Ghodssi R., Lin P., MEMS Materials and Processes Handbook

Подождите немного. Документ загружается.

774 A.D. Raisanen

Thermal budget Activation of doped regions will generally require anneal or diffusion

temperatures >900

◦

C. Te mperature-sensitive components can be fabricated

on the wafer once the dopants have been activated, but will need to be

protected from t he etch solution by passivation layers or ﬁxturing.

Selectivity Selectivity between n and p silicon-doped regions is reported at 200:1 in KOH

[43] and as high as 3000:1 in EDP [31].

Viable etchants KOH, EDP, and TMAH all exhibit this effect.

Material

compatibility

Same general materials requirements as the underlying KOH, EDP, and TMAH

etch solution. Good electrical contacts must be made to the n-doped layer,

and these contacts must be protected from the etchant.

10.3 In Situ Doping

One of the most effective methods of doping a semiconductor material is during the

growth of that material. Incorporation of dopant atoms during the high-temperature

growth phase results in an undamaged crystal structure with activated dopants,

avoiding the need for subsequent annealing processes and large thermal budgets.

Unfortunately, only bulk material or blanket ﬁlms can be easily formed in this man-

ner, so complex device structures will generally require some additional processing

technique.

10.3.1 Chemical Vapor Deposition

Polysilicon is a polycrystalline form of silicon commonly used as a thin-ﬁlm

mechanical material in MEMS and as an electronic gate for MOS electronic devices.

Polysilicon is straightforward to deposit on a wafer surface with chemical vapor

deposition techniques, and is easily etched with a variety of wet chemical and dry

plasma processes. It is compatible with high temperatures, thermally matched to the

underlying silicon wafer, and can be doped to achieve a wide range of conductivity

values.

As stated above, polysilicon is typically deposited using a chemical vapor depo-

sition process in a low-pressure three-zone tube furnace, although the process is also

compatible with single-wafer deposition equipment such as used in cluster tools. A

common arrangement is illustrated in Fig. 10.16. Wafers stored vertically in a car-

rier are centered in a three-zone quartz hot wall furnace that is evacuated with a

vacuum pump. A throttle valve controlled by a pressure sensor is normally placed

at the output port to allow continuous closed loop control of the process pressure.

Reactants are introduced at the front of the tube. Silane (SiH

) is the most com-

mon reactant gas, but dichlorosilane (SiH

) and silicon tetrachloride (SiCl

)are

occasionally used as well. At the wafer surface, silane is broken down to produce

elemental silicon and hydrogen gas.

Minor process details, such as reactor vacuum integrity, carbon and oxygen

contamination, and wafer ﬁxturing details can produce signiﬁcant variations in

10 Doping Processes for MEMS 775

Fig. 10.16 Schematic of low-pressure chemical vapor deposition system

deposition rate and morphology [7], making quantitative comparison of literature

results difﬁcult. A typical undoped polysilicon process r ecipe used in the author’s

laboratory uses a deposition temperature of 650

◦

C, with 100 sccm SiH

ﬂow at 300

mTorr total pressure, which produces deposition rates of 110 Å/min. Thickness uni-

formities for a 5000 Å ﬁlm of <5% within wafer and 10% along the length of the

boat are typical. Three to four sacriﬁcial “dummy” wafers are placed at the front

and back of the boat to improve the thickness uniformity of the device wafers.

A small temperature gradient of 10–15

◦

C is generally added to each zone, with

the coolest zone closest to the gas inlet, and the hottest zone closest to the exhaust

port. The increase in reaction rate at higher temperatures along the wafer load helps

compensate for the reduction in SiH

partial pressure at the wafer surface as reactant

is consumed. Increasing silane gas ﬂow or pressure will increase the deposition rate

to some degree, but altering these parameters will also modify the uniformity within

the boat and within each wafer, so it is usually necessary to optimize temperature,

pressure, and gas ﬂow simultaneously to produce a viable process in each individual

reactor.

Polysilicon ﬁlm morphology and deposition rate depend strongly on deposition

temperature, with deposition temperatures below 575

◦

C producing ﬁne-grained or

amorphous ﬁlms and low deposition rates. Deposition temperatures above 625

◦

produce large column-shaped crystallites [44]. Intermediate temperatures produce

small microcrystallites or amorphous ﬁlms, depending on reactor conﬁguration,

impurities, and other details. Figure 10.17 illustrates the typical morphologies

observed in most reactors, with the transition between amorphous low-temperature

depositions and columnar crystallite depositions occurring gradually somewhere

between 525 and 625

◦

As long as sufﬁcient silane is delivered to the process, deposition rates follow

an Arrhenius relationship with an activation energy of about 1.7 eV [7]. The pro-

cess is then driven by the reaction rate at the wafer surface rather than by mass

transport effects, and good thickness uniformity can be maintained over all wafers.

776 A.D. Raisanen

Fig. 10.17 Polysilicon undoped morphology and in situ doping resistivity trends as a function

of deposition temperature. Minimum resistivities generally observed for high doping densities are

listed for in situ boron and phosphorus-doped ﬁlms

Lower silane ﬂows result in the deposition process being governed by mass trans-

port rates, and deposition rate, uniformity, and morphology then tend to be much

more sensitive to reactor geometry details.

By adding a suitable source gas to the input reactant stream during chemi-

cal vapor deposition, it is practical to dope the polysilicon grains during growth.

Phosphorus oxychloride (POCl

) is a liquid material that can be added to the reac-

tor using an inert gas bubbler or vapor draw system, and is used as a phosphorus

dopant in CVD systems. Diborane (B

), arsine (AsH

), and phosphine (PH

)are

gaseous materials that can be added to the reactor to provide boron, arsenic, and

phosphorus dopants, respectively. Typical dopant ﬂows are no more than a few per-

cent of the silane ﬂow, and using a dilute (e.g., 2% PH

in N

) s ource gas is not

uncommon for safety reasons [45].

It should be noted that every one of these materials (SiH

,POCl

,AsH

and PH

) is signiﬁcantly hazardous, with SiH

being pyrophoric, POCl

being cor-

rosive, and the rest being very toxic. Safety equipment including gas detectors and

waste scrubbers are critical for these processes. Most workers use in situ doped

polysilicon processes when a heavily doped polysilicon ﬁlm is desired, such as

that used for comb drives, local interconnect, gates, and other electrostatic devices.

Resistance control of in situ doped material for lighter-doping levels tends to be

poor between wafers and from run to run, often making simpler diffusion or ion

implant processes more attractive for these applications.

At low doping levels, dopants tend to segregate to the grain boundaries between

crystallites, producing little effect on the conductivity of the polycrystalline ﬁlm.

Resistivity of polysilicon at low dopant levels is then very sensitive to the exact mor-

phology of the ﬁlm. Dopants added to amorphous ﬁlms collect at the large numbers

of grain boundaries between the numerous small crystallites, where they are rela-

tively ineffective compared to the same dopant density added to more crystalline

ﬁlms grown at higher temperature, where more of the dopant is incorporated in the

10 Doping Processes for MEMS 777

crystallites themselves. This leads to a relatively abrupt decrease in the resistivity

observed for a polysilicon ﬁlm saturated by dopants as the morphology of the ﬁlm

goes from amorphous to columnar crystalline at increased deposition temperatures.

Addition of phosphorus (PH

) dopant during the polysilicon deposition pro-

cess has the effect of suppressing large crystallite growth until higher temperatures

are achieved relative to an undoped deposition, as illustrated schematically in

Fig. 10.17. The minimum r esistivity achievable tends to be in the range 10–

100  cm below about 625

◦

C, which corresponds to the range where amorphous

deposition occurs. At temperatures higher than about 625

◦

C, the columnar growth

mode predominates, and the minimum resistivity obtainable drops to about 0.001–

0.01  cm. Total deposition rates of polysilicon grown with phosphine and arsine

dopant gases are strongly suppressed [7, 46], with up to an order of magnitude

decrease in deposition rate observed with dopant ﬂows of more than a few percent

of the silane ﬂow.

Addition of boron (B

) dopant during the polysilicon deposition process

increases the deposition rate of polysilicon by as much as two to three times [46],

with little effect on the overall morphology. Low-resistivity polysilicon with large

microcrystallites can be obtained at temperatures as low as 525–550

◦

10.3.2 Crystal Growth and Epitaxy

Conventional crystal growth techniques are used to produce boules and wafers of a

given desired dopant concentration. A wide range of dopant concentrations is pos-

sible but only bulk material can be produced. Much of the silicon produced in the

world today is grown by the Czochralski technique illustrated in Fig. 10.18 [47, 48].

A quartz crucible containing very high-purity polysilicon and dopant materials such

as antimony or phosphorus is heated by a set of electric heaters until it is above the

melting point of silicon at 1414

◦

C. A small single-crystal seed of silicon is lowered

into the melt on a support rod, and then slowly withdrawn. The crystal orientation

of the boule can be selected by proper orientation of the seed crystal.

Fig. 10.18 Czochralski

silicon crystal growth method

778 A.D. Raisanen

New silicon grows on the seed from the melt, using the seed crystal structure

as a template, and cools as the boule is withdrawn. The crucible and the growing

boule of single-crystal silicon are continuously rotated in opposite directions as the

crystal is drawn. The entire apparatus is housed in an inert gas environment during

the process, which takes many hours depending on the size of the crystal boule being

grown.

This growth technique produces silicon of a doping level set by the amount of

dopant material included in the original melt. Oxygen, from crucible quartz dis-

solved into the melt by the molten silicon, and carbon, from sublimation of the

carbon heater elements used to heat the melt, are incorporated in the boule dur-

ing growth [48], with typical levels >10

−3

for oxygen and >10

−3

for

carbon. These impurities are often beneﬁcial in trapping and immobilizing trace

metallic impurities, but can also cause problems for some devices. For example,

oxygen precipitates in silicon used for anisotropic KOH etching can induce small

pitting defects that degrade the mirror ﬁnish of etch facets.

An alternative popular silicon crystal growth method is the ﬂoat-zone technique

schematically illustrated in Fig. 10.19. This method avoids use of a crucible or other

container that can introduce contaminants into the boule. A large polysilicon rod is

lowered gradually through a zone heated inductively by radio frequency energy. A

seed crystal at the lower end provides the template for single-crystal growth of the

larger boule. Silicon is melted by the RF power, and surface tension of the liquid

allows it to ﬂow down through the hot zone and recrystallize on the seed crystal.

Impurities in the initial polysilicon rod tend to stay segregated in the liquid zone

rather than incorporate in the growing single-crystal zone, so boules of extremely

high purity can be grown this way. Dopants may be introduced by incorporating

them in the original polysilicon rod, exposing the hot zone to a ﬂow of suitable

dopant gas such as phosphine or diborane, or even by feeding a rod of dopant mate-

rial into the molten zone and letting it melt [49]. Silicon with resistivities exceeding

Fig. 10.19 Silicon

single-crystal growth by the

ﬂoat-zone technique

10 Doping Processes for MEMS 779

-cm can be grown by this method for special applications such as radio

frequency integrated circuits, which require very low substrate conductivities.

Single-crystal epitaxial ﬁlms of silicon or other semiconductors are often grown

on wafers that have been prepared by bulk silicon growth techniques such as the

Czochralski or ﬂoat-zone techniques. Unlike the polycrystalline or amorphous mate-

rials grown with simple polysilicon deposition methods, these epitaxial growth

techniques replicate the crystal structure of the underlying wafer, producing a thin

layer of single-crystal material with optimized properties. Epitaxial techniques are

widely used to grow semiconductor layers with speciﬁc dopant levels, stacks of

layers with different composition or dopant levels (multilayers), and atomically

abrupt junctions not obtainable with ex situ doping techniques. Alloys such as SiGe

and compound semiconductors can be grown on substrates with the proper lattice

structure using heteroepitaxial growth methods.

A variety of epitaxial growth techniques exists such as liquid phase epitaxy, vapor

phase epitaxy (VPE), molecular beam epitaxy (MBE), atomic layer epitaxy (ALE),

and metal–organic vapor phase epitaxy (MOCVD or OMVPE). Most techniques

operate by heating a substrate to a large fraction of its melting temperature and

exposing i t to a ﬂow of reactants. High substrate temperatures allow for a high

mobility of adatoms on the surface, allowing deposited materials to conform to the

underlying crystal structure.

Vapor phase and metal–organic vapor phase epitaxial reactors are probably the

most common systems used for producing commercial quantities of material, as

these systems can achieve a fairly high throughput. See Fig. 10.20. Substrates

may be heated by simple electrical heating elements, optical, or radio frequency

energy, and the chamber walls are generally cooled (“cold wall” reactor), which

minimizes deposition on the walls. Horizontal and vertical conﬁgurations exist,

with some reactors accepting one wafer at a time, and others handling multiple

wafers. Subatmospheric pressure processes are typical, but full atmospheric pressure

processes are not unknown.

Fig. 10.20 Metal–organic vapor phase epitaxy (MOCVD) system schematic

780 A.D. Raisanen

A common conﬁguration in the microelectronics industry is the silicon epitax-

ial reactor, which ﬂows hydrogen and silane, dichlorosilane, or silicon tetrachloride

over multiple wafers at temperatures above 900

◦

C. Deposition kinetics is a com-

plex interaction between deposition of Si and etching from the Cl species in the

gas stream. Dopants may be introduced by adding phosphine or other dopant gases.

This produces a high-quality single-crystal ﬁlm at deposition rates that can approach

1 μm/min. This type of reactor would be used to place a low doping density sil-

icon epilayer over a heavily doped boron etch stop layer, for example, to allow

subsequent fabrication of transistors or other active devices on top of a released

diaphragm.

III-V epilayers for optoelectronic devices are commonly grown using liquid pre-

cursor materials delivered by bubbling hydrogen through a heated ﬂuid bath. Vapor

is entrained in the gas ﬂow and delivered to the reactor. Examples of common liq-

uid precursors for GaAs epitaxy include trimethylgallium (Ga(CH

)

) and tertiary

butyl arsine AsH

(t-C

), but a vast array of chemicals exists for aluminum, gal-

lium, arsenic, phosphorus, and boron sources for MOCVD growth. II-VI materials

such as Hg

1−x

Te, used in infrared detector devices, can also be proﬁtably grown

using MOCVD epitaxy techniques.

Molecular beam epitaxy (MBE) is a high vacuum deposition technique that

grows an epilayer on a substrate by exposing it to an atomic or molecular beam

of reactant. See Fig. 10.21. For example, elemental Si is used for deposition of Si

epilayers, and Al, Ga, and As are used to deposit Al

1−x

As alloy semiconduc-

tor epilayers. Beams of dopant atoms can also be added to the ﬂux of material

impinging on the substrate, allowing very precise control of dopant level and

alloy composition. Beams of solid-source material are generated by temperature-

controlled collimated evaporation cells (“effusion” or “Knudsen” cells) equipped

Fig. 10.21 Molecular beam epitaxy (MBE) system

10 Doping Processes for MEMS 781

with shutters to start and stop the molecular beam, and liquid and gaseous species

are introduced by special cracking cells or gas injectors.

Deposition rates are generally low, on the order of 1 μm/h or less. The vacuum

system is maintained at a low pressure (10

−7

−10

−11

torr) ensuring that chemical

reactions only occur at the substrate when the beams impinge on it, and to prevent

contaminants from being included in the growing ﬁlm. Large vacuum pumps and

liquid nitrogen cooled cryoshrouds help maintain low pressure and adsorb stray

gas molecules before they can be incorporated in the epilayer. Most MBE systems

can only deposit an epilayer on a single substrate at a time. This, coupled with the

low deposition rate, makes the MBE growth technique less popular for production

of commercial quantities of epitaxial material, but they are capable of producing

very precise structures with material composition and doping density modulated

at the monolayer level, which is often not practical with MOCVD or other high-

throughput techniques.

For example, “delta doping” (Fig. 10.22) is a technique used to produce an

extremely abrupt layer of dopants with very high dopant levels [50] that can actu-

ally exceed the solid solubility limits in bulk material. The dopants are conﬁned to

a narrow region of material, often as thin as a single atomic layer in the crystal.

This thin dopant layer alters the electronic structure of the semiconductor, forming

a narrow, deep quantum well in the conduction band that can trap electrons in the

form of a highly mobile two-dimensional “electron gas.” If the well is sufﬁciently

deep and narrow, a set of electronic subbands is also formed which have unique

optical and electronic properties. High-performance transistors can be fabricated by

this approach which takes advantage of the high mobile charge density in the con-

duction channel and resonant tunneling effects. Optical devices such as lasers and

detectors are also made possible by transitions between the well subbands.

Fig. 10.22 Delta doping

concept. A layer of very

heavy dopant density is

grown between two layers of

much lighter dopant density.

Conduction and valence

bands form a potential well

that can conﬁne electrons in a

two-dimensional layer

10.4 Diffusion

Diffusion processes to fully dope a thin-ﬁlm or produce a desired dopant pro-

ﬁle are common in the semiconductor and MEMS ﬁelds. These processes rely on

782 A.D. Raisanen

straightforward thermal diffusion of dopant species from areas of high concentra-

tion to areas of low concentration at high temperature, usually >900

◦

C in silicon

materials.

Diffusion of dopants in Si and other materials can be described by Fick’s law

of diffusion [50], where the concentration of dopant at time t and depth x below a

surface is usually expressed in simpliﬁed form as

∂C(x, t)

∂t

= D

∂

C(x, t)

∂x

, (10.2)

where D (in cm

/s) is the diffusivity of the dopant species in Si at a speciﬁed tem-

perature. More advanced models incorporate the fact that this diffusivity is actually

a function of dopant concentration and is affected by other factors such as oxidizing

or nitridizing ambient environments in the diffusion tube, but for reasonably low

concentrations and inert environments it can be regarded as a constant.

In the simplest possible model, diffusivity varies with temperature according to

an Arrhenius relationship,

D = D

exp



−



, (10.3)

where D

is the frequency factor and E

the activation energy for diffusion. More

complex models for D

incorporate corrections for concentration, crystal damage,

oxidation, and other factors [52, 53]. Values for D

and E

for a few typical silicon

dopants in single-crystal silicon are given in Table 10.1. Diffusivity will depend on

the concentration of interstitials, vacancies, grain boundaries, and other defects in

the crystal [54, 55], so it is clear that polycrystalline material will have a signiﬁcantly

higher diffusivity than single-crystal materials.

Table 10.1 illustrates that the diffusion constant for As in polysilicon is more

than 10

higher than in single-crystal silicon even at relatively low temperatures.

Diffusion through polysilicon ﬁlms is strongly inﬂuenced by the morphology of

the ﬁlm, but because the diffusivities are high and the ﬁlm thicknesses are low, in

practice even a short anneal will evenly distribute any dopant evenly throughout the

thickness of the ﬁlm. Lateral diffusion of a micron or more can be expected with

polysilicon even during short high-temperature anneals.

Table 10.1 Frequency factor D

and activation energy E

for calculating diffusion constants as

a function of temperature (Kelvin) for common dopants in Si. Values for single crystal (c-Si) and

polysilicon (poly) are listed

Dopant D

(cm

/s) E

(eV) Ref

B (c-Si) 2.64 3.6 900–1200

◦

C [Mathiot]

As (c-Si) 35.3 4.11 950–1150

◦

C [Ishikawa]

P (c-Si) 3.62 3.61 950–1150

◦

C [Ishikawa]

As (poly) 8.6 × 10

3.9 800

◦

C [Swaminathan]

10 Doping Processes for MEMS 783

Two solutions for Equation (10.2) are broadly useful, corresponding to the case

of (1) constant dopant concentration, and (2) ﬁxed total dopant quantity. Case 1 is

commonly encountered in gas and solid-state diffusion processes where the dopant

is continuously introduced during the process or is available from a large reservoir,

so the dopant is never depleted by the quantity that diffuses into the silicon surface.

The solution that satisﬁes (Equation (10.2)) under these conditions is

C(x, t) = C

erfc



√



, (10.4)

where x is the depth below the silicon surface in cm, C

is the ﬁxed dopant concen-

tration in the ambient (atoms/cm

−3

), D is the diffusivity from Equation (10.3), and t

is the amount of time that the sample is held at the diffusion temperature in seconds.

The erfc is the complementary error function, which is tabulated in many reference

works [56] and is even a native function supported in spreadsheet applications such

The second case is appropriate for instances where the total amount of dopant at

the start of the diffusion process is ﬁxed at some value and not replenished as it is

consumed by diffusion into the silicon. This is common f or materials such as spin-on

dopants or deposited glass thin-ﬁlm dopants, which have a ﬁnite amount of dopant

in a small reservoir. The solution to Equation (10.2) under this set of conditions is

then given as

C(x, t) =

√

πDt

exp



−

4Dt



, (10.5)

where Q

is the total quantity of dopant at the beginning of the anneal at the surface

in atoms/cm

10.4.1 Gas Phase Diffusion

Gas phase diffusion processes were widely used at the beginning of semiconductor

technology, but tend to be less popular now simply due to the toxic gases required.

Dopant atoms are introduced by ﬂowing a suitable gas such as phosphine, diborane,

POCl

, or BCL

into a furnace that maintains the substrates at a temperature sufﬁ-

cient to produce indiffusion. See Fig. 10.23. The constant gas ﬂow sets a constant

dopant concentration in t he tube ambient, providing appropriate conditions for the

application of Equation (10.4). A burn box or scrubber of some type is required

for dealing with the toxic efﬂuent from the tube, and the furnace must be sealed or

otherwise isolated from the laboratory environment to prevent even ppm levels of

dopant gas from escaping. Nitrogen or argon is usually admitted to the furnace to

provide an inert environment.