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

Подождите немного. Документ загружается.

194 In situ characterization of thin ﬁlm growth

refraction effects become dominant. EDS detectors have a large sensitive

area and accept radiation over a large solid angle. The acceptance solid

angle must be reduced in order to block X-rays emitted from the chamber

walls by X-ray uorescence or BSE impact. A collimator system consisting

of successive apertures is inserted between sample and detector. The Si(Li)

detector crystal must be cooled in order to reduce the thermal background

noise level. Cooling can use liquid nitrogen contained in a Dewar tank or,

more simply, stacked Peltier cooling stages. Most detectors are sealed with

a Be window in order to keep the crystal free of moisture condensation and

contamination. In addition, the foil removes and blocks the light and BSE.

Unfortunately, the Be window also acts as a lter absorbing the lower energy

part of X-ray photons. The sensitivity of the detector is progressively reduced

for photon energies below about 1 keV, cutting off, for instance, the oxygen

Ka radiation. The Be foil can be replaced with lower Z materials, such as

sapphire and C (in the form of polymer materials), or totally removed by

carefully keeping the detector under controlled vacuum conditions. Because

low Z window materials do not efciently block fast BSE, a strong permanent

magnetic can be used to dump the electrons before reaching the detector.

An additional requirement is to keep the detector crystal and the window

foil free of material deposition. This is very important for quantitative analyses

because any deposit will cause a nonlinear, selective absorption of X-rays

and will selectively modify the intensities of measured peaks. A solution for

materials with a low evaporation temperature is to use a heat control able to

keep the foil clean during the process. This technique is used for elimination

of the deposition of As during GaAs/AlGaAs by Pellegrino et al. (1998).

Another approach is the use of an easily replaceable thin Be foil (Sun et al.,

2009). In case the detector head cannot be easily accessed, moving a lm like

Mylar in front of the detector is another possible approach. Finally, X-ray

energy dispersive detectors are not bakeable and must be placed far enough

from the chamber wall or mounted on a mechanical retraction to move the

detector far enough to stay at ambient temperature.

Quantication of characteristic X-ray line spectra

The signal intensity I

(x)

of an element A is given by general Formula 7.7.

The matrix factor is given by the product

M = w

· d/cos (q

) B · F [7.11]

of the uorescence yield w

for transition i, the backscattering factor B, the

uorescence factor F, and the attenuation length, corrected for the takeoff

angle q

. the factor B becomes more signicant when both incident and take

off angles are small and is calculated by Monte Carlo simulations (CASINO).

the standard method for quantitative calculation of atomic concentrations



195Spectroscopies combined with RHEED

is described by Reimer (1985, p365) and referred to as the ZAF procedure.

The method is based on the use of reference spectra from standards and

followed by a linear combination of the standards to t the measured signals.

The accuracy of this simple method is much improved by introducing a

correction factor for the Z dependence. It is assumed that the concentration

is proportional to the ratio between the measured and the standard signal

(corresponding to 100% concentration). Then the correction due to matrix

effects, the ZAF factor, is introduced, taking into account variations of the

BSE backscattering factor for different Z, of the absorption coefcient, and

of the uorescence factors between the actual sample and the standards.

EDS-XRF combined with RHEED was used by Pellegrino et al. (1998)

for quantitative monitoring of the surface composition of epitaxially grown

InGaAs on GaAs. The stability of the signal over a long period of time

is very reproducible and the accuracy of the mole fraction of In is <0.1%

after background correction. Quantitative analyses are performed by tting

Gaussian peaks for each element. The full width half maximum (FWHM)

of the distribution is known to be the energy resolution of the detector and

deconvolution of multiple, overlapping lines is therefore accurate and reliable

(Hashimoto et al., 2009).

The surface sensitivity of EDS-XRF was tested by Ino et al. (1980) for the

deposition of thin Ag layers on Si(111) surface. The detection limit is very

good, less than 1% of a monolayer. The measurement of a high-Z material

deposited on low-Z substrate is favorable for obtaining the highest detection

limit because the bremsstrahlung background from the substrate is low and

the PE backscattering and uorescence effects are minimized.

7.4.3 Increasing the surface sensitivity with RHEED-

TRAXS

The signal in XRF is dominated by the bulk emission and has a limited

surface sensitivity. When the take-off angle of the X-ray signal is limited to a

small angular range very close to the surface plane, the X-ray signal becomes

highly surface sensitive (Hasegawa et al., 1985). TRAXS is achieved when

the geometrical conditions in Fig. 7.7 are fullled (Wang, 1996, p356). The

X-rays emitted from a surface atom A at decreasing angle with the surface,

labeled (1) and (2), are weakly refracted. The refractive index for X-rays is

slightly smaller than unity and the emerging angle is larger than the incidence

angle. X-rays emitted from A parallel to the surface (3) are refracted into

vacuum at an angle q

. This angle is the critical angle for total refraction in

XRF. Reversing the beam direction, X-rays with an incidence angle equal

or smaller than q

will be totally reected from the surface. There is no

radiation emerging from the surface layer in the angular range between q

and the surface plane. The signal from deeper atomic layers B will not be



196 In situ characterization of thin ﬁlm growth

emitted under the critical angle because the path to the surface is large and

they will be absorbed. As absorption will cause uorescence, a part of the

signal from deeper layers may be re-emitted by the surface atoms, a process

similar to the backscattering of electrons. The high surface sensitivity of

this technique relies rst on the small angular range selected by the detector

close to the total reection angle suppressing contributions from larger angles

(and deeper layers) and second on the strong absorption of X-rays emitted

at the critical angle but originating from deeper atomic layers. The depth

of information is related to the decay length of the evanescent wave and is

estimated in the order of 2 to 3 nm (Hasegawa et al., 1985).

Measurement of the critical angle in RHEED-TRAXS

The values for the critical angle are in the milliradian range and afford a

high mechanical stability and accuracy of the positioning of the collimator

unit. In principle, the take-off angle must be adjusted for each line when

working under critical angle conditions. The critical angle for a compound

is given by the simplied formula

= 7.35 ¥ 10

–5

l (r SZ/SM)

1/2

[7.12]

for an X-ray of wavelength l and a material of density r, mass number M,

and atomic number Z. The critical angle depends both on the X-ray line

energy and on the composition of the matrix. The angle is larger for lower

X-ray energies (longer wavelength), for instance q

= 1.77° for Mg Ka

(1254 eV) in MgO and only q

= 0.33° for Fe Ka ( 6403 eV) in BaFe

(Sun et al., 2009). Accurate measurements of q

are given by Chandril et

al. (2009) and Tompkins et al. (2006).

Measurement of layer thickness using attenuation by RHEED-TRAXS and

in situ XPS

A different way to determine the thickness of a lm during deposition is

to measure the attenuation of a signal originating from the substrate. An

7.7 Beam geometry for critical angle X-ray spectroscopy (TRAXS).



197Spectroscopies combined with RHEED

example is given by Sun et al. (2009) for the deposition of MgO onto SiC.

The relative intensity changes were observed under xed RHEED electron

energy (12.5 keV). Monitoring substrate Si Ka intensity decrease with lm

growth provides a real-time lm thickness measurement. The lm thickness

is calibrated using in situ XPS measurements of the attenuation of the Si 2p3

photoelectron line at different angular incidence (angular resolved XPS).

Measurement of multilayer thickness using TRAXS and X-ray reectivity

Chandrill et al. (2009) demonstrated that the angular dependence of XRF

near the critical angle can be analyzed within the kinematic approach to yield

useful structural information not only about a single thin lm but also for a

multilayered structure. Hence, it is possible to perform in situ quantitative

structural characterization without the use of any reference samples. The X-ray

reectivity of a multilayer is measured ex situ using an X-ray beam of the

same energy and scanned over the same angular range as the detected waves

in TRAXS. This is like reversing the path of the X-ray waves and, according

to the reciprocity theorem, the intensity of the X-rays emitted from a depth

z inside a medium and observed at a grazing exit angle is proportional to

the X-ray intensity, which is impinging at the same grazing incident angle

and observed at the depth z inside the medium. Angular intensity proles

measured from multilayers of single layers of Mn and Y as well as bilayers

containing Y on Mn and Mn on Y allow the determination of the layered

structure and of the smoothness of the interface surfaces.

7.4.4 In situ reﬂection energy loss spectroscopy

Electron energy loss spectroscopy (EELS) is a familiar technique used in

transmission electron microscopes for measuring the ionization losses suffered

by the PE beam (Disko et al., 1992). The PE beam of high energy E

(larger

than 50 keV) is transmitted through a thin sample foil and is energy ltered by

a high-energy resolution analyzer. The energy distribution of the transmitted

electrons is measured and shows a strong direct transmitted peak at E

electrons transmitted with no energy loss. The intensity decreases very sharply

with increasing loss energies and shows step-like structure corresponding

to the ionization threshold energies E

of core level electrons. As not only

core level ionizations but also CEL losses like plasmons are possible, the

observed step can be shifted by one or more plasmon energies depending on

the IMFP for plasmon excitation and lm thickness (Egerton, 1992, p41).

The thickness of the sample is selected to optimize the ionization yield and

to keep the plasmon losses to a minimum.

REELS is a similar technique that is used in reection instead of

transmission. There is a major difference between the transmission and



198 In situ characterization of thin ﬁlm growth

reection modes. In transmission and under optimal thickness conditions,

the PE direct beam is straight and its angular extension is limited by the

small analyzer acceptance angle (typically 50 milliradian). The ionization

process leads to a smaller scattering angle, especially at larger PE kinetic

energies, and is accepted by the lter whereas elastically scattered PE and

BSE are rejected. In reection mode and under RHEED conditions, the

total scattering angle between the PE beam and detector is large. Elastic

scattering collisions are needed to backscatter the PE beam and consequently

it is mixed with the ux of BSE emerging from the surface. In contrast with

the transmission mode, the ionization loss structures sit on top of a larger

distribution of BSE (see Fig. 7.1). Unlike for the transmission mode, there

is little latitude to optimize the length traveled by PE in the surface region

in order to optimize the probability of excitation. The only factors which

can be used are the incidence angle and the beam energy E

. The signal to

background ratio (S/B) is therefore not as good as in transmission mode.

Experimental set-up for REELS analyses

A modied TEM (transmission electron microscope) energy lter for

energy losses spectroscopy is used in reection mode by Nikzad et al.

(1992). A small pencil of diffracted or diffused beam is selected and enters

the analyzer. The beam is energy ltered by a stigmatic imaging magnetic

sector followed by a set of magnifying projection lenses. The focal plane

contains a range of energies which are collected simultaneously. Another

analyzer for REELS is a device that combines energy ltering and imaging.

The full diffraction diagram is accepted and, by means of a set of electron

lenses, is energy ltered and projected onto a uorescent screen similar to

the usual RHEED screen (Staib et al., 1999). The angular distribution is

preserved and RHEED diagrams are energy ltered as shown in Fig. 7.8.

The ltering progressively removes the inelastic BSE part and, at the cut

off energy E

, only the elastic diffracted spots remain and energy losses are

ltered out. REELS distributions are achieved by measuring the intensity

of a selected area of the diffraction diagram and lowering the energy lter

voltage from above the elastic peak down to a few 100 eV loss energy. The

signal is measured using a lock-in detection technique in order to improve

the signal to noise ratio.

Examples of in situ REELS distributions

The surface sensitivity of REELS is demonstrated by Nikzad et al. (1992)

for the deposition of Ge onto Si. The growth of the Ge L

ionization edge

is recorded at a PE energy of 30 keV. A 0.15 nm deposited layer is clearly

detected after normalization of the spectra. Weaker structures can be detected



199Spectroscopies combined with RHEED

after differentiation of the signal. A more detailed discussion of REELS in

MBE is given by Wang (1996, p334).

Figure 7.9 shows the REELS distribution of SrTiO

recorded at E

14.5 kV measured using the imaging energy lter described by Staib et al.

(1999). The energy loss distribution includes the elastic peak and multiple

CEL features. The losses are strongly apparent without background subtraction.

The electron binding energy levels, available from Sevier (1972) and updated

by Williams (2006), are used to identify the loss structures. The structures

(6) correspond to Ti 3s (58.7 eV), (4) to Ti 3p1/2 3p3/2 (32.6 eV), (3) to Sr

4p1/2 (21.6 eV) and Sr 4p3/2 (20.1 eV). Loss energy (5) corresponds to O

2s (41.6 eV) but is shifted to about 47 eV as a result of a chemical shift.

Use of RHEED-REELS for monitoring the surface growth and roughness

The ratio between surface and volume plasmon is used to characterize,

in situ during growth, the surface roughness of a deposited Al lm on

sapphire (Strawbridge et al., 2006). The layer thickness was calibrated ex

situ by Rutherford backscattering (RBS) and the surface roughness was

measured by atomic force microscopy (AFM). The change of the energy

(a)

(b)

(c)

(d)

Cut off

7.8 Energy ﬁltered RHEED diagrams of (100) SrTiO

at ﬁlter voltages

(a) 300 eV, (b) 100 eV, (c) 30 eV, and (d) 10 eV from the cut-off voltage.



200 In situ characterization of thin ﬁlm growth

loss distribution from the bare sapphire substrate to deposited Al layer is

very sensitive to the Al atomic concentration. For thin layer deposits, only

the surface plasmon is observed. For larger deposits in the range 3–14 nm,

two behaviors are found; either the surface plasmon remains dominant

correlating to a smooth surface deposit, or the volume plasmon dominate

witnessing a rougher surface as conrmed by AFM scanning images. It is

worth noticing that RHEELS data can be acquired in spite of the fact that

sapphire is an excellent insulator. RHEELS is far less sensitive to charging

effects than other analysis methods because the gain or loss in kinetic energy

of the PE due to the surface charges compensates after backscattering. The

PE enters and leaves the surface within a nanometer scale offset and the

surface potential is homogeneous over this range. The same holds for the

characteristic energy and ionization losses.

Monitoring the ionization loss for quantitative analyses

Quantitative analyses can be performed using the technique developed for

transmission EELS as described by Leapman (1992). The intensity of the EEL

structures is measured after background subtraction using a simple power law

formula. The signal is then integrated over an energy width that includes the

CEL region. A similar calculation can be used for reection EELS, but an

additional correcting factor is needed to account for backscattering effects

–140 –120 –100 –80 –60 –40 –20 0 20

Energy loss (eV)

Signal N(E) arb. units

100

¥5

¥1

RHEED energy 10 kV

7.9 REELS of (100) SrTiO

near the elastic reﬂected peak measured at

a PE energy of 14.5 keV.



201Spectroscopies combined with RHEED

(which are not present in transmission). The extraction of structures near

the elastic peak, where ionization losses are superimposed on multiple CEL

losses, requires a Gaussian background tting.

7.4.5 In situ monitoring by Auger electron spectroscopy

(RHEED-AES)

The escape depth of Auger electrons is given by the IMFP for a specic Auger

line energy and material composition. A few relevant IMFP data are plotted in

Fig. 7.10 as function of the kinetic energy using tabulated data from Powell

(2000, NIST database 71). The plot shows that the IMFP depends on Z, but

is even more dependent on the chemical state of the surface. Oxides like SiO

and ZnO have larger IMFPs than their pure components. The MFP is minimum

for kinetic electron energies in the range 30–50 eV with values around 0.5 nm

and increases to values 2–3 nm at 1000 eV (Ferguson, 1989, p25; Tanuma,

2003). When working with very thin deposited or adsorbed layers, the IMFP

can alternatively be expressed in terms of monolayer coverage.

Experimental set-up for RHEED-AES

For in situ operation, the energy analyzer must fulll the specic requirements

listed previously. These constraints rule out the classical spectrometer designs

NIST database 71

SiO

GaAs

ZnO

200 400 600 800 1000 1200 1400 1600 1800 2000

Electron energy (eV)

ImFP (nm)

5.0

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

0.5

0.0

7.10 ImFP for several elements and compounds in the energy range

of Auger electrons.



202 In situ characterization of thin ﬁlm growth

in favor of specially designed systems consisting of an electron optical lens

system collecting the Auger electrons in front of the sample followed by

an energy lter located further away from the sample, possibly behind the

chamber wall. The angular position of the analyzer with respect to the surface

is important. The analyzer should not be mounted at grazing take-off angle

because the angular distribution of Auger electrons roughly follows Lambert’s

law of a cosine distribution. The maximum peak intensity is normal to the

sample surface; see Fig. 7.1. In contrast, the BSE peak toward the direction

of specular reection. Thus, an Auger analyzer mounted with its axis normal

to the sample will collect most of the Auger electrons, less BSE, and thus

have the best sensitivity. The normal position has the advantage of not being

sensitive to crystalline channeling effects affecting the angular distribution

of Auger electrons, especially marked for low energy lines. The azimuthal

rotation of the sample during acquisition is then possible. An additional

electron gun, positioned far from grazing incidence angle (about 45°) and

able to work over a wide energy range, is useful for calibration purpose in

quantitative analyses and for measuring REELS distributions in a lower

beam energy range.

In situ monitoring of oxygen during ZnO oxide growth on GaN substrates

ZnO is grown on top of a GaN substrate using a radio frequency (RF) plasma

source for oxygen deposition. The growth process can be followed in real

time using a fast Auger probe analyzer mounted at normal incidence angle

using the experimental set-up given by Staib (2011). The Auger electron

lines of Ga LMM (1066 eV), oxygen O KLL (503 eV) and Zn LMM (990 eV)

are monitored. The growth rate is about 10 ML/min and at an acquisition

speed of 6 V/s, the measuring time for an Auger line is 3–10 s. The Ga signal

abruptly vanishes and the Zn signal grows rapidly when the Zn shutter opens;

see Fig. 7.11(a). The oxygen line is present on the surface, provided by the

residual gas pressure in the chamber. The oxygen K LVV line grows further

on opening the oxygen source shutter (see Fig. 7.11b), and reaches a stable

value. The normalized intensity ratio of O KLL to Zn LMM is a measure

of the ratio of atomic concentrations.

In situ monitoring of growth of MgO and of Cr

(1–x)

on MgO substates

Chambers et al. (1996) used RHEED combined with AES and in situ XPS to

study the homoepitaxy growth of MgO on MgO(001) substrates. The sample

temperature is kept at 750 °C during deposition and oxygen is provided by

an electron cyclotron resonance (ECR) plasma source. The Auger lines of

Mg KLL and O KLL are converted into atomic densities by cross-calibration

using in situ XPS intensities of the Mg 2p and O 1s photoelectron lines.



203Spectroscopies combined with RHEED

Lmm Auger lines

800 850 900 950 1000 1050 1100 1150

Energy (eV)

(a)

Signal (mV)

1600

1500

1400

1300

1200

1100

1000

900

800

700

600

O KVV Auger line

450 500 550

Energy (eV)

Normalized O KLL signal

350

300

250

(b)

7.11 (a) Growth of ZnO on GaN substrate. The Ga Lmm Auger line

disappears as the Zn Lmm line grows rapidly to a steady state

value. (b) Growth of the oxygen O KLL Auger line under the same

conditions. Curves labelled ‘a’ correspond to the initial substrate,

curves labelled ‘d’ indicate the ﬁnal state.

