research papers
Demonstration of a ring-FEL as an EUV lithography tool
aPohang Accelerator Laboratory, POSTECH, Pohang, Gyungbuk 37673, Republic of Korea, bDepartment of Physics, POSTECH, Pohang, Gyungbu 37673, Republic of Korea, and cSLAC National Accelerator Laboratory, Menlo Park, CA 94025, USA
*Correspondence e-mail: tlssh@postech.ac.kr
This paper presents the required structure and function of a ring-FEL as a radiation source for extreme ultraviolet radiation lithography (EUVL). A 100 m-long straight section that conducts an extremely low emittance beam from a fourth-generation storage ring can increase the average power at 13.5 nm wavelength to up to 1 kW without degrading the beam in the rest of the ring. Here, simulation results for a ring-FEL as a EUVL source are described.
Keywords: fourth-generation storage ring; EUV; free-electron laser.
1. Introduction
Extreme ultraviolet radiation lithography (EUVL) is a candidate to succeed 193 nm immersion lithography to pattern critical layers during the manufacture of integrated circuits. Therefore, radiation sources for EUVL are currently being developed. Advances in the power and reliability of laser-produced plasma (LPP) sources have enabled their application to next-generation technology development programs, but slow progress in LPP sources has delayed the adoption of EUV technology in high-volume manufacturing. In addition, increasing source requirements are driving the source community to consider pushing LPP technology beyond its current target of 250 W, or even higher.
A free-electron laser (FEL) (Kim, 1986) that uses a storage ring (ring-FEL) (Murphy & Pellegrini, 1985) is a candidate high-brightness radiation source for EUVL. A ring-FEL has the advantage of high average power, but electron bunches in the storage ring have a low and a high energy spread. Therefore, many techniques have been developed to obtain FEL lasing by using the storage-ring electron beam. Methods to overcome the low include the use of a transverse gradient undulator (TGU) (Smith et al., 1979) and the application of bunch compression to increase the (Feng & Zhao, 2017). However, those techniques must bypass the electron beam from the ring; this strategy reduces the FEL repetition rate, whereas EUVL requires a high average radiation power, so the high repetition rate of a storage ring is a great advantage. Recent work on steady-state self-amplified FELs proved that a bypass is not necessary for the FEL in a storage ring at EUV and soft X-ray wavelengths (Huang et al., 2008). Here we show that sufficient EUV FEL lasing for EUVL can be obtained using the electron bunch of a low-emittance ring without beam manipulation or a bypass beamline.
A fouth-generation storage ring (4GSR) that uses a multi-bend achromat (MBA) lattice concept may be able to surpass the et al., 2013; Steier et al., 2016; see also MAX IV Conceptual Design Report, https://www.maxlab.lu.se/maxlab/max4/index.html; ESRF-EBS, https://indico.psi.ch/conferenceDisplay.py?confId=5589; APS Upgrade, https://www1.aps.anl.gov/aps-upgrade; Spring-8-II Conceptual Design Report, https://rsc.riken.jp/eng/index.html).
and coherence that are attained using present third-generation storage rings (3GSRs). 4GSRs have decreased beam emittance to a few hundred picometers, or even to ≤100 pm (LiuA ring-FEL that uses a 4GSR has extremely low electron-beam emittance, and can therefore generate high average power that is suitable for EUVL. However, uses for EUVL are impeded by relatively large energy spread and low introduces the characteristics of the PAL-4GSR lattice. Section 3 describes the ring-FEL output characteristics. Section 4 presents conclusions.
By applying a 10 nC camshaft bunch, a low-emittance electron beam with 250 A is successfully generated. As a result, the average power at 13.5 nm wavelength from a ring-FEL has been increased to 1 kW. In this paper, we describe the start-to-end simulation of a ring-FEL as a beam source for EUVL. Section 22. PAL-4GSR lattice design for a ring-FEL
The PAL-4GSR (Kim et al., 2019) storage ring (Table 1) is a hybrid seven-bend achromat (H7BA) lattice with a horizontal emittance of 90 pm. The ring is 570 m in circumference and is composed of 20 symmetrical cells. The straight section is 6.5 m long to accommodate two SCRF modules. The PAL-4GSR lattice contains a 2 T super-bend magnet in the central dipole to produce radiation that has a of 12 keV.
|
ESRF-II and APS-U lattices were adopted in the PAL-4GSR lattice. The dispersion was deliberately enlarged between the first and second dipoles, and between the sixth and seventh dipoles. All chromatic sextupoles were located in this dispersion bump region to reduce the strength required to control the chromaticity. The betatron phase advances between the two dispersion bumps were set to ≃ 3π in the horizontal plane and ≃ π in the vertical plane. As a result, non-chromatic effects of the sextupoles are cancelled out naturally. To minimize natural emittance, five-step longitudinal gradient dipoles and reverse bending magnets were used.
The storage-ring RF system includes two SC cavities, each driven by a klystron that delivers a continuous wave of up to 240 kW at 500 MHz. The accelerating voltage in the main RF cavity is 2.15 MV to realize a bucket height of ∼5%. The stored beam has RMS bunch length of 8 mm with the assumption of a third harmonic cavity for bunch prolongation, and natural energy spread of 0.1%. When the beam current is 400 mA in multi-bunch operation mode, the emittance increase due to intra-beam scattering is 10% in round-beam mode.
The ring-FEL option of PAL-4GSR requires two 100 m-long straight sections for FEL undulators (Fig. 1). Inserting the two long straight sections in the PAL-4GSR nominal lattice changes the PAL-4GSR shape from ring to racetrack; the circumference becomes 770 m and the revolution time increases to 2.57 µs. To obtain high without any bunch compressing, we increase the bunch charge to 10 nC and turn off the third-harmonic cavity. Bunch length is slightly increasing along the bunch charge, so for the ring-FEL option we use bunches with a length of 4.2 nm and total charge of 10 nC. is ∼280 mA and energy spread is 0.1%. Due to intra-beam scattering (IBS) (Kubo et al., 2005), a high-charge bunch has rather high emittance (Fig. 2). To reduce IBS, we use round-beam mode. The emittance of a zero-charge beam is set to 58 pm for the round-beam mode. The emittance of the 10 nC round beam is ∼133 pC and we set the beam emittance as 200 pC to allow margin for error. A charge of 10 nC translates to a bunch current of 4 mA. We set the filling pattern for the ring-FEL option as 80 camshaft bunches, so then the total current is 320 mA. Considering the heating effect of the camshaft bunches, we choose a total current that is lower than the nominal operation mode. As a result, the FEL repetition rate is 31.2 MHz, which greatly exceeds that of an XFEL, which is ∼100 Hz for normal conducting RF and 1 MHz for superconducting RF, so a ring-FEL has a great advantage.
We should consider that using a 100 m-long undulator consumes large RF power. The total power from the undulator spontaneous radiation is calculated as
where E [GeV] is electron energy and B0 [T] is undulator peak field, L [m] is the total undulator length and I [A] is the beam current. In our case, the FEL undulator has Po = 382.8 kW. Considering that one superconducting RF module provides 240 kW, a ring-FEL option requires two additional superconducting RF modules.
3. Ring-FEL output characteristics
The required FEL power to achieve 1 kW average EUV power can be calculated from the measured bunch length and repetition rate. A bunch length of 4.2 mm and repetition rate of 31.2 MHz yield a bunch averaged FEL power of 0.916 MW. In a conventional linac X-ray FEL, bunch-averaged FEL power is on the order of gigawatts, so the required FEL power for the EUVL in a ring-FEL is relatively low because of the high repetition rate of the storage ring. Despite the large energy spread and low ).
of the electron beam in the storage ring, a megawatt-level FEL is easily achievable in the EUV regime if a 100 m undulator is used (Fig. 3Due to the high repetition rate of the ring-FEL option of PAL-4GSR, its parameters (Table 2) enable an average FEL power that is sufficiently high at the saturation point. To reduce degradation of electron beam quality and the undulator length, we will stop the FEL lasing before the saturation point. Currently, the demanding laser wavelength for the EUVL is 13.5 nm. However, in the near future a reduction in wavelength will be needed; therefore, we consider use of a variable gap undulator to change the FEL wavelength from 13.5 nm to 7 nm. This decrease in wavelength will decrease FEL power and increase the saturation length, but the average power will remain >1 kW. Assuming that the electron beam parameters of the storage ring are invariant including the FEL undulator, an average power of 100 kW is achieved before 100 m. However, the electron beam parameter is affected by the FEL radiation and we should identify the equilibrium condition of the ring-FEL configuration.
|
The FEL simulation was conducted using the code Genesis 1.3 (Reiche, 1999). Time-dependent simulation shows the evolution of the microbunching. The electron beam began to form microbunches at >135 m from the undulator (Fig. 4). Microbunching is the key to FEL lasing, but can damage a storage ring.
Energy loss and increase in energy spread also degrade the electron beam of the FEL lasing. The energy loss of the 3 GeV electron beam is 1.5 keV per turn, which is negligible compared with the energy loss of the FEL (∼1 MeV per turn) (Fig. 5). Of this 3 GeV electron beam, 343 keV per turn is caused by bending-magnet radiation except the insertion devices. The relative energy spread is increased 0.0001% per one turn by FEL generation (Fig. 6). Therefore, we will reduce the harmful effect on the storage ring by extracting the electron beam in the undulator before saturation. The bunch-averaged FEL power along the undulator begins to increase exponentially at ∼40 m; the energy spread also begins to increase, but remains low (Fig. 7). Undulator length is determined by the required EUV radiation power, not by the saturation length.
To examine the effect of the FEL undulator on the storage ring, a full particle tracking is needed for several thousand turns. In addition, at least 1000 particles per 13.5 nm must be tracked to deliver the microbunching property. The RMS bunch length is 4 mm and the FEL wavelength is 13.5 nm, so more than one million microbunches occur. If we track 1000 particles per microbunch, then one billion particles must be tracked. Therefore, a full particle tracking for several thousand turns requires too much computing power. We rather examine the microbunching effect of one microbunch in the
position and examine the energy and the energy spread effect of the FEL to the rest of the ring. We also examine the emittance change by the FEL through the damping wiggler theory.We used the code Elegant (Borland, 2000) to conduct a particle-tracking simulation of the ring, excluding the FEL part. The microbunching is washed out before the first turn (Huang et al., 2008), so we focused on the energy and energy spread effect. The energy-loss effect that was described in the previous section is negligible compared with that in a normal wiggler. The energy spread is the key to the interaction between the ring and the FEL; this effect has been well investigated previously (Huang et al., 2008). The maximum radiation power is determined by the electron beam power, the Pierce parameter, the longitudinal damping time, and the energy spread without the FEL. In our case, the maximum radiation power is ∼1 kW which is our target FEL power. Storage-ring theory suggests that equilibrium energy spread is determined when the quantum excitation effect and the longitudinal damping effect have the same strength but opposite directions. The equilibrium energy spread considering the FEL effect is estimated as (Huang et al., 2008)
where is the relative energy spread in the absence of the FEL and is an increase of the square of the relative energy spread by single FEL lasing, [ms] is the longitudinal damping time and T0 [µs] is the revolution time. In our case, = 10-3, = 2×10-9, = 10 ms, T0 = 2.57 µs. The calculated equilibrium relative energy spread was = 2.2×10-3, which noticeably degrades the beam property, and thereby reduces the FEL power and increases the saturation length. Because of this energy spread increase, we introduce the seeded FEL scheme to compensate for the increase in energy spread.
To achieve 1 kW power, we conducted two FEL simulations. Both started with 10 kW pulsed seeding power, but the energy spread was 0.1% in the first simulation and 0.2% in the second. With 0.1% energy spread (Fig. 7), the average FEL power was 1 kW at about z = 63 m and energy spread increased at z = 63 m to ∼0.0001%, which is the same as in the no-seeding case. In contrast, with 0.2% energy spread (Fig. 8), which is expected for the equilibrium energy spread, 1 kW power was achieved at z = 80 m but the energy spread increased by only 0.00003%, and the energy spread started to decrease after z = 60 and became close to the initial value at z = 100 m. After z = 60 m, the energy spread decreased but the FEL power increased linearly. This result indicates that we could obtain 1 kW EUV FEL power while the equilibrium energy spread is below 0.2% with 10 kW pulsed seeding power and a 100 m-long undulator.
Many 3GSRs implement a damping wiggler to decrease the emittance (Jang et al., 2020). A wiggler with a strong field in a non-dispersive section helps to reduce the equilibrium emittance. One can expect that the FEL undulator effects the storage ring similar to the damping wiggler. In our case, the undulator period is 4.4 cm and the peak magnetic field is 1.54 T; therefore the emittance reduction rate is ∼0.3 and the equilibrium emittance will change from 200 pm to 60 pm (Fig. 9).
In this simulation, we achieved 1 kW EUV power by using the seeded FEL scheme. A seeded FEL requires a source to seed the radiation. The required seeding radiation is 1 kW pulsed radiation that is synchronized with the electron beam. The average power of the seeding radiation is ∼40 W. The seeding methods of the seeded FEL are external seeding or self-seeding. For external seeding, a laser is generally used. If the external seeding source is difficult to obtain, the self-seeding scheme can be used (Feldhaus et al., 1997). When a self-seeding scheme is used, the requirements for repetition rate and pulse length are easily achievable. Requirements for the external seeding laser are: 31 MHz repetition rate, 10 kW peak power, and 4 mm RMS pulse length. Currently, these requirements are not achievable. Therefore, two self-seeding schemes are considered to generate seeding radiation satisfying the requirement.
The traditional self-seeding scheme can be applied by adding another long undulator in the opposite long straight section of the race-track structure; one long undulator is for seeding radiation and the other is for the seeded FEL. The seeding FEL undulator is 60 m long and generates 10 kW FEL without degrading the energy spread.
Another scheme is to use a part of the FEL radiation as a seeding source. This approach is similar to the regenerative amplifier scheme (Nguyen et al., 1999), but our scheme reflects only a small fraction of the FEL radiation. FEL peak power is >2.5 MW and the required seeding peak power is ∼10 kW, so the mirror reflects only 0.4% of the FEL radiation power and transmits the rest. To achieve timing, the number of camshaft bunches should be an integer divisor of the harmonic number of the storage rings. To further increase the FEL power, the undulator tapering method can be considered.
4. Conclusion
The ring-FEL option of the PAL-4GSR is investigated for use in EUV lithography. The PAL-4GSR is a 3 GeV, 570 m, 58 pm low-emittance ring that is in its design stage. The ring-FEL option requires a change from ring structure to race-track structure with two 100 m-long straight sections. It also requires a dedicated operation mode that has 80 camshaft bunches with 10 nC single-bunch charge. The beam emittance of the camshaft bunch is increased by intra-beam scattering, but the damping wiggler effect of the FEL undulator reduces the beam emittance. The equilibrium beam emittance is <200 pm, so the ring-FEL dedicated mode is usable from the other storage-ring beamlines. Considering the energy spread increase from the FEL, we conclude that the seeding power is critical. With 10 kW seeding power, we achieved an average 1 kW EUV radiation as a EUV lithography source.
Funding information
This research was supported by the the Basic Science Research Program through the National Research Foundation of Korea (NRF-2019R1C1C1003412) and the Basic Science Research Program through the National Research Foundation of Korea (NRF-2019R1A2C1004862).
References
Borland, M. (2000). Elegant: A Flexible SDDS-Compliant Code for Accelerator Simulation. Technical Report LS-287. Advanced Photon Source, Argonne, IL, USA. Google Scholar
Feldhaus, J., Saldin, E. L., Schneider, J. R., Schneidmiller, E. A. & Yurkov, M. V. (1997). Opt. Commun. 140, 341–352. CrossRef CAS Web of Science Google Scholar
Feng, C. & Zhao, Z. (2017). Sci. Rep. 7, 4724. Web of Science CrossRef PubMed Google Scholar
Huang, Z., Bane, K., Cai, Y., Chao, A., Hettel, R. & Pellegrini, C. (2008). Nucl. Instrum. Methods Phys. Res. A, 593, 120–124. Web of Science CrossRef CAS Google Scholar
Jang, G. et al. (2020). J. Synchrotron Rad. Submitted. Google Scholar
Kim, J., Jang, G., Yoon, M., Oh, B., Lee, J., Ko, J., Parc, Y., Ha, T., Kim, D., Kim, S. & Shin, S. (2019). Phys. Rev. Accel. Beams, 22, 011601. Web of Science CrossRef Google Scholar
Kim, K. J. (1986). Phys. Rev. Lett. 57, 1871–1874. CrossRef PubMed CAS Web of Science Google Scholar
Kubo, K., Mtingwa, S. K. & Wolski, A. (2005). Phys. Rev. ST Accel. Beams, 8, 081001. Web of Science CrossRef Google Scholar
Liu, L., Milas, N., Mukai, A. H. C., Resende, X. R., Rodrigues, A. R. D. & de Sá, F. H. (2013). Proceedings of the Fourth International Particle Accelerator Conference (IPAC2013), 12–17 May 2013, Shanghai, China, pp. 1874–1876. TUPWO001. Google Scholar
Murphy, J. B. & Pellegrini, C. (1985). Nucl. Instrum. Methods Phys. Res. A, 237, 159–167. CrossRef Web of Science Google Scholar
Nguyen, D. C., Sheffield, R. L., Fortgang, C. M., Goldstein, J. C., Kinross-Wright, J. M. & Ebrahim, N. A. (1999). Nucl. Instrum. Methods Phys. Res. A, 429, 125–130. Web of Science CrossRef CAS Google Scholar
Reiche, S. (1999). Nucl. Instrum. Methods Phys. Res. A, 429, 243–248. Web of Science CrossRef CAS Google Scholar
Smith, T. I., Madey, J. M. J., Elias, L. R. & Deacon, D. A. G. (1979). J. Appl. Phys. 50, 4580–4583. CrossRef CAS Web of Science Google Scholar
Steier, C., Byrd, J. M., De Santis, S., Nishimura, H., Robin, D., Sannibale, F., Sun, C., Venturini, M. & Wan, W. (2016). Proceedings of the 7th International Particle Accelerator Conference (IPAC2016), 8–13 May 2016, Busan, Korea, pp. 2956–2958. WEPOW049. Google Scholar
© International Union of Crystallography. Prior permission is not required to reproduce short quotations, tables and figures from this article, provided the original authors and source are cited. For more information, click here.