Bright Visible Light Extraction from Amorphous Silicon Nitride Heterojunction Pin Diode

The fabrication of monolithic optoelectronic devices requires a silicon (Si) based efficient light source compatible with the Si electronics.¹ Plasma enhanced chemical vapor deposited (PECVD) hydrogenated amorphous silicon nitride (a-SiN:H) is one of the candidates: it exhibits high radiative recombination rates owing to the relaxation of the momentum conservation law² and its band gap engineering from 1.8 to 4.8 eV renders color adjustments from near infrared to UV.³ This material can be deposited over large surfaces at very low costs below 523 K, making it suitable for both nanophotonics and large area electronics.

Although photoluminescence (PL) analyses indicate that a-SiN:H is a potentially efficient light source,^4–6 the electroluminescence (EL) intensity measured from the corresponding thin film light emitting diodes (TFLEDs) remains below the detection limit of the naked eye due to the poor current injection efficiency and the conduction difficulty through the a-SiN:H layer.^{3, 7} As a solution, Si nanocrystallites can be formed within the a-SiN:H matrix to increase the density and the mobility of the charge carriers.⁸ The creation of nanocrystalline Si (nc-Si) islands within the luminescent active region is usually managed by post-deposition annealing treatments via thermal and/or optical stimuli,^9–13 which lead to size limitation and cost penalty. Another way of nanocrystallization is the Joule heating induced forming process (FP) achieved by the application of sufficiently high forward bias to a fresh (i.e. as grown, virgin) diode.

FP and the accompanying strong EL emission was first observed in the early 1960s in Al/AlO/Au devices.¹⁴ In 1975, this phenomenon was reported for a Si based device, Al/SiN/Al, for the first time.^{15, 16} More recently, similar observations were only partially mentioned in the works focusing on the dielectric properties of gate insulators.^17–20 The fabrication of TFLEDs was never considered since EL (if observed) was not uniformly emitted from the whole diode surface, but from several spots. Lately, Pelant et al. reported considerable lateral propagation of the formed region in their a-Si:H homojunction diodes, and they clearly demonstrated by Raman measurements that the formed regions include nc-Si islands.^{21, 22} Stenger et al. presented very similar Raman results from their polymorphous silicon carbon (pm-SiC:H) homojunction diode and they photographed strong orange-red EL after FP.²³ In this work, FP has been applied to a-SiN:H based diode for the first time.

The trilayer structure was fabricated on Cr-coated glass substrate in a single PECVD cycle. The optimized deposition conditions of the and nc-Si:H doped layers of thicknesses 100 and 50 nm, respectively, may be found elsewhere.^24–27 The luminescent intrinsic a-SiN:H layer of thickness 30 nm was sandwiched between the carrier injecting doped layers using 45 sccm of ammonia and 5 sccm of silane under 0.5 Torr at 22 mW/cm and at 523 K. Next, 300 nm thick indium tin oxide (ITO) dot contacts of surface area 7.8 ×10 cm were coated as window electrodes through a shadow mask by sputtering at room temperature. Finally, the dots were separated from each other by reactive ion etching to form mesas (inset of Fig. 1).

Zoom In Zoom Out Reset image size

Time-resolved current density was measured during FP by a stabilized DC power supply. The lateral propagation of the formed region was photographed by a commercial camera. The EL spectra were taken by both photomultiplier tube (PMT) and Si photodiode (Si PD), and were matched after the responsivity corrections for each detector. The absorbance of the emitted light by the doped layers was analyzed using their spectral absorption coefficient determined from transmittance measurements. To discuss the luminescence mechanism after FP, PL spectrum of a fresh a-SiN:H film, deposited on crystalline Si wafer under the same conditions with the intrinsic layer of the diode at hand, was measured using the 325 nm line of a HeCd laser as the excitation source.

Figure 1 presents the time evolution of the current density during the application of 13 V forward bias stress on the diode schematically depicted in the inset. At the very moment the bias stress is switched on, visible light emission emerges from an arbitrary point on the diode surface. Within the first few seconds after FP begins, the current density increases from 4 to 6 A/cm together with the lateral propagation of the light emitting region. Under 13 V, the major part of the diode is formed in 10 s and the electroforming of the whole diode lasts around 90 s. The process is irreversible, stable under mechanical, electrical and optical disturbances, and identically repeatable on each fresh diode. Once the transformation is accomplished, the diode becomes a 'thin film solid state lamp' which can be switched on and off rapidly.

The time duration until FP finishes depends mainly on the magnitude of the applied forward bias stress, e.g. the diode is completely electroformed within few seconds when 14 V is applied, whereas, it takes more than half an hour when the process is run under 9 V.

To present the lateral propagation of the formed region by real-time photography, one of the fresh diodes shown in Fig. 2a was subjected to 10 V bias stress and its photographs were taken periodically. In Fig. 2b, the forming initiates at an arbitrary point not necessarily under the measurement probe. Then, the formed region expands starting from this initial small area until the forming of the whole diode is completed as given in Figs. 2c, 2d, 2e, 2f, 2g and 2h. Figure 2i demonstrates the zoomed-out appearance of this TFLED in its ON state. The brightness of EL is strong enough to be seen by naked eye even in a moderately illuminated room.

Zoom In Zoom Out Reset image size

Although FP and the accompanying light emission are also observed in the diodes with other window electrodes like Al and Cr, the formed region is successfully propagated along the whole diode surface only when ITO is used as the window electrode.²⁴ This is believed to be due to the difference between the thermal conductivities of metals and ITO. It has been verified by Raman and X-ray diffraction measurements.^{21–23, 26} that FP is a temperature related phenomenon leading to quasi-melting of a-Si:H clusters and their simultaneous crystallization.¹⁷ Therefore, the quality of FP depends on whether the heat is stored in the system or dissipated towards the surroundings. In the case of metal top electrodes which have very high thermal conductivities, e.g., 94 W/K·m for Cr and 234 W/K·m for Al, the heat produced within the structure during FP cannot be stored. The formed region remains point-like and further forming is prevented by the dissipation of the excess heat. As for the ITO electrode, it is a transparent semiconductor with high electrical conductivity. Considering that its thermal conductivity of 11 W/K·m is much lower than that of metals, the heat may be stored in the system and laterally distributed along the whole diode instead of its vertical dissipation towards the air. Consequently, the excess heat of the system can be used to carry on FP until the diode is completely formed.

The EL spectrum of the formed diode can be seen in Fig. 3. It spans the whole visible region together with near infrared emission. The interference fringes are due to the multiple reflections of the emission prior to its escape from the diode. As interpreted from the spectral absorption coefficient of the doped layers provided in Fig. 3, some fraction of the emitted light is, unfortunately, re-absorbed. Discussion on the luminescence mechanism is rather proper after the correction of the EL spectrum for the absorbance of the emitted light by the doped layers. With the assumption that the intensities of the original emissions directed to the and layers are the same, the corrected EL spectrum is plotted as shown in Fig. 3. There seems to be two main emission bands located at 530 and 800 nm. The band at shorter wavelengths almost overlaps with the PL spectrum of the a-SiN:H film identical to that used as the intrinsic layer of the structure, i.e. is 1.3 and optical gap is 4.8 eV. This band corresponds to the well-known band tail luminescence of a-SiN:H.^{4–6, 28, 29} The emission at longer wavelengths may stem from several sources. First, the interfaces of our heterojunction diode probably include a broad distribution of energy levels. In addition, FP may induce new energy states within the forbidden gaps of both intrinsic and doped layers. These states may be utilized as radiative recombination centers under efficient current injection.

Zoom In Zoom Out Reset image size

In conclusion, a-SiN:H based heterojunction diode has been transformed into a high brightness TFLED via application of a sufficiently high forward bias stress at room temperature. The method is cost-effective and lasts at most few seconds until the whole diode is electroformed under a calibrated voltage of 14 V. The thickness and the surface area of an eventual optoelectronic device fabricated by the integration of these TFLEDs with either thin film transistors^{30, 31} or metal-oxide-semiconductor field effect transistors would be determined only by the dimensions of the substrate chosen.