Presented at the COMSOL Conference 2010 India

Use of COMSOL Multi-Physics in Modeling Galvanic Corrosion

Kiran B. Deshpande
Senior Researcher, Material Characterization and Modeling Group India Science Lab, General Motors Global Research and Development GM Technical Centre India Pvt Ltd, Creator Building, International Tech Park Ltd. Whitefield Road, Bangalore - 560 066, INDIA Email: [Link]@[Link]

COMSOL CONFERENCE 2010, BANGALORE October 29 - 30, 2010

Motivation
Why Magnesium in Automobiles?

Light weight increases fuel economy High strength to weight ratio makes it a good engineering material Good castability enables high production rate

Challenges
Poor corrosion resistance

Mg is not very suitable for automotive applications which are exposed to corrosion media
High galvanic corrosion

Mg is very anodic to other metals such as Fe, Al poor durability in dissimilar material
joints and applications

Research Focus

Understanding macro-galvanic and micro-galvanic corrosion

Comsol Conference, Bangalore, October 29 - 30, 2010

India Science Lab

What is galvanic corrosion?

Electrolyte Mild Steel 2H+ H2 Electric field Mg++ Zn Al Mg Anodic

Al

2e-

Mg

3 conditions for galvanic corrosion

Potential difference between dissimilar materials Electrical contact between dissimilar material for electron transport Exposure to conductive medium for ionic transport

Cathodic

India Science Lab

Comsol Conference, Bangalore, October 29 - 30, 2010

Modeling of Corrosion Phenomena

Micro-galvanic corrosion Macro-galvanic corrosion

AM50 Die Cast Microstructure a Phase

Mild Steel (cathode)

AE44 (Mg alloy)

After 3 days immersion in 1.6 wt% NaCl solution b Phase

Objective: To optimize the processing route to obtain the desirable microstructure for the better corrosion resistance

Objective: To develop a numerical tool to predict galvanic corrosion rate, which can provide design specifications

India Science Lab

Comsol Conference, Bangalore, October 29 - 30, 2010

Microstructure corrosion correlation

Alloy AM50 AM50 AM50 Casting Process Die Cast Permanent Mold Sand Cast Ecorr (V) -1.48 -1.5 -1.5 b phase content % 18.4 1.08 0.61 icorr (A/m2) 0.081 0.182 0.269

Ecorr - Corrosion Potential

icorr Corrosion current density

AM50 DC

AM50 PM

AM50 SC

Microstructure effect on corrosion b phase fraction b phase Distribution Al content in the a phase India Science Lab
Comsol Conference, Bangalore, October 29 - 30, 2010

Reference: Sundarraj et al., Magnesium Technology 2008, TMS 2008.

Modeling approach
Model Schematic n 0
Electrolyte

Assumptions
Well mixed, incompressible and electroneutral electrolyte solution Uniform Al content throughout alpha phase No interfacial resistance due to corrosion product is considered here The dissolution reaction is considered at the anode surface

n 0

2 0
b phase (Cathode) a phase (Anode)

n 0

where is the electric potential

j k n f ( )

Corrosion rate calculation

Corrosion rate in [Link]-1 [Link]-1 Equivalent weight in [Link]-1 Density in kg.m-3 anodic current density in Amp.m-2

The model is capable of explicitly tracking the corroding phase. The model predictions are based on non-linear polarization data.

India Science Lab

Comsol Conference, Bangalore, October 29 - 30, 2010

Modeling Approach: moving boundary formulation

Arbitrary Lagrangian Eulerian (ALE) Method COMSOL Multi-Physics
Initial mesh

Electrolyte

Cathode

Anode

Deformed mesh

India Science Lab

x
Comsol Conference, Bangalore, October 29 - 30, 2010

Effect of b phase fraction (area ratio)

Model Schematic
10% b phase
Electrolyte a phase (Anode) 90 b phase (Cathode) a phase (Anode) 20 m 90 m

50% b phase

90% b phase

The maximum of anodic current density (corrosion rate)

India Science Lab
Comsol Conference, Bangalore, October 29 - 30, 2010

as b phase fraction

Scanning Vibrating Electrode Technique

Equipment
Electrometer

Schematic
Vin RE

Potentiostat WE CE RE

Scanning probe Test sample Electrolyte

Sample specimen

SVET output Current density calculation

j
b a b

E A

Current density, Amp/m2 Conductivity, S/m Potential difference at the two extremes of amplitude of probe vibration, V Amplitude of probe vibration, m

E
SVET experiments are performed on a galvanic couple where individual components are in direct physical and electrical contact, which eliminates IR drop.

India Science Lab

Comsol Conference, Bangalore, October 29 - 30, 2010

Model - SVET Comparison

10% b phase

50% b phase

90% b phase

Model and SVET experiments : Corrosion rate as b phase fraction The corrosion rate AM50 dc was the lowest despite the highest b phase fraction This behavior motivates us to investigate the effect of b phase distribution
India Science Lab
Comsol Conference, Bangalore, October 29 - 30, 2010

Level set formulation

Microstructure of AM50 die cast along the depth

A single boundary is used to specify the anode surface and the cathode surface using the level set function The above formulation can capture the varying ratios of the a phase and the b phase along the depth of the alloy
Acknowledgement: Shashank Tiwari for microstructure

Reference: K. B. Deshpande, Electrochimica Acta (2010), doi:10.1016/[Link].2010.09.044

India Science Lab

Comsol Conference, Bangalore, October 29 - 30, 2010

Continuous b phase network

Schematic of microstructure Model prediction

Electrolyte

b phase

Dissolved a phase

a phase preferentially dissolves in electrolyte solution until continuous network of b phase is exposed to electrolyte solution after which corrosion is halted.
India Science Lab
Comsol Conference, Bangalore, October 29 - 30, 2010

a phase dissolution

b phase
India Science Lab

Dissolved a phase

Comsol Conference, Bangalore, October 29 - 30, 2010

Continuous b phase network

Surface b phase fraction evolution

Average anodic current density evolution

Average anodic current density as surface b phase fraction The corrosion activity is however halted after the a phase present on the surface is dissolved into the electrolyte solution. The total charge passed per square meter area = 1.3844105 C/m2 The total mass of material removed per square meter area = 17.22 g/m2

India Science Lab

Comsol Conference, Bangalore, October 29 - 30, 2010

Discrete b phase
Model prediction Schematic of microstructure Electrolyte

b phase

Dissolved a phase

a phase continuously dissolves in the electrolyte solution as discontinuous b phase network assists corrosion. As discrete b phase is not well supported, b phase eventually spatters off.
References: Ambat, et al., Corrosion Science, 42, 1433-1455 (2000); Song et al., Corrosion Science, 41, 249-273 (1999).

India Science Lab

Comsol Conference, Bangalore, October 29 - 30, 2010

Discrete b phase
Surface b phase fraction evolution
a b a b

Average anodic current density evolution

Neck formation and neck thinning leading to b phase spattering is captured using the model. The corrosion activity lasts for a lot longer for discrete b phase than in continuous b phase network. The model captures the scenario leading to b phase spattering, but not the actual spattering as the model cannot capture topological changes.

The total charge passed per square meter area = 2.5060 10 C/m2
5

The total mass of material removed per square meter area = 31.17 g/m2

India Science Lab

Comsol Conference, Bangalore, October 29 - 30, 2010

Modeling of Corrosion Phenomena

Micro-galvanic corrosion Macro-galvanic corrosion

AM50 Die Cast Microstructure a Phase

Mild Steel (cathode)

AE44 (Mg alloy)

After 3 days immersion in 1.6 wt% NaCl solution b Phase

Objective: To optimize the processing route to obtain the desirable microstructure for the better corrosion resistance

Objective: To develop a numerical tool to predict galvanic corrosion rate, which can provide design specifications

India Science Lab

Comsol Conference, Bangalore, October 29 - 30, 2010

Modeling approach
Model Schematic n 0
Electrolyte

Assumptions
Well mixed, incompressible and electro-neutral electrolyte solution No interfacial resistance due to corrosion product is considered here Anode surface is assumed to be corroding and cathode surface is assumed to be non-corroding

n 0

2 0
Mild Steel (Cathode) Magnesium (Anode)

n 0

where is the electric potential

j k n f ( )

Corrosion rate calculation

Corrosion rate in [Link]-1 [Link]-1

Equivalent weight in [Link]-1

The model is capable of explicitly tracking the corroding phase. The model predictions are based on non-linear polarization data.

Density in kg.m-3 anodic current density in Amp.m-2

India Science Lab

Comsol Conference, Bangalore, October 29 - 30, 2010

Animation capturing moving Mg surface

Mild Steel

AE44 (Mg alloy)

2 mm

India Science Lab

Photograph after cross-sectional cut of AE44 MS couple after 3 days of immersion

Comsol Conference, Bangalore, October 29 - 30, 2010

SVET and model prediction

AE44 Mild steel couple AE44 AA6063 couple

The corrosion rate predicted using the numerical model is within +9% of that estimated using SVET analysis for AE44 mild steel couple. The corrosion rate predicted using the numerical model is within -29% of that estimated using SVET analysis for AE44 AA6063 couple.
References: K. B. Deshpande, Corrosion Science, 52 (2010) 2819 2826 K. B. Deshpande, Corrosion Science, 52 (2010) 3514 - 3522 Comsol Conference, Bangalore, October 29 - 30, 2010

India Science Lab

Immersion and model prediction

AE44 Mild steel couple AE44 AA6063 couple

The corrosion rate predicted using the numerical model is within -20% of that estimated from immersion technique for AE44 mild steel couple. The corrosion rate predicted using the numerical model is within -47% of that estimated from immersion technique for AE44 AA6063 couple.
Acknowledgement: Sampath Vanimisetti (Image digitization)

India Science Lab

Comsol Conference, Bangalore, October 29 - 30, 2010

Summary
Corrosion rate (mm/y) Galvanic Couples SVET experiments 197 42

Immersion experiments
243 52

Mixed Potential Theory 231 26

ALE method 210 29

AE44 - MS AE44 AA6063

The corrosion rate predicted using the numerical model is in good agreement with that estimated from the two experimental techniques.

The galvanic corrosion behavior is demonstrated at both macro and micro scale using the numerical model developed using COMSOL Multi-Physics.

India Science Lab

Comsol Conference, Bangalore, October 29 - 30, 2010