Anda di halaman 1dari 14



The ever-increasing demand for faster information transport and processing capabilities is
undeniable. Our data-hungry society has driven enormous progress in the Si electronics industry
and we have witnessed a continuous progression towards smaller, faster, and more efficient
electronic devices over the last five decades. The scaling of these devices has also brought about
a myriad of challenges. Currently, two of the most daunting problems preventing significant
increases in processor speed are thermal and signal delay issues associated with electronic
interconnection. Optical interconnects, on the other hand, possess an almost unimaginably large
data carrying capacity, and may offer interesting new solutions for circumventing these
problems. Optical alternatives may be particularly attractive for future chips with more
distributed architectures in which a multitude of fast electronic computing units (cores) need to
be connected by high-speed links. Unfortunately, their implementation is hampered by the large
size mismatch between electronic and dielectric photonic components. Dielectric photonic
devices are limited in size by the fundamental laws of diffraction to about half a wavelength of
light and tend to be at least one or two orders of magnitude larger than their nanoscale electronic
counterparts. This obvious size mismatch between electronic and photonic components presents
a major challenge for interfacing these technologies. Further progress will require the
development of a radically new chip-scale device technology that can facilitate information
transport between nanoscale devices at optical frequencies and bridge the gap between the world
of nanoscale electronics and microscale photonics.

Plasmons are being considered as a means of transmitting information on computer chips,

since plasmons can support much higher frequencies (into the 100 THz range, while
conventional wires become very lossy in the tens of GHz). For plasmon-based electronics to be
useful, the analog to the transistor, called a plasmonster


Metal nanostructures may possess exactly the right combination of electronic and optical
properties to tackle the issues outlined above and realize the dream of significantly faster
processing speeds. The metals commonly used in electrical interconnection such as Cu and Al
allow the excitation of surface plasmon-polaritons (SPPs). SPPs are electromagnetic waves that
propagate along a metal-dielectric interface and are coupled to the free electrons in the metal
It is important to realize that, with the latest advances in electromagnetic simulations and current
complementary metal-oxide semiconductor (CMOS)-compatible fabrication techniques, a variety
of functional plasmonic structures can be designed and fabricated in a Si foundry right now.
Current Si-based integrated circuit technology already uses nanoscale metallic structures, such as
Cu and Al interconnects, to route electronic signals between transistors on a chip.

This mature processing technology can thus be used to our advantage in integrating plasmonic
devices with their electronic and dielectric photonic counterparts. In some cases, plasmonic
waveguides may even perform a dual function and simultaneously carry both optical and
electrical signals, giving rise to exciting new capabilities.

Fig. 1 An SPP propagating along a metal-dielectric interface. These waves are transverse magnetic in nature. Their
electromagnetic field intensity is highest at the surface and decays exponentially away from the interface. From an engineering
standpoint, an SPP can be viewed as a special type of light wave propagating along the metal surface.


In order to study the propagation of SPPs, we constructed a photon scanning tunneling
microscope (PSTM) by modifying a commercially available scanning near-field optical
microscope. PSTMs are the tool of choice for characterizing SPP propagation along extended
films as well as metal stripe waveguides Fig. 2a shows how a microscope objective at the heart
of our PSTM can be used to focus a laser beam onto a metal film at a well-defined angle and
thereby launch an SPP along the top metal surface. This method of exciting SPPs makes use of
the well-known Kretschmann geometry that enables phase matching of the free space excitation
beam and the SPP.

Fig. 2 (a) Schematic of the operation of a PSTM that enables the study of SPP propagation along metal film surfaces.
The red arrow shows how an SPP is launched from an excitation spot onto a metal film surface using a high
numerical aperture microscope objective.

A sharp, metal-coated pyramidal tip (Figs. 2b and 2c) is used to tap into the guided SPP
wave locally and scatter light toward a far-field detector. These particular tips have a nanoscale
aperture at the top of the pyramid through which light can be collected. The scattered light is
then detected with a photomultiplier tube. The signal provides a measure of the local light
intensity right underneath the tip and, by scanning the tip over the metal surface, the propagation
of SPPs can be imaged.

Fig 2(b) Scanning electron microscopy (SEM) image of the near field optical cantilever probe used in our
experiments. The tip consists of a micro fabricated, hollow glass pyramid coated with an optically thick layer of Al.
Light can be collected or emitted through a ~50 nm hole fabricated in the Al film on the top of the pyramid.
(c) A cross-sectional view of the same hollow pyramidal tip after a large section was cut out of the sidewall with a
focused ion beam (FIB). In close proximity to the surface, the pyramidal tip can tap into the propagating SPP and
scatter out a little bit of light through the ~ 50 nm hole (shown pictorially). The scattered light is detected in the
far-field, providing a measure of the local field intensity right underneath the tip. By scanning the tip over the
sample and measuring the intensity at each tip position, images of propagating SPPs can be created.

The operation of the PSTM can be illustrated by investigating the propagation of SPPs on
a patterned Au film (Fig. 2d). Here, a focused ion beam (FIB) was used to define a series of
parallel grooves, which serve as a Bragg grating to reflect SPP waves. Fig. 2e shows a PSTM
image of an SPP wave excited with a 780 nm wavelength laser and directed toward the Bragg
grating. The back reflection of the SPP from the grating results in the standing wave interference
pattern observed in the image. From this type of experiment the wavelength of SPPs can be
determined in a straightforward manner and compared to theory.

Fig 3(d) SEM image of a Au film into which a Bragg grating has been fabricated using a FIB.(e) PSTM image of an
SPP wave launched along the metal film toward the Bragg grating. The back reflection of the SPP from the Bragg
grating results in the observation of a standing wave interference pattern.

The PSTM can also be used to image SPP propagation directly in plasmonic structures
and devices of more complex architecture to determine their behavior. This is quite different
from typical characterization procedures for photonic devices in which the device is seen as a
black box with input and output ports. In such cases, the device operation is inferred from
responses measured at output ports to different stimuli provided at the input ports. The PSTM
provides a clear advantage by providing a direct method to observe the inner workings of
plasmonic devices, offering a peek inside the box.



Based on the data presented above, it seems that the propagation lengths for Plasmonic
waveguides are too short to propagate SPPs with high confinement over the length of an entire
chip (~1 cm). Although the manufacturability of long-range SPP waveguides may well be
straightforward within a CMOS foundry, it is unlikely that such waveguides will be able to
compete with well-established, low-loss, high-confinement Si, Si3N4, or other dielectric
waveguides. However, it is possible to create new capabilities by capitalizing on an additional
strongpoint of metallic nanostructures. Metal nanostructures have a unique ability to concentrate
light into nanoscale volumes. This capability has been employed to enhance a diversity of
nonlinear optical phenomena. For example, surface-enhanced Raman scattering (SERS) is
widely used in the field of biology.
This technique makes use of the enhanced electromagnetic fields near metallic
nanostructures to study the structure and composition of organic and biological materials.
Enhancement factors on the order of 100 have been predicted and observed for spherical
particles. Even greater enhancements can be obtained near carefully engineered metal optical
antenna structures that basically resemble scaled-down versions of a car antenna. Recently, such
antennas have even enabled single molecule studies by SERS and white-light super continuum
Despite the numerous studies on antennas in the microwave and optical regimes, their
application to solve current issues in chip-scale interconnection has remained largely unexplored.

The field concentrating abilities of optical antennas may serve to bridge the large gap between
microscale dielectric photonic devices and nanoscale electronics .This diagram shows a detail of
a chip on which optical signals are routed through conventional dielectric optical waveguides.
The mode size of such waveguides is typically one or two orders of magnitude larger than the
underlying CMOS electronics. An antenna can be used to concentrate the electromagnetic
signals from the waveguide mode into a deep sub wavelength metal/insulator/metal waveguide
and inject it into a nanoscale photo detector. The small size of the detector ensures a small
capacitance, low-noise, and high-speed operation. By using metallic nanostructures as a bridge
between photonics and electronics, we play to the strengths of metallic nanostructures
(concentrating fields and subwavelength guiding), dielectric waveguides (low-loss information
transport), and nanoscale electronic components (high-speed information processing).

Fig 3. Schematic of how a nanoscale antenna structure can serve as a bridge between microscale dielectric components and nanoscale
electronic devices.



Surface-plasmon polaritons (SPPs), light waves coupled to free-electron oscillations in metal ,
can be laterally confined below the diffraction limit using subwavelength metal structures .
Plasmonic components open an appealing perspective of combining the high operational
bandwidth of photonic components with the subwavelength dimensions of SPP waveguides .
Recently developed dielectric loaded SPP waveguides (DLSPPWs), utilizing high effective
indexes of SPP modes guided by dielectric ridges on smooth metal films , represent an attractive
alternative to other plasmonic technologies by virtue of being naturally compatible with different
dielectrics and large-scale industrial fabrication using UV lithography . Preliminary
investigations indicated that DLSPPW-based components feature relatively low bend and
propagation losses , but their potential for wavelength selection, a crucial functionality for any
photonic circuit, has so far not been explored.

Wavelength-selective plasmonic components generally feature more complicated

geometry , thereby imposing additional requirements to the accuracy of fabrication. In our case,
one should also bear in mind a (diffraction) limited resolution of UV lithography employed for
the DLSPPW fabrication. Taking into account a rather good performance of recently investigated
DLSPPW-based S bends ,we suggest making use of wavelength-dependent behavior of
directional couplers (DCs) [for the selection and spatial separation of radiation channels at
different wavelengths. Note that the wavelength dispersion of directional coupling using
DLSPPWs has not been considered in previous publications . Here we report on the fabrication,
characterization, and modeling of DLSPPW-based wavelength-selective DCs operating at
telecommunication wavelengths.

All waveguide structures were fabricated using deep UV lithography (wavelength of _250 nm)
with a Süss Microtech MJB4 mask aligner in the vacuum contact mode and a _550 nm thick
layer of polymethyl- methacrylate (PMMA) resist spin coated on a 60 nm thin gold film, which
was supported by a thin glass substrate. Typically, the width of the produced waveguides,
inspected with scanning electron microscopy (SEM), was close to 500 nm ensuring the
singlemode (and close to optimum) DLSPPW operation . The performance of the fabricated
components was characterized using a scanning near-field optical microscope (SNOM),
operating in collection mode with an uncoated fiber tip used as a probe, and an arrangement for

SPP excitation (at _=1500–1620 nm) in the Kretschmann–Raether configuration, as described in
detail elsewhere . All waveguide structures were connected to funnel structures facilitating
efficient excitation of the DLSPPW mode , with the further improvement in that the DLSPPW
mode was excited directly inside the taper by matching the excitation angle (under total internal
reflection), resulting in SNOM images of high quality . We found that all structures exhibited the
DLSPPW propagation length of _50 _m, slightly increasing with the wavelength .

Several DLSPPW-based 45-_m-long DCs, comprising 3 _m offset, 10-_m-long (input

and output) Sbends, having different center-to-center separations, S=800, 900, and 1000 nm,
between 25-_m-long parallel waveguides were fabricated and characterized with the SNOM at
different wavelengths
It was observed that the coupling length lc (i.e., the interaction length needed to
completely transfer the power from one waveguide to another) depends strongly on the
separation Figs. 4(c) and 4(e)and varies noticeably with the wavelength, influencing the power
level in the output waveguides Figs. 4(c) and 4(d). At the same time, the total DC transmission
evaluated using the input and output waveguide cross sections (separated by L=45 _m) is
practically wavelength independent maintaining the level of _0.23, a value that is consistent with
the loss incurred by 25-_m-long propagation __40%_ and two S bends (_35% per bend) .One can
thereby conclude that the performance of the fabricated DCs is not degraded owing to fabrication
errors, an important circumstance that is achieved because the DC design does not contain
critical elements requiring high resolution lithography.

Fig. 4. (Color online) (a) Scanning electron microscope image of the fabricated DC showing the funnel structure facilitating
the DLSPPW excitation. (b) Topographical and (c)–(e) near-field optical [_=_c_ 1500, (d) 1620, and (e) 1500 nm] SNOM
images of 45 _m long DCs with the separations (b)–(d) S=1000 nm along with an inset showing
SEM image of the coupling region and (e) S=900 nm.

To get further insight into the DLSPPW-based DC operation, we conducted full three-
dimensional finite-element method (3D-FEM) simulations of the DC operation [6] that allowed
us to retrieve the coupling length dispersion. It is seen that the simulation results agree well with
the experimental values obtained directly from the SNOM images (Fig. 2). This dispersion was
used to calculate the normalized output signals, i.e., straight Ts and cross Tc transmissions, with
only one waveguide being excited at the input, where the first factor reflects the S-bend
transmission, the second one reflects the power loss incurred by the propagation through the

section of parallel waveguides (of length Lp), and Li denotes the effective interaction length;
typically, Li_Lp due to the additional mode coupling in the S bends . Using
the already known parameters, Tbend_0.65 (evaluated using the available experimental data ),
Lp =25 _m and lc___ for S=1000 nm (Fig. 5), we found that the interaction length Li_34 _m
results in good agreement (Fig. 6)between the calculated output signals and the values obtained
from SNOM images, similar to those shown in Fig. 4. Using the obtained results one can design
a DC structure that would ensure spatial separation of radiation corresponding to different bands
used in optical telecommunications. For example, an increase of the length of parallel section by
_5.5 _m for the DC with S=1000 nm would enable spatial separation of the wavelengths of 1400
and 1620 nm (belonging to E and L bands) as seen from the calculations with Lp=30.5 _m and
Li_39.5 _m (Fig. 6).

Fig. 5. (Color online) DC coupling length evaluated for different wavelengths and separations directly
from SNOM
images similar to those shown in Figs. 1(c)–1(e) and calculated with 3D-FEM simulations.

we have considered the usage of DC wavelength dispersion for the selection and spatial
separation of radiation channels at different wavelengths.Using industrially compatible large-
scale UV-lithography-based fabrication and exploiting the principles of DLSPPW-based
plasmonic technology, we have fabricated and characterized wavelengthselective DCs operating
at telecommunications wavelengths. We have demonstrated a DLSPPW-based 45-_m-long DC
comprising 3 _m offset S bends and 25-_m-long parallel waveguides that changes from the
through state at 1500 nm to 3 dB splitting at 1600 nm and have shown that a 50.5-_m-long DC
should enable complete separation of the radiation channels at 1400 and 1620 nm (belonging to
E and L bands). The performance of the considered structures could be further improved by fine
tuning the DC parameters so as to achieve overall optimization (e.g., with the 3D-FEM
simulations) with respect to the wavelength selection and insertion loss. Taking into account the
fact that this technology is naturally compatible with different dielectrics, one can envisage the
development of ultracompact plasmonic components utilizing thermo-optical, electro-optical,
magneto-optical, acousto-optical, and nonlinear optical effects as well as being integrated with
electrical circuits.

Fig. 6. (Color online) DC

output signals
normalized with respect
to the input when exciting
only one waveguide,

experimentally from SNOM images similar to those shown in Figs. 4(c)–4(e) and the DC coupling
length dispersion shown inFig. 6.


1. Wide variety of Bio-technology applications, which include:

. Medical Diagnostics.
. Environmental Sensing.

. Ailmentary Emergency and Hygiene.

2. Less time consuming and cost effective, in comparison to many other similar applications.

3. Industrial Process Control.

4. Plasmon’s can support much higher frequencies into the 100 THz range
and it support
extremely small wavelengths.

Plasmonics has the potential to play a unique and important role in enhancing the
processing speed of future integrated circuits. The field has witnessed an explosive growth over
the last few years and our knowledge base in plasmonics is rapidly expanding. As a result, the
role of plasmonic devices on a chip is also becoming more well-defined and
is captured in Fig. 7. This graph shows the operating speeds and critical dimensions of different
chip-scale device technologies. In the past, devices were relatively slow and bulky. The
semiconductor industry has performed an incredible job in scaling electronic devices to
nanoscale dimensions. Unfortunately, interconnect delay time issues provide significant
challenges toward the realization of purely electronic circuits operating above ~10 GHz. In stark
contrast, photonic devices possess an enormous data-carrying capacity (bandwidth).
Unfortunately, dielectric photonic components are limited in their size by the laws of diffraction,
preventing the same scaling as in electronics. Finally, plasmonics offers precisely what
electronics and photonics do not have: the size of electronics and the speed of photonics.
Plasmonic devices, therefore, might interface naturally with similar speed photonic devices and
similar size electronic components. For these reasons, Plasmonics may well serve as the missing
link between the two device technologies that currently have a difficult time communicating. By
increasing the synergy between these technologies, plasmonics may be able to unleash the full
potential of nanoscale functionality and become the next wave of chip-scale technology.

Fig. 7 Operating speeds and critical dimensions of various chip-scale device technologies, highlighting the strengths of the different


1. W. L. Barnes, A. Dereux, and T. W. Ebbesen, Nature 424, 824 (2003).

2. T. W. Ebbesen, C. Genet, and S. I. Bozhevolnyi, Phys.Today 61, 44 (May 2008).

3. C. Reinhardt, S. Passinger, B. N. Chichkov, C.Marquart, I. P. Radko, and S. I. Bozhevolnyi,
Opt. Lett.31, 1307 (2006).

4. B. Steinberger, A. Hohenau, H. Ditlbacher, A. L.Stepanov, A. Drezet, F. R. Aussenegg,

A.Leitner, and J. R. Krenn, Appl. Phys. Lett. 88, 094104 (2006).

5. T. Holmgaard and S. I. Bozhevolnyi, Phys. Rev. B 75, 245405 (2007).

6. A. V. Krasavin and A. V. Zayats, Appl. Phys. Lett. 90, 211101 (2007).

7. B. Steinberger, A. Hohenau, H. Ditlbacher, F. R. Aussenegg, A. Leitner, and J. R. Krenn,

Appl. Phys. Lett. 91, 081111 (2007).

8. T. Holmgaard, S. I. Bozhevolnyi, L. Markey, and A. Dereux, Appl. Phys. Lett. 92, 011124

9. T. Holmgaard, Z. Chen, S. I. Bozhevolnyi, L. Markey, A. Dereux, A. V. Krasavin, and A. V.

Zayats, Opt. Express 16, 13585 (2008).