Evidence for band overlap metallization of hydrogen.

The report "Optical studies of hydrogen above 200 gigapascals: Evidence for metallization by band overlap" by H. K. Mao and R. J. Hemley (1) in my opinion does not present evidence that conclusively demonstrates the metallization mentioned in the title. The most demanding test of such metallization is to show that the dc electrical conductivity remains finite as the temperature goes to zero. One could also show that the sample possesses the optical properties of a metal. Visual darkening might be expected for a thin metallic film, but by itself the darkening of a sample is not sufficient evidence of metallization, just as a lustery metallic reflectance is not. Good examples are the semiconductors germanium and silicon, which as thin films are dark in the visible, while bulk samples are metallic in appearance. Darkening ofa sample can arise from any number of physical mechanisms that cause absorption throughout the visible spectral region. At zero pressure hydrogen has a strong broadabsorption band at 11 eV due to electronic excitons and a valence-conduction band gap absorption at 15.4 eV. When the band gap goes to zero, the solid becomes a metal. As the pressure is increased, both the exciton band and the band gap shift to lower energy. The exciton, if it remains a stable state, is always lower in frequency than the band gap. Thus, when the exciton is in the visible the sample darkens; likewise when the band gap passes through the visible the sample absorbs. Darkening of a sample has been observed in a number of pressurized insulators, which require still higher pressures to become metallic (2). To demonstrate metallization, one could also show that the optical absorption and reflection behave as expected for free electrons with relaxation, including interband transitions, using a Drude type model for analyses of the data. This evidence was presented in recent metallization experiments on xenon at pressures of 100 to 200 GPa (3). In the Drude model, the plasma frequency vp depends on the square root of the free carrier (electron or hole) density. For optical wavelengths longer than Ap = cdvp, the sample is highly reflective and absorptive, whereas for shorter wavelengths the reflectivity decreases, as does the absorption coefficient (that is, the transmission of light through the sample increases). On the other hand, the band gap of an insulator is not closed and it is expected that the transmission decreases for shorter wavelengths if the band gap energy is in the spectroscopic region of study. In the report by Mao and Hemley, the transmission as a function of wavelength shows little structure. There is a weak decrease with increasing frequency. This is the behavior of an insulator, not a metal. A uniform decrease in transmission as a function of wavelength is not expected for a metal, although for the limited spectral region they present, it is difficult to classify the material. Mao and Hemley state that "[i]n some cases . .. a weak increase in absorptivity was observed above 800 nm." Accepted scientific methodology demands reproducibility, especially to show metallization of hydrogen. The reflectivity spectrum, measured over a limited spectral range, which is shown in figure 2b (curve C of Mao and Hemley) could be that of a metal, but for evidence of metallization, absorption and reflectivity data should be measured as a function of pressure. The region of increasing reflection in a metal should shift to higher frequency with increasing pressure. This is because the band overlap increases with pressure so that the number of free carriers increases, resulting in an increase in the plasma frequency. In our experience in doing microspectroscopy (3), the background normalization of spectra is necessary to obtain correct results. This is especially true for samples with dimensions on the order of the wavelength of light, where frequency-dependent diffraction effects are important. This is exactly the measurement geometry of the samples of Mao and Hemley. We found that normalization of reflectivity spectra to the gasket reflectivity spectrum, as was done by Mao and Hemley was unreliable. Spectra taken from different regions of the gasket differed from each other because of surface inhomogeneities and ruby on the gasket surface. A procedure of normalization of the transmission to that of ruby is also questionable when one does not know the ruby absorption spectrum at the pressures used. For a weakly absorbing indirect band gap insulator, an incorrect normalization can change the slope of transmission/reflection as a function of frequency, so that an insulator appears as a metal and vice versa. Mao and Hemley discuss the possible red shift of the resonance-enhanced intensity of the vibron Raman scattering signal with increasing pressure as being indicative of a red shift of electronic transitions. They state that it is "clearly associated with changes in the electronic structure of hydrogen"; however, one would actually expect enhancement first at the 488-nm wavelength of laser excitation, before the 514.5-nm wavelength-the opposite of what Mao and Hemley report. Their explanation of the enhanced vibron frequency may be correct; however, it is not evidence of metallization. Mao and Hemley present the opacity, or darkening of the sample, as evidence of metallization and show a color photograph ofa sample in their figure 1. The dark area in the high-pressure central culet region is apparent. But the photograph also shows a large dark region to the right, in the much lower pressure diamond bevel region of the hydrogen-ruby mixture. Although a uniform darkening in this region may be due to the diamond bevel angle, the localized darkening cannot be. One is thus presented with dark regions in both highand low-pressure regions, without explanation. The observations ofMao and Hemley are suggestive, but are also compatible with the narrowing down of the band gap in an insulator. I do not believe they fulfill the scientific criteria for establishing metallization. IsAAc F. SILVERA Lyman Laboratory ofPhysics, Harvard University, Cambridge, MA 02138