Baca A.G., Ashby C.I.H. Fabrication of GaAs Devices

Подождите немного. Документ загружается.

Ohmic contacts

conduction band in the bulk of the semiconductor. The bending of

the conduction band at the metal/semiconductor interface sets the

barrier to current ﬂow across the interface, as in FIGURE 6.3.

Electrons may cross the potential or Schottky barrier by one of

two mechanisms: thermionic emission or ﬁeld emission (tunnel-

ling). A combination of the two mechanisms is termed thermionic-

ﬁeld emission. In thermionic emission, a fraction of the electrons

have sufﬁcient thermal energy to cross the barrier. The fraction is

related to the factor

exp(−qφ

/kT) (6.4)

and the current due to thermionic emission is given by

J = (const) exp(−qφ

/kT)[exp(qV/nkT) − 1]

J = J

[exp(qV/nkT) − 1]

(6.5)

where φ

is the barrier height, q is the electron charge, k is

Boltzmann’s constant, V is the applied voltage across the bar-

rier and n is the ideality factor (discussed in Section 7.2.1). The

purpose of an ohmic contact is to provide a low-resistance con-

tact. EQN (6.5) indicates that low currents arise with low voltages

unless the Schottky barrier is fairly low.

The second mechanism for current ﬂow across a barrier, ﬁeld

emission, allows for quantum mechanical-tunnelling through

the Schottky barrier. The equation governing current from ﬁeld

emission is given by

J = exp(−qφ

) (6.6)

where

= qh (N/εm

∗

)

1/2

/2 (6.7)

and h

is Planck’s constant dividedby2π, ε is the dielectric constant,

N is the doping concentration and m

∗

is the electron effective mass.

Field emission works by electrons tunnelling through the Schottky

barrier, whose width is roughly the same as the thickness of the

depletion region. The tunnelling current increases exponentially

with the doping concentration. Practically speaking, this current

becomes signiﬁcant enough to make low resistance contacts to

GaAs when N exceeds 10

−3

In reality, both thermionic and ﬁeld emission can come into play

for contacts at room temperature. This type of conduction is termed

thermionic ﬁeld emission and it works by electron tunnelling from

thermally excited electrons that tunnel through a somewhat thinner

barrier by virtue of the electron energy being somewhat higher than

the Fermi level.

182

Ohmic contacts

As seen from EQNS (6.5)–(6.7), lowering the barrier height and

increasing the doping concentration are the two basic means for

creating an ohmic contact to any semiconductor. Simple as this

seems, GaAs presents a somewhat difﬁcult case for barrier height

lowering. Surface or interface state densities are fairly high for

GaAs, compared with the SiO

/Si interface, for example, which

has the practical effect of pinning the Fermi level to midgap and

constraining the barrier heights of most metal/GaAs contacts to

a fairly narrow range (Section 7.4). If barrier-height-lowering is

to be a practical method, the creation of lower bandgap materials or

the creation of lower-barrier-height interfaces are the only avail-

able approaches. Some metallurgical approaches along this line

will be presented in the next section. Alternatively, the strategy

for designing an ohmic contact can rely on achieving high doping

levels.

6.2.3 Metallurgy of ohmic contact formation

Although Ge/Au/Ni is the dominant contact metallurgy for n-type

GaAs, literally hundreds of recipes exist for ohmic contact forma-

tion to GaAs and no mechanism is expectedtoexplain all thedetails

of the various ohmic contact processes. It is somewhat surprising,

therefore, that most of the contacts to GaAs seem to follow the

broad outlines of a general mechanism. Studying this mechanism

can help the practising engineer understand the importance of the

various elements of the recipe and relate them to the otherwise con-

fusing folklore that is sometimes told about GaAs contacts. Prior

to a description of this mechanism, a few historical highlights of

GaAs ohmic contacts will be presented.

Some folklore about GaAs contacts

(and reply). See text for more

mechanistic explanation.

AuGe eutectic: Eutectic chosen so that

contacts will melt at chosen

temperature; molten alloy allows Ge to

diffuse better into GaAs.

Reply: The mechanisms for Ge and

other constituents to diffuse into GaAs

can work well at temperatures well

below the eutectic melting of GeAu.

Role of Ge: Ge acts as n-type dopant

in GaAs.

Reply: Ge reactions with Ni are also

important.

Role of Ni: Ni acts as a wetting agent so

that the melted AuGe will not “ball up.”

Reply: Ni plays an important role at

temperatures well below the AuGe

eutectic.

Sn was used as an early contact to GaAs, while GeAuNi became

the most widely used contact after its report in 1967 [1]. Inter-

estingly, Ge and Au were used in their eutectic composition,

88% Au and 12% Ge by weight, which is the composition of low-

est melting point. Metal alloys with eutectic compositions have

many uses, including the attachment of semiconductor chips to

their package by eutectic melting and bonding. Ohmic contacts

with low-bandgap surface regions obtained by epitaxial growth

of InGaAs were ﬁrst reported in 1981 [2] and many variations

on this theme were later reported. Contacts based on solid-state

regrowth using the Pd/Ge metal system were introduced in 1985 as

a way of improving the reliability of ohmic contacts by understand-

ing and controlling their metallurgical reactions [3]. Although

these contacts were never widely adopted, the subsequent research

on solid-state regrowth solidiﬁed our understanding of basic

mechanisms in contact formation.

The metallurgy of ohmic contact formation to n-type GaAs will

include some or all of the following steps: 1) an element of the

183

Ohmic contacts

contact reacts with the native oxide present on the GaAs surface,

2) a low temperature process occurs where one element of the con-

tact, the Ni in Ge/Au/Ni alloys, diffuses into GaAs and reacts to

form intermediate X-GaAs complexes, “decomposing” the GaAs

or otherwise facilitating interdiffusion between the contact and

GaAs, 3) an element of the contact capable of doping GaAs diffuses

into the GaAs aided by defects created in step 2 and dopes GaAs,

4) an element of the contact capable of forming a low bandgap

alloy with GaAs diffuses into the GaAs, 5) Ga outdiffuses from

GaAs and reacts with a contact element, often Au, 6) elements

of the contact continue to react and form their thermodynamically

favoured alloys until the reaction is complete or the thermal budget

is exhausted, 7) excess As diffuses to the surface where it resides

or vapourises.

The most commonly used contact is GaAs/Ge/Au/Ni. The

metallurgy of this contact probably includes six of the steps out-

lined above. The Ge/Au/Ni contact will be described in terms of

these steps of the metallurgy as a speciﬁc example of the reactions.

In step 1, the Ni reacts with the native oxide. The native oxide is

not easily characterised on GaAs and may vary in unrepeatable

ways due to variations in processing, so it is useful for the contact

to have a constituent such as Ni that will reliably react with differ-

ent types of surface oxide. GeAu contacts without Ni are known

to have poor morphology and large variability. The native oxide of

GaAs may have regions that allow easy penetration of the GeAu

and the alloy formation occurs rapidly in these areas. Other areas

may never react because the GeAu does not penetrate the native

oxide and poor morphology results.

Summary: 7 common steps of GaAs

ohmic contact alloy reactions

1. An element of the contact reacts

with the GaAs native oxide.

2. An element of the contact reacts

with GaAs at low temperatures to

form a X-GaAs complex.

3. An element of the contact diffuses

into GaAs mediated by defects

created in step 2 and dopes GaAs.

4. An element of the contact diffuses

into GaAs and forms a low bandgap

semiconductor phase.

5. Ga outdiffuses from the contact and

reacts, usually with Au.

6. Elements of the contact react and

form their thermodynamically

preferred compounds.

7. As diffuses from the GaAs to the

contact surface where it resides or

vaporises.

In step 2, the Ni forms intermediate complexes with GaAs. The

Ni-GaAs complexes later give way to more stable products such

as AuGa intermetallic reaction products. The important role of the

intermediate complexes is to disturb the crystal lattice to make way

for diffusion of other atoms and further reactions. In the case of

GeAuNi contacts, the absence of Ni would result in more thermal

energy required for the Au and Ge reactions to get started. Once

started they quickly get out of control leading to severe spiking and

poor morphology. The Ni-based complexes are thought to form at

temperatures below 100

◦

Once the Ni-GaAs complexes have disturbed the crystal lattice,

the way is prepared for the diffusion of Ge into the GaAs in

step 3. Normally, diffusion of dopants such as Ge in GaAs requires

a substitutional-interstitial mechanism where the Ge can diffuse

along interstitial sites until it ﬁnds a vacancy in the GaAs lattice.

This type of mechanism has high activation energy and normally

occurs to a signiﬁcant degree at temperatures of 600

◦

C and above.

In a GaAs ohmic contact, the diffusion process is thought to occur

184

Ohmic contacts

at much lower temperatures because of the driving forces of the

Ni-GaAs complex formation and the driving force of the next

reaction which further disturbs the crystal lattice. The high activ-

ation energies of creating and ﬁlling Ga vacancies are overcome

by complex reaction intermediates involving Ni-Ge-GaAs rather

than by thermal effects that are necessary during substitutional-

interstitial diffusion. The diffusion of Ge into GaAs at tem-

peratures between 100 and 400

◦

C leads to very high doping

levels that make tunnelling contacts possible for doping levels

above 10

It is not presently believed that ohmic contact formation with

Ge/Au/Ni in GaAs participates in step 4 through the formation of

a low bandgap alloy, though NiAs phases were considered pos-

sible low-barrier-height phases in the past. The second part of

the driving force for steps 2 and 3 comes from a step 5 reaction,

the Au-Ga reaction. The Au-Ga alloy is more thermodynamically

stable than any other compound between the ﬁve elements that

can form. At temperatures at and above 250

◦

C, this reaction is

kinetically favoured and with sufﬁcient time will continue until

the supply of Au is exhausted.

In step 6, other reactions between the GaAs and the contact

constituents continue. Some of these reactions form intermediate

compounds such as NiAs while others such as NiGe are more

stable.

The last step is the transport of As to the surface of the con-

tact and its desorption from the contact. Among the ﬁve reacting

elements, As is most stable in the elemental phase and leaves the

GaAs. The signiﬁcance of this step along with step 5 is the con-

sumption of GaAs. Voids formed by the loss of As are formed

and ﬁlled with AuGa, NiGe and other stable compounds. In the

ideal case with good processing technique, the reaction occurs

uniformly. Because the ideal case is usually only approximately

achieved and because the reactions occur very fast, spikes form

if there are any process non-uniformities. Visual inspection of an

alloyed GaAs/Ge/Au/Ni contact will show a textured surface for

a properly alloyed contact (FIGURE 6.4(a)) and a much greater

morphology change for over-alloyed contacts (FIGURE 6.4(b)).

Surface proﬁlometry or atomic force microscopy measurements

will show depressions in the areas where the contact has spiked

and excess As has been consumed. Transmission electron micro-

scopy directly images the regions where spiking has occurred and

has been used to identify some of the reaction products.

(a)

(b)

FIGURE 6.4 Optical micrograph

of an alloyed GeAuNi ohmic

contact: a) slight texture for

properly alloyed contact, b) greater

texture for over-alloyed contact.

Most contact metallurgies in use today follow some of the seven

steps outlined in this section. Some of these steps have resulted in

problematic characteristics for a GaAs ohmic contact and have

stimulated further research. Much of the research has resulted

185

Ohmic contacts

in optimisation and better understanding of the mechanism of

the GaAs/Ge/Au/Ni contact. Some alternative metallurgies have

been developed and the more common ones will be reviewed in

Section 6.3.

(a)

(b)

(c)

FIGURE 6.5 Illustration of the

steps required for forming an

ohmic contact.

6.3 FABRICATION AND TESTING OF

OHMIC CONTACTS

6.3.1 Fabrication of ohmic contacts

The fabrication of an ohmic contact involves the steps of pattern-

ing the GaAs with resist or another suitable mask, preparing or

cleaning the surface, depositing the metal, and alloying the metal,

if needed. The most common process is to pattern the sample ﬁrst

and to use a liftoff process to produce the patterned contact. The

basic steps are illustrated in FIGURE 6.5. A photoresist is applied

to the wafer and spun typically between 1500 and 6000 rpm to

uniformly coat the wafer. The resist thickness should be uniform

so that good control of the ﬁnal linewidth can be achieved. Photo-

lithography is discussed in greater detail in Chapter 4. A glass or

quartz mask is then aligned to the pattern on the wafer so that light

can pass through the areas to be exposed and be blocked in the areas

to be protected from the light. Any of the available lithography

tools may be used, but the optical stepper is the tool of choice

for most high-performance and high-yield processes in HBT and

FET fabrication. Once the mask is aligned with the wafer, the res-

ist can be exposed to the light and the pattern can be developed

(FIGURE 6.5(a)). A re-entrant cross-sectional proﬁle is desired so

that the surface of the resist shadows the sidewalls and prevents

the evaporated metal from coating them. Re-entrant proﬁles can be

achieved with either positive or negative (image reversal) resists

(Section 4.4.1). The choice of resists depends on the type of optical

aligner used, the wavelength of the light and factors relating to the

device morphology.

Prior to depositing the ohmic contact, a surface treatment may

be used. This may include a plasma oxygen treatment to remove

monolayers or less of organic material that remain after the resist

develop process. The surface treatment may also include oxide

removal or other steps. The ohmic metal is then evaporated over

the whole wafer where it forms the contacts in the patterned areas

and covers the resist over the rest of the wafer (FIGURE 6.5(b)).

The unwanted metal is removed from the wafer when the resist

is dissolved, typically by means of a solvent spray. At this point,

the ohmic metal (FIGURE 6.5(c)) can be tested electrically if a

non-alloyed ohmic contact process was employed. Otherwise, the

186

Ohmic contacts

contact will be alloyed so that the metallurgical reactions needed

for ohmic contact formation can proceed.

Surface treatments for removing native

oxides of GaAs were described in

Section 3.2.2.

The preceding description is an overview of the fabrication pro-

cess. Several of the fabrication critical steps will now be described

in more detail. The pre-clean step can be critical for achieving

low electrical resistance in the contact. One common procedure

used with Ge/Au/Ni contacts employs a plasma oxygen clean fol-

lowed by an oxide removal step. The GaAs native oxide is about

1–2 nm thick. Exposure of a clean GaAs surface to the ambient

results in the formation of a consistent native oxide. This native

oxide is easily removed by dilute acids such as HF or HCl or bases

such as an NH

OH:H

O mixture (Section 3.2.2). During routine

semiconductor processing the GaAs comes into contact with other

chemicals such as hydrocarbons in a photoresist formulation that

may leave residues on the surface of the GaAs. In such cases, both

the formation of the native oxide and the reaction of a deposited

ohmic metal are complicated by the presence of hydrocarbon or

ionic residues. It is found that ohmic contacts can have excel-

lent electrical characteristics but are not repeatable. Consequently,

various attempts have been made to improve the repeatability of

ohmic contacts. Several of these involve methods to remove the

hydrocarbon residues.

A common procedure to accomplish the hydrocarbon removal is

to expose the patterned GaAs wafer to an oxygen plasma. Typic-

ally RF radiation at 13.56 MHz is used to create a plasma with

and maybe a carrier gas such as Ar. Equipment speciﬁc-

ally designed for this purpose is often available and one type is

commonly referred to as an “asher”. A popular geometry is the cyl-

indrical shape with a powered metal cylinder enclosing an interior

region that contains the plasma and the samples, which mount to

grounded rods that are placed at positioned breaks in the powered

electrode design. The operation of a barrel asher was described

in Section 5.6 and FIGURE 5.4. This particular design is useful

for hydrocarbon residue removal because the geometry leads to

a lower plasma sheath voltage than the parallel electrode design.

Such a lower self-biasing of the plasma is important for a gen-

eral purpose plasma oxygen process because it is potentially used

multiple times in a typical device fabrication and some of the steps

will have requirements of low or no plasma damage.

A newer method of hydrocarbon removal uses UV-generated

ozone exposure. Ozone can react with hydrocarbon residues by

releasing molecular oxygen and an oxygen radical that quickly

reacts with the hydrocarbon to form carbon oxides and other

organic molecules such as alcohols and ketones. The reactive

ozone is a neutral molecule which forms in the absence of a plasma

and cannot cause the types of damage caused by the acceleration of

187

Ohmic contacts

charged particles. It is therefore ideally suited as a general purpose

hydrocarbon removal tool. Commercial ozonation tools are among

the least complex clean room equipment in that they only need a

source of electrical power. In contrast to ashers, which are designed

to remove several microns of photoresist per hour, ozonators typic-

ally remove resist at a rate of several monolayers per minute. They

are especially valuable for stubborn submonolayer hydrocarbon

removal.

The removal of hydrocarbon residues prior to ohmic contact

deposition is a good way to ensure repeatability of the surface

condition. In addition to removing hydrocarbon deposits that can

block interface reactions, it also can create repeatable conditions

for oxidising the GaAs surface. Although oxides themselves are

potential impediments to interface reactions between the GaAs

and the ohmic contact, they are easily removed prior to contact

deposition and only partially regrow if samples are loaded within

minutes (up to ten is a reasonable compromise between unrealistic-

ally short times and times long enough to fully regrow the oxide) to

the deposition system. Common methods of native oxide removal

by acids such as HCl and HF or bases such as NH

OH also work

for the plasma or UV-ozone oxides. A further beneﬁt of incorpor-

ating an oxide removal step in the ohmic contact process is that any

low-level impurities that settle on the GaAs surface are dissolved

with the oxide.

Following the surface preparation methods comes the actual

contact deposition. Evaporation is the technique of choice for most

metal depositions on GaAs because it provides a method of high

purity deposition with minimal damage to the GaAs. This is espe-

cially so for Au-based contacts because an evaporation process

with liftoff patterning works well with Au, while etch processes

are limited to ion milling. Evaporation is carried out in a high-

vacuum chamber with pressure in the mid to low 10

−7

torr pressure

range. This pressure level is required so that the residual gases

in the vacuum chamber will not be a source of contamination

as the deposition proceeds. As previously noted, electron beam

evaporation results in minimal damage to most semiconductors

compared to the main alternative, sputtering. With electron-beam

deposition, no energetic ions are produced which can be accel-

erated to the sample surface and cause lattice damage. The main

concern for damage with electron-beam deposition is the effect

of energetic electrons. For cases where this type of damage is a

concern, thermal evaporation is used. For GaAs ohmic contacts,

lattice damage is not usually a concern because lattice disruption

is an intended consequence of the contact reaction.

A variety of evaporation equipment is available depending on

the degree of throughput and automation needed. For high-volume

188

Ohmic contacts

production, extremely large evaporators are available with the

capability of depositing on dozens of wafers per run. These large

vacuum systems are equipped with heavy duty vacuum pumps that

can bring about the required vacuum conditions in considerably

less than 1 h. On the other end of the spectrum, research tools have

much smaller vacuum chambers and pumps and may require sev-

eral hours for pump down after sample loading. A brief description

of the steps in an electron beam deposition follows.

An electron beam typically accelerated to 10 kV is directed at

one of the hearths or crucibles containing the high-purity solid.

The energetic electrons enter the solid where collisions and scat-

tering events with valence and core electrons from the solid result

in energy transfer from the accelerated electron to the solid source.

Ultimately, the source heats up in the localised area where the elec-

tron beam is focused and begins to melt and evaporate. It becomes

a source of atomic ﬂux throughout the chamber. The acoustic wave

properties of a quartz crystal are used as a ﬂux monitor to provide

feedback to the electron beam source and to stabilise the ﬂux to a

preset value. Once the ﬂux is stable, a shutter is opened to allow the

sample to receive a thin ﬁlm deposition. The process is ended when

the desired ﬁlm thickness is reached and is then repeated for any

other depositions that are needed in a multimetal stack. A thermal

evaporation follows a similar procedure, except that instead of an

electron beam being used to heat up the source material, electrical

current is passed through a refractory metal holder that contains a

small amount of the source material.

The various components of the ohmic contact are typically

deposited layer by layer. For example, deposition of a Ge/Au/Ni

ohmic contact will proceed with the deposition of the Ge layer

ﬁrst, followed by the Au layer, and ﬁnish with the deposition of

the Ni layer. On occasion two constituents will be co-deposited.

This occurs frequently with Au and Ge due to the fact that an alloy

of 12% by weight of Ge and 88% of Au forms the eutectic com-

position and melts at a lower temperature (361

◦

C) than other alloy

compositions. After deposition is completed the shutter is closed

and the evaporator is vented for sample removal.

Source material deposition is completed by the removal of the

resist material and the unwanted metal through a liftoff process. At

this point the contact formation is completed by means of a thermal

annealing step, if required. This step is often termed an ohmic alloy

for contacts that undergo an alloy process. For Ge/Au/Ni contacts

the thermal anneal is typically carried out near 400

◦

C for 10 s to

several minutes. Usually the shorter times require slightly higher

temperatures than the longer times.

A rapid thermal annealer (RTA) is commonly used for this

step, although furnaces and strip heaters are also used. An RTA

189

Ohmic contacts

is composed of several necessary components such as an inert

chamber (often quartz) which will contain the samples, a heater

such as an infrared lamp, a network of plumbing for inert gas

introduction and pumping, and a control system. It is important to

exclude oxygen during the anneal. Oxygen can cause several dele-

terious effects. First, it can oxidise the exposed GaAs surfaces.

Because this oxidation will be occurring at elevated temperat-

ures, the As-oxides are volatile and the process can continue to

the point of consuming a considerable amount of As from the

GaAs. Second, oxygen can react with the contact materials. For the

Ge/Au/Ni ohmic contact, Ge and Ni oxides would seriously inhibit

the intended reactions. The chamber containing the samples is usu-

ally purged with a high purity inert gas to drive out the oxygen.

Some equipment is designed for pump and purge cycles to speed

up this process while other equipment is designed for vacuum

annealing.

Other considerations also come into play for the RTA pro-

cess. One of these is the method of sample placement, which

will affect the thermal loading. One common type of annealer

uses a Si wafer mounted to quartz rods attached to the door of

the chamber with a thermocouple embedded in the wafer. In

such an arrangement, the sample is mounted directly on top of

the Si wafer. Another method involves placing the sample in a

specially designed susceptor, as illustrated in FIGURE 6.6. An

example of a sample susceptor for RTA is a SiC disk that is

coated with graphite for chemical inertness and as a radiation

absorbing ﬁlm. A central circular area is recessed a distance

slightly deeper than the typical sample thickness and is the region

where the sample will be placed. A second disk with a matched

diameter to the recessed region acts as a cover to the sample. This

type of sample arrangement adds thermal loading to the sample

and more uniformity to the temperature proﬁle, and makes for

a more accurate temperature calibration. However, this method

is less amenable to automated handling of the wafers and may

be disadvantageous in production environments that require high

throughput.

FIGURE 6.6 Illustration of a

sample susceptor for an RTA.

Another consideration relates to the efﬁciency of the heating

process. To reach 400

◦

C, the lamps do not require very much

power and will emit more infrared than visible radiation, much

of it below the GaAs bandgap and not efﬁciently absorbed by

GaAs. However, the metals in the ohmic metal do absorb the

infrared radiation effectively. In such a situation, the temperat-

ure calibration becomes difﬁcult to achieve, as one must take into

account the quantity and type of thin ﬁlm deposited on the wafer.

Obviously, the results can be ohmic-pattern dependent. Because

of the difﬁculty of accounting for the pattern in the temperature

190

Ohmic contacts

calibration, the second type of sample placement within an IR

absorbing susceptor is often preferable.

6.3.2 Measurements of ohmic contacts

The purpose of an ohmic contact is to provide a linear relationship

between current and applied voltage in a semiconductor device. It

is usually desirable to have low access resistance from the device to

the outside world. Consequently the measurement of ohmic con-

tacts to accurately determine their small but ﬁnite resistance is an

important part of semiconductor engineering and one which can be

the subject of detailed treatments. For the purpose of this chapter,

only two practical methods of measurement will be introduced.

More detailed treatment of other methods and other models can be

found in [4]. Both methods treated in this chapter will be based on

the planar contact common to FETs and the base and collector of

HBTs. The two methods differ in the test structure chosen, either

a linear contact array or a circular contact test structure.

FIGURE 6.7 Illustration of

linear contact array.

From a linear contact array illustrated in FIGURE 6.7 with

resistances deﬁned in FIGURE 6.2, the contact resistance R

given by

R = 2R

+ R

L/w (6.8)

where R

is the sheet resistance of the semiconductor between the

contacts, L is the length between the contacts and w is the width

of the contacts. In practice, the linear array contains contact pairs

with differing separations. The resistance of a contact pair is meas-

ured and plotted as a function of L, as illustrated in FIGURE 6.8

for a contact width of 100 μm. The slope of the resistance gives the

sheet resistance (201 /square after normalising length/width of

y = 2.0082x+0.5585

=0.9998

0102030

contact spacin

(

)

resistance (

)

FIGURE 6.8 Plot of resistance versus contact separation to extract contact

resistance.

191