Johannes Fiedler presented NanoLace at FOMO, 20 September 2022

Title: LACENET —A machine learning approach for mask generations for matter-wave lithography

Abstract:

Recent progress in matter-wave experiments led to technical applications, particularly for acceleration
sensing, single-particle detectors, quantum microscopes or matter-wave lithography.
Thus, they act on the nanometre length scale. Consequently, the quantised nature of the object
is not neglectable. In particular, the quantum vacuum has to be taken into account. Hence, the
interactions between the objects are dressed by the vacuum polarisability leading to dispersion
forces.
The diffraction of matter waves is based on the wave-particle duality and has the advantage
that waves with sub-nanometre wavelengths can be created and thus strongly increases
the resolution compared to optical devices [1]. However, the additional interactions between
the matter-wave particles and the diffraction object dramatically influence the propagation of
the wave [2]. We will illustrate the impact of dispersion forces on the results of diffraction
experiments and demonstrate possibilities for their manipulation to enhance the contrast for
matter-wave lithography applications [3].
hotolithography is a commonly applied method to create, among others, semiconductor devices.
The current use is extreme-ultraviolet (EUV) photolithography that uses electromagnetic
radiation with a wavelength of 13.5 nm, corresponding to an energy of 92 eV.
The ability to pattern materials at ever-smaller sizes using photolithography is driving advances
in nanotechnology. When the feature size of materials is reduced to the nanoscale, individual
atoms and molecules can be manipulated to dramatically alter material properties.However,
the secondary electron blurring from extreme-ultraviolet photons hinders the creation patterns
with a resolution below around 8 nm. An alternative approach is the use of matter waves
which reaches much smaller wavelengths with a lower amount of kinetic energy. Lithography
with metastable atoms has been suggested as a cost-effective, less-complex alternative to EUV
lithography. In binary holography, a pattern of holes is used to approximate a Fourier transform
of the desired target pattern [4]. This simple approach cannot be applied to matter-wave lithography
with dielectric masks due to the additional dispersion forces. To overcome this issue, we
will introduce a machine learning approach trained on the relation between mask design and
interference pattern allowing an efficient estimation of a mask for a given target pattern [5].
This is of particular relevance for metastable atom lithography with binary holography masks,
currently pursued in the FET-Open project Nanolace [6].
[1] C. Brand, et al. Ann. Phys. (Berlin) 527, 580–591 (2015).
[2] N. Gack, et al. Phys. Rev. Lett. 125, 050401 (2020).
[3] J. Fiedler, B. Holst, J. Phys. B: At. Mol. Opt. Phys. 55, 025401 (2022).
[4] T. Nesse, I. Simonsen, B. Holst, Phys. Rev. Applied 11, 024009 (2019).
[5] J. Fiedler, et al. in preparation.
[6] https://www.nanolace.eu/