Octane: Chemical, Fuel Metrics & Computational Insights

• Octane is an eight‐carbon alkane with multiple isomers (e.g., n‑octane and iso‑octane) that serve as model compounds in chemical research and fuel metrology.
• As a fuel metric, octane quantifies resistance to autoignition via RON and MON, driving advances in combustion science and surrogate fuel design.
• In various studies, octane functions as a nonpolar solvent in thermodynamics, a benchmark in graph-theoretic QSPR, and a namesake in computing frameworks.

Searching arXiv for relevant papers on "octane" to ground the article. Octane denotes several distinct but technically important concepts across chemistry, combustion science, thermodynamics, catalysis, graph-theoretic QSPR, benchmarking, and machine learning. In molecular chemistry, it refers to $\mathrm{C_8H_{18}}$, including linear $n$-octane and branched isomers such as 2,2,4-trimethylpentane; in fuel science, “octane” often denotes a macroscopic antiknock metric, especially Research Octane Number (RON), defined relative to blends of $n$-heptane and iso-octane; in thermodynamics and interfacial science, octane serves as a nonpolar solvent or model alkane in liquid–liquid equilibrium, dielectric, and thin-film studies; in computer science, Octane also names a JavaScript benchmark suite and, separately, an autoencoder training framework acronymized as OCTANE. These usages are related by name rather than by a single underlying theory, but each has become a stable technical term in its own field (Zhu, 2022, Haji-Akbari et al., 2015, Vekris et al., 2016, Khatri et al., 9 Sep 2025).

1. Molecular identity and isomerism

In the chemical sense, octane is an eight-carbon alkane, $\mathrm{C_8H_{18}}$. The data distinguish at least two specific structural realizations. First, $n$-octane is a linear, saturated alkane and is explicitly described as a “linear, saturated alkane” and as a “nonpolar $n$-alkane” in several studies (Arribas et al., 26 Jan 2026, Alonso et al., 2024). Second, 2,2,4-trimethylpentane is identified as iso-octane, a highly branched C8 isomer that serves as the high-reference component in octane-number definitions (Zhu, 2022, Daly et al., 2016).

The structural diversity of octane isomers is central in graph-theoretic and QSPR work. One study explicitly uses the 18 structural isomers of octane as hydrogen-suppressed carbon-skeleton graphs, listing $n$-octane, 2-methylheptane, 3-methylheptane, 4-methylheptane, 3-ethylhexane, 2,2-dimethylhexane, 2,3-dimethylhexane, 2,4-dimethylhexane, 2,5-dimethylhexane, 3,3-dimethylhexane, 3,4-dimethylhexane, 2-methyl-3-ethylpentane, 3-methyl-3-ethylpentane, 2,2,3-trimethylpentane, 2,2,4-trimethylpentane, 2,3,3-trimethylpentane, 2,3,4-trimethylpentane, and 2,2,3,3-tetramethylbutane (Mondal et al., 2019). In that representation, all octane isomers are simple connected trees with $n=8$ vertices and $m=7$ edges, and maximum degree $\leq 4$ (Mondal et al., 2019).

This graph-theoretic treatment makes explicit that “octane” is not a single topology but an isomer class. A plausible implication is that the term functions differently across subfields: in physical chemistry it usually denotes the linear solvent $n$0-octane unless otherwise qualified, whereas in combustion metrology it is inseparable from the linear/branched contrast between $n$1-heptane and iso-octane (Zhu, 2022, Daly et al., 2016).

2. Octane as a fuel metric: RON, MON, sensitivity, and antiknock behavior

In combustion science and engine research, “octane” commonly refers not to the molecule octane but to a standardized measure of resistance to autoignition, or “knock.” Research Octane Number (RON) is determined in a CFR engine and is defined by matching the knock behavior of the test fuel to that of a primary reference fuel blend of $n$2-heptane and iso-octane, where $n$3-heptane has RON $n$4 and iso-octane has RON $n$5 (Zhu, 2022). A fuel with RON $n$6 behaves, under RON test conditions, like a mixture containing 90 vol% iso-octane and 10 vol% $n$7-heptane (Zhu, 2022). The same reference pair is also used in infrared-chemometric work on gasoline-like fuels (Daly et al., 2016).

Motor Octane Number (MON) is measured under more severe CFR-engine conditions, and octane sensitivity is defined as

$n$8

This identity is used explicitly in surrogate-fuel optimization and graph-ML fuel design (Daly et al., 2018, Rittig et al., 2022). For practical fuel specification, the $n$9 index is also used, although some studies emphasize RON because it is more representative of typical spark-ignition engine operation conditions (Zhu, 2022).

The antiknock interpretation of octane is linked in the data to microscopic combustion descriptors. Reactive molecular dynamics simulations show that the time of the turning point in the potential-energy profile exhibits a strong linear correlation with RON, while the equilibrium number of hydroxyl radicals per molecule has a clear negative correlation with RON (Zhu, 2022). For $n$0-paraffins and olefins, the turning-point correlation has $n$1 and $n$2, respectively, at 2500 K; for iso-paraffins the correlation remains positive with $n$3 (Zhu, 2022). At 2750 K, the duration of Stage I retains correlations $n$4 for $n$5-paraffins and $n$6 for olefins (Zhu, 2022). The same study notes that branched hydrocarbons invariably have higher RON than their linear isomers and that $n$7-octane lies near the low end of the surrogate set, with RON approximately $n$8, whereas toluene is near the high end at approximately 118 (Zhu, 2022).

Other computational fuel-design studies formalize “high octane” as an optimization target. A graph-ML framework uses

$n$9

as a proxy for high-octane-index fuels and reports GNN test-set mean absolute errors of approximately 4.5 for RON and 4.4 for MON (Rittig et al., 2022). A generative latent-space inverse-design framework chooses RON as the target property of interest and reports cross-validated CatBoost performance of $\mathrm{C_8H_{18}}$0, MAE $\mathrm{C_8H_{18}}$1, and RMSE $\mathrm{C_8H_{18}}$2 for RON prediction (Yalamanchi et al., 16 Apr 2025). These results suggest that in contemporary computational fuel science, “octane” has shifted from a purely empirical engine-test label to a target in surrogate modeling, generative design, and graph representation learning (Rittig et al., 2022, Yalamanchi et al., 16 Apr 2025).

3. Octane as a solvent and thermodynamic component

In mixture thermodynamics, octane usually denotes $\mathrm{C_8H_{18}}$3-octane as a nonpolar solvent. Several 2024 studies use it as one component in binary systems exhibiting upper critical solution temperatures (UCSTs), dielectric nonideality, or both (Alonso et al., 2024, Alonso et al., 2024, Alonso et al., 2024, Alonso et al., 2024).

In the binary system $\mathrm{C_8H_{18}}$4-caprolactam + octane, octane is the nonpolar alkane paired with a strongly polar, hydrogen-bonding cyclic secondary amide (Alonso et al., 2024). The measured liquid–liquid equilibrium coexistence curve spans $\mathrm{C_8H_{18}}$5 to $\mathrm{C_8H_{18}}$6, with a UCST of

$\mathrm{C_8H_{18}}$7

critical composition

$\mathrm{C_8H_{18}}$8

and fitted parameters $\mathrm{C_8H_{18}}$9, $n$0, $n$1, $n$2, and $n$3 (Alonso et al., 2024). Across the $n$4-alkane series, the UCST increases almost linearly with carbon number: 352.13 K for heptane, 354.51 K for octane, 358.61 K for nonane, and 363.43 K for decane (Alonso et al., 2024). The same paper states that UCST(2,2,4-trimethylpentane) $n$5 K exceeds UCST(octane) $n$6 K by about 7.8 K (Alonso et al., 2024).

In 2-phenoxyethanol + octane, the coexistence curve also shows a UCST and a nearly horizontal top (Alonso et al., 2024). The critical parameters are

$n$7

with empirical-fit parameters $n$8, $n$9, $n$0, and standard deviation $n$1 (Alonso et al., 2024). Near criticality, the same system yields a critical exponent $n$2, regarded as close to the 3D Ising or renormalization-group values and clearly non-classical (Alonso et al., 2024). Relative to other 2-phenoxyethanol + alkane systems, octane gives the highest UCST among the mixtures studied in that paper and lies above heptane while remaining far above cyclohexane and its alkyl-substituted derivatives (Alonso et al., 2024).

Octane also appears as a reference nonpolar diluent in dielectric and refractive-index studies. In dibutyl ether + octane, the data show small, slightly negative $n$3, $n$4, linear $n$5, and the explicit conclusion that “the dibutyl ether + octane system does not show meaningful structure” (Alonso et al., 2024). In TEGDME + octane, pure octane has $n$6 at 293.15 K and the mixture displays negative $n$7 over the whole composition range, with a minimum near $n$8 at $n$9 (Alonso et al., 2024). That study states that comparison of $n$0 and $n$1 for TEGDME + octane and DBE + octane shows that the polyether is a more structured liquid (Alonso et al., 2024).

Across these thermodynamic contexts, octane functions as the paradigmatic nonpolar component. The recurring interpretation is that it does not provide strong specific interactions, so deviations from ideality largely expose the self-association or structural organization of the polar co-component (Alonso et al., 2024, Alonso et al., 2024, Alonso et al., 2024).

4. Interfacial, catalytic, and nanoscale transport behavior of octane

Octane is also a model system in interfacial physics and surface catalysis. In thin-film simulations, $n$2-octane nanofilms confined between substrates exhibit strong thermodynamic and kinetic anisotropies (Haji-Akbari et al., 2015). Complete freezing is observed at low temperatures, while at intermediate temperatures a frozen monolayer emerges at both interfaces (Haji-Akbari et al., 2015). The effective melting temperature of the film is estimated as $n$3 K, with complete freezing for $n$4 K and surface freezing for $n$5 (Haji-Akbari et al., 2015). Two dynamical regimes occur near substrates: loose substrates accelerate dynamics, while sticky substrates decelerate dynamics, sometimes by as much as two orders of magnitude (Haji-Akbari et al., 2015). The same work reports no noticeable difference between free-surface and bulk regions in the ability to explore the potential-energy landscape, unlike model atomic glass-formers (Haji-Akbari et al., 2015).

In surface catalysis, $n$6-octane on Pt(111) is investigated as a model aliphatic hydrocarbon for the activation of inert $n$7 and C–C bonds (Arribas et al., 26 Jan 2026). When deposited at 300 K, $n$8-octane physisorbs intact as an all-trans chain; above about 330 K, a terminal C–H bond is activated and partially dehydrogenated chemisorbed chains form (Arribas et al., 26 Jan 2026). At high temperature, two major channels are identified. One is intramolecular cyclization and aromatization to adsorbed benzene plus a two-carbon fragment, with overall exothermicity of approximately $n$9 eV and an estimated rate-limiting barrier of about $n=8$0 eV (Arribas et al., 26 Jan 2026). The second is intermolecular dehydrogenative lateral homo-coupling of two fully unsaturated $n=8$1-octa-1,3,5,7-tetraenyl chains, proceeding in a zipper-like fashion to anthracene-family polycyclic products, with an estimated rate-limiting barrier of about $n=8$2 eV (Arribas et al., 26 Jan 2026). At still higher temperatures, extended nanographene patches form (Arribas et al., 26 Jan 2026).

At the nanoscale transport level, octane derivatives serve as benchmark molecular junctions in inelastic electron tunneling spectroscopy. A semi-analytical DFT–CPKS method for first-order electron–vibration coupling is applied to octane-dithiol and octane-diamine single-molecule junctions to discuss the influence of the anchoring group and mechanical stretching on the IETS (Bürkle et al., 2013). The paper defines the linear electron–vibration interaction through

$n=8$3

with coupling matrices

$n=8$4

(Bürkle et al., 2013). In this context, octane is not a solvent or fuel metric but a molecular backbone in a transport-active nanoscale device.

5. Octane in QSPR, chemical graph theory, and molecular design

Octane isomers are a standard testbed in chemical graph theory because they combine manageable size with nontrivial structural diversity. One 2019 paper introduces four neighbourhood-degree-based indices—$n=8$5, $n=8$6, $n=8$7, and $n=8$8—and evaluates them on the 18 octane isomers (Mondal et al., 2019). The central quantity is the neighbourhood degree

$n=8$9

from which the new indices are constructed (Mondal et al., 2019). For the octane isomer set, the reported Pearson correlations with acentric factor are $m=7$0 for $m=7$1, $m=7$2 for $m=7$3, $m=7$4 for $m=7$5, and $m=7$6 for $m=7$7; correlations with entropy are $m=7$8, $m=7$9, $\leq 4$0, and $\leq 4$1, respectively (Mondal et al., 2019). The same study reports sensitivity $\leq 4$2 for $\leq 4$3 and $\leq 4$4, indicating non-degeneracy over the 18 octane isomers (Mondal et al., 2019).

A later graph-polynomial study does not itself compute octane values but states that the hyperbolic Sombor index (HSO), proposed in 2025, “shows its chemical applicability for octane isomers and the structure sensitivity and abruptness for octane, nonane, and decane isomers, respectively” (Barman et al., 16 Feb 2026). In that paper, HSO is defined as

$\leq 4$5

and expressed through the M-polynomial by

$\leq 4$6

(Barman et al., 16 Feb 2026). The octane-specific numerical values are deferred to another publication, but the present paper makes explicit that octane isomers remain a benchmark family for testing the structure sensitivity of degree-based indices (Barman et al., 16 Feb 2026).

Octane is also central to inverse molecular design for fuels. Graph machine learning for high-octane fuels uses RON and octane sensitivity as the optimization targets and reports rediscovery of well-established high-octane components such as ethanol, MTBE, and ETBE, together with new candidate molecules (Rittig et al., 2022). A Co-optimized variational autoencoder plus CatBoost regression is trained on a C/H/O subset of GDB-13 enriched with a curated RON database and uses differential evolution in latent space to identify promising high-RON molecules (Yalamanchi et al., 16 Apr 2025). The latter reports 1189 valid SMILES strings with predicted RON $\leq 4$7, corresponding to 1185 unique chemical species, of which 921 are novel relative to the VAE training set (Yalamanchi et al., 16 Apr 2025). These studies use “octane” in the property sense, but the structural motifs they highlight—branching and oxygenated functional groups—connect directly back to the longstanding contrast between low-octane linear chains and high-octane branched or oxygenated molecules (Rittig et al., 2022, Yalamanchi et al., 16 Apr 2025).

6. Octane in computation and software systems

Outside chemistry, Octane has a distinct meaning in computer science. In programming-language verification, Octane denotes a JavaScript benchmark suite developed by Google (Vekris et al., 2016). A refinement-type system for TypeScript, Refined TypeScript (RSC), is evaluated on parts of this suite, specifically navier-stokes, richards, splay, and raytrace (Vekris et al., 2016). These benchmarks are described as performance-critical JavaScript using arrays, numeric code, and control flow, making them stringent tests for static verification of array bounds, null safety, and related invariants (Vekris et al., 2016). Quantitatively, the reported benchmark sizes and timings are: navier-stokes, 366 LOC and 473 s; splay, 206 LOC and 6 s; richards, 304 LOC and 7 s; raytrace, 576 LOC and 15 s (Vekris et al., 2016). In this usage, Octane has no relation to fuels or hydrocarbons beyond the name.

A second computational usage is the acronym OCTANE, standing for “Optimal Control for Tensor-based Autoencoder Network Emergence” (Khatri et al., 9 Sep 2025). This 2025 framework models encoder and decoder as coupled differential equations, formulates training as an optimal-control problem, and solves state and adjoint dynamics on low-rank tensor manifolds using a rank-adaptive explicit Euler scheme (Khatri et al., 9 Sep 2025). The forward encoder and decoder dynamics are

$\leq 4$8

with cost

$\leq 4$9

(Khatri et al., 9 Sep 2025). On MNIST denoising, reported average memory savings range from 7.10% for $n$00 to 16.21% for $n$01; on deblurring, savings range from 46.74% to 57.46% (Khatri et al., 9 Sep 2025). The authors recommend $n$02, $n$03, and

$n$04

as practical hyperparameter ranges (Khatri et al., 9 Sep 2025). This OCTANE is therefore an acronymic reuse of the term rather than an octane-related scientific concept.

The coexistence of Octane as benchmark suite and OCTANE as optimal-control autoencoder framework illustrates a common terminological pattern in computing: chemically suggestive names are frequently repurposed as project titles or acronyms. This suggests that, in interdisciplinary literature searches, “octane” is highly polysemous and requires context-sensitive disambiguation (Vekris et al., 2016, Khatri et al., 9 Sep 2025).

7. Conceptual synthesis and disambiguation

Across the sources, “octane” has at least four stable technical senses. The first is a molecular species or isomer class, centered on $n$05-octane and iso-octane (Alonso et al., 2024, Mondal et al., 2019). The second is a macroscopic fuel metric, primarily RON and related quantities such as MON, sensitivity, and octane index (Zhu, 2022, Daly et al., 2018). The third is a model component in physical chemistry, where octane serves as a nonpolar solvent, thin-film material, catalytic reactant, or molecular junction backbone (Alonso et al., 2024, Haji-Akbari et al., 2015, Arribas et al., 26 Jan 2026, Bürkle et al., 2013). The fourth is nominal or acronymic reuse in computation, as in the Octane JavaScript benchmark suite and OCTANE autoencoder framework (Vekris et al., 2016, Khatri et al., 9 Sep 2025).

A frequent source of confusion is the conflation of the molecular name with the fuel metric. The data make the distinction explicit: $n$06-octane is itself a specific hydrocarbon, but “octane number” is defined with respect to $n$07-heptane and iso-octane and does not denote the concentration of octane in a fuel (Zhu, 2022, Daly et al., 2016). Another potential misconception is that “octane” in materials or thermodynamic studies automatically refers to fuel behavior; in fact, many such studies use $n$08-octane because it is a simple nonpolar alkane or a well-defined linear chain, not because its octane number is relevant (Haji-Akbari et al., 2015, Alonso et al., 2024).

Taken together, the literature portrays octane as an exemplary cross-disciplinary term. In chemistry it anchors ideas of branching, polarity contrast, and alkane structure; in combustion it anchors the practical language of knock resistance; in catalysis and nanoscience it serves as a tractable C8 hydrocarbon model; in graph theory it provides a canonical finite family of isomers; and in computer science it survives as a benchmark name and acronym. The technical meaning of “octane” is therefore determined less by the word itself than by the surrounding formalism: phase equilibria, CFR metrology, EV-coupling Hamiltonians, graph invariants, or neural-ODE optimal control.