Letter pubs.acs.org/NanoLett
Temperature-Dependent Resonance Energy Transfer from Semiconductor Quantum Wells to Graphene Young-Jun Yu,*,†,○ Keun Soo Kim,‡,○ Jungtae Nam,‡,○ Se Ra Kwon,§ Hyeryoung Byun,§ Kwanjae Lee,∥ Jae-Hyun Ryou,⊥ Russell D. Dupuis,# Jeomoh Kim,# Gwanghyun Ahn,∇ Sunmin Ryu,∇ Mee-Yi Ryu,*,§ and Jin Soo Kim*,∥ †
Creative Research Center for Graphene Electronics, Electronics and Telecommunications Research Institute (ETRI), 218 Gajeong-ro, Yuseong-gu, Daejeon 305-700, Korea ‡ Department of Physics and Graphene Research Institute, Sejong University, Seoul 143-747, Korea § Department of Physics, Kangwon National University, Kangwon-Do 200-701, Korea ∥ Division of Advanced Materials Engineering, Research Center of Advanced Materials Development (RCAMD), Chonbuk National University, Jeonju 561-756, Korea ⊥ Department of Mechanical Engineering and Texas Center for Superconductivity at the University of Houston (TcSUH), University of Houston, Houston, Texas 77204-4006, United States # Center for Compound Semiconductors and School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332-0250, United States ∇ Department of Chemistry, Pohang University of Science and Technology, Pohang 790-784, Korea S Supporting Information *
ABSTRACT: Resonance energy transfer (RET) has been employed for interpreting the energy interaction of graphene combined with semiconductor materials such as nanoparticles and quantum-well (QW) heterostructures. Especially, for the application of graphene as a transparent electrode for semiconductor light emitting diodes, the mechanism of exciton recombination processes such as RET in graphene-semiconductor QW heterojunctions should be understood clearly. Here, we characterized the temperature-dependent RET behaviors in graphene/semiconductor QW heterostructures. We then observed the tuning of the RET efficiency from 5% to 30% in graphene/QW heterostructures with ∼60 nm dipole− dipole coupled distance at temperatures of 300 to 10 K. This survey allows us to identify the roles of localized and free excitons in the RET process from the QWs to graphene as a function of temperature. KEYWORDS: Resonance energy transfer, graphene, semiconductor quantum well, free exciton, localized exciton, temperature dependence
F
LED will play a significant role in the optical performance characteristics of the devices. It has been shown that the RET from semiconductor QWs to graphene occurs by acting as a donor and an acceptor, respectively, with sufficient overlap of spectra between the absorption of graphene and the photoemission of the semiconductor QWs.5−8 Especially, studies have been carried out on the RET process between graphene and semiconductor QWs with a focus on dipole−dipole coupled distances, i.e., the thickness of the cap layer that isolates the graphene from the QWs.5−15 However, despite this dipole−dipole interaction dramatically influenced by temper-
örster resonance energy transfer (RET) that reflects the transferring of the absorbed photon energy in the donor to the acceptor has been demonstrated in dipole−dipole coupled systems such as two chromphores, semiconductor−nanoparticles, and graphene−semiconductor junctions.1−20 In particular, the energy transfer between dipoles in one- and two-dimensional materials has been studied for the coupling of graphene with semiconductor materials such as nanocrystals and quantum wells (QWs). Owing to significant transparency (>90% transparency to visible light), high conductivity (∼4 e2/ h), and high flexibility (foldable), graphene has been proposed for the application of a transparent electrode in semiconductor light-emitting diodes (LEDs) to replace indium tin oxide.21−30 In the case of visible LEDs with graphene electrodes, the energy interaction between the graphene electrode and the QWs of the © XXXX American Chemical Society
Received: September 21, 2014 Revised: December 30, 2014
A
DOI: 10.1021/nl503624j Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
Figure 1. (a) Schematic illustration of transferred graphene layers on an InGaN/GaN QW structure with a GaN cap layer. Here we adjusted the number of the graphene layers from 1 to 4, and the cap layer thicknesses were 10, 20, 30, and 50 nm. (b) Raman spectra of SLG, BLG, TLG, and QLG on 10 nm-thick cap-layered InGaN/GaN QW and Raman spectrum of bare InGaN/GaN QW (dashed line). Inset: Optical images of the bare InGaN/GaN QWs on sapphire substrate and the QLG transferred MQWs. The scale bar is 1 cm. (c) The represented PL spectra of bare QW and QLG/QW under different temperatures. Here QW is covered with Wcap = 10 nm. (d) PL intensity variation (ΔI/I10) of bare QW (open circles and dashed lines) and QLG/QW (filled circles and dashed lines) with Wcap= 10 nm (black), 20 nm (blue), and 30 nm (red) as a function of temperature. Black and red solid lines are fitted with activation energy of Ea = 90 and 65 meV, respectively. Here although there are missing data at 230 and 270 K for bare QW and QLG/QW with Wcap = 30 nm, respectively, the trend of ΔI/I10 as a function of temperature is not influenced. Inset: PL intensity variation (ΔI/I10) of bare QW (open circles and dashed lines) and QLG/QW (filled circles and dashed lines) with Wcap = 50 nm as a function of temperature. Black and red solid lines are the guided lines with activation energy of Ea= 75 and 65 meV, respectively.
QW structures covered with graphene were fabricated, as shown in the optical images of bare QWs and QWs with quadruple-layer graphene (QLG) in the inset of Figure 1b. The number of graphene layers was confirmed by Raman spectra, exhibiting a fingerprint of increased intensity ratio between the G and 2D peaks (IG/I2D) from 0.5 for single-layer graphene (SLG) to 2 for QLG as shown in Figure 1b.31 Here, since the Raman peaks of bare InGaN/GaN QW appear at 418.6, 569.2, and 735.4 cm−1 for sapphire substrate, E2H phonon mode, and A1(LO) mode, respectively, we can observe clear Raman spectrum of graphene in the range of ∼1500−3000 cm−1 (see Supporting Information, Figure S1.)32,33 The signal offset between the Raman G and 2D peaks of the SLG appears to be due to the embedded tail of the photoluminescence (PL) spectrum of the QW, and this offset fades away with increasing number of graphene layers, resulting in both the screening of PL emission from the QWs and scaling of IG and I2D. Since we did not observe a broadening in the line-width of the Raman 2D peaks fitted by a single Lorentzian peak, which is associated with the interactions of phonon variation modes between the stacked graphene layers for bi-layer graphene (BLG), triplelayer graphene (TLG), and QLG, the stacked graphene layers on the QWs in this work show a random stacking condition rather than Bernal stacking (see Supporting Information, Figure S1).31,34 Also, from the analysis of Raman 2D frequency (ω2D) vs G frequency (ωG),35 we estimated an accumulated hole
ature variation, the dependence of RET on temperature in such structures is yet to be investigated. Upon comprehending the temperature dependence of exciton recombination mechanism on these hybrid structures, the avenues could be opened for better control and enhancement of LED−graphene electrode junction performance. In this work, we characterized the RET behaviors influenced by exciton recombination in graphene/ QW heterostructures at different temperatures. The RET efficiency from the QW with low population of localized excitons to graphene was enhanced with decreasing temperature (80%) of QLG,36 we can excite carriers and measure the PL of QWs through the QLG. Although different influences occur from graphene to each upper and lower QW, the LED structure consisting of five-period multiple QWs can be simplified as a single QW band structure with an effective potential barrier width (dϕ) and a well width (Lw) by focusing on the PL peak energy from the five-period QWs band structure ensemble (see Supporting Information, Figure S4). The energy variation of the dominant PL peaks that is related to the effective band gap variation, ΔEG = EG0 − EG1, of QWs (where EG0 and EG1 are the PL peak energies of bare QW and graphene/QW, respectively), with different Wcap as a function of the number (N) of graphene layers was studied. The peak variation shows a red-shift whereby the ΔEG increases by ∼70 meV with escalating N. Especially, the highest variation of ΔEG was observed from Wcap = 10 nm because this cap is thinner than the others and the QLG that supports a sufficient electric field larger than the influences of the residues at the interfaces of the graphene layers led to the obvious PL-peak red-shift of QW compared with SLG, BLG, and TLG (see Supporting Information, Figure S3). This result suggests that rearranging the Fermi energy level by different work functions between hole-accumulated graphene and InGaN/GaN QWs37−42 leads to an increase in quantum-confined Stark effect (QCSE), which causes the reduced band gap energy of the QW.43 We confirm that the accumulated charge carrier density (assumed in the range of 4.0−6.0 × 1012 cm−2)30 in QLG yields sufficient electric field to induce strong coupling between graphene and
QW, leading to the additional QCSE (see Supporting Information, Figure S4). The charge carrier transfer between graphene and QWs does not occur due to the sufficient thickness (Wcap = 10−50 nm) of the cap layer; however, the RET with the InGaN/GaN QWs will occur, which is associated with an energy corresponding to the PL emissions at around 2.75−2.85 eV (wavelength λ = 435−450 nm) overlapped with a broad absorption spectrum of graphene from the near-infrared to ultraviolet spectral range.7 Since the behavior of the multilayered graphene that is stacked randomly on QWs is similar to that of the SLG for RET (except for the induced additional QCSE), we focused on a QLG/QW sample for the RET studies in this work. In Figure 1c for QLG/QW in comparison to bare QW, we observed not only the red-shifted peak (ΔEG) but also the variation in the PL intensities. Figure 1d shows the relative PL intensity variation (ΔI/I10) of bare QW and QLG/QW. Employing the Arrhenius equation as ΔI/I10 = 1/(1 + A exp[−EA/kBT]) to fit the thermal quenching data, the activation energies of bare QW (EA) and QLG/QW (EA′) as a function of temperature (T) are extracted, where ΔI, I10, A, and kB are the variation of PL intensity, the PL intensity at 10 K, rate constant, and Boltzmann constant, respectively.44−49 In the case of the QWs with Wcap = 10, 20, and 30 nm, the EA of 90 meV decreases to the EA′ of 65 meV. For Wcap = 50 nm, EA and EA′ are 75 and 65 meV, respectively. The activation energy mainly originates from the localized exciton energy due to band-edge fluctuations caused by the variation in the indium mole fraction.42,43,46 This result suggests that the population of the localized excitons leading to EA in QW with Wcap = 10, 20, and 30 nm is higher than that in QW with Wcap = 50 nm, and that the localized potential states leading to EA′ affects the RET in QLG/QW. For the investigation of the RET contribution from the QW to graphene at different T, we employed PL decay dynamics. C
DOI: 10.1021/nl503624j Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
Figure 3. (a) Energy transfer efficiency between QLG and QW under different T and cap-layer thickness of Wcap = (I) 10 nm (black dots) and (II) 50 nm (red dots). (b) Averaged RET efficiency (ηET) at high T = 170−300 K (red dots) and low T = 40−170 K (black dots) of QLG/QW with different Wcap evaluated from panel a and Supporting Information, Figure S7. The effective potential barrier distance dϕ between graphene and QW could be assumed as dϕ = Wcap + LB with Wcap = 10, 20, 30, and 50 nm and barrier width (LB) = 11 nm. The red solid line indicates the variation of ηET evaluated with the Förster radius of d0 = 16 nm and n = 2. The blue and black dashed lines show the ηET as a function of dϕ obtained by employing d0 = 6 and 31 nm, respectively, with n = 2.
increase as τR < τR′ at 170 K < T < 300 K, and at T < 170 K, the offset between τR and τR′ are decreased for Wcap = 10 nm and reversed (τR > τR′) for Wcap = 50 nm, as shown in Figure 2e,f, respectively (see also Supporting Information, Figure S5). Employing the internal quantum efficiency (IQE), the ratio of the number of collected charge carriers in QW to a given excitation energy, the IQE of QLG/QW evaluated by IQE = τNR′/(τR′+τNR′) exploits the suppression of 10−30% compared with that of bare QW by the RET from QW to graphene, and it exceeds ∼96% at 40 K (see Supporting Information, Figure S6).47,48 This demonstrates that, while the free and localized excitons maintain saturation at T = 40 K, an unexpected radiative recombination variation occurs between τR and τR′ under T < 40 K (see Figure 2c,e) due to the supplemented extra carriers in the QW from the undesired energy states such as residues at the interface between graphene and QW at under T < 40 K. Since the subtracting rate for QLG/QW versus QW could ignore the contribution of the undesired residues at interface between graphene and QW,55−57 anomaly behaviors such as negative value of RET appear. Thus, we neglect not only the PL decay times of all QLG/QW under T < 40 K but also any τGQW that are longer than τQW for RET discussion as shown in Figure 3. Note that no opposite energy transfer occurs from graphene to QW due to the very fast carrier relaxation time (∼10−100 fs) in graphene with the absence of a band gap58 compared to the RET time of τET = 10−250 ns. To verify the RET variation between graphene and QW under different T, we derived the RET efficiency (ηET) with ηET = τET−1/(τQW−1 + τET−1) = 1 − (τGQW/τQW).7−15 By plotting ηET as a function of T, as shown in Figure 3a, we determine two trends for ηET, whereby as T decreases, ηET for Wcap = 10 nm (I) reduces steadily from the maximum value at 230 K, but for Wcap = 50 nm (II), ηET increases gradually from the minimum efficiency at 170 K. In Figure 3b, employing the relation between RET efficiency and dϕ as ηET = 1/[1 + (dϕ/d0)n], the Förster radii of QLG/QW are extracted with different Wcap and T, where d0 is the Förster radius, which is a distance between two dipoles leading to the ηET of 50%, and n = 2 is a parameter representing the interaction between two-dimensional dipoles.3−15 The tuned ηET from 30% to 5% with increasing dϕ from 21 to 61 nm agrees well with the evaluated ηET (red solid line in Figure 3b) using the Förster radius d0 =16 nm and n = 2
Figure 2a shows a comparison of PL decay dynamics between bare QW and QLG/QW with Wcap = 50 nm at 10 and 300 K. The PL decay contains a fast and a slow component, and thus, the PL decay curves can be expressed by a two-exponential function, A1 exp(−t/τ1) + A2 exp(−t/τ2), where τ1 and τ2 are the fast and slow decay times, respectively, and A1 and A2 are the contributions of the corresponding parts to the total PL intensity. The slow decay times of both bare QW and QLG/ QW are ∼18−30 ns at 10 K and ∼13 ns at 300 K. These slow decay times are due to an impurity-related transition, while the fast PL decay times are mainly related to the recombination of the excitons in QW.50,51 Although there could be other multiple processes involved,52−54 RET between band structures of QW and graphene can be addressed by fast decay time, and the contribution of impurity-related transition leading to slow decay time will be discussed later. Therefore, our following discussion will focus on the changes in fast decay times for bare QW (τQW) and QLG/QW (τGQW). The PL decay dynamic behavior of bare QW and QLG/QW with Wcap = 50 nm represents a decrease in the decay times from τQW ≈ 9.3 ns and τGQW ≈ 7.8 ns at 10 K to τQW ≈ 2.3 ns and τGQW ≈ 2.1 ns at 300 K, respectively, as shown in Figure 2b,d. Figure 2c,d shows the τQW and τGQW as a function of T and the escalation of both τGQW and τQW with decreasing T due to the increased trapping probability of the excitons. The figures also show the offset between τGQW and τQW due to the RET from QWs to graphene. Furthermore, we observe the longer τGQW than τQW at low T ≈ 10 K for the thin cap layer (Wcap = 10 nm) (see also Supporting Information, Figure S5). Reconstructing the PL decay time with radiative (τR) and nonradiative (τNR) recombination lifetimes, as well as the characteristic time of the RET process (τET), τQW and τGQW can be expressed as 1/τQW = 1/τR + 1/τNR and 1/τGQW = 1/τR′ + 1/τNR′ = 1/τR + 1/τNR + 1/τET, respectively, where τR, τNR and τR′, τNR′ are the radiative and nonradiative decay times for bare QWs and QLG/QWs, respectively.7−12,48,49 Figure 2e,f shows the radiative and nonradiative decay times for the bare QW and the QLG/QW as a function of T. For nonradiative recombination, while maintaining the rapid elevation of both τNR and τNR′ for decreasing T, we could also observe the diminished lifetime (τNR > τNR′) of the QWs covered with graphene. However, the radiative decay times lead to a slight D
DOI: 10.1021/nl503624j Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters in the high-temperature range (170 K < T < 300 K), while in the low-temperature range (40 K < T < 170 K), the ηET decreases to 6.7% and 4.9% for dϕ = 21 and 31 nm, respectively. The ηET for dϕ = 21 and 31 nm at low T can be well fitted with d0 = 6 nm and n = 2 as shown in the blue dashed line in Figure 3b. The reduced ηET at low T is attributed to the suppression of the RET process for QLG/QW with Wcap = 10 nm, which is case (I) in Figure 3a, due to the increased number of localized excitons at low T. The probability of exciton localization at potential traps in InGaN QW resulting from indium segregation is escalated at low T due to the decreasing thermal energy.9 Namely, in the case of (I), while the relatively short distance of dϕ leads to a strong RET from QW to graphene at high T (>170 K), the large population of point-like localized excitons at low T (40 K < T < 170 K) reduces the RET rate. Moreover, owing to the elevation of probability for exciton dissociation at room temperature, a decrease of RET could be observed.9 On the contrary, in the case of (II), the increase of ηET exceeded ∼20% for dϕ = 61 nm at low T (40 K < T < 170 K) as shown in Figure 3a, which agrees well with d0 ≈ 30 nm and n = 2 as shown in the black dashed line in Figure 3b. This reveals that although the relatively long distance of dϕ blocks the RET (ηET ≈ 6%) between QW and graphene at high T (>170 K), the escalation of ηET for decreasing T is attributed to the elevated contribution of free excitons to the RET process rather than to a radiative recombination process leading to a shorter τR′ than τR at low T (≤100 K), as shown in Figure 2f. Rindermann et al. also reported that for decreasing T, RET could be enhanced by modifying the Wannier−Mott exciton dimensionality with in-plane wave vectors by free excitons that are more dominant than localized excitons.9 Therefore, we can confirm that the optimum wave vector with the suppression of the population of localized excitons in a graphene−semiconductor QW heterostructure results in the significant improvement of the RET radius, d0 ≈ 30 nm, even under a long dipole distance dϕ = 61 nm at T < 170 K. Note that as aforementioned with IQE, the negative values of ηET for Wcap = 10, 20, and 30 nm extracted with longer τGQW than τQW as shown in Figure 2a,c, corresponding to longer τR′ than τR at T < 40 K as shown in Figure 2e, are overlooked in this work. Furthermore, negative values of ηET (τGQW > τQW) for slow lifetime (τ2) reflected impurity-related transitions are extracted under T < 150 K for all specimens (see Supporting Information, Figure S7). This indicates that impurities in QLG/QW reflected with slow lifetime contribute to the negative value of ηET extracted by fast lifetime (τ1) at T < 150 K. Therefore, the anomalies in Figures 1d and 2c can be attributed to the undesired residues and impurities at low T. In conclusion, we characterized the variation of the exciton recombination process in semiconductor QWs covered with graphene layers. The RET efficiency from the QWs to graphene with the effective barrier width of 21−61 nm is determined to be ∼30−5%, respectively, at high T. Especially, the reverse RET efficiency tendencies such as the reduction and increase of large and small populations of localized excitons in QLG/QWs, respectively, are observed at low T (