The broadening of atomic emission lines by high-velocity motion of gas near accreting supermassive black holes is an observational hallmark of quasars. Observations of broad emission lines could potentially constrain the mechanism for transporting gas inwards through accretion disks or outwards through winds. The size of this broad-line region has been estimated by measuring the light travel time delay between the variable nuclear continuum and the emission lines - a method known as reverberation mapping. In some models the emission lines arise from a continuous outflow, whereas in others they are produced by orbiting gas clouds. Directly imaging such regions has not hitherto been possible because of their small angular sizes (< 0.1 milli-arcseconds). Here we report a spatial offset (with a spatial resolution of ten micro-arcseconds or about 0.03 parsecs for a distance of 550 million parsecs) between the red and blue photo-centres of the broad Paschen-{\alpha} line of the quasar 3C 273 perpendicular to the direction of its radio jet. This spatial offset corresponds to a gradient in the velocity of the gas and thus implies that the gas is orbiting the central supermassive black hole. The data are well fitted by a broad-line-region model of a thick disk of gravitationally bound material orbiting a black hole of 300 million solar masses. We infer a disk radius of 150 light days; a radius of 100-400 light days was found previously using reverberation mapping. The rotation axis of the disk aligns in inclination and position angle with the radio jet. Our results support the methods that are often used to estimate the masses of accreting supermassive black holes and to study their evolution over cosmic time.

We compare two observations of gamma-rays before, during, and after lightning flashes initiated by upward leaders from a tower during low-altitude winter thunderstorms on the western coast of Honshu, Japan. While the two leaders appear similar, one produced a terrestrial gamma-ray flash (TGF) so bright that it paralyzed the gamma-ray detectors while it was occurring, and could be observed only via the weaker flux of neutrons created in its wake, while the other produced no detectable TGF gamma-rays at all. The ratio between the indirectly derived gamma-ray fluence for the TGF and the 95% confidence gamma-ray upper limit for the gamma-ray quiet flash is a factor of $1\times10^7$. With the only two observations of this type providing such dramatically different results -- a TGF probably as bright as those seen from space and a powerful upper limit -- we recognize that weak, sub-luminous TGFs in this situation are probably not common, and we quantify this conclusion. While the gamma-ray quiet flash appeared to have a faster leader and more powerful initial continuous current pulse than the flash that produced a TGF, the TGF-producing flash occurred during a weak gamma-ray "glow", while the gamma-ray quiet flash did not, implying a higher electric field aloft when the TGF was produced. We suggest that the field in the high-field region approached by a leader may be more important for whether a TGF is produced than the characteristics of the leader itself.

The method developed by Stock and Stock (1999) to derive absolute magnitudes and intrinsic colors is applied to simulated low-resolution spectra. The simulation is made by convolving real spectra with a Gaussian function, $\sigma$ (the full width at half maximum) related to the final spectral resolution. The accuracy with which the stellar parameters are determined indicates that the method may be applied to typical objective-prism spectra. We show that changes in the spectral resolution do not significantly affect the stellar parameters obtained with this method for early-type stars, whereas for later-type stars an improved approach is necessary.

The method developed by Stock and Stock (1999) for stars of spectral types A to K to derive absolute magnitudes and intrinsic colors from the equivalent widths of absorption lines in stellar spectra is extended to B-type stars. Spectra of this type of stars for which the Hipparcos Catalogue gives parallaxes with an error of less than 20% were observed with the CIDA 1-meter reflector equipped with a Richardson spectrograph with a Thompson 576x384 CCD detector. The dispersion is 1.753 A/pixel using a 600 lines/mm grating in the first order. In order to cover the spectral range 3850 A to 5750 A the grating had to be used in two different positions, with an overlap in the region from 4800 A to 4900 A. A total of 116 stars was oberved, but not all with both grating positions. A total of 12 measureable absorption lines was identified in the spectra and their equivalent widths were measured. These were related to the absolute magnitudes derived from the Hipparcos Catalogue and to the intrinsic colors (deduced from the MK spectral types) using linear and second order polynomials and two or three lines as independent variables. The best solutions were obtained with polynomials of three lines, reproducing the absolute magnitudes with an average residual of about 0.40 magnitudes and the intrinsic colors with an average residual of 0.016 magnitudes.