One quasar, several images
A quasar is the bright centre of a distant galaxy, where gas falls into a giant black hole and glows. It is so far away that it looks like a single point of light. Sometimes another galaxy sits almost exactly in front of it. That galaxy's gravity bends the quasar's light the way a glass lens bends sunlight, and the light can reach us along two or four different paths. A telescope then sees two or four copies of the same quasar, about an arcsecond apart. An arcsecond is the size of a coin seen from four kilometres away.
The paths are not equal. Some are longer because they swing wider around the galaxy. Some pass deeper through the galaxy's gravity, and clocks there run slightly slow. Both effects add up to an arrival time for each path. Gravity also changes how bright each copy looks.
A flare that arrives on different days
Quasars are not steady. Their brightness drifts and sometimes jumps over days. When the quasar flares, the flare travels down every path at once, so each image brightens in turn. Astronomers watch the images night after night for years and line up the light curves, the records of brightness against time. The shift that makes the curves match is the time delay. Among the lenses Christopher Kochanek collected in 2002, measured delays ran from about 12 days in PG1115+080 to 417 days in Q0957+561. Kochanek 2002, Table 1
You can see this in the Lab. With the default lens, the four images light up over about two weeks: the second image follows the first within a day, and the last one trails by 15.5 days.
From a delay to a distance
The delay grows with the size of the whole set-up. Put the same galaxy and quasar twice as far away, with the same angles on the sky, and every path gets twice as long, so the delay doubles. That is why a delay can work as a ruler. In symbols,
Here Δt is the delay, c is the speed of light and Δϕ is a number fixed by the image positions and the galaxy's mass, measured in square arcseconds. DΔt is a combination of distances to the galaxy and the quasar, called the time-delay distance.
That distance tells us about the expansion of the universe. The Hubble constant, H0, is how fast distant galaxies recede for each megaparsec of distance (a megaparsec is about 3.3 million light-years). For given redshifts, a faster expansion means smaller distances, so the time-delay distance is proportional to 1/H0. Sjur Refsdal proposed the method in 1964. TDCOSMO IV, Sec. 1 and 2.1
The catch: a mass you cannot see
The images fix where the light goes, but they do not quite fix how much mass the galaxy has. Imagine spreading a thin, even sheet of extra mass over the galaxy and shrinking the galaxy's own mass to compensate. The images stay exactly where they were. The delays, however, shrink or stretch in proportion. Astronomers call this the mass-sheet degeneracy. With image positions alone, the same lens can give quite different values of H0. Peter Schneider and Dominique Sluse gave an example in 2013 where two lens models fit the same images but disagree on H0 by about 20 percent. Schneider and Sluse 2013, abstract
The Lab has a preset for this. "Same images, hidden sheet" keeps all four images in place but moves the flares closer together by 12 percent. Someone who assumed there was no sheet would read H0 as 79.5 instead of the true 70 used in the Lab.
Two kinds of evidence help. Galaxies near the line of sight add a real sheet of mass, and counting them estimates it. The speed of the stars inside the lensing galaxy weighs the galaxy independently of the light paths, because the sheet changes the star speeds in a different way from the delays.
What astronomers have found
In 2019 the H0LiCOW team combined six lensed quasars and found H0=73.3 km/s per megaparsec, with an uncertainty of 2.4 percent, assuming the galaxies' mass follows standard shapes. H0LiCOW XIII, abstract
In 2020 the TDCOSMO team dropped that assumption and let the hidden sheet float, held in place only by the stellar speeds. With seven lenses the answer became 74.5, now uncertain by about 8 percent. Adding 33 more lens galaxies that have measured star speeds but no quasar behind them moved it to 67.4, with an uncertainty of about 5 percent. That last step assumes the two groups of galaxies are alike. TDCOSMO IV, abstract
The spread between those numbers is the mass sheet at work. It matters beyond lensing. Supernova distances and the cosmic microwave background, the afterglow of the early universe, give different values of H0, and time delays are independent of both. H0LiCOW XIII, abstract
The course continues with the arrival-time surface and the delay of the simplest lens: arrival time, the Fermat potential and the isothermal lens.