Investigation of boundary scattering in dry etched quantum wires by artificial manipulation of the wire boundaries

Investigation of boundary scattering in dry etched quantum wires by artificial manipulation of the wire boundaries

Microelectronic Elsevier Engineering Investigation 23 (1994) 429-432 of boundary artificial manipulation R. Bergmann, A. Menschig, D. Griitzmache...

374KB Sizes 2 Downloads 22 Views

Microelectronic Elsevier

Engineering

Investigation

23 (1994) 429-432

of boundary

artificial manipulation R. Bergmann, A. Menschig, D. Griitzmacherb a4. Physikalisches bInstitut

Institut,

An optimized fabrication observe

boundaries. damage

1.

process

electron

beam

Aachen,

D-5100

roughness,

using a low damage

Germany

used magneto

the distribution

transport

and density

by an artificial electron

cyclotron

process

as a sensor

on the quality

of etch induced

manipulation resonance

needs the understanding

measurements

scattering

of the wire boundaries.

reactive

ion beam etching

to

of the wire centers

due

Side wall process

and

step for the etch mask exposure.

The reduction of the size of semiconductor devices is one of the major aims in semiconductor technology, because of the improvements of device characteristics for low dimensional systems [l]. Using high resolution lithography systems and dry etching it is possible to reduce the lateral dimensions of the devices into the submicron range. Electron transport in submicron semiconductor devices is influenced strongly by scattering of electrons at the boundaries. This scat,tering mechanism can be investigatet in long quantum wires (length > elastic mean free path), where an anomalous magneto resistance peak (aMRP) occurs at low magnetic fields (< 2T). This effect can be related to the strong influence of the wire side walls if the cyclotron radius corresponds to the wire width [2-41. An optimization of process parameters for the wire fabrication can be realized by using the aMRP as a sensor for the quality of the wire side walls. A variation of the process parameters for the etch mask and the dry etching process will gain more insight in boundary scattering processes with respect for symmetry, distribution and scattering strength. Therefore we have investigated on the one hand the influence of the dry etching process using different techniques such as reac-

0

Aachen,

Germany

and the etch mask definition

INTRODUCTION

c)167-9317/94/$07.00

Stuttgart,

wires with less side wall scattering

We have

process

was investigated

scanning

wires by

V. Harle, F. Scholz, H. Schweizer,a

D-70550

for quantum

mechanisms.

The effect of surface

can be minimized

a minimal

technology

J. Hommel,

Stuttgart,

RWTH

of the dry etching

to the wire definition

in dry etched quantum

of the wire boundaries

Universitst

the scattering

the influence

scattering

N. Lichtenstein,

fiir Halbleitertechnik,

how to minimize

429

1994 - Elsevier

Science

B.V.

tive ion beam etching (RIBE) and electron cyclotron resonance reactive ion etching (ECR-RIE). On the other hand the influence of the electron beam lithography process for the etch mask on the quality of the wire is investigated by variation of the exposing beam scanning step. In order to get more information about the wire side walls we have introduced a well defined chain of microscopic holes at the wire edges to simulate etch induced scattering centers. With this technique we take advantage of the well defined configuration of these scattering centers.

2. TECHNOLOGY RATION AND

OF WIRE PREPAMEASUREMENTS

The wire structures were fabricated from Ino,5sGao,47As/InP modulation doped heterostructures. We used two different samples with mobilities of 170 000cm2/Vs and 115 000cm2/Vs and and carrier densities of 8x 1011cm-2 1.1 x 10’2cm-2, respectively, at 4.2K. With optical lithography we fabricated ohmic contact pads consisting of Au/Ge-Ni on large mesa areas. The contacts were alloyed at 45O“C for 3 min in a Hz/N2 atmosphere. The wire masks were defined by high resolution electron beam lithography at an acceleration voltage of 50kV and a beam current of 100pA on a JEOL JBX 5D2(U) using the negative tone reAll rights reserved.

et al. i investigation

R. Bergmann

430

sist AZPNl14 a 250nm thick at a temperature

[5]. Th e samples were coated by resist layer and prebaked for 120s of 12O’C. Different wire widths

were realized by multiple line scans in combination with different line doses (0.3-1.3 nC/cm). After exposure on a hot developer

the mask

plate and

and Hz0

wed by a Hz0

is baked

developed dilution

two

with

different

Ar/Oa

at

dry

defects the resist the heterostructure

etching

processes:

RIBE

an acceleration voltage of 250V (30V) using [6] with low DC-bias

and ECR-RIE CClzFz/Ar. The magneto transport measurements have been performed in a Be4 bath cryostat at 4.2K where the magnetic field is oriented perpendicular to the 2DEG. The measurements out, with standard lock-in technique terminal configuration of the contact

3.

MAGNETO

were carried using a two pads.

TRANSPORT

ETCHED

QUANTUM

IN

DRY

WIRES

Magneto transport measurements are a sensitive tool to analyse boundary scattering in quantum wires thus process induced Typical

giving the defects.

magneto

possibility

resistance

to control

curves

of

180nm

scattering

tuations lead to additional anisotropic scattering at the wire boundaries. The wire side walls are described by the surface roughness Aw and the correlation length X. II) For isotropic scattering due to etch induced defects a surface layer is induced at the wire boundaries with a width b and a high defect density nd. In the following section influence

rinse.

To compare etch induced pattern was transfered into by

at 90°C for llmin in a standard AZ(1:l) for 60 s, follo-

of boundary

definition formation

process

the

and the mask

process on these parameters. This in gives the possibility to optimize the fa-

brication 4.

we will investigate

of the dry etching

technique.

ARTIFICIAL THE WIRE

MANIPULATION SIDE WALLS

OF

Variation of scattering strength for different dry etching processes Typical dry etching processes for pattern trans-

4.1.

fer of the

mask

into

the

semiconductor

produce

damage at the wire boundaries. In order to get information about the influence of the dry etching process on the wire quality we used two different techniques. the samples, DC-bias

One is RIBE, known for and the other ECR-RIE

(3OV),

a dry etching

ses very low damage at the The magneto transport

process

damaging with low which

wire surface. measurements

cauin fi-

gure 1 compare the effect of the dry etching processes on 180nm wide wires. Although there is

wide wires are shown in figure 1. When the magnetic field is increased an anomalous magneto resistance peak appears at magnetic fields below 2T. The experiments that in wider systems

for InGaAs/InP (Ly >lOOnm)

wires show the resistance

reaches the maximum at L,/r, x 0.55 where L, is the wire width and r, is the cyclotron radius. At this ratio the interaction of the electrons with the

wire

side

walls

reaches

a maximum

and

14 g

-I?

2 12-1-”

c

I -ECR-RIE

cau-

ses an increase of backscattering due to diffuse boundary scattering [2, 31. We recently have developed a quantum mechanical description of the magneto resistance trace in quantum wires on the basis of the Kubo formula in self consistent Born approximation [4]. A microscopic description of the boundary scattering the

of electrons following

in a narrow models

channel

[4, 71: I) Wire

is made width

0

I 2

* ,

.

magnetic

I.

I.

kid

B [i

t

.

.

*

8

by fluc-

Figure (dashes)

180nm wide 1. and by ECR-RIE

wires

etched

(solid).

by

RIBE

431

R. Bergmann et al. / Investigation of boundary scattering no change in the carrier density, the ECR-RIE etched wire shows a significantly smaller resistance as the RIBE etched wire over the whole magnetic field range. The amplitude of the aMRP is by a factor of 5 smaller for the ECR-RIE process, the zero field resistance has dropped by a factor 1.5. The constant carrier density in both wires indicates that the difference in the magneto resistance can be attributed to effects of the wire boundary. The decrease of the amplitude of the aMRP shows that there is a strong decrease of boundary scattering for the ECR-RIE process compared to the RIBE process. The number nd of etch induced scattering centers is significantly smaller for ECR-RIE etched wires. The magneto transport measurements proove that the ECR-RIE process causes very low damage in submicron semiconductor devices. 4.2.

Controlled manipulation of the wire boundary by changing the beam scanning step for etch mask definition We have investigated the influence of the electron beam lithography process on the quality of the wire boundaries by changing the beam scanning step from 2.5nm up to 25nm. The wire masks of the 1OOnm wide wires were defined by three parallel line scans at a distance of 1Onm and for 180nm wires by 12 line scans. Each line scan is performed by exposing points with a user defined pitch of 2n times 2.5nm ( n = 0, 1, 2 ,... ). Figure 2 shows 180nm wide wires with different mask definitions using several beam scanning steps between 2.5nm and 20nm. In all cases the electron density remains unchanged at a value of 1.1x1012cm-2. We conclude that the difference in the magneto resistance is due to the influence of the wire boundary. With increasing beam scanning step the resistance in the magnetic field range after the aMRP is increased. This can be related either to a decrease of the mobility or to a decrease of the effective wire width. Due to the position of the aMRP, which is unchanged, we attribute this effect to a decrease of the mobility of the wire. The amplitude of the aMRP increases with increasing beam scanning step as shown the inset

0

1

4 ma&c

5

fields B [T]

Figure 2. Magneto transport measurements 180nm wide quantum wires with increasing tron beam scanning step in the lithography cess. Inset: The amplitude of the aMRP function of the beam scanning step.

on elecproas a

in figure 2 as well as the zero field resistance. For the 1OOnm wide wires the same effect has been observed. From theoretical calculations we know that both effects can be related to an increase of electron scattering due to an increase of surface roughness Aw, to a change in the correlation length X of the scattering centers or to an increase of the number of scattering centers nd at the wire boundaries. Several parameters are changed simultanously by increasing the beam scanning step in this experiment. Therefore it is not possible to distinguish the influence of the different parameters. These results show that the electron beam lithography process affects the quality of the wire boundaries. Even a change of the beam scanning step in a 2.5nm raster can be detected by magneto resistance measurements. We conclude that it is necessary to reduce the beam scanning step as far as possible to obtain high quality submicron devices with a minimum of boundary scattering effects. This contradicts the aim to minimize exposure time by increasing the beam current and therefore the beam scanning step.

R. Bergmann

332

4.3.

Anti&t

arrxlgenlctnt

at

the

et al. 1 Investigation

wire

siclcl

walls

it1

1111,

a 3OOiirri wide wire wit II ii11 ai11 iat (lit, wiw side walls (dian1~~tc~r

Pigun, 3 shows tlot arrai~gc~nwiit.

(Niitii, pwiotl 200111r1). ‘Ttic niagncto resist ailcc’ l.racm of t Ii<‘ reff,rence wirt> and the wire with thta ailtidot,s show again t hr haiii(‘ carrier densily foi hol.11 wirm. ‘I‘lir:rcxforc the, iricrc,asc> of 1he ;IIIII)li(ude of t.llc, a!VlRl’ I)y iI factor of 1.7 autl I II<, iricreasc, of th(, rt‘sist.;lllcr~ after tII(> ah1RT’ t!y ;t factor of 1 .I can 1)~ relatcvl ‘l‘hc~ adtlitionall~ iiit,roclriwtl

to Imundary f,ff;,c,th sca.tkriiig cCiift>rh at

1IIC wire houndaries

aii

Ilary scattuirig rcsist,arrce

which

lead

to

increase

is rc+lcct cd iii the

t,ra.cc by a11 increase

of holiilmagiic~tcr

of the, aRlR J’. ‘1‘11~~

mcrcase of the i-esist.ancc aft,er the aR/IItJ’ cati ho relat,ed to an increase of etcll induced scat.tering ccnt,crs becalm, t,llc position of t,he ahlR.P r~~rnains uilchanged which excludes a gcomct,ry cffcct.. ‘1‘11~ I)<,riod of the ant,idot st riicf.urt~s can tic rclntul t,o t,hc correla.t,ioii lcmgt II X of the scxt,t,eririg (‘<‘iIt.ers and the dot dialn&~r to t.he surface, rougllncss 1tl:. Furt.her work will hc done t.o corrc,lat,c, t,he experimental observations with the tlieorr~tical calcli1a.t ioiis.

01 boundary

.scuttrring