We have detected the microwave spectrum of the smallest cyclic peptide-diketopiperazine-in the frequency range 48-72 GHz, demonstrating that the molecule does not adopt in isolation the highly symmetric (C-2h) planar-ring structure obtained in the solid state via X-ray crystallography. From a comparison of the derived rotational constants (MHz), A = 4906.4098(44), B = 1582.1420(37), C = 1239.4218(44), with those obtained from an ab initio molecular orbital. calculation [MP2/6-311++G(d,p) level], the stable form is a boat configuration having C-2 symmetry. Exploration of the ring puckering potential energy surface indicates that this "methylene" boat conformer is the only stable conformer of diketopiperazine. The microwave spectrum deviated from that of a rigid rotor in that all of the measured transitions were members of doublets in which the separation was similar to 2 GHz. This is attributed to tunneling between two equivalent conformations through a relatively low barrier on the potential energy surface. Our exploration of the ring puckering possibilities via ab initio molecular orbital calculations indicates that the minimum energy pathway linking the two boat (C-2) enantiomeric conformers passes over a barrier of about 470 cm(-1). The chair (C-i) conformer is involved at the summit of the barrier. This barrier is significantly lower in energy than the planar ring (C-2h) species which appears to be a higher saddle point on the potential energy hypersurface. The calculated energy barrier is plausibly consistent with the tunneling splitting found in the spectrum. A simple empirical modeling of the ring puckering energy of diketopiperazine in terms of peptide linkage torsion and ring-angle deformations represents the ab initio ring flexure energies surprisingly accurately. The fitted torsional energy function is in close agreement with the comparable ab initio omega-torsion in N-methyl acetamide and is predominantly quartic. This has implications for protein modeling since this appears to deviate in detail from the form of potential currently included in the molecular mechanics computational models employed for the cis peptide linkage in the theoretical study of protein folding.