No Data Left Behind Modeling Colorful Compounds in Chemical Equilibria Mike DeVries D. Kwabena Bediako Prof. Douglas A. Vander Griend Outline • What is chemical equilibrium? • What makes color data good or not-so-good? • How does matrix algebra work again? • What is Sivvu and how does it work? What is Chemical Equilibrium? When chemicals react, they ultimately form a balance between products and reactants such that the ratio is a constant: [products]/[reactants] = Kequilibrium Log(products) – Log(reactants) = LogKeq = -G/RT G is called free energy. Example Seal .04 mole of NO2 in a 1 liter container. 2NO2 N2O4 G = -5.40 kJ/mol (@ 25°C) [N2O4]/[NO2]2 = Keq = exp(-G/RT) = 8.97 Let x amount of NO2 that reacts. (0.5x)/(.04-x)2 = 8.97 x = .01304, or 32.6% reacts Demonstration AB Secretly choose A or B and submit choice. As directed, choose A or B again: 1. 2. If last time you submitted A, now submit B. If last time you submitted B, 3. Submit B again if coin flip shows tails Submit A if coin flip shows heads Repeat #2 as directed. Multiple Equilibria Solving one equilibrium equation can be tricky. Solving simultaneous equilibria requires a computer. Exhaust Example Calculate the equilibrium amounts of CO2, N2, H2O, CO, O2, NO, and H2 after burning 1 mole of C3H8 in air. 4 mass balance equations, 1 for each element: • Carbon = 3, Hydrogen = 8, Nitrogen = 40 and Oxygen = 10 3 equilibrium equations: • • • 2CO + O2 2CO2 N2 + O2 2NO 2H2 + O2 2H2O G = -187.52 kJ/mol G = 125.02 kJ/mol G = -247.86 kJ/mol Exhaust Example 4 mass balance equations: • • • • Carbon = 3 Hydrogen = 8 Nitrogen = 40 Oxygen = 10 3 equilibrium equations: • • • [CO2]2/[CO]2/[O2] = 41874 [NO]2/[N2]/[O2] = 0.001075 [H2O]2/[H2]2/[O2] = 1134096 Species Amount CO2 2.923 N2 19.99 H2O 3.980 CO 0.07667 O2 0.03471 NO 0.02732 H2 0.02006 Metal Complexation Reactions [Ni]2+ + py [Nipy]2+ G1 [Nipy]2+ + py [Nipy2]2+ G2 [Nipy2]2+ + py [Nipy3]2+ G3 [Nipy3]2+ + py [Nipy4]2+ Ni G4 [Nipy4]2+ + py [Nipy5]2+ G5 [Nipy5]2+ + py [Nipy6]2+… Pyridine and Nickel Ni2+ and BF4- in methanol equivalents of pyridine added: 1 2 3 4 5 UV-visible Spectroscopy 0.1046 M Ni(BF4)2 in methanol 0.9 0.8 Ni2+ Absorbance 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 400 500 600 700 800 900 1000 Wavelength(nm) Absorbance = ·Concentration (Beer-Lambert Law) Color is Additive py 1.2 Absorbance 1 0.8 0.6 0.4 0.2 0.0997 M Ni(BF4)2 w/ 0.585 pyridine 0 400 500 600 700 800 900 Wavelength(nm) Abs = 0[Ni]2+ + 1[Nipy]2+ + 2[Nipy2]2+ + 3[Nipy3]2+ + … 1000 What we know is not very much [Ni]2+ + py [Nipy]2+ G1 [Nipy]2+ + py [Nipy2]2+ G2 [Nipy2]2+ + py [Nipy3]2+ G3 [Nipy3]2+ + py [Nipy4]2+ Ni G4 [Nipy4]2+ + py [Nipy5]2+ G5 [Nipy5]2+ + py [Nipy6]2+… The Problem We don’t know the wavelength-dependent colors or the equilibrium constants! We can’t measure the independent color (absorptivities) because all the compounds are present together. We don’t know the amounts of the compounds because they have equilibrated. Almost all the data is composite. The Solution It is possible, using advanced mathematical computations, to isolate information about pure species without chemically isolating them. How? Generate more composite data by making more mixtures with differing amounts of reactants. Model all the data according to chemical equilibria and the Beer-Lambert law for combining absorbances. Why it Works Abs = 0[Ni]2+ + 1[Nipy]2+ + 2[Nipy2]2+ + 3[Nipy3]2+ + … Each data point corresponds to a single equation. For each point on the same curve, the [concentrations] are the same. For each point at the same wavelength, the molar absorptivities, n, are the same. With enough solution mixtures then, there will be more equations than unknowns. This is known as an overexpressed mathematical system which can theoretically be solved with error analysis. Matrix Algebra Refresher 2x + 3y = 8 3x – y = 5 2 3 x 8 3 1 y 5 Matrix Multiplication C = AB (n x p) (n x m) (m x p) Matrix Algebra Abs = 0[Ni]2+ + 1[Nipy]2+ + 2[Nipy2]2+ + 3[Nipy3]2+ + … Abs 0 1 Ni2 2 2 Nipy Nipy 2 2 Matrix Form of Beer-Lambert Law Abs p C n n m m p m C Absorbances (n x p) Abs component Molar Absorbtivity Abs (n x m) Concentration C (m x p) component m n # of wavelengths m # of chemical species p # of mixture solutions Measured Absorbances Every column is a UV-vis curve. Every row is a wavelength Absorbances (n x p) n = number of wavelengths p = number of solution mixtures So there are a total of np absorbance data points, each associated with a distinct equation. The total absorbance at any particular point is the sum of the absorbances of all the chemical species in solution according to Beer-Lambert Law. Molar Absorptivities Every column represents one of the m chemical species. Every row is a wavelength (n x m) This smaller matrix contains all of the molar absorbtivity values for each pure chemical species in the mixtures at every wavelength. Component Concentrations Every column corresponds to one of the p solution mixtures (UVvis curve). Concentration (m x p) Each row represents one of the m chemical species. This smaller concentration matrix contains the absolute concentration of each chemical species in each of the solution mixtures. Matrix Absorbance Equation Abs = C Concentration (m x p) C Absorbances (n x p) Abs component (n x m) So the problem is essentially factoring a matrix. …Or solving np equations for mn + mp unknowns. With Residual Error! Abs = C + R Concentration (m x p) Absorbances (n x p) (n x m) Residual (n x p) Given data matrix of absorbances, find the absorptivity and concentration matrices that result in the smallest possible values in the residual matrix. Factor Analysis What is m? • • How many pure chemical species? How many mathematically distinct (orthogonal) components are needed to additively build the entire data matrix? The eigenvalues of a matrix depict the additive structure of the matrix. Requires computers. Matlab is very nice for this. Random Matrix Structure Eigenvalue Significance 1.4 1.2 1 0.8 (16 x 461) matrix of random numbers 0.6 0.4 All 16 eigenvalues contribute about the same to the structure of the matrix 0.2 0 1 2 3 4 5 6 7 8 9 Component 10 11 12 13 14 15 16 Non-random Data 50 data curves of Ni2+ solution with zero to 142 equivalents of pyridine. How many additive factors exist in this data? -0.15 -0.2 400 500 600 700 800 Wavelength (nm) 900 1000 Data Matrix Structure Relative Significance 7 absorbance data random numbers 6 5 m=6 4 3 6th eigenvalue is relatively small, but still possibly significant. 2 1 0 10 20 30 Factor Ratios 40 50 Equilibrium-Restricted Factor Analysis Factoring a big matrix into 2 smaller ones does not necessarily give a positive or unique answer. We also the force the concentration values to adhere to equilibrium relationships. Sivvu Inputs • raw absorbance data Absorbance Data n = 305 wavelengths p = 50 solution mixtures Sivvu Inputs • • raw absorbance data composition of solutions for mass balance equations Composition of Solutions 1 Ni++ Log(Concentration) 0.5 Py 0 From 0 to 142 equivalents of pyridine. -0.5 -1 -1.5 Pure pyridine (12.4 Molar) is dripped into 0.10 M Ni(BF4)2 Solution -2 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 Solution number Sivvu Inputs • • • raw absorbance data composition of solutions for mass balance equations chemical reactions for equilibrium equations Inputs Sivvu Inputs • • • • raw absorbance data composition of solutions for mass balance equations chemical reactions for equilibrium equations guesses for G’s Inputs Sivvu Inputs • • • • raw absorbance data composition of solutions for mass balance equations chemical reactions for equilibrium equations guesses for G’s Process • Calculates concentrations from G’s Sivvu Inputs • • • • raw absorbance data composition of solutions for mass balance equations chemical reactions for equilibrium equations guesses for G’s Process • • calculates concentrations from G’s solves for wavelength dependent colors Sivvu Inputs • • • • raw absorbance data composition of solutions for mass balance equations chemical reactions for equilibrium equations guesses for G’s Process • • • calculates concentrations from G’s solves for wavelength dependent colors calculates root mean square of residuals Sivvu Inputs • • • • raw absorbance data composition of solutions for mass balance equations chemical reactions for equilibrium equations guesses for G’s Process • • • • calculates concentrations from G’s solves for wavelength dependent colors Calculates root mean square of residuals Searches for G’s that minimize rms residual Structure of Residual Residual Factors in Measured Data 0.3 1 2 3 4 5 0.2 0.1 0 -0.1 -0.2 -0.3 -0.4 400 500 600 700 800 900 1000 Ni Now we know a lot! [Ni]2+ + py [Nipy]2+ G1 = -6.76(2) kJ/mol [Nipy]2+ + py [Nipy2]2+ G2 = -3.52(2) kJ/mol [Nipy2]2+ + py [Nipy3]2+ G3 = 0.64(3) kJ/mol [Nipy3]2+ + py [Nipy4]2+ G4 = 5.8(5) kJ/mol [Nipy4]2+ + py [Nipy5]2+ G5 = ? [Nipy5]2+ + py [Nipy6]2+ G6 = ? So what color is [Nipy]+2? Ni++ NiPy++ NiPy2++ NiPy3++ NiPy4++ Ni++ NiPy++ NiPy2++ NiPy3++ NiPy4++ What’s going on in 25th solution? 1.2 0.0997 M Ni(BF4)2 w/ 0.585 pyridine Asorbance 1 0.8 0.6 0.4 0.2 0 400 500 600 700 800 Wavelength (nm) 900 1000 Summary Conclusions Factor Analysis extracts much useful information about complex systems from easy experiments. Forcing the concentrations to satisfy chemical equilibria greatly enhances stability and sensibleness. SIVVU is Uv-vis in reverse. Acknowledgements American Chemical Society Petroleum Research Fund Research Corporation Cottrell College Science Award Pleotint L.L.C. Calvin College Research Fellowship Jack and Lois Kuiper Mathematics Fellowship