History of the CMB Dipole Anisotropy

The Solar System is moving at 370 km/sec relative to the Universe and we can measure this using the dipole anisotropy of the Cosmic Microwave Background (CMB). This was recognized as soon as the CMB was discovered, so experimenters went to work to take data immediately. I was not working on the CMB at the relevant time, but I did hear some of the first announcements of results.

By the time Peebles' Physical Cosmology was published in 1971, the velocity of the Solar System was already in the textbooks. This is the textbook that George Smoot used to learn about cosmology when he started working on the Alvarez proposal for an anisotropy satellite. Given this, how can Smoot (see Horgan, July 1992 Scientific American, pp 34-41)[1] and the LBL press office maintain that Smoot discovered the dipole? It's a lie, so make it a big lie! Charley Lineweaver has published a list of dipole determinations. Smoot et al. (1977) is the 10^th entry on that list.

The first claimed detection was by Edward (Ned) [2] Conklin, using a ground-based differential radiometer working at 8 GHz (1969, Nature, 222, 971). He observed at 32^o declination, and found a peak at 13^h right ascension, with an amplitude of 1.6 mK. In 1971 he had (1.9+/-0.8) mK peaking at 10^h in right ascension. In 1972 he reported more data giving an amplitude of (2.28+/-0.92) mK with the peak at 10^h 58^m right ascension. The modern value of the dipole from WMAP would predict an amplitude of 2.57 mK at this declination and a right ascension (epoch 1970) of 11^h 10^m. The actual declination (epoch 1970) was -7^o, so observing at 32^o reduces the expected amplitude. So Conklin was correct to within 12% in amplitude and 3^o.

The declination of the dipole peak was first measured by Paul Henry (1971, Nature, 231, 516). This was his PhD thesis work, supervised by David T. Wilkinson. Princeton at the time required that PhD work be published as single authored papers. Henry used a balloon-borne radiometer at 10.15 GHz and got an amplitude of (3.2+/-0.8) mK, a right ascension of (10.5+/-4)^h, and a declination of (-30+/-25)^o. The declination is harder to measure because a balloon floats at nearly constant latitude, so only a small range of declinations is sampled.

The Conklin (1969, 1972) and Henry (1971) results are what allowed Peebles to include the velocity of the Solar System in his 1971 textbook. The velocity is given as (-220+/-75,124+/-75,-200+/-100) km/sec in Eqns (40) and (41).

Brian Corey and David Wilkinson reported results from a 19 GHz balloon-borne radiometer at the June 1976 AAS meeting in Haverford, PA. I was in the room during the talk and heard the results, which are printed in (1976, BAAS, 8, 351). They gave the velocity of the Solar System as 270+/-70 km/sec toward right ascension (13+/-2)^h, declination (-25+/-20)^o.

Later in the summer of 1976 I heard George Smoot give a talk about the observations he was going to make using the U2. So it is important to remember that Corey and Wilkinson had reported their results before Smoot made any flights with the U2. Smoot then made rapid progress and reported results at the April 1977 APS meeting and in Smoot, Gorenstein and Muller (1977, PRL, 39, 898). Smoot et al found an amplitude of (3.5+/-0.6) mK at right ascension (11.0+/-0.6)^h, declination (6+/-10)^o.

So there can be no question that Smoot came along after Conklin, after Henry, and after Corey and Wilkinson. It is thus quite bizarre to read in Smoot's book: "Confident that our data were valid, I planned to announce our results at the American Physical Society meeting in Washington, D.C., in April 1977. Our competitors, particularly the Princeton team of Brian Corey and David Wilkinson, were hot on the track of the dipole with balloon-borne observations, but I wasn't sure how close they were." This is just one example of fictional history from Wrinkles in Time.

Hence the only question that remains is whether the dipole was "discovered". One way to judge this is to compute the likelihood ratio for two hypotheses: no dipole, or the true dipole. A table of these ratios gives:

Observer          Date  L(true)/L(null)
Conklin           1969       3.5
                  1971      12.2
                  1972      20.1

Henry             1971    1900.7

Peebles           1971     145.5

Corey & Wilkinson 1976     280.1

Smoot et al.      1977      107

The graph at right shows a joint fit to the Henry (1971) data in red and the Corey & Wilkinson (1976) data taken from the plots in Muller's 1978 SciAm article. It gives a dipole amplitude of 2.83 mK, a right ascension of 11.2 hours, and a declination of 14^o.

But Smoot still repeats his claim to have discovered the dipole. In his Nobel lecture [3], he said there had been earlier results but that they had found comparable dipole and quadrupole terms. This is a continuation of the big lie, since neither Henry nor Corey & Wilkinson made a quadrupole claim. The erroneous quadrupole claims by Fabbri et al. (1980, PRL, 44, 1563) and by Boughn, Cheng & Wilkinson (1981, ApJL, 243, L113) were made only after the dipole was definitely discovered. And neither of these papers had "comparable" quadrupole and dipole amplitudes. The variance of the dipole pattern on the sky is 3.7x10⁶ μK², while Fabbri et al. got a quadrupole with a variance of 2x10⁵ μK² and Boughn et al. got 1.6x10⁵ μK². A quadrupole variance 18 times smaller than the dipole, if due solely to noise, would imply a 5.5 sigma detection of the dipole. Smoot's second point against the earlier detections is that they had large position errors, while his U2 result had 10 degree uncertainty. Ten degrees is what you would expect from a 6 sigma result like Smoot et al (1977). A 3 or 4 sigma result like Henry (1971) and Corey & Wilkinson (1976) would have 15 or 20 degree position uncertainty if the whole sky was sampled, but due to limited sky coverage they had somewhat larger uncertainties. But this is what you expect for a first detection. The combined fit above to the Henry and Corey & Wilkinson data gives a dipole that is perfectly consistent with the Smoot et al. dipole. So Smoot is clutching at straws to justify his long standing false claim to have discovered the dipole.

In his Nobel lecture, Smoot also said that his group had made a plot of the anisotropy, and then showed a map of the dipole model from Muller's 1978 article in Scientific American. This was not a map of the observations, and because the U2 made discrete pointings instead of scanning the whole sky, Smoot's U2 experiment never made a map of the temperature of the sky. It is just a pictorial representation of the dipole model on the sky, suitable for a popular science magazine like Scientific American. Lubin, Villela, Epstein & Smoot, flying a 90 GHz radiometer on a balloon, did make a map of the sky, but published it only in 1985. This map is shown at right. Click on the image for a color version in galactic coordinates.

Smoot's U2 experiment came along too late to discover the dipole, but he led the team that built the COBE DMRs, which are very similar to the U2 experiment, and I was able to tell him in August 1991 that the DMRs had discovered a signal with a quadrupole variance of only 300 μK².

[3] Smoot also used 3 of my slides without asking for permission or giving any credit (at about 31:27, 31:34 and 38:00). He even went so far as to crop off part of the x-axis label on the third of these slides to avoid showing my name. Here is a photo taken of the online video of Smoot's lecture, and here is original image. The original context was meant to show that the BOOMERanG experiment published data with serious systematic errors in 2000. Smoot greatly distorted this message in his lecture.

© 2006-2009 Edward L. Wright. Last modified 14 Mar 2009