 James Underwood and Eric Gullikson used the Calibration and Standards Beamline at the Advanced Light Source to measure scattering from multilayer mirrors. | Photo by Roy Kaltschmidt |
By stacking ring-shaped mirrors of different curvatures--sections of paraboloids and hyperboloids--x-rays can be coaxed through successive grazing angles to a tight focus. To get normal-incidence reflection at short wavelengths, however, requires a trick: interference.
Interference is a fundamental consequence of the wave nature of light. Waves that are out of phase interfere destructively and are canceled; waves that are in phase interfere constructively and are amplified.
When materials with different properties of refraction (bending) and absorption are layered together, some light is reflected from the interface and some is transmitted. Where the glass of a camera lens meets the air, for example, light is reflected and does not reach the film, but an optical coating of the right thickness can cancel the reflection.
A stack of layers of two different materials can also be designed to boost reflectivity instead of canceling it, so that light reflected at each interface constructively interferes with light reflected from other interfaces. The amplified reflection offsets losses due to absorption.
"Most good camera lenses have antireflective coatings," says Underwood. "In visible light there's a whole cookbook of designs for thin-film reflective coatings."
The idea of mirrors that could reflect x-rays and EUV at normal incidence by using constructive interference to amplify weak reflections was suggested in the early 1920s, long before their fabrication was practical.
"Finally in 1940 a multilayer mirror was fabricated from gold, which is dense, and from less-dense copper," Underwood says. "It had layers about 10 nanometers thick and could reflect x-rays with a wavelength of about seven nanometers. Unfortunately gold and copper rapidly diffuse into each other, and the mirror's reflective power decayed in a few days."
Then, in the late 1970s, Troy Barbee and his colleagues at Stanford University developed a precise method of depositing alternating layers of materials only a few atoms thick on a substrate. Atoms are sputtered off separate sources and mounted over a turntable which alternately moves the substrate under one source, then the other. By controlling such variables as power to the sources and distance to the substrate, uniform layers are built up.
Meanwhile Eberhard Spiller at IBM was developing a technique using electron beams to evaporate the materials. Both methods are now used to manufacture mirrors from materials such as tungsten and carbon, molybdenum and beryllium, and molybdenum and silicon. The mirrors typically have 30 to 100 layers, each no more than a dozen atoms thick. CXRO began making, testing and using multilayer mirrors in 1984.
"We make molybdenum and silicon multilayer mirrors that are capable of reflecting over 60 percent of 13.5-nanometer EUV," says Underwood. "Recently we've gotten up to 68 percent."
An area of great practical promise for multilayer mirrors lies in microlithography for the computer chips of the future. Manufacturers of integrated circuits continually seek to pack more and more features on chips, which requires better resolution, which in turn requires shorter wavelengths--much shorter than the 248-nanometer deep-ultraviolet light now used commercially.
"EUV light at 13.5 nanometers could etch features as small as 100 nanometers across," says Eric Gullikson, also a member of CXRO, "The path to high resolution leads to shorter wavelengths, but as wavelengths get shorter, it's the errors in the optical surface that will limit performance."
Gullikson and his colleagues have recently used one of the two Calibration and Standards beamlines built by CXRO at the Advanced Light Source (ALS) to measure scattering from multilayer mirrors directly for the first time. No matter how uniform the layers, roughness in the underlying mirror substrate can cause EUV light to be scattered at angles away from the reflected beam, leading to a loss of contrast and brightness. "Scattering increases with decreased wavelength," Gullikson says, "and it's a big jump to go from a couple of hundred nanometers down to 13.5."
To count scattered photons at angles up to 40 degrees from the main reflected beam, Gullikson uses a channel electron multiplier mounted on a moving arm. Because the incoming beam of EUV light from the ALS is so intense and so tightly focused, scattering can be measured over angles as tiny as a hundredth of a degree.
In another technique, the beam is shone directly onto the test mirror through a hole in a circular plate. Most of the beam is reflected straight back through the hole, but scattered light hits the plate, which is riddled with microchannels that produce showers of electrons when struck by photons. These excite a phosphor screen immediately behind the plate to create a pattern that can be viewed directly.
"One of the main outcomes of our work was to show that we could use conventional optics and atomic-force microscopes to predict the pattern of scattering that we actually observed," says Gullikson. "This lends confidence to the design, manufacture, and quality control of multilayer mirrors for EUV lithography."
Multilayer coatings have many other applications. They can improve the performance of glancing-incidence mirrors; used on diffraction gratings, they can sharply select certain wavelengths and suppress others; and lens-like zone plates (concentric clear and opaque rings) etched into multilayers can focus x-rays of highly specific wavelengths.
In less than two decades multilayer technology has become an indispensable tool for the scientific investigation of the physical world--and the creation of a new world of computer technology to come.
Results of multilayer-mirror scattering measurements can be seen on the ALS website at http://www-als.lbl.gov/als/science/sci_archive/EUVmirrors.html.
