As in last week’s post with TMCC2021, this week I will also be live-blogging the “A (Hubble) Tension Headache” conference “from” Southampton. As a clarification note, the conference features four plenary live talks, and a bunch of pre-recorded talks, both invited (including mine) and contributed. The conference will feature summaries and discussions of the pre-recorded talks, alongside panel discussions led by the speakers (plenary and non) of the day. Most of the pre-recorded talks will later be available on YouTube. As such, I will live-blog the plenary talks and the subsequent summaries/discussions/panels of the day, i.e. basically what is really going on during the conference. And the usual reminder of taking whatever is written here with a grain of salt always applies. Hope you enjoy it!
a (hubble) tension headache DAY 1
WENDY FREEDMAN (CHICAGO) - LOCAL MEASUREMENTS OF H0: IS THERE A CRISIS IN COSMOLOGY?
Current situation reminds Wendy of earlier discrepancies in H0 (50 vs 100), people had strong feelings about cosmological constant (did anyone really believe 50 or 100??? sounds crazy to me now haha!), finally converged towards 70
Wendy doesn’t have a favorite value for H0 (appreciate the honesty)
Didn’t known Eddington found H0~900 in the 1930s??? (the Cambridge neighborhood where I live takes its name from him) Well Hubble was also pretty off
Cepheid value has been quite steady across time, it’s really the CMB which has gone down a bit
Perhaps tension is less strong than we thought as Colin Hill advocates, or stronger (6 sigma) as Eleonora Di Valentino advocates (I really like her “brave analysis” paper!)
Interesting new test of SH0ES photometry (from Javanmardi et al. 2021, see TMCC conference diary) -> excellent agreement between TRGB and Cepheid distance to NGC 5584 -> will this stand as we probe out farther???
Double-blind work within Wendy’s group (definitely thumbs up!!!)
Comparison between TRGB and Cepheid distances is excellent
(Wendy’s dog is clearly in a lot of distress!)
JAGB method - Carbon stars as means of measuring very accurate relative distances in LMC
summaries and discussions of pre-recorded talks
Mickael Rigault
Progenitor-age step biases SNeIa luminosities (younger SNeIa fainter?). Consequences for H0, no drift of SNeIa properties, but how you construct your sample is important
Look out for Pantheon+ (>1700 SNeIa!), individual SN covariance matrix
Joint Pantheon+SH0ES analysis with joint covariance matrix, account for different DE models
Dan Scolnic
GAIA EDR3 parallaxes improve results a lot
Potential problems: Cepheid physics different locally vs more distant not a problem, Cepheid crowding not a problem, Cepheid photometry is fine (see Javanmardi et al.)
Nandita Khetan
SBF calibration of SNeIa, H0=70.50±2.37
72.8±1.7 if they use population from SH0ES sample (connect to Rigault’s talk?)
Take-away message: H0 tension could still hide systematics of astrophysical interest - make sure there is nothing under the carpet! (needless to say I agree)
Q: Replace SBF as SNeIa and get high H0? (use SBF as indicators rather than calibrators)
Suhail Dhawan
Hubble-flow rung of distance ladder (low-z SNeIa with ZTF) - usually assume value of q0 and j0, but here have consistent covariance matrix between calibrator and Hubble flow SNeIa, look at impact on assumed DE model on H0 (also mentioned by Dan). No big impact on H0, though I guess you can’t be 100% sure there doesn’t exists a model which will have a big effect. General take-away message: local H0 stable against assumed DE model
Testing new standardization techniques (rising part of light curve), can test role of SNeIa environment on inferred H0, can also do direction-dependent DE studies
Can do lots of interesting science with iPTG16geu
Can ZTF help with environmental effects?
Martin Millon
TDCOSMO, most conservative result assume SLACS and TDCOSMO lenses have same anisotropy and radial mass density properties, H0~67.4 +4.1 -3.2
Correlations between H0 and redshift due to underestimating convergence or lensing? (very interesting trend, I think there is real physics there) Correction for external convergence just reduces scatter, can’t be responsible for trend
What’s main reason for revised lower value? Mass-sheet degeneracy and/or galaxy evolution effect
Tom Collett
Discuss SLACS vs TDCOSMO lenses, not necessarily from the same family -> Tom thinks this is a very dangerous/bold assumption (not surprising that H0 with SLACS constraints as prior is low)
Tom thinks 74.5 +5.6 -6.1 or 73.3 +5.8 -5.8 are much safer values, else you thA cprow away a lot of the valuable work of H0LiCOW assuming things that are not necessarily true about lens families, he wouldn’t have trusted the lower values even before unblinding
JO DUNKLEY (princeton) - the hubble constant from the cosmic microwave background
H0 tension really started to appear after WMAP when higher acoustic peaks were measured
Two questions to ask from the CMB in the future according to Jo: 1) is the Planck H0 robust assuming LCDM? 2) Are there non-LCDM models which can increase H0 and still fit the CMB? Jo says 2) is a significant challenge, true, but I think it’s even more challenging to fit CMB+BAO etc.
Is the Planck H0 robust assuming LCDM? (important analysis by Efstatiou & Gratton), important to do alternative analyses to Planck, measure polarization better
Future is from ground-based data: Atacama desert and South Pole, can measure many acoustic peaks in polarization which improve the inferred value of H0
ACT: H0=67.6±1.1, great consistency with Planck (surprise when unblinding)
A completely independent CMB analysis (different data, different methods) gives consistent result with Planck, this answers question 1: the Planck H0 robust assuming LCDM. SPT consistent too
Why doesn’t Planck want H0=74? Residuals would be too big, bad fit in both temperature and polarization between 500 and 2500, lower matter density significantly (Omegam~0.24 too low), but you also need to change Omegabh2 and ns, gets peak height wrong, everything wrong, fails!!!
How to get high H0 from CMB? Keep sound horizon and distance to CMB fixed, change contents or geometry; shrink both sound horizon and distance to CMB, gives higher H0 without changing late-time ingredients. Difference is in whether you keep triangle fixed or shrinked
Interacting neutrinos change polarization quite a lot (which is why the model doesn’t work)
Will ACT data have a low enough noise or do we need to wait for CMB-S4? By 2021 with ACT sigma_H0 should be 0.4, according to Jo will gradually exclude alternative models
What does Jo think about A_lens, statistical fluke or real physics? Probably a fluke
summaries and discussions of pre-recorded talks
Colin Hill
Early dark energy inconsistent with LSS data
Is this due to prior volume effects? No
Potential biases for future H0 due to NL CMB lensing -> NL evolution and baryonic effects
Oliver Philcox
H0 tension is no longer just CMB vs distance ladder, really CMB+LSS vs distance ladder
Matter P(k) has two interesting scales: equality scale and sound horizon (both standard rulers)
Only information from equality scale k_eq: H0=65.1 +3.0 -5.4
Comparison of equality and BAO constraints can test new physics at high z!
Given what we know about Omegabh2 today, Euclid constraints would mostly come from k_eq
Leonardo Senatore
H0=68.7±1.5 (LSS+BBN prior)
Strong reassurance about lack of systematic errors in CMB and LSS
Clustering quintessence can do w<-1, gives low H0 (known from inverse distance ladder)
EDE gives low H0, doesn’t agree with response by Kamionkowski’s group
Interesting analysis with A_LSS (similar to A_lens, but for CMB vs LSS)
Nikki Arendse
Phrase H0 tension in terms of sound horizon
BAO+SNeIa+SH0ES+CCHP+H0LiCOW: cosmology-independent analysis (is combination fine?)
Focusing on H0 obscures the real picture (I think one should go beyond H0-rd e.g. Omegam, tU)
Pablo Lemos
What features in Planck prefer a low H0?
(Note to self: never register to a conference with your old institutional email if you are moving!)
Take Planck and fix H0=74 (of course usually you wouldn’t want to do this, see Colin’s talk above!!!)
Really low Omegach2, ns=1 (coincidence?), but not a disaster by eye (although Delta chi^2=+42)
Mainly affects damping tail (obvious a posteriori? Since with WMAP there is no H0 tension)
panel discussion 1
Mat Smith & Lisa Kelsey moderating discussion
Mat was impressed by consistency between TRGB and Cepheids - so where is the issue at low redshifts if there is any?
Wendy: it’s a combination of factors (e.g. Cepheid distance gets high weight in analysis), there definitely is a systematic offset which has to be solved
Dan: agree with Wendy but main systematics on SNeIa side at ~0.3-0.5 sigma level, to break SNeIa enough to fix LCDM you actually have to break LCDM (SNeIa are not the key out of here)
I think we all agree that the only way to get to high H0 is to break LCDM or systematics in calibration which don’t affect SNeIa alone
Dan to Thomas: why don’t you just quote one single H0? Simply we didn’t really have all systematics under control, not all the collaboration is agreeing on the details of the lenses (I guess the point is that we are human so there is some spread in what results we believe in)
Colin: strong lensing has high enough error bars that they don’t distinguish between high and low H0 (so they don’t arbitrate the tension)
Thomas’ standard view: yes this is true, but this is not the standard view (very conservative position)
Wendy: kudos to TDCOSMO folks for having the courage to relax their assumptions (something I am definitely going to agree with)
Unknown unknowns? Bad question according to Dan, just give us the systematic and we’ll check for it (which is something the SH0ES team has been doing for a long time, think about z-dependent SNeIa luminosities, voids, Cepheid crowding, etc. - Wendy disagrees and thinks that we have to keep in mind possible systematics we don’t know about, and the only way is to make use of other methods which are completely independent
Jo: tension between Cepheid and TRGB is just 2.4 sigma, so why all this fuss? Wendy: we need to compare with all other late-time measurements, and there the tension is >3 sigma
Thomas: as a non-SNeIa person, how do I choose the calibration (Cepheid vs TRGB) to go for? Dan: average the two, you get 71.5, you can pick your favorite, but nobody likes picking kids. Thomas: neither of you knows what you’re doing (laughs too)? Dan: we’re different teams, we’re allowed to disagree with each other (laughs too), you’ll just have to keep watching. Wendy: we’re actually converging, especially considering where we came from, we’ll solve this
Saurabh: what is covariance between whole calibration of Cepheid vs TRGB, what is the real sigma?
Mat: is there a model on the horizon which can solve the H0 tension?
Jo: interacting neutrinos (doesn’t work with polarization, at least in the minimal implementation)
Leonardo: in the mathematical space it might be possible to find this model
(was very hungry at this point, went to have dinner)
a (hubble) tension headache DAY 2
ULF DANIELSSON (UPPSALA) - WHAT IF THE HUBBLE TENSION IS FOR REAL?
(It’s been so long since I heard such a strong Swedish accent as Ulf’s, love it, reminds me of the “good old” years in Stockholm!)
Historical background: perturbations in orbit of Uranus -> Neptune, perturbations in orbit of Neptune -> measurement errors (wasn’t aware of this), motion of mercury -> GR, no static solution for Einstein or Newton (Newton thought about a divine solution??? That was a big surprise to me)
Of course Ulf point out Knut Lundmark’s discovery of Hubble’s law :) H0~100 (better than Hubble’s value which was ~500!!!)
Running cosmological constant? w<-1, bad since violates NEC (but CC violates ECs anyway…)
Is it Uranus, Neptune, or Mercury? Provocative question asked by Ulf: is there any change in the fundamental laws or new component of matter (regardless how crazy it is) that can fit all data?
Ulf: fundamental physics ? string theory :) AdS5 with negative CC, unstable AdS is used to get AdS to decay into dS -> DE from decay of dark bubbles? Gives dark radiation and positive curvature
Conclusion: bet 10$ that the Hubble tension is Neptune -> systematics
summaries and discussions of pre-recorded talks
Eleonora Di Valentino
LCDM based on three (non-motivated?) dark components/assumptions: DM, DE, inflation
How robust is BAO to beyond-LCDM cosmologies?
A_lens>1 can solve H0 tension (but remember A_lens is not phenomenological)
Nima Khosravi
(Zoom is forbidden in Iran?!?)
How to think about a tension? Think of more temporal and spatial anomalies at the same time
DE experiences phase transition -> Ginzburg-Landau EFT -> DE emerges at some redshift
Eoin Ó Colgáin
(always wondered if the Ó bit is part of the surname? Anyone from Ireland has any idea? Yes)
Interesting argument on H0 being an integration constant, running H0 sign of beyond-LCDM???
Throw away models where H0 is lower, true for many quintessence models from string theory
Yours sincerely
Won’t take notes for this obviously, but you can watch my talk on YouTube
Tanvi Karwal
I think it’s the first time I heard the phrase “LSS tension” - I bet it will stick! :)
EDE worsens LSS tension, but this tension already exists within LCDM (A_s)
Split PPS amplitude between CMB and LSS (reminds me of A_lens, so it is unphysical?)
Why can’t EDE address H0 tension without turning on perturbations?
Answer: EDE has huge ISW effect, perturbations suppress these so don’t change CMB so much
Shouvik Roy Choudhury
Massive neutrino self-interactions (mediated by heavy scalar or vector, >1 keV)
Key: including small-scale polarization, previously solution came from Neff-H0 correlation
Take-away message: self-interacting neutrinos no longer can solve the H0 tension
Anirban Das
Neutrino self-interactions in conflict with laboratory data (new decay channels) and BBN
Need flavor-specific interactions (only 1 or 2 neutrino states interact), not necessarily unmotivated
Only tau neutrinos couple, still tension (didn’t vary Neff and Mnu)
Shao-Jiang Wang
Evade necessity for early-time solutions through inhomogeneities
Couple chameleon field to local inhomogeneities, need ~20% overdensity (Great Wall)
Possible interesting connections to H0 vs z descending trend seen in H0LiCOW
Lu Yin
Proca field (massive spin-1 field) is possible DE candidate (action non-analytical?)
DE EoS different values at different redshifts depending on free parameter s (-1-4s/3 -> -1-s -> -1)
Once include BAO, H0 is low, so usual problems as other models!
panel discussion 2
(not taking notes for this as I am one of the panelists, not too good at multitasking!)
(the following are brief notes written ~15 hrs after the discussion, based on my recollection)
Interesting question about BAO model-dependence. My answers: real issue is if differences in the distance-redshift relationship between true and fiducial cosmology don’t scale linearly, then usual AP scaling breaks down, can happen if have large metric gradients (e.g. large late-time inhomogeneities, backreaction) -> in this case can use FS power spectrum, but still need to assume fiducial cosmology to convert redshifts to distances -> Oliver: redo things self-consistently beyond linear level (true, but difficult in practice, you need mocks to estimate covariances, how do you get mocks for a crazy theory?). Take-away message: important to get BAO experts in the H0 tension discussion!!!
Where do we draw line for not combining datasets? Clearly subjective. 2 sigma for me, ~2.5 sigma for Tanvi. For Jo (and the ACT team) 2.6 sigma is consistency, for Eleonora it’s tension. My thought: real question is whether something in grey zone between 2 and 3 sigma is fluke or not. Try to find out using either more data (more sky) or alternative experiment looking at same sky (e.g. ACT vs Planck), or take empirical tension metrics Jeffreys-like scale, combine only if mild tension
Question from Kostas: should we think bigger from the theory side? Example by Nima and myself: inflation (works by Shafieloo, Starobinsky, Sahni, Keeley, L’Huillier). Colin: difficult to match polarization. Ulf: can we push early modifications very early so they don’t disrupt CMB? Tanvi: difficult, think of kernel quantifying response of sound horizon to changes in H(z) as a function of redshift. I’d like to think about this more, maybe something super-crazy super-early?
Is there something in the H0LiCOW H0 vs lens redshift trend? Thomas: probably not, most likely a systematic, something with evolution of galaxy profiles as a function of redshift (?), but if it were real it would be intriguing (e.g. G_matter vs G_light? See recent works by Eric Linder)
My general thoughts: important to focus on right quantities, so far mostly incremental leaps, huge leaps in understanding H0 tension when realized focusing only on H0 obscures the real story, e.g. Bernal, Verde & Riess (later Aylor et al.) told us to look at rs as well, now Jedamzik, Pogosian & Zhao tell us to look at omegam too, I think next important quantity will be tU (age of Universe), once we nail down right quantity we will make the next big leap in understanding the H0 tension
a (hubble) tension headache DAY 3
BERNARD SCHUTZ (CARDIFF) - WILL GRAVITATIONAL WAVES HELP RESOLVE THE TENSION?
Binary inspiral phase very well understood, and distance measurements comes from inspiral phase, not merger phase, so it is a very clean measure, plus it is a one-step measure which doesn’t need a ladder to calibrate in between
Pro: no absorbers, scatterers, dispersers, reddening (a bit of lensing but phase doesn’t change)
Cons: few events
Dark sirens will probably be the best tools for H0
How do standard sirens work? Can track phase and amplitude since GWs emitted coherently, most of information comes from phase. GR scale-free, everything scales with M. Set c=G=1, luminosity dimensionless, but timescales and total energy proportional to M (but luminosity can depend on dimensionless numbers like mass ratios)
Complications: need to measure orbital inclination, Malmquist bias, errors biased to large values
GW190521 gives very low H0 ~40-50????? Maybe highly eccentric. Otherwise other values ~70
GW190814 first event with measured higher-order radiation modes, helps to measure inclination (note that it was a dark siren), get H0~69 (can one do anything more with higher-order radiation modes? Not sure if one can learn anything about fundamental physics)
Future: LISA, Taiji (?), Einstein Telescope, Cosmic Explorer, we should see all BNS mergers!
Bernard bets on dark sirens in Advanced LIGO-Virgo era before 2025, will still be useful
Stephen Feeney
(it’s the first time I see someone giving a talk - or rather a talk summary - while driving!)
How much does likelihood-free inference rely on good prior choices? Not too much
Why single-GW posterior asymmetric? Depends on distance prior tail
Rachel Gray
No EM counterpart, use gal catalogs to provide all possible hosts z (problem: gal incompleteness)
How much does this depend on BBH tracing stellar mass vs SMF? (not sure I caught the answer)
Why on average posteriors overestimate H0 as you add DSs? High H0 tail, more info at low z
Rupert Croft
Measure H0 from direct galaxy parallaxes (cosmic secular parallax), Rubin: 2.8%, Roman: 0.8%
There was some H0 tension in the early 1990s? (but just because errorbars underestimated)
Interesting historical discussion about accuracy and confirmation bias (relevant today too???)
Asta Heinesen
General luminosity distance for generic space-time (don’t assume FLRW or field equations)
How big corrections with respect to FLRW? Not clear, depends on many things
What observations do we need to test this? Large catalogs of standardizable objects
panel discussion 3
(probably won’t be not taking notes for this as I am one of the panelists - replacing Eleonora Di Valentino - and I’m not good at multitasking!)
(the following are brief notes written almost 24 hrs after the discussion, based on my recollection)
Basically the panelists want/look forward to/agree on:
-better CMB polarization data
-more public modified Boltzmann solvers
-more systematics checks
-more H0 measurements independent of distance ladders
-agreeing on the level of significance of the tension
-find new important quantities to look at (rs, omegam, tU,…)
-agree that this is an opportunity, not a headache
-think bigger in terms of theory space to come up with the next big idea
-this is definitely a healthy discussion, probably more so than H0 = 50 vs 100 in the past
-there is definitely a fair bit of sociology involved (ambulance chasing, big names,…)
-important to understand how model-dependent BAO are wrt possible very exotic late-time changes
-important to not consider H0 tension in isolation, but consider other tensions/anomalies too
-related to above, investigate S8 tension in a model-independent way as did with H0 tension
-important to talk to particle physicists/experimentalists searching for DM
-more interactions between theorists and observers
As with TMCC2021, this too was a fantastic conference (especially the discussion panels)!