Flash Graphene from Oil-Sands Asphaltenes: Yield, Energy, Cost and Life-Cycle Figures (Saadi et al., 2022)

"Sustainable valorization of asphaltenes via flash joule heating" was published in Science Advances in November 2022. It was received on 7 June 2022 and accepted on 30 September 2022. M. A. S. R. Saadi is listed first, and the paper notes that some of its authors contributed equally. The corresponding authors are Md Golam Kibria, James M. Tour (James M. Tour), Pulickel M. Ajayan and Muhammad M. Rahman. The paper's question is whether asphaltenes can be made into graphene by Flash Joule Heating (FJH). Asphaltenes are the heavy, carbon-rich residue of crude-oil and bitumen refining, which the authors say is usually burned as fuel or dumped in tailing ponds and landfills. The paper then tests the product in epoxy composites. Its most useful contribution to this dossier is an itemised account of how much electricity the flash uses and what the authors calculate the graphene would cost.

What was done

The feedstock was asphaltene powder from Alberta oil sands, supplied by InnoTech Alberta. Asphaltene conducts electricity too poorly to flash by itself. The authors therefore ball-milled it with 20 wt% carbon black (Cabot BP-2000), packed the blend into a quartz tube between copper electrodes, and discharged a capacitor bank through it. The sequence used four pulses at 185, 250, 250 and 370 V. According to the paper, the first three pulses drive off volatiles and non-carbon material, and the last does the conversion. The paper reports temperatures above 3000 K reached in under 100 ms, with the whole conversion over in less than a second. The product is called asphaltene-derived flash graphene (AFG).

The authors report two kinds of yield, and they should be kept separate. The process yield, AFG mass as a share of the starting carbon black and asphaltene mass, was about 45%. The AFG yield, given as over 95%, is a different measure: the share of Raman spectra that passed the paper's graphene criteria. Those criteria were an I2D/IG ratio of at least 0.3, a signal-to-noise ratio above 5 in the 2D band, and a 2D-band width under 100 cm⁻¹. By that test, 97.9% of spectra qualified. In other words, more than half the input mass did not end up as graphene, and the 95%-plus figure describes the quality of what remained.

What was measured

The authors characterised AFG with several methods:

• Raman: the average spectrum had I2D/IG of 0.941 and ID/IG of 0.244 (N = 100).
• XPS: carbon rose from about 93% to about 99%, oxygen fell from about 4% to about 1%, and sulfur fell from about 2.6% to almost zero.
• XRD: the (002) peak sat at 26.1°, broadened and with a low-angle tail, which the authors read as Turbostratic Graphene.
• TEM: flakes averaged about 25.4 nm across (SD 7.4 nm), with interlayer spacing of about 0.359 nm, compared with about 0.335 nm for AB-stacked graphene.

A ReaxFF molecular-dynamics simulation of ten asphaltene molecules was ramped to 3000 K and held for about 0.62 ns. In it, a graphitic carbon surface standing in for the carbon black acted as a nucleation site, and the system released CO, H₂ and C₂H₂.

In epoxy, 1 wt% AFG gave the best tensile results: strength up to about 37% higher and Young's modulus up to 12% higher than neat epoxy. Loadings of 3 and 5 wt% made the composite weaker, which the authors attribute to agglomeration. Thermal conductivity was not measured directly. Instead, the authors fitted infrared-camera temperature profiles to a finite-element model and inferred values from 0.18 W/(m·K) for neat epoxy to 0.46 W/(m·K) at 5 wt%. On mild steel in 3.5% NaCl, a coating with 10 wt% AFG gave a Tafel corrosion rate of 8.06 × 10⁻⁴ mpy, against 1.23 mpy for plain epoxy and 98.46 mpy for bare steel. Adding 7.5 wt% fumed silica made the ink printable, and the authors 3D-printed honeycomb and rectilinear structures with it.

The energy and cost figures

The energy accounting uses E = (V₁² − V₂²)C/2M, where C is the bank's capacitance of 0.624 F and M is the 4 g batch mass. The four pulses work out to 0.91455, 4.095, 4.8672 and 10.67625 kJ/g, about 20.55 kJ/g in total. The authors say the final 370 V pulse accounts for about 47% of energy use. By the paper's own per-pulse numbers, however, 10.676 of 20.55 kJ/g is closer to 52%. Because M is the mass of the whole batch, the 20.55 kJ/g is energy per gram of feedstock, not per gram of graphene recovered at 45% yield. The paper does not spell out how its cost model handles this difference.

The techno-economic model assumes electricity at USD 0.05/kWh and asphaltene at USD 0.52/kg. On those inputs it puts AFG production cost at 2.15 USD/kg. The abstract presents this as cheaper than other flash feedstocks, but the comparison in the results is narrower. According to the paper, plastic waste and rubber come out more expensive because of higher precursor costs or lower yields. Flash graphene from carbon black, coke or coal is also described as low-cost, owing to high yield and cheap precursors, so the advantage does not hold against the cheapest carbon sources. The authors also set AFG against CVD graphene, which costs far more, but they call this "not a fair comparison": CVD grows monolayer or few-layer films over large areas, whereas AFG is synthesised in bulk. The cost counts only precursor and electricity. The paper does not say whether the 20% carbon-black additive, equipment, labour or purification are included.

The life-cycle assessment is gate-to-gate, and its functional unit is 1 kg of asphaltene. It compares the flash route's electricity input with the energy content of burning the same asphaltene, estimated from Dulong's formula at 39,641 kJ/kg (11.01 kWh/kg). It covers six ILCD midpoint categories under two grid scenarios, natural gas and hydropower. The authors report lower impacts for the flash route in every category, and the lowest with hydropower. The inventory comes from laboratory conditions, literature and the Ecoinvent database. Rather than measuring uncertainty, the authors assigned a flat 20% error band to each impact category.

What the paper does and does not establish

This is a peer-reviewed primary study, and its methods section is detailed enough to check the energy arithmetic. That makes it stronger evidence than the patent abstracts and news items listed in Flash Joule Heating Claims vs. Evidence: The Dossier's Figures and the Absence of Independent Verification. Its limits are still substantial:

• The work is laboratory-scale, and the energy accounting assumes 4 g batches. The cost and LCA figures are models built on that scale, not on a production line.
• Conductivity depended on 20 wt% carbon black. AFG from an earlier run can replace it, but only at a higher share: a 50:50 AFG–asphaltene blend.
• The LCA compares burning asphaltene for energy with turning it into a material. The paper describes its boundary as covering only pulsed AFG synthesis and asphaltene combustion as fuel, so the graphene's use phase and any credit for displaced energy are not part of it.
• The paper has internal inconsistencies. The 47% versus 52% share is one. The corrosion section states that the current densities for AFG-coated steel were one and three orders of magnitude higher than the controls, while also reporting far lower corrosion rates. The methods give two process-yield values (67.6% and 94.8%) for the recycled-AFG route. The thermal-test laser power appears as 45 mW in a figure caption and 25 mW in the methods.
• The paper discloses that Rice owns intellectual property on flash conversion of carbon sources, including asphaltenes, licensed to Universal Matter Inc. and Universal Matter Ltd., in which Tour owns stock.

No independent group's replication appears in the paper.

Place in the dossier

This paper supplies a costed energy budget that can be set against other figures in the dossier. Its 20.55 kJ/g contrasts with the roughly 7.2 kJ/g that The Granted Flash Graphene Patent (US12054391B2): What Rice University Claimed for Pulse-Heated Carbon reports for high-carbon sources such as carbon black, calcined coke and anthracite, where the patent gives yields of 80 to 90% against about 45% here. The two are not directly comparable: the feedstocks differ, and the paper's figure covers a four-pulse sequence that includes three pretreatment pulses. The gap is consistent with the energy cost depending on the feedstock, but because the pulse schedules and per-gram bases also differ, the pair cannot establish that.