Introduction

In March 2019 the ALD/Database was introduced to the ALD community on www.atomiclimits.com. The ALD/Database is an initiative by Harm Knoops, Bart Macco, Vincent Vandalon and Erwin Kessels and offers a centralized platform for the ALD community to provide an overview of papers reporting on ALD processes and their characteristics. New papers can be added to the database by visitors – and as such the site is a true community-driven effort – which will be reviewed to keep the database consistent and ensure that only papers with an appropriate scope are included.

We would like to seize the end of 2019 as an opportunity to look back and reflect on the first year of the database. In this blog post we will also provide context of its inception, insight into how the ALD/Database has been growing, and explain the review process of the new submissions in more detail.

Why did we start the database?

We identified the need for a central and up-to-date “database” of papers and publications a couple years ago. At the start of 2019, more concrete plans were formulated and the website was built / programmed by Jan Buiter. Practically speaking, we wanted the ALD/Database to provide: (1) A “periodic table” of the materials showing at a glance which materials have been synthesized by ALD. (2) Allow a visitor of the site to quickly track down the relevant papers describing how to perform a specific ALD process. (3) Enable visitors to submit new entries for the database.

The new entries submitted by the community to the ALD/Database are reviewed before they are shown on the website. The main acceptance criteria are the correctness / consistency of the data entered in the database (e.g. naming of the reactants) and whether there are sufficient processing details in the paper. Currently the entries are reviewed by Harm Knoops, Joseph P. Klesko (joining the team mid 2019), Bart Macco, and Vincent Vandalon.  See the inset for a more detailed description of the type of publications we want to collect in the database.

How does the ALD/Database work?

An overview of all the processes in the ALD/Database are shown graphically using an annotated periodic table. This user interface was inspired by a figure in the review paper by Puurunen [J. Appl. Phys. 97, 121301 (2005)]. By clicking on an element in the ALD/Database, a list is shown of papers reporting on the synthesis of films containing this element. This allows a visitor to quickly determine if a material has already been grown by ALD and to easily find the appropriate references describing the process (if any).

We encourage the use of the ALD/Database in your presentations and papers. You can download an image of the periodic table (as shown above) for use under the creative commons agreement from the website itself. See the so called “hamburger” menu on the website for the download button; you can cite the ALD/Database using its DOI:10.6100/ALDDatabase. Also let us know if you are using the ALD/Database in your presentation! This is of course fully optional but it allows us to quantify its usage better and it is of course very good for our motivation.

The ALD/Database: setting-up the database and some statistics

To make the ALD/Database a success, we recognized that the website should offer value for the visitors from day one of its introduction. This meant that we wanted to start with a good collection of references. Apart from writing most of the code, Jan Buiter also did some data mining to fill the database. For this purpose, several excellent review papers were used:

• Puurunen [J. Appl. Phys. 97, 121301 (2005)]
• Miikkulainen et al. [J. Appl. Phys. 113, 021301 (2013)]
• Knoops et al. [J. Vac. Sci. Technol. A 37, 030902 (2019)]
• Mackus at al. [Chem. Mater. 31, 1142 (2019)].

Combined these review papers contained 2970 references. On top of that, during the year 1293 entries were contributed to the database by the ALD community (see also inset). The figure below shows the distribution of the number of ALD entries in the database per year of publication (data taken on 27-11-2019). Because of the large number of entries from the review papers, a jump in the data is seen at the cut-off date of the Miikkulainen paper which is a superset of the references in the paper by Puurunen. Over time, we expect that this aberration will fade as more-and-more papers will be added.

In Q1 of 2020 we will release a more comprehensive analysis of publications from 2019 reporting on ALD processes and their resulting material properties mainly based on the entries in the ALD/Database. In this report, we will take a closer look at the materials synthesized by ALD in 2019 reported for the first time as well as trends in new precursor and much more.

Our experiences running the ALD/Database

When starting an initiative such as the ALD/Database you always encounter interesting and unexpected factors. For example, we anticipated a sharp increase in the number of submissions around the AVS, MRS, and ALD conference. When preparing a presentation, you typically go through some literature and if you are presenting a recently published paper you might take the chance to add it to the ALD/Database. On the conference itself, people mentioning the ALD/Database in talks or during the coffee break might trigger people to have a look at it and (hopefully) submit their own works to the database. In practice these peaks in submissions are less prominent than expected and we get a steady stream of new submissions (10’s a week). The review process itself turned out to be more straightforward than expected. With a good set of rules, most papers either clearly meet the requirements or fall well outside the type of papers we are looking for. Keeping the database consistent was expected to be a major challenge and it indeed turns out to be just that. Finding and removing “near” duplicate entries is difficult: Do you submit a nearly stoichiometric materials using integer numbers (e.g. TiO₂) or variables (e.g. TiO_x)? In practice, people do both and these duplicates can be hard to track down. However, we feel that just forcing either style is too limiting and both have their use. Naming schemes in general are complex – see also the inset – with the main challenges being: keeping the naming consistent and correctly rendering complex chemical equations from plain text chemical equations.