Koster G., Rijnders G. (Eds.) In situ Characterization of Thin Film Growth

Подождите немного. Документ загружается.

112 In situ characterization of thin ﬁlm growth

–5

–10

–15

–20

1.0 1.5 2.0 2.5 3.0 3.5 4.0

Photon energy (eV)

(a)

d = 264.7 Å

d = 104.1 Å

d = 76.7 Å

d = 57.7 Å

d = 47.9 Å

d = 37.2 Å

d = 24.2 Å

–2

–4

–6

–8

1.0 1.5 2.0 2.5 3.0 3.5 4.0

Photon energy (eV)

(b)

d = 317.2 Å

d = 124.5 Å

d = 79.6 Å

d = 52.7 Å

d = 39.7 Å

d = 31.8 Å

d = 20.8 Å

5.12 Optical constants for thin metal ﬁlms determined from

combined in situ SE and transmission intensity measurements.

Optical properties for (a) cobalt and (b) titanium thin ﬁlms change

with thickness within the measured range up to 265 Å and 317 Å,

respectively (adapted from Pribil et al., 2004, with permission).



113In situ spectroscopic ellipsometry (SE)

with either high refractive indices or strong absorption. Thus, it fails with

transparent layers common to optical coating stacks. Other VI approaches

have been proposed using multiple time-points to allow a direct solution.

Urban and Tabet (1993) offer a calculation involving ve time-points to

balance the number of unknowns and measured values. More recently, Johs

(2004) proposed a general virtual interface (GVI) algorithm applicable with

any type of material using a minimum of three time-points. Johs demonstrated

the GVI approach during deposition of a diamond-like carbon lm on metal

in a plasma-enhanced chemical vapor deposition (PECVD) chamber.

Exp 350 nm

Exp 475 nm

Exp 600 nm

56 58 60 62 64

Time (min)

Psi (degrees)

5.13 In situ SE measurement of AlGaAs ﬁlm deposition shows

oscillations due to coherent interference when light can travel

through the ﬁlm and return to the surface. Light at 350 nm is strongly

absorbed by the ﬁlm, resulting in dampened interference oscillations.

Data at 475 nm and 600 nm continue to oscillate as thickness

increases, due to lower absorption at these wavelengths.

Actual sample

Layer #4

Layer #3

Layer #2

Layer #1

Substrate

Virtual

interface

Simpliﬁed optical model

Layer #4

Layer #3

‘Pseudo-substrate’

5.14 Virtual interface approach can simplify in situ SE data analysis

by approximating the underlying structure as a single interface.



114 In situ characterization of thin ﬁlm growth

5.3.3 Multilayer and graded ﬁlm characterization

Multilayers, with their many unknown properties, are often too complex for

ex situ characterization. In situ SE provides two key advantages: rst, each

layer of the structure can be studied individually, and second, multilayer

modeling can often be simplied with a VI approach. Graded lms are

modeled in a similar manner to multilayers, with a large number of layers

with varying optical constants to approximate the lm variation. For both

situations, a VI avoids the complex thin lm stack calculation for the entire

sample structure.

Kim and Collins (1995) demonstrated in situ SE analysis during PECVD

of compositionally graded amorphous silicon–carbon alloys. The H

-dilution

ratio was adjusted to produce void fraction variations through the lm, which

are measured by their effects on optical constants using a CPA approach.

Amassian et al. (2002) used GVI analysis to determine the refractive index

prole (Fig. 5.15) of ‘an apodized rugate lter deposited by PECVD from a

continuously varying TiCl

/SiCl

mixtures’, which produces mixed TiO

–SiO

lms.

5.3.4 Process control

Because in situ SE measurements can be collected in a fraction of a second,

they offer an opportunity to implement real-time process control, which is

important in many industrial applications. Real-time ellipsometry control

was rst demonstrated by Aspnes et al. (1992) for a parabolic quantum

well structure. In this study, a single wavelength was used to determine

composition of the outermost surface during AlGaAs deposition with feedback

500 1000 1500 2000

Thickness (nm)

Index, n at 550 nm

2.2

2.1

2.0

1.9

1.8

1.7

1.6

1.5

1.4

5.15 Index of refraction proﬁle versus ﬁlm thickness measured by in

situ SE for an apodized rugate ﬁlter (reprinted from Amassian et al.,

45th Annual Technical Conference Proceedings-SVC, pp. 250–255,

2002, with permission from Society of Vacuum Coaters).



115In situ spectroscopic ellipsometry (SE)

to the aluminum source for composition control. Though the composition

prole was controlled, well-width was established assuming growth rates

from previous calibration.

Multi-wavelength ellipsometry expanded the ability to control processes

in real-time. Growth-rate control of CdTe from metal-organic vapor phase

epitaxy (MOVPE) was demonstrated using a 12-wavelength ellipsometer

(Johs et al., 1993). The feedback of voltage to the Cd mass ow controller

was limited to every 3 seconds due to data acquisition and computer speed at

the time. kildemo, et al. (1996) used a three-wavelength in situ ellipsometry

measurement to control PECVD deposition of optical coating multilayers

with 1% accuracy. Because the coatings were transparent, a standard VI

approach was not successful. Instead, they predicted data trajectories in

advance of deposition for the measured structure. During the process, the

layer deposition was stopped when real-time experimental data approached

the calculated end-points.

Herzinger et al. (1996) demonstrated multilayer growth control of an

8-period Bragg reector structure with a VI approach that considers multiple

measurement times to adjust the model interface. Although analysis on a

90 MHz Pentium computer required 3 to 8 seconds per measurement time-

point, precise shutter control was achieved through stop-point predictions

based on the most recent measured thicknesses, as shown in Fig. 5.16.

Successful control to 1.3% thickness accuracy was achieved using a 44-

wavelength ellipsometer under non-ideal conditions (beam wobble, window

birefringence, and non-optimum angle) with an intentionally drifting process.

Today, modern computers and advanced SE systems with hundreds or even

thousands of wavelengths make in situ SE control signicantly better and

easier.

SE data

points

Target thickness

Shutter

time

9.9 10.2 10.5 10.8 11.1 11.4 11.7

Time (min)

Layer thickness (Å)

600

500

400

5.16 Demonstration of MBE growth control for a 583 Å GaAs layer

(adapted from Herzinger et al., 1996, with permission from the

Materials Research Society).



116 In situ characterization of thin ﬁlm growth

5.3.5 Process-dependent optical constant libraries

Material properties such as crystallinity, composition, conductivity, and strain

affect the optical constants of many materials. in addition, process conditions

such as temperature, pressure, and growth rate can also affect optical constants.

Because in situ Se can precisely measure optical constants, this technology

can also be used to monitor material and process conditions.

Volintiru, et al. (2008a) studied the effects of process conditions on the

refractive index of aluminum oxide thin lms. They observed that as substrate

temperature increased during remote plasma-enhanced metal-organic chemical

vapor deposition (MOCVD), lm density also increased which resulted in

higher refractive index. They also observed further index increases with

increasing bias voltages, as shown in Fig. 5.17.

Where the same process is applied to the same material (e.g. commercial

production), in situ SE is often used to create process-dependent or material-

dependent optical libraries. A series of materials with varying properties

are measured to determine optical constants related to each condition. A

critical point shifting algorithm (Snyder et al., 1990) is widely used to build

material libraries which vary the optical constants versus material property or

process condition. This approach can also be applied to materials with two

varying properties, such as quaternary compound semiconductors or ternary

semiconductors with both composition and temperature variation. Figure

5.18 shows the complex dielectric function for a series of Hg

1–x

Te thin

lms with varied composition, x, determined in situ at growth temperature

of 180 °C.

Refractive index

Deposition rate (nm/min)

1.65

1.60

1.55

1.50

1.45

1.40

0 10 20 30 40 50 60 70 80

–Bias voltage (V)

in situ

ex situ

5.17 Refractive index at 633 nm (measured with both in situ SE and

ex situ SE) and deposition rate versus bias voltage for aluminum

oxide thin ﬁlms measured by in situ SE (reprinted from Volintiru et

al., 2008a, with permission from Wiley-VCH).



117In situ spectroscopic ellipsometry (SE)

Material libraries can be created from a series of samples measured ex

situ. However, this is a tedious and slow process, often with sample exposure

to lm-changing environments like oxygen or moisture. The real benets

of in situ SE are the ability to measure (i) samples under varying process

conditions not available ex situ, (ii) virgin surfaces with little to no oxide, (iii)

surfaces exposed to controlled atmosphere, and (iv) a large number of thin

lms within a single process run. For the latter, virtual interface modeling

can help determine lm properties without concern for the underlying sample

structure. Figure 5.19 shows in situ Se data from a complex process run

with a total of 18 different thin lms. Process conditions, such as sputter

gun current and oxygen ow, were varied for each layer to produce series

of different Ta, a-Si, and SiO

layers.

x = 0.15

x = 0.19

x = 0.23

x = 0.27

1.8 2.1 2.4 2.7 3.0 3.3 3.6

Photon energy (eV)

5.18 Optical constant library developed for Hg

1–x

Te thin ﬁlms at

180 °C.

250 nm

500 nm

950 nm

Layer number

1 2 3 4 5 17 18

0 5 10 15 20 110 115 120 125 130 135 140 145

Time (min)

Psi (degrees)

5.19 In situ SE data from a sputter-process run with 18 various Ta,

a-Si, and SiO

layers. Process conditions were varied for each layer

to calibrate their effect on ﬁlm properties.



118 In situ characterization of thin ﬁlm growth

Data analysis of each layer is facilitated with a virtual interface to avoid,

in some cases, tens of under-layers in the model. From this single run, growth

rates for both Ta and a-Si lms were calibrated, as shown in Fig. 5.20. In

addition, the evolution of optical constants for each material was determined

in relation to process condition.

5.3.6 Surface and interface quality

The surface sensitivity of SE measurements encourages study of both interface

and surface quality. Surface quality can be an important measurement before

and after growth. In situ Se can also monitor the surface development during

processing. Johs et al. (1997) demonstrated in situ Se monitoring of a CdZnTe

substrate surface before Hg

1–x

Te lm deposition, as shown in Fig. 5.21.

SE measurements determined both surface quality and substrate temperature

during the heat-clean procedure. As temperature increased, amorphous Te

desorbed from the surface, followed by oxide desorption to produce a clean

underlying surface for lm growth.

Surface evolution can also be studied during growth. Fujiwara et al.

(2003) monitored the formation of interface and surface roughness during

microcrystalline silicon lm growth. Figure 5.22 shows the surface, interface,

and bulk layer thickness evolution for two microcrystalline silicon lms on

different substrates. An interface of 18 Å forms on the ZnO:Ga surface, but

not on SiO

, which was conrmed with cross-sectional transmission electron

microscopy (TEM) on the same samples. Evolution of bulk lm and surface

are similar for both samples, after interface formation is complete.

In situ Se measurements during the process are sensitive to interface

development. Johs et al. (1998a) compared two interfaces for lm growth

Ta rate

a-Si rate

Linear (Ta rate)

Linear (a-Si rate)

0 0.1 0.2 0.3 0.4 0.5

Gun current (A)

Growth rate (Å/s)

3.5

3.0

2.5

2.0

1.5

1.0

0.5

5.20 Ta and a-Si growth rate calibration versus gun current

developed from in situ SE measurements of multiple layers.



119In situ spectroscopic ellipsometry (SE)

Thickness (Å)

Amorphous Te desorbs

Temperature (°C)

Oxide

desorbs

Thickness

Temperature

0 10 20 30 40 50

Time (min)

390

350

310

270

230

190

150

5.21 In situ SE during the preparation of CdZnTe substrate before

HgCdTe ﬁlm growth provides surface layer thickness and substrate

temperature. As the substrate is heated, an amorphous Te layer

is desorbed, followed by the native oxide desorbtion – ultimately

producing an ideal surface to start ﬁlm growth (reprinted from Johs

et al., III-Vs Review, vol. 10, no. 5, ‘Real-time process control with in

situ spectroscopic ellipsometry’, pp. 42, 1997, with permission from

Elsevier).

on inP. The in

0.53

0.47

As layer produced an ideal interface, while the poor

data t quality for In

0.52

0.48

As layer suggested a rough interface. This was

substantiated by successfully modeling the rough interface with an effective

medium approximation (EMA).

5.4 In situ considerations

In this section, we review basic integration issues specic to in situ Se

measurements, including:

∑ mechanical integration;

∑ sample alignment;

∑ rotating samples with beam wobble;

∑ window concerns;

∑ environmental conditions;

∑ software integration.

5.4.1 Mechanical integration

One of the most important requirements for in situ ellipsometry is optical

access to the sample during processing. The ellipsometer source and receiver

‘heads’ are typically external to the process area, so multiple holes are

needed in the chamber at appropriate locations for incoming and outgoing

light beams. A clear path for the light beam within the process chamber is



120 In situ characterization of thin ﬁlm growth

also necessary. Optical ports should be positioned to allow an appropriate

angle of incidence between the measurement light beam and sample surface.

A common limitation of in situ measurements is a xed angle of incidence,

so angle choice can be important. Preferred angles are between 60° and

75° relative to the sample normal. However, angles from 35° to 85° may

be acceptable, although with reduced sensitivity, depending on the sample

to be studied.

There are three considerations for appropriate in situ angles. first, is

5.22 Evolution of interface, surface roughness, and bulk thickness

measured by in situ SE during growth of microcrystalline silicon

ﬁlms (reprinted from Fujiwara et al., 2003, with permission from the

American Institute of Physics).

800

600

400

200

800

600

400

200

Layer thickness (Å)

Surface roughness

Interface

Bulk

µc-Si:H

Interface

layer

18 Å (ﬁxed)

(a) µc-Si:H/SiO

(b) µc-Si:H/ZnO:Ga

0 10 20 30 40 50

Time (min)



121In situ spectroscopic ellipsometry (SE)

adequate optical access to the process chamber available for in situ Se?

When necessary, beam steering via prisms or mirrors may be necessary to

manipulate the measurement beam to an appropriate angle of incidence.

Second, does the angle provide measurement sensitivity to the material

properties of interest? This is less critical with modern ellipsometers, which

collect accurate SE measurements even at non-optimal angles. However, the

polarization change is reduced as angle of incidence moves signi cantly away

from the Brewster condition and angles below 45° are rarely useful. Third,

does the angle produce a large spot size on the sample? As angle of incidence

increases, so does the projected length of the light beam on the substrate.

This can be important for small or laterally non-uniform samples. The spot

size on the sample surface will remain the width of the beam diameter, but

with a projected length along the plane of incidence dependent on angle of

incidence (f) as:

lengt

length lengt

Diameter

cos(

)

pot

beam

Diameter

beam

Diameter

)

[5.5]

Thus, the spot length projects to 2¥, 3¥, and 4¥ the beam diameter near 60°,

70°, and 75°, respectively.

Figure 5.23 shows the integration of an in situ Se on sputter deposition

chamber (Woollam, et al., 1996). An existing chamber was retro t with

ports positioned at 75° to monitor above one of four sputter guns. In the

molecular beam epitaxy (MBE) system used by Herzinger et al. (1996),

in situ SE ports were adapted to existing ports for re ection high-energy

Spectroscopic

ellipsometer

‘source head’

Spectroscopic

ellipsometer

‘receiver head’

Plasma

Sample

Light

beam

5.23 Mechanical integration of in situ SE to Kurt J. Lesker sputter

deposition chamber (adapted from Woollam et al., SPIE Proc., vol.

2873, pp. 140–143, 1996, with permission from SPIE).

