The study focused not merely on which alternative fuels receive attention, but on the engineering problems that inventors are trying to solve once a fuel reaches a vessel. Patent abstracts were analyzed with BERTopic, a semantic topic-modeling framework that groups documents according to contextual meaning rather than simple word overlap. Cleaned patent families were embedded with a Sentence-Transformers model, reduced in dimension with UMAP and clustered with HDBSCAN. Class-based term frequency-inverse document frequency was then used to identify representative terms in each topic. After related clusters were consolidated, the researchers organized the non-outlier topics into 14 interpretable themes and manually checked their labels against keywords and representative patent abstracts.

Those themes separated into pathway-specific innovations, transferable engineering modules and auxiliary maritime functions. Hydrogen patents centered on fuel-cell conversion, including reactant supply, water and thermal management, purge strategies and safety-oriented control. Ammonia patents focused on liquefaction, pressurized transfer, tank interfaces and ammonia-slip control. Methanol innovation was concentrated in shipboard tank and compartment design intended to accommodate safety separation without imposing excessive cargo-space penalties. LNG patents emphasized cryogenic storage, pressure management, boil-off handling, cooling and pumping. Biofuel patenting appeared more supply-oriented, with biomass, lignin conversion, renewable liquid-fuel production and fuel upgrading more visible than distinctive shipboard handling systems.

The semantic structure also revealed that fuels can remain technically distinct while sharing many ship-level requirements. Pairwise topic embeddings had a median off-diagonal cosine similarity of about 0.69, with an interquartile range of approximately 0.61 to 0.79. LNG and ammonia supply topics were the most similar pathway pair, at roughly 0.87, followed by LNG and methanol at about 0.84. Methanol and biofuel production were also closely linked at approximately 0.84, reflecting shared language around liquid-fuel processing and conditioning. By contrast, ammonia supply and biofuel production had the weakest pathway-level relationship, at about 0.61. These similarities describe relationships among terms and documents in the corpus; they are not performance measurements or direct indicators of commercial maturity.

Hierarchical clustering produced three broad technological branches. One centered on electrification, joining data-driven and hybrid-electric control with onboard battery management and hydrogen fuel-cell systems. Another covered shipboard handling and auxiliary functions, including fluid filtration, cryogenic LNG storage, emissions mitigation, lubrication and ammonia supply. A third linked methanol integration, biomass-based production and marine-fuel upgrading. In relative terms, the liquid-fuel and fuel-handling branches sat closer to each other than to the electrification branch. The analysis indicates where recurring engineering concepts appeared together in patent language; low similarity was treated as evidence of weaker association, not automatically as an empty technological field.

Patent activity over time showed hydrogen maintaining the highest intensity through most of the observation period, increasing markedly after 2012–2014 and remaining dominant in the latest time slices. Ammonia patenting stayed comparatively low for much of the earlier record but rose sharply after 2021. Methanol followed a similar, though more concentrated, pattern, with limited activity before 2021 and a pronounced increase in 2024–2025. LNG grew more gradually, reaching a moderate level by 2012–2017 and strengthening again after 2021. Biofuel activity fluctuated: it increased during 2009–2014, weakened during 2015–2020 and recovered in the most recent slices. These counts were assigned at the patent-family level, and a family could contribute to more than one fuel pathway when it contained explicit terms for several fuels.

The literature analysis showed a parallel progression from basic fuel properties toward integrated deployment questions. CiteSpace keyword timelines divided research into an emergence phase in 2006–2010, centered on combustion, flammability and early alternative-fuel testing; an expansion phase in 2010–2015, when natural gas, hydrogen, methanol and biomass appeared alongside engine adaptation; an optimization phase in 2015–2020, emphasizing systems, cost, efficiency and integration; and a deployment phase from 2020 to 2025, with greater attention to green hydrogen, green ammonia, water electrolysis, mission profiles and modeling. Keyword-burst analysis found the strongest recent signal for ammonia, with a burst strength of 11.32 during 2023–2025. Hydrogen reached 6.74 in 2024–2025, green hydrogen 6.17 in 2023–2025 and methanol 4.11 in 2024–2025.

The authors combined those publication signals with patent activity in a five-pathway matrix. Its horizontal axis represented sustained patenting across the full 2000–2025 period, its vertical axis captured recent literature-frontier signals during 2023–2025, and bubble area represented each pathway’s cumulative patent-family count. Hydrogen reached the maximum normalized sustained-activity score of 1.000 and a frontier score of 0.595. Ammonia scored 1.000 for recent literature salience but 0.000 for sustained patent activity, while methanol scored 0.363 and 0.044, respectively. LNG recorded normalized scores of 0.377 for sustained patenting and 0.275 for frontier attention; biofuels scored 0.296 and zero because no eligible recent keyword burst was identified for that pathway.

The resulting picture supports differentiated rather than universal innovation strategies. Hydrogen’s strongest opportunity appears to lie in fuel-cell systems, hybrid-power coordination and shipwide energy management, making it an architecture-level opportunity rather than a simple fuel substitution. Ammonia’s recent attention centers on storage, transfer, safety management and exhaust-side control at the ship–fuel interface, while methanol’s opportunity extends across onboard adaptation, fuel upgrading and renewable production. LNG innovation is concentrated in cryogenic and pressure-management interfaces, whereas biofuel innovation is more visible in feedstocks, production and conditioning. Across all five pathways, storage, fuel supply, emissions control and energy management recur as common engineering bottlenecks. The evidence does not establish which fuel will dominate shipping or demonstrate that any pathway is commercially deployable; it maps where inventive activity and research attention have accumulated and where shared technical obstacles are repeatedly expressed.

Several limitations constrain how far the patent-led findings can be generalized. Patent records may lag industrial activity because of disclosure delays, and searches of the Derwent Innovations Index and Web of Science may omit relevant work. Fuel tagging depended on explicit terms and allowed multi-pathway overlap, while some topics were more directly linked to maritime use than others. The positioning matrix also omitted patent quality, technology readiness, cost, life-cycle emissions, infrastructure conditions and policy feasibility. It therefore compares relative research attention and disclosed invention activity rather than direct measures of adoption, efficiency or environmental performance. The study’s central implication is methodological as well as substantive: identifying opportunities in green shipping may require looking beyond fuel rankings to the recurring interfaces through which new fuels become workable aboard ships.

Subject of Research: Semantic patent mining and bibliometric analysis of technological innovation in five green marine fuel pathways from 2000 to 2025.

Article Title: Technological opportunity identification for green marine fuels: a BERTopic-based patent mining study

Article References: Li, Z., Xiao, G., Biancardo, S. A., Zhang, M., & Wang, T. (2026). Technological opportunity identification for green marine fuels: a BERTopic-based patent mining study. Frontiers in Marine Science, 13, Article 1898136. https://doi.org/10.3389/fmars.2026.1898136

Image Credits: AI Generated

DOI: 10.3389/fmars.2026.1898136

Keywords: green marine fuels, patent mining, BERTopic, maritime decarbonization, hydrogen propulsion, ammonia fuel, methanol fuel, LNG, biofuels, shipping innovation, Technological, opportunity