Development of functional sub-100 nm structures with 3D two-photon polymerization technique and optical methods for characterization

Two-photon polymerization (TPP) technique is one of the common and widely used techniques to realize the fabrication of high quality 3D microstructures.

A tightly focused laser beam mostly within a pulse duration range between several tens and hundreds of femtoseconds is used to produce a focal spot inside the material. Polymerization takes place inside the photosensitive polymer resin just in a tiny region where the accumulated energy exceeds the polymerization threshold resulting in arbitrary 3D structures with subwavelength resolution.^1–5 Different groups have reported that the TPP technique allows reaching a structuring resolution of sub-100 nm in different materials. Haske et al. reported up to 65 nm thin lines,⁶ Juodkazis et al. reached 30 nm structures in SU-8 photoresist,⁷ and Tan et al. even sub-25 nm lines in SCR500.⁸ Due to the two-photon absorption and the threshold behavior⁹ of the used material, it is possible to reach higher resolution than with the often used laser direct writing process.^10,11 Also the precise control of the processing parameters such as the number of pulses, pulse duration, or pulse energy, for example, allow increasing the resolution. Such small structures produced by TPP suffer from destruction and deformation caused by surface tension during the development process in the post-treatment. One of the possibilities to improve the resolution and increase the physical strength of the structures is the optimization of the chemical composition of the material.

In this paper, we demonstrate TPP structuring with sub-100 nm resolution using ultrashort laser pulses and controlling the chemical composition of the material.

In the last years, miniaturization is pushing technical structure sizes to its limits. Not only in the research field of TPP but also in semiconductor industry, structures well below 100 nm are produced.¹² For that the characterization of such structures is a demanding and very important task.¹³ Classical microscopy methods do not allow direct imaging in this regime because of the Abbe diffraction limit. At the same time, other methods capable to fulfill this task as are electron or atomic force microscopy are too sophisticated or expensive. A well-established method for the characterization of such kind of structures is the scatterometry method. We have analyzed the feasibility of a combination of a scatterometric method and white light interferometry for the characterization of periodic sub-100 nm structures with special emphasis on structures produced with TPP. For this, we have performed a simulation based sensitivity analysis and also built an experimental setup to proof the simulation results.

We have developed a new photosensitive material, which is synthesized by adding a cross-linker to the zirconium based organic–inorganic hybrid material (Zr-hybrid material). The cross-linker leads to an enhanced stability of the polymerized material and increases the survival odds of the sensitive structures during the post-treatment process.

For the fabrication of the structures in this project, the Zr-hybrid material was synthesized using the sol-gel base technique (refractive index: 1.502).¹⁴ In this material, methacryloxypropyl tri-methoxysilane (MAPTMS) and methacrylic acid were used for photo-polymerizable methacrylate moieties. Zirconium n-propoxide (ZPO, 70% in propanol) was used as an inorganic network former. The molecular ratio of MAPTMS to ZPO was set to 5:1 (20%). After the synthesis of above organic/inorganic material, 0.5%(w/w) of 4,4-bis(diethylamino)benzophenone and 15%(w/w) of 2 dipentaerythritol penta-meth-acrylate (DPMA) was added as a photo-initiator and cross-linker, respectively. The cross-linker DPMA is a molecule containing multiple methacrylate groups, which is able to participate in photo-polymerization reaction along with the Zr-hybrid. The cross-linker “reinforces” the hybrid material by increasing the achievable cross-linking density. In this way, the resulting TPP structures with much smaller features can survive, i.e., the achievable resolution is improved. The sample was prepared by drop casting onto a cover glass substrate. The sample was dried on the hotplate at 100 °C for 1 h before the polymerization.

Figure 1 shows a schematic illustration of the experimental setup. The used light source is a Ti:Sapphire based high-peak power ultrafast fs laser oscillator¹⁵ with a repetition rate of 5 MHz, a central wavelength of 800 nm, and a pulse width smaller than 50 fs. A 100 immersion microscope objective lens¹⁶ was used to focus the laser beam into the volume of the photosensitive material. The laser beam was focused from the backside of the cover glass plate through the refractive index matching oil (⁠$noil=1.515$⁠) filled between the objective lens and the surface of the glass plate. The photo-polymerized structure was drawn by a computer controlled three-axis linear stage. The scanning speed of the laser beam was kept at 1 mm/s. A complementary metal–oxide–semiconductor (CMOS) camera was mounted behind a dielectric mirror for on-line monitoring. After the polymerization process, all nonirradiated photosensitive material is removed by dipping the sample into 1-propanol and dried with the critical point drying technique to avoid the deformation of the polymerized structure due to surface tension generated during the drying.

In order to study the influence of the cross-linker on the stability of the polymerized material, 3.2 μm long free-hanging lines were fabricated. Figure 2 shows scanning electron microscopy (SEM) images of the free-hanging lines fabricated with standard Zr-hybrid material and with added cross-linker. For the Zr-hybrid material with the cross-linker, a minimum lateral resolution of 82.5 nm was obtained using an average input laser power of 0.5 mW. Without the addition of the cross-linker, only a minimum resolution of 150 nm was achieved. The aspect ratio of height to width for the Zr-hybrid material with cross-linker is smaller than that of the standard Zr-hybrid material. This fact can be attributed to material characteristics itself, since intensity distribution in the focused laser beam does not change in the experiment.

Figure 3 shows the dependence of the structural resolution (width and height) on the input laser power. Both dimensions of the polymerized lines increase when the average input laser power is increased. Also the material with added cross-linker has a higher TPP threshold than that of the standard material, and smaller minimum resolutions can be achieved.

TPP with few-cycle pulses from an ultrabroadband laser (white light source) is a yet unexplored regime promising the potential to shrink the TPP structures to a value considerably below the structures produced with pulses in the regime of several tens or hundreds of femtoseconds. Few-cycle laser pulses allow for reduced average power for TPP because of their much higher peak intensity and structuring without thermally damaging the samples at the same time.

In our experiments, we employed a home-built NOPA (Ref. 17) which supports pulses with up to 300 nm of optical bandwidth and duration of a few optical cycles. This system should provide the above mentioned benefits for sub-100 nm TPP structuring in the polymer material. The pulsed radiation is focused into the material, wherein a nonlinear photo-chemical reaction takes place and leads to the polymerization process.¹ Because of the threshold behavior and the nonlinear type of interaction, it is possible with precise control of the pulse parameters like number and pulse energy to reach a spatial resolution markedly below the diffraction limit of the used laser wavelength. Even with laser radiation around 800 nm (diffraction limit around 400 nm), structure sizes below 100 nm are achievable, since the polymerization occurs only in a very small area around the focal volume as a result of its cubic dependence.¹⁸ 

The light source for the experiments is an amplifier system consisting of a laser oscillator,¹⁹ a single pass rod-type fiber amplifier,²⁰ and the NOPA as white light source.¹⁷ The usable power spectrum of the NOPA output (420 mW average at 1 MHz) covers a spectral range from 700 to 980 nm, which supports Fourier-limited pulse duration of 5.9 fs. For more detailed information about this system, see the corresponding Refs. 17 and 19, and 20.

In 3D nano-fabrication, the voxel size depends on 1/(NA).⁴ This is why we choose a microscope objective with large numerical aperture (NA) (Zeiss microscopic lens 100 ×, 1.25 NA (oil-immersion)) for realizing a high fabrication resolution. The material used is the same as described in Sec. II A (20% Zr-hybrid material without addition of cross-linker) developed by the Laser Zentrum Hannover e.V. The experiments are realized in dried polymer droplets on a cover glass substrate, which are moved under the objective lens with translation stages (Physik Instrumente M-511.DD in x-direction and M-505.2DG in y- and z-directions) with minimum step size of 50/100 nm and a maximum velocity of 3/50 mm/s, respectively.

For pulse duration as short as possible inside the material, the pulse chirp is precompensated using dispersive mirrors (DCMs: double chirped mirrors) and a CaF₂ wedge pair before the light propagates through the glass material of the microscopic lens (see Fig. 4). The pulse duration behind the microscopic lens was measured with SPIDER technique to approximately 10 fs. The postprocessing of the polymerized material was performed as described in Sec. II B.

To determine the optimum processing parameters for producing sub-100 nm direct written TPP structures, pulse power and velocity (equivalent to the repetition rate) are changed independently, while the other parameters stay constant. The resulting dependencies of the width and height are shown in Fig. 5(a).

It should be mentioned that the power for all results within this article was measured in front of the microscope objective, because of a lack of a standard method to precisely measure the power inside the focal spot, particularly, when the laser is tightly focused with an oil-immersion microscopic lens. This difficulty arises from the very small average power inside the focal spot and the high divergence of the beam. Even calculations from lens transmission could not give precise values since the power losses in microscopic lenses are caused by the losing at the microscope aperture, which lead to a relation between net output power and incident beam shape and size.¹⁸ 

The depth of recording was set to approximately 10 μm beyond the surface of the glass substrate to have aberration-free structuring conditions inside the material.¹⁸ 

Figure 5(a) reveals that the cross section of the produced structure has an axial extent larger than the lateral. This behavior is typical and depends on the high NA value of the focusing optics. The aspect ratio (axial/lateral) was approximately 2:3 for the NA = 1.25 focusing. In the axial dimension, it was possible to realize structure sizes of 90 nm, whereas in the lateral extend, the minimum dimensions were approximately 240 nm [see Fig. 6(b)]. All measurements include a not exactly known amount of gold layer thickness, which is needed for SEM-observation.

Sub-100 nm structures are also produced by varying the writing speed [see Fig. 5(b)], which leads to a different amount of applied pulses and deposited pulse energy along the structure. SEM images of the smallest structures out of Fig. 5(a) are shown in Fig. 6.

Scatterometry has established itself as one of the most important methods for the technical characterization of periodic structures with feature sizes below the diffraction limit.²¹ Scatterometry is used as generic term for a variety of nonimaging optical methods for the reconstruction of periodic structures even below the Abbe diffraction limit. The main principle is based on the parameter reconstruction of the analyzed structures by a comparison of measured and simulated spectra containing intensity as well as polarization information.²² 

An example of such a scatterometric measurement setup is the so called Fourier-Scatterometry (FS). The sample to be analyzed is illuminated through a high numerical aperture microscope objective, and the back focal plane of this objective, also called the pupil plane, is imaged with a CCD-camera. Each point in the pupil plane contains the diffraction information for its corresponding incident direction (see Fig. 7). For that, in contrast to other scatterometric configurations, it allows to access the diffraction information for a wide range of incident angles with one shot and without any mechanical angle scanning. The parameter reconstruction of the structures is then performed as for other scatterometry types by comparison of measured and simulated spectra.

To increase the sensitivity toward the depth of the structures, in this work, we have expanded the Fourier-Scatterometry by combining it with a white light interference setup containing a white light source and an optical reference path. In this Linnik-type interferometry setup, both the object path pupil plane and the one in the reference path interfere and are imaged with a Bertrand-lens on a CCD-camera.

Using this combination, not only the diffraction information of the sample can be recorded but also phase information; in this way, the high lateral resolution of a scatterometric measurement is combined with the supreme topographic information and depth sensitivity of a white light interferometry setup, allowing a high-sensitivity characterization of the analyzed structures. The measurement principle is depicted schematically in Fig. 8.

We have performed a simulation based sensitivity analysis for the characterization of different line gratings: photosensitive (TPP-structured) material on a glass substrate, the same material on a silicon substrate, as well as electron-beam structured resist on silicon. The width (critical dimension, CD) of the lines was 50 and 100 nm, the period (pitch) was 100 and 200 nm, the height was also 100 nm, and the side-wall angle was chosen to be 87°.

For the simulations, we used our software package MICROSIM, which is a Maxwell solver based on the rigorous coupled wave analysis method.²³ 

The sensitivity of the Fourier-Scatterometry methods toward changes in the mentioned structure parameters was analyzed. For the reconstruction of structure parameters, it is not only important to be sensitive, but also the changes in signal for the variation of different parameters need to show small correlations. A more detailed definition of the sensitivity (uncertainty) values and the corresponding covariance-matrix (correlation values) is given in Ref. 24.

The Tables I and II contain the results for the scanning white light interference Fourier-Scatterometry (SWL, italic) and for the classical FS for a simulation of a line-grating of photosensitive material (TPP-structured) on a silicon substrate. The results for the other grating types we have investigated are comparable and are omitted for the sake of brevity. The SWL-FS mainly shows comparable sensitivities to the FS, but is more sensitive to variations of the height and also shows smaller correlation values of CD/height and CD/SWA (marked bold). However, for transparent substrates, there is no sensitivity gain because of the reduced reflected intensity of the sample.

There are, to the authors knowledge, no other studies analyzing the sensitivity and correlations of the combination of Fourier-Scatterometry and white light interference. For this, the obtained results could not be directly compared to other results. A very closely connected theoretical work²⁵ based on simulations of Fourier-Scatterometry using a spatially coherent light source and scanning over the object of interest also shows that in some circumstances a comparable improved sensitivity is achieved. A very similar experimental realization has also been presented by de Groot et al.²⁶ 

We have built a setup for the white light interference Fourier-Scatterometry measurements to verify the simulation results. The schematic configuration which the experimental setup is based on is depicted in Fig. 8. Additional information can be found in Ref. 24.

In a first step, we recorded the pupil plane images for different reference mirror positions and compared them to the pupil images obtained from the simulations. The analyzed structure was an e-beam written line-grating with a CD of 200 nm and a period of 400 nm.¹² The images show good agreement (see Fig. 9),²⁷ but the differences between simulation and measurement still does not allow a reconstruction from a library search. At the moment, no aberrations in the optical path are taken into account in the simulations. The beam splitter, objectives, and reference mirror are modeled with their corresponding ideal Jones matrices. Measurement and inclusion of the transfer function of each of these optical elements is desirable and probably will allow getting better agreement in future. As soon as better agreement is reached probably, also other standard optimization algorithms²⁸ instead of the mentioned library search should be taken into account.

We have demonstrated TPP structuring with sub-100 nm resolution via optical and chemical approaches, respectively. The TPP structure fabricated with Zr-hybrid with cross-linker material showed smaller minimum resolution compared to that of standard Zr-hybrid material. The adding of cross-linker material enhanced physical strength of TPP structure resulting in an achievement of the resolution as small as 82.5 nm. Thus, cross-linker material has a potential to improve the resolution to sub-100 nm concerning with the material characteristics. On the other hand, we also gain experiences in TPP with few-cycle laser pulses showing that the reduction of the used pulse duration allows for even thinner feature size. It should be mentioned that due to different structure geometries this comparison cannot bear finally, but we showed to our knowledge for the first time 90 nm resolutions in TPP process initiated by a NOPA system supplying sub-10 fs pulse duration using the standard sol-gel Zr-hybrid material. In the end, we believe that a combination of few-cycle laser pulses with the further developed material (addition of cross-linker) could be a promising route toward resolutions of 50 nm or less.

With regard to the characterization of such small structures, the combination of white light interferometry and Fourier-Scatterometry proofs to be a good alternative to other scatterometric methods. The combination shows significantly higher sensitivities toward the structure topography and height. At the same time, it has smaller parameter correlations than the classical Fourier-Scatterometry. The actual verification of the structure reconstruction capabilities of the method for the sub-100 nm TPP structures is work in progress. The next step is the reconstruction of structure features by performing a library search.

This work was supported by the German DFG-funded priority program (No. SPP1327) on optically generated sub-100 nm structures for technical and biomedical applications, within the subproject development of functional sub-100 nm 3D two-photon polymerization technique and optical characterization methods.