Here are 3:
1) gene knockout. Just delete the gene from the genome. The function is gone - useful for demonstrating a direct involvement of the gene in the phenotype. As a phenotype, the microarray will register all sorts of reactions to the loss of the gene in addition to the RNA in question being gone.
2) use selection to find mutants for the gene. Expose the organism to a mutagen or high amounts of UV radiation. Useful mutants vary by what the function of the genes are. Temperature sensitivity - the gene doesn't function above 30 deg C is an example of a phenotype for a mutant that has been used classically. In this case the functioning of the gene itself is a phenotype. This is an example of a weaker perturbation. Can be used when knockouts are fatal or when the system is too sensitive for a knockout.
3) change the environmental conditions. increase the concentration of a substrate or lower the availability of nutrients. Phenotypes can be rate of growth/division, production of a given metabolite from the cells, but it goes on and on.
The affect on whole genome microarray data can be quite complex in such perturbation experiments because the coupling of many genes, even in bacteria can be broad and significant.