1 Introduction

Carbon nanotubes (CNTs) have attracted considerable interest owing to their outstanding mechanical, thermal, electrical, and structural properties [1,2,3]. The global demand for CNTs has increased significantly [4,5,6]. The catalytic decomposition of methane (CDM), as one of the typical chemical vapor deposition processes, is the most promising method for large-scale CNTs preparation. However, the purity of as-prepared CNTs (AP-CNTs) cannot be ensured because of the real-time fluctuation in the catalyst structure and the reaction conditions. The impurities are of three types: residual metal catalyst (1–10 wt.%), carbon impurities (CIs, such as amorphous carbon (AC), carbon black, graphitic fragments (GFs), and imperfections in the CNTs skeleton (such as pentagons, heptagonal carbon rings, and oxygen-containing groups). All these impurities significantly affect the CNTs properties and must be eliminated before used as a qualified product [7, 8].

Various methods have been adopted to purify the AP-CNTs [8]. The physical methods, including size exclusion chromatography, filtration, centrifugation, and high-temperature annealing, can protect the MWCNTs from damage but are usually inefficient and complex [1]. Chemical methods are more applicable, classified by liquid treatment or gas phase oxidation.

Firstly, the liquid phase treatment can typically dissolve the residual metal. A key strategy for removing the residual metal embedded in the CNTs or CIs is to damage the graphitic shells [9] or CIs and then etch the metal away using liquid oxidizers such as HNO3, H2SO4, H2O2, or KMnO4 solution or their mixtures. Although this approach can decrease the residual metal content to below 1000 ppm, the CNTs structure is inevitably damaged by the strong oxidizability, resulting in the chemical modification of the CNTs, specifically, the implantation of oxygen-containing groups on the CNTs skeleton [10, 11]. Maintaining the pristine structure of the CNTs is a crucial priority for any refinement process. Based on this goal, hydrochloric acid (HCl), a non-oxidizing acid, is an excellent choice for removing residual metal without extensive CNTs destruction [12]. However, its effect is unsatisfactory because the metal particles are usually embedded in CIs and CNTs, making it difficult for HCl to access the metals [1, 8, 13]. Therefore, it is necessary to expose the embedded metals to HCl so that the residual metal can be removed.

Secondly, the oxidation gas (such as air/O2 [1, 14, 15] or CO2 [16]) oxidizes the CIs. However, it is difficult to remove CIs without excessive loss of the CNTs, owing to the highly similar chemical reactivity of the CIs and CNTs and their entanglement in AP-CNTs [8]. The oxidized rate of sp2 carbon by CO2 is lower than that by air/O2 [16], which results in fewer defects. Therefore, CO2 is considered as a better mild oxidizer than air/O2 for removing CIs. However, the complete removal of CIs by CO2 oxidation is difficult because CIs are located either outside or inside the CNTs skeleton. It is relatively easy to remove external CIs because they are exposed to the oxidant directly; by contrast, it is difficult to remove internal CIs because they are within the CNTs structure. Thus, Separating the internal CIs from the CNTs facilitates the purification of the AP-MWCNTs.

Based on the analysis above, the separation of residual metal, CIs from the CNTs skeleton is a priority for the AP-CNTs purification process. Ultrasonication can functionalize the surface of the CIs and the CNTs, enhancing their detachment effect [17,18,19]. High-energy ultrasonication is advantageous for more effective disruption of the van der Waals interaction of the CIs, residual metal with the CNTs, but it causes more severe damage to the CNTs structure. The severity of the damage is strongly affected by the dispersant [20], operating parameters of the ultrasonic device [19, 21], and natural characters of the AP-CNTs. In particular, the structural deformation of the CNTs is solvent-dependent. Less damage has been observed after sonication in water or ethanol, and high-energy ultrasonication has been found to cause more severe damage to carbon nanoparticles than to CNTs [20]. However, the current researches on ultrasonication have focused on commercial CNTs dispersion [22, 23]. The effect of ultrasonic dispersion on the CIs, residual metal, and the CNTs during the purification process has not been investigated.

Although many high-quality investigations on MWCNTs purification have been published, few were focused on the root that influence the yield and damage degree of MWCNTs. In this work, a mild three-step cascade purification process is proposed to purify the MWCNTs from the raw soot synthesized by the CCVD method. This treatment procedure mainly involves three steps: the ultrasonication of the AP-MWCNTs in ethanol to separate the CIs and residual metal, mild gas oxidation by CO2 to eliminate the CIs, and non-oxidative etching by HCl to remove residual metal. It is found that this purification process maintains a higher yield with few defects on the pristine CNTs skeleton.

2 Experimental

2.1 Materials

Pristine AP-MWCNTs were prepared by the CDM according to the reaction process described in the Supplementary Material or in the literature [24]. First, the collected AP-MWCNTs were ground to the powder (> 200 mesh). Compressed air (21 ± 0.5 vol.% oxygen, balance nitrogen), carbon dioxide (99.2 vol.%), and nitrogen (99.999 vol.%) were purchased from Chengdu Guangdu Gas Co. Ltd. All other chemicals, including ethanol, isopropanol, methanol, HCl, HNO3, and H2O2, were purchased from Chengdu KeLong Chemical Reagent Co. Ltd. and used without further purification.

2.2 Ultrasonic dispersion

AP-MWCNTs (100 mg) were placed in a 250 mL conical flask, and 100 mL of pure ethanol, isopropanol, methanol, or deionized (DI) water was added as a dispersant. The conical flask was sealed and placed in an ultrasonic cleaner (BELON10-300, Shanghai Belon Instrument Manufacturing Co. Ltd.) for 5–25 h of ultrasonication at 50–70 °C. Then the sample was filtered by a microporous hydrophobic polytetrafluoroethylene (PTFE) membrane (pore size, 0.22 μm) and then dried at 80 °C.

2.3 Etching and oxidation

AP-MWCNTs (100 mg) and 100 mL of an acid aqueous solution (5 mol·L−1 HCl, 5 mol·L−1 HNO3, or 5 mol·L−1 HCl + HNO3 at a ratio of 3:1, or 5 mol·L−1 HCl + H2O2 at a ratio of 3:1) or 100 mL of a 5 mol·L−1 HCl ethanol solution (named as HCl (E)) were mixed in a 250 mL conical flask, which was soaked in an oil bath with a condensation reflux device at 120 °C with continuous magnetic stirring at 300 rpm for 10 h. The sample was filtered by a PTFE membrane, washed with ethanol, and dried at 80 °C for 12 h.

The CIs in the AP-MWCNTs were selectively oxidized by CO2. The sample was placed in a fixed-bed reactor and heated to an arbitrary temperature based on the TGA results under N2 atmosphere. After the temperature became stable, 50 mL·min−1 CO2 and 50 mL·min−1 N2 were passed through the sample for 4 h, and the sample was cooled naturally to room temperature to obtain purified MWCNTs.

2.4 Mild cascade purification process

The details of CO2 oxidation and HCl etching are listed in the SI. The MWCNTs purification steps (dispersion in ethanol, oxidation by CO2, and etching with HCl) were combined to construct various mild cascade purification processes, as shown in Fig. 1.

Fig. 1
figure 1

Descriptions of mild cascade purification processes

Full size image

P1 was designed as a comparison experiment. In P1-1, AP-MWCNTs were oxidized by CO2 at 560 °C and then etched in HCl. In P1-2, AP-MWCNTs were oxidized by 60 mL·min−1 air at 500 °C and etched in HCl. In P1-3, AP-MWCNTs were refluxed in HCl with subsequent CO2 oxidation at 740 °C. In P2, AP-MWCNTs were ultrasonicated in ethanol for 10 h at 50 °C; they were then oxidized by CO2 for 4 h and refluxed in HCl for 10 h. P3 was designed to simplify the procedures of P2. In P3-1, AP-MWCNTs were refluxed in HCl (E) and oxidized by CO2 at 640 °C. In P3-2, AP-MWCNTs were ultrasonicated in HCl (E) and thermally oxidized in CO2 at 660 °C.

2.5 Characterization

Thermogravimetric analysis (TGA, STA449F3, NETZSCH) was used to determine the structural relationship among CIs, residual metal and MWCNTs. AP-MWCNTs were thermally oxidized at 30–1100 °C under 30 mL·min−1 CO2 and 30 mL·min−1 N2 (or 60 mL·min−1 air) at the constant heating rate of 5 °C·min−1 with/without a thermostatic during for 48 h (or 8 h) at 500 °C; The MWCNTs were examined by field-emission scanning electron microscopy (SEM, JSM7610F, JEOL Co. Ltd.) at 10 kV; Transmission electron microscopy (TEM, JEM-F200, JEOL Co. Ltd.) was used to observe the fine structure and flaws of the MWCNTs; in addition, the CIs content was evaluated at more than four random scanning positions to reduce the analytical error [25] and energy-dispersive spectroscopy (EDS) mapping was used to determine the elemental distributions; X-ray photoelectron spectroscopy (XPS, Thermo Scientific K-Alpha, Thermo Fisher) was used to determine the types of carbon in the MWCNTs; The changes in the crystallinity of the MWCNTs at 500 °C over time in air or CO2 were recorded by in-situ Raman spectroscopy (λ = 455 nm, DXR Raman Microscope, Thermo Fisher); Ex-situ Raman spectroscopy was also used to examine the crystallinity; The zeta potentials of the MWCNTs at 25 °C were determined (Malvern Zetasizer Nano ZS90, Malvern Panalytical Ltd.) after the MWCNTs were dispersed in a mixed solution of water and alcohol at a 1:1 volume ratio; Finally, the functional groups of the MWCNTs were analyzed qualitatively by Fourier transform infrared (FT-IR) spectroscopy (Spectrum Two, PerkinElmer).

3 Results and discussion

3.1 Entanglement of AP-MWCNTs

TGA (Fig. 2A) and DTG (Fig. 2B) results confirm that the CIs and MWCNTs in AP-MWCNTs cannot be oxidized at different temperatures. First, the temperature at which AP-MWCNTs mass loss began in CO2 (550 °C) was higher than that in air (475 °C), because CO2 is a weaker oxidizer [16]. Secondly, the residual mass ratios [25] were 9.95 ± 2.42 wt.% in CO2 and 10.47 ± 2.56 wt.% in air, with deviations that cannot be eliminated by either oxidant [12, 14]. Finally, two wide oxidation peaks appear in the CO2 derivative thermogravimetry (DTG) curve. The peak at 779 °C represents the oxidation of AC and some CNTs, and that at 978 °C is attributed to the oxidation of MWCNTs and GFs, owing to the common oxidation sequence of the AC, single-walled CNTs, MWCNTs, and GFs [26, 27]. These results are consistent with those of the standard carbon samples (Fig.S1). By contrast, the air-DTG curve shows only a single wide oxidation peak at 450–720 °C, which can be also divide into two peaks at 613 °C and 644 °C, representing the oxidation of AC/CNTs and MWCNTs/GFs, respectively.

Fig. 2
figure 2

Characteristics of AP-MWCNTs: A TGA results, B DTG results, C TGA at 30 mL·min−1 CO2 and 30 mL·min−1 N2, 5 °C·min−1, 30–1100 °C followed by 48 h at 500 °C, D In-situ Raman results with temperature held at 500 °C for 500 min in CO2, E SEM and TEM images, F EDS mapping images

Full size image

In addition, the results of thermostatic TGA and in-situ Raman spectroscopy also show that the CIs and CNTs cannot be oxidized separately completely. Although approximately 80 wt.% of the AP-MWCNTs has been oxidized when the temperature was held at 500 °C for 48 h (in CO2, Fig. 2C), no separate stage is observed. Moreover, the in-situ Raman results (Fig. 2D) also show that the oxidation behavior of the AP-MWCNTs in CO2 or in air did not differ greatly excepting that the oxidation rate was relatively low in CO2. The D-band peak at 1335–1345 cm−1 is ascribed to morphological defects in the CNTs caused by the symmetry breakdown of carbon atoms close to the edges of graphite sheets [28]. The G band at 1590 cm−1 is assigned to the sp2-hybridized carbon networks in ordered graphite [29]. Regardless of the oxidizer, the area ratio of these two bands (IG/ID) is almost the same, 1.1.

SEM, TEM and EDS-mapping images confirm that the CIs and MWCNTs in AP-MWCNTs are highly entangled and interwoven. Compared to fresh catalyst (Fig. 5A), SEM images (Fig. 2E) of AP-MWCNTs represent that a large amount of carbon nanofibers accumulate on the catalyst, and few CIs cover on their skeletons. TEM images (Fig. 2E) demonstrate that they are MWCNTs with a wide estimated diameter range of 5–60 nm (Fig.S1), resulting in a wide oxidation temperature range [26, 27]. Furthermore, the corresponding EDS mapping images (Fig. 2F) indicate that the aggregations in the AP-MWCNTs are the combination of carbon (AC, CNTs, or GFs) and residual metal (Fe, Ni), but residual metal also appears inside the MWCNTs.

The results above confirm that the oxidation temperatures of different carbon allotropes cannot be clearly distinguished dynamically, although their CO2 oxidation thermodynamics differ (Fig. 2B). The rationale is the very poor gas–solid diffusion owing to the interweaving of CIs and CNTs, which hinders gas diffusion to the CIs. It is common that AP-MWCNTs synthesized by catalytic chemical vapor decomposition (CCVD) have a state of entanglement and interweaving among CIs, residual metals and CNTs [30]. Thus, overcoming the limited mass transfer is a crucial task in AP-MWCNTs purification.

3.2 Dispersion of AP-MWCNTs

Methanol, ethanol, and isopropanol were used as the dispersants to separate CIs, residual metal, and CNTs in the AP-MWCNTs prior to CO2 oxidation, with DI water serving as a control. The temperatures at which rapid weight loss occurs (Fig. 3A and Table 1) in the treated AP-MWCNTs increase to approximately 790–837 °C from only 534 °C. This confirms the separation of residual metal from carbon and the removal of CIs from the CNT skeleton [8, 31], which greatly decreases the mass transfer resistance of CO2 in oxidation. Such effective purification of CNTs by ultrasonic operation is in agreement with the results reported, removing catalysts and graphitic particles adhered to single-walled CNTs [32], and separating the amorphous carbon enwrapped inside CNTs [33]. However, the multi-walled structure of AP-MWCNTs increases the difficulty of removing impurities and requires higher ultrasonic intensity and longer treatment time, which could lead to a significant increase in the damaging risk of CIs structure [34]. Thus, a suitable solvent condition for ultrasonic treatment of AP-MWCNTs is important.

Fig. 3
figure 3

AP-MWCNTs after ultrasonication for 10 h at 50 °C: results of (A) TGA and (B) DTG in CO2

Full size image
Table 1 TGA and DTG results summary of initiation, oxidation temperature after different treatment
Full size table

A new peak appears at 639 °C in the DTG curves of AP-MWCNTs after ultrasonication (Fig. 3B). This heat release should originate from the oxidation of AC released from the CNT skeleton by the combined action of ultrasonication and each dispersant. However, when water was used as the dispersant, the linkage between AC and CNTs was not broken owing to the hydrophobicity of AP-MWCNTs [35]. Thus, water cannot easily enter the pores in the primary agglomerates in the AP-MWCNTs and come into closer contact with carbon to separate the AC from the CNTs under ultrasonication [36].

The second oxidation peak, which is attributed mainly to MWCNT oxidation, increased by approximately 60 °C compared with that of the untreated AP-MWCNTs, indicating a sharp increase in the purity of the MWCNTs (Table 1). The reason is that ultrasonication detaches AC and residual metal from the MWCNT skeletons [25], causing this DTG peak to become sharper owing to the relatively pure carbon content [26, 27]. The third peak, which is ascribed to the mixture of MWCNTs and GFs, remains the same in all cases, suggesting that ultrasonication cannot separate them.

Thus, it can be concluded that ultrasonication in alcohols can effectively separate CIs and residual metal from MWCNTs, improving the effectiveness of the oxidation process for AP-MWCNT purification. Alcohols are necessary during ultrasonication as a dispersant to decrease the hydrophobicity of MWCNTs. In the experiments described below, ethanol is used as the dispersant.

The effect of operating temperature and duration of ultrasonication in ethanol was investigated, as shown in Fig. 4A. After ultrasonication at 50 °C, the temperature at which oxidation began increased to 800–838 °C for most MWCNT samples, except for that after 5 h of treatment. However, three separate DTG peaks were observed for all samples regardless of the duration of ultrasonication (Fig. 4B and Table 1). The mass loss of the sample treated for 5 h in the low-temperature zone (548–700 °C) was 15.5 wt.%, which is much larger than those of the other samples (approximately 3.0 wt.%), and its oxidation peak was 25 °C higher. This first peak is attributed to a mixture of AC and MWCNTs with small diameters or many defects [24, 25] because CIs cannot be adequately separated by a short period of ultrasonication. Extending the ultrasonication time facilitates the complete separation of CIs from MWCNTs and thus results in sharp discrete peaks.

Fig. 4
figure 4

TGA and DTG results of AP-MWCNTs in CO2 after ultrasonication in ethanol for 5–25 h at (A, B) 50 °C, (C, D) 70 °C, (E, F) 60 °C

Full size image

When the ultrasonication temperature is increased to 70 °C, MWCNTs are destroyed regardless of the duration. Samples ultrasonicated for different durations all begin to exhibit rapid mass loss at approximately 600 °C (Fig. 4C), and the DTG peak of MWCNTs disappears (Fig. 4D and Table 1). This notable change is ascribed to the strong destructive effect of ultrasonication at high temperatures, which breaks the MWCNTs and generates many highly defective MWCNTs and even carbon fragments (see the TEM images in Fig.S2), which can be oxidized at lower temperatures. It can be concluded that the ultrasonication of MWCNT at 70 °C is highly destructive.

Therefore, an intermediate temperature of 60 °C was adopted; the TGA results in Fig. 4E show an intervening regularity partly similar to both cases. The results obtained at 50 °C and 70 °C indicate that a short treatment did not fully remove the initial CIs, but high temperature and long duration damaged the MWCNTs. A treatment time between 10 and 15 h is thus considered suitable for removing CIs without destroying the MWCNT skeleton at 60 °C. The DTG results in Fig. 4F and Table 1 confirm this result, because only two main isolated peaks, at 827 °C for the MWCNTs and at 934 °C for GFs, appear after 10 and 15 h of ultrasonication.

Thus, it can be concluded that the treatment temperature is a decisive factor affecting MWCNT damage, and a long treatment time is suitable for CI separation. After AP-MWCNTs were treated at 50 °C for more than 10 h or at 60 °C for 10 to 15 h, the desired separation of CIs and MWCNTs can be achieved, in preparation for oxidation in CO2 to obtain qualified MWCNTs.

The structure of AP-MWCNTs before and after ultrasonication was observed by SEM. As shown in Fig. 5A–C, many CNTs grew densely on the catalyst surface, and entangled CNTs completely covered the surface 37 After 10 h of ultrasonication at 50 °C, the compact entanglement was maintained, and more CIs (yellow circles) appeared on the CNT surface (Fig. 5D). However, when the ultrasonication time was extended to 25 h, the MWCNT aggregations were looser, and fewer exposed CIs adhered to the surface, suggesting that impurities were successfully separated (Fig. 5E). When the temperature was increased to 60 °C (Fig. 5F and Fig. 5G) or 70 °C (Fig. 5H and Fig. 5I) the CNTs were partially or completely stripped from the aggregations. However, no destruction of MWCNTs is observed owing to the limitations of SEM. The SEM observations visually indicate that ultrasonication loosened the AP-MWCNT aggregations and exposed CIs.

Fig. 5
figure 5

SEM images of (A) fresh catalyst, (B, C) AP-MWCNTs, and AP-MWCNTs after ultrasonication at (D) 50 °C for 10 h, (E) 50 °C for 25 h, (F) 60 °C for 10 h, (G) 60 °C for 25 h, (H) 70 °C for 10 h, and (H) 70 °C for 25 h

Full size image

Qualitative analysis by FT-IR spectroscopy (Fig. 6A) shows that ultrasonication had no notable effect on AP-MWCNTs, excluding sp3 carbon [38]. For the sample ultrasonicated for 25 h at 50 °C, the peak at 1386 cm−1, which is attributed to the bending vibration of –CH3 [2], is enhanced, and that at 1397 cm−1, which is attributed to –C–H– deformation [39], is weakened, indicating that the sp3 carbon content increased and the sp2 carbon content decreased. Moreover, the C–OH peak at 1050 cm−1 and C–O peak at 1160 cm−1 are weak, and the –COOH peak at 1740 cm−1 is negligible [16], demonstrating that ultrasonication had little effect on the generation of oxygen-containing groups.

Fig. 6
figure 6

A FT-IR spectra, (B) C 1 s XPS spectra, and (C, D) zeta potentials of AP-MWCNTs before and after ultrasonication

Full size image

The XPS spectrum also confirms that few MWCNTs were destroyed by ultrasonication. As shown in Fig. 6B, the C 1 s peak can be deconvoluted into three peaks at 284.8, 285.5, and 286.5–288.0 eV, which are assigned to sp2 carbons, sp3 carbons, and C–O/C = O, respectively [38]. A perfect CNT skeleton contains only sp2 carbon. However, after ultrasonication in ethanol at 50 °C for 25 h, the sp3 carbon content increased from 13.8% to 16.3%, and the C–O/C = O content decreased slightly, from 8.9% to 8.8%. By contrast, the sp3 carbon content increased to 16.3% (60 °C for 25 h) and 18.3% (70 °C for 25 h), and C–O/C = O increased sharply to 12.6% (60 °C for 25 h) and 16.7% (70 °C for 25 h). These results confirm that ultrasonication strongly affected the conversion of sp2 carbon to sp3 carbon.

The Raman spectrum (Fig.S3) shows that R (IG/ID) increases from 1.26 (AP-MWCNTs) to 1.36 (after ultrasonication at 50 °C for 25 h), 1.35 (after ultrasonication at 60 °C for 25 h), and 1.46 (after ultrasonication at 70 °C for 25 h). The enhanced graphite degree indicates the removal of CIs.

The zeta potentials show that surface electrification breaks the van der Waals force between CIs and CNTs. As shown in Fig. 6C, the zeta potential decreased from 5.52 ± 0.09 mV (AP-MWCNTs) to − 3.62 ± 0.10 mV (in methanol), − 6.32 ± 0.38 mV (in ethanol),  − 4.12 ± 0.24 mV (in isopropanol), and − 4.26 ± 0.35 mV (in water) after ultrasonication at 50 °C for 10 h. When the AP-MWCNTs were dispersed in ethanol, the zeta potential decreased to approximately − 6.0 mV after ultrasonication at 50 °C, but it increased to 9.0 mV after ultrasonication at 70 °C (Fig. 6D).

The level of the zeta potential is not only related to the amount of surface charge but also to the distribution position of surface charge [14]. The zeta potential is positive at 70 °C because the positive and negative charges are distributed in the vacancy position, resulting in more positive C exposed than negative O, so the overall external positive charge. The surface damage at 60 °C is less than that at 70 °C, but there is also damage leading to external charge instability. At 50 °C, the surface is basically not damaged, so that C exists inside and O is outside, resulting in an external negative charge.

These changes are attributed to less ionization of –COOH or –OH groups owing to the acceleration of the hydrolysis reaction at high temperatures [40]. In addition, during ultrasonication at 60 °C, the zeta potential first decreased and then increased, confirming that CIs were first removed to expose the MWCNT surface, and then the MWCNT skeleton was destroyed.

The above analysis indicates that the ultrasonication of AP-MWCNTs in ethanol can separate CIs from MWCNTs, but it damaged the MWCNT skeleton to some extent. The ultrasonication temperature is a decisive factor affecting the damage to MWCNTs, and a longer duration can worsen the damage but enhance the dispersion degree.

3.3 Selectivity of gas oxidation and liquid etching

After ultrasonic treatment, the residue CIs and metal impurities originated from catalysts remained in AP-MWCNTs should be further removed for quality improvement. For CIs removing, three possible gases, CO2, O2 and H2, are compared for determining the feasibility based on the selectivity simulation on different carbon allotropes at the thermodynamic level.

As shown in Fig. 7A, the conversion rate of graphite carbon (GC) by O2 always pertains 100% in the whole temperature range from 0 to 1200 ºC, indicating its excessive strong oxidizability without enough space for selective tuning, achieving CIs removal. However, H2 can completely reduce GC to methane below 200 ºC and partially reduce below 800 ºC, and CO2 can partially oxidize GC to CO below 1000 ºC and achieve complete oxidization after beyond this temperature point. This simulation confirms that CO2 is a relatively mild oxidizer and can be used partially removing CIs from AP-MWCNTs if the temperature raising sufficiently.

Fig. 7
figure 7

A Thermodynamic equilibrium of graphitic carbon oxidation in O2, CO2 and H2; B Thermodynamic equilibrium of carbon allotropes oxidation in CO2; C Derivative results of CO2 oxidation thermodynamic equilibrium of carbon allotropes

Full size image

Investigating the thermodynamic equilibrium conversion rate of four main carbon allotropes contained in AP-MWCNTs, including GC, C60, AC, and Cdiamond, reacting with CO2, a significant difference among four species can be observed (Fig. 7B and C). It can be determined that GC is the most difficult to be oxidized by CO2 with the highest starting temperature of the reaction. The peak positions of conversion rate after derivation of C60, AC, Cdiamond, and GC, are 484 ºC, 595 ºC, 684 ºC, and 713 ºC, respectively. This occurrence exhibits a thermodynamic feasibility by using CO2 as the oxidizer to selectively remove CIs since the least temperature gap of oxidization with CO2, between DC and GC, has reached 29 ºC. However, the actual removal effect of CIs by CO2 oxidization should be testified and evaluated due to the aggregation status of different carbon species in AP-MWCNTs which evokes strong mass transfer resistant and converts the process into kinetical controlling. Ultrasonic dispersion is necessary to expose CIs, reducing the negative effect of aggregation and promoting the purification according to thermodynamic.

The removal of metal impurities of MWCNTs usually is acid etching [1]. Four etching liquid is used for achieving this purpose, including HCl, HNO3, HCl + HNO3, and HCl + H2O2. The TEM images of the acid-treated samples shown in Fig. 8 confirms the different removing the effect of the etching solution, including the metal particle residual amount and structure destruction degree of MWCNTs. The residual metal of the sample treated by HCl is the most, but no observable damage of MWCNTs is detected, which is ascribed to the non-oxidizability of HCl [41, 42]. However, HNO3 can effectively eliminate residual metal, but severely damages the main structure of MWCNTs simultaneously due to strong oxidizability [43]. Using a hybrid solution of HCl and HNO3, the metal remove the effect and MWCNT structure damaging is intermediate between the former two. Some metal embedded in MWCNTs and CIs on tube walls fail to remove although the perfection of MWCNTs structure is basically maintained in treatment. The effect of HCl + H2O2 is similar to HCl + HNO3 [44].

Fig. 8
figure 8

TEM images of AP-MWCNTs treated in acid: A HCl; B HNO3; C HCl + HNO3; D HCl + H2O2

Full size image

Therefore, the post-treatment procedures for purifying AP-MWCNTs after ultrasonic treatment are determined to be gas oxidation by CO2 for eliminating CIs due to its mild oxidizability and usable selectivity, and liquid etching by HCl for dissolving metal impurities due to its least damage effect of MWCNTs structure.

3.4 Mild purification process

In the P1 process, ultrasonic dispersion in ethanol is omitted, and oxidation in CO2 or air and acidification with HCl are applied. As shown in Table 2, the total MWCNT yields of all three processes are relatively low: 65.74 wt.% for P1-1, 60.86 wt.% for P1-2, and 76.49 wt.% for P1-3. The metal content after treatment is high: 2.90 ± 0.68 wt.% for P1-1, 2.61 ± 0.67 wt.% for P1-2, and 1.93 ± 1.45 wt.% for P1-3.

Table 2 Yields of each purification process
Full size table

TEM images confirm that the P1 processes (P1-1, P1-2, and P1-3) cannot simultaneously ensure the structural integrity of the MWCNTs at a high yield and a high CI removal rate. The P1-1 process caused little damage, because CO2 is a weak oxidant, and HCl is a non-oxidative acid, but many CIs remained interwoven with MWCNTs (Fig. 9A and Fig.S7A). When air was used instead of CO2 (P1-2), the damage to MWCNTs was more severe, although most of the CIs were removed (Fig.S9B and Fig.S7B) owing to the strong oxidation capacity of air in the first step, and this process had the lowest yield, 71.22 wt.%. When the order of the two steps of P1-1 was reversed, in P1-3, the obtained MWCNTs exhibited greater structural integrity than those obtained using P1-2, but less structural integrity than those obtained using P1-1. The CI removal rate (Fig.S9C and Fig.S7C) was also intermediate between those of P1-1 and P1-2.

Fig. 9
figure 9

Products TEM images of A P1-1, B P1-2, C P1-3, D P2, E P3-1, and F P3-2

Full size image

The P2 process has an attractive yield and CI/residual metal removal rates. No weight loss occurred in the ultrasonic dispersion step, and 94.33% and 90.80% of the weight was retained during oxidation and etching; these values are much larger than those of the other processes. The total MWCNT yield was 85.65 wt.%, and only 1.34 ± 0.06 wt.% remained as residual metal (Tab.2). This significant improvement is ascribed to the excellent dispersion effect of ultrasonication, which increases the selectivity of CO2 oxidation for CIs, and the accessibility to HCl of residual metal inside MWCNTs or CIs. The TEM images in Fig. 9D and Fig.S8 show that the MWCNTs are highly dispersed and show the least damage and lowest CI and residual metal contents, indicating that the three-step process exhibits excellent purification performance.

P3 did not perform as well as P2 but outperformed any P1 process. The total MWCNT yield of P3-1 was 86.15 wt.%, and that of P3-2 was 88.62 wt.%; the residual metal contents were 1.72 ± 0.22 wt.% and 5.77 ± 0.20 wt.%, respectively (Table 2). The total yield in both cases was slightly higher than that of P2, but the MWCNTs were significantly more degraded (Fig. 9E, F, and Fig.S9). The MWCNTs purified by P3-1 had excessive defects, and those purified by P3-2 contained many CIs inside MWCNTs.

The Raman results in Fig. 10A and Fig.S10 show that all the R (= IG/ID) values are higher than that of the AP-MWCNTs (1.26), except for that of the MWCNTs purified by P3-1, which suggests that their crystallinity and graphitization degrees were higher after purification. The R value of each step varies depending on the trade-off between CI removal and MWCNT damage, but the overall improvement indicates effective purification. Typically, the R value of the P2 process increases from 1.26 to 1.34 after ultrasonication owing to the removal of CIs, but it decreases slightly (to 1.33) after CO2 oxidation and decreases further (to 1.29) after refluxing in HCl owing to the mildly damaging effect of the latter two steps.

Fig. 10
figure 10

A Raman results obtained after each step of various processes, B XPS results of CIs

Full size image

Furthermore, the XPS results (Fig. 10B and Fig.S11) confirm that the C–O/C = O content of MWCNTS purified by P2 is lower (16.9%) than that of MWCNTs purified by the other processes, although it is higher than that of pristine AP-MWCNTs (8.9%). The MWCNTs purified by P1-2 have the highest C–O/C = O content (21.6%) because of the strong oxidation capacity of air. The next-highest C–O/C = O contents were found in the MWCNTs purified by P3-1 (17.7%) and P3-2 (18.7%) owing to the combined effects of ethanol and HCl. It is interesting that the sp3 carbon content of MWCNTs purified by P2 decreased slightly (by 0.8%), whereas those of MWCNTs purified by the other processes increased (to 14.9% for P1-1, 14.9% for P1-2, 15.9% for P1-3, 18.9% for P3-1, and 18.5% of P3-2, from 13.8% for pristine AP-MWCNTs). The results of the P2 process demonstrate that ultrasonic dispersion not only increases the selectivity of CO2 oxidation for CIs but also prevents the transformation of sp2 carbon by non-oxidative destruction.

In summary, the most promising process is P2, which affords the lowest residual metal content, the most well-formed MWCNT structure with the fewest CIs, and a relatively high yield. As shown in Fig. 11, the dispersion operation in the first step is essential for separating CIs and even residual metal from the MWCNTs, exposing them and thus improving the effect of mild CO2 oxidation and HCl non-oxidative etching. Consequently, the total yield of MWCNTs obtained by the CDM was 85.65 wt.%, and 87.20% of the residual metal was removed. The MWCNT yield and residual metal removal rate are 19.91% and 15.68% higher, respectively, than those obtained without the dispersion step (P1-1). This yield is also higher than those reported in the literature [1, 14] (79.2% for H2, air oxidation, and HCl etching; 69.62% for air oxidation and HNO3 and HCl etching). Most of the CIs were removed, and the MWCNT skeletons were preserved intact without notable damage. This mild three-step purification process for MWCNTs produced by CDM shows promise for large-scale industrial applications.

Fig. 11
figure 11

Flow diagram of optimal process

Full size image

4 Conclusion

This work proposed a mild purification process based on the gas–solid and liquid–solid reaction dynamics for raw MWCNTs produced by CDM; this process exhibited increased selectivity of CO2 gas phase oxidation, HCl liquid phase etching and a higher MWCNT yield.

A novel ultrasonic pretreatment method in ethanol was devised to separate and expose CIs and residual metal from MWCNTs, particularly those embedded within graphitic shells. This process effectively relieves mass transfer resistance encountered in the purification of AP-MWCNTs. The damage to MWCNTs depended on the operating temperature and duration of ultrasonication. The least damage to the MWCNT skeletons occurs, and only 2.5% of the sp2 carbon is converted to sp3 carbon, during 25 h of ultrasonic dispersion at 50 °C. CO2 was used as a mild oxidant to remove CIs by the stepwise oxidation of carbon allotropes according to the thermodynamics, and HCl was used as an etching agent to remove residual metal because it causes the least damage to the MWCNT skeleton.

By combining the three operations into a cascade process, the purification of AP-MWCNTs was achieved, with a final one-way MWCNT yield of 85.65 wt.% and remaining residual metal content of 1.34 ± 0.06 wt.%. The MWCNT yield and residual metal removal rate were 19.91% and 15.68% higher, respectively, than those obtained without ultrasonic dispersion. Moreover, this process prevents the conversion of sp2 carbons by non-oxidative destruction, and the C–O/C = O content increased by only 8%.

This work describes an efficient method of disentangling raw MWCNTs; this mild purification process shows promise for application in qualified MWCNT production.