# Fileset

[Advanced Materials - 2024 - Hunter - Elucidating the Mechanism of Iron‐Catalyzed Graphitization  The First Observation of (1).pdf](https://mdr.nims.go.jp/filesets/912edb71-a963-4e4b-a27a-b6ea3f1c21b9/download)

## Creator

Robert D. Hunter, [Masaki Takeguchi](https://orcid.org/0000-0002-0282-6020), Ayako Hashimoto, Kannan M. Ridings, Shaun C. Hendy, Dmitri Zakharov, Nils Warnken, Jack Isaacs, Sol Fernandez‐Muñoz, Joaquín Ramirez‐Rico, [Zoe Schnepp](https://orcid.org/0000-0003-2171-067X)

## Rights

[Creative Commons BY Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/)

## Other metadata

[Elucidating the Mechanism of Iron‐Catalyzed Graphitization: The First Observation of Homogeneous Solid‐State Catalysis](https://mdr.nims.go.jp/datasets/6f72b092-90b6-4092-a28f-5b5967f5520a)

## Fulltext

Elucidating the Mechanism of Iron‐Catalyzed Graphitization: The First Observation of Homogeneous Solid‐State CatalysisRESEARCH ARTICLEwww.advmat.deElucidating the Mechanism of Iron-Catalyzed Graphitization:The First Observation of Homogeneous Solid-State CatalysisRobert D. Hunter, Masaki Takeguchi, Ayako Hashimoto, Kannan M. Ridings,Shaun C. Hendy, Dmitri Zakharov, Nils Warnken, Jack Isaacs, Sol Fernandez-Muñoz,Joaquín Ramirez-Rico,* and Zoe Schnepp*Dedication: In the memory of Professor Roy L. JohnstonCarbon is a critical material for existing and emerging energy applications andthere is considerable global effort in generating sustainable carbons. Aparticularly promising area is iron-catalyzed graphitization, which is theconversion of organic matter to graphitic carbon nanostructures by an ironcatalyst. In this paper, it is reported that iron-catalyzed graphitization occursvia a new type of mechanism that is called homogeneous solid-state catalysis.Dark field in situ transmission electron microscopy is used to demonstratethat crystalline iron nanoparticles “burrow” through amorphous carbon togenerate multiwalled graphitic nanotubes. The process is remarkably fast,particularly given the solid phase of the catalyst, and in situ synchrotron X-raydiffraction is used to demonstrate that graphitization is complete within a fewminutes.1. IntroductionIn chemical and biochemical processes, catalysts are typically de-fined as homogeneous or heterogeneous. Homogeneous cata-lysts are molecular species that drive reactions of other molec-ular species in the gas or liquid phase. In contrast, heteroge-neous catalysis involves the reaction of liquid, solution, or vaporR. D. Hunter, J. Isaacs, Z. SchneppSchool of ChemistryUniversity of BirminghamBirmingham B152TT, UKE-mail: z.schnepp@bham.ac.ukM. Takeguchi, A. HashimotoCenter for Basic Research on MaterialsNational Institute for Materials ScienceTsukuba, Ibaraki 305-0047, JapanK. M. Ridings, S. C. HendyDepartment of PhysicsThe University of AucklandAuckland 1010, New ZealandThe ORCID identification number(s) for the author(s) of this articlecan be found under https://doi.org/10.1002/adma.202404170© 2024 The Author(s). Advanced Materials published by Wiley-VCHGmbH. This is an open access article under the terms of the CreativeCommons Attribution License, which permits use, distribution andreproduction in any medium, provided the original work is properly cited.DOI: 10.1002/adma.202404170phase molecular species at a solid catalystsurface. In the realm of materials science,the concept of catalysis expands to includegrowth of solid materials from nanopar-ticles. Here, precursor species dissolve ina liquid or solid nanoparticle and a newmaterial grows from the nanoparticle, typ-ically in the form of a nanowire or nan-otube. The function of the catalyst is toprovide a surface for nucleation of the1D nanostructure and to facilitate diffu-sion of atoms from the precursor to thegrowing material surface. Catalytic growthis used widely in research and industryto produce metal, chalcogenide and oxidenanowires,[1] and carbon nanotubes.[2] Thistype of catalytic outgrowth is called vapor–liquid–solid (Figure 1a), vapor–solid–solid,solution-liquid–solid, solution-solid–solid, or solid–liquid–solid(Figure 1b) growth, depending on the physical phase of the pre-cursors, catalyst, and product, respectively.[3]Whatever the type of catalyst, there is a need for movement. Re-actants need to diffuse to the active site on homogeneous or het-erogeneous catalysts and products need to move away. In the caseof 1D nanostructure growth, precursors need to be transported toD. ZakharovCenter for Functional NanomaterialsBrookhaven National LaboratoryUpton, NY 11973-5000, USAN. WarnkenSchool of Metallurgy and MaterialsUniversity of BirminghamBirmingham B152TT, UKS. Fernandez-Muñoz, J. Ramirez-RicoInstituto de Ciencia de Materiales de Sevilla (ICMS)Universidad de Sevilla-CSICSevilla 41092, SpainE-mail: jrr@us.esAdv. Mater. 2024, 36, 2404170 2404170 (1 of 10) © 2024 The Author(s). Advanced Materials published by Wiley-VCH GmbHhttp://www.advmat.demailto:z.schnepp@bham.ac.ukhttps://doi.org/10.1002/adma.202404170http://creativecommons.org/licenses/by/4.0/http://creativecommons.org/licenses/by/4.0/mailto:jrr@us.eshttp://crossmark.crossref.org/dialog/?doi=10.1002%2Fadma.202404170&domain=pdf&date_stamp=2024-07-16www.advancedsciencenews.com www.advmat.deFigure 1. Schematic of a) vapor–liquid–solid and b) solid–liquid–solid catalytic growth of 1D nanostructures with arrows showing the direction oftransport of atoms. c) Scanning electron microscope (SEM) image of nanotubes formed from pyrolysis of cellulose with iron nitrate (reproduced withpermission from reference 5).the catalyst interface, atoms need to diffuse through the catalyst,and the nanowire or nanotube needs to grow out from the cata-lyst. Because of the multiple transport mechanisms required, itis reasonable to assume that at least one phase in a catalytic pro-cess must be a liquid or gas. In this paper, we demonstrate thefirst catalytic system where the precursor, catalyst and productare all solids.The system in question is catalytic graphitization.[4] When or-ganic matter (e.g., biomass or biopolymers) is mixed with ironsalts and heated in an inert atmosphere, the organic moleculesdecompose into amorphous carbon and the iron salts form ironor iron carbide (Fe3C) nanoparticles. On further heating (≈700–800 °C), the catalytic nanoparticles convert the amorphous car-bon matrix to a dense network of nanotubes made up of lay-ered graphene sheets (Figure 1c).[5] These are analogous to mul-tiwalled carbon nanotubes but are irregular in shape. Fractureplanes of graphitized biomass show that the nanotubes pene-trate through the sample, indicating that the catalyst particlestravel through the whole amorphous carbon matrix. Transmis-sion electron microscopy (TEM) has been used to show thatthe catalyst particles move along irregular pathways through theamorphous carbon, leaving a graphitic nanotube trail behindthem.[6] The fast movement and liquid-like appearance of thecatalyst particles have led many authors, including ourselves, toassume that the catalyst is in the liquid state.[7–10] We now re-port that graphitization actually occurs via rapid diffusion of acrystalline catalyst particle through a solid amorphous carbonmatrix. This detailed insight into the mechanism of graphiti-zation will enable much better understanding and predictionof structure–property relationships in carbons from catalyticgraphitization.2. Results and Discussion2.1. In Situ Transmission Electron MicroscopyIron-doped amorphous carbon powder (prepared from iron ni-trate and cellulose) was heated to 900 °C in a transmission elec-tron microscope (TEM). Video footage at 800 °C (Video S1, Sup-porting Information; and Figure 2a) shows that the graphiti-zation process is characterized by rapid movement of liquid-like particles through the amorphous carbon matrix. The move-ment is irregular, with particles continuously stopping and start-ing, and the average speed was estimated as 4 nm s−1. Theparticles travel along irregular pathways and closely follow theedge of the carbon sample, showing no evidence of outgrowthfrom the carbon surface. As the particles move, they dissolveamorphous carbon from the matrix and deposit trails of graphi-tized carbon in the form of nanotubes (Video S2, SupportingInformation; and Figure 2b,c). The nanotubes are irregularlyshaped with walls made up of stacked layers of straight orcurved graphene sheets, (Figure 2d). During continued heat-ing, the particle movement becomes more sporadic. Particles ap-pear to stop moving when they have exhausted their supply ofamorphous carbon and become surrounded by graphitic mate-rial. A helpful analogy for this is the old mobile phone gameof snake, where the head of the snake can become trappedin coils of its own body (Figure 2e). Many particles continueto show liquid-like behavior after their fast movement stopsand can be seen to move slightly within the cavity they aretrapped in (Figure 2f). Selected area electron diffraction andhigh-resolution imaging of particles that had stopped movingshowed patterns characteristic of crystalline iron (Figure S1,Supporting Information) or iron carbide (Figure S2, SupportingInformation).Dark field TEM footage was collected by selecting diffractionspots using the objective aperture and using electrons diffractedalong that angle to generate images.[11] This method is used toshow bright contrast only for crystalline particles that are diffract-ing electrons along the selected angle range. The dark field TEMvideo footage of this sample (Video S3, Supporting Information)shows that fast moving catalyst particles alternate between brightand dark as they move (Figure 3a). The observation of particles re-maining continuously bright for sections of movement throughthe carbon matrix (Figure 3b) confirms that the particles are crys-talline, while they are catalyzing graphitization. The alternatingbetween bright and dark indicates that the particles are chang-ing crystallographic orientation as they move. It could be arguedAdv. Mater. 2024, 36, 2404170 2404170 (2 of 10) © 2024 The Author(s). Advanced Materials published by Wiley-VCH GmbH 15214095, 2024, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202404170 by National Institute For, Wiley Online Library on [09/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons Licensehttp://www.advancedsciencenews.comhttp://www.advmat.dewww.advancedsciencenews.com www.advmat.deFigure 2. a) Sequence images from TEM video at 800 °C showing catalyst particles moving rapidly through an amorphous carbon matrix. b) Imagesfrom a video at higher magnification showing a single catalyst particle generating a multiwalled graphitic nanotube (dashed line indicates the edge of theamorphous carbon). c) Schematic of catalyst particle movement with dissolution of amorphous carbon at the front edge of the catalyst and depositionof a graphitic nanotube at the back edge. d) TEM image showing the multiple graphitic walls of a nanotube. e) Schematic of the mobile phone game“snake” showing the snake becoming trapped. f) Images from TEM video of a catalyst particle that has become trapped but continues to move slightly.that the dark/bright alternation is a result of sequential meltingand solidification. However, this would be energetically a lot lessfavorable than particle rotation as it would involve consumptionand release of latent heat. Crystalline grains are known to ro-tate readily in polycrystalline systems during deformation andannealing.[12] Similarly, Wang et al. observed that Co-W-C alloycatalyst particles would rotate during the formation of carbonnanotubes from a C2H4 feedstock and were able to measure thechanges in orientation of the catalyst nanoparticles.[13] However,in this work, the rotation of the catalyst particles appears to beextremely fast, making it challenging to quantify in the samemanner.2.2. In Situ Synchrotron X-Ray DiffractionIn situ synchrotron X-ray diffraction (XRD) data supports the insitu TEM observations. Iron-doped amorphous carbon was pre-pared from cellulose and Fe(NO3)3 and heated to 400 °C in N2 toremove volatile organic pyrolysis products. The sample was thenheated under N2 at 5 °C min−1 to 800 °C in the synchrotron beam.Heat maps of the diffraction data (Figure S3, Supporting Infor-mation) show the appearance of very broad peaks for iron oxide.The broad nature of the peaks makes refinement difficult butthe sudden appearance of the peaks at 600 °C suggests they arecaused by carbothermal reduction of Fe3O4 to FeOx (wustite).[14]Adv. Mater. 2024, 36, 2404170 2404170 (3 of 10) © 2024 The Author(s). Advanced Materials published by Wiley-VCH GmbH 15214095, 2024, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202404170 by National Institute For, Wiley Online Library on [09/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons Licensehttp://www.advancedsciencenews.comhttp://www.advmat.dewww.advancedsciencenews.com www.advmat.deFigure 3. a) Images from dark field TEM video at 900 °C showing a catalyst particle moving through the amorphous carbon matrix and alternatingbetween dark and bright appearance. b) Images from dark field TEM showing a particle remaining bright for a section of movement, proving that thecatalyst is crystalline during movement.An expanded section of the heat map (Figure 4a) shows thatsharp peaks for Fe/Fe3C emerge around 750 °C. A peak corre-sponding to the interplanar spacing of graphitic carbon (Q =1.8 Å−1) evolves alongside the Fe/Fe3C peaks (Figure 4b), indi-cating that graphitization commences immediately after forma-tion of the catalyst. A plot of normalized graphite peak intensity(Figure 4c) shows that graphitization is largely complete within10 min. This confirms that the in situ TEM observations of fastcatalyst movement are consistent with the bulk behavior of thesystem.Figure 4. a) Heatmap showing diffracted intensity as a function of scattering vector magnitude Q and measurement time for in situ diffraction measure-ments in the range ≈700–800 °C. The right panel shows the temperature program used for the experiment. b) Scattered intensity recorded at differenttemperatures during in situ synchrotron X-ray diffraction showing evolution of the (002) graphitic peak (left panel) and Fe-containing phases (rightpanel). c) Normalized graphite peak intensity and d) mass fraction of 𝛼-Fe, 𝛾-Fe, and Fe3C obtained from Rietveld refinement of diffraction data asa function of measurement time and plotted alongside reaction temperature. e) Differential scanning calorimetry data from bare cellulose fibers (redline) and cellulose fibers impregnated with Fe(NO3)3 solution (blue line). The inset shows a small endothermic peak at ≈605 °C due to carbothermalreduction of the oxide nanoparticles.Adv. Mater. 2024, 36, 2404170 2404170 (4 of 10) © 2024 The Author(s). Advanced Materials published by Wiley-VCH GmbH 15214095, 2024, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202404170 by National Institute For, Wiley Online Library on [09/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons Licensehttp://www.advancedsciencenews.comhttp://www.advmat.dewww.advancedsciencenews.com www.advmat.deRietveld refinement was used to calculate phase compositionsfor iron-containing crystalline phases as a function of temper-ature. Three crystalline phases were considered for the refine-ments: Fe3C (cementite), 𝛼-Fe (ferrite), and 𝛾-Fe (austenite). It isimportant to understand that weight fractions are calculated bydetermining the relative contribution (scale factor) of each phaseto the total diffracted intensity, normalizing to 100%, and that theabsolute contribution of each phase is unknown. This means thatduring the early stages of graphitization and Fe/Fe3C crystalliza-tion, the values for the scale factors are relatively low and there-fore the weight fractions have large uncertainties (more detailscan be found in the Supporting Information). Figure S4 (Support-ing Information) shows the evolution of the crystalline domainsize for each phase, indicating an increase in crystallinity of allthe iron-containing phases alongside the growth of the graphitepeak. An example of a single refinement showing all three Fe-Cphases is shown in Figure S5 (Supporting Information) and fullrefinement details are available in the Supporting ExperimentalInformation. Figure 4d shows the weight fraction of each of thethree iron phases during the graphitization step. At the onset ofgraphitization, the primary crystalline phase is 𝛼-iron. Duringgraphitization, this phase largely disappears, with 𝛾-iron becom-ing dominant and Fe3C (cementite) also increasing. This is con-sistent with the phase diagram for iron/carbon.[15] The complexmixture of phases makes it impossible to conclude which is re-sponsible for catalytic graphitization and it is possible that morethan one iron phase is catalytically active for graphitization. Inchemical vapor deposition (CVD) synthesis of carbon nanotubes,there is evidence that Fe and Fe3C can both act as catalysts but fordifferent types of carbon nanotube (straight and bamboo-like).[16]The rapid emergence of 𝛾-Fe and Fe3C alongside graphitizationin the cellulose system could indicate that both phases are cat-alytic here.When the sample is held at 800 °C for 2 h, the crystallographiccomposition stays constant (Figure S6, Supporting Information).On cooling, 𝛾-iron is replaced by the lower temperature 𝛼-ironphase (Figure S7, Supporting Information). This is accompaniedby a small increase in the graphite peak intensity (Figure S8, Sup-porting Information). Carbon is more soluble in 𝛾-iron than 𝛼-iron,[17] so it is likely that the additional graphitic carbon is precip-itated during the iron phase transformation. The fact that carbonis precipitated from the 𝛾-iron phase when it transforms to thelower temperature 𝛼-iron phase indicates that the 𝛾-iron nanopar-ticles contain an appreciable amount of dissolved carbon. Thissupports the proposed dissolution-precipitation mechanism ofgraphitization.Thermogravimetric analysis (TGA) and differential scanningcalorimetry (DSC) offer further insight into the formation andnature of the active catalyst. A sample of cellulose/Fe(NO3)3 washeated to 1350 °C under Ar. A rapid mass loss around 330 °C(Figure S9, Supporting Information) corresponds to the main de-composition step of cellulose, where carbonization of the polysac-charide releases CO2, H2O, and CO. The small mass loss at605 °C corresponds to carbothermal reduction of Fe3O4 to FeOx,as observed in the XRD data. This is confirmed by the releaseof CO and CO2 at this temperature (Figure S10, SupportingInformation) and the appearance of a small endothermic peakin the DSC data (Figure 4e). The TGA/DSC data do not showthe graphitization step, as this is a transformation from disor-dered, mainly sp3-hybridised carbon to ordered sp2-hybridizedgraphitic carbon. The absence of a melting transition in theDSC data at the graphitization step provides further evidencefor the bulk catalyst being solid during graphitization. Furtherheating of the sample reveals that the Fe-C catalyst undergoesa melting transition at 1282 °C. This is consistent with the Fe-Cphase diagram, where there is a phase boundary between 𝛾-ironand a liquid/𝛾-iron mixture at 1282 °C for an Fe-C alloy of ≈1.3wt% C.[15]2.3. Molecular Dynamics Simulations of the CatalystThe observation of solid catalyst nanoparticles in our system isconsistent with the bulk iron-carbon phase diagram, where liquidphases only exist above 1150 °C. However, the observed liquid-like behavior of catalysts in nanostructure growth is often ratio-nalized with reference to the Gibbs–Thomson effect of meltingpoint depression in nanoparticles. Therefore, we conducted sim-ulations of Fe and Fe-C clusters to determine the effect of par-ticle size on melting temperature. Melting of Fe nanoparticleshas been modeled, but only with older potentials based on thesecond moment approximation of the tight binding model.[18,19]To the best of our knowledge, Fe3C nanoparticle melting has notbeen modeled. Here, a classical molecular dynamics method us-ing LAMMPS (Large-scale Atomic/Molecular Massively ParallelSimulator) was employed to study the melting behavior of freeand substrate-supported iron and Fe3C nanoparticles. Modifiedembedded atom model (MEAM) potentials were used to describethe Fe-Fe, Fe-C, and C-C interactions and a Lennard–Jones po-tential function was used to describe interactions between thenanoparticles and a structureless, mean-field substrate. Full de-tails can be found in the Experimental Section and Support-ing Information. Figure 5a shows snapshots of a heating sim-ulation of a freestanding 𝛼-iron cluster, constructed from 8393atoms using MEAM potentials reported by Asadi et al.[20] Thebody-centered cubic (BCC) structure (blue atoms) breaks downwith increasing temperature to give a disordered cluster (grayatoms), indicating the melting transition. Figure 5b shows caloriccurves for heating and cooling of freestanding BCC iron clus-ters of various sizes. The heating curves show a steady shift tohigher melting points with increased cluster size, as would beexpected. The cooling simulations also show clear transitionscorresponding to solidification. However, there is less correla-tion between solidification temperature and cluster size, with allsolidification occurring in the 1000–1200 K temperature range.The hysteresis observed in all the simulations is due to the veryfast heating and cooling rates required for molecular dynam-ics simulations.[21] Equilibrium melting temperatures (Tm) forall the cluster sizes were calculated from the heating and cool-ing simulations (details in the Supporting Information). A plotof Tm versus 1/radius (Figure 5c) shows a linear trend, consis-tent with Pawlow’s formula[22] and with older simulations us-ing potentials based on the second moment approximation ofthe tight binding model. The data indicate that a liquid ironphase could be possible below 800 °C, but only for particles thatare a lot smaller (≈1 nm diameter) than we observe in the cat-alytic graphitization (50–100 nm diameter). Simulations wereattempted for freestanding 𝛾-iron clusters but at the startingAdv. Mater. 2024, 36, 2404170 2404170 (5 of 10) © 2024 The Author(s). Advanced Materials published by Wiley-VCH GmbH 15214095, 2024, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202404170 by National Institute For, Wiley Online Library on [09/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons Licensehttp://www.advancedsciencenews.comhttp://www.advmat.dewww.advancedsciencenews.com www.advmat.deFigure 5. a) Snapshots of a heating simulation of a freestanding 𝛼-iron cluster constructed from 8393 atoms. b) Caloric curves calculated from heatingand cooling simulations of 𝛼-iron clusters of various sizes. c) Equilibrium melting temperature (Tm) versus 1/radius for 𝛼-Fe clusters with a fit linecalculated using Pawlow’s formula.[22] d) Caloric curves calculated from heating and cooling simulations of Fe3C clusters of various sizes. e) Snapshotsof a heating simulation of 𝛼-Fe cluster consisting of 5065 atoms with a cluster-substrate interaction strength of 1.0 eV. f) Plot of equilibrium meltingtemperature versus cluster-substrate interaction strength (𝜖) for BCC Fe clusters of various sizes. The numbers in the legends in panels (b), (d), and (f)indicate the number of atoms in the cluster.temperature (1100 K), the face-centered cubic (FCC) structuretransformed instantly to BCC, as expected from the phasediagram.Simulations of freestanding Fe3C clusters were carried outusing MEAM potentials reported by Liyanage et al.[23] In heat-ing simulations, a clear jump in energy was observed in thecaloric curves (Figure 5d), corresponding to a melting transition.However, the freezing transition in cooling simulations was lesswell defined, making it difficult to assign a specific solidifica-tion point. This could be due to the metastable nature of theFe3C phase, which will decompose into iron and carbon. Theonly previous simulations of Fe3C melting and solidification in-volved a box half-filled with solid Fe3C atoms and half with liq-uid atoms.[21] This provided a nucleation surface for solidifica-tion. In the freestanding Fe3C clusters, the energy barrier to ho-mogeneous nucleation could be too high, or the rate of nucle-ation is too slow to be captured on the short timescale of the MDsimulation.The nanoparticles involved in catalytic graphitization are inconstant contact with the carbon. Therefore, we also evaluatedthe impact of particle size on melting point for particles in contactwith a substrate. Figure 5e shows a selection of snapshots of a typ-ical simulation, showing the BCC structure of 𝛼-iron (blue atoms)breaking down on melting. Caloric curves were constructed forclusters of a range of sizes and for a range of cluster-substrateinteraction strengths. While melting point decreases withAdv. Mater. 2024, 36, 2404170 2404170 (6 of 10) © 2024 The Author(s). Advanced Materials published by Wiley-VCH GmbH 15214095, 2024, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202404170 by National Institute For, Wiley Online Library on [09/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons Licensehttp://www.advancedsciencenews.comhttp://www.advmat.dewww.advancedsciencenews.com www.advmat.deFigure 6. Images from in situ TEM footage showing a) a moving particle accumulating smaller Fe-C nanoparticles. b) Two moving particles coalescing.c) A particle dividing. d) A particle appearing to squeeze through a small gap and e) a particle sending out a “tongue” of material before retracting.decreasing cluster size, it increases with a higher cluster-substrate interaction strength (Figure 5f). This is consistent withthe concept of effective radius of curvature, where a particlestrongly wetting a surface has a surface curvature equivalent toa much larger particle and displays a melting point closer to thatof the equivalent larger particle. Similar simulations for Fe3Cclusters were conducted. Only heating simulations were consid-ered, given the lack of an observable solidification transition inthe freestanding clusters. A similar trend was observed, with alarger cluster-substrate interaction strength giving a higher melt-ing point (Figure S11, Supporting Information). All the molecu-lar dynamics simulations together suggest that size effects wouldnot result in a significant melting point depression for eitherFe3C or Fe catalyst particles. This is consistent with our exper-imental TEM observations.2.4. Insight into the Catalyst MovementFurther insight into the graphitization mechanism can be gainedfrom closer examination of the in situ TEM footage. The movingcatalyst particles collect smaller stationary particles (Figure 6a)and merge with other moving particles (Figure 6b) as they travelthrough the amorphous carbon matrix, resulting in a general in-crease in size of the particles. The fact that the very small particlesare stationary suggests that there may be a critical size that par-ticles need to reach before they become catalytically active andable to move. In some cases, particles split into two smaller par-ticles (Figure 6c). This indicates that the forward driving force forparticle movement is very strong. As noted above, many catalyststhat appear to be stationary are in fact moving within a confinedregion, presumably because they have been trapped betweenAdv. Mater. 2024, 36, 2404170 2404170 (7 of 10) © 2024 The Author(s). Advanced Materials published by Wiley-VCH GmbH 15214095, 2024, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202404170 by National Institute For, Wiley Online Library on [09/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons Licensehttp://www.advancedsciencenews.comhttp://www.advmat.dewww.advancedsciencenews.com www.advmat.deFigure 7. Schematic of proposed homogeneous solid-state catalysis mechanism, including (surface and/or vacancy) diffusion of iron through the amor-phous carbon matrix and interstitial diffusion of carbon atoms through the iron in the opposite direction, with subsequent precipitation of a graphiticnanotube.regions of graphitic carbon. In some cases, the catalysts discoversmall channels of amorphous carbon and will squeeze throughnarrow gaps to continue graphitization (Figure 6d). This againsuggests there is a very strong driving force for catalyst move-ment. In other cases, catalyst particles send out tongues of cata-lyst material to consume small pockets of residual amorphouscarbon (Figure 6e), again indicating that dissolution of amor-phous carbon is energetically favorable.2.5. Mechanism of Iron-Catalyzed GraphitizationThe remarkable observation of solid-state homogeneous cataly-sis can be rationalized by considering atomic diffusion. Atomsdiffuse through solids via several mechanisms. In iron or ironcarbide, the carbon atoms will be present in the close-packediron lattice on interstitial sites and will diffuse through the lat-tice by moving between interstices. Diffusion of carbon in theiron lattice depends on temperature and carbon concentrationbut is faster in 𝛼-iron (Dc ≈ 10−6 cm2 s−1 at 800 °C) than 𝛾-iron (Dc ≈ 10−8 cm2 s−1 at 800 °C) because of the lower den-sity body-centered-cubic structure.[24] Diffusion of carbon is slow-est in the stoichiometric Fe3C phase (Dc ≈ 10−10 cm2 s−1 at800 °C).[25] Diffusion of iron atoms in iron (self-diffusion) occursby movement of atoms between lattice vacancies. This is by na-ture a slower process than interstitial carbon diffusion (Dc ≈ 10−12cm2 s−1 for Fe in 𝛼-iron at 800 °C),[26] but faster at grain bound-aries or on surfaces.[27] Diffusion of atoms through nanoscalesolids can cause rapid structural transformations. An exampleof this is the diffusion of bismuth atoms outward from a bis-muth nanoparticle through a Bi2O3 shell to form a hollow Bi2O3nanoparticle through the Kirkendall Effect.[28] The diffusion co-efficient of Bi in the Bi2O3 shell was calculated to be 0.75 ×10−14 cm2 s−1 and the formation of hollow Bi2O3 nanoparti-cles from Bi nanoparticles (≈100 nm diameter) was completedwithin 2–5 min. This is a similar timescale to the graphitiza-tion process observed in this paper. Self-diffusion of iron (mostlikely via surface diffusion) is therefore a reasonable mechanismfor the movement of the catalyst particle during graphitization(Figure 7).3. ConclusionsIron-catalyzed graphitization is a fascinating process. A simpleand abundant transition metal can convert raw biomass to adense network of multiwalled carbon nanotubes at moderatetemperatures. On a fundamental level, the observation that thegraphitization catalyst is crystalline is remarkable and raises thequestion of what is meant by the word solid. The ordered struc-ture of the crystalline lattice is consistent with microscopic chem-ical and mechanical definitions of solids.[29] Furthermore, the ev-idence of a melting transition in the DSC well above the graphi-tization temperature indicates that the catalyst has not under-gone a solid-liquid phase transformation when it starts to move.While the observations of the catalyst fit within accepted defini-tions of solids, the behavior certainly challenges our everyday un-derstanding of solids as rigid materials which retain their shape.This paper also opens up many fascinating research questions.What is the driving force for catalyst movement? Why is the cata-lyst movement so erratic? Does the nature of the organic precur-sor impact the way that the catalyst moves? Is Fe3C or 𝛼-Fe theactive catalyst? If we can answer these questions, we may be ableto control catalyst movement and use this simple method to pro-duce carbons with tailored graphitic structure and properties. Itmay even be possible to control the chirality of the resulting nan-otubes. Given the importance of carbon in energy technologies,Adv. Mater. 2024, 36, 2404170 2404170 (8 of 10) © 2024 The Author(s). Advanced Materials published by Wiley-VCH GmbH 15214095, 2024, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202404170 by National Institute For, Wiley Online Library on [09/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons Licensehttp://www.advancedsciencenews.comhttp://www.advmat.dewww.advancedsciencenews.com www.advmat.dethis would represent a huge leap forward in sustainable materialsproduction.4. Experimental SectionIron-Doped Amorphous Carbon Sample Preparation: Cellulose fibers(C6288) and iron (III) nitrate nonahydrate were sourced from Sigma-Aldrich. 0.27 g (0.00 068 moles) of Fe(NO3)3.9H2O was dissolved 15 mLof DI water and the resulting solution added to a beaker containing 5 g ofcellulose fibers (powder). The mixture was manually stirred until the solu-tion had been absorbed. The sample was dried in an air oven at 70 °C togive an orange/yellow solid. The sample was pyrolyzed in an alumina boatcrucible in a tube furnace to 400 °C under nitrogen. The heating rate was5 °C min−1.TEM Measurements: Small portions of iron-doped amorphous carbonsample (≈50 mg) were dispersed in ethanol (≈1 mL) by sonication for10 min. One drop of the dispersion was pipetted on to a Protochips Ther-mal E-chip (E-FHDC). In situ TEM footage was collected on a JEOL JEM-ARM200F equipped with a Schottky field emission gun. The sample washeated at 1 °C min−1 with a Protochips heating holder, FUSION, insidethe microscope. Dark field TEM footage was collected at 900 °C inside themicroscope.In Situ Synchrotron Powder X-Ray Diffraction: In situ X-ray diffractionexperiments were carried out at the materials science powder diffraction(MSPD) beamline of the ALBA-CELLS synchrotron (Cerdanyola del Vallés,Spain). Beam energy was ≈30 keV (wavelength 𝜆 = 0.4127 Å as deter-mined using a Si standard) and scattering was measured in the range ≈1°to ≈60° in 2𝜃 using an array of linear position sensitive detectors (DectrisMythen). Powdered samples were contained in open-ended quartz capil-laries (1 mm inner diameter) and held in place using quartz wool. Heatingwas performed in a capillary flow-cell[30] using a hot-air blower (OxfordFMB), while N2 flowed through the sample at a nominal rate of 20 mLmin−1, which was controlled at the inlet and monitored at the outlet usingmass flow controllers.The samples had been previously heated ex situ in a laboratory furnaceup to 400 °C to release the most volatile products of pyrolysis. During insitu experiments, the samples were heated at 20 °C min−1 up to 400 °C,then at 5 °C min−1 up to 800 °C, held at 800 °C for 2 h and then cooleddown to 100 °C at a rate of 20 °C min–1. Diffraction patterns were mea-sured continuously every minute. Real temperature at the sample was cal-ibrated previously using a K-type thermocouple in place of the sample andlater checked by observing several phase transitions.TGA/MS: Thermogravimetric analysis (TGA) coupled with differen-tial scanning calorimetry (DSC) experiments were carried out using a dualsystem (SDT Q600, Thermal Advantage Instruments). Samples consistingof raw cellulose and cellulose impregnated with Fe(NO3)3 solution weremeasured from 100 to 1350 °C using a heating rate of 5 °C min−1 underflowing Ar at 100 mL min−1. An initial mass of ≈10 mg was used, andsamples were contained in brand-new alumina pans that were previouslywashed in 1 m HCl solution to discard the presence of any metal contam-inants. The exhaust of the system was connected to a mass spectrometer(GSD 320 Omnistar, Pfeiffer Vacuum) to qualitatively evaluate the com-position of the gases evolved during pyrolysis. A secondary electron mul-tiplier detector polarized at 970 V was used, and m/Z ratios of 10 to 50u.m.a were measured at a scan rate of 200 ms/u.m.a.Molecular Dynamics Simulations: Free Standing Clusters: Clusters ofBCC-Fe, FCC-Fe, and Fe3C were constructed in LAMMPS using lattice pa-rameters extracted from crystallographic data and a spherical cluster witha certain radius was defined. For pure Fe clusters, the Fe–Fe interactionswere modeled using a modified embedded atom model (MEAM) poten-tial reported by Asadi et al.[20] For Fe-C clusters, the MEAM potential re-ported by Liyanage et al. was used to describe the Fe-Fe, Fe-C, and C-Cinteractions.[23]The classical equations of motion were integrated using a timestep of2.0 fs and the temperature was controlled using the Langevin thermostat.Periodic boundary conditions were employed in the x, y, and z directions.The clusters were initially annealed at the starting temperature for 10 nsbefore the start of a simulation. Heating and cooling simulations were thencarried out at a rate of 16 K ns−1 and statistical averages were generated.Trajectories were visualized using OVITO.Molecular Dynamics Simulations: Substrate-Supported Fe Clusters: Thesame procedure was carried out to construct clusters with a BCC-Fe,FCC-Fe, or Fe3C structure and the same Asadi and Liyanage potentialswere used to describe the interactions within the cluster. A fixed wall sub-strate was introduced along the z-plane of the simulation box and periodicboundary conditions were only employed in the x and y directions. Thecluster-substrate interaction was described using a Lennard–Jones 12–6potential. The distance of closest approach (𝜎) in the Lennard–Jones po-tential was fixed at 2.91 Å. The well depth (𝜖) was varied from 0.05 to 1.0 eVto model the variable strength of interaction between the clusters and thesubstrate.The classical equations of motion were also integrated using a timestepof 2.0 fs. The angular momentum of the clusters in the x-y plane were fixedat each step to prevent movement of the cluster along the substrate dur-ing the simulation. To reduce computational cost, only the atoms withinthe nanoparticles were heated in the NVE ensemble using a Langevin ther-mostat. The clusters were initially annealed at a fixed distance of 2𝜎 fromthe substrate at the starting temperature of the simulation for 10 ns. Thecluster was then relaxed at the initial temperature for 5 ns to allow for thecluster to wet onto the substrate. Heating and cooling procedures werethen carried out as with the freestanding clusters.Molecular Dynamics: Calculation of Melting/Freezing Transitions: Themelting/freezing transition of the clusters was characterized by plottingcaloric curves (E(T) vs T). The phase transition temperature was also char-acterized from the heat capacities, which were calculated from the fluctu-ations of energy using Equation (1)CV =⟨E2⟩− ⟨E⟩2kBT2(1)In general, the combination of the two methods were used to confirmthe temperature at which melting or freezing occurred. To avoid effects ofsuperheating/supercooling, melting, and freezing simulations were runfor each cluster at the same heating/cooling rate. For most of the clusterstested, there was some level of hysteresis, where the freezing transition oc-curs at a lower temperature than the melt. This is a common phenomenondue to the fast heating and cooling rates required for MD and were ac-counted for by applying Equation (2) to calculate an equilibrium meltingtemperature, TmTm = Tc+ −√Tc+Tc− + Tc− (2)where Tc+ = T of melting transition and Tc− = T of freezing transition.Supporting InformationSupporting Information is available from the Wiley Online Library or fromthe author.AcknowledgementsZ.S. acknowledged the Leverhulme Trust (No. RPG-2020-076) and the Uni-versity of Birmingham for funding. J.R.-R. acknowledged funding fromthe Spanish Ministry of Science, Innovation and Universities (Grant No.PID2019-107019RB-I00) and from the Junta de Andalucía regional govern-ment (Grant Nos. P20-01186 and US-1380856, with funds from FEDER).In situ x-ray diffraction experiments were performed at the MSPD beam-line at ALBA synchrotron with the collaboration of ALBA staff (ExperimentNo. 2021024916).Conflict of InterestThe authors declare no conflict of interest.Adv. Mater. 2024, 36, 2404170 2404170 (9 of 10) © 2024 The Author(s). Advanced Materials published by Wiley-VCH GmbH 15214095, 2024, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202404170 by National Institute For, Wiley Online Library on [09/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons Licensehttp://www.advancedsciencenews.comhttp://www.advmat.dewww.advancedsciencenews.com www.advmat.deData Availability StatementThe data that support the findings of this study are available from the cor-responding author upon reasonable request.Keywordsgraphitization, homogeneous solid-state catalysis, nanotubeReceived: March 21, 2024Revised: June 24, 2024Published online: July 16, 2024[1] L. Güniat, P. Caroff, A. Fontcuberta i Morral, Chem. Rev. 2019, 119,8958.[2] J. Pang, A. Bachmatiuk, I. Ibrahim, L. Fu, D. Placha, G. S. Martynkova,B. Trzebicka, T. Gemming, J. Eckert, M. H. Rümmeli, J. Mater. Sci.2016, 51, 640.[3] K. W. Kolasinski, Curr. Opin. Solid State Mater. Sci. 2006, 10, 182.[4] R. D. Hunter, J. Ramírez-Rico, Z. Schnepp, J. Mater. Chem. A. 2022,10, 4489.[5] R. D. Hunter, J. L. Rowlandson, G. J. Smales, B. R. Pauw, V. P. Ting,A. Kulak, Z. Schnepp, Mater. Adv. 2020, 1, 3281.[6] S. Glatzel, Z. Schnepp, C. Giordano, Angew. Chem., Int. Ed. 2013, 52,2355.[7] O. P. Krivoruchko, V. I. Zaikovskii, Mendeleev Commun. 1998, 8, 97.[8] O. P. Krivoruchko, N. I. Maksimova, V. I. Zaikovskii, A. N. Salanov,Carbon 2000, 38, 1075.[9] H. Rastegar, M. Bavand-vandchali, A. Nemati, F. Golestani-Fard,Phys. E 2018, 101, 50.[10] E. Thompson, A. E. Danks, L. Bourgeois, Z. Schnepp, Green Chem.2015, 17, 551.[11] N. D. Klein, K. R. Hurley, Z. V. Feng, C. L. Haynes, Anal. Chem. 2015,87, 4356.[12] L. Wang, J. Teng, P. Liu, A. Hirata, E. Ma, Z. Zhang, M. Chen, X. Han,Nat. Commun. 2014, 5, 4402.[13] Y. Wang, L. Qiu, L. Zhang, D.-M. Tang, R. Ma, C.-L. Ren, F. Ding, C.Liu, H.-M. Cheng, Sci. Adv. 2022, 8, eabo5686.[14] M. S. Chambers, R. D. Hunter, M. J. Hollamby, B. R. Pauw, A. J.Smith, T. Snow, A. E. Danks, Z. Schnepp, Inorg. Chem. 2022, 61,6742.[15] J. Chipman, Metall. Trans. 1972, 3, 55.[16] Z. He, J.-L. Maurice, A. Gohier, C. S. Lee, D. Pribat, C. S. Cojocaru,Chem. Mater. 2011, 23, 5379.[17] H. K. D. H. Bhadeshia, J. Mater. Sci. 2004, 39, 3949.[18] F. Ding, A. Rosén, S. Curtarolo, K. Bolton, Appl. Phys. Lett. 2006, 88,133110.[19] V. Rosato, M. Guillope, B. Legrand, Philos. Mag. A 1989, 59, 321.[20] E. Asadi, M. A. Zaeem, S. Nouranian, M. I. Baskes, Phys. Rev. B. 2015,91, 024105.[21] S. N. Luo, A. Strachan, D. C. Swift, J. Chem. Phys. 2004, 120, 11640.[22] P. Pawlow, Z. Phys. Chem. 1909, 65, 1.[23] L. S. I. Liyanage, S.-G. Kim, J. Houze, S. Kim, M. A. Tschopp, M. I.Baskes, M. F. Horstemeyer, Phys. Rev. B. 2014, 89, 094102.[24] J. Cermak, L. Kral, J. Alloys Compd. 2014, 586, 129.[25] A. Schneider, G. Inden, Calphad 2007, 31, 141.[26] Y. Iijima, K. Kimura, K. Hirano, Acta Metall. 1988, 36, 2811.[27] A. Inoue, H. Nitta, Y. Iijima, Acta Mater. 2007, 55, 5910.[28] K.-Y. Niu, J. Park, H. Zheng, A. P. Alivisatos, Nano Lett. 2013, 13,5715.[29] S. Alexander, Physica A 1998, 249, 266.[30] P. J. Chupas, K. W. Chapman, C. Kurtz, J. C. Hanson, P. L. Lee, C. P.A. Grey, J. Appl. Crystallogr. 2008, 41, 822.Adv. Mater. 2024, 36, 2404170 2404170 (10 of 10) © 2024 The Author(s). Advanced Materials published by Wiley-VCH GmbH 15214095, 2024, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202404170 by National Institute For, Wiley Online Library on [09/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons Licensehttp://www.advancedsciencenews.comhttp://www.advmat.de Elucidating the Mechanism of Iron-Catalyzed Graphitization: The First Observation of Homogeneous Solid-State Catalysis 1. Introduction 2. Results and Discussion 2.1. In Situ Transmission Electron Microscopy 2.2. In Situ Synchrotron X-Ray Diffraction 2.3. Molecular Dynamics Simulations of the Catalyst 2.4. Insight into the Catalyst Movement 2.5. Mechanism of Iron-Catalyzed Graphitization 3. Conclusions 4. Experimental Section Supporting Information Acknowledgements Conflict of Interest Data Availability Statement Keywords