Elimination of TiN peeling during exposure to CVD tungsten deposition process…

Allan Quadros

Madhav Dhital

Department of Statistics

Introduction

Problem



  • Increasing silicon wafers from 150mm to 200 mm caused the peeling of the TiN layer


  • Maintain film uniformity (150mm) vs prevent TiN peeling

Solution



  • DoE + RSM

  • 3 experiments + 1 Verification run


    • … using a multivariate approach!

Experiment 1

Experiment 1 - Design

  • Factors: Temperature, Pressure, and Backside Argon

  • Design: \(2^3\) factorial with 3 center points (but only one was performed due to lack of wafers - 9 runs total)

  • 2 models: one with Uniformity as response variable and another one with Stress as response

  • both models obtained using stepwise regression beginning from the model with 3 main effects and 3 two-factor interactions

Factor Low High
Temperature 440˚C 500˚C
Pressure 0.8 torr 4.0 torr
Backside Argon 0.8 sccm 300 sccm

Experiment 1 - Model 1



Model 1 - Uniformity Response

Term Coeff. Std.Error t.value Signif.
Average 6.698889 0.229599 NA NA
Pressure -0.435000 0.243526 -1.79 0.1243
Back Argon 1.657500 0.243526 6.81 0.0005 
No. cases = 9, R2 = 0.8919, R2 adj. = 0.8559, RMS Error = 0.688, Resid. df = 6

Experiment 1 - Model 2



Model 2 - Stress Response

Term Coeff. Std.Error t.value Signif.
Average 8.938333 0.088301 NA NA
Temp -1.205000 0.093658 -12.87 0.0001
Pressure -0.358500 0.093658 -3.83 0.0087 
No. cases = 9, R2 = 0.9678, R2 adj. = 0.6570, RMS Error = 0.2649, Resid. df = 6

Experiment 1 - Results



  • Uniformity is improved by increasing pressure and decreasing backside argon


  • Stress is improved (lowered) by increasing both temperature and pressure

Experiment 1 - Contour plot

  • For the contour plot, the authors fixed the temperature at 480˚C since it is relatively difficult to change the reactor temperature. This is the same temperature used by other processes.

  • No TiN peeling occured in any treatment combination

  • The contour plot shows that the high pressure/low backside argon corner is good in terms of both stress and uniformity

  • But still have not met the goals of reaching the results of the old method for 150 mm wafers

  • The bottom-right corner suggests that is a region for improving the process using an increased range of pressure

Experiment 1 - Contour plot (2)

Experiment 2

Experiment 2 - Design

  • Conducted on a broader range of pressure

  • Factors: Pressure, and \(H_2/WF_6\) ratio

  • The authors had evidence that the \(H_2/WF_6\) ratio should produce curvature (local maximum) in the deposition rate, and a consequent local minimum in Uniformity. Therefore, …

  • Design: CCI - Central Composite Inscribed design

  • Due to limited supply of resources, the authors chose to include only 3 centerpoints resulting in 11 runs
Factor Low High
Pressure 4 80
\(H_2/WF_6\) Ratio 2 10

The CCI design

CCI

  • Use only points within the factor ranges originally specified

  • Do not provide the same high quality prediction over the entire space compared to the CCC

  • Uses the factor settings as the star points

  • Useful in situations which the limits specified for a factor are truly operational limits

CCI

Experiment 2 - Models



  • 2 models: one with Uniformity as response variable and another one with Stress as response + visual inspection of TiN peeling


  • both models obtained using stepwise regression beginning from quadratic models for physical reasons

Experiment 2 - Model 1



Model 1 - Uniformity Response

Term Coeff. Std.Error t.value Signif.
Average 5.927273 0.185831 NA NA
Pressure -1.912335 0.308142 NA NA
\(H_2/WF_6\) -0.224966 0.308142 NA NA
Pressure*\(H_2/WF_6\) 1.699487 0.616144 2.76 0.0282 
No. cases = 11, R2 = 0.8695, R2 adj. = 0.8136, RMS Error = 0.66163, Resid. df = 7

Experiment 2 - Model 2



Model 2 - Stress Response

Term Coeff. Std.Error t.value Signif.
Average 7.733551 0.037042 NA NA
Pressure -0.737459 0.044084 NA NA
\(H_2/WF_6\) 0.497877 0.044084 11.29 1e-04
Pressure\(^2\) -0.494690 0.070923 -6.97 2e-04 
No. cases = 11, R2 = 0.9849, R2 adj. = 0.9784, RMS Error = 0.08817, Resid. df = 7

Experiment 2 - Results



  • Uniformity is improved at high pressure and low \(H_2/WF_6\)


  • Stress is controlled by pressure and the \(H_2/WF_6\) ratio, with significant curvature in pressure. The lowest stress occurs at low pressure and low \(H_2/WF_6\) ratio.

Experiment 2 - Contour plot

  • The TiN peeling (Yes/No) was also observed for this experiment. Treatment combinations where peeling occurred are circled in the contour plot (*)

  • The results indicate that the \(H_2/WF_6\) Ratio clear controls TiN peeling. The authors considered the results a pivotal finding, since they expected temperature and pressure to be the controlling factors

  • Although low levels of \(H_2/WF_6\) result in better stress and uniformity, TiN peeling occurs at low \(H_2/WF_6\) levels, rendering this region of process space unusable.

  • Due to tight deadline circumstances and the previous results, the team decided to compromise stress in favor of uniformity.

Experiment 2 - Contour plot (2)

Experiment 3

Experiment 3 - Design

  • Experiment 2 results indicated Uniformity (response variable) could be traded off against TiN peeling by setting the pressure at 80 torr and considering several \(H_2/WF_6\) ratios against the peeling threshold

  • Single factor experiment in the \(H_2?WF_6\) ratio to find the peeling threshold, holding pressure constant at 80 torr.

Pressure H2WF6_Ratio TiN_Peeling
80 3 YES
80 4 YES
80 5 NO

Experiment 3 - Results

  • The TiN peeling threshold appears to be at a \(H_2/WF_6\) ratio somewhere between 4 and 5.

Verification Run

New parameters

Parameter Value
Backside Ar 0 sccm
Temperature 480˚C
Pressure 80 torr
\(H_2/WF_6\) ratio 5

Conclusions and Recommendations



Tungsten film uniformity on 200mm wafer can match uniformity on 150mm wafers by:

  • increasing pressure from 800 mtorr to 80 torr; and

  • decreasing the \(H_2/WF_6\) ratio from 23 to 5.

However, this results in a worse film Stress.