Superluminous Supernovae

What am I being asked to do?

We are asking you to help us look for potential superluminous supernovae in data gathered by the Zwicky Transient Facility (ZTF). The data is a mixture of archival lightcurves gathered over the past 3 years of ZTF operations along with daily updates of new lightcurves gathered within the previous 24 hours.

How do I tell if the lightcurve has been 'rising for more than 20 days'?

The x-axis units are in days and the lightcurve is plotted such that the first detection associated with the transient is at 0 days. This is to try to make the determination of the rise easier. However, figuring out the duration of the rise can still be tricky because, for example, we might not have detected the supernova until several days after the explosion. We have detailed instructions on how to resolve this and similar questions in the Field Guide under the 'Common challenges and how to solve them' section.

How do I tell if it is 'close to a faint, fuzzy galaxy'?

Good question and it is a little subjective. Although we have done our best to ensure that the lightcurves you are shown are of supernovae, other types of variable sources (things that vary in brightness over time) will get into the project. This question is partly designed to weed out those objects. Fuzzy is designed to remove any variable stars since those will appear as sharp point sources in the images and other supernovae types are expected in brighter galaxies. Faint is to focus on the fact that superluminous supernovae usually only appear to explode in faint galaxies whereas some other supernovae types are more often observed in spiral galaxies. We further require that the crosshairs are close to the faint, fuzzy galaxy since a superluminous supernova can explode anywhere in the galaxy, but variable sources like active galactic nuclei and tidal disruption events are only observed near the core of a galaxy. For more details check out the 'Common challenges and how to solve them' section of the Field Guide.

What are the numbers on the graph axes showing?

The graphs plot time from left to right, so more recent observations appear on the right. The observation that triggered our systems to pay attention will be in the middle, at day zero. The higher the point is, the brighter the object appeared. The numbers give the brightness in magnitudes; smaller numbers mean a brighter object. Magnitudes are explained here.

Where can I see more examples?

To see more examples and additional explanations, check out the Field Guide to the right of your screen in the classification interface. There are also some examples available by clicking the "Need some help with this task?" button in the classification interface.

Why can't a computer do this?

There are many nuances in the data that can make the identification of the lightcurve rise time difficult. Answering the second question also requires human reasoning about the close, faint and fuzzy criteria. To automate this process there are two approaches to we could try. We could either attempt to write a traditional computer program or try to train a machine learning algorithm. For our research goal, it is important for us to try and find all the SLSNe in this data set. But to write a computer program we would need to be able to write down a set of rules that filter out the SLSNe from all the other types. You can see how it could be difficult to determine these rules such that we capture all the SLSNe but don't incorrectly characterise other types as SLSNe. In fact, to a certain extent, we have used computer programs to whittle down the sample we are asking you to review to only contain those lightcurves we think are likely supernovae. We do not currently have a good handle on how to narrow the search further without the risk of discarding potential SLSNe. But what about machine learning? Machine learning is an attractive alternative, it would circumvent the need to manually write down the rules for identifying SLSNe. The 'rules' would instead be learned by an algorithm as it inspects the lightcurves and host galaxies of supernovae and SLSNe. The problem for this project is that the algorithms available today tend to require large numbers of examples in order to learn the differences, but SLSNe are very rare and to date, we have only discovered about 100 examples. One aim of this project is to use your classifications to help us towards automating the process and narrowing the search.

Why do you show the results from both red and blue filters?

Different objects may show different behaviour when viewed with different colours. If a variable is surrounded by dust, for example, then it may appear brighter in the red than the blue, so we'd like you to consider both curves independently. If either graph matches what we're looking for, then record it as a supernova.

What are supernovae?

Supernovae are stars that have exploded. There are different types of supernovae and each type has different observational characteristics and can tell us about the evolution of the star that exploded. They disburse into space all of the chemical elements that were produced inside their progenitor stars, including the elements essential for making planets and life. You can learn more about supernovae here.

What are superluminous supernovae?

Superluminous supernovae (SLSNe) are stellar explosions that are at least ten times more powerful than the typical supernovae, making them the brightest explosions in the known universe. They are much rarer than other types of supernovae making finding them harder and very scientifically valuable. You can read more about them on the project Research page.

Where do the lightcurves come from?

The lightcurves are from the Zwicky Transient Facility(ZTF). ZTF observes large areas of the sky every night producing many detections of supernovae of the past 3 years of operations.

Where do the images come from?

The images are taken from the Panoramic Survey Telescope and Rapid Response System (Pan-STARRS). Pan-STARRS has produced a large, high-quality image of the entire sky north of -30 degrees declination meaning we can always access an image of the location of a variable source if it lies in this area. This can help us determine the host galaxies and type of the transients (anything that brightens then fades away) we find.

Why do we want to find superluminous supernovae?

For starters, we aren't exactly sure how they form. Over the years, multiple models have been proposed to describe what scenarios may produce an SLSN. These include core-collapse in very massive stars, millisecond magnetars, interaction with dense circumstellar material, or pair-instability supernovae. As of yet no single model can explain every SLSNe found, but the models have suggested that stars 40 times more massive than our Sun are likely to form SLSNe. The aim of this project is to find more examples of SLSNe that we can use to help us determine which of these models are most likely to explain what we observe.

What happens if I find a superluminous supernova?

Since ZTF is an active survey, the project contains both historical lightcurves from the past 3 years of operations and a continuous stream of new or updated lightcurves generated each night. Any SLSN discovered during the search of the historical lightcurves will feed into our analysis of SLSN rates informing our understanding of how often we should expect to find SLSNe in the future. If an SLSN is discovered in the newer data then it will also feed into the analysis of the rates, but depending on the phase (time since the initial explosion) of the SLSN its discovery could trigger followup observations where we use other telescopes to gather additional information about the event. We will keep a track record of SLSNe discovered on the project's Results page.

I've been asked to classify a lightcurve with the same name more than once. What's going on?

We expect this. Since ZTF is an active survey, it is constantly generating new data. ZTF is a high cadence survey meaning that on average the same patch of will be observed about 300 days a year. This means that the lightcurves we add one day may have additional lightcurve points the next day, or a few days later. If a new lightcurve point is received we will add a new subject to the project with the updated lightcurve. These additional lightcurve points might prove crucial for the SLSN or not classification. By leaving the older lightcurves in the project until they are retired, we will be able to track how these new lightcurve points influence volunteer classifications. This project will be a pathfinder for future lightcurve projects and this information will help inform how we should approach these types of project in the future.

How do I know if I'm doing this right?

Humans are really good at recognising patterns. We expect most lightcurves are not superluminous supernovae. If you are worried then have a look in the Field Guide to the right of your screen for some guidance. If you are still unsure make your best guess then click "Done & Talk" and let us know how you classified the image and we can try and give you guidance for similar cases in the future.

What if I made a mistake?

We ask multiple people for classifications of each lightcurve. So if you really think you made a mistake don't worry, this is the beauty of crowdsourcing, we all make mistakes that others can compensate for while we can compensate for their mistakes.

Didn't find an answer? Ask our community of volunteers and researchers on Talk.