Evidence for microlensing by primordial black holes in quasar broad emission lines

With the detection of black hole mergers by the LIGO gravitational wave telescope, there has been increasing interest in the possibility that dark matter may be in the form of solar mass primordial black holes. One of the predictions implicit in this idea is that compact clouds in the broad emission line regions of high redshift quasars will be microlensed, leading to changes in line structure and the appearance of new emission features. In this paper the effect of microlensing on the broad emission line region is reviewed by reference to gravitationally lensed quasar systems where microlensing of the emission lines can be unambiguously identified. It is then shown that although changes in Seyfert galaxy line profiles occur on timescales of a few years, they are too nearby for a significant chance that they could be microlensed, and are plausibly attributed to intrinsic changes in line structure. In contrast, in a sample of 53 high redshift quasars, 9 quasars show large changes in line profile at a rate consistent with microlensing. These changes occur on a timescale an order of magnitude too short for changes associated with the dynamics of the emission line region. The main conclusion of the paper is that the observed changes in quasar emission line profiles are consistent with microlensing by a population of solar mass compact bodies making up the dark matter, although other explanations like intrinsic variability are possible. Such bodies are most plausibly identified as primordial black holes.


INTRODUCTION
The recent detection of black hole mergers by the LIGO gravitational wave observatory has in the first instance been attributed to the merging of massive stellar black hole remnants (Abbott et al. 2016).However, it has also been seen as adding considerable weight to the idea that dark matter is in the form of primordial black holes (Bird et al. 2016).The idea was that the detections would form part of a high mass tail to an expected broad mass function peaking at around a solar mass.From a theoretical perspective, a mechanism for the formation of primordial black holes in the early Universe was discussed by Carr & Hawking (1974), with the additional suggestion (Chapline 1975) that such objects might make up the dark matter.Constraints appearing to rule out this idea on the basis that it implied excessive, unseen variations in quasar brightness due to microlensing by the black holes acting as lenses (Schneider 1993) were shown to be insecure by Zackrisson & Bergvall (2003), on the basis that an unrealistically small source size had been assumed for the quasar accretion disc.
The first claim that primordial black holes had actually been detected (Hawkins 1993) made the case that the observed brightness variations in samples of quasars could only be explained by the effects of microlensing by a large population of compact bodies, most plausibly primordial black holes.These bodies would have to make up a large fraction of the dark matter.In the ensuing years, a number ★ E-mail: mrsh@roe.ac.uk of varied observations suggested that optical variability in quasars was at least partly the result of microlensing by a population of stellar mass compact bodies.These results were summarised by Hawkins (2011) as a case for primordial black holes as dark matter.
The idea that dark matter is in the form of compact bodies has a long and controversial history.An early review by Trimble (1987) largely focussed on a variety of elementary particles as dark matter candidates, although she did include a number of compact objects including cosmic strings, quark nuggets and primordial black holes as alternative possibilities.It seems fair to say that the consensus view at that time was that dark matter was in the form of elementary particles, which would soon be detected by one of a number of ongoing experiments.In the event, particle dark matter has not so far been detected, and as detection limits approach the neutrino floor, the prospects for such detections are not good.There have nonetheless been a number of attempts to detect or put limits on any population of compact bodies which might contribute to or account for the dark matter.Perhaps the most significant of these was the large scale survey by the MACHO collaboration (Alcock et al. 2000) to detect microlensing events in the light of Magellanic Cloud stars by compact bodies along the line of sight in the Galactic halo.The results of this project were controversial, as although the number of events detected exceeded any known stellar population, it was less than that expected for a halo entirely composed of compact bodies.
Since the publication of the results of the MACHO collaboration (Alcock et al. 2000), there have been a number of attempts to constrain any population of compact bodies making up the dark matter.
In the first instance the microlensing observations were repeated by two other groups, the EROS and OGLE collaborations (Tisserand et al. 2007;Wyrzykowski et al. 2011).The results of these two surveys were in significant ways inconsistent with those of the MACHO collaboration (Hawkins 2015), which led to a variety of new approaches to constraining dark matter in the form of compact bodies, and in particular primordial black holes (Carr et al. 2010(Carr et al. , 2017)).The constraints included so-called pixel-lensing in M31, brightness changes in Type Ia supernovae due to microlensing, the disruption of wide binary star systems by compact bodies, the depletion of stars in the centre of dwarf galaxies due to mass segretation, excess Xray luminosity in the Galactic centre due to interaction of compact bodies with the inter-stellar medium and distortions in the Cosmic Microwave Background due to accretion onto primordial black holes in the early Universe.However, all these constraints are based on assumptions which have been vigorously challenged in the literature (Byrnes et al. 2018).
A more direct approach to the question of whether the dark matter in galaxy halos can be in the form of compact bodies is based on analysing photometric variations in the multiple images of gravitationally lensed quasars (Mediavilla et al. 2009;Pooley et al. 2012).The idea is that although intrinsic variations in the quasar will be observed in all the quasar images, subject to time delays approropriate to the light travel time to the individual images, it is also the case that the images vary independently of each other.This is widely interpreted to be due to microlensing by a population of stellar mass compact bodies, where the light from each quasar image traverses a different amplification pattern on its trajectory to the observer.The mass estimate is derived from the characteristic timescale of the events, and the question of interest is whether the lenses are stars in the lensing galaxy, or compact bodies making up the dark matter halo.
The conventional approach to determining the stellar fraction of the lensing galaxy halo has been to use a maximum likelihood estimate of the ratio of mass in compact bodies to that in smoothly distributed particles, based on the observed microlensing amplifications.This procedure has tended to give low values for the ratio of compact to smoothly distributed matter, consistent with the stellar population of the lensing galaxy acting as the lenses (Mediavilla et al. 2009;Pooley et al. 2012).However, this result does not agree with direct measurements of the stellar population in the vicinity of the quasar images from the distribution of starlight in a sample of wide separation systems where the quasar images lie well clear of the stellar population, and yet are strongly microlensed (Hawkins 2020a,b).The problem seems to be that the maximum likelihood estimates are based on large samples of mostly compact lens systems, where the quasar images are buried deep within the stellar distribution of the lensing galaxy and the disc stars form a large optical depth to microlensing.In this case there is no reason to doubt that the observed microlensing can be produced by the stellar population.This is clear from the work of Pooley et al. (2012), where in a subsample of wide separation lensed quasars the observed variations are not consistent with microlensing by the relatively sparse stellar population.
A direct measurement of star light in wide separation lensed quasar systems where the images lie well clear of the stellar population of the lensing galaxies (Hawkins 2020a) has shown that stars in the galaxy halos are far too sparse to account for the observed microlensing amplifications.The most convincing evidence comes from the cluster lens SDSS J1004+4112 (Hawkins 2020b) where strong microlensing is observed in the light curves of the quasar images some 60 kpc from the cluster centre.As the optical depth to microlensing  * by the stellar population has already dropped to negligible levels 25 kpc from the cluster centre, it seems clear that the microlenses must be part of some other population of compact bodies.In addition, for quasars in the general field it has been shown that the observed variations in quasar brightness cannot be accounted for by intrinsic changes in luminosity (Hawkins 2022).The additional contribution of the microlensing amplification predicted for a population of solar mass compact bodies making up the dark matter is required to provide a good match to the data.There are a number of constraints on the identification of a population of compact bodies making up the dark matter (Hawkins 2020a).For a start they must be non-baryonic, as well as sufficiently compact to act as lenses, and the timescale of microlensing events implies that the mass of the lenses must peak at around a solar mass.These constraints appear to rule out all known candidates for the compact bodies apart from primordial black holes (Hawkins 2020a).In addition, there is a strong theoretical framework for the creation of primordial black holes in the early Universe (Byrnes et al. 2018), with a mass function peaking at around a solar mass.
For a quasar accretion disc to be microlensed, it must be more compact than the Einstein ring associated with the lensing objects.For most cosmological situations this implies a lens of around a solar mass for a typical quasar accretion disc.The question of whether the broad emission lines are microlensed is less clear, and depends on the structure of the broad emission line region (BLR).At the time of early work on microlensing (Schneider & Wambsganss 1990) it was generally believed that the BLR was large, of the order of a light year, and an order of magnitude larger than a typical Einstein ring for a solar mass lens.On this basis, Schneider & Wambsganss (1990) concluded that the flux from the BLR would not be significantly affected by microlensing unless the internal structure of the BLR was nonuniform.The idea that the BLR might not be uniform, but confined by magnetic stresses (Rees 1987) was developed by Bottorf & Ferland (2001) to argue that the clouds within the BLR are not isolated individual entities embedded within a confining medium, but transient knots of higher density within an overall turbulent BLR.This idea was developed by Lewis & Ibata (2006) who showed that although the integrated microlensing effect of a fractal structure would result in an overall constant light curve, significant magnification of substructures could alter the emission line profiles in the BLR.
One of the first gravitational lens systems to be analysed for microlensing effects is the quadruply imaged Q2237+0305, also known as the Einstein Cross.A relative change in the brightness of two of the images was definitively observed by Irwin et al. (1989), and was in fact the first detection of a microlensing event.Q2237+0305 has continued to be a valuable laboratory for exploring the effects of microlensing of the continuum source associated with the accretion disc, and was an obvious choice for early investigations into the possible microlensing of the BLR.The first such study (Wayth et al. 2005) was largely focussed on measuring the ratio of the size of the C iii]/Mg ii regions.In the process they convincingly showed that the broad lines were being microlensed.They noted that the flux ratios for the two BLRs were consistent with each other, but not with that for the continuum.From this they concluded that the two emission regions were of the same size, and located along the same line of sight.
Since this early work there have been extensive efforts to look for the effects of microlensing on broad line regions in gravitationally lensed quasar systems (Sluse et al. 2012;Guerras et al. 2013;Fian et al. 2021).From these studies it is clear that in most of these systems the broad line region is being microlensed.This conclusion is primarily based on the observation that after allowing for time delays beween the images, the structure of the emission lines is significantly different.This is interpreted as a consequence of the differing amplification patterns due to a population of stellar mass lenses traversed by light rays from each image to the observer.The question of the nature of the lenses is not addressed in these papers, but the implication seems to be that they must be stars in the lensing galaxy, as advocated in earlier work on microlensing of the continuum light from the quasar accretion disc (Mediavilla et al. 2009;Pooley et al. 2012).This orthodoxy has recently been challenged (Hawkins 2020a,b), where it is shown that in wide separation lens systems the stellar population of the lensing galaxy in the vicinity of the quasar images is far to sparse to be responsible for the observed microlensing.
The idea behind the present paper is as follows.If the dark matter is indeed made up of stellar mass primordial black holes, then these compact bodies should microlens the broad emission lines of a substantial fraction of quasar spectra.This microlensing effect would result in a change in the structure of the emission lines over a period of a few years, by analogy with changes observed in the images of gravitationally lensed quasars.If no such changes are seen then this would be inconsistent with the view that the dark matter is largely composed of compact bodies.If changes in emission line structure are observed, then although these observations would be consistent with a compact body component of the dark matter, there remains the question of whether they can be attributed to intrinsic changes in the velocity structure of the broad line region.studies, a few general points can be made about microlensing of the BLR.As mentioned above, if the entire emission region is sufficiently compact, that is smaller than the Einstein disc of the lenses, then microlensing will be seen as an increase in broad line flux, but with little or no associated distortion of the line profile.However, for a larger and more complex structure of the BLR, individual knots with non-systemic radial velocities can act as lenses to create composite time-varying line profiles.The change in shape of the Si iv and C iv emission lines illustrated in Fig. 1 for SDSS J1004+4112 would appear to be an example of this.
These observations raise the question of the timescale of structural changes in the BLR.In their analysis of broad line microlensing in SDSS J1004+4112 Richards et al. (2004), as well as noting differences in line profile between images A and B, also find that for image A the profile of the C iv line changes by a large amount over a period of 6 months.This very short timescale should be compared with the dynamical or cloud crossing timescale (Peterson 1993) for the BLR in quasars, given by where  is the radius of the BLR as measured in reverberation mapping experiments, and is typically about a year for quasars (Lira et al. 2018), and  FWHM is the Doppler width of the broad line with a typical value of 6000 km sec −1 .Given that by comparing the spectra of different images it is established that the BLR is being microlensed, it seems reasonable to conclude that the very short timescale event in image A is also the result of microlensing.This would be in line with results from monitoring of the accretion disc of SDSS J1004+4112 for microlensing variations, where image C was observed to increase in brightness by 0.7 magnitudes in 200 days (Hawkins 2020b).The implication here is that the accretion disc and BLR are being microlensed by the same population of compact stellar mass bodies.

INTRINSIC VARIABILITY OF QUASAR BROAD EMISSION LINE PROFILES
It has for some time been well-established that changes in continuum flux from the accretion disc of low luminosity Active Galactic Nuclei (AGN) or Seyfert galaxies can produce variations in the strength of quasar broad emission lines (Peterson et al. 2002).In this case, the changes in emission line strength typically follow the continuum flux changes by a few days, representing the light travel time from the accretion disc to the broad line clouds.This reverberation effect has also been observed by Kaspi et al. (2000) in relatively low luminosity quasars from the PG sample of nearby quasars (Schmidt & Green 1983).More recently, reverberation mapping has been extended to samples of luminous quasars (Kaspi et al. 2007;Lira et al. 2018), and some interesting trends have emerged.It is clear that as AGN become more luminous, the associated BLR becomes more fragmented, and the response of the emission lines to changes in continuum flux becomes more patchy.
Although most changes in emission lines can be explained by a simple increase in brightness in response to changes in continuum flux from the accretion disc, changes in the line profiles of Seyfert galaxies are also observed on timescales of around 5 years (Wanders & Peterson 1996), consistent with the dynamical timescale for Seyfert galaxies,  dyn ≈ 3 − 5 years (Peterson 2001).These changes in line profile are well illustrated in Figure 39 from Peterson (2001), and in Fig. 2 the superpositon of two H line profiles from AGN Watch2 data illustrates changes over the course of ∼ 5 years.However, an important point to make is that there is no evidence for significant changes in the shape or profile of the emission lines in response to variations in the luminosity of the accretion disc.This is consistent with results from Peterson et al. (1999), who conclude on the basis of an 8 year survey that broad line profile changes are not reverberation effects, but are due to mass motions within the BLR on the dynamical timescale.
Following on from the success of the International AGN Watch programme (Peterson et al. 2002) which focused on the Seyfert galaxy NGC 5548, attention turned to investigating reverberation in more luminous AGN where the BLR is expected to be larger.This challenge was addressed by Kaspi et al. (2000) with a programme to monitor a well-defined sample of 28 Palomar-Green (PG) nearby AGN (Schmidt & Green 1983) to look for reverberation effects in the broad emission lines.The very large number of spectra which make up this survey are available on line3 .Although it is true, as pointed out by Kaspi et al. (2000), that there were changes in emission line profiles over the course of the 10 years of their survey, examination of the spectra has shown that this only applies to the low luminosity subset of their sample.For quasars with   < −23 and reverberation timescales of around 200 days, the changes in emission line shape are very small compared with those observed in Seyfert galaxies, as shown in Fig. 2.This difference between changes in emission line structure for high and low luminosity AGN is illustrated in Fig. 3, which shows the largest observed variation in emission line shape for quasars over the 10 year monitoring period of PG quasars by Kaspi et al. (2000).It may be seen from Fig. 3 that the only difference between the two profiles is a small enhancement of flux in the blue wing of the H line, very different from the large changes in shape illustrated in Fig. 2.This result is not surprising, as the dynamical timescale for quasars or luminous AGN from Eq. 1 is of the order of 50 years, far longer than the 10 years of the PG spectroscopic monitoring programme (Kaspi et al. 2000).

MICROLENSING IN HIGH REDSHIFT QUASAR SPECTRA
An interesting development in the study of emission line changes in quasar spectra came from a search for evidence for binary supermassive black holes in quasars (Liu et al. 2014).The investigation took the form of a search for quasars with bulk velocity offsets in the broad Balmer lines with respect to the systemic redshift of the host galaxy.This resulted in the compilation of a catalogue of 399 quasars from the Sloan Digital Sky Survey (SDSS) with offset broad H lines, and a mean redshift  = 0.43, around twice the mean redshift of the PG quasars.Second epoch spectra of 50 of the candidates showed that for the most part any changes in the spectra were limited to additional velocity offset, with little change in the structure of the emission line profiles.However, the authors flagged the case of SDSS J0936+5331 (illustrated in Fig. 4) which shows a strong additional red feature which disappeared on a timescale of 10 years.This short timescale suggests that some other process unrelated to the dynamic timescale may be involved, and provides a useful comparison with the PG quasars with a mean redshift  = 0.20.The possibility that the rapid changes in quasar emission line profiles are associated with the presence of a binary supermassive black hole (BBH) is addressed by Liu et al. (2014), who argue that most BBHs will not exhibit double-peaked broad lines due to limitations in parameter space.By implication, this rules out profile changes similar to those illustrated in Fig. 4 for SDSS J0936+5331.On this basis, Liu et al. ( 2014) rejected quasars with double peaked broad Further interest in changes in quasar emission lines was focussed on the emergence or disappearance of broad emission lines in socalled 'changing look' (CLQ) quasars (MacLeod et al. 2016).Subsequent surveys for CLQs were based on finding quasars for which large changes were observed in the broad H line flux (MacLeod et al. 2019;Green et al. 2022).Between these two surveys some 30 new CLQs were discovered satisfying the adopted search criteria, which meant that the recorded changes in the H line were confined to total flux and not variations in line width or profile.The first attempt to find CLQs at high redshift ( > 2) focussed on changes in the C iv 1549 emission line flux (Ross et al. 2020), resulting in the discovery of three quasars with large changes to the C iv flux.Ross et al. (2020) also found the C iv profile to be approximately constant, with the line flux responding to changes in continuum luminosity.
The selection criteria adopted by Ross et al. (2020) for their sample of high redshift CLQs appear to leave open the question of the extent to which changes in emission line profiles occur in the spectra of quasars with  > 2. To address this issue, the SDSS Time Domain Spectroscopic Survey (TDSS)4 provides a very good basis for selecting a sample of quasars for the investigation of changes in emission line profiles.TDSS comprises a number of sub-samples focusing on different categories of variable objects, and of particular relevance for the study of quasar emission line variability is the sample associated with the bitmask TDSS_FES_HYPQSO of variable QSOs chosen for repeated observation.In order to identify profile changes in high redshift quasars, a subsample of the HYPQSO sample was selected with redshift 1.5 <  < 3.0, and the further requirement that there were at least two spectra included in the HYPQSO sample for each quasar.This resulted in a final set of 53 quasars for further study.Spectra of these candidates were then plotted out and examined for obvious changes in emission line structure.The idea was not to compile a complete sample, but to establish whether profile changes do occur in high redshift quasars.As expected, most of the quasar emission lines showed little or no significant changes in emission profile.As a general rule the C iii] line lies in the part of the SDSS spectra with the best signal-to-noise for the redshift range 1.5 <  < 3.0, and Fig. 5 illustrates three typical examples showing little or no change of structure in the C iii] line.However, detailed examination of the spectra and superposition of emission lines from different epochs revealed 9 quasar spectra, illustrated in Fig. 6, with unmistakeable changes in the C iii] emission line profile.These make up around 20% of the sample and show changes on a timescale of 10 years.The changes typically take the form of a broad emission line feature emerging in the blue wing of the C iii] line.Although it is possible that in some cases such changes could be produced by misalignment of the object in the aperture, where several observations are available new line structures typically persist between two epochs separated by a short timescale.
The most natural explanation for the observed changes in line structure is that they are intrinsic to the broad line region, resulting from knots in a turbulent BLR emitting at non-systemic velocities in an analagous way to that observed in Seyfert galaxies (Wanders & Peterson 1996), and illustrated in Fig. 2 above.Rapid changes in broad line shape are only very rarely observed in low redshift quasars (Liu et al. 2014), and the dynamical timescale given in Eq. 1 for luminous quasars would seem to make such changes unlikely.This appears to leave room for a mechanism for variation external to the BLR for high redshift quasars where rapid broad line changes are observed, and on the basis of the discussion in Section 2 above, microlensing of the BLR by a population of stellar mass compact bodies would appear to be a possibility.It is also worth pointing out that the microlensing of other blended lines may be responsible for the change in shape of C iii].In particular, Al iii lies close to the appearance of new features in Fig. 6, and microlensing of this line may well contribute to the overall change in the structure of the C iii] line.This of course does not mean that the broad emission lines are being microlensed, but that if the conclusions of Hawkins (2020aHawkins ( ,b, 2022) ) are correct, then the expected microlensing of quasar broad emission lines is plausibly observed.
Rapid changes in emission line shape due to microlensing are well illustrated by considering the differences in emission line profile due to microlensing in the cluster lens SDSS J1004+4112, illustrated in Fig. 1.The difference in light travel time between the two images is very short, image B leading image A by 41 days (Fohlmeister et al. 2008), which implies that the observed change in emission line structure cannot be intrinsic to the emission line region.After a careful study of various possibilities Richards et al. (2004) conclude that the observed variations in emission line profile must be attributed to microlensing of part of the broad line region of the quasar, resolving structure in the source plane on a scale of ∼ 10 16 cm.
As mentioned above, an important indication that high redshift quasars may be microlensed is the contrast between line profile changes in low and high redshift quasar samples.Although the quasar sample of Liu et al. (2014) provides a useful low redshift sample for comparison, a more direct control sample was obtained from the low redshift HYPQSO quasars.In the parent sample there were 10 members with redshift  < 0.4, corresponding to a probability of microlensingg of around 1%.Here the H line is prominent in the high S/N part of the SDSS spectra, and the 10 quasars in this subsample were examined in a similar way to the high redshift objects, but in this case there was no evidence for significant changes in emission line structure for any of the quasars.This is consistent with the results of Liu et al. (2014) for changes in the H line for a sample of 50 low redshift quasars.
Despite the dynamical timescale given by Peterson (1993) in Eq. 1 of around 50 years for intrinsic changes to broad emission line structure in quasars, there are some caveats with the argument that this completely excludes short term intrinsic variability of emission line structure.For example, a high ionization emission line blended with C iii] can be globally magnified by intrinsic variability, implying that changes in the broad emission lines of quasars can take place over relatively short timescales and induce an apparent change in the shape of the C iii] line (Shen et al. 2023).A related possibility is that regions with different velocities can respond with different time lags to intrinsic variability causing differences in the shape of emission  line profiles observed at different epochs (Grier et al. 2013;De Rosa et al. 2018).Richards et al. (2004) appeared to accept the widely held view at the time that the compact bodies acting as microlenses were stars in the lensing cluster.Given the apparent absence of starlight at such a great distance (60 kpc) from the cluster centre of SDSS J1004+4112, this somewhat implausible hypothesis was examined in detail by Hawkins (2020b).The result of this study, as dicussed above, showed that from measurements of starlight in the vicinity of the quasar images, the optical depth to microlensing from stars was far too small to explain the observed differential changes in brightness of the quasar images.The main conclusion was that to account for the observed microlensing, the lenses must make up at least a large part of the dark matter.This conclusion was reinforced by a more general study of quasar variability on a cosmological scale (Hawkins 2022), where it was concuded that to account for the distribution of quasar lightcurve amplitudes it was necessary to include the microlensing effects of a cosmologically distributed population of stellar mass compact bodies.These results suggest that such a population of compact bodies might also be responsible for microlensing the broad line regions of quasars in the general field, and thus produce the observed rapid changes in broad line profile.

DISCUSSION
The idea behind this paper has been to review evidence that quasar broad emission lines are being microlensed.For this to occur, there must be a substantial optical depth to microlensing  of lenses along the line of sight to the quasar.For quasars in multiply lensed systems there remains the possibility that the population of lenses is associated with the dark matter in the lensing galaxy or cluster halo, but for isolated quasars in the general field where the lenses are assumed to make up the dark matter, the expectation of microlensing events will depend on the redshift of the quasar.
It has been demonstrated above that low luminosity AGN or Seyfert galaxies at low redshift show changes in emission line profile on a timescale of a few years.This is compatible with the dynamical timescale of 3-5 years for such small broad line regions, and there is no need to invoke microlensing which would be very unlikely at such low redshift with a corresponding low value of .On the other hand, the low redshift sample of quasars show no significant changes in emission line profiles.These quasars are certainly too nearby for there to be any significant chance of microlensing, and the dynamical timescale for the large associated BLR means that any intrinsic changes to the emission line profiles would occur on a timescale of ∼ 50 years, far longer than the length of the monitoring programme (Kaspi et al. 2000).There are however some caveats to this broad picture.A high ionization line such as Al iii blended with the C iii] line can reverberate differently to changes in continuum flux, which can occur on short timescales.The resulting changes in the flux ratio from the two lines centered on different wavelenghts can thus plausibly result in significant changes in broad line structure over a relatively short timescale.Another possibility is that regions with different velocities can respond with different time lags to changes in the continuum source.This can then result in apparent changes in the structure of the broad line region when observed at different epochs.
A new approach to changes in quasar broad lines was proposed by Liu et al. (2014), with the focus on measuring bulk offsets in line velocity relative to the systemic redshift of the quasar.The idea was to look for evidence of binary supermassive black holes in quasars, but their data have turned out to be very useful for studying changes in quasar emission line profiles.In the sample of 50 candidates selected by Liu et al. (2014), only the quasar SDSS J0936+5331 showed unmistakeable evidence for changes in the H line profile, with a marginal additional candidate SDSS J1345+1144.Given that the target emission line was H, the typical redshift of the sample members was inevitably small, with a median value  ≈ 0.4 corresponding to an optical depth to microlensing  = 0.01 in a standard ΛCDM Universe (Fukugita et al. 1992).This implies a probability of microlensing for an average sample member of ∼ 1%, which is not inconsistent with the microlensing of just one sample member.However, the important thing to note is that there is no evidence for widespread changes in emission line profile in such a low redshift sample.This is consistent with the long dynamical timescale expected in quasars for changes in the structure of the broad line region, as well as the small probability of microlensing in such a low redshift sample.
The question of emission line profile changes in high redshift ( > 2) quasars was first addressed by Ross et al. (2020).Their candidates were selected from the SDSS archives on the basis of optical variability and the availability of a second spectrum.Subsequent visual inspection revealed 3 quasars showing interesting emission line behaviour, and follow-up spectroscopy confirmed the changing nature of the emission lines, with particular focus on the Ly , C iv and Mg ii lines.The main changes observed in the emission lines were line emergence and collapse, but no evidence was reported for significant changes in the shapes of the line profiles.The authors further conclude that the main driver for emission line variability is the broad-band continuum itself, but declare themselves 'agnostic' as to the underlying physical processes.
An outstanding question remaining after the investigation by Ross et al. ( 2020) is whether structural changes in broad emission line profiles can occur in high redshift quasar spectra on timescales of a few years, as occasionally observed in low redshift quasar samples (Liu et al. 2014).To answer this question, the HYPQSO sample of quasars earmarked for repeated observations has provided a useful start.Given the rarity of changes in emission line profiles for low redshift quasars (Liu et al. 2014), where emission line changes are largely confined to increase or decrease in line flux with no associated change in the shape of the line profile, it was surprising to find that profile changes in the HYPQSO sample with 1.5 <  < 3.0 were readily identified, as illustrated in Fig. 6.The implied dependence of changes in line profile on redshift rather than BLR size suggests that some external mechanism may be involved.Given that compact bodies are well known to microlens broad emission lines in multiply imaged quasar systems (Richards et al. 2004;Sluse et al. 2012;Fian et al. 2021), microlensing of the broad emission lines may provide a solution to the observed changes in emission line structure in luminous quasars.
The idea that quasar emission lines are being microlensed raises some immediate questions.Firstly, what is the origin of the lenses?This can be answered on the basis of the original motivation for this paper, which arose from evidence (Hawkins 2020a(Hawkins ,b, 2022) that microlensing observations implied at least a major component of dark matter to be in the form of stellar mass compact bodies, most plausibly primordial black holes.One of the consequences of this would be the microlensing of the BLR in quasars of sufficiently high redshift, thus providing an explanation for the observed changes in emission line profiles.To put this on a firmer statistical basis the probability of microlensing can be estimated from the optical depth to microlensing .Using the equations of Fukugita et al. (1992), the value of  for a quasar with  = 2 in a ΛCDM Universe is  ≈ 0.2, implying a probability of around 20% that the quasar will be significantly microlensed.This is interestingly close to, and certainly consistent with, the proportion of quasars found to show changes in emission line structure for the TDSS subsample described in Section 4. In fact the probability is somewhat higher than this, as for values of ≳ 0.1, the amplification patterns of the individual lenses will start to combine in a non-linear way to form a pattern of high-amplification caustics (Kofman et al. 1997).However, at a redshift  ∼ 2 this effect will be small.By a fortunate circumstance, one of the high redshift quasars described in Section 4 and illustrated in Fig. 6, where changes in emission line profile were detected, was also included in the SDSS Legacy photometric monitoring programme in Stripe 825 , providing the opportunity to look for any unusual features in the light curve which might be connected with changes in emission line profile.The SDSS lightcurve for SDSS J0151+0100 is plotted in Fig. 7, together with additional measures from the Pan-STARRS1 data archive6 (Flewelling et al. 2020).The light curve shows achromatic variation with amplitude of a magnitude over a timescale of around 10 years.There is no indication of short term fluctuations in brightness which might be associated with changes in emission line strength.

CONCLUSIONS
This paper has set out to investigate whether changes in the structure of quasar broad emission lines are consistent with the expected microlensing by a population of stellar mass primordial black holes making up a large fraction of dark matter.Evidence for such a population has been published recently (Hawkins 2020a(Hawkins ,b, 2022)), which raises the possibility that these compact bodies may also be detected by microlensing the broad emission line clouds associated with quasars.Microlensing effects are expected to change the shape of the broad line profile, typically resulting in the appearance of new emission features offset from the systemic velocity of the quasar.Although such changes can happen as a result of random motions of the clouds in the broad emission line region, they would be expected to occur over very long dynamical timescales of the order of 50 years or more, as opposed to microlensing timescales of around 5 years.A further discriminant comes from the redshift of the quasar.Low redshift quasars are very unlikely to be microlensed due to the predicted small optical depth to microlensing of the lenses, but for redshift  > 2 the probability of microlensing rises to around 20%.
The main results of the paper are as follows: (i) The paper starts by characterising the changes in quasar emission line structure due to microlensing, based on spectroscopy from the literature of quasar spectra from gravitational lens systems where microlensing is known to occur.
(ii) Changes in emission line structure in low redshift Seyfert galaxies are illustrated, and attributed to cloud motions in the BLR.These changes occur on short timescales of the order of 5 years, corresponding to the dynamical timescale of the BLR.
(iii) Luminous low redshift quasars show no changes in emission line structure.This is attributed to the much larger size of the BLR, with a dynamical timescale of the order of 50 years.
(iv) Luminous quasars at high redshift ( > 2) exhibit different behaviour, with around 20% showing strong changes in emission line structure on a timescale of 5 years.This is consistent with the expected microlensing from a population of solar mass compact objects comprising at least a large fraction of the dark matter, although intrinsic variability is an alternative explanation.The most plausible candidates for the compact bodies are primordial black holes.

Figure 1 .
Figure 1.Emission lines from Keck spectra of the wide separation gravitational lense system SDSS J1004+4112.The left hand panel shows the Si iv 1400 line for the A image (blue), and the B image (red).The dashed vertical line shows the systemic wavelength.The right hand panel shows similar data for the C iv 1549 line.

Figure 2 .
Figure 2. Emission line profiles for the H line in the Seyfert galaxy NGC 5548 from International AGN Watch data showing variations on a timescale of ∼ 5 years.

Figure 3 .
Figure 3. Emission line profiles for the H line of the low redshift quasar PG1351+640 from the spectroscopic monitoring programme undertaken by Kaspi et al. (2000).

Figure 4 .
Figure 4. Emission line profiles for the H line from SDSS archive spectra of the the quasar SDSS J0396+5331.

Figure 5 .
Figure 5. Emission line profiles for the C iii] line from SDSS archive spectra of quasars in the redshift range 1.5 <  < 3.0, chosen for showing small changes in structure on a timescale of 10 years in the quasar rest frame.

Figure 6 .
Figure 6.Emission line profiles for the C iii] line from SDSS archive spectra of quasars in the redshift range 1.5 <  < 3.0, showing changes in structure on a timescale of 10 years in the quasar rest frame.

Figure 7 .
Figure 7.Light curve for the quasar SDSS J0151+0100 in the g-band (blue filled circles) and the r-band (red open circles).Data for the years 1998 to 2007 are from the SDSS stripe 82 archive, and for 2010 to 2013 from the Pan-STARRS1 data archive.The green arrows mark the epochs of the two spectra shown in Fig. 7