In-situ AFM Tensile Testing ME 395 Professor Horacio D. Espinosa Chris Biedrzycki, Matthieu Chardon, Travis Harper, Richard Neal

Micron Scale Mechanical Properties Once structures enter microscale, laws of macro mechanics no longer dominate mechanical response Samples approach size scale of grains and dislocations, also dominated by surface effects Multimaterial interfaces, sharp corners in MEMS devices require in-situ characterization Devices in market require a high amount of reliability, necessitating direct fatigue testing methods

Advantages of In-Situ Testing Model entire range of behavior Differentiate elastic and plastic regimes Understand more than ultimate tensile strength (UTS)

Similar Works 1 3 1 =2 μm No testing results 2 1) Mechanical Measurements at the Micron and Nanometer Scales, Knauss, Huang 2) Microtensile Tests with the Aid of Probe Microscopy for the Study of MEMS Materials, Knauss, Chasiotis 3) MEMS Fatigue Testing to Study Nanoscale Response, Fisher, Labiossiere

Testing stage Polysilicon specimen Integrated during fabrication Pulled by mass Actuators move mass when voltage applied Force measured by capacitive sensor Strain measured by DIC From Zhu et. al., 2003

ANSYS Modeling of System The maximum load deliverable by the loading device is ~100 mN From Zhu et. al., 2003

Polysilicon Sample Properties 4μm 10μm Young's Fracture Method Modulus Strength (Gpa) (GPa) MUMPs21 136-174 1.3-2.8 tensile MUMPs19 132 - Stress Concentrations 1.3-1.5 Size Effect 154.1-159.6 1.51-1.67 Size Effect-doped 2.0-2.8 Microcantilever 174±20 2.8±0.5 bending AFM deflection 173±10 2.6±0.4 Bulk 181-203 indentation Doped-Undoped 95-175 4μm 10μm From Prorok, et al. Vol 5, 2004.

Integrated in Chip Yong’s Chip Chris’s Quarter

Testing Setup Difficult to land the cantilever on the test specimen without repeatedly breaking tips

Atomic Force Microscopy (AFM) Benefits Very fine resolution (~0.02nm) Variable scanning size/rate Risks Noise sensitivity, drift Constraints on sample size/mass From Espinosa, “Introduction to AFM and DPN”, ME395 Winter 2004

AFM Resolution Resolution Dimension 3100: 512x512 pix JEOL: 256x256 pix Digital Image Correlation (DIC) can interpret change in position of a pixel by change in grayscale pixel strength (from 1=black to 255=white). Ideally, DIC can understand a resolution of 1/255th of a pixel

AFM Resolution Scan size (say, 1.0 μm) corresponding to 512 pixels DIC resolution: 1/255th of a pixel Theoretical (maximum) displacement resolution (a) If length of sample is 10m, theoretical resolution of strain is (b) (a) (b)

Polysilicon pad imaged using D3100 AFM AFM Imaging Types Amplitude Topography Polysilicon pad imaged using D3100 AFM

Polysilicon pad imaged using JEOL AFM AFM Imaging Types Amplitude Topography Polysilicon pad imaged using JEOL AFM

AFM Problems Adjustment of voltage, frequency for tip Sensitivity to dust/scratches Breakage of tip Crash of tip into sample Sensitivity of feedback control system

Digital Image Correlation No automatic compensation for drift Correlation on stationary samples Drift strain is only strain Used to generate correction factor

Digital Image Correlation (DIC) Example Illustrates DIC Capabilities Sample Prior to Deformation Error X-X Tensile Strain Y-Y Tensile Strain Shear Strain

DIC Problems Out-of-plane Deformation Displacement Gradients Scanning Noise

DIC Resolution Displacement resolution of DIC is 1/255pix, much smaller than the 0.512 pix minimum displacement in pixels for the polysilicon at failure. AFM/DIC should be able to display 0.512/(1/255), or 130 images for the range of displacement to failure based on ideal resolution/largest scan area. *Note: actual resolution smaller than ideal, fewer images result. Smaller scan areas will provide more images, and thus finer understanding of the change in strain over time.

Digital Image Correlation (DIC) Sample DIC images from 10x10μm sample using D3100 AFM

Digital Image Correlation (DIC) Sample DIC images from 10x10μm sample using JEOL 5200 AFM

The EPIC AFM, scanned in amplitude mode, yields the best results AFM Comparison The EPIC AFM, scanned in amplitude mode, yields the best results

Experimental Design Optimal Designs Computer generated Best with respect to a particular criterion D-Optimal minimizes l(X’X)-1l Places points at regions of greatest standard error

Experimental Design Robust Design All relevant information obtained using less than half the required experiments in a full factorial study H0: μ1 = μ2 H1: μ1 ≠ μ2 α = P(type I error) = P(reject H0 l H0 true) β = P(type II error) = P(fail to reject H0 l H0 is false) Power = 1 - β

Experimental Design Standard Order #19 Removed (Above 3.5) Outliers distort the analysis of variance Outlier when residuals +/- 3σ

Significant Factors: A, B, AB Experimental Results Significant Factors: A, B, AB R-Squared: .776

Experimental Results

In Conclusion… Optimal testing conditions for drift reduction: JEOL 5200 AFM (Forced Feedback) Amplitude Scan Tapping Mode High Scan Area Sample choice and scan rate effect drift rate very little

In Conclusion… Optimized drift condition remains too large to distinguish polysilicon strains from error 1% Strain at failure for polysilicon – minimized drift remains 100% of experimental strain With current minimized strain values, brittle samples cannot be assessed

Future Work Drift Effects Reduction Opportunities Appraiser Variation in DIC Analysis Environmental Effects Electromechanical Noise Temperature, Pressure, Humidity Variation Feedback Mechanism - Consider Image Repeatability Sample Ductility

Thank You For Your Time Questions? In-Situ AFM Tensile Testing Group Chris Biedrzycki, Matthieu Chardon, Travis Harper, Richard Neal