On the CVD Growth of C Nanotubes over Fe-Loaded Montmorillonite Catalysts

2.1 Catalysts

2.2 Carbonaceous deposits

2.2.1 Synthesis and yield evaluation

CVD growth reactions were carried out in i-C₄H₁₀+H₂ atmosphere at 700 °C. The reduced catalyst (0.5 g) was placed in a quartz boat inside the quartz reactor, located in a horizontal electric furnace, and preliminarily heated upon 120 cc/min 1:1 He+H₂ flow up to synthesis temperature. Then helium was replaced with isobutane keeping constant flow ratio and total flow rate. After 2 h the reaction was stopped. The raw products were cooled down to room temperature (RT) in He atmosphere and weighed (0.1 mg balance sensitivity) for yield evaluation.

The mass of deposited carbon (m_C = m_T-m_R) was estimated subtracting the mass (m_R) of reduced catalyst employed from the mass (m_T) of materials (reaction products+catalyst) discharged from the reactor at the end of each synthesis. Two trials were carried for each parameter setting. Yield

Y C = 100 ⋅ m C / m R

(1.a)

and specific yield

S Y C = 100 ⋅ m C / m Fe

(1.b)

OPEN IN VIEWER

Table 1.

Codes, reduction temperature (T_R) and specifics of all the considered catalysts. Catalyst codes, NKw_L and Kw_L, summarize information concerning the type of clay (NK and K stand for Na⁺-exchanged and un-exchanged K10 clay, respectively) and iron extra-load (w_L = 5.0-25.0 wt%). NK0 and K0 denote unloaded catalysts. The overall metal content (w_Fe) is the sum of iron already contained in NK0 (1.7 wt%) and K0 (2.0 wt%) and iron added by impregnation. d_Fe and A_BET denote mean size of Fe-particles and measured specific surface area of the overall catalysts (inclusive of metal and clay-support). A star marks cases in which Fe⁰ is not detected in XRD patterns. Mean catalytic yields (Y_C) and specific yields (SY_C) are also reported, as well as calculated dispersion (D) and specific surface area of iron (A_Fe).

Sample code	wFe (wt%)	TR (°C)	ABET (m2/g)	dFe (nm)	D	AFe (m2/g)	YC (wt%)	SYC (wt%)
NK0	1.7	500	222	*	—	—	3.0	176.5
NK5	6.7	500	164	*	—	—	29.5	440.3
NK15	16.7	500	157	22.4	5.2 × 10−2	5.1	179.9	1077.2
NK20	21.7	500	170	24.4	4.7 × 10−2	6.3	217.8	1003.7
NK25	26.7	500	109	27.6	4.2 × 10−2	6.9	207.0	775.3
K0 [7,8]	2.0	700	201	*	—	—	0.4	10.0
K5 [7,8]	7.0	700	115	32.0	3.6 × 10−2	1.2	14.7	210.0
K15 [7,8]	17.0	700	69	45.1	2.6 × 10−2	2.5	31.7	186.5

of the catalytic process were calculated by normalizing m_C to the mass (m_R) of the whole catalyst (clay-support included) [7,8,10,11] and to the mass (m_Fe) of its sole active-component. Mean values obtained are reported in Tab. 1.

4.1 Catalysts

The diffraction peaks of Na⁺-K10-montmorillonite clay [8] are detected in all the catalysts (Fig. 1). Structural analysis further reveals the presence of muscovite and quartz as clay's impurities. The peak of Fe0 is not detected in NK0 and NK5 catalysts [8], likely due to the insufficient amount of iron and/or to the smaller particle size. Instead, in the remaining catalysts (Fig. 1) its intensity progressively increases with the amount of metal loaded.

Also d_Fe slowly increases with Fe-load (Tab. 1), causing also dispersion and exposed surface of the metal to vary. Fe dispersion, namely the molar fraction of exposed iron, is calculated, assuming a spherical particle approximation, as [20]

OPEN IN VIEWER

Table 2.

Results of TEM, TG, DTG and MRS analyses carried out on purified reaction products. w_I is the relative amount of Fe-impurities present in C deposits. T_P is peak temperature of DTG profiles giving a measure of thermal stability. The G'/D intensity ratio (I_G'/I_D) monitors crystalline quality. The G/D intensity ratio (I_G/I_D) allows estimating the average size (L_C) of graphitic crystallites (the values calculated following both refs. [18,19] (a) and refs. [16,17] (b) are reported).

Sample code	TEM	TG	DTG	MRS
	Morphology	wI (wt%)	TP (°C)	IG'/ID	IG/ID	LC (a) (nm)	LC (b) (nm)
NK0	CNFs+DGFs	16.6	475
NK5	MWCNTs	4.5	535	1.34	1.17	19.4	5.2
NK15	MWCNTs	10.0	536	2.43	1.93	32.0	8.5
NK20	MWCNTs	4.2	514	1.13	0.88	14.6	3.9
NK25	MWCNTs	5.5	515	0.99	0.88	14.6	3.9

D = d VA / d Fe = 6 τ Fe / ( a Fe ⋅ d Fe ) ,

(2)

where d_Fe is the mean metal particle size calculated via the Scherrer equation, d_VA=6τ_M/a_M is the volume-area mean diameter, a_M = 6.13×10⁻²⁰ m² is the effective average area occupied by a metal atom in the surface [20] and τ_M=1.18×10⁻²⁹m³ is the volume per metal atom in the bulk, calculated as w/(N_A·ρ), with N_A=6.023×10²³/mol standing for the Avogadro's number, w=55.845 a.m.u. denoting the atomic weight of iron and ρ=7.86×10⁶ g/m³ being the mass density of bulk iron.

By this procedure, D is found to decrease from 5.2×10⁻² to 4.2×10⁻² going from sample NK15 to NK25 (Tab.1). The same calculation is applied to the formerly studied un-exchanged K10-catalysts reduced at 700 ×C. Lower D values (Tab. 1) are obtained for catalysts K5 and K15, based on d_Fe data inferred from previous XRD measurements [8]. Specific surface area

A Fe = D ⋅ a M ⋅ N A ⋅ n Fe

(3.a)

(where n_Fe is the overall number of moles of added iron per gram of catalyst) and surface area

S Fe = A Fe ⋅ m R

(3.b)

of the exposed iron are found to slightly increase, respectively from 5.1 m²/g to 6.9 m²/g (Tab.1) and from 2.6 m²/g to 3.5 m²/g, going from sample NK15 to NK25 (Tab. 1). Smaller A_Fe (Tab. 1) and S_Fe values are obtained for catalysts K5 and K15.

It is worthwhile noting that in all the cases the calculated specific surface area of the metal is always much smaller than the overall specific surface area, A_BET, of the catalysts (Tab. 1), demonstrating that iron does not completely cover the clay surface, either at higher loads.

4.2 Carbon deposits

TEM analysis had previously evidenced [7,8] that only carbon nanofibers (CNFs) and disordered graphite flakes (DGFs) form on the unloaded clay. Present study (Fig. 3) reveals that MWCNTs are, instead, obtained over all the Fe-loaded catalysts, regardless their load. SEM analysis (Fig. 2) evidences no substantial change in their morphology, but increased entanglement of the tubes and less uniform and larger diameters, in accord with the increase of d_Fe [21,22].

Disordered carbon nanostructures formed over NK0 are here found to contain the highest amount of Fe-impurities (w_I = 17 wt%). In addition, as lattice defects act as initial oxidation sites [23], they also possess the lowest thermal stability (T_P = 475 °C).

Among MWCNTs, the highest oxidative resistance (T_P = 536 °C) pertains to sample NK15, where TEM analysis reveals the presence of straight and smooth tubes consisting of regular sequence of perfectly ordered graphitic layers (Fig. 3). Very intense G- and G'-bands dominate the Raman spectrum of this sample (Fig. 5), whose crystalline quality (I_G'/I_D = 2.43) largely exceeds that of remaining samples (Tab. 2), as well as that of MWCNTs reported as the best crystallized ever obtained over clay-based catalysts [5,7]. The high crystalline perfection of MWCNTs NK15 is proved also by the size of crystallites (32 nm), larger by a factor of 1.5-2.0 than in remaining NK_wL samples (Tab. 2).

Recent studies have then demonstrated that, thanks to the high graphization-degree of these MWCNTs, the incorporation of small amounts (1-5 wt%) of hybrid NK15 (MWCNTs + clay) within an insulating polymeric matrix outstandingly enhances its electrical conductivity (up to 9 orders of magnitude) with respect to the pristine polymer [9].

4.3 Growth process over clay-based catalysts

As shown in Tab. 1, at lower loads, Y_C rapidly increases with Fe addition. The maximum value (∼2.2 grams of deposited carbon per gram of catalyst) is reached for w_L = 20 wt%. Above this load Y_C decreases, in spite the larger amount of iron. This is because SY_C progressively drops above w_L = 15 wt%, load for which the largest amount (∼12 grams) of carbon per gram of iron is produced.

As initially mentioned, previous studies demonstrated that lowering the acidity of the clay support, via Na⁺ exchange reaction or reduction-temperature enhancement, inhibits the formation of highly disordered carbonaceous nanostructures, responsible for metal deactivation at small iron loads (w_L = 5 wt%), producing beneficial effects on yield [7].

For w_L≤15 wt%, Y_C was shown to effectively increase exponentially with the product w_Fe·A_BET in un-exchanged K10 catalysts reduced at 700 °C [7], as well as in Na⁺-exchanged K10 catalysts reduced at lower temperature (500 °C) as in present study. A saturating trend is here observed at higher loads [Fig. 6(a)].

OPEN IN VIEWER

Figure 6.

(a) Comparison between plots of Y_C vs. the product w_Fe·A_BET and of m_C vs. S_Fe for Na⁺-K10 catalysts (present study) and K10 catalysts [7] reduced at different temperature (arrows indicate which axes the data refer to; length of error bars equals the difference between Y_C values obtained in the two trials). The dependence of (b) A_BET and (c) SY_C on d_Fe is also shown.

The same trend is observed when m_C is plotted against the exposed surface of loaded iron S_Fe [Fig. 6(a)]. Note that this regards exclusively samples (NK15-NK25 and K5-K15) for which d_Fe data are available. As a fixed amount of reduced catalyst is here used, based eq. (1.a), m_C is proportional to Y_C. Therefore, the similarity between Y_C vs w_Fe·A_BET and m_C vs S_Fe trends hints at the existence of a linear relationship between S_Fe and w_Fe·A_BET, as actually found. In fact, A_BET is here found to coarsely scale as 1/d_Fe [Fig. 6(b)]. For fixed m_R=0.5 g, S_Fe is proportional to A_Fe [eq. (3.b)]. In turn, according to eq. (3.a), A_Fe ∞ D·n_Fe, where of course n_Fe ∞ w_Fe and, based on eq. (2), D ∞ 1/d_Fe just alike A_BET.

Both these findings deserve a detailed and attentive discussion in order to avoid drawing erroneous conclusions.

An important consequence of the existence of the effective dependence of Y_C on w_Fe·A_BET is that, in Kw_L catalysts reduced at 700 °C, a higher w_Fe may compensate for a lower A_BET to finally give the same (or comparable) w_Fe·A_BET product as in NKw_L catalysts reduced at 500 °C, hence resulting in the same (or comparable) Y_C. This is immediately evident by considering the values of w_Fe, A_BET and Y_C relative to samples NK5 and K15 reported in Tab. 1 and/or by comparing the corresponding data in Fig. 6(a). As can be seen, as strictly concerns catalytic yield, in spite of different w_Fe and T_R (and resulting n_Fe and d_Fe) the two catalysts behave similarly.

As for the found dependence of m_C on S_Fe, it can be argued that, for catalysts NK15-NK25 and K5-K15, the amount of produced carbon is ultimately determined exclusively by the exposed surface of added iron, as evaluated [through eqs. (3.a) and (3.b)] on the basis of merely geometric factors (d_Fe and resulting D) besides to, of course, the iron content (n_Fe) and the catalyst amount (m_R).

This is not a surprising result because all measured d_Fe values largely exceed 10 nm (Tab. 1), value above which the relative amount of the diverse types of iron atoms (saturated ones, present in the crystal face, and unsaturated ones, present at edges, corners and terraces), endowed with different catalytic activity, hardly changes with size [20].

Based on the similar trends of the curves Y_C (W_Fe·A_BET) and m_C(S_Fe) [Fig. 6(a)], one might be brought to extend considerations made for NK15-NK25 and K5-K15 to the remaining samples. Indeed, the lack of information about d_Fe impedes any straight widening of the above discussion. In fact, for d_Fe < 10 nm, also the different catalytic activity and relative amount of the diverse types of iron atoms should be taken into account.

By fitting 10 available experimental data, it is found that the mass m_C of nanotubes, synthesized at 700 °C upon (120 cc/min) 1:1 i-C₄H₁₀+H₂ flow over (0.5 g) Na⁺-K10 and K10 iron catalysts, increases with the exposed surface of loaded iron S_Fe according to the relationship

m C = 1.09 - 1.07 ⋅ exp ( - ( ( S Fe + 5.50 ) / 7.74 ) 12.9 ) .

(4)

Although the choice of the function to be utilized to reproduce m_C data of Fig. 6(a) is to some extent arbitrary, that here used is theoretically inferred from the simple physical model below.

To this purpose, it is worthwhile noting that, far from establishing a general model of the growth process, the discussion below aims exclusively at accounting for the changes of m_C with S_Fe observed in the samples under study.

Therefore, a stringent simplifying assumption is introduced, consisting in attributing the same active surface to both the kinds of unloaded catalysts (Na⁺-exchanged K10 reduced at 500 °C and un-exchanged K10 reduced at 700 °C).

The main reason for this choice is that very large errors affect lower Y_C and m_C values (relative error is well above 75 % in case of K0 catalyst), which renders them hardly distinguishable, at least as concerns the present aim of reproducing the changes of m_C with S_Fe.

Actually, unloaded catalysts exhibit comparable iron contents (Tab. 1), but the diverse T_R might introduce a relevant size-difference. However, the very small reduction of the pristine iron content (δw_Fe =−0.3 wt%, Tab. 1), produced by the ion exchange reaction with a NaCl solution, suggests that Fe cation is not in an exchangeable position, but it is mainly present as a structural ion and/or as iron oxide formed from the dissolution of octahedral iron sites upon acid treatment [8]. This hints at a strong iron/support interaction, which likely hampers, or at least severely limits, possible sintering effects and size-variations at higher T_R. Indeed, A_BET, which, as shown [Fig. 6(b)], is sensitive to the changes of d_Fe, exhibits quite similar values in the two unloaded catalysts (Tab. 1).

Thus, indicated with S₀ the active surface in unloaded catalysts and with S_Fe the surface increase brought about by iron addition, let S = S₀+S_Fe be the overall iron exposed-surface and m_C the mass of carbon correspondingly synthesized.

Because of the low fraction of supplied carbon that is transformed to carbon deposits (maximum: 6.8 wt% after 2 h of reaction for w_L = 20 wt%), the incident C flux does not suffer significant impoverishment owing to the formation of carbonaceous products. Thus, in principle, it should be expected that the increase of S with w_L would produce a linear increase of m_C. Nonetheless, it has to be borne in mind that the increase of S with w_L is the result of the simultaneous variation of manifold influential factors (amount, size and related dispersion of Fe-nanoparticles, possibly involving also textural effects) and that an optimal size-range relative to C supply [24,25] does exist, as demonstrated by the dependence of SY_C on d_Fe shown in Fig. 6(c).

Based on all these factors, a non-linear dependence of m_C on S can be reasonably hypothesized.

Therefore, if the variation, d_mC, undergone by m_C for an increment of the effective-variable d(S^α), is taken as proportional, according to the simplest possibility, besides to d(S^α), to the residual “distance” (m_C^max-m_C) from the largest mass amount, m_C^max =1.09±0.04 g (where the error is evaluated considering the different Y_C values obtained in the two trials carried out upon fixed conditions), which is produced under the considered growth conditions, namely

d m C ∝ ( m C max - m C ( S α ) ) ⋅ d ( S α ) ,

by setting X = m_C^max-m_C, the differential equation dX = -a·X·d(S^α) is obtained that can be solved in terms of X to give

X = A ⋅ exp ( - ( S / S * ) α ) .

This finally becomes

m C = m C max - A ⋅ exp ( - ( ( S 0 + S Fe ) / S * ) α ) ,

where S* is a critical S value. This expression, besides accounting for the positive curvature of the m_C(S_Fe) curve at small S_Fe (Fig. 3), is also able to reproduce the negative curvature and saturating trend at larger S_Fe. As the experimental data are fitted to this function, A = 1.07 ± 0.02 g, S₀ = 5.50±0.04 m², S* = 7.74±0.04 m² and α=12.9±0.6 are found, leading to eq. (4) and validating the quite simple assumptions made. Empirical law (4) describes how the amount of grown MWCNTs increases with the overall exposed surface of iron, and includes the effect of the simultaneous variation of number, size and dispersion of Fe-nanoparticles available for their growth, as well as of the existence of an optimal size-range relative to C supply [24,25].

In case of mono-dispersed iron (namely, D=1), eq. (2) gives d_Fe = d_VA = 1.2 nm that is the diameter of “only-surface-atom cluster” containing ca 100 atoms [20]. Interestingly, assuming that in the unloaded catalyst iron is nearly mono-dispersed (D≤1) and accordingly using d_Fe≥d_VA as mean Fe-nanoparticle size value to calculate specific surface area of the exposed iron, from eq. (3.a) A_Fe≤11.24 m²/g is obtained. According to eq. (3.b), being m_R=0.5 g, this corresponds to S₀≤5.62 m², value which is fairly compatible with the one (5.50 m²) derived from the fitting procedure.

Also the large value of the exponent α (12.9) needs a brief comment. Combining eqs. (1.a) and (1.b) catalytic yield can be written as Y_C = w_Fe·SY_C/100, so as m_C = w_Fe·SY_C·m_R. Thus, the observed (non linear) dependence of m_C on S_Fe [i.e. eq. (4)] is through w_Fe and SY_C. Different from w_Fe, whose changes cause a smooth variation of d_Fe (22.4-27.6 nm at 500 °C and 32.0-45 nm at 700 °C) and, hence, of S_Fe, SY_C is strongly sensitive to the changes of d_Fe and drastically declines at larger d_Fe [Fig. 6(c)].