We have reexamined the relationship between ''U-shaped'' peak flux density microwave spectra and solar proton events for ~200 large (Sp(≥2 GHz)≥800 solar flux units (sfu) microwave bursts (1965-1979). The radio spectra fell into two basic classes: U-shaped, with two maxima (≥800 sfu) in the range from 200 MHz to ≥10 GHz (59% of all events), and cutoff spectra, with a maximum ≥800 sfu at f≥2 GHz and Sp (200 MHz)10-MeV proton events of any size (≥0.01 protons cm-2 s-1 sr-1) are as follows: U-shaped: type II/IV (90% of large microwave bursts with U-shaped spectra are associated with type II/IV events), protons (77%); intermediate: type II/IV (78%), protons (73%); and cutoff, type II/IV (22%), protons (33%). These statistics affirm various lines of evidence linking coronal shock waves and interplanetary proton events. They also suggest that the meter wavelength branch of the U-shaped spectrum may be attributable to second-phase (versus flash phase) accelerated electrons.

We have examined this latter supposition and find that it cannot be true in general. In our sample a type II event was in progress at the time of the peak of the low-frequency branch for only about half of the bursts with U-shaped spectrum (U bursts). For these events we cannot rule out a possible contribution to the peak 200-MHz flux from either the second harmonic of the type II burst or from flare-continuum of the type FC II, provided that the starting frequency of the fundamental type II burst is >100 MHz. The low-frequency branch of the U burst appears to be more closely related to impulsive phase type III emission. We note that the small sample of U bursts that lacked type II/IV association is also poorly associated with proton events. We conclude that the observed association between U bursts and proton events probably results from the big flare syndrome rather than a close physical link between these two phenomena. If the current National Oceanic and Atmospheric Administration prediction threshold of J(>10 MeV)≥10 protons cm-2 s-1 sr-1 had been in effect during the period covered by our data base (1965-1979), the U burst ''yes or no'' forecast tool would have had a false alarm rate of 50-70% and would have failed to provide warning for 40-50% of the significant prompt proton events attributable to disk flares. We note that several (eight of 46) of the prompt events with J(>10 MeV)≥10 protons cm-2 s-1 sr-1 observed from 1965 to 1979 originated in flares that had relatively weak (≤300 sfu) burst emission at 200 MHz.