This paper presents an empirical way to find type 3 Kardashev civilizations, which is to identify outliers in the Tully-Fisher relation for spiral galaxies and in the fundamental plane for elliptical galaxies.
The author begins by arguing that it is possible to observe galaxy scale interruption of starlight. Then the author derives the scaling relation for galaxies argues that the scatter seen in empirical data around the scaling relation is as low as 10 percent.
This tight relation could be used to identify outlier galaxies which have very low surface brightness, probably with a infrared excess if the type 3 Kardashev civilizations use technologies such as Dyson sphere.
Then the author describes the empirical effort to find such outliers in 31 spiral galaxies and 106 elliptical galaxies. However, no significant outlier has been detected in those galaxies. The author argues that this could be due to setting a too strict cut on the surface brightness deviation. Newer samples could reduce the cut down to 50 percent. However, the author leaves it for another time. Further, the author discusses the limitation of photographic plates not being able to detect low surface brightness galaxies which could give us biases towards regular galaxies than galaxies harnessed by type 3 Kardeshev civilizations.
Finally, the author discusses how long it would take for a type 3 Kardashev civilization to rise and the upper limit is 304 billion years which is much longer than the age of the universe. There are several reasons which could lower this upper limit. First, the sample is incomplete and they do not have any very low surface brightness galaxies. Second, not all type 3 Kardashev civilizations have to be star-fed. They could harness other types of energy such as dark matter and dark energy. Third, the time for those civilizations to develop may be much shorter than 10 billion years. One thing I do not understand is that why it is an upper limit than a lower limit because if those civilizations do not exist, the upper limit should be infinite.