*2.4. Tailored Electrochemical Growth Conditions Producing Thinner CNTs from CO2*

Figure 6 demonstrates additional electrochemical conditions that yielded high-purity carbon nanotubes by CO2 molten electrolysis. In the top row, panels H, as with high current density (Figure 2 panels A), a moderate current density of 0.4 A/cm2 (with the same electrolyte, a Muntz Brass cathode and a Nichrome C anode) yielded high-purity (96%) CNTs that were somewhat longer (100–200 μm) at a coulombic efficiency approaching 100%. Switching the cathode material to Monel, as shown in the second row (Figure 6, panels a), yielded shorter 20–50 μm CNTs with 97% purity and coulombic efficiency again approaching 100%. Not shown in the figure, but included in Table 1 (Electrolysis D), is the fact that a switch from Nichrome C to a pure nickel anode (while retaining the Monel Cathode, and with electrolyte additives at J = 0.2 A/cm2) led to a substantial drop in CNT purity to 70% with the remainder of the product consisting of nano-onions. A drop of current density from 0.4 A to 0.1 A/cm2, as shown in in Figure 6 panels K, yielded 97% purity CNTs of 30–60 μm length with only a small drop of coulombic efficiency to 97%. In a single panel of L located in the lower left corner of Figure 6, it can be seen that an overabundance of Fe2O3 was added, which was previously observed to lead to a loss of control of the synthesis specificity [45]. In this case, the total purity of the CNTs remained high at ~95%, but this consisted of two distinct morphologies of CNT in the product. The majority product, at ~75%, was twisted CNTs, and the minority product, at ~20%, was straight CNTs. Finally, in panels M on the middle and right lowest row of Figure 6, a noble metal, iridium, was used as the anode (along with the Monel Cathode) at a low 0.08 A/cm<sup>2</sup> current density. Transition metals released from the anode, during its formation of a stable oxide over layer, can contribute to the transition metals ions that are reduced at the cathode and serve as nucleation points for the CNTs. This was not the case here due to the high stability of the iridium. Instead, as a single, high concentration transition metal, 0.81 wt% Cr was made as the electrolyte additive. The product was highly pure (97%) CNTs that were the thinnest shown (<50 nm diameter), were 50–100 μm long for an aspect ratio > 1000, and formed at a coulombic efficiency of 80%.

**Figure 4.** SEM of the synthesis product of high-aspect-ratio (and high-purity and high-yield) carbon nanotubes prepared by electrolysis F in Table 1, splitting CO2 in 770 ◦C Li2CO3. Moving left to right in the panels, the product is analyzed by SEM with increasing magnification. Scale bars in panels Fa-Ff (clockwise from top) are 500, 400, 100, 5, 5, and 10 μm.

**Figure 5.** TEM and HAADF of the synthesis product of high-purity, high-yield carbon nanotubes under the Electrolysis F (Table 1) electrochemical conditions by electrolytic splitting of CO2 in 770 ◦C Li2CO3. In the top row, the product is analyzed by TEM with scale bars of 1 μm (left panel) or 100 nm (right). Scale bars in the middle right moving left to right are of 50, 20, and 1 nm. HAADF measurements in the bottom panel each have scale bars of 200 nm. Panels: (**A**), A-1, A-2, (**B**), B-1, B-2, B-3, B-4: TEM; B-2-1 TEM with measured graphene layer thickness. (**C**): Elemental HAADF elemental analysis with (right side) elemental profile.

**Figure 6.** SEM of the synthesis product of high-purity, high-yield carbon nanotubes under a variety of electrochemical conditions by electrolytic splitting of CO2 in 770 ◦C Li2CO3. The washed product was collected from the cathode subsequent to the electrolysis described in Table 1. Scale bars (starting from left) are for panels H: 100, 4, and 2 μm; for panels I: 100, 10, and 4 μm; for panels K: 40, 5, and 3 μm; for panels L: 40, 50, and 5 μm.

SEM analysis of several of the synthesis products, specifically Electrolyses H, B, and C, revealed evidence of nodules that appeared as "buds" attached to the CNTs. These buds were the most consistent in Electrolysis H and were explored by TEM and HAADF, as shown in Figure 7. It is fascinating, as seen in the top row of the figure, that the buds generally had a spherical symmetry, and while not prevalent in the structure, appeared in a structure comparable to grape bunches growing on a vine. The buds generally contained a low level of the transition metal nucleating metal, such as the 0.3% Fe shown, and the rest of the structure was generally pure carbon, with an occasional metal core. This low level of metal used was easily removed by an acid wash. Previously introduced higher levels of Ni or Fe could lead to magnetic carbon nanotubes with useful properties of recyclability, filtration, and shape-shifting materials among other applications [41]. As seen in the left side of the second row, the carbon nanotube walls continued to exhibit the regular 0.33 to

0.34 graphene interwall separation, as seen on the right side of the row, and the adjacent CNTs may have had merged or distinct graphene structures. Similarly, as seen in the third row of Figure 2, the adjacent buds on the CNTs could have graphene walls that bended to join, and were shared, or, as seen in the fourth row, appeared instead to be distinct (intertwined, not merged) structures.

**Figure 7.** TEM and HAADF of the synthesis product of carbon nanotubes that exhibited nodules or buds under the Electrolysis H (Table 1 and SEM on top row of Figure 6) electrochemical conditions by electrolytic splitting of CO2 in 770 ◦C Li2CO3. In the top row, the product is analyzed with scale bars from left to right of 200, 100, 20 and 100 nm. Scale bars in the second right row are of 1μm, then 20, 5, 200, and 1 nm. The third row scale bars are 200, 20, 1, 5, and 200 nm. The bottom row scale bars are 50, 1, 1, and 200 nm. TEM Panels: (**A**) A-1, A-2-1, (**B**), B-1, B-1-1, (**C**), C-1, C-2, D-1-1, D-1-2, (**E**), E-1, E2: TEM; A-2, (**D**,**E**) also include Elemental HAADF elemental analyses.

#### *2.5. Electrochemical Conditions to Synthesize Macroscopic Assemblies of CNTs*

In addition to individual CNTs, the final series of electrolyses generated useful macroscopic assemblies of CNTs. There has been interest in densely packed CNTs for nanofiltration, and also, due to their high density of conductive wires, as an artificial neural net [52–62]. CNTs' aerogels have been reported as being formed by CVD and/or also reported as being formed within molds. Their sorbent properties have been investigated for applications such as the cleanup of chemical leakages under harsh conditions. Those studies noted that such aerogel matrices, consisting of highly porous, intermingled CNTs, can be repeatedly compressed to a small fraction of their initial volume without damaging the structure of the carbon nanomaterials [52–56].

The term "Nanofiltration" was proposed in 1984 to solve the terminology problem for a selective reverse osmosis process that allows ionic solutes in a feed water to permeate through a membrane [57]. In addition to low energy consumption, with respect to alternative unit operations such as distillation and evaporation, thermal damage of heat-sensitive molecules can be minimized during the separation due to the potential for low operating temperatures of nanofiltration [58]. Small-diameter CNTs were demonstrated due to their ability to be used as a nanofiltration or molecular sieve to selectively remove larger size molecules from smaller size molecules, such as the selective removal of cyclohexane from n-hexane [57]. There is a need to control the fabrication of macroscopic assemblies of CNTs to optimize their capabilities for nanofiltration, and CNT assemblies synthesized by CVD and laser ablation have been investigated [52–56,58–61].

An artificial neural network is a collection of interconnected nodes that loosely model the neurons in a biological brain. Estimates of biologic neuronal density (rat) are in the range of 100 in a 100 μm cube (105/mm3) on each side. The fabrication of an artificial neural network with a structure that mimics that number of nanowires and nodes presents a challenge. However, this is in the same size domain as macroscopic assemblies of CNTs. For example, Gabay and co-workers explored the engineered self-organization of neural networks using carbon nanotube clusters [38], emphasizing the need for improved pathways to fabricate and control macroscopic assemblies of CNTs.

The macroscopic assemblies observed in this study are referred to as nano-sponge, densely packed parallel CNTs, and nano-web CNTs, in Table 3 and Figure 8. The Nanosponge assembly was formed by Electrolyses N with Nichrome C serving as both the cathode and the anode, with 0.81% Ni powder added to the 770 ◦C Li2CO3 electrolyte, the initial current ramped upwards (5 min each at 0.008, 0.016, 0.033 and 0.067 A/cm2), then a 4h current density of 0.2 A/cm<sup>2</sup> generating a 97% purity nanosponge at 99% coulombic efficiency. As previously described, long, densely packed, parallel carbon nanotubes were produced in Electrolysis F with a 0.1 wt% Fe2O3 additive to the Li2CO3 electrolyte, a Muntz Brass cathode and an Inconel 718 anode, and two layers of Inconel 600 screen at 0.15 A/cm2.



**Figure 8.** SEM of the synthesis product consisting of carbon nanotubes arranged in various packed macroscopic structures that are amenable to nano-filtration. The washed product was collected from the cathode subsequent to the electrolysis described in Table 3. Moving from left to right in the panels, the product is analyzed by SEM with increasing magnification. Scale bars in panels These include nano-sponge, dense packed straight, and nano-web CNTs. Moving from left to right in the panels, the product is analyzed by SEM in the top two rows (N), and subsequent rows P and Q, with increasing magnification. Scale bars in panels (starting from left) are for top panels N: 500, 40, 20, and 8 μm; for lower panels N: 400, 10, and 5 μm; for panels P: 300, 40, and 5 μm; for panels Q: 500, 40, 20, and 8 μm.

As opposed to the parallel assembly produced in Electrolysis F, nano-web aptly describes the interwoven carbon nanotubes from Electrolyses P and Q, presented in the lower rows of Table 3 and Figure 8. Two different routes to the nano-web assembly are summarized. The first uses an 0.1% Fe2O3 additive, a Muntz Brass cathode, and an Inconel 718 anode with three layers of Inconel 600 screen, at 0.08 A/cm2, generating a nano-web with a purity of 97% at a coulombic efficiency of 79%. The second pathway uses an 0.81% Ni powder additive, a Monel cathode, and Nichrome C anode, at 0.28 A/cm2, generating a nano-web with a purity of 92% at a coulombic efficiency of 93%.

The densely packed straight CNTs had an inter-CNT spacing ranging from 50 to 300 nm; moreover, the CNTs were highly aligned, providing unusual nanofiltration opportunities for both this size domain, and for an opportunity to filter 1D from 3D morphologies. The nano sponge did not have this alignment feature, and as shown in Figure 8, provided nanofiltration pore sizes of 100 to 500 nm, while the nano-web product provided nanofiltration with pore sizes of 200 nm to 1 μm. Future studies will investigate the effectiveness of this portfolio of macroscopic CNT assemblies for nanofiltration.

#### *2.6. Raman and XRD Characterization of the CNTs and Their Macro-Assemblies*

Figure 9 presents the Raman spectra effect of variation of the CNT electrolysis conditions on the CNT assembly products from CO2 electrolysis in 770 ◦C Li2CO3. The Raman spectrum exhibits two sharp peaks ~1350 and ~1580 cm<sup>−</sup>1, which correspond to the disorder-induced mode (D band) and the high-frequency E2G first order mode (G band), respectively, and an additional peak, the 2D band, at 2700 cm<sup>−</sup>1. In the spectra, the graphitic fingerprints lie in the 1880–2300 cm-1 and are related to different collective vibrations of sp-hybridized C-C bonds.

**Table 4.** Raman spectra of a diverse range of carbon allotropes and macro-assemblies formed by molten electrolysis.


**Figure 9.** Raman spectra of the synthesis product consisting of various labeled carbon nanotube assemblies synthesized by the electrolytic splitting of CO2 in 770 ◦C Li2CO3 with a variety of systematically varied electrochemical conditions described in Table 4.

Interpretation of the Raman spectra provides insights into the potential applications of the various carbon allotropes. As shown in Figure 9, the intensity ratio between the D band and the G band (ID/IG) was calculated; this ratio, and the observed shift in the IG frequency, are useful parameters to evaluate the relative number of defects and degree of graphitization, and are presented in Table 4. Note in particular that, of the nano-sponge, nano-web, and densely packed CNT assemblies described in Figure 8 and Table 3, the nano-web CNT assembly exhibited low disorder, with ID/IG = 0.36, as shown in Table 4, the densely packed CNT assembly exhibited intermediate disorder with ID/IG = 0.49, and

the nano-sponge exhibited the highest disorder with ID/IG = 0.62 while accompanied by a shift in IG frequency.

That is, for the assemblies with increasing ID/IG ratio: CNT nano-web < Dense packed CNT < CNT nano-sponge.

Previously, increased concentrations of iron oxide added to the Li2CO3 electrolyte were correlated with an increasing degree of disorder in the graphitic structure [38]. It should be noted that these defect levels each remain relatively low as the literature is replete with reports of multiwalled carbon nanotubes made by other synthetic processes with ID/IG > 1. Lower defects are associated with applications that require high electrical power and strength, while high defects are associated with applications that permit high diffusivity through the structure, such as those associated with increased intercalation and higher anodic capacity in Li-ion batteries and higher charge supercapacitors.

Along with the XRD library of relevant compound spectra, XRD results of the CNT assembly products are presented in Figure 10, prepared as described in Figure 8 and Table 3. Each of the spectra exhibited strong diffraction peaks at 2*θ* = 27◦, characteristic of graphitic structures. The nano-sponge XRD spectra was distinct from the others, with a dominant peak at 2*θ* = 43◦, indicating the presence of iron as Li2Ni8O10 and chromium as LiCrO2 by XRD spectra match. The XRD result of this nano-sponge exhibited little or no iron carbide. On the other hand, both the nano-web and densely packed straight CNTs exhibited additional significant peaks at 2*θ* = 42 and 44◦, indicative of the presence of iron carbide, Fe3C. The diminished presence of defects previously noted by the Raman spectra for the other densely packed CNTs, along with the XRD presence of Li2Ni8O10, LiCrO2, and Fe3C, provide evidence that the co-presence of Ni, Cr, and Fe as nucleating agents can diminish defects in the CNT structure compared to Ni and Cr alone.

**Figure 10.** XRD of the synthesis product consisting of various labeled carbon nanotube assemblies synthesized by the electrolytic splitting of CO2 in 770 ◦C Li2CO3 with a variety of systematically varied electrochemical conditions described in Table 4.

#### **3. Materials and Methods**

## *3.1. Materials*

Lithium carbonate (Li2CO3, 99.5%), lithium oxide (Li2O, 99.5%), lithium phosphate Li3PO4 (Li3PO4, 99.5%), iron oxide (Fe2O3, 99.9%, Alfa Aesar), and boric acid (H3BO3, Alfa Aesar 99+%) were used as electrolyte components in this study. For electrodes, Nichrome A (0.04-inch-thick), Nichrome C (0.04-inch-thick), Inconel 600 (0.25-in thick), Inconel 625 (0.25-in thick), Monel 400, Stainless Steel 304 (0.25-in thick), Muntz Brass (0.25-in thick), were purchased from onlinemetals.com (accessed on 14 December 2021). Ni powder was 3–7 μm (99.9%, Alfa Aesar). Cr powder was <10 μm (99.2%, Alfa Aesar). Co powder was 1.6 μm (99.8%, Alfa Aesar). Iron oxide was 99.9% Fe2O3 (Alfa Aesar). Co powder was 1.6 μm (99.8%, Alfa Aesar). Inconel 600 (100 mesh) was purchased from Cleveland Cloth. The electrolysis was a conducted in a high form crucible >99.6% alumina (Advalue).
