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

Подождите немного. Документ загружается.

224 In situ characterization of thin ﬁlm growth

this means that the absorbed fraction (Nsl) is 1% for a density of 10

over a 10 cm path length. Molecules with strong bands at >200 nm can

be detected by using standard UV quartz optics, while those that are only

<200 nm require vacuum UV techniques and are usually not practical, except

in special cases to be discussed later.

There are a number of different light sources that can be used to measure

absorption. For example, a tungsten lamp in the optical and a deuterium lamp

in the UV can be used, or alternately a Xe arc lamp. Of particular interest

are atomic resonance lamps known as ‘hollow cathode lamps’ (HCL) that

are available for most elements. They emit resonance light in the visible

and near-UV that can excite their ground state atoms in the vapor. In the

next section we discuss the sensors based on these lamps. Another possible

light source is the tunable diode laser (TDL). These lasers are able to tune

through a narrow range of wavelengths (typically 2–10 nm) and if there is a

fortuitous match with the atomic resonance of the species of interest then a

sensor can be obtained. Some atomic gas species such as halogens, O, and

N absorb in the vacuum ultraviolet (VUV) (110–160 nm) which requires

vacuum enclosures of the optical paths.

Figure 8.5 shows schematically what happens during the vapor absorption

process for the two light sources discussed here. The HCL produces a broad

lineshape caused by Doppler broadening from the lamp emission prole

(~2 GHz) compared to the evaporated atoms (~0.5 GHz). The HCL-AA

systems give a number that is equal to the integral of the signal lineshape

Source

Hollow cathode lamp

Tunable diode laser

Vapor absorption Signal

From HCL

From TDL

8.5 Comparison of the hollow cathode lamp (HCL) source and

tunable-diode laser (TDL) source AA. With an HCL-AA system the

source is broader than the absorption line and one only gets an

absorption value that does not reﬂect pure Beer’s law. With TDL-

AAS the source is much narrower and one can obtain a complete

absorption spectrum that can be ﬁtted to Beer’s law.



225In situ deposition vapor monitoring

after absorption. Absolute calibration of the atomic density is complicated

by deviations from Beer’s law in HCL-AA systems. TDL light sources, on

the other hand, produce an extremely narrow lineshape (0.3 MHz) that can

be swept across the vapor absorption line; typical tuning range is 100 GHz.

We call this technique TDL-atomic absorption spectroscopy (TDL-AAS),

emphasizing that it is spectroscopic in nature, unlike HCL-AA which only

measures an absorption integral. Figure 8.5 shows a comparison of the TDL

with HCL sources. The TDL-AAS signal can be tted to Beer’s law exactly

as long as the light intensity is not too high causing absorption saturation,

also known as bleaching.

8.5.1 HCL-based AA vapor monitors

AA vapor sensing using HCL has been used in several types of deposition

processes including sputtering, plasma-assisted deposition, laser-assisted

deposition, and more common evaporation. The rst AA commercial

instruments were developed by ULVAC (Japan) starting in the late 1970s.

Following that, Lu et al. demonstrated an HCL-AA instrument for deposition

rate monitoring in the late 1980s (Lu et al., 1989; Missert et al., 1989). They

used AA for evaporation rate control during co-deposition of Y, Ba, and

Cu for in situ growth of the high temperature superconducting lms. Figure

8.6 shows a schematic of the MBE system that was used with a simple

Beam splitter

Lamp

Lens

Filter

Computer

Si photodiode

Shutter

controller

8.6 Schematic of the apparatus to monitor ﬂux by absorption and

consequently MBE growth rate. Reprinted with permission from

Chalmers and Killeen, Applied Physics Letters, 63, 3131–3133 (1993).



226 In situ characterization of thin ﬁlm growth

AA vapor monitoring by Chalmers and Killeen somewhat later (Chalmers

and Killeen, 1993). The authors used the optical-based ux monitoring to

control the MBE source shutter operation by timing the shutter opening for

the desired amount of deposited material.

Klausmeier-Brown et al. (1992) used a similar system and measured the

atomic ux of Bi, Sr, Ca, and Cu by AA during atomic layer-by-layer MBE

growth of high-temperature superconducting lms. Benerofe et al. (1994)

developed a dual-beam atomic absorption system to measure Sr and Ru vapor

during co-evaporation of SrRuO

. The reference arm path compensates for

light uctuations from the lamp, and the normalized absorption signal could

be fed back to the source for stable control of evaporant ux, as shown in

Fig. 8.7. These methods are stable and fast (~200 ms), but sensitivity depends

on the element.

Lu and Guan (1995) have described a different dual beam scheme, as

shown in Fig. 8.8, that improves the baseline stability by correcting for

viewport coating and other optical alignment shifts with a reference xenon

lamp (ltered by a monochromator) through the same reactor and reference

paths as the hollow cathode lamp. The two lamps are individually controlled

by a timing circuit.

Hollow

cathode lamp

e-beam

feedback

2.45 GHz

ECR

Chamber

Low-gain

ampliﬁer

Lock-in

ampliﬁers

Sig out Ref out

Absorption rate

Divider

rate = 10 – 10

sig

ref

Reference freq.

Signal freq.

Transimpedance

ampliﬁer

Spectrometer/

PMT

8.7 Dual-beam HCL-AA system with a reference path used to correct

the lamp ﬂuctuations (ECR = electron cyclotron resonance). Reprinted

with permission from Benerofe et al., Journal of Vacuum Science and

American Vacuum Society.



227In situ deposition vapor monitoring

Receiving ﬁber

optic cable

Sampling

channel

Gated

integrator

Gated

integrator

Signal

processor

and

controller

Control

output

Hollow cathode

lamp drive

Xenon ﬂash

lamp drive

Reference

channel

L1 HCL

Emitting ﬁber

optic cable

XFL

Evaporation

source power

supply

Substrate

RS-485

Monitor

Keyboard

8.8 Dual beam HCL-AA system with a reference xenon ﬂash lamp to correct for viewport coating and optical alignment

shifts. The system uses a common optical path for automatic correction of transmission. In addition to the HCL, a

xenon ﬂash lamp (XFL) is used for determining optical transmission. Light from both sources is combined in a beam

splitter (BS) and then focused onto the ﬁber optic cable by quartz lenses (L1 and L2). The ﬁber optic cable guides

the light into the chamber and on the opposite side a receiving probe collects the light to a second ﬁber optic cable.

A monochromator (M1) is used to isolate the speciﬁc atomic emission line. A silicon photodetector (D1) is used to

measure the light intensity. A similar photodetection arrangement (M2 and D2) is used to analyze a portion of the light

from each source coming off of the beam splitter and determines the reference channel. All the signal intensities are

compared in a microprocessor-based signal processing unit. Reprinted with permission from Lu and Guan, Journal of



228 In situ characterization of thin ﬁlm growth

8.5.2 TDL absorption spectroscopy

As already described, tunable diode lasers offer very narrow coherent

light sources that can be swept across the absorption line of the vapor to

obtain a complete absorption spectrum. This technique was demonstrated

extensively in the 1990s, in particular in a collaboration between groups at

Stanford University and New Focus, Inc. (Wang et al., 1995b, 1996, 1997).

They demonstrated frequency modulation, frequency doubling of the laser

wavelength using second harmonic generation in a waveguide, and atomic

velocity measurements using two light beams.

sensitivity in measuring small absorbed fractions can be improved

dramatically by frequency modulation (FM) of the laser frequency and

homodyne detection of the signal photocurrent at the modulation frequency

or a harmonic. This technique has been employed for TDL-AAS by Wang

et al. (1995a), as shown in Fig. 8.9. Laser source noise becomes very small

with detection at these high modulation frequencies. Frequency modulation

can be achieved by adding an FM component to the DC injection current

in the diode laser. When w

is less than the frequency half-width of the

absorption line, this method is commonly called wavelength-modulation

(WM) spectroscopy or derivative spectroscopy. In that case the modulation

is usually applied by use of an external electro-optic modulator.

Since the laser lineshape is very narrow, it can exactly reproduce the AA

Isolator

Diode

Laser

QPM-SHG

Waveguide

HCL

CIG QCM

Detector

Aperture

Evaporation

source

Function

generator

Lock-in

ampliﬁer

Feedback

controller

8.9 Frequency doubled-diode-laser-based atomic absorption monitor

(QPM-SHG = quasi-phase matched second harmonic generation

device). Reprinted with permission from Wang et al., Applied Physics

Physics.



229In situ deposition vapor monitoring

in the vapor plume and an accurate spectrum can be obtained. By measuring

the absorption along the direction of the expanding vapor, one can obtain a

component of the directional velocity and measure the corresponding Doppler

shift. The real atomic plume has a nite angular spread in space, and the

velocity has a distribution, resulting in a spectral spread in the absorption

prole due to the Doppler broadening. By measuring two spectra, one with a

light beam with a component in the direction of the velocity and one against

it, two peaks can be measured whose separation will be the average velocity

times the cosine of the angle at which the light intersects the beam. Figure

8.10 shows two different schemes: cross-beam and counter-propagating-beam

scheme (Wang et al., 1999). The cross-beam scheme is better in principle

as separate spectra that do not need to be tted are obtained. The counter-

propagating beam scheme is simpler to implement in a chamber and is

usually sufcient if the peaks are well separated. Both schemes have been

demonstrated to work in implementing ux measurements (Matijasevic and

Slycke, 1998; Wang et al., 1999).

AA vapor monitoring techniques have no limitation on process operating

pressure, are non-intrusive, i.e., they can measure right in front of the sample

surface, and are atomic species specic. With proper vacuum system design

the sampling region can be extremely close to the sample surface for precise

control of lm composition and thickness, something that is not possible with

other techniques. These are strong advantages of AA techniques. However,

the effectiveness of a particular AA monitor still depends on the details of

the specic atomic transition (absorption coefcient) that is used and on the

application.

Substrate/QCM

Laser

beam 1

Source

8.10 Two different schemes for measuring the Doppler shift with

two laser beams traversing the vapor at an angle. Reprinted with

permission from Wang et al., Journal of Vacuum Science and

American Vacuum Society.

Laser

beam 2



230 In situ characterization of thin ﬁlm growth

8.6 Summary of techniques and resources

Table 8.1 summarizes the characteristics of the six sensors discussed in

detail in this chapter. Note that only the QCM measures a true mass ux.

Furthermore, only the optical absorption techniques are non-intrusive.

The following is a partial list of commercial suppliers of deposition rate

instruments:

∑ QCM controllers: Incon (www.incon.com), East Syracuse, New York,

supplies a whole range of QCM controllers including IC/5 and IC6

instruments that are capable of doing auto-Z correction and accurate

measurement of uxes down to 0.01 Å/s.; Maxtech and Sigma Instruments

QCM controllers now sold through Incon; Sycon, (www.sycon.com),

East Syracuse, New York, supplies crystal heads and controllers; SRS

Stanford Research Systems, supplies low-cost QCM controllers; ColnaTec

(www.colnatec.com), Gilbert, AZ, supplies sensor heads including a

24-crystal head; Tangidyne (tangidyne.us), formerly Virginia Sensors,

supplies sensor heads, including a high-temperature (up to 900°C) sensor

head that uses a gallium orthophosphate crystal; Intellimetrics (www.

intellimetrics.com); Ulvac (www.ulvac-vacuum.com).

∑ Chopped ion gauge monitors: no longer available or supported, but many

are in use using newer greases on the bearings.

∑ EIES: Incon (www.incon.com), East Syracuse, New York, supplies

the EIES system Guardian, formerly sold as Sentinel.

Table 8.1 Comparison table of the deposition vapor sensors

Sensor Measures Characteristic Sensitivity Sampling

rate

QCM Flux Material non-speciﬁc

Intrusive

0.001 nm/s 1–10 Hz

Ion gauge Vapor density Material non-speciﬁc

Intrusive

0.0001 nm/s kHz

QMS Vapor density Material speciﬁc

Intrusive

0.0001 nm/s kHz

EIES Vapor density Material (atomic)

speciﬁc

Intrusive

0.0001–0.01 nm/s

depending on the

element

10 Hz

HCL-AA Vapor density Material (atomic)

speciﬁc

Non-intrusive

0.01–10 nm/s

(depends on the

atomic absorption

line)

1–10 Hz

TDL-AAS Vapor density

and velocity

(hence can

obtain ﬂux)

Material (atomic)

speciﬁc

Non-intrusive

0.0001 nm/s 10 Hz



231In situ deposition vapor monitoring

∑ QMS: Hiden Analytical Ltd (www.hidenanalytical.com), UK, supplies

QMS systems designed for vapor monitoring and source control, control

of up to 16 signals.

∑ Atomic absorption: SVTA (www.svta.com), Eden Prairie, MN, supplies

an HCL-AA system, AccuFlux Process Controller. A unique optical-

electronic design is claimed to correct for drift. UV (~200 nm) optics

is available as needed for atomic Se, Te, Sb, and As. The system

originally developed by Lu and Guan, described in section 8.5.1 was

sold commercially under the trade name ‘Atomicas’ for over a decade.

However, the instrument is no longer manufactured or supported. TDL-

aas systems are not available commercially, but tdLs are available

for a number of wavelength ranges from the following suppliers, among

others: New Focus (www.newfocus.com), Santa Clara, CA; Toptica

(www.toptica.com), Graefelng (Munich), Germany.

8.7 Case studies

We discuss some specic thin lm processes and their rate monitoring

implementations.

8.7.1 MBE growth of oxide ﬁlms with AA rate

monitoring

MBE is a very precise evaporation technique usually practiced in ultra-high

vacuum, although more recently oxides and nitrides are deposited by MBE

at slightly elevated pressures of activated gas species. In order to achieve

very accurate control of deposition rates one usually employs effusion cells

that have very stable rates due to a stable temperature in the cell. Even

so, deposition rates drop over time as the source material is depleted. To

compensate for these slow changes in evaporation rates one can periodically

adjust the source temperature (typically raise it) by doing a calibration.

Conventionally such a calibration is commonly done with a simple ion gauge

that is placed in the sample position. Typically an MBE system is equipped

with shutters for each source, so each source can be calibrated separately.

However, this requires moving the sample out while the calibration is done

and the process is time consuming.

Several research groups have implemented their own versions of an HCL-

AA system. As already discussed Klausmeier-Brown et al. and Chalmers and

Killeen were the rst to report such implementations for their MBE systems.

More recently the Santa Barbara group (Jackson et al., 1999) has reported

improvements in reduction of long-term drift by additional chopping of the

atomic beam at low frequencies. Note that the shuttered growth in MBE

is ideal for aa as one can continuously correct for baseline drifts during



232 In situ characterization of thin ﬁlm growth

periods when there is no ux. Bozovic and Matijasevic (2000) implemented

a similar system in the Oxxel MBE that is now used at Brookhaven National

Laboratory. Typical rates are very low in MBE deposition so one is not so

concerned about a constant rate as for the integrated ux, or thickness of

deposit. Thus the AA monitor signal can be integrated and the shutter can

be controlled to attain a precise thickness of the deposit in each layer.

8.7.2 Co-deposition of Y, Ba and Cu during reactive

co-evaporation of YBa

Much research has been devoted in the last two decades to making high-

quality lms of YBa

(YBCO) high-temperature superconductor;

see for example the review by Matijasevic and Bozovic (1996). YBCO

material is considered to be the most practical for many applications of high

temperature superconductivity. Of the various approaches for making YBCO

thin lms, one particularly attractive and inexpensive method is co-evaporation

from elemental sources. This method has three major challenges: uniform

large-area heating, relatively high oxygen pressure during evaporation, and

accurate atomic ux control. The large-area uniform heating and high oxygen

pressure have been resolved rather elegantly with a method developed by

Kinder and coworkers with a mechanical rotation of the substrate that allows

for a black body heater with an oxygen pocket to be used (Berberich et

al., 1993). As for the elemental ux control several approaches have been

employed.

The rst and still most widely used method is to use QCMs with

‘collimators’, or tubes, in front of them (Kinder et al., 1997). Collimators

can select the source for the vapor to sample and block out the adjacent

sources. This method works very well in high vacuum. However, at pressures

above 10

–5

torr there is a signicant pressure effect. Figure 8.11 shows the

pressure dependence of the QCM with a collimator. The comparison is for

a QCM that is held at the sample position and has no collimator as pictured

in Fig. 8.11, while only one source is used for evaporation. There is a fairly

signicant effect in ux for the pressure of interest in this process, about

5 ¥ 10

–5

torr. This effect is explained by scattering from the background

gas. Some atoms are scattered away from the control QCM but cannot be

scattered back due to the collimator, whereas atoms can scatter into the

sample QCM. An additional problem is that each atomic species pressure

effect is somewhat different and hence the atomic ratio will change as the

pressure varies. The pressure effect can be mitigated if one can maintain

the pressure in the chamber exactly during deposition and from run to run.

However, maintaining a constant pressure is difcult to do in practice for

this process since there are a number of different pressure regions in the

system that all need to be kept constant.



233In situ deposition vapor monitoring

During the 1990s as atomic absorption sensors were being developed, both

based on HCL and TDL, as described in Section 8.5, there was an industrial

effort as well to implement these new sensors in manufacturing of YBCO

wafers for microwave applications (Matijasevic and Slycke, 1998). This

resulted in detailed comparisons of the various sensors that can be used in

this process. Figure 8.12 shows a comparison of the pressure dependences

for these sensors. QCMs were used with collimators. For the TDL-AAS

sensor, a measurement of vapor density and velocity was used to yield a

ux. Only the TDL-AAS ux sensor was deemed to be capable of attaining

1% composition control at pressures of 5 ¥ 10

–5

torr.

Ratio of deposition rates (sample/control)

2.0

1.5

1.0

0.5

–7

–6

–5

–4

–3

Chamber pressure (torr)

Sample QCM Control QCM

Source

8.11 Pressure dependence of the control QCM for evaporation from

a Cu source. The schematic shows the scattering effect presumed to

cause the pressure dependence. Atoms can scatter into the sample

QCM from higher angles, but they cannot scatter into the control

QCM due to the collimator.

