• No results found

On the switching speed of SOI LEDs

N/A
N/A
Protected

Academic year: 2021

Share "On the switching speed of SOI LEDs"

Copied!
1
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

On the switching speed of SOI LEDs

Jurriaan Schmitz, Roel de Vries, Cora Salm, Tu Hoang, Ray Hueting, and Jisk Holleman

MESA+ Institute for Nanotechnology, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands

(2)

1. Abstract

Recently, we presented a novel design for a silicon LED in

SmartCUT™ SOI

wafers. It exhibits a

record quantum

efficiency for SOI-based silicon LEDs and opens the way to the integration of light emitters in a VLSI process on SOI. In this paper, we present first

experimental and

modeling results

showing that this new design has the potential to switch on and off faster than 1 MHz – sufficiently fast for several commercial applications such as an integrated optocoupler.

2. Introduction

The design of our LED is presented in Figure 1.

The light-emitting

central region is a lowly p-type doped silicon volume of 5 x 60 x 0.15 µm3.

Fig.1: Schematic

cross-section of the SOI LED under study in this work. Reprinted from [1].

Two highly doped (n+ and p+) regions provide the carriers to be recombined in the lowly doped central region. The highly doped regions are separated

from the central region by gated, ultra-thin (down to 5 nm) silicon slabs. These allow the injection of one type of carriers, while they form a barrier for the other type of carrier. As a result, carriers injected in the central region have no other option than to recombine. Through the use of high-purity silicon with a low doping level, an injection level around 1017 cm-3, and

well-passivated silicon surfaces (using thermal oxidation), the non-radiative recombination rate in the central volume is not too high, and the probability for radiative recombination becomes significant. The device exhibits an

external quantum

efficiency of 1.4 x 10-4

at room temperature, for emission around 1150 nm (the silicon B-B peak) – comparable to the best bulk-silicon LEDs when no special surface modifications are made to enhance the outcoupling [2]. This device exhibits a power output up to 0.25 µW [1].

3. Switching

speed

For integrated optics, a switching light source has great advantages over a continuous light source that requires an external switching device. Our LED has been designed to be fast-switching (apart from the rather large dimensions of the present prototypes, due to our lab capabilities). However, the LEDs compactness, and its

unconventional

emission wavelength complicate a direct measurement of the

on/off switching

behavior of the LED with the instruments at hand. In this section, we

present indirect

estimates of the switching speed.

On-switching

The on-switching speed

is intrinsically

determined by the RC-delay occuring in the device, when the diode is switched to forward-bias. (That is, if V in figure 1 is switched from low to high and back to create light modulation;

alternatively, the light can be modulated with Vb1 and/or Vb2!) It is not

correct to assume that the device will start emitting light as soon as current starts to flow, so electrical measurements are not conclusive to determine the on-switching behaviour. However, we can assume that, once the excess carriers are inside the central region, they will mix almost instantaneously given the device dimensions, the low field in the central region and the thermal velocity. The total capacitance of the diode sketched in figure 1 consists of three major components: the diffusion capacitance, the gate capacitances, and the capacitance towards the substrate. (The latter is an order of magnitude less than the gate capacitance.) A rough calculation of the capacitances and resistances involved can

give an indication of on-switching speed. When the positive terminal V is switched from 0 to 1 V, charge needs to be supplied to charge up the gate capacitance Vb1

of 1 pF. The external resistance of the device is approximately 1 kΩ. Then, charge is to be supplied to the central region to raise the

carrier density.

Assuming that the central region has np = ni2 in off-state, and n = p

= 5x1017 cm-3 in the

on-state, caused by a change of 1 V in the forward bias, the diffusion capacitance is first-order estimated at 7 pF with the device dimensions mentioned in Section 2. Combined

with the gate

capacitance, this would lead to an on-switching RC-time of 8 ns.

Off-switching

The off-switching can be studied with electrical measurements. As a diode is switched from forward-bias to reverse, a short transient current will be observed originating in the excess carriers flowing out of the depletion region. This so-called reverse recovery current [4,5] of the diode gives a direct indication how long it takes to remove the excess carriers from the central region of our LED. It can be readily assumed that light emission will stop instantly when the pn-product has dropped back to ni2 [6].

Therefore, the reverse

recovery current

indicates the speed of off-switching of our LED.

(3)

Figure 2 shows the recovery current of our LED, when the forward-voltage of the diode is switched from high to zero volts. The

measurement was

carried out using a Keithley 4200 with pulsed-IV extension, in oscilloscope mode. The fast decay of the

recovery current

indicates that the LED will switch off within ~200 ns.

Fig.2: The off-switching

behaviour of an SOI LED.s and |Vb1| = Vb2 = 1 V. The

diodes are pumped by a voltage pulse: pulse width of 250 ns; pulse decay of 10 ns. The on-state bias was 3.5 V, off-state 0 V.

4. Device

modeling

The switching behaviour can be systematically investigated using a modeling package such as ATLAS [3]. At the time of writing, we can only report the first results of SOI pin-diodes without the gated, thinned-down access regions. The simulator reproduces the dc I-V characteristics of this diode (fig. 3).

Fig. 3: Simulated and

measured I-V characteristics of a lateral SOI pin diode with 30 µm intrinsic region.

The overall picture of on- and off-switching of such a device is in line with our findings reported in Section 3 (Fig. 4). In accordance with gated-pin-diodes reported earlier [7] the low voltage regime the current of our SOI diode is determined by Shockley-Read-Hall recombination in the bulk and at the Si/SiO2

interfaces. The high voltage regime is determined by series resistance of the charge plasma in the intrinsic region.

We expect to report considerable further progress with both

simulations and

measurements at the workshop.

Fig. 4: Simulated

switching behaviour of a lateral SOI pin diode with a 30 µm intrinsic region, showing a 1/e on-switching time of about 90 ns and 1/e off-switching time of 50 ns. The on-state

bias is 2.0 V, off-state 0 V.

5. Conclusions

In this paper we address the switching speed of an SOI-based LED with emission efficiency comparable to bulk Si LEDs. On the basis of calculations, simulations

and electrical

measurements, we made plausible that the device can reach a switching speed around 1 MHz. Further work will include the direct optical switching measurement, and device improvement (shorter access regions) to further increase the speed.

6.

Acknowledgement

s

We would like to thank Tom Aarnink for assistance with the device manufacturing, Gert ‘t Hooft, Martin Kittler and Alfred Driessen for fruitful discussions, and Lee Stauffer and coworkers for their assistance with the recovery current measurement. This work is supported by the Dutch Technology Foundation STW (project 6159 HELIOS). References [1] T. Hoang, J. Holleman and J. Schmitz, EDL 28 (5) pp. 383-385, May 2007.

[2] M. A. Green, J. Zhao, A. Wang, P. J. Reece, and M. Gal, Nature, vol. 412, no. 6849, pp. 805–808, Aug. 2001. [3] ATLAS device simulation software, Silvaco Int., v 5.13.0.C, 2007

[4] W. Gerlach et al., Phys. Stat. Sol. (a), 13, 277 (1972).

[5] B. J. Baliga, Modern power devices, Wiley, New York 1987; pp.228-232. [6] E. F. Schubert, Light emitting diodes, 2nd edition, Cambridge University Press, 2006. [7] T. Ernst, et al., IEEE TED (46) 7, pp. 1503-1509, 1999.

(4)

Referenties

GERELATEERDE DOCUMENTEN

My hypothesis is that, contrary to Mafouné, Tansila has a spreading rule which causes the tone of the first syllable of the noun to spread to its second syllable.. The tone of

The transition of prophecy from an oral to a written form did not merely lead to the disappearance of oral prophets, but it also led to the emergence of the

Pagina 4 van 5 Zorginstituut Nederland Bedrijfsdiensten Automatisering Onze referentie 2020029926 trastuzumab-emtansine (Kadcyla®), tweede bespreking. 27

vlekken? Bij bemonstering aan het begin en aan het eind van de lichtperiode moet dit verschil duidelijk worden. Dit is onderzocht bij gewas, verzameld in januari 2006. Bij de

Maar wat het verkeersaanbod betreft lijkt er in Noord-Holland slechts weinig verschil tussen die twee weekendnachten te bestaan, zodat er in de vrijdagnacht ook in

Twee wandfragmenten met zandbestrooiing in scherven- gruistechniek zijn mogelijk afkomstig van een bui- kige beker met een lage, naar binnen gebogen hals (type Niederbieber 32a),

Het kunnen foto’s zijn van mensen en gebeurtenissen uit het eigen leven, bijvoorbeeld van een cliënt of medewerker, maar ook ansicht- kaarten of inspiratiekaarten zijn hier

Hogere waterpeilen leiden tot hogere kosten, omdat de landbouw schade ondervindt en omdat bij de situatie van plasdras fors moet worden geïnvesteerd in het drooghouden van bebouwing