Bill's comments in these text boxes Project 1 : Introduction to MSE 510 Projects Data Analysis Using Example Diffusion Data for Aluminum Ions in Aluminum Oxide Student Name MSE 510 Due 01/24/2012 Ÿ Project Objectives 1. Familiarize yourself with projects in this course 2. Familiarize yourself with Mathematica 3. Encourage you to review chapter 1 in the textbook Ÿ Summary Diffusion data is selected from a peer-reviewed journal article and imported, plotted, and analyzed using Mathematica. The source of the data is the diffusion studies of Paladino and Kingery A1E and the following Mathematica functions are utilized: ® The raw data is imported from a .csv file using "Import" and plotted using "ListPlot" ® A non-linear mathematical fit is performed using "NonlinearModelFit" ® "Manipulate" is used to show the effects of changing two relevant parameters on the fit of the data ® Relevant parameters for diffusion are extracted from the data fit and compare to the published data ® Statistical error of the non-linear fit is analyzed ® Integration and differentiation are perfomed on the non-linear fit equation Relevant discussion and comments are included where appropriate. Add an Abstract no more than 150 words long that explain your findings. Import, Plot, and Fit Data Import the data to a list designated as "data" : 2 MSE 510 - Project 1 - Excellent Example.nb In[1]:= Out[1]= data = Import@"D:\\Bill\\Bsu\\Courses\\MSE 510 ElecOptDielProps\\Projects Exams Quizzes @MSE 510D\\Projects\\2013 Projects\\Project 1 @Ch 1D\\Student Projects\\AaronWoodard\\Aaron Woodard diffdata.csv"D 992178., 1.1 ´ 10-10 =, 92103., 4. ´ 10-11 =, 92093., 3.5 ´ 10-11 =, 92073., 2.8 ´ 10-11 =, 92043., 1.9 ´ 10-11 =, 92003., 1.6 ´ 10-11 =, 91993., 1.1 ´ 10-11 =, 91943, 1.9 ´ 10-12 == Paladino and Kingery measured the amount of aluminum that diffused across an interface joining two cylinders of aluminum oxide. A plot of the experimental results is generated showing the diffusion coefficient as a function of anneal temperature. The diffusion coefficent describes the rate of atomic motion in the material. In[2]:= dataplot = ListPlotAdata, Frame ® True, GridLines ® Automatic, PlotStyle ® 8Red, AbsolutePointSize@5D<, PlotRange ® 881900, 2200<, 80, 1.4 * 10 ^ - 10<<, FrameLabel ® 9"T HKL", "D Hcm2 •sL"=, PlotLabel ® "Diffusion Rate vs. T For Al in Al2 O3 "E Diffusion Rate vs. T For Al in Al2 O3 1.4 ´ 10-10 1.2 ´ 10-10 D Hcm2•sL 1. ´ 10-10 8. ´ 10-11 Out[2]= 6. ´ 10-11 4. ´ 10-11 2. ´ 10-11 0 1900 1950 2000 2050 T HKL 2100 2150 Label your plot. That is, give it a figure number and caption. e.g., "Figure 1: Diffusivity of Al in Al2O3 as a function of temperature. Data taken from reference [1]." You can then get rid of the title at the top. Do this for all your figures. You can then refer to the figure in the text 2200 by its title (i.e., figure 1). From the plot, it is apparent that the diffusion rate increases with increasing temperature. It is well known that the equation for the diffusion coefficient has the form of the Arrhenius rate equation, and is given as D = Do expI In[3]:= -E A M kT where Do is the pre-exponential consant, E A is the activation energy for diffusion, k is the Boltzmann constant, and T is temperature @2D . A model of this form is now fitted to the data. PrintA"D = ", fitexp = NonlinearModelFitAdata, a * ãb•x , 8a, b<, xEE H*Calculates the fit model and outputs the resulting equation*L D = FittedModelB 47.0619 ã-58 346.2•x F MSE 510 - Project 1 - Excellent Example.nb In[4]:= 3 plotdf = PlotAfitexp@xD, 8x, 1900, 2200<, Frame ® True, GridLines ® Automatic, PlotStyle ® Blue, PlotRange ® 881900, 2200<, 80, 1.4 * 10 ^ - 10<<, FrameLabel ® 9"T HKL", "D Hcm2 •sL"=E; Show@dataplot, plotdf, PlotLabel ® "Diffusion Rate vs. T For Al in Al2 O3 Blue Line = Fit, Red Dots = Data"D Diffusion Rate vs. T For Al in Al2 O3 Blue Line = Fit, Red Dots = Data 1.4 ´ 10-10 1.2 ´ 10-10 Label your plot. That is, give it a figure number and caption. Do this for all your figures. You can then refer to the figure in the text by its title (i.e., figure #). Out[5]= D Hcm2•sL 1. ´ 10-10 8. ´ 10-11 6. ´ 10-11 4. ´ 10-11 2. ´ 10-11 0 1900 1950 2000 2050 2100 2150 2200 T HKL The exponential model appears to fit the data well, as expected based on the known diffusion coefficient equation. In general, the fit appears to be better at higher temperatures and diffusion rates. There is less agreement of the fit with the observed data at lower temperatures. The accuracy of the fit will be analyzed during the error analysis performed in a subsequent section. The constants in the diffusion coefficent equation (excluding the Boltzmann constant) are dependent on the material properties and conditions under which diffusion takes place. For example, crystal structure, grain structure, atomic bonding, and defect characteristics affect the value of the pre-exponential constant and activation energy for a particular situation. Using the built-in "Manipulate" function, the effects of changes in these values on the diffusion rate can be investigated. 4 MSE 510 - Project 1 - Excellent Example.nb In[6]:= Clear@d, g, h, TD H*initialize variables; g = pre-exponential, h = activation energy*L k = 8.617 * 10 ^ - 5;H*Boltzmann constant, eV•K*L -h d@g__, h_, T_, kD := g * ã k*T H*define the function for the diffusion coefficient equation*L ManipulateAShowAPlotAd@g, h, T, kD, 8T, 1900, 2200<, Frame ® True, GridLines ® Automatic, PlotStyle ® Blue, PlotRange ® 881900, 2200<, 80, 1.4 * 10 ^ - 10<<, FrameLabel ® 9"T HKL", "D Hcm2 •sL"=, PlotLabel ® "Diffusion Rate vs. T For Al in Al2 O3 "E, dataplotE, 99g, 47, "Pre-Exponential Hcm2 •sL"=, 0, 2000, 1=, 88h, 5.03, "Activation Energy HeVL"<, 4.0, 8.0, .01<E Pre-Exponential Hcm2 •sL Activation Energy HeVL 47 Very nice that you show values before creating the pdf file 5.03 Diffusion Rate vs. T For Al in Al2 O3 1.4 ´ 10-10 Out[9]= 1.2 ´ 10-10 D Hcm2•sL 1. ´ 10-10 8. ´ 10-11 Label your plot. That is, give it a figure number and caption. Do this for all your figures. You can then refer to the figure in the text by its title (i.e., figure #). 6. ´ 10-11 4. ´ 10-11 2. ´ 10-11 0 1900 1950 2000 2050 2100 2150 2200 T HKL Relevant diffusion parameters can be extracted from the fit of the data. The resulting equation for the non-linear fit gives the estimated values of the pre - exponential and activation energy for the material and conditions studied by Paladino and Kingery. These values and their corresponding units are listed in the table below. In[10]:= GridA98"Parameter", "Symbol", "Value", "Unit"<, 8"Activation Energy", "EA ", "5.03", "eV"<, 9"Pre-Exponential", "DO ", "47.06", "cm2 •s"==, Alignment ® Left, Frame ® AllEH*generates a custom table of values*L Out[10]= Nice. Do provide a label and title for your table. That is, give it a table number and Pre-Exponential DO 47.06 cm2 •s Title. That is, "Table 1: Extracted Diffusions Parameters" and place it at the top or By comparison, Paladino and Kingery found values for the pre-exponential activation of 28 4.94, beginning of theandtable. Do energy this for all and your respectively. By examination, using the manipulation chart presented above, these values appear to also produce tables. You can then refer to the table in a good fit for the data. In fact, by further manipulation of the chart, there is an infinite number of combinations that will the text by its title (i.e., Table #). Parameter Symbol Value Unit Activation Energy EA 5.03 eV produce a line that fits the data. In order to determine the correct parameters, the data must be plotted as Ln D vs. 1/T. This is the general convention by which diffusion data is plotted. MSE 510 - Project 1 - Excellent Example.nb 5 By comparison, Paladino and Kingery found values for the pre-exponential and activation energy of 28 and 4.94, respectively. By examination, using the manipulation chart presented above, these values appear to also produce a good fit for the data. In fact, by further manipulation of the chart, there is an infinite number of combinations that will produce a line that fits the data. In order to determine the correct parameters, the data must be plotted as Ln D vs. 1/T. This is the general convention by which diffusion data is plotted. Import the new data set for Ln D and 1/T : In[11]:= Out[11]= data2 = Import@"D:\\Bill\\Bsu\\Courses\\MSE 510 ElecOptDielProps\\Projects Exams Quizzes @MSE 510D\\Projects\\2013 Projects\\Project 1 @Ch 1D\\Student Projects\\AaronWoodard\\Aaron Woodard diffdata2.csv"D 880.459137, - 22.9<, 80.475511, - 23.9<, 80.477783, - 24.1<, 80.482393, - 24.3<, 80.489476, - 24.7<, 80.499251, - 24.9<, 80.501756, - 25.2<, 80.514668, - 27.<< By taking the natural logarithm of both sides of the diffusion coefficient equation, the equation becomes Ln D = Ln Do - E A /kT which is in the form of a straight line, y = mx + b @3D . Therefore, the pre-exponential and activation energy can be obtained by a linear fit of the Ln D vs. 1/T data plot. The slope of the fit line is -E A /k and the intercept is Ln Do . Plot the new data and perform a linear fit: In[12]:= dataplot2 = ListPlotAdata2, Frame ® True, GridLines ® Automatic, PlotStyle ® 8Red, AbsolutePointSize@5D<, PlotRange ® Automatic, FrameLabel ® 9"1000•T HK-1 L", "Ln D Hcm2 •sL"=, PlotLabel ® "Diffusion Rate vs. T For Al in Al2 O3 "E; Print@"D = ", fitline = LinearModelFit@data2, x, xDD plotfitline = PlotAfitline@xD, 8x, .45, .52<, Frame ® True, GridLines ® Automatic, PlotStyle ® Blue, PlotRange ® Automatic, FrameLabel ® 9"1000•T HK-1 L", "Ln D Hcm2 •sL"=E; Show@dataplot2, plotfitline, PlotLabel ® "Fit = Blue Line; Data = Red Points"D D = FittedModelB 7.19797 - 65.2783 x F Fit = Blue Line; Data = Red Points -23 -24 Ln D Hcm2•sL Out[15]= Label your plot. That is, give it a figure number and caption. Do this for all your figures. You can then refer to the figure in the text by its title (i.e., figure #). -25 -26 -27 0.46 0.47 0.48 0.49 1000•T HK -1L 0.50 0.51 6 MSE 510 - Project 1 - Excellent Example.nb Fit = Blue Line; Data = Red Points Slope = -23 EA k Ln D Hcm2•sL -24 In[16]:= Increasing diffusion coefficient -25 -26 Increasing Temperature -27 0.45 0.46 0.47 0.48 0.49 0.50 0.51 0.52 1000•T HK -1L Fit = Blue Line; Data = Red Points Slope = -23 Nice EA k Label your plot. That is, give it a figure number and caption. Do this for all your figures. You can then refer to the figure in the text by its title (i.e., figure #). Ln D Hcm2•sL -24 Out[16]= -25 Increasing diffusion coefficient -26 Increasing Temperature -27 0.45 0.46 0.47 0.48 0.49 0.50 0.51 0.52 1000•T HK -1L Good observation on the variation of the data at lower temperatures. What statistical test can be performed to determine the validity of removing data outliers? Again it is seen that the linear fit agrees well with the data at higher temperatures, but not as well at relatively lower temperatures. The data point corresponding to the lowest temperature could be an outlier; it appears that by omitting this point the accuracy of the fit could be improved. This point also corresponds to the lowest rate of diffusion measured by Paladino and Kingery. It is possible that the measurement device used was more accurate for higher diffusion rates. The resulting equation for the linear fit gives the estimated values of the pre-exponential and activation energy as shown in the table below. In[17]:= GridA98"Parameter", "Symbol", "Value", "Unit"<, 8"Activation Energy", "EA ", 65.2783 * 1000 * k, Do "eV"<, provide a label and title for 9"Pre-Exponential", "DO ", ã7.19797 , "cm2 •s"==, Alignment ® Left, Frame ® AllE Out[17]= Parameter Symbol Value Unit Activation Energy EA 5.62503 eV Pre-Exponential DO 1336.71 cm2 •s your table. That is, give it a table number and Title. Place it at the top or beginning of the table. Do this for all your tables. You can then refer to the table in the text by its title (i.e., Table #). The values reported here are not in agreement with those reported by Paladino and Kingery. The difference could be due to the method of data extraction from the paper--values plotted here were extracted visually from the data chart in the paper and the error could be significant. Raw data values for the paper were not available. Your arguments to the reasons that your parameter values are not in agreement with Paladino and Kingery are In summary,method the data has importedthe and validity plotted in of its raw form. A non-linearis equation was fit the to theliterature data, but thefor fit the fairly sound. A common tobeen examine your parameters to search was not adequate to obtain estimates of relevant diffusion coefficient equation parameters. The data was then plotted in work of others that have extracted the values. One way to do so is to see who has referenced the work of an alternate format, consistent with scientific convention for diffusion data. The resulting linear fit provided estimates Paladino and Kingery. In addition, calculating the percent error using the values you extracted and the values for the pre-exponential and activation energy parameters. An error analysis of the linear fit is performed in the next given by Paladino and Kingery provide a value of accuracy of the data. section. MSE 510 - Project 1 - Excellent Example.nb The values reported here are not in agreement with those reported by Paladino and Kingery. The difference could be7 due to the method of data extraction from the paper--values plotted here were extracted visually from the data chart in the paper and the error could be significant. Raw data values for the paper were not available. In summary, the data has been imported and plotted in its raw form. A non-linear equation was fit to the data, but the fit was not adequate to obtain estimates of relevant diffusion coefficient equation parameters. The data was then plotted in an alternate format, consistent with scientific convention for diffusion data. The resulting linear fit provided estimates for the pre-exponential and activation energy parameters. An error analysis of the linear fit is performed in the next section. Error Analysis The accuracy of the linear model is now analyzed by examining the statistical parameters associated with the fitting equation. Recall the linear model used to fit the Ln D vs. 1/T data : In[18]:= Print@"D = ", fitline = LinearModelFit@data2, x, xDD D = FittedModelB 7.19797 - 65.2783 x F Ÿ Fit Parameters In[19]:= Print@"Model Parameter Table = ", fitline@"ParameterTable"DD Estimate Model Parameter Table = 1 x Standard Error t-Statistic P-Value 7.19797 3.7537 -65.2783 7.69559 1.91757 0.103615 -8.48256 0.000146805 The Estimates represent the calculated values for the intercept HLn Do L and the acivation energy H - E A • kL. The Standard Error represents how precisely the estimated the parameter was computed and estimates the Dovalue you ofmean "Estimated value"? standard deviation of the sample mean based on the population mean @4D . The t Statistic is the ratio of the Estimate to the Standard Error. It is used to calculate the P-Value @5D . The P-Value represents the probability of getting a t Statistic greater than the calculated value by chance. A very low value indicates that the parameter estimate is significantly different from zero; a rule of thumb is that a P-Value below 0.05 is significant @5D . Good. Based on the model parameters, the linear fit gives a good estimate of the activation energy for diffusion, since the PValue is very small. The estimate for Ln Do may not be accurate, as the P-Value indicates that there is at least a 10% chance of obtaining the estimate by chance. You stated above that Standard Error is based on precision and the P-value Ÿ Coefficients of Determination In[20]:= is based on the Standard Error. In this sentence, you mention accuracy. between "accurate" and "precise"? PrintA"R2 = ", fitline@"RSquared"DE What is the difference PrintA"Adjusted R2 = ", fitline@"AdjustedRSquared"DE R2 = 0.923031 Adjusted R2 = 0.910203 The Coefficient of Determination represents how well the model fits the data. The value varies between 0 and 1, with 1 being a perfect fit, meaning that the model can accurately describe every data point. The Adjusted Coefficient of Determination also represents how well the model fits the data, but also takes into account the degrees of freedom. Here, the high values indicate that the linear model is a relatively good fit for the data. This agrees with the theoretical Arrhenius behavior of the diffusion coefficient with temperature, which indicates that Ln D vs. 1/T should be a linear relationship. As noted earlier, the data point corresponding to the lowest anneal temperature could be the reason that R^2 is not higher @6D . 8 MSE 510 - Project 1 - Excellent Example.nb The Coefficient of Determination represents how well the model fits the data. The value varies between 0 and 1, with 1 being a perfect fit, meaning that the model can accurately describe every data point. The Adjusted Coefficient of Determination also represents how well the model fits the data, but also takes into account the degrees of freedom. Here, the high values indicate that the linear model is a relatively good fit for the data. This agrees with the theoretical Arrhenius behavior of the diffusion coefficient with temperature, which indicates that Ln D vs. 1/T should be a linear relationship. As noted earlier, the data point corresponding to the lowest anneal temperature could be the reason that R^2 is not higher @6D . Ÿ Residuals What are the degrees of freedom? Please define. The residuals represent the difference between the value predicted by the model and the value of the corresponding obeserved data point. In[22]:= Print@"Residuals = ", fitline@"FitResiduals"DD ListPlot@fitline@"FitResiduals"D, PlotRange ® Automatic, Filling ® Axis, Frame ® True, PlotLabel ® "Ln D vs. 1•T Linear Fit Residuals"D Residuals = 8- 0.126285, - 0.0574176, - 0.109105, - 0.0081723, 0.0541939, 0.492289, 0.355811, - 0.601315< Ln D vs. 1•T Linear Fit Residuals 0.4 Label your plot. That is, give it a figure number and caption. Do this for all your figures. You can then refer to the figure in the text by its title (i.e., figure #). 0.2 Out[23]= 0.0 -0.2 -0.4 -0.6 0 2 4 Good. Unknown factors could be other competing kinetic mechanisms including other diffusion mechanism or reactions 6 8 From the residual plot, the values least well described by the model are the three data points that correspond to measurements taken at the lowest anneal temperatures. This supports the theory that the measurement device used may not be sensitive enough to detect these lower amounts of diffused particles. It is also possible that, as temperature increases, it becomes the dominant mechanism controlling the diffusion rate. At lower temperatures, there could be unknown factors controlling the rate of diffusion. Ÿ Confidence Intervals A confidence interval represents a range where the true value of the estimated parameter lies according to a given probability. The confidence intervals for the parameters estimated by the linear fit are presented below. MSE 510 - Project 1 - Excellent Example.nb In[24]:= 9 Print@"80% Confidence Interval = ", fitline@"ParameterConfidenceIntervals", ConfidenceLevel ® .8DD Print@"85% Confidence Interval = ", fitline@"ParameterConfidenceIntervals", ConfidenceLevel ® .85DD Print@"90% Confidence Interval = ", fitline@"ParameterConfidenceIntervals", ConfidenceLevel ® .90DD Print@"95% Confidence Interval = ", fitline@"ParameterConfidenceIntervals", ConfidenceLevel ® .95DD Print@"99% Confidence Interval = ", fitline@"ParameterConfidenceIntervals", ConfidenceLevel ® .99DD ListPlot@Table@fitline@"ParameterConfidenceIntervals", ConfidenceLevel ® pD, 8p, 8.8, .85, .9, .95, .99<<D, Frame ® True, GridLines ® Automatic, PlotLabel ® "Confidence Intervals: 80%, 85%, 90%, 95%, 99%"D 80% Confidence Interval = 881.79356, 12.6024<, 8- 76.3581, - 54.1985<< 85% Confidence Interval = 881.00372, 13.3922<, 8- 77.9774, - 52.5793<< 90% Confidence Interval = 88- 0.0961403, 14.4921<, 8- 80.2322, - 50.3244<< 95% Confidence Interval = 88- 1.987, 16.3829<, 8- 84.1087, - 46.4479<< 99% Confidence Interval = 88- 6.71859, 21.1145<, 8- 93.8091, - 36.7475<< Confidence Intervals: 80%, 85%, 90%, 95%, 99% Label your plot. That is, give it a figure number and caption. Do this for all your figures. You can then refer to the figure in the text by its title (i.e., figure #). 20 10 0 -10 Out[29]= -20 -30 -40 -50 -80 -60 -40 -20 0 What are the title & units of the axes? Can you define them? The confidence intervals at all levels are relatively large. The range of values in which the true values may lie would have a significant effect on the calculation of Do and E A . Using the 95% confidence interval as an example, the high and low true values of Do and E A are shown in the following table: 10 MSE 510 - Project 1 - Excellent Example.nb In[30]:= GridA98"Parameter", "High Value", "Low Value", "Unit"<, 8"EA ", 84.1087 * 1000 * k, 46.4479 * 1000 * k, "eV"<, 9"DO ", ã16.3829 , ã-1.987 , "cm2 •s"==, Alignment ® Left, Frame ® AllE Out[30]= Parameter High Value EA 7.24765 DO Low Value Unit 4.00242 eV 1.30318 ´ 107 0.137106 cm2 •s Do provide a label and title for your table. That is, give it a table number and Title. Place it at the top or beginning of the table. Do this for all your tables. You can then refer to the table in the text by its title (i.e., Table #). The high range of values indicates that the estimates of the parameters are poor. To obtain more accurate estimates, more data points could be measured or a more accurate measurement device could be used. It is also possible that there is an unknown factor controlling the rate of diffusion that influences the behavior of the diffusion coefficent in response to a temperature change. In that case, the diffusion coefficient might not completely obey the Arrhenius relationship. Good. Integration and Differentiation Since the data plotted as Ln D vs. 1/T fit a straight line, differentiation of the fit line will be a constant. However, as noted above, the resulting constant is a useful parameter for diffusion--the activation energy. The value is actually -E A • H1000 *kL, so E A is found by multiplying by Boltzmann's constant and by 1000 (since T was plotted as 1000/T). The activation energy found for the analyzed diffusion data was presented earlier, however its calculation by differentiation is demonstrated here. Since the second derivative would be zero, it is not considered. Calculate the activation energy by differentiation of the linear fit : In[31]:= Print@"The activation energy for diffusion is = ", D@fitline@xD, xD * - 1000 * k, " eV."D The activation energy for diffusion is = 5.62503 eV. The derivative can also be plotted. In this case, it is simply a straight line at 5.625 eV, as expected. MSE 510 - Project 1 - Excellent Example.nb 11 In[32]:= Needs@"PlotLegends`"D dfitline@x_D = fitline '@xD; Plot@dfitline@xD * - 1000 * k, 8x, 1900, 2200<, PlotRange ® 881900, 2200<, 80, 10<<, Frame ® True, GridLines ® Automatic, PlotStyle ® Blue, FrameLabel ® 8"T HKL", "dD•dT HeVL"<, PlotLegend ® 8"Activation Energy"<, PlotLabel ® "Derivative of Diffusion Rate vs. T For Al in Al2 O3 "D General::obspkg : PlotLegends` is now obsolete. The legacy version being loaded may conflict with current Mathematica functionality. See the Compatibility Guide for updating information. Derivative of Diffusion Rate vs. T For Al in Al2 O3 10 Label your plot. That is, give it a figure number and caption. Do this for all your figures. You can then refer to the figure in the text by its title (i.e., figure #). dD•dT HeVL 8 6 Out[34]= 4 2 0 1900 1950 Activation Energy 2000 2050 T HKL 2100 2150 2200 Good. I like the legend. There is no value in performing integration on the diffusion data, since it will not yield useful information. However, the built-in Mathematica integration function is used for demonstration purposes. In[35]:= Out[35]= intfitline@x_D = Integrate@dfitline@xD, xD H*define a function that equals the integral of the derivative of the fitline*L - 65.2783 x 12 MSE 510 - Project 1 - Excellent Example.nb In[36]:= PlotAintfitline@xD, 8x, .45, .52<, Frame ® True, GridLines ® Automatic, PlotStyle ® Blue, PlotRange ® Automatic, PlotLegend ® 8"Original Linear Fit Line HIntercept = 0L"<, PlotLabel ® "Integration Example", FrameLabel ® 9"1000•T HK-1 L", "Ln D Hcm2 •sL"=E Integration Example Label your plot. That is, give it a figure number and caption. Do this for all your figures. You can then refer to the figure in the text by its title (i.e., figure #). -30 Ln D Hcm2•sL -31 -32 Out[36]= -33 -34 0.45 0.46 Original Linear Fit Line HIntercept = 0L 0.47 0.48 0.49 0.50 0.51 0.52 1000•T HK L -1 Since the integral of the derivative of the fitline was calculated, the fitline is also the result. As expected, the integration does not give the intercept; however the slope is the same as that of the linear fit line. This indicates that the integration function was used properly. Applications This type of analysis of diffusion data is not limited to the diffusion of aluminum in aluminum oxide. It can be applied to the diffusion of any type of atom in any material and will show whether a particular diffusion process can be characterized by an Arrhenius relationship to temperature. Specifically for aluminum diffusion in aluminum oxide, one example application is in characterization of the oxidation rate of aluminum alloys. By performing this type of analysis, the ion diffusion rate can be determined over different operating temperatures, which assists in the determination of the rate of oxide growth. Aluminum oxide is typically a protective oxide layer, so the data can be used to determine both the temperature and length of time at which the alloy should be annealed in order to achieve the desired oxide thickness, or the rate at which the alloy will oxidize in service. In order to further characterize the process, the diffusion of oxygen in aluminum oxide could be investigated. Based on a comparison of the respective diffusion rates of oxygen and aluminum in the oxide, the ion which controls the overall oxide formation rate can be determined. If this is known, the rate of formation of oxide can be further controlled by inhibiting or enhancing the rate of diffusion of the rate limiting ion. Even further experiments to determine the mechanism by which diffusion occurs can allow for extensive control over the oxide formation rate. For example, diffusion through a single crystal of aluminum can be compared to that through polycrystalline aluminum to determine the effects of the presence of grain boundaries, which can be conduits by which diffusion occurs. In summary, the analysis presented here can be repeated across different material types and structures to characterize the rate of formation of any type of alloy, ceramic, polymer or phase for a given situation. MSE 510 - Project 1 - Excellent Example.nb References 13 Nice reference list. [1] Paladino, A.E.; Kingery, W.D. Aluminum ion Diffusion in Aluminum Oxide. J.Chem.Phys. Vol 37, 5. 1962. [2] Kasap, S.O. Principles of Electronic Materials and Devices. 3rd Ed. McGraw Hill, 2006. [3] Atkins, P.; de Paula, J. Physical Chemistry. 7th Ed. Oxford University Press, 2002. [4] Press,W.H.;Flannery,B.P.;Teukolsky,S.A.;and Vetterling,W.T. Numerical Recipes in FORTRAN:The Art of Scientific Computing. 2nd Ed. Cambridge University Press, 1992. [5] Bivariate Scatterplot and Fitting. JMP7: Statististical Discovery. Computer Software. SAS Institute, 2007. [6] Interpreting Regression Results, OriginLabs - Origin Software - http://www.originlab.com/. Comments: Excellent work. Very thorough. Writing style very clear. Nice citing of references. Improvements: - I encourage you to move beyond the depth of the material in the textbook. For the first project, your analysis is fine. But I'm looking for a greater level of critical and creative thinking. Mathematica is a tool to be used to analyze the data. Leverage it to analyze the data in a critical and creative way and to reveal interesting and/or peculiar phenomena. - Clarity between precision and accuracy - Determining if data is/are outlier data - Compare your results to other work in the literature - Can extend the discussion to the diffusion mechanisms in Al2O2.