The etching of polycrystalline Cu surfaces by Br-2 was studied using temperature programmed desorption and scanning tunneling microscopy, and the results were compared with a recent study of the interaction of Br-2 with Cu(100). For both Cu surfaces, the etching mechanism could be described as a two step process : adsorption rate limited CuBr formation followed by CuBr desorption. The adsorption rate was found to be structure sensitive. At 325 K, the sticking coefficient for CuBr formation was higher on Cu(100) than on polycrystalline Cu. For both surfaces, increasing the temperature during dosing decreased the CuBr formation rate. The decrease was larger for polycrystalline Cu and so the structure sensitivity increased with increasing temperature. The removal of CuBr by sublimation was found to depend more strongly on surface structure. For low CuBr coverages on Cu(100), a CuBr desorption peak was observed at 450 K, a temperature lower than anticipated from CuBr vapor pressure data. When the CuBr coverage was increased, the peak at 450 K shifted to higher temperatures and eventually disappeared; the remaining CuBr desorption peaks were consistent with sublimation of bulk CuBr. A CuBr desorption peak at 450 K was also observed on polycrystalline Cu. On the polycrystalline surface, however, this peak saturated at high coverages but did not shift to higher temperatures. The intensity of the 450 K desorption peak depended on the treatment of the polycrystalline surface. Annealing was found to increase the peak intensity while sputtering and Br-2 etching were found to decrease the intensity. Scanning tunneling microscopy images following each of these treatments suggested that the 450 K desorption peak is associated with narrow terraces on the surface. The images also revealed that etching greatly increases terrace widths suggesting that steps act as Cu atom sources for CuBr formation, and thus that Br-2 etching of Cu can be considered the reverse of step flow growth.