Materials science domain is facing the fourth paradigm of science, i.e., data-driven science, which also encompasses the first three paradigms based on theory, experiment, and simulation. The semiconductor domain is one of many sub-domains of materials science, involving both mathematical models-based simulations and conventional experiments to study materials. A significant challenge in the semiconductor domain is the lack of interoperability between materials simulation data and experimental data. While there is existing work, such as the Materials Design Ontology, that enhances the interoperability of simulation data, there remains a need for representing experimental data with rich semantics. To improve the findability, accessibility, interoperability, and reusability of semiconductor experimental data, we present the initial steps in developing a semiconductor domain ontology, SemicONTO.
CEUR ceur-ws.org
The materials science domain, aiming at better understanding and discovering materials characteristics and properties, has a typical workflow that relies on both experiments and simulations [1]. Taking the semiconductor field as a sub-field of materials science as an example, researchers may first use computational methods to model semiconductor materials, and then compare simulation results with experimental results, such as from spectroscopy experiments to learn materials’ properties. Along such a workflow, massive materials data may be generated from computational simulations or experiments. Therefore, the materials science domain faces the big data challenges of volume, variety, velocity as well as variability [2, 3]. To organize materials data in a structured way and to share such data in a FAIR (Findable, Accessible, Interoperable and Reusable) [4] manner can not only help users better understand the data, but also enable the data to be used eficiently in diferent applications (e.g., machine learning-based materials design [5]). Several global eforts focus on dealing with data challenges in the domain, SeMatS 2024: The 1st International Workshop on Semantic Materials Science co-located with the 20th International http://huanyuli.se (H. Li); https://www.ida.liu.se/~patla00/ (P. Lambrix) © 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). e.g., the Materials Genome Initiative (MGI),1 and the European Materials Informatics Network (EuMINe).2 MGI aims at assembling and curating databases combining materials property data from both experiments and simulations. EuMINe targets at harmonizing resources and approaches within the domain of materials science.
To make materials data FAIR, ontologies has been realized as a way to represent semantics
of materials data and thus alleviate the heterogeneity issue among diferent data sources [ 6,
p. 22]. Essentially, ontologies contain: (
In this short paper, we present initial work of developing the SemicONductor onTOlogy (SemicONTO) version 0.1. The remainder of the paper is as follows. We introduce the related work in Section 2. Then, the development and content of SemicONTO is presented in Section 3. We maintain SemicONTO in a public GitHub repository,3 and publish the ontology with a permanent URI4 through the w3id service. In Section 4, we present its initial usage. Finally, in Section 5, we present concluding remarks and future work.
Our prior work presented in [3, 6, 7] has studied existing ontologies that are relevant to the materials science domain. For instance, the top-level ontology, Elementary Multiperspective Material Ontology (EMMO)5 aims at developing a standard ontology framework according to knowledge of materials modeling and characterization. Given that materials science is a broad domain that has diverse sub-domains, existing work also focuses on knowledge representation for these sub-domains. For instance, MatOWL [8], NanoParticle Ontology [9], MMOY [10], and Dislocation Ontology [11] focus on representing materials. Furthermore, the study and research on materials involve two basic activities that are materials simulations and experiments. The former is to run computational models, while the latter is to conduct particular experiments on materials, for studying and investigating materials’ structures, properties, etc. There has been some work focusing on these two directions. For instance, our prior work, Materials Design Ontology (MDO) [12, 7] is the first materials design ontology focusing on formally representing calculated materials data. The Platform MaterialDigital Ontology (PMD) [13] and its extended version (PMDco) [14] focus on describing materials science and engineering (MSE) 1https://www.mgi.gov/ 2https://www.cost.eu/actions/CA22143/ 3https://github.com/huanyu-li/SemicONTO 4http://w3id.org/SemicONTO 5https://github.com/emmo-repo/EMMO processes. However, to our knowledge, there is no existing work on formally representing domain knowledge for the semiconductor field.
In this work, we choose the Linked Open Terms (LOT) [15] methodology for developing our SemicONductor onTOlogy (SemicONTO). LOT is a lightweight methodology aiming at aligning the ontology development with software development agile practices. There are also other ontology development methodologies such as NeOn [16], “Ontology Development 101” [17] and eXtreme Design (XD) [18]. Similar as LOT, they all include general ontology development steps such as requirements analysis, ontology implementation, publication and maintenance. All the above methodologies have been used in various applications. In addition, LOT is the first methodology that has a focus on publishing ontologies in accordance with the FAIR principles. Therefore we choose LOT for developing SemicONTO.
To develop SemicONTO, knowledge engineers and a domain expert (i.e., the second author)
discussed domain interests regarding knowledge representation and data management for
semiconductors. During the discussions, we identified some use cases and competency questions.
Use cases. SemicONTO aims to capture knowledge regarding: (
Competency questions. After discussions between domain experts and knowledge engineers, we formulate four competency questions that the developed ontology should be capable to answer.
• CQ1: What are the diferent kinds of semiconductors? • CQ2: What is the composition information of a semiconductor material? • CQ3: Does a semiconductor have donors or acceptors? If it does, what are the donor and acceptor materials? • CQ4: What are the diferent steps for a semiconductor experiment?
The ontology requirements specification process aims to clarify why the ontology is being developed, and identify requirements in the form of use cases and competency questions. For developing SemicONTO, we identified use cases and competency questions as introduced in Section 3.1. Then, the implementation process includes conceptualization based on the identified requirements, reusing existing ontologies and design patterns, encoding the conceptualization rdfs:subClassOf results, and evaluation. Tools including the OntOlogy Pitfall Scanner (OOPS!) [19] and OOPS! for FAIR (FOOPS!) [20], recommended by LOT, are used in evaluation.6
The core concepts and relatiships of SemicONTO are shown in Figure 1 of which the formulation details are introduced as follows.
Semiconductor and structure. In our current ontology, a semiconductor is distinguished as either of the type ExtrinsicSemiconductor or IntrinsicSemiconductor (Axioms 1 and 2). In terms of structural composition, an extrinsic semiconductor has composing particles called acceptors and donors. An acceptor is a material (e.g., an atom or a molecule) which can bind to or accept an electron, therefore is able to form a positive hole in an extrinsic semiconductor (Axiom 3). A donor is a material (e.g., an atom or a molecule) which can provide an electron, therefore contributes a conducting electron in an extrinsic semiconductor (Axiom 4). Moreover, we define a relationship, hasStructure to represent that a material can be associated with some structural information in terms of chemical compositions (Axioms 5 and 6), and reuse the Structure and Composition concepts from MDO.
⊔ ⊑
⊓ = ∅ ⊑ ∃ ℎ. ⊓ ∀ ℎ. ⊑ ∃ ℎ. ⊓ ∀ ℎ.
⊤ ⊑ ∀ℎ.
∃ ℎ.⊤ ⊑ ∶
(
Experimental Step as both the Experiment and Experimental (PRO V-O). 7 dom Step ain as and range (Axioms 8, 9 sub-concepts of Activity ⊑ ∃ ℎ . ⊓ ∀ ℎ . ℎ ∘ ⊤ ∃ ℎ
⊑ ℎ .
⊑ ∀ ⊤ ℎ ℎ . ⊑ Contextual information of experiments and experimental steps. To represent contextual information of experiments and experimental steps, we use the InformationObject based on DOLCE [21]. Each experiment or experimental step is described by a specialized information object in which such information object captures information to describe an experiment or experimental step.
⊑ ∃ . ⊓ ⊑ ∀ . ∃ . ⊓ ∀
.
O
N
T
O
We identified four relevant experiments for organic semiconductor materials as uses cases
to be represented by the developed ontology. These four experiments are (
In accordance with competency questions outlined in Section 3, we formulate four SPARQL
queries. In Listing 1, we provide an example SPARQL quer y corresponding to CQ4. For each
competency question (CQ1-CQ4), we formulated one or more SPARQL queries.
oncluding
re
m
arks
and
future
w
ork
This paper introduces our initial work on SemicON TO,
domain. The ontology captures basic semantics that can
ductors. We showcase usage scenarios in which SemicO
diferent kinds of experiments on semiconductor materials,
an ontology for the semiconductor
represent experiments for
semiconN TO is utilized: (
NTO/demo/
(
rdf:type hasSubStep hasSubStep
rdf:type hasEquipment isDescribedBy experimentsOn
hasDonor
Semiconductor hasAcceptor Material experimentsFor
rdf:type SemiconductorExperiment
OPV_semiconductor_1
hasDonor experimentsFor rdf:type experimentsOn hasAcceptor rdf:type rdf:type Y6_1
PM6_1 "Preparation of Active Layer Solution"^^xsd:string hasExperimentalStepDescription
hasEquipment exp_0_step_0_infoobj "1"^^xsd:integer exp_0_step_0 hasNextStep
hasExperimentalStep hasExperimentalStep experiment_0 experimentsOn isDescribedBy exp_0_step_1_infoobj hasExperimentalStep isDescribedBy hasID "Ba ery Device Preparation and Parameter Characterization"^^xsd:string hasExperimentalStepDescription exp_0_step_1 "Substrate Cleaning"^^xsd:string rdf:type exp_0_infoobj annotated data. Since this paper presents the initial work of SemicONTO (version 0.1), we will continue working on SemicONTO towards introducing new concepts and relationships. For instance, representing properties and quantities, annotating more material experiments, and aligning SemicONTO with general ontologies such as EMMO. Moreover, we will discuss with domain experts in terms of describing experimental steps in a more structured way. For instance, to represent inputs, outputs and condition details of an experimental step.
This work has been financially supported by the Swedish National Graduate School in Computer Science (CUGS) and the EU Horizon project Onto-DESIDE (Grant Agreement 101058682).