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

Подождите немного. Документ загружается.

132 In situ characterization of thin ﬁlm growth

5.5 Further in situ SE examples

In situ applications have been an important eld of ellipsometry for at

least half a century. The importance of in situ SE is reected by topics of

papers given at international conferences on ellipsometry from 1963 to 2010

(Section 5.9.1). At recent conferences, as many as 10% of a few hundred

papers involved in situ Se. Current in situ studies can be loosely grouped

into topics such as (i) electrochemistry, (ii) compound semiconductor growth

and interfaces, (iii) processes related to uses of silicon in electronics, (iv)

biological interfaces, and (v) fundamental surface science. The category

we call ‘processes related to development of uses of silicon’ includes, for

example, studies of rapid thermal anneal of ion-implanted crystalline silicon,

growth of silicides, control of amorphous to polycrystalline silicon volume

composition ratio, or plasma etching dynamics of photoresists on silicon

surfaces.

The material systems benetting from in situ diagnostics, and thus topics

discussed at conferences, have evolved over the years due to changing

commercial interests. for example, in situ SE has now been used to study

SiGe for circuits; III–V semiconductors for optoelectronics; GaN and

related compounds for wide bandgap optical devices; amorphous Si, mc-Si,

CdTe or CIGS for photovoltaics; conducting polymers for electronics and

photovoltaics; biointerfaces for bio-implant materials and drug testing; and

numerous other material systems of commercial importance.

In this section we choose a few specic material and technology areas to

describe in detail:

∑ compound semiconductor growth and process control;

∑ nanomaterials, nanostructures, and processing for nanotechnology;

∑ optical coatings and multi-layer structures;

∑ biomaterials and bio-interfaces;

∑ in situ studies of materials and devices for photovoltaics.

5.5.1 Compound semiconductors

Compound semiconductors exhibit strong absorption above their bandgap

energies. Typical wavelengths for band-edge absorption range from the

ultraviolet to near-infrared. Optical constant dispersion shapes are affected

by temperature, composition, crystallinity, and other material properties.

Thus, in situ SE can be a useful tool to perform real-time diagnostics of

compound semiconductor properties and obtain layer thickness. In fact, in

the mid-1990s the promise of real-time monitoring and control for compound

semiconductors helped push development of fast, wide spectral range SE

systems that could cover appropriate photon energies of semiconductor

critical points.



133In situ spectroscopic ellipsometry (SE)

Much of this early work focused on Hg

1–x

Te, as maintaining correct

detector bandgap and optimum performance for infrared detectors required

composition control to better than ±0.002. In situ Se has demonstrated this

level of performance, as shown in Fig. 5.32 (Phillips et al., 2001) where

in situ SE composition is compared with ex situ fourier transform infrared

(FTIR) spectroscopy measurements for 50 MBE growth runs. The in situ

SE composition for 90% of the layers in this study fell within ±0.002 of the

desired composition, with a 5¥ improvement in standard deviation for runs with

SE control compared with runs without SE control. Composition measured

by in situ SE depends on well-calibrated optical libraries (Section 5.3.5),

which can be produced on the same process chamber later used for wafer

production. Phillips et al. (2001) went on to demonstrate real-time feedback

control of a linear composition prole (Figure 5.33) by adjusting the Te

effusion cell temperature based on in situ Se composition measurements.

5.5.2 Nanomaterials

Nanomaterials offer novel properties of interest for electronic, optic, photonic,

and sensor applications. The high surface sensitivity of SE makes it useful

in the study of thin lms and structures with nanometer dimensions. Optical

0.200 0.205 0.210 0.215 0.220 0.225 0.230 0.235

x (SE)

FTIR

= x

+ 0.0025

Dx = ± 0.002

x (FTIR)

0.240

0.235

0.230

0.225

0.220

0.215

0.210

0.205

0.200

5.32 Results from 50 growth runs show excellent agreement in Cd

composition (x) for Hg

1–x

Te thin ﬁlms determined with in situ SE

measurements when compared to ex situ Fourier transform infrared

(FTIR) transmission results. Ninety percent of runs have an accuracy

(Dx) within ±0.002 (reprinted with permission from Phillips et al.,

2001, J. Vac. Sci. Technol. B, vol. 19, no. 4, pp. 1580–1584. Copyright

2001, American Vacuum Society).



134 In situ characterization of thin ﬁlm growth

properties of these mixed materials have been described using EMA theory

(Aspnes, 1982). Here, the volume fraction and shape of each material affect

the polarizability of the mixture.

Oates and Christalle (2007) used the Maxwell Garnett EMA to model

silver nanoparticles in a poly(vinyl alcohol) (PVOH) host matrix. The optical

response from in situ SE, specically a plasmon resonance due to silver

nanoparticles, was used to estimate nanoparticle radius. Particle formation

for two curing temperatures is compared in Fig. 5.34. The EMA model only

considers silver particles as they become large enough to exhibit metallic

behavior, which explains why the increasing silver percentage coincides

with increasing particle radius.

Some nanostructures, such as nanorods, can exhibit shape anisotropy. EMA

theory can describe many of these structures, but must include a directional-

dependent parameter to give different polarizabilities parallel and perpendicular

to the structures. Hsu et al. (2008) demonstrated characterization of silicon

nanorods using an anisotropic EMA with variation in void percent as a

function of depth into the lm. Similar modeling during sputter deposition

of GaSb nanopillars led to determination of pillar height and both bottom

and top diameters for the pillars (Nerbø et al., 2009). In situ SE thickness

matched ex situ AFM thickness over a range of nanopillar heights from

about 20 to 80 nm.

ALD is another candidate for nanotechnology processing; implemented

as two self-limiting processes it offers the ability to grow thin lms one

monolayer at a time. Many researchers are demonstrating the importance

of in situ SE characterization for ALD processes. Langereis et al. (2009)

Target proﬁle

50 100 150 200 250 300 350

Time (min)

Cd fraction in HgCdTe

0.222

0.220

0.218

0.216

0.214

0.212

0.210

0.208

5.33 Demonstration of real-time feedback control to grow a linearly

graded HgCdTe ﬁlm (reprinted with permission from Phillips et al.,

2001, J. Vac. Sci. Technol. B, vol. 19, no. 4, pp. 1580–1584. Copyright

2001, American Vacuum Society).



135In situ spectroscopic ellipsometry (SE)

provide an excellent review of in situ SE applications for ALD processes and

their considerable efforts will not be repeated here; instead we will limit our

discussions to a few cases. For example, although ALD is considered an ideal

layer-by-layer growth process, nucleation effects can inhibit growth. Figure

5.35 shows ALD growth of a TiN lm with poor nucleation on surfaces with

low –OH densities, while the TaN shows immediate nucleation (Langereis

et al., 2009).

In situ SE monitoring also offers the benet of tracking variations in

growth rates as the lm changes. In situ SE measurement during ALD of

TiO

layers show the phase transition between amorphous and crystalline

(anatase) lm growth, as demonstrated in Fig. 5.36 (Langereis et al., 2009).

With these and other similarly important applications, the utility of in situ

SE continues to expand into new areas of processing for nanostructures.

5.5.3 Optical coatings

Optical coating production often relies on optical monitoring (R/T) or quartz

crystal monitoring during thin lm deposition. Quartz-crystal monitoring

is an inexpensive method of tracking growth rate, but it does not directly

measure the sample of interest. R/T sensitivity to very thin layers is greatly

150 °C

120 °C

0 10 20 30 40 50 60

Time (min)

Radius (nm)

% Silver

1.2

0.8

0.4

0.0

5.34 In situ SE measurements during curing of silver nanoparticles in

PVOH host at 120 °C (circles) and 150 °C (triangles) show the increase

in silver for the EMA model as the size of nanoparticles increases

(adapted from Oates and Christalle, 2007, J. Phys. Chem. C, vol. 111,

pp. 182–187, with permission from American Chemical Society).



136 In situ characterization of thin ﬁlm growth

diminished, although this is not typically a problem for optical coatings

as they often employ quarter-wave optical thickness layers. In situ Se can

supplement these methods to accurately determine refractive index, detect

index gradients, and monitor advanced multi-layer coating structures.

In situ Se measurements are often sensitive to refractive index values to

±0.001, which allows calibration of optical coating processes under various

TaN on HF-last Si

TiN on SiO

(1000 °C)

0 50 100 150 200

Number of cycles

Film thickness (nm)

5.35 Film thickness measured by in situ SE during ALD growth of TiN

and TaN ﬁlms demonstrating a nucleation effect for the former, but

not the latter (reprinted from Langereis et al., 2009, J. Phys. D: Appl.

Phys., vol. 42, 073001 (19 pp) with permission from IOP Publishing).

Amorphous Anatasez (nm) z (nm)

0 10 20 30 40 50

Thickness (nm)

1 µm

Growth rate (nm/cycle)

0.10

0.08

0.06

0.04

0.02

0.00

5.36 Growth rate determined from in situ SE during plasma-assisted

ALD of TiO

ﬁlm showing the shift from amorphous to anatase

crystal phase (reprinted from Langereis et al., 2009, J. Phys. D: Appl.

Phys., vol. 42, 073001 (19 pp) with permission from IOP Publishing).



137In situ spectroscopic ellipsometry (SE)

conditions. for example, larouche et al. (2001) studied the index of refraction

for TiO

/SiO

mixtures with varying composition, deposited by PECVD.

Figure 5.37 shows the non-linear trend of increasing refractive index with

increasing TiO

composition. Accurate calibration of this deposition behavior

allowed the authors to design and fabricate optical coating structures with

graded-index.

An example of a graded index optical lter was demonstrated by Amassian

et al. (2002), as shown in Fig. 5.15. The authors also utilized the sensitivity of

in situ SE to detect index gradients so they could (i) detect inhomogeneities,

(ii) study their origin, and (iii) nd process conditions to mitigate their effect

on optical design.

In situ SE has also been applied to multilayer stacks. Vergöhl et al. (1999)

demonstrated control of a four layer anti-reective coating stack using thin

lm calculations for all layers in the optical model. As mentioned earlier, this

approach is often impractical for graded or multilayer stacks where a large

number of layers need to be calculated. Nonetheless, Dligatch et al. (2004)

applied multi-wavelength in situ ellipsometry and photometric monitoring to

a multilayer optical coating with 79 layers. They used real-time monitoring

to correct for any deviation from the optical design as the layers were being

deposited.

5.5.4 Biological ﬁlms

Use of spectroscopic ellipsometers for biological lms has been well-

reviewed by Arwin (1998, 2000, 2005, 2011), and is an important topic for

0 0.2 0.4 0.6 0.8 1.0

[Ti]/([Si] + [Ti])

2.6

2.4

2.2

2.0

1.8

1.6

1.4

5.37 Calibration of refractive index for PECVD deposited thin ﬁlm

mixtures of TiO

and SiO

(reprinted from Larouche et al., 2001, 44th

Annual Technical Conference Proceedings-SVC, pp. 277–281, with

permission from the Society of Vacuum Coaters).



138 In situ characterization of thin ﬁlm growth

overview in this chapter. Most experiments are done in liquids, usually with

water as the solvent for proteins and other biomolecules. Common proteins

are biotin, avidin, g-globulin, ferritin, bronectin, bovine (or human) serum

albumin (BSA or HSA), and hemoglobin. Other biomaterials (molecularly

more complex) include antigen–antibody bound multilayers, enzymatic

catalysists, and enzymes (Berlind et al., 2010).

Most studies have been conducted on smooth interfaces with substrates

of crystalline silicon or various metals. Studies have also involved porous

silicon surfaces, where the pores offer an increased surface area for reaction,

with protein adsorption at the surface as well as penetration of protein into

the pores (Karlsson et al., 2005). Berlind et al. (2011) are also testing new

surfaces, such as carbon and carbon nitride, in their quest ‘for improved

materials for life science applications like biomaterials and biosensors’.

Unfortunately, the surfaces that are ideal for Se characterization are not

often used in practical application such as hip joint implants, heart valves,

and stents.

Metal surfaces present a few additional challenges as they are often rough

and/or oxidized, and their optical properties vary with both microstructure

and surface condition. A common method of addressing these issues is by

measuring the substrate prior to bio-lm interaction (preferably in liquid

solution) to determine reference optical constants for the current surface.

Gold, for instance, has no oxide which simplies analysis, but roughness

and microstructure still affect its optical constants. An interface layer can

be induced in the near-surface region of gold in response to interaction with

an adsorbed biological lm, which is speculated to be a region depleted of

free electrons (Mårtensson and Arwin, 1995; Mårtensson et al., 1995). This

region is optically modeled using an EMA layer to mix the gold and bio-lm

optical properties.

Gold easily reacts with thiols to allow systematic, fundamental scientic

studies of nominally well-oriented proteins with a wide range of chemical

species of self-assembled monolayers. Much research to date has been

dedicated to fundamental studies of substrate surface chemistry effects,

solution composition and concentration, solution pH, temperature, surface

hydrophobicity, diffusion of biomolecules from solution, competition between

multiple biomolecules simultaneously in solution, and other controllable

experimental conditions (Arwin, 2005; Berlind et al., 2008). Most of these

studies use ‘ideal biomolecules’ such as simple or common proteins.

SE measurements are perfectly suited to thin bio-lms, as the phase (D)

is very sensitive to sub-nanometer surface layers. Figure 5.38 shows an in

situ Se measurement through a liquid cell during cetyltrimethylammonium

bromide (CTAB) adsorption on a gold surface (Tiwald, 2007). CTAB was

added to the liquid solution twice during the measurement, followed each time

by multiple rinse steps. Kinetic ‘D’ data are shown for a single wavelength of



139In situ spectroscopic ellipsometry (SE)

600 nm, while the measurement consisted of 435 simultaneous wavelengths

across the visible and near-infrared. Data analysis of the complete spectrum

results in a smooth thickness curve (Fig. 5.38), as a result of averaging data

from hundreds of simultaneous wavelengths for each time-point.

Because many bio-lms are only a few nanometers thick, there is strong

correlation between thickness and refractive index which limits the ability to

determine both values independently. SE measurements are still very sensitive

to the thickness-index product (equivalent to surface mass per unit area), so

measurements are often considered without separating these two values. It is

most common to x the index of refraction at measured values from similar

thick lms. The xed index allows calculation of an approximate thickness,

Add CTAB

Rinse 1, 2 Rinse 3, 4

Add CTAB

0 10 20 30 40 50

Time (min)

CTAB thickness (nm)

Delta (degrees)

3.0

2.5

2.0

1.5

1.0

0.5

0.0

102.0

101.8

101.6

101.4

101.2

101.0

100.8

5.38 Measured D at wavelength of 600 nm and SE-determined

thickness during an in situ liquid-cell experiment. CTAB is added to

the liquid solution twice, followed by multiple rinse steps (adapted

with permission from Tiwald, 2007).



140 In situ characterization of thin ﬁlm growth

as with the CTAB in Fig. 5.38 where index was xed at 1.5. Arwin and

Aspnes (1984) successfully separated index and thickness from very thin

layers using the substrate spectral response to indicate the correct thickness.

However, many bio-lms do not meet the ‘ideal’ requirements for this

method to work. Biolayer lms at interfaces can (i) lack lateral uniformity

because fractional coverage is common over surfaces and complicates data

analysis, (ii) be inherently anisotropic, and (iii) congure themselves in

various geometries depending on interface surface chemistry, solution pH,

surface hydrophobicity, and other factors.

Several promising approaches for bio-lm interface studies using in situ

SE have recently evolved. Measurements at new wavelengths are exploring

bio-lms from the ultraviolet to the infrared range, where resonant absorptions

reveal chemical bonding information (Arwin et al., 2008). Total internal

reection ellipsometry (TIRE), also known as surface plasmon resonance

enhanced ellipsometry, uses a cell conguration such as shown in Fig. 5.29(c)

to provide increased sensitivity to very thin lms (Poksinski and Arwin,

2007). Unlike conventional surface plasmon resonance (SPR), which measures

reected intensity, SE uses the full polarization measurement over a wide

range of wavelengths and angles to increase sensitivity for a broad range

of experimental and surface conditions. Poksinski and Arwin (2004, 2007)

have demonstrated very sensitive measurements of protein adsorption using

TIRE with thickness resolution as low as 1 picometer, which corresponds to

a surface mass density of 100 pg/cm

. Nabok et al. (2005) compared TIRE

and conventional SPr for the detection of environmental toxins and found

signicantly better sensitivity for TIRE measurements.

In the past few years scientists have combined in situ SE with the use

of quartz crystal monitors (QCM), a long-established method for bio-lm

diagnostics. QCM technology has been commercialized and widely used for

decades. In combinations, in situ SE and QCM provide a host of synergistic

advantages, since both techniques can simultaneously sense the same lm at

the same time. This allows researchers to compare results without having to

worry about repeatability of lm properties sample-to-sample. In addition,

SE and QCM report a different thickness (surface mass density); SE reports

the total density of attached molecules while QCM reports the total mass

coupled to the surface, including both adsorbed molecules and any trapped

water. Rodenhausen and Schubert (2011) have proposed methods to interpret

the SE and QCM data simultaneously to better understand the formation of

adsorbed lms. Consider the simultaneous in situ SE and QCM measurements

during adsorption and rinse for a CTAB lm on gold shown in Fig. 5.39

(Rodenhausen et al., 2011). The adsorbed lm thickness quickly rises upon

injection of CTAB with a peak near 14 minutes. Note that, the SE and QCM

thicknesses are not equal due to entrapped water, as discussed above. The

authors also calculate an absorbent mass fraction, f

, to relate the QCM and



141In situ spectroscopic ellipsometry (SE)

SE thicknesses – which is useful when watching the phases of lm evolution.

During the rinse step, there is a pause suggesting a separation between the

removal of weakly bound and strongly bound CTAB molecules at the surface

(Rodenhausen et al., 2011).

Another promising area of bio-lm interface studies with SE is development

of biosensors for rapid testing of new drugs (Arwin, 2005). In sensing

applications, surfaces are systemically functionalized with specic chemistries.

On-chip optical sensing enables rapid medical analysis using bio-chip arrays.

Imaging ellipsometers may also play a role in bio-sensing, but the technique

is still in its infancy, especially for practical applications. lateral resolution

has been demonstrated to better than 5 nm (Jin et al., 1996), but most work

to date involves single-wavelength ellipsometry. Recent development of

imaging SE systems (Jin et al., 2011; Meng and Jin, 2011) may advance the

capabilities of such devices, but sampling rates, especially for spectroscopic

measurements, are still a limiting factor.

Inject CTAB Rinse

QCM-D

thickness

10 20 30 40

Time (min)

d (nm) or G (mg/m

)

1.0

0.8

0.6

0.4

0.2

0.0

1.0

0.5

0.0

5.39 SE and QCM thickness measured during CTAB adsorption and

rinse process. An absorbent mass fraction, f

is calculated from both

thicknesses to help interpret the ﬁlm phase evolution (adapted from

Rodenhausen et al., 2011, with permission from Elsevier).

