Dr Philipp Jordt | Finding the Best Way for Large Research Facilities to Handle All Their Data

As technology advances, physical experiments are performed by bigger, more powerful, and more sophisticated machines. Research conducted using these large-scale facilities (LSF) typically involves collaboration between the operating staff, instrument scientists, and external research groups, who propose specific experiments to be performed. With many people involved and such large volumes of data generated by each experiment, it has become challenging to make sure that results are collected, catalogued, stored, and distributed efficiently.

The German consortium DAPHNE4NFDI has been working to integrate Electronic Laboratory Notebooks (ELNs) into the Photon and Neutron (PaN) research community. One of their main goals has been to develop the decision-making process behind how research facilities, universities, and researchers can evaluate the range of ELNs available and decide on which solution to integrate into their operations.

The Age of Big Data

We live in the Age of Data. It’s estimated that over 90% of all the world’s data have been generated just in the last 2 years, and this exponential increase shows no signs of slowing down. Among some of the most impressive data generators in the world are LSFs. PETRA III at DESY, Hamburg, is one of the world’s biggest and brightest storage ring X-ray sources and produces nearly 5 petabytes of data every year. In nearby Schenefeld, at the European X-ray Free Electron Laser, 30 petabytes of scientific data were acquired in 2024, with data rates up to hundreds of gigabits per second for a single large area X-ray detector. These two facilities combined have generated enough data to fill up the hard drives of nearly 20,000 typical home computers —and these are just two examples from large scale facilities. If you include all the other experiments at the many other LSFs across Europe, the amount of data physical science produces each year is almost unfathomable.

With so much data being generated, it’s imperative that research facilities have the appropriate software in place to deal with such volume. Gone are the days where pen and paper could do an adequate job. This is where ELNs come in. In the last few years, the number of ELNs available has considerably increased but many are aimed at individual researchers. This means that LSFs now have a range of solutions to choose from, but it leads to a new question: how does one decide which ELN is the most appropriate for a LSF research community?

Making Data More FAIR

The Data from Photon and Neutron Experiments (DAPHNE4NFDI) consortium led by Dr Bridget Murphy is a member of the German National Research Data Infrastructure (NFDI), whose aim is to develop and improve the research data management tools available to LSFs. The DAPHNE4NFDI team have developed a workflow which facilities can use to decide which of the many ELNs out there would be most suitable to adopt.

The main goal of implementing an ELN is to help record experiment details and results according to ‘FAIR’ principles: these letters stand for Findable, Accessible, Interoperable, and Reusable. Data is Findable if it’s straightforward for a human or machine to locate it; this is especially important when such large quantities of it are generated daily. Once it’s found, that data should be easily Accessible by those who need to access it. The ideal is then for the data to be Interoperable — that it can be straightforwardly integrated with other data or applications. Finally, it’s desirable for data to be as Reusable as possible. It shouldn’t just vanish into the ether after experiments are performed.

Successful Deployment in X-Ray Reflectivity

The DAPHNE4NFDI team have already successfully engineered the implementation of ELNs – in the field of X-ray reflectivity. This is the area in which Kiel University researchers at DAPHNE4NFDI work. In these experiments, X-rays are used to study the properties of surfaces (or, more technically, interfaces) of solid materials and liquids, such as magnetic materials for sensors or lipid membranes as potential biological switches. Understanding the character of the interface on the nanoscale is critical to understanding how the material or device is going to behave. When X-rays hit the interface, if the angle of incidence is smaller than a certain critical angle and the surface is flat, the X-rays are almost totally reflected. By changing the angle of incidence, one can vary the depth to which the X-rays can penetrate into the sample and analyse their thickness, density, and roughness on the nanoscale. To extract this information from the measured data some advanced analysis tools have to be used, eg. neural networks, as developed within DAPHNE4NFDI by Prof. Frank Schreibers group at Tübingen university.

One of the key advantages of using an ELN is the potential for supervisors, collaborators, and research students to collaborate during experiments, even when the team is at diverse locations. With the enhanced productivity ELNs have shown for X-ray reflectivity experiments, Dr Jordt and the DAPHNE4NFDI team are now looking to extend the use of ELNs into other communities in collaboration with German LSFs, such as DESY in Hamburg, and MLZ in Munich.

How to Choose an ELN Using Specifications

One of the key questions the DAPHNE4NFDI team wanted to address was how to choose what ELN to use out of the currently available options. Initially, they wanted to assess the ‘suitability’ and ‘maturity’ of the available solutions, including two different ELN approaches such as the science driven, bottom-up, snip logbook by Dr. Markus Osterhoff from Göttingen University, and the facility provisioned MLZ-ELN tested by Dr. Yuliia Tymoshenko from the neutron scattering group at KIT. An ELN is suitable if it’s a good fit for the workflows and procedures at the facility, and it’s mature if it’s a well-developed, technically proficient software package. In general, many ELNs will be mature in that they operate effectively, but not all of them may be suitable for use in a LSF due to the demanding environment. Out of the 12 ELNs identified, 82% were considered to be mature and 65% were suitable. Overall, this led to 53% of ELNs that were both suitable and mature, meaning they would be satisfactory for potential deployment at PaN facilities.

But the DAPHNE4NFDI team were interested in performing a more detailed analysis, one that took more carefully into account the individual features characteristic of a good ELN. They drew up a list of 31 different specifications; these are individual features that are necessary for an ELN to improve productivity at the facility. Some of the most important specifications included: intuitive user interfaces, auto-saving, parallel editing, and the ability to manage access rights.

With these specifications in place, Dr Jordt and his colleagues asked some of the senior scientists to provide a ‘satisfaction’ rating. They then devised a formula to calculate an overall rating for each ELN –the Figure of Merit (FOM). The FOM would be a number between 0 and 1, with a FOM of 0 indicates that one or more essential features are missing or unsatisfactory from a given ELN. A cutoff value of 0.6 was pre-selected to determine ELN suitability and only 2 out of 12 ELNs passed the cutoff – that’s a pass rate of 17% compared with the 53% from the previous overall survey due to the lack of critical components.

Future Integration into the Photon and Neutron Community

One of the most intriguing results from the DAPHNE4NFDI team’s research is that ELNs are much more likely to be deemed suitable when evaluated wholesale than when evaluated by rating each specification one at a time. This suggests that researchers providing evaluations may not have all the important necessary features in mind when they’re asked to rank just the overall suitability of the ELN. As a result, the DAPHNE4NFDI team suggest that rating each individual specification and computing a FOM score may provide a less biased, more accurate way to determine whether an ELN is suitable for use in a large research facility.

But the FOM shouldn’t be the one and only arbiter of an ELN’s suitability. The DAPHNE4NFDI team suggest that, while the calculation of an FOM is a good guide and an essential step in the decision-making process, once a selection of ELNs have passed, a series of detailed discussions should follow, considering critical functionality and requirements before deciding on which ELN to finally adopt. For the PaN community, the DAPHNE4NFDI team’s work has helped to narrow down the options and get closer to making a final decision. But beyond the PaN field, it also offers a model workflow for the adoption of ELNs.