the experimental results
can be fitted, as shown in Fig. 4, A and B.
Lines have been drawn for various values of vs.
There is good agreement between the theoret-
ical calculation and the experimental results
for vs0 1.2 ? 103cm/s. This is a factor of 10
lower than typical values for GaInAsP (22) at
room temperature and is consistent with the
effects of sample cooling. We believe that
similar values of vswill be attainable at room
temperature, with special treatment of the
surface (29). Theuseofhigh-qualitycrystalline
silicon, with surface-recombination velocities
of less than 1 cm/s (30), might also be
advantageous for the suppression of nonradia-
tive processes, because it will lead to the
realization of high-quality silicon-based pho-
control of thermal (or black body) emission
would no longer be affected by surface-
We believe that our demonstration of simul-
taneous inhibition and redistribution of sponta-
neous light emission in photonic crystals is an
important step toward the evolution of photonic
devices and systems in various fields, including
photonics, illuminations, displays, solar cells,
and even quantum-information systems.
References and Notes
1. B. E. A. Saleh, M. C. Teich, Fundamentals of Photonics
(Wiley, New York, 1991).
2. D. A. Steigerwald et al., IEEE J. Sel. Top. Quantum
Electron. 8, 310 (2002).
3. K. Ziemelis, Nature 399, 408 (1999).
4. M. Gra ¨tzel, Nature 414, 338 (2001).
5. H. Mabuchi, A. C. Doherty, Science 298, 1372 (2002).
6. The term ‘‘redistributing’’ is used here to indicate
that the energy, which is otherwise utilized (or
distributed) for undesirable spontaneous emission, is
(re)distributed in more useful forms. In light-emitting
diodes, this energy, which is otherwise distributed in
spontaneous emission that cannot be extracted from
a device, is (re)distributed in emission that can be
extracted from the device. In lasers, the energy is
redistributed into the lasing mode. In solar cells, the
energy is redistributed for use as electrical energy.
7. E. Yablonovitch, Phys. Rev. Lett. 58, 2059 (1987).
8. S. John, Phys. Rev. Lett. 58, 2486 (1987).
9. S. Noda, K. Tomoda, N. Yamamoto, A. Chutinan,
Science 289, 604 (2000).
10. A. Blanco et al., Nature 405, 437 (2000).
11. S. Ogawa, M. Imada, S. Yoshimoto, M. Okano, S. Noda,
Science 305, 227 (2004).
12. M. Qi et al., Nature 429, 538 (2004).
13. P. Lodahl et al., Nature 430, 654 (2004).
14. S. Fan, P. R. Villeneuve, J. D. Jonnopoulos, Phys. Rev.
Lett. 78, 3294 (1997).
15. O. Painter et al., Science 284, 1819 (1999).
16. M. Boroditsky et al., J. Lightwave Technol. 17, 2096 (1999).
17. S. Noda, A. Chutinan, M. Imada, Nature 407, 608 (2000).
18. S. Noda, Y. Yokoyama, M. Imada, A. Chutinan, M.
Mochizuki, Science 293, 1123 (2001).
19. Y. Akahane, T. Asano, B. S. Song, S. Noda, Nature
425, 944 (2003).
20. H. Y. Ryu, Y. H. Lee, R. Sellin, D. Bimberg, Appl. Phys.
Lett. 79, 3573 (2001).
21. R. Loudon, The Quantum Theory of Light (Oxford
Univ. Press, New York, 2000).
22. L. A. Coldren, S. W. Corzine, Diode Lasers and Pho-
tonic Integrated Circuits (Wiley, New York, 1995),
pp. 148–153, 518–525.
23. J. K. Hwang, H. Y. Ryu, Y. H. Lee, Phys. Rev. B 60,
24. R.K.Lee,Y.Xu,A.Yariv,J.Opt.Soc.Am.B17,1438 (2000).
25. The details of the FDTD method are presented as
supporting material on Science Online.
26. M. Fujita, A. Sugitatsu, T. Uesugi, S. Noda, Jpn. J.
Appl. Phys. Part 2 43, L1400 (2004).
27. The PBG region is identified as follows. At the edges
of the PBG (the upper and lower boundaries of the
PBG region), the group velocity of light approaches
zero and waves propagating in different directions in
the slab couple in a type of 2D cavity resonator,
resulting in standing-wave formation. Under suffi-
ciently strong excitation, above the population-
inversion threshold, a resonant peak is observed at
the photonic band edge as a result of stimulated
emission. In these experiments, we used an excita-
tion intensity of 10 W/cm2, which is 20 times as
high as that used in the spontaneous emission-
lifetime measurement, to obtain such a population
inversion. In this way, we detected resonant peaks
corresponding to the photonic band edges over a
broad wavelength range. It was possible to fit the
relation between the band-edge wavelength and the
photonic-crystal period using a linear function,
obtaining lS0 2.28a þ 350 (nm) for the shorter
wavelength band edge and lL0 2.90a þ 357 (nm)
for the longer wavelength band edge. These results
are consistent with the 3D FDTD photonic band-
28. D. O’ Connor, D. Phillips, Time-Correlated Single Photon
Counting (Academic Press, London, 1985).
29. E. Yablonovitch, C. J. Sandroff, R. Bhat, T. Gmitter,
Appl. Phys. Lett. 51, 439 (1987).
30. E. Yablonovitch, D. L. Allara, C. C. Chang, T. Gmitter,
T. B. Bright, Phys. Rev. Lett. 57, 249 (1986).
31. This work was partly supported by a Grant-in-Aid
(no. 15GS0209) and an IT project grant from the
Ministry of Education, Culture, Sports, Science and
Technology of Japan, and by the Japan Science and
Technology Corporation (CREST). M.F. was supported
by a Research Fellowship of the Japan Society for the
Promotion of Science (no. 15004417).
Supporting Online Material
Figs. S1 and S2
References and Notes
31 January 2005; accepted 4 April 2005
Fig. 4. Plots of the experimental results
(open circles). (A) Overall spontaneous-
emission decay rates. (B) Emission effi-
ciency in the direction normal to the
crystal. The results were normalized by
those for the reference sample without
photonic crystals. The solid lines repre-
sent theoretically fitted results for a
range of values of vs. The experimentally
observed inhomogeneous broadening of
the emission spectra is accounted for in
the calculation. The closest correspon-
dence between the experimental results
and theoretical curves is obtained for vs0
R E P O R T S
27 MAY 2005 VOL 308 SCIENCEwww.sciencemag.org