Students and others looking for a place to start often ask, "If I could read just one book about xxxx, which would you suggest?" When it comes to cities, I have a new favorite. It is Shlomo Angel's Planet of Cities. The author studies an amazing amount of data covering the world's 3,646 cities (usually, metropolitan areas) with population of 100,000 or more in 2000. This is one amazing data base and Angel makes the best of it, addressing the interesting questions surrounding growth, spread, "sustainability", etc. Yes, cities spread out everywhere and "containing" them is undesirable as well as implausible.
Angel evokes the "Goldilocks Principle" more than once and wonders if there is a population density which is not too low and not too high. As I have often mentioned, the question obscures the real story. Metropolitan areas are very big and diverse. The Los Angeles-Orange area covers more than 3,250 square miles. In 2000, the population residing in that area was 12,366,000 and the number of jobs was 5,292,000. Annual sub-metropolitan data are available for Public Use Microsample Areas (PUMAs) of which there are 84 in LA-Orange with an average size of about 40 square miles. The average PUMA density was 8,748 per square mile but the standard deviation of the PUMA densities was almost as large, 8,456.
Angel argues (as have many others) that there is no "optimal" urban size. There is, likewise, no "optimal" urban density. There are complex density distributions in all of the metropolitan areas. These are endogenous and it takes flexible land markets to hone in on the most suitable (congenial to growth) ones in each case.