Article pubs.acs.org/JPCC
Determination of the Energy Band Gap Depending on the Oxidized Structures of Quantum Dots Eunseog Cho, Hyosook Jang, Shinae Jun, Hyun A Kang, Jae Gwan Chung, and Eunjoo Jang* Advanced Material Research Center, Samsung Advanced Institute of Technology, Samsung Electronics Co., Mt.14-1, Nongseo-Dong, Giheung-Gu, Yongin-Si, Gyeonggi-Do 449-712, Korea S Supporting Information *
ABSTRACT: Theoretical and experimental studies on the changes of the optical properties of CdSe/CdS/ZnS (core/double-shell) quantum dots (QDs) during the oxidation process were first performed. An effective medium approach using the modified Khon−Sham equation presents a new method to predict the effects of the oxidation and to determine the oxidized ratio of nanoscale materials by a quantitative comparison with the experimental photoluminescence (PL) changes. As the oxidation progressed from the CdSe/CdS/ZnS nanocrystal surface, the PL peak shifted to longer wavelength and the quantum efficiency (QE) continuously decreased. It was also found that such changes were accelerated when the thickness of the outermost ZnS shell became thinner than a monolayer. The radial wave functions showed that the electron carriers rapidly extended into the shell region while the hole carriers spread very little into the core region. This indicates that the electrons are the key carriers to induce the changes in the energy band gap and the QE.
Q
using an effective medium approach to estimate the optical band gap of QDs with multishells and present a quantitative prediction of the PL wavelength shifts resulting from gradual oxidations. The UV exposure experiment and the X-ray photoelectron spectroscopy (XPS) analysis were carried out to clarify the calculations. By the close interplay between the modeling and experimental result, the nanocrystalline oxidized structure was predicted quantitatively and the physical origin causing the optical band gaps was suggested. Even though there exists a fair amount of theoretical understanding on the quantum confinement effect (QCE) as a function of size of the core QD particles,7,8 little is known about the effect on the QCE of the QD structures with multi passivation layers created by either designed syntheses or inadvertent oxidation. Inerbaev et al. investigated the oxidation effect of small (CdSe)n clusters (n = 6−26) based on the density functional theory (DFT) and suggested that the PL peak could be blue-shifted or red-shifted depending on the oxygen atom adsorbed on two-coordinated or three-coordinated Cd atoms on the surface.9 The DFT method, however, imposes a tight limitation on the number of atoms in dealing with the nanostructures. A CdSe QD with a radius of 2.0 nm is composed of about 1200 atoms, and if the CdSe core is coated with shells, the number of atoms increases to several thousands.10 In addition, colloidal QDs contain complex organic surfactants coordinated with dangling atoms on the surfaces. For these reasons, we devised an efficient method
uantum dots (QDs) have been proposed to be applicable as an active luminescent material for display applications due to their high quantum efficiency (QE), easy color tunability, and highly saturated color.1 In particular, down converting light emitting diodes (LEDs) with QD color converting layers have shown superior luminous efficacy and color gamut compared to those of the conventional phosphors.2,3 Such application requires that QDs should endure high initial QE and maintain photoluminescence (PL) spectrum under the pertinent conditions such as high radiation dose and high LED junction temperature. However, the carriers trapped in QDs’ defect states give rise to nonradiative relaxation inducing the decrease of the QE. In addition, the highly reactive dangling atoms on the surface make QDs vulnerable to oxidative degradation, altering the PL spectrum.4 In order to resolve such undesirable surface conditions, passivation methods using multishell structures have been suggested.5,6 Previously, we disclosed a continuous and scalable method without a core isolation step to synthesize CdSe/CdS/ ZnS core/double-shell QDs in one pot.6 The prepared CdSe/ CdS/ZnS QDs were applied on a 390 nm UV light emitting diode (LED) as a red color converter, and we obtained 48% of external quantum efficiency (EQE) at 5 mW with a narrow spectrum width of 20 nm. However, since the LED packages for display back-light units or general illuminations typically require 300 mW to 1 W, the optical properties such as the luminescence intensity and the emission spectrum can be seriously deteriorated during the operation even for the QDs with multipassivation layers. To the best of our knowledge, neither the reliabilities nor the corresponding structural changes of the QDs under the oxidative environment have been reported yet. Here, we propose a new calculation method © 2012 American Chemical Society
Received: March 13, 2012 Revised: April 21, 2012 Published: May 2, 2012 11792
dx.doi.org/10.1021/jp3024217 | J. Phys. Chem. C 2012, 116, 11792−11796
The Journal of Physical Chemistry C
Article
the same energy band gap of oleic acid (5.0 eV).7 Figure 1b represents the change of the optical gap as a function of the size of the CdSe core by using our calculation. The QCE diminished as the radius of CdSe core increased, and the band gap at the radius of 4.5 nm became 1.81 eV. Since this value was similar to the experimental energy gap of 1.76 eV for the bulk CdSe crystal, it was speculated that the QCE was negligible over the radius of 4.5 nm in the CdSe QD. The calculated optical gap was slightly larger (∼0.14 eV) than the experimental value of the CdSe core with a radius of 2.0 nm. This can be rationalized by two possibilities. First, our calculation was based on the ground state energy, which means all the data represent the band gaps at the absolute temperature, 0 K. However, the experimental gaps in Figure 1b were measured at the ambient condition around 298 K. Generally, semiconductors have more reduced band gaps at higher temperature, and this effect is expressed by an empirical Varshni relation13
based on the effective medium approach by solving the modified Khon−Sham equation to deal with a core/multishell structure; we calculated the Khon−Sham equation selfconsistently by using the effective masses and energy band offsets in the three-dimensional real space. Real-space approaches have been specially designed to overcome difficulties in the plane-wave pseudopotential method, which requires a large number of basic functions and usually adopts the periodic boundary conditions even for a localized system like a QD.11 The real- space approach is particularly useful to calculate the total energies of charged systems because the compensating background charge that would be required for supercell geometries is not necessary. The Ceperley and Alder form was employed as an exchange-correlation function, and a higher-order finite difference method with multigrid technique was adopted for the self-consistent calculation. The optical band gap (Eopt g ) was obtained by the following equation. Egopt = E( +1) + E( −1) − 2E(0) − ECoul
(1)
Eg (T ) = Eg 0 − α
where E(+1), E(−1), and E(0) are the total energies of the charged structures with the charge of +1, −1, and 0, respectively, and ECoul represents the Coulomb energy, which is treated in the first-order perturbation theory. The detailed calculations and related parameters are disclosed in the Supporting Information. The CdSe/CdS/ZnS QDs were prepared by previously disclosed method,6 and they showed higher than 60% of QE with uniform size and shape. From the inductively coupled plasma−atomic emission spectroscopy (ICP−AES) and transmission electron microscopy (TEM, Figure S3) analysis, the radius of CdSe core was measured as 2.0 nm and the thicknesses of CdS and ZnS shells were calculated as 0.75 and 0.65 nm, respectively. For the calculation of the optical gaps, it was assumed that the multishell QD was a perfect sphere and that the core and shells shared same centric position (Figure 1a). Interestingly, CdSe QDs have been experimentally verified that their energy band gap shows no measurable dependence on the nature of the coordinating organic ligands.12 Therefore, it was also assumed that all the ligands on the QD surface had
T2 T+β
(2)
where Eg0 is the energy gap at 0 K and the parameters α and β are taken as 370 μeV/K and 150 K, respectively from the CdSe bulk crystal data.14 Considering the temperature effect, the band gap obtained by the modeling could be reduced by 0.07 eV at 300 K, but the calculated gap was still 0.07 eV larger than the experimental value. Second, in the calculation, the external potential was adopted as a finite hard wall in the core−ligand interface. In real circumstances, however, charge transfer happens to some extent at the interface even at 0 K, which leads to the interface potential being less hard. Moreover, the elevated temperature facilitates such transfer and makes the interface potential deviate from the hard wall more than at 0 K. As a result, the calculated QCE tends to be exaggerated compared to the real condition. Thus the calculated energy gaps become larger than the experimental values. Our purpose was to elucidate the structural changes quantitatively by comparing the calculated value with the experimental properties, which were measured at the ambient condition and innately exerted as the less hard potential at the interface. Therefore, the following calculations were performed simply by treating the core size as a little larger to alleviate the QCE and to mimic the experimental situations as depicted in Figure 1c. The synthesized CdSe with a radius 2.0 nm showed the optical gap of 2.13 eV, but the calculation renders almost an identical gap of 2.14 eV when the radius of the core was assumed as 2.25 nm. The optical band gaps of the core, core/shell, and core/ multishells QDs are presented in Table 1. According to our Table 1. Optical Band Gaps of the Core (CdSe), Core/Shell (CdSe/CdS), and Core/Multishells (CdSe/CdS/ZnS) QDs band gap (eV)
CdSe
CdSe/CdS
CdSe/CdS/ZnS
experiment6 modeling
2.13 2.14
2.07 2.05
2.07 2.05
calculation, the first CdS shell with the thickness of 0.75 nm reduced the band gap by 0.09 eV. However, the second ZnS shell with the thickness of 0.65 nm did not diminish the band gap more. This trend perfectly agreed with the experiments. This can be explained by the fact that both electron and hole carriers are confined in the ZnS shell more strongly than in the
Figure 1. (a) Schematic structure of CdSe/CdS/ZnS. (b) Band gap energy as a function of the radius of CdSe core: (■ modeling values, red ● experimental data) (c) Potential walls () and wave functions (blue ). 11793
dx.doi.org/10.1021/jp3024217 | J. Phys. Chem. C 2012, 116, 11792−11796
The Journal of Physical Chemistry C
Article
ZnO (CdSe/CdS/ZnO), and finally some part of the first shell can be oxidized to CdO (Figure S2-b, CdSe/CdS/CdO/ZnO) after a long-time exposure in the oxidation environment. Table 2 shows the PL peak wavelength as a function of the oxidation
CdS shell because ZnS has heavier carrier effective masses and larger band gap energy than CdS has. The changes of optical properties in the CdSe/CdS/ZnS QD toluene solution with the same optical density of 0.1 were measured when the QDs were exposed under 365 nm UV irradiation with 90 W, and measured PL spectra and binding energy changes are shown in Figure 2. The peak wavelength
Table 2. PL Peak Wavelength as a Function of the Oxidation Ratio on the Basis of the ZnS Shell Thicknessa oxidation ratio
0%
25%
50%
75%
100%
125%
λmax (nm)
606
607
609
617
633
672
a
0%∼100% represents the thickness ratio of ZnS shell changed to ZnO; 125% is the 25% of the CdS shell oxidized to CdO after the entire ZnS oxidation to ZnO.
ratio on the basis of the shell thickness which represents the degree of the formation of ZnO from the ZnS shell. The PL wavelength shows a dramatic shift to a longer wavelength as the oxidation progresses. Until the 25% of the ZnS thickness was oxidized, the wavelength hardly changed, but when the oxidation ratio of the ZnS shell reached 50%, the wavelength was red-shifted by 3 nm. The 75% oxidation of ZnS shell led to an 11 nm red-shift, and finally the 100% of ZnS change to ZnO resulted in a 27 nm red-shift. Further oxidation after the entire oxidation of ZnS shell to ZnO makes the next CdS to CdO. If the 25% of CdS is oxidized to CdO, the PL wavelength will be excessively red-shifted by 66 nm from the initial wavelength. From the experimental red-shift of 8 nm under the UV irradiation of 16 h, we concluded that the final oxidation ratio observed in this experiment lay between 50% and 75% of ZnO formed in the ZnS outer shell. For the hole carrier, since it has a lower valence band maximum (VBM) and heavier effective mass in the ZnO than in the ZnS, the oxidation leads to more strongly confined wave function, and the resultant PL wavelength is expected to become a blue-shift. On the contrary, for the electron carrier, it has a much lower conduction band minimum (CBM) and lighter effective mass in the ZnO than in the ZnS. Thus, it is expected that after the oxidation, the electron easily spreads toward the outer region and results in a red-shifted wavelength. The red shift of the PL wavelength represents that the electron exerts stronger influence on the optical gaps than the hole carrier. At the initial stage of the oxidation, the ZnS shell is thick enough to confine the carriers, so the portion of ZnO in the outer shell did not affect the PL. At the 50% oxidation, the ZnS shell becomes thinner to 0.325 nm from 0.65 nm, which is almost the same as the ZnS monolayer (0.31 nm),20 and consequently the carriers are capable of tunneling through the thin shell to confront with the energy potential of ZnO. If the oxidation proceeds more than 50% of the ZnS, the ZnS shell becomes thinner than a monolayer. In that case, the capability of confining the electron becomes so weak that the PL peak sharply shifts to the longer wavelength. For the CdS oxidation, since the CdO has the lower VBM than the CdS, the hole carriers are more tightly confined after the oxidation. On the other hand, electrons are likely to extend toward the shell region, since the CBM for CdO is 1.41 eV lower than that of CdS. Just like the ZnS oxidation, the rapid red-shift of the PL wavelength represents that the spreading of the carrier is much superior to the confining of the carrier. Our prediction can be clarified in Figure 3, which represents the changes of the radial probabilities and the lowest wave functions for both holes and electrons as the gradual oxidation proceeds. In Figure 3a, the oxidation draws the hole wave
Figure 2. (a) PL spectra of CdSe/CdS/ZnS solution under 365 nm UV irradiation. (b) XPS analysis of the binding energies of Zn 2p3/2 (left) and S 2p3/2 (right) electrons before (red) and after (blue) the UV irradiation.
was shifted from 603 to 611 nm after the continuous UV exposure for 16 h, and the QE dropped severely from 63% to 1%. Also, the QDs tended to make precipitates over time. It was expected that the wavelength shifts and QE drops resulted from the oxidation of the outer shell. The XPS analysis was done to confirm the oxidation state of the QDs given in Figure 2b. The binding energy of the electrons of Zn 2p3/2 shifted to 0.4 eV higher state from 1022.0 to 1022.4 eV; the 1022.0 eV is almost identical to Zn 2p3/2 electron binding energy (1021.9 eV)15 for the ZnS crystal, and the 1022.4 eV corresponds to ZnO (1022.2−1022.4 eV).16 For the S 2p3/2 electron, the binding energy appeared around 170 eV after the irradiation, which represents the formation of new oxide phase; the S 2p3/2 electron binding energy for ZnS is 161.5 eV.15 This demonstrates that the oxidation of both Zn and S happened after UV irradiation. Although the change of the optical properties could be measured, it was difficult to identify the local structure changes and the degree of the oxidation in the core/multishells nanostructure. Although intensive studies on the oxidation of the various ZnS structures have been carried out previously, 17−19 the exact chemistry and reaction mechanisms are considerably complicated and disputable depending on the temperature, pressure, or oxidation regime. But, the primary product after the oxidation has been suggested as ZnO. Therefore, we presumed that some part of the outermost ZnS shell becomes first oxidized to ZnO (Figure S2a, CdSe/CdS/ZnS/ZnO) in the process of photo-oxidation, then by further oxidation the entire ZnS can be oxidized to 11794
dx.doi.org/10.1021/jp3024217 | J. Phys. Chem. C 2012, 116, 11792−11796
The Journal of Physical Chemistry C
Article
Figure 3. Radial probabilities and the lowest wave functions (right bottom) for (a) the holes and (b) the electrons. The upper and lower wave functions for both carriers show the S-orbitals of the CdSe/CdS/ZnS (0%: nonoxidized CdSe/CdS/ZnS structure) and the CdSe/CdS/CdO/ZnO (125%: 25% of the CdS shell oxidized to CdO after the entire oxidation of ZnS to ZnO).
■
function to the core region, but the effect is not great over the entire oxidation. This confirms that the hole carrier does not give a significant influence on the optical band gap. For the electron (Figure 3b), however, as the oxidation progresses, the wave function noticeably extends to the shell region. For the 50% oxidation, the radial probability considerably spreads to the shell region, and when the entire ZnS shell is oxidized to ZnO, the electron carrier is totally delocalized between the CdSe core and the CdS shell. In the case of the 25% of CdS being oxidized to CdO, the most probable carrier region moves to the shell from the CdSe core, and consequently the electron is not confined in the core region any more. The changes of the radial probabilities and wave functions by the oxidation confirm that the electron, not the hole carrier, is the origin to trigger the red-shift of the PL peak observed in the experiment. Moreover, exciton binding energy gets weakened during the oxidation because the electron spreads into the shell region while the hole moves into the core region. Thus, the oxidation brings about the QE reduction because of the less probable recombination of the electron and hole carriers. In conclusion, we studied the changes of the optical properties as a function of the oxidation of the CdSe/CdS/ ZnS QDs by the close comparison between the theoretical modeling and the experiments. The red-shift of the PL peak from 603 to 611 nm was observed under the 365 nm UV irradiation for 16 h, and the QE also dropped from 63% to 1%. The XPS analysis confirmed that the ZnS shell was oxidized after the UV irradiation. For the theoretical investigations, we devised an efficient method based on the effective medium approach by solving the modified Khon−Sham equation in three-dimensional real space. Our modeling showed the similar PL peak shifts to the longer wavelength with the progress of the oxidation. By the quantitative comparison with the experimental PL shift, it was found to correspond that 50−75% of the initial ZnS shell was oxidized. It was also found that such red-shift effect was accelerated when the thickness of ZnS shell became thinner than a monolayer. The radial probabilities and the lowest wave function changes for both electron and hole carriers elucidated that electron carrier is the key factor to result in the red-shift. Moreover, the electron and the hole carriers move in the opposite direction during the oxidation, which led to the QE reduction.
ASSOCIATED CONTENT
* Supporting Information S
Calculations details, effective masses, and parameters for the calculations, energy band offsets for CdSe and multishell structures, oscillator strength, and TEM images of CdSe, CdSe/ CdS, and CdSe/CdS/ZnS. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail
[email protected], tel. 82-31-280-6753, fax 8231-280-9349. Notes
The authors declare no competing financial interest.
■
REFERENCES
(1) Steckel, S. J.; Snee, P.; Coe-Sullivan, S.; Zimmer, P. J.; Halpert, E. J.; Anikeeva, P.; Kim, L.-A.; Bulovic, V.; Bawendi, G. M. Angew. Chem., Int. Ed. 2006, 45, 5796. (2) Jang, H. S.; Yang, H.; Kim, S. W.; Han, J. Y.; Lee, S.-G.; Jeon, D. Y. Adv. Mater. 2008, 20, 2696. (3) Jang, E.; Jun, S.; Jang, H.; Lim, J.; Kim, B.; Kim, Y. Adv. Mater. 2010, 22, 3076. (4) Kloepfer, A. J.; Bradforth, E. S.; Nadeau, L. J. J. Phys. Chem. B 2005, 109, 9996. (5) Reiss, P.; Protière, M.; Li, L. Small 2009, 5, 154. (6) Lim, J.; Jun, S.; Jang, E.; Baik, H.; Kim, H.; Cho, J. Adv. Mater. 2007, 19, 1927. (7) Pellegrini, G.; Mattei, G.; Mazzoldi, P. J. Appl. Phys. 2005, 97, 073706. (8) Wang, L. W.; Zunger, A. Phys. Rev. B 1996, 53, 9579. (9) Inerbaev, M. T.; Masunov, E. A.; Khondaker, I. S.; Dobrinescu, A.; Plamada, A.-V.; Kawazoe, Y. J. Chem. Phys. 2009, 131, 44106. (10) The number of atoms (y) for the CdSe sphere is calculated in each diameter (x: 2.0∼4.4 nm with 0.2 nm interval), and the equation is obtained by exponential fitting, y = 54.90*exp(x/1.28) − 54.90. The sphere strucutres were built by cutting out of the bulk zinc blend strucutre, centered in the middle of the Cd−Se bond. (11) Jin, Y.-G.; Chang, J. K. J. Korean Phys. Soc. 2002, 40, 406. (12) Yu, W. W.; Qu, L.; Guo, W.; Peng, X. Chem. Mater. 2003, 15, 2854. (13) Valerini, D.; Creti, A.; Lomascolo, M.; Manna, L.; Cingolani, R.; Anni, M. Phys. Rev. B 2005, 71, 235409. (14) Al Salman, A; Tortschanoff, A.; Mohamed, B. M.; Tonti, D.; van Mourik, F.; Chergui, M. Appl. Phys. Lett. 2007, 90, 093104. 11795
dx.doi.org/10.1021/jp3024217 | J. Phys. Chem. C 2012, 116, 11792−11796
The Journal of Physical Chemistry C
Article
(15) Briggs, D.; Seah, M. P. Practical Surface Analysis, 2nd ed.; John Wiley and Sons: New York, 1993; Vol. 1. (16) Wagner, D. C.; Moulder, F. J.; Davis, E. L.; Riggs, M. W. Physical Electronics Division, Perking-Elmer Corporation, end of the book, 1979. Onyiriuka, C. E. J. Non-Cryst. Solids 1993, 163, 268. (17) Dimitrov, R.; Bonev, I. Thermochim. Acta 1986, 106, 9. (18) Siriwardane, V. R.; Woodruff, S. Ind. Eng. Chem. Res. 1995, 34, 699. (19) Yang, H.; Holloway, P. H.; Cunningham, G.; Schanze, K. S. J. Chem. Phys. 2004, 121, 10233. (20) Xie, R.; Kolb, U.; Li, J.; Basché, T.; Mews, A. J. Am. Chem. Soc. 2005, 127, 7480.
11796
dx.doi.org/10.1021/jp3024217 | J. Phys. Chem. C 2012, 116, 11792−11796