The most obvious difference between CDM and MOND in the context of the microwave background anisotropies is the baryon fraction. CDM is thought to outweigh ordinary baryonic matter by a factor of (Evrard, Metzler, & Navarro 1996EMN). The precise value depends on the Hubble constant and on the type of system examined (McGaugh & de Blok 1998aMBa). If instead MOND is the cause of the observed mass discrepancies, there is no CDM. The difference between a baryon fraction and unity should leave a distinctive imprint on the microwave background.

The main impact of varying the baryon fraction
is on the relative amplitude of the peaks in the angular power spectrum of the
microwave background (as expanded in spherical harmonics). In general,
increasing *f*_{b} increases the baryon drag, which enhances the amplitude of
compressional (odd numbered) peaks while
suppressing rarefaction (even numbered) peaks
(Hu, Sugiyama, & Silk 1997HSS).
The precise shape of the power spectrum is thus very sensitive to *f*_{b}.

In order to investigate this aspect of the problem, I have used CMBFAST
(Seljak & Zaldarriagacmbfast 1996)
to compute the expected microwave background power spectrum in
several representative cases. These have reasonable baryon fractions for
each case and baryon-to-photon ratios consistent with primordial
nucleosynthesis. Several specific cases are illustrated in Figure 1.
These have , 0.02, and 0.03 with or 0 (so *f*_{b} = 0.05, 0.1, 0.15 or 1). Other model parameters are held
fixed (*h* = 0.75, *T*_{CMB} = 2.726 K, *Y*_{p} = 0.24, = 3), and
adiabatic initial conditions are assumed. As a check, models with
and 0.4 were also run with the same baryon fractions
and *H _{0}* scaled to maintain the same baryon-to-photon ratio. As expected,
these resulted in power spectra which are indistinguishable in shape.

I am interested in the shape of the power spectrum, not the absolute positions of the peaks. The latter depends mostly on the scale and geometry of the universe. For purposes of computation, I assume the universe is flat, with . This results in a CDM universe close to the current ``concordant'' model (e.g., Ostriker & SteinhardtOS 1995). In the case of MOND, the resulting model is very close to the de Sitter case. This is a plausible case for a MOND universe (indeed, the relation of inertial mass to a finite vacuum energy density has been suggested as a possible physical basis for MOND: MilgromM99 1999), but is by no means the only possibility. A model with no cosmological constant and is plausible, but would be very open if the geometry were Robertson-Walker. The position of the first peak in the power spectrum moves to smaller angular scales in open universes because of the dependence of the angular diameter distance on . For such low with ,the position of the first peak occurs at . This is inconsistent with recent observations which constrain to be near 200 (Miller et al.MAT 1999). However, the geometry in MOND might not be Robertson-Walker, so the position of the first peak is not uniquely specified. It is important to realize that while the position of the first peak provides an empirical constraint on the geometry traversed by the microwave background photons, in the context of MOND this does not necessarily translate into a measure of .

The test is therefore not in the absolute positions of the peaks, but in
the shape of the spectrum. As the baryon fraction becomes
very high^{}
(), the even numbered peaks are suppressed to the point
of disappearing. One is left with a spectrum that looks rather
like a stretched version of the standard CDM case.

The difference between the CDM and MOND cases is obvious by inspection
(Figure 1).
However, from an observer's perspective, it is not so easy to distinguish them.
The second peak has disappeared in the MOND case, so what would have been
the third peak we would now count as the second peak. The absolute positions
of the peaks are not specified *a priori* by either theory. The
absolute amplitude in the CDM case is constrained by the need to match
large scale structure at *z*=0. The mechanics to do a similar exercise
with MOND do not currently exist, so the absolute amplitude is also not
specified *a priori*. We must therefore rely on the relative
amplitudes and positions of the peaks to measure the
difference. Since the third peak becomes the second peak in MOND, the
observable difference is rather more difficult to perceive than one might
have expected, at least for the assumptions made here.

The ratios of the positions and amplitudes of the peaks are given in Table 1.
The peak position ratios depend on the sound horizon at recombination,
which should not depend on MOND (for constant *a _{0}*) because this is well before
the universe approaches the low acceleration regime.
Other parameters do matter a bit, which can complicate matters.

One difference we could hope to distinguish is in the ratio of the positions of the first and second peaks. In the CDM models, , while in the case of MOND . This requires a positional accuracy determination of beyond , no small feat.

If we can recognize that second peak is actually missing, so that what we called the second peak in MOND actually corresponds to the third peak in CDM, then the distinction is greater: for CDM, , which should be compared to MOND's 2.66. It is not clear how to do this observationally. Once the position of the first peak is tied down, the given ratio predicts the expected position of the second observable peak (under the assumptions made here). This is not very different in the two cases.

The ratios of the positions of the next observable peaks help not at all. For CDM, . For MOND, .

The ratio of the absolute amplitudes of the peaks can also distinguish the two
cases, but require comparable accuracy. In CDM,
,while in MOND .This may appear to be a substantial difference, but recall that what is measured
is the temperature anisotropy. Since ,one requires accuracy to distinguish the two cases at the
level. The
amplitude ratios of the second and third peaks have a bit more power to
distinguish between CDM and MOND, but are more difficult to measure.
The precise value of this ratio is very sensitive to *f*_{b} in the CDM case.
In CDM, for *f*_{b} > 0.05,
while in MOND .

Using the absolute amplitude of the peak heights does not untilize all the
information available. In the purely baryonic MOND cases, there is a longer
drop from the first peak to the first trough, and a shorter rise to the
second peak than in the CDM cases. Therefore, measuring the peak heights
relative to the bottom of the intervening trough may be a better approach.
To do this, we define to be the ratio of the amplitudes at maxima *n* and *n*+1 less the amplitude
of the intervening minimum. This does indeed appear more promising.
The purely baryonic MOND cases all have , while the CDM cases have (Table 1). This is a nice test, for in most cases this ratio
falls well on one side or the other (for ,).

By inspection of Figure 1, one might also think that the width of the first peak could be a discriminant, as measured at the amplitude of the first minimum. This is a bit more sensitive to how other parameters shift or stretch the power spectrum. It is also very sensitive to the neutrino mass. Baryonic models with zero neutrino mass have perceptibly broader peaks than the equivalent CDM model, but zero CDM models with finite neutrino mass have peaks which are similar in width to those in the CDM models.