Middleton W.M. (ed.) Reference Data for Engineers: Radio, Electronics, Computer and Communications

Подождите немного. Документ загружается.

REFERENCE

DATA

FOR ENGINEERS

NONINVASIVE

PROGYOSIS

REAL

TIME

Fig.

57.

Hierarchy of

process

monitoring

and

control sensors for semiconductor manufacturing.

example, most contact-type sensors that require direct

mechanical contact to the wafer surface (e.g., thermo-

couples for wafer temperature measurements)

are

invasive. On the other hand, the noninvasive in situ

sensors

are

the most suitable type for device fabrica-

tion process control needs. These sensors usually

employ noninvasive probe beams or characteristic

detected beams of energy with well-defined physical

signatures (e.g., optical, acoustical, electronic, and

electromagnetic) to monitor the desired parameters

based on a known physical phenomenon. The ex situ

sensors

are

essentially for off-line preprocess and post-

process monitoring applications; however, the in situ

sensors

are

designed for real-time as well as on-line

pre- and postprocess monitoring and process control.

indicated in Fig.

57,

the real-time in situ sensors

can be employed for process end-point detection, real-

time feedback control, and process/equipment diagno-

sis and prognosis. For instance, an in situ real-time

sheet resistance sensor can

used within a metal

CVD process module to provide feedback process

control (for instance, to control the deposition rate)

and to achieve precise process end-point detection.

The same sensor may

also

provide data to feed-forward

links and to expert systems for equipment and process

prognosis/diagnosis purposes. The in situ preprocess

and postprocess sensors

are

mounted within a load-

lock chamber or a dedicated sensor (metrology)

mod-

ule interconnected to the vacuum-integrated cluster

equipment. These sensors

are

useful for next-wafer

(delayed) process end-point control, feed-forward

links, and equipment/process prognosis/diagnosis. The

specific choice of various types of in situ sensors

depends on various factors such as the application and

cost. For instance, depending on the process applica-

tion, the task of in situ dielectric thickness measure-

ment may either

performed

real time (for real-

time rate control and/or end-point detection) or follow-

ing the process by a postprocess sensor (for delayed

end-point control and feed-forward links).

The in situ sensors

are

intended to monitor various

parameter sets within three domains. These domains

are associated with the equipment, process, and wafer

parameters. Depending on a specific application, the

critical equipment parameters to be monitored may

include gas showerhead temperature (e.g., in PECVD

TEOS

oxide), background contamination (e.g.,

epi-

taxial Si

growth

and PVD processes), quartz window

temperature (in lamp-heated

RTP

systems), chamber

wall deposits, and energy source status (e.g., status of

INTEGRATED CIRCUITS

DIRECT FEEDBACK CONTROLLER

20-6

lamps and the total lamp power in RTP). Additional

equipment parameters are the gas flow rate and pres-

sure settings in a processing reactor. However, the pro-

cess parameters are directly governed by the

equipment parameters (both controlled and uncon-

trolled parameters). Depending

a specific applica-

tion, the process parameters of interest to monitor may

be wafer temperature, process gas flow field, plasma

density, and plasma electron energy.

general, a

semiconductor wafer exposed to a process contains the

most important parameters that have

be controlled.

This is due to the fact that the wafer parameters have a

direct impact

device performance and yield.

For

example, in a CVD or PVD (sputter deposition) pro-

cess used for deposition of dielectric or conductive

films, the important wafer-level parameters include

film thickness, refractive index, stress, conductive-

layer sheet resistance, and thickness uniformity.

Among the above-mentioned three parameter

domains, the equipment parameters

are

the easiest to

monitor. The equipment parameters can be directly

adjusted and controlled, whereas the process and wafer

parameters

are

dependent variables. Monitoring and

control of the desired process and wafer parameters

require specialized noninvasive in situ sensors and

advanced control techniques. Proper adjustments of

process and wafer parameters

are

usually performed

by varying the equipment parameters. The process

parameters are functions of the equipment variables

via equipment models, and the wafer parameters are

determined by the process parameters via process

models.

Sensors for Direct and Indirect Feedback Process

Control-Fig.

shows the direct and indirect meth-

ods of parameter measurement and device manufactur-

ing process control. The conventional methods of

manufacturing process control employ equipment-

level sensors to monitor the equipment parameters.

These measurements (e.g., lamp power, mass-flow

controller settings, etc.) provide the necessary real-

time signals for an indirect feedback control loop that

maintains the critical equipment parameters at or near

the desired predetermined set-point values. The use of

closed-loop control based

measured equipment

parameters usually provides a significant improvement

(in terms of process parameter spread) over the open-

loop mode of operation; however, this type of process

control relies

two levels of equipment and process

models in order to relate the equipment parameters to

the wafer parameters that are the ultimate parameters

of interest for manufacturing process control. These

models or the process-vs.-equipment and wafer-vs.-

process relationships can be affected by various

INDIRECT FEEDBACK CONTROLLER

EQUIPMENT

PARAMETERS

MEASUREMENT

EQUIPMENT

NOISE

PROCESS

PROCESS PARAMETERS

SENSORS

DIRECT

MEASUREMENT

SIGNALS

NOISE

SENSORS

PARAMETERS

Fig.

58.

Schematic illustration

direct and indirect process control techniques.

20-62

RF PHASE ANGLE

BIAS

VOLTAGE

PLASMA

EQUIPMENT

MODEL

TEMPERATURE

GAS SHOWER

HEAD

TEMPERATURE

sources of drift and noise resulting from the uncon-

trolled variables. For instance, the optical/vacuum

quartz window temperature in an RTP system behaves

as an uncontrolled variable that can cause process

drift. Although the uncontrolled equipment parameters

are not controlled directly by an operator, most of

them should be monitored to further refine the equip-

ment model for improved process control. For exam-

ple, real-time monitoring of the optical/vacuum quartz

window temperature in an RTP system can provide a

useful signal for improved closed-loop wafer tempera-

ture control and for equipment diagnosis.

As illustrated in Fig.

58,

the direct process control

techniques employ suitable in situ sensors to directly

measure the important process and/or wafer parame-

ters. These direct measurement signals are fed to a

controller, which continuously adjusts the equipment

parameters in real time

minimize the differences

between the measured signals and the process and/or

wafer parameter set points. It is evident that the direct

control methods should result in a much more robust

process contxol and smaller process and/or wafer

parameter spread. This is because the direct feedback

control loops contain the drift/noise sources within the

overall control loop. When in situ noninvasive sensors

INDIRECT SIGNALS

MICROWAVE POWER

RF CURRENT

are used for real-time monitoring of the desired wafer

parameters, the process control feedback loop will

minimize the error signal representing the difference

between the measured and set-point wafer parameter

values. As a result, the equipment-related drift effects

will be essentially eliminated.

As described earlier, the indirect equipment-related

process control signals

are

related to the direct process/

wafer-related process control parameters via the process

and equipment models.

an example, Fig.

illus-

trates some representative equipment and process/wafer

parameters

a typical plasma deposition or etch pro-

cess. The plasma process and wafer parameters (direct

process control signals such as plasma density, ion flux,

wafer temperature, and processing rate) are related to

the plasma equipment parameters (indirect process con-

trol signals such

RF/microwave power, dc bias volt-

age, and susceptor temperature) via the specific plasma

equipment model. Of course, the functional relation-

ships between the wafer and process parameters are

established by the physics of plasma and

the

plasma

process model.

general, a good knowledge of

the

equipment and process models is necessary to ensure

optimal and fast real-time process control based on

either direct or indirect feedback control techniques.

DIRECT SIGNALS

PLASMA

DENSITY

Fig.

59.

Typical wafer/process and

equipment

parameters

a plasma process

20-63

Critical Sensors for Device Fabrication Process

Control-The critical in situ process control sensors

required for a given fabrication process

are

determined

based on various considerations such as the impact of

the specified process on the overall device perfor-

mance and fabrication yield. The process requirements

for a given process indicate if any real-time and/or pre-

and postprocess in situ sensors must be utilized to

meet the target specifications for the process and wafer

parameters. Fig. 60 illustrates various single-wafer

processes and their associated “critical” in situ process

control sensors. For example, consider an RTP module

used for CMOS gate dielectric formation by rapid

thermal oxidation and nitridation (RTO/RTN) pro-

cesses. Thickness, growth temperature, and uniformity

of gate dielectrics are the main critical wafer and pro-

cess parameters that can influence the overall device

performance and fabrication yield. As a result, one

important sensor is a real-time multipoint temperature

sensor for closed-loop temperature control to improve

process repeatability. Another very useful sensor is an

in situ ellipsometer for postprocess measurement of

gate dielectric thickness and refractive index. This sen-

sor will help to further reduce the overall gate dielec-

tric thickness spread via delayed (next-wafer)

feedback. An additional postprocess in situ sensor may

be used for monitoring the gate dielectric thickness

uniformity. The in situ postprocess uniformity sensor

will assist in meeting the process uniformity specifica-

tion and can provide delayed feedback process control

information for process uniformity control, assuming

the RTP module employs a multizone illuminator for

dynamic and real-time uniformity control. If the tem-

perature sensor monitors multiple points on wafer for

dynamic uniformity control during the gate dielectric

formation process, the postprocess uniformity sensor

will be rather noncritical.

Various blocks shown in Fig. 60 illustrate some

selected critical in situ sensors for CVD processes

(metals, dielectrics, amorphous and polycrystalline sil-

icon), epitaxial growth, plasma etch (anisotropic and

isotropic

resist processing, ion implant processing,

dry surface cleaning, glass reflow, and silicide forma-

tion. As another example, consider the critical sensor

needs for metal CVD processes such as tungsten CVD

using rapid thermal CVD (RTCVD) modules. Again, a

temperature sensor is needed for real-time wafer tem-

perature control. Precise control of wafer temperature

and its uniformity in the CVD tungsten temperature

range (300-550

“C)

are critical requirements because

of their impact on the deposition rate and process uni-

formity. Another very useful sensor is a noninvasive

sheet resistance sensor for real-time monitoring of

tungsten sheet resistance during the RTCVD process.

This sensor will provide the process end-point infor-

mation and will help to reduce the process parameter

spread. Besides these real-time sensors, the process

can benefit from two postprocess sensors for

situ

monitoring of the metal film thickness and/or its sur-

face roughness.

Intelligent Cluster-Tool-Based Device Manufac-

turing-The earlier discussions have indicated that

the use of low-cost in situ sensors in conjunction with

single-wafer vacuum-integrated cluster equipment can

provide a fast-cycle-time processing medium to manu-

facture advanced semiconductor devices. This type of

factory relies on intelligent real-time process control

based on in situ sensors to minimize the device param-

eter spread and to enhance the overall equipment reli-

ability and process up-time.

Fig. 61 shows the simplified block diagram

CMOS process flow and its associated cluster equip-

ment for a fully cluster-based fabrication sequence.

Each cluster equipment contains three or four process

modules.

is expected that increasing the number of

process modules on a cluster equipment much beyond

four to six may result in reduced equipment reliability

and increased cost. The process flow starts with the

CMOS tank (or well) formation module and is fol-

lowed by the isolation, pregate, CMOS gate, source/

drain, silicide, planarized interlevel dielectric, contact,

metal, and passivation modules (assuming one metal

level). As shown, the CMOS tank module can be com-

pleted using two cluster tools: one for microlithogra-

phy marker pattern formation, and another one to

complete the CMOS well formation. The device isola-

tion module employs three cluster tools: one to com-

plete the oxidation mask stack deposition, another

cluster equipment for completion of the patterned oxi-

dation mask, and a third module for field oxidation and

oxidation mask removal. Other process modules

employ suitable cluster tools, as indicated. This

CMOS process flow employs a total of about

clus-

ter tools (each with three or four individual process

chambers) besides some additional equipment require-

ments such as the wet cleaning stations, ion implanter,

and microlithography stepper. Appropriate sensors are

included in each cluster equipment to allow real-time

process control and device manufacturing based on

intelligent equipment and factory control.

Figs. 62 and 63 show schematic diagrams of two dif-

ferent cluster tools with some real-time and pre- and

postprocess in situ sensors for the two CMOS flow

modules shown

Fig.

61.

Fig. 62 illustrates a three-

module cluster tool for CMOS gate structure forma-

tion. The first step in the process sequence uses the pre-

clean module (such as vapor-phase clean)

remove

surface contaminants including native oxide. Then, the

RTP module is used to form the gate dielectric by

RTO. The subsequent step

silicon gate deposition

LPCVD. All three modules require real-time tempera-

ture sensors (critical for RTP and LPCVDBTCVD but

less critical for the preclean module) for wafer temper-

ature measurements and process uniformity control.

The preclean module may also employ an ultraviolet-

light intensity sensor for photochemical cleaning pro-

cess monitoring and control. Moreover, the RTP mod-

ule uses a real-time thickness sensor for process end-

point detection. An alternative option would be to use a

postprocess in situ thickness sensor. Similarly, the

SHEET

RESISTANCE

(END-POINT)

TEMP.

(REAL-TIME)

METAL

FILM

THICKNESS

(POSTPROC.)

SURFACE

ROUGHNESS

(POSTPROC.)

THICKNESS

(ENDPOINT)

(REAL-TIME) (REAL-TIME)

REFRACTIVE

RESISTANCE

EPITAXIAL

GROWTH

THICKNESS

pk-l

(REAL-TIME)

REFLECTANCE

lS0TROPIC

Lfi

DIMENSION

RESISTANCE

FORMATION

(REAL-TlME)

SURFACE

REFLECTANCE

(END-POINT)

RESISTANCE

THICKNESS

SURFACE OXIDATION/

NITRIDATION

(POSTPROC.)

SUPS

(POSTPROC.)

ACTIVATION

DRIVEAN

ELECTRICAL

ACIlVATION

(ENDPOINT)

RESISTANCE

(POSTF'ROC.)

SURFACE

TOPOGRAPHY

(ENDPOINT)

THICKNESS

(P/P-PROC.)

(POSTPROC.)

RECOMB. UNIFORMITY

(PP-PROC.) (POSTPROC.) (P/P-PROC.)

Fig.

60.

Useful

situ

sensors

for process control

various semiconductor device fabrication processes.

PLASMA

EMISSION

(END-POINT)

PLASMA

DENSITY

(POSTPROC.)

PROCESSING

SURFACE

CONDITION

(END-POINT')

EMISSION

(END-POINT)

RESIST

THICKNESS

(F'REPROC.)

CRITICAL

DIMENSION

(POSTPROC.)

20-66

FLOWMODULES

r-€

PHFWTF

GATE

SILICIDE

MLO

FOR

COSTACT

MI.TA1.

PASSIVATION

LPCVD silicon process module also

uses

a silicon

thickness sensor for process end-point detection. The

load-lock

(or

metrology) chamber contains a slip sen-

sor

for postprocess monitoring of possible RTO-

induced slip dislocations. A sheet resistance sensor

may

used for postprocess monitoring of LPCVD in

situ-doped silicon sheet resistance. Another load-lock

sensor can

used to monitor the effectiveness of the

surface preclean process.

Fig.

illustrates the example of a four-module

cluster tool for self-aligned silicide formation. The

process sequence employs preclean for native oxide

removal, refractory metal sputtering,

RTP

silicide for-

mation, plasma etch for removal of unreacted refrac-

tory metal and metal nitride, vapor-phase clean, and

finally another

RTP

cycle for silicide annealing. All the

critical real-time and pre- and postprocess sensors

are

also shown.

RTP for Integrated Device

Processing

Conventional and state-of-the-art semiconductor

factories employ batch furnaces for many thermal fab-

rication steps.

Various

applications include gate and

field oxidations, low-pressure chemical-vapor deposi-

tion

(LPCVD)

(e.g., for polysilicon and silicon

nitride), and thermal anneals (e.g., for junction forma-

tion and glass reflow). A significant number of other

fabrication processes, however,

are

already based

SWP.

Examples include plasma etch and deposition,

ion implantation, and microlithography. The current

trend toward replacement of batch processes with

SWP

is evident from the increasing market share for

cluster tool equipment with

SWP

modules.

RTP

is an

obvious choice to replace batch furnaces for thermal

processing in a semiconductor minifactory environ-

ment. Advanced RTP tools

are

based on modular

designs compatible with vacuum-integrated cluster

equipment. High-performance multizone illuminators

with cylindrical symmetry have been applied to vari-

ous

processes including RTO, CVD

tungsten and

amorphous/polycrystalline

silicon, silicide formation,

sinter, epitaxy, as well as high-temperature rapid ther-

mal annealing (RTA).* Concurrent use of multizone

illuminators and multipoint temperature sensors pro-

vides a capability for real-time wafer temperature con-

trol and process uniformity optimization.? Moreover,

INTEGRATED CIRCUITS

20-67

LOAD-LOCK/METROLOGY CHAMBER

Fig.

62.

Schematic illustration

of a

cluster equipment for

CMOS

gate structure formation.

numerous in situ process monitoring sensors have been

developed to meet the RTP monitoring and control

requirements. Some process applications of advanced

load-locked single-wafer RTP techniques and associ-

ated sensors will be described.

Multizone

RTP

Illuminators-The known RTP

systems mostly employ single-zone tungsten-halogen

or arc lamp modules for wafer heating. The illuminator

is usually designed to provide relatively uniform

steady-state wafer heating over a narrow range of

wafer temperatures. The conventional RTP illumina-

tors provide real-time lamp power control capability

for only a single zone. This has been known to impose

major limitations in terms of dynamic wafer tempera-

ture uniformity and slip dislocation control.$ For a

given optical flux distribution on a semiconductor

wafer, temperature uniformity can be affected by vari-

ous

additional parameters such as gas flow rates and

process pressure. Moreover, the transient heat-up and

cool-down temperature uniformity of a wafer can be

Reference

50.

Reference

52.

quite different

from

the steady-state uniformity.$

These observations imply that the conventional single-

zone lamp RTP systems may not be able to meet the

stringent wafer temperature and process uniformity

requirements over an extended domain of process

parameters. Some developments in the area of RTP

equipment have addressed these requirements. One

approach has been to use a crossed lamp configuration,

which employs two parallel crossed banks of linear

tungsten-halogen lamps above and below the wafer.$

This technique is superior to the conventional single-

zone RTP designs; however, owing to lack of cylindri-

cal symmetry, only a limited real-time uniformity con-

trol can be achieved.$ This limitation is due to the

circular shape

semiconductor wafers and linear/

crossed configuration of

the

lamps.

Another approach has been to employ multizone

illuminators with overall cylindrical symmetry. As an

example, Fig.

shows a schematic illustration

Texas Instruments' Advanced/Automated Vacuum Pro-

cessor (to be called TI-AVP) using a three-zone illumi-

nator. The TI-AVP three-zone illuminator employs a

Reference

53.

20-68

REFERENCE

DATA

FOR

ENGINEERS

RTP

MODULE

OXIDE

THICKNESS

SURFACE

CONDITION

PVD

MODULE

METAL

THICKNESS

LOADLOCK/METROLCGY CHAMBER

Fig.

63.

Schematic picture of a cluster equipment for self-aligned silicide formation.

single center lamp, a first ring of 12 lamps, and an

outer ring of

lamps. The combination of

the

point-

source lamps in each ring simulates a continuous opti-

cal flux ring. All the zones

are

concentric with the cen-

tral axis of the reactor process chamber and the center

of the semiconductor wafer. The multizone lamp mod-

ule heats the semiconductor wafer from the rear. Each

lamp zone can

controlled independently in real time

via its dedicated remote and computer-controlled

power supply. The illuminator design provides multi-

ple hollow light-pipe holes through the illuminator

housing and associated reflector assembly for multi-

point real-time wafer temperature measurements and

control. This feature is important because various RTP

fabrication processes are performed over an extended

process parameter domain (e.g., 0.25-760 torr pres-

sure for

LPCVD

and

RTO

processes). Recent model-

ing results have shown that the process pressure and

gas flow rates can have significant effects on wafer

temperature uniformity, particularly near the wafer

edge.* In the

RTP

examples presented here, wafer pro-

cessing and heating is performed face-down. Fig.

shows a representative open-loop response of four

radially distributed temperature sensors over a

150-

mm wafer using a three-zone RTP illuminator

torr/

250

sccm [standard cubic centimeters per minute] flow

nitrogen).

With

optimum zone power ratios, this

illuminator provides temperature nonuniformity of

less than

f1.5

"C.

Multipoint Temperature Sensor-RTP applica-

tions require accurate and repeatable real-time control

of wafer temperature. The most commonly used non-

contact temperature measurement technique is optical

pyrometry.

The

advantages of optical pyrometry for

RTP include its noninvasive nature and relatively fast

measurement speed. Both of these requirements are

critical for proper control of transient heat-up and

cool-down temperature ramps and accurate and repeat-

able temperature control. The fundamental disadvan-

tage with conventional pyrometry, however,

its

dependence on

the

wafer surface roughness and emis-

sivity. Surface roughness is typically wafer dependent,

while emissivity depends on a range of parameters,

including temperature and chamber reflectivity, as well

wafer-doping characteristics and surface optical

condition.? Further, the direct reflection of heating

lamp light from

the

wafer surface must be eliminated

from the pyrometer measurement. Similarly, the heat-

Reference 54.

References 55-57.

INTEGRATED CIRCUITS

20-69

Fig.

64.

Schematic illustration

a Texas Instruments

single-wafer

(TI-AW) reactor

with

multizone lamp heating

module