Rare earth implanted MOS structures: Advantages and drawbacks for optoelectronic applications

Integrated photonics is a key technology of the 21st century, and Si-based photonic components are of special interest as they propose an easy integration into the CMOS platform. An essential building block is the electrically driven light emitter (LE) which, however, is difficult to realize. Among the different approaches rare earth (RE) implanted MOS structures feature a high conformity with standard CMOS processes combined with the excellent optical properties of RE elements.
However, despite an intense research for more than 20 years in the field of Si-based light emission, the results are mixed. With respect to important key parameters like power efficien-cy, Si-based LE are not yet able to compete with their counterparts based on compound or organic semiconductors. The following contribution is focused on RE implanted MOS structures and discusses the different problems to turn them into long-living, efficient LE.
Such a LE typically consists of a Si substrate, a dielectric stack and a transparent electrode (Abb. 1). The layers of the dielectric stack fulfil several tasks like hosting the luminescence centres or providing a buffer against early electrical breakdowns. Usually, the injected charge carriers have to overcome substantial potential barriers requiring high electric fields for operation. The electroluminescence (EL) is based on radiative 4f intrashell transitions within the trivalent RE ion, although there are exceptions involving a 5d electron (e.g. Ce). For this reason oxidic materials, especially SiO2, are suitable as host matrices, but there are also successful reports where RE elements are embedded in other materials, e.g. silicon nitride.

Fig. 1: Basic scheme of a RE-implanted MOS device

At the beginning, a short view to the EL excitation mechanism may be helpful. From the physical point of view, there are two main types of excitation. At first, under favourable conditions electrons can be accelerated to high energies which can be transferred by inelastic scattering to the lumines¬cence centres. This mode can be fairly efficient, depending on the fraction of hot electrons in the average electron energy distribution, but inherently contains the seed of destruction: hot electrons are also efficient in creating new defects and degrading oxides. Different to this, charge carrier recombination releases the band gap energy of the host semiconductor at maximum, which prevents degradation and limits the different types of potential luminescence centres. In case of Si-based LE the efficiency is often less than in case of hot electron excitation, but there are ongoing research activities to improve the situation.
Starting the discussion with the quest for the most efficient RE-implanted MOS device, there are at least three strategies to be pursued: (i) to increase the excitation cross section, (ii) to increase the internal quantum efficiency, and (iii) to increase the outcoupling efficiency.
In most cases the first strategy was pursued, and indeed, this strategy probably offers the greatest potential if the efficiency is far away from reasonable levels. Among the most popular ideas is the use of Si nanoclusters to enhance the excitation cross section of RE ions, especially of Er. Although this recipe was quite successful in case of photoluminescence, it was found that often only a small percentage of Er is excited [1]. The problem was partially solved in the last years by substantial changes in the composition of the dielectric stack. The use of Ge nanoclusters gave the unexpected result that the RE ions were pumping the Ge nanoclusters instead vice versa [2]. At present, in case of EL the use of a second RE element pumping the first one was more efficient, as demonstrated for Gd pumping the EL of Ce [3].
In case of hot electron excitation another problem appears, namely the existence of a dark zone close to the injecting interface. This is the region where electrons are already accelerated but not yet have enough energy to excite the RE ions. In case of Tb-implanted MOS devices it was found that this zone extends up to 20 nm into the gate oxide [4] which limits the scalability to thinner devices and to lower operation voltages.
To improve operation lifetime, the use of LOCOS (LOCal Oxidation of Silicon) structures and dielectric buffer layers made of SiON [5] can prevent early electric breakdowns. However, these strategies cannot solve the fundamental problem of oxide degradation by hot electrons. In addition, the EL often exhibits degradation over lifetime due to the charging of defects and RE clusters [6]. Experiments with MOS structures, where SiO2 is replaced by materials with less hot electrons, show that in fact the operation lifetime is strongly enhanced, but unfortunately at the expense of efficiency. A possible solution could be the shift of the excitation mode to charge carrier recombination, e.g. if electrons and holes are both injected during AC excitation.
In summary, the advantages and drawbacks of RE-implanted MOS devices were discussed. At present, the current status of these devices allows only a few applications where efficiency and operation lifetime is not of major importance, as for some sensor applications. However, some possible solutions to reduce or overcome these problems were addressed.
References
[1] B. Garrido, C. García, P. Pellegrino, D. Navarro-Urrios, N. Daldosso, L. Pavesi, F. Gourbilleau, and R. Rizk, Appl. Phys. Lett. 89 (2006) 163103
[2] L. Rebohle, A. Kanjilal, W. Skorupa, and M. Helm, Opt. Mat. 33, Issue 7 (2011) 1075.
[3] J.M. Sun, S. Prucnal, W. Skorupa, M. Helm, L. Rebohle, and T. Gebel, Appl. Phys. Lett. 89, 091908 (2006)
[4] L. Rebohle, J. Lehmann, S. Prucnal, J. M. Sun, M. Helm and W. Skorupa, App. Phys. B 98:2 (2010) 439
[5] J.M. Sun, L. Rebohle, S. Prucnal, M. Helm, W. Skorupa, Appl. Phys. Lett. 92, 071103 (2008)
[6] A. N. Nazarov, S. I. Tiagulskyi, I. P. Tyagulskyy, V. S. Lysenko, L. Rebohle, J. Lehmann, S. Prucnal, M. Voelskow, and W. Skorupa, J. Appl. Phys. 107, 123112 (2010)