Historically, studies of galaxy shape alignments have focused on the role of the environment on the formation and evolution of galaxies. However, the diversity of results in the literature also highlight the difficulty of measuring the observational signatures. Closer examination shows that the findings are expected to depend on the methods used to quantify galaxy shapes. The importance of intrinsic alignments for the interpretation of weak lensing measurements has renewed interest in this field of research, with larger data sets expected to become available soon.
Shape estimation itself continues to be an area of active development because improvements on current shear measurement pipelines (e.g. Mandelbaum et al., 2014) are necessary to cope with the statistical accuracy of next-generation surveys. The determination of the intrinsic alignment signal depends critically on the adopted algorithm for shape measurement. Consequently, results from early measurements of galaxy shape alignment cannot be readily applied to current studies. On the other hand, it is also not clear that the algorithms used in lensing studies are optimal for the study of environment-dependent galaxy shape alignments. Regardless, a careful accounting for a range of observational biases is essential, and progress in either field will benefit from advances in the other. The last decade has also seen the development of a sophisticated set of statistics and estimators that allow for the measurement of ellipticity-density correlation on large scales and the development of a theory which matches the data at linear scales. Results have so far largely been limited to two-point statistics, but with the advent of larger data sets, which can prove smaller physical scales, we expect that the study of higher-order statistics will gain further interest.
The existence of intrinsic galaxy alignments is now well established, to a large extent thanks to modern wide area surveys, such as the SDSS, where we need to distinguish between galaxy type and the physical scales involved. For early-type galaxies significant detections have been reported: they show a clear intrinsic alignment signal, in good agreement with the linear alignment model at scales > 10 Mpc / h. At intermediate scales the non-linear alignment model, which uses the non-linear power spectrum to boost the amplitude of the predicted signal, provides a good fit to data. We stress, however, that there is currently no theoretical justification for this phenomenology and more work on these quasi-linear scales is required. At smaller scales, < 2−6 Mpc / h, these models fail to explain the strength of the observed correlation and attempts to model local alignments through, for example, the halo model are required. The halo model tends to match the observations well, though the number of free parameters provides a good deal of freedom in the fit. There is no significant evidence for redshift-dependence of the signal beyond that which is already included in the linear alignment model, but there exists strong evidence for a dependence of intrinsic alignment on luminosity, with the brightest galaxies exhibiting stronger alignments.
In contrast, for late-type galaxies the situation is less clear. Observational constraints are more limited because of the limited spectroscopic coverages for large samples, and no statistically significant evidence for shape alignment has been detected from such surveys. This null signal is consistent with the quadratic alignment model at linear scales, though a higher number density in observations would be desirable to reduce the statistical errors of current measurements. Some papers that studied the alignment of disc galaxy spin vectors, as an alternative to direct measurement of the ellipticity correlation of spirals, did see evidence, albeit at low significance, of correlation at small scales ≲ 1 Mpc / h.
Hence, the picture at large scales, where correlations between large, diverse galaxy samples of galaxies are considered, is starting to become clear as powerful datasets become available. The results at scales where the morphology of the local large-scale structure becomes important are more ambiguous. This is not surprising as the intricacies of astrophysics, galaxy formation physics and galaxy evolution history are complicated and far from perfectly understood. The existing literature tends to classify the alignment of galaxy shapes on small scales by reference to the local morphology of the large-scale structure: galaxies located in voids, sheets, filaments or knots (groups and clusters) are expected to exhibit different shape alignment properties. In addition, the influence of galaxy type and history remains relevant at small scales.
These complications hamper a clear interpretation and comparison of the results. For instance the study of alignments for galaxies on the surfaces of voids has produced measurements that are consistent with no alignment, a significant alignment parallel to the void surface and the opposite, alignment perpendicular to the void surface. This highlights a need for more observations aimed at determining environment-dependent alignment. Even for well-defined structures such as groups and clusters of galaxies contradictory results have been reported, while we note that all the most recent studies in galaxy clusters find no evidence for shape alignments, neither from studies using spectroscopic redshifts (Sifón et al., 2015) nor those using photometric redshifts (Chisari et al., 2014). In such high-density environments the presence of nearby galaxies can affect shape measurements, especially those based on the shapes of isophotes, which may have biased earlier measurements.
A main driver of current research into galaxy alignments is the promise of cosmic shear as a powerful probe of cosmology. For this applications, the intrinsic alignments compromise a straightforward interpretation of the measurements and thus represent a dominant astrophysical source of bias. If ignored, the biases in the resulting cosmological parameter estimates are much larger than the statistical uncertainties afforded by future wide-field cosmic shear surveys using photometric redshifts such as Euclid, LSST and WFIRST (Laureijs et al., 2011, LSST Science Collaboration et al., 2009, Spergel et al., 2015). Consequently, the desire to constrain the nature of dark energy will drive much development in this field in terms of observations and the development of shape measurement and analysis pipelines.
These data will not be optimal for the study of environment-dependent alignments because of the relatively crude redshift precision. To clear up the uncertainties about the relation of alignment to morphology, and hence learn about galaxy formation and evolution, we need surveys with (near-)spectroscopic redshifts that have high galaxy density down to relatively faint magnitudes. Such data will allow a good determination of the local morphology required for unambiguous measurements of shape alignments. Many of the future spectroscopic surveys plan to survey the brightest, easiest target galaxies, which may not provide the type of data we need for environment-dependent intrinsic alignment studies but there is some reason to expect progress in the right direction. For instance, the Dark Energy Spectroscopic Instrument (DESI) bright survey (Levi et al., 2013) or Subaru Prime Focus Spectrograph (PFS) (Takada et al., 2014) may extend the sample to fainter magnitudes and higher redshifts. Wide-area observations using a large number of narrow-band filters, such as PAUCam (Martí et al., 2014) provide another avenue. These yield photometric redshifts that are much more precise than those obtained using broad-band observations, and do so for a wide range of galaxy types.
Despite the expected progress in measuring alignments, the much smaller uncertainties from future surveys with larger area will require significantly improved strategies for mitigation if we are to produce unbiased measurements of cosmology. To this end, much effort is spent on exploring general approaches that seek to exploit the different redshift dependencies of the GG, II and GI contributions. The most promising alternative involves the use of a flexible model of the intrinsic alignment contribution that includes a number of variable “nuisance parameters” which can be explored in tandem with the cosmological parameter space under consideration. The nuisance parameters are then marginalised over. If the model is sufficiently flexible, then the resulting cosmological constraints will be unbiased, albeit with a loss in overall constraining power. It has been demonstrated that, even after marginalising aggressively over uncertainties in intrinsic alignments, useful cosmological information can be gained from a photometric cosmic shear survey (Joachimi & Bridle, 2010). In this context, the goal of observational studies of intrinsic alignments, as they relate to cosmic shear systematics, can be thought of as applying more rigorous priors to the intrinsic alignment nuisance parameters. The large- and small-scale observations quoted in this review are an excellent start to this process, though it is worth noting that no paper seeking to marginalise over intrinsic alignments in the pursuit of cosmic shear has, thus far, explicitly employed priors derived from dedicated observations of intrinsic alignments.
For this reason it is also worthwhile to examine the value of complementary data to reduce or calibrate the intrinsic alignment signal. An interesting application in the near term is the cross-correlation between galaxy cosmic shear surveys and weak lensing of the CMB. Further ahead, the large density of sources at radio wavelengths that will arrive with the forthcoming SKA survey offers a number of exciting possibilities for the study of intrinsic alignments. In this case information from radio polarisation or rotational velocity measurements could allow the intrinsic alignment and weak lensing information to be separated cleanly. These constitute extremely powerful datasets, complementary to those from optical weak lensing surveys, and, if the many practical difficulties of radio weak lensing can be surmounted, they will provide important advances in our understanding of intrinsic alignments across all galaxy types, particularly much improved statistical uncertainties for late-type galaxies.
In the meantime a number of important open questions remain, including: are late-type galaxies really free of intrinsic alignment? How do intrinsic alignments evolve with redshift? Can we predict the alignments for a mix of morphological type? Addressing these questions requires additional measurements with larger number densities, covering higher redshifts. They should be a priority, as answering these questions also sheds light on the relative importance of the physical mechanisms that give rise to the alignments. The greater uncertainty at small scales is somewhat offset by the presence of competing systematics like non-linear clustering and the influence of baryon physics on the matter power spectrum. On the one hand, this means that the pressure to fully understand intrinsic alignments at these scales is reduced, as the other sources of uncertainty may make these scales less useful regardless of our intrinsic alignment knowledge. Nevertheless, future weak gravitational lensing surveys such as Euclid or LSST aim to exploit cosmic shear down to scales of 1.5 Mpc / h (Kitching et al., 2014a), so a dedicated programme of intrinsic alignment measurements at small scales would be beneficial and may require auxiliary datasets in addition to those planned for standard cosmic shear analysis. For example, good spectroscopic redshifts for reasonably well-sampled galaxies representative of cosmic shear survey galaxies would be invaluable in making accurate measurements of the relevant intrinsic alignment contamination signal.
In this review we have provided an overview of the current status of observations of intrinsic alignments, perhaps with a bias towards the impact on cosmic shear. It is clear, however, that the data that are due to become available over the next decade offer exciting opportunities to test methods for intrinsic alignment measurement and mitigation much more rigorously. The larger number density will allow us to measure intrinsic alignments in large shear catalogues with more precision, particularly for late-type galaxies, while deeper surveys will push our baseline for intrinsic alignment measurements to higher redshift. Despite all the progress, the most unclear part of the current intrinsic alignment observational landscape is certainly still the dependence on environment of the alignments on quasi- and non-linear scales. Although planned surveys may help in this regard, dedicated efforts to resolve this situation will be needed.
We acknowledge the support of the International Space Science Institute Bern for two workshops at which this work was conceived. We thank S. Bridle and J. Blazek for stimulating discussions.
MLB is supported by the European Research Council (EC FP7 grant number 280127) and by a STFC Advanced/Halliday fellowship (grant number ST/I005129/1).
HH, MC and CS acknowledge support from the European Research Council under FP7 grant number 279396.
BJ acknowledges support by an STFC Ernest Rutherford Fellowship, grant reference ST/J004421/1.
TDK is supported by a Royal Society URF.
RM acknowledges the support of NASA ROSES 12-EUCLID12-0004.
MC was supported by the Netherlands organisation for Scientific Research (NWO) Vidi grant 639.042.814.
AC acknowledges support from the European Research Council under the EC FP7 grant number 240185.
AK was supported in part by JPL, run under a contract by Caltech for NASA. AK was also supported in part by NASA ROSES 13-ATP13-0019 and NASA ROSES 12-EUCLID12-0004.
AL acknowledges the support of the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement number 624151.