High temperature annealing leads to graphitization and will improve properties of your CNT.. but I am skeptical about removal of metal impurities by heating CNTs and in your case temperature is pretty low too,  600-700C.. 

Acid treatment is best way to remove metal impurities from CNT..  Though it is know that acid treatment will damage CNTs, but its the best and simplest method.

if your impurities are magnetic you can use magnetic filtration technique

3M Nitric acid reflux for few hours could avoid the damge to CNT but still would remove most of the carbon impurites.

I would not refer to what you have in the shoulder as 'carbon nanoparticles'. For sure those are not metal impurities (as said in some previous answers, since those will not burn at 600 oC.

Most likely what you have is some form of carbon as an impurity byproduct during synthesis. Since it burns at 600 oC I would expect to be some carbon not entirely graphitized. Of course high temperature annealing will make the peak disappear, but because the impurities will graphitize so they will appear at higher temperature.

Have you checked literature about purification methods for CNTs? Maybe centrifugation steps will remove some of this (at the cost of losing some CNT too) instead of only transforming the impurity.

Thanks for everybody's answer!

Fernando, may I ask that based on what you don't think they are not carbon nanoparticles? In my working place, there is no high temperature oven. I am not able to use high temperature annealing proposed by Sridhar Vadahanambi. But I found that I can remove the shoulder after putting my sample into a 200 oC oven for 48-72 hous.

Check Datsyuk paper in Carbon 2008, vol. 46, pp. 833-840, where he studied different oxidation agents.

His conclusion:

Use of non-acidic treatments

such as the mixture of NH4OH and H2O2, facilitates the complete

removal of disordered carbon from the MWCNT specimens

I think you should use acidic treatment. It will damage the MWCNTs. But these defects are useful for further processes like functionalization, since the MWCNTs will gain the reactivity due to the defects.

All the best

Nanotubes were probably synthesised at a temperature above 200 oC. I do not see why the shoulder would disappear after heating just 200 oC.

Besides, the sample during TG is being heated already at 200 oC and more, and the shoulder does not disappear (nothing happens in your TG plot at 200 oC).

Anyway, acid treatment may help (as said above) or treatment with a solvent. Also, as Farooq above said, I remember that I could remove that peak from single wall carbon nanohorns only by mixing the sample with n-hexane and sonicating at room temperature.

I might suggest to follow the purification protocol introduced by Clascídia A. Furtado more than 10 years ago in J. Am. Chem. Soc., 2004, 126 (19), pp 6095–6105

DOI: 10.1021/ja039588a. The Authors went to great lenghts to study the best way to purify carbon nanotubes soots from amorphous carbon and metallic catalysts. They characterized all the purification steps by Raman, AFM, TEM and they included TGA measurements. In particular, they found that heating a carbon nanotubes sample in air (dry oxidation) results in a faster oxidation of amorphous carbon than carbon nanotubes and thus allowing the selective oxidation of amorphous carbon soot. The conditions of dry oxidation depend on each carbon soot batch but are usually in the range 300-400 °C for 60-90 minutes in air.

As reported in the abstract: "The SWNTs studied were produced by the arc-discharge process. Dry oxidation was used in an initial step to remove amorphous carbon. Subsequently, two acid purification schemes were investigated (HCl- and HNO3-reflux) to remove the metal growth catalyst (Ni-Y). Finally, ultrasonic dispersion of isolated tubes into either N,N-dimethylformamide (DMF) or N-methyl-2-pyrrolidone (NMP) was carried out. Raman scattering, atomic force microscopy (AFM), and electron microscopy were used to study the evolution of the products. Raman scattering was used to probe possible wall damage during the chemical processing. We found that both HCl and HNO3 could be used to successfully remove the Ni-Y below ∼1 wt %. However, the HNO3-reflux produced significant wall damage (that could be reversed by vacuum annealing at 1000 °C). In the dispersion step, both amide solvents (DMF and NMP) produced a high degree of isolated tubes in the final product, and no damage during this dispersion step was observed."

Hi, Patrizio! Thanks for your kind help! That article is of great help!

 I agree with Debmalya Roy ·

Hi, Debmalya Roy! I also agree with your explanation! 

One of method will be sulfidative purification of carbonaceous impurites on CNTs.

See the paper entitled “Sulfidative Purification of Carbon Nanotubes Integrated in Transistors”, J. Am. Chem. Soc. 2005, 127, 8300.

You can consult with the following references in which a separation solution for carbon nanotubes possible at an industrial scale was discussed:

-C.A. Howard et al, J. Am. Chem. Soc., 2004, 126,13228

-S. Fogden et al, ACS Nano, 2012, DOI:10.1021/nn2041494

Try to do an isothermal TGA at 200oC. Hold the samples at 200oC in the TGA instead of the oven. If Dr. Roy is right and there is combustion of carbon nanoparticles then the TGA should show weight loss over the long (48-72 hr) time period.

Also, for acid treatment, I would suggest the use of HCl for removing metal impurities as it is a non-oxidative acid. The damge to the MWNTs will be minimal even when using concentrated HCl as compared to using HNO3 and/or H2SO4.

Another way to enhance the effect of HCl treatment would be to heat the samples in air to 200-300oC prior to putting it in HCl. 

Sometimes ago I did this experiment with good results, which is a combination of several methods. First in a CVD reactor,  I put CNT at 400 to 420 °C under air for 30 min (Thermal oxidation) . After that, air was replaced by Ar, and temperature increased until 750 to 800 °C and kept for 30 min (Annealing). After that TGA showed no amorphous carbon and XRD and Raman also showed an improvement in their structure (Crystallinity).

Thanks a lot for your answer, Jose!