Advances in semiconductor materials and device metrology CIICT 2007 - - PowerPoint PPT Presentation

1

Advances in semiconductor materials and device metrology

CIICT 2007 DCU, 28 Aug 2007.

Patrick J. McNally Nanomaterials & Processing Laboratory School of EE, DCU

2

Co-Workers

• Dr. Lisa O’Reilly, Lu Xu, Ken Horan, Jennifer

Stopford, Dr. Donnacha Lowney (DCU).

Nick Bennett, Prof. Brian Sealy, University of Surrey. Dr Jim Greer, Dr. Nicolás Cordero, Yan Lai, Tyndall

Natioinal Institute, Cork.

• Prof. Nick Cowern, University of Newcastle.
• Dr. Gabriela Dilliway, University of Southampton.

Acknowledgements: Science Foundation Ireland,

Enterprise Ireland.

3

Metrology Challenges

Length scales

Lateral (X-Y) Vertical (Z).

Requirements:

Non-destructive. In situ. In line. “Nano” sensitivities: small volumes, areas (nm-

scale); impurity sensitivites (e.g. 1012 - 1013 cm-3); nano-void and nano-pore detection; etc.

4

Low-k metrology challenges

New low-k dielectrics have different

mechanical/physical properties compared to SiO2.

Pores in the material. Fragile – delamination; stress-induced

fracture.

Back end of the line (BEOL): problems

with assembly and packaging.

No convenient and competent metrology

tools.

Source: ITRS 2005

5

Cu metallisation metrology challenges

Measuring barrier layer(s) under seed copper. Detection of voids in copper lines after CMP and

anneal processes.

Thick Cu lines mask this voiding. Detection through multi-layer structures e.g.

individual layer thicknesses.

Delamination of Cu from e.g. low-k layers before and

after CMP.

Local stress vs. wafer stress. Adhesion strength measurements are still done using

destructive methods.

Detection of killer pores and voids is not yet possible. Source: ITRS 2005

6

A Selection of DCU’s metrology approaches

Gas Cell Photoacoustic Microscopy Micro-Raman Spectroscopy

7

Photoacoustic Microscopy (PAM)

Source: N. George, Cochin Univ. Sci. & Tech., India.

(incl. sample)

DCU proprietary PA cells

8

Automated PAM system for Si wafer analysis

9

Analysis of multi-layer structures on Si wafers

Si Si

SiO2

Si

Cu

Si

SiO2 Cu Cu, SiO2 layer thicknesses = 500nm

10

IC Chip Cracking & Delamination

• 236.1
• 226.4
• 216.8
• 207.1
• 197.4
• 187.8
• 178.1
• 168.5
• 158.8
• 150.8

PA Phase [Deg.]

Optical Micrograph Photoacoustic Phase Image Delamination

11

Wafer bonding defects (Phase Contrast)

• 10,000 pixel image shown is
• btained in 8 minutes.
• The bonding defects acted as extra

thermal barrier, shown as extra time delay in phase images.

ICPAM Phase image ( f =216 Hz).

2 4 6 8 10 2 4 6 8

Position [mm]

Position [mm]

• 87.00
• 86.10
• 85.20
• 84.30
• 83.40
• 82.50
• 81.60
• 80.70
• 79.80
• 78.90
• 78.00

Phase [Deg.] Wafer Bonding defects confirmed by OM Wafer edge effects Inhomogeneous bonding interference effects ??

Doppler interference effect at bonded interface Wafer bonding defects confirmed by IR and Optical Microscopy

• L. Xu, P McNally, DRIP XII, 9 - 13 September 2007

Berlin (Germany).

12

After Image Processing

DELAMINATION

13

Ongoing PAM Developments

Upgrade to 200 mm & 300 mm wafer

capability (end-2007 – Enterprise Ireland Proof of Concept Fund)

Porous dielectric measurements – early

results promising.

Wafer edge sub-surface cracks. Measure nm-scale delamination. Technology licensing underway.

14

Summary for Photoacoustic Microscopy

Can see through opaque (metallic)

layers.

Multi-layer characterisation :

thicknesses, delamination, porosity.

Nanometric vertical scale sensitivities. Whole wafer scanning.

15

Micro-Raman Spectroscopy (µRS)

Incident light excites

vibrational modes in the

• sample. Subsequently

scatter the light.

Some light is scattered at a

different energy (wavelength).

Energy exchange between

incident photons and semiconductor phonons (internal vibrational modes).

Raman light intensity is very

weak.

Typically about one photon

• ut of 10 12.

Ephonon = Eincident - Escattered

• Probe regions ~ 1µm diam.
• 325nm laser Si

penetration depth ~ 9nm

• True nanometric scale

depth metrology.

16

• Strained/deformed crystal.
• Vibrations of crystal lattice altered.
• “Spring constant(s)” between atoms changed.
• Shifts frequency of inelastically scattered Raman photons.
• A plot this shifted light ouput intensity vs. frequency is a

Raman spectrum.

Source: www.kosi.com

17

JY-Horiba LabRam 800 µRS System

Notch filter Laser source CCD detector Confocal hole Mirror Mirror Microscope XY stage holding sample Lens and mirror Sample

488nm or 325nm

18

“The industry cannot live without strain-engineering for enhanced mobility”

Strained Silicon CMOS Technology

(Source: Valencia, IBM) (Source: Thompson et al., Intel)

Semiconductor International, 2006

How does strain affect heavily doped device regions?

source drain

• Strained Si channels improve electron and hole mobilities.
• Greater current drive for less power.

19

Research Motivation

ITRS identifies 3 key requirements… 1) Increasingly shallow junction depth (xj) 2) Increasingly steep junction profile 3) Maintain low resistance (Rs)

Gate Stack Oxide Channel

source drain

Single nMOS transistor

I mportantly… Sb has the edge

• ver As

Source/drain extensions are one example

20

Sample Information

17 nm strained Si layer

grown on a graded Si0.83Ge0.17 virtual substrate

43 nm strained Si layer

grown on Si0.80Ge0.20 virtual substrate

Antimony and Arsenic

Ion Implantation

2keV, 4e14cm-2 Sb 2keV, 4e14cm-2 As

Annealed @ 600, 700,

8000C in N2 ambient

Comparison to bulk

unstrained Si

600 650 700 750 800 500 600 700 800 900 1000 1100 1200 1300 1400 1500

Sheet resistance (Ω/sq.) Anneal temperature (

• C)

Sb in Bulk Si As in Strained Si Sb in Strained Si

Large Rs reduction for strain vs bulk for Sb doping Lower Rs for Sb doping compared to As in strained Si. Sb more highly activated than As in the presence of strain

• N. S. Bennett et al., Appl. Phys. Lett. 89,

182122 (2006)

21

325 nm HeCd UV laser. dp ≈ 9 nm.

UV Laser penetration in SiGe structure Strained Si Si0.8Ge0.2 Si0.9Ge0.1 ~ 40nm ~ 200nm ~ 1000nm Si Substrate UV

Raman Spectra: 325nm laser

200 300 400 500 600 1000 2000

Raman Intensity (a.u.) W avenum ber (cm

• 1)

R eference Si Strained Si, 20% G e

500 510 520 530 1000 2000

Raman Intensity (a.u.) W avenum ber (cm

• 1)

Reference Si Strained Si, 20% G e

Si-Si Si-Si

Tensile

RED SHIFT

22

Raman Spectra: UV laser

Sb, 2keV, 4e14cm-2 As, 2keV, 4e14cm-2

Red-shift of Si peak indicates the presence of tensile strain in the

Si cap layer

Spectra of Sb and As implanted samples show similar behaviour

with clear intensity variation with heat treatment.

500 510 520 530 600 1200

Wavenumber (cm

• 1)

500 510 520 600 1200

Intensity (a.u.) Wavenumber (cm

• 1)

Ref Si SSi 0.7% SSi-Sb SSi-Sb600 SSi-Sb700 SSi-Sb800

Si Ref Imp. 600 700 800 εSi

23

UV Raman Spectra Analysis

Si Raman Peak Intensity Si-Si peak intensity variation is consistent with lattice disorder introduced by ion implantation which recovers with heat treatment. Lattice recovery may not be complete following RTA at 8000C for 10 sec.

Imp. 600 700 800 Strain. Si Ref. Si

300 600 900 1200 1500 1800 2100 2400

SSi-Sb SSi-As Average Si Intensity (a.u.)

24

UV Raman Si Peak Shift

Peak shift Relative to

Unstrained Reference Si Sb implant

• 8.0
• 7.5
• 7.0
• 6.5
• 6.0
• 5.5
• 5.0

Raman shift (cm

• 1)

17% Ge 20% Ge

Sb Implant Sb, 600 Sb, 700 Sb, 800 Strained Silicon

Biaxial stress σxx = σyy = -∆ωSiUV/4 GPa Strained Si, 17% Ge:

∆ωSiUV = -5.64 ± 0.3 cm-1 ⇒ σ = 1.4 ± 0.1 GPa ε=0.77 ±0.06%

Strained Si, 20% Ge:

∆ωSiUV = -6.27 ± 0.2 cm-1 ⇒ σ = 1.57 ± 0.1 GPa ε=0.86 ±0.06%

Ref: I. De Wolf, Semicond. Sci. Technol. 11, 139(1996)

But peak red-shift clearly increases

following ion implantation and RTA

Strain relaxation would be detected by

a blue shift (higher wavenumber) of the Si Raman peak.

25

What is causing the net red shift of the Si peak position observed in the UV Raman spectra?

0.0 6.0x10

13

1.2x10

14

1.8x10

14

2.4x10

14

0.0 6.0x10

13

1.2x10

14

1.8x10

14

2.4x10

14

• 1.8
• 1.6
• 1.4
• 1.2
• 1.0
• 0.8
• 0.6
• 0.4
• 0.2

Linear fit, R=0.80 Sheet Carrier Concentration (cm

• 2)

Linear fit, R=0.85 Normalised Raman shift (cm-1) Sheet Carrier Concentration (cm

• 2)

Sb Doping As Doping

Clear linear dependence between the normalised Raman shift

and the sheet carrier concentration!

26

Why does doping cause a Raman shift?

Cerdeira and Cardona* observed carrier-concentration

related frequency shifts in the Raman spectra of both p- type and n-type Si.

N-doping of Si alters the lattice deformation potential,

effectively “softening” the lattice

⇒ lower phonon vibrational frequencies ⇒ Raman red-shifts

Usually a very small effect in n-type Si

Only significant when doping concentration is large. Which it is here!!! 1020 – 1021 cm-3 effective doping!!** Independent of dopant type.

• * F. Cerdeira & M. Cardona, Phys. Rev. B. 5, 1440 (1972).
• ** L. O'Reilly et al., INSIGHT-2007, Napa, California, U.S.A., May 6-9, 2007.

27

Conclusions from Micro-Raman Spectroscopy

Caution needed when using UV Raman for Si

strain metrology in highly doped ultra shallow junction structures.

For all implanted samples (strained and bulk

substrates) there is a net red shift in the position

• f the Si-Si phonon mode.

Confinement, stress and carrier concentration

effects contribute to this shift.

The observed anomalous Raman shift originates

from the high levels of doping achieved in the samples.

28

Summary

DCU’s combined suite of technologies

provides versatile methodologies for advanced IC metrology.

Virgin wafer through to completed

circuit.

Nanometre to mm probe depths. Virtually any materials combination!!!

29

Confocal µRS microscope

Allows rejection of radiation originating away from the focal point conjugate to the confocal aperture. This radiation from the blue and red planes does not pass through the aperture, because they are not focussed in the confocal plane. Raman radiation originating away from the sample depth of interest never reaches the entrance to the spectrograph Acquired spectrum is specific to the depth of the sample in focus.