Observation of In-Band Lattice Solitons
Xiaosheng Wang1and Zhigang Chen1,2
1Department of Physics and Astronomy, San Francisco State University, San Francisco, California 94132, USA
2TEDA Applied Physical School, Nankai University, Tianjin 300457, China
Jiandong Wang and Jianke Yang
Department of Mathematics and Statistics, University of Vermont, Burlington, Vermont 05401, USA
(Received 4 August 2007; published 14 December 2007)
We report the first experimental and theoretical demonstrations of in-band (or embedded) lattice
solitons. Such solitons appear in trains, and their propagation constants reside inside the first Bloch band
of a square lattice, different from all previously observed solitons. We show that these solitons bifurcate
from Bloch modes at the interior high-symmetry X points within the first band, where normal and
anomalous diffractions coexist along two orthogonal directions. At high powers, the in-band soliton can
move into the first band gap and turn into a gap soliton.
DOI: 10.1103/PhysRevLett.99.243901 PACS numbers: 42.65.Tg, 42.65.Sf, 42.65.Wi, 42.82.Et
Solitons are nonlinear wave phenomena ubiquitous in
nature , from fluid dynamics to biological systems, and
from nonlinear optics to Bose-Einstein condensates
(BECs). Such phenomena can exist in homogeneous or
periodic nonlinear systems. In optics, a spatial soliton
formed in periodic nonlinear media (such as waveguide
arrays) is often called a ‘‘discrete soliton’’  if its propa-
gation constant resides in the semi-infinite gap, or a ‘‘gap
soliton’’ (GS) [3–5] if its propagation constant resides in a
photonic band gap. A common belief is that lattice solitons
must reside in a gap, but not inside Bloch bands. In fact,
‘‘in-band’’ solitons in periodic systems have never been
demonstrated. A soliton whose propagation constant lies
inside the continuous spectrum of radiation modes is also
called an ‘‘embedded soliton’’ (ES) . Although it has
been shown theoretically that ESs could exist in homoge-
neousmedia orindiscrete (periodic) mediawith special
nonlinearities , such solitons have never been observed
in any physical systems.
In this Letter, we report the first experimental and theo-
retical demonstrations of in-band (embedded) lattice sol-
itons. Suchsolitonsare spatially localized in onetransverse
dimension, but extended as periodic wave trains in the
other transverse dimension, propagating longitudinally
without any change in shape. They are experimentally
realized with a unique excitation of a quasi-one-
dimensional (1D) stripe beam into a 2D optically induced
photonic lattice. Different from all previously observed
GSs [4,5], which typically bifurcate from the lowest or
the highest edges of a Bloch band, the ESs established here
bifurcate from the interior high-symmetry X points of the
first band [denoted as X1in Fig. 1(a)], with their propaga-
tion constants remaining inside the 1st band. We also show
that, at high powers, the propagation constants of ESs can
move into the 1st band gap (between the 1st and 2nd Bloch
bands), thus turning into GSs. This provides direct experi-
mental evidence that GSs can bifurcate not only from the
lowest or highest band edges but also from other locations
inside a Bloch band. Our work could have direct impact on
the study of nonlinear wave phenomena in other periodic
systems such as condensed matter physics or BECs [8,9].
The experimental setup used for this work is similar to
that we used earlier [10,11]. The lattice is induced in a
10 mm-long photorefractive SBN crystal by a spatially
modulated partially coherent light beam (488 nm) sent
through an amplitude mask. The mask is appropriately
imaged onto the input face of the crystal, creating a peri-
odic input intensity pattern [Fig. 1(b)] for optical induction
of 2D lattices. With a negative bias voltage, a ‘‘backbone’’
waveguide lattice (23 ?m spacing) is established as the
crystal turns into a defocusing nonlinear medium [4,5].
The lattice beam is ordinarily polarized; thus, it experien-
ces only a weak nonlinearity, while the probe beam is
extraordinarily polarized and thus undergoes strong non-
linear propagation due to the anisotropic photorefractive
nonlinearity. An incoherent background illumination is
employed to fine-tune the nonlinearity [4,11]. To excite
the X1points, the probe beam (cylindrically focused into a
narrow stripe beam) is launched into the crystal collinearly
with the lattice beam (i.e., without tilting angle), and the
FIG. 1 (color online).
square lattice with high-symmetry points marked by red dots.
(b),(c) Lattice and probe beams at crystal input with correspond-
ing k-space spectra shown in the insets. Dashed lines mark the
(a) Structure of the 1st Bloch band for a
PRL 99, 243901 (2007)
14 DECEMBER 2007
© 2007 The American Physical Society
stripe orientation is in parallel to one of the lattice principal
axes [Fig. 1(c)]. Indeed, its spatial (k-space) power spec-
trum contains only wave numbers along the y axis, and it
covers the two X points at the edge of the Brillouin zone
Under proper nonlinear conditions (i.e., a bias field of
?1:5 kV=cm, lattice intensity of about 1, and probe inten-
sity of 0.5 as normalized to background illumination), a
soliton train is observed whose intensity profile is shown in
Fig. 2(a). Interferograms between the soliton beam and a
tilted plane wave reveal that this soliton train has a phase
structure characteristic to Bloch modes at the X1points
, as the adjacent stripes are out of phasewith each other
[Fig. 2(b)] while the peaks along the same stripe are all in
phase [Fig. 2(c)]. This phase structure differs significantly
from that of the fundamental soliton train in the semi-
infinite gap bifurcating from the ? point of the 1st band
in focusing lattices , and from that of the GS train
bifurcating from the M points of the 1st band in defocusing
lattices .In fact, the k-space spectrum ofthe soliton train
in Fig. 2(d) shows that the spectrum is mostly located near
two X points. Since these X points are embedded inside the
1st band, the soliton we obtained here is thus an ES. It is
important to note that the distinctive spectrum in Fig. 2(d)
is due to the nonlinear self-action of the stripe beam. For
comparison, the spectrum of the same stripe beam prop-
agating linearly (before the self-action takes place) through
the lattice is also recorded in Fig. 2(e), which resembles the
input spectrum in Fig. 1(c) but differs greatly from the
soliton spectrum in Fig. 2(d). While experimentally it is
difficult to monitor the beam propagation through the
crystal, numerical results obtained with conditions corre-
sponding to those of the experiment shows perfect agree-
ment with results presented in Figs. 2(a)–2(e). Further-
more, simulations show that the ES can maintain its struc-
ture at propagation distances much longer than the crystal
length [Fig. 2(f)].
The formation of in-band solitons is intriguing, as usu-
ally a soliton is not expected to exist inside the continuous
spectrum . This impels us to do further theoretical
study, and to confirm what we observed is truly an ES in
this periodic system. Our theoretical model is a ?2 ? 1?D
NLS equation with a saturable defocusing nonlinearity and
a periodic lattice potential [4,15]:
iUz? Uxx? Uyy?
1 ? IL?x;y? ? jUj2U ? 0;
where U is the envelope function of the optical field, z is
the propagation distance, ?x;y? are the transverse coordi-
nates, E0is the applied dc field, and IL?x;y? ? I0sin2??x ?
(with peak intensity I0). All variables have been properly
normalized . For I0? 1:5 and E0? ?10 (close to our
experimental parameters), the lattice pattern and band gap
structure are shown in Figs. 3(a) and 3(d). Soliton solutions
in Eq. (1) are sought in the form of U?x;y;z? ? u?x;y??
exp?ikzz?, where kzis the propagation constant (if kzlies
inside a Bloch band, i.e., the continuous spectrum, then the
soliton is in-band or embedded). Using the modified
squared-operator iteration method developed in , we
do find a family of ES trains inside the 1st Bloch band
bifurcating from the X1points. The power curve of these
solitons is plotted in Fig. 3(d), where the power is defined
over one period along the train direction. A typical solution
of the ES train is shown in Figs. 3(b) and 3(c), which
?sin2??x ? y?=
? is the lattice intensity pattern
FIG. 2 (color online).
soliton. (a)–(c) Output pattern of the soliton train and its inter-
ferograms with a plane wave tilted from two different directions.
(d),(e) k-space spectrum for ES and for linear propagation,
respectively. (f) Numerical simulation under nonlinear (top)
and linear (bottom) propagation up to 30 mm.
Experimental observation of embedded
FIG. 3 (color online).
gap soliton trains in a square lattice of (a). Shown are the
intensity and spectrum of an embedded (b),(c) and gap (e),(f)
soliton. (d) The power verses propagation constant (kz) diagram,
where the red star (circle) in the 1st Bloch band (gap) marks the
kzlocation of the embedded (gap) soliton illustrated in the right
panels. The dashed line in (d) corresponds to the location of the
X point in the first band.
Theoretical solutions for embedded and
PRL 99, 243901 (2007)
14 DECEMBER 2007
consists of several periodically modulated stripes. Close
examination shows that the phase structure and spatial
spectrum of this soliton solution are in agreement with
that from the experiment of Fig. 2. More importantly, this
solution exists with its propagation constant located inside
the 1st Bloch band [marked by a red star inside the shaded
1stbandin Fig.3(d)]rather than inthe 1stgap, revealing its
nature of the ES.
The mechanism for formation of ESs merits further
discussion. These solitons bifurcate from the interior
high-symmetry X points in the 1st Bloch band, where the
diffraction surface has a saddle shape; i.e., the diffraction
coefficients have opposite signs along two orthogonal di-
rections [Fig. 1(a)]. Specifically, the diffraction at the X1
point marked in Fig. 1(a) is normal along the kxdirection
but anomalous along the kydirection. This saddle shape of
the diffraction surface serves for two purposes. First, it
makes the solitons existing near the saddle point to be
embedded. Second, it dictates that these solitons must be
localized along the y direction (where anomalous diffrac-
tion can be balanced by self-defocusing nonlinearity), but
delocalized along the x direction. This existence mecha-
nism is analogous to that for line solitons U?x;y;t? ?
face also has a saddle shape. In addition, the ESs form a
continuous family, similar to the above line solitons. We
point out that line solitons have never been observed, as
natural materials are not endowed with such saddle-shaped
iUt? Uxx? Uyy? jUj2U ? 0, where the dispersion sur-
rsech?ry?exp?ir2t? found in the ?2 ? 1?D NLS equation
dispersion.Even in ourinduced photonic lattices, a circular
or elliptical ES localized in both transverse dimensions
would not exist near the X1points. The fact that ESs reside
inside the continuous spectrum of the periodic system has
important implications for their dynamical evolutions
under perturbations. For instance, one of the key features
of all ESs is the resonant continuous-wave radiation under
Our analysis in Fig. 3(d) also shows that, when the
power of the ES is high, its propagation constant can
move into the photonic gap (between the 1st and 2nd
bands) to form a ‘‘true’’ GS. A typical solution of the GS
is presented in Fig. 3(e), whose intensity and phase struc-
tures are similar to those of the ES [Fig. 3(b)], but the
location of the propagation constant is now in-gap rather
than in-band. The spectrum of this GS [Fig. 3(f)] is also
different from that of the ES [Fig. 3(c)], with two bright
spots near X points merging towards the center, indicating
that inthegap the soliton poweris moreconcentrated in the
central stripe. Unlike the conventional GSs [4,5], the GS
train created here bifurcates from the interior X1points.
This illustrates that GSs can bifurcate not only from the
lowest or highest band edges but also from other locations
inside a band. Asimilar feature hasbeen discussedrecently
in the study of self-trapped gap waves in BECs .
Experimentally, it poses a challenge to identify the
transition between an ES and a GS, since there is no phase
transition between the two types of solitons, and the soliton
existence curve is continuous from the band to the gap
FIG. 4 (color online).
a GS at higher power, (c) restoration of an ES at reduced lattice potential, and (d) deterioration of the soliton when the lattice potential
is too low (see text for details). From top to bottom are the output intensity patterns, spectra, and interferograms.
Observation of transition between embedded and gap solitons. (a) An ES formed at low power, (b) transition to
PRL 99, 243901 (2007)
14 DECEMBER 2007
[Fig. 3(d)]. Nevertheless, we perform a series of experi- Download full-text
ments under different conditions in order to observe the
subtle difference between an ES and a GS. An example
from experimental results is presented in Fig. 4. First, with
a low-power beam (about 0.5 probe-to-lattice intensity
ratio), an ES is established [Fig. 4(a)] under conditions
similar to those of Fig. 2. Second, the intensity ratio is
increased to about 1.0 with all other conditions unchanged,
then a new soliton state is achieved as shown in Fig. 4(b).
While the intensity and phase structure appears to be
similar, the change in the spectrum is clearly visible as
the two bright spots tend to merge towards the center at the
higher power. This experimental observation agrees well
with the theoretical results of Fig. 3, indicating that an ES
can move into the gap as its power is increased.
Furthermore, after the GS train is formed, we decrease
the lattice potential (by reducing the lattice intensity by a
factor of 1=3 while keeping the same probe-to-lattice
intensity ratio), and we observe that the output of the
soliton [Fig. 4(c)] again resembles an ES with a distinctive
feature around the X points in the spectrum. This suggests
that the reduced lattice potential (accompanied by narrow-
ing of the gap and widening of the band) favors the
formation of the ES when its power is not increased
accordingly. Of course, if the lattice potential is too low
to open a gap, the ES can no longer sustain as seen in
Fig. 4(d), where the output spectrum becomes similar to
that of the linear propagation in Fig. 2(e), and the interfero-
gram no longer has an out-of-phase relation as shown in
Figs. 4(a)–4(c). Numerical simulations under these con-
ditions show similar behaviors. Despite this transition be-
tween in-band and in-gap solitons, we emphasize that the
ESs established here are fundamentally different from all
previously observed spatial GSs [4,5] or reduced-
symmetry GSs . In particular, the reduced-symmetry
GSs arise from Bloch modes at the X points (band edge) of
the second band, whereas the ESs arise from Bloch modes
at the interior X points (subband edge) of the first band.
In summary, we have demonstrated a new type of spatial
solitons, namely, in-band (or embedded) solitons. We have
shown that an embedded soliton can move into a photonic
band gap and turn into a gap soliton. These findings may
have direct impact on the study of nonlinear waves and
soliton phenomena in other branches of physics of periodic
systems. For instance, we expect such solitons to exist also
in BECs loaded in 2D optical lattice potential, now that
nonlinear self-trapping of matter waves have been demon-
strated in a number of experiments [9,17]. It would also be
interesting to explore if similar phenomena can exist in
other periodic systems such as condensed matter physics
where gap solitons have also been demonstrated , and if
such phenomena are related to persistent spin waves dis-
cussed for neutron-scattering experiments  or those
electronics and molecular bound states embedded in the
continuum , which have intrigued scientists for
This work was supported by NSF, AFOSR, PRF, the 973
Program and PCSIRT.
 M.J. Ablowitz and H. Segur, Solitons and the Inverse
Scattering Transform (SIAM, Philadelphia, 1985); Y.S.
Kivshar and G. Agrawal, Optical Solitons: From Fibers to
Photonic Crystals (Academic, San Diego, 2003).
 D.N. Christodoulides and R.I. Joseph, Opt. Lett. 13, 794
(1988); H.S. Eisenberg et al., Phys. Rev. Lett. 81, 3383
Siberberg, Nature (London) 424, 817 (2003).
 Y.S. Kivshar, Opt. Lett. 18, 1147 (1993).
 J.W. Fleischer et al., Phys. Rev. Lett. 90, 023902 (2003);
J.W. Fleischer et al., Nature (London) 422, 147 (2003); D.
Mandelik et al., Phys. Rev. Lett. 92, 093904 (2004); D.N.
Neshev et al., ibid. 93, 083905 (2004).
 C. Lou et al., Phys. Rev. Lett. 98, 213903 (2007); Z. Shi
and J. Yang, Phys. Rev. E 75, 061302 (2007).
 J. Yang et al., Phys. Rev. Lett. 83, 1958 (1999); J. Yang,
ibid. 91, 143903 (2003).
 K. Yagasaki et al., Nonlinearity 18, 2591 (2005).
 W.P. Su et al., Phys. Rev. Lett. 42, 1698 (1979); M. Aı ¨n
et al., ibid. 78, 1560 (1997).
 B. Eiermann et al., Phys. Rev. Lett. 92, 230401 (2004).
 Z. Chen and K. McCarthy, Opt. Lett. 27, 2019 (2002).
 H. Martin et al., Phys. Rev. Lett. 92, 123902 (2004).
 A.A. Sukhorukov et al., Phys. Rev. Lett. 92, 093901
(2004); D. Tra ¨ger et al., Opt. Express 14, 1913 (2006).
 Z. Chen et al., Phys. Rev. Lett. 92, 143902 (2004).
 J.P. Boyd, Weakly Nonlocal Solitary Waves and Beyond-
All-Orders Asymptotics (Kluwer, Boston, 1998).
 N.K. Efremidis et al., Phys. Rev. Lett. 91, 213906 (2003);
J. Yang, New J. Phys. 6, 47 (2004).
 J. Yang and T.I. Lakoba, Stud. Appl. Math. 118, 153
 Th. Anker et al., Phys. Rev. Lett. 94, 020403 (2005); T.J.
Alexander et al., ibid. 96, 040401 (2006).
 R. Fischer et al., Phys. Rev. Lett. 96, 023905 (2006).
 Y.J. Uemura et al., Phys. Rev. Lett. 51, 2322 (1983).
 L.S. Cederbaum et al., Phys. Rev. Lett. 90, 013001
F. Lederer, and Y.
PRL 99, 243901 (2007)
14 DECEMBER 2007