Optical switching in metal tunnel-insulatorn-p+ silicon devices

S. Moustakas, J.L. Hullett, R.B. Calligaro, A.G. Nassibian and D.N. Payne

Indexing terms: Metal-insulator-semiconductor devices, Photoelectric devices, Semiconductor switches

Abstract: This paper considers the mechanism of optical switching and the possible utilisation of the metaltunnel-insulator n-p* silicon device in optical communication systems. The pertinent design approaches aredescribed. Under optical excitation, photo holes and electrons generated in the surface depletion region, orwithin diffusion range, will eventually be separated by the electric field and produce an increment in theforward current. Those hole-electron pairs generated in the junction region, or within diffusion range,produce a photovoltaic increase in the p*-n junction bias. Switching is induced optically, as it is electrically,by the build up of holes at the insulator-semiconductor interface. This paper employs the 1-dimensionaldiffusion equation to derive the light-generated minority carrier distributions and diffusion currents in theneutral n and p* regions, together with the currents in the surface and p*-n junction depletion regions.The calculated values of both the drift and diffusion currents compare favourably with those observedexperimentally.

1 Introduction

The electrical switching mechanism of the metal-tunnelinsulator-«-p+ silicon structure has recently been describedby Simmons, El Badry and Chik.1"3 Such a device iscompatible with i.e. fabrication techniques1 and hence haspotential in commercial i.e. digital circuits. The addedability to switch under optical excitation may lead tofurther applications in optical communication systems.4 Inthis paper we consider the mechanism responsible foroptical switching.

Those devices with a lightly doped n -section switchelectrically from the high-impedance off state to the lowimpedance on state when, with increasing forward bias, thedepletion region of the w-section reaches through to thep*-n junction. Up to the point of switching, the normaltendency to form an inversion layer is prevented by holestunnelling through the thin oxide from the semiconductorto the metal electrode. Switching can also be inducedoptically and, depending on the design of the structure, thedevice can be made quite optically sensitive.

The 1-dimensional diffusion equation is employed toderive the light-generated minority carrier distributions andhence diffusion currents in the neutral n and p+ regions.The drift photocurrents in the surface and p+n junctiondepletion regions are also considered. These light-generatedcurrent components are shown to be responsible for both aphotovoltaic increase in the p+n junction voltage and acollapsing of the surface depletion region. Depending onthe applied external bias, the device will then either remainin its off state or switch to its on state.

Experimental verification of the photocurrent levels andthe photovoltaic increase in the p+n junction voltage wasundertaken. For ease of experimentation, an He-Ne redlaser was used in the tests.

Possible improvements in the sensitivity of the device asan optical switch are outlined.

Paper T386S, first received 6th February, and in revised form 4thMay 1979

The authors, with the exception of Dr. Payne, are with the Depart-ment of Electrical & Electronic Engineering, University of WesternAustralia, Nedlands, Western Australia, 6009. Dr. Payne is withthe Department of Electronics, University of Southampton,Southampton, England $09 5NH

SOLID-STATE AND ELECTRON DEVICES, JULY 1979, Vol. 3, No. 4

2 Electrical switching mechanism

A schematic diagram of the physical structure of the deviceis shown in Fig. 1. It comprises a number of layers: analuminium cathode layer evaporated onto a thin tunneloxide, followed by an epitaxially grown «-region on a p+

substrate. Electrical contact is made to the p+-region bymeans of a metal anode. Furthermore, a metal gate elec-trode has been evaporated onto the device adjacent to thecathode. Separation between the two terminals consists ofa thick (nontunnelling) oxide.

metalgate

fieldoxide

metal" cathode

to outercontactfor bonding

gate cathode gate

field oxide

junctiondepletion

region

Fig. 1 Schematic of the physical structure of the device

a Top viewb Cross-sectional view(Not to scale)

85

0308-6968/79/030085 + 09 $01-50/0

The forward I/V characteristic of the silicon structurewith the gate electrode floating is shown in Fig. 2a. Itcomprises three parts: a high-impedance low-current offstate, a negative-resistance switching state, and a low-impedance high-current on state. Furthermore, as thecharacteristic resembles that of a thyristor, the device isknown as a metal-insulator-semiconductor thyristor orm.i.s.t.

Forward biasing of the m.i.s.t. is effected by operatingthe cathode negative with respect to the anode (Fig. 2b).Under this condition, free electrons are swept away fromthe surface and a depletion region is established in the^-section under the cathode. Conversely, holes enteringor generated within this region are swept away to theoxide-silicon interface where they tunnel through the thinoxide into the metal. Thus, as the bias is increased, thenormal tendency to invert the surface is prevented and thesurface depletion region continues to grow towards thep+-n junction. The semiconductor surface is then said to bein deep depletion. The off state current in this mode islimited by generation in the depleted surface region,1"3

and by recombination in the junction depletion region.At a sufficiently high forward bias Vs, the surface

depletion region punches through to the p+-n junctionspace-charge region. Any further increase in the appliedbias voltage will effectively lower the potential-barrierheight of the p+-n junction, causing a rush of holes to beinjected from the p+ substrate into the n-type region, wherethey are swept to the oxide-silicon interface. However, justbefore punchthrough, the voltage across the oxide is onlysufficient to pass the small off-state current. Thus, imme-diately after punchthrough, the oxide field is insufficient

current I

load line

switchingpoint

Voltage V

Vc V,S vapp

-V,vapp

gate

Fig. 2

to allow the relatively large injected hole current to tunnelthrough the thin insulator. Consequently, accumulationof the injected holes takes place at the oxide-silicon inter-face. Furthermore, electron tunnelling from the metal tothe semiconductor condition band has a strong dependenceon the free hole density at the semiconductor surface.5

Those electrons which tunnel through the oxide are thenswept by the surface electric field to the p+-substrate, thuscausing a correspondingly larger number of holes to beinjected from the p+-region to the oxide-silicon interface.The free hole density at the semiconductor surface increasesfurther, thereby causing an ever larger electron tunnelcurrent to flow. Hence, a regenerative positive-feedbackloop with loop gain greater than unity is established in thedevice, giving rise to the following two interacting effects:1

(a) The w-layer begins to move from deep depletiontowards inversion, causing the surface potential and, hence,the voltage across the m.i.s.t. to decrease, and the voltageacross the load resistor to increase.

(b) The field in the oxide begins to increase, allowing alarger current to pass. Thus, a negative resistance region isallowed to develop as the device switches from the off tothe on state.

A steady-state operating point is reached when the loadresistor current, the p*-n junction current and the oxidetunnelling current are equal.

Electrical control of the switching process can be realisedby either injecting or extracting charged minority carriersthrough the gate terminal, or by applying a voltage directlyto the gate.

3 Optically generated currents in the m.i.s.t.

Absorption of light in the m.i.s.t. produces hole-electronpairs at a rate G(x), where6

G(x) = (1)

where 17 is the effective quantum efficiency, $0 is thetotal incident photon flux per unit area per second(photons/cm2 s), and a is the absorption coefficient incm"!. Both T? and a are functions of the optical wavelengthand the semiconductor material.6

Shown in Fig. 3 is a plot of the rate of optical generationof hole-electron pairs as a function of distance from thecathode of the mi.s.t. Here, dox, Wd, Wn and Wj denotethe widths of the tunnel oxide, surface-depletion region,neutral n-region and junction depletion region, respectively.

We now consider the regions of the m.i.s.t. where light-generated current components are produced. These are thesurface and junction depletion regions where drift currentsexist, and the neutral n-region and p+-substrate wherediffusion currents flow.

neutralp*- region

surface I neutraletionj n-region

egion j

metalanode

a I/V characteristic of the deviceb Biasing the m.i.s.t.

p*n-junction

Fig. 3 Rate of generation of hole-electron pairs due to absorptionof light in the m.i.s.t. (not to scale)

86 SOLID-STA TE AND ELECTRON DE VICES, JULY 19 79, Vol. 3, No. 4

3.1 Surface depletion region

Hole-electron pairs created in the surface depletion regionwill immediately be separated by the depletion-layer electricfield. Electrons are swept to the neutral «-region wherethey diffuse towards the p+-n junction space-charge regionand recombine with holes injected into the junctiondepletion region by the small forward bias, while holes areswept to the oxide-silicon interface where they tunnelthrough the thin SiO2 layer. If the device is not triggeredinto the switching mode (see Section 4 for discussion onoptical switching), the generated photocurrent contributesdirectly to the m.i.s.t. off-state current. For a maximumincrease in drive current the surface depletion region widthis made large relative to the light penetration depth a"1.

The current reaching the oxide-silicon interface, due tooptical excitation of hole-electron pairs in the surfacedepletion region, is given by:6

rWrfG(x)dx

JO

he(2)

Here, the total incident photo flux $ 0 was written in termsof the total incident optical power P:

PXhcAL

where h is Planck's constant, c and X are the speed andwavelength of the incident light, respectively, and AL is theilluminated area. For

Wd >1

then

he

and a maximum current increase is achieved.

3.2 Junction depletion region

Holes and electrons generated optically in the junctiondepletion region are swept to the p+-substrate and epitaxialn-layer, respectively. The resulting current component doesnot contribute directly to the m.i.s.t. off-state current.However, a photovoltaic increase in the p*-n junction biasis brought about, and this leads to an increased forwardjunction current that:

(i) supports the optically generated junction depletionregion current

(ii) supports the surface depletion region photocurrentThe photocurrent produced in the junction depletionregion is given by

Wj

oJG(z)dz

where G(z) has been defined in this region as

PX

(3)

G(z) = n 0 < z <hcAL

SOLID-STATE AND ELECTRON DEVICES, JULY 1979, Vol. 3, No. 4

Here, z = 0 is taken to be at the edge of the junctiondepletion region and the p+-substrate (Fig. 3).

Substituting for G(z) in eqn. 3 and evaluating theintegral gives the current produced by optical excitation ofhole-electron pairs in the p+-n junction depletion region as

_ _qrjXP

(4)

3.3 Neutral n-region

Optical generation of hole-electron pairs in the neutral^-regions results in two current components flowing inopposite directions. The components are due to the dif-fusion of photoholes to either the surface or junctiondepletion regions. Photoholes reaching the surface depletionregion are swept to the oxide-silicon interface where theytunnel through the thin SiO2 layer. If the device is nottriggered into the switching mode, this generated photo-current contributes directly to the m.i.s.t. off-state current.Photoholes diffusing to the junction depletion region areswept to the p+-substrate by the electric field presentwithin the junction space-charge region. They then con-tribute to a photovoltaic increase in the p+n junction bias.

As linear superposition applies to the junction biascurrent and the photocurrent,7 then the two opticallygenerated current components can be determined fromthe photohole (or minority carrier) gradient on either sideof the neutral «-layer. The distribution of photoholes inthe neutral «-region is obtained by solving the steady-state1-dimensional diffusion equation:6'7

DT G(y) = 0

or

,<x(y-Wn-wd)by'' D

(5)

where Dp is the hole diffusivity, TP is the minority carrierlifetime, Lp is the minority carrier diffusion length, andthe rate of optical generation of hole-electron pairs in theneutral w-layer is given by

G(y) ~ T?<Poae vy n d>

Here, .y = 0 is taken to be at the edge of the junctiondepletion region and the H-type epitaxial layer, as shown inFig. 3.

The general solution to eqn. 5 is

p . (y) = AeylLp + Be~ylLp + v-eaiy-wn-wd)Xpe

where xP is

XP = (7)

The constants A and B can be determined from thefollowing set of boundary conditions:

(a) Because of the polarity of the electric field in thejunction depletion region, photoholes that diffuse to theedge of the junction depletion region will run down apotential energy hill to a more stable lower energy levelin the p+-substrate of the m.i.s.t. Thus, a perfect sink exists

87

for minority carriers, so that

PnL(0) = 0 (8)

(ft) Similarly, the edge of the surface deep depletionregion at y = Wn is another sink for minority hole carriers,so that

= 0 (9)

Application of these boundary conditions to eqn. 6, andsolving for A and B, gives the distribution of photoholes inthe neutral w-region as

<*{y-wn-wdy (10)

According to the boundary conditions (eqns. 8 and 9) twolight-generated current components may be seen to exist- a smaller component IpL flowing across the p+-n junction,and a larger component IpL flowing into the surfacedepletion region. Each is determined by the minority carriergradient on either side of the neutral w-layer. That is:

IpL = -dPnL

dy y=oAL

cosh " *inh ̂ ~LP

(11)

PL = ~<lDt

dpnL

dy y=wn

qrj\Pe-aWdle-aWn - cosh ^ + ccLp sinhy1

sinh —-

(12)

However, when no surface depletion region exists (zero biasapplied to the cathode of the m.i.s.t.), the boundary con-dition given by eqn. 9 no longer applies. Thus, IpL and IpL

given by eqns. 11 and 12, respectively, are similarly nolonger valid. Under this condition of zero bias, photoholescreated in the «-type epitaxial layer of the m.i.s.t. will nowonly diffuse towards the p+-n junction depletion region.

That is, there is no photocurrent flowing away from thejunction, and hence

= 0 (13)

For the current IpL flowing across the junction depletionregion and into the p+-substrate, the steady-state diffusionequation must be solved subject to the new boundaryconditions:

PnL(0) = 0

and

= G(Wn)Tp

(8)

(14)

where Wd — 0.The second boundary condition is valid only if the hole

diffusion length Lp is much less than the width of theneutral w-region Wn. Otherwise, holes created at the surfacewill diffuse to the edge of the junction depletion regionwhere they will immediately be swept to the p+-substrate.Consequently, the steady-state concentration of photoholeswill be less than that suggested by eqn. 14. However, them.i.s.t. is fabricated on wafers where, typically, the widthof the epitaxially grown w-layer (and hence the neutralw-region for Wd = 0) is greater than the hole diffusionlength. Therefore, the special case of Lp > Wn is nottreated.

Application of the boundary conditions (eqns. 8 and 14)to eqn. 6, and solving for the constants A and B, gives thedistribution of photoholes in the neutral w-region for Wd =0,as

WnlLp -1)y/Ln

2swhWjLp

-y/LP +

(15)

The resulting light-generated current flowing across thejunction into the p+-substrate is then determined by theminority carrier gradient at the edge of the junctiondepletion region:

_ —grjaXPLp>L Wd=° ~ he sinh Wn/Lp

qy\P(e-aW" cosh Wn/Lpn sinh Wn/Lp - 1)

ahcLp 1 - I sinh WJLl

(16)

3.4 Neutral p*-region

A similar situation exists in the p+-substrate as in theneutral «-section. Optically generated minority carriers(electrons) within diffusion range of the p+-n junctionwill eventually be swept by the electric field toward then-type epitaxial layer. Thus, a further contribution ismade to the photovoltaic increase in the junction bias.Photoholes, on the other hand, will diffuse away fromthe junction and eventually recombine with the minoritycarrier electrons.

88 SOLID-STATE AND ELECTRON DEVICES, JULY 1979, Vol. 3, No. 4

As linear superposition of the junction bias current andthe photocurrent applies,7 the optically generated currentcan be determined by the minority carrier photoelectrongradient at the edge of the junction depletion region.The distribution of photoelectrons in the p+-substrate isobtained by solving the steady-state diffusion equation

+ G(u) = 0

or

Dr

(17)

where Dn is the electron diffusivity, rn is the minoritycarrier lifetime, Ln is the minority carrier diffusion length,and the rate of optical generation of hole-eiectron pairs inthe neutral p+-substrate is given by

G(u) = v*oae-aifl+w"*w**wi>

Here, u = 0 is taken to be at the edge of the junctiondepletion region and the p+-substrate, as shown in Fig. 3.

The general solution to eqn. 17 is

npL(u) = AeulLn

where the term \n *s defined as:

Xn =

(18)

(19)

The constants A and B are determined from the followingboundary conditions:

(a) Any photoelectrons reaching the junction depletionregion will run down a potential energy hill to a more stablelower energy level in the neutral w-layer of the m.i.s.t. Thusa perfect sink exists for minority carriers, so that

«PL(0) = 0 (20)

(b) Photoholes diffuse away from the junction andeventually recombine with the minority carrier electrons, sothat

= 0 (21)

Application of these boundary conditions to eqn. 18, andsolving forv4 and B, gives the distribution of photoelectronsin the p+-substrate as:

npL(u) = Xne-«Wn+w*+w» [e-^-e-^n] (22)

The resulting light-generated current flowing across the p+-njunction is then determined from the photoelectron gradientat the edge of the junction space-charge region:

dn

u=0

he 1aL,

(23)

4 Discussion of the optical switching mechanism

We now consider the mechanism responsible for opticallyswitching a given device biased as shown in Fig. 2.

To switch the m.i.s.t. optically, an inversion layer muststill be formed at the surface, as is the case for electricalswitching. However, the mechanism responsible for in-version is not punchthrough but rather the generation ofphotoholes within diffusion range of, or in, the surfacedepletion region. This creates the current componentsIdL and IpL that flow towards the oxide-silicon interface.Furthermore, optical generation of hole-electron pairswithin diffusion range of, or in, the junction depletionregion result in reverse current components IpL, IjL andInL which give rise to a photovoltaic increase in the p+-njunction voltage. Consequently, there is an increase in theinjection of holes from the p+-substrate towards the surfacedepletion region where they are then swept to the oxide-silicon interface by the surface electric field.* However, atthe interface, the oxide voltage is sufficient to pass only thesmall off-state current, and thus accumulation of both thephotoholes and injected holes takes place at the surface. Asis the case for electrical switching, the build up of holesresults in three interacting effects:

(i) The w-layer begins to move from deep depletiontowards inversion, causing the surface potential, and hencethe voltage across the m.i.s.t., to decrease, and the voltageacross the load resistor to increase.

(ii) The oxide field begins to increase, allowing a largercurrent to pass.

(iii) Electron tunnelling from the metal to the semi-conductor conduction band increases because of its strongdependence on the free hole density at the semiconductorsurface.

Increased electron tunnelling results in the electronsbeing swept by the surface field to the p+-substrate, thusinitiating a further increase in the injection of substrateholes to the oxide-silicon interface. If the oxide voltage ischarging up at a rate sufficient to accommodate the secondhole injection mechanism, the surface free hole density willbegin to decrease, causing the device to remain in theoff-state. A steady-state operating point is established inthe m.i.s.t. when the currents flowing through the loadresistor, the p+-n junction and the tunnel oxide are equal,and there is a sufficient free-hole density at the surface tosustain the increased oxide voltage. Under these conditions,there is an increase in the m.i.s.t. off-state current with acorresponding decrease in the voltage across the device, asdepicted in Fig. 4a. This has been observed experimentally(Fig. 4b).

As the incident optical power is increased there is anapproximately linearly related increase in the forwardcurrent components IdL and IpL, and the reverse currentcomponents IpL, IjL andInL. Consequently, the free holedensity at the oxide-silicon interface is increased, causinga larger electron tunnel current to flow, which in turncauses an even greater increase in the injection of substrateholes to the semiconductor surface. Although the oxidefield is increasing in response to the surface free-holedensity, it becomes increasingly more difficult for the oxidefield to accommodate the increased flow of holes to the

Recombination of injected holes will take place in the neutral/i-region. However, as this does not affect greatly the overall dis-cussion of the optical switching mechanism, it is assumed for thesake of clarity that recombination is negligible.

SOLID-STATE AND ELECTRON DEVICES, JULY 1979, Vol. 3, No. 4 89

surface. When the rate at which these holes reach the oxide-silicon interface becomes greater than the rate at whichholes are tunnelling through the thin oxide, enhancedaccummulation takes place at the surface. Hence, a regener-ative positive-feedback loop with loop gain greater thanunity is established and the device moves into a switchingtransient. In this mode the surface potential continues todecrease, and the oxide voltage and device current toincrease, thus allowing the formation of the negativeresistance region in the m.i.s.t. output characteristic asshown in Fig. 4. When the surface becomes inverted, thesurface potential no longer decreases but remains fixed attwice the potential difference between the semiconductorFermi level and intrinsic Fermi level. The oxide voltage,however, continues to increase, thus allowing the lowimpedance on-characteristic to develop. A steady-stateoperating point is then reached when there is currentcontinuity through the load resistor, the p+-n junction andthe tunnel oxide.

anode-cathode current IAC, pA

dark

V0N JMDFF2|V0FFdVapp

VDFF3 VOFF1

anodecathodevoltagevA C.v

Fig. 4

a I/V characteristic of the m.i.s.t. under different intensities ofirradiation (not to scale)

b Experimental observations (vertical scale: 1 MA/division, hori-zontal scale: 1 V/division)

90

Thus, the processes involved in both electrical andoptical switching are basically identical, differing only inthe mechanism responsible for initiating inversion of thesemiconductor surface. Electrically, and without gatecontrol, inversion is brought about by the surface depletionregion punching through to the junction depletion region,whereas, under optical excitation, it is the generation ofboth photoholes in and around the surface depletion region,and of hole-electron pairs in and around the junctiondepletion region that is responsible.

The sensitivity of the m.i.s.t. to a constant input opticalpower can be improved if more holes can be collected atthe surface and at a faster rate. These effects can be achievedby implementing the following simple structural changesto the device:

(a) increasing the epitaxial layer depth to above theabsorption depth of the incident radiation, thus ensuringa near optimum collection efficiency of the generatedphotoholes

(b) decreasing the p+-n junction area, and hence capaci-tance, which increases the rate at which the junctionvoltage charges up to the value fixed by the photovoltaiceffect.

However, increasing the epitaxial-layer depth will resultin a smaller photovoltaic increase in the junction bias andhence a smaller injection of substrate holes. A compromisesituation is therefore created.

5 Optical current continuity in the m.i.s.t. in theoff state

Optical switching of the m.i.s.t. takes place when the rateof collection of free holes at the surface is sufficient toinitiate a positive regenerative feedback loop with loop gaingreater than unity such that inversion of the surface isallowed to occur. However, it was seen in Section 4 that thedevice may remain in the off state if the rate at which holesare tunnelling through the thin oxide is greater than therate at which they are being replenished at the surface. Asteady-state operating point is then established in the offstate when these two rates become equal. We now examinethe components of current that bring about this currentcontinuity throughout the device.

'CTl dL

'PL

'diff.p

neutraln-region

IjL

InLanode

p*-substrate

tunnel surfaceoxide depletion

region

junctiondepletion

region

Fig. 5 Off-state current components in the m.i.s.t. in equilibrium

All components indicate a positive flow of holes

SOLID-STATE AND ELECTRON DEVICES, JULY 1979, Vol. 3, No. 4

Shown in Fig. 5. are the off-state current componentsflowing in the m.i.s.t. when in equilibrium. Here, IdL,IpL, IpL, IJL and InL are the optical current componentsdescribed in Section 3, Ir is the junction recombinationcurrent, Id\ffP is the junction hole diffusion current, whileICT and IVT denote the electron and hole tunnellingcurrents, respectively. All the current components shownindicate a positive flow of holes, although ICT and InL aredue to the flow of electrons. Furthermore, as the p+-substrate is more highly doped than the «-type epitaxiallayer, the junction electron diffusion current Idtffn has beenneglected.8

With the device in the dark, the junction voltage is of theorder of 0-1 V,2"4 and thus the junction recombinationcurrent dominates the junction hole diffusion current.8 Thesource of electrons that feed this recombination current arederived from both electron tunnelling (ICT) fr°m the metalto the semiconductor conduction band and from thethermal generation of electron-hole pairs in the epitaxiallayer (IVT)- For large junction area devices, it is the electrontunnelling mechanism that dominates. The total currentthrough the device may then be written as:

lACd = Ir = ICT

With the m.i.s.t. excited optically, we may distinguishbetween two cases. These are when:

(i) the junction hole diffusion current is much less thaneither the junction recombination current or the light-generation current components IdL and/pL. This will occurfor junction voltages of approximately less than 0-3 V.Under these circumstances there will be no injection ofholes from the p+-substrate to the oxide-silicon interfacewhereby the surface free hole density will consist solely ofphotoholes. The corresponding photoelectrons that areswept out of the surface depletion region help feed thejunction recombination current, which must also overcomethe optical current components IpL, IJL and InL. Thedevice current is therefore

UCL

Here, there is a negligible change in the off-state electrontunnelling current due to the accumulation of holes at thesurface.5 Hence, the increase in device current due tooptical excitation is (IdL + I^L)-

(ii) the junction hole diffusion current is of the sameorder of magnitude as either the junction recombinationcurrent or the light generated current components IdL andIpL. This will occur for junction voltages of approximatelygreater than 0-3 V. In this case, hole injection from thep+-substrate to the oxide-silicon interface takes place wherebythe surface free-hole density will now consist of bothphotoholes and substrate holes. Assuming a negligibleincrease in the off-state electron tunnelling current,5 theincrease in the device current due to optical excitation willbe the sum ofIdL,IpL and/di/yp.t

t Recombination of injected holes in the neutral n-region must betaken into account when calculating Idiffp.

SOLID-STATE AND ELECTRON DEVICES, JULY 1979, Vol. 3, No. 4

That is:

*ACL = Idiffp +Ir~ (IpL + IjL + InL)

= IcT + (IdL + IpL) + Idiffp

The optically generated current components can be seen tofit in well with the general concept of continuity of currentflow throughout the device. It now remains to show experi-mentally the existence of these light current components.

6 Experimental results

A number of experimental devices have been fabricated onsilicon wafers with a 20-6 //m epitaxial w-layer of 34 £2 cmresistivity, grown on a p+-substrate of 001 Cl cm resistivity.The wafers were cleaned in a solution based on hydrogenperoxide.9 This gave an oxide thickness of 13 A beforefurnacing. The thin oxide was grown at 800° C for 5-5 minin dry oxygen and was then nitrogen annealed at the sametemperature for 15 min. A final oxide thickness of 29 A,as measured with an ellipsometer, was obtained. Aluminiumcontacts were deposited on the m.i.s.t. in a vacuum systemof pressure less than 10~6 torr. The thin oxide cathode areawas 40jum x 40/xm while the p*-n junction area was1 -4 mm x 1 -4 mm.

In order to verify the existence of the light-generatedcurrent components, the following experimental measure-ments were taken:

(a) The increase in the m.i.s.t. off-state current at a fixedanode-cathode voltage was measured for various intensitiesof light. By ensuring that the p+-n junction bias does notexceed 0-3 V, then these increases will be equal to the sum°f IdL and IpL, as discussed in Section 5.

(b) The photovoltaic increase in the junction voltage atzero anode-cathode bias was measured for various intensitiesof light. This increase should then be equal to the junctionvoltage required to produce the three photo currentsIpL I wd = 0 JJL and InL.

All measurements were taken using a Tektronix type525 curve tracer. For ease of experimentation, a He-Ne redlaser (optical wavelength X = 0-6328 fim and absorptiondepth a"1 =2-5/xm6 was used. Furthermore, an opticaltest bench was set up in order to concentrate the incidentoptical power onto the most sensitive area of the m.i.s.t.,which was found to be the thick oxide just outside themetal cathode. Here, the illuminated area was of the orderof 5 urn x 5 /zm.

Shown in Fig. 6 is a comparison of the experimentallymeasured increase in the off-state current of the m.i.s.t.biased at VAC = 5 V, with the theoretical increase pre-dicted by the sum of IdL and lLL for three values of thedevice quantum efficiency r\}°' Here the incident opticalpower was concentrated onto the thick oxide just off themetal cathode. During the course of the measurements thep+-n junction voltage was not allowed to exceed 0-3 V.Under no incident optical power, the depth Wd of thesurface depletion region at VAC = 5 V was calculated to be615/im.8 Under illumination Wd decreased at a rate ofapproximately 0-017/xm//iW. (See Appendix 11). Finally,as the device off-state current in the dark is recombinationdominated (lAcd - Ir = ICT), a value for the minoritycarrier hole diffusion length LP may be calculated. Thisgave Lp =6jum. In estimating Lp, the following points

91

were taken into account:(i)Lp<Wd + Wn

(ii) the p+-n junction recombination and hole diffusioncurrents are strongly dependent on Lp, and are of the sameorder of magnitude at a junction voltage of approximately0 - 4 - 0-5 V.8

The theoretical and experimental curves illustrated inFig. 6 exhibit an approximately linear dependence on theincident optical power. This is sufficient to indicate that forjunction voltages less than 0-3 V, the increase in the m.i.s.t.off-state current is due to the optical current componentsIdL and IpL. The best fit between experiment and theoryoccurs for a quantum efficiency of approximately 0-85. Itremains to be shown experimentally that as P, and hencethe junction voltage, is increased, a third current com-ponent, namely IdiffP, becomes a major contributor to theincrease in the device off-state current. Furthermore, itshould be pointed out that if the diffusion photocurrentIpL is of the same order or greater than the drift photo-current IdL, then lateral diffusion effects must be con-sidered. In our measurements however, the width of thesurface depletion region is much greater than the absorp-tion depth of the incident radiation. Hence IdL is muchgreater than IpL so that lateral diffusion effects are mini-mised to such an extent as to be considered negligible.

If the incident radiation is concentrated just outsidethe cathode of the m.i.s.t. for the case of zero surfacedepletion region, then lateral diffusion effects will becomedominant. To offset this, the device was uniformly illumi-nated whereby the illuminated area is equal to the junctioncross-sectional area. The photocurrents IPL\wd.o hh an(^InL will flow in the mi.s.t. which cause a photovoltaic volt-age to appear across the otherwise zero biased p+-n junction.The resulting junction current flows in opposition to thethree photocurrents so that the external device current iszero. For junction voltages less than 0-3 V the junctioncurrent is recombination dominated. That is:8

or

A 6optical power P, JJW

Fig. 6 Experimental and theoretical increases in m.is.t. currentI AC at VAC = 5V and VAG < 0-3 V, for different intensities ofHe-Ne light

Lower three lines indicate theoretical predictions

lnL

2*T \2Ll(IpL\Wd = o+IjL+InL)In I 1—_. . _ h 1 (24)

where nt is the intrinsic carrier concentration, Aj is thejunction cross-sectional area and VAG is the p+-n junctionor anode-to-gate voltage.

Illustrated in Fig. 7 is a comparison of the experimentallymeasured photovoltaic voltage of the m.i.s.t. on opencircuit, with the theoretical increase predicted by eqn. 24for three values of the device quantum efficiency T?.10' U

Both the experimental and theoretical curves exhibit alogarithmic dependence on the incident optical power P.This is sufficient to indicate that the optical current com-ponents /PLljyd = o, IJL and InL are responsible for theobserved photovoltaic effect. The discrepancies betweenthe curves can be explained as being a result of uncertaintyin the magnitude of the illuminated area t and the absolutevalue of the incident optical power P. The best fit betweenthe experiment and theory occurs for a quantum efficiencyof approximately 0-8 (cf. rj = 0-85 for best fit in curvesshown in Fig. 6).

7 Application as an optical threshold detector

Provided the sensitivity can be improved, and the switchingspeed is of the order of a few nanoseconds, the m.i.s.t.could be applied to optical communication systems as anoptical threshold detector. Turn-on times of less than 2 ns,including delay, and turn-off times of less than 1 ns havebeen reported13 on electrically switched devices. Similaroptical switching speeds are envisaged, although for devicesdescribed in this paper, no attempt was made to optimisethe fabrication process and geometry in order to producehigh-speed devices.

At present, in high-speed fibre systems p-i-n or avalanchephotodiodes are employed as photodetectors. However, asthese diodes do not exhibit any switching property, extracircuitry is required to extract information from thereceived signal. The circuitry introduces unwanted noisewhich would not be present if an optical threshold detectorwere used. Thus, if the m.i.s.t. also proves to be no noiser

02r

6 8 10optical power P pW

12

Fig. 7 Experimental and theoretical increases in the functionvoltage of the m.Us.t. at VAc = 0V, for different intensities ofHe-Ne light

+ The devices tested were large-area devices and were not isolatedduring the fabrication process. Hence the junction cross-sectionalarea was not well defined.

92 SOLID-STATE AND ELECTRONDEVICES, JUL Y1979, Vol. 3, No. 4

than a p-i-n or avalanche photodiode, then a reduction inreceiver noise as well as circuit complexity can be obtainedwith its use in optical transmission systems.

Finally, we note that as fibre systems incorporate opticalsources operating in the 0-8—0-9 jum wavelength region,future experimental work on the m.i.s.t. as an opticalswitching device must be geared to this wavelength region.

8 Conclusion

In this paper we have considered the mechanism responsiblefor optically switching metal tunnel-insulator n-p+ silicondevices. The 1-dimensional diffusion equation was employedto derive the light-generated minority carrier distributionsand hence diffusion currents in the neutral n and p+ regions,together with the photocurrents in the surface and junctiondepletion regions. For junction voltages less than approxi-mately 0-3 V, the increase in the device off-state current isdue to the optical current components IdL and IpL, whereasfor junction voltages above 0-3 V, a third current com-ponent, namely the junction hole diffusion current, becomesa major contributor to the increase in the device off-statecurrent. Photocurrents IPL, IJL and InL were shown tocontribute to a photovoltaic increase in the p*-n junctionvoltage.

Experimental verification of the photocurrents andthe photovoltaic effect was undertaken. The agreementobtained between the experimental and predicted resultswas sufficient to indicate that the optical current com-ponents derived in this paper exist and hence play animportant role in the operation of the device.

Finally, the m.i.s.t. was shown to have potential as anoptical threshold detector in high-speed optical-fibre systemsprovided certain levels of sensitivity, switching speed andnoise performance are achieved.

9 Acknowledgments

The authors wish to thank the Radio Research Board ofAustralia for their financial support of the project.

10 References

1 SIMMONS, J.G., and EL-BADRY A.: 'Theory of switchingphenomena in metal/semi-insulating/«-p+ silicon devices', SolidState Electron., 1977, 20, pp. 955-961

2 EL-BADRY, A., and SIMMONS, J.G.: 'Experimental studies ofswitching in metal semi-insulating n-p+ silicon devices', ibid.,1977, 20, pp. 963-966

3 SIMMONS, J.G., and CHIK, D.K.: 'The metal-insulator-siliconthyristor' (to be published)

4 NASSIBIAN, A.G., CALLIGARO, R.B., and SIMMONS, J.G.:'Digital optical metal insulator silicon thyristor (o.m.i.s.t.)',IEEJ. Solid-State & Electron. Devices, 1978, 2, pp. 149-154

5 GREEN, M.A., and SHEWCHUN, J.: 'Current multiplication inmetal-insulator-semiconductor (MIS) tunnel diodes', Solid StateElectron, 1974,17, pp. 349-365

6 SZE, S.M.: 'Physics of semiconductor devices' (Wiley, 1969)7 WOLF, M.: 'Limitations and possibilities for improvement of

photovoltaic solar energy converters. Pt. I: Consideration forearth's surface operation'., Proc. IRE, 1960,48, pp. 1246-1263

8 GROVE, A.S.: 'Physics and technology of semiconductordevices' (Wiley, 1967)

9 KERN, W, and PUOTINEN, D.A.: 'Cleaning solutions based onhydrogen peroxide for use in silicon semiconductor technology',RCA Rev. 1970, 31, pp. 187-206

10 BRUGLER, J.S.: 'Optoelectronic nomenclature for solid-stateradiation detectors and emitters', IEEE J. Solid-State Circuits,1970, SC-5, pp. 276-283

11 SCHNEIDER, M.V.: 'Schotty barrier photodiodes with anti-reflection coating', Bell. Syst. Tech. J., 1967, 45, pp. 1611-1638

12 NASSIBIAN, A.G., and CALLIGARO, R.B.: 'Surface statecharge in thin oxide mi.s.t. devices', IEE J. Solid-State &Electron. Devices, 1979, 3, pp. 6-10

13 YAMAMOTO, T., KAWAMURA, K., and SHIMIZU, H.: 'Siliconp-N insulator-metal (p-N-I-M) devices', Solid-State Electron,1976,19, pp. 701-706

11 Appendix

The width Wd of the surface depletion region is given by:8

(25)

where e8 is the semiconductor permittivity, ND is thedensity of electrons in the neutral n-layer, and i//s is thesurface potential.

The surface potential can be calculated by consideringthe total voltage across the device. That is, with the m.i.s.t.not illuminated, we have

FB (26)

where VAC, Vox, VAG and VFB refer to the anode-cathodevoltage, oxide voltage, anode-gate (p+-n junction) voltageand flat-band voltage, respectively. The subscript d indicatesthat measurements are taken in the dark.

The oxide voltage in the dark can be written as:8

v = &-Voxd r

1/2*• 0-0082 (i//Sd)

where the surface space charge per unit area is

Qs = qNDWd = (2qNDeJ8dy/2C/m2

and the oxide capacitance is:

Cox = f* F/m2

Neglecting the oxide voltage, eqn. 26 may be rewritten as

FB AGd

where VACd is kept constant at 5 V, and VAGd is measuredexperimentally. The flatband voltage for this particulardevice was approximately 0-5 V.12 Hence \pSd and Wd maybe calculated, giving \ps — 4-38 V and Wd = 4-35 /im.

Keeping the anode-cathode voltage constant whileilluminating the m.i.s.t., the p+-n junction voltage increases,thus decreasing both the surface potential and the surfacedepletion region width. Ignoring the oxide voltage, theoff-state surface potential under illumination becomes

K = vACd - vFB - VAGL

where

or

AGL

K =

AGd

vAGd - VAGL

The corresponding depletion width is then obtained fromeqn. 25. It was found experimentally that Wd decreased ata rate of approximately 0-017jum/juW.

SOLID-STA TE AND ELECTRON DEVICES, JULY 19 79, Vol. 3, No. 4 93