III. RESULTS AND DISCUSSION

Figure S3(a) (supplementary information) shows a drawing of the schematic view of a section of the quartz tube in which substrates 1, 2, and 3 were placed. The figure also illustrates a possible mechanism of how the feedstock flow on substrates 1, 2, and 3, labeled F1, F2, and F3, respectively. From very simple area calculations (total cross-sectional area of quartz tube minus the transversal area below the substrate), it is possible to estimate that the flow under the substrate is approximately 0.034 L/min, around 70 times smaller than the flow passing on top. From the cross-section shown in Fig. S3(b), it can be seen that the lower part of the substrate was exposed to relatively low amounts of the gaseous precursor, simply because there is less effective cross-sectional area. Figure S3(b) also shows the substrate and the reduced space located under the substrate, where r is the radius of the quartz tube (1.27 cm); L, the width of the substrate (1 cm); and the A ₁ and A ₂, the cross-sectional areas calculated from the radius and the size of the substrate. The geometry and size of the darker semicircular zone located at the edge of the substrate could be attributed to the effective cross-sectional area formed between the substrate and the arc of the quartz tube. However, it is quite possible that if substrates with very similar geometries were used, highly uniform growth zones could be obtained. Therefore, we could state that small changes in the overall pressure of the experiment could significantly affect the nanotube growth beneath the substrate. Similarly, it has been reported that a limited gas supply on catalytic substrates leads to dendritic structures. ^{Reference Liu and Kong38}

Figure 1 depicts the SEM images of vertically aligned CN _x MWNTs with pine-tree-like morphologies grown on Si/SiO₂ substrates. Figure 1(a) shows a 45° tilted view of a large CN _x MWNTs array. Figure 1(b) shows a higher magnification of Fig. 1(a) in which each “pine tree” is formed by the vertically aligned CN _x MWNTs. Individual nanotubes are clearly noticeable in Fig. 1(c), in which a “pine tree” of 40 µm height and 20 µm width has been transversally cut. Each pine-tree-like feature is formed by a bundle of CN _x MWNTs arranged in a conical cluster. A high magnification SEM image reveals CN _x MWNTs exhibiting diameters around 34 nm. All the average diameters were calculated by carrying out 100 individual measurements obtained from different SEM images [Fig. 1(d)]. For substrate 1, it was difficult to observe the pine-tree-like structure due to an abundant growth of nanotubes in addition to the formation of an amorphous carbon layer on the surface of the array (see Fig. S4 in the supplementary information). The morphology of the CN _x MWNTs in substrate 3 was very similar to that observed on substrate 2, but the formation of “pine trees” was favored on substrate 2 (see Fig. S5 in the supplementary information).

FIG. 1.

(a–c) SEM images of pine-tree-like morphologies formed by CN _x MWNTs tubes grown on the Si/SiO₂ substrates. These pines were found upside down on substrate 2, named CN _x -P2. The “pine tree” width average is 29.42 ± 5.86 μm and the average length is 39.47 ± 5.73 μm (the averages were obtained after measuring numerous structures from multiple SEM images). The cross-sectional cut of the sample shown in (c) reveals the tube composition inside the “pine trees,” and (d) shows a higher magnification image of the CNTs that compose a pine-tree-type bundle.

To study the degree of crystallinity of the pine-tree-like CN _x MWNT structures, Raman spectroscopy measurements were carried out. Figure 2(a) shows the Raman spectra recorded along one representative pine-tree-like structure, as depicted in the inset (see points 1–5 in the SEM image). The spectra were monitored specifically at the first-order Raman G-band; the disorder-induced band known as D-band; and the G′-band, which is the overtone of the D-band. We focused on analyzing the changes observed in the intensity of the D-, G-, and G′-bands from the bottom to the top of the “pine tree” (from point 1 to point 5), and the position of G-band. Hence, the degree of crystallinity of the material was monitored as a function of: (i) the degree of disorder or the intensity I _D/I _G and I _G′/I _G ratios and (ii) the band shift of D, G, and G′ as a function of the position along the pine structures. We noted that the intensity ratios I _D/I _G for the five points are greater than 0.8, thus indicating a relatively high degree of disorder along the pine-tree-like structure. The I _D/I _G ratios are 0.83, 0.85, 0.82, 0.83, and 0.83 for the points 1, 2, 3, 4, and 5 respectively [see Fig. 2(b)]. Therefore, it is not possible to conclusively prove that CN _x MWNTs are more crystalline from point 3 to point 5 with respect to point 1. Nevertheless, I _G′/I _G is an alternative parameter that can be used for quantifying disorder, since the G′-band feature also provides a very sensitive probe for characterizing specific sp ² hybridized nanocarbons. The G′-band is more sensitive to shifts and thus useful for determining the level of defects. ^{Reference Soin, Roy, Ray and McLaughlin39} Figure 2(c) exhibits the plot of I _G′/I _G intensity ratio, which increases as it approaches the tip. After point 3, the intensity of the G′-band shows a small increment from point 1 to 5 [see Fig. 2(a)]. The G′-band intensity appears to be higher in pure graphitic materials with low level of defects and decreases as the amount of defects increases. ^{Reference Dresselhaus, Jorio, Hofmann, Dresselhaus and Saito40} The increment of the I _G′/I _G intensity ratio indicates a low degree of crystallinity on the base of the pine-like structures (points 1 and 2) and slightly higher crystallinity at the tips (from point 3 to point 5). These differences in crystallinity could be produced because at the initial growth process, below the substrates, a pattern was generated by the flow of the precursors ^{Reference Tian, Li, Jiang, Li and Zuo41} ; in such a case, the initial CN _x MWNTs growth was faster in some specific areas (the pine-tree top). However, due to the limited space beneath the substrates, some additional flow turbulences were probably generated producing more defects at the final part of the growth (root part). Figure 2(d) shows the position of the G-band in the spectra probed from point 1 to point 5 (1596.43 and 1585.8 cm⁻¹, respectively). The observed tendency of the G-band position shows a shift toward larger wave numbers with respect to point 5 [see Fig. 2(a)]. The position of the G-band for sp ² hybridized carbon appears at approximately 1582 cm⁻¹ and it is in agreement with the values observed in doped nanocarbons. ^{Reference Dresselhaus, Jorio, Hofmann, Dresselhaus and Saito40} For undoped MWCNTs, the metallic behavior could be enhanced by the introduction of a small quantity of N atoms. ^{Reference Jeong, Lee, Shin, Choi, Shin, Kang and Choi26,Reference Wang, Shao, Matson, Li and Lin27} For example, Koos et al. showed that for undoped MWCNT fabricated with ferrocene 5 at.% in toluene, I _D/I _G is around 0.40 and for CN _x MWNTs produced with 7.5 at.% in benzylamine, I _D/I _G is close to 0.88. ^{Reference Koos, Nicholls, Dillon, Kertesz, Biro, Crossley and Grobert42} Liu et al. also measured the I _D/I _G ratios of three different CN _x MWNTs samples fabricated with melamine; their results indicated that CN _x MWNTs with 1.3, 3.2, and 4.5% of nitrogen content exhibit I _D/I _G ratios of 0.897, 1.333, and 1.553, respectively. ^{Reference Liu, Zhang, Li, Sun, Desilets, Abou-Rachid, Jaidann and Lussier43} These previous results imply a high defect concentration in CN _x MWNTs. The N-doping breaks the hexagonal symmetry, thus leading to shifts of the G-band toward values higher than 1582 cm⁻¹. ^{Reference Dresselhaus, Jorio, Hofmann, Dresselhaus and Saito40} The position of the G-band in our pine-tree-like structures is higher than the values reported for pure sp ² hybridized carbon, ^{Reference Dresselhaus, Jorio, Hofmann, Dresselhaus and Saito40} and the differences in the positions could be correlated with an increment of the nitrogen-doping level within the CNTs. ^{Reference Gupta and Patel44} We have fabricated CN _x MWNTs and undoped MWCNT using 2.5 at.% of ferrocene in toluene and benzylamine, respectively. The I _D/I _G ratios were 0.306 and 0.7769, respectively, for the undoped MWCNT and CN _x MWNTs. These values are close to those reported by Koos et al. ^{Reference Koos, Nicholls, Dillon, Kertesz, Biro, Crossley and Grobert42} (see Fig. S6 of supplementary information).

FIG. 2.

(a) Raman spectra from different zones of a pine-tree-like morphology (inset: SEM image showing labeled locations of the cone using 1, 2, 3, 4, and 5). The D-, G-, and G′-bands appear around 1362, 1596, and 2717 cm⁻¹, respectively; (b) Intensity ratios of I _D/I _G (square symbols in black color) and (c) I _G′/I _G (circle symbols in red color). (d) Second-order G′-band shifts for the different locations.

Figure 3 depicts the 45° tilted SEM images of CN _x MWNTs grown on substrate 2, and Fig. 3(a) shows the morphology of CN _x MWNTs tubes grown on top of the substrate, which was exposed directly to the feedstock [see Fig. S2(a)]. The resulting CNT array is formed by nanotubes of uniform length that contain an amorphous carbon layer on top. Figs. 3(b)–3(d) exhibit the structures formed below the substrate, which follow a pattern represented by the map drawn in the inset of Fig. 3(a). Such a map could help to identify the different zones below substrate 2, in which different structures, including “pine trees,” were found. In the map [see inset of Fig. 3(a)], the yellow zone represents the CN _x MWNTs nanotube domains exhibiting polyhedral shapes [Fig. 3(b)]. Each polyhedron is 50–70 µm in size and exhibits a small zone in the center where the nanotubes are longer than the rest of the structure. The red zone indicates the region where the CN _x MWNTs are arranged in columns of different sizes and diameters with nonuniform distribution (curled tips are also observed). Finally, the green zone is where the pine-tree-like structures were found abundantly [Fig. 3(d)]. We believe that CN _x MWNTs tubes grew at different rates leading to these pine-tree-like morphologies. For an easy identification of samples with different structures, we named CN _x MWNTs that grew on substrate 2 as CN _x -P0 and the polyhedral, columns, and pine-tree shapes as CN _x -P1, CN _x -P2, and CN _x -P3, respectively.

FIG. 3.

SEM images of CN _x MWNTs arrays: (a) Well-aligned CN _x MWNTs tubes preserving the same length with a deposited layer of amorphous carbon on top (labeled CN _x -P0). Inset: schematic distribution of the pine-tree zones observed below substrate 2. (b) Aligned CN _x MWNTs with certain resemblance to polygon-like morphologies (labeled CN _x -P1/yellow zone in the map in (a). These structures were grown preferentially near the substrate edge. (c) Nanotube bundles grown on the substrate edge interacting directly with the feedstock input (labeled CN _x -P2/red zone in the inset map). (d) Pine-tree-like morphologies of nanotubes arrays [labeled CN _x -P3/green zone in the map in (a)].

An XPS analysis was performed on the CN _x samples (see Fig. 4). High resolution C1s, N1s, and O1s line scans were acquired. The presence of 1.9 at.% nitrogen in the sample was confirmed. Concentrations of carbon and oxygen were found to be 95.7 and 2.4%, respectively. The N1s line scan was carefully analyzed and curve fittings were performed using a Shirley background. Two main different nitrogen environments were detected at approximately 399 and 401 eV. These signals have been linked to N substitutionally immersed in the graphitic lattice and to pyridinic N. ^{Reference Elias, Ayala, Zamudio, Grobosch, Cruz-Silva, Romo-Herrera, Campos-Delgado, Terrones, Pichler and Terrones45,Reference Lv, Li, Botello-Mendez, Hayashi, Wang, Berkdemir, Hao, Elias, Cruz-Silva, Gutierrez, Kim, Muramatsu, Zhu, Endo, Terrones, Charlier, Pan and Terrones46} Therefore, we can be sure that the nitrogen signal detected by XPS belongs to N that has been successfully introduced within the graphitic lattice of the tubes. The inset of Fig. 4 show the XPS C1s, O1s, and N1s line scans obtained from control-undoped MWNT samples. The nitrogen content found was absent, C1s and O1s were found at 98 and 2 at.%, respectively.

FIG. 4.

(a) XPS N1s line scan obtained from the CN _x MWNT samples. The signal can be attributed to two main nitrogen-bonding environments: substitutional N and pyridinic N; (b) N1s line scan obtained from the control samples of undoped MWNT samples.

The electron FE properties of different regions of the substrates containing arrays of pine-tree-like structures, as well as the other structures, were determined by measuring the current versus the applied field [Fig. 5(a)]. The current was collected by a 1 cm² copper anode, placed at 200 µm above a substrate containing the different CN _x MWNTs morphologies explained above (Fig. 3). The FE properties of the samples were analyzed using the Fowler–Nordheim (FN) theory, which provides a quantitative description of the emission process. ^{Reference Fursey47} Experimentally, the FN plot could be achieved by plotting (I/E ²) as a function of 1/V [see Fig. 5(b)]. ^{Reference Sharma, Kaushik, Girdhar, Singh, Shukla and Vankar48} From this plot, it is possible to determine the field enhancement factor by measuring the slope s in the linear region using the expression: β = − Bϕ^3/2 d/s, ^{Reference Perea-Lopez, Rebollo-Plata, Briones-Leon, Morelos-Gomez, Hernandez-Cruz, Hirata, Meunier, Botello-Mendez, Charlier, Maruyama, Munoz-Sandoval, Lopez-Urias, Terrones and Terrones14} where B is a constant (6.83 × 10⁹ V eV^3/2 m⁻¹), ϕ is the work function (5 eV, the same as graphite), and d is the electrode separation distance (∼200 µm). Figures 5(a) and 5(c) show the current–field (I–E) curves and FN plots, respectively, for different micromorphologies.

FIG. 5.

(a) Electron FE curves for four different samples labeled CN _x -P0, CN _x -P1, CN _x -P2, and CN _x -P3; (b) the turn-on field measured from the FE curves as a function of different micromorphologies. Note that the FE currents are higher and an enhancement in the turn-on field are obtained with pine-tree-like morphologies; (c) F–N plot, and (d) the β factor value plot with different micromorphologies; the pine-tree-like structures exhibit the best FE performance.

From Fig. 5(a), it can be observed that the I–E curve for CN _x MWNTs grown on the top side of the substrate (CN _x -P0) yields to few microamperes of high field values, thus revealing a poor performance. This effect could be attributed not only to the accumulation of amorphous carbon “crust” on top of the forest but also to the flat shape of the substrate. ^{Reference de Jonge and Bonard49} In this case, a turn-on voltage of 2.20 V/µm was needed to produce 100 nA of current. Bai et al. ^{Reference Bai, Zhang and Zhang50} performed similar experiments using MWCNT arrays grown on silicon substrates, and found turn-on values above 2.11 V/µm. For the CN _x -P1 sample (polyhedral structure), the FE curve exhibited a better response (turn-on field of 1.93 V/µm, 13% lower than that of CN _x -P0). The reduction of the turn-on field for CN _x -P1 with respect to CN _x -P0 could be attributed to the electrons localized in the small bundles that outstand from the center of polyhedrons [see Fig. 3(b)]. These protuberances allow electrons to be emitted easily. For CN _x -P2 (cylindrical bundles), the emission properties are considerably better than those of the polyhedral structure, and the turn-on field is around 1.55 V/µm, which is even lower than that of the individual bundles. ^{Reference Perea-Lopez, Rebollo-Plata, Briones-Leon, Morelos-Gomez, Hernandez-Cruz, Hirata, Meunier, Botello-Mendez, Charlier, Maruyama, Munoz-Sandoval, Lopez-Urias, Terrones and Terrones14} In addition, we observed that the arrays of cylindrical bundles are not uniform and some of them bend and curl, thus affecting the FE performance. The screening effect between adjacent CNT bundles could also affect the emission properties. ^{Reference Zeng and Ren51} For CN _x -P3, bundles with pine-like morphology, the I–V curve has an exponential behavior showing turn-on fields of 0.59 V/µm. Figure 5(b) shows the turn-on field (E _ON) values as a function of the structure type. One can clearly observe that the lowest turn-on fields correspond to the pine-tree-like structures, attributed to the conical geometrical structure, ^{Reference de Jonge and Bonard49,Reference Eletskii and Bocharov52} which has more emitting sites than the polygonal structures (CN _x -P1) but have enough separation to avoid the screening effects observed for cylindrical bundles (CN _x -P2). The observed value of E _ON for the CN _x MWNTs with pine-like morphology (0.59 V/µm) is lower than those reported for other materials, such as Ga-doped ZnO nanopins (1.92 V/µm). ^{Reference Sun53} It is also lower than those of CN _x MWNTs arrays (1.80 V/µm), ^{Reference Ghosh, Kumar, Maruyama and Ando28} but is in the range of pure nonaligned MWCNT (0.75 V/µm). ^{Reference Lahiri, Seelaboyina, Hwang, Banerjee and Choi54} However, it is higher than the turn-on voltage of long individual bundles of CNTs (0.18 V/µm). ^{Reference Perea-Lopez, Rebollo-Plata, Briones-Leon, Morelos-Gomez, Hernandez-Cruz, Hirata, Meunier, Botello-Mendez, Charlier, Maruyama, Munoz-Sandoval, Lopez-Urias, Terrones and Terrones14} Yu and Zheng studied the effect of nitrogen doping on the electronic and FE properties of CNTs, carbon nanocones, and graphene nanoribbons. ^{Reference Yu and Zheng55} They demonstrated that the doping of N atoms into CNT can improve the FE properties of these carbon-based materials. Killian and collaborators also reported FE properties of CNT pillar arrays. ^{Reference Killian, Zuckerman, Niemann, Ribaya, Rahman, Espinosa, Meyyappan and Nguyen56} They found emission at fields as low as 0.9 V/µm (much lower than those of the continuous arrays ^{Reference Jo, Tu, Huang, Carnahan, Wang and Ren57} ) and stable current densities of 10 mA/cm² at an average field of 5.7 V/µm with a very low percentage of standard deviation. However, it is noteworthy that our results demonstrate that the pine-tree morphology is optimal for FE and the presence of nitrogen can enhance these properties further.

We also calculated the field intensification factor (β) from the FN plots shown in Fig. 5(c). Figure 5(d) depicts the plot of β factor as a function of the different bundle morphologies. The β factor value was lower for CN _x -P0 (3.66 × 10³), CN _x -P1 (4.64 × 10³), and CN _x -P2 (5.55 × 10³), when compared to the β factor of pine-tree-like structures (7.67 × 10³). From these results, it was clear that the “spiky” structures showed the best FE properties, and the conical clustering and increased separation between pines improve the turn-on fields and the β factor, ^{Reference Bonard, Salvetat, Stockli, Forro and Chatelain13} which is superior than that reported for aligned arrays of CN _x MWNTs (2000). ^{Reference Ghosh, Kumar, Maruyama and Ando28} The current stability versus time for pine trees measured during approximately four days is included in the SI (Fig. S7). In Table S1 (see supplementary information) a description of different results obtained in this work and their relationship with the processing parameters is shown. In the same table other results published in the literature are included. From Table S1, it is now possible to observe that very less information is available about the relationship between the electron FE properties of CN _x MWNTs and their N-concentration and the type N-incorporation within the CNT matrix. Additionally, it can be seen that although the bundles of CNTs exhibit enhanced FE properties, ^{Reference Perea-Lopez, Rebollo-Plata, Briones-Leon, Morelos-Gomez, Hernandez-Cruz, Hirata, Meunier, Botello-Mendez, Charlier, Maruyama, Munoz-Sandoval, Lopez-Urias, Terrones and Terrones14} the mats of CN _x MWNTs fabricated with camphor and 5.2% of nitrogen content can also display a good FE performance, ^{Reference Ghosh, Kumar, Maruyama and Ando28} similar to the pine-tree-like CN _x MWNTs.

To explain the growth mechanism of the pine-tree-like morphologies in CN _x MWNTs tube forests, it is important to realize that the transport of the nebulized cloud containing ferrocene (Fe(C₅H₅)₂) and benzylamine (C₇H₉N) was carried out by a flux of Ar. Figure 6 schematically shows how the process may occur during CVD. Figure 6 depicts a drawing of the pyrolysis process involving ferrocene and benzylamine. In short, the aerosol mixture (shown in orange color) is carried by the Ar flux. This cloud is then decomposed inside the furnace zone. As the cloud enters the furnace, the decomposition of ferrocene occurs at temperatures higher than 500 °C, thus releasing Fe atoms, small iron clusters and nanoparticles, and hydrocarbon side-products (Fe(C₅H₅)₂ → Fe + H₂ + CH₄ + C₅H₆ + C₉H₈ + etc.). ^{Reference Barreiro, Hampel, Rummeli, Kramberger, Gruneis, Biedermann, Leonhardt, Gemming, Buchner, Bachtold and Pichler58,Reference Leonhardt, Hampel, Muller, Monch, Koseva, Ritschel, Elefant, Biedermann and Buchner59} Simultaneously, the benzylamine gets decomposed at temperatures >600 °C, leading to benzene (C₆H₆), toluene (C₇H₈), benzonitrile (C₇H₅N), hydrogen cyanide (CHN), and ammonia (NH₃). This thermal decomposition yields to the gas phase feedstock (see burgundy color inside the quartz tube; Fig. 6). In this process, the iron clusters and nanoparticles, which are chemically reactive, tend to form spherical agglomerates covered with carbon multishells. However, an excess of carbon leads to the formation of single- or multiwall carbon nanotubes instead of multishell-covered nanoparticles. ^{Reference Koprinarov, Konstantinova and Marinov60} Such nanoparticle agglomerates can reach 50 nm of diameter and are composed of pure iron, iron carbide, or both in a core–shell structure; these phases are formed between 400 and 800 °C. This process is carried out using standard CVD synthesis conditions (e.g., atmospheric pressure and uniform gas flow), where the resulting material grows uniformly and the CNTs start growing in a vertically aligned manner after a few seconds. ^{Reference Hiraki, Jiang, Wang and Hiraki22,Reference Zhang, Li, Tan, Lolli, Sakulchaicharoen, Requejo, Mun and Resasco61,Reference Bower, Zhu, Jin and Zhou62} For the different substrate locations, we noted that the flux of the incoming gaseous species traveling inside the quartz tube is affected significantly beneath the substrates due to reduced space between the substrate and the quartz tube. These conditions locally affect the distribution of the agglomerated catalytic metallic nanoparticles, thus leading to the nucleation and growth of CNTs at different rates.

FIG. 6.

Schematic representation of the CVD process involving ferrocene (Fe(C₅H₅)₂) and benzylamine (C₇H₉N) mixtures.

To visualize the different morphologies grown on substrate 2, Fig. S8 shows a hypothetical drawing representing the SEM micrograph of substrate 2 after the CVD experiment [see Fig. S2(b)]. Here, the representation of different microstructures produced with CNTs are colored; the small blue circles represent the regions where the cylindrical bundles grow; the big blue circles with a black point in the center correspond to small columns with an almost complete pine-tree-like morphology. Next to these structures is a red zone where pine trees with different diameters are formed (green circles with a dark green point in the center). The red regions correspond to aligned CNTs with a similar size. Next to the “pine trees,” polygonal structures were formed. At the center of these polygonal objects, there is a small bundle emerging as seen in the image (yellow point). Additionally, the diameter of the bundles that grow at the entrance matches with the sizes of the bundles located at the center of the pine trees and the polygons.