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

Подождите немного. Документ загружается.

122 In situ characterization of thin ﬁlm growth

electron diffraction (RHEED). This forced a non-ideal angle of incidence

above 80°, but avoided additional chamber modications to allow access

for SE. A general-purpose angle of incidence is 70° and many commercial

process chambers are now offered with existing ports at this angle to allow

in situ Se.

Modern SE ‘heads’ are compact and allow convenient attachment to most

chambers. However, space on a busy chamber may still be difcult to nd.

Beam steering has been implemented, with the use of prisms or mirrors, to

manipulate the light beam in and out of a chamber. Figure 5.24 shows a

special ‘prism-assisted’ method to allow optical access when optical ports are

not available, through a chamber lid positioned parallel to the sample surface

(Johs, 1999). Special care is required with extra reections to calibrate the

polarization effects, track possible changes to the plane of incidence, and

determine the actual angle of incidence at the sample surface. Fiber optics

also occupy a role in moving light from one point to another for optical

measurements. However, ber optics do not maintain light polarization over

a wide spectral range, so SE polarizing components remain between the ber

and sample.

The source head is positioned to send the incoming light beam to the

sample. Some tip-tilt adjustment may be necessary. The receiver head is

positioned to collect the measurement beam after interaction with the sample.

This may require translation of the receiver position; depending on chamber

geometry, sample angle precision, and measurement beam size. Large chamber

dimensions exacerbate the need for receiver translation. Tip-tilt adjustment

of the receiver head may also be required to align the collected beam within

the receiver optical path.

Small process chambers (e.g. liquid and temperature cells) may be mounted

to an ex situ Se system. Here, the alignment steps typically involve movement

(tip–tilt–translation) of the process chamber relative to the SE system. The

Sample

Right angle

prisms

69°

Windows

Chamber lid

5.24 Mechanical integration with ‘prism-assisted’ access to a vacuum

chamber through a chamber lid parallel to the sample surface

(adapted from Johs, 1999).



123In situ spectroscopic ellipsometry (SE)

process chamber is often aligned such that incoming light beam enters normal

to the entrance window. A well-machined design can mitigate many of the

alignment issues.

With proper alignment, the in situ SE will capture the reected light beam

from the sample. However, SE is typically sensitive to angle of incidence

to ±0.01° and thus requires accurate knowledge of correct angle to maintain

high-accuracy measurements. Angle of incidence can be altered by alignment

and should be determined prior to measurement. This is commonly achieved

by measuring a known sample, such as SiO

on Si. The optical constants

for both crystalline Si and thermal SiO

are well-known (Herzinger et al.,

1998), which allows ellipsometric characterization of the SiO

thickness and

angle of incidence. The measured angle will remain accurate provided the

sample position is retained. Alternately, the angle can often be determined

during in situ measurements, avoiding the need to determine angle from a

standard sample.

The SE alignment depends on the reproducibility of sample position. If the

sample is displaced along the sample normal (d

) as shown in Fig. 5.25(a),

it will lead to an offset of the reected measurement beam at the receiver

receiver

), according to:

receiver

= 2d

sin(f) [5.6]

where f is the angle of incidence. Thus, the beam displacement at the receiver

Sample

normal

receiver

(a)

(b)

5.25 The reﬂected beam position at the receiver plane is displaced

due to (a) sample offset normal to the surface or (b) sample tilt.



124 In situ characterization of thin ﬁlm growth

will equal the sample displacement at 30° angle of incidence and exceed

1.8¥ the sample displacement for angles above 65°.

If the sample is tilted, the resulting beam displacement depends on tilt

direction relative to the plane of incidence. Figure 5.25(b) shows a sample

tilted in the plane of incidence (q

), which deviates the reected beam angle

) by:

= 2q

[5.7]

The angular beam deviation for sample tilt perpendicular to the plane of

incidence (q

) can be calculated from (Liphardt, 2011):

cos(q

) = sin

(f) + cos

(f) cos(2q

) [5.8]

This produces an offset beam position at the receiver (d

receiver

), dependent

on the distance (D) between the sample and receiver, as:

receiver

= D tan(q

) [5.9]

Small tilt angles in the p- and s-planes produce beam displacement in the

p- and s-planes, respectively. At 75° angle of incidence, a sample tilt of 1°

will deviate the reected beam by 0.5° to 2.0°, depending on tilt direction

relative to the plane of incidence. This can produce signicant offsets in

the reected beam position at the receiver, particularly with large chamber

dimensions. For ‘D’ equal to 500 mm, a 1° tilt in p- and s-planes would

offset the beam location by 17.5 mm and 4.5 mm, respectively.

Many SE systems integrate alignment detectors in the receiver head to

allow alignment to the measurement beam. These detectors can also be used

to track alignment changes from one sample to the next, or even during a

process. This becomes increasingly important when considering beam wobble

due to substrate rotation, which is covered in the next section.

5.4.2 Sample rotation and beam wobble

To maintain lm uniformity, it is common to rotate the sample during

processing. However, any misalignment between sample surface normal and

the rotation axis will produce wobble and affect the reected beam position,

as shown in Fig. 5.26. The sample wobble produces a nearly ellipsoidal beam

trajectory at the receiver plane and affects the overall accuracy of in situ Se

measurements. The length and width of the ellipsoidal beam trajectory can

be calculated based on Equations [5.7]–[5.9].

length = 2D tan(2q) [5.10a]

Width = 2D tan{cos

–1

[sin

(f) + cos

(f) cos(2q)]} [5.10b]

Further, for small tilt angles, the width is approximately:

Width ª 2D tan(2q)cos(f) [5.10c]



125In situ spectroscopic ellipsometry (SE)

Maximum beam displacement is generally of most concern, and will occur

in the plane of incidence, according to Equation [5.10a].

Preferred levels of beam wobble are less than ±0.1°, but this can be

difcult to maintain, even with excellent mechanical design. Johs et al.

(1998b) demonstrated that wobble as high as ±0.5° can be mitigated with the

following methods: detector overll, data synchronization, special alignment

and calibration practices, and zone-averaging.

It is critical that the reected light reach the detector, but this can be

difcult to maintain and depends on chamber dimensions and the amount

of wobble. A beam with a large diameter will overll the detector aperture

and thus ensure that signal reaches the SE receiver, as shown in Fig. 5.26.

In this scenario, only a portion of the light beam is collected, which reduces

the signal-to-noise ratio. Fortunately, the measurements are still accurate

because SE data are a ratio of two polarization directions.

Both angle of incidence and plane of incidence can change as the beam

precesses. if measurement time is a fraction of the sample rotation speed, each

measurement will collect a different time-averaged range of beam locations.

To improve data accuracy, Johs et al. (1998b) suggest synchronizing the

measurement averaging time and sample rotation speed. This eliminates

‘beating’ effects in the data, caused by consecutive measurements that collect

different portions of the beam precession.

Tilt angle

Rotation

Beam

trajectory

Receiver

aperture

Sample

normal

Axis of

rotation

Beam trajectory

at receiver plane

Ellipsometer

beam

Ellipsometer

beam

5.26 Substrate rotation can produce beam wobble when the sample

normal is not aligned with the axis of rotation. This produces an

elliptical trajectory of the reﬂected beam at the receiver plane. A

large beam diameter can maintain overlap with the receiver aperture

as the beam precesses (adapted from Johs et al., Thin Solid Films,

vol. 313–314, pp. 490–495, 1998b, with permission).



126 In situ characterization of thin ﬁlm growth

It is also important to align the sample and calibrate the ellipsometer

while the sample is rotating under measurement conditions. This optimizes

SE alignment, calibration, and calculation of the average angle of incidence.

For high-accuracy measurements, zone-averaging is also recommended as a

means to cancel the rst-order ellipsometry errors due to misaligned plane of

incidence. Zone-averaging requires multiple SE measurements with polarizing

optics at positive and negative orientations. This does increase measurement

time which may not be feasible for rapid processes.

An alternate method to mitigate beam wobble has been demonstrated by

both Haberland et al. (1998) and Johs and He (2011). A spherical mirror

which reects the beam back onto the sample is attached opposite the source

head. The receiver head is located near the source head, collecting the return-

beam. This ‘return path’ design can cancel beam precession at the receiver,

which improves signal-to-noise. Haberland et al. (1998) concluded that this

method can easily compensate for wobble of nearly 2°. The optical model

used with such ‘return-path’ SE must consider that the measurement beam

reects from the sample twice. In addition, any mirrors or prisms that direct

the light beam to the receiver head can produce polarization effects, which

also need to be considered. Synchronization of the measurement time and

rotation speed is benecial, as the angle of incidence and plane of incidence

will still vary on the sample even as the beam wobble is mitigated. Figure

5.27 shows one realization of this method using a prism to direct the source-

beam into the chamber using the same window as the reected beam traveling

back to the receiver (Johs and He, 2011). Under standard conditions, tests

using ±0.8° substrate wobble and 1 m distance between sample and receiver

produced about 28 mm beam precession at the receiver. Implementing the

‘return-path’ approach reduced this precession to less than 1 mm.

To receiver head

From

source

head

Rotating

sample

Spherical

mirror

5.27 ‘Return-path’ ellipsometry method to compensate for substrate

wobble uses a spherical mirror to reﬂect light back on the sample.

Both source and receiver heads use the entrance window, with a

prism to direct the source beam in this implementation (adapted

from Johs and He, 2011).



127In situ spectroscopic ellipsometry (SE)

5.4.3 Window concerns

Windows often separate in situ SE optics from the process. However,

windows can affect the light beam through birefringence, reduced intensity,

and possible beam deviation or displacement. Window birefringence will

modify the light beam polarization and affect SE measurements; because of

this, anisotropic window materials, such as sapphire, are avoided. Amorphous

materials, such as fused silica glass, can have strain-induced birefringence

of a couple of degrees. Fortunately, calibration methods can successfully

correct the small window birefringence effects of common vacuum windows

(Jellison, 1999; Johs, 2000; Johs et al., 2001). Strain-induced birefringence

often depends on window mounting, so calibration is recommended when

windows are repositioned, remounted, placed under vacuum, or modi ed

by cleaning or heating.

Window retardance effects depend not only on the window material

but also on its dimensions. Standard fused silica  ats can have 3° to 5°

retardance, which is manageable through proper calibration. For very sensitive

applications, ‘strain-free’ windows, such as those from BOMCO Inc., exhibit

only 0.1° to 0.5° retardance. Special care should be taken when studying

anisotropic samples, as traditional window calibrations are designed for

isotropic measurements and may not correctly anticipate cross-polarization

possible from anisotropic samples.

Another undesirable condition occurs if windows become coated with an

absorbing material due to their close proximity to the deposition process. This

will affect signal-to-noise, but not SE accuracy as both p- and s-polarizations

are equally absorbed. Window coating can be a more signi cant problem for

in situ R/T, which measure light intensity, not polarization. To reduce window

coating, optical ports can be placed as far from the process as possible, gated,

heated, or even purged to reduce window exposure. Eventually, windows

may need to be cleaned or replaced.

Windows can also deviate the beam direction if the light beam is incident

non-normal to a window or the window is wedged. Figure 5.28 shows an

offset in beam position, d

window

, for a planar window, which can be calculated

for a given window thickness, d

, as:

ndow

–

)

[5.11]

where the refracted angles (f

, f

) can be calculated from the incident angle

) using Snell’s Law (Hecht, 1987, p. 84):

sinf

= n

sinf

= n

sinf

[5.12]

This effect is generally trivial unless the ambient index (n

, n

) differs on

each side of the window, as is common for liquid cells, where the window



128 In situ characterization of thin ﬁlm growth

is in contact with air on one side, and with the liquid on the other. The

light beam is redirected with both the displacement, d

window

, and an angular

deviation, q

, as shown in Fig. 5.28. Final beam location can be determined

by adding window displacement of Equation [5.11] with further displacement

through the liquid, calculated as:

liquid

= tan(q

)D = tan(f

– f

)D [5.13]

for a path length through the process (D). Of course, this is only an estimate,

as beam deviation may also occur at the exit window. The key to reduce

beam deviation is by either aligning the measurement beam normal to

the entry window or reducing path length within the liquid environment

(Hilker, 1995).

With a short path length, beam deviation may not be a concern. Byrne

et al. (2009) demonstrated in situ SE with a tubular ow cell that allows

variable angle of incidence measurements. While the light beam does not

enter a at window, the short path length within the cell minimizes beam

deviation. Their tubular design provides the additional capacity to map the

sample along a linear direction, a feature that was utilized to study protein

adsorption on binary graded metals, such as Al

1–x

5.4.4 Process condition

SE measurements are compatible with most process conditions. However,

background light or ambient absorption can have an effect on certain

measurements. For example, some plasma processes produce bright light, which

Window

liquid

window

Ellipsometer

beam

Sample

normal

5.28 Light beam is offset (d

window

) through a window when entering

at non-normal incidence. If ambient index is different on each side of

the window (e.g. liquid cell), an angular deviation (q

) can also occur.



129In situ spectroscopic ellipsometry (SE)

may affect the DC component of the SE signal. Fortunately, most modern

SE designs are chopped or optically modulated by rotation or modulation

of the polarizing optics. Using the AC signal components isolates the Se

measurement from uctuations in ambient light level.

Since SE is also a non-invasive measurement, it is not generally affected

by process environment. Light absorption is not typically a problem in

vacuum or air, although oxygen and/or water vapor will strongly absorb

vacuum ultraviolet (VUV) wavelengths. Thus, VUV ellipsometry requires

vacuum or purge throughout the entire optical path (Hilker, 2005). Various

gases, such as CO

, also absorb some bands in the infrared spectrum.

Light absorption in liquid ambient is more common. For this case, cell

designs reduce path length within the liquid or avoid traveling within the

liquid altogether. A standard liquid cell design is shown in Fig. 5.29(a).

Here, light travels through the liquid to reach the sample surface. Zudans, et

al. (2003) demonstrated in situ liquid interface studies with a measurement

beam entering the backside of a glass substrate, as shown in Fig. 5.29(b).

Poksinski and Arwin (2004) used a prism to accommodate internal reection

ellipsometry, which also bypasses the liquid, as shown in Fig. 5.29(c). Of

course, these designs place specic requirements on substrate and thin lm

choices.

5.4.5 Software integration

In situ SE can involve communication with the process to provide feedback

control. For example, end-point detection can be achieved by sending a

signal to close a shutter. Control of temperature or composition may involve

feedback to a heater or sputter gun current supply. Software integration

can also provide synchronization between the SE and a moving process.

Section 5.4.2 already described the benets of measurement averaging that

coincides with beam wobble. Another application of software integration is

found in processes that involve movement of the sample within the process.

Johs (2004) demonstrated in situ monitoring within a PECVD deposition

process that rotated both the samples and the planetary table. Due to the

non-specular nature of actual parts to be coated, such as gears, a NiCr-plated

cylinder was used as witness sample for in situ SE. The witness sample

rotated around the planetary table at three revolutions per minute, allowing

measurement for about 0.5 seconds during each 20 second revolution. High-

speed SE measurements at hundreds of wavelengths every 21 milliseconds

collected adequate information during this brief encounter with the sample.

Measurement triggering was congured to collect data only when adequate

detected light intensity reached the SE detector, signifying the witness sample

passing through the SE beam. Figure 5.30 shows a schematic of this set-up.

Film growth of 5 nm or greater would occur between sample rotations into



130 In situ characterization of thin ﬁlm growth

the SE beam position, but in situ Se could still monitor the process and

witness optical constant variations in the deposited diamond-like carbon

(DLC) lms at different stages of lm growth.

Software synchronization has also been used to communicate between

multiple ellipsometers along the same process chamber used to deposit

multilayer coatings. Peros and Kuester (2007) used multiple SE systems

to study each layer in multi-layer deposition for solar thermal applications.

5.29 Three liquid cell geometries to study the interface between

a surface and liquid: (a) standard geometry where light enters

windows and travels through the liquid to reach the surface, (b)

reverse-side ellipsometry through a glass substrate where light

does not have to travel through the liquid (adapted from Zudans et

al., 2003) and (c) total internal reﬂection ellipsometry set-up where

light enters a prism and does not have to travel through the liquid

(adapted from Poksinski and Arwin, 2007).

Window

Liquid

Sample

(a)

Light in

Air

Glass

Liquid

(b)

Thin

ﬁlms

Back reﬂection

Front reﬂection

Light in

Prism

Liquid out

Liquid in

Metal coated

glass slide

(c)



131In situ spectroscopic ellipsometry (SE)

Figure 5.31 depicts their roll-to-roll process, where results from the rst

ellipsometer are passed to the second ellipsometer, and so on. This allows

implementation of dynamic process control changes as needed for overall

optical coating performance.

SE beam

SE witness sample

(NiCr-plated ring)

Coated components

(gears, rollers, etc.)

Planetary table

(rotates at 3 rpm)

Magnetron

5.30 Top-view of the dual magnetron PECVD sputter chamber with

planetary rotation used to coat DLC on gears, rollers, and an SE

witness sample. The planetary table rotates at 3 rpm, which positions

the witness sample at in situ SE measurement location for a short

time-period each 20 seconds (adapted from Johs, Thin Solid Films,

vol. 455–456, pp. 632–638, 2004, with permission).

SiO

, CrO

, TiO

e-beam PVD

Glow discharge

Thickness from SE #1

Thickness from SE #2

Thickness from SE #3

Substrate

Thickness – SE #1

Thickness – SE #2

Substrate

Thickness – SE #1

Substrate

Unwinder Upwinder

Ellipsometer

Aluminum

sputter

5.31 Roll-to-roll coater incorporating multiple in-line SE systems

to determine multi-layer optical coating thicknesses (adapted from

Peros and Kuester, 2007, with permission).

