A measurement of the temperature-density relation in the intergalactic medium using a new Ly alpha absorption-line fitting method

The evolution of the temperature in the intergalactic medium is related to the reionization of hydrogen and helium and has important consequences for our understanding of the Ly alpha forest and of galaxy formation. We measur e the temperature-density relation of intergalactic gas from Ly alpha fores t observations of eight quasar spectra, using a new line fitting technique to obtain a lower cutoff on the distribution of line widths from which the temperature is derived. Using a numerical simulation, we examine the detail s of this kind of measurement at different densities, finding that the temp erature may be difficult to measure for gas with Delta (g) less than or sim ilar to 1 (Delta (g) is the density of the gas in units of the mean density ) because the velocities due to expansion always dominate the widths of the corresponding weak lines, and that the temperature measurement is increasi ngly ambiguous for gas with Delta (g) greater than or similar to 5 because the dispersion in temperature at fixed density is high. From our observed s pectra, the temperature is most precisely determined at densities slightly above the mean: T-* = (20,200 +/- 1300, 22,600 +/- 1900) K (statistical err or bars) for gas densities Delta (*) = (1.42 +/- 0.08, 1.37 +/- 0.11, 1.66 +/- 0.11) at redshift (z) over bar = (3.9, 3.0, 2.4). Systematic errors in T-* should be less than 2000 K. The power-law index of the temperature-dens ity relation, defined by T = T-*(Deltag/Delta (*))(gamma -1), is gamma - 1 = (0.43 +/- 0.45, 0.29 +/- 0.30, 0.52 +/- 0.14) for the same three redshift s. The temperature at fixed overdensity Delta = 1.4 is T-1.4 = (20,100 +/- 2800, 20,300 +/- 1400, 20,700 +/- 1900) K. This unchanging temperature is h igher than expected for photoionized gas in ionization equilibrium with a c osmic background. If the heat from the He II reionization is responsible fo r the high measured temperature, then the temperature should not be constan t but should have a maximum at the end of the reionization epoch. We update the lower limit to the baryon density implied by the observed mean flux de crement.