We present a detailed observational and theoretical study of a D3 hr long X-ray burst (the "" superburst ÏÏ) observed by the Rossi X-Ray T iming Explorer (RXT E) from the low-mass X-ray binary 4U 1820[30. This is the longest X-ray burst ever observed from this source and perhaps one of the longest ever observed in great detail from any source. We show that the superburst is thermonuclear in origin. Its peak luminosity of D3.4 ] 1038 ergs s~1 is consistent with the helium Eddington limit for a neutron star at D7 kpc as well as the peak luminosity of other, shorter, thermonuclear bursts from the same source. The superburst begins in the decaying tail of a more typical (B20 s duration) thermonuclear burst. These shorter, more frequent bursts are well-known helium Ñashes from this source. The level of the accretion-driven Ñux as well as the observed energy release of upward of 1.5 ] 1042 ergs indicate that helium could not be the energy source for the superburst. We outline the physics relevant to carbon production and burning on helium-accreting neutron stars and present calculations of the thermal evolution and stability of a carbon layer and show that this process is the most likely explanation for the superburst. Ignition at the temperatures in the deep carbon "" ocean ÏÏ requires more than 30 times the mass of carbon inferred from the observed burst energetics unless the He Ñash is able to trigger a deÑagration from a much smaller mass of carbon. We show, however, that for large columns of accreted carbon fuel, a substantial fraction of the energy released in the carbon-burning layer is radiated away as neutrinos, and the heat that is conducted from the burning layer in large part Ñows inward, only to be released on timescales longer than the observed burst. Thus, the energy released during the event possibly exceeds that observed in X-rays by more than a factor of 10, making the scenario of burning a large mass of carbon at great depths consistent with the observed Ñuence without invoking any additional trigger. A strong constraint on this scenario is the recurrence time : to accrete an ignition column of 1013 g cm~2 takes g s~1) yr. Spectral analysis during the superburst reveals the D13/(M 0 /3 ] 1017 presence of a broad emission line between 5.8 and 6.4 keV and an edge at 8È9 keV, likely due to reÑec-tion of the burst Ñux from the inner accretion disk in 4U 1820[30. We believe that this is the Ðrst time such a signature has been unambiguously detected in the spectrum of an X-ray burst.