Documente Academic
Documente Profesional
Documente Cultură
The MacDiarmid Institute for Advanced Materials and Nanotechnology, Department of Electrical and Computer
Engineering, University of Canterbury, Christchurch, New Zealand
2
Department of Chemical Engineering, The University of Texas at Austin, Austin, Texas, USA
3
Current Address: Department of Physics, University of Otago, New Zealand
*richard.blaikie@otago.ac.nz
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13710
13. IBM, A Testbed for 193 nm Interferometric Immersion Lithography. [Online] Available:
http://www.almaden.ibm.com/st/chemistry/lithography/immersion/NEMO/ [3 May 2012].
14. C. H. Chang, The MIT Nanoruler: A Tool for Patterning Nano-Accurate Gratings. [Online] Available:
http://nanoweb.mit.edu/Annual%20Reports%202005/sec.10.ms.pdf [3 May 2012].
15. I. Wathuthanthri, K. Du, W. Xu, and C.-H. Choi, Simple Holographic Patterning for High-Aspect-Ratio ThreeDimensional Nanostructures with Large Coverage Area, Adv. Funct. Mater. 23(5), 608618 (2013).
16. C. W. Holzwarth, J. E. Foulkes, and R. J. Blaikie, Increased process latitude in absorbance-modulated
lithography via a plasmonic reflector, Opt. Express 19(18), 1779017798 (2011).
17. S. A. Maier, Plasmonics: Fundamentals and Applications (Springer Science + Business, 2007).
18. H. Raether, Surface plasmons on smooth and rough surfaces and on gratings (Springer-Verlag, 1988).
19. A. R. Zakharian, J. V. Moloney, and M. Mansuripur, Surface plasmon polaritons on metallic surfaces, Opt.
Express 15(1), 183197 (2007).
20. J. Lagois and B. Fischer, Experimental Observation of Surface Exciton Polaritons, Phys. Rev. Lett. 36(12),
680683 (1976).
21. F. Z. Yang, G. W. Bradberry, and J. R. Sambles, Long-Range Surface-Mode Supported by Very Thin Silver
Films, Phys. Rev. Lett. 66(15), 20302032 (1991).
22. F. Z. Yang, G. W. Bradberry, and J. R. Sambles, Experimental-Observation Of Surface Exciton-Polaritons On
Vanadium Using Infrared Radiation, J. Mod. Opt. 37(9), 15451553 (1990).
23. M. D. Arnold and R. J. Blaikie, Subwavelength optical imaging of evanescent fields using reflections from
plasmonic slabs, Opt. Express 15(18), 1154211552 (2007).
24. J. B. Pendry, Negative refraction makes a perfect lens, Phys. Rev. Lett. 85(18), 39663969 (2000).
25. R. J. Blaikie and D. O. S. Melville, Imaging through planar silver lenses in the optical near field, J. Opt. A
7(2), S176S183 (2005).
26. D. O. S. Melville and R. J. Blaikie, Super-resolution imaging through a planar silver layer, Opt. Express 13(6),
21272134 (2005).
27. D. O. S. Melville and R. J. Blaikie, Experimental comparison of resolution and pattern fidelity in single- and
double-layer planar lens lithography, J. Opt. Soc. Am. B 23(3), 461467 (2006).
28. N. Fang and X. Zhang, Imaging properties of a metamaterial superlens, Appl. Phys. Lett. 82(2), 161163
(2003).
29. N. Fang, H. Lee, C. Sun, and X. Zhang, Sub-diffraction-limited optical imaging with a silver superlens,
Science 308(5721), 534537 (2005).
30. P. Mehrotra, High Aspect Ratio Lithographic Imaging at Ultra-high Numerical Apertures: Evanescent
Interferometric Lithography with Resonant Reflector Layers, PhD thesis, (University of Canterbury,
Christchurch, New Zealand, 2012).
31. E. Hecht, Optics, 4th ed. (Addison Wesley, 2001).
1. Introduction
Advances in super-resolution optical imaging are being aggressively pursued for applications
in imaging, data storage, and lithography. In biological imaging, for example, recent advances
have seen reports of molecular-scale resolution using a range of nonlinear and stochastic
processes [1,2], and commercial instruments are now available. For other applicationssuch
as lithography imaging that is the subject of this paperanalogous techniques are not readily
applied and alternative means of achieving super-resolved imaging must be developed.
The semiconductor industry has historically reduced the feature sizes used in integrated
circuits by a factor of 2 every 4 years or so. The most important feature to reduce is the pitch
of a repeating pattern, since this determines the packing density of transistors in the circuit.
The resolution (R), defined as the minimum half-pitch of a pattern than can be imaged in a
single exposure in a particular optical lithography system, is limited by the ratio of the
wavelength () to the numerical aperture of the imaging tool (NA):
R k1
(1)
, k 1 0.25.
NA
where k1 is a process variable that takes into account factors relating to the particular
instrument configuration and imaging conditions [3,4].
To further improve resolution, one must lower the wavelength or increase NA. Today, a
wavelength of 193 nm is common for semiconductor lithography and lowering it has proven
difficult in manufacturing; for example, extreme ultraviolet (EUV) lithography, using 13.5
nm wavelength, is on the very near horizon for manufacturing but has taken considerable time
and cost to develop. Increasing NA is also difficult, being limited by the lowest refractive
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13711
index in the optical path on the wafer-side of the imaging tool. Today, water is used as the
immersion fluid, giving a maximum practical NA of ~1.35. Higher index materials are
employed in what is known as a solid-immersion system [510]. The limiting refractive index
is then that of the photoresist, constraining NA to ~1.6. Higher NAs are extremely attractive
and feasible but put the system into the evanescent regime. Unfortunately, evanescent images
decay exponentially into the resist, leading to images that are practically useful only for
extremely thin resists, making pattern transfer difficult.
Here we report new principles, designs, and experimental demonstration for a method
whereby coupling of the evanescent component to a resonant surface state on an underlying
medium can be used to significantly enhance the depth of focus of evanescent images. We
analyze and experimentally verify our new techniques using a simple interference lithography
(IL) arrangement, but the principles are adaptable to more-conventional lithography or other
ultra-high resolution imaging techniques.
The rest of the paper is structured as follows. In the next section we review conventional
IL techniques and introduce a solid-immersion Lloyds mirror interference lithography
(SILMIL) technique. We have already developed a SILMIL testbed for performing
evanescent interferometric lithography (EIL) and have reported its basic performance
characteristics elsewhere [11]. Following this we analyze the use of surface states on an
underlying interface to enhance the usable image depth of EIL, showing that there are three
distinct regimes worthy of exploration. Idealized and practical examples are then presented,
followed by experimental results demonstrating the technique for pattering 55-nm half-pitch
lines with 100 nm image depth using 405 nm wavelength light.
2. Interference Lithography in the Propagating and Evanescent Regimes
Interference lithography is a straightforward way to achieve the lowest possible value for the
resolution factor k1 (k1 = 0.25), with the simple principle to interfere two plane waves
travelling at the opposite ends of the available NA in a given optical system [3]. The aim of
research in IL is to be able to fabricate high resolution structures with high aspect ratios for
large area patterning; this is a well-developed field in which many groups have made their
contributions to fulfilling either one or more of these requirements in a variety of
configurations [1215].
Figure 1(a) illustrates a simple scheme for two-beam IL carried out in air upon photoresist
in which case the NA of the interference pattern is less than one; the resultant periodic pattern
has a period p = /2sin = /2NA (not super-resolved) and infinite image depth provided that
the resist is lossless. This illustration is for transverse electric (TE) polarization, as defined in
Fig. 1(a), which results in deep nulls in the interference pattern, but the use of transverse
magnetic (TM) polarization is also possible [16]. In either case the image depth will be
determined by the Lambert-law absorption decay length of the interference pattern in the
photoresist, which may be many microns or tens of microns.
We are interested in exploring the ultimate limits of resolution at a particular wavelength,
for which using liquid- or solid-immersion schemes as depicted in Figs. 1(b) and 1(c) are
advantageous. Figure 1(b) demonstrates what we call the high-NA regime in which the
numerical aperture exceeds that of air (NA > 1), but is less than the refractive index of the
photoresist (nr)in this case imaging is through propagating waves and image depth is also
unlimited as in the ideal case. The practical disadvantage is the need to use a prism with a
liquid immersion layer in intimate contact with the resist, or to use careful gap control (in the
nanometer range) in a solid-immersion system [5]. For the ultra-high NA scheme (NA greater
than the resist refractive index, but less than np, the prism index) of Fig. 1(c) resolution is
further improved, however now the fields in the resist are evanescent and image depth
becomes a severe practical constraint [5, 711].
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13712
Fig. 1. Schematic representations of (a) basic interference lithography carried out with air as
the ambient medium, (b) inside a prism at a high numerical aperture, and (c) in a prism at an
ultra-high numerical aperture (NA), with beyond the critical angle at the prism-resist
interface. These illustrations are for transverse electric (TE) polarisations, with the electric and
magnetic field directions defined in part (a). Electric and magnetic field orientations are
exchanged in the transverse magnetic (TM) configuration.
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13713
Fig. 3, which depicts how an evanescent field permeates the photoresist film (blue longdashed line) and how this field may then induce a reflected image (green dashed line) using
a carefully chosen substrate, so that the resultant image field (red solid line) has higher
intensity and less variation with depth compared to the original evanescently decaying image.
The advantage in enhancing the otherwise decaying field in such a way is the ability to
expose the photoresist to a greater depth and with a higher contrast and hence create taller or
higher aspect ratio structures, which are needed to allow pattern transfer in lithographic
applications.
Fig. 3. Schematic diagram showing field formation within resist and motivation to consider
evanescent field interaction with an underlying substrate.
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13714
TM polarization of light, this is achieved using metals or lossy dielectrics as the substrate
materials, which support SPP surface states. The case of a high-loss dielectric underlayer is
also interesting to consider for TM polarization, as it supports the so-called surface exciton
polaritons (SEP) surface state [2022]. Although this surface state has not received the same
attention as the SPP surface state for super-resolution imaging, we note this point here for
future considerationwe have undertaken preliminary studies on SEP-enhanced EIL, but will
report those elsewhere.
Evanescent-wave reflections from metal underlayers can provide strong field
enhancements, particularly close to surface-plasmon resonances, and hence allow the
evanescent decay of the original imaging field to be countered. An analytical case study is
presented here using a fictitious metal to illustrate the point. It is to be noted that silver has
been proposed as a reasonable candidate for such enhancement at 365 nm wavelength [11,
23]. In fact, the superlens suggested by John Pendry [24], and demonstrated experimentally
using thin silver films [2529], also utilizes this principle; enhancement is achieved via field
redistribution by coupling to surface states on both sides of the superlens allowing energy
extraction and redistribution from the source apertures to photoresist.
Two SPP-enhanced EIL scenarios are illustrated in Fig. 4, alongside the non-enhanced
intensity decay of a transverse-magnetic (TM) polarized evanescent-wave image [Fig. 4(a)].
Finite-element method (FEM) numerical simulations, using a fictitious metal underlayer
(permittivity = 29.8 + j0.0), is used to set the scene for the approach we present here.
Considering Fig. 4, the evanescent decay in the simple index-matched system of Fig. 4(a) is
seen to be overcome, or even overwhelmed, by resonant coupling to SPP states [Fig. 4(b) and
4(c)]. The difference between Figs. 4(b) and 4(c) is a small change in the NA of the imaging
condition (1.79 compared with 1.85 respectively) illustrating how sensitive this resonant
phenomenon is to NA, and how this tuning can be used effectively to tailor the final field
profile in the imaging (resist) layer.
Fig. 4. Evanescent image (TM light) in 82.5-nm thick resist, = 193 nm with (a) resist
underlayer, fictious metal ( = 29.8) underlayer for (b) optimal off-resonant enhancement
(NA = 1.85), and (c) non-optimal resonant enhancement (NA = 1.79).
The underlying principle is energy extraction which would otherwise totally internally
reflect back into the prism and its redistribution in the resist cavity. The principle is evident
by reference to the attenuated total reflectance (ATR) spectra of Fig. 5, shown for the
fictitious metal case of Fig. 4; setting the operating NA around the resonant dip in the ATR
spectrum (NA = 1.79) allows fine control of the field profiles to produce a symmetrical
intensity distribution [Fig. 4(b)] at NA = 1.85.
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13715
Lossy metals,
Lossless metals
Lossy dielectrics
Gain medium
Gain medium
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13716
The high index on low index stack is in fact an EGM provided we concern ourselves with
only reflection at the particular NA. The resonant spatial frequency here is the effective
surface state. It should be noted that while the EGM presented here mimics the workings of a
waveguide, other EGMs may also be created for the TM polarization of light using several
layers of metals, lossy dielectrics and dielectrics. Such a setup would then support a
composite surface state that would create an effective gain medium for the purposes of an
enhanced reflection.
7. Design and implementation
Proposed design for industry standard = 193 nm
Figure 7(a) illustrates imaging of 26-nm half-pitch features where the image depth is only 20
nm into thick photoresist (n = 1.7 + i0.02). Figure 7(b) illustrates the significant enhancement
when a tuned stack of 43 nm layer of Al2O3 (sapphire) with index n3 = 2.08 [11] upon 50nm
SiO2 with index n4 = 1.56 is used. The resulting image is 82.5 nm deep. We have also
included photoresist loss in this design and with this the new reflection from the real-world
stack is calculated as 6.77ei0.212 instead of the optimum 7.39. This design, however, still
allows ultra-high NA patterning at an aspect ratio of ~3.2. For the purposes of evanescent
wave coupling at an ultra-high NA of 1.85, the stack behaves the same way as a hypothetical
gain medium. The ATR spectrum of the system is shown in Fig. 8, together with the ATR
spectrum for the fictitious-metal case of Fig. 5. As well as being a much sharper resonance,
the ATR spectrum for the EGM case has a resonant dip at a spatial frequency corresponding
to a NA of ~1.83 indicative of the off-resonance operation mentioned earlier.
Implementation at = 405 nm
An experimental demonstration of EIL image enhancement effect is presented in this section,
using a 405 nm wavelength that will be more readily accessible in research laboratories. A
solid immersion Lloyds mirror interference lithography (SILMIL) [11] system was used with
a Yttrium Aluminium Garnet (YAG) prism at = 405 nm and a NA = 1.824. Since the index
of the photoresist at this wavelength is n = 1.68 + j0.031, the NA is high enough to create
evanescent images.
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13717
Fig. 7. TE imaging of 26-nm (half-pitch) evanescent features into (a) semi-infinite lossy resist
giving 20-nm image depth, and (b) 82.5 nm thick lossy resist on an effective gain medium
made up of 43 nm of Al2O3 (Sapphire) on SiO2, giving an image depth of 82.5 nm.
Two experiments are compared, with the imaging layers for each one shown in Fig. 9. The
first experiment involves imaging into a single 450-nm thick layer of resist [Fig. 9(a)], which
acts as a control. The thickness of the resist in the control sample was large enough to prevent
reflections of the evanescent wave. The results are then compared with an optical stack that
uses hafnium oxide (HfO) as the high index dielectric stacked upon silicon dioxide (SiO 2) as
the low index dielectric [Fig. 9(b)]. The HfO upon SiO2 stack is carefully designed to result in
an enhanced reflectivity and hence serves as our EGM. Both imaging stacks were created on
quartz substrates to ensure that the mechanical properties of the optical stacks were the same,
as the stacks must deform similarly to the mechanical stress that is applied within the SILMIL
system.
Fig. 9. Optical stacks for imaging at = 405 nm and NA = 1.824, including a water-soluble
poly vinyl acetate (PVA) layer that is used as a barrier between the index matching liquid and
the photoresist. (a) Without an EGM underlayer for conventional EIL, and (b) with an EGM
underlayer (characterized by ATR earlier) for high aspect ratio imaging.
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13718
exposure time in seconds, resulting in a PD measured in mJ/cm2. This PD does not take
account reflection losses at prism-air and prims-resist interfaces, however it does allow
images to be compared between control and reflection-enhanced scenarios (Fig. 9) for
identical exposure conditions.
An example scanning electron microscope (SEM) scan of a control image exposed at NA
= 1.824 and a PD = 257 mJ/cm2 is presented in Fig. 10, showing that deep sub-wavelength
resolution can be achieved using the SILMIL solid-immersion scheme in the ultra-high NA
regime. The 55-nm half-pitch spaces, compared to the 405-nm exposure wavelength,
corresponds to better than /7 resolution.
Fig. 10. SEM plan views, at pseudo-dosage of 257 mJ/cm2, NA = 1.824, = 405 nm,
conventional EIL using SILMIL, resulting pitch ~111 nm (55.5 nm half-pitch)
Representational cross-sections through this image are shown in Fig. 11 illustrating the
shallow nature of such non-enhanced images, of 30-40 nm, as has previously been reported
[11]. Control images were obtained over a wide range of exposure doses, with PD varied
from 0 to 1540 mJ/cm2 in 86 mJ/cm2 steps, and the images presented here are the cleanest and
deepest that were obtained. For lower PD the image depth was shallower than that of Fig. 11,
and for higher doses image fidelity was lost (with no increase in image depth) due to
unintended exposure in the imperfect nulls in the interference patterns. The limits that the
evanescent nature of the imaging imposes on the maximum aspect ratio of the final resist
lines (0.72 height-width ratio in this case) presents severe limitations for subsequent use of
these lines for pattern transferaspect ratios of greater than unity are normally required for
this.
Fig. 11. Cross-sectional views, at NA = 1.824, = 405 nm, conventional EIL using SILMIL,
resulting pitch ~111 nm (55.5 nm half-pitch) at pseudo-dosage (PD) of 257 mJ/cm2, giving 3040 nm image depths. The (a) and (b) SEM scans are at different positions on the sample.
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13719
Fig. 12. Finitie-element simulation of imaging with the stacks in Fig. 9 for NA = 1.824, =
405 nm, into (a) semi-infinite resist, and (b) 105 nm resist with an EGM underlayer.
It is clear from Fig. 12(b) that the EGM is not perfect, as the image intensity is not
symmetrically distributed vertically through the resist layer. However it suffices to
demonstrate proof of concept as we see from the experimental AFM and SEM scans that are
depicted below. Figure 13 depicts AFM scans of structures that have been created using the
imaging stack in Fig. 9(b) at a pseudo-dosage of 214 mJ/cm2. These scans demonstrate a
considerably large image depth of 90 - 105 nm.
Fig. 13. AFM scan demonstrating EIL with SILMIL using the imaging stack of Fig. 9 (b),
high aspect ratio (~1.8) structure imaged at a NA of 1.824 at = 405 nm with pattern half-pitch
~55.5 nm. The average depth measured using AFM software was 100 nm. (a) a 2 m by 1 m
AFM scan, and (b) a corresponding 2D-like perspective view to depict the tall standing
structures.
To compare with the control samples (Fig. 11) cross-sectional SEM images are presented
in Fig. 14, again showing image depth of more than 90 nm compared with the weak, shallow
images from the control samples. The underlying HfO layer can be seen in these SEM images
(marked in Fig. 14), and it is noted that full clearance of the resist down to the underlying
substrate has not been achieved for all the lines presented in this image. Difficulties with
sample cleavage and charging on the non-crystalline, insulating substrates used for these
studies make this cross-sectional SEM imaging challenging, but it is acknowledged that this
imperfect development may be a result of imperfect coupling into the HfO resonant
underlayer in this part of the sample. Further optimization of the imaging stack will be
required to improve image quality and overcome this issue.
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13720
Fig. 14. SEM cross-sectional views showing tall standing structures, at PD of 214 mJ/cm2, NA
= 1.824, = 405 nm, EIL using SILMIL using the imaging stack of Fig. 9 (b), resulting halfpitch ~55.5 nm, with pattern depth ~96 nm .
There is clear indication that the use of an EGM has made a significant difference. This is
due to energy extraction and redistribution right down to the bottom of the resist which was
facilitated through near-field interaction with the EGM, i.e., evanescent coupling of the image
with the EGM.
Dose variation and resist collapse proof of high aspect ratio imaging
As with the control images, EGM-enhanced image exposures were performed over a range of
pseudo dosages (PD = 86-684 mJ/cm2 in 86 mJ/cm2 steps). Image depth, as measured from
AFM scans, increased linearly for PD from 86 to 257 mJ/cm2, at which point full
development through the imaging resist layer was achieved. For higher doses the resist depth
decreased due to background development in the imperfect interference nulls, accompanied
by the appearance of a phenomenon previously unreported for evanescent-field imaging
resist pattern collapse. This is illustrated in Fig. 15, which shows AFM scans for samples
exposed at PD = 342 mJ/cm2, where there is clear indication of resist collapse in some areas
of the sample.
Resist collapse is a side-effect of over-exposure/development, which is generally
undesirable. Nevertheless it is very interesting to report here, as resist collapse requires a very
high aspect ratio (typically greater than 1.5:1) and hence such collapse with EIL has never
been thought possible earlier. The AFM scans in Fig. 13 demonstrate aspect ratios as high as
1.8:1, where better process control has resulted in lines that have not collapsed.
It is necessary to also point out here that resist collapse is also due to the amplification of
the imperfect minima (nulls) in the interference pattern. Conventionally with interference
lithography, imperfect nulls only contribute to the background exposure; however, as the
effective reflection coefficient from the EGM amplifies/enhances all signals (even the
imperfect minima of the interference pattern), and redistributes energy to the bottom of the
resist, collapse takes places. For future work, and for application of the method for
nanofabrication, resist collapse should be avoidedthis can be achieved by careful control of
the interfering beams to improve the nulls in the interference pattern, together with good
process control (exposure dose, development time and low surface-tension solvent
evaporation).
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13721
Fig. 15. AFM scans demonstrating resist collapse using the imaging stack of Fig. 9(b), for an
exposure at PD = 342 mJ/cm2 at a NA of 1.824 and = 405 nm with pattern half-pitch ~55 nm.
(a) A 5 m by 5 m AFM scan showing some resist collapse, and (b) a large area scan cropped
to 5 m by 2.5 m demonstrating greater resist collapse, a result of overdosage/exposure/development.
8. Conclusion
Ultra-high NA patterning has long been a critical outstanding problem due to the physical
limits presented by evanescent fields. We have suggested coupling of the evanescent fields to
surface states as a means to control the field intensity and profile in a photoresist cavity.
Through the use of the Fresnel reflection equation for TE and analysis of fields, we have
presented a practical way to couple the evanescent field into a waveguide-like underlayer to
image super-resolved structures with high aspect ratios. We show how a high aspect ratio
may be achieved at NA = 1.85 and = 193 nm in photoresist by using a layer of sapphire
deposited on SiO2. The method encourages use of solid-immersion systems and paves the
way for feasible ultra-high-NA imaging and patterning.
An experimental demonstration of the method at = 405 nm shows imaging at an ultrahigh NA of 1.824. While the imaging depth for a thick resist was measured to be only ~30-40
nm maximum for a set dosage, the use of a hafnium oxide on silicon dioxide stack served to
increase this to the range of 90-105 nm. Careful monitoring of the refractive index and
thickness of the hafnium oxide film is expected to result in a better design that is likely to
give image depths of ~160 nm under the same circumstances.
Such results are the first of their kind to be reported, and with the use of a more robust dry
or immersion interference lithography system such as the Amphibian TM [10] this could also
be demonstrated at = 193 for a NA of 1.85, resulting in sub-30 nm high-aspect-ratio imaging
at low cost over large areas.
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13722
9. Appendix
In this section an analysis is presented of the Fresnel reflection equations for both the TM and
TE polarizations of light in the near-field reflection regime relevant to this paper. The
analysis is used to deduce which substrates would allow a reflection larger than unity in the
evanescent regime. Further details and derivation can be found in Ref [30].
The various terms and relationships used or mentioned in this section are listed in Table 2,
where l is used to refer to a particular layer. The analysis is limited to non-magnetic media
due to the lack of availability of magnetic materials at optical frequencies and is also limited
to the magnitude of the Fresnel reflection, and not its phase. Media 1, 2 and 3 are indicated in
Fig. 2 and these will also be referred to in the text. The equations relating the transmission
and reflection coefficients of the x, y and z components of the E-field for the TM and TE
polarizations at the interface between two media a and b are tabulated in Table 3 and
Table 4 [31]. Here, transmission denotes the transmitted amplitude into medium b from
medium a and reflection denotes the reflected amplitude back into medium a from medium b.
Table 2. Definitions and identities used in this chapter
k0
nl nl inl
l l i l
n l in l
NA n 2
k x NA k 0
k z ,l k z ,l ik z,l
k 0 l i l NA 2
x component, Ex
E t ,x
Transmission
Reflection
Et
2k
t ab ,TM a k zbza ba kbza
Ei
Er
k k
rab ,TM bb k zaza aa k zbzb
Ei
E i ,x
E t ,z
t ab ,TM x
1 NA
b
E r ,x
E i ,x
z component, Ez
E i ,z
2
rab ,TM
t ab ,TM z
t ab ,TM
E r ,x
E i ,x
NA
rab ,TM
Et Et ,y
2k za
t ab ,TE
Ei Ei ,y
k za k zb
Reflection
k k zb
E r E r ,y
rab ,TE za
Ei Ei ,y
k za k zb
TM Fresnel reflection
Firstly, the TM Fresnel reflection equation is analysed for conditions that will give rise to a
reflectivity r > 1, which is required for the resonance-assisted lithography technique to be
effective. The TM Fresnel equation can be derived as (see [31] for details)
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13723
r23,TM
3 k z ,2 2 k z ,3
.
3 k z ,2 2 k z ,3
(2)
r23,TM
3 k z ,2 2 k z ,3
3 k z ,2 2 k z ,3
1.
(3)
A finite loss in the photoresist is first assumed, such that 2 has both non-zero real and
imaginary components. Expanding Eq. (4) in its complex form and rearranging gives [30],
k z ,3 3k z,2 2 3k z ,2 2 3k z ,2 2 3k z,2 2
(4)
Equation (4) is a general form including resist losses, but is not sufficiently simple to
allow us to get any idea of the material properties required of substrate 3 to achieve
resonance-assisted enhancement, which is the primary goal of this exercise. However, note
that the loss or extinction coefficient in conventionally used photoresists is more than 10
times smaller than its refractive index, and close to a 100 times smaller than the resist index in
several cases, so we can make the reasonable assumption that 2 2 to proceed further.
Under this assumption, which also implies that k z ,2 0 , since operation is in the evanescent
regime of the photoresist, Eq. (4) then simplifies to
3k z ,3 3k z,3 .
(5)
Interestingly, Eq. (5) is now independent of the optical properties of the photoresist.
However, this is so only if it is a lossless photoresist operating in the evanescent regime (so
that n2 < NA). If the conditions in the following paragraph are satisfied, then a method to
achieve a reflection greater than unity in the evanescent regime will be found.
If 3 0 and 3 0 , then the right hand side of Eq. (5) is always negative and its left
hand side is always positive, therefore satisfying it. If 3 0 and 3 0 , then Eq. (5) may
still be satisfied if material 3 is sufficiently lossy. This indicates that the underlying enhancer
could either be a plasmonic material (i.e. a metal, with 3 0 ) or a positive dielectric with
significant loss. These are important solutions and reinforce the prior knowledge that
plasmonic metals [1719] are possible candidates, but that other surface states such as the
SEP [2022] may also be used.
However, one may go beyond the well-known SPP and SEP solutions and a closer look
reveals that Eq. (5) is also satisfied by a material with 3 0 and 3 0 provided k z ,3 0
and k z,3 0 . This requires a material with a negative loss (i.e. a gain medium requirement).
Figure 16 shows the use of a complex plane plot with relevant quantities to illustrate this
particular scenario. The relevant quantities are labeled and depiction of the figure is explained
as necessary in the caption of Fig. 16, for example 3 is labeled as Vector C and denoted by a
dotted arrow.
The quantity k 02 3 i 3 NA 2 (Vector B) is the argument for which the roots are to be
determined to compute kz,3 so that the gain-medium requirement may be satisfied; this
argument is denoted as a solid arrow. Like any complex number this quantity also has two
possible roots (Vector A and Vector D), only one of which satisfies the gain medium
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13724
requirements. The desired root is Root 2 as labeled in Fig. 16 as it is the only correct solution
that allows Eq. (5) and the gain medium requirements to be satisfied.
Fig. 16. Complex-plane plot showing that Eq. (5) is also satisfied if the substrate 3 has a
positive real part 3 and a negative imaginary part 3 , depicted by Vector C. Although, this
results in an argument (Vector B) that has two possible Roots (to compute the z-wave number
kz), namely Root 1 (Vector D) and Root 2 (Vector A); the correct solution is in fact Root 2
(Vector A) as this allows
k z ,3 . This figure indicates the relative positions of the vectors required to achieve the desired
solution.
The enhancement shown above using metals (or lossy dielectrics) is limited to the TM
polarization. The TM polarization suffers from an increasingly lower contrast at higher NAs
[3]. In addition, as we will show in a later section, the material requirements for NAs as high
as 1.85 and possibly larger are met much more easily for the TE polarization of light, which is
the industrys preferred polarization for lithography. With this we proceed to solve the TE
Fresnel reflection for a reflectivity > 1.
TE Fresnel reflection
It is well known that metals and lossy dielectrics are not a possible solution to support TE
surface waves for evanescent field enhancement as the TE polarization lacks a normal
component of the electric field. The following analysis confirms this result but we also seek
new gain-like solutions. The TE reflection coefficient at the 2 : 3 boundary is [31]
r23,TE
k z ,2 k z ,3
k z ,2 k z ,3
1.
(6)
Rearranging with real and imaginary terms and simplifying [30] yields
k z ,2 k z ,3 k z,2 k z,3 0
(7)
(8)
Of course it is known that k z,2 must be positive owing to the fact that the photoresist is a real
medium operating in the evanescent regime. This directly implies that for Eq. (8) to be
satisfied k z,3 must be negative. This is the solution for a growing evanescent wave in the
substrate which again implies gain in the underlying medium. While a gain medium below
the photoresist may seem at first unfeasible, there is a practical effective-gain-medium
solution, involving the use of composite or multi-layered media, thanks to the evanescent
nature of the waves, as used in this paper.
#189972 - $15.00 USD Received 3 May 2013; revised 20 May 2013; accepted 23 May 2013; published 31 May 2013
(C) 2013 OSA
3 June 2013 | Vol. 21, No. 11 | DOI:10.1364/OE.21.013710 | OPTICS EXPRESS 13725