Claire
Blaga
a,
Ángel
Labordet Álvarez
b,
Akshay
Balgarkashi
a,
Mitali
Banerjee
c,
Anna
Fontcuberta i Morral
ad and
Mirjana
Dimitrievska
*ab
aLaboratory of Semiconductor Materials, Institute of Materials, School of Engineering, École Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland
bNanomaterials Spectroscopy and Imaging Group, Transport at Nanoscale Interfaces Laboratory, Swiss Federal Laboratories for Material Science and Technology (EMPA), Ueberlandstrasse 129, 8600 Duebendorf, Switzerland. E-mail: mirjana.dimitrievska@empa.ch
cLaboratory of Quantum Physics, Topology and Correlations, Institute of Physics, School of Basic Sciences, Ecole Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland
dInstitute of Physics, School of Basic Sciences, Ecole Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland
First published on 30th July 2024
Tungsten diselenide (WSe2) is a 2D semiconducting material, promising for novel optoelectronic and phononic applications. WSe2 has complex lattice dynamics and phonon structure. Numerous discrepancies in the literature exist regarding the interpretation and identification of phonon modes. This work presents a complete investigation of the vibrational properties of 1L to 5L flakes and bulk WSe2 using multi-wavelength Raman spectroscopy. We especially highlight measurements using 785 nm excitation, which have not been performed before. These allow us to solve inconsistences in the literature in terms of defect-activated non-Γ point single phonon modes and Breit–Wigner–Fano type resonance. We identify 35 Raman peaks per flake thickness, which we attribute to either one-phonon or multi-phonon modes, including two-phonon scattering due to a van Hove singularity (vHs). The measurements are in excellent agreement with the theoretical predictions. Using photoluminescence measurements, we identify photon-exciton coupling leading to resonant Raman scattering, suggesting wavelength laser excitations best suited for further investigations of specific WSe2 flake thicknesses. Finally, we report the observation of phonon-cascades for all WSe2 flake thicknesses, indicating strong phonon–electron interactions during early carrier relaxation processes in WSe2. This research provides a solid foundation and reference for future investigations of the vibrational properties of WSe2, paving the way for further development of this material towards applications.
Understanding the vibrational (phononic) properties of thin layer TMDs is crucial for applications requiring the engineering of thermal conductivity.6–8 It is also essential for leveraging the interlayer charge transfer dynamics in van der Waals heterostructures,9,10 particularly for optoelectronic and photovoltaic applications, and the investigation of Moiré phonons in twisted layers of TMDs for novel phononic devices.11
Raman spectroscopy is a routinely used technique for the characterization of 2D materials, as it provides information on the structure,12,13 defect type and density,14,15 and strain of the lattice.16–18 It also gives access to a variety of electronic parameters such as the charge carrier density,19,20 maximum charge carrier mobility,17,21 phonon–electron coupling and electron–electron scattering.22,23 Considering the plethora of information that Raman spectroscopy can provide, a precise comprehension of the Raman spectra and the lattice dynamics is crucial for their further development and utilization in future applications.
Vibrational properties of few-layered WSe2 have been explored in several studies before, where most of the effort was made on the proper interpretation of Raman peaks of WSe2.24–32 Many of these studies show significant discrepancies in the analysis of the Raman spectra and assignment of phonon modes. For example, the interpretation of Raman peaks and assigned phonon modes vary across the literature, from 17 to 42 phonon modes observed.24–32 Interestingly, several studies report activation of non-Γ point single phonon modes in the Raman spectra of WSe2 flakes.30,33,34 These modes are usually not expected in the Raman spectrum of high crystal quality materials, as they require the presence of defects in order to be activated.35–37 Some studies report the observation of the Breit–Wigner–Fano (BWF) type resonance of the low intensity Raman peak at 302 cm−1 assigned to WSe2.30 This is very unusual, as theoretically calculated phonon frequencies do not predict the presence of any single or multi-phonon mode at this position. Finally, one finds different interpretations of the Raman peak at 260 cm−1, from the identification as a broad 2LA(M) mode, to the deconvolution into a dozen peaks assigned to one-phonon and multi-phonon modes.33,34,38 Such inconsistencies in reported results require clarification by a systematic review of published data and an in-depth investigation of vibrational properties of multilayer WSe2 systems.
In this work, we investigate vibrational properties of multilayer (1L to 5L) and bulk WSe2 using multiwavelength excitation Raman spectroscopy. Using a rigorous deconvolution process of the Raman spectra measured with 488, 532 and 785 nm laser excitations, we provide a complete analysis of all Raman active modes of WSe2 systems. To the best of our knowledge, this is the first comprehensive analysis of the Raman spectra of WSe2 measured at 785 nm, along with a concise comparison of theoretical and experimental results in terms of phonon symmetries and frequencies. We also shed new light on the inconsistencies reported in the literature regarding the activation of non-Γ point single phonon modes and BWF type resonance. Finally, we highlight the observation of the phonon cascades in 1L to 5L and bulk WSe2 as well as suggest the optimal laser excitations under which different numbers of layers could be further explored. These results can be used as a reference for future studies of vibrational properties of WSe2, paving the way for successful integration of WSe2 in phononic and optoelectronic devices.
Atomic force microscopy (AFM) measurements were performed on a Bruker FastScan AFM, using a ScanAsyst_Fluid+ tip. The AFM mapping is performed in tapping-mode to ensure that the flake does not incur any damage from contact with the tip.
The optical microscopy images were taken with a BX53M Olympus microscope, using a 6.4 megapixel DP23 Olympus digital color camera and an LMPLFLN100x objective.
In order to confirm the homogeneity and high crystal quality of the WSe2 flake, we have performed high-resolution Raman imaging using 532 nm laser excitation with a laser spot diameter of 1 μm, as shown in Fig. 2. It should be noted that before any Raman measurements, a laser power study was performed on the small area of the monolayer, which was used to determine the maximal laser power. A consequence of the power study is structural damage to the small area of the monolayer, which is noticeable by the dark spot in Fig. 2a. This area was excluded from any further analysis that was performed in this study.
Fig. 2b shows a representative Raman spectrum of bilayer (2L) WSe2, highlighting some of the most intense Raman peaks at 138, 250, 260, 309 and 375 cm−1 of WSe2, as well as the 520 cm−1 peak corresponding to the Si substrate. Variations in the Raman intensity of these peaks over the WSe2 flake are shown in the Raman mapping images in Fig. 2c. It can be observed that there is no notable change in the intensity of Raman peaks within each area of a single thickness, for all presented Raman maps. This signals structural uniformity of layers. Additionally, we can observe that the Raman intensity variation between different layers changes depending on which Raman peak is used for imaging. For example, Raman maps corresponding to the intensity of 250 and 260 cm−1 peaks show clear differences in the 1L and 2L areas. The difference between 3L, 4L and 5L is harder to distinguish. On the other hand, the Raman intensity maps corresponding to 309 and 375 cm−1 peaks show the difference among 2L, 3L, 4L and 5L better, while making the identification of 1L areas difficult. Indeed, in the case of 1L, the 375 cm−1 peak has very low intensity and the 309 cm−1 peak is not active. In contrast, Raman imagining with low frequency peaks, such as 138 cm−1, does not show any major differences in the areas corresponding to different numbers of layers. Interestingly, the best imaging of the differences in the number of layers is achieved with Raman mapping of the Si peak at 520 cm−1, enabling clear differentiation between 1L and 5L areas. We attribute this to the absorbance of the Si Raman emission by the layers, similarly to the work of Negri et al.39 This illustrates the potential of using Raman imaging as a complementary method for the identification of the number of layers, when AFM measurements are not possible. However, care should be taken when selecting the Raman peaks used for imaging, depending on the number of layers that need to be identified.
Γtotal = A1g + 2A2u + 2B2g + B1u + E1g + 2E1u + 2E2g + E2u. |
Raman and infra-red active modes are:
ΓRaman = A1g + E1g + 2E2g |
ΓIR = 2A2u + 2E1u |
In contrast to the bulk, multilayer WSe2 systems are considered quasi 2D, due to the lack of translation symmetry along the c-axis direction. Their space group symmetry is determined by the parity of the number of layers. An odd number of layers is classified within the D3h point group symmetry (without inversion symmetry), while an even number of layers belongs to the D3d point group symmetry (with a center of inversion). The labeling of phonon modes is different depending on the number of layers, and the irreducible representations at the Γ point of the Brillouin zone are:40–42
Therefore, strictly speaking, the notations of the modes for the multilayer WSe2 systems differ, with A1g and E2g modes for bulk, translating to and E′ modes for an odd number of layers and A1g and Eg modes for an even number of layers.
The one-phonon modes described above, arising from the Γ point, are a result of the first-order Raman scattering process. Higher-order Raman scattering processes are also possible. These involve the combination of multiple phonon modes from Γ and non-Γ points of the Brillouin zone for which the conservation of the momentum rule is preserved. This means combining two (or more) phonons from the Γ point which have zero momentum, or combining phonons from M and K points (or anywhere else in the Brillouin zone) whose total momentum is zero (for example pairs of phonons with equal and opposite momentum). Higher-order Raman scattering is less probable than the first-order, and therefore the combination modes in the Raman spectrum have lower intensities and higher widths of peaks when compared to the first-order Raman modes. These intensities can be significantly enhanced under Raman resonant conditions, when the excitation wavelength is close to the energy of some interband transition in the material. Therefore, for a complete interpretation of the Raman spectrum, it is important to consider the full phonon and energy band dispersion of the material, over the whole Brillouin zone.
The band structure and phonon dispersions of bulk and multilayer WSe2 systems have been extensively studied and reported in the literature.25,27–29,33,34,43–45 For convenience, Fig. S2 in the ESI† presents the density functional theory (DFT) calculated phonon dispersions of monolayer (1L) WSe2 adapted from ref. 33. Monolayer WSe2 has three atoms per unit cell, resulting in a total of nine phonons. The three acoustic phonons correspond to the out-of-plane (ZA) vibrations, in-plane transverse (TA) vibrations, and in-plane longitudinal (LA) vibrations. These phonons are located at 0 cm−1 at the Γ point and correspond to the lowest frequency phonon branches across the Brillouin zone in Fig. S2.† The E′′ and E′ phonon modes are located at 176 and 250 cm−1 at the Γ point, and as doubly degenerated modes, they are split into two optical branches each, over the Brillouin zone. The lower frequency phonon branches correspond to transverse (TO1 and TO2) vibrations, while the higher frequency phonon branches correspond to longitudinal (LO1 and LO2) vibrations of E′′ and E′ modes, respectively. The and
modes are located at 250 and 310 cm−1 at the Γ point and correspond to two out of plane (ZO1 and ZO2) vibrations, respectively. They are represented by the highest frequency phonon branches in Fig. S2.†
In comparison, the phonon dispersions of the multilayer WSe2 systems (2L, 3L and higher) slightly differ from those of the monolayer WSe2. The main difference is in the number of expected phonons, i.e. phonon branches, with the total number of phonons being calculated as 3N, where N is the number of atoms per unit cell. Therefore, bilayer (2L) WSe2, having 6 atoms per unit cell, will have a total of 18 phonon branches, while trilayer (3L) WSe2, with 9 atoms per unit cell, will have a total of 27 phonon branches, etc. The new phonon branches in the multilayer systems arise from the splitting of the monolayer phonon branches. Therefore, the frequencies of these new phonons are very close to the one of the original monolayer phonon branch. This is the main reason why the overall phonon dispersion of WSe2 looks very similar regardless of the number of layers.33 Another difference is the appearance of the low frequency shear and breathing modes, which arise due to the interlayer van der Waals (vdW) coupling between the layers. Their frequencies depend on the number of layers present.43
It should be noted that the DFT predicted frequencies of the phonon modes, especially at Γ, M and K points of the Brillouin zone, are very important and will be used for the identification of the combination phonon modes in the Raman spectra of WSe2 multilayer systems.
Deconvolution of the Raman spectra measured with 488, 532 and 785 nm excitation for each individual layer system of WSe2 was performed using the minimum number of Lorentzian components, allowing the identification of around 35 Raman peaks for each system. The representative deconvolution results corresponding to Raman spectra of 2L WSe2 measured at 488 nm, 3L WSe2 measured at 532 nm and 5L WSe2 measured at 785 nm are shown in Fig. 4a–c. Each peak was modeled with a Lorentzian curve characterized by peak position, peak width, and intensity. As the fitting procedure includes a large number of variables, rigorous restrictions were imposed on the fitting parameters in order to avoid correlation among the parameters and obtain meaningful results, as explained here below and in ref. 36,46–48. This includes leaving the intensity and peak position as free parameters, while the widths of peaks were restricted. As the peak widths are mostly dependent on the phonon lifetime, which is determined by the crystal quality of the material, it is expected that all fundamental one-phonon Raman modes have similar widths, regardless of the symmetry of the mode. This results in allowing only a narrow interval of change for the one-phonon peak widths during the deconvolution process. Possible two-phonon or multi-phonon modes are similarly modeled with higher widths compared to the one-phonon modes. This allows unambiguous interpretation of the phonon nature of the peaks, rendering the identification procedure more accurate.
![]() | ||
Fig. 4 Lorentzian deconvolution of the Raman spectra for (a) 2L WSe2 measured at 488 nm, (b) 3L WSe2 measured at 532 nm and (c) 5L WSe2 measured at 785 nm excitation wavelengths. |
The Raman frequencies of all peaks obtained from the deconvolution process are listed in Table 1, along with their symmetry assignment and a comparison with the frequency of the DFT calculated phonon modes and experimentally reported values from the literature. All peaks are identified as either one-phonon or multi-phonon modes based on the results from the deconvolution procedure. Overall, we observe excellent agreement (on-average 2% difference) between the experimentally observed peaks and the theoretically predicted Raman frequencies. Minor disagreement in the Raman peak positions between the experimental and the theoretical results is expected, due to approximations applied during the calculations, such as the three-body and long-range interactions.
This work | Literature | |||||||
---|---|---|---|---|---|---|---|---|
λ ext (nm) | ν exp (cm−1) | Symmetry assignment | ν theory (cm−1) | ν exp (cm−1) | ||||
1L | 2L | 3L | 5L | Bulk | ||||
a Labels: λext – laser excitation under which the Raman mode is present; νexp – Raman peak frequency experimentally determined from the Raman spectra; νtheory – theoretically calculated Raman mode frequency; “ – ” Raman peak not observed in the spectra; “/” – additional phonon mode assignment for the Raman peak with the same frequency. Fundamental one-phonon Raman modes arising from the Γ point are highlighted in bold. | ||||||||
488 | — | 28 | 19 | 13 | — |
A
1g
(even),
![]() |
27.7 (2L), 19.4 (3L), and 14.8 (5L)43 | 28 (2L), 19 (3L), and 12 (5L)43 |
488 | — | 17 | 21 | 22 | 24 | E g (even), E′ (odd), and E2g(bulk) | 17.8 (2L), 21.6 (3L), 23.5 (5L), and 24.6 (bulk)43 | 17 (2L), 21 (3L), 23 (5L), and 24 (bulk)43 |
488 | — | — | — | 32 | — |
![]() |
30.7 (5L)43 | 32 (5L)43 |
488 | 32 | 36 | 36 | 36 | 35 | LA(M) – TA(M) | 31.7 (ref. 33) | — |
488 | 75 | 75 | 75 | 75 | 75 | ZO2(M)–TO1(M) | 76.3 (ref. 33) | 74.4 (ref. 33) |
All | 97 | 97 | 97 | 97 | 97 | LO1(M)–ZA(M)/TO1(M)–TA(M) | 94.2/95.3 (ref. 33) | 96.0 (ref. 33) |
All | 115 | 114 | 114 | 111 | 111 | TO2(K)–TA(K)/ZO1(K)–LA(K)/LO2(M)–LA(M)/LO1(K)–TA(K)/LO2(K)–LA(K) | 114.5/114.9/115.1/116.3/116.5 (ref. 33) | 114.3 (ref. 33) |
All | 119 | 119 | 119 | 119 | 119 | ZO2(K)–LA(K) | 119.5 (ref. 33) | 120.1 (ref. 33) |
All | 128 | 128 | 128 | 127 | 127 | TO2(M)–TA(M) | 130.7 (ref. 33) | 130.5 (ref. 33) |
All | 138 | 138 | 138 | 138 | 138 | LO2(K)–ZA(K)/ZO1(Γ)–TO/LO1(Γ) | 135.2/136.2 (ref. 33) | 137.4 (ref. 33) |
488, 532 | 147 | 147 | 147 | 147 | 147 | LO2(M)–TA(M) | 146.8 (ref. 33) | 145.0 (ref. 33) |
All | 153 | 153 | 153 | 153 | 153 | ZO1(K)–TA(K) | 157.4 (ref. 33) | 158.4 (ref. 33) |
488, 532 | — | 176 | 175 | 175 | 176 | E g (even), E′ (odd), and E1g(bulk) | 176.3 (2L)27 | 176.3 (2L)27 |
488, 532 | 209 | 209 | 209 | 209 | 209 | 3- Or 4-phonon mode (example: TO2(M)–TA(M) + ZO2(M)–TO1(M)) | 209.6 (ref. 33) | 208.1 (ref. 33) |
All | 219 | 219 | 219 | 218 | 219 | TA(M) + ZA(M) | 220.4 (ref. 33) | 219.2 (ref. 33) |
532, 785 | 229 | 229 | 229 | 232 | 232 | TA(M) + LA(M) | 228.5 (ref. 33) | 231.2 (ref. 33) |
All | 240 | 240 | 240 | 241 | 241 | TA(K) + LA(K) | 242.3 (ref. 33) | 239.9 (ref. 33) |
All | 250 | 250 | 250 | 250 | 250 |
E
g
& A
1g
(even), E′ &
![]() |
250.8 & 251.3 (ref. 33) | 249.9 (ref. 33) |
All | 258 | 258 | 258 | 257 | 257 | 2vHs | 256 (ref. 33) | 258.7 (ref. 34) |
All | 261 | 261 | 261 | 261 | 261 | 2LA(M) | 260.2 (ref. 33) | 261.0 (ref. 33) |
488, 532 | 283 | 283 | 284 | 284 | 284 | 2LA(K) | 284.8 (ref. 33) | 274.1 (ref. 34) |
All | — | 309 | 309 | 309 | 309 |
A
1g
(even),
![]() |
311.3 (ref. 27) | 310 (ref. 27) |
488, 532 | 331 | 330 | 330 | 331 | 331 | TO1(K) + ZA(K) | 333.3 (ref. 33) | 331.3 (ref. 33) |
488, 532 | 340 | 341 | 341 | 341 | 341 | TO2(M) + ZA(M) | 339.9 (ref. 33) | 341.5 (ref. 33) |
All | 351 | 350 | 350 | 352 | 352 | 2![]() |
352.3 (ref. 33) | 351.2 (ref. 33) |
488, 532 | 359 | 359 | 359 | 359 | 359 | ZO1(M) + TA(M)/ZO2(K) + TA(K) | 361.7/361.8 (ref. 33) | 360.0 (ref. 33) |
All | 374 | 373 | 373 | 373 | 373 | LO2(M) + LA(M) | 373.3 (ref. 33) | 375.4 (ref. 33) |
All | 396 | 394 | 394 | 394 | 394 | ZO2(M) + ZA(M) | 392.0 (ref. 33) | 393.0 (ref. 33) |
532 | 444 | 444 | 444 | — | — | LO1(M) + TO2(M)![]() |
445.3 (ref. 33) | — |
488, 532 | 456 | 456 | 456 | 456 | 456 | TO1(M) + ZO1(M)/2TO2(M)/LO2(M) + TO2(M)![]() |
457.0/458.2/461.4 (ref. 33) | — |
488, 532 | 469 | 469 | 469 | 469 | 469 | TO1(K) + ZO2(K)/TO2(K) + ZO1(K) | 471.5/471.7 (ref. 33) | — |
All | 488 | 488 | 488 | 488 | 488 | LO1(Γ) + ZO1(Γ)/ZO1(M) + TO2(M) | 488.29/492.4 (ref. 33) | 491.3 (ref. 33) |
488, 532 | 496 | 496 | 496 | 496 | 496 | ZO2(M) + TO2(M) | 499.1 (ref. 33) | 499.2 (ref. 33) |
488, 532 | 510 | 511 | 511 | 512 | 512 | ZO1(M) + LO2(M)/2ZO1(K) | 508.5/514.6 (ref. 33) | 507.1/514.4 (ref. 33) |
All | 533 | 532 | 532 | 532 | 532 | ZO1(M) + ZO2(M) | 533.3 (ref. 33) | 533.8 (ref. 33) |
Finally, we note the appearance of three peaks at 302, 432 and 520 cm−1 in the Raman spectra of all WSe2 systems. These peaks arise from the oxidized Si substrate, as shown in the Raman spectra of the bare substrate in Fig. S3(a) in the ESI.† The peak at 302 cm−1 should be especially mentioned, as it has been previously attributed to WSe2,30 and explained as experiencing Breit–Wigner–Fano (BWF) type resonance for Raman measurements with 785 nm excitation. We disagree with this explanation as (1) there is no theoretically predicted one-phonon or multi-phonon mode of WSe2 at this frequency, (2) the Raman intensity of this peak scales exactly the same as the Raman intensity of the bulk Si 520 cm−1 peak (Fig. S3(b) in the ESI†), and (3) no other Raman modes of WSe2, especially fundamental one-phonon modes, experience BWF resonance for measurements with 785 nm, which makes it highly unlikely that only the 302 cm−1 mode would have such behavior. In conclusion, we attribute the Raman peak at 302 cm−1 to the TA phonon of Si and not to WSe2.
![]() | ||
Fig. 5 Exciton energies of WSe2 and Raman resonance. (a) Room-temperature photoluminescence spectra of 1L–5L of WSe2 measured at 488 nm excitation wavelength. (b) Exciton energies in 1L–5L WSe2 determined from the PL peak positions corresponding to the indirect band gap, A, B, A′ and B′ excitons. It should be noted that the energy values of the B′ exciton have been taken from ref. 51. Energies of the laser excitations used for the Raman measurements are indicated with horizontal lines labeled 488, 532 and 785 nm. (c) Comparison of the Raman intensity of the 250 cm−1 peak with the energy difference between WSe2 excitons and laser excitation, allowing identification of Raman resonance conditions for certain WSe2 flake thicknesses. |
The comparison of the Raman scattering intensity of the A1g/E2g mode located at 250 cm−1 with the energy difference between a certain exciton and laser excitation is plotted in Fig. 5c. In the case of the 488 nm laser excitation, we can observe that the Raman intensity of the A1g/E2g mode is high for 1L, decreases for 2L and 3L and again increases for 5L WSe2. Considering the band structure of WSe2, amplification effects due to allowed electronic transitions are observed in the case of 1L. They are due to the coupling with the A′ exciton, and in the case of 5L, due to coupling with the B′ exciton, as shown by the line indicating the laser energy in Fig. 5b. For Raman measurements performed with 532 nm excitation, the highest intensity of the A1g/E2g mode is observed in the case of 2L WSe2. This is followed by slightly lower intensities for 1L and 3L and the lowest intensity for 5L WSe2. The Raman amplification in this case is due to coupling with the A′ exciton, whose energy in the case of 2L WSe2 (2.32 eV) ideally coincides with the green laser excitation (2.34 eV). Finally, the Raman amplification in the case of 785 nm laser excitation is observed for 5L WSe2, as the intensity of the A1g/E2g mode is the highest in comparison to the other thicknesses. The Raman amplification in this case is due to coupling with the A exciton, whose energy for 5L WSe2 is around 1.59 eV, which is very close to the energy of the red laser excitation (1.58 eV). This implies that further Raman characterization, where, for example phonon–phonon or electron–phonon interactions would be investigated, would be best achieved when using 488 nm excitation for 1L and 3L, 532 nm excitation in the case of 2L and 785 nm excitation for 5L and higher WSe2 thicknesses.
Fig. 6a and b present the Raman spectra of 1L and 5L WSe2 respectively, measured with 488 and 532 nm excitations, along with the phonon cascade fits. We attribute these peaks to phonon cascades (PCs) in the Raman spectra of 1L–5L and bulk WSe2. These oscillations are not present under 795 nm excitation, as can be clearly seen in Fig. 3f, upon comparison with Fig. 3c and d. We have fit the spectra with three periodic Gaussian peaks, labeled as NPC = 1, 2 and 3, at about 16, 31 and 46 meV, respectively. Fittings of Raman spectra for 2L and 3L layers are presented in Fig. S4 in the ESI.†
Phonon cascades59,60 are due to the electron–phonon interaction, where electrons undergo phonon-induced transitions with a radiative recombination at each step, as illustrated in Fig. 6c. Phonon cascades start when incoming photons with energy (Eexc) excite electrons from the valence band to energy states corresponding to the in-plane wavevector cone of k ≤ ωexc/c, where ωexc is the frequency of the incoming photons (this cone is indicated with blue dashed lines in Fig. 6c). Due to phonon–electron interaction, the excited electrons undergo successive transitions between vibrational states (labeled with dashed black lines in Fig. 6c). At each step, one phonon with an energy corresponding to the difference between the vibrational states (Ephonon) is created. The creation of phonons is followed by a possible radiative recombination of the electron with the hole in the valence band and emission of a photon with an energy Eexc – NPCEphonon, where NPC corresponds to the number of phonons that were created in the phonon cascade process. As a result, periodic emission peaks are observed in the energy region close to the laser excitation (i.e. the Raman region, also called the hot photoluminescence region). According to the theory,61 similar behavior of phonon cascades is expected for involvement of excitons, as well as for free electrons or holes. Therefore, it is not possible to distinguish between free-carrier and exciton cascades from hot photoluminescence experiments. However, considering that excitonic behavior is dominant in WSe2, it makes sense that phonon cascades involving excitons are more probable.
The excitation energy of the incoming photons needs to be above the free-carrier gap excitation for phonon cascades to occur. This is the reason why phonon cascades in WSe2 are not observed with 785 nm laser excitation, as the energy of the A exciton for 1L–5L WSe2 is always higher than the 785 nm laser energy (1.58 eV), which in turn is insufficient to excite the carriers above the A exciton level.
Previously, phonon cascades were reported only in monolayer WSe2.62 Here, we report phonon cascades up to 5L and bulk WSe2. While the fine structure of the spectra is slightly different as a function of the number of ML, PC fitting did not reveal any significant shift in the energy of the PC peaks with the change in the WSe2 layer thickness, or with the change in the laser excitation, as indicated in Fig. 5d. Linear fitting of the PC peak energy (EPC) with the number of peaks in the cascade (NPC), allows for the determination of the energy of the phonon (Ephonon) involved in the cascade process, as shown in Fig. 5d. This also corresponds to the average energy difference between successive PC peaks. In this case, the obtained phonon energy of around 15.4 meV is in agreement with the reports for monolayer WSe2.62 This energy corresponds well to the out-of-plane (ZA) acoustic phonon (around 125 cm−1) at the M and K points of the Brillouin zone (Fig. S2 in the ESI†). The involvement of the ZA phonon is further confirmed by its phonon branch being relatively flat in the region between the M and K points, which is necessary for the generation of a high number of PC peaks.
The observation of phonon cascades in WSe2 indicates that an exceptionally strong electron–phonon interaction occurs in the initial steps of carrier relaxation that prevails until 5ML, which may provide design-relevant insights for emerging optoelectronic applications.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4na00399c |
This journal is © The Royal Society of Chemistry 2024 |