Multiscale design of nanofibrous carbon aerogels: Synthesis, properties and comparisons with other low-density carbon materials

Accepted Manuscript Multiscale design of nanofibrous carbon aerogels: Synthesis, properties and comparisons with other low-density carbon materials

Mark A. Atwater, Roger J. Welsh, David S. Edwards, Laura N. Guevara, Christopher B. Nelson, Ben T. Stone PII:

S0008-6223(17)30916-8

DOI:

10.1016/j.carbon.2017.09.041

Reference:

CARBON 12371

To appear in:

Carbon

Received Date:

09 June 2017

Revised Date:

31 August 2017

Accepted Date:

11 September 2017

Please cite this article as: Mark A. Atwater, Roger J. Welsh, David S. Edwards, Laura N. Guevara, Christopher B. Nelson, Ben T. Stone, Multiscale design of nanofibrous carbon aerogels: Synthesis, properties and comparisons with other low-density carbon materials, Carbon (2017), doi: 10.1016/j. carbon.2017.09.041

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Multiscale design of nanofibrous carbon aerogels: Synthesis, properties and comparisons with other low-density carbon materials Mark A Atwater1,*, Roger J Welsh1, David S Edwards1, Laura N Guevara1, Christopher B Nelson1, Ben T. Stone2 1

Department of Applied Engineering, Safety & Technology, Millersville University,

Millersville, PA, USA 2

Department of Chemistry, Millersville University, Millersville, PA, USA

Abstract By leveraging decades of research on the catalytic synthesis of carbon nanofibers and combining it with unique processing methodology, low density monolithic carbon can be created. The process is simple, scalable and controllable. Analysis of deposition kinetics reveals that the final density of centimeter-scale carbon monoliths can be determined through basic process parameters, and the synthesis can be performed with a single step in as little as one hour. In general, this material can be created over a large range of density (20-700 mg/cc), possesses high surface area (~215 m2/g), is comprised of a mixture of straight and twisted carbon nanofibers (~150-500 nm in diameter) with low crystallinity, which results in low electrical conductivity (1.32-2.83 S/m) and low thermal conductivity (~0.03 W/mK), even at the highest densities. The materials are mechanically robust, with density-dependent elasticity values of 60-600 kPa. The methods and properties are discussed in context with other low-density carbon materials. The principal advantages of the process used in this work include the simplicity, direct control of density, and most notably, the ability to create this material with specific bulk geometry.

*Corresponding Author. E-Mail: [email protected] (Mark Atwater); Tel.: +1-717871-7217; Fax: +1-717-871-7931.

ACCEPTED MANUSCRIPT

1

Introduction

Advanced carbon materials continue to be an important area of development as their performance and versatility enable a diversity of applications. These applications are often enhanced in low density, high surface area structures, and this is exemplified by aerogels. The prevalence and diversity of aerogel materials has grown rapidly in the last few decades. As shown in Fig. 1, the overall rate of publication regarding aerogels has accelerated over the past 5 years, and that is partially attributable to the increasing publication rate concerning carbon aerogels (CAs). For instance, CA publications comprised approximately 23% of total publications in 2006, but in 2016 they accounted for 44%. Therefore, CAs are an important contributor to the scientific field and will likely provide a proportional contribution to commercial applications as the technology develops.

Fig. 1 - Trends in the publication rate regarding aerogel materials on general, and carbon aerogels in particular. Publication frequency determined using Web of Science search for articles on topic of “aerogel” and refined with addition of “carbon.” Commercial applications are satisfied by material properties, not by material classifications, so a variety of comparable structures will be considered alongside “traditional” aerogels, which generally have a monolithic structure with an open network of pores and possess 99% or greater open volume (a final density of ~20 mg/cm3 or less for carbon-based aerogels). Recent examples

ACCEPTED MANUSCRIPT

of CAs and related materials include the use of carbon nanotubes (CNTs) [1-3] and graphene [47], and although CAs can be processed by sol-gel methods [8, 9], as used with traditional silica aerogels [10], other processes such as carbide-derived carbons [11, 12], cellulose-derived carbon [13-15] and chemical vapor deposition techniques [1, 16] are being developed. The processing and resulting properties of these materials vary greatly, as do the complexity, scalability and efficiency of the methods. The material analyzed in this work is produced in a fundamentally different fashion. It leverages decades of work on carbon nanofiber (CNF) synthesis to create and control bulk, monolithic structures. The process results in a nonwoven nanofibrous carbon material that can vary in density from the upper limit of aerogels (~20 mg/cc) to an order of magnitude greater, and that density is directly controlled through process conditions. Based on simple modifications to the process, the material’s morphological, mechanical, electrical and thermal properties can also be controlled. For the purpose of describing this work and drawing comparisons to prior work, the properties and applications of the materials will be the primary consideration, and the structural differences will be highlighted within that context. Comparable materials include low-density carbon sponges [1, 17], mats [18, 19], foams [20, 21], fibrous [16, 22] or tubular [23] networks, and other nontraditional aerogel structures. The simplicity, scalability and functional benefits of these carbon aerogels and analogous materials are described below.

2 2.1

Experimental Materials and Methods

CNFs were synthesized through the catalytic decomposition of ethylene over a nickel-copper catalyst. The catalyst was prepared by mechanically alloying elemental powders of nickel (Sigma Aldrich, 99.7% pure, < 50 µm) with copper (Sigma Aldrich, 99% pure, < 75 µm) in a 7:3 atomic ratio, as described elsewhere [24]. To prevent cold-welding of the material, 1 wt% stearic acid was added. 5 g of the powder was milled for 4 h in a SPEX 8000D high-energy ball mill using a ball-to-powder mass ratio of 10:1.

ACCEPTED MANUSCRIPT

CNF deposition was conducted in a stainless steel mold with an internal diameter of 5 cm and height of 1.9 cm, and catalyst was evenly distributed by carefully sprinkling it over the bottom of the mold before placing it in the furnace. The mold was sealed by polished mating surfaces and secured with threaded fasteners. This mold was then heated in the center of a 15 cm diameter, three-zone, Lindberg/Blue M tube furnace. The reaction conditions include a processing temperature of 550 °C reached after a 35 min ramp time under constant 4:1flow ratio of ethylene (C2H4) to hydrogen (as 5% H2 in Ar) controlled by digitally programmed MKS G-series mass flow controllers. A regulator was placed upstream to limit the maximum system pressure, which was monitored over the duration of the reaction using an Additel ADT 680 wireless pressure gauge and accompanying software to log the data. The flow rate was monitored at the exit of the mold using a 1000 mL/min Omega rotameter. The carbon was extracted from the mold at the conclusion of each reaction and weighed to determine the geometric density. To accommodate thermal testing, a separate mold was created (internal diameter of 14.4 cm and a height of 1.3 cm), but otherwise the process was the same. The general process flow is given schematically in Fig. 2.

Fig. 2 - Basic method to produce and test bulk nanofibrous carbon aerogels. 2.2

Material Characterization

Surface analysis was performed using a Micromeritics ASAP 2420 to conduct nitrogen adsorption/desorption measurements. Samples were degassed at 200 °C (with 10 °C ramp rate)

ACCEPTED MANUSCRIPT

for 4 h. Surface area was determined by Brunauer–Emmett–Teller (BET) analysis, and pore volume and size were assessed using Barrett-Joyner-Halenda (BJH) analysis. Scanning electron microscopy (SEM) was performed using a Zeiss Auriga 60 Crossbeam microscope operated at 3 kV (for general morphology) or a Hitachi SU-5000 operated at 5 kV (for ion beam cross-sections). Broad beam ion milling was performed on as-synthesized material using cross-sectional Ar ion milling with a Hitachi IM4000Plus system. Transmission electron microscopy (TEM) was performed using an FEI Talos 200C microscope operated at 200 kV. TEM samples were prepared by breaking up the monolith, sonicating loose fibers in ethanol and distributing dropwise on a lacey carbon grid. Compression testing of the monoliths was performed using a Shimadzu AGS-X load frame equipped with a 1 kN load cell. The maximum force applied during compression testing was limited to 950 N to avoid reaching the load cell limit. Compressive force was applied at a displacement rate of 0.1 cm/min until the maximum force was reached. After reaching the maximum force, it was held constant for 1 min to assess any viscoelastic response. Cyclic compression testing was also performed with this setup using a load/unload displacement rate of 1 cm/min between 2.5 kPa and 25 kPa over 10 cycles. Each sample was pre-loaded to 2.5 kPa before beginning the tests. Electrical testing was conducted with a Keysight U3606B power supply/digital multimeter. Samples were placed between copper plates 5 cm in diameter and ~ 0.3 cm thick. Wire leads were soldered directly to these plates and current was applied across the entire setup while measuring voltage using the four probe method. The resistance of the copper plates and wire was found to be 12 mΩ, which was negligible compared to the material being measured. As noted by Mikhalchan et al. [16], the contact resistance can be reduced by adding a conductive paste, but that paste can be absorbed uncontrollably and result in errant electrical resistance values. To avoid this completely, we applied a 50 g weight to the top plate before testing. The top copper plate mass was 90 g, for a total applied load of 140 g. On all samples this was found to create a uniform contact surface. The copper plates were polished smooth on the contact surfaces, and the surfaces were cleaned again using #0000 steel wool directly before measuring to remove surface

ACCEPTED MANUSCRIPT

oxidation. The electrical conductivity of the material was then measured using a four probe method by stepping up the applied current to a maximum of 1 A in 0.05 A increments. Each step was held constant for 5 s and the resulting voltage was recorded. Resistance was determined by the slope of the I-V curve. Resistivity was based on the dimensions of the carbon monoliths (5 cm diameter, 1.9 cm height), and conductivity was taken as the inverse. The electrical resistance of the carbon was measured directly during cyclic compression testing (described above) using a Keysight four-point Kelvin probe set in conjunction with the multimeter described. Thermal analysis could not be completed with available equipment using 5 cm x 1.9 cm (diameter x height) monoliths, as the thermal conductivity was too low. Instead, testing was performed via guarded hot plate analysis using a TA Instruments Fox 200 heat flow meter on 10 cm x 10 cm x 1.3 cm samples with a lower plate temperature of 10 °C and top plate temperature of 35 °C. Average temperature was ~ 22 °C during testing, and 10 measurements were made over the span of 1 h. Reported values are the average of these measurements. Variation between measurements was below 2.5% in all cases.

3 3.1

Results Deposition Kinetics

The synthesis method used here, known as the constrained formation of carbon nanofibers (CoFFiN) process [25], fills a mold with CNFs until gas flow is eventually restricted by the deposited carbon. The maximum pressure is dictated by a pressure regulator or pressure relief mechanism, and that pressure will determine the reaction duration. To track growth kinetics, the pressure at the mold inlet was monitored over the duration of the reaction. It was found that the maximum final pressure is roughly proportional to the initial catalyst load, and the time before pressure builds is longer for low catalyst charges (e.g., less than 10 mg). For instance, in Fig. 3 it can be seen that most reactions begin to build pressure after approximately 90 min, but when 9.8 mg of catalyst is used, the pressure begins to increase after 116 min, and 5.6 mg of catalyst does not build pressure at all.

ACCEPTED MANUSCRIPT

A series of pressure spikes can be seen soon after the pressure begins to build, and in some cases carbon is seen at the exhaust of the furnace, evidently caused by the movement of the nanofibers within the mold. The lowest catalyst loadings result in lower pressure being developed later in the reaction, but above a critical value (~20 mg, as discussed next) the additional carbon deposition can prematurely block flow. For instance, in Fig. 3B, the reaction with highest catalyst load (51.8 mg) takes longer to build pressure than lower loadings, even though the initial spikes in pressure occur sooner than other runs. This indicates the carbon has filled the volume sufficiently to interfere with gas flow sooner, but the density of the deposited carbon is low enough to permit the carbon to be moved from the gas flow path or to be ejected with the exhaust. Therefore, where sufficient catalyst is added to rapidly fill the mold, the material dynamics within the mold (i.e., deposition and flow-induced carbon movement) will have the most distinct effect on the pressure.

Fig. 3 – Representative pressure curves (A) over the 755 min reaction time (including temperature ramp) and (B) magnified view of marked inset in (A) for CNF monoliths formed using the indicated initial catalyst loads. The overall carbon deposition is also proportional to the initial catalyst loading as shown in Fig. 4. The carbon to catalyst deposition ratio (Fig. 4A) shows that there are two limitations to maximizing this ratio. For very small catalyst loads, the final mass of carbon is limited by catalyst deactivation. This is supported by the finding that the lowest catalyst loads did not build as much pressure as higher loadings (see Fig. 3), and that is attributed to the carbon deposition stopping before the mold is thoroughly filled and reaching a density where gas flow is blocked.

ACCEPTED MANUSCRIPT

As the catalyst load increases, the carbon deposition causes substantial growth which prevents further deposition by blocking gas flow before the catalyst is deactivated. The gradual decline in final deposition ratio occurs at increasing catalyst load results in a 50% drop in catalytic efficiency (gram carbon /gram catalyst), but it also results in different structural properties, as discussed later. The final densities of the material for these 12 h reactions range from 28.8 mg/cc to 169.4 mg/cc. Greater amounts of catalyst result in higher density monoliths, and, as shown in Fig. 4B, the increase in density approximately follows a logarithmic trend (𝑦 = 60.63𝑙𝑛 𝑥 ‒ 76.006; 𝑅2 = 0. 8895). This deposition behavior, then, can be used to create materials of a given density through initial catalyst load. The uniformity of the density is also controlled by the initial placement of the catalyst, movement of the material during processing and the growth uniformity when the mold fills [26]. While the process is dependent on these factors, it is not highly sensitive, and uniform distribution can be achieved without precise attention.

Fig. 4 – (A) The ratio of carbon deposition to catalyst charge is limited by catalyst deactivation or the blockage of gas flow by the deposition. (B) The final density of the monolith is controlled by several process factors, but the initial catalyst load is the most influential and follows a logarithmic relationship (𝒚 = 𝟔𝟎.𝟔𝟑𝒍𝒏 𝒙 ‒ 𝟕𝟔.𝟎𝟎𝟔; 𝑹𝟐 = 𝟎. 𝟖𝟖𝟗𝟓). As shown in Fig. 4A, the maximum deposition rate is achieved near 20 mg of catalyst. To avoid a reaction being limited by catalyst activity, a catalyst loading above 20 mg is desirable, but

ACCEPTED MANUSCRIPT

excessive catalyst loading is less efficient over the 12 h reaction. Therefore, a fixed catalyst load of 23 mg (±0.5 mg) was used to produce monoliths with reaction times up to 12 h in 1 h intervals, and the resulting carbon deposition rates and final material densities are shown in Fig. 5.

Fig. 5 – Deposition kinetics and final density of nanofibrous components at reaction times up to 12 h using 23 mg (±0.5 mg) catalyst The increase in density is highly linear up to 8 h of reaction time (𝑦 = 15.95𝑥; 𝑅2 = 0.974 ), after which the rate of increase lowers. This trend is expected to be driven by fast, unconstrained growth in the early stages (1-2 h reactions), a slowing of gas flow as pressure builds in the midrange durations (3-8 h reactions) and slower, restricted growth through the end of process (9-12 h reactions). Pressure was found to reach its maximum in 2 h for most reactions (see Fig. 3), after which the deposition rate drops sharply. From that point, the deposition rate remains fairly steady until hour 9. That indicates that pressure alone is not indicative of the growth rate. Since the pressure is regulated to approximately 3 psig or below, the flow continues without building additional pressure, and it is likely that once the maximum pressure is reached, remaining flow continues until it exceeds the limit of the setup. Specifically, gases may be diverted out of the mold between the mating surfaces of the lids and body, especially given the rudimentary nature of the design. It is feasible, then, that additional growth at comparable rates can be achieved by raising the maximum pressure and improving the sealing capability of the mold.

ACCEPTED MANUSCRIPT

3.2

Carbon Nanofiber Characteristics

The material created in this work is entirely comprised of vapor-deposited CNFs, so their properties are foundational to the bulk properties and applications. As a basic assessment of the material, the monoliths were sectioned using a razor blade to examine the fiber properties in various regions. This is potentially important as the fiber growth starts at the bottom where the catalyst is distributed and later reaches the top of the mold and pressure begins to build. Because the catalyst has been noted to continuously break up during the deposition process [27, 28], the fibers generated and their characteristics may also change. As shown in Fig. 6, the surface properties were characterized for the bottom, middle and top thirds of a carbon monolith. The BET surface areas for these sections were 212 m2/g (bottom), 218 m2/g (middle), and 215 m2/g (top), indicating no significant difference. Furthermore, the BJH pore size distributions and adsorption/desorption behavior were also consistent between areas (Fig. 6A and Fig. 6B, respectively). Fig. 6B shows that the isotherms follow an IUPAC Type II adsorption/desorption behavior with an H4 hysteresis loop, which are indicative of being nonporous or having a large proportion of micropores, which is consistent with BJH analysis (Fig. 6A).

ACCEPTED MANUSCRIPT

Fig. 6 – Surface analysis for carbon nanofibers from separate sections of a monolith regarding (A) BJH pore size distribution and cumulative pore volume, (B) adsorption (solid line) and desorption (dashed lines) isotherms, and (C) schematic representation of the volumes analyzed. As shown in Fig. 7, the spacing of fibers imaged from the top (Fig. 7A, B), middle (Fig. 7C, D) and bottom (Fig. 7E, F) of the monolith decreases, indicating the point at which growth starts will have a slightly higher density. To quantify the difference, carbon monoliths of varying initial density were carefully sectioned to separate the bottom, middle and top volumes, and the density of each was measured separately. This revealed that the middle and top sections had relative densities of 67.4% (±8.7%) and 63.5% (±3.8%), respectively, when compared to the bottom section. Variation in section density does not appear to be related to the initial bulk density of monolith, but in prior work [25, 26] it has been noted that the placement of catalyst plays a decisive role in determining the growth characteristics and final bulk density. Therefore, the density gradient is more likely controlled by the initial catalyst distribution (in three-dimensions) than by the amount of catalyst. This includes the aspect ratio of the mold cavity such that increasing height or decreasing diameter will likely increase the gradient. The initial catalyst distribution is also affected by the method used to apply the catalyst powder and any jostling of

ACCEPTED MANUSCRIPT

the mold during handling which may shift the powder. From higher magnification imaging (Fig. 7B, D, F), the fiber characteristics, such as diameter and morphology, are consistent through the material, being comprised of a combination of larger fibers with a relatively straight and smooth morphology and smaller, highly twisted fibers.

Fig. 7 – A 131.6 mg/cc monolith was cross-sectioned to examine the (A, B) top, (C, D) middle and bottom (E,F). The fiber morphology and spacing was examined using SEM, and the fiber density decreases slightly from bottom to top, but the fiber size and morphology is consistent throughout. The material cuts cleanly with a razor, but the cutting process is destructive to the nanoscale structure. The appearance of the undisturbed structure was of interest, so broad beam ion milling was applied to create a large, damage-free surface as shown in Fig. 8. Although the section is much nearer to the as-produced condition, some disturbance is still possible as the sample for ion milling had to be razor cut from a larger monolith and the cutting may have some proximity effect. Despite this, the appearance is markedly different from Fig. 7, as the fibers are sectioned at varying orientations. The images reveal the fibers are closely spaced with variable bulk

ACCEPTED MANUSCRIPT

density over distances of 10s of microns, but the random fiber orientations and nonlinear nature of the fibers prevent tight packing over larger distances.

Fig. 8 - Ion-milled section of nanofibrous nonwoven carbon. As shown in Fig. 9A, TEM examination confirms the co-existence of twisted and straight fibers, with twisted fibers being more numerous in samples prepared from the top section of the material. Although catalyst particles (Fig. 9B and Fig. 9C) were found in all sections, they were more prevalent in fibers from the bottom. This indicates that the majority of the catalyst remains near the origin, which is consistent with the higher density observed there. The catalyst particles within twisted fibers (Fig. 9B) appear highly faceted so that their structure is not apparent from the profile, but particles in the straight fibers (Fig. 9C) were consistently diamond-shaped. Twisted morphology is attributed to uneven deposition of solid carbon due to differing crystallographic orientations or alloy distribution, and the correlation between fiber characteristics and particle structure has been described elsewhere [29, 30]. No long-range

ACCEPTED MANUSCRIPT

crystallinity is observed in any fibers (see Fig. 9D), and catalyst particles are present along the length of the fibers, rather than at the tip or base as sometimes observed (e.g., [31]).

Fig. 9 – TEM images of (A) carbon nanofibers with catalyst particles typical of (B) twisted fibers and (C) straight fibers. (D) High magnification of carbon near a catalyst particle reveals a disordered structure. 3.3

Mechanical Testing

The mechanical properties of the as-produced materials were assessed through compression testing (see Fig. 10). Each monolith was compressed to near the load cell limit (950 kN), and the strain at that applied force varied with the density of the material. For instance, stress-strain curves for a range of densities are shown in Fig. 10B. It can be seen that higher density samples exhibit more linear stress-strain curves and reach the maximum stress at lower strain values. For example, the maximum strain for an initial density of 169.4 mg/cc is 49.3% and 71.6% for an initial density of 34.1 mg/cc. The elasticity was assessed for each sample using the initial slope of the curve and the values are plotted in Fig. 10C. As density increases, the elasticity increases

ACCEPTED MANUSCRIPT

exponentially (𝑦 = 0.0455𝑒0.0158𝑥; 𝑅2 = 0. 9483), and the elasticity was found to range from 58 kPa for a 28.8 mg/cc sample to 591 kPa for the 169.4 mg/cc sample. Low density samples, as shown in Fig. 10A, maintain their integrity throughout the compression test, and they recover some of their height after removing the stress, but higher density monoliths tend to crumble some at the edges during compression. Despite differences in stress-strain response, all samples exhibited a similar, and very limited, viscoelastic response indicated by the vertical drop at the end of each compression curves shown in Fig. 10B. Force was held for 1 min after reaching the maximum, and variation in loading rate and relaxation time produced no significant difference. Although there is some response, the magnitude of this response is relatively minor.

Fig. 10 – (A) Compression testing of monoliths was performed, and the materials were able to hold together under strain exceeding 70%. (B) As the material density increases, the stress-strain curves become more linear and reach lower strain values at the maximum stress. (C) The elasticity of the materials was determined as a function of density and exhibits an exponential trend with increasing density (𝒚 = 𝟎.𝟎𝟒𝟓𝟓𝒆𝟎.𝟎𝟏𝟓𝟖𝒙; 𝑹𝟐 = 𝟎. 𝟗𝟒𝟖𝟑).

ACCEPTED MANUSCRIPT

Cyclic compression testing was performed on samples of varying density as shown in Fig. 11. The test was cycled between 2.5 kPa and 25 kPa, and the samples were pre-loaded at 2.5 kPa (~5 N) at the beginning of the test. From Fig. 11A it is evident the magnitude of strain increases as density decreases, as also demonstrated in Fig. 10A. With cyclic testing it is additionally determined that the maximum strain (Fig. 11B) increases significantly with each cycle for samples below 100 mg/cc, but those above are relatively stable, indicating they recover elastically after each cycle. The amount of strain for each cycle (Fig. 11C) also indicates that the lowest density sample (41.8 mg/cc) shows a distinct drop with each cycle, likely attributable to an increasing densification as the entire structure is compressed permanently (compare to Fig. 10A). At the applied stress levels, the higher density samples (> 100 mg/cc) are largely unaffected.

Fig. 11 - Cyclic compression testing of samples showing (A) the stress-strain curves collected over 10 cycles for samples of varying density, (B) the magnitude of strain between each load-

ACCEPTED MANUSCRIPT

unload cycle (samples between 131.1mg/cc and 157.0 mg/cc not shown due to excessive overlap) and (C) the maximum strain after each loading cycle. 3.4

Electrical and Thermal Testing

The electrical conductivity of samples was measured using the four point probe method as described in Section 2.2. The relationship between applied current and measured voltage is shown in Fig. 12A for samples ranging between 41.8 mg/cc and 157.0 mg/cc. The density does not directly determine the electrical conductivity as measured in this work. Since a 140 g load was applied to ensure proper electrical contact without the use of silver paste, a greater level of strain in lower density samples may raise the conductivity of those samples. Despite these differences, the conductivity showed little variation compared to other carbon materials, ranging from 1.32 S/m to 2.83 S/m. This is an order of magnitude lower than comparable carbon materials (see Fig. 12B), and this is attributed to the crystallographic disorder and disconnected fibrous nature of the material.

Fig. 12 – Electrical properties are shown by (A) current-voltage diagrams for samples of various densities (not all samples shown due to excessive overlap), and (B) the electrical conductivities of other low density carbon materials are compared to values determined in this study. Literature values are from Mikhalchan 2016 [16], Chen 2011 [7], Zhang 2011 [5], Zou 2010 [2], and Lu 1993 [32].

ACCEPTED MANUSCRIPT

Electrical resistance was monitored during the cyclic compression testing described in Section 3.3, and the changes in resistance with applied strain are given for the lowest density (41.8 mg/cc, Fig. 13A) and highest density (157.0 mg/cc, Fig. 13B) samples. The resistance drops with increasing compressive strain, and the variation in resistance is similar (~ 1.5 – 3.0 Ω) despite the large differences in strain. To better compare the effect of density on resistance, the change in resistance was compared to the applied strain, ε, and the resulting data is provided in Fig. 13C. It can be seen that strain sensitivity is greatest in the samples of highest density, and those below 100 mg/cc respond similarly. All samples show a relatively stable response over the 10 cycles, but the higher density samples display a steady reduction in sensitivity. This occurs even though the mechanical response is more stable (see Fig. 11), and it may be partially attributed to the lower overall strain at the start of testing or the possible degradation of the more tightly packed fibers during testing.

Fig. 13 - Cyclic compression and electrical testing curves from (A) the lowest density and (B) highest density materials (note the order of magnitude difference in strain between (A) and (B) and the corresponding difference in time), and (C) the variation in resistance with strain over the 10 cycles tested. As described in Section 2.2, thermal conductivity was measured on samples produced in a different mold in order to achieve the necessary ratio of cross-sectional area to thickness, and this resulted in materials of higher density (487-700 mg/cc). Despite the increased density, the thermal conductivity of 0.030 W/mK is among the lowest measured in literature regarding lowdensity carbon materials (see Fig. 14). This result is an order of magnitude lower than the next

ACCEPTED MANUSCRIPT

lowest materials in this density range, including “mats” created from CNFs [19]. The remarkably low values for electrical and thermal conductivity are consistent with one another, and the hierarchical disorder in the structure is considered the primary contributor.

Fig. 14 - Thermal conductivity for material produced in this work as well as literature values for other low-density carbon materials. Literature values are from Jia 2016 [33], Lv 2016 [34], Mikhalchan 2016 [16], Feng 2012 [35], Mahanta 2010 [19], Worsley 2010 [6], and Lu 1993 [32].

4

Discussion

The carbon nanomaterial described here does not fit a traditional aerogel definition, but in many respects it can serve in comparable applications that justify similar assessment. Indeed, because of its unique processing and properties, the nanofibrous nonwoven carbon described here may have particular advantages in simplicity, scalability and versatility. For the sake of providing context, this method and material are contrasted with other traditional and non-traditional carbon aerogels (CAs), and particular attention is given to other fibrous CAs. For instance, Lin et al. [1] recently published a comprehensive review of low-density CNT materials, including true aerogels and analogues such as “sponges” produced through floating catalyst CVD (FC-CVD). Mikhalchan et al. [16] have also recently published work on a FC-CVD method to create bulk,

ACCEPTED MANUSCRIPT

nanofibrous carbon aerogels, and the interested reader is directed to those efforts for a more thorough description of processing, properties and applications. Although the work described here utilizes carbon nanofibers, not carbon nanotubes, the governing principles of synthesis and properties will be largely similar. Specifically, the qualities, interactions and alignment of the constituent fibers will control the bulk properties. 4.1

Processing Methodology

The general sol-gel process for creating aerogel materials requires three steps: 1) creation of a solution containing the aerogel structural material (traditionally silica), 2) gelation of the solution and aging to strengthen the matrix around liquid-filled pores, and 3) exchange of the liquid in pores with air to create an aerogel [36]. The exchange process must be done carefully to avoid the collapse of the structure, and this is often achieved through supercritical drying or freeze drying, although recent developments have allowed for more robust processing, including ambient pressure [33]. The sol-gel process has been successfully demonstrated in a variety of studies, and is often among the best choices for creating exceptionally low density materials. Modifications and alternative methods are becoming more prevalent. These include CVD [1, 16, 17], carbide-derived carbons [11, 12] and cellulose derived carbons [13-15]. With FC-CVD methods, the carbon is synthesized at the exhaust end of a tube furnace through which gaseous precursors (e.g., ferrocene, thiopehene, xylene, dichlorobenzene, methane, etc.) are passed at high temperature (e.g., 1200 °C). The carbon collects on a substrate (e.g., sapphire) where it assembles into a bulk, scaffold-like structure. The size of deposited carbon monolith can be centimeters in size, the orientation and properties of the fibers can be modified through processing conditions, and the ability to scale the process or operate continuously is often suggested as an asset of the method. A cost-effective advancement in aerogel preparation is the use of natural precursors such as cellulose or cotton. These methods convert preexisting structures into carbon through pyrolysis and this eliminates the need for a synthesis step. Even so, the process still requires high temperature to complete the pyrolysis and careful control of the process to avoid damaging the

ACCEPTED MANUSCRIPT

structure. For instance, Bi et al. [37] describe the drying of natural cotton at 60 °C for 12 h, pyrolyzing under partial argon pressure at 800 °C for 2 h with a heating rate of 5 °C/min and cooling in the furnace to room temperature. Even without considering the cooling time, the process requires approximately 16.5 h to complete. Of course this is largely balanced by the miniscule cost of the precursor materials and the ability to process a collection of samples together. Another example of natural fibrous material is bacterial cellulose. Wu et al. [14] pyrolized preexisting cellulose nanofibers at temperatures ranging from 700-1300 °C, with a notable increase in hydrophobicity with pyrolysis temperature. In this process the cellulose precursor was first converted to an aerogel using freeze-drying, thereby increasing the complexity of process somewhat. In addition to the three day soak time in deionized water and freeze-drying, the process of pyrolysis utilizes a heating cycle requiring more than 12 h, not including the final cooling step. The process used here is based on the constrained formation of nanofibrous structures (CoFFiN) process [25], and it produces similarly structured, fibrous CAs. The low density of this material can be ascribed to the tortuous growth behavior and rigidity of the constituent fibers, which prevents long-range fiber alignment and results in a highly open but self-supporting bulk structure. The process combines a long history of study in catalytic CNF synthesis where the kinetics and properties of fiber deposition are controlled by a combination of catalyst form, composition, deposition temperature and gas mixture [38, 39]. This basic framework is then applied within a constrained environment to increase the degree of entanglement. Not only does constraint produce mechanical integrity, the constraint geometry provides a direct means to produce components of desired shape and size in a single step. This eliminates the need to further process the CA for application, as the material can be produced with exact dimensions for the intended use. In prior work, rectangular and tubular molds of varying dimensions have also been used [25, 26], and this process allows for application-specific development. In addition to the ability to create varied sizes and shapes, the relative simplicity of the process also makes it efficient enough to produce many components of repeatable shape and quality.

ACCEPTED MANUSCRIPT

This is demonstrated throughout the results described above, where more than 80 samples were produced and tested as part of this study. This is an order of magnitude greater than most carbon aerogel studies. That is only practical with a simple process which produces samples in finished form every time. This concept is demonstrated in Fig. 15A, where only 23 samples are shown (about a quarter of the total produced and tested here). The ability to produce very low density monoliths is shown in Fig. 15B, but higher density materials are a particular competency as they more accurately replicate the mold dimensions, including sharp corners (e.g., those in Fig. 15A). To our knowledge, no other process creates material in a “finished” form, though the ability to replicate container geometry [7] and machinable material [33] have been reported.

Fig. 15 - More than 80 samples were produced in this study, and this was made possible by the ease of processing and repeatability demonstrated across (A) 23 samples shown here. (B) The nanofibrous CA produced in this work can range from 10s to 100s of mg/cc. The sample shown has a bulk density of 20 mg/cc. The closest method to that investigated here mentions a CNF monolith being created from a Ni 25% Cu catalyst, but it is only described in abstract form [40]. Although the methods used to prepare the material are not described, surface area (117 m2/g) and elasticity (1.7 kPa) are listed for a density of 280 mg/cc, each of which is lower than those determined here. Alternatively, fibrous mats have been produced from commercially available CNFs by, “dispersing the carbon nanofiber in a proprietary solvent and filtering the solution,” [19] such that the fibers were held together by mechanical interlock and Van der Waals forces. It is not evident that threedimensional control of geometry and density are possible or practical with these methods.

ACCEPTED MANUSCRIPT

4.2

Structural and Morphological Properties

Although the density of aerogels is a benchmark achievement in many studies, we suggest that this is of secondary importance to the resulting properties and alignment with desired applications. For many applications, the strength, surface area, pore size and accessibility, electrical and thermal conductivity, or other properties are the most important aspects. Low density is only valuable when it enhances these. Indeed, as the density reaches ~200 mg/cc and below, further reductions cannot exceed a 1% difference, and another order of magnitude reduction (i.e., 20 mg/cc) leaves only 0.1% change in density remaining. Although the production of ultralight carbons is an interesting an important endeavor, many commercial applications are not significantly benefitted, especially where processing must become more intricate. Here, we have produced robust structures from ~20-700 mg/cc, which is generally in the mid to upper range of other methods. This density range is quite large when compared to most processes, and the ability to control it through basic process variables is a distinct advantage. The mold geometry and catalyst placement are decisive in the outcome. Because the mold used here is relatively deep, a density gradient develops where the carbon near the catalyst origin is of a higher density. This density gradient may be inconsequential or even desirable in certain applications, but by reducing the mold aspect ratio or relocating the catalyst, the growth behavior may be controlled with greater precision. For instance, the catalyst powder may be distributed on the mold walls, placed on a sacrificial template or otherwise introduced at varying times and locations to allow for unique possibilities. Concepts, challenges and strategies for scaling this process and applying it more broadly are the topic of a recent article [26]. The nanofibrous CA described here is structurally distinct from traditional aerogels, but comparable to fibrous carbon materials described above. In particular, CNT aerogels tend to have smaller diameter fibers and those fibers are intermittently linked together through junctions which create a three-dimensional network that adds to the stability of the structure. Hashim et al. [41] describe the importance of elbow junctions in CNTs as preferential sites for doping by nitrogen, boron or sulfur to create these networks. The generation of these network nodes is

ACCEPTED MANUSCRIPT

dependent on processing conditions, and Shan et al. [42] report on the synergistic effects of sulfur and nitrogen on the formation of elbow and welded junctions. Although our material has a different structure in the as-produced state, CNFs may also be covalently linked [1, 42]by doping [43] or through the creation of composite carbon structures [44]. The characteristics of the deposited fibers are dependent on catalyst composition and form. For instance, in previous work on Pd powder, foil and sputtered films [45, 46], sputtered films and sub-micron and nanoscale powders readily produced fibers, but foils resulted in microns thick carbon films being deposited. Additionally, concerns with sintering and strategies for controlling fiber diameter with bulk powder, as used here, have been described [47]. In this work powder was distributed very finely, which reduces the likelihood of sintering, and the fiber characteristics were found to be similar throughout the monoliths. This is likely because the fibers to quickly fill the mold and later build density. That results in more uniform fiber characteristics as “early” growth fibers are distributed throughout. Also, limited branching was observed, which indicates the catalyst particles do not rapidly disintegrate into smaller units which ted to produce fibers of varying diameter. The individual nanofibers which comprise the CAs produced here are highly amorphous and exhibit a combination of straight and twisted morphologies. Because the CoFFiN process constrains catalytically deposited CNFs, the catalyst and deposition environment control the properties of the bulk material. Fibers of varying morphology, crystallinity, size, etc. can be created using the same basic CVD process by modifying the catalyst composition, gas chemistry, temperature and other parameters [30, 38], so this method may achieve various mechanical, electrical, thermal and other properties. Additionally, the material may be graphitized after synthesis to increase the crystallinity [19, 48]. Post-production treatment may also include activation to increase the surface area and/or modify the pore size. The adsorption isotherms, with relatively small hysteresis loops, indicate a nonporous or microporous surface structure which allows for rapid desorption of nitrogen. The baseline surface area of ~215 m2/g is mid-range for CAs in general, and may be increased by activation using CO2 or KOH, as is common in these materials [49-51]. In addition to increasing

ACCEPTED MANUSCRIPT

the overall surface area by activation, the pore size may be tuned as well [11]. Because the catalyst was not removed from the monoliths, it may affect the final properties, especially functional properties. The remaining catalyst in this work is typically less than 0.01 vol% of the final structure and can be removed by acid dissolution if desired. Conversely, composite materials could be made by adding additional catalyst to create a carbon structure “decorated” with catalyst particles, thereby creating and using the CNFs as a catalyst support in a single-step process. 4.3

Mechanical Properties

The mechanical properties of the monoliths are highly density dependent, with lower density materials (< 100 mg/cc) being more elastic and substantially less stiff. This is evidenced in the exponential fit of the measured stiffness (see Fig. 10C), but it is also apparent in the individual stress-strain compression curves, where the stress increases rapidly for low density samples at high strain as a consequence of the densifying of the material. Nonwoven materials can be described through the van Wyk model [52] which has been applied to other nonwoven CNFs with good agreement [25]. This model considers the bending of individual fibers and is dependent on fiber properties such as diameter length, volume fraction, etc. Cyclic strain creates a shift in total strain as the material compacts more with each cycle, but the change decreases for low density samples and is negligible for higher density samples at the same stress levels. Gui et al. [17] report a similar trend in strain (approximately logarithmic) during cyclic testing of a 5-10 mg/cc CNT “sponge,” though at larger strain and number of cycles. The lowest density material produced here is not as elastic as some CAs which can be reversibly cycled to as much as 80% strain (e.g., [15, 17, 34, 53]). The lack of cross-linking between fibers is considered the primary cause, as any reorientation of the fibers at high strain is not reversible. Despite their nonwoven structure, the nanofibrous CAs produced here are robust, and they can be handled and processed without particular care. This makes them simple to extract from the mold and to integrate into other structures as components for applications such as filtration [54].

ACCEPTED MANUSCRIPT

4.4

Thermal and Electrical Properties

The monoliths produced here have very low electrical conductivity. The electrical conductivity measured on CNFs produced from the same catalyst composition in other, non-aerogel work is reported as 32.3 S/m [55]. Given the density range of samples measured for conductivity (1.907.14% solid volume), the expected value would be 0.61-2.31 S/m. The actual values (1.322.83 S/m are fairly consistent with that assessment. The conductivity increases with compressive strain, and the electrical strain response is consistent during cycling, with the magnitude of the response being greatest at higher densities. Sensing applications have been explored in a variety of nanofibrous and aerogel materials due to their strain-dependent resistance (e.g., [2, 56, 57]). Since density and geometry can be readily controlled using the methods in this study, specific properties may be engineered on an applications basis. The thermal conductivity of these samples created here is very low (~0.03 W/mK), especially given the higher density of the samples tested (487-700 mg/cc). Many other CAs have higher thermal conductivity for lower density, and this is often cited to be the result of the carbon form (e.g., graphene or CNTs), which contributes to the solid conductivity, and the radiative conductivity which is inversely proportional to density [35]. If desired, the thermal conductivity of these nanofibrous CAs may be increased by graphitization to increase crystallinity [19].

5

Conclusions

Nanofibrous carbon aerogels have been synthesized by constraining the growth of catalytically deposited carbon nanofibers. The material can be precisely controlled to achieve a specified density and external geometry using a one-step direct synthesis method known at the constrained formation of fibrous nanostructures (CoFFiN) process. The deposition kinetics and properties of these materials have been studied by producing and testing more than 80 samples, which is only practical given the simplicity and repeatability of the process. By modifying basic process parameters, the material can be created over a large range of density (20-700 mg/cc, possibly higher). The nanofibers, which are a mixture of straight and twisted morphologies with varying diameters (~150-500 nm), possess high surface area (~200 m2/g) characterized by micropores.

ACCEPTED MANUSCRIPT

The materials are mechanically robust, with density-dependent elasticity values of 60-600 kPa. The materials can be reversibly cycled to modest strain, which is dependent on the initial density, and the electrical resistance scales proportionally. These materials exhibit low electrical conductivity (1.32-2.83 S/m) and low thermal conductivity (~0.03 W/mK), even at the highest densities, and these properties are attributed to the highly amorphous structure of the nanofibers and the lack of linkage between individual fibers. Acknowledgements We would like to thank the organizations which assisted through characterization facilities and analysis: The University of Delaware Keck Center for Advanced Microscopy and Microanalysis (nanofiber SEM and TEM), Pennsylvania State University Materials Characterization Laboratory (BET and BJH surface analysis), TA Instruments (Wakefield, MA, thermal conductivity) and Hitachi High Technologies (Clarksburg, MD, ion milling and imaging of milled CNFs). This work was supported by the National Science Foundation (Award # 1436444). References [1] Lin Z, Zeng Z, Gui X, Tang Z, Zou M, Cao A. Carbon Nanotube Sponges, Aerogels, and Hierarchical Composites: Synthesis, Properties, and Energy Applications. Advanced Energy Materials. 2016;6(17). [2] Zou J, Liu J, Karakoti AS, Kumar A, Joung D, Li Q, et al. Ultralight Multiwalled Carbon Nanotube Aerogel. ACS Nano. 2010;4(12):7293-302. [3] Gui X, Zeng Z, Cao A, Lin Z, Zeng H, Xiang R, et al. Elastic shape recovery of carbon nanotube sponges in liquid oil. J Mat Chem. 2012;22(35):18300-5. [4] Wu Z-S, Yang S, Sun Y, Parvez K, Feng X, Müllen K. 3D Nitrogen-Doped Graphene Aerogel-Supported Fe3O4 Nanoparticles as Efficient Electrocatalysts for the Oxygen Reduction Reaction. J Am Chem Soc. 2012;134(22):9082-5. [5] Zhang X, Sui Z, Xu B, Yue S, Luo Y, Zhan W, et al. Mechanically strong and highly conductive graphene aerogel and its use as electrodes for electrochemical power sources. J Mat Chem. 2011;21(18):6494-7. [6] Worsley MA, Pauzauskie PJ, Olson TY, Biener J, Satcher JH, Baumann TF. Synthesis of Graphene Aerogel with High Electrical Conductivity. J Am Chem Soc. 2010;132(40):14067-9. [7] Chen W, Yan L. In situ self-assembly of mild chemical reduction graphene for threedimensional architectures. Nanoscale. 2011;3(8):3132-7. [8] Lim MB, Hu M, Manandhar S, Sakshaug A, Strong A, Riley L, et al. Ultrafast sol–gel synthesis of graphene aerogel materials. Carbon. 2015;95:616-24.

ACCEPTED MANUSCRIPT

[9] Horikawa T, Hayashi Ji, Muroyama K. Controllability of pore characteristics of resorcinol–formaldehyde carbon aerogel. Carbon. 2004;42(8–9):1625-33. [10] Soleimani Dorcheh A, Abbasi MH. Silica aerogel; synthesis, properties and characterization. J Mater Process Tech. 2008;199(1–3):10-26. [11] Oschatz M, Boukhalfa S, Nickel W, Hofmann JP, Fischer C, Yushin G, et al. Carbidederived carbon aerogels with tunable pore structure as versatile electrode material in high power supercapacitors. Carbon. 2017;113:283-91. [12] Oschatz M, Borchardt L, Thommes M, Cychosz KA, Senkovska I, Klein N, et al. Carbide‐Derived Carbon Monoliths with Hierarchical Pore Architectures. Angewandte Chemie International Edition. 2012;51(30):7577-80. [13] Zu G, Shen J, Zou L, Wang F, Wang X, Zhang Y, et al. Nanocellulose-derived highly porous carbon aerogels for supercapacitors. Carbon. 2016;99:203-11. [14] Wu Z-Y, Li C, Liang H-W, Chen J-F, Yu S-H. Ultralight, Flexible, and Fire-Resistant Carbon Nanofiber Aerogels from Bacterial Cellulose. Angewandte Chemie International Edition. 2013;52(10):2925-9. [15] Wan Y-J, Zhu P-L, Yu S-H, Sun R, Wong C-P, Liao W-H. Ultralight, super-elastic and volume-preserving cellulose fiber/graphene aerogel for high-performance electromagnetic interference shielding. Carbon. 2017;115:629-39. [16] Mikhalchan A, Fan Z, Tran TQ, Liu P, Tan VBC, Tay T-E, et al. Continuous and scalable fabrication and multifunctional properties of carbon nanotube aerogels from the floating catalyst method. Carbon. 2016;102:409-18. [17] Gui X, Wei J, Wang K, Cao A, Zhu H, Jia Y, et al. Carbon Nanotube Sponges. Adv Mater. 2010;22(5):617-21. [18] Weng B, Xu F, Salinas A, Lozano K. Mass production of carbon nanotube reinforced poly(methyl methacrylate) nonwoven nanofiber mats. Carbon. 2014;75:217-26. [19] Mahanta NK, Abramson AR, Lake ML, Burton DJ, Chang JC, Mayer HK, et al. Thermal conductivity of carbon nanofiber mats. Carbon. 2010;48(15):4457-65. [20] Wang X, Luo R, Ni Y, Zhang R, Wang S. Properties of chopped carbon fiber reinforced carbon foam composites. Materials Letters. 2009;63(1):25-7. [21] Wenmakers PWAM, van der Schaaf J, Kuster BFM, Schouten JC. "Hairy Foam": carbon nanofibers grown on solid carbon foam. A fully accessible, high surface area, graphitic catalyst support. J Mat Chem. 2008;18(21):2426-36. [22] Huang Y, Lai F, Zhang L, Lu H, Miao Y-E, Liu T. Elastic Carbon Aerogels Reconstructed from Electrospun Nanofibers and Graphene as Three-Dimensional Networked Matrix for Efficient Energy Storage/Conversion. Scientific Reports. 2016;6. [23] Mecklenburg M, Schuchardt A, Mishra YK, Kaps S, Adelung R, Lotnyk A, et al. Aerographite: Ultra Lightweight, Flexible Nanowall, Carbon Microtube Material with Outstanding Mechanical Performance. Adv Mater. 2012;24(26):3486-90. [24] Guevara L, Wanner C, Welsh R, Atwater M. Using Mechanical Alloying to Create Bimetallic Catalysts for Vapor-Phase Carbon Nanofiber Synthesis. Fibers. 2015;3(4):394-410 [25] Atwater MA, Mousavi AK, Phillips J, Leseman ZC. Direct Synthesis of Nanoscale Carbon Nonwovens by Catalytic Deposition. Carbon. 2013;57:363-70. [26] Atwater MA, Welsh RJ, Edwards DS. Direct Synthesis of Nanofibrous Nonwoven Carbon Components: Initial Observations, Capabilities, and Challenges. J Micro Nano Manuf. 2016;4(4):041004-1--8.

ACCEPTED MANUSCRIPT

[27] Jablonski GA, Geurts FW, A. Sacco J. Carbon Deposition Over Fe, Ni, and Co Foils From CO-H,-CH4-C02-H20, CO-C02, CH4-H2, and CO-H2-H20 Gas Mixtures: II: Kinetics. Carbon. 1992;30(1):99-106. [28] Guinot J, Audier M, Coulon M, Bonnetain L. Formation and characterization of catalytic carbons obtained from CO disproportionation over an iron nickel catalyst—I. Fragmentation and rates of carbon deposition. Carbon. 1981;19(2):95-8. [29] Baker RTK, Harris PS, Terry S. Unique form of filamentous carbon Nature. 1975;253(5486):37-9. [30] Rodriguez NM, Chambers A, Baker RTK. Catalytic engineering of carbon nanostructures. Langmuir. 1995;11(10):3862-6. [31] Klein KL, Melechko AV, Rack PD, Fowlkes JD, Meyer HM, Simpson ML. Cu–Ni composition gradient for the catalytic synthesis of vertically aligned carbon nanofibers. Carbon. 2005;43:1857-63. [32] Lu X, Nilsson O, Fricke J, Pekala RW. Thermal and electrical conductivity of monolithic carbon aerogels. J Appl Phys. 1993;73(2):581-4. [33] Jia X, Dai B, Zhu Z, Wang J, Qiao W, Long D, et al. Strong and machinable carbon aerogel monoliths with low thermal conductivity prepared via ambient pressure drying. Carbon. 2016;108:551-60. [34] Lv P, Tan X-W, Yu K-H, Zheng R-L, Zheng J-J, Wei W. Super-elastic graphene/carbon nanotube aerogel: A novel thermal interface material with highly thermal transport properties. Carbon. 2016;99:222-8. [35] Feng J, Feng J, Zhang C. Thermal conductivity of low density carbon aerogels. Journal of Porous Materials. 2012;19(5):551-6. [36] Du A, Zhou B, Zhang Z, Shen J. A Special Material or a New State of Matter: A Review and Reconsideration of the Aerogel. Materials. 2013;6(3):941. [37] Bi H, Yin Z, Cao X, Xie X, Tan C, Huang X, et al. Carbon Fiber Aerogel Made from Raw Cotton: A Novel, Efficient and Recyclable Sorbent for Oils and Organic Solvents. Adv Mater. 2013;25(41):5916-21. [38] Rodriguez NM. A review of catlytically grown carbon nanofibers. J Mater Res. 1993;8(12):3233-50. [39] deJong KP, Geus JW. Carbon nanofiber: synthesis and applications. Catal Rev. 2000;42(4):481-510. [40] Ge X, Wu X-l, Wang J-t, Long D-h, Qiao W-m, Ling L-c. Synthesis of carbon nanofiber monoliths by chemical vapor deposition. Carbon. 2015;86:372. [41] Hashim DP, Narayanan NT, Romo-Herrera JM, Cullen DA, Hahm MG, Lezzi P, et al. Covalently bonded three-dimensional carbon nanotube solids via boron induced nanojunctions. Scientific reports. 2012;2:363. [42] Shan C, Zhao W, Lu XL, O’Brien DJ, Li Y, Cao Z, et al. Three-Dimensional NitrogenDoped Multiwall Carbon Nanotube Sponges with Tunable Properties. Nano Letters. 2013;13(11):5514-20. [43] Qie L, Chen WM, Wang ZH, Shao QG, Li X, Yuan LX, et al. Nitrogen‐Doped Porous Carbon Nanofiber Webs as Anodes for Lithium Ion Batteries with a Superhigh Capacity and Rate Capability. Adv Mater. 2012;24(15):2047-50. [44] Panomsuwan G, Saito N, Ishizaki T. Nitrogen-Doped Carbon Nanoparticle–Carbon Nanofiber Composite as an Efficient Metal-Free Cathode Catalyst for Oxygen Reduction Reaction. ACS Applied Materials & Interfaces. 2016;8(11):6962-71.

ACCEPTED MANUSCRIPT

[45] Atwater MA, Phillips J, Doorn SK, Luhrs CC, Fernandez Y, Menendez JA, et al. The production of carbon nanofibers and thin films on palladium catalysts from ethylene–oxygen mixtures. Carbon. 2009;47(9):2269–80. [46] Atwater MA, Phillips J, Leseman ZC. Formation of carbon nanofibers and thin films catalyzed by palladium in ethylene-hydrogen mixtures. J Phys Chem C. 2010;114(13):5804-10. [47] Atwater MA, Phillips J, Leseman ZC. The effect of powder sintering on the palladiumcatalyzed formation of carbon nanofibers from ethylene–oxygen mixtures. Carbon. 2010;48(7):1932-8. [48] Ramos A, Camean I, Garcıa AB. Graphitization thermal treatment of carbon nanofibers. Carbon. 2013;59:2-32. [49] Ma C, Song Y, Shi J, Zhang D, Zhai X, Zhong M, et al. Preparation and one-step activation of microporous carbon nanofibers for use as supercapacitor electrodes. Carbon. 2013;51:290-300. [50] Yoon S-H, Lim S, Song Y, Ota Y, Wenming Qiao, Tanaka A, et al. KOH activation of carbon nanofibers. Carbon. 2004;42:1723-9. [51] Wang G, Pan C, Wang L, Dong Q, Yu C, Zhao Z, et al. Activated carbon nanofiber webs made by electrospinning for capacitive deionization. Electrochim Acta. 2012;69:65-70. [52] Wyk CMv. 20—NOTE ON THE COMPRESSIBILITY OF WOOL. Journal of the Textile Institute Transactions. 1946;37(12):T285-T92. [53] Guan L-Z, Gao J-F, Pei Y-B, Zhao L, Gong L-X, Wan Y-J, et al. Silane bonded graphene aerogels with tunable functionality and reversible compressibility. Carbon. 2016;107:573-82. [54] Li H, Gui X, Zhang L, Wang S, Ji C, Wei J, et al. Carbon nanotube sponge filters for trapping nanoparticles and dye molecules from water. Chemical Communications. 2010;46(42):7966-8. [55] Dresselhaus MS, Dresselhaus G, Sugihara K, Spain IL, Goldberg HA. Graphite fibers and filaments: Springer Science & Business Media; 2013. [56] Luhrs CC, Daskam CD, Gonzalez E, Phillips J. Fabrication of a Low Density Carbon Fiber Foam and Its Characterization as a Strain Gauge. Materials. 2014;7(5):3699-714. [57] Wang M, Anoshkin IV, Nasibulin AG, Korhonen JT, Seitsonen J, Pere J, et al. Modifying Native Nanocellulose Aerogels with Carbon Nanotubes for Mechanoresponsive Conductivity and Pressure Sensing. Adv Mater. 2013;25(17):2428-32.