The average energies of neutrons emitted from a graphite column at 22 degrees C were compared by measurement of the cross section of boron for neutrons which are stopped by cadmium. At a distance from the neutron source great enough to insure that the neutrons were in thermal equilibrium the average energies of the emerging neutrons were found to be proportional to the temperature within the limits of the experimental error. A measurement made with boron absorbers which had been thus standardized in the graphite column indicated neutrons emerging from the chain reacting pile to have an average temperature approximate 60 +- 50 degrees above that of thermal neutrons emerging from the graphite column at 22 degrees C. Such a measurement made inside the chain reacting pile indicated the average temperature of neutrons therein to be about 65 degrees +- 15 degrees above the average temperature of neutrons in the graphite column.

The temperature coefficient is calculated for various lattice arrangements, taking into account the variation of [formula], suggested by Fermi. Four contributions are included: leakage, levelling of the dip in thermal neutron density in the lump, resonance absorption, and hardening of the neutrons as they penetrate a metal lump. The departure of neutron temperature from lattice temperature decreases the total coefficient. Values are given for 3 typical piles; in general, the larger the uranium elements, the less stable the pile. A rod lattice tends to be more stable. A pile with metal lumps over 50 lbs. will be unstable.

Radial flattening of activity in the cores of spherical and cylindrical piles is discussed in connection with pile control and power improvement. Partial flattening as a result of k loss from temperature rise is also considered.