Algorithm-improved high speed and non-invasive confocal
Raman imaging of two-dimensional materials
Sachin Nair
†, Jun Gao
†, Qirong Yao
§, Michael H.G. Duits
†, Cees Otto
‡, Frieder Mugele
††
Physics of Complex Fluids, University of Twente,
‡Medical Cell BioPhysics, University of Twente,
§Physics of Interfaces and Nanomaterials, University of Twente.
EMCCD m n wave number m× n pixels m× n pixels PC2 PC3 PC4 PC5 PC6 PC1 PC8 PC7 PC1570 PC1571 ... reconstruct decompose ... ... wave number ... ... discard (a) (b) (c) (d) (e) (f) 590 585 580 1000 v (cm-1) 2000 v (cm-1) 580 600 620 640 1000 2000
normal CRM (without denoising) 12.5 µW/µm2, 20 ms
ai-CRM (with denoising)
Spectro-meter D G I (a.u.) I (a.u.) 1L 2L 3L AFM topography 10 ms, ai-CRM
(a)
(c)
500 ms, normal CRM(e)
0 4 nm 0 200a.u. 0 13000 a.u.(b)
integration time (ms) SNR 0 0 4 8 20 40 500 ai-CRM normal CRM (d) S (a) (b) (c) laser dose = 4.9 x 10 8 J/m 2 (d) I(G)/I 0 (G) 0.2 0.4 1.0 0.8 0.6 0 time (s) 500 0 time (s) 500 (h) time time (s) first scan laser dose = 4.9 x 10 5 J/m 2second scan third scan
0 6500a.u. 0 6500a.u. 0 6500a.u. (e) (f) (g) 0 300a.u. 0 300a.u. 0 300a.u. 0 50 90 50 90 500 FWHM(G) (cm -1 ) 5 cm-1 0.2 0.4 1.0 0.8 0.6 time time (s) 0 500 5 cm-1 S S I(G)/I 0 (G) FWHM(G) (cm -1 ) 50 µm 10 µ se cills 02 , m s 1000 ν (cm-1) 2000 2600 ν (cm-1) 3200 (a) (b) rGO PAA 50 µm 6000 0 a.u. 100 ν (cm )-1 1000 0 250a.u. 100 -1 1000 0 160a.u. ν (cm ) 4000 0 a.u. 3000 0 a.u. 100 -1 1000 0 70 a.u. ν (cm ) 50 ms, 50 µW/µm2, ai-CRM
(a) (b) 50 ms, 50 µW/µm2, ai-CRM (c) 50 ms, 1.25 mW/µm2, ai-CRM
1L
MoS2 WS2
BN
Motivation
• Confocal Raman Microscopy (CRM) is an imoportant tool for characterizing 2D materials, but its low throughput significantly hinders its applications. • For metastable materials like graphene oxide (GO), this problem is aggravated by the requirement of very low laser power dose per unit time to avoid laser induced reduction.
• Reduction of the laser power and exposure time often results in insuffiicient signal to noise ratio (SNR).
Fast mapping of GO
• ai-CRM falicitates fast mappingof GO (10 ms/pixel), at a low laser power of 5 µW.
• GO can be imaged at a laser dose 2 to 3 orders of tude lower than previously reported.
Reliable mapping of GO
• Previous reports suggest that GO is reduced even at a low laser power of 48 µW/s. • ai-CRM can falicitate GO mapping
at a laser power of 4 µW and integration time of only 20 ms.
• The analysis shows that the laser induced damage can be minimized to a much greater extent.
ai-CRM method
• The principle of ai-CRM is based on SNR improve-ment by using an EMCCD camera and principal
com-ponent analysis (PCA) guided data denoising.
Fast volumetric mapping
• Fast volumetric mapping of
a reduced GO-PAA composite. • Laser power of 0.75 mW and integration time of 1 ms!
• a 3D reconstruction of resolution
100 x 100 x 20 pixels took 12 mins.
Fast mapping of other
2D materials
•ai-CRM can also be applied to other 2D materials like MoS2, WS2 and BN. • The tehcnique can be used for GO and graphene , on arbitrary substrates (not shown).Conclusions
• We introduce algorithm-improved Confocal Raman Microscopy (ai-CRM), which increases the Raman scan-ning rate by one to two orders of magnitude with respect to state-of-the-art works for a variety of 2D materials. • GO can be imaged at a laser dose that is 2 to 3 orders of magnitude lower than previously reported, such that laser-induced variations of the material properties can be avoided.
•Since ai-CRM is based on general mathematical principles, it is cost-effective, facile-to-implement and univer-sally applicable to other hyperspectral imaging methods.
