Info

0h 2h 4h 6h 8h

10h 12h 14h Right ascension

16h 18h 20h 22h

Fig. 4.25. Results of the scan of a circle at declination 6 = —8° in the experiment at Princeton University. The fractional temperature fluctuations are 6T/T ~ 3 x 10~3 (Wilkinson and Partridge 1967). ©1969 Nature Publishing Group.

0h 2h 4h 6h 8h

10h 12h 14h Right ascension

16h 18h 20h 22h

Fig. 4.25. Results of the scan of a circle at declination 6 = —8° in the experiment at Princeton University. The fractional temperature fluctuations are 6T/T ~ 3 x 10~3 (Wilkinson and Partridge 1967). ©1969 Nature Publishing Group.

I and a handful of undergraduate students working with us then read the output of the chart recorder by hand to determine the differences between the declination 6 = —8° circle and our constant calibration point, the north celestial pole. We ran this experiment for substantially more than a year to help average out diurnal effects. Some of those results appear in Figure 4.25.

It soon became clear that atmospheric noise was completely dominating the signal. By late 1966 we were planning improvements. It would have helped, for instance, if we had been able to switch the main beam more rapidly, but we were aware that the ferrite devices used for switching are themselves a source of noise and potential systematic error, a problem later encountered in another anisotropy experiment by Dave Wilkinson and Paul Henry (Henry 1971). So we took another approach to doing a better experiment, trying to find a place where the atmosphere is more benign. We probably should have leapt immediately to the conclusion that we needed to get above the atmosphere altogether, as Dave later did in his pioneering balloon experiments, and as George Smoot and his colleagues later did with their U-2 experiments (Smoot, Gorenstein and Muller 1977). But we were a frugal pair, so we decided instead to find the place in the United States with the least cloud cover. Dave discovered that that is southwestern Arizona, and I found out that there is an Army base at Yuma, smack in the middle of this relatively cloudless zone. Through my father's Army connections, I got us permission to move an improved isotropy-measuring device to the Army's Yuma Proving Ground.

We faced some constraints in designing the equipment. Both of us were busy teaching and could not spend much time in Yuma. So we needed to design a fully automated station that would take data and record it and that needed no daily maintenance. The equipment was designed with the main horn antenna pointed down, to prevent the collection of dust, rain, dead moths, etc. We also designed the equipment to scan two circles in the sky as well as the constant reference point, the north celestial pole (the reference horn also pointed there). Finally, we also took much greater care to prevent radiation from the ground entering the main antenna through its side lobes - see the ground screens identified in Figure 4.26. I took charge of designing the structure to support the main antenna, as well as the rotating beam-switching device, a tilted, elliptical mirror. I recall bringing my designs to Bob Dicke, who took a brief look at them and said, "Well, it is certainly sturdy." By that he meant that I had overdesigned the strength of the contraption by several orders of magnitude - I suspect it was at least as "sturdy" as the Army's top line tank!

Fig. 4.26. A refined experiment to look for anisotropy in the CMBR. Note the inverted horn and the use of ground screens to minimize stray radiation from the ground (and the sturdy construction).

Now, if you're designing a remote experiment, you need to have it in a place where casual hikers or hunters are not likely to poke around in it. The management at the Army's Yuma Proving Ground suggested that we use a securely fenced area at the outer edge of the base. It was securely fenced because it was the site at which the Army tested the integrity of nerve gas shells. There were racks and racks of nerve gas shells of various sorts lying about in the desert, left out to see whether or when they would leak. Needless to say, the area was both securely fenced and patrolled.

So, in the summer of 1967 we packed the monstrosity I had designed plus some additional equipment (see below) into a large U-Haul truck, and set out for the west. Dave used the trip as a family vacation; I got to drive the U-Haul. When we arrived at the Yuma base, Dave was appointed a temporary captain, and I got to be a lieutenant. The U-Haul was costing us a fortune to drive back and forth to our remote site, so we bought an item designated as a "personal transport device" to the NSF, otherwise known as a moped, for me to commute to the instrument.

We soon had the equipment up and running. Since useful computers were still a ways in the future, the basic control mechanism for the experiment was derived from a rewired digital clock, and the data were printed out on a line printer (whose values still needed to be recorded and sorted by hand).

While we were installing the Yuma apparatus, Dave and I were finishing up two papers on the results of the first anisotropy experiment at Princeton. One of those was written in a crummy motel room in Yuma using the only available horizontal surface, the top of a beer cooler. I remember sitting on the dirty green shag carpet, drinking Coors, and the excitement of reaching millikelvin levels in anisotropy.

We also had some time to learn about the nerve gas from soldiers on duty near our site. They explained the use of gas masks - we were issued with them in case the nerve gas really did leak - and told us that the first way to detect leaks of nerve gas was to check out the rabbits. Near each stack of nerve gas shells there was a hutch containing several standard laboratory rabbits. These were placed there since rabbits are highly sensitive, it appears, to nerve gas. So the first alarm for leakages was increased rabbit mortality. Well, in our brief time in Yuma, the rabbits started to die. There was considerable consternation, not least on our part, until a wise veterinarian pointed out that all the rabbits had been bought at the same time, and all of them appeared, entirely naturally, to be reaching the rabbit equivalent of three score and ten. Since we had escaped the nerve gas, Dave and I joined the soldiers' favorite game of sitting in a cargo container while someone set off a military-strength tear gas grenade. The game was to see who could last the longest before bolting for fresh air. We were young then.

As the Yuma experiment came on line, Dave and Bob Stokes left for the second main leg of the summer's work, a refined, multi-frequency measurement of the spectrum of the CMBR. I later joined them, traveling through the desert on my trusty "personal transportation device."

The idea here was to measure the temperature of the CMBR at three different frequencies - later extended to four by Paul Boynton - using very similar apparatus, so that the temperature measurements could be securely intercompared. In particular, the hope was to show that the spectrum of the radiation we were studying is not an exact Rayleigh-Jeans form, with energy density that varies with frequency v as uv rc v2, but instead shows some curvature as the peak of a blackbody spectrum at temperature about 3K is approached.

So we designed radiometers having similar beam sizes, all able to couple to a common calibration cold load. To prevent systematic errors, we designed the main horn antenna to look downward at an angle, making it easy to couple to a tilted dewar containing the cold load without moving the apparatus (Figure 4.27; Stokes, Partridge and Wilkinson 1967). Thus, to deflect the beam to the zenith in order to measure the CMBR, we needed to use an oversize reflector. We also arranged the reflector to be movable, so that the main beam could be cast through different zenith angles, enabling us to measure the atmospheric emission with the same equipment used for the absolute temperature measurements. One of the three radiometers used in the 1967 campaign is shown in Figure 4.27, along with the experimental setup.

Fig. 4.27. Photo of one of the three radiometers used on White Mountain, California, to measure the spectrum of the CMBR (Stokes, Partridge and Wilkinson 1967). The horn antenna is coupled to a large-diameter cold load. On the right is a schematic of the radiometers (Wilkinson 1967). ©1967 American Physical Society.

Fig. 4.27. Photo of one of the three radiometers used on White Mountain, California, to measure the spectrum of the CMBR (Stokes, Partridge and Wilkinson 1967). The horn antenna is coupled to a large-diameter cold load. On the right is a schematic of the radiometers (Wilkinson 1967). ©1967 American Physical Society.

It is worth mentioning the care we took to avoid systematic error. Dave Wilkinson, as all who knew him will attest, was extraordinarily careful about finding and eliminating, or at least modeling, sources of systematic error. We took great precautions, for instance, to control emission from the ground leaking into the side lobes of the antennas we used. We were conscious that emission from the walls of the calibration cold load could present a problem, and for that reason we expanded the beam and used a large "over-moded" cold load immersed in liquid helium. I have already mentioned quasisimulta-neous measurements of the atmospheric emission. And we also took account of the possible emissivity of the reflecting surface.

The result of this work was to produce temperature measurements at three wavelengths with substantially smaller error bars than previous workers had been able to obtain. The error bars were small enough to show rather convincingly that the spectrum of the CMBR does indeed begin to turn over at high frequencies, as expected for a 3K blackbody (Figure 4.28). And the final temperature we derived from combining observations at the three frequencies gave a value T = 2.68+g;0| K, in remarkably good agreement with the COBE satellite results that came along nearly two decades later (Stokes, Partridge and Wilkinson 1967; Wilkinson, 1967).

Fig. 4.28. Results of the Princeton measurements on White Mountain, showing a departure from the Rayleigh-Jeans law uv x v2 (Partridge 1969). The crucial 0.33-cm measurement was made by Boynton, Stokes and Wilkinson (p. 312). ©1969 American Scientist.

The spectral observations were carried out at the highest place in the United States with electrical power, the White Mountain Research Station maintained by the University of California. Not surprisingly, other groups had figured out that this was an excellent place from which to observe the microwave background. When we arrived, we discovered Bernie Burke and his colleagues busy assembling apparatus that looked an awful lot like that shown in Figure 4.27 (Ewing, Burke and Staelin 1967). Our group and his agreed to work entirely independently, so as not to influence one another's results. Yet another group, Welch et al. (1967), also recognized the value of high-altitude observations. However, they encountered problems with the design of their cold load calibrator, and perhaps as a consequence came up with too low a value for the CMBR temperature. On page 295 Welch describes how that turned out.

Even for these measurements, there was a constant struggle against the atmosphere. Uncertainties in the amount of emission from the atmosphere, particularly from water vapor, dominated the error budget. Throughout the experiment, we were worried about possible frequency-dependent systematic errors that could bias our results. I suspect it was at this stage that Dave came to recognize the value of balloon experiments, and even more of a satellite experiment to get out of the atmosphere altogether. Nevertheless, working with Bob Stokes and Paul Boynton, Dave went on to do one more ground-based temperature measurement in these years, the measurement carried out at 0.33-cm wavelength at the High Altitude Observatory in the Colorado Rocky Mountains (Boynton, Stokes and Wilkinson 1968). They found T = 2.4610 .44 K, which within the errors is consistent with the modern value. And, as a footnote, I went on to join an Italian-Berkeley-Haverford team that returned to White Mountain 15 years later to make refined spectral measurements at five wavelengths, 0.33-12 cm (Smoot et al. 1985).

What were we trying to accomplish with these early experiments? With the wisdom of hindsight, it is clear that we were beginning the process of mining the CMBR for cosmological clues. But in the years 1965-1968, the full value of spectral and anisotropy measurements was far from appreciated. The beautiful and influential theoretical work on the power spectrum of CMBR fluctuations, for instance, lay years in the future. So what were we really trying to accomplish?

First and foremost, we were trying to establish that the microwave radiation detected by Penzias and Wilson (1965a) is indeed cosmic, and not coming from more local sources in the Solar System, the Milky Way, or galaxies or from some other class of extragalactic objects. In the mid-1960s there were plenty of skeptics, and numerous noncosmological explanations of the "excess noise" reported by Penzias and Wilson. We recognized that strong proof of cosmic origin lay in two fundamental tests: the blackbody shape of the spectrum of the radiation and its isotropy on both large and small scales.

Electromagnetic radiation pervades the universe. At radio wavelengths it is dominated by the emission from galaxies and quasars. Could the "excess noise" detected by Penzias and Wilson simply be the high-frequency tail of this background? The spectrum holds the key to the answer. Emission from radio galaxies is typically dominated by the synchrotron process, producing a power-law spectrum uv <x v-a with a generally in the range 0.5-1.0. This is very different from a thermal spectrum where uv rc v2 at long wavelengths. Another possibility is "free-free" emission from a thin plasma with non-relativistic electrons, which typically produces a power-law spectrum with a ~ 0.1. Such a spectrum, too, is easy to distinguish from the truly thermal or blackbody spectrum expected from radiation left over from a hot, dense state of the early universe.

More difficult to distinguish from a true blackbody spectrum is gray-body - emission from an optically thin but higher temperature source. At wavelength A » 0.3/T cm, with temperature T measured in kelvin, gray-body emission can have the same v2 dependence as true blackbody emission, but the spectrum peaks at shorter wavelengths. To confirm true blackbody emission at T ~ 3K, we needed both to confirm the v2 dependence at long wavelength and find evidence for the peak expected near 0.1-cm wavelength.

It is worth repeating how unlikely it is to find a purely thermal spectrum in the cosmic setting, where densities tend to be very low. Only if the universe were many orders of magnitude denser than it is now, could true thermal equilibrium have been established. If the microwave background radiation truly does have a thermal spectrum, it not only establishes the cosmic origin, it also shows that the early properties of the universe were radically different from those prevailing today.

By 1967 we had the answer: we were seeing curvature in the spectrum consistent with a peak at a wavelength of about 1 mm (Stokes, Partridge and Wilkinson 1967; Wilkinson 1967).

Isotropy is the second test. An observer not moving with respect to the comoving coordinates of the universe discussed in Chapter 2 (p. 9) would see that radiation left over from the big bang appears isotropic (apart from the disturbances caused by the departure from a smooth mass distribu-tion).13 On the other hand, a Solar System origin would be expected to

13 Parenthetically, this would not be the case if the universe itself were expanding in an anisotropic way. Ellis (p. 380) notes that this idea was of considerable interest in the 1960s.

produce intensity variations tied to coordinates fixed with respect to the direction to the Sun in the sky. Sources in our Galaxy would presumably produce radiation that peaks in the direction of the galactic plane, in an anisotropic distribution akin to the concentration of bright stars in the plane of the Milky Way. Such a distribution would introduce a large dipole moment, and particularly a quadrupole moment, into the distribution of CMBR intensities. If the radiation were somehow produced by a myriad of extragalactic radio sources, as suggested for instance by Wolfe and Bur-bidge (1969), it would be "grainy" on a small scale. More precise limits on anisotropy on both large and small scales would, we hoped, kill off these noncosmological explanations. This hope motivated our work, and we soon showed (Partridge and Wilkinson 1967; Smith and Partridge 1970) that the radiation is indeed highly isotropic on both large and small angular scales.

Nor were challenges to the cosmic origin of the CMBR mounted solely by inventive theorists. At least one experimental result, the pioneering rocket measurement of Shivanandan, Houck and Harwit (1968; see Harwit's piece here) seemed to favor a graybody spectrum. The results naturally raised doubts about the cosmic origin of the microwave background.

All of these results, attacks on the very notion of the "primeval fireball," were very much on our minds as we mounted the experiments described above and wrote up our results.

One anecdote encapsulates the skeptical air of the times. In 1969, as I recall, I gave a talk on our Yuma experiment at a meeting at Caltech. In the question period, a formally dressed, middle-aged man in the back asked, in effect, "Given that you see no change in emission as the sky passes overhead each day, how do you know your equipment is even switched on?" Fortunately - since the questioner was Charles Townes -1 gave an appropriate answer, describing in detail the care we took to calibrate the instrument.

So I would say that in the 1960s, we were on a mission to convince the skeptics, an attitude that strongly colors an early review of the primeval fireball published in the spring 1969 issue of the American Scientist (Partridge 1969). I would suggest, however, that there was another influence at work. We were, after all, working for Bob Dicke, acknowledged as the master of the beautiful null experiment. These are experiments designed to test the absence of some physical effect by establishing more and more stringent upper limits on the magnitude of the effect. Dicke's ultrasensitive version of the Eotvos experiment, for instance, showed that there are no differences in the way gravity acts on different chemical elements to a level of roughly one part in 1011. I will speak only for myself here, but part of the motivation driving me was to do better and better null experiments on the CMBR, in particular to establish lower and lower upper limits on possible fluctuations in the CMBR (see Partridge 2004). In other words, I am confessing to having been driven less by theoretical concerns or predictions than by an experimenter's lust to do the best possible experiment, and to cover as much parameter space - in this case sensitivity and angular scale - as possible. As Dave and I were planning better experiments to push down limits on the amplitude of the temperature fluctuations on degree scales, I was also thinking about ways to limit anisotropies on small angular scales. This required the use of larger aperture devices, since the angular scale goes approximately inversely as the diameter. Others - Ned Conklin (1969), Eugene Epstein (1967) and Penzias, Schraml and Wilson (1969) -had already set upper limits on arcminute-scale fluctuations; Paul Boynton and I realized we could reach both smaller angular scales and higher sensitivity using a 36-ft telescope operated by the NRAO. Our results turned out to be only mildly interesting, and I mention them simply because they reflect at least one person's motivation in these early years - to set the best possible limits on fluctuations at all angular scales.

That I was not alone in this aim is reflected in the way in which anisotropy measurements were presented in these early years, and for at least a decade afterward (Figure 4.29). What is shown is basically a plot of upper limits on

-3 r-|—i—i—i—|—i—i—i—i—|—i—i—i—i—|—i—i—i—i—|—i—i—i—i—|—i—i—i—r

0 0

Post a comment