Key Parameters Demonstrated for High-Volume EUV Lithography Sources

Extreme ultraviolet (EUV) lithography is currently the leading option to replace ArF immersion lithography for high-volume manufacturing below the 32 nm node. Development tools will be required as early as 2009 to meet the timeline for introduction of EUV lithography at these nodes. The critical challenges for the technology are source power and lifetime, resist resolution and sensitivity, and photomask defect density and protection.

High source power is required to support high wafer throughput and meet cost-of-ownership targets. The requirements for source power have increased over time as it has become clear that high resist dose is needed to simultaneously meet resolution and linewidth-roughness (LWR) targets. It is estimated that EUV power of >115 W is required for 5 mJ/cm² photoresist speed to enable >100 wph scanner throughput, and 180 W is needed for 10 mJ/cm². Photoresist sensitivities above 20 mJ/cm² could drive power requirements well above the 200 W level.

It is clear that a scalable EUV source architecture is needed to enable the evolution of EUV lithography during the lifecycle of the technology. Although experimental EUV exposure tools currently use low-power discharge-produced plasma (DPP) sources, it is unlikely that these will allow scaling to the power levels needed for high-volume manufacturing. The laser-produced plasma (LPP) source is the most promising technology to deliver the high power output needed for the production of ICs in the post-193 nm immersion era.

LPP sources generate EUV (13.5 nm) radiation by depositing laser energy into a source element, such as xenon, tin or lithium, creating a highly ionized plasma. The energetic radiation generated during de-excitation and recombination of these ions is emitted in all directions, but collected and focused to an intermediate focus (IF) point at the entrance to the scanner optical system.

Radiation is emitted over a wide range of wavelengths, but high conversion efficiency (CE) of the laser energy into in-band EUV energy is critical to meeting the required power levels. Studies of CE for several combinations of laser wavelength and source element have shown that the combination of CO₂ laser radiation and tin droplet targets provides the optimum efficiency. A production-worthy source requires development and integration of several technologies, such as the CO₂ laser, tin droplet generator, collection optics and debris mitigation technology, and progress in these areas is described below.

Laser

We use a CO₂ laser in an RF-pumped master oscillator power amplifier (MOPA) configuration with multiple stages of amplification. The seed pulse is initiated by a Q-switched master oscillator with low energy and high repetition rate, capable of 100 kHz operation. The laser pulse is amplified, shaped and focused before entering the LPP chamber. The laser pulses are timed to impinge on an incoming liquid tin droplet stream flowing from a target delivery system at a repetition rate comparable to or larger than the laser. Steering is provided to an exact point in the vacuum chamber where the droplet is irradiated by the laser beam. The peak emission of the plasma occurs at 13.5 nm, and matches well to the reflectivity curve of multilayer-coated mirrors designed for this wavelength. In the EUV region, the coating acts as a spectral purity filter, reflecting light mainly near 13.5 nm.

The size of the EUV emitting region of the micro plasma has been measured using a pinhole camera. The plasma has overall cylindrical symmetry with respect to the optical axis with on-average ellipsoidal, near-spherical shape. This results in a small value of etendue (~0.1 mm² sr), which can simplify the design of optical illumination devices, thus enabling the implementation of high numerical aperture (NA) and high-throughput power systems. This is a significant advantage of LPP sources over DPP sources, which typically have much larger etendue.

Droplet generator

The main requirements for the droplet generator are the capability to produce tin droplets at a controllable frequency with high droplet uniformity and stability, and to provide reliable operation over long periods of time. We currently have a third-generation version of droplet generator in operation. Stable droplets have been obtained over a range of frequencies of 20–500 kHz, with corresponding droplet diameters of 150–20 μm.

Droplets at the minimum size contain only ~10¹⁴ atoms, approaching the condition of the minimal mass required for efficient EUV generation and minimal debris generation. The total tin consumption in this high-frequency regime of operation is estimated to be only 120 mL/day. When equipped with a small-diameter nozzle, the droplet generator can produce stable tin droplets over extended periods of time. So far, 55 hours of continuous generator operation have been demonstrated, limited only by the currently installed volume of tin supply. The next-generation droplet generator will have continuous operation capability.

The stability of the droplets produced by the droplet generator is determined by the nozzle design, the parameters of the droplet generator (e.g., applied pressure and actuator driving voltage) and the overall mechanical stability of the system. Using a new nozzle design, substantial progress has recently been achieved with respect to the timing jitter of the droplets. A timing jitter of 25 nsec (0.2% of the droplet period) has been demonstrated without using any active stabilization techniques. At the plasma location, this corresponds to a spatial uncertainty of the droplet along the jet trajectory of ±0.6 μm. The short-term positional (transverse) stability of the droplets is on the order of 5 μm, whereas the long-term (~5 min) stability is in the 100 μm range. The latter is routinely compensated by the active stabilization steering system. Operation of the droplet generator in both vertical and horizontal orientations has been demonstrated, which provides the maximum flexibility for integration with the scanner illuminator.

Collector

EUV radiation emitted by the plasma is collected and directed to the IF point by a large normal-incidence ellipsoidal multilayer mirror (MLM). In our light-source configuration, the laser beam is focused through a central opening in the collector mirror onto the droplet target for plasma generation. The EUV radiation emitted in the backwards direction is reflected at near-normal incidence by the mirror and focused to the IF. An advantage of this configuration is that a massive collector shell with large thermal load capacity and low deformation potential can be employed with thermal management from the backside. A schematic view of this configuration is shown in Figure 1.

1. This schematic shows a light collection configuration with a sub-aperture (1.6 sr) or full-aperture (5 sr) mirror.

The infrastructure for manufacturing these large normal-incidence collector mirrors has been developed. The light collection capability of such mirrors has been demonstrated using a 1.6 sr sub-aperture version (320 mm optical diameter) of the full high-volume design (with 5 sr collection angle). After machining and super-polishing, the mirror surface is coated using a deposition tool with DC magnetron sputtering technology capable of depositing a graded high-temperature-stable coating with interface-engineered multilayers. This coating was designed to provide high EUV reflectivity at various angles of incidence. Multiple sub-aperture mirrors for lifetime testing have been completed, and the full >5 sr aperture manufacturing process has begun. Mirrors with good peak reflectivity and uniformity, close to the target of 50% average reflectivity, have been obtained.

Debris mitigation

LPP EUV sources generate debris in the form of high-energy ions, neutral atoms and clusters of target material, which impacts the surface of the collector. Of these three types of debris, the most hazardous for the collector mirror coating is the ion flux. The amount of neutral atoms and clusters from the droplet target impinging onto the collector is small because most of the target material moves in a direction pointing away from the collector surface. The normal-incidence mirror must be protected from the plasma by advanced debris mitigation technology to maintain the reflectivity of the collector mirror and enable long lifetime for this component and low cost of operation.

The interaction of ions (with energies of a few keV) with the mirror surface results in erosion of the material of the MLM coating. For lifetime extension of the collector, a mitigation technique for the energetic ions has been developed and tested. The technique provides a suppression of the ion flux by four orders of magnitude (Fig. 2). Moreover, the maximum observed ion energy in the spectrum is reduced from 3 keV to <300 eV. Based on a measured erosion rate of 0.2 layers/Mpulses without debris mitigation, and the suppression factor of 10⁴, it is estimated that a collector mirror coating fabricated with ~500 sacrificial layers will have a lifetime exceeding 10¹² pulses. This corresponds to about one year of operation of the collector mirror.

2. Debris mitigation can suppress ion flux by four orders of magnitude.

System integration

Presently, two LPP development systems using a high-power, high-repetition-rate CO₂ laser are operational. Over the past year, one of these chambers has been used to demonstrate substantial increases in power output from droplet targets obtained primarily by continuous improvement of the laser intensity generated on the target droplet. This has been achieved by a number of advances in the areas of laser output power, laser beam quality, beam and pulse shaping, and focusing optics. For this data, the IF-equivalent power is determined from the in-band EUV power measured at the plasma using EUV monitors with small input apertures at well-defined distances from the plasma, and extrapolating to a collection angle of 5 sr, 50% collector average reflectivity and 90% transmission along the optical path from the plasma to IF.

The second chamber (Fig. 3) can handle testing under high-duty-cycle conditions, and integration of the collector and the advanced debris mitigation system to allow long-term measurements at IF. The key challenge is to maintain high CE by optimizing the droplet and laser beam characteristics as duty cycle and running time are increased. This chamber also allows for inline connection to illuminator test modules for initial exposure system illuminator testing.

3. This LPP development chamber can handle testing under high-duty-cycle conditions, and integration of the collector and debris mitigation system.

Figure 4 shows the most recent data obtained in the new chamber during burst-mode operation at a 50 kHz repetition rate on tin droplets. A sequence of 300 bursts of 1 msec length with the system in a free-running operation mode without active control is shown, producing IF-equivalent output power of ~100 W. The equivalent average power of ~5 W realized at IF, obtained during a period of 100 sec at 50 kHz pulse frequency, is also shown. The next phase of development will focus on increasing average power to 100 W at IF, primarily through duty-cycle improvements, and measuring this power directly behind the IF.

4. The most recent data from the new chamber shows burst power (IF-equivalent) at 50 kHz on tin droplets of ~100 W (left), and an equivalent average power of ~5 W (right).

Our LPP source roadmap is shown in the Table. The Generation 1 product is designed to meet requirements for pre-production or beta-generation scanners in 2009. For this product, the in-band EUV output power is targeted to be 100 W, using a 10.8 kW CO₂ laser system with tin droplets delivering 3.0% CE. This level of performance has already been demonstrated in burst mode (Fig. 4). The normal-incidence collector will have 5 sr collection angle and a coating with an average EUV reflectivity of ~50%. Sub-aperture collectors have already been fabricated with performance close to this target. Transmission losses caused by obscurations, absorption and debris mitigation techniques are projected to be <10%.

In line with the expected requirements for increased in-band EUV power at IF, the second and third generation of LPP EUV sources will be brought to market using higher-power CO₂ laser technology and moderate improvements in conversion and collection efficiencies. It is expected that the time between insertion of each product generation will be about 24 months.

Summary

EUV lithography is the leading option for critical-layer imaging beyond 193 nm immersion and double patterning. One of the key challenges for the successful introduction of EUV lithography is the development of a high-power, long-lifetime source. LPP EUV light sources have been shown to be the most promising source technology to scale power to meet the throughput requirements as the lithography tools evolve over their life cycle. We have chosen an LPP architecture using the high-CE combination of 10.6 μm CO₂ laser radiation and a tin source element, which has shown CE well in excess of 4%.

High CE is the key to a cost-effective solution for chipmakers. High-power, high-repetition-rate CO₂ laser technology has been validated and is operational above the 12 kW level. High-CE operation has been demonstrated on droplet targets, and measured. EUV burst power of 300 W at the plasma and 100 W at IF has recently been achieved. This is the target power level needed for first-generation source products to support pre-production EUV exposure tools. Additional development over the next year will focus on increasing average power to the 100 W level by increasing duty-cycle scaling.

Debris mitigation techniques have been developed and shown to suppress energetic ions, which can erode collection optics and reduce reflectivity. It is estimated that the lifetime of the collector coating can reach 1012 pulses, or one year of operation. A sub-aperture normal-incidence collector mirror of 320 mm diameter with high-temperature graded multilayer coating has been developed, including the fabrication infrastructure. Over the next year, the collection optics, together with debris mitigation technology, will be integrated and tested. Early versions of the integrated system are targeted for delivery by the end of 2008.