temperature compensation by embedded temperature variation … · 2019. 12. 14. · temperature...
TRANSCRIPT
Temperature Compensationby Embedded Temperature Variation Method
foran AC Voltammeric Analyzer of Electroplating Baths
Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
Aleksander Jaworski, Hanna Wikiel and Kazimierz WikielTechnic, Inc., Cranston, RI, USA
2
Typical Copper Plating Bath Composition
Copper Copper Sulfate 0.25-1 mol/L
Acid Sulfuric Acid 0.1-2 mol/l
Chloride Chloride Ion 20-100 ppm
Suppressor
Accelerator
Leveler N+
N+
N+
ppm(s)
Wide (from sub-ppm to g/L)
Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
100s ppm
3
Superfilling Mechanisms of Submicron Featuresadsorption-based, temperature dependent
Strong adsorbing Leveler inhibits plating (by deactivating accelerator) in the field and at the mouth of the feature.Diffusion-Consumption Model
Suppressor: adsorption instantaneous but weak, diffuses slowly, but moderately concentrated: adequate initial supply.
Accelerator: adsorption of moderate pace but strong, diffuses fast, but low concentration: insufficient initial supply, gradual displacement of suppressor, bottom-up platingCurvature Enhanced Accelerator Coverage Model
P.M. Vereecken et al., IBM J. Res. & Dev. Vol. 49 No 1, January 2005, pp.3-18
Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
4Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
P.M. Vereecken at al., IBM J. Res. & Dev. Vol.49 No 1 January 2005, p.3-18
“Simple System”:Cu2+ , H2S04, Cl-, suppressor, accelerator
5Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
Multitask Electrochemical Probe
6Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
Leveler, NLev
Suppressor,NSupp
Accelerator, NAcc
Temperature o C
0.67 0.50 0.67 19.0
0.67 1.50 0.83 20.0
0.67 1.25 1.00 21.0
0.67 1.00 1.17 22.0
0.67 0.75 1.33 23.0
0.83 1.00 1.33 19.0
0.83 0.75 0.67 20.0
0.83 0.50 0.83 21.0
0.83 1.50 1.00 22.0
0.83 1.25 1.17 23.0
1.00 1.50 1.17 19.0
1.00 1.25 1.33 20.0
1.00 1.00 0.67 21.0
1.00 0.75 0.83 22.0
1.00 0.50 1.00 23.0
1.17 0.75 1.00 19.0
1.17 0.50 1.17 20.0
1.17 1.50 1.33 21.0
1.17 1.25 0.67 22.0
1.17 1.00 0.83 23.0
1.33 1.25 0.83 19.0
1.33 1.00 1.00 20.0
1.33 0.75 1.17 21.0
1.33 0.50 1.33 22.0
1.33 1.50 0.67 23.0
Two Training Sets: at constant temperature and with embedded temperature variation
7
Fundamental Frequency AC Cyclic Voltammogram: Dependence on Leveler Concentration
f=50 Hz, ϕ=0o, A=50 mV, v=50 mV/s, Eini=0.8, Evertex=0 V vs. ECu2+ /Cu
Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
0
0.5
1
1.5
2
2.5
3
3.5
4
0 500 1000 1500
AC
cu
rre
nt
/ m
A
Index point of voltammogram, variable j
0.67 N_lev, CC3
0.83 N_lev, CC8
1.00 N_lev, CC13
1.17 N_lev, CC18
1.33 N_lev, CC23
8
Variable Selection Based on Leveler ImpactRegression analysis of voltammetric data
𝑿 𝐼×𝐽 voltammetric data matrix
𝒄 𝐼×1 leveler concentration vector
𝒕 𝐼×1 temperature
Ƹ𝑐𝑖,𝑗 = 𝛽0,𝑗 + 𝛽1,𝑗𝑥𝑖,𝑗 LSR equation
Ƹ𝑐𝑖,𝑗 = 𝛽0,𝑗 + 𝛽1,𝑗𝑥𝑖,𝑗 + 𝛽2,𝑗𝑡𝑖 trivariate regression eq.
𝑅𝑗2 =
σ𝑖=1𝐼 𝑐𝑖ො𝑐𝑖,𝑗− Τσ𝑖=1
𝐼 𝑐𝑖 σ𝑖=1𝐼 ො𝑐𝑖,𝑗 𝐼
2
σ𝑖=1𝐼 𝑐𝑖
2− ൗσ𝑖=1𝐼 𝑐𝑖
2𝐼 σ𝑖=1
𝐼 ො𝑐𝑖,𝑗2 − ൗσ𝑖=1
𝐼 ො𝑐𝑖,𝑗2
𝐼squared correlation coefficient
Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
9Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 500 1000 1500
R2
Index point of voltammogram, variable j
Training set CC, 21C
Training set CV, no T compensation
Training set CV, with T compensation
Variable Selection Based on Leveler ImpactLeveler calibrations: R2 calculated individually for points of voltammograms of training sets CC and CV
10Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
0.5
1
1.5
2
2.5
542 562 582 602 622 642 662
AC
cu
rre
nt
/ m
A
Index point of voltammogram, variable j
0.67 N_lev, CC3
0.83 N_lev, CC8
1.00 N_lev, CC13
1.17 N_lev, CC18
1.33 N_lev, CC23
Variable Selection Based on Leveler ImpactA selected portion of AC voltammogram for a range corresponding to applied DC potential of 260 to 134 mV vs. E Cu2+ /Cu , respectively recorded at 21°C for different concentrations of leveler additive.
11Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
Leveler, NLev
Suppressor,NSupp
Accelerator, NAcc
Temperature o C
0.67 0.50 0.67 19.0
0.67 1.50 0.83 20.0
0.67 1.25 1.00 21.0
0.67 1.00 1.17 22.0
0.67 0.75 1.33 23.0
0.83 1.00 1.33 19.0
0.83 0.75 0.67 20.0
0.83 0.50 0.83 21.0
0.83 1.50 1.00 22.0
0.83 1.25 1.17 23.0
1.00 1.50 1.17 19.0
1.00 1.25 1.33 20.0
1.00 1.00 0.67 21.0
1.00 0.75 0.83 22.0
1.00 0.50 1.00 23.0
1.17 0.75 1.00 19.0
1.17 0.50 1.17 20.0
1.17 1.50 1.33 21.0
1.17 1.25 0.67 22.0
1.17 1.00 0.83 23.0
1.33 1.25 0.83 19.0
1.33 1.00 1.00 20.0
1.33 0.75 1.17 21.0
1.33 0.50 1.33 22.0
1.33 1.50 0.67 23.0
Variable Selection Based on Temperature Impact
Five subsets of the training set with parametrized leveler concentration
12
Variable Selection Based on Temperature Impact Relationship between temperature and univariate voltammetric data
within 1/5 subsets of the training set
Ƹ𝑡𝑖 = 𝛼0,𝑗 + 𝛼1,𝑗𝑥𝑖,𝑗 regression equation
𝑅𝑡,𝑗2 =
σ𝑖=1𝐼/5
𝑡𝑖 Ƹ𝑡𝑖,𝑗 − σ𝑖=1𝐼/5
𝑡𝑖 ൗσ𝑖=1𝐼/5 Ƹ𝑡𝑖,𝑗 Τ𝐼 5
σ𝑖=1𝐼/5
𝑡𝑖2 − ൗσ
𝑖=1𝐼/5
𝑡𝑖2
Τ𝐼 5 σ𝑖=1𝐼/5 Ƹ𝑡𝑖,𝑗
2 − ൗσ𝑖=1𝐼/5 Ƹ𝑡𝑖,𝑗
2Τ𝐼 5
squared correlation coefficient
Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
13
Variable Selection Based on Temperature ImpactSquared correlation coefficients between selfpredicted and actual temperature values calculated individually for
each point of voltammogram, subsets of matrix CV with parametrized leveler concentration
Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 500 1000 1500
R2
Index point of voltammogram, variable j
0.67 N_lev
0.83 N_Lev
1.00 N_Lev
1.17 N_Lev
1.33 N_Lev
14
Variable Selection Based on Temperature ImpactAC voltammogram for leveler, selected range 542-668, dependence on temperature at parametrized
leveler concentration of 1.33 N_Lev
Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
0.5
0.7
0.9
1.1
1.3
1.5
1.7
542 562 582 602 622 642 662
AC
cu
rre
nt
/ m
A
Index point of voltammogram, variable j
19.0 C
20.0 C
21.0 C
22.0 C
23.0 C
15
Calibration Calculation by Principal Component Regression (PCR)
𝑿 = 𝑺𝑽𝑇 + 𝑬 PCA decomposition into scores S and loadings V𝜷 = 𝑺𝑇𝑺 −1𝑺𝑇𝒄 Inverse Least Squares Regression on scoresƸ𝑐𝑢 = 𝒙𝑢𝑽𝜷 Regression equation
𝑺𝑡 = 𝑺 𝒕 PCA scores augmented with temperature𝜷𝑡 = 𝑺𝑡
𝑇𝑺𝑡−1𝑺𝑡
𝑇𝒄 Inverse Least Squares Regression on scores augmented with temperature
Ƹ𝑐𝑢 = 𝒙𝑢𝑽 𝑡𝑢 𝜷𝑡 Regression equation with embedded temperature variance
Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
16Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
0.60
0.80
1.00
1.20
1.40
1 2 3 4 5 6 7 8
22.5o C
0.60
0.80
1.00
1.20
1.40
1 2 3 4 5 6 7 8
21.5o C
0.60
0.80
1.00
1.20
1.40
1 2 3 4 5 6 7 8
20.5o C
0.60
0.80
1.00
1.20
1.40
1 2 3 4 5 6 7 8
19.5o C
n Actual n Model at 21o C n Embedded temp. var.
No
rmal
ized
leve
ler
con
c. N
_lev
Prediction of Leveler Concentration in Validation Set Samples
17
Conclusions
Jaworski et al., 17th ESEAC, Rodos, Greece, June 3-7, 2018
General, rigorous routine for the development of the analytical method using a chemometric model with temperature variation embedded in regression is introduced for exemplary determination of leveler additive concentration by AC voltammetry.
Chemometrics is critical in mitigating the adverse effect of temperature variation on accuracy of concentration prediction by an on-line AC voltammetric analyzer.
Accurate calibration can be calculated for experimental conditions where hard-models do not exist.
Chemometrics promotes an interest in AC-based electroanalytical techniques for industrial applications.