▎ 摘　　要

The Raman spectrum of epitaxial graphene on SiC is generally obtained by simply subtracting a SiC spectra from the experimental data, which results in noisy spectrum and negative intensity. By using a Non-negative Matrix Factorization (NMF) method, we obtain pure graphene spectra, even for monolayer graphene and sub-micron size patterned features, as well as in spatial mapping and depth profile. We show that the NMF method is efficient in data smoothing and for signal deconvolution with no assumption required for the functional form of the signals. (C) 2013 AIP Publishing LLC.