CosmoCoffee

The authors compare how much information there in the BAO peak compared to the overall shape of the large-scale structure power spectrum. They basically conclude that at present, the information in the LSS data (when combined with the CMB) is dominated by the BAO scale.

For me, the model-independent extraction of the BAO scale via spectral analysis was particularly interesting.

There is one statement which I did not understand. On page 6 the authors explain the positive correlation between the dark energy equation of state [tex]\omega[/tex] and the primordial spectral index by saying that as [tex]\omega[/tex] becomes closer to zero (i.e. grows), the late ISW becomes larger. I would have thought that as [tex]\omega[/tex] goes to zero, the late ISW effect would vanish.

They're using CAMB, which if I remember correctly defaults to always using a sound speed of 1 for dark energy, regardless of what [tex]w[/tex] is set to. So I don't think dark energy quite behaves like CDM as [tex]w\to0[/tex], which seems to be what you're thinking of.

But I think what's more important for late ISW is that as long as [tex]w<0[/tex], a larger [tex]w[/tex] (closer to zero) means dark energy remains dominant back to higher redshift than for a more negative [tex]w[/tex], and this longer duration of dark energy dominance leads to more ISW. To put it another way: if [tex]w\to -\infty[/tex], then you get zero late ISW, because [tex]\Omega_{de}[/tex] was zero up until a fraction of a second ago.

I see, that makes sense.

Syksy Rasanen wrote:The authors compare how much information there in the BAO peak compared to the overall shape of the large-scale structure power spectrum. They basically conclude that at present, the information in the LSS data (when combined with the CMB) is dominated by the BAO scale.

I would just point out that this is a model-dependent as well as dataset-dependent statement. In the SDSS DR7 analysis, we found that using the shape information improved constraints on both [tex]m_{nu}[/tex] and [tex]N_{eff}[/tex] when combined with the CMB alone (and considering these two parameters separately). However, in this paper they're allowing both parameters to vary simultaneously (which I would guess are highly degenerate in P(k)); moreover, they've included the Riess et al. [tex]H_0[/tex] constraint, which already buys you a lot in terms of breaking degeneracies with the CMB on these parameters.

In any case, it's a very interesting paper and I think they've made good improvements to how the likelihoods are implemented and clarified some confusion I had about generalizing BAO constraints to models with [tex]N_{eff} \neq 3.04[/tex].

Right, I should have specified that for extended models, there is extra information in the overall shape. The authors point this out for the example of a variable number of neutrino species together with a variable dark energy equation of state. In this case, the neutrino masses, equations of state and the spectral index all benefit from the shape information.

How important do you think is the [tex]H_0[/tex] prior?