COMMENTARY

What diffraction limit?

Nikolay I. Zheludev
is at the Optoelectronics Research Centre, University of Southampton, Highfield, Southampton SO17 1BJ, UK.
e-mail: [email protected]

Several approaches are capable of beating the classical ‘diffraction limit’. In the optical domain,
not only are superlenses a promising choice: concepts such as super-oscillations could provide
feasible alternatives.

R
esearch on artificial photonic Modern microscopes push the boundaries near-field optical resolution for the first
materials engineered on the time. Here, an image of a structure is
subwavelength scale was The wide interest in super-resolution created by scanning a physical probe
stimulated a few years ago by was not in itself a surprise, because with a subwavelength aperture in close
the intriguing opportunity to the increasing finesse of observational proximity to the illuminated specimen. In
develop media that refract light in the instruments has for many centuries fact, the stethoscope, a near-field acoustic
opposite direction to normal media, and been one of the main engines of science imager, has been an indispensable tool of
that therefore were branded ‘negative- and technology. The invention of the medical practitioners since its invention
index materials’. The term ‘metamaterial’ compound optical microscope by Hans in 1816 by René Laennec. It also seems
quickly came into widespread use to and Zacharias Janssen in 1590 and that high-resolution near-field scanning-
describe this general class of artificial its improvement by Robert Hooke, aperture optical imaging is a much
media. Today its meaning encompasses Anthony van Leeuwenhoek and earlier invention by E. H. Synge, who
not only negative-index materials1,2 but Ernst Karl Abbe revolutionized all aspects described a method “which makes the
also manmade media with all sorts of of science, especially biology, when it attainment of a resolution of 0.01µ and
unusual functionalities that can be achieved became possible to see bacteria and blood even beyond” possible. He published the
by artificial structuring smaller than the cells, for instance. Since then the Nobel idea in 1928 (ref. 14), being encouraged by
length scale of the external stimulus (such Prize has been awarded several times Albert Einstein. Suggested implementation
as the optical wavelength). This includes for the development of new imaging involved “a miniature aperture, whose
metamaterials with exceptionally high3 techniques: in 1903 to Richard Zsigmondy diameter is approximately 10−6 cm, …
or zero4 refractive indices and high- for the slit optical ultramicroscope and the that is illuminated intensely from below,
permeability materials, or optical-frequency study of colloids, in 1953 to Frits Zernike and is placed immediately beneath” the
‘superconductors’ that repel the magnetic for the phase-contrast biological imaged sample so the transmitted light
field of optical waves in the same way microscope; and in 1986 to Ernst Ruska is than detected by a photoelectric cell
as conventional superconductors repel for the electron microscope and to and registered by a telephotographic
static magnetic fields5. Metamaterials have Gerd Binnig and Heinrich Rohrer for the apparatus, synchronized with the motion
proved to show fairy-tale properties such as scanning tunnelling microscope. of the aperture. Recently, the wide
‘invisibility’ (as introduced by H. G. Wells application of lasers has also led to the
in The Invisible Man)6,7, the ability to The increasing finesse of development of several high-resolution
‘hide’ objects in the miraculous manner nonlinear optical techniques such as two-
of Harry Potter’s cloak8,9, and puzzling observational instruments has photon luminescence microscopy and
transmission properties rivalling that of for many centuries been one stimulated emission depletion microscopes.
Lewis Carroll’s looking-glass10. They can of the main engines of science Unfortunately these optical microscopy
act as science-fiction-like electromagnetic techniques work only with narrow classes
force shields, as recently proposed by and technology. of specimens.
Graeme Milton, and in flying ‘magic carpets’ Indeed, it has not been possible so
with the use of quantum levitation11. Recent advances in high-resolution far to look inside a living cell, small
However, it was the incredible promise of a X-ray microscopy with sub-15-nm biological objects or other specimens
Veselago–Pendry optical negative-refraction spatial resolution were underpinned non-destructively with subwavelength
superlens12, capable of resolving features by the development of a sophisticated resolution by using low-intensity light
beyond the wavelength limit, and possibly zone plate for X-ray applications13. In and without dependence on specific
even revealing the structure of individual the domain of optical instruments, a key molecular absorption resonances. Electron
molecules, that was the impetus to mobilize step was the invention in 1984 of the microscopy techniques can provide such
the best research laboratories in the world scanning near-field optical microscope resolution, but living cells cannot survive
to work at the interface between ‘nano’ and (SNOM) by Dieter Pohl, Aaron Lewis the required vacuum, exposure to intense
‘meta’ photonics. and co-workers, allowing subwavelength electron beams or the often necessary

420 nature materials | VOL 7 | JUNE 2008 | www.nature.com/naturematerials

© 2008 Nature Publishing Group
 COMMENTARY
sample metallization. High-resolution Highest harmonic
tunnelling and optical scanning 1.0
microscopes are not capable of seeing
internal sections of a living object — or
indeed any object — without destroying it:

f (x)
0.5
their operation depends on the presence of

© 2007 I0P
probes a few nanometres from the feature Superoscillating
being imaged. Optical subwavelength 0
function
resolution is not possible for objects more –0.02 0 0.02
x
than about a fraction of a wavelength away
from the probe of a scanning instrument.
This is why so much interest is being shown Hot
spot
in the development of new concepts of
optical super-resolution in which it will
be possible to image an object located Nano-hole array
at a somewhat remote distance from the
imaging instrument, perhaps even a few

© 2007 I0P
tens of micrometres away.

The superlens

The remarkable superlens proposed Figure 1 Focusing light with a nanohole array. a, An array of nano-holes in a screen as a generator of a
by John Pendry and Victor Veselago superoscillating field. It can create a subwavelength hot-spot when illuminated by a plane monochromatic
is exactly this kind of device allowing wave. b, A function (equation (1)) superoscillating at x = 0. c, An example of a ‘photonic carpet’ generated by
super-resolution imaging of a distance quasi-periodic array of holes when subwavelength superoscillating hot spots were observed. Parts b and c
object into a far-field image. The reprinted with permission from ref. 26.
superlens is based on the recovery of the
quickly fading evanescent fields close
to the object by amplifying them in a
slab of a negative-index material. Such resolution in the optical part of the far-field16. Although the ‘NIM detour’
evanescent, non-propagating fields are spectrum by the groups of Richard Blaikie16 superlens has demonstrated the unique
commonly believed to be the necessary and Xiang Zhang17. However, this ability of overcoming the ‘diffraction limit’
components for forming subwavelength remarkable achievement had limited the main limitation of all such designs is
field concentrations and to achieve practical importance: the object and the that the object still has to be in the near-
subwavelength resolution. Indeed, it is image would have to be restrictively close, field of the superlens.
accepted by the photonics community that within nanometres, to the metal film. This
the resolving power of optical instruments is why much effort is being concentrated Lessons from microwaves
imaging objects located in the far-field, on the development of proper negative-
where evanescent waves have faded, cannot index materials — ones that show negative However, there is a solution that can
be far from that given by the well-known values of both ε and µ. Despite recent provide subwavelength concentrations
Abbe–Rayleigh rule, according to which numerous successful demonstrations of of light beyond the near-field. As noted
the smallest distance between two points such double-negative optical materials, by G. Toraldo di Francia as early as 1952,
that can be distinguished with a lens is there is still substantial scepticism that “fortunately it appears that microwave
about the wavelength of light. such materials can, in the near future, be researchers were not very much
A review by Xiang Zhang and developed for use in the manufacturing of concerned, or perhaps even acquainted,
Zhaowei Liu in this issue15 gives a practical superlenses. The resonance nature with the old well-established theorems
representative and captivating account of the negative index that is coupled to the of wave optics [on the Abbe/Rayleigh
of the current status of optical superlens problem of losses inheritably limiting the resolution criterion] … As a result, an
research, from the negative-index optical bandwidth and transmission of the entirely new theory has been set up,
superlens and beyond. The bulk negative- superlens is the main fundamental obstacle. which contains many revolutionary
index material required for the superlens For this reason, many researchers are now implications”18. For several decades the
should simultaneously exhibit negative seeking alternatives to the negative-index microwave community contemplated the
permeability µ and a negative electric superlens. For instance, the group of idea of constructing antennas that beat
permittivity ε. Apparently, achieving super- Xiang Zhang came up with an ingenious the diffraction limit for directivity. In fact
resolution when the object and its image idea of decoupling evanescent waves on the as early in 1922, Oseen, with reference
are placed in the subwavelength vicinity image side of the “poor-man’s superlens” to Einstein’s radiation ‘needle stick’,
of the ‘lens’ is a much simpler task than with a grating to achieve a near-field to far- proved that an arbitrarily large fraction
imaging a remote object. In this case only field imaging device16. Another detour from of the emitted energy can be sent into an
one material property (ε or µ) of the ‘lens’ the use of bulk negative-index materials arbitrarily small solid angle19.
needs to be negative, an eased constraint is to use anisotropic materials with In 1943, S. A. Shelkunoff published
that led John Pendry to brand this ‘lens’ hyperbolic dispersion: when evanescent an analysis of the radiation pattern of a
a “poor-man’s superlens”. It was the silver waves enter such anisotropic media, their linear array of dipoles and proved that, by
nanolayer-film “poor-man’s superlens” wavevectors are gradually compressed properly adjusting the individual radiating
that was used in the independent until they become propagating waves that elements, it is possible to achieve a much
demonstrations of subwavelength could project a magnified image into the narrower radiation pattern than that of a

nature materials | VOL 7 | JUNE 2008 | www.nature.com/naturematerials 421

© 2008 Nature Publishing Group
Commentary
conventionally uniform array20. Soon after superoscillating function. It is plotted in metal film25. When illuminated with a
that Bouwkamp and Bruijn21, and then Fig. 1b (solid curve) alongside the highest- coherent light source it creates a complex
Woodward and Lawson22, were able to frequency component (dashed curve). diffraction pattern on the other side
prove that there were no theoretical limits At x = 0 the function has a feature that of array — a few tens of micrometres
to directivity whatsoever. oscillates nearly nine times faster than its away. At certain distances these patterns
However, in spite of several cunning highest-frequency component. show well-defined, sparsely distributed
suggestions for super-gain antenna subwavelength light localizations.
designs, the initial enthusiasm for the One day, thanks to Moreover, because such subwavelength
development of extremely narrow-beam localizations are formed by propagating
antennas was soon replaced by sober nanotechnology, a schoolboy far-fields, they can be projected to the
scepticism. The sceptics argued that the will be able to screw a nano- far-field by a conventional lens26 or used
extreme sensitivity of the super-directive as a subwavelength source in a scanning
illumination function to changes in the
array lens to his science imaging device for imaging an object
array design and feed characteristics class microscope and see located far beyond the near-field area. The
would make the necessary manufacturing a DNA molecule. question now is whether such a pattern,
tolerance difficult to achieve and would or for that matter any superoscillating
drastically limit the antenna’s bandwidth. Aside from the fact that the super-gain grating-type field generator, could be used
Moreover, the sharp increase in the antenna aims at creating a narrow beam as a proper far-field to far-field super-
proportion of reactive to radiated power of electromagnetic radiation while the resolution lens and whether it can achieve a
that would be required to achieve super- superoscillation generator seeks to achieve subwavelength resolution.
directivity could mean that the gain subwavelength localization of light at In conclusion, as is often the case,
improvement might well be offset by the a distance from the grating, both ideas science develops in circles: what the
need to provide an even higher increase have the same underlying physics: the microwave community has understood
in power to the antenna to maintain tailored interference of several coherent about loopholes in wave theory but failed
the signal level, thus rendering the sources (see Fig. 1a,c). However, the task of to apply in practical super-directive
concept impracticable. designing superoscillation in optics could antenna designs, could well be a sensible
be a much easier problem than designing proposition for optical microscopy: one
Superoscillations a super-gain microwave antenna. An array day, thanks to nanotechnology, a schoolboy
of nano-holes may be used in such a way will be able to screw a nano-array lens
However, the idea of achieving super- that superoscillation is achieved a few tens to his science class microscope and see a
resolution without evanescent fields had of micrometres away from it by the tailored DNA molecule.
an independent revival recently in the interference of light penetrating through
domain of optics: Berry and Popescu, the holes. References
starting from earlier studies on quantum Indeed, microscopy can tolerate 1. Veselago, V. G. Sov. Phys. Usp. 10, 509–514 (1968).
2. Smith, D. R. et al. Phys. Rev. Lett. 84, 4184–4187 (2000).
mechanics, predicted that diffraction much higher losses than communications 3. Riikonen, S., Romero, I. & García de Abajo, F. J. Phys. Rev. B
on a grating structure could create applications can. If the photon throughput 71, 235104 (2005).
subwavelength localizations of light inefficiency of the system is the price to 4. Alù, A., Silveirinha, M. G., Salandrino, A. & Engheta, N.
that propagate farther into the far field pay for improved resolution, one can Phys. Rev. B 75, 155410 (2007).
5. Schwanecke, A. S. et al. J. Opt. A 9, L1–L2 (2007).
than more familiar evanescent waves23 reasonably work with only a few detected 6. Alù, A. & Engheta, N. Phys. Rev. E 72, 016623 (2005).
(Fig. 1). They related this effect to the photons per second, giving about 19 orders 7. Fedotov, V. A., Mladyonov, P. L., Prosvirnin, S. L. &
fact that band-limited functions are able of magnitude to play with (a 1-watt laser Zheludev, N. I. Phys. Rev. E 72, 056613 (2005).
to oscillate arbitrarily faster than the generates about 1019 photons per second). 8. Leonhardt, U. Science 312, 1777–1780 (2006).
9. Schurig, D. et al. Science 314, 977–980 (2006).
highest Fourier components that they Such a power reserve will in fact be needed 10. Fedotov, V. A. et al. Phys. Rev. Lett. 97, 167401 (2006).
contain, a phenomenon now known as because the cost of a decrease in the 11. Leonhardt, U. & Philbin, T. G. New J. Phys. 9, 254 (2007).
superoscillation24. The superoscillation hot-spot size is a polynomial increase in 12. Pendry, J. B. Phys. Rev. Lett. 85, 3966–3969 (2000).
13. Chao, W., Harteneck, B. D., Liddle, J. A., Anderson, E. H. &
idea challenges the well-established belief the power going into the sidebands. The Attwood, D. T. Nature 435, 1210–1213 (2005).
that a function whose Fourier spectrum relative intensity and phase stability of 14. Synge, E. H. Phil Mag. 6, 356–362 (1928).
is bounded can vary no faster than its emitters coherently excited by one light 15. Zhang, X. & Liu, Z. Nature Mater. 7, 435–441 (2008).
highest-frequency component. source are easy to maintain in the optical 16. Blaikie, R. J. & Melville, D. O. S. J. Opt. A
7, S176–S183 (2005).
This astonishing claim is clearly system. The only serious barrier to the 17. Fang, N. et al. Science 308, 534–537 (2005).
counterintuitive to many and goes against development of optical superoscillation 18. Toraldo di Francia, G. Suppl. Nuovo Cim. 9, 426–438 (1952).
all common experience. However, many generators is manufacturing accuracy. 19. Oseen, C. W. Annln Phys. (Leipz.) 374, 202–204 (1922).
examples of simple superoscillating Today, a hole in a thin metal film on a silica 20. Schelkunoff, S. A. Bell Syst. Tech. J. 22, 80–107 (1943).
21. Bouwkamp, C. J. & de Bruijn, N. G. Philips Res. Rep.
functions have been identified. For or silicon substrate can be fabricated with 1, 135–158 (1945/1946).
instance, a limited series with five a positional accuracy of a few nanometres, 22. Woodward, P. M. & Lawson, J. D. J. Inst. Elect. Eng. III
components only which should be sufficient to create optical 95, 363–370 (1948).
23. Berry, M. V. & Popescu, S. J. Phys. A 39, 6965–6977 (2006).
superoscillation generators that beat 24. Ferreira, P. J. S. G. & Kempf, A. IEEE Trans. Signal Process.
f(x) = ∑an cos(2π × nx) (1) the diffraction limit several times over. 54, 3732–3740 (2006).
An improvement by a factor of only a 25. Huang, F. M., Zheludev, N., Chen, Y. & Garcia de Abajo, F. J.
can generate superoscillating functions fewfold would be a revolutionary step for Appl. Phys. Lett. 90, 091119 (2007).
26. Huang, F. M., Chen, Y., Garcia de Abajo, F. J. & Zheludev, N. I.
relevant to optical scattering and observational instrumentation. J. Opt. A 9, S285–S288 (2007).
microwave emission. Thus, f(x) with An optical generator of superoscillating
a0 = 1, a1 = 13295000, a2 = −30802818, fields has recently been demonstrated Acknowledgements
a3 = 26581909, a4 = −10836909, with the use of a Penrose type quasi- I thank Nader Engheta, Fumin Huang, Mark Dennis,
a5 = 1762818 and an = 0 is a crystal array of nano-holes in a thin Kishan Dholakia and Javier Garcia de Abajo for discussions.

422 nature materials | VOL 7 | JUNE 2008 | www.nature.com/naturematerials

© 2008 Nature Publishing Group