Jpn. J. Appl. Phys. Vol. 41 (2002) pp.L326–L328 Part 2, No. 3B, 15 March 2002
c
2002 The Japan Society of Applied Physics
The Roughness-Enhanced Light Emission from Metal-Oxide-Silicon Light-Emitting Diodes
Using Very High Vacuum Prebake
Min-Hung LEE, Kuan-Fu CHEN, Chang-Chi LAI, Chee Wee LIU∗, Woei-Wu PAI1, Miin-Jang CHEN2and Ching-Fuh LIN2
Department of Electrical Engineering and Graduate Institute of Electronics Engineering, National Taiwan University, Taipei, Taiwan, R.O.C.
1Center for Condensed Matter Sciences, National Taiwan University, Taipei, Taiwan, R.O.C.
2Department of Electrical Engineering and Graduate Institute of Electro-Optical Engineering, National Taiwan University, Taipei, Taiwan, R.O.C.
(Received November 5, 2001; accepted for publication December 20, 2001)
The oxide roughness of metal-oxide-silicon diodes can be intentionally controlled by the very high vacuum pre-bake and the growth conditions during rapid thermal oxidation. Both surface and Si/oxide interface have the similar magnitude of roughness measured by atomic force microscopy, indicating the conformal growth of oxide. At accumulation bias (positive gate bias), the holes tunnel from gate electrode to n-type Si through the ultrathin oxide, and recombine with the electrons in the accumulation region radiatively if phonon scattering and roughness scattering provide the necessary momentum. The light emission intensity increases with increasing oxide roughness. Strong electroluminescence with an external quantum efficiency of∼ 2 × 10−6at room temperature was observed from a rough metal-oxide-silicon tunneling diode. [DOI: 10.1143/JJAP.41.L326]
KEYWORDS: metal oxide silicon diode, roughness, electroluminescence, electron-hole plasma recombination, external quantum efficiency
The Si-based optoelectronic devices attract great attention to process optical signal on the silicon chips1) with poten-tial applications such as optical interconnects2) and optical communication. Based on pn junctions, two different Si light emitters have been reported. A narrow-band infrared emitter with photon energy of silicon bandgap was implemented us-ing a pn junction under forward bias with an external quantum efficiency of∼ 10−4.3)Very recently, the band-edge emission from a simplified structure of pn junction with similar effi-ciency is also reported.4) Another broad (450–850 nm) visi-ble light emitter was realized using an avalanche pn diode at reverse bias with the reported external quantum efficiency of ∼ 10−8 3)and∼ 10−6.5)For metal-oxide-silicon (MOS) struc-tures, both the infrared band-edge emission2) and the broad visible emission6) are also observed at accumulation bias. Note that the same MOS tunneling structure biased at deep depletion region can be also used as a photodetector.7) Al-though there is a concern of the reliability of utrathin oxide in the MOS structure, the incorporation of deuterium into the oxide has been demonstrated to improve the degradation of light emission intensity.8)Due to the indirect bandgap nature of Si, additional momentum is required for the light emis-sion process. The phonon scattering9)and the oxide roughness scattering10)can provide the additional momentum for the ra-diative electron-hole recombination at oxide/Si interface. In this letter, an effective way to increase the oxide roughness by very high vacuum pre-bake before rapid thermal oxidation is demonstrated. The light emission intensity increases with the magnitude of oxide roughness up to the external quan-tum efficiency of 2× 10−6for the MOS light-emitting diode (LED).
The ultrathin gate oxide was grown by rapid thermal ox-idation on 1–10-cm n-type wafers at the 800–900◦C. The gas flows were 500 sccm nitrogen and 500 sccm oxygen at a reduced pressure. After a HF dip, the wafer was transferred to the process chamber through the load-lock chamber, and was in situ baked in the very high vacuum (< 3 × 10−6Torr, maintained by a turbo pump) at 1000◦C, and then, the wafer
∗Corresponding author. E-mail address: [email protected]
was baked in hydrogen at 1000◦C for 1 min. Note that this very high vacuum bake should yield a rough and clean Si sur-face.11, 12) The oxide growth was conducted after the hydro-gen bake, and subsequently nitrohydro-gen post-oxide-anneal was performed at 900◦C for 10 min. The oxide thickness is mea-sured by ellipsometry. The roughness is meamea-sured by atomic force microscopy (AFM) and its magnitude is defined as the root-mean-square of height variation. The surface roughness on the oxide is measured right after the oxide growth, and the Si/oxide interface roughness is measured immediately after the oxide removal of the same samples by HF dip. The ellip-sometry confirms little native oxide growth after AFM mea-surements for both conditions. The surface roughness and in-terface roughness have similar magnitude (Fig. 1). This indi-cates the conformal growth of the rapid thermal oxidation.13) Note that our AFM can not resolve the spatial spectrum vari-ation of these samples. However, ultrathin thermal oxide has similar bandwidth in the interface roughness spectra as sured by atomic force microscopy(AFM) and mobility mea-surement.14) The PMOS (p-channel) LED has Al gate elec-trodes with circular areas defined by photolithography.
An-Fig. 1. The plot of surface oxide roughness vs interface roughness between Si/oxide. The linear relationship indicates the conformal growth of rapid thermal oxidation.
Jpn. J. Appl. Phys. Vol. 41 (2002) Pt. 2, No. 3B M.-H. LEEet al. L327
Fig. 2. The current components of a PMOS tunneling diode. The hole cur-rent is smaller than the electron curcur-rent. The hole curcur-rent tunneling to the interface states cannot contribute the radiative recombination. Neither does the electron current component.
Fig. 3. The EL spectra of PMOS tunneling diodes with different oxide roughness. The oxide thickness and voltage are indicated in the parenthe-ses.
other Al contact is on the back of the wafer.
At the positive gate bias, the holes in the Al electrode tun-nel to the n-Si through the ultrathin oxide, and meanwhile the electrons in the accumulation region of n-Si tunnel to the Al electrode (Fig. 2). Note that the holes tunneling to the inter-face states as shown in the Fig. 2 cannot contribute the ra-diative band-edge recombination. Due to barrier height dif-ference (3.1 eV for electrons and 5.8 eV for holes15, 16)), the hole tunneling current of PMOS devices is smaller than the electron tunneling current. The tunneling holes from Al to n-Si in PMOS can be scattered by oxide roughness and ra-diatively recombine with electrons in the accumulation layer with the assistance of phonons.10)Figure 3 shows the electro-luminescence from the PMOS devices with the device area of 4× 10−2cm2 for different oxide roughness. The light emission of the Al gate MOS devices is collected from the edge of the gate electrode at the drive current of 100 mA. All the emission spectra can be fitted by the
electron-hole-Fig. 4. The external quantum efficiency vs oxide roughness. The error bar is the standard deviation of external quantum efficiency for a set of devices.
plasma recombination model, which is the convolution be-tween electron and hole populations by considering the en-ergy conservation.17)Note that the surface roughness and the interface roughness have similar magnitude due to the con-formal growth of the oxide. The emission intensity of the MOS devices increases with the oxide roughness, while the oxide thickness of each device is different. Both the phonon scattering and oxide roughness scattering can conserve the momentum during the electron-hole radiative recombination at oxide/Si interface, but at a fixed device temperature, the phonon population should be constant and oxide roughness is the main factor to affect the emission efficiency. Note that the similar line widths in the electroluminescence spectra (Fig. 3) indicate the similar device temperatures.9)To confirm this ob-servation, a set of PMOS devices have been measured for each oxide roughness, and the statistical results of external quantum efficiency at the current of 100 mA vs oxide rough-ness are shown in Fig. 4. The error bar in Fig. 4 is the stan-dard deviation of the device set for efficiency measurements. Although there are some variations of external quantum ef-ficiency, the trend definitively shows that the efficiency in-creases with increasing roughness. The spectral distribution of roughness cannot be accurately measured in our AFM due to the lateral resolution of the AFM tips. In the litrature,14)the spectral distribution of the roughness have no consistent re-sults between AFM measurement and mobility measurement, but the spectral distributions are similar for different rough-ness magnitudes in both measurements. It is reasonable to as-sume that the spatial spectra in our samples are similar and the spectral distribution of oxide roughness has smaller effect on the enhancement of emission intensity than the magnitude of the roughness.
A particular PMOS device with oxide roughness of 1.5 nm shows an external quantum efficiency of 2.5 × 10−6 at a 100 mA, and the efficiency increases as drive current in-creases (Fig. 5). Based on the model calculation18) of tun-neling current (Fig. 2), the ratio of hole current to electron current increases with large gate bias, and only the hole tun-neling current can contribute the light emission. This may be responsible for the increase of emission efficiency with the increasing gate current.
We have demonstrated that the correlation between oxide roughness and electroluminescence intensity in the PMOS tunneling diodes. The high vacuum pre-bake seems to be an
L328 Jpn. J. Appl. Phys. Vol. 41 (2002) Pt. 2, No. 3B M.-H. LEEet al.
Fig. 5. The external quantum efficiency vs drive current of a particular PMOS LED.
effective method to produce oxide roughness, and the MOS LED emits more light after such very high vacuum pre-bake process.
This work is supported by TSMC and National Science Council, Taiwan, R.O.C. (002-017, 89-2218-E-002-054).
1) K. D. Hirschman, L. Tysbekov, S. P. Duttagupta and P. M. Fauchet: Nature 384 (1996) 338.
2) C. W.Liu, M. H. Lee, C. F. Lin, I. C. Lin, W. T. Liu and H. H. Lin: Int. Electron Devices Meet. Tech. Dig., 1999, Washington D.C., p. 749. 3) J. Kramer, P. Seitz, E. F. Steigmeier, H. Auderset and B. Delley: Sens.
& Actuat. A 37/38 (1993) 527.
4) W. L. Ng, M. A. Lourenco, R. M. Gwilliam, S. Ledain, G. Shao and K. P. Homewood: Nature 410 (2001) 192.
5) L. W. Snymandu, M. Plessis, E. Seevinck and H. Aharoni: IEEE Elec-tron Device Lett. 20 (1999) 614.
6) C. W. Liu, M. H. Lee, S. T. Chang, W. T. Liu, M.-J. Chen and C.-F. Lin: Appl. Phys. Lett. 77 (2000) 4347.
7) C. W. Liu, W. T. Liu, M. H. Lee, W. S. Kuo and B.-C. Hsu: IEEE Elec-tron Device Lett. 21 (2000) 307.
8) C. W. Liu, C.-H. Lin, M. H. Lee, S. T. Chang, Y. H. Liu, M.-J. Chen and C.-F. Lin: Appl. Phys. Lett. 78 (2001) 1397.
9) C. W. Liu, M.-J. Chen, I. C. Lin, M. H. Lee and C.-F. Lin: Appl. Phys. Lett. 77 (2000) 1111.
10) C. W. Liu, M. H. Lee, M.-J. Chen, C.-F. Lin and M. Y. Chern: IEEE Electron Device Lett. 21 (2000) 601.
11) X. Xu, R. T. Kuehn, M. C. Ozturk, J. J. Wortman, R. J. Nemanich, G. S. Harris and D. M. Maher: J. Electronic Mater. 22 (1993) 335. 12) F. W. Smith and G. Ghidini: J. Electrochem. Soc. 129 (1982) 1300. 13) V. Tsai, X.-S. Wang, E. D. Williams, J. Schneir and R. Dixson: Appl.
Phys. Lett. 71 (1997) 1495.
14) A. Pirovano, A. L. Lacaita, G. Ghidini and G. Tallarida: IEEE Electron Device Lett. 21 (2000) 34.
15) W.-C. Lee and C. Hu: VLSI Tech. Dig., 2000, Hawaii, p. 198. 16) Y. Taur and T. H. Ning: Fundamentals of Modern VLSI Devices
(Cambridge University Press, Cambridge, 1998) p. 59.
17) C. W. Liu, M. H. Lee, M.-J. Chen, I. C. Lin and C.-F. Lin: Appl. Phys. Lett. 76 (2000) 1516.
18) S. Keeney, F. Piccini, M. Morelli, A. Mathewson, C. Lombardi, R. Bez, L, Ravazzi and D. Cantarelli: Int. Electron Devices Meet. Tech. Dig., 1990, San Francisco, p. 201.
