The Frontier of Cosmic Cataclysms

Ryan Ridden-Harper

A thesis submitted for the degree of Doctor of Philosophy of The Australian National University Research School of Astronomy and Astrophysics May 2020

c Ryan Ridden-Harper, 2020. ⃝

Iherebydeclarethattheworkinthisthesisisthatofthecandidatealone,except where indicated below or in the text of the thesis. The work was undertaken between the 20th of May, 2016 and the 15th of October, 2019 at the Australian National University (ANU), Canberra. It has not been submitted in whole or in part for any other degree at this or any other university.

Statement of Contribution

This thesis is submitted as a Thesis by Compilation in accordance with https: //policies.anu.edu.au/ppl/document/ANUP_003405. Ideclarethattheresearchpresentedinthisthesisrepresentsoriginalworkthat I carried out during my candidature at the Australian National University, except for contributions to multi-author papers incorporated in the thesis where my contributions are speciﬁed in this Statement of Contribution.

Title and authors: Capability of detecting ultraviolet counterparts of gravitational waves with GLUV. Ridden-Harper, R.,Tucker,B.,Sharp,R.,Gilbert, J., Petkovic, M. Current status of paper:Published. Contribution to paper: I was the primary contributor to this paper. As such, I was responsible for the content, while coauthors provided feedback.

Title and authors: Kepler/K2: Background Survey. Ridden-Harper, R.,Tucker, B. E., Gully-Santiago, M., Barensten, G., Rest, A., Garnavich, P., Shaya, E. J. Current status of paper: Accepted with revisions. Contribution to paper: I was the primary contributor to this paper. As such, I was responsible for the content, while coauthors provided feedback. iii

Title and authors: Discovery of a New WZ Sagittae Type Cataclysmic Variable in the K2/Kepler Data. Ridden-Harper, R., Tucker, B. E., Garnavich, P., Rest, A., Margheim, S., Shaya, E. J., Littleﬁeld, C., E., Barensten, G., Hedges, C., Gully- Santiago, M. Current status of paper: Accepted. Contribution to paper: I was the primary contributor to this paper. As such, I was responsible for the content, and text. Garnavich, P., provided analysis of the K2/Kepler light curve; Margheim, S., amd Rest, A., provided follow-up observations. Other coauthors provided feedback.

Ryan Ridden-Harper 14/05/2020 Candidate Signature Date

Endorsed

Brad Tucker 14/05/2020 Chair of Supervisory Panel Signature Date

Delegated Authority Signature Date The surface of the Earth is the shore of the cosmic ocean. On this shore, we’ve learned most of what we know. Recently, we’ve waded a little way out, maybe ankle-deep, and the water seems inviting.

– CARL SAGAN Acknowledgements

When I started my Ph.D. 3 years seemed like an eternity away, but it quickly passed by. Over these years challenges arrived and were surmounted in large part due to those I will mention here.

First, I would like to thank my primary supervisors: Brad Tucker, for teaching me how to be an independent researcher and the amazing opportunities he has provided; and Rob Sharp, for his patience in teaching me telescope design and his ever entertaining emails.

Next, I would like to thank my family for their support. They have supported me from the very beginning, nurturing my interest in science and helping me through challenges. This entire endeavour began with my Mum helping me write an introduction email and my Ph.D. application; without her support I would never have made it this far. My Brother, Dr. Andrew, was the first to fall in love with the Universe, but I soon followed after he dragged me out to see the night sky. So it’s only fitting that I finish my Ph.D. in astronomy shortly after he did, albeit on the other side of the world. He’s been my endless source of encouragement in everything I’ve done, and will continue to inspire me.

Often it’s teachers who make or break a students interest in a subject. Thank- fully, I was fortunate enough to have Ian Chinnery as my physics teacher. His passion for physics and teaching was remarkable, conversations I had with him about physics and the Universe would always leave me inspired and motivated.

I’ve been lucky to work with incredible collaborators. My visits to the Kepler Guest Observer Oﬃce were always a pleasure, their expertise and enthusiasm pushed me forward in the analysis of the often-times messy Kepler/K2 data. Likewise, the

v vi Acknowledgements

KEGS group welcomed and helped me a great deal. In particular I would like to thank Peter Garnavich for his incredible knowledge of transients, and Armin Rest for his support and giving me something to do after this Ph.D.! IwouldalsoliketothankthosewhokeptMt.StromloObservatorymoving forward. In particular Michelle Cicolini, for knowing the answer to all questions I had; Howard Cole, for maintaining the Observatory with his lovely dogs; and the RSAA IT group, for holding the servers together. Over the past 3 years I’ve had the pleasure of teaching many fantastic future scientists. I’m forever astounded with what they know, and how fast they learn. I’m grateful to have had the opportunity to help them reach their bright futures. Finally, I would like to thank my dear colleagues and friends from the past 3 years. No Ph.D. is easy, but their support and friendship made it all the easier. For those yet to ﬁnish their Ph.D., I wish you luck! My oﬃce mate, Alec Thomson, was the best I could have asked for, always happy to help me with my dumb code mistakes, and restyling my life. I’m grateful for his friendship. At the end of these seemingly short 3 years, I’m sorry to be leaving my new family, but I look forward to what awaits! Abstract

The short time domain ( 1day)isthefrontieroftransientastronomy.Inthis ≤ frontier, new phenomena are waiting to be discovered that may hold the answers to crucial questions, particularly for progenitors of extreme events e.g., supernovae, gravitational waves/kilonovae, and gamma-ray bursts. Telescope systems across the world, such as Pan-STARRS, ASAS-SN, ATLAS, and ZTF are pushing towards shorter cadences, however, they are still limited by the diurnal cycle and weather. Although these telescope systems are successful at discovering transients that evolve over many days, such as supernovae, they are currently unable to systematically explore the very rapid time domain of transients and features that evolve on time scales 1day. ≤ In this thesis we seek to explore the short time domain to discover and understand short phenomena. To begin this exploration, we consider two pathways into the frontier of rapid transients: a high altitude balloon borne ultra-violet survey telescope, known as GLUV,optimisedtodetecttypeIasupernovashockinteractions and core collapse supernovae shock breakouts; and we developed the Background Survey, to conduct a systematic a transient survey of high cadence data obtained by the Kepler Space Telescope (Kepler). With these two pathways, we were able to actively explore the short time domain, while developing a purpose built instrument.

We ﬁnd that for GLUV to routinely detect SN Ia shock interactions, it must have a limiting magnitude of 21. GLUV could meet this sensitivity if it features ∼ a30cmprimarymirror,allowingittodeterminetheprogenitorsystemsresponsi- ble for cosmological SN Ia. The same telescope design would also provide valuable

vii viii Abstract diagnostic information on counterparts of gravitational wave events, such as kilonovae and possible emission from binary black hole mergers. Through the Kepler/K2 Background Survey,wefindthattherearestilltransientstobediscoveredinthe Kepler/K2 data. In initial tests, we discover a new WZ Sagittae type dwarf nova, named KSN:BS-C11a. Through the superb high cadence photometry afforded by Kepler,weareabletoconstrainanumberofphysicalpropertiesofthesystem,and discover that the rise of dwarf novae superoutbursts can be characterised by a broken power law. The presence of the broken rise suggests new physics in how the superoutbursts begins. We also observe the broken rise phenomena in another dwarf nova discovered and observed by Kepler and Z Chamaeleontis, which was observed by TESS. The discovery of KSN:BS-C11a and its subsequent properties was only possible through high-cadence observations, and is indicative of future discoveries that can be made in the short time domain frontier. The short time domain of transient astronomy offers the potential for many new and novel discoveries. Many of these discoveries will come from the unique high cadence data obtained by Kepler and by TESS,ofwhichtheBackground Survey is the best tool to make these discoveries. As we learn more about the short time domain, dedicated transient telescope systems such as GLUV will be poised to make full use of the advancing knowledge, as we explore the new frontier of astronomy. Contents

Acknowledgements v

Abstract vii

1 Introduction 1 1.1 Time domain astronomy ...... 1 1.1.1 Transients ...... 3 1.1.2 Exotic Events ...... 11 1.1.3 Exoplanets ...... 14 1.2 Ultra-violet astronomy ...... 17 1.3 Aim of this thesis ...... 18 1.3.1 Overview of chapters ...... 19

2 Google Loon Ultra-Violet Telescope 21 2.1 Overview ...... 21 2.2 Loon platform ...... 23 2.3 Ozone layer ...... 24 2.3.1 Ozone dynamics ...... 26 2.4 Atmospheric transmission ...... 26 2.5 GLUV –Pathﬁnder Spectrograph ...... 30 2.5.1 GLUV –PS ﬂight ...... 31 2.6 Sky brightness ...... 36 2.7 GLUV signal-to-noise calculator ...... 37

ix x Contents

3 GLUV Science Cases 41 3.1 Supernovae shocks ...... 41 3.1.1 SN Ia shock interaction ...... 42 3.1.2 CC SN shocks ...... 43 3.2 Survey strategy & rates ...... 45 3.2.1 Preliminary Survey Conﬁguration ...... 45 3.2.2 SN Ia shock interaction detection rates ...... 46 3.2.3 CC SN shock breakout detection rates ...... 47 3.3 Gravitational wave counterparts ...... 47 3.4 Exoplanet atmospheres ...... 48 3.4.1 Hot-Jupiters ...... 48 3.4.2 M-star planetary systems ...... 50 3.4.3 Transit signal-to-noise ...... 52 3.5 Conclusion ...... 54

4 Detecting Gravitational Wave UV Counterparts with GLUV 55 4.1 Abstract ...... 55 4.2 Introduction ...... 56 4.3 UV signatures from mergers ...... 58 4.3.1 Binary black hole merger ...... 58 4.3.2 Binary neutron star merger ...... 61 4.3.3 Black hole–neutron star merger ...... 66 4.3.4 Gamma-ray bursts ...... 68 4.4 GLUV: a UV Survey Telescope ...... 70 4.4.1 Sky background ...... 70 4.4.2 Instrument throughput ...... 71 4.4.3 Fiducial Survey Strategies & Detection Rates ...... 71 4.4.4 Survey Strategies ...... 73 4.5 Conclusion ...... 76 4.6 Appendix: Detection rates ...... 78

5 Kepler/K2: Background Survey 83 5.1 Abstract ...... 83 Contents xi

5.2 Introduction ...... 84 5.3 Kepler/K2 data ...... 86 5.4 Methods ...... 88 5.4.1 Science target mask ...... 88 5.4.2 Telescope drift correction ...... 89 5.4.3 Short event identiﬁcation (< 10 days) ...... 92 5.4.4 Long event identiﬁcation (> 10 days) ...... 96 5.4.5 Variable stars ...... 97 5.4.6 Asteroids ...... 97 5.4.7 Event sorting ...... 98 5.4.8 Event ranking ...... 99 5.4.9 Visual inspection ...... 99 5.5 Survey Characteristics ...... 101 5.5.1 Magnitude limits ...... 102 5.6 Detected events ...... 104 5.7 Expected rates ...... 106 5.7.1 Volumetric rates ...... 107 5.7.2 Expected detection rates ...... 108 5.8 Conclusions ...... 111

6 Discovery of a new WZ Sagittae type cataclysmic variable in Ke- pler/K2 data 113 6.1 Abstract ...... 113 6.2 Introduction ...... 114 6.3 Data ...... 116 6.3.1 Search for Transients in K2/Kepler ...... 116 6.3.2 Discovery of KSN:BS-C11a ...... 117 6.3.3 Full Frame Images ...... 118 6.3.4 K2/Kepler Light Curve ...... 120 6.3.5 Gemini Spectra ...... 121 6.3.6 DECam images ...... 122 6.4 Analysis ...... 123 6.4.1 Power Spectrum ...... 123 xii Contents

6.4.2 Early Rise ...... 127 6.4.3 Optical Spectrum ...... 128 6.5 Discussion ...... 131 6.5.1 Orbital Period and Mass Ratio ...... 131 6.5.2 Rebrightening Phase ...... 133 6.5.3 Early Rise ...... 135 6.5.4 Distance ...... 136 6.6 Conclusion ...... 137 6.7 Appendix ...... 138 6.7.1 GMOS-N spectrum ...... 138

7 Conclusion 141 7.1 Summary of GLUV ...... 141 7.2 Summary of GLUV science cases ...... 143 7.3 Summary of the Background Survey ...... 145 7.4 Summary of this thesis ...... 146 7.5 Future work ...... 147 7.5.1 The future of GLUV ...... 147 7.5.2 The future of the Background Survey ...... 147 7.6 Final remarks ...... 149

Bibliography 151 1 Introduction

1.1 Time domain astronomy

Astronomical phenomena can occur over the course of billions of years, to just a fraction of a second. The enormous time scale range can be split into two categories; the static Universe, for phenomena that evolve over millions to billions of years; and the dynamic Universe, for phenomena that change rapidly on time scales less than ayear.Somerapidphenomena,likeexoplanettransits,canrepeat,butothers,such as supernovae, are transient, appearing only once, before disappearing again forever. In this thesis we push the boundaries of the time domain to search for phenomena that evolve over time scales of a day to minutes.

Transients, such as supernovae, form a key aspect of time domain astronomy. To detect any transient, the frequency, or cadence, of observations must be high enough to catch the transient during its lifetime. This often requires dedicated surveys with purpose built instruments. Through modern robotic observatories many

1 2 Introduction instruments are now exploring the time domain, with high-cadence observations, in a range of wavelengths. Key modern examples of such dedicated systems are SkyMap- per (Scalzo et al. 2017), Pan-STARRS (Chambers et al. 2016), ATLAS (Tonry et al. 2018), ASAS-SN (Kochanek et al. 2017), PTF, ZFT (Bellm 2014). Collectively, these systems detect thousands of transients a year, but are all restricted by daily observation gaps. To search for very rapid transients, The Deeper Wider Faster (DWF) program networks more than 30 telescopes around the world for intensive observation campaigns that run for 4–6 nights per semester (Andreoni & Cooke ∼ 2019). Although DWF achieves high-cadence observations, it is limited to the campaign window, and is also interrupted by the diurnal cycle. Exoplanet detection also depends greatly on the time domain, requiring rapid and frequent observations to detect exoplanets. The Kepler Space Telescope (Kepler, Basri et al. 2005b; Howell et al. 2014)andtheTransitingExoplanetSurveySatellite (TESS, Ricker et al. 2015), utilise high-cadence observations to detect exoplanets via the transit method. Although exoplanet transits occur with a regular period, high- cadence observations are required to observe the full transit and identify features, such as the ingress and egress, that can last only minutes. Kepler is capable of 1minuteand30minutecadenceobservations,whileTESS is capable of 2 minute and 30 minute cadences1.Therapidcadenceofthesesystemsmakesitpossibleto not only detect faint Earth sized planets around distant stars, but also explore short scale astrophysical phenomena. As many fascinating phenomena lie within the short time domain, a global push is under way to explore this frontier. Within this frontier there may lie answers to fundamental questions in astronomy, for example;“what are the progenitor systems behind supernovae, and can we really treat type Ia supernovae as standardizable candles?”; “are there other Earth-like planets that may contain life?”; and simply “what is yet to be discovered?”. The following sections will explore the foundations of these questions, and outline what we expect to gain from exploring the time domain’s frontier.

1As the work in this thesis was primarily conducted before the TESS data releases, we primarily focus on Kepler/K2 data. 1.1 Time domain astronomy 3

1.1.1 Transients

Transients form a key aspect of time domain astronomy, driving advances in new telescope systems to discover short lived phenomena. Over the past decades, these transients have become crucial in understanding nucleosynthesis (e.g., Hoyle & Fowler 1960) and the evolution of the Universe (e.g., Riess et al. 1998; Perlmutter et al. 1999). All transients share two main characteristics: peak brightness and lifetime, as shown in Fig. 1.1.Theparameterspaceoftransientswithlifetimesthatlastfor days has been well explored, and contains many noteworthy phenomena, such as cataclysmic variables, and supernovae. Although the long lifetime transients have been well explored, little is known about transients that evolve on time scales of a day or less. Many transients that exist in this time domain will exceed the maximum brightness possible from synthe- sised 54Ni, as described by Arnett’s rule, so are expected to be powered by relativistic sources and shocks (Arnett 1979, 1982). Such phenomena provide unique insight into relativistic processes, such as gamma-ray bursts, and provide crucial information on the progenitor systems that lead to supernovae. It is also possible that there are time scales for which optical transients do not exist, however, we must search to know if extremely short duration do or do not exist. 4 Introduction

Figure 1.1: The current state of the transient time domain in R-band. The parameter space of events with lifetimes (decay time) greater than a day is well explored and constrained, events with lifetimes of a day and shorter are unexplored. Figure modiﬁed from Copperwheat et al. (2015).

Type Ia supernovae

Type Ia supernovae (SN Ia) are a vital tool in cosmology, due to their ability to be standardised. The standardisation of SN Ia is possible due to the strong theoretical and observational constraints that show SN Ia are the product of C/O white dwarfs (WDs) undergoing thermonuclear explosion in a binary system (e.g., Hoyle & Fowler 1960; Colgate & McKee 1969; Woosley et al. 1986; Bloom et al. 2012). The explosion is triggered when the white dwarf accretes enough mass from the binary companion to reach the Chandrasekhar mass of 1.4M (Chandrasekhar 1931). Although the ⊙ physical mechanism is understood, SN Ia vary in intrinsic brightness, and so must be standardised. The Philips relationship standardises SN Ia by determining the absolute magnitude of a SN Ia, from the peak brightness and the initial rate of decline in the light curve (Phillips 1993).

As SN Ia are bright and standardizable, they have been instrumental for measur- ing distances in the Universe. This culminated with the discovery of the accelerated expansion of the Universe through a property known as Dark Energy (Riess et al. 1998; Perlmutter et al. 1999). Following the discovery of Dark Energy, numerous 1.1 Time domain astronomy 5 supernova projects have sought to constrain its equation of state, such as the Pan- theon Sample (Scolnic et al. 2018b), the SDSS-II Supernova Survey (Sako et al. 2018), and the Dark Energy Survey (Abbott et al. 2019).

Each cosmological survey has produced stronger constraints on the nature of

Dark Energy, and the value of the Hubble parameter, H0.Theseconstraintshowever, have lead to a signiﬁcant tension in H0 between early and late Universe probes. Two key examples of this are the Riess et al. (2019)SNmeasurement,whereH =74.03 0 ± 1 1 1.42 km s− Mpc− ,andthePlanck Collaboration et al. (2018)CosmicMicrowave

1 1 Background measurements, where H =67.4 0.5kms− Mpc− .The> 3σ tension 0 ± between H0 measurements of the early and late Universe has lead to a number of ideas such as interacting dark energy (Di Valentino et al. 2019), decaying dark matter (Pandey et al. 2019; Vattis et al. 2019), new neutrino physics (Kreisch et al. 2019; Barenboim et al. 2019), new particles (D’Eramo et al. 2018), and new physics (e.g., M¨ortsell & Dhawan 2018). The tensions has also raised fundamental questions about how cosmological parameters are measured with SN Ia and their standardisation.

The origin of systematics in the SN Ia H0 determination can arise from instrumental differences, and a lack of understanding in the physics behind SN Ia mechanisms. Instrumental systematics are being refined through single instrument surveys, such as the Dark Energy Survey (Abbott et al. 2019). The reduction in instrumental systematics has brought attention to the systematics that may be present in the SN Ia explosion mechanism and their standardisation. Childress et al. (2013) identified a number of correlations between SN Ia Hubble residuals and properties of the host galaxy, which suggests the standardisation of SN Ia is incomplete. Further- more, Rigault et al. (2018)identifiedastrongdependenceofSNIastandardisation and the local specific star formation rate. The incomplete understanding of progenitors to “normal” SN Ia may play a significant role in these correlations and dependencies.

SN Ia pregenitor There are two possible progenitor scenarios to produce a SN Ia, being the single- degenerate (SD) scenario and double-degenerate (DD) scenario. In the SD scenario, as seen in Fig. 1.2 (left), the WD accretes material from a companion star until it reaches the Chandrasekhar mass and detonates (e.g., Whelan & Iben 1973). In the 6 Introduction

DD scenario, as seen in Fig. 1.2 (right), two WDs merge, triggering the explosion (e.g.; Iben & Tutukov 1984). Although both progenitor systems are consistent with the SN Ia population, it is unclear if both contribute to the SN Ia population, and if so, in what proportions. So far no progenitor system has been dete cted (e.g., Goobar et al. 2014; Kelly et al. 2014), likewise searches for surviving companion stars in Galactic supernova remnants were unsuccessful (e.g. Kerzendorf et al. 2012; Schaefer & Pagnotta 2012).

Figure 1.2: Artist impressions of the two SN Ia progenitor systems. Left: the single-degenerate (SD) scenario, where a companion star feeds material onto a white dwarf until it reaches the Chandrasekhar mass and produces a SN Ia. Right: the double-degenerate (DD) scenario, where two white dwarfs collide, reaching the Chandrasekhar mass, producing a SN Ia. Credit: SD scenario NASA/JPL-Caltech, DD scenario NASA and Sky Works Digital.

Kasen (2010)suggestedawaytoidentifySDprogenitorsystemsintheearlyrise of the SN Ia light curves. As ejecta from the SN Ia expands, it will collide with the companion, or donor star. This collision results in a shock interaction producing excess emission, changing the rise of a SN Ia light curve so that it no longer follows the expanding fireball model of L t2,whereL is the luminosity, and t is the ∝ time since explosion (Riess et al. 1999). The emission from the shock interaction model is brighter for shorter wavelengths and larger companion stars, while being heavily dependent on the viewing angle, as seen in Fig. 1.3.Fromthesemodels,itis clear that if SD SN Ia are observed within the first day of explosion, with sufficient photometric precision, the progenitor system could be inferred. Complications arise in searching for early flux from SN Ia to identify the progenitor, as the shock interaction is not the only physical mechanism that can cause such an early excess flux. Piro & Nakar (2014)suggestedthatanoverdensityof56Ni 1.1 Time domain astronomy 7

Figure 1.3: Shock models for ejecta from a SN Ia interacting with companion stars of different masses and evolutionary stages. For these models the viewing angle is fixed to θ =0o,suchthatthecompanionliesbetweenthewhitedwarfandEarth, giving the maximum model flux. Figure taken from Kasen (2010). near the ejecta surface could also produce excess early flux, potentially reconciling such observations with the expanding fireball model. As a result, both the SD and DD could produce signatures that resemble the shock interaction signature. Fur- thermore, Piro & Morozova (2016)showedthattheinteractionbetweenejectafrom aDDSNIaandthecircumstellarmaterialcanproducesignaturesintheearlylight curve shape.

All models make it clear that early observations of SN Ia are critical to understanding the progenitor systems. This requires a push into the sub-day to hour time domain with precision photometry to detect these signatures. Precise high-cadence surveys from the ground are diﬃcult to achieve, so only two normal SN Ia, being SN 2012cg and SN 2017cbv, were observed to have excess ﬂux early in the light curve (Marion et al. 2016; Hosseinzadeh et al. 2017). Analysis of these early light 8 Introduction curves favour the SD shock scenario, however, this is challenged by a lack of swept up Hα from the companion star, found in the nebula phase spectra (Sand et al. 2018; Shappee et al. 2019).

Kepler SN Ia The need for precise high-cadence observations of SN Ia led to the Kepler Extra- Galactic Survey (KEGS). Using the Kepler prime mission high-cadence data, of observations every 30 minutes, for tens of thousands of galaxies, the rise of 12 ∼ SN Ia were observed (Villar et al. prep). Olling et al. (2015)analysedthelight curves of 3 “normal” SN Ia, KSN 2011b, KSN 2011c, and KSN 2012a, observed with Kepler,andfoundnoearlyexcess.Theresultsuggeststhatthese3SNIa are products of DD systems, however, it does not rule out SD systems from also producing normal SN Ia.

Kepler did detected an early excess ﬂux in SN 2018oh in Kepler/K2 Campaign 16 (Dimitriadis et al. 2019a; Shappee et al. 2019). As seen in Fig. 1.4,thepreci- sion high-cadence photometry of Kepler/K2 clearly shows excess ﬂux early in the light curve. Furthermore, ground based observations show the colour evolution from blue to red (Dimitriadis et al. 2019a). These observations favour a SD progenitor system with a companion separation of 2 1012 cm. However, as with observa- ∼ × tions of SN 2012cg and SN 2017cbv, SN 2018oh also lacks hydrogen in the nebula phase spectra, suggesting it is more consistent with an overdensity of 56Ni on the surface (Dimitriadis et al. 2019b; Tucker et al. 2019). An analysis is in progress of the remaining SN Ia observed by Kepler to identify the likely progenitors for those systems (Villar et al. prep).

TESS SN Ia Although the Kepler mission has ended, the search for high-cadence SN Ia light curves is still possible through the Transiting Exoplanet Survey Satellite (TESS). TESS will tile most of the sky at 30 minute cadence, covering sectors for 27 days. ∼ Although TESS is not as sensitive as Kepler, it has already observed many SN Ia. Fausnaugh et al. (2019)analyses18SNIaobservedbyTESS and ﬁnds no conclusive evidence for SD shock interactions, placing upper limits on the companion sizes to be < 25 R for 6 SN Ia and < 4R for 4 SN Ia. ∼ ⊙ ∼ ⊙ 1.1 Time domain astronomy 9

Figure 1.4: Kepler/K2 light curve of SN 2018oh with the excess early flux shown in the image cutout. As seen in the cutout, there is a significant residual shortly after explosion from the expanding fireball model fit. Figure taken from Dimitriadis et al. (2019a).

Although nearby SN can now be observed in high-cadence with TESS,theshock phenomena are most prominent at short wavelengths. To conduct a comprehensive survey for SN Ia shocks and identify progenitor systems, a high-cadence ultra violet survey is required. Understanding the progenitors of “normal” SN Ia is critical to understanding the tension in H0 and the nature of dark energy.

Core collapse supernovae

Core collapse supernovae (CC SN), or type II/Ibc SN, mark the end of star with masses 8M (for review, see Heger et al. 2003). The collapse begins shortly ≥ ⊙ after iron is produced at the stars core. Since energy is not generated by fusing iron nuclei, the hydrostatic equilibrium of the star is disrupted as the gravitational force is no longer balanced by radiative pressure from the centre. How energy from gravitational collapse is converted to explosive energy is not fully understood, and may require core accretion instabilities (Blondin et al. 2003)orenergydeposition from neutrinos (Bethe & Wilson 1985). Due to the origin of CC SN the light curves of sub-types exhibit great variation (Pritchard et al. 2014). Despite this variation all CC SN are expected to generate bright shocks in hard X-ray, and UV radiation as the shock from the core-collapse 10 Introduction reaches the stellar surface (Falk 1978a; Klein & Chevalier 1978). Observing these shocks provides an indication of the progenitor system as the timescale of the shock is approximately the light travel time across the stellar radius (Nakar & Sari 2010). AstandardisationmethodhasbeenproposedforSNII-Pbymeasuringtheshocks, and from that properties of the system, however, this technique is more involved than that of SN Ia standardisation (Eastman et al. 1996; Poznanski et al. 2009). Typical shocks from CC SN are expected to last for only < 1hour,sohigh- ∼ cadence observations are critical (Garnavich et al. 2016). To date several shocks have been observed from the CC SN subclass SN II-P by the GALEX satellite (Schawinski et al. 2008; Gezari et al. 2015)andKepler (Garnavich et al. 2016). The two shocks observed by GALEX lasted signiﬁcantly longer than an hour, suggesting that supergiants have extremely large radii or there is circumstellar material that prolonged UV emission (Ofek et al. 2010; Chevalier & Irwin 2011). Modelling the shocks observed by Kepler, as seen in Fig. 1.5,foundthattheprogenitorradiiwere 280 20 R and 490 20 R .TheseradiiobservedbyKepler support the idea of ± ⊙ ± ⊙ circumstellar material around the SN II-P observed with GALEX. Observing more CC SN at high-cadence is crucial to understanding their progenitor systems. The cadence required to observe these phenomena requires a push into the sub-hour; a time domain that has not yet been systematically explored. As with SN Ia, understanding the progenitor systems behind these explosions requires high-cadence observations. 1.1 Time domain astronomy 11

Figure 1.5: The rising LCs of type II-P SNe KSN 2011a and KSN 2011d observed by Kepler (Garnavich et al. 2016). The blue points are the Kepler 30 minute observations, with the red being the 6 hour median. The thin line is the predicted rise from an analytical model. The residual at the onset of KSN 2011d is the shock breakout.

1.1.2 Exotic Events

In newly explored time domains, new and unique phenomena have been discovered. This has been true for modern surveys, that are now capable of exploring the time domain of 1day. Drout et al. (2014)presentsonesuchexample,where10Fast ∼ Evolving Luminous Transients (FELTs) were detected in PS1, that evolved over 12 days. These events were of unknown origin and shared similar features, such as strong blue continuum consistent with hot, optically thick ejecta. Furthermore, these events showed no signs of being powered by radioactive decay of 56Ni, rather by envelope emission from stellar explosions, or shock breakouts from stars encased in an optically thick wind. The mechanism behind these mysterious events remain unknown. Similarly, a rapidly evolving transient was discovered with Kepler/K2,known as KSN 2015K. As seen in Fig. 1.6,KSN2015Krisesinjust2daysmakingitthe fastest rising SN ever seen (Rest et al. 2018). As with the objects discovered in PS1, 12 Introduction

Figure 1.6: Left: The K2 light curve of KSN 2015K, blue dots are individual 30 min cadence observations while the red points are 3 hr binned data. Right: The rise time of KSN 2015K compared to other transients, it features the fastest rise of any discovered SN. Figures taken from Rest et al. (2018).

KSN 2015K does not ﬁt SN models, and is best described by either ejected stellar material interacting with a dense circumstellar medium or a relativistic event. To date no event similar to KSN 2015K has been observed.

Kilonova

Another newly discovered event that falls into the day time domain are kilonova. These events are produced through the merger of a binary neutron star system (BNS) or a neutron star and black hole system (BHNS). Kilonova have become a focal-point of multi-messenger astronomy, as the merger event produces gravitational waves (GW) as was detected with GW170817 (Abbott et al. 2017d). An example observation of GW170817 is shown in Fig. 1.7. In optical wavelengths kilonovae are short lived, lasting only < 5days(Drout ∼ et al. 2017), as a result only a few potential kilonovae have been observed (e.g. Yang et al. 2015; Evans et al. 2016). Observing more of these events is critical as they are thought to be the primary source of many heavy elements (e.g., Kasen et al. 2017), and may provide an independent cosmological distance measure (Abbott et al. 2017c). 1.1 Time domain astronomy 13

Figure 1.7: Collapsed data cube of GW170817 observed with the ANU 2.3 m WiFeS integral ﬁeld unit spectrograph at +0.93 days from discovery. Figure taken form Andreoni et al. (2017).

Stellar ﬂares

For the first time it is possible to systematically explore the time domain from days to minutes, with Kepler and TESS.Oneknowntransientthatisubiquitousatminute to hour time scales is M-dwarf flares, which can take on a plethora of outburst shapes. The high-cadence observations of Kepler and TESS has enabled detailed study of these phenomena, which strongly impact the habitability of exoplanets around these stars (e.g., Davenport 2016; Yang et al. 2017; Günther et al. 2020).

Gamma-ray bursts

The short time domain contains relativistic explosions, such as gamma-ray bursts (GRBs). These events feature diﬀerent populations and a variety of possible progenitor systems (for a review, see Levan et al. 2016). Although the energy from GRBs is mostly contained in gamma-rays, if the jet interacts with surrounding circumstellar medium, it can produce an optical afterglow that lasts from days to months (Rees &Meszaros1992). So far GRB afterglows have been detected from Type Ic core collapse supernovae (e.g., Galama et al. 1998), kilonovae (e.g. Tanvir et al. 2013)and abinaryneutronstarmerger(Abbott et al. 2017e). Further detections are crucial for understanding GRBs and the distinct populations. 14 Introduction

Unknown transients

Along with the predicted short transients, there may be currently undiscovered transient types in the short time domain. This parameter space is poorly explored, with only limited observations from programs such as DWF (Andreoni & Cooke 2019), and so may contain yet to be discovered transient types. To ﬁnd, or rule out these events, a systematic search of the rapid time domain is required.

1.1.3 Exoplanets

The first extra-solar planet (exoplanet) around a sun-like star was a hot-Jupter, discovered in 1995 orbiting 51 Pegasi (Mayor & Queloz 1995). In the decades following, many thousands of exoplanets have been discovered, primarily with Kepler and the transit method (Batalha 2014). Statistical studies have shown that exoplanets are ubiquitous, with microlensing revealing that there is at least one bound planet per star (Cassan et al. 2012). Furthermore, from analyses of Kepler data, it is expected that 11–34% of Sun-like stars have Earth sized planets within the habitable zone (Cassan et al. 2012; Pintr et al. 2014). Now that the relative populations of planets are understood, focus has shifted from detecting exoplanets to understanding the alien worlds. The majority of exoplanets have been detected through the transit method, where the stellar flux temporarily drops as the exoplanet occults it. As seen in Fig. 1.8,thetransitiscomprisedoftwoperiods,theingress(oregress),wherethe exoplanet is beginning (or ending) the transit; and the main transit, in which the full exoplanet disk covers the star. Observing the structure of these transits requires high-cadence observations, for example those provided by Kepler and TESS . The transit depth or dimming of stellar flux can be found by considering the emitting area. The flux received from a star can be thought of being emitted from a circle with an area 2πRs,whereRs is the stellar radius. If an exoplanet is transiting the stellar disk, it will obstruct flux according to the relative size of the exoplanet to the host star. In transit the stellar flux, FT ,isgivenby;

2 Rp F = F 1 eff (1.1) T s − R ! " s # $ 1.1 Time domain astronomy 15

Figure 1.8: Cartoon of the time evolution of a transiting exoplanet system. Credit: NASA Ames.

where Fs is the intrinsic stellar flux, Rpeff is the effective radius of the transiting exoplanet, and Rs is the radius of the star. If the exoplanet has an atmosphere then the transit depth may vary with wavelength, as R λ.Thisproportionalityis peff ∝ due to the transmission properties of atoms and molecules present in an exoplanet’s atmosphere (Brown 2001; Lecavelier Des Etangs et al. 2008). Fig. 1.9 shows three possible scenarios for exoplanet atmospheres and the effect it would have on the transit depth.

Although a high level of precision is required to observe exoplanet spectra, it is achievable with modern instruments (for a review, see Crossfield 2015). Primarily, atmospheric properties, such as potassium, sodium, water, Rayleigh scattering and clouds/haze, have been observed for gas giants close to their host stars, known as hot-Jupiters (e.g., Turner et al. 2016; Mallonn & Strassmeier 2016; Sing et al. 2011, 2015; Sing et al. 2016a; Sing et al. 2016b). Despite the increased challenges in observing atmospheres of Earth sized planets, Southworth et al. (2017)detect the presence of an atmosphere for the 1.6 M transiting planet GJ 1132 b. All of ⊕ these measurements rely on detecting slight differences in transit depths at different wavelengths.

The eﬀect of an atmosphere on the transit depth can be understood by separating

Rpeff into two components, Rpeff = RA + Rp;consistingoftheplanet’strueradius

Rp,whichisdeﬁnedbyarockysurface,forterrestrialplanets,oranopticallythick cloud deck for gas giant planets; and RA,theheightabovetheplanetaryradiusat 16 Introduction

Figure 1.9: Cartoon of the effects of planetary atmospheres on transit depth. Top: Extended hydrogen atmosphere, which blocks UV, leading to a larger UV transit depth. Middle: Water dominated atmosphere, which has no scattering. So the effective radius is the same in all wavelengths. Bottom: The presence of clouds can produce the same transit depth for all wavelengths. Credit: NAOJ. which the atmospheric optical depth is < 0.56 for a given wavelength (Lecavelier ∼ Des Etangs et al. 2008). The two components contribute to the total transit depth, however, only RA is wavelength dependent and can provide information on the atmospheric properties of the exoplanet. The impact a planetary atmosphere has on the transit depth can be seen by calculating the area of the atmospheric annulus. The decrease in stellar flux due to an exoplanet atmosphere, FTA ,isgivenby;

R + R 2 R 2 F = F 1 A p p (1.2) TA s − R − R % !" s # " s # $& R2 +2R R = F 1 A A p (1.3) s − R2 " s # thus, observations of the same transit at diﬀerent wavelengths will yield diﬀerent

RA and ﬂux measurements in transit. By comparing observations at multiple wavelengths, the atmospheric properties of an exoplanet can be examined at a range of diﬀerent altitudes. With a diverse set of data from infra-red to UV wavelengths, exoplanet atmospheres can be modelled and properties such as scale height, temperature, and atmospheric escape rate can be constrained. 1.2 Ultra-violet astronomy 17 1.2 Ultra-violet astronomy

Ultra-violet (UV) astronomy is the study astrophysical processes which interact at wavelengths between 100 and 320 nm. All wavelengths of light offer different information about the nature of objects and processes in the Universe. As UV has ahigherenergyperphotonthanopticallight,itisoftenusedasatracerofhigh energy processes, such as star formation, supernovae and GRBs. Although UV astronomy offers a unique window to the Universe, it is limited by atmospheric opacity. Unlike optical wavelengths, where the atmosphere is transpar- ent, UV light is absorbed by a layer in the atmosphere known as the ozone layer.

Comprised of O3,theozonelayerexistsinthelowerstratosphereatanaltitude range of 15–30 km and blocks almost all UV radiation incident on the Earth’s atmosphere (for a review, see McElroy & Fogal 2008). Due to the ozone layer, the UV flux at ground level is negligible, far below a level necessary for ground based UV astronomy to be viable. Due to the ozone layer, UV time domain astronomy is relatively unexplored, leaving many aspects yet to be investigated. Some of the earliest UV measurements were conducted using balloon based telescopes, which would float above the ozone layer, for a time, and act as a temporary UV observatory. Examples of these early flights can be seen in Herse (1979), where a 20 cm UV telescope was flown at 33 km to image the sun and in Staath & Lemaire (1995), where a 30 cm telescope was flown with a UV spectrograph, to observe the sun. Since the pioneering balloon missions, a number of space based UV telescopes have been established. Notable space based UV telescopes include the Hubble Space Telescope (HST), the Ultraviolet Imaging Telescope (Cornett et al. 1992), SWIFT (Gehrels et al. 2004)andGALEX (Martin et al. 2005). Currently only the HST and SWIFT are still operational, with the HST set for decommissioning in the early 2020s, leaving UV astronomy extremely limited. In efforts to expand UV astronomy, a number of proposals have been put forward for space based UV telescopes of various sizes. The largest of these telescopes is the World Space Observatory (WSO–UV ), which will feature a 2 m primary mirror (Shustov et al. 2018). WSO–UV aims to fill the void that will be left by the HST and will have a narrow field of view. 18 Introduction

Alongside the large space telescope proposals, there are a number of proposals for small cubesat UV space telescopes. These smaller UV telescopes aim to explore the UV time domain, searching for high energy transient phenomena, such as those discussed in 1.1.TheULTRASAT mission is planned to feature a 13.3 cm aperture, § 2 alarge802deg field of view, with a spatial resolution of 19.3′′ and a limiting magnitude of 21 (Sagiv et al. 2014). Despite the strong science case, ULTRASAT does not have a launch date. The Cubesat Ultraviolet Transient Imaging Experiment (CUTIE), shares the same science case as ULTRASAT. CUTIE will have a 2.4 cm aperture and a large 121 deg2 field of view, and a limiting magnitude of 19 (Cenko et al. 2017). As with ULTRASAT, CUTIE is yet to receive a launch date. Although the science cases for UV astronomy are clear, particularly in the time domain, the shortage of instruments has limited the field. Any new UV telescope system will have a plethora of key projects to pursue, such as those described in Sagiv et al. (2014).

1.3 Aim of this thesis

In this thesis we outline two new methods for exploring the short time domain in astronomy. We aim to search for events and phenomena that evolve on times- scales less than a day. As many fascinating and crucial questions are tied to what could be found in the short time domain, this thesis was developed keeping key science questions such as the nature of dark energy, detecting oxygen in exoplanet atmospheres, and discovering new short scale phenomena, in mind. Since the short time domain is yet to be thoroughly explored, another key aspect of this thesis was in understanding the occurrence rate of events that might exist in that time, such as optical afterglows from GRBs. To reach the short time domain we have worked on developing a new telescope system alongside analysing existing high-cadence data. The telescope system is described in Chap. 1,withtheprimarysciencecasesdiscussedinChap.3 and 4. To explore the very short time domain (< day lifetime) we utilise public data from ∼ the Kepler/K2 campaigns to conduct a search for transients in background pixels. 1.3 Aim of this thesis 19

This search is described in Chap. 5 with the ﬁrst object discovered by the survey discussed in Chap.6.

1.3.1 Overview of chapters

Chapter 2: Google Loon Ultra-Violet Telescope

We present the design of the Google Loon Ultra-Violet Telescope (GLUV )telescope system. This system is designed to be lightweight, compact and observe at wavelengths currently poorly explored (Sharp et al. 2016). As this system is expected to ﬂy at altitudes that place it within the ozone layer, we analyse what wavelengths are accessible, and develop instruments to directly measure the atmospheric transmission and sky brightness. Future work on this system will build upon the work presented here.

Chapter 3: GLUV Science Cases

The development of GLUV was informed by the science cases and goals covered in this chapter. The primary science goal of GLUV is to identify the progenitors of SN Ia, as there are multiple pathways to produce “normal” cosmological SN Ia that may lead to systematic errors in cosmological measurements. We show that the proposed GLUV system should meet this science goal by routinely detecting shocks generated by SN Ia donor stars, if they exist. Furthermore we show it is possible for GLUV to study exoplanet atmospheres through detailed observation of hot-Jupiters around sun-like stars, and super-Earths around M-dwarf stars.

Chapter 4: Detecting Gravitational Wave UV Counterparts with GLUV

While investigating the transient science case for GLUV,wefoundthatsucha system could provide valuable information on gravitational wave electromagnetic counterparts. We analyse theorised electromagnetic emission mechanisms for all gravitational wave progenitor systems: binary black hole, black hole – neutron star, and binary neutron star. For binary black hole mergers, we ﬁnd that in some limited cases GLUV may be able to detect emission from a fossil accretion disk. In the black hole – neutron star and binary neutron star mergers, we ﬁnd that 20 Introduction

GLUV could provided valuable diagnostic information on system parameters and the merger remnant.

Chapter 5: Kepler/K2: Background Survey

In this chapter we present the Kepler/K2: Background Survey (KS: BS), a systematic search for transients serendipitously discovered by Kepler.The1minute high-cadence and 30 minute slow-cadence observations of Kepler allow us to search for rapid transients that evolve on time scales of 1day.Thesurveyisdefinedby ≤ the characteristics of Kepler/K2,withalimitingmagnitudeof 21 K ,asurvey ∼ p area of 50 deg2,andobservationwindowslasting 80 days. We find that the ∼ ∼ K2: BS survey is capable of detecting transients known to existing transient surveys, such as the Kepler Extra-Galactic Survey (KEGS), and identifying other transients, such as cataclysmic variable outbursts and AGN flares.

Chapter 6: Discovery of a new WZ Sagittae in Kepler/K2 data

We present the first discovery from the K2: BS, a WZ Sagittae type dwarf nova called KSN:BS-C11a. The new dwarf nova KSN:BS-C11a was discovered while in asuperoutburstinK2campaign11,whichwasbothSunfacingandagalactic field, so had minimal concurrent ground based follow-up. Following the discovery of KSN:BS-C11a, we confirmed the nature of the object with a GMOS-N spectrum and imaging with DECam. From the Kepler light curve we find a broken power law rise in KSN:BS-C11a, and in other similar dwarf novae superoutbursts, observed with Kepler and TESS. This discovery indicates new physics in the superoutbursts of dwarf novae, only accessible though high-cadence observations.

Chapter 7: Conclusions and future work

We summarise the work presented in this thesis and present extensions that will carry this work into the future. Translating the K2: BS to the TESS: BS allows for a high-cadence transient search on a much larger scale. Through this search so far, we have identiﬁed numerous ﬂare stars, dwarf novae outbursts and are working towards identifying potentially rare phenomena. 2 Google Loon Ultra-Violet Telescope

2.1 Overview

UV astronomy is a field that, although shows great promise for providing valuable data, suffers from a lack of resources. This limitation stems from the atmospheric opacity at UV wavelengths, and the subsequent cost of space based instruments. The Google Loon Ultra Violet survey telescope (GLUV ), aims to fill the current UV instrumentation void with a low cost, but high impact system. GLUV is a high altitude balloon based UV survey telescope, with the primary science case of performing a high-cadence ( daily) search for transients, such as supernovae. This ∼ project is being developed as part of a collaboration between the Australian National University (ANU), Google X, and NASA Ames/Goddard.

As GLUV is intended to be balloon borne, weight is a key factor in the system design. The telescope will have a large ﬁeld of view of 7o,andaprimarymirror ∼ diameter of 30 cm. To conserve mass, the telescope will follow a compact ﬁve

21 22 Google Loon Ultra-Violet Telescope element catadioptric design, to correct for the large ﬁeld of view (Sharp et al. 2016). With the catadioptric design, shown in Fig. 2.1,itisexpectedtoachieveanangular

1 resolution of 4.64′′ pixel− . Although the bandpass is not currently defined, we expect the effective wavelength to be 300 nm. The planned flight altitude of 20– ∼ 30 km poses some challenges for the system, such as a wavelength cut-offdiscussed in 2.4 and the need to be resistant to the reactivity of ozone. § The development of GLUV has been driven by the following three primary science cases:

1. Detection of supernovae type Ia shock interactions and type II shock breakouts (Kasen 2010; Rabinak & Waxman 2011; Arcavi et al. 2017).

2. Detection of gravitational wave counterparts (Ridden-Harper et al. 2017).

3. Exoplanet atmosphere composition and surface habitability (e.g., France et al. 2013; O’Malley-James & Kaltenegger 2017).

These three cases share design requirements and show promise to provide valuable insight into each of the encapsulating fields. The science cases are further explained in Chap. 3 and 4.TomeetthesesciencecasesGLUV aims to be a low-cost, scalable and long flight duration project that provides the first high-cadence survey of the near-UV to UV spectrum. 2.2 Loon platform 23

Figure 2.1: Preliminary CAD design of the GLUV telescope. Credit R. Sharp.

2.2 Loon platform

The premise of GLUV is to conduct a low cost high-cadence UV survey by utilizing an agreement in which Google X will provide Loon platforms. Project Loon has been under development for some years by Google X to provide wireless internet access to remote areas. To fulfil the commercial requirements, Loon was developed to have a flight duration of 6monthsataltitudesbetween20–30km.TheLoon ∼ flight paths are also highly controllable due to extensive test flights and detailed atmospheric modelling, enabling them to circle a region or fly between continents. Experiments on payload stability have also proven promising for arc-second pointing accuracy. As this is a commercial enterprise, the Loon payloads are recovered after flight to reduce cost.

All of the previous factors make the Loon an ideal platform for a small UV survey telescope. Since the platform is provided, the main cost of this project is in developing the telescope, substantially reducing the running cost of this project, placing it far below the cost of proposed space based UV survey telescopes.

As the agreement stands, the project will not be limited to a single GLUV,rather it is expected that a small fleet of telescopes will be flown. Thus if the project is successful a constellation of GLUV smaybeflownatanygiventime,providing 24 Google Loon Ultra-Violet Telescope high-cadence coverage. Through a ride sharing agreement with Loon, GLUV will have a significantly reduced cost per launch or unit compared to other proposed space telescope missions. Furthermore, it is possible that a small fleet of GLUV smaybeflownatanygiven time. Unlike typical high altitude UV telescopes, GLUV will fly within the ozone layer, presenting unique challenges. We must carefully examine the behaviour of the ozone layer to determine the optimal flight plan considering altitude, latitude and season, to maximise atmospheric transmission given the operational constraints.

2.3 Ozone layer

The ozone layer plays a crucial role in protecting the Earth’s surface from deadly UV irradiation, however, it has made UV astronomy impractical. The ozone layer is comprised of ozone molecules and exists between 10–30 km above the Earth’s surface. The composition of ozone allows for strong and broad absorption feature that dominates the UV to near–UV spectrum, from 200–350 nm (e.g., Serdyuchenko ∼ et al. 2014). For a comprehensive analysis of atmospheric transmission within the ozone layer we analyse spatial and temporal variation in the ozone density. Since a balloon platform has large freedom in its global position it is worthwhile to analyse an atmospheric cross section of ozone density. For this analysis we use the high-resolution vertical ozone profiles from the Global Ozone Monitoring Experiment (GOME-2), O3M SAF dataset (Hassinen et al. 2016). The GOME-2 ozone density profiles are generated by the OPERA algorithm developed at the Royal Netherlands Meteo- rological Institute, and features a spatial resolutions of 80 40 km and a vertical × resolutions of 7–15 km, depending on altitude. Although GOME-2 does not have the highest vertical resolution of ozone detecting satellites, it is still active and can provide recent ozone profiles. GOME-2 measures the ozone profile in a region below the satellite, so over the course of an orbit the ozone profile is measured over a range of latitudes and longitudes. To construct the profile we assume that the primary variation in the ozone structure is latitudinal, thus the longitudinal aspect is neglected. The data are then 2.3 Ozone layer 25

Figure 2.2: Ozone density proﬁles for January, April, July and October obtained from GOME-2 data. The ozone layer is prone to seasonal variation, such as raising and lower of the ozone altitude and density, as evident in the ﬁgures, most notably with the ozone hole forming in October.

averaged together based on altitude and latitude bins. Data from a single orbit, or even a day is insufficient to fill all latitude bins. In order to fill in the latitude bins 10 days of data were averaged together, which has the added benefit of reducing ∼ noise. Although the ozone layer exhibits variation, it occurs on the scale of months, thus averaging over 10 days does not average out short scale features. ∼

The ozone profiles generated from this analysis method are shown in Fig. 2.2. Following the earlier assumption that longitudinal variations are insignificant, these latitudinal profiles are assumed to hold for all longitudes. These profiles are used for all atmospheric transmission calculations and plots in 2.4. § 26 Google Loon Ultra-Violet Telescope

2.3.1 Ozone dynamics

Although the ozone layer exists in the highly stable and stratified stratosphere, variation still occurs in the layer. Due to the nature of the stratosphere, the variations occur on a time-scale of months, driven by seasonal changes. As evident in Fig. 2.2 there are latitudinal variations alongside temporal variations. The main variations both latitudinally and temporally in the ozone layer profile are a result of Brewer–Dobson circulation (Brewer 1949). This is a global circulation cell driven by temperature differences between the equator and poles, primarily the winter hemisphere. The high solar radiation at the equator causes air to rise (upwell) between the tropics ( 23o), move poleward and subducts at high latitudes. This ± process results in the ozone layer being inflated and sparse at the equator, while depressed and dense at high latitudes. These features can be seen in Fig. 2.2. The variation induced by Brewer–Dobson circulation are accentuated by seasonal variations. During winter months the stratosphere cools due to the reduced solar flux. The cooling both acts to strengthen the Brewer–Dobson transport and lower the ozone layer altitude. Thus the winter hemisphere has a dense and lower altitude ozone layer. The peak ozone density nominally occurs through spring to early summer1. During summer the stratosphere is warmed and slows the Brewer–Dobson circulation. With the ozone transport effectively halted, the layer is no longer replenished, resulting in the ozone density decreasing due to destructive reactions with UV pho- tons. Thus, the ozone layer is at its lowest density, but highest altitude during autumn and the beginning of winter. The seasonal variation of ozone layer could provide a noticeable advantage to the atmospheric transmission. Since GLUV is intended to fly near the top of the ozone layer, understanding these variations is crucial to developing the strongest survey.

2.4 Atmospheric transmission

Using the ozone data discussed in 2.3 we show the atmospheric transmission calcu- § lated for a range of wavelengths. Most short wavelength balloon–borne missions have

1The ozone hole in the Southern Hemisphere counteracts this cycle. 2.4 Atmospheric transmission 27 previously ﬂown at altitudes around 80 km, so the wavelength range or bandpass can be freely chosen. In the case of GLUV,ideallyconventionalCCDsensitivitywill set the minimum wavelength to 250 nm, however, the system will likely be limited by atmospheric transmission.

To identify the atmospheric limit to the minimum wavelength we generate an atmospheric transmission model. The model is built oﬀthe data shown in Fig. 2.2 and uses the Beer–Lambert Law to calculate the transmittance, T ,alongalineof sight:

σ l n(z)dz T = e− 0 , (2.1) ! where σ is the absorbing cross section, l is the path length, and n is the number density. As ozone is the primary absorber, we calculate the transmission with the ozone cross sections listed in Serdyuchenko et al. (2014).

To maximise atmospheric transmission, short path lengths are preferred. This places a preference on minimising the angle from zenith, however, the presence of the balloon above the telescope obstructs angles around 0o from zenith. The minimum angle from zenith, θ,iscalculatedasfollows;

1 rb FoV θ =tan− + , (2.2) h 2 ' ( where rb is the balloon radius, h is the tether length, or distance of telescope beneath the balloon, and FoV is the ﬁeld of view in degrees. Putting the expected values

o of rb =7.5m,h =12m(Nagpal & Samdani 2017)andFoV =5,weﬁndthe minimum possible pointing angle from zenith to be 35o.Wetakethemaximum angle from zenith to be 70o,asthisistheconventionalairmasslimit.

With this information we can calculate the expected transmission for UV wavelengths at different altitudes and latitudes. We calculate the atmospheric transmission for the pointing limits, with altitude and latitude resolutions of 100 m and 1o, respectively. As seen in Fig. 2.3,fortheflightaltitudesofGLUV the ozone enforces alowerwavelengthlimitof290nm,witha< 20% atmospheric throughput at 35o. ∼ It is also apparent that transmission increases for latitudes further from the pole, so flights should be conducted at least 30o from the equator. 28 Google Loon Ultra-Violet Telescope

In order to obtain realistic signal-to-noise estimates of the GLUV system, we use these transmission models in the signal-to-noise calculations. 2.4 Atmospheric transmission 29

(a) Fixed viewing angle of 35o from zenith

(b) Fixed viewing angle of 70o from zenith

Figure 2.3: Transmission profiles using the Beer-Lambert Law along line of sights pointing towards the hemispheric pole at a fixed viewing angle through the O3 profiles shown in Fig. 2.2.Contoursoutsidetheverticalredlinesarederivedthrough extrapolation. 30 Google Loon Ultra-Violet Telescope 2.5 GLUV –Pathfinder Spectrograph

The GLUV –Pathfinder Spectrograph (GLUV –PS) is a low cost, UV optimised spectrograph. GLUV –PS was designed to determine if observing UV flux at the planned GLUV flight altitudes of 20–30 km is possible. As discussed in 2.3,modellingin- § dicates that atmospheric transmission should be vastly improved for near-UV wavelengths at flight altitude. GLUV –PS was designed to verify these models. GLUV –PS was designed with off-the-shelf UV optimised components installed in a 3D-printed housing. The GLUV –PS housing shown in Fig. 2.4 accommodates two Thorlabs 15 mm lenses; a Thorlabs 30 mm lens; a 100 µmopticalslit;a600 grooves/mm Edmund optics UV transmission grating beamsplitter; and a Toshiba- TCD1304AP linear CCD array2.Thethroughputsforeachopticalelementare shown in Fig. 2.5 (left). The expected shape of GLUV –PS spectra is shown in Fig. 2.5 (right), where a 1.5 airmass solar spectrum3 is convolved with the known GLUV –PS throughput. The CCD was controlled via an Arduino Duo system4. As the sky brightness that GLUV –PS was to observe was unknown, we included a variable exposure time calculator in the control systems. This system would linearly scale the exposure time of the CCD, based offthe maximum counts received from the last exposure. This reduced the likelihood that spectra would be under- or over-exposed during a flight.

Figure 2.4: 3D model of the GLUV UV spectrograph pathﬁnder. The model is designed to be 3D printed in two sections and joined through interlocking ridges. The design contains holders for 5 optical elements and a strip CCD.

2The optical system was designed by Sharp, R.. 3https://rredc.nrel.gov/solar//spectra/am1.5/ 4Control systems were developed by Gilbert, J.. 2.5 GLUV –Pathfinder Spectrograph 31

1.0 CCD 1.0 Sun Grating GLUV –PS Lens 0.8 0.8 Total

0.6 0.6

0.4 0.4 Throughput (%) Throughput (%)

0.2 0.2

0.0 0.0 200 400 600 800 1000 200 400 600 800 1000 Wavelength (nm) Wavelength (nm)

Figure 2.5: Left: transmission values for the optical elements used in GLUV –PS. Right: the solar spectrum at 1.5 airmass convolved with the GLUV –PS response function. The spectra is truncated due to an unknown QE for the CCD for wavelengths < 400 nm.

2.5.1 GLUV –PS ﬂight

GLUV –PS flew on YerraLoon 1 from the Temora airfield on the 5th of December 2017 (UTC). On this flight GLUV –PS reached a maximum altitude of 32 km and ∼ recorded spectra for the duration of the 2hourflight.TheYerraLoon1payload ∼ also included telemetry instrumentation and an automated camera. The payload and flight image taken at 30 km are shown in Fig. 2.6. ∼ Throughout the flight altitude was recorded, shown in Fig. 2.7. With the altitude for each observation known, we bin spectra into corresponding 5 km altitude bins. Over the total duration of the flight 7989 spectra were collected with the adaptive exposure time, however, many of these spectra have poor data quality. The lack of platform stability, and clouds present at low altitudes (0–15 km) produced a highly variable sky brightness observed by GLUV –PS. The changing sky brightness and angle to the Sun resulted in many spectra shifting position and/or being unusable due to being under- and over-exposed as seen in Fig. 2.8.

For the spectra to be analysed we must first define a wavelength scale for GLUV – PS. We find the wavelength scale by fitting the convolved 1.5 airmass solar spectrum with the theoretical GLUV –PS throughput to each spectra observed during the

YerraLoon 1 ﬂight. The wavelength scale, λs,innmisdeﬁnedlinearlyasfollows,

λs = λ0 + ps.x (2.3) 32 Google Loon Ultra-Violet Telescope

Figure 2.6: Left: the YerraLoon 1 ﬂight payload. Right: an image from YerraLoon 1atanaltitudeof 30 km. ∼

15 Altitude (km) 10

0 00:00 01:00 02:00 Time (5 Dec 2017 UTC)

Figure 2.7: GLUV –PS flight profile, the total flight time was 2hours. ∼

where λ0 is the wavelength corresponding to the first pixel of GLUV –PS in nm, ps 1 is the pixel scale in nm pix− ,andx is the pixel number. The parameters λ0 and ps are identified through χ2 fitting with each GLUV –PS spectra and the convolved solar spectra. As the performance of GLUV –PS is unknown past 400 nm and the solar spectra can change drastically for λ<500 nm at altitude, we only fit the solar spectrum from 450–850 nm. An example of a solar fit is shown in Fig. 2.9,witha GLUV –PS spectrum.

The results of ﬁtting all GLUV –PS spectra obtained in the YerraLoon 1 ﬂight

1 are shown in Fig. 2.10.Weﬁndthatλ0 =200nmandps =0.35 nm pix− . With 2.5 GLUV –Pathfinder Spectrograph 33

1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4 Normalised ﬂux Normalised ﬂux

0.2 0.2

0.0 0.0

0 500 1000 1500 2000 0 500 1000 1500 2000 Pixel Pixel

Figure 2.8: Examples of bad spectra recorded on the GLUV –PS ﬂight. Left: the detector saturated. Right: ashortexposuretimeandstraylight. awavelengthscaleestablished,wenowimposestrictvettingofthespectraunder following conditions:

1 7 1. The maximum counts s− must be > 3 10 . × 2. The normalised counts at 300 nm must be < 0.8.

3. The normalised counts at 800 nm must be < 0.5.

4. The spectrum must be recorded before balloon burst at 1:00 UTC. ∼ These strict conditions cut all under- and over-exposed spectra, and spectra dominated by spurious electrical signals. Following the cuts we reduce the number of usable spectra from 7989 to 65, with 48 between 0–5 km and 15 between 20– 30 km, as seen in Tab. 2.1.Thelackofusablespectrabetween5–20kmislikely due to the presence of cloud, and turbulence causing large changes in the pointing of GLUV –PS. We average the remaining data in altitude bins of 0–5 km and 20–30 km. The resulting spectra presented in Fig. 2.11.WefindthatnotonlyisGLUV –PS sensitive to short near UV wavelengths, it detected higher relative UV flux at/within the ozone layer compared to the ground. From Fig. 2.11 (right), it is clear that compared to the normalised ground spectrum, the 20–30 km spectrum has 2timestheflux ≥ at λ 300. The high flux at short wavelengths indicates that even though GLUV ≤ 34 Google Loon Ultra-Violet Telescope

1.0

0.8

0.6

0.4 Normalised ﬂux

0.2

GLUV –PS 0.0 Solar spectrum

200 300 400 500 600 700 800 Wavelength (nm)

Figure 2.9: Example of the wavelength ﬁt to a random GLUV –PS ﬂight spectra (blue). A solar spectrum convolved with known throughputs of optical elements in GLUV –PS is shown with the orange dashed line.

Table 2.1: Observations taken by GLUV –PS in 5 km altitude bins. After strict quality cuts spectra were only available at 0–5 km and 20–30 km.

Altitude (km) Observations Usable 0–5 5321 48 5–10 721 0 10–15 582 0 15–20 446 0 20–25 382 9 25–30 536 6 Total 7989 63 is expected to ﬂy within the ozone layer, it will out perform ground based telescopes at short wavelengths. Further analysis, such as calculating the atmospheric transmission and comparing it to models shown in 2.4 requires further work. To make such a compari- § son, we must have a well deﬁned response function for GLUV –PS, particularly for λ<400 nm, and pointing stability. Despite these shortcomings, the apparent signal observed by GLUV –PS at 250 nm suggests that the atmospheric transmission ∼ models may underestimate atmospheric transmission. If the previous two limitations are resolved, then the atmospheric transmission models can be tested and a directly observed atmospheric transmission can be used in signal-to-noise calculations discussed in 2.7. § 2.5 GLUV –Pathfinder Spectrograph 35

100

1 10

2 10 2 10 Occurence (%) Occurence (%)

3 3 10 10

4 4 10 10 100 150 200 250 300 0.2 0.3 0.4 1 Initial wavelength (0) Pixel scale (nm pixel )

Figure 2.10: Distributions of the wavelength scale parameters derived for each spectrum from GLUV –PS. Left: the wavelength corresponding to the ﬁrst pixel, with amedianvalueλ0 =200nm. Right: the pixel scale of GLUV –PS, with a median 1 value of 0.35 nm pixel− .

1.0 6 0–5 km 20–30 km 0.8 5 U band

0.6 4

0.4 3 Normalised ﬂux

0.2 0–5 km 2 20–30 km Normalised ﬂux to 0–5 km spectrum U band 1 250 300 350 400 450 500 250 300 350 400 450 500 Wavelength (nm) Wavelength (nm)

Figure 2.11: Left: normalised spectra at 0–5 km (blue) and 20–30 km (orange), with the U band overlaid. Right: the spectra shown left are normalised to the 0–5 km spectra, with the U band overlaid. These plots show a substantially higher UV fraction at/within the ozone layer compared to the ground. 36 Google Loon Ultra-Violet Telescope 2.6 Sky brightness

To produce an accurate representation of the expected GLUV signal-to-noise, we must know the UV sky background. Since there are no direct measurements of the UV flux at the planned flight altitude of 20–30 km we must establish a presumptive value, based offsimilar measurements. As a reference point, the U-band sky brightness for Mauna Kea is included alongside the space based UV sky brightness presented in Waller & Stecher (1998). As discussed in Waller & Stecher (1998), at 250 nm, the background is produced from OI emission and the galactic and extragalactic backgrounds. As GLUV will still be subject to the atmospheric background and will likely not reach 250 nm, we take the Waller & Stecher (1998)backgroundof

2 26 mag arcsec− to be the theoretical maximum. The guiding sky brightness values are shown in Tab. 2.2.

As the sky brightness is crucial to understanding the sensitivity of GLUV,we must conduct measurements to determine it. Currently a new robust spectrograph is under development to measure the sky background. The new spectrograph is designed to ﬂy on a Loon and collect direct measurements of the UV sky background at the expected ﬂight altitudes of 20–30 km.

Table 2.2: Sky brightness for selected sites in U-band, the upper UV limit and a presumptive value. It is expected that the real value will lie between the Guess and Upper limit values. Mauna Kea Presumptive Waller & Stecher (4.2 km) value 250 nm space (20 km) Sky brightness 23.2 25 26 2 [mag arcsec− ] 15 16 16 Flux 2.2 10− 5.0 10− 2.4 10− 1 2 2 × × × [erg s− cm− arcsec− ] 2.7 GLUV signal-to-noise calculator 37 2.7 GLUV signal-to-noise calculator

The development of GLUV,likeanyotherinstrument,mustassessthenoiseinherent to the system. During the preliminary system design phase, signal-to-noise calculations (S/N) are crucial in determining what physical properties are required to meet the science goals discussed in Chap. 3 and 4.ThesciencecasesforGLUV can be sorted into two observational categories; surveying large volumes at a S/N ! 5; and conducting detailed observations for individual targets, such as transiting exoplanets, requiring S/N ! 100. Furthermore, for the survey observation mode we expect to achieve a S/N ! 5withlimitingmagnitudesofmAB =19. With the S/N and magnitude requirements defined, we can analyse the likelihood that a telescope system built to the constraints imposed by the Loon platform can be successful. Due to weight constraints imposed by the platform, GLUV is expected to be a small telescope with a 30 cm diameter primary mirror. To meet the requirement that GLUV features a large field of view (FoV) 7deg2,thesecondarymirror ∼ must have a large diameter, 45% that of the primary. As discussed in 2.4,the ∼ § expected flight altitude will impose an atmospheric transmission of 40 80%, at ∼ − the expected wavelength range of 210 350 nm. It is not expected that platform − stability and image blurring will significantly impact, as it is possible to stabilise balloon platforms to high precision (e.g., Kraut et al. 2008). Outside of the platform induced constraints, the instrument itself will strongly influence the S/N. The preliminary telescope design, presented in Sharp et al. (2016) and shown in Fig. 2.12,containstwomirrorsand3lenses.Inthepreliminarysensi- tivity calculations it is assumed that each optical element has a 95% efficiency, which leads to an overall throughput efficiency of 77% for the optical system. Furthermore the filter, which is currently undefined, will have an imperfect transmittance. For the purpose of these preliminary calculations an efficiency of 80% is selected, which is on par with the HST filter efficiencies, however, it is noted that the SWIFT uvw1 filter has an efficiency of 20%. Asignificantportionoftheinstrumentnoisewillarisefromthedetector.Cur- rently the detector for GLUV is not defined, however, for the purposes of these calculations, the Finger Lakes Instruments Proline 4710 w/e2v Technologies 4710 CCD is

1 1 used. The detector is a 1024 1024 pixel array, with 13 µmpixels,a0.05 e− s− pix− × 38 Google Loon Ultra-Violet Telescope

Figure 2.12: The GLUV ﬁve element catadioptric design was informed by reference to the CSTAR telescope concept (Yuan et al. 2008). Figure taken from Sharp et al. (2016).

1 dark current, 9 e− pix− readout noise, and an average quantum efficiency of 55%. Crucially, the dark current of this detector is less than the expected sky background, so for sufficiently long exposures the images can be background limited. Taking account of all the known losses in transmittance, we expect the GLUV transmittance from the optical system to be 34%. In this proposed system, ∼ the main limiting factor is the low detector quantum efficiency. If atmospheric transmission from 2.4 is taken into account, the total system efficiency is expected § to be between 14 27%. − When considering the signal-to-noise ratio of the instrument, all sources of noise must be accounted for. The sources of noise and how they contribute to the total noise, N,isasfollows;

2 N =(D + Rsky + Rsource) t Ap + Read Ap, (2.4)

1 1 where D is the dark current in e− pix− s− , Rsky is the sky background noise in 1 1 1 1 e− pix− s− , Rsource is the source noise in e− pix− s− , t is the exposure time in s, 1 Ap is the aperture in pix and Read is the RMS readout noise in e− pix− .Thedark current and readout noise are speciﬁc to a camera, thus are changeable based on availability. Since the sky background noise is uncontrollable, the system should be optimised such that the sky background noise is the dominant noise source. To ensure the images are noise limited by the sky, the target dark current must be 2.7 GLUV signal-to-noise calculator 39

104 ) e 103 Readout noise Dark current 2 25 mag/00 2 24 mag/00 102 Sky background Noise per 5” apeture (

101 0 50 100 150 200 250 300 Exposure time (s)

Figure 2.13: Noise sources for GLUV per 5” aperture. smaller or equivalent to the sky background noise. Similarly, the minimum exposure time is set by the time at which D + Rsky > Read. The individual contributions of noise sources, excluding source noise, can be seen in Fig. 2.13. As the sky background is not exactly known, we allow a range of values between the Mauna Kea and the hypothesised UV sky brightness. This background range can lead to a large diﬀerence for the time at which sky noise dominates the readout noise, thus the optimal exposure time. To reduce the likelihood of telescope drift interfering with the images, we take the maximum exposure time to be 900 s (15 mins), when for all scenarios observations will be dominated by the sky background noise. With the preliminary S/N calculator we have developed for GLUV we can calculate the limiting magnitude of the system. For the primary science case of detecting transients, such as SN Ia, a S/N 5 is required. Fig. 2.14 shows the expected S/N ≥ for sources with apparent magnitudes of 10, 21, and 22, at atmospheric transmission of 40% and 80%. We ﬁnd that for the presumptive sky brightness range GLUV should be capable of reaching a limiting magnitude of 22 for S/N 5. It is possible ≥ that neglecting atmospheric scattering and other factors may lead to this being an overestimate, so further testing is required. 40 Google Loon Ultra-Violet Telescope

70 m = 19 m = 21 m = 22 16 7 60 14 6 50 12 5 40 10 4 30 8 3

Signal to noise 2 25 mag/00 6 20 2 2 24 mag/00 4 10 Keck 2 1 S/N=5 0 0 0 0 250 500 750 1000 0 250 500 750 1000 0 250 500 750 1000 Exposure time (s)

(a) T = 40%

100 25 m = 19 m = 21 m = 22 10 80 20 8

60 15 6

40 10

Signal to noise 2 4 25 mag/00 2 24 mag/00 20 5 Keck 2 S/N=5 0 0 0 0 250 500 750 1000 0 250 500 750 1000 0 250 500 750 1000 Exposure time (s)

(b) T = 80%

Figure 2.14: GLUV signal-to-noise for a variety of sky backgrounds, sources at 19, 21, and 22 mag at diﬀerent atmospheric transmissions (T). The base signal-to-noise requirement of 5 is marked with dashed horizontal lines. 3 GLUV Science Cases

3.1 Supernovae shocks

Although UV observations of SN are valuable, the primary objective with GLUV is to observe SN shocks. With these shock observations it will become possible to understand progenitors and underlying physics behind the explosions. For this reason we will focus on the detection of SN Ia shock interactions and CC SN shock breakouts. As these events are often much fainter and more challenging to detect, assuring that GLUV detects a suﬃcient number of these events will also guarantee that at least an equivalent number of SN are observed.

41 42 GLUV Science Cases

3.1.1 SN Ia shock interaction

For this science case, we use the SN Ia shock interaction models presented in Kasen (2010). In this model, the shock interaction produces the following luminosity,

43 1/4 7/4 3/4 1/2 1 Lc,iso =10 a13 Mc v9 κe− tday− ergs s− , (3.1)

13 where a13 is the binary separation in units of 10 cm; Mc is the ejecta mass in units 9 1 of the Chandrasekhar mass (1.4 M ); v9 is the ejecta velocity in units of 10 cm s− ; ⊙ 2 1 κe is the ejecta opacity in units of 0.2cm g− ;andtday is the time from explosion in units of days. It is worth noting that the shock luminosity is insensitive to the ejecta mass. The spectrum of this model is given through black body radiation, with the eﬀective temperature, Teff ,deﬁnedas,

4 1/4 35/36 37/72 T =2.5 10 a κ− t− K. (3.2) eff × 13 e

The radius, r,oftheshockandthereforethesizeoftheemittingareaisgiven by,

Lc,iso r(Lc,iso,Teff )= 4 m, (3.3) )4πκBTeff where κB is the Stefan Boltzmann constant. With the effective temperature and radius defined, the spectrum of can be calculated and the brightness at different filters defined. As emission is directional, an additional viewing angle factor, f,mustbein- cluded,

f(θ)=(0.5cos(θ)+0.5)(0.14θ2 0.4θ +1), (3.4) − where θ is the viewing angle of the system, for θ =0thecompanionstarliesbetween the WD and the observer. Finally, the shock interaction ﬂux is given by,

2hc2 4πr2 f(θ) F = 5 hc . (3.5) λ λκT e eff 1 − 3.1 Supernovae shocks 43

Figure 3.1: Kasen (2010)shockinteractionmodelsfor4SNIaprogenitorcasesfor the GLUV ﬁlter (290–350 nm), integrated with the SN 2011fe SWIFT uvw1 light curve, shown by the black points. The viewing angle dependence creates a wide range of possible magnitudes for each model.

As the shock interaction model described above does not model other processes such as 56Ni decay, we append the shock model to an observed SN Ia. We choose SN 2011fe to be the “normal” SN Ia template, as early UV observations with SWIFT showed no deviation from the expanding ﬁreball rise model (Brown et al. 2012). The shock models for a variety of progenitor systems are shown in Fig. 3.1, where some relationships are clear. Larger companion stars produce brighter shocks and longer shock durations, while extreme viewing angles can hide signatures of the shock. To fully investigate if “normal” SN Ia are produced by single degenerate systems, GLUV must be capable of regularly detecting the early phase of SN Ia. As most shock emission decays after 1day,GLUV must feature at least daily cadence ∼ on extragalactic survey ﬁelds. Furthermore, since the shocks for 1 M companions ⊙ only have a maximum absolute magnitude of 14.5, GLUV must be as sensitive ∼− as possible, supporting the case for a 30 cm diameter.

3.1.2 CC SN shocks

For the CC SN shock, we use the Arcavi et al. (2017) SW16 Red Super Giant (RSG) model that builds on the Rabinak & Waxman (2011) model to calculate UV ﬂux. For this preliminary study of detectability, we only consider the base case for RSG, which are expected to lead to SN II-P. The SW16 RSG shock luminosity, LRSG,is 44 GLUV Science Cases given by,

2 0.086 2 v t − v R13 L =1.88 1042 s,8.5 , s,8.5 RSG × f Mκ κ " ρ 0.34 # 0.34 1.67t exp erg/s, (3.6) 1 0.5 × !− %(19.5κ0.34Mevs,−8.5) &$

8 where vs,8.5 is the shock velocity in units of 10 .5cm;t is the time since explosion in 13 days; R13 is the progenitor star radius in units of 10 cm; κ0.34 is the ejecta density 2 1 in units of 0.34 cm g− ; M = Me + M ,whereMe is the ejecta mass in units of ⊙ 0.5 solar mass; ﬁnally, fρ =(Me/M ) .Theeﬀectivetemperature,TRSG,is given by, ⊙

2 2 0.027 0.25 4 vs,8.5t R13 0.5 T =2.05 10 t− K. (3.7) RSG × f Mκ κ " ρ 0.34 # 0.34

With the eﬀective temperature and luminosity, we calculate the radius of the photosphere, rph,usingEq.3.4, as in the SN Ia case. Finally, the ﬂux, F ,iscalcu- lated with,

2 2 2hc 4πrph F = 5 hc . (3.8) λ e λκTRSG 1 −

In this preliminary study we set vs,8.5 = κ0.34 =1andonlyconsidertheimpact of progenitor mass and radius on shock detectability. We note that increasing the shock velocity shortens the lifetime of the shock breakout and increases the peak brightness of the shock, particularly for short wavelengths. An example RSG shock model for a M =10M and R =500R progenitor is shown in Fig. 3.2 (left). It ⊙ ⊙ is apparent that the peak brightness is higher for shorter wavelengths. With the SW16 model we calculate a range of peak magnitudes that GLUV should expect to observe for RSG shocks. We take the bounds on RSG progenitors, that lead to SN II-P from Smartt (2015a), where M (8, 18) M and R (3.5, 7) ∈ ⊙ ∈ × 1013 cm. The range of peak magnitudes is shown in Fig. 3.2 (right), where the peak magnitude varies by only 0.4mag.Forratecalculations,wetaketheabsolute ∼ magnitude for SN II-P shocks to be -18 mag, furthermore the shock spends ! 1day near peak magnitude. 3.2 Survey strategy & rates 45

18 18.20 14 18.15 17

) 18.10 16 13

M 18.05 15 12 18.00 14 11 17.95 GLUV 13 10 Progenitor mass ( 17.90 Absolute magnitude (AB) b band

g band Peak absolute magnitude (AB) 12 9 17.85 r band 8 17.80 0 5 10 15 20 4 5 6 7 Time (days) Progenitor radius (1013 cm)

Figure 3.2: Arcavi et al. (2017) SW16 model for Red Super Giant (RSG) shock breakouts. Left: shock breakout model for a 8 M and 3 1013 cm radius progenitor ⊙ RSG. The shock is brightest at shorter wavelengths, with× the GLUV band of 300 nm receiving the highest ﬂux. Right: peak magnitudes of GLUV band RSG shocks for progenitors ranging from 8 15 M in mass and (3.5 7.1) 1013 cm in radius, ⊙ to represent the expected population− of RSGs that lead− to type× II-P supernovae (Smartt 2015b).

3.2 Survey strategy & rates

Equipped with an understanding of the transient science cases, a survey strategies can be considered. Due to the nature of the GLUV project, it is possible to run different surveys. The first mission will be limited to a single GLUV limiting survey variations, however, if a constellation of several GLUV sareflowninagiven campaign, numerous strategies may be available. In the case where a constellation of GLUV sareflying,atradeoffmustbemadebetweencadenceandsurveyarea. Each strategy will have its merits and drawbacks for the individual science case. The following outlines the strategies most useful to each science case.

3.2.1 Preliminary Survey Conﬁguration

For both the supernova and GW science cases, a cadence of a day or less is required. The cadence requirements must also be balanced with the observation depth and survey area, this will be discussed for the individual science cases. Given the daily cadence requirement, a maximum survey area can be calculated for a GLUV instrument. Assuming that the available observational time per night is 8 hours, the 46 GLUV Science Cases maximum survey area can be calculated as follows;

8 SA = .F oV (3.9) te where SA is the survey area, te is the exposure time plus 10% overhead and FoV is the telescopes field of view. From Fig. 2.14,itisclearthatGLUV is capable of reaching a 22 limiting magnitude of for a 10 min exposure. If we take 10 min to ∼ be the default exposure time, the maximum survey area would be 300 deg2.In ∼ the event that a constellation of 10+ GLUV sisflownatanygiventime,atrade offcan be made between increasing the cadence by staggering the constellation longitudinally, and increasing survey area. As GLUV is intended to search for rapid transients, we must account for the probability that an event will be observed in the survey area for a given cadence,

C,andeventlifetime,Elt,

Elt if Elt < 1, P (E C)= C C (3.10) lt ∩ ⎧ ⎪1ifElt 1. ⎨ C ≥ ⎩⎪ With the probability of detection and survey area bounds established, the expected detection rates for transients can be calculated.

3.2.2 SN Ia shock interaction detection rates

For the SN Ia calculations we use the Li et al. (2011b)volumetricrateforSNIaand

5 3 1 the fraction of “Normal” SN Ia from Li et al. (2011a): (3.01 0.62) 10− SN Mpc− yr− ± × at z = 0. We use shock the models from Fig. 3.1 as the source of shock absolute magnitudes and detection time window. Since the shock magnitude depends on viewing angle, we calculate the accessible volume for each of the simulated viewing angles. Finally, for each volume we calculate the number of expected detections over a 6 month observation campaign, and average it to reach an expected detection rate. We ﬁnd that with the sensitivities presented in Chap. 2.7, GLUV is capable of fulﬁlling the primary science case. As seen in Tab. 3.1,eveninthe40%atmospheric transmission scenario GLUV is expected to detect 1.0 0.2shockinteractionsfrom ± 3.3 Gravitational wave counterparts 47

Table 3.1: Detection rates for SN Ia shock interactions for given models over a hypothetical 6 month GLUV campaign, including atmospheric transmission and signiﬁcance. For these calculations the Li et al. (2011b)normalSNIarateisused. We assume no preference for SN Ia progenitor system geometry, allowing for all possible angles.

Atmospheric transmission Progenitor Model 40% 80% (Kasen 2010) 3σ 5σ 10σ 3σ 5σ 10σ 1M 2 11.0 0.20.4 0.1 4 11.8 0.40.7 0.1 ⊙ 2M 4 ± 12± 10.7 ± 0.1 7 ± 13± 11.2 ± 0.2 ⊙ 6M 20± 410± 23± 1 33± 716± 36± 1 ⊙ 1M RG 260 ± 50 130 ± 30 50 ± 10 460 ± 90 210 ± 40 80 ± 20 ⊙ ± ± ± ± ± ±

1 M companion systems at a conﬁdence of 5σ.IfGLUV meets this expectation, ⊙ then it will be capable of determining the progenitors of “normal” SN Ia.

3.2.3 CC SN shock breakout detection rates

For a preliminary analysis of detecting CC SN shocks with GLUV we calculate the expected detection rate of SN II-P shocks. For the SN II-P rate, we use 70% (Li et al. 2011a)ofthetotalSNIIvolumetricratepresentedinLi et al. (2011b), giving

5 3 1 the SN II-P rate to be (3.113 0.973) 10− SN Mpc− yr− .Wealsotakethepeak ± × magnitude of SN II-P shocks from 3.1.2 to be M = 18, and duration to be § shock − 3days. ∼ Due to the bright nature of SN II-P shocks, a large volume is accessible to GLUV.Weﬁndforthislimitedplausibilitystudythatovera6monthcampaign GLUV should detect 400 100 SN II-P shocks at a conﬁdence of 5σ. Although this ± large number is likely unphysical due to the simplifying assumptions made in this rate calculation, it is a strong indicator that GLUV would be capable of observing numerous SN II-P shocks.

3.3 Gravitational wave counterparts

Anewandactiveareaoftransientastronomyisindetectinggravitationalwave counterparts and kilonovae. As there are many possible outcomes and models for gravitational wave sources, the development of this science case led to a paper presented in Chap. 4 (Ridden-Harper et al. 2017). In this paper we review the diﬀerent 48 GLUV Science Cases emission mechanisms for binary neutron star, black hole neutron star, and binary black hole mergers, and calculate the expected detection rates from hypothetical surveys with a full deployment of a constellation of GLUV s.

3.4 Exoplanet atmospheres

As GLUV will be unique in the ability to oﬀer near-UV measurements, initially via photometry, there are critical exoplanet science objectives it could contribute to. The utility of GLUV will be in providing near-UV data points for transmission spectroscopy of gas giants close to their host star, Earth sized exoplanets orbiting M-dwarf stars, and provide a way to monitor the UV environment of M-dwarf stars. In this section we conduct a preliminary analysis to assess if indeed GLUV could be capable of providing valuable insights into these science cases, using the design optimised for transient science.

3.4.1 Hot-Jupiters

Soon after searches for exoplanets began, Jupiter sized bodies orbiting close to their host star were discovered. These massive planets were collectively labelled hot- Jupiters. The high detection rate of hot-Jupiters was due to the large relative size of the exoplanet to its host star. In the case of detecting hot-Jupiters via the transit method, the relative size of the exoplanet is 10% R ,whichleadstoa 1% dip ∼ s ∼ in stellar flux during transit. For gas giants close to their host star, the transmission spectra for wavelengths 700 nm will likely be dominated by Rayleigh scattering, the cause of which is ≤ thought to originate from extended H2 atmospheres about the gas giants. With simultaneous measurements across the spectrum, the Rayleigh scattering profile can be determined, as shown in Fig. 3.3,withwhichthescaleheightandeffective temperature of the atmosphere can be calculated (Southworth et al. 2012; Pont et al. 2013; Sing et al. 2015, 2016b). Currently, all undisputed observations of transit depths around 300 nm are from the Hubble Space Telescope (HST ). As the HST mission is drawing to a close, and with it sensitive space based UV observations, it is imperative that a new 3.4 Exoplanet atmospheres 49

Figure 3.3: Transmission spectra of WASP-31b obtained with the HST.Inpurple is the Rayleigh scattering model with a cloud deck and in red is a cloud free solar abundance model. Taken from Sing et al. (2015).

method is identiﬁed. Currently, there is no replacement for the HST in near-UV to UV capabilities. The upcoming James Webb Space Telescope is unable to observe wavelengths < 600 nm, so the development of a new UV telescope is critical to maintain and extend current capabilities in analysing exoplanet atmospheres.

For GLUV to be successful it must be able to achieve high precision photometry. In the case of hot-Jupiters dominated by Rayleigh scattering, as seen in Fig. 3.3, the effective radius is expected to increase by 1% at UV wavelengths. Thus from ∼ 5 Eq. 1.3,thechangeinstellarfluxduetoanatmosphereis 10− .Therefore, ∼ the cumulative signal-to-noise required is on the order of 105,toderiveacredible effective planetary radius. This high signal-to-noise requirement can be obtained by averaging observations over multiple transits, with equal observation time in and out of transit.

Turner et al. (2016)presentsthestate-of-the-artinU -band photometric measurements of transiting exoplanets for 15 hot-Jupiters. Although this study obtains asignal-to-noiseoftheorder104 at 365 nm, it is still capable of identifying instances of strong Rayleigh scattering. So if GLUV can achieve a signal-to-noise an order of magnitude greater than that of Turner et al. (2016)at 300 nm, GLUV will provide ∼ a precise and unique dataset for identifying Rayleigh scattering in hot-Jupiters. 50 GLUV Science Cases

3.4.2 M-star planetary systems

M-stars have become a crucial target for observing exoplanet systems. Since M-stars are close to the same size as Jupiter, detecting Earth sized exoplanets (exo-Earths) orbiting these stars, inside the habitable zone, becomes near equivalent to detecting hot-Jupiters orbiting Sun-like stars. Although it is possible to detect exo-Earths in the habitable zone, it is unknown if such planets could support Earth-like life, due to the high UV output and flares emitted by M-stars. For GLUV,M-starsoffer two science cases: 1) monitoring exoplanet transits, as with hot-Jupiters; and 2) extended monitoring of M-star flare activity for exo-Earth habitability analysis.

Exo-Earth atmospheres

The discovery of Earth sized planets orbiting dwarf stars such as Proxima Centauri (Anglada-Escudéet al. 2016) and TRAPPIST-1 (Gillon et al. 2017), have presented the best possibility for analysing atmospheres of Earth sized exoplanets. The principle behind the high detection rate of hot-Jupiters applies to Earth sized exoplanets orbiting dwarf stars, as the relative size of the exoplanet and star, Rp 10%, is Rs ≈ the same for both systems. ' ( There is a wide range of atmospheres that many be present on Earth sized exoplanets. Unlike the case of hot-Jupiters, the atmospheric density of exo-Earths can range from a dense atmosphere 100 times that of Earth, to no atmosphere. ∼ Alongside the added complexity of atmospheric density is its composition, which can lead to transmission spectra vastly different to hot-Jupiters. Bétrémieux & Kaltenegger (2013)calculatethetransmissionspectrumoftheEarth,whichshows that O2 and O3 dominate the spectrum at wavelengths < 300 nm, obscuring the underlying Rayleigh scattering, as seen in Fig. 3.4.TheO3 signature provides a strong case for GLUV observations. O’Malley-James & Kaltenegger (2017) calculate the UV surface habitability of the TRAPPIST-1 planets from an atmospheric simulation which produced a range of ozone density profiles. In cases for aerobic atmospheres the ozone layer exhibits similar concentration and extent to that of the Earth’s ozone layer. Using the Earth transmission spectrum from Bétrémieux & Kaltenegger (2013)asananaloguefor TRAPPIST-1 planets, we can explore the plausibility of detecting an ozone layer 3.4 Exoplanet atmospheres 51

Figure 3.4: The expected transmission spectra of Earth. The dot-dashed and dashed lines show the individual contribution of Rayleigh scattering with and without the effect of refraction, respectively. Taken from Bétrémieux & Kaltenegger (2013). with GLUV. As seen in Fig. 3.4 the presence of an ozone layer increases the effective radius by 20 km more than for visual wavelengths. Similarly, the presence of an Earth- ∼ like ozone layer produces an effective radius 15 km greater than that of Rayleigh ∼ scattering alone at 300 nm. The variation in planetary radius from 300 nm to 400 nm with an Earth-like ozone layer is 3% RN ,whilethevariationdueto ∼ ⊕ Rayleigh scattering (without an ozone layer) is 0.08% RN .Thus,GLUV must be ∼ ⊕ sensitive to a variation of 0.01% in the effective radius of an exo-Earth, by Eq. 1.3 ∼ 5 this corresponds to a decrease in stellar flux of 10− to determine the existence ∼ of an ozone layer. So to detect the presence of an ozone layer in the atmosphere of an exo-Earth orbiting an M-dwarf star, GLUV requires a signal-to-noise of 105, ∼ identical to the hot-Jupiter case.

Monitoring M-dwarf stars, the hosts of exo-Earth systems

An added challenge to the detection of ozone layers and classifying habitability for planets orbiting M-dwarfs is the UV variability. Although M-dwarfs are far 52 GLUV Science Cases cooler and less massive than the Sun, they are prone to strong, UV bright flares (e.g., Hawley & Pettersen 1991; France et al. 2013). The potential for high UV emission and variability threatens the possibility for stable planetary atmospheres and habitability (for a review, see Shields et al. 2016). Understanding the flare rate of M-dwarfs is crucial for the search of habitable worlds. Loyd et al. (2018)foundthroughHST far-UV observations that young (< 40 Myr) M-dwarfs produce significant UV flares that may disrupt the formation of planetary atmospheres. Furthermore, Tilley et al. (2019)findsthattheflarerate and intensity of active M-dwarfs, would be enough to destroy ozone layers around un-magnetized exo-Earths. Without an ozone layer, exo-Earths orbiting M-dwarfs would be subject to dangerous levels of surface UV radiation. While the high UV levels may not exclude the presence of life on these worlds (i.e., O’Malley-James & Kaltenegger 2019), it does reduce habitability. While there are numerous systems that can monitor for M-dwarf flares, i.e., TESS and EVRYFLARE (Howard et al. 2019), there is limited survey capability in the UV. Even through a SN search GLUV would detect numerous M-dwarf flares, due to the large field of view, at daily cadence. GLUV could also conduct directed observations, with < 10 minute cadence of interesting M-dwarfs. ∼

3.4.3 Transit signal-to-noise

To investigate the plausibility of the exoplanet science case, we use the signal-to- noise (S/N) calculator presented in 2.7 to calculate the expected S/N for stars § ranging from mGLUV =8to14.Inthisplausibilitystudy,weneglecteffectsof detector saturation and inter/intra-pixel variability, but note they will introduce noise thus making this analysis an upper bound. From the GLUV S/N calculator we find that the sensitivity needed for the transient science cases lends well to the exoplanet science case. In this preliminary analysis, we use the S/N achieved after a 5 minute exposures. For hot-Jupiters, the transit time is taken to be 2hours,leadingto 20 exposures per transit, ∼ ∼ that are averaged to increase the S/N. In the case of smaller systems, such as exo- Earths orbiting M-dwarfs the transit time is taken to be 1hour,allowingfor ∼ 10 exposures per transit. The S/N can be further improved by stacking multiple ∼ 3.4 Exoplanet atmospheres 53 transits, observed on different occasions with a single GLUV,orsimultaneouslywith multiple GLUV s.

The number of transits and time required to reach the a S/N of 105 is shown in Fig. 3.5.Formtheseinitialcalculations,itplausiblethatGLUV could meet the requirements for this science case, especially if multiple GLUV ssimultaneously observe transits. With this positive initial result, more work is required to fully investigate the plausibility of the exoplanet science case. Neglected eﬀects such as telescope pointing, the detector saturation time, and/or detector variability may present insurmountable systematic noise.

If GLUV does perform close to the performance shown here, it will provide valuable and unique data on exoplanets. Combining GLUV transit depths with those from longer wavelength observations could identify Rayleigh scattering in the atmospheres of hot-Jupiters and potentially identify the presence of oxygen and an ozone layer in exo-Earth atmospheres.

20 80 30 60

18 70 27 50 16 24 60 14 21 40 12 50 18 10 40 15 30

8 30 12 20

6 time (hours) 9 time (hours) 20 4 6 Number of transits Number of transits 10

10 Exo-Earth observation

2 Hot-Jupiter observation 3 0 0 0 0 8 9 10 11 12 13 14 8 9 10 11 12 13 14 mGLUV mGLUV

(a) Hot-jupiters (b) Exo-Earths

Figure 3.5: Time required to reach a given signal-to-noise of 105 for hot-Jupiters and exo-Earths through simple stacking of transits. The shaded region is the range between the best-case atmospheric transmission of 80% and the worst-case of 40%. Including noise sources not accounted for here, i.e. inter/intra-pixel variations and pointing stability, will increase the number of transits required. Future work will introduce and analyse the impact of all noise sources. 54 GLUV Science Cases 3.5 Conclusion

From this preliminary analysis of the GLUV science cases, it is clear to see the a UV survey system would have substantial impact on numerous fields. In particular, the theorised GLUV system would be capable of observing supernovae shocks to determine the progenitor pathways of SN Ia and constrain the progenitor star properties for CCSN. Determining the SN Ia progenitor pathways would significantly impact modern SN Ia cosmology, allowing for an improved standardisation of SN Ia. For the planetary/stellar science cases explored here, GLUV shows promise as being able to provide a unique and valuable dataset. Directed observations of exoplanet transits, could identify atmospheric features, such as the presence of an extended hydrogen envelope and ozone. Further work is needed to fully determine if GLUV would have the sensitivity required for this science case. GLUV will be capable of monitoring M-dwarf flares, as part of directed observations for planetary systems, or through a transient survey. Such observations could provide further insights into the habitability of planets orbiting M-dwarfs. 4 Detecting Gravitational Wave UV Counterparts with GLUV

This chapter is published in Monthly Notices of the Royal Astronomical Society as Ridden-Harper, R., Tucker, B., Sharp, R., Gilbert, J., Petkovic, M., “Capabil- ity of detecting ultraviolet counterparts of gravitational waves with GLUV”,2017, MNRAS, 472, 4521-4531

4.1 Abstract

With the discovery of gravitational waves (GW), attention has turned towards detecting counterparts to these sources. In discussions on counterpart signatures and multi-messenger follow-up strategies to GW detections, ultra-violet (UV) signatures have largely been neglected, due to UV facilities being limited to SWIFT, which

55 56 Detecting Gravitational Wave UV Counterparts with GLUV lacks high-cadence UV survey capabilities. In this paper, we examine the UV signatures from merger models for the major GW sources, highlighting the need for further modelling, while presenting requirements and a design for an effective UV survey telescope. Using u′-band models as an analogue, we find that a UV survey telescope requires a limiting magnitude of m (AB) 24 to fully complement the u′ ≈ aLIGO range and sky localisation. We show that a network of small, balloon-based UV telescopes with a primary mirror diameter of 30 cm could be capable of covering the aLIGO detection distance from 60–100% for BNS events and 40% for ∼ ∼ BHNS events. The sensitivity of UV emission to initial conditions suggests that a UV survey telescope would provide a unique dataset, that can act as an effective diagnostic to discriminate between models.

4.2 Introduction

The historic discovery of the first gravitational waves (GWs) from the coalescence of binary black hole systems (BBH; Abbott et al. 2016b; Abbott et al. 2016c), has drawn attention to a new way of investigating the Universe. The first event +4 +4 GW150914 was inferred to have initial masses of 29 4 M and 36 5 M ,afinal − ⊙ − ⊙ +4 +160 merged mass of 62 4 M ,andoccurredataluminositydistanceof410 180 Mpc − ⊙ − (Abbott et al. 2016b). While the second event GW151226 occurred at a luminosity +180 distance of 440 190 Mpc, it was a lower mass event, with a final merged mass of − +6.1 20.8 1.7 M (Abbott et al. 2016c). The third event GW170104 is the most distant − ⊙ +450 event detected, so far, at a luminosity distance of 880 390 Mpc, with a final merged − +5.9 mass of 50.75.0 M (Abbott et al. 2017a). ⊙ Although a seminal moment in history, the data obtained from the GW events was limited. Without any complimentary information to the GW detection, the discovery is unable to test underpinning aspects to General Relativity, such as the propagation velocity of gravitational waves (Nishizawa 2016). In an effort to detect the GW progenitor systems, a new era in multi-messenger astronomy is forming to complement GW detectors (Coward et al. 2011; Kelley et al. 2013; Chu et al. 2016; Abbott et al. 2016a,d; Evans et al. 2016). Theories regarding counterparts to GW sources have become increasingly important as they inform observation strategies 4.2 Introduction 57 and provide testable predictions. The primary sources of aLIGO GW detections are expected to be binary systems of compact objects such as coalescing binary neutron stars (BNS), coalescing black hole and a neutron star (BHNS), and coalescing binary black holes (BBH).

Currently the UVOT instrument on SWIFT is the only UV telescope that has participated in multi-messenger follow-up observations (Evans et al. 2016). The narrow 17 17 arcmin field of view of SWIFT heavily limits the telescope’s effec- × tiveness for both follow-up and serendipitous detections of GW counterparts. Given the present lack of UV survey capability, it is unsurprising that detailed UV models for GW counterparts have not been developed. However, it leaves any future UV survey missions largely uninformed on expected signals, particularly for fast UV transients that GW events may produce. To begin the discussion on UV transients from GW events, we review relevant models, using the u′-band as a proxy, due to the limitations of UV models, with which we form a case for a capable and versatile UV survey telescope. One reason short wavelengths are poorly modelled is due to unquantified uncertainties produced by lanthanide absorption in ejecta from mergers involving neutron stars (NSs; Kasen et al. 2013, 2015).

The motivation for this analysis of UV counterparts to GWs stems from a UV survey telescope under development, known as GLUV. Described in Sharp et al. (2016), GLUV will be a balloon–based near-UV survey telescope, with the primary objective of high-cadence, early UV observations of supernova. During the ﬁrst hours to days of a supernova, interactions between the ejecta and outer layers or companions are expected to produce UV bright shock emissions (Falk 1978b; Klein &Chevalier1978; Kasen 2010).

The instrument will be low-cost, high-cadence, and able to provide UV observations from both hemispheres. As this telescope will present a unique opportunity in UV astronomy, we analyse the beneﬁt it would provide in the study and character- isation of GW sources.

A number of different UV survey missions are currently being considered. As previously mentioned GLUV will be balloon–based, however other proposed systems, such as ULTRASAT intend to be space based, utilising cubesat technology (Sagiv 58 Detecting Gravitational Wave UV Counterparts with GLUV et al. 2014). Being space based, ULTRASAT is not limited by atmospheric transmission, thus, the bandpass is expected to extend into the far-UV (200–240 nm). The proposed wavelength ranges of GLUV and ULTRASAT are complementary, and if both instruments are successful significant survey capabilities would be available for far-UV and near-UV wavelengths. The review and analysis that follows are complimentary to any future UV survey system. The UV signatures of GW mergers are discussed in Section 4.3.InSection4.4 we present the preliminary GLUV telescope specifications, survey configurations, expected detection rates and consider the use of GLUV as a complementary data source to the upcoming Large Synoptic Survey Telescope (LSST). LSST will feature awidefieldofview,withabaselinecadenceof 3daystomaptheSouthern ∼ sky in optical to near infra-red wavelengths (Ivezic et al. 2008). Since GLUV will be operating at UV wavelengths, a case is made for how its observations would complement those of LSST to characterise GW events. All magnitudes presented are AB.

4.3 UV signatures from mergers

In this section we will explore the luminosity of GW counterparts in the u′-band as a proxy for the UV. Although the models presented in the section feature larger uncertainties for shorter wavelengths, they provide a powerful guide to beneﬁts oﬀered by UV observations.

4.3.1 Binary black hole merger

Optical counterparts to BBH mergers are widely unexpected, however, the possible detection of a short gamma-ray burst (sGRB) by the Fermi Gamma-ray Burst Monitor 0.4s after GW150914 (Connaughton et al. 2016)spurreddiscussionofmulti- messenger counterparts. A comprehensive study of the multi-messenger follow-up survey for GW150914 in Abbott et al. (2016d)foundthesGRBwasunrelatedand found no electromagnetic counterparts. Despite this non-detection, theory developed to support the initial sGRB claim suggests it may be possible to detect an optical and UV counterpart if an accretion disk is present in the BBH system. 4.3 UV signatures from mergers 59

The origin of an accretion disk may stem from either the accretion of dense interstellar dust if the BBH formed through direct collapse as discussed in Belczynski et al. (2016), or via the formation of a “fossil disk” from fall-back material of a failed supernova, (Perna et al. 2014). A fossil disk is expected to be a product of super- Eddington accretion winds in the fall-back accretion disk. Such a process would produce a bright UV transient, reaching M (AB) = 16 for a Wolf-Rayet star and u′ − M (AB) = 17.5forabluesuper-giant(Kashiyama & Quataert 2015). A UV u′ − survey telescope could be able to spot the formation of fallback BHs and potentially the systems in which BBH mergers with a fossil disk could occur.

As described in Murase et al. (2016), a fossil disk would be long-lived and un- detectable until it is disrupted and ionised by the BH-BH merger. The idea of a long-lived fossil disk, has been questioned by Kimura et al. (2017), who show that the inclusion of tidal torque causes the disk to become excited and active, thousands of years before the BBH merger. The fossil disk hypothesis is likewise examined by Ioka et al. (2017)whoarguethataccretionoftheinterstellarmediumwillheat the disk, greatly reducing its lifetime. These two extensions to the fossil disk model highlight the uncertainty in current understand of EM counterparts to BBH mergers.

5 In the case of GW150914, Murase et al. (2016)showafossildiskof10− 1M − ⊙ is heated and ionised during the merger to a thermal emission temperature of 1.1 × 4 10 K. By integrating the black body distribution over the u′-band, it is found that 17% of the accretion disk emission is within the bandwidth. Using the bolometric luminosity, the accretion disk UV luminosity is calculated to be:

1 39 1 3 L 6.12 10 erg s− M R− UV ≈ × BH ,1 .78 d,8 1 3 2 1 1 r κ /0.34 cm g− − , (4.1) × w,10.5 T . / where MBH is the BH mass, Rd is the disk radius, rw is the radius at which the disk wind is no longer a continuous outﬂow, κT is the Thompson scattering opacity and the subscript numbers denote the power of 10 with which the variable is normalised. The luminosity of the accretion disk is linearly dependent on the ﬁnal merged BH mass, so larger BH mergers will have larger luminosities, thus easier to detect.

The emission models for the first two BBH detections are shown in Fig. 4.1.Itis 60 Detecting Gravitational Wave UV Counterparts with GLUV apparent that the disk winds for this model is half a magnitude brighter in B-band than u′-band, but the disk winds are not the only emission source expected. Perna et al. (2016)andMurase et al. (2016)showthatsuper-Eddingtonaccretionwill likely produce a GRB, which is expected to produce a prominent, UV afterglow. The likelihood of detecting disk wind emission from a BBH merger is small. Detection is challenging due to the short event lifetime of 3hours,definedbythe ∼ time required for the outflow to reach the photospheric radius. The short lifetime is then compounded by three other factors: 1) a faint absolute magnitude, preventing the aLIGO range from being covered; 2) poor event localisation, inhibiting rapid follow-up, which is required for events with short lifetimes; and 3) the unknown rate of a BBH systems forming with a fossil accretion disk. As GW localisation improves, follow-up surveys will be able to effectively cover the search area. However, this will not improve the aforementioned points 1 and 3, which will dominate the detectability of such BBH configurations. The model developed in Murase et al. (2016)requires,rapid,half-hourlyfollow- up will be required to detect EM emission from a BBH merger. The disk wind emission model also indicates that large survey telescopes operating in optical wavelengths, such as the LSST, may be better suited for detecting BBH counterparts. However, expanding the fossil disk model to include tidal torques and interstellar medium accretion raises doubt on the fossil disk model, leaving it unknown if BBH mergers produce EM counterparts, thus, observations in all wavelengths are essential to test all possibilities. 4.3 UV signatures from mergers 61

24 u0-band model for GW150914 GW150914 model B-band model for GW150914 u0-band model for GW151226 Apparent magnitude (AB) 22 GW151226 model u0-band model for GW151226 GW170104 model 20 101 102 103 104 Distance (Mpc)

Figure 4.1: Distance-magnitude relation for both BH-BH merger events detected by aLIGO, calculated using the bolometric luminosity in Murase et al. (2016). The thick dashed lines indicate the distance uncertainty in the measurements.

4.3.2 Binary neutron star merger

The merger of a binary NS (BNS) system is predicted to generate a relatively weak GW with an appreciable optical counterpart known as a kilonova (Metzger & Piro 2014). The detection of BNS mergers, both through electromagnetic and gravitational radiation, is expected to answer long standing questions regarding the structure and composition of NSs (Takami et al. 2014). Currently the aLIGO oﬀ-line analysis is expected to detect GW from BNS out to a range of 70Mpc for a pro- ∼ genitor with a component mass distribution of 1.35 0.13M (Abbott et al. 2016e). ± ⊙ The detection range is expected to increase to 200Mpc in the future (Abadie et al. 2010). Despite aLIGO not realising its full detection range, BNS mergers are expected to be detected in the upcoming 3rd aLIGO run. If no BNS events are discovered, anumberofmodels,willbeatoddswithobservation(Abbott et al. 2016e). At peak capability, aLIGO is expected to detect on the order of 40 events per year. With the potential of BNS GW events in the near future, multi-messenger follow-up observations will be crucial to understanding the events. A BNS merger is expected to emit light in three ways over diﬀerent time frames;

1. aneutronjetpoweredprecursor 62 Detecting Gravitational Wave UV Counterparts with GLUV

2. accretion disk outﬂows following the merger

3. sGRB afterglow as different physical processes produce these three sources, detection would provide valuable insight into the merger dynamics. During the merger, it is predicted that at the point of collision jets of mildly relativistic free neutrons could be emitted. Such jets would heat the medium as the neutrons decay, leading to a UV bright precursor (Metzger et al. 2015). The neutron-powered precursor (NPP) is expected have a lifetime of only 1hour,as ∼ seen in Fig. 4.2. The primary source of kilonova emission is produced by disk wind outflows from tidally disrupted material (Kasen et al. 2015). As the disk contains a high fraction of neutrons, it is predicted that the r-process will dominate the ejecta increasing the disk opacity at short wavelengths. Since the kilonova luminosity is linked to the disk mass, it depends on both the NS mass ratio and orbital eccentricity (Bauswein et al. 2013; Gold et al. 2012). During a BNS merger, strong magnetic fields on the order of 1016 Gcaneasily ∼ be produced (Giacomazzo et al. 2015). Interactions between the magnetic fields and plasma may drive the emission of strong post merger electromagnetic signals and possibly even a sGRB. The nature of sGRBs produced by BNS mergers and their UV afterglow will be discussed in Sec. 4.3.4. UV emission from BNS mergers are highly model- and remnant- dependent. Short wavelengths offer modelling challenges as the r-process, which takes place in the neutron rich ejecta, can quickly form lanthanide series elements which strongly absorb short wavelengths Kasen et al. (2013). The r-process can be suppressed by aneutrinoflux,whichisdependentonthemodel,thussotooisthespectra.The strong model variation suggests UV observations may be a useful to distinguish between merger pathways. All models, however, show the same general trend that after 1daythekilo- ∼ nova ejecta becomes dominated by the r-process, suppressing all emission at short wavelengths. Conversely, red wavelengths will benefit from the r-process leading to alongemissionfalltime. Unlike BBH or BHNS mergers, BNS mergers have the potential to produce one 4.3 UV signatures from mergers 63 of four remnants, being: 1) an intermediary high-mass NS (HMNS) that collapses to a BH; 2) a HMNS; 3) a magnetar; and, 4) direct collapse to a rapidly rotating BH. As each of the remnants will produce different neutrino fluxes, the r-process will proceed at different rates, making UV light curves sensitive to the remnant and its pathway, more so than longer wavelengths.

Kasen et al. (2015) found that if the merger produces a HMNS, the lifetime of the remnant strongly inﬂuences the blue emission. For longer lifetimes a HMNS is expected to produce a larger neutrino ﬂux which suppresses the r-process, thus the production of high opacity lanthanide elements is delayed. However, it is also found that a shell of neutron rich ejecta is present around the merger which will act like a lanthanide curtain and heavily reduce UV and blue emission. The variation that is present in the scenarios previously outlined suggest that UV observations would provide powerful diagnostic information on the merger pathway and r-process.

If the remnant produced is a magnetar, strong X-ray and UV emission is expected. After the merger, the strong magnetic ﬁelds of the magnetar will interact with the surrounding medium and drive an X-ray and UV shock emission which is estimated to be 100 times brighter than a kilonova and can last upwards of days ∼ (Yu et al. 2013; Metzger & Piro 2014). Li & Yu (2016)foundthatthespin-down of a magnetar could drive out a wind producing an early peak UV magnitude of M (AB) = 21 a day after the merger. Longer wavelengths have a similar magni- u′ − tude peak, but days after the merger (Li & Yu 2016).

To detect a magnetar event within the full aLIGO range of 200Mpc, a UV survey telescope requires a limiting magnitude of m (AB) 15. The production of u′ ≥ magnetars from BNS mergers would be an excellent candidate for detections in UV, as it should be far brighter in UV than in longer wavelengths.

For less exotic remnants, the peak luminosity is expected to be much lower than that of a magnetar. These remnants are expected to have two main sources of UV emission as previously mentioned – the neutron-powered precursor (NPP) shown in Fig. 4.2 and the disk outﬂows shown in Fig. 4.3.

Further analysis and modelling is required to test the robustness of the NPP since observation angles have yet to be included (Metzger et al. 2015; Fernández & Metzger 2016). The kilonova disk wind model is also known to underestimate ejecta 64 Detecting Gravitational Wave UV Counterparts with GLUV temperatures, therefore luminosities of shorter wavelengths (Kasen et al. 2015). With the aforementioned sources of error in kilonova emission, the models are taken as the optimal scenario. In Fig. 4.3 six scenarios developed in Kasen et al. (2015)areshownalongside the NPP. The disk wind model begins at 2hoursafterthemerger,bywhichtime ∼ the NPP is expected to have faded. Since the kilonova model doesn’t overlap with the NPP, the total early emission from a BNS merger is expected to be larger than the NPP predicts. In the cases of an encompassing shell of neutron matter, the NPP offers a baseline of emission as seen in Fig. 4.3 (d,e). Although the NPP is short lived, it is expected to be a guaranteed source of UV emission that distinguishes a kilonova produced by a BNS merger rather than a BHNS merger. Thus, early UV observations would be critical to identifying the merger system photometrically (Metzger et al. 2015). If a UV survey telescope were to complement the full range of aLIGO (200Mpc), the limiting magnitude must be mu′ (AB) ! 23 to detect the neutron-powered precursor. However, for some disk wind models, such as those shown in Fig. 4.3 (c) and (f), a limiting magnitude of mu′ (AB) ! 23 should be adequate to detect disk outflow winds for most, if not all, viewing angles. Since the UV emission peaks early, a UV survey telescope would be well suited for rapid follow-up soon after a GW trigger to localise and initiate follow-up observations at longer wavelengths. As the light-curves appear unique at UV wavelengths, early UV observations could be crucial in characterising the merger pathway and understanding the composition of NSs. Although all BNS mergers, except those leading to magnetars, will be brighter and long lived at optical wavelengths, the variation and sensitivity of models to UV wavelengths provides an excellent case for discriminating between models and prob- ing elemental abundances. Thus, a UV survey telescope would be able to provide a complementary and perhaps necessary dataset to optical observations. 4.3 UV signatures from mergers 65

16 u-band g-band 15 i-band

Absolute magnitude (AB) 11

10 1 10 100 Hours since merger

Figure 4.2: The ﬁducial neutron-powered precursor, recreated with data provided by Metzger et al. (2015).

18 19 19 19 17 (a) NS tl = 100ms (b) NS tl = 300ms (c) NS tl = 20 20 1 20 16 21 21 21 15 22 22 22 14 23 23 23 13 24 24 24 12 25 25 25 11 26 26 26 10 1 10 100 1 10 100 1 10 100 18 19 19 19 3 5 17 (d) NS tl = 100ms + 10 M ejecta (e) NS tl = 100ms + 10 M ejecta (f) BH ↵=0.8 20 20 20 16 21 21 21 15

Absolute magnitude (AB) 22 22 22 14 23 23 23 13

24 24 24 Apparent magnitude at 200Mpc (AB) 12 25 25 25 11 26 26 26 10 1 10 100 1 10 100 1 10 100 Hours since merger

Figure 4.3: Kilonova u′-band light curves for both the neutron-powered precursor (dotted blue) and the disk outflow winds, for viewing angles ranging from 0 180o (solid blue), encased in a factor of 2 error (grey). Figures (a-c) are models− which form a HMNS that collapses to a BH after a HMNS lifetime tl. Figures (d,e) are figure (a) with a spherical shell of neutron rich ejecta. Finally figure (f) is a direct collapse to a rapid rotating BH. 66 Detecting Gravitational Wave UV Counterparts with GLUV

4.3.3 Black hole–neutron star merger

Another promising candidate for both GW and electromagnetic observations is the merger of a BH and NS system. As there are no known and studied BHNS systems, the detection of GWs or a kilonova associated with such a system would confirm such configurations can occur. The lack of knowledge of BHNS systems gives rise to a rather uncertain rate, however, at full capacity aLIGO is expected to detect BHNS to a distance of 400Mpc and at a rate of 10 per year (Abadie et al. 2010). Similar to the BNS case in Sec. 4.3.2,modelssuggesttheremaybeseveral processes during the merger that drive EM emission for BHNS. The luminosity of a BHNS merger is dependent on a number of parameters, including the mass and spin of the BH and NS, NS equation of state (Kawaguchi et al. 2015), the NS magnetic field strength (Paschalidis et al. 2013; Kiuchi et al. 2015; D’Orazio et al. 2016)and orbital eccentricity (Stephens et al. 2011). Shortly before the merger, precursor emission is expected to be generated via interactions between the NS’s magnetic field and the BH (Paschalidis et al. 2013; D’Orazio et al. 2016; Paschalidis et al. 2015). Close to the merger, a fireball of γ and hard X-rays is expected to be emitted (D’Orazio et al. 2016). Such a fireball and gamma-ray burst, discussed further in Sec. 4.3.4,couldproduceanexcitingtarget for UV survey telescopes. BHNS mergers are also expected to produce kilonova from the disk outflows and radioactive decay. As the NS becomes tidally disrupted an accretion disk will form the outflows of which will drive emission similar to that of the BNS kilonova, due to the r-process synthesis of lanthanides (Surman et al. 2008; Just et al. 2015). It was found in Tanaka et al. (2014) that BHNS kilonova could, under the right conditions, be brighter than a BNS kilonova. Representative light curves of a BHNS merger from Fernández et al. (2017)andKawaguchi et al. (2016) are shown in Fig. 4.4. It can also be seen in Fig. 4.4 that the NSBH kilonova light curve is similar that of the BNS kilonova. Both light curves feature an initial peak soon after the merger, that rapidly falls off 1dayafterthemergerasr-processelementsdominate ∼ the ejecta. However, unlike the BNS merger, the BHNS merger is not expected to produce a NPP, so UV observations could be crucial in early classification of kilonova (Metzger et al. 2015). 4.3 UV signatures from mergers 67

An afterglow is also possible if some of the NS’s magnetic field lines are frozen into the BH. Rotational energy could be extracted from the BH to power a Blandford- Znajek process, this is expected to produce an afterglow, close to the time at which the fireball expands (D’Orazio et al. 2016). As suggested in D’Orazio et al. (2016), if the BH mass is large enough ( 6M ), ! ⊙ it could potentially swallow the NS whole, greatly diminishing or even preventing a kilonova. However, the precursor and afterglow emissions will be unaffected as they are generated by the NSs magnetosphere. Currently, both the fireball and Blandford-Znajek process afterglows are not modelled in the UV. The lack of afterglow and early kilonova emission heavily limits the ability to asses the ability of a UV survey telescope in BHNS follow-up observations. From Fig. 4.4 it can be seen that for a UV survey to completely complement the aLIGO range from the “late time” (> 10 hours) kilonova emission alone, a limiting magnitude of mu′ (AB) ! 24 is required. Further modelling is required for early time UV emissions, as all BHNS kilonova models, start ! 1 day after the merger. At such late times the UV emission is expected to be small as the r-process is expected to produce a high concentration of lanthanides which will block most UV emission, as with the BNS case (Kasen et al. 2013; Tanaka & Hotokezaka 2013). 68 Detecting Gravitational Wave UV Counterparts with GLUV

16 22

15 23

14 24

13 25

12 26

Absolute magnitude (AB) 11 27 Apparent magnitude at 400Mpc (AB) 10 28 1 10 100 Hours since merger

Figure 4.4: u′-band light curves for the two BHNS merger models. In blue the Fernández et al. (2017) fiducial F0 model, with viewing angles, and in green is the H4Q3a75 BHNS merger model produced in Tanaka et al. (2014)andrecreatedin Kawaguchi et al. (2016). These models exclude early emission sources, such as GRB and fireball afterglows.

4.3.4 Gamma-ray bursts

As mentioned in the BNS and BHNS sections, mergers of such objects are expected to produce sGRBs (Tanvir et al. 2013; Tanaka 2016; Lazzati et al. 2017). This idea is supported in Tanvir et al. (2013)whereakilonovawaslinkedwithGRB130603B, however, it is unclear if the event was produced by a BNS or BHNS merger. The GRBs from these mergers will be highly directional and energetic with afterglows that are visible in the UV (Roming et al. 2009).

Lazzati et al. (2017)investigatethenatureofsGRBafterglowsproducedby BNS and BHNS mergers. Currently, simulations are not capable of distinguishing a GRB generated by a BNS merger, or a BHNS merger. However they should be diﬀerent due to variations in mass ratio and cocoon energetics. In Fig. 4.5 we show the u′-band afterglow of a GRB produced from a BNS or BHNS merger at diﬀerent viewing angles based on models in Lazzati et al. (2017).

From the model, it is apparent that a sGRB produced in a merger involving a NS will be bright for many days, even at a distance of 400 Mpc for viewing angles close to the emission axis. Despite the high on-axis luminosity of these events, the magnitude of the afterglow decreases substantially with increasing viewing angles. 4.3 UV signatures from mergers 69

26 12 24 14 14 22 16 16 20 18 18 18 20 20 16 22 22 14 24

Absolute magnitude (AB) 24 12 26 26 Apparent magnitude at 200Mpc (AB) Apparent magnitude at 400Mpc (AB) 10 28 1 0 1 2 3 4 5 10 10 10 10 10 10 10 Hours since merger

Figure 4.5: u′-band model for a sGRB afterglow produced from a BNS or BHNS merger event, as described in Lazzati et al. (2017). A range of diﬀerent viewing angles from 0o to 90o is shown in steps of 5o. Although the sGRB model is capable of producing UV data, u′-band data is shown for consistency with the other models.

The strong angle dependence of emission, coupled with the preferential sensitivity of gravitational wave detectors to unaligned systems, joint detections may be unlikely (Lazzati et al. 2017).

Despite the potential rarity of sGRBs produce by mergers involving NSs, the luminosity of these events make them excellent targets for UV survey missions. For atelescopefeaturingalimitingmagnitudeof22,itwouldbeabletodetectan afterglow at a viewing angle of up to 35o at 200 Mpc and 30o at 400 Mpc for some days after the merger.

SWIFT has been successful in detecting UV light curves from GRBs. Although the UV imager, UVOT, has a small ﬁeld of view, SWIFT utilises the wide ﬁeld BAT for detection and triggering follow-up observations. This method has lead to a catalogue of UV light curves for GRBs (Roming et al. 2009), however it is unlikely that UVOT could detect afterglows when there is a BAT non-detection. In the case of follow-up observations of an aLIGO GW trigger, it becomes crucial to have UV survey capabilities to detect potential sGRB afterglows, due to the likelihood the SWIFT BAT would be inactive or pointing elsewhere. 70 Detecting Gravitational Wave UV Counterparts with GLUV 4.4 GLUV: a UV Survey Telescope

The authors are developing a balloon-based high cadence UV survey to address a key question in supernova physics. During the early stages of a supernova, shock interactions occur which are UV bright and provide crucial information on the progenitor system (Kasen 2010). Operating at UV wavelengths, GLUV is expected to produce routine discoveries of type Ia supernova shock interaction. As the future of GW science develops, it has become apparent that a UV survey telescope could provide novel data in understanding GW events and their progenitors, using the existing design parameters. Following a brief overview of the instrument, three fiducial observation strategies are presented, along with preliminary rate calculations. By utilizing advancements in long-duration high altitude balloons, the GLUV project aims to fly a network of UV telescopes at altitudes of 20–30 km for months at a time (Sharp et al. 2016). The telescopes will be modest, with a diameter of 30 cm, recoverable, steerable and feature a 7 deg2 field of view. Although the filter band-pass is yet to be determined, the initial system design employs UV coated CCDs setting the lower limit on the wavelength range of 250 nm. Further technical ∼ details are presented in Sharp et al. (2016). Initial evaluations of the system’s stability and limiting magnitude appear promising, being 5arcsec,m (AB) 22 for a signal-to-noise of 5 after a 10 minute ex- ∼ uv ≈ posure, respectively. The bulk characteristics of the system sensitivity are outlined in the following subsections.

4.4.1 Sky background

The faint limiting magnitude is due to the low UV sky background at the expected flight altitude. At expected flight altitude, the effects of scattered light are reduced, resulting in sky glow being the prime contributor to the sky background. The sky background at the intended flight altitudes is currently ambiguous as no current measurements exist. Currently a pathfinder mission is in development to measure the sky background at the intended flight altitudes, the success of which will provide a firm description of the sky background GLUV will experience. 4.4 GLUV: a UV Survey Telescope 71

For a preliminary value of sky background we refer to Waller & Stecher (1998)

2 which identiﬁes the sky-glow from Oxygen I as m (AB) 26 mag arcsec− . Waller uv ≈ &Stecher(1998)conductedtheirmeasurementsfromSpaceShuttleColumbia at a higher altitude than GLUV will ﬂy and at a shorter wavelength of 250 nm, we therefore anticipate GLUV will experience a brighter sky background. For the preliminary

2 calculations, a sky background 2 magnitudes brighter, m (AB) 24 mag arcsec− , uv ≈ is adopted.

4.4.2 Instrument throughput

Sharp et al. (2016) outlines the preliminary design of GLUV. The system will follow afiveelementcatadioptricdesigncontainingthreecorrectivelensesalongsidethe primary and secondary mirrors. The lens elements are expected to have a throughput of 99%, while the mirror reflectance is taken as 95%. As a filter is not currently defined, the filter profile is unknown, so a peak throughput of 80% is implemented. The detector will be a UV coated CCD with a QE 55%. The resulting instrument ≈ throughput is taken to be 37%. ∼ Since GLUV is expected to fly close to the ozone layer, atmospheric transmission must be considered. As the flight altitude and therefore atmospheric transmission is yet to be determined, we perform the preliminary calculations assuming the atmospheric transmission to be 40% at 300 nm. The wavelength limitation by atmospheric transmission is a necessary trade offto ensure that the telescope has maximum flight time. Future work will identify the true value of atmospheric transmission GLUV will experience. With the inclusion of atmospheric transmission we expect GLUV, with a30cmdiameterprimarymirror,toreachadepthofm (AB) 22 for 10 minute uv ≈ exposures.

4.4.3 Fiducial Survey Strategies & Detection Rates

Before considering ﬁducial survey strategies, it is worth considering how the capabilities of GLUV could complement and compare to upcoming survey telescopes. Two such telescopes are the LSST, an optical survey telescope, and ULTRASAT, a space based UV survey telescope. 72 Detecting Gravitational Wave UV Counterparts with GLUV

The LSST (LSST Science Collaboration et al. 2009), will feature a 9.6deg2 field of view and work toward a limiting magnitude of 24.5forallbands,witha ∼ 3 day cadence. Although LSST is working towards a deep magnitude, the opacity ∼ of the atmosphere limits the capabilities in the u′-band, with a mean filter efficiency of 20%. When taking the atmospheric effects into account, the LSST u′-band ∼ limiting magnitude is close to what GLUV is expected to achieve. Thus, GLUV would provide an excellent complementary dataset to LSST, due to the sensitivity gain from being at altitude, coupled with a shorter operational wavelength. ULTRASAT plans to utilise cubesat technology to develop a space-based UV survey telescope. As described in Sagiv et al. (2014), ULTRASAT is expected to feature a large 800 deg2 field of view, with a limiting magnitude of 21 with 12 ∼ minute exposures, and have a bandpass of 200-240 nm. In comparison to GLUV, ULTRASAT will have a significantly higher survey rate, however, GLUV is expected to be a magnitude more sensitive, and operate in the near-UV wavelengths ( 250– ∼ 290 nm). Therefore, these two systems would provide comprehensive, high cadence coverage of the ultraviolet. The detection capabilities of GLUV come as a trade-offbetween cadence and survey area. For the NPP, a cadence of 30 minutes is required while the kilonova ∼ disk winds are longer lived and require daily-cadence. Daily-cadence observations ∼ limit the maximum survey area to the area one GLUV telescope can observe in a given night. For one GLUV telescope with an on-sky time of 8 hours, the maximum survey area can be calculated as follows;

8 SA = FoV (4.2) τ where SA is the survey area, τ is the exposure time in hours, plus 10% overhead and FoV is the telescope field of view. From signal-to-noise calculations it appears the system will achieve m (AB) 22 and a signal-to-noise of 5, for a 10 minute expo- uv ≈ sure. Thus, the maximum survey area for daily-cadence at a depth of m (AB) 22 uv ≈ would be 300 deg2. ∼ In the event that a constellation of >10 GLUVs is flown at any given time, a trade-offcan be made between increasing the cadence and increasing survey area. Increasing the cadence, by staggering the constellation longitudinally, would not 4.4 GLUV: a UV Survey Telescope 73 only provide higher quality light curves, but may also suite the requirements for NPP detection. To develop a realistic detection rate, the event rates are weighted by an observation probability, constructed as follows. The event rates used in these rate calculations are volumetric, so by using the absolute magnitudes shown in Tab. 4.1 the rates can be converted into a rate per telescope pointing (7 deg2) (see Fig. 4.7 for rate comparison). With a rate per pointing, the probability that an event location,

EL,occurswithinthesurveyarea,SA,isgivenby;

SA P (E SA)= (4.3) L ∩ Area of sky

We must also account for the probability that an event will be observed in the survey area for a given cadence, C,andeventlifetime,Elt,

Elt if Elt < 1, P (E C)= C C (4.4) lt ∩ ⎧ ⎪1ifElt 1. ⎨ C ≥ ⎩⎪ As each of the conditions stated are independent, the total probability of detecting an event is given by;

P (Detection)=P (E SA).P (E C) (4.5) L ∩ lt ∩

With the probability of detection established, the detection rate can be calculated for a number of ﬁducial survey strategies.

4.4.4 Survey Strategies

A GW/kilonova survey features two aspects: a follow-up campaign for GW observatories and a “blind” transient survey. We will consider both aspects, and their requirements. The survey rate may also be comparable between GLUV and LSST. Although GLUV will be a small instrument requiring 15 minute exposures to reach m (AB) 22 for a signal-to-noise of 5, provided the success of initial flights, it is uv ≈ 74 Detecting Gravitational Wave UV Counterparts with GLUV expected that a constellation of GLUVs will be flown. At this time it seems likely that > 10 GLUVs could fly during a campaign. Individually, the survey area of a GLUV will be far less than LSST, however, if 10 GLUVs are flown they will achieve

2 1 asurveyrateof0.080 deg s− ,whichis 12% that of the expected LSST survey ∼ rate. The potential of a large collective survey rate will assist GLUV in providing complementary near-UV observations to LSST.

The follow-up survey requires rapid-response to any GW triggers. From the models presented, it appears that a UV survey telescope with a limiting magnitude m (AB) 22 will be capable of covering 60 100% and 40% of the aLIGO de- u ≈ − ∼ tection distance for BNS and BHNS events, respectively. The challenging aspect is likely to come from coordinating observations between a constellation of telescopes. If a GW event is localized to a part of the sky viewable during the night, then a constellation of GLUVs may be adequate for covering the survey area. For localisation on the order of 400 deg2, a constellation of 10 GLUVs surveying diﬀerent ∼ patches would cover the survey area to a depth of m (AB) 22 in 1.4hours. uv ≤ ∼ The “blind” kilonova survey requires a trade-oﬀbetween the survey area and cadence. As merger events that produce kilonova are rare, a large survey area is required. Conversely, a primary case for UV observations of kilonova is to detect the NPP, which occurs on short time-scales, requiring high cadence. To address these two aspects we calculate the number of detections per 6 month campaign with a constellation of 10 GLUVs.

In the ﬁrst case the GLUV constellation operates independently, surveying different areas. With each GLUV surveying to a depth of m (AB) 22, 3000 deg2 uv ≈ ∼ can be surveyed in a given night. As seen in Fig. 4.6,withtheopenpoints,itappears likely that such a survey would detect the luminous UV BNS merger pathways such as NS HMNS, NS Magnetar on a regular basis. The NS BH pathway also → → → has a relatively high detection rate which will lead to detections or constrain rates following multiple campaigns.

For the second case, the GLUV constellation monitors 5 independent survey areas with a 12-hour cadence. Again, each GLUV is surveying to a depth of 22. However, they are only covering 1500 deg2 which is observed every 12 hours. ∼ From looking at the closed points in Fig. 4.6,itisapparentthatincreasingthe 4.4 GLUV: a UV Survey Telescope 75

Table 4.1: Peak absolute AB magnitude of GW counterparts, in a range of wavelengths. Cells are left blank if no model data is currently available.

NPP BNS (HMNS BH) BNS (HMNS) BNS (Magnetar) BNS (BH) BHNS BBH (60M ) → ⊙ GALEX FUV 9 GALEX NUV − − − − − − 11−.5 − − − − − − − u′-band 14.7 13.4 14 15.5 21.2 14 14.8 12 g-band − 14 −12.5 →−15 −15.6 − 21 −13.6 −14.6 −12 r-band −13.8 −12.6 →−15.4 −15.8 −20.9 −13.8 −15.2 −12 J-band − −14.2 →−15.6 −16.8 − −15.4 − 16 −11.4 −−→− − − − − − cadence by a factor of 2 does not compensate for loss in equivalent survey area. From preliminary calculations, it appears that survey area is favoured over cadence. A limit of daily-cadence is not required, although, since most events have a lifetime 1day,acadencelessthanthatwouldlikelyresultinsingledetectionsof ∼ events. It is also apparent that NPP are unlikely to be observed through a ‘blind survey’ due to their short lifetime. It may be that the best strategy for NPPs is low latency follow-ups of triggers, rather than relying on serendipitous detections. 76 Detecting Gravitational Wave UV Counterparts with GLUV

aLIGO 2010 rate compendium

NPP BNS HMNS BH ! ! BNS HMNS Fong et al. GRB ! BNS Magnetar ! BNS BH ! BHNS BH !

Jin et al. kilonova

6 5 4 3 2 1 0 1 2 3 4 10 10 10 10 10 10 10 10 10 10 10 Number of events per 6 month campaign

Figure 4.6: Expected number of GW event detections for a constellation of 10 GLUVs over a 6 month campaign, with limiting magnitude muv(AB) = 22 and two observing strategies. The solid points are for a survey area of 1500 deg2 with 12 hourly cadence, while the hollow points are for a 3000 deg2 with daily-cadence. For these calculations, the full BNS rate is used for each BNS subclass, as it is currently unknown which pathways permitted and at what rates. The three rates used are, from kilonova (Jin et al. 2015), sGRB (Fong et al. 2015)andtheaLIGO2010rate compendium (Abadie et al. 2010). See the appendix for a breakdown of all rates examined in Abbott et al. (2016e).

4.5 Conclusion

We have examined the utility of a UV survey telescope for studying gravitational wave sources. Surprisingly, no models of kilonova emission have been developed for short wavelengths including UV. The lack of model data is highlighted in Tab. 4.1, where entries for NUV and FUV are blank and u′-band features unquantiﬁed uncertainty.

Although models predict that UV emission will be fainter, than optical for kilonova, it may present valuable diagnostic information to discriminate between models. The sensitivity of UV light curves with respect to models and viewing angles, will be 4.5 Conclusion 77 greater than that of u′-band, thus, it will provide useful data to test and constrain models, alone and in conjunction with optical wavelengths. Rapid, high energy processes, which are expected to occur in most compact object mergers, will be UV bright. An example is the presence of a neutron-power precursor in a BNS merger, which may be critical in distinguishing kilonova from BNS and BHNS events.

With the limited models available, it appears that a small UV survey telescope could cover a large portion of the predicted aLIGO detection range. Through weighted rate calculations, we ﬁnd that the detection of BNS mergers may be likely for a constellation of >10 GLUVs. Although the kilonova associated with BHNS mergers are expected to be brighter than the BNS counterparts, the low expected rate makes detecting BHNS mergers unlikely.

The fascinating science cases available to UV survey telescopes has prompted the development of such systems. In this paper we have focused on GLUV, which is being developed by the authors with plans to launch in 2019 to conduct a high cadence near-UV survey. Other systems, such as ULTRASAT, aim to provide high cadence UV survey with space based telescopes. If successful the two aforementioned missions would provide a high cadence coverage of a large portion of the UV, opening up new possibilities for studying energetic phenomena and explore short time–scale events.

We have also compared the eﬀectiveness of GLUV as a complementary near-UV dataset to LSST. Although LSST is expected to feature a faint limiting magnitude, atmospheric transmission renders the limiting magnitude of LSST u-band to be close to that of GLUV. Therefore, there is an ideal opportunity for GLUV to provide a complementary dataset at short wavelengths. The survey rate of LSST is far larger than a single GLUV, however, it is expected that in the future, on a 5yeartime– ∼ scale, a constellation >10 GLUVs will be ﬂying during any given campaign. As the constellation grows, the collective GLUV survey rate may become comparable to that of LSST, while providing a unique, complementary wavelength.

Overall this analysis of UV emission from GW sources has shown a lack of modelling. Without precise models for emissions in NUV wavelengths, we are unable to eﬀectively constrain the utility of a UV survey telescope for detecting GW events. Within the coming years, new UV survey system will be operational, that will open 78 Detecting Gravitational Wave UV Counterparts with GLUV up the possibility of observing and testing models of high energy processes.

Acknowledgements

We thank Daniel Kasen for discussions and comments on models cited for both BNS and BHNS mergers, and Davide Lazzati for information on the sGRB afterglow models that were cited. This research was conducted by the Australian Research Council Centre of Excellence for All-sky Astrophysics (CAASTRO), through project number CE110001020 and supported by an Australian Government Research Training Program (RTP) Scholarship.

4.6 Appendix: Detection rates

Here we present the detection rates for all models examined in Abbott et al. (2016e). Figures 6 and 7 from Abbott et al. (2016e) have been recreated in Fig. 4.7,topro- vide a direct comparison between the models.

The detections rates for each of the survey conﬁgurations are shown in Fig. 4.8 and Fig. 4.9.ThespreadbetweenthemodelsisencapsulatedintheaLIGO2010 compendium rate. 4.6 Appendix: Detection rates 79

aLIGO 2010 rate compendium aLIGO 2010 rate compendium

Kim et al. pulsar Fong et al. GRB Fong et al. GRB Coward et al. GRB Siellez et al. GRB

Coward et al. GRB Petrillo et al. GRB

Petrillo et al. GRB Jin et al. kilonova

Jin et al. kilonova Vangioni et al. r-process Vangioni et al. r-process de Mink & Belczynski pop syn de Mink & Belczynski pop syn

Dominik et al. pop syn Dominik et al. pop syn

0 1 2 3 4 2 1 0 1 2 3 4 10 10 10 10 10 10 10 10 10 10 10 10 3 1 3 1 BNS Rate (Gpc yr ) BHNS Rate (Gpc yr )

(a) Rates for black hole-neutron star (BHNS) (b) Rates for black hole-neutron star (BHNS) mergers. mergers.

Figure 4.7: The merger rates are obtained from a variety of models including: population synthesis (de Mink & Belczynski 2015; Dominik et al. 2015), elemental abundance (Vangioni et al. 2016), kilonova rate (Jin et al. 2015), sGRB rate (Petrillo et al. 2013; Coward et al. 2012; Siellez et al. 2014; Fong et al. 2015), pulsar rate (Kim et al. 2015)andtheaLIGO2010ratecompendium(Abadie et al. 2010). 80 Detecting Gravitational Wave UV Counterparts with GLUV

aLIGO 2010 rate compendium

Kim et al. pulsar

Fong et al. GRB

Siellez et al. GRB NPP BNS HMNS BH Coward et al. GRB ! ! BNS HMNS ! BNS Magnetar ! Petrillo et al. GRB BNS BH ! BHNS BH ! Jin et al. kilonova

Vangioni et al. r-process

de Mink & Belczynski pop syn

Dominik et al. pop syn

5 4 3 2 1 0 1 2 3 4 10 10 10 10 10 10 10 10 10 10 Number of events per 6 month campaign

Figure 4.8: Expected number of GW event detections for a constellation of 10 GLUVs over a 6 month campaign, with limiting magnitude muv(AB) = 22, survey area of 3000 deg2, and daily-cadence. The full rate and rate errors for BNS mergers are used for the BNS subclasses, as it is currently unknown which pathways are permitted and at what rates. 4.6 Appendix: Detection rates 81

aLIGO 2010 rate compendium

Kim et al. pulsar

Fong et al. GRB

Siellez et al. GRB NPP BNS HMNS BH Coward et al. GRB ! ! BNS HMNS ! BNS Magnetar ! Petrillo et al. GRB BNS BH ! BHNS BH ! Jin et al. kilonova

Vangioni et al. r-process

de Mink & Belczynski pop syn

Dominik et al. pop syn

5 4 3 2 1 0 1 2 3 4 10 10 10 10 10 10 10 10 10 10 Number of events per 6 month campaign

Figure 4.9: Expected number of GW event detections for a constellation of 10 GLUVs over a 6 month campaign, with limiting magnitude muv(AB) = 22, survey area of 1500 deg2, and 12 hourly cadence. The full rate and rate errors for BNS mergers are used for the BNS subclasses, as it is currently unknown which pathways occur and at what rates. 82 Detecting Gravitational Wave UV Counterparts with GLUV 5 Kepler/K2: Background Survey

This chapter is submitted to the Monthly Notices of the Royal Astronomical Society Ridden-Harper, R., Tucker, B. E., Gully-Santiago, M., Barensten, G., Rest, A., Garnavich, P., Shaya, E. J., “Kepler/K2: Background Survey”, 2019, MNRAS, submitted

5.1 Abstract

The K2 mission of the Kepler Space Telescope oﬀers a unique possibility to examine sources of both Galactic and Extra-galactic origin with high-cadence photometry. Alongside the multitude of supernovae and quasars detected within targeted galaxies, it is likely that Kepler has serendipitously observed many transients throughout K2.Sucheventswilllikelyhaveoccurredinbackgroundpixels,coincidentally surrounding science targets. Analysing the background pixels presents the possibility to conduct a high-cadence survey with areas of a few square degrees per

83 84 Kepler/K2: Background Survey campaign. We demonstrate the capacity to independently recover key K2 transients such as KSN 2015K and SN 2018oh. With this survey, we expect to detect numerous transients and determine the ﬁrst comprehensive rates for transients with lifetimes 1day. ≤

5.2 Introduction

Launched in 2009, the Kepler Space Telescope (Kepler)allowedforhighprecision and high cadence stellar photometry (Basri et al. 2005a), with the goal of detecting thousands of exoplanets (Batalha 2014), until the failure of two reaction wheels led to a new mission profile – K2.TheK2 mission extended Kepler observations along the ecliptic, presenting the opportunity to apply high precision and cadence photometry to a multitude of galactic and extragalactic sources (Howell et al. 2014). The reduced telescope stability present in K2 affords it a lower photometric precision than the original Kepler mission. Despite the drop in photometric precision, K2 campaigns C01 to C19 were successful in obtaining unparalleled photometric data that lead to among other things, the detection of new exoplanets (e.g., Pearson et al. 2018), numerous supernovae (SN) (e.g., Garnavich et al. 2016; Olling et al. 2015; Rest et al. 2018; Dimitriadis et al. 2019a; Shappee et al. 2019; Li et al. 2019), and variability in quasars (Aranzana et al. 2018). The high cadence K2 observations make it possible to study short-duration struc- tures in supernovae light curves and detect short-duration transients. Analysis of K2 data has shown potential evidence for SN Ia shock interaction, as described in Kasen (2010), in SN 2018oh (Dimitriadis et al. 2019a), however, not in 3 other SN Ia (Olling et al. 2015). K2 has also shown evidence of shock breakouts in core-collapse supernovae (CCSN) (Garnavich et al. 2016), and lead to the discovery of KSN 2015K, a rapid transient that is thought to be powered by stellar ejecta producing a shock in the circumstellar medium (Rest et al. 2018). Short cadence observations were crucial to these results and offer the possibility to analyse a time domain that has previously been technically infeasible. All known K2 transients were detected through targeted observations of pre- selected galaxy targets, however, this may not be all the transient events detected 5.2 Introduction 85 with Kepler. As seen in Barclay et al. (2012)andBrown et al. (2015)itispossible that Kepler has serendipitously detected numerous events in background or sky pixels, such as super-outbursts of dwarf novae. Such background events will likely be faint and will require an analysis that will be heavily influenced by detector noise and telescope stability.

The high cadence observations of Kepler,presentauniquepossibilitytoprobe anewparameterspaceunavailabletoprevioustransientsurveys.Currenttransient surveys cover larger areas than Kepler,however,theyhavelongercadences.The Pan-STARRS Medium Deep Survey had an nominal cadence of 3 days in any given ﬁlter to a depth of 23.3mag(Rest et al. 2014; Chambers et al. 2016); ZTF ∼ surveys the the sky North of δ = 31o every 3 nights to 20.4mag,throughthe − ∼ public survey (Bellm et al. 2019); ASAS-SN covers the entire sky to 17 mag, at a ∼ nominal cadence of a few days (Kochanek et al. 2017); and ATLAS covers the sky North of δ = 30 every two nights to 20.2mag(Tonry et al. 2018). Previous − ∼ surveys, such as the SNLS, ESSENCE, and SDSS-II also had a cadence of a few days, before taking into consideration bad weather, so would typically miss transients that evolve on timescales of days (Sullivan et al. 2011; Kessler et al. 2009; Miknaitis ∼ et al. 2007). Although these surveys sample volumes much larger, the comparatively high cadence observations of Kepler/K2 mean its legacy data can provide a unique wide-ﬁeld transient survey.

Aworldwidepushisbeingmadetoextendsurveysintohighcadencestoex- plore a new time domain for transients. Andreoni et al. (2020)presentonesuch eﬀort, known as Deeper, Wider, Faster (DWF), which coordinates multi-messenger observations from 40+ telescopes for short 1hourcampaigns,withacadence ∼ of 1.17 minutes. From this unique data set, an upper-bound on extragalactic fast

2 1 transients was found to be ReF T < 1.625 deg− d− . Although Kepler has a longer cadence of 30 minutes, a ﬁeld is examined for 74 days. Similarly, the TESS mission ∼ is an ongoing space-based mission, with 30-minute cadence and a wider ﬁeld, albeit alowersensitivitythanKepler/K2 (Ricker et al. 2015). Sharp et al. (2016)and Ridden-Harper et al. (2017)presentafuturehighaltitudeballoon-basedtelescope system that aims to explore the time domain of days at ultraviolet wavelengths. 86 Kepler/K2: Background Survey

Marshall et al. (2017)presentsthecaseforhighcadenceobservations,highlighting the unique short duration events expected to be seen at Kepler-like cadences, such as GRB afterglows, other relativistic events, and supernovae shocks. As the time domain at the timescale of hours is relatively unexplored for transients, there is potential to discover exotic new transients and provide well-defined rates for current and upcoming high cadence transient surveys. Although Kepler/K2 data can probe this parameter space, the classification of transients will be particularly challenging, since only observation in the Kepler filter will be available for most candidates. Although challenging, examining every pixel in K2 data presents an opportunity to conduct a unique, high cadence survey over an appreciable volume. With this volume, Kepler can be used to examine expected rates of exotic short duration events, such as binary neutron star mergers, similar to the analysis Scolnic et al. (2018a)conductedonexistingsurveydata. In this paper, we present the details of the “K2 :BackgroundSurvey(K2 :BS)” along with examples of positive detections. A companion paper Ridden-Harper et al. (2019) presents the analysis of KSN-BS:C11a, the first transient discovered in K2 :BS. The analysis method is presented in 5.4,followedbythesurveychar- § acteristics in 5.5 and example detections of known objects in 5.6.In 5.7 we § § § present the extragalactic volume surveyed and the expected detection rates for an assortment of transients in K2.

5.3 Kepler/K2 data

The Kepler/K2 data provides high cadence photometry on a wide range of targets. In each of the 19 campaigns between 50–100 targets were observed in short cadence mode with frames every 1minute,whilebetween10,000–20,000targetswere ∼ observed in long cadence mode, with frames every 30 minutes1.Inthispaper,we will focus on data from the long cadence mode, however, this analysis technique is also applicable to short cadence data. Observation campaigns nominally lasted 80 days, with some campaigns shortened, due to technical diﬃculties. ∼ 2 1 Kepler had a 116 deg ﬁeld-of-view (FOV), with a plate scale of 4” pixel− .Due

1https://keplerscience.arc.nasa.gov/data-products.html 5.3 Kepler/K2 data 87 to memory constraints, the entire Kepler FOV could not be telemetered so instead, science targets were pre-selected each observing campaign and allocated a pixel mask that extended several pixels around the science target, both for background subtraction and to account for telescope drift. These pixel masks are known as target apertures in the downloaded data cube, the Target Pixel Files (TPFs), and all targets have unique identiﬁers, known as the EPIC number for K2 (Huber et al. 2016)andKICnumberforKepler (Brown et al. 2011). As the TPFs record both spatial and temporal information they are the data we analyse in this project.

AkeydifferencebetweenK2 data and that of the original Kepler mission is the poor telescope stability. Following the failure of 2 reaction wheels, leaving Kepler with only 2 functional reaction wheels, the K2 mission was developed, using solar pressure on the solar array, in conjunction with the two remaining reaction wheels (Howell et al. 2014). This configuration was subject to drift up to 1pixel(4”) ∼ every 6 hours, so the pointing was corrected by thruster firings every 6 hours, and are recorded as quality flags in the TPF. The drift is recorded in the TPF through two displacement parameters labelled POS CORR, which give the local image motion, calculated from fitting motion polynomials to the centroids of bright stars in each detector channel (DAWG (Data Analysis Working Group) 2012). The POS CORR values give a valuable reference to telescope stability throughout campaigns, with the total image displacement given by D2 =POSCORR12 +POS CORR22.

All K2 fields are pointed along the ecliptic, enabling a range of galactic and extra-galactic fields to be observed. A consequence of pointing along the ecliptic is that a large number of asteroids cross through the K2 fields and contaminate data.

Since Kepler only had a single broadband ﬁlter all magnitudes presented will be in Kepler magnitudes, or Kp. We use the AB magnitude system zeropoint determined in Garnavich et al. (2016),

K = 2.5log(C) + 25.47, (5.1) p − where C is the counts. 88 Kepler/K2: Background Survey 5.4 Methods

The unique dataset from K2 presents a number of challenges for event detection. Spacecraft drift, throughout K2 data, presents the largest challenge, requiring spe- cial treatment. Conventional methods such as image subtraction proved to be in- effective with the K2 data. The up to 4” motion in K2 images results in poor ∼ subtractions between images using a template image and numerous false candidates. To counteract this we trialled subtracting images with similar displacements, while it achieved cleaner subtractions, it could not work over all images and introduced temporal biases in event detection. As these subtraction methods failed to be generally applicable, we developed new methods for event detection which still contain elements of conventional image subtraction. Two main methods were developed from this principle to identify events of different duration. Short event (< 10 days) detection, operates on the principle that the peak brightness of an event will be unique and noticeably larger than the scatter present in a light curve. For long events (>A10 days), the short detection method fails as the event strongly influences the scatter present in the light curve, thus more care is required. For this method we employ heavy smoothing of the data identify specific reference times of the data, which proves successful at detecting events like type Ia supernovae, where the short event detection method fails. The following section will describe the main data manipulation methods, followed by a detailed description of the detection methods. The analysis presented here is applicable to Kepler, K2,andTESS data.

5.4.1 Science target mask

The TPF is separated into science targets or field stars, and background pixels. This distinction assists with both event detection methods as discussed in 5.4.4 § and event sorting. The spatial extent of the science targets is identified through an iterative process of taking sigma cuts on pixel brightness. The mask is created from a median frame. The median frame is generated by averaging a set of successive frames that have a total displacement less than 0.2 pixels from nominal telescope pointing. A sigma cut is performed on the median frame, 5.4 Methods 89 where all pixels that are brighter than median + σ of all pixels in the median frame are added to the science target mask. The previous sigma cut is repeated with the newly defined science target mask applied to the median frame, all pixels identified in this sigma cut are added to the science target mask. To prevent transients from being included in the science target mask, we create two masks, one at the start and another at the end of the campaign. All science target mask pixels that appear in both the start and end masks are included in afinalsciencetargetmask.TransientsthatcanbeidentifiedbyK2 :BS must be shorter than the K2 campaign duration ( 80 days), so this method ensures that ∼ detectable events aren’t included in the science mask.

The large (4′′)pixelsofKepler make is possible that multiple sources blend together into a single mask. In an eﬀort to separate these possible science targets, we separate the science target mask into multiple masks, through a watershed algorithm (Barnes et al. 2014). All of the science targets are checked against NASA/IPAC Extra-galactic Database (NED)2 and the Simbad (Wenger et al. 2000)database,to identify the object within each science target.

5.4.2 Telescope drift correction

During K2 telescope drift introduced significant instrumental artefacts to the data. During campaigns, Kepler drifts to a maximum of 1pixelfromnominaltelescope ∼ pointing, over 6hours.Forbrighttargets,thiseffectisoftennegligibleandcan ∼ be offset with the motion correction tools present in Pyke (Still & Barclay 2012; Vin´ıcius et al. 2017), Lightkurve (Lightkurve Collaboration et al. 2018), K2SFF (Vanderburg & Johnson 2014), and EVEREST (Luger et al. 2016). The techniques used with K2SFF and EVEREST were developed to primarily correct stellar light curves, that don’t feature strong variability. As a result these techniques fail to correct the background motion, without removing the transient signal. The Lightkurve PRF photometry tool can correct for telescope motion, while preserving transient signal, however, PRF photometry currently requires tailoring

2The NASA/IPAC Extra-galactic Database (NED) is operated by the Jet Propulsion Labora- tory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. 90 Kepler/K2: Background Survey priors, such as position and flux, for each object in the TPF, and has a long com- putation time. These issues make PRF photometry infeasible for this analysis. As no existing detrending methods are suitable for a transient search, we develop a method to fit and remove motion induced noise for each pixel. Telescope drift generates discontinuous light curves for many background pixels. During a campaign the system geometry changes and alters the preferred roll direction, this causes bright sources to periodically spill over into neighbouring pixels. The impact this has on a light curve can be seen with the blue points in Fig. 5.1. These motion induced signals increase the detection threshold and can masquerade as real transient signals, so it is critical that they are corrected. The data reduction we developed relies on accurate quality flags for the thruster resets. These thruster resets are identified through the quality flag 1048576. Util- ising the thruster resets, we employ a data correction method that analyses each segment between thruster resets. The procedure acts on every pixel individually, altering the data with the following steps:

1. All images/frames that have < 0.3pixeldisplacementinacampaignareiden- tiﬁed.3

2. A1Dsplineisfittedtothefluxvaluesoftheimagewiththeleastdisplacement between thruster resets of the previously identified frames for each pixel.

3. The 1D splines are subtracted from each of the pixel light curves for the entire campaign, leaving residual ﬂux that is largely the product of telescope motion.

4. The residual light curve is broken into segments deﬁned by thruster resets.

5. To avoid real transients from being removed in the motion correction, we perform a 2σ sigma clip on the data.

6. For each residual light curve segment, we ﬁt a cubic polynomial through linear regression with scikit-learn (Varoquaux et al. 2015).

7. The cubic polynomial ﬁts are subtracted from the residual ﬂux calculated in step iii.

3This choice is made based on data quality and the number of observations available at high precision. 5.4 Methods 91

1.0 130 Raw data Cubic ﬁt Linear spline 0120.8 Thruster reset Corrected data 110 0.6 100

Counts 0.904

80 0.2 70

0.600 0.0 5.5 06.20 6.5 0.47.0 0 0.6 20 400.8 60 1.0 Time (days)

Figure 5.1: An example of the motion correction procedure on a pixel containing SN 2018ajj (Smith et al. 2018). Left: The components of the fitting procedure are shown: the thruster resets (red dashed line) divide the light curve into segments; a linear spline (orange line) is fit to the most stable points in each light curve segment; acubicpolynomials(bluelines)arefittotherawdata(blackpoints).Right: The full light curve is shown for SN 2018ajj, with raw data (black points), and the motion corrected data (magenta line). This process dramatically reduces noise induced by telescope motion.

8. Finally the the 1D spline is added back to the data, resulting in motion corrected data.

This procedure is applied to all pixels, where the correction is most prevalent in pixels near the science target, or field stars. A dramatic example of this correction can be seen in Fig. 5.1,wherethroughthismotioncorrectionwesuccessfully reconstruct the light curve of SN 2018ajj (Smith et al. 2018). No extrapolation is included in this method, so the data becomes truncated to the first and last instance where the pointing accuracy is < 0.3pixels. We find this procedure is successful at removing almost all motion artefacts from pixel light curves. The only residual artefacts that have been encountered are ones that persist between multiple thruster resets. Although we do not remove these trends, the false positives they produce in the event detection are removed through vetting steps discussed in 5.4.3. § 92 Kepler/K2: Background Survey

5.4.3 Short event identiﬁcation (< 10 days)

The core aspect of K2 :BS is identifying real events from noise or other contaminants (e.g. asteroids). For each campaign every pixel is analysed independently to identify apotentialeventandanassociatedeventtime.Adetectionlimitiscalculatedfor each pixel as Limit =median+3σ from all images. This cut-offvalue sets the magnitude limit for each pixel. Due to telescope drift, pixels close to science targets or stars have brighter limits than pixels with large separations. This highly variable level of counts, as seen in the magnitude limits shown in Fig. 5.6,isthemotivation for analysing pixels individually. The process of identifying potential events utilises Boolean arrays and indexing. The K2 target pixel file flux is converted to a Boolean array, conditioned on pixels having counts greater than the aforementioned limit. To avoid anomalous detections based on spacecraft operation all pixels in frames that coincide with nonzero quality flags, indicating anomalous behaviour, are set to false. Residual telescope motion and noise can cause false breaks in the Boolean array, as the light curve may periodically drop below the detection limit. These breaks will produce false event durations and may lead to transients being processed incorrectly and lost. To prevent noise from truncating event duration the Boolean detection array is smoothed by an iterative process of convolving each pixel, through time, with a 1d kernel of zeros with ones at the start and end positions. The smoothing process iterates from alength9kernel,toalength3kernel,witheachiterationactingontheproduct from the previous step. The convolution process is as follows,

C = D k (5.2) ∗ 1, if C =2 A(C)=⎧ (5.3) ⎨⎪0, if C<2 R = D + A(C), (5.4) ⎩⎪ where D is the initial Boolean detection array, k is the 1d smoothing kernel described above, A(C)istheconstructionofanewBooleanarrayfrom,C,theconvolvedarray, and R is the ﬁnal smoothed Boolean array used for detection. An example of this 5.4 Methods 93 process with a length 3 kernel is shown in Fig. 5.2. We use the smoothed Boolean array to identify candidate events with durations longer than a chosen baseline. The event identiﬁcation process analyses each pixel independently and operates as follows:

1. We identify the indices of all False values in the Boolean array.

2. The diﬀerence between neighbouring False indices is calculated.

3. Diﬀerences that are greater than a predeﬁned baseline are selected as candidate events.

4. The start and end times of the candidate event are recorded along with the pixel position. we set the minimum length to be 3 frames, or 1.5 hours, this requirement avoids detecting spurious noise or cosmic rays. This method proves successful in identifying transients that evolve rapidly over 1.5 hours to 10 days. ∼ Many of the potential events selected are false detections and contamination from known variable sources, that must be removed by subsequent checks. Sources of contamination can either be astrophysical (e.g., variable stars, asteroids), or instrumental (e.g., residual telescope motion). We impose a multi stage vetting procedure to limit false detections. First we match all coincidental events. For each candidate event we collapse the Boolean detection array along the time axis from the beginning to the end of the candidate events duration. We then construct a Boolean mask array and set the pixel position of the candidate event to be True. The mask is then iterative convolved with a 3 3TruearrayandanyTrueelementsthatappearinboth × collapsed Boolean array and the convolved mask are added to the mask, for the next iteration. This process terminates when no new pixel is added to the Boolean mask. Candidate events that are included in this new mask are merged into a single candidate. 94 Kepler/K2: Background Survey

Following the event matching procedure we vet the candidates based on their the uniqueness. We smooth the light curve by with a Savitzky-Golay filter, setting the width to be twice the event duration. All peaks in the light curve are then identified using the Scipy find peaks algorithm, which identifies local maxima through neighbour comparisons, and we condition on the peaks. We assume a real event will be significant and unique, so candidate events are rejected if: 1) the largest peak occurs outside the identified event time; 2) all peaks within a time frame less than twice the event duration from the candidate event must be less than 80% the brightness of the maximum peak. We find that in areas of prolonged poor pointing, aforestofsimilarpeaksemerge,whichcondition2isinstrumentalinvetting. Although these conditions are successful in eliminating almost all false detections, they introduce complexity into the true limiting magnitude. A thorough analysis of the magnitude limits for each pixel will be presented in future work. 5.4 Methods 95

1 0 1 0 0 1 * 1 0 1 Eq. 2 = 0 2 0 1 1 0 Eq. 3 0 1 0 0 0 0 + 1 0 1 0 0 1 Eq. 4 = 1 1 1 0 0 1

Figure 5.2: An example Boolean detection array being smoothed by a length 3 kernel. The smoothing prevents events from being truncated and rejected due to pixel count drops from instrument error. 96 Kepler/K2: Background Survey

5.4.4 Long event identiﬁcation (> 10 days)

This method closely follows that of the short detection method, however, greater care is taken in the construction of the pixel limits. As the K2 campaigns nominally last for 80 days, long transients, such as supernovae (SN) can strongly impact the ∼ light curve throughout the campaign, raising the detection limit to a point where the transient is no longer detectable. In order to construct an accurate limit we must examine trends in the light curve.

First, we smooth the light curve with a Savitzky-Golay ﬁlter with a window of 5 days. Although this process truncates the light curve by 10 days, it removes the signatures of short events (e.g., asteroids) from the light curve. We then identify the largest peak in the data using the Scipy find peaks algorithm (Virtanen et al. 2019), and break the light curve into 3 zones: the event zone, extending 10 days before to 25 days after the peak; and zones 1 and 2, before and after the event zone respectively. As some transients, such as SN can have long declines, we compare the light curve properties of zones 1 and 2 with the following;

µ2 >µ1 +3σ1, (5.5)

where µ1,2 are the light curve means of zones 1 and 2, and σ1 is the standard deviation of zone 1. If this condition is met then only zone 1 is used to calculate the limit, otherwise, both zones are used to calculate the limit. As with the short detection method, this limit is determined for each pixel individually as Limit = median+3σ.

With limits for each pixel, candidate long events are selected through a similar process to short events. Without smoothing the Boolean array, we locate events by the positions of False values. In this case we require the event duration to last longer than 10 days. For these events, we assume that they are unique and have well behaved peaks, as such candidate events are vetted by the following conditions: 1) the largest peak occurs inside the identified event time; 2) the peak can be well fit by a 3rd degree polynomial and poorly fit by a 1st degree polynomial, requiring the coefficient of determination (R2)tobe> 0.95 and < 0.5respectively. 5.4 Methods 97

5.4.5 Variable stars

Variable stars and their associated overﬂow into columns of pixels can lead to false detections. These false detections are mitigated by checking the pixels neighbouring the event pixel, if one of the neighbours contain > 100, 000 counts the event is considered false and discarded. This value is chosen as it encompasses the apparent variability in the saturation level of the pixels. If the bleeding of the target occurs at a lower count rate than the cut-oﬀ, then it may also be contained in a “Probable” event category. This category is discussed further in 5.4.7. §

5.4.6 Asteroids

Due to K2 observing along the ecliptic, there are many asteroids that pass through the data. As most asteroids remain in the TPF FOV for < 1day,theyarepredom- inantly detected by the short event method. The asteroid detection method utilises the fact that asteroids tend to move with a near constant speed through the TPF, thus they produce a parabolic light curve, as seen in Fig. 5.3. K2 :BS identifies asteroids by fitting a parabola to frames 2fromthebrightestpointinthepotential ± event light curve. The generated parabola is then fitted and subtracted from the event light curve, if the residual is small then the potential event is categorised as an asteroid. All asteroids are recorded and can be examined independently. It is possible that real short transients may fall within the asteroid criteria, however, such events could be recovered at a later point following an analysis of asteroid-like events4.

4Although uncommon, some asteroids will move through an epicycle within the TPF. These asteroids do not have uniform motion, and so can be missed by the asteroid ﬁlter. We are currently working on adaptive masks that are capable of tracking asteroids. These masks would extract all asteroid information and separate asteroids from transients. 98 Kepler/K2: Background Survey

36 Counts

34 Parabola ﬁt 32 Event LC Event duration 30 348 350 352 354 356 358 360 Time (frames)

Figure 5.3: An example of a potential event identiﬁed as an asteroid in C01, EPIC 201735583. Since asteroids travel at a relatively constant speed through the aperture, the light curves they produce are largely parabolic. Thus parabolic light curves of short events are classiﬁed as asteroids.

5.4.7 Event sorting

For each object detected, we must search for a potential host or source of the event. To simplify visual vetting of candidate events, they are split into categories based on several event aspects such as duration, brightness, detection method, source type, and relation to masked objects. Identifying the host of a potential event can follow different pathways. For each detection, we query NED and Simbad to identify potential sources or host galaxies. If the event occurs within the mask of a science target, the host is defined to be the science target, and the candidate event is sorted into the “In” category and subcategory based on the host type. Similarly, events that occur next to a science target mask are sorted into the “Near” category and host type subcategory. Candidate events that occur in the background pixel, not associated with a science target, are identified through querying the coordinates corresponding to the brightest pixel, and sorted into a category corresponding to the likely host type. If a candidate event’s coordinates do not correspond to an object in NED or Simbad database, it is assigned to the “Unknown” category. This category often 5.4 Methods 99 contains false detections form the K2 electronic noise sources, however it can also contain previously unseen events and some of the most exciting events promised by this analysis. An example of a new object found through an “Unknown” classification can be found in (Ridden-Harper et al 2019).

5.4.8 Event ranking

As a further diagnostic to assist in the vetting of candidate events is a series of quality rankings. Each candidate receives ranking for the brightness, duration, mask size, and source/host type. These rankings, as outlined below, provide a way of sorting candidate events into prioritised lists, simplifying the vetting process.

The brightness ranking is taken as the significance of the event. The sig- • nificance is found by calculating the peak flux of the candidate event and comparing it to the median and standard deviation of the entire light curve, excluding points from two days before the candidate event starts, to ten days after it ends.

The duration ranking is simply the calculated duration of the event in days. • The mask size ranking is calculated from the number of pixels included in the • candidate event mask. This ranking is normalised such that three or more pixels in a mask produce a maximum rank of 1.

The host ranking is based on classiﬁcation given to the candidate event, from • the NED object classiﬁcation system. Objects that aren’t of interest to this survey, such as stars, are ranked 0, while objects of interest, such as galaxies, quasars, etc., and unknown objects, are ranked 1.

5.4.9 Visual inspection

If an event passes through the K2 :BS conditions it is finally checked through visual inspection. K2 :BS generates an event figure, video, and event positions. An example detection figure is shown in Fig. 5.4 for a short outburst from quasar [HB89] 1352- 104. The top of the figure contains diagnostic information on the object and position; left shows the event light curve, with diagnostic information, such as thruster firings 100 Kepler/K2: Background Survey

Figure 5.4: Example of a vetted, real event detected by K2 :BS. This event was detected in K2 object 212595811, which is a short 3 day outburst from the quasar [HB89] 1352-104. During the outburst the apparent magnitude increases from 17.2 Kp to 15.1 Kp.Thelightcurveontheleftpresentstheeventlightcurve along with diagnostic information such as when a thruster reset occurs and quality flags. The sub-figures on the right show the reference and peak brightness event frame, with the event mask overlaid in red points. and quality flags, the background and nearest science target light curves, and the identified event duration in orange; right shows the reference image at the top and the brightest frame from the event at the bottom. Selected frames from the corresponding event video for [HB89] 1352-104 are shown in Fig. 5.5.Leftistheeventlightcurvewiththeverticalredlineshowing the time stamp of the K2 frame shown on the right. These figures, alongside candidate event sorting provide enough information to accurately assess the validity of acandidateevent. 5.5 Survey Characteristics 101

12 a b c a 6

1000) 9

⇥ 4 Row 6 2 Counts ( 3 0 60 65 70 0 2 4 6 8 Time (days) Column b c 6 6

4 4 Row Row 2 2

0 0 0 2 4 6 8 0 2 4 6 8 Column Column

Figure 5.5: The simpliﬁed light curve used in K2 :BS event video for the short outburst from the quasar [HB89] 1352-104 (top left), alongside selected frames. Pannels a, b,andc are K2 images corresponding to the times indicated by red lines.

5.5 Survey Characteristics

As this survey is using existing data, survey limits and characteristics are set by the original data, as discussed in 6.3.Keysurveyparametersarethecadence,survey § time, area, and depth. Operating with K2 long cadence data ﬁxes the cadence to 30 minutes and survey time per ﬁeld to 80 days. Events can only be detected ≤ if they are longer than the minimum event duration of 8 frames or 4 hours (set to prevent contamination from short scale noise or cosmic rays, and short enough to exhibit variation over a campaign, so that the long detection method can be successful. 102 Kepler/K2: Background Survey

Table 5.1: Number of background pixels, and the equivalent on sky area for K2 extragalactic pointing campaigns.

Campaign Pixels Area (deg2)Duration(days) C01 3894581 4.8 83 C06 1957708 2.4 79 C12 2120713 2.6 80 C13 1645855 2.0 81 C14 2095376 2.6 81 C16 1408794 1.7 81 C17 2590027 3.1 69

Since two distinct detection methods are used that probe two different time domains, it is necessary to define the limits and recovery rates for both. Although the two methods target different time domains, there is an overlap for events with peaks or durations that last 2days.Thisoverlappreventsdetectiongapsinthe ∼ time domain between the two. As both detection methods scan every pixel, they feature the same survey area. For this analysis, we will focus on extra-galactic pointing fields, which have the best chance at detecting transients. As seen in Tab. 5.1,thebackgroundpixelscovera significant area. C01 features the highest number of background pixels, due to each science target being assigned larger pixel masks. Although C16 is an extra-galactic field, the area of background pixels is significantly lower than that of the other due to the field containing the Beehive cluster, and the Earth.

5.5.1 Magnitude limits

K2 :BS is limited to the area shown in Tab. 5.1,however,ithasahighlyvariable magnitude limit. In this analysis the limiting magnitude is set by the detection limit that is imposed and discussed in 5.4.3 and 5.4.4.Theinterplayofthescience § target and telescope drift sets diﬀerent magnitude limits, and therefore volumes, for each pixel. An example of the variable magnitude limit for a target pixel ﬁle can be seen in Fig. 5.6.Ingeneral,pixelsclosetothesciencetargethavebrighterlimits, than those further away. This limit provides a simple way to calculate the expected volumetric rate of events. Due to the mission design of K2,themajorityofdownloadedpixelsdonot contain a target, and have faint limiting magnitudes. The distribution of pixel 5.5 Survey Characteristics 103

14 7 15 6 16 5 17 4 18 p Row

3 K 19 2 20 1 21 0 22 0 1 2 3 4 5 6 7 8 Column

Figure 5.6: Magnitude limit for EPIC 212787678 from C17. Pixels near the original science target (centre frame) have a brighter limiting magnitude due to the science target’s brightness, variability and residual telescope drift, with the pixels far from the target have a faint limiting magnitude. Each ﬁeld has an associated magnitude limit from which the observed volume can be calculated. limiting magnitudes are shown for C06 in Fig. 5.7,where78%ofpixelshavelimiting magnitudes 18th,whichareidealfortransientdetection. ≥ Although the limit provided by K2 :BS is representative of the expected limiting magnitude, the candidate event vetting process will introduce complexities. As discussed in 5.4.3 and 5.4.4 both the short and long detection methods undergo § vetting to remove false detections. These checks have the potential to discard real transients if data quality is sub-optimal. As determining accurate limiting magnitudes are critical to event rates, we will present a robust analysis accounting for the detection eﬃciency and contamination for a variety of key transients. 104 Kepler/K2: Background Survey

105

104

103 Number of pixels 102

12 14 16 18 20 22 Limiting magnitude

Figure 5.7: Distribution of pixel magnitude limits for C06, where 78% of pixels have limiting magnitudes 18th.Thehighproportionofpixelswithfaintlimitsareideal ≥ for transient searches. Although the limits of some pixels are fainter than Kp =21, it is unlikely that Kepler can achieve such sensitivity, so the limit is set to Kp =21 in rate calculations.

5.6 Detected events

The pilot program of K2 :BS identiﬁed a number of known and undiscovered events. Initial runs have yielded promising results, by independently recovering key transients discovered in the Kepler Extra-galactic Survey, and recovering a superoutburst of a dwarf WZ Sge nova which is presented in Ridden-Harper et al. (2019). Here we will present example detections of known events, independently recovered by K2 :BS, that we use as method veriﬁcation. Two examples of short transient events detected by K2 :BS are variability in quasar [HB89] 1352-104, and KSN 2015K. During C06, quasar [HB89] 1352-104 experienced an outburst that lasted 3 days, as seen in Fig. 5.4.Variabilitypre- ∼ and post- outburst are also visible. The short transient KSN 2015K (Rest et al. 2018) was also recovered, as seen in Fig. 5.8.Thepositionoftheeventwasdeterminedto within to within 4” of the accepted value, which is the resolution limit of Kepler. Both of these events are key examples of short events that may be detected by 5.6 Detected events 105

Figure 5.8: K2 :BS independent detection of KSN 2015K, presented in Rest et al. (2018). This event was recovered through the “short” event detection method.

K2 :BS. K2 :BS is also capable of detecting type Ia supernovae, such as SN 2018oh, as seen in Fig. 5.9. This event was detected through the long detection method. As well as detecting SN 2018oh, the excess light at rise is visible, as discussed in Dimitriadis et al. (2019a); Shappee et al. (2019); Li et al. (2019). 106 Kepler/K2: Background Survey

Figure 5.9: K2 :BS independent detection of SN 2018oh, presented in Dimitriadis et al. (2019a); Shappee et al. (2019); Li et al. (2019). This event was recovered with the “long” event detection method.

5.7 Expected rates

The motivation for the K2 :BS is to detect transient events in K2 data, so its useful to determine the likely number of events the survey should detect. Since each K2 campaign is unique, the survey area and therefore event rates vary between campaigns. In the following sections we present the expected detection rate for SN Ia and CCSN, gamma ray burst (GRB) afterglows, kilonova, and the fast-evolving luminous transients (FELTs) detected by PAN-STARRS1 (Drout et al. 2014). FELTs, kilonova, and GRB afterglows evolve rapidly, so are detectable by the short detection method and we therefore use rates from this method, while SN such as SN Ia and CCSN are detected through the long event detection method and we use respective rates.

There are few known types of transients that have lifetimes on the order of days. As K2 :BS will probe the 1daytimedomain,theresultsofthissurveywillbe ∼ used to deﬁne rates for day to sub-day duration transients and presented in future work. 5.7 Expected rates 107

5.7.1 Volumetric rates

For K2 :BS the volumetric rates are derived for each pixel and summed to reach the ﬁnal rate. The motivation for this segmented approach is inﬂuenced by the large pixel size and, more importantly, the high variability in sensitivity between pixels. By knowing the magnitude limit per pixel we can calculate the volume, V ,asfollows,

3 θ 1 (m M+5) V = 10 5 − 3 ' ( where θ is the solid angle of a pixel in steradians, m is the magnitude limit of a pixel, and M is the absolute magnitude of the transient. As these calculations are dependent on an absolute magnitude, the volumes are model dependent. For supernovae we use the Li et al. (2011b)typeIasupernovarateof(0.301 ± 4 3 1 0.062) 10− Mpc− yr− and an absolute magnitude of 19.0. For the kilonova rate × − +3200 3 1 we use 1540 1220 Gpc− yr− from Abbott et al. (2017b), and take the absolute mag- − nitude to be 16, based oﬀGW170817 light curves presented in Villar et al. (2017). − 3 1 The FELT rate is taken as 6400 2400 Gpc− yr− with an absolute magnitude ± range of 16.5to 20 K from Drout et al. (2014). − − p Since GRB light curves are highly model dependent, we will explore their rates in the following subsections.

GRB rates

For this analysis we use the afterglowpy package described in Ryan et al. (2019)for GRB afterglow light curves. In this exploratory case, we generate GRB afterglows using the afterglowpy top hat jet with the following set of ﬁducial GRB parameters;

52 isotropic-equivalent energy, E0 =10 erg; half-opening angle, θj =3◦ (Nicuesa 3 Guelbenzu et al. 2012); circumburst density, n =1.0cm− ;energyequipartition factors of the electrons and magnetic ﬁeld, ϵe =0.1andϵB =0.01 respectively. The apparent magnitude of a GRB afterglow depends on both the viewing angle,

θobs,andtheredshift,z.WecalculatetheapparentmagnitudeoftheGRBafterglow independently varying θ from 0 90◦ in steps of 0.5◦ and z from 0.01 5insteps obs − − of 0.01. By comparing the limiting magnitude of a pixel to the GRB magnitude array, we can determine the maximum z apixelcoulddetectaGRBforagiven 108 Kepler/K2: Background Survey viewing angle.

Following Guetta et al. (2005), we approximate the GRB formation rate with the star formation rate from Rowan-Robinson (1999);

100.75z,z<1 RGRB = ρ0 ⎧ , (5.6) ⎪100.75,z1 ⎨ ≥ ⎪ 3 3 1 ⎩ where ρ 33 h Gpc− yr− (Guetta et al. 2005). From Japelj & Gomboc (2011) 0 ≈ 65 the number of GRBs per year is given by

zmax R dV N = GRB dz, (5.7) 1+z dz 00 where the co-moving volume element, dV/dz,isgivenby

dV c 4πχ2(z) c z′ dz = ,χ= ′ , (5.8) dz H E(Ω ,z) H E(Ω ,z) 0 i 0 00 i ′

3 and E(Ωm,z)= Ωm(1 + z) +ΩΛ.Forthesecalculationsweadoptastandard 1 1 ΛCDMcosmological1 model where H0 =72kms− Mpc− ,Ωm =0.3, and ΩΛ =0.7. With the ﬁducial GRB light curves and occurrence rate, we calculate the expected number of detections. The number of detections are calculated per pixel, accounting for all possible viewing angles. In these calculations we take “On-Axis” GRBs to be θ θ and “Oﬀ-Axis” GRBs, or Orphan Afterglows, to be θ >θ. obs ≤ j obs j

5.7.2 Expected detection rates

Tab. 5.2 shows the expected number of events from various K2 extra-galactic campaigns. We find that it is unlikely that K2 serendipitously observed any kilonova or GRBs during the extragalactic pointing campaigns, however, it is likely that K2 serendipitously observed many SN Ia and FELTs. Although the expected rates for SN Ia and FELTs appear high, they are likely overestimates, due to intricacies in the K2 field selections. The Kepler Extragalactic Survey (KEGS) survey will bias K2 :BS towards observing galaxies, as most large galaxies were selected, while the star targets may bias K2 :BS against galaxies. These subtle, but important, biasing 5.7 Expected rates 109 factors will be incorporated into a follow-up paper where we present the comprehensive transient rates from K2 :BS, following completion of the survey. 110 Kepler/K2: Background Survey GRB 0.080.080.06 0.02 0.08 0.02 0.05 0.01 0.10 0.02 0.02 0.02 -axis ff )On-axisO P 30.140.03 K 1 2 1 2 1 2 ± 0 ± ± ± ± ± ± . 6 6 5 6 4 8 10 20 → → → → → → → →− 01 01 01 01 01 02 02 ...... FELT 0 0 0 0 0 0 P 0 ± ± ± ± ± ± K ± 5 . 05 05 04 05 03 06 08 ...... 0 0 0 0 0 0 16 0 − ( campaigns for SN Ia, kilonovae, and FELTs. C01 has the largest rate, due K2 ) 3 − 8 4 3 4 5 5 6 +20 − +9 − +8 − +11 − +12 − +13 − +15 − 10 5 4 6 6 6 7 10 × ( 2 1 2 1 2 1 2 ± ± ± ± ± ± ± 7 7 6 8 5 9 12 Campaign SN IaC01 Kilonova (NS-NS) C06 C12 C13 C14 C16 C17 Table 5.2:to Volumetric larger pixel masksC16, rates around although objects, an for extra-galactic and the field, therefore contains more extra-galactic the background Beehive pixels cluster and and the Earth, pointing largest which both number of severely reduce expected the detections. sensitivity. Conversely 5.8 Conclusions 111 5.8 Conclusions

The Kepler K2 mission has provided a wealth of information on transient events. Through a directed survey of over 40,000 galaxies KEGS found a multitude of transients, including a rare fast transient KSN 2015K (Rest et al. 2018)andthepoten- tial detection of a SN Ia shock interaction in SN 2018oh (Dimitriadis et al. 2019a; Shappee et al. 2019; Li et al. 2019). The transients detected by KEGS are not the only transients detected by Kepler,anumberofeventsmayhavebeendetected serendipitously. The precedent on this was set by the discovery of a dwarf cataclysmic variable outburst in the original Kepler mission (Barclay et al. 2012). Through the K2 :BackgroundSurvey,weproposetoanalyseallpixelsinK2 data for serendipitous transient detections. This survey will be blind, with final candidates vetted. The survey is not limited to specific types of transients, as all events that meet the detection threshold will be selected for final vetting. This unconstrained search, coupled with the rapid cadence of Kepler will allow us to search for transients in the time domain of hours, which is poorly understood. This survey will cover 3deg2 per campaign, with a total area of 50 deg2 ∼ ∼ through all K2 fields. The total area, combined with the depth of up to 22 Kp makes it likely that a number of unknown transients have been observed during K2. This is made evident by the first reported detection of a WZ Sge dwarf nova superoutburst presented in Ridden-Harper et al. (2019), discovered by K2 :BS. The analysis method presented here is not only applicable to K2 data. As Kepler and TESS share the same data structure as K2,theytoocanbeanalysedthrough the background survey. The search strategies will also be more effective in Kepler and TESS data, due to the comparative high stability. Following the analysis of K2,thebackgroundsurveywillrunonthesetwoadditionaldatasets.

Acknowledgements

We thank Geoﬀrey Ryan for their assistance in implementing afterglowpy.This research was supported by an Australian Government Research Training Program (RTP) Scholarship and utilises data collected by the K2 mission. Funding for the K2 mission is provided by the NASA Science Mission directorate. We also make 112 Kepler/K2: Background Survey signiﬁcant use of the NASA/IPAC Extragalactic Database (NED), which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. 6 Discovery of a new WZ Sagittae type cataclysmic variable in Kepler/K2 data

This chapter is accepted for publication to the Astrophysical Journal Ridden-Harper, R., Tucker, B. E., Garnavich, P., Rest, A., Margheim, S., Shaya, Littleﬁeld, C., E., Barensten, G., Hedges, C., Gully-Santiago, M., “Discovery of a New WZ Sagittae Type Cataclysmic Variable in the K2/Kepler Data”, 2019, MNRAS, accepted.

6.1 Abstract

We identify a new, bright transient in the K2/Kepler Campaign 11 ﬁeld. Its light curve rises over seven magnitudes in a day and then declines three magnitudes over a month before quickly fading another two magnitudes. The transient was still detectable at the end of the campaign. The light curve is consistent with a WZ Sge type dwarf nova outburst. Early superhumps with a period of 82 minutes

113 Discovery of a new WZ Sagittae type cataclysmic variable in 114 Kepler/K2 data are seen in the first 10 days and suggest that this is the orbital period of the binary which is typical for the WZ Sge class. Strong superhump oscillations develop ten days after peak brightness with periods ranging between 83 and 84 minutes. At 25 days after the peak brightness a bump in the light curve appears to signal a subtle rebrightening phase implying that this was an unusual type-A outburst. This is the only WZ Sge type system observed by K2/Kepler during an outburst. The early rise of this outburst is well-fit with a broken power law. In first 10 hours the system brightened linearly and then transitioned to a steep rise with a power law index of 4.8. Looking at archival K2/Kepler data and new TESS observations, a linear rise in the first several hours at the initiation of a superoutburst appears to be common in SU UMa stars.

6.2 Introduction

Cataclysmic variable stars (CVs) consist of a white dwarf star accreting material from a companion star. These accreting binary systems show a rich array of physical processes that are displayed over many diﬀerent timescales (for a review, see Warner 1986). Mass transfer from the companion star is often facilitated through an accretion disk around the white dwarf star. Thermal instability in the disk drives outbursts known as dwarf novae (DN; Osaki 1974; H¯oshi 1979). The character of these DN outbursts depends on the binary orbital period. Accretion disks in systems with orbital periods less than two hours can experience periodic “superoutbursts” that are more luminous and last longer than normal outbursts. This class of DN are named for their prototype, SU Ursa Majoris (SU UMa). During a superoutburst, the disk can be excited by a 3:1 orbital resonance with the binary orbit that generates strong optical oscillations known as “superhumps” (SHs) (Whitehurst 1988; Osaki 1989; Lubow 1991a,b). The frequency and amplitude of the SHs evolve over the outburst in speciﬁc ways that are good diagnostics of the physics in the disk and binary system. SU UMa DN are thought to be evolving to shorter orbital periods through energy loss by gravitational radiation. As the orbital periods shorten, the secondary mass decreases and the mass ratio of the systems become extreme. At periods around 6.2 Introduction 115

80 minutes, the separation between the components begins to increase in what is known as the period bounce (Kolb & Baraﬀe 1999; Knigge et al. 2011,andreferences therin). CVs near the bounce period have mass ratios less than 0.1 and these are known as WZ Sge type DN. Kato (2015)presentsanextensivereviewofthese binaries. WZ Sge stars have a number of distinctive properties, such as very large amplitude outbursts, brightening by 7 or 8 magnitudes. The time between outbursts can be many years to decades, meaning that catching a particular system in a bright state is rare. Unlike other SU UMa DN, WZ Sge DN feature double-peak variations in their light curves known as “early superhumps”. These early superhumps are a result of the high mass ratio of WZ Sge systems, that allows the accretion disk to reach the 2:1 orbital resonance (Osaki & Meyer 2002). Early superhumps are a valuable diagnostic tool, featuring periods that correspond closely with the orbital period (Kato 2002; Ishioka et al. 2002).

Another feature common to WZ Sge DN is a late time rebrightening event at the end of a superoutburst. As described in Kato (2015), these events have been classi- fied into 5 categories: type–A/B outbursts, long duration or multiple rebrightenings; type–C outbursts, single rebrightening; and type–D outbursts, no rebrightening, all identified by (Imada et al. 2006); and type–E, double superoutburst, identified by Kato et al. (2014). Using the rebrightening types Kato (2015)constructsanevo- lutionary sequence that can be used to identify the age of the system and the configuration, with the evolutionary sequence type C D A B E. Type–A → → → → outbursts were found to be close to the period minimum, making them good period bouncing candidates.

The Kepler space telescope has proven to be a valuable instrument in detecting and understanding DN. With the 1 minute short cadence, and 30 minute long cadence data, Kepler has obtained exquisite light curves of known and previously unknown DN. Kato & Osaki (2013b) presents the analysis of three SU UMa DN, V585 Lyr and V516 Lyr and the serendipidously discovered KIC 4378554 presented in Barclay et al. (2012). With the high photometric precision of Kepler data and long baseline aﬀorded by the Kepler prime mission, Kato & Osaki (2013b)presented evidence that supported the thermal–tidal instability theory as the origin of superoutbursts. Discovery of a new WZ Sagittae type cataclysmic variable in 116 Kepler/K2 data

Like Kepler, K2 data offers the unparalleled high-cadence, 30 minute observations of thousands of targets (Borucki et al. 2010; Koch et al. 2010; Howell et al. 2014). K2 differs from Kepler in that each science target is allocated a larger detector area to compensate for telescope drift, as a result fewer science targets were observed. Although drift reduces the photometric precision of K2,theincreaseincontiguous background pixels improves the chances that transients can be detected. Here, we present observations of a new WZ Sge type CV discovered as part of asystematicsearchfornewtransientsintheK2/Kepler campaigns, known as the K2: Background Survey (K2:BS). Our new CV is the first WZ Sge type system to be observed with the high-cadence and continuous monitoring of Kepler.InSec.6.3 we describe the detection method used to discover KSN:BS-C11a. Furthermore, we outline the K2/Kepler light curve processing and spectroscopy of the system. In Sec. 6.4 we analyse the K2/Kepler light curve and the unique high-cadence coverage of the rise to peak brightness. Finally in Sec. 6.5 we compare the light curve of KSN:BS-C11a with other WZ Sge stars to derive fundamental parameters of the binary system.

6.3 Data

6.3.1 Search for Transients in K2/Kepler

Through the K2/Kepler Extra-Galactic Survey (KEGS), K2/Kepler observed thousands of galaxies and provided a wealth of information on transients that were detected in scheduled galaxy targets (e.g. Garnavich et al. 2016; Rest et al. 2018; Dimitriadis et al. 2019a). Although many transients were detected in the directed search, there are more transients hidden in the K2/Kepler data. Each science target in K2/Kepler has numerous background pixels that are observed at high-cadence. During a campaign these background pixels may serendipitously collected transient signals, which have gone previously undetected. Ridden-Harper et al. (2020)presentstheK2:BackgroundSurvey(K2:BS),which conducts a systematic search for transients in K2/Kepler background pixels. K2:BS independently analyses each pixel to detect abnormal behavior. This is done by searching for pixels that rise above a brightness threshold set from the median 6.3 Data 117 brightness and standard deviation through a campaign. Telescope motion presents achallengeincandidatedetectionassciencetargetsmaydriftintobackground pixel, triggering false events. False triggers are screened by vetting of events that last < 1day,chosenforcandidateswiththe6hourlytelescoperesets.Coincident pixels that pass the vetting procedure are grouped into an event mask. All candidate events are checked against the NASA/IPAC Extragalactic Database (NED)1 and the SIMBAD database (Wenger et al. 2000)toidentifypotentialhosts.K2:BSproduces event ﬁgures, videos and relevant detector information, which are used for manual vetting.

6.3.2 Discovery of KSN:BS-C11a

A bright transient, KSN:BS-C11a, was detected in the K2/Kepler archived target pixel ﬁle (TPF) for EPIC 203830112 Campaign 11 (C11). The science target EPIC 203830112 was proposed for observation by several programs in C02 and C11 shown in Table 6.1. The science target is a F-type dwarf that was selected for magnitude limited surveys of FGK dwarf stars with the objective of understanding the occurrence rate and properties of near-Earth sized planets. The location of KSN:BS-C11a was observed in both C02 and C11, however, it was in quiescence in C02.

Table 6.1: Proposals to observe EPIC 203830112, all are independent of this paper.

Proposal ID Campaign PI GO2054 C02 Sanchis-Ojeda et al. GO2104 C02 Petigura et al. GO11071 C11 Charbonneau et al. GO11122 C11 Howard et al. GO11902 C11 GO Oﬃce, K2

KSN:BS-C11a was discovered in K2/Kepler C11 galactic pointing campaign, during a preliminary search of K2:BS. This event went undetected as transient surveys (e.g. Rest et al. 2018)avoidedthecrowdedgalacticﬁelds.Furthermore, KSN:BS-C11a was unlikely to be detected from ground based surveys as Kepler

1The NASA/IPAC Extragalactic Database (NED) is operated by the Jet Propulsion Labora- tory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. Discovery of a new WZ Sagittae type cataclysmic variable in 118 Kepler/K2 data

2457666.59 BJD 2457695.01 BJD

510 510

508 508

506 506

504 504

502 502 Row Row 500 500

498 498

496 496

494 494

492 492 0 2 4 6 8 10 12 14 16 18 0 2 4 6 8 10 12 14 16 18 Column (+989) Column (+1021)

Figure 6.1: K2/Kepler Full Frame Images (FFIs) for C11, the white circle indicates the science target for EPIC 203830112, the orange solid box indicates the chosen event aperture, and the black dashed boxes indicate the K2 target pixel file boundaries for the C11 sub-campaigns (C111 and C112). The C111 image is shown on the left and the C112 on the right, which contains KSN:BS-C11a 3dayspostmax. ∼ was “backward facing” during C11, meaning that the target field appeared in the evening twilight as viewed from the Earth. The discovery of KSN:BS-C11a shows the capability of K2:BS to rapidly analyse K2/Kepler data for transients, even in galactic pointing fields.

6.3.3 Full Frame Images

In each K2/Kepler campaign, at least one full frame image (FFI) is taken. Many campaigns, such as C11, have two FFIs: one at the start, and one midway through the campaign. In rare cases when a transient occurs during an FFI, it can be shown in relation to all other objects in the region. Such occurrences are useful to establish the reality of the transient and determine its total brightness.

The two FFIs for C11 were taken at 2457666.59 BJD and 2457695.01 BJD, where the second FFI occurs 3 days after KSN:BS-C11a reached peak brightness. As ∼ shown in Fig. 6.1 the event is well separated from the science target, by 3pixels, ∼ or 12′′. The presence of KSN:BS-C11a in the FFI is strong conﬁrmation that ∼ KSN:BS-C11a is not an artefact of the data. 6.3 Data 119

2.0 Time of FFI 1.5 x displacemnt y displacemnt 1.0

0.5

0.0

Displacement (Pixels) 0.5

1.0 690 700 710 720 730 BJD (-2457000)

Figure 6.2: Kepler’s displacement during the C112 sub-campaign. Although there is large x-displacement (across pixel columns), there is negligible y-displacement (across pixel rows), only deviating by 0.1 pixels during the event. It seems reasonable that the TPF consistently receives 96% of the event ﬂux. ∼

Because the transient is at the edge of the TPF, the wings of the point-spread- function (PSF) are cutoffin the individual images. We can use the FFIs to estimate the fraction of event flux that the TPF missed. Aperture photometry was preformed with the event apertures shown in Fig. 6.1 with the C111 sub-campaign aperture for background subtraction. The FFI counts were CFFI =20586e/s,whiletheevent light curve recorded lower counts of C =17556e/s.Thus,theTPFreceived 96% lc ∼ of the total event flux. As noted below, we find that there is little spacecraft motion that would move KSN:BS-C11a offthe TPF, so it likely that a high level of flux was captured for the entire event. Discovery of a new WZ Sagittae type cataclysmic variable in 120 Kepler/K2 data

Figure 6.3: The K2/Kepler light curve of the transient KSN:BS-C11a shown with the 30-minute cadence. The time axis is shown in barycentric Julian days and the ﬂux has been converted to Kepler magnitudes.

6.3.4 K2/Kepler Light Curve

The light curve for KSN:BS-C11a was constructed through simple aperture photometry. Due to telescope motion across the TPF, as seen in Fig. 6.2,thecustom aperture was chosen to extend across many columns, to encapsulate all flux. Al- though a larger mask introduced larger background noise, this was inconsequential due to the high event count rate, and a quiescent period at the beginning of sub- campaign C112, which is ideal for background subtraction. AlesssuccessfulmethodofdatareductionwasattemptedusingPointResponse Function (PRF) model fitting photometry. This data reduction method is available though the Lightkurve python package (Lightkurve Collaboration et al. 2018), in which a PRF is fit to each frame to determine telescope properties and recover the flux, free from telescope effects. Although this method proved successful in accounting for telescope motion, it was unable to recover the total counts of the event, thus the aperture photometry method was favoured. 6.3 Data 121

The light curve from the K2/Kepler data is shown in Fig. 6.3.Duetothehigh count rate of this event, and the rapid cadence of K2/Kepler data, the light curve of the event is extremely well deﬁned. The event had a rapid rise of > 7magover ∼ 1day,followedbyaslowdecaylastingaboutamonth.Largeamplitudeoscillations began about 10 days after maximum light. Overall, the light curve appears to be that of a short period CV in superoutburst. In particular, its characteristics are similar to outbursts of GW Librae (Vican et al. 2011), a WZ Sge type CV.

6.3.5 Gemini Spectra

Following the discovery of KSN:BS-C11a, Pan-STARRS1 (PS1) images of the field were searched to identify the system in quiescence. Given the 4” pixel-scale of the K2/Kepler data, however, there were several possible progenitors within the uncertainty ellipse. Spectra of the first candidate was obtained on 2018 July 20 (UT), the remaining two candidates were observed on 2018 July 25 (UT) (MJD58324.3) with Gemini Multi-Object Spectrograph North (GMOS-N), using the R400 grating with a wavelength range of 3960–8687 Aand˚ a 1” longslit. The data were processed using standard routines within the Gemini IRAF package. The faintest candidate revealed a blue continuum with a broad, double-peaked Hα emission feature which confirmed this source as the dwarf nova. Using positions for nearby stars from the USNO-B1.0 catalog, we measure the astrometric position of the progenitor to be RA = 16 : 53 : 50.67 DEC = 24 : 46 : − 26.50 (J2000) with an uncertainty of 0.24 arcsec. The location of the star in the quiescent state is shown in Fig. 6.4. Discovery of a new WZ Sagittae type cataclysmic variable in 122 Kepler/K2 data

Figure 6.4: A pseudo-color image of KSN:BS-C11a constructed from g, r,andi ﬁlter images obtained with DECam on the CTIO 4.0 m telescope. The images were taken on MJD 58362.9 when KSN:BS-C11a was in quiescence. The variable is the the blue object marked by dashes. North is toward the top and east is to the left in the image.

6.3.6 DECam images

Following spectroscopic confirmation with Gemini, deep images were taken with CTIO 4-meter telescope and the DECam instrument on 2018 September 1 (UT) (MJD58363.0). KSN:BS-C11a was imaged in four SDSS filters, with exposure times of 120 s for g and r,and90sforz and i.Theimageswerereducedfollowing standard procedure with the DECam pipeline. Photometry in the crowded field was performed by point-spread-function (PSF) fitting with DAOPHOT implemented in IRAF2. An average PSF was constructed

2IRAF is distributed by the National Optical Astronomy Observatory, which is operated by 6.4 Analysis 123 from isolated stars in the ﬁeld and this function was scaled to match the brightness of the two stars just west of the variable. After subtracting these stars, aperture photometry was performed on KSN:BS-C11a and on nearby stars calibrated in the PS1 Survey DR1 catalogue (Chambers et al. 2016). We ﬁnd the quiescent brightness of KSN:BS-C11a is r =21.46 0.08 and g =21.52 0.07. From these estimates ± ± we approximate the magnitude in the Kepler system as Kp =21.5. Given that the Kepler brightness at the peak of the outburst was K 13.4, the full amplitude of p ∼ the outburst was 8.1 mag.

6.4 Analysis

6.4.1 Power Spectrum

During a superoutburst, oscillations occur in the light curve caused by orbital res- onances so that the power spectrum of an outburst can probe the physics of the disk. The high-cadence observations of K2/Kepler are capable of identifying the short scale superhump oscillations and measure their time evolution (Barclay et al. 2012; Kato & Osaki 2013c,a; Brown et al. 2015). To study the oscillations during the outburst we have removed the slow decline from maximum by subtracting a median-ﬁltered light curve with using a 36 hour wide kernel. The resulting residual light curve is shown in Fig. 6.5. The evolution of a WZ Sge outburst begins with “early” superhumps that tend to have a low amplitude and a period close to the orbital period of the system. An increasing amplitude signals the Stage A superhumps that rapidly transition to the Stage B. The outburst ﬁnally shows variations on the time scale of a few days that we associate with the rebrightening phase (Kato 2015). The power spectrum of KSN:BS-C11a over the outburst is seen in Fig. 6.6,and

1 1 reveals two broad peaks at frequencies around 17.2 day− and 14.4 day− .Anarrow

1 peak at 4 day− is due to the regular 6 hour spacecraft drift plus thruster correction cycle.

1 The power at 14.4 day− is likely the second harmonic of the primary peak the Association of Universities for Research in Astronomy (AURA) under a cooperative agreement with the National Science Foundation. Discovery of a new WZ Sagittae type cataclysmic variable in 124 Kepler/K2 data

Figure 6.5: The outburst light curve starting at maximum light with the fading trend subtracted. The initial section (red) shows the Early SHs. The development of the standard SHs, or Stage A is shown in green. The Stage B SHs decline in amplitude and frequency (blue). Finally, a rebrightening stage is shown in magenta.

sampled beyond the Kepler long cadence Nyquist frequency (Barclay et al. 2012). This results in a folded alias of that harmonic and this is not a true periodicity in the light curve. To test this, we took a power spectrum of a periodic function

1 with a frequency of 17.25 day− .Thepowerspectrumoftheperiodicfunctionhad

1 signiﬁcant harmonics at 34.5 day− when sampled at a fast cadence. When re-

1 sampled at the K2/Kepler cadence a folded alias appeared at 14.4 day− ,andthis alias plus the true power peak are shown in blue in Fig. 6.6. The variation in the superhump frequencies can be seen time-resolved power spectrum shown in Fig. 6.7. Here the early superhumps are seen to have a nearly constant frequency before the transition to Stage A around day 702. The Stage B superhumps drift to lower frequencies before fading in amplitude around day 715. 6.4 Analysis 125

Figure 6.6: The power spectrum of the light curve over the entire outburst (red line). For comparison, the power spectrum of a periodic function with a frequency 1 1 of 17.25 day− is plotted in blue to demonstrate that the power near 14.4 day− 1 results from a harmonic beyond the Nyquist frequency. The peak near 4 day− comes from the regular K2/Kepler thruster ﬁrings. Discovery of a new WZ Sagittae type cataclysmic variable in 126 Kepler/K2 data

Figure 6.7: The time-resolved power spectrum of the light curve starting at the peak of the outburst and plotted against BJD ( 2457000). Each power spectrum is calculated from a moving 2.7 days of data. The− strongest peak shows an Airy pattern (sidelobes) due to the sharp edges of the sample box. Early superhumps are most prominent in the ﬁrst few days, but extend into the Stage A superhumps phase. Between day 703 and 715 the Stage B superhumps decrease in frequency. 6.4 Analysis 127

6.4.2 Early Rise

The photometric cadence of K2/Kepler permits the study of the earliest phases in DN outbursts. K2/Kepler has observed only a handful of DN superoutbursts and no outbursts of a WZ Sge star until now. The fixed aperture K2/Kepler photometry of KSN:BS-C11a shows sawtooth pattern with an amplitude of 100 e/s. The sawtooth before the outburst is due to ± the drifting motion of Kepler spacecraft bringing the wings of nearby stars into the aperture. For this target the K2/Kepler motion results in image movement mainly along the lines of the CCD (see Fig. 6.2). So there is an excellent correlation between the flux variations in the aperture and the x-direction image displacement. We fit the correlation with a third order polynomial and subtract the fit from the light curve. After removing a slow trend in the flux, the corrected light curve has astandarddeviationaboutzeroofσ =10.5 e/s over the seven days before the outburst. First light of the outburst is detected at BJD2457690.71 when the flux within the aperture rises above three standard deviations of the background. As shown in Fig. 6.8,thebrightnessrisesslowlyovertheinitial10hoursafterdetection. When the system has reached about 0.5% of peak luminosity the rate of brightening dramatically increases and it takes just 23 hours to brighten by the remaining 99.5%. Atwo-componentor“broken”powerlawfitprovidesanexcellentdescriptionof the rising light curve up to 50% of the peak flux. As expected, the initial linear phase corresponds to a power law with index α =0.95 0.09 followed by a quick transition 1 ± to a steeper index of α =4.82 0.07 after 10 hours. The rate of brightening begins 2 ± to slow and deviate from the steepest power law rise after 25 hours. This initial linear rise has never been observed in a WZ Sge type star before. The transition to the steep rise occurs around 5.7 magnitudes fainter than the peak, or approximately a Kepler magnitude of 19.1 mag. For a quiescent r =21.5magnitude, the star rose by a factor of 10 in brightness during the slow linear segment. Discovery of a new WZ Sagittae type cataclysmic variable in 128 Kepler/K2 data

Figure 6.8: The rising part of the light curve plotted as a fraction of the peak flux. Right: The outburst begins slowly before a rapid rise 10 hours after first light. Extrapolating the nearly-linear early light gives a time for the onset of the outburst as BJD2457690.66. The dotted lines indicate the flux corresponding to 3 standard deviations of the pre-outburst background counts. Left: The first day± of the outburst the transient reaches 50% of its peak luminosity and the light curve is well fit by a broken power law. Initially the power law index is 1.0, but after ten hours the index steepens to an index of 4.8. At the beginning of the second day, the rise rate slows from the steep power law and the outburst finally peaks 33 hours after the onset of the burst.

6.4.3 Optical Spectrum

The GMOS spectra revealed a double-peaked Hα emission line, as seen in Fig. 6.9. Other spectral features could not be recovered from the GMOS-N spectrum (see

Fig. 6.14). The double-peaked Hα emission line is produced by orbiting gas in the accretion disk seen at a signiﬁcant inclination to our line of sight. To recover the observed disk velocity, we ﬁt the data using χ2 minimisation with a model consisting of a double Gaussian added to a linearly varying continuum. As the 900 s integration time is only a fraction of the expected 80 minute orbital period, we can not ∼ separate the systemic velocity of the binary from the orbital motion of the WD. As

1 seen in Fig. 6.9 the redshifted H component was found to peak at +450 60 km s− α ± 1 and the blue at 460 60 km s− relative to the centre of the emission. The full − ± 6.4 Analysis 129 width at half maximum (FWHM) of each of the Gaussian components was found to

1 be 740 45 km s− . ± From the spectrum, we can roughly constrain system properties, such as inclination and WD mass. As discussed in Warner (1986), the equivalent width (EW) of the Hβ emission line is inclination dependent. Expanding on this, Casares (2015) has calibrated the EW and FWHM of the Hα emission line in CVs and X-ray transients to permit the estimate of inclination and mass. To ﬁrst order, Casares (2015) make the following approximation:

B EW (6.1) ≈ cos i where EW is the equivalent width, B =9 8 A,˚ and i is the inclination. The ± value of the B parameter is calibrated from a set of well-observed CVs. From our spectrum, we ﬁnd the H EW = 88 4, which leads to an inclination i =84 5◦. α − ± ± We conclude that the inclination is fairly high, although it must be on the low side of this range as the system is not eclipsing. Discovery of a new WZ Sagittae type cataclysmic variable in 130 Kepler/K2 data

4.0

3.5

3.0

2.5

2.0 Relative ﬂux

1.5

1.0

0.5 4000 3000 2000 1000 0 1000 2000 3000 4000 1 Velocity (km s )

Figure 6.9: A Double Gaussian ﬁt to the rotationally broadened Hα proﬁle, obtained with the Gemin-Ni GMOS spectrograph on MJD 58324.3. The velocity range corresponds to a wavelength range of 6476–6651 A.˚ The two Gaussian components have 1 1 means of +460 70 km s− and 450 70 km s− . ± − ± 6.5 Discussion 131 6.5 Discussion

6.5.1 Orbital Period and Mass Ratio

Early superhumps are known to have a period within 0.1% of the binary orbital period (Kato 2015). The power spectrum starting from the peak of the light curve to day 702.4 (the onset of Stage A superhumps) provides a good estimate of the orbital period of 82.13 0.03 minutes (Fig. 6.10). As compiled by Kato (2015), the ± median orbital period of WZ Sge stars is 81.94 minutes. Ordinary superhumps are seen to develop at day 702 as the amplitude of oscillations rises from a few percent up to 8% on day 703. The delay in the start of true superhumps is 9.8 0.2 days, which is typical for WZ Sge stars with an ± 82 minute period. Stage A superhumps last only 20 to 30 cycles and are difficult to measure due to our 30 minute cadence. We apply a power spectrum analysis to the data between day 702.4 and 703.6 (20 cycles) and find a peak centred at P =84.40 0.05 minutes. Although, it is clear from the time-resolved power SH ± spectrum shown in Fig. 6.7,thattheStageAsuperhumpsevolverapidlyfromlower frequencies toward the Stage B frequencies. Fig. 6.10 also shows that the early superhumps and the Stage A oscillations are present simultaneously in the system just before the onset of Stage B. The difference in period between the Stage A superhumps and the orbital period is related to the binary mass ratio through the superhump excess parameter ϵ, where ϵ = P /P 1. We do not have an independent measurement of the SH orb − orbital period, so we use the frequency of the early superhumps to estimate ϵ.From P =84.40 minutes and P =82.13 minutes we find ϵ =0.0276 0.0005, meaning SH orb ± the precession rate of the disk is ϵ∗ = ϵ/(1 + ϵ)=0.0269. Kato & Osaki (2013b) determined a relation between the superhump excess and the components mass ratio and from their Table 1 the mass ratio for KSN:BS-C11a is q =0.070 0.005. The ± orbital period and mass ratio places KSN:BS-C11a very close to the period bounce (Fig. 6.11). The rate that the oscillation period changes with time during Stage B superhumps is also a key diagnostic of the binary parameters. Fig. 6.10 shows how the power peak trends to lower frequencies during Stage B superhumps. We estimate Discovery of a new WZ Sagittae type cataclysmic variable in 132 Kepler/K2 data

Figure 6.10: The frequency of the peak of power spectrum binned in two day time steps. The power spectrum centred on day 700.5 shows two strong peaks, so both frequencies are plotted. The ﬁrst 9 days after peak brightness display “early superhump” oscillations that generally match the binary orbital period. Stage A superhumps transition quickly to the Stage B oscillations. The superhump period over Stage B shifts to lower frequencies The measured rate of period change is 5 P˙ =6.4 10− . × 6.5 Discussion 133 the rate at which the superhump power spectrum peak shifts to lower frequencies to

2 5 beν ˙ = 0.0191 day− corresponding to a period change rate of P˙ =6.4 0.5 10− . − ± × For a collection of WZ Sge stars with mass ratios between 0.06

5 5 sured period derivative ranges between 2 10− < P<˙ 7 10− with a mean of × × 5 4 10− (Kato 2015). × Casares (2016)hasidentiﬁedarelationbetweenthequiescentHα emission line properties and the binary mass ratio for systems with q<0.25. These extremely low mass ratios are found in black hole X-ray binaries as well as CVs near the period bounce. Casares (2016)showedthattheratiobetweentheseparationinthedouble peaked Hα emission (DP) and the full width at half maximum (FWHM) of the two components is independent of the inclination of the disk, but sensitive to the mass ratio. The model predicts the relation,

1 0.49 (1 + q)− f(q)= 2/3 1/3 (6.2) 0.6+q ln (1 + q− ) DP =31/3 (1 + q)2/3 β αf (q) (6.3) FWHM 1 where α and β are free parameters. Casares (2016)estimatedthatα 0.42 and ≈ β 0.83 by ﬁtting Equation 6.3 to a set of known CVs. ≈ From our Gemini spectrum of KSN:BS-C11a we measure DP/FWHM =0.60 ± 0.02, consistent with a mass ratio of q<0.2. Using our estimated mass ratio of q =0.070 0.005, we can plot KSN:BS-C11a on the Casares (2016) relation for CVs ± as shown in Fig. 6.12. The small mass ratio we have estimated for KSN:BS-C11a is consistent with this analysis, although, the scatter in the CV relation is large.

6.5.2 Rebrightening Phase

WZ Sge stars show a variety of behaviours toward the end of a superoutburst and are generally referred to as “rebrightening”. These behaviours have been divided into classes by Imada et al. (2006)andreviewedbyKato (2015). The rebrightening phase for KSN:BS-C11a begins with a slight rise followed by a shallow dip and then keeps a nearly constant brightness over five days before a final fade. The lack of large oscillations rules out the rebrightening classes of Type B and C. KSN:BS- C11a also does not fit the Type D rebrightening class characterised by GW Lib. Discovery of a new WZ Sagittae type cataclysmic variable in 134 Kepler/K2 data

Figure 6.11: The estimated mass ratio, q, versus orbital period for the known WZ Sge stars from Kato (2015) (red points). KSN:BS-C11a is shown as a green point just after bounce. The dashed line is the standard evolutionary track for CVs from Knigge et al. (2011)whilethesolidlineistheiroptimalbinarytrack.

The behaviour of KSN:BS-C11a at the end of its outburst is probably closest to the Type A class, like AL Com, but with only a shallow dip before the plateau phase. The slight rise around day 717 suggests the rebrightening may have started before fading from the superoutburst began. This odd situation suggests that the fading begins in a diﬀerent part of the disk than the rebrightening. Superhumps are seen to reform and intensify during the rebrightening phase. 6.5 Discussion 135

0.64 Casares (2016) CVs 0.62 KSN–BS C11a

0.60

0.58 DP/FWHM

0.56

0.54

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 q

Figure 6.12: The Hα line ratio versus binary mass ratio, q, based on Fig. 2 from Casares (2016) for CVs. We add KSN:BS-C11a to the ﬁgure, which is consistent with the model parameters presented in Casares (2016)ofα =0.42 and β =0.83.

6.5.3 Early Rise

The continuous monitoring and fast cadence of the K2/Kepler and TESS missions allows us to study the earliest phases of the superoutburst in SU UMa stars. We see in Fig. 6.8 that KSN:BS-C11a shows a nearly linear precursor before the very fast rise to maximum light begins. This prompted us to look at other short period systems for evidence of precursor behaviour. We found slow linear precursors are seen in the serendipitiously discovered system KIC 4378554 (Barclay et al. 2012)as seen in Fig. 6.13.WeﬁndasimilarbrokenriseinandZChafromTESS (Court et al. 2019). The detection of slow, linear precursors in these three objects suggests that such rises may be a common feature of SU UMa outbursts.

The linear precursor in KSN:BS-C11a transitions to the fast rise at only 0.5% of the peak ﬂux of its outburst and lasts for 10 hours, or only 7 binary orbits. The linear phase of KIC 4378554 reaches 4% of the peak ﬂux but lasts for only 5.5 hours. The linear precursor in Z Cha is most impressive as it reaches to 5% of peak and lasts for 17.5 hours. Discovery of a new WZ Sagittae type cataclysmic variable in 136 Kepler/K2 data

Figure 6.13: Early rises of superoutbursts from Barclay et al. (2012)(left)and ZCha(right). Since this feature is present in diﬀerent dwarf novae types and diﬀerent instruments, we expect this to be an indication of new physics in dwarf novae superoutbursts.

If we assume the Kepler bandpass captured most of the emission during the rise (bolometric ﬂux), then the linear rise might suggest the temperature increased with time as t1/4.Atemperatureincreaseoflessthanafactoroftwocouldexplainthe factor of ten increase in ﬂux during the early phase, pushing the disk to the critical ionization temperature for the neutral hydrogen. The temperature of quiescent disks vary with radius, but typical estimates range between 3000 K to 5600 K (Rutkowski et al. 2016). So a rise of a factor of two in temperature is reasonable at the onset of an outburst.

6.5.4 Distance

The typical luminosity of WZ Sge stars in quiescence is not well determined because these objects are so intrinsically faint and only a handful have apparent magnitude brighter than 19 in quiescence. The physical origin of the ordinary superhumps suggests a fairly uniform disk luminosity for WZ Sge systems when these oscillations begin (Kato 2015). Patterson (2011) estimates the absolute magnitude for SU UMa stars near the time of superhump onset to be M =5.5 0.2mag,sowecanuse V ± this as an indicator of the distance to KSN:BS-C11a.

The apparent Kepler magnitude at superhump appearance is Kp =15.4 mag. The quiescent mag was 21.5, so the amplitude at the time at the onset of ordinary superhumps is 6.1 mag, typical of WZ Sge stars (Kato 2015). We do not know a 6.6 Conclusion 137 precise inclination for KSN:BS-C11a, but it is not an eclipsing system so we will assume no signiﬁcant inclination correction is required. The distance modulus to KSN:BS-C11a is m M =9.9mag,or 900 pc. The Galactic coordinates KSN:BS- − ∼ C11a are l =357◦ and b =12◦,extremelyclosetotheGalacticcenter.Thetotal dust extinction in the direction of KSN:BS-C11a is a substantial AV =1.03 (Schlaﬂy & Finkbeiner 2011), arguing for a distance as close as 600 pc. Given the relatively blue colour of the star in quiescence, we suspect that much of the extinction lies beyond KSN:BS-C11a, and the true distance is closer to 900 pc.

6.6 Conclusion

Searching through the K2/Kepler pixels ﬁles as part of the K2:Background Survey, we have identiﬁed a bright transient. The continuous monitoring and rapid cadence of K2/Kepler showed that the transient increased in brightness by 8 mag in about one day and faded slowly over a month. Oscillations with varying frequencies strongly suggest that the transient is a WZ Sge type dwarf nova with a binary orbital period of 82 minutes. This type of close binary lose energy by emitting gravitational radiation. The stars get closer together driving continued mass transfer mass to the WD. But mass transfer forces the secondary out of thermal equilibrium and around a period of 80 min the period “bounces” so that evolution continues with an increasing period (Knigge et al. 2011). We identify the progenitor system in quiescence using spectroscopy and deep

DECam imaging. The spectra reveal a double-peaked Hα emission line coming from the quiescent accretion disk. Its equivalent width suggests a high inclination system, although, eclipses are not seen in the outbursting light curve. The line properties are consistent with a very low mass ratio as expected for a WZ Sge system. We also ﬁnd that SU-UMa DN superoutbursts exhibit a broken power law rise. This phenomena has been observed in KSN:BS-C11a, KIC4378554, and Z Cha. The mechanism behind this broken rise is unknown, however, we expect it indicates new physics behind DN superoutbursts. Discovery of a new WZ Sagittae type cataclysmic variable in 138 Kepler/K2 data

Table 6.2: KSN:BS-C11a Characteristics

Parameter Value Uncertainty Units Outburst type A RA 16:53:50.67 0.24” J2000 DEC -24:46:26.50 0.24” J2000 Time of outburst 2457690.66 0.01 BJD Break in rise 10.0 0.2 hours Full Outburst Amplitude 8.1 0.2 mag Orbital Perioda 0.05704∼ 0.1% days Stage A Period 0.0586 0.2% days 2 Stage B,ν ˙ 0.0191 2.6% day− SH Excess, ϵ 0.0276− 1.8% Mass Ratio, q 0.070 0.01 1 Disk velocity, v 460 70 km s− Orbital inclination, i 84 5 deg Distance 900 pc ∼ a From Early Superhumps

Acknowledgements

This research weas supported by an Australian Government Research Training Pro- gram (RTP) Scholarship. This paper includes data collected by the K2 mission. Funding for the K2 mission is provided by the NASA Science Mission directorate. Based on observations obtained at the Gemini Observatory, as part of the GN- 2018A-LP-14 program, which is operated by the Association of Universities for Re- search in Astronomy, Inc., under a cooperative agreement with the NSF on behalf of the Gemini partnership: the National Science Foundation (United States), National Research Council (Canada), CONICYT (Chile), Ministerio de Ciencia, Tecnolog´ıa e Innovación Productiva (Argentina), Ministèrio da Ciência, Tecnologia e Inova¸cão (Brazil), and Korea Astronomy and Space Science Institute (Republic of Korea).

6.7 Appendix

6.7.1 GMOS-N spectrum

The full range of the GMOS-N spectrum is shown in Fig. 6.14. As the spectrum was taken while KSN:BS-C11a was in quiescence, the source is faint, with only the Hα emission clearly resolved. The A band telluric absorption line is seen from 7600–7650 A.˚ 6.7 Appendix 139

H↵

2 Relative ﬂux

4000 5000 6000 7000 8000 Wavelength (A)˚

Figure 6.14: The entire wavelength range of the GMOS-N 900 s spectrum taken on MJD58324.3. The Hα emission is indicated by the dashed vertical line, and the A band telluric absorption line is a seen from 7600–7650 A.˚ Discovery of a new WZ Sagittae type cataclysmic variable in 140 Kepler/K2 data 7 Conclusion

In this thesis I explored the short time domain of transient astronomy through developing a high-cadence UV survey telescope and by systematically searching high-cadence Kepler data for previously undiscovered transients. I summarise the conclusions of these projects and draw conclusions for the entire body of work. Finally we discuss the future for the work presented in this thesis.

7.1 Summary of GLUV

At the beginning of this thesis we analyse the feasibility of a 30 cm telescope on ahighaltitudeballoonfromatechnicalperspective,inChap.2,andascientific perspective in Chap. 3,and4. GLUV was proposed to operate from long duration balloons that float 20–30 km in the atmosphere, located within the ozone layer. The prospect of flying above a significant portion of the ozone layer presents the possibility of conducting an all sky survey at UV wavelengths to search for rare and UV bright rapid transients. If feasible, this project will open up a new window in

141 142 Conclusion astronomy to study rapid transients.

In Chap. 2 IpresentmyworkonGLUV telescope system. Conventional UV telescopes operate beyond the ozone layer, so GLUV presents a unique challenge as it is planned to fly within the ozone layer. Since the ozone layer is highly influential to the success or failure of GLUV,Ianalysethenatureoftheozonelayerandproducea model for the expected atmospheric transmission. By utilising ozone density profiles generated GOME-2 in the O3M SAF dataset (Hassinen et al. 2016), I find that seasonal variability in the ozone layer can play a key role in the determining optimal flight paths of GLUV.ThecombinationofsolarirradiationandtheBrewer–Dobson Circulation (Brewer 1949)leadstotheozonelayerrisingandreducingindensity through summer to autumn, and the converse occurring from winter to spring. Furthermore, the mean height of the ozone layer peaks at the equator and decreases towards the poles. With knowledge of these dynamics, it is clear that GLUV flights should take place between 30–90o in latitude of both hemispheres. Depending on the flight altitude there will also be a preference for time of year, with low altitudes (20–25 km) flying in summer months, and high altitudes (25–30 km) flying in winter months.

Ialsogeneratedatmospherictransmissionmapstoexplorewhatwavelengths would be accessible to GLUV.Thetransmissionprofilesweregeneratedusingthe GOME-2 ozone density profiles in conjunction with ozone cross sections from Serdyuchenko et al. (2014)andtheBeer–LambertLaw.Transmissionprofileswere generated for two pointing angles of 35o and 70o from zenith. From this analysis I find that the ozone layer should truncate the GLUV bandpass at 290 nm as the ∼ atmospheric transmission drops 20%. This suggests that GLUV will operate with an extreme U-band like filter, peaking at 300 nm. ∼ In conjunction with GLUV collaborators we developed a UV optimised spectrograph to determine if UV flux was accessible at the planned GLUV flight altitudes. The first GLUV –Pathfinder Spectrograph (GLUV –PS) was flown on the 5th of De- cember 2017 (UTC), on the YerraLoon 1 flight. Over two hours the spectrograph reached an altitude of 32 km, gathering usable spectra between 0–5 km and ∼ 20–30 km. After determining the wavelength scale for GLUV –PS, by fitting a 1.5 airmass solar spectrum to each spectra, I found that the UV flux in the 20–30 km 7.2 Summary of GLUV science cases 143 normalised spectrum was > 2 times that of the 0–5 km normalised spectrum. Al- though issues such as no pointing stability and a lack of absolute calibration limit the conclusions that we can draw from the GLUV –PS data, the result is promising. While no direct comparison can be drawn, from the GLUV –PS result, it indicates that the atmospheric transmission models discussed earlier underestimates the atmospheric transmission. Finally, in this Chapter I present a preliminary signal to noise calculator for GLUV. As many factors still remain unknown about the GLUV system, such as the detector and its environment, atmospheric transmission and sky brightness, representative values were chosen from literatures. We find that even for the modest size of 30 cm, GLUV is expected to have a limiting magnitude of 22 for a 5σ detection ∼ from a 15 min exposure. The faint limiting magnitude is promising, indicating that GLUV will be able to achieve the primary science case of detecting rare and rapid UV transients.

7.2 Summary of GLUV science cases

Throughout the development process of GLUV three science cases have informed its system requirements. The two science cases that initially drove development are discussed in Chap. 3.Thethirdsciencecasedevelopedrapidlyandformsthe basis of Chap. 4 which is presented in Ridden-Harper et al. (2017). These three science cases provide a strong case for the development of a high-cadence UV survey telescope such as GLUV. The primary science case is to detect bright shocks in supernovae to determine their progenitor systems. For SN Ia in particular, two possible systems could produce “normal” SN Ia used in cosmological analysis, which may introduce an unaccounted for systematic error. Kasen (2010)presentsawaytoobservationallydeterminethe nature of a SN Ia progenitor through the prediction of a UV shock within the first few days from explosion if it is a single degenerate system. The brightness of the UV shock produced depends on the size of the progenitor size, with 1 M producing ⊙ faint signals. With a theoretical GLUV bandpass and the signal to noise calculator, IcalculatetheexpectednumberofSNIashocksthatGLUV should expect to 144 Conclusion observe over a standard 6monthobservingcampaign.IfindthatassumingtheLi ∼ et al. (2011a)SNIarate,andallowingforallobservationangles,thatwithadaily cadence over 300 deg2, GLUV should expect to observe 1shockinteractionfrom ≥ a1M companion at a confidence of 5σ.Theresultfromthisanalysissuggeststhat ⊙ asystemlikeGLUV would quickly determine the nature of SN Ia progenitors by exploring the rapid UV time domain.

IalsooutlinethecaseforhowGLUV would be useful in exoplanet studies. With the discovery of numerous hot-Jupiters and exo-/super-Earths, there is key interest in understanding the atmospheres of these worlds. With sufficient system stability, GLUV could be used in differential photometry measurements of exoplanet atmospheres, providing a measurement of the transit depth and apparent radius at UV wavelengths. I find that it is plausible for GLUV to reach the photometric precision required for detected extended hydrogen envelopes in hot-Jupiters (e.g., Sing et al. 2015), and potentially detect ozone around exo-/super-Earths in M-dwarf systems, if they have a similar atmospheric composition to the Earth (Bétrémieux &Kaltenegger2013). Furthermore, as many exo-/super-Earths have been found orbiting M-dwarfs, GLUV would be able to provide a valuable insight into the habitability of such systems by monitoring flare activity in promising systems.

The final science case for GLUV is detecting electromagnetic counterparts of gravitational waves, and forms the basis of Chap. 4 (Ridden-Harper et al. 2017). In Ridden-Harper et al. (2017)Iexploreemissionmechanismsforallgravitational wave progenitor systems and identify areas where early UV observations can aid in constraining merger models. The weakest case for EM emission comes from binary black hole mergers, where it is predicted that emission will only be produced under specific conditions where a “fossil disk” can be reactivated by the merger (Murase et al. 2016). In this model u’-band observations have no distinct advantage over B-band observations for detecting an EM counterpart to binary black hole mergers. I did, however, find that early UV observations of kilonovae produced by binary neutron star or neutron star black hole mergers contains crucial diagnostic information that can distinguish between both the merging system and the remnant, based off models. The diagnostic capability arises from semi-relativistic jets (Metzger et al. 2015)andastrongsensitivitytotheejectamassaslanthanideserieselementsmade 7.3 Summary of the Background Survey 145 in post merger ejecta will strongly absorb UV flux. With the current emission models we predict that with a constellation of 10 GLUV sandover6months,weshould not expect to detect any counterparts from a blind search. The low detection rate is primarily a product of the low occurrence rate of such events, however, rapid follow-up of aLIGO events would allow GLUV to provide valuable information on the merging system.

These three science cases I investigated provide an excellent scientiﬁc basis for a high-cadence UV survey telescope. A system such as GLUV would have access to many key areas of astronomy by exploring the rapid time domain with high-cadence observations.

7.3 Summary of the Background Survey

In Chap. 5 IoutlinetheK2:BackgroundSurvey,whichconductsthefirstsystematic search of Kepler data to detect rapid transients. Over the course of the Kepler/K2 mission, Kepler observed 50 deg2 background pixels at 30 minute cadence and ∼ depth of 21 K .Ifindthatwithinthebackgroundpixelsalone,weshouldexpect ∼ p to detect numerous transients serendipitously observed by Kepler. With volumetric rates I find that there should be 6SNIaandbetween 0.5–6 FELTs per extra- ∼ ∼ galactic pointing K2 campaign. The analysis methods developed here are applicable to all Kepler data, allowing me to also analyse crowded galactic campaigns that transient surveys avoid. Finally, this survey is independent of transient type, so if rare or unique transients were observed by Kepler that have no analogue, K2:BS is capable of detecting them.

In Chap. 6 I present the first transient discovered by K2:BS, a WZ Sge type dwarf nova called KSN:BS-C11a. I discovered KSN:BS-C11a in superoutburst during Campaign 11 of K2,agalacticfieldthatwasSun-facingsoreceivedfewconcur- rent ground based observations and was not searched for transients. The discovery was confirmed as a galactic WZ Sge type dwarf nova with a GMOS-N spectra on 2018 July 25 (UTC) and a subsequent DECam image on 2018 September 1 (UTC), where KSN:BS-C11a was in quiescence. With the help of collaborators I found a 146 Conclusion number of intriguing properties of KSN:BS-C11a, such as: it has a quiescence magnitude of 25.1 Kp,brighteningby8.1magnitudesduringtheC11superoutburst;an 1 accretion disk with a rotational velocity of 460 70 km s− ;andanorbitalperiod ± of 80 minutes. Furthermore, I found that the rise of KSN:BS-C11a can be well ∼ characterised by a broken power law. Similar broken power law rises were also found in the dwarf novae presented in Barclay et al. (2012)thatwasobservedbyKepler and in Z Cha observed by TESS (Court et al. 2019). I conclude that broken power law rises are a common phenomena in dwarf novae superoutbursts, indicating new physics. This discovery was only possible with high-cadence and high photometric precision afforded by Kepler and TESS.

High cadence instruments, such as Kepler and TESS open up a new frontier in time domain astronomy. With K2:BS we will run the largest high-cadence survey and will place strong constraints on the rates of rapid transients, that evolve over 1day,andidentifynewphenomenaashasbeenshownwiththediscoveryof ∼ KSN:BS-C11a.

7.4 Summary of this thesis

In this thesis I explore the frontiers of time domain astronomy through the development of a new telescope system and a high-cadence transient survey. From the literature, there are many compelling science cases for exploring the short time domain, especially for transients. Many key questions within the time domain occur at UV wavelengths, which motivated the development of GLUV. Within the analysis of GLUV and the primary science cases I find that even a small UV telescope system could have profound implications for supernovae and cosmology. I also began the exploration of the rapid time domain for transients with lifetimes < 1daywiththe ∼ Kepler Space Telescope.Thissearchwillresultinmanysurprisingdiscoveries,the first of which being new physics in the onset of superoutbursts from dwarf novae. Such discoveries are only possible in the short time domain afforded to us by Kepler and now TESS. 7.5 Future work 147 7.5 Future work

Science is never completed as one answer will open many more questions, such as with this thesis. The work presented in this thesis does not conclude the topics and projects discussed within it, in this section I will discuss the future for these projects.

7.5.1 The future of GLUV

The work presented on GLUV in this thesis covers the preliminary design and science cases of the system. Through fascinating science cases and preliminary design analysis I have seen that not only is GLUV possible, it could have a substantial impact in astronomy. Following on from GLUV –PS, a robust and more sensitive spectrograph is under development at the ANU. The improved spectrograph will fly aboard the intended balloon platform for GLUV recording sky spectra during the flight duration. The data from this system will provide us with the expected sky brightness at flight altitudes and be used to test atmospheric transmission models. If successful, the GLUV project can begin construction of the telescope alongside developing the control systems and operations. The goal for GLUV places many identical instruments on balloons simultaneously, which will allow us to conduct an all sky high-cadence UV survey. The results of such a survey would provide answers to key questions in astrophysics.

7.5.2 The future of the Background Survey

The Background Survey still contains many projects. For this thesis I developed the Background Survey for the Kepler/K2 data, however, it is applicable to all Kepler and TESS data. Despite the detection fo KSN:BS-C11a, K2:BS is not completed, with much of the K2 left to be analysed. Alongside KSN:BS-C11a other promising candidates were detected, which require follow-up. One such example is shown in Fig. 7.1,wherethecandidateevolvesover 3days.Thisnewcandidateislikely ∼ galactic, as it was discovered in the galactic pointing campaign 11. The location and lifetime of this candidate suggests that it is an outburst from an undiscovered nova or dwarf nova. Follow-up images from DECam have been taken, which should 148 Conclusion

Figure 7.1: A promising K2:BS candidate discovered in C11. The fast lifetime of 3dayscombinedwiththegalacticpointingofKeplerduringC11,suggeststhat this∼ is an outburst from an undiscovered nova or dwarf nova. confirm the nature of this promising candidate discovered through K2:BS. Following the completion of K2:BS on each Campaign and all transients identified, I will calculate occurrence rates. Currently occurrence rates for transients that evolve over < 1 day are poorly defined. With the characteristics of K2:BS being well ∼ known, I will be able to define the best rates for short duration transients, galactic and extragalactic. Understanding the rates of short transients would provide valuable insight into how to construct surveys to target such phenomena. Following the K2:BS, I will also conduct Kepler Prime:BS (KP:BS). This survey will have a small survey area, due to the small pixel stamps used in the Kepler Prime mission, however, the data will have high photometric precision. The results of this survey are not expected to be as impactful as K2:BS, but it will also make use of a unique and valuable dataset. Alongside the K2:BS and KP:BS, I have developed the TESS: Background Survey (TESS:BS). As the successor to Kepler, TESS features an extensive survey area, high-cadence observations, and superb photometric precision. TESS:BS can operate on both the short and long cadence data products from TESS, allowing it to cover 7.6 Final remarks 149

50 deg2 per sector at 2 minute cadence, to m 17 and 3200 deg2 per sector ∼ ≈ ∼ at 30 minute cadence to m 19. This vast data volume will provide a wealth ≈ of objects waiting to be discovered. Although there are numerous ground based transient survey telescopes, TESS has a unique capability to discover transients that evolve over a day, or less. The discovery capability of TESS can be seen in Fig. 7.2 where TESS:BS detects an outburst of AD Mensae in the TESS short cadence data that was missed by ASAS–SN, despite near daily cadence. Despite the brighter limit of TESS compared to Kepler/K2,theenormoussurvey area of TESS means it will cover 3timesthevolumecoveredbyKepler/K2. With ∼ this large volume, I expect that numerous unique transients will be detected and will allow for TESS:BS to improve on the rate of rapid transients established by K2:BS.

14 14 TESS 15 ASAS-SN 15

16 16

17 17

18 18 Apparent magnitude Apparent magnitude

19 19

20 20 8340 8360 8380 8400 8360 8361 8362 8363 8364 8365 Time (MJD - 50000) Time (MJD - 50000)

Figure 7.2: An outburst from the nova AD Mensae, detected in the TESS short cadence mode. Despite the event being too short for high-cadence ground based surveys to detect, TESS recovers the entire outburst light curve at unparalleled time resolution.

7.6 Final remarks

Many questions remain in the short time domain, that are crucial to key questions in astronomy. Many surveys, such as Pan–STARRS, ASAS–SN, PTF, ZTF, DWF and ATLAS, are pushing to shorter cadences to explore the short time domain, however, they are not optimised for the discovery of events in the rapid time domain. In this 150 Conclusion thesis I begin the search for extremely short duration transients, with lifetimes less than a day. Many short transients are predicted to emit strongly at UV wavelengths, so I developed the concept of a low cost UV survey telescope in GLUV.Ialsoutilise data from Kepler/K2 to perform a transient survey with a unique dataset that is capable of detecting rapid transients. The steps that I have taken with Kepler into the short time domain, will be followed and superseded by TESS.Theimmensevolumeofdatathathasbeen,and will be, produced by TESS will have many transients hidden within it. With this work we began exploring the frontiers of cosmic cataclysms, and the science seems inviting. Bibliography

Abadie J., et al., 2010, Classical and Quantum Gravity, 27, 173001

Abbott B. P., et al., 2016a, Living Reviews in Relativity, 19, 1

Abbott B. P., et al., 2016b, Phys. Rev. Lett., 116, 061102

Abbott B. P., et al., 2016c, Physical Review Letters,116,061102

Abbott B. P., et al., 2016d, ApJS, 225, 8

Abbott B. P., et al., 2016e, ApJ, 832, L21

Abbott B. P., et al., 2017a, Phys. Rev. Lett., 118, 221101

Abbott B. P., et al., 2017b, Phys. Rev. Lett., 119, 161101

Abbott B. P., et al., 2017c, Nature, 551, 85

Abbott B. P., et al., 2017d, ApJ, 848, L12

Abbott B. P., et al., 2017e, ApJ, 848, L13

Abbott T. M. C., et al., 2019, ApJ, 872, L30

Andreoni I., Cooke J., 2019, in Griﬃn R. E., ed., IAU Symposium Vol. 339, Southern Horizons in Time-Domain Astronomy. pp 135–138 (arXiv:1802.01100), doi:10.1017/S1743921318002399

Andreoni I., et al., 2017, PASA, 34, e069

Andreoni I., et al., 2020, MNRAS, 491, 5852

Anglada-Escud´eG., et al., 2016, Nature, 536, 437

Aranzana E., K¨ording E., Uttley P., Scaringi S., Bloemen S., 2018, MNRAS, 476, 2501

Arcavi I., et al., 2017, ApJ, 837, L2

Arnett W. D., 1979, ApJ, 230, L37

Arnett W. D., 1982, ApJ, 253, 785

Barclay T., Still M., Jenkins J. M., Howell S. B., Roettenbacher R. M., 2012, MNRAS, 422, 1219

151 152 Bibliography

Barenboim G., Denton P. B., Oldengott I. M., 2019, Phys. Rev. D, 99, 083515

Barnes R., Lehman C., Mulla D., 2014, Computers and Geosciences, 62, 117

Basri G., Borucki W. J., Koch D., 2005a, New A Rev., 49, 478

Basri G., Borucki W. J., Koch D., 2005b, The Kepler Mission: A wide-ﬁeld transit search for terrestrial planets, doi:10.1016/j.newar.2005.08.026

Batalha N. M., 2014, Proceedings of the National Academy of Science, 111, 12647

Bauswein A., Goriely S., Janka H. T., 2013, ApJ, 773, 78

Belczynski K., Holz D. E., Bulik T., O’Shaughnessy R., 2016, Nature, 534, 512

Bellm E., 2014, in Wozniak P. R., Graham M. J., Mahabal A. A., Seaman R., eds, The Third Hot-wiring the Transient Universe Workshop. pp 27–33 (arXiv:1410.8185)

Bellm E. C., et al., 2019, PASP, 131, 068003

Bethe H. A., Wilson J. R., 1985, ApJ, 295, 14

B´etr´emieux Y., Kaltenegger L., 2013, ApJ, 772, L31

Blondin J. M., Mezzacappa A., DeMarino C., 2003, ApJ, 584, 971

Bloom J. S., et al., 2012, ApJ, 744, L17

Borucki W. J., et al., 2010, Science, 327, 977

Brewer A. W., 1949, Quarterly Journal of the Royal Meteorological Society, 75, 351

Brown T. M., 2001, ApJ, 553, 1006

Brown T. M., Latham D. W., Everett M. E., Esquerdo G. A., 2011, AJ, 142, 112

Brown P. J., et al., 2012, ApJ, 753, 22

Brown A., et al., 2015, AJ, 149, 67

Casares J., 2015, ApJ, 808, 80

Casares J., 2016, ApJ, 822, 99

Cassan A., et al., 2012, Nature, 481, 167

Cenko S. B., et al., 2017, American Astronomical Society, AAS Meeting #229, id.328.04, 229

Chambers K. C., et al., 2016, arXiv e-prints, p. arXiv:1612.05560

Chandrasekhar S., 1931, ApJ, 74, 81

Chevalier R. A., Irwin C. M., 2011, ApJ, 729, L6 Bibliography 153

Childress M., et al., 2013, ApJ, 770, 108

Chu Q., Howell E. J., Rowlinson A., Gao H., Zhang B., Tingay S. J., Bo¨er M., Wen L., 2016, MNRAS, 459, 121

Colgate S. A., McKee C., 1969, ApJ, 157, 623

Connaughton V., et al., 2016, ApJ, 826, L6

Copperwheat C. M., et al., 2015, Experimental Astronomy, 39, 119

Cornett R. H., et al., 1992, ApJ, 395, L9

Court J. M. C., et al., 2019, MNRAS, 488, 4149

Coward D. M., et al., 2011, MNRAS, 415, L26

Coward D. M., et al., 2012, MNRAS, 425, 2668

Crossﬁeld I. J. M., 2015, PASP, 127, 941

DAWG (Data Analysis Working Group) 2012, Kepler: A Search for Terrestrial Planets Kepler Archive Manual

D’Eramo F., Ferreira R. Z., Notari A., Bernal J. L., 2018, J. Cosmology Astropart. Phys., 2018, 014

D’Orazio D. J., Levin J., Murray N. W., Price L., 2016, Phys. Rev. D, 94, 023001

Davenport J. R. A., 2016, ApJ, 829, 23

Di Valentino E., Melchiorri A., Mena O., Vagnozzi S., 2019, arXiv e-prints, p. arXiv:1908.04281

Dimitriadis G., et al., 2019a, ApJ, 870, L1

Dimitriadis G., et al., 2019b, ApJ, 870, L14

Dominik M., et al., 2015, ApJ, 806, 263

Drout M. R., et al., 2014, ApJ, 794, 23

Drout M. R., et al., 2017, Science, 358, 1570

Eastman R. G., Schmidt B. P., Kirshner R., 1996, ApJ, 466, 911

Evans P. A., et al., 2016, MNRAS, 460, L40

Falk S. W., 1978a, The Astrophysical Journal,225,L133

Falk S. W., 1978b, ApJ, 225, L133

Fausnaugh M. M., et al., 2019, arXiv e-prints, p. arXiv:1904.02171

Fern´andez R., Metzger B. D., 2016, Annual Review of Nuclear and Particle Science, 66, 23 154 Bibliography

Fern´andez R., Foucart F., Kasen D., Lippuner J., Desai D., Roberts L. F., 2017, Classical and Quantum Gravity, 34, 154001

Fong W., Berger E., Margutti R., Zauderer B. A., 2015, ApJ, 815, 102

France K., et al., 2013, ApJ, 763, 149

Galama T. J., et al., 1998, Nature, 395, 670

Garnavich P. M., Tucker B. E., Rest A., Shaya E. J., Olling R. P., Kasen D., Villar A., 2016, ApJ, 820, 23

Gehrels N., et al., 2004, ApJ, 611, 1005

Gezari S., et al., 2015, ApJ, 804, 28

Giacomazzo B., Zrake J., Duﬀell P. C., MacFadyen A. I., Perna R., 2015, ApJ, 809, 39

Gillon M., et al., 2017, Nature, 542, 456

Gold R., Bernuzzi S., Thierfelder M., Br¨ugmann B., Pretorius F., 2012, Phys. Rev. D, 86, 121501

Goobar A., et al., 2014, ApJ, 784, L12

Guetta D., Piran T., Waxman E., 2005, ApJ, 619, 412

G¨unther M. N., et al., 2020, AJ, 159, 60

Hassinen S., et al., 2016, Atmospheric Measurement Techniques, 9, 383

Hawley S. L., Pettersen B. R., 1991, ApJ, 378, 725

Heger A., Fryer C. L., Woosley S. E., Langer N., Hartmann D. H., 2003, ApJ, 591, 288

Herse M., 1979, Sol. Phys., 63, 35

H¯oshi R., 1979, Progress of Theoretical Physics, 61, 1307

Hosseinzadeh G., et al., 2017, ApJ, 845, L11

Howard W. S., Corbett H., Law N. M., RatzloﬀJ. K., Glazier A., Fors O., del Ser D., Haislip J., 2019, ApJ, 881, 9

Howell S. B., et al., 2014, PASP, 126, 398

Hoyle F., Fowler W. A., 1960, ApJ, 132, 565

Huber D., et al., 2016, ApJS, 224, 2

Iben I., Tutukov A. V., 1984, The Astrophysical Journal Supplement Series, 54, 335 Bibliography 155

Imada A., Kato T., Monard L. A. G., Retter A., Liu A., Nogami D., 2006, PASJ, 58, 383

Ioka K., Matsumoto T., Teraki Y., Kashiyama K., Murase K., 2017, MNRAS, 470, 3332

Ishioka R., et al., 2002, A&A, 381, L41

Ivezic Z., et al., 2008, arXiv.org, 0805, 2366

Japelj J., Gomboc A., 2011, PASP, 123, 1034

Jin Z.-P., Li X., Cano Z., Covino S., Fan Y.-Z., Wei D.-M., 2015, ApJ, 811, L22

Just O., Bauswein A., Ardevol Pulpillo R., Goriely S., Janka H. T., 2015, MNRAS, 448, 541

Kasen D., 2010, ApJ, 708, 1025

Kasen D., Badnell N. R., Barnes J., 2013, ApJ, 774, 25

Kasen D., Fern´andez R., Metzger B. D., 2015, MNRAS, 450, 1777

Kasen D., Metzger B., Barnes J., Quataert E., Ramirez-Ruiz E., 2017, Nature, 551, 80

Kashiyama K., Quataert E., 2015, MNRAS, 451, 2656

Kato T., 2002, PASJ, 54, L11

Kato T., 2015, PASJ, 67, 108

Kato T., Osaki Y., 2013a, PASJ, 65, 97

Kato T., Osaki Y., 2013b, PASJ, 65, 115

Kato T., Osaki Y., 2013c, PASJ, 65, L13

Kato T., et al., 2014, PASJ, 66, 90

Kawaguchi K., Kyutoku K., Nakano H., Okawa H., Shibata M., Taniguchi K., 2015, Phys. Rev. D, 92, 024014

Kawaguchi K., Kyutoku K., Shibata M., Tanaka M., 2016, The Astrophysical Journal,825,52

Kelley L. Z., Mandel I., Ramirez-Ruiz E., 2013, Phys. Rev. D, 87, 123004

Kelly P. L., et al., 2014, ApJ, 790, 3

Kerzendorf W. E., Schmidt B. P., Laird J. B., Podsiadlowski P., Bessell M. S., 2012, ApJ, 759, 7

Kessler R., et al., 2009, ApJS, 185, 32

Kim C., Perera B. B. P., McLaughlin M. A., 2015, MNRAS, 448, 928 156 Bibliography

Kimura S. S., Takahashi S. Z., Toma K., 2017, MNRAS, 465, 4406

Kiuchi K., Sekiguchi Y., Kyutoku K., Shibata M., Taniguchi K., Wada T., 2015, Phys. Rev. D, 92, 064034

Klein R. I., Chevalier R. A., 1978, ApJ, 223, L109

Knigge C., Baraﬀe I., Patterson J., 2011, ApJS, 194, 28

Koch D. G., et al., 2010, ApJ, 713, L79

Kochanek C. S., et al., 2017, PASP, 129, 104502

Kolb U., Baraﬀe I., 1999, MNRAS, 309, 1034

Kraut A., et al., 2008, Sub-arcsecond pointing for balloon-borne telescopes. p. 70123Q, doi:10.1117/12.789138

Kreisch C. D., Cyr-Racine F.-Y., Dor´eO., 2019, arXiv e-prints, p. arXiv:1902.00534

LSST Science Collaboration et al., 2009, arXiv e-prints, p. arXiv:0912.0201

Lazzati D., Deich A., Morsony B. J., Workman J. C., 2017, MNRAS, 471, 1652

Lecavelier Des Etangs A., Vidal-Madjar A., D´esert J. M., Sing D., 2008, A&A, 485, 865

Levan A., Crowther P., de Grijs R., Langer N., Xu D., Yoon S.-C., 2016, Space Sci. Rev., 202, 33

Li S.-Z., Yu Y.-W., 2016, ApJ, 819, 120

Li W., et al., 2011a, MNRAS, 412, 1441

Li W., Chornock R., Leaman J., Filippenko A. V., Poznanski D., Wang X., Gane- shalingam M., Mannucci F., 2011b, MNRAS, 412, 1473

Li W., et al., 2019, ApJ, 870, 12

Lightkurve Collaboration et al., 2018, Lightkurve: Kepler and TESS time series analysis in Python, Astrophysics Source Code Library (ascl:1812.013)

Loyd R. O. P., Shkolnik E. L., Schneider A. C., Barman T. S., Meadows V. S., Pagano I., Peacock S., 2018, ApJ, 867, 70

Lubow S. H., 1991a, ApJ, 381, 259

Lubow S. H., 1991b, ApJ, 381, 268

Luger R., Agol E., Kruse E., Barnes R., Becker A., Foreman-Mackey D., Deming D., 2016, AJ, 152, 100

Mallonn M., Strassmeier K. G., 2016, A&A, 590, A100

Marion G. H., et al., 2016, ApJ, 820, 92 Bibliography 157

Marshall P., et al., 2017, Lsst Science Collaborations Observing Strategy White Paper: “Science-Driven Optimization Of The Lsst Observing Strategy”, doi:10.5281/zenodo.842713

Martin D. C., et al., 2005, ApJ, 619, L1

Mayor M., Queloz D., 1995, Nature, 378, 355

McElroy C., Fogal P., 2008, Atmosphere-Ocean,46,1

Metzger B. D., Piro A. L., 2014, MNRAS, 439, 3916

Metzger B. D., Bauswein A., Goriely S., Kasen D., 2015, MNRAS, 446, 1115

Miknaitis G., et al., 2007, ApJ, 666, 674

M¨ortsell E., Dhawan S., 2018, J. Cosmology Astropart. Phys., 2018, 025

Murase K., Kashiyama K., M´esz´aros P., Shoemaker I., Senno N., 2016, ApJ, 822, L9

Nagpal L., Samdani K., 2017, in 2017 2nd IEEE International Conference on Re- cent Trends in Electronics, Information & Communication Technology (RTEICT). IEEE, pp 1778–1784, doi:10.1109/RTEICT.2017.8256905, http://ieeexplore. ieee.org/document/8256905/

Nakar E., Sari R., 2010, ApJ, 725, 904

Nicuesa Guelbenzu A., et al., 2012, A&A, 548, A101

Nishizawa A., 2016, Phys. Rev. D, 93, 124036

O’Malley-James J. T., Kaltenegger L., 2017, MNRAS, 469, L26

O’Malley-James J. T., Kaltenegger L., 2019, MNRAS, 485, 5598

Ofek E. O., et al., 2010, ApJ, 724, 1396

Olling R. P., et al., 2015, Nature, 521, 332

Osaki Y., 1974, Publications of the Astronomical Society of Japan,26,429

Osaki Y., 1989, Publications of the Astronomical Society of Japan, 41, 1005

Osaki Y., Meyer F., 2002, A&A, 383, 574

Pandey K. L., Karwal T., Das S., 2019, arXiv e-prints, p. arXiv:1902.10636

Paschalidis V., Etienne Z. B., Shapiro S. L., 2013, Phys. Rev. D, 88, 021504

Paschalidis V., Ruiz M., Shapiro S. L., 2015, ApJ, 806, L14

Patterson J., 2011, MNRAS, 411, 2695

Pearson K. A., Palafox L., Griﬃth C. A., 2018, MNRAS, 474, 478 158 Bibliography

Perlmutter S., et al., 1999, ApJ, 517, 565

Perna R., Duﬀell P., Cantiello M., MacFadyen A. I., 2014, ApJ, 781, 119

Perna R., Lazzati D., Giacomazzo B., 2016, ApJ, 821, L18

Petrillo C. E., Dietz A., Cavagli`aM., 2013, ApJ, 767, 140

Phillips M. M., 1993, ApJ, 413, L105

Pintr P., Peˇrinov´aV., LukˇsA., Pathak A., 2014, Planet. Space Sci., 99, 1

Piro A. L., Morozova V. S., 2016, ApJ, 826, 96

Piro A. L., Nakar E., 2014, ApJ, 784, 85

Planck Collaboration et al., 2018, arXiv e-prints, p. arXiv:1807.06209

Pont F., Sing D. K., Gibson N. P., Aigrain S., Henry G., Husnoo N., 2013, MNRAS, 432, 2917

Poznanski D., et al., 2009, ApJ, 694, 1067

Pritchard T. A., Roming P. W. A., Brown P. J., Bayless A. J., Frey L. H., 2014, ApJ, 787, 157

Rabinak I., Waxman E., 2011, ApJ, 728, 63

Rees M. J., Meszaros P., 1992, MNRAS, 258, 41

Rest A., et al., 2014, ApJ, 795, 44

Rest A., et al., 2018, Nature Astronomy, 2, 307

Ricker G. R., et al., 2015, Journal of Astronomical Telescopes, Instruments, and Systems, 1, 014003

Ridden-Harper R., Tucker B. E., Sharp R., Gilbert J., Petkovic M., 2017, Monthly Notices of the Royal Astronomical Society,472,4521

Ridden-Harper R., et al., 2019, MNRAS, 490, 5551

Ridden-Harper R., Tucker B. E., Gully-Santiago M., Barentsen G., Rest A., Gar- navich P., Shaya E., 2020, Monthly Notices of the Royal Astronomical Society, submitted

Riess A. G., et al., 1998, The Astronomical Journal,116,1009

Riess A. G., et al., 1999, AJ, 118, 2675

Riess A. G., Casertano S., Yuan W., Macri L. M., Scolnic D., 2019, ApJ, 876, 85

Rigault M., et al., 2018, arXiv e-prints, p. arXiv:1806.03849

Roming P. W. A., et al., 2009, ApJ, 690, 163 Bibliography 159

Rowan-Robinson M., 1999, Fireworks in a Dark Universe,1,3

Rutkowski A., Waniak W., Preston G., Pych W., 2016, MNRAS, 463, 3290

Ryan G., van Eerten H., Piro L., Troja E., 2019, arXiv e-prints, p. arXiv:1909.11691

Sagiv I., et al., 2014, AJ, 147, 79

Sako M., et al., 2018, PASP, 130, 064002

Sand D. J., et al., 2018, ApJ, 863, 24

Scalzo R. A., et al., 2017, PASA, 34, e030

Schaefer B. E., Pagnotta A., 2012, Nature, 481, 164

Schawinski K., et al., 2008, Science, 321, 223

Schlaﬂy E. F., Finkbeiner D. P., 2011, ApJ, 737, 103

Scolnic D., et al., 2018a, ApJ, 852, L3

Scolnic D. M., et al., 2018b, ApJ, 859, 101

Serdyuchenko A., Gorshelev V., Weber M., Chehade W., Burrows J. P., 2014, Atmospheric Measurement Techniques, 7, 625

Shappee B. J., et al., 2019, ApJ, 870, 13

Sharp R., Tucker B., Ridden-Harper R., Bloxham G., Petkovic M., 2016, GLUV: aballoon-bornehigh-cadenceultravioletmonitoringtelescopeforsupernovashock breakouts and exoplanet atmospheres. p. 99080V, doi:10.1117/12.2231555

Shields A. L., Ballard S., Johnson J. A., 2016, Phys. Rep., 663, 1

Shustov B., et al., 2018, Ap&SS, 363, 62

Siellez K., Bo¨er M., Gendre B., 2014, MNRAS, 437, 649

Sing D. K., et al., 2011, MNRAS, 416, 1443

Sing D. K., et al., 2015, MNRAS, 446, 2428

Sing D. K., et al., 2016a, Nature,529,59

Sing D. K., et al., 2016b, Nature, 529, 59

Smartt S. J., 2015a, PASA, 32, e016

Smartt S. J., 2015b, Publications of the Astronomical Society of Australia,32,e016

Smith K. W., et al., 2018, The Astronomer’s Telegram, 11661, 1

Southworth J., Bruni I., Mancini L., Gregorio J., 2012, MNRAS, 420, 2580

Southworth J., Mancini L., Madhusudhan N., Molli`ere P., Ciceri S., Henning T., 2017, AJ, 153, 191 160 Bibliography

Staath E., Lemaire P., 1995, Astronomy & Astrophysics, 295, 517

Stephens B. C., East W. E., Pretorius F., 2011, ApJ, 737, L5

Still M., Barclay T., 2012, PyKE: Reduction and analysis of Kepler Simple Aper- ture Photometry data, Astrophysics Source Code Library (ascl:1208.004)

Sullivan M., et al., 2011, ApJ, 737, 102

Surman R., McLaughlin G. C., Ruﬀert M., Janka H. T., Hix W. R., 2008, ApJ, 679, L117

Takami K., Rezzolla L., Baiotti L., 2014, Phys. Rev. Lett., 113, 091104

Tanaka M., 2016, Advances in Astronomy,2016,1

Tanaka M., Hotokezaka K., 2013, ApJ, 775, 113

Tanaka M., Hotokezaka K., Kyutoku K., Wanajo S., Kiuchi K., Sekiguchi Y., Shibata M., 2014, ApJ, 780, 31

Tanvir N. R., Levan A. J., Fruchter A. S., Hjorth J., Hounsell R. A., Wiersema K., Tunnicliﬀe R. L., 2013, Nature, 500, 547

Tilley M. A., Segura A., Meadows V., Hawley S., Davenport J., 2019, Astrobiology, 19, 64

Tonry J. L., et al., 2018, PASP, 130, 064505

Tucker M. A., Shappee B. J., Wisniewski J. P., 2019, ApJ, 872, L22

Turner J. D., et al., 2016, MNRAS, 459, 789

Vanderburg A., Johnson J. A., 2014, PASP, 126, 948

Vangioni E., Goriely S., Daigne F., Fran¸cois P., Belczynski K., 2016, MNRAS, 455, 17

Varoquaux G., Buitinck L., Louppe G., Grisel O., Pedregosa F., Mueller A., 2015, GetMobile: Mobile Computing and Communications,19,29

Vattis K., Koushiappas S. M., Loeb A., 2019, Phys. Rev. D, 99, 121302

Vican L., et al., 2011, PASP, 123, 1156

Villar V. A., et al., 2017, ApJ, 851, L21

Villar A., Rest A., Garnavich P., Shaya E., Narayan G., Ridden-Harper R., in prep, The Astrophysical Journal, in prep

Vin´ıcius Z., Barentsen G., Gully-Santiago M., Cody A. M., Hedges C., Still M., Bar- clay T., 2017, KeplerGO/PyKE, doi:10.5281/zenodo.835583, https://doi.org/ 10.5281/zenodo.835583

Virtanen P., et al., 2019, arXiv e-prints, p. arXiv:1907.10121 Bibliography 161

Waller W., Stecher T. P., 1998, Symposium - International Astronomical Union, 179, 147

Warner B., 1986, MNRAS, 222, 11

Wenger M., et al., 2000, A&AS, 143, 9

Whelan J., Iben Icko J., 1973, ApJ, 186, 1007

Whitehurst R., 1988, MNRAS, 232, 35

Woosley S. E., Taam R. E., Weaver T. A., 1986, ApJ, 301, 601

Yang B., et al., 2015, Nature Communications, 6, 7323

Yang H., et al., 2017, ApJ, 849, 36

Yu Y.-W., Zhang B., Gao H., 2013, ApJ, 776, L40

Yuan X., et al., 2008, Chinese Small Telescope ARray (CSTAR) for Antarctic Dome A. p. 70124G, doi:10.1117/12.788748

de Mink S. E., Belczynski K., 2015, ApJ, 814, 58