0100020003000400050006000700080009000

10000

100 200 300 400 500 600 700 800 900010020030040050060070080090010001100120013001400

Goal: To characterize chemistries and residual stresses in PZT thin films for MEMS by Raman SpectroscopyWhat is Raman spectroscopy? Raman spectroscopy is the study of the inelastic scattering of photons by molecules. The energy of the scattered radiation that is more (Stokes) or less (anti-Stokes) than the incident radiation (laser beam) is related to the vibrational energy.

Several tests were run where the only thing changed was the polarization of the laser beam from horizontal to vertical, on thick (30 layers) and thin (6 layers) samples, as well as 40:60 and 52:48 samples. While some were more dramatic than others, all the tests showed that the polarization did not effect the size or shape of the peaks, but that setting the polarization to horizontal gave a stronger signal.

Polarization matters for thin films, but not so much for thicker films. This is possibly due to the fact that the thin films are on the scale of 600nm thick (6 layers), which is a lot closer to the wavelength of the light (5145Å) we are using than the 3000 nm (30 layers) thick wafers.

Acknowledgements: Financial support was provided by the National Science Foundation's Division of Materials Research REU program under grant number 0139125.

References:

Gerald Burns and B. A. Scott, Physical Review Letters, v. 25, p. 167, (1970)

http://carbon.cudenver.edu/public/chemistry/classes/chem4538/raman.htm

When defining the peaks on our graphs, we choose to use the same method as Burns, in applying ten Gaussian curves to fit our datacurve. Burns only used nine within our data range. However, wefound it advantageous to use ten in order to take care of some background on the tail of our last peak (only the last peak is different than the basic peaks of Burns).

Chemistry and Spectroscopy of PZT:

Pb(ZrxTi1-x)O3

52:48 PZT takes on the rhombohedral structure of lead zirconate.

40:60 PZT takes on the tetragonal structure of lead titanate.

Because of the two distinct structure differences, the vibrational energies are also distinctly different. Therefore the intensities of different Raman spectra peaks are unique from each other.

52:48

40:60

Different Compositions of PZT thin films.

Wave numbers (cm-1)

Arb

itrar

y in

tens

ity o

f the

52:

48 c

ompo

sitio

n

Arb

itrar

y in

tens

ity o

f the

40:

60 c

ompo

sitio

n

An additional small peak on the shoulder of the 40:60 line appears.

The 40:60 peak is much bigger

proportionally and has a sharper

peak that is shifted slightly to

the right.

0100020003000400050006000700080009000

10000

100 200 300 400 500 600 700 800 900

Gaussian curve fit of a 12 layer 52:48 sample

Wave numbers (cm-1)

Arb

itrar

y in

tens

ity

115.58

273.74

202.82 287.83

323.00 554.30

597.70 705.49744.36

796.29

GaussiansBackground

Sum of Gaussians

Real data line

1000

1500

2000

2500

3000

3500

4000

4500

5000

100 200 300 400 500 600 700 800 900

Raman Spectra of a PZT thin film

Wave numbers (cm-1)

Arb

itrar

y in

tens

ity

Stresses in the sample change the positions of the peaks because the bonds are stretched. Compositions, temperature of the sample, annealing temperature, and crystal structure can all impact the peak position.

As far as what kinds of samples can be used, we must consider that we must be able to collect spectra that is reflected off the sample we are testing. Therefore, it is easiest to test solid samples, of which the desired compound is the top-most material on the sample. Powders incased in glass tubes or other transparent casings are often tested. It is possible, but much more difficult to test samples in the liquid state and nearly impossible to test samples in the gas phase.

What can Raman tell you about a material?When the laser light (incident radiation) hits the sample, each bond between elements emits a specific energy in response, creating a“molecular fingerprint.” The molecular fingerprint IDs the sample.

anti-Stokes

RayleighStokes

Therefore the difference between the energy of the scattered radiation and the incident radiation is a direct measurement of the vibrational energy. In order to identify the vibrational energy of molecules in samples, Raman spectroscopy measures the differencebetween the incident radiation and the scattered radiation, by exciting the bonds in the molecules with light waves of certain energy and collecting the data.

Laser interactions with sample:

0

5000

10000

15000

20000

25000

30000

35000

40000

100 200 300 400 500 600 700 800 900

Area with no well

Edge of well

Center of well

Wave numbers (cm-1)

Arb

itrar

y in

tens

ity

Three different positions on the same 40:60 wafer

Conclusions:

Testing with the laser set on horizontal polarization generally works best.

The relative intensity of a peak and the thickness of the sample are not directly proportional.

The fourth peak (about 610 cm-1)is the largest proportionally on the 40:60 graphs and has a sharper peak that is shifted to the right of that of the 52:48 composition.

Samples of the same composition (or even the same sample) will have dramatically different spectra near wells versus on a non-well area. Further study is needed to determine why.

0.6

0.7

0.8

0.9

1

1.1

0 5 10 15 20 25 30

Number of layers of PZT

I~210

I~610

Intensity versus thickness1000

1500

2000

2500

3000

3500

4000

4500

5000

100 200 300 400 500 600 700 800 900

Wave numbers (cm-1)

Arb

itrar

y in

tens

ity

I~210 I~610

The two peaks being comparedOne possibility was that the thickness of the sample was effecting the relative intensities of the peaks.

The relative intensities do not appear to correlate to thickness

Argon Ion Laser ~ 5145 Å line

Prisms

PolarizerSample Compartment

Monochrometer

0

5000

10000

15000

20000

25000

30000

35000

40000

100 200 300 400 500 600 700 800 900

Thin (6 layer) 40:60 PZT wafer

Wave numbers (cm-1)

Arb

itrar

y in

tens

ity Horizontal polarization

Vertical polarization

Examining the above graph and replotting it with a second y-axis, it’s clear that the spectra is the same, but the intensity is dramatically different.

0

5000

10000

15000

20000

25000

30000

35000

40000

100 200 300 400 500 600 700 800 9000

500

1000

1500

2000

2500

3000

3500

Arb

itrar

y in

tens

ity

Arb

itrar

y in

tens

ity

Wave numbers (cm-1)

Thin (6 layer) 40:60 PZT wafer

05000

1000015000200002500030000350004000045000

100 200 300 400 500 600 700 800 900

Thick (30 layer) 40:60 PZT wafer

Wave numbers (cm-1)

Arb

itrar

y in

tens

ity Horizontal polarization

Vertical polarization

The back of a PZT wafer

Crossection of a wafer

PZTPt

Si

well

In order to determine if the stress was the same at all points on agiven wafer, we tested several points on one single wafer. We testedin the very center, the center of a well and the edge of a well. As you

can see from the figure below, the portion ofwafer that we tested where there was no wellshows up as a normal looking 40:60 graph, butas we move to the edge of the well, it looksmore like a 52:48 graph. The center of thewell looks identical to a 52:48 graph. After

making this discovery, we compared these results to the stresses measured using x-ray diffraction on the 40:60 and the 52:48 wafersand found that the 52:48 wafers are stressed at 400MPa whereas the 40:60 wafers are only stressed at 190MPa. The Raman suggests that the wafers are stressed nearly twice as much over a well as they are in a non-well area.

The shoulder is flattening out to resemble a 52:48 wafer, rather than the 40:60 wafer that it is.

These two peaks are becoming less and less defined. The smoother the peaks get, the more like a 52:48 wafer it seems and the less like a 40:60 wafer.

The peak on the right is disappearing and the point of the

peak is shifting dramatically to the

left.