SCIENCE ADVANCES | RESEARCH ARTICLE

CONDENSED MATTER PHYSICS Copyright © 2018

The Authors, some

Anderson localizations and photonic band-tail states rights reserved;

exclusive licensee
observed in compositionally disordered platform American Association
for the Advancement
of Science. No claim to
Myungjae Lee,1,2,3 Jeongkug Lee,1,3* Sunghwan Kim,4 Ségolène Callard,5 original U.S. Government
Christian Seassal,5 Heonsu Jeon1,2,3† Works. Distributed
under a Creative
Anderson localization in random structures is an intriguing physical phenomenon, for which experimental verifications Commons Attribution
are far behind theoretical predictions. We report the first experimental confirmations of photonic band-tail states and a NonCommercial
complete transition of Anderson localization. An optically activated photonic crystal alloy platform enables the acqui- License 4.0 (CC BY-NC).
sition of extensive experimental data exclusively on pure eigenstates, revealing direct evidence of band-tail states and
Anderson localization transition within the band-tail states. Analyses of both experimental and simulated data lead to a
comprehensive picture of photon localization that is highly consistent with theories by Anderson and others. We be-
lieve that our results provide a strong experimental foundation upon which both the fundamental understandings and
application possibility of Anderson localization can be promoted significantly.

INTRODUCTION structures in an optically active medium and tracing the spectral evolu-
It has been nearly 60 years since Anderson published the theory of elec- tion of resultant lasing modes, we clarify that the eigenstates of the sys-
tron localization in disordered lattice structures (1), which is now re- tem are photonic band-tail states, which are speculated to be intimately
ferred to as Anderson localization. Anderson suggested that random related to strong photon localizations (7). In addition, direct visualiza-
potential fluctuations superimposed in a periodic lattice structure can tion of the wavelength-scale modal profiles, using a scanning near-field
produce a wide variety of electronic eigenstates that span from diffusive optical microscopy (SNOM) technique (28, 29), reveals that the band-
(thus, extended) to localized states. Numerous theoretical studies to un- tail states are Anderson-localized, with their localization strengths
derstand and predict the physics of the phenomenon have followed; the monotonically dependent on the degree of disorder as well as the state
studies of Abou-Chacra et al. (2), Abrahams et al. (3), Thouless (4), Lee energy. Our results demonstrate that photon localization can be ma-
and Ramakrishnan (5), John et al. (6, 7), and Mieghem (8) are only a nipulated over a wide spectro-spatial range in a controlled and predict-
selection of the most notable ones. Contrarily, experiments began only able manner, which should stimulate further studies to understand and
after it was found that localization could also occur in classical waves use the light-matter interactions via Anderson-localized modes (30, 31).
(9), such as electromagnetic waves (photons) (10–23) or acoustic waves
(24, 25) [later, this phenomenon was also confirmed in quantum waves
(26, 27)]. Previous experimental evidence regarding photon localization RESULTS AND DISCUSSION
is, however, predominantly based on macroscopic and collective data, PhC alloy system
such as backscattering (10–12) or transmission (13, 14); sometimes, The device system studied here consists of a set of 2D PhC alloys, where
their proofs can be indirect and therefore contested. Furthermore, most each device is fabricated by perforating an InAsP/InP multiple quantum
of these works used lattice-disordered platforms, probably influenced well (MQW) slab waveguide bonded on a fused quartz substrate with a
by the thermally excited disorder in solid-state systems; this kind hexagonal lattice array of air holes [see Materials and Methods for de-
of disorder, however, quickly destroys the crystalline symmetry, re- tails and section S1 for scanning electron microscopy (SEM) images].
sulting in systems very different from Anderson’s original model. To comply with the philosophy of the original Anderson model, in
A notable experimental breakthrough was made by Schwartz et al. which disorder is introduced only through potential fluctuations, we
(17); they succeeded in visualizing the localized photon modes in a adopt the compositional disorder (32) that preserves the crystalline
perturbed periodic potential by exploiting the effectively enlarged struc- symmetry regardless of disorder strength (see section S2). The compo-
tural dimension in transverse directions, albeit permitting the photon sitional disorder, which mimics that of compound semiconductor al-
propagation in longitudinal direction. They also refreshed the impor- loys, implies a random distribution of multiple (quaternary in this
tance of crystalline symmetry in Anderson localization, leading to suc- study) kinds of basis atoms at regular lattice sites, where the composi-
cessive research on disorder-enhanced photon transport (21–23). As a tional ratios among the multiple basis atoms remain the same through-
research platform for another experimental milestone, we propose a out the structure (Fig. 1A). We quantify the disorder strength using the
two-dimensional (2D) photonic crystal (PhC) alloy system that enables scheme of Conti et al. (33): The radius of each basis atom is given by r =
complete photon trapping. By realizing such Anderson-type disordered r0(1 + gx), where g (≥0) is the parameter that controls the degree of
disorder and x is a deviation that uniformly varies within the range of
1
Department of Physics and Astronomy, Seoul National University, Seoul 08826, −1/2 ≤ x ≤ 1/2. Because we design our PhC alloys to be quaternary, we
Republic of Korea. 2Institute of Applied Physics, Seoul National University, Seoul fix x at four equally spaced values: −1/2, −1/6, +1/6, and +1/2. r0 is the
08826, Republic of Korea. 3Inter-university Semiconductor Research Center, Seoul
National University, Seoul 08826, Republic of Korea. 4Department of Energy
air-hole radius of the reference PhC composed of homogeneous air
Systems Research and Department of Physics, Ajou University, Suwon 16499, Re- holes (g = 0). Figure 1B is a photonic band structure calculated for
public of Korea. 5Université de Lyon; Institut des Nanotechnologies de Lyon, UMR the reference PhC, where two band-edges (BEs) [K1 and M2, at both
CNRS 5270, CNRS, Ecole Centrale de Lyon, Ecully F-69134, France. ends of the band-gap (BG)] are denoted. Figure 1C schematically shows
*Present address: Display Research Center, Samsung Display, Yongin, Gyeonggi
17113, Republic of Korea. the photonic density of states as a function of photon energy. In an anal-
†Corresponding author. Email: [email protected] ogy to band-tail states in electronic disordered systems (such as semi-

Lee et al., Sci. Adv. 2018; 4 : e1602796 5 January 2018 1 of 7

SCIENCE ADVANCES | RESEARCH ARTICLE

Fig. 1. Compositionally disordered PhCs. (A) SEM image of a fabricated PhC alloy device. The device is composed of four kinds of photonic atoms with different air-
hole sizes, distributed randomly at otherwise regular hexagonal lattice sites. Multiple hexagons are intentionally overlapped with the SEM image to illustrate that the
hexagonal symmetry is strictly preserved. (B) Calculated photonic band structure of the reference PhC composed of homogenous air holes in a slab waveguide, ar-
ranged in the hexagonal lattice: r0 = 0.30a and h = 0.51a, where a = 450 nm is the lattice constant and h is the MQW slab thickness. (C) Schematic of photonic density of
states (PDOS) versus photon energy with an illustration of band-tail states expected as a consequence of disorder.

conductor alloys) (6, 8, 34, 35), we speculate that photonic band-tail (8). We note that, unlike in the simulation results, the experimentally
states develop inside the photonic BG upon the introduction of dis- determined width of the K1 band-tail states increases only to g = 0.5 and
order, which was theoretically predicted earlier by John (7). then decreases thereafter. However, this discrepancy is simply caused by
the reduction in optical gain overlap, not any intrinsic change in the
Laser spectra and band-tail states properties of the band-tail states (see section S5); that is, not all band-
We first measured microphotoluminescence (mPL) spectra from the tail states are subject to lasing.
fabricated PhC alloy devices in the far-field regime (see Materials and Figure 2F (as a function of g) shows the width of the energy range
Methods for details). Some representative emission spectra are shown over which the K1 band-tail states spread, which is defined by the
in Fig. 2A. Note that every sharp peak in the mPL spectra is an individual energy difference between the deepest and the shallowest band-tail
lasing mode (see section S3) and therefore corresponds to an eigenstate states, |ED – ES|. The penetration depth of the K1 band-tail states is
of the system. When the system is ordered (g = 0), the spectrum is domi- super-linearly proportional to g. This phenomenon is similar to an
nated by a single lasing mode, which we identify as the K1 BE mode (for Urbach tail, which often appears in the absorption edge of dis-
reasons that will be explained later). For intermediate disorders (0.2 ~ ordered solids (36). We note that the disorder-induced localized
g ~ 0.6), lasing actions occur at multiple closely spaced modes; the spectral states were found to be largely responsible for the Urbach tail (6).
width over which multiple lasing peaks appear becomes broadened as The similarity between electronics and photonics encourages us to
the degree of disorder increases. At high disorders (0.6 ~ g ~ 0.8), an- examine the mode profiles of the band-tail states in our system with
other set of lasing peaks develops from the short-wavelength side. For the expectation that they are spatially localized eigenstates.
theoretical confirmation, we performed finite-difference time-domain
(FDTD) simulations (see Materials and Methods for details). Figure Eigenmode profiles measured by near-field
2B shows simulation results that not only are consistent with the exper- optical microscopy
imental spectra but also predict that the two groups of lasing modes Figure 3A shows several representative near-field images measured for
eventually merge at high disorders. the K1 band-tail states (see Materials and Methods for details, section S6
Panels C and D are replots of panels A and B of Fig. 2, but are shown for the M2 band-tail states, and section S7 for the excitation power den-
as direct functions of the degree of disorder to elucidate the evolution of sity dependence). For a given value of g, the shallowest band-tail state
lasing modes with respect to the BG. The photonic BG is found to grad- should be nearly identical to the BE itself (see section S8). The images
ually narrow and eventually vanish as the degree of disorder increases. are taken from a single set of devices such that their disorder configura-
We claim that this is the first experimental observation that disorder- tions are identical; these are the devices for which mPL spectra are
induced eigenstates (or band-tail states) fill the photonic BG, as implied presented in Fig. 2A. Therefore, one can make direct comparisons
in theories by John (7). These observations also indicate that the groups not only among the SNOM images for different g values but also be-
of lasing modes at long- and short-wavelength sides originate from the tween emission spectra and mode images. When g = 0, for which only a
two BEs, K1 and M2, respectively. This argument is also supported by single K1 BE mode exists (see Fig. 2A), the SNOM image exhibits a pe-
modal profiles in momentum space (see section S4). riodic 2D Bloch mode profile throughout the entire excitation region;
To ensure generality, we apply statistical treatments to the measured the minor intensity variations are due to structural inhomogeneity inev-
results by taking an ensemble average over different PhC alloy config- itably introduced during device fabrication. In the presence of disorder,
urations. In this way, we can exclude (or minimize) the effect of spec- however, the observed near-field profiles are drastically changed.
ificity in disorder configurations. Figure 2E presents both the averaged Whereas the shallowest band-tail states are commonly characterized by
spectral positions of the upper (upward-pointing triangles) and lower a fairly large mode profile composed of multiple loosely connected
(downward-pointing triangles) bounds (in terms of wavelength) of segments, the deepest band-tail states exhibit a spatially isolated single-
scattered lasing modes, as functions of the degree of disorder, along with enveloped mode pattern. The localized mode shrinks in size as g in-
the ranges of band-tail states estimated by simulations (areas shaded in creases; the mode is as small as approximately 1 mm (only a small number
color). The width of the K1 band-tail states gradually increases as g in- of PhC lattices) for g = 0.6. Again, the FDTD simulation method is used
creases, which is followed by the situation in which the K1 and M2 band- for comparison, and the results shown in Fig. 3B agree very well with the
tail states merge and their discrimination becomes difficult (g ≥ 0.7). SNOM images. In particular, note that even the physical locations of lo-
The overall trend is similar to that of electronic disordered systems calized spots are correctly reproduced by the simulations.

Lee et al., Sci. Adv. 2018; 4 : e1602796 5 January 2018 2 of 7

SCIENCE ADVANCES | RESEARCH ARTICLE

Fig. 2. Band-tail states in lasing spectra. (A) Measured mPL spectra from the PhC alloy devices with different degrees of disorder (g = 0.0, 0.2, 0.4, 0.6, and 0.8). Each
spectrum is constructed by overlapping at least five individual spectra taken from different locations on the same device. The excitation power density was 6.2 kW/cm2
throughout the measurements. Each peak in the spectra corresponds to an individual laser line with a well-defined threshold. (B) Simulated |E|2 spectra for the model
structures that correspond to those shown in (A). (C and D) Measured (C) and simulated (D) spectra, plotted as a function of the degree of disorder. Note that simulations
are performed over an extended spectral range that includes the M2 BE as well. (E) Statistical summary of all the band-tail states identified from mPL measurements. For
each degree of disorder, the average (▲,▼) and SD (I) are determined from an ensemble of 10 PhC alloy devices with different disorder configurations. The shaded areas
are from simulations. (F) Energy range occupied by the K1 band-tail states, presented as a function of the degree of disorder. Results from mPL measurements and FDTD
simulations are presented as red bars and white bars, respectively. The solid line is an exponential fitting to the simulation result.

The modal patterns observed using SNOM (and confirmed by calization. The near-field images from the deepest band-tail states strong-
FDTD) are consistent with the theories by Anderson (1) and John (7), ly support this conceptual picture. Our observations also indicate that the
which explain the localization phenomenon via the hopping probability. strong and weak localized states can coexist in the same disordered struc-
According to the theory, the introduction of random potential fluctua- ture, which was also theoretically predicted (5).
tions breaks up a homogeneous Bloch mode because of the loss of To better understand modal evolutions, we compare the SNOM
long-range phase coherence, resulting in a mixture of multiple split field images (and the corresponding FDTD simulation results) of the K1
domains represented by an envelope function. When the state energy is band-tail states of various intermediate (between the shallowest and
shallow, as compared to potential fluctuations, the amplitude of the en- the deepest) energy levels from a single device with the degree of dis-
velope function between domains is finite, allowing photons to transfer by order g = 0.6 (Fig. 3C). Results show that the mode size monotonically
hopping between domains spread over the entire structure, which is shrinks, which indicates that the eigenstate gradually changes its
called weak localization. We believe that this is what we observed from character from a weaker to stronger localization, as the band-tail state
the shallowest band-tail states. For a deep state, however, the envelope becomes deeper. Therefore, eigenstates of the intermediate energy levels
function decays completely, far before it reaches other domains. This sit- are neither fully extended nor fully localized, but are instead in gradual
uation results in a completely isolated envelope function and thus no transition, qualitatively consistent with the discussion by Edwards and
hopping probability between domains, leading to a state called strong lo- Thouless (37).

Lee et al., Sci. Adv. 2018; 4 : e1602796 5 January 2018 3 of 7

SCIENCE ADVANCES | RESEARCH ARTICLE

Fig. 3. Mode patterns of the K1 band-tail states. (A) Representative SNOM images taken from devices with different degrees of disorder: g = 0.0, 0.2, 0.4, and 0.6. For
each degree of disorder, near-field images for the shallowest and deepest K1 band-tail states are shown in the left and right columns, respectively. All the SNOM images
were taken at a fixed excitation power density (~7 kW/cm2) well above laser thresholds of the modes. (B) Simulated |E|2 mode patterns for the band-tail states that
correspond to those shown in (A). (C) Measured (SNOM) and simulated (FDTD) images of various K1 band-tail states for a fixed degree of disorder (g = 0.6) but at
different band-tail state energies. The images are arranged in the order of state energy (from the shallowest at the top to the deepest at the bottom): DE ≡ E – ES = 0.0,
2.5, 4.9, 6.3, 9.5, and 13.0 meV. The inset is a schematic SNOM measurement configuration.

Effective widths and the Ioffe-Regel criterion affected “false” ones when making claims about Anderson localization.
The physical length scale over which a localized mode extends is a direct In our PhC alloy system, at least the deepest band-tail states of g ≥ 0.3
measure to quantify the strength of localization. From measured near- are candidates for Anderson localization.
field images, we can extract the effective width for a band-tail state, which Finally, we evaluated our devices in terms of the Ioffe-Regel factor,
2
12 kl, where k is the wave number and l is the mean free path. It is a the-
is defined byweff ≡P
2 ¼ f½∫Iðx; yÞ2 dxdy=½∫Iðx; yÞ dxdy g for a 2D
1

oretical notion that when the relative scales of l (=2p/k) and l become
system, where P is the participation ratio (17, 38). Results obtained from comparable, wave interference modifies transport properties drastically
all of the measured SNOM images are summarized in Fig. 4A. The ef- and a transition from extended to localized modes takes place (40, 41).
fective width is dependent on both the degree of disorder and the depth By analyzing the SNOM images of the deepest band-tail states (only for
of the band-tail state; that is, as g increases and the band-tail state g ≥ 0.3 as explained in the previous paragraph; see section S9), we ob-
becomes deeper, the effective width tends to be smaller, and thus, the tained 1.2 < kl < 2.0, which is close to the Ioffe-Regel condition (kl ≈ 1).
corresponding mode becomes more localized. FDTD simulations not However, it was pointed out that the criterion should be modified for
only confirm the experimental results with excellent agreement but also a periodic arrangement of coherent scatterers to k*l ≈ 1, where k* ≡
reveal that the M2 band-tail states exhibit a similar behavior (Fig. 4B). |k – kBE| is the wave number of the envelope wave function (41). We
Intuitively, increasing the area of the PhC pattern will result in the summarize the results in Fig. 4D, where the deepest band-tail states
expansion of the BE mode. However, such a simple size effect cannot are indicated in the (k*, k*l) plane. The modified Ioffe-Regel factors
occur in a strongly localized mode because the mode is already confined now become 0.15 < k*l < 0.45 (shaded area in Fig. 4D), well below 1,
within a local spot and, thus, is insensitive to the PhC pattern bound- and therefore, the Ioffe-Regel condition is fully satisfied. The corre-
aries. Figure 4C, which is obtained using FDTD simulations, shows how sponding mean free path is l ≈ 150 nm, which is only a fraction of the
the effective widths of the deepest band-tail states change as the PhC lattice constant a = 450 nm. This result explains why localized pho-
pattern size increases. When we enlarge the PhC domain size (from ton states are readily available in our devices, confirming that our
A0 to A) by appending additional photonic atoms around the initial PhC alloy system is an ideal platform on which to study the phenom-
PhC domain, the effective width for low disorder (0 ≤ g ≤ 0.2) mono- enon of Anderson localization.
tonically increases, starting from a relatively large value (>6 mm). How-
ever, the situation for higher disorder (g ≥ 0.3) is very different: The
effective width is independent of the PhC domain size. These results CONCLUSION
imply that the PhC pattern boundaries may or may not influence the Our 2D PhC alloy system constructed in an MQW slab waveguide not
formation of localized modes (39). Therefore, one should carefully dis- only conforms to Anderson’s original model structure but also has sev-
criminate the disorder-induced “true” localizations from the boundary- eral unique advantageous features: precision controllability of the degree

Lee et al., Sci. Adv. 2018; 4 : e1602796 5 January 2018 4 of 7

SCIENCE ADVANCES | RESEARCH ARTICLE

Fig. 4. Spatial extents of band-tail states. (A) Effective spatial widths of band-tail states, deduced from SNOM images for various degrees of disorder, band-tail state energies,
and disorder configurations. (B) Simulated effective widths of various band-tail states. Simulations were performed over an extended spectral range to include the M2 band-tail
states. (C) Effective width versus PhC pattern size, simulated for various degrees of disorder (0.0 ≤ g ≤ 0.7). Throughout simulations, the disorder configuration is fixed so as to
exclude configuration-specific fluctuations. (D) Modified Ioffe-Regel factors (k*l) estimated for the deepest band-tail states of g ≥ 0.3. The results are plotted as a function of the
envelope wave number k* such that the slope directly reflects the mean free path l.

of disorder and direct imaging capability for the internal mode profiles. proximately 1530 nm. The MQW section consists of four InAsP
Furthermore, the optically active nature of the system, achieved by con- MQWs separated by InP barriers, totaling 230 nm in nominal thickness,
structing the crystal symmetry–preserving disordered structure in an which corresponds to ~l/2n, where n is the effective refractive index
MQW slab waveguide, enables us to exclusively manipulate lasing modes of the slab waveguide. Grown underneath the MQW section was an
over a wide range of disorder, from weakly to strongly disordered regimes. InGaAs etch-stop layer for selective chemical etching. The grown
We have visualized localized modes in real scales and established a com- wafer was flip-bonded onto a fused silica substrate using a wafer fusion
prehensive picture of the localization transition in the band-tail states. technique (42), resulting in an InAsP/InP MQW layer that is directly
The evolution of localized modes in our system is highly reproducible in contact with the silica substrate. The InP substrate and InGaAs etch-
and can be tailor-designed, which we have proved to be inherent to the stop layer were then selectively removed in dilute HCl and FeCl3 solu-
band-tail states distributed over wide spectral (across the entire BG) and tions, respectively. A 50-nm-thick silicon nitride hard mask layer was
spatial (from only two to three lattices to the entire structure) ranges. deposited on the exposed surface of the MQW epilayer using the plasma-
Furthermore, all these observations are made from lasing modes, indicat- enhanced chemical vapor deposition method (310PC, Surface Technology
ing that the localized eigenstates in our platform constitute well-trapped Systems) at a process temperature of 200°C. Electron-beam lithography
photon modes with high-quality factors, which makes them attractive for (JBX-6300FS, JEOL) was then used to generate 2D PhC patterns, which
not only the fundamental research but also photonic device technology. was followed by reactive ion etching (RIE 80 Plus, Oxford Instrument)
We believe that these results can accelerate our understandings on dis- to sequentially transfer the PhC patterns onto the silicon nitride hard mask
order physics and also shed light on sophisticated next-generation pho- layer and the MQW layer. Finally, the hard mask layer was removed using
tonic devices in the areas where localized states can be engineered to the RIE to complete device fabrication.
multiplex their spectral and spatial degrees of freedom for enhanced de-
vice performance and new functionalities, such as biosensing, photovol- mPL spectra
taics, optical data storage, and cavity quantum electrodynamics. mPL measurements were performed using a homemade fiber-based
setup (43), in which a cleaved butt-end fiber tip with a 62.5-mm core
diameter is connected to a 1 × 2 multiplexing multimode fiber coupler.
MATERIALS AND METHODS The pump beam from a 1064-nm pulsed laser diode (PSL10, Multiwave
Device fabrication Photonics), operating at a 500-kHz repetition rate with a 20-ns pulse
An InP-based MQW epilayer structure was grown on an InP substrate duration, was coupled to the input port of the coupler. The cleaved fiber
using molecular beam epitaxy, with a center emission wavelength at ap- tip was brought in close proximity to the desired location on the PhC

Lee et al., Sci. Adv. 2018; 4 : e1602796 5 January 2018 5 of 7

SCIENCE ADVANCES | RESEARCH ARTICLE

sample surface. Scattered light output emitted from the PhC was then monochromator was tuned to the corresponding lasing mode
detected via the same fiber tip and delivered to the output port of the wavelength. The amount of spectral shift caused by probe perturbation
coupler, which was, in turn, fed to an optical spectrum analyzer was measured to be less than 1 nm. The spectral resolution of the mono-
(Q8381A, Advantest) for spectral analysis. chromator was set to ±1 nm for half width at half maximum, which is an
optimal compromise between signal-to-noise ratio and mode distin-
FDTD simulation guishability. For a single SNOM scan, more than 2.6 × 105 pixel data
Numerical simulations based on the FDTD method were performed points were collected, which corresponds to a spatial resolution of
with a commercial FDTD software package (FDTD solutions, Lumerical 17 pixels/mm.
Solutions). For each model structure, the simulation domain size was
32 mm × 28 mm, including the padding. The spatial resolution was set
to 45 pixels/mm, which is sufficiently high that we can distinguish SUPPLEMENTARY MATERIALS
individual modes. We used the so-called 2.5D FDTD method where Supplementary material for this article is available at http://advances.sciencemag.org/cgi/
a 3D structure is collapsed into a set of effective 2D materials, with content/full/4/1/e1602796/DC1
effective refractive indices determined from the in-plane wave vector section S1. SEM images of fabricated devices
section S2. Ordered and disordered components in compositional disorder
component. This method is known to offer 3D accuracy using only
section S3. Lasing properties of band-tail states
2D computing resources, although it is applicable for a specific section S4. Mode profiles in momentum space
simulation bandwidth (44). In our simulations, the effective index section S5. Gain overlap factors of band-tail states
of neff = 2.7 was used for our PhC waveguide slab. The model PhC section S6. Eigenmode profiles of the M2 band-tail states
structures used in the simulations are nominally identical to the real section S7. Excitation strength dependence of modal properties
section S8. Estimation of the K1 BE positions in disordered PhCs
devices examined experimentally, because the same CAD files were section S9. Localization length and mean free path
shared by both electron-beam lithography and FDTD simulations. section S10. The Bloch states in simulated spectra
To mimic real experimental conditions and to excite as many reso- section S11. Effective width in momentum space
nant modes existing in the model structure as possible, we applied the section S12. Resolution dependence of effective widths
fig. S1. Fabricated devices.
dipole cloud approach, in which many dipoles with random positions,
fig. S2. Ordered versus disordered components.
polarizations, and phases are incorporated within a virtual pump spot. fig. S3. Lasing properties of band-tail states.
Likewise, we placed many randomly positioned time monitors over the fig. S4. Momentum space profiles of the band-tail states.
entire PhC slab such that all of the resonant modes can be identified fig. S5. Gain overlap factors of band-tail states.
from the resultant spectrum. In the analysis step immediately before fig. S6. Mode profiles of the M2 band-tail states.
fig. S7. Excitation power density dependence of modal properties.
the Fourier transformation to extract the modal spectra, we applied fig. S8. K1 band edges in disordered PhCs.
the apodization of time signals with a Gaussian profile to eliminate fig. S9. Localization length and mean free path.
the start and end effects. For the simulated |E|2 distribution of each res- fig. S10. Bloch states.
onant mode, the same approach was taken, except that time monitors fig. S11. Effective widths in momentum space.
fig. S12. Resolution dependence of effective widths.
were replaced with 2D frequency-domain field monitors.
References (45–48)

Near-field imaging
To directly obtain the modal pattern for an individual lasing mode, we
custom-designed and built a SNOM apparatus operating in the near- REFERENCES AND NOTES
infrared range (28). The scanning part of the apparatus was built by in- 1. P. W. Anderson, Absence of diffusion in certain random lattices. Phys. Rev. 109,
1492–1505 (1958).
tegrating a scanning controller system (SMENA, NT-MDT) with an 2. R. Abou-Chacra, D. J. Thouless, P. W. Anderson, A self-consistent theory of localization.
inverted optical microscope (Eclipse Ti-S, Nikon). The detecting part J. Phys. C Solid State Phys. 6, 1734–1752 (1973).
was assembled from a monochromator (MicroHR, Horiba), an InGaAs 3. E. Abrahams, P. W. Anderson, D. C. Licciardello, T. V. Ramakrishnan, Scaling theory of
photodiode (DSS-IGA025T, Horiba), and a lock-in amplifier (SR830, localization: Absence of quantum diffusion in two dimensions. Phys. Rev. Lett. 42,
673–676 (1979).
Stanford Research Systems).
4. D. J. Thouless, Electrons in disordered systems and the theory of localization. Phys. Rep.
The SNOM apparatus was operated in a collection geometry in 13, 93–142 (1974).
which a laser diode, which is identical to that used for the mPL mea- 5. P. A. Lee, T. V. Ramakrishnan, Disordered electronic systems. Rev. Mod. Phys. 57, 287–337
surements (except for an additional 1-kHz modulation to provide a (1985).
reference signal to the lock-in amplifier), delivered a pump beam from 6. S. John, C. Soukoulis, M. H. Cohen, E. N. Economou, Theory of electron band tails and the
Urbach optical-absorption edge. Phys. Rev. Lett. 57, 1777–1780 (1986).
the back side of the sample (through the transparent silica substrate). 7. S. John, Strong localization of photons in certain disordered dielectric superlattices.
The near-field signal from the sample was recorded by the SNOM Phys. Rev. Lett. 58, 2486–2489 (1987).
probe while the probe tip scanned the front surface of the sample. 8. P. Van Mieghem, Theory of band tails in heavily doped semiconductors. Rev. Mod. Phys.
We used the scanning-by-probe scheme, in which only the probe 64, 755–793 (1992).
9. P. W. Anderson, The question of classical localization: A theory of white paint?
mounted in the scan head moves during the scanning process. This
Philos. Mag. B 52, 505–509 (1985).
scheme is particularly useful in our study because the relative positions 10. M. P. Van Albada, A. Lagendijk, Observation of weak localization of light in a random
of pump beam and sample are preserved during scanning such that a medium. Phys. Rev. Lett. 55, 2692–2695 (1985).
high-quality near-field image over a large scan area (30 mm × 30 mm) 11. P.-E. Wolf, G. Maret, Weak localization and coherent backscattering of photons in
can be obtained. disordered media. Phys. Rev. Lett. 55, 2696–2699 (1985).
12. S. Etemad, R. Thompson, M. J. Andrejco, Weak localization of photons: Universal
The near-field intensity profile was obtained for each of the lasing fluctuations and ensemble averaging. Phys. Rev. Lett. 57, 575–578 (1986).
peaks seen in the mPL spectra using a commercial dielectric SNOM 13. D. S. Wiersma, P. Bartolini, A. Lagendijk, R. Righini, Localization of light in a disordered
probe (MF115_NTF/WA, NT-MDT). The center wavelength of the medium. Nature 390, 671–673 (1997).

Lee et al., Sci. Adv. 2018; 4 : e1602796 5 January 2018 6 of 7

SCIENCE ADVANCES | RESEARCH ARTICLE

14. M. Störzer, P. Gross, C. M. Aegerter, G. Maret, Observation of the critical regime near 37. J. T. Edwards, D. J. Thouless, Numerical studies of localization in disordered systems.
Anderson localization of light. Phys. Rev. Lett. 96, 063904 (2006). J. Phys. C Solid State Phys. 5, 807–820 (1972).
15. R. Dalichaouch, J. P. Armstrong, S. Schultz, P. M. Platzman, S. L. McCall, Microwave 38. R. J. Bell, P. Dean, D. C. Hibbins-Butler, Localization of normal modes in vitreous silica,
localization by two-dimensional random scattering. Nature 354, 53–55 (1991). germania and beryllium fluoride. J. Phys. C Solid State Phys. 3, 2111–2118 (1970).
16. A. A. Chabanov, M. Stoytchev, A. Z. Genack, Statistical signatures of photon localization. 39. D. M. Jović, Y. S. Kivshar, C. Denz, M. R. Belić, Anderson localization of light near
Nature 404, 850–853 (2000). boundaries of disordered photonic lattices. Phys. Rev. A 83, 033813 (2011).
17. T. Schwartz, G. Bartal, S. Fishman, M. Segev, Transport and Anderson localization in 40. A. F. Ioffe, A. R. Regel, Non-crystalline, amorphous, and liquid electronic semiconductors.
disordered two-dimensional photonic lattices. Nature 446, 52–55 (2007). Prog. Semicond. 4, 237–291 (1960).
18. J. Fallert, R. J. Dietz, J. Sartor, D. Schneider, C. Klingshirn, H. Kalt, Co-existence of strongly 41. S. John, Localization of light. Phys. Today 44, 32–40 (1991).
and weakly localized random laser modes. Nat. Photonics 3, 279–282 (2009). 42. C. Seassal, C. Monat, J. Mouette, E. Touraille, B. B. Bakir, H. T. Hattori, J.-L. Leclercq,
19. T. Sperling, W. Bührer, C. M. Aegerter, G. Maret, Direct determination of the transition to X. Letartre, P. Rojo-Romeo, P. Viktorovitch, InP bonded membrane photonics
localization of light in three dimensions. Nat. Photonics 7, 48–52 (2013). components and circuits: Toward 2.5 dimensional micro-nano-photonics. IEEE J. Sel. Top.
20. A. Crespi, R. Osellame, R. Ramponi, V. Giovannetti, R. Fazio, L. Sansoni, F. D. Nicola, Quant. Electron. 11, 395–407 (2005).
F. Sciarrino, P. Mataloni, Anderson localization of entangled photons in an integrated 43. Y. Park, S. Kim, C. Moon, H. Jeon, H. J. Kim, Butt-end fiber coupling to a surface-emitting
quantum walk. Nat. Photonics 7, 322–328 (2013). G-point photonic crystal band edge laser. Appl. Phys. Lett. 90, 171115 (2007).
21. L. Levi, Y. Krivolapov, S. Fishman, M. Segev, Hyper-transport of light and stochastic 44. M. Hammer, O. V. Ivanova, Effective index approximations of photonic crystal slabs:
acceleration by evolving disorder. Nat. Phys. 8, 912–917 (2012). A 2-to-1-D assessment. Opt. Quant. Electron. 41, 267–283 (2009).
22. M. Leonetti, S. Karbasi, A. Mafi, C. Conti, Observation of migrating transverse Anderson 45. H. Cao, Y. G. Zhao, S. T. Ho, E. W. Seelig, Q. H. Wang, R. P. H. Chang, Random laser action in
localizations of light in nonlocal media. Phys. Rev. Lett. 112, 193902 (2014). semiconductor powder. Phys. Rev. Lett. 82, 2278–2281 (1999).
23. P. Hsieh, C. Chung, J. F. McMillan, M. Tsai, M. Lu, N. C. Panoiu,C. W. Wong, Photon 46. E. Yablonovitch, T. J. Gmitter, R. D. Meade, A. M. Rappe, K. D. Brommer, J. D. Joannopoulos,
transport enhanced by transverse Anderson localization in disordered superlattices. Donor and acceptor modes in photonic band structure. Phys. Rev. Lett. 67, 3380–3383
Nat. Phys. 11, 268–274 (2015). (1991).
24. R. L. Weaver, Anderson localization of ultrasound. Wave Motion 12, 129–142 (1990). 47. S. Kim, S. Yoon, H. Seok, J. Lee, H. Jeon, Band-edge lasers based on randomly mixed
25. H. Hu, A. Strybulevych, J. H. Page, S. E. Skipetrov, B. A. van Tiggelen, Localization of photonic crystals. Opt. Express 18, 7685–7692 (2010).
ultrasound in a three-dimensional elastic network. Nat. Phys. 4, 945–948 (2008). 48. T. Quang, M. Woldeyohannes, S. John, G. S. Agarwal, Coherent control of spontaneous
26. U. Gavish, Y. Castin, Matter-wave localization in disordered cold atom lattices. Phys. Rev. emission near a photonic band edge: A single-atom optical memory device. Phys. Rev.
Lett. 95, 020401 (2005). Lett. 79, 5238–5241 (1997).
27. J. Billy, V. Josse, Z. Zuo, A. Bernard, B. Hambrecht, P. Lugan, D. Clément,
L. Sanchez-Palencia, P. Bouyer, A. Aspect, Direct observation of Anderson localization of Acknowledgments
matter waves in a controlled disorder. Nature 453, 891–894 (2008). Funding: This work was supported by the National Research Foundation (NRF) grant funded
28. N. Louvion, D. Gérard, J. Mouette, F. de Fornel, C. Seassal, X. Letartre, A. Rahmani, by the Korea Government (Ministry of Education, Science and Technology) (no.
S. Callard, Local observation and spectroscopy of optical modes in an active 2014R1A2A1A11051576). The collaboration between Seoul National University and Institut
photonic-crystal microcavity. Phys. Rev. Lett. 94, 113907 (2005). des nanotechnologies de Lyon/Ecole Centrale de Lyon was carried out in the frame of the
29. F. Intonti, S. Vignolini, F. Riboli, A. Vinattieri, D. S. Wiersma, M. Colocci, L. Balet, C. Monat, French-Korean Laboratoire International Associé “Center for Photonics and Nanostructures.”
C. Zinoni, L. H. Li, R. Houdré, M. Francardi, A. Gerardino, A. Fiore, M. Gurioli, Spectral Author contributions: M.L. conducted the majority of experimental work, including design,
tuning and near-field imaging of photonic crystal microcavities. Phys. Rev. B 78, 041401 fabrication, measurements, and simulations. J.L. and S.K. constructed the SNOM setup and
(2008). performed initial data collection. S.C. and C.S. provided the bonded InAsP MQW wafers and
30. L. Sapienza, H. Thyrrestrup, S. Stobbe, P. D. Garcia, S. Smolka, P. Lodahl, Cavity quantum involved SNOM measurements. H.J. conceived and directed the research. All authors
electrodynamics with Anderson-localized modes. Science 327, 1352–1355 (2010). contributed to the scientific discussions and the preparation of the manuscript. Competing
31. A. Cazé, R. Pierrat, R. Carminati, Strong coupling to two-dimensional Anderson localized interests: The authors declare that they have no competing interests. Data and materials
modes. Phys. Rev. Lett. 111, 053901 (2013). availability: All data needed to evaluate the conclusions in the paper are present in the
32. S. Kim, J. Lee, H. Jeon, S. Callard, C. Seassal, K.-D. Song, H.-G. Park, Simultaneous paper and the Supplementary Materials. Additional data related to this paper may be
observation of extended and localized modes in compositional disordered photonic requested from the authors.
crystals. Phys. Rev. A 88, 023804 (2013).
33. C. Conti, A. Fratalocchi, Dynamic light diffusion, three-dimensional Anderson localization Submitted 11 November 2016
and lasing in inverted opals. Nat. Phys. 4, 794–798 (2008). Accepted 30 November 2017
34. R. Zimmermann, E. Runge, Excitons in narrow quantum wells: Disorder localization and Published 5 January 2018
luminescence kinetics. Phys. Stat. Sol. (A) 164, 511–516 (1997). 10.1126/sciadv.1602796
35. H. F. Hess, E. Betzig, T. D. Harris, L. N. Pfeiffer, K. W. West, Near-field spectroscopy of the
quantum constituents of a luminescent system. Science 264, 1740–1745 (1994). Citation: M. Lee, J. Lee, S. Kim, S. Callard, C. Seassal, H. Jeon, Anderson localizations and
36. F. Urbach, The long-wavelength edge of photographic sensitivity and of the electronic photonic band-tail states observed in compositionally disordered platform. Sci. Adv. 4,
absorption of solids. Phys. Rev. 92, 1324 (1953). e1602796 (2018).

Lee et al., Sci. Adv. 2018; 4 : e1602796 5 January 2018 7 of 7