Method and system for etch pattern generation

Abstract

An etch pattern generation system (1) and method comprising-an incident particle beam (21) having a most probable wavelength (lambda) from a particle source (2); and-a reticle (4) comprising through-holes (3) arranged in the incident particle beam (21). The etch pattern generation system (1) generates a single nano feature size pattern (51), and wherein the through-holes (3) are arranged aperiodically in the reticle. A computer-implemented method for generating a model of an etch reticle (4) is also disclosed.

Description

Technical Field

[0001] This invention relates to micro-etching and the fabrication of integrated circuits and other devices, such as quantum devices that cannot be mass-produced using existing technologies. More specifically, it relates to exposing a mask using a particle beam to generate a particle pattern based on information encoded in the mask, and using the particle pattern to expose a target (such as a resist on a silicon wafer). Background Technology

[0002] Integrated circuit manufacturing is currently based on pattern generation (using photolithography based on photomasks): an enlarged version or portion of the desired chip pattern is printed on a substrate (photomask). The pattern on the photomask can be distorted due to aberrations, etc. Light (photons) is transmitted through or reflected off the photomask using optical components (refractive or reflective lenses), thus creating a reduced image of the pattern on a silicon wafer. The silicon wafer is coated with a photosensitive material (resist), which reacts with the photons, creating a permanent imprint of the image in the resist. This resist is then used as a physical mask in subsequent manufacturing steps. The two main light sources used in the semiconductor industry today are deep ultraviolet (DUV) light with a wavelength of 193 nm and extreme ultraviolet (EUV) light with a wavelength of 13.5 nm.

[0003] The long-term goal is to produce patterns with increasingly smaller feature sizes at ever-increasing information densities, often referred to as minimum pitch—the minimum distance between two feature centers. For standard optical systems, when processing in air or vacuum, the minimum pitch achievable is half the wavelength of the projected beam (refractive index 1, Abbe criterion). Minimum pitches of approximately 30% of the wavelength are achieved by further advancing DUV performance using wavefront engineering techniques such as phase-shifting masks, off-axis illumination, or optical proximity correction. The final pitch in DUV etching is achieved using immersion lithography, where a high-refractive-index fluid is introduced between the final lens and the wafer. Combining these techniques allows DUV to operate at an effective wavelength of approximately 135 nm, resulting in a minimum pitch of approximately 38 nm. Further reductions in pitch require multiple exposures to create a single chip pattern, thus reducing throughput and increasing complexity.

[0004] The most advanced technology is extreme ultraviolet (EUV) lithography, which uses photons (electromagnetic waves) with a wavelength of 13.5 nm. According to the Abbe criterion with a refractive index of 1 (vacuum), it should be able to produce patterns with a minimum spacing or minimum feature size of about 6.75 nm.

[0005] However, due to the high energy of photons in EUV etching, the patterning process in the resist is mediated by photogenerated secondary electrons, which can travel several nanometers before inducing the reaction. Current experimental and theoretical indicators suggest that the secondary electron blurring radius of EUV is approximately 3 nm, which limits the achievable feature size to about 6 nm. This means that it is impossible to produce quantum devices based on small quantum dots and individual atoms and molecules using EUV. Shifting to wavelengths even shorter than 13.5 nm will only exacerbate the secondary electron problem.

[0006] In Berggren, KK, Bard, A., Wilbur, JL, Gilaspy, JD, Helg, AG, McClelland, JJ, Rolston, SL, Phillips, WD, Prentiss, M., Whitesides, GM: Microlithography by Using Neutral Metastable Atoms and Self-Assembled Monolayers, Science 269(5228), 1255–1257(1995) https: / / arxiv.org / abs / https: / / www.science.org / doi / pdf / 10.1126 / science.7652572. https: / / doi.org / 10.1126 / science.7652572, etching using metastable atoms was proposed as an alternative to photolithography. Pattern generation in thiol-based resists was demonstrated using a beam of metastable argon atoms transmitted through a grating in a proximity etching setup. In other experiments, photomasks have been used to generate patterns using metastable atoms, and in some cases, atoms have been deposited directly onto the substrate. However, it has not yet been demonstrated that atomic beams can be used to generate any desired complex patterns with small minimum spacing.

[0007] Some experiments have been conducted to focus atomic beams onto a small point using solid-state lenses, mirrors, or fields. However, this can only be applied to serial writing and is therefore not suitable for mass production.

[0008] In Fujita, J., Morinaga, M., Kishimoto, T., Yasuda, M., Matsui, S., Shimizu, F.: Manipulation of an atomic beam by a computer-generated hologram, Nature 380(6576), 691-694(1996), https: / / doi.org / 10.1038 / 380691a0, metastable neon atoms are used to generate a desired pattern on a screen by transmitting an atomic beam through a solid mask (composed of approximately circular vias with a diameter of 30 nm etched into a silicon nitride film). All the vias are of the same size and are positioned on the film according to a regular, square grid structure. We call this method grid-based binary holography. This setup operates with all system dimensions within the Fraunhofer regime (far field), where the Fresnel number F << 1. The mask is designed using the Fourier transform of the desired pattern, such that the diffraction process occurring during transmission through the mask results in multiple copies of the desired pattern appearing on each side of the 0th-order peak. The pattern copy on one side of the 0th-order peak is rotated 180° relative to the pattern copy on the other side (see the original article). Figure 4 ).

[0009] Nesse, T., Simonsen, I., Holst, B.: Nanometer-Resolution Mask Lithography with Matter Waves: Near-Field Binary Holography, Phys. Rev. Applied 11, 024009 (2019), https: / / doi.org / 10.1103 / PhysRevApplied.11.024009 discloses a theoretical study on grid-based binary holography, which uses small masks with square and hexagonal grid structures. The distance between the mask and the pattern plane is only 40 μm, and it has extremely small holes of less than 1 nm.

[0010] The prior art grid-based binary holography disclosed in some of the above disclosures and shown in Figure 1 is not suitable for rapidly generating patterns with feature sizes smaller than those of current state-of-the-art technologies. Specifically, it is not suitable for generating small features using the large masks required for the large write-fields (typically 29mm x 33mm) used in the semiconductor industry. Summary of the Invention

[0011] There is a need to rapidly generate patterns with feature sizes lower than those of previous technologies in order to improve the performance of semiconductor devices and enable the mass production of a range of quantum devices.

[0012] The objective of this invention is to disclose a system and method for rapidly generating patterns with feature sizes down to the nanometer range.

[0013] One of the main advantages of this invention compared to prior art is its ability to generate nano-features with nanometer spacing on a large writing field, which was not possible with prior art. Furthermore, nano-features can be written using apertures larger than the feature size, something not yet demonstrated in binary holography. This simplifies the production of photomasks.

[0014] A further advantage of this invention is that the exposure time can be more precisely defined than in grid-based binary holography, where the pattern intensity is distributed unevenly across and between patterns. Furthermore, in this invention, only one copy of the desired pattern is produced during exposure, which simplifies the production process before and after pattern generation.

[0015] The present invention that solves the above problems is an etching pattern generation system and method according to the independent claims, which generates patterns by etching. Attached Figure Description

[0016] The above and other advantages of the systems and methods provided according to the present invention will now be explained in more detail by way of example and with reference to the accompanying drawings. The drawings are not necessarily drawn to scale. Rather, certain features may be exaggerated in scale or shown in a slightly simplified or schematic manner, wherein certain known elements may be omitted in order to illustrate the principles of the invention without making the drawings confusing due to details that do not contribute to the understanding of these principles.

[0017] Figure 1 illustrates pattern generation based on prior art grid-based binary holography. Multiple copies of the target pattern (510A, 510B) are generated via diffraction of atomic waves (210) through a binary holographic mask (400) having apertures (300) of all the same size. The apertures are placed at positions corresponding to the grid structure. The grid has a pixel-like structure, for example, 512x512 pixels, where each pixel can be fully or partially open (through-hole) or fully closed. The periodicity of the grid structure results in several copies of two rotating versions of the target pattern with different intensities at a pattern-target distance (L2) behind the mask.

[0018] Figure 2An etching pattern generation system (1) according to an embodiment of the present invention is schematically shown. The etching pattern generation system (1) includes an incident particle beam (21) with the most probable wavelength (λ) from a particle source (2), and a mask (4) disposed between the particle source (2) and a pattern target (5). The pattern (51) generated on the pattern target is generated by particles diffracted by the mask (4). The source distance (L1) from the particle source (2) to the mask (4) and the pattern target distance (L2) from the mask (4) to the pattern target (5) are indicated, respectively. Furthermore, the mask (4) includes a through-hole (3) with an aperture width (w) and a mask extension (a), i.e., the maximum distance between the through-holes (3) exposed to the incident particle beam (21).

[0019] Figure 3 The main components and parameters of an etching pattern generation system (1) according to an embodiment of the present invention are schematically shown in one dimension. Component symbols and Figure 2 The similarity used in [the text].

[0020] Figure 4 The Fraunhofer diffraction patterns of light waves (A) and metastable helium matter waves (B) of the same wavelength passing through identical SiN double-slit masks with a thickness of 5 nm and 8 nm openings separated by 8 nm walls are illustrated graphically. The vertical axis represents normalized intensity, while the horizontal axis represents position in μm. The matter wave interacts with the dielectric mask using dispersion forces (which do not apply to light waves), resulting in a broadened diffraction pattern with an increased distribution of higher diffraction orders.

[0021] Figure 5.1 , Figure 5.2 , Figure 5.3 , Figure 5.4 , Figure 5.5 and Figure 5.6 The atomic flux [atoms / s / m] in one dimension is shown for each sequentially corresponding instance in Tables 1 and 2 in the vertical direction. 2 The horizontal line represents the position in nm. The lower curve shows the flux considering the dispersion forces (Di), while the upper curve does not, and therefore corresponds to the pattern that would be seen if the mask were illuminated with a beam of similar wavelength to the atomic beam. It can be seen that, for atomic beams, by selecting appropriate system parameters to adjust the dispersion forces, a strong reduction in the 0th-order background and higher feature contrast can be obtained, as further explained below.

[0022] Figure 6a , Figure 6b and Figure 6c Three examples of 2D pattern generation and the mask applied to the right of each corresponding pattern are shown in cross-sectional views. Horizontal and vertical dimensions are in nm. Figure 6aThe obtained feature size is 1.9 nm. Figure 6b The medium is 2.5nm and Figure 6c The wavelength is 1.1nm.

[0023] Figure 7 An example of a mask with mask material indicated by black squares is shown. The shaded areas are absorption areas of the physical apertures where particles do not transmit light, as explained below. Only the remaining white areas contribute to the diffraction pattern. The attached figure corresponds to the case where the center of the mask is aligned with the center of the source opening. Detailed Implementation

[0024] In the following description, various examples and embodiments of the invention are set forth to enable those skilled in the art to gain a more thorough understanding of the invention. The specific details described in the various embodiments and with reference to the accompanying drawings are not intended to be construed as limiting. Rather, the scope of the invention is defined in the appended claims.

[0025] The terminology used in this specification is intended to be interpreted in its broadest and most reasonable manner, although it is used in conjunction with detailed descriptions of certain specific embodiments of the invention.

[0026] For clarity, the following clarifies the terminology used in the document:

[0027] - Binary holograms and binary holography: Traditionally, a binary hologram refers to a hologram having a region (aperture) that is either completely transparent or completely opaque to an incident beam of light (scalar wave). For matter waves, the situation is slightly different, as the dispersion forces between the matter wave and the dielectric mask may prevent some or all of the matter wave from penetrating through the aperture in the mask. Furthermore, the phase is perturbed at the aperture. For the purposes of this document, the terms binary hologram and binary holography also apply to such holograms (masks) exposed to matter waves. Previous literature (theoretical and experimental) on binary holography did not consider dispersion force interactions. Experimental papers (such as NATURE cited above) include target patterns with large features that make the effects of dispersion forces less noticeable.

[0028] -Contrast ratio (c):(I 最大 -I 背景 ) / (I 最大 +I 背景 ), where I 最大 and I 背景 Represents the highest and lowest intensity.

[0029] - Characteristics: The smallest piece of information in a pattern. In etching, it is usually the identifiable objects in the pattern, such as lines or dots.

[0030] - Feature size: The dimensions of the feature. Feature size may vary in different directions, but in the case of this document, the minimum feature size is the most important.

[0031] - Fresnel number (F): where,

[0032] -Half Height Full Width (d 1 / 2 ): The distance between points where the intensity reaches half of its maximum value.

[0033] - Mesh Period: (g). The minimum distance between two apertures in a mesh-based binary holography. The mesh period will be equal to or greater than the minimum aperture width.

[0034] - Mask extension (a): The maximum distance between two through holes (3) in the plane of the mask (4).

[0035] -Minimum hole width (w): The minimum width of the through hole (3) in the mask (4). The through hole (3) may have different shapes and sizes depending on the situation.

[0036] - Opening diameter (da): The diameter of the opening of the particle source (2), which for a supersonic source is defined as the portion of the virtual source retained by the collimating aperture. The virtual source is defined in the literature as the intensity distribution that can be obtained by tracing back the atoms away from the exit surface to the minimum region. The exit surface is the point where the particles can be said to move in a free molecular flow without interacting with each other. See, for example: https: / / www.researchgate.net / publication / 356510630_Neutral_Helium_Micros copy_SHeM_A_Review. In the case of an ejection source, the opening of the particle source is simply a hole located between the particle reservoir and the vacuum, or a collimating aperture located downstream of that hole.

[0037] - Patterned target (5): An object set in a diffractive particle pattern opposite to the particle source (2) relative to the mask (4). In the context of integrated circuit manufacturing, this can typically be a silicon wafer coated with resist. Formulations for resists suitable for helium atoms are presented in the literature, for example in the paper by Berggren et al. cited above.

[0038] - Spacing: The distance between the centers of two adjacent features.

[0039] - Resolution: The minimum pitch that can be transferred to the substrate.

[0040] -Source distance (L1): The distance between the particle source (2) and the mask (4). More specifically, it can be described as the distance from the opening of the particle source (2) to the front of the holographic mask (4).

[0041] - Pattern target distance (L2): The distance from the mask (4) to the pattern target (5). More specifically, it can be specified as the distance from the back of the mask (4) to the pattern target (5).

[0042] - Production volume: This is the number of substrates or wafers that can be exposed per hour, and therefore a measure of the efficiency of the etching process.

[0043] - Wavelength (λ): The most likely wavelength of a particle in the incident particle beam (21) from the particle source (2). The particle source may have a skewed wavelength distribution.

[0044] The prior art conventional binary holography, initially mentioned and shown in Figure 1, used in conjunction with an atomic beam, is based on a holographic mask with uniformly sized apertures set according to an underlying grid structure. This has several effects on the resulting pattern.

[0045] First, different diffraction orders have different intensities, which are determined by the size and shape of the aperture and the grid period, meaning that the exposure time cannot be optimized for all parts that produce the pattern.

[0046] Second, the target patterns (510A, 510B) shown in Figure 1 are oriented differently, which complicates the production steps before, during and after the pattern generation.

[0047] Third, the pattern-target distance (L2) must be large enough to prevent the 0th order beam and target pattern from overlapping.

[0048] The minimum pattern target distance (L2) can be derived from the angular position θ of the first-order peak in the pattern target plane, relative to the center of the first-order peak as determined by the underlying mesh structure. The angle θ is measured relative to the surface normal of the mask plane, see Figure 1.

[0049] sinθ=λ / g

[0050] The center position of the target pattern closest to the 0th-order peak in the target plane (the center position of the first-order peak) is (approximately using a small angle):

[0051] (λ / g)L2

[0052] To prevent the first-order diffraction from overlapping with the 0th-order diffraction, regardless of the distance to the mask, the center of the first-order peak must be at a distance *a* from the center, so as not to overlap with the 0th-order peak, which will have the mask extension (*a*), since the 0th-order peak consists of all atoms passing through in a straight line. Therefore, we have

[0053] (λ / g)L2≥a→L2≥ga / λ≥wa / λ

[0054] We see that the larger the mask, the larger the minimum pattern target distance L2 must be.

[0055] In the first independent system implementation ES1-1, the present invention is an etching pattern generation system (1), comprising:

[0056] - An incident particle beam (21) with wavelength (λ) from particle source (2), and

[0057] - A mask (4) including a through hole (3) is set in the incident particle beam (21), wherein the etching pattern generation system (1) is configured to generate a pattern (51) on a pattern target (5).

[0058] In the system implementation ES1-2 of ES1-1, particles in the incident particle beam (21) arrive at the mask (4) as incident particle waves, wherein the mask (4) is configured to induce a phase shift in the particle wavefront as it propagates through the mask.

[0059] In the system implementation ES1-3 of ES1-1 or ES1-2, the induced phase shift is at least π, 3 / 2π or 5 / 3π for a distance of 0.016 times the minimum aperture width (w) of the through hole (3) from the mask plate (4).

[0060] The minimum hole width (w) is the minimum hole width of the through hole (3) in the mask (4).

[0061] In any of the system implementations ES1-1 to ES1-3, ES1-4, particles in the incident particle beam (21) arrive at the mask (4) as incident particle waves, wherein the mask (4) is configured to adjust the amplitude of the particle wavefront as it propagates through the mask.

[0062] In any of the system implementations ES1-1 to ES1-4, ES1-5, the etching pattern generation system (1) is configured to generate a single copy of the pattern (51).

[0063] In the system implementation ES1-6 of ES1-1 or ES1-5, the etching pattern generation system (1) is configured to generate a pattern (51) on the opposite side of the incident particle beam (21) at a target distance (L2) from the mask (4).

[0064] In any of the system implementations ES1-1 to ES1-6, ES1-7, the through-hole (3) has been adjusted to have an effective opening size (w) smaller than the physical aperture width (w) of the through-hole (3) by adjusting the dispersion force interaction between the incident particle beam (21) and the mask (4). e ).

[0065] In any of the system implementations ES1-1 to ES1-7, ES1-8, the parameter values ​​of physical aperture width (w), mask material, and mask thickness are selected to induce a phase shift on the incident particle wavefront by modifying the dispersion force interaction between the incident particle beam (21) and the mask (4).

[0066] In any of the system implementations of ES1-1 to ES1-8, ES1-9, the mask (4) is doped or implanted with any of the neutral particles, ions and electrons to alter the dispersion force interaction and / or Debye force interaction, thereby altering the phase shift on the wavefront.

[0067] In any of the system implementations ES1-10 of ES1-1 to ES1-9, the etching pattern generation system (1) is configured to reduce the background signal (the portion of the signal that passes straight through) such that the portion of the diffracted beam is larger than the light wave that propagates through the same mask (4) in the same direction along the axial direction of the via (3) to, for example, increase the contrast (c) of the pattern (51).

[0068] Generally speaking, the overall transmitted signal of particles is reduced more than that of light. However, the reduction in background signal (the portion of the signal that passes straight through) is greater than the reduction in diffraction signal.

[0069] In any of the system implementations of ES1-11 to ES1-10, the via (3) has a minimum aperture width (w) and a mask extension (a), which is the maximum distance between two vias exposed by the incident particle beam (21) in the mask (4), wherein;

[0070]

[0071] In any of the system implementations ES1-11 to ES1-11, ES1-12, the particle source (2) has an opening diameter (da) smaller than the source distance (L1) between the particle source (2) and the mask (4).

[0072] In any of the system implementations ES1-13 of ES1-1 to ES1-12, the mask extension (a) is less than the source distance (L1) between the particle source (2) and the mask (4).

[0073] In any of the system implementations ES1-14 of ES1-1 to ES1-13, the mask and the pattern target (5) are arranged in parallel.

[0074] In any of the system implementations ES1-15 of ES1-1 to ES1-13, the mask and the pattern target (5) are not arranged in parallel.

[0075] In any of the system implementations of ES1-11 to ES1-15, ES1-16, the mask is curved.

[0076] In any of the system implementations of ES1-1 to ES1-16, ES1-17, either the mask (4) or the patterned target (5) is tilted relative to the particle source (2).

[0077] In any of the system implementations ES1-18 of ES1-1 to ES1-17, the center of the particle source (2) is displaced relative to the center of the mask (4).

[0078] A tilted mask will introduce asymmetry into the diffraction pattern, and a curved mask can be configured to equalize the distribution of diffraction orders and thus support the illumination of the entire patterned target (5).

[0079] In any of the system implementations of ES1-1 to ES1-18, ES1-19, the minimum aperture width (w) is at least two, three, or four times the minimum feature size of the pattern (51).

[0080] In any of the system implementations ES1-1 to ES1-19, ES1-20, the mask (4) is a binary holographic mask.

[0081] In any of the system implementations ES1-21 of ES1-1 to ES1-20, the mask (4) is made of dielectric material.

[0082] In any of the system implementations ES1-1 to ES1-21, ES1-22, the mask (4) includes silicon nitride, Si3N x Or Si3N4.

[0083] In any of the system implementations ES1-1 to ES1-22, the via (3) is not set according to a predetermined underlying periodic grid structure on the mask (4).

[0084] In any of the system implementations ES1-24 of ES1-1 to ES1-23, the through hole (3) has a non-uniform minimum hole width (w).

[0085] In any of the system implementations ES1-1 to ES1-24, ES1-25, the via (3) has a minimum aperture width (w) of 5 nm, 4 nm, 3 nm, 2 nm or 1 nm.

[0086] Hole sizes smaller than 5 nm can be achieved in 2D materials.

[0087] In any of the system implementations ES1-1 to ES1-25, ES1-26, the via (3) has a maximum aperture width (w) of 100 nm.

[0088] In any of the system implementations ES1-1 to ES1-26, ES1-27, the through hole (3) has a different hole width (w).

[0089] In any of the system implementations ES1-1 to ES1-27, ES1-28, the through hole (3) has a different shape.

[0090] In any of the system implementations ES1-1 to ES1-28, ES1-29, the center-to-center distance (c) between adjacent through holes (3) is non-uniform.

[0091] In any of the system implementations ES1-1 to ES1-29, ES1-30, the mask thickness is less than 500nm, 100nm, 50nm, 20nm, 10nm, or 7nm.

[0092] In any of the system implementations ES1-31 from ES1-1 to ES1-30, the mask thickness is less than the minimum aperture width (w).

[0093] In any of the system implementations ES1-1 to ES1-31, ES1-32, the particle beam (21) comprises metastable atoms or molecules.

[0094] In any of the system implementations ES1-1 to ES1-32, ES1-33, the particle beam (21) comprises metastable inert gas atoms.

[0095] In any of the system implementations of ES1-1 to ES1-33, ES1-34, the particle beam (21) comprises metastable helium atoms.

[0096] In any of the system implementations of ES1-1 to ES1-34, ES1-35, the wavelength (λ) is less than any of the minimum feature size (df), minimum pitch, minimum aperture width (w), and mask thickness.

[0097] In any of the system implementations of ES1-1 to ES1-35, ES1-36, the wavelength (λ) is less than 1.0 nm, 0.5 nm, or 0.2 nm.

[0098] In any of the system implementations ES1-1 to ES1-36, ES1-37, the etching pattern generation system (1) is configured to generate a pattern on the pattern target (5) in a single exposure.

[0099] In any of the system implementations of ES1-1 to ES1-37, ES1-38, the etching pattern generation system (1) includes a particle source (2).

[0100] In any of the system implementations of ES1-1 to ES1-38, ES1-39, the etching pattern generation system (1) includes a pattern target (5).

[0101] In any of the system implementations ES1-1 to ES1-39, ES1-40, the particle source (2) is supersonic or ejected.

[0102] In any of the system implementations ES1-1 to ES1-40, ES1-41, the mask (4) is manufactured based on a mask model generated according to any of the computer-implemented methods EC1-1 to EC1-6.

[0103] In the first independent method implementation EM1-1, the present invention is a method for generating a pattern (51) by etching, which includes;

[0104] - Expose the mask (4) including the through hole (3) to the incident particle beam (21) from the particle source (2).

[0105] In the method implementation EM1-2 of EM1-1, particles in the incident particle beam (21) arrive at the mask (4) as particle waves, wherein the mask (4) is configured to induce a phase shift in the particle waves as they propagate through the mask.

[0106] In the method implementation EM1-3 of EM1-1 or EM1-2, the induced phase shift is at least π, 3 / 2π or 5 / 3π for a distance of 0.016 times the minimum aperture width (w) of the through hole (3) from the mask plate (4).

[0107] In method implementation EM1-4 of any one of EM1-1 to EM1-3, the method includes:

[0108] - Place the mask (4) in front of the pattern target (5) and generate a pattern (51) on the pattern target (5).

[0109] In any of the method embodiments EM1-1 to EM1-4, EM1-5, the patterned target (5) is provided at a target distance (L2) from the mask, the through hole (3) has a minimum aperture width (w) and the mask (4) has a mask extension (a) located between the through holes (3) exposed by the incident particle beam (21), and wherein;

[0110]

[0111] In any of the method implementations EM1-1 to EM1-5, EM1-6, the through-hole (3) has an effective aperture width (w) smaller than the minimum aperture width (w) due to the dispersion force between the incident particle beam (21) and the mask (4). e ).

[0112] In the method implementation EM1-7 of any of EM1-1 to EM1-6, the method is performed on the etching pattern generation system (1) of any of ES1-1 to ES1-40.

[0113] In method implementation EM1-8 of any of EM1-1 to EM1-7, the method includes generating a mask model according to any of EC1-1 to EC1-6.

[0114] In method implementation EM1-9 of any of EM1-1 to EM1-8, the method includes manufacturing a mask based on a mask model generated in any of the computer-implemented method implementations EC1-1 to EC1-6.

[0115] Now refer to Figure 2 Specific independent embodiments of the present invention are disclosed.

[0116] In this embodiment, the present invention is an etching pattern generation system (1) comprising an incident metastable helium atom beam (21) with the most probable wavelength (λ) from a particle source (2), and a mask (4) comprising a through hole (3) disposed in the incident particle beam (21).

[0117] Unlike the mask (400) of a prior art etching pattern generation system in Figure 1, where the vias (300) are set independently of the target pattern according to a predetermined underlying mesh structure, the vias (3) according to the present invention are not set on the mask according to a predetermined underlying mesh structure (4). Furthermore, the vias (3) can have different shapes and different minimum aperture widths (w). A single pattern (51) with a minimum feature size as low as 1 nm is generated on a patterned target (5) located at a target distance (L2) from the mask (4).

[0118] This invention utilizes the perturbation of particle waves caused by the dispersion force interaction between a dielectric mask and the particle itself when a particle passes through a hole in the mask. The dispersion force, specifically the Casimir-Polder force, is caused by the quantum mechanical ground-state fluctuations of an electromagnetic field in the absence of charge. Due to the field fluctuations, metastable helium atoms will be polarized for a short time. The resulting induced dipole moment then interacts with the mask through dipole-dipole interactions. These forces decay with distance (r... -3While exhibiting a power-law-like interaction, it plays a crucial role at the nanoscale. The interaction is stronger closer to the pore edge and weaker closer to the center. This results in a reduction in the number of atoms passing through in a straight line. In the Fraunhofer diffraction system, this has the effect of reducing the 0th-order diffraction peak and broadening the diffraction pattern, such as... Figure 4 As shown.

[0119] In this embodiment, metastable helium atoms pass through pores in a mask (which is a silicon nitride film). Dispersion forces prevent a very large majority of metastable helium atoms with a wavelength of 0.1 nm from penetrating through pores with a diameter of less than 2 nm in a 5 nm thick silicon nitride mask.

[0120] The interaction of dispersion forces has a positive effect on the contrast of the resulting pattern.

[0121] The difference in contrast can be illustrated by examining a one-dimensional combination of key system parameters. Table 1 below shows examples of parameter values ​​for a system implementation considering dispersion forces (assuming metastable helium and a 5nm thick silicon nitride mask) and a scalar wave system implementation. Table 2 shows the obtained parameter values ​​for both systems. It is noteworthy from Table 1 that all examples are in the range L2 ≤ wa / λ, where the first-order diffraction of the prior art overlaps with the 0th-order diffraction.

[0122]

[0123] Table 1: Parameters for the feature size used to compare systems considering and not considering dispersion forces. Dispersion force distributions of metastable helium atoms were used with a standard 5nm thick SiN wafer. The pattern consists of three slits, one at each edge of the mask and one in the middle. Furthermore, although the Fresnel number varies significantly, it can be seen from the last column that the condition L² ≤ wa / λ is satisfied for all instances.

[0124]

[0125] Table 2: Parameter results of the system in Table 1, for the full width at half maximum (FWHM) without dispersion: (d 1 / 2 (c)(-), and contrast: (c)(-) and the corresponding parameter (d) with dispersion force. 1 / 2 )(Di) and (c)(Di).

[0126] The two columns for contrast deserve special attention. It can be seen that, when considering dispersion forces, the contrast of all instances with comparable feature sizes has increased significantly. These results are also shown in... Figures 5.1 to 5.6 .

[0127] This invention allows the most probable (λ) particle in an incident particle beam to be compared with other key dimensions (such as, for example, feature size (df), minimum distance between features (d)). 1 / 2The distances are smaller compared to the mask extension (a) and the source distance and pattern target distance (L1, L2).

[0128] Since the most probable wavelength (λ) can be chosen to be small compared to the desired feature size (df), patterns can be generated using a velocity-broadened beam with a supersonic or jet source. The slight blurring introduced into the pattern is not a problem. The wavelength is given by the velocity, and conversely, by the de Broglie equations.

[0129] The contrast in a diffraction pattern can be modified by altering the effect of dispersion forces in one of the following ways: changing the size of the aperture, changing the shape of the aperture, changing the thickness of the mask material (film), changing the material of the mask, doping the mask by electronic or ion implantation or surface functionalization, or changing metastable atoms or molecules.

[0130] Figure 6a , Figure 6b and Figure 6c Cross-sections of three 2D patterns generated in the settings listed in Examples 1, 2, and 3 of Table 3 below, according to embodiments of the present invention, are shown. The right side of each figure is an illustration of the applied mask, which is 10 μm × 10 μm in size and has 10 nm circular vias. A standard 5 nm thick SiN wafer and a dispersion force distribution of metastable helium atoms are used. The repeating patterns are generated by vias arranged in a 3 × 3 grid with a hole spacing of 6 μm. Notably, the hole width (w) in the mask is 4 to 9 times the feature size in the generated patterns.

[0131]

[0132] Table 3: Regarding Figure 6a , Figure 6b and Figure 6c Information on the features of the 2D patterns set and generated in the examples shown. The parameters are the same as those in Tables 1 and 2.

[0133] The desired 2D pattern is periodic, which gives the following... Figure 6a , Figure 6b and Figure 6c The periodic mask structure seen in the image.

[0134] The diffraction pattern is determined by Kirchhoff's diffraction formula, which uses the transmission function of a neutral matter wave.

[0135]

[0136] Among them, the wave amplitude and the wave propagator (wave propagation operator) g X (r2-r1) from the source point r1=s,s along the distance X=L1,L2.m To the final point r2 = s m This equation must be read from right to left: the matter wave propagates along a distance L1 from a point s in the source plane to a point s in the mask plane. m The phase shift on the wavefront collected due to dispersion forces (in this case, the Casimir-Pold interaction in the non-delay limit) depends on the interaction strength C3 between the mask material and the particles, the mass m of the beam particles, and the wavelength λ. o The mask thickness d, Planck's constant h, and the mask shape as represented by the integral over the mask surface are given; the propagation then continues to a point r in the patterned target plane along the propagation length L2. The interference is described by the superposition of all light paths from the source s to the patterned target r, determined by the square of the absolute value. Since each particle is only subject to its own interference, different initial conditions, such as other source points s or different wavelengths (or particle velocities), will produce incoherent superpositions. Therefore, the interference of all states must be incoherently averaged, thus affecting the source region and the wavelength distribution f(λ) with the most probable wavelength λ and wavelength distribution Δλ. o Integrate λ, Δλ.

[0137] For tilted or curved masks and / or patterned targets, the above formulas need to be modified.

[0138] There are two integrals in the mask plane: one is for the coherent superposition s m And one is for phase shift s m '. The first integral (for s) m ) describes the optical path and is for the aperture region. Another (for s) m ') describes the interaction with the mask material and thus represents the bulk material. Figure 7 These two distinct regions of a mask cross-section with regular spherical apertures are shown. The mask material is depicted by the dark regions and the phase shift integral s is represented. m The physical aperture is represented by white and gray areas, where the latter is the absorption region, defined as the area where atoms do not transmit light and therefore do not contribute to the light path. Only the white area is associated with coherent superposition.

[0139] Kirchhoff's diffraction formula describes the most general interference case. The propagator has different approximations, leading to efficient simplifications in different systems. These are summarized in Table 4. It can be seen that, by eliminating dependencies in the denominator, the propagator simplifies to a Fourier transform in Fraunhofer and Fresnel systems, which is used in all previous publications on matter-wave etching. We consider a mask and a target (g... L2 A wave propagator between the source and the mask (g) is used to achieve resolution, taking into account the source and the mask (g) L1The Fraunhofer system between the two is used to achieve stability regarding source extension and wavelength variation.

[0140]

[0141] Table 4: Overview of wave propagators in different systems (François system, Fresnel system, wave propagator).

[0142] The properties of the mask (4), such as aperture width (w), mask extension (a), aperture arrangement, and mask material and thickness, will determine the resulting pattern. However, in real-world situations, the resulting pattern or desired pattern will be determined by the circuit design of the manufactured article (such as, for example, an integrated circuit). Therefore, a mask needs to be generated based on the desired pattern. This is often referred to as the reverse engineering problem.

[0143] As discussed above, existing binary holographic theories of scalar waves use a grid of apertures of the same size as the basis for the mask, and assume that the system size and wavelength satisfy the well-known optical conditions of the Fraunhofer approximation. In the Fraunhofer approximation, the inverse problem is simplified to a Fourier transform of the desired pattern, which produces the required mask. Sampling the mask according to the Nyquist-Shannon sampling theorem provides the bandwidth of its patterns, and therefore the minimum possible spacing. In one dimension, this can be done analytically as described in prior art. This can also be applied to matter waves if the apertures in the mask are large enough that the dispersion forces can be neglected.

[0144] However, the above does not apply to matter waves and masks with smaller apertures required for small spacing, because the dispersion force interaction between particles and diffracting objects induces a complex phase distribution of matter waves as explained above, which was not explained in the mask generation theory of the prior art.

[0145] According to an embodiment of the present invention, an alternative method for manufacturing mask structures based on machine learning is disclosed below.

[0146] A general solution to the inverse problem is approximated using a deep convolutional neural network or any model representing a differentiable, nonlinear function. This neural network is trained on a large dataset of randomly generated instances that map a mask to a pattern by applying Kirchhoff's diffraction formula as explained above. The mask can be represented as a finite-length binary sequence or array, where 1 represents a through-hole in the mask and 0 represents a blocking space. In one implementation, the pattern can also be represented by a similar binary sequence or array, where the pattern target can be defined as a fine mesh, wherein the distance between the mesh lines is smaller than the desired feature size.

[0147] Once an approximate solution (i.e., a binary mask) is obtained, it can be further improved until a solution with the desired accuracy is generated using a genetic algorithm or any other optimization algorithm based on maximizing fitness or a reward function. Both gradient-free and non-gradient-free algorithms can be used.

[0148] Genetic algorithms are stochastic classical evolutionary algorithms, meaning they dynamically change to optimize the fitness function F. Inspired by Darwin's theory of evolution, which describes how genes in a population evolve according to their ability to reproduce and mutate. In genetic algorithms, the population is not an actual biological population, but rather a set of solutions (called chromosomes) to a specific optimization problem. Furthermore, mutations are not necessarily the result of a natural random process, but occur according to different rules that can be determined during implementation. For each chromosome in the population, a fitness value can be calculated using a predetermined set of formulas chosen by the researchers. Finally, the process of life's reproduction is simulated by combining different chromosomes (parents) into the next generation of offspring. Genetic algorithms are highly robust optimization algorithms, particularly well-suited for handling sequences of categorical variables, such as, for example, the bases present in DNA. The mask used here is entirely limited to sequences or arrays of binary values.

[0149] The fitness function is the average absolute error of the generated diffraction pattern plus the reciprocal of a small numerical constant α.

[0150]

[0151] It is the absolute square of the wave function discretized on a grid of radial coordinates ri, which measures the distance to the center of the diffraction pattern. Ψ(ri; x; d) is a wave function that depends on the position of the mask opening x = (x1; x2; :::) and the thickness d = (d1; d2; :::).

[0152] Deep neural networks can be implemented in available software libraries, such as TensorFlow with Keras modules available at https: / / www.tensorflow.org / , or PyTorch.

[0153] Genetic algorithms can be implemented using pygad: https: / / pygad.readthedocs.io / en / latest / #. Other optimization algorithms can be implemented, for example, using nevergrad: https: / / facebookresearch.github.io / nevergrad / .

[0154] In the standalone implementation EC1-1, the present invention is a computer-implemented method for generating an etched mask model, comprising:

[0155] - Generate a training mask that includes vias, where the properties of the training mask are output variables of the machine learning algorithm.

[0156] - Based on the generated training mask, training patterns are produced, where the properties of the training patterns serve as input parameters for the machine learning algorithm.

[0157] - An etch mask structure is generated based on the desired pattern by feeding the desired pattern into a machine learning algorithm.

[0158] The properties of the training mask can depend on the properties of the particle source, such as wavelength, aperture diameter, velocity broadening, etc.

[0159] The etched mask model generated by this method is the model of the real mask (4) that will be generated later based on this model, while the training mask is a mask model used only for training machine learning algorithms. The desired pattern is, for example, the result of a circuit design that will occur when the real mask (4) is exposed to an incident particle beam (21) to produce the generated pattern (51). The generated pattern (51) should be as similar as possible to the desired pattern.

[0160] In the supplementary implementation EC1-2, which can be combined with EC1-1, the position, width, and shape of the vias are properties of the training mask that are generated randomly or partially randomly. Therefore, this method allows for a high degree of freedom in generating the training mask. This will also be reflected in the resulting mask model and the final mask, because meshes or repetitive structures are generally not required to generate the desired pattern, provided the pattern itself is not periodic.

[0161] In the subsidiary implementation EC1-3, which can be combined with EC1-1 or EC1-2, the machine learning algorithm includes a supervised machine learning algorithm.

[0162] In the subsidiary implementation EC1-4, which can be combined with any of EC1-1 to EC1-3, the supervised machine learning algorithm includes a deep neural network.

[0163] In an adjunct implementation EC1-5, which can be combined with any of EC1-1 to EC1-4, the computer-implemented method for generating an etch mask model includes manufacturing a mask (4) included in any of ES1 to ES40 based on the generated etch mask model.

[0164] After the mask structure has been generated, the physical mask can be manufactured based on the parameters obtained from the aforementioned machine learning process. Mask production can be performed using prior art, such as by etching holes, and is not within the scope of this application. However, as stated above, the present invention allows for the use of relatively larger hole sizes compared to prior art to obtain small feature sizes, which makes mask production easier, resulting in faster and cheaper mask production.

[0165] In the subsidiary implementation EC1-6, which can be combined with any of EM1-1 to EM1-5, the computer-implemented method for generating an etch mask model includes manufacturing a mask (4) included in any of ES1 to ES40 based on the generated etch mask model.

[0166] In illustrative embodiments, various features and details are shown in combination. The fact that several features are described for a particular instance should not be construed as implying that those features must be included together in all embodiments of the invention. Rather, features described with reference to different embodiments should not be construed as mutually exclusive. As will be readily understood by those skilled in the art, the inventors have considered any subset of features described herein that are not explicitly interdependent and are intended to be part of the disclosure. However, explicitly describing all these embodiments does not contribute to understanding the principles of the invention, and therefore, for simplicity or brevity, certain arrangements of features have been omitted.

Claims

1. An etching pattern generation system (1), comprising: - An incident particle beam (21) with wavelength (λ) from particle source (2), and - A mask (4) including a through-hole (3) disposed in the incident particle beam (21), wherein, The etching pattern generation system (1) is configured to generate a pattern (51) on a pattern target (5), wherein particles in the incident particle beam (21) arrive at the mask (4) as incident particle waves, and wherein the mask (4) is configured to induce a phase shift on the wavefront of the particle waves as the particle waves propagate through the mask.

2. The etching pattern generation system (1) according to claim 1, wherein, For a distance of 0.016 times the minimum aperture width (w) of the through hole (3) from the mask plate (4), the induced phase shift is at least π, 3 / 2π or 5 / 3π.

3. The etching pattern generation system (1) according to claim 1 or 2, wherein, The etching pattern generation system (1) is configured to generate a pattern (51) on the opposite side of the incident particle beam (21) at a target distance (L2) from the mask (4).

4. The etching pattern generation system (1) according to any one of claims 1 to 3, wherein, The parameters of aperture width (w), mask material, and mask thickness are selected to induce a phase shift in the incident particle wave by modifying the dispersion force interaction between the incident particle beam (21) and the mask (4).

5. The etching pattern generation system (1) according to any one of claims 1 to 4, wherein, The mask (4) is doped or implanted with any one of neutral particles, ions and electrons to alter the dispersion force interaction, thereby changing the phase shift on the wavefront.

6. The etching pattern generation system (1) according to any one of claims 1 to 5 is configured to reduce the background signal more than the diffraction order compared to the diffraction of light waves propagating through the same mask (4) in the same direction along the axial direction of the via (3), thereby increasing the portion of the beam that passes through the diffracted rather than straight path and increasing the contrast (c) of the pattern (51).

7. The etching pattern generation system (1) according to any one of claims 1 to 6, wherein, The through-hole (3) has a minimum aperture width (w) and the mask (4) has a mask extension (a), the mask extension (a) being the maximum distance between two through-holes in the mask (4) exposed by the incident particle beam (21), wherein:

8. The etching pattern generation system according to any one of claims 1 to 7, wherein, The particle source (2) has an opening diameter (da) smaller than the source distance (L1) between the particle source (2) and the mask (4).

9. The etching pattern generation system according to any one of claims 1 to 8, wherein, The mask extension (a) is less than the source distance (L1) between the particle source (2) and the mask (4).

10. The etching pattern generation system (1) according to any one of claims 1 to 9, wherein, The minimum aperture width (w) is at least two, three, or four times the minimum feature size of the pattern (51).

11. The etching pattern generation system (1) according to any one of claims 1 to 10, wherein, The mask (4) is a binary holographic mask.

12. The etching pattern generation system (1) according to any one of claims 1 to 11, wherein, The mask (4) is made of dielectric material.

13. The etching pattern generation system (1) according to any one of claims 1 to 12, wherein, The mask (4) includes silicon nitride and Si3N. x Or Si3N4.

14. The etching pattern generation system (1) according to any one of claims 1 to 13, wherein, The through-hole (3) is not located at a non-periodic position on the mask (4) according to a predetermined underlying periodic grid structure.

15. The etching pattern generation system (1) according to any one of claims 1 to 14, wherein, The through hole (3) has a non-uniform minimum hole width (w).

16. The etching pattern generation system (1) according to any one of claims 1 to 15, wherein, The through-hole (3) has a minimum hole width (w) of 5nm, 4nm, 3nm, 2nm or 1nm.

17. The etching pattern generation system (1) according to any one of claims 1 to 16, wherein, The through hole (3) has a maximum hole width (w) of 100 nm.

18. The etching pattern generation system (1) according to any one of claims 1 to 17, wherein, The through hole (3) has a different shape.

19. The etching pattern generation system (1) according to any one of claims 1 to 18, wherein, The center-to-center distance (c) between adjacent through holes (3) is not uniform.

20. The etching pattern generation system (1) according to any one of claims 1 to 19, wherein, The mask thickness is less than 500nm, 100nm, 50nm, 20nm, 10nm or 7nm.

21. The etching pattern generation system (1) according to any one of claims 1 to 20, wherein, The mask thickness is less than the minimum aperture width (w).

22. The etching pattern generation system (1) according to any one of claims 1 to 21, wherein, The particle beam (21) comprises metastable atoms or molecules.

23. The etching pattern generation system (1) according to any one of claims 1 to 22, wherein, The particle beam (21) comprises inert gas atoms.

24. The etching pattern generation system (1) according to any one of claims 1 to 23, wherein, The particle beam (21) includes helium atoms.

25. The etching pattern generation system (1) according to any one of claims 1 to 24, wherein, The wavelength (λ) is less than any of the minimum feature size (df), minimum spacing, minimum aperture width (w), and mask thickness.

26. The etching pattern generation system (1) according to any one of claims 1 to 25, wherein, The wavelength (λ) is less than 1.0 nm or 0.5 nm.

27. A method for generating a pattern (51) by etching, comprising: - Expose a mask (4) including a through hole (3) to an incident particle beam (21) from a particle source (2), wherein particles in the incident particle beam (21) arrive at the mask (4) as particle waves, and wherein the mask (4) is configured to induce a phase shift of the particle waves as the particle waves propagate through the mask.

28. The method for generating a pattern (51) by etching according to claim 27, comprising: - The mask (4) is placed in front of the pattern target (5) and a pattern (51) is generated on the pattern target (5).

29. The method for generating a pattern (51) by etching according to claim 27 or 28, wherein, A patterned target (5) is positioned at a target distance (L2) from the mask, the through-hole (3) has a minimum aperture width (w) and has a mask extension (a) located between the through-holes (3) exposed by the incident particle beam (21), wherein:

30. The method for generating a pattern (51) by etching according to any one of claims 27 to 29, wherein, The through-hole (3) has an effective aperture width (w) smaller than the minimum aperture width (w) due to the dispersion force between the incident particle beam (21) and the mask (4). e ).

31. A computer-implemented method for generating an etched mask model, comprising: - Generate a training mask including vias, wherein the properties of the training mask are output variables of a machine learning algorithm. - A training pattern is generated based on the resulting training mask, wherein the properties of the training pattern are input parameters of the machine learning algorithm. - The etching mask model is generated based on the desired pattern by feeding the desired pattern into the machine learning algorithm.

32. The computer-implemented method for generating an etched mask model according to any one of claims 31 to 32, wherein, The location, width, and shape of the via are properties of the training mask that are randomly or partially randomly generated.

33. The computer-implemented method for generating an etched mask model according to any one of claims 31 to 33, wherein, The machine learning algorithms include supervised machine learning algorithms.

34. The computer-implemented method for generating an etched mask model according to any one of claims 31 to 34, wherein, Supervised machine learning algorithms include deep neural networks.

35. A computer-implemented method for generating an etched mask model according to any one of claims 31 to 34, comprising manufacturing the mask (4) included in any one of claims 1 to 22 based on the generated etched mask model.

36. The method of generating a pattern (51) by etching according to any one of claims 27 to 30, comprising manufacturing a mask based on a mask model generated in any one of claims 31 to 35.

Images