Mathematical Methods and the BornOppenheimer Approximation

The “degree of difficulty” is given in parentheses after each problem. Your total score on each problem will be the total number of points you earn on the problem (out of twenty) times the degree of difficulty. The average degree of difficulty will usually be about 1.0 (and is occasionally a bit more). You must choose two problems off this set.


The Dirac Delta Function (.8)

On the first day of class, I talked about the Dirac Delta function. The defining relation of the Dirac delta function is that

f (x ) δ (x − x 0 )dx ≡ f (x 0 )



Show that the Dirac delta function is normalized. That is, show that
∞ −∞

∫ δ (x − x


)dx ≡ 1


Next, recall that any function can be expanded in terms of a complete set of orthonormal basis functions— for instance, the eigenfunctions of the harmonic oscillator Hamiltonian (or, for that matter, any Hamiltonian).


∞ Let {ψn (x )}n =0 be a complete set of orthonormal functions. Show that the Dirac delta function is given by

n =0


(x 0 ) ψn (x ) ≡ δ (x − x 0 ) .


The left-hand-side of Eq. (I.3) is called the Poisson kernel. If the sum doesn’t extend all the way to infinity, it is called the Christoffel-Darboux kernel.


Use integration by parts to show that
∞ −∞

f (x )

d (δ(x −x 0 )) dx

dx ≡ − dfdxx

( )

x =x 0

= −f ′ (x 0 )


In three dimensions, we have δ (r − r0 ) ≡ δ (x − x 0 ) δ (y − y 0 ) δ (z − z 0 )



Confirm that

∫ f (r ) δ (r − r )dr = f (r ) .
0 0


(Here, and in the following, integrations without limits are assumed to be over all space. So
∞ ∞ ∞


f (r )dr ≡

−∞ −∞ −∞


f (x , y, z )dxdydz


In Spherical coordinates, 1 δ (r − r0 ) = r sin θ δ (r − r0 ) δ (θ − θ0 ) δ (φ − φ0 )



Explain why the

1 r 2 sin θ

term is needed.

When I introduced the Dirac delta function, I said that ∫ ρ (r ) δ (r − r0 )dr = ρ (r0 )

represented the probability of observing an electron at r0 , if ρ (r ) is the electron density—the distribution function for electrons.


Let P (r ) be the charge density for a single classical particle. What sort of physical system is characterized by P (r ) ≡ Zeδ (r − r0 ) ? (I.10)

The Dirac delta function is usually constructed using “delta function sequences”. The simplest such sequence is the function ⎧0 ⎪ x < −a ⎪ ⎪1 ⎪ fa (x ) = ⎨ 2a −a ≤ x ≤ a (I.11) ⎪ ⎪0 x >a ⎪ ⎪ ⎩


Show that
a → 0+

lim fa (x ) → δ (x )


by showing that this function satisfies the key relation, Eq. (I.1).
From this result, do you see why I said that Eq. (I.9) represents the “expectation value” for observing an electron at the point r0 ? Hint for part G: Remember, from calculus, that the average value of a function on an interval is

average value of f in the interval [a,b] ≡ f ≡

1 b −a

∫ f (x )dx


Equation (I.13) is easy to remember: it is just the continuous version of the formula for the mean of a data set,

f =

1 N

n =1




Eigenfunctions, Eigenvalues, and the Variational Principle (.7)
A. Show that if ψi (x ) and ψj (x ) are eigenfunctions of this operator with different eigenvalues, then they are orthogonal. That is,

ˆ ˆ Let A be a Hermitian operator. For the purposes of this exercise, you may assume that A is a real-valued operator, whence all its eigenfunctions can be chosen to be real.

∫ ψ (x ) ψ (x )dx = 0
i j



=0} ˆ ΨH Ψ Ψ Ψ (II.4) 3 .8) In class.4) ˆ where ψi (r ) are the eigenfunctions of H . (III.B. and that ψi (x ) ≠ ψj (x ) . look it up.3) ∫ ∇ ⋅ F (r )dr = ∫∫ A ⋅ nda R →∞ If you don’t remember what a surface integral (that thing on the right side of Eq. E first excited state = min {Ψ Ψ Ψ g .5) III. Namely. The key idea is to parametrize the surface as S (u. Chief among these is Gauss’s divergence theorem.2) +2h (r ) ∇f (r ) ⋅ ∇g (r ) + 2g (r ) ∇f (r ) ⋅ ∇h (r ) + 2 f (r ) ∇g (r ) ⋅ ∇h (r ) There are some very useful identities from multivariate calculus that you should remember. Derive Eq. (III. E. Why? ˆ ΨH Ψ Ψ Ψ A very important result in chemistry is the variational principle.1) A. D. v ) i + y (u.2) From A and B. (III. v ) ≡ x (u. Prove that E ground state = min Ψ ˆ ΨH Ψ Ψ Ψ (II. I said that ∇2 f (r ) g (r ) ≡ f (r ) ∇2g (r ) + 2∇g (r ) ⋅ ∇f (r ) + g (r ) ∇2 f (r ) (III. χa (x ) = cai ψi (x ) + caj ψj (x ) and χb (x ) = cbi ψi (x ) + cbj ψj (x ) so that ∫χ C.3)) is.s . show that we can find the energy of the first excited state by minimizing the energy with respect to all wave functions that are orthogonal to the ground state. Suppose ψi (x ) and ψj (x ) are normalized eigenfunctions with the same eigenvalue. v ) j + z (u.1) The generalization to three function is sometimes useful ∇2 f (r ) g (r ) h (r ) ≡ g (r ) h (r ) ∇2 f (r ) + f (r ) h (r ) ∇2g (r ) + f (r ) g (r ) ∇2h (r ) (III. That is. v )k (III. Now. a (x ) χ (x ) dx b =0 (II. the ground state wave function is found by minimizing with respect to all wave functions. we can prove that the eigenvectors of a Hermitian operator can always be chosen so that they are orthogonal. Show that we can find a two linear combinations of these two functions. Some Identities in Vector Calculus (.3) Hint: any wave function can be written as Ψ (r ) = ∑ ci ψi (r ) i =0 ∞ (II.

respectively).5) where the double integral is performed over the entire region of space. the wave function must decay faster than Ψ (r ) ∼ 1 r2 3 for large r . and z directions. v )dudv . using we have that Ψ (r ) dr → ∞ 2 2 ∂ Ψ(r .3). v ] ≡ ∫∫ A ⋅ nda ≡ ∫∫ F (u. define the spherical average of (III. Show that each of the following formulae gives the correct kinetic energy for a one-particle system ∫ D. The main case we’ll use in this course is when integration is over all of space. and k are the unit vectors in the x.φ) ∂r ˆ ≡ ∇ Ψ (r ) ⋅ r . (III. then (III. C.14) ∫ Then. Then N [u. Because 2 (III. j. so that Ψ * (r ) = Ψ (r ) .8) are the three-dimensional analogues of integration by parts.7) (III.9) Similarly. show that T [Ψ] = 2 8m ∫ 2 ∇ Ψ (r ) ⋅ ∇ Ψ (r ) dr 2 Ψ (r ) 2 2 (III. 2 ∇Ψ* (r )⋅∇Ψ(r ) 2m dr = ∫− 2 Ψ * (r )∇2 Ψ(r ) 2m dr = T [ Ψ ] (III. Show that if Ψ (r ) decays as for large r .10) = − 8m ∫ Ψ (r ) ∇ (ln Ψ (r ) )dr 2 2 2 In parts C and D.8) ∫∫ (φ (r ) ∇ψ (r ) − ψ (r ) ∇φ (r )) ⋅ nda B. y. if the wave function is real. you’ll find it helpful to assume that the integral extends over all space and that the wave function is normalized.12) Ψ (r ) with Ψ (r ) ≡ ∫∫ Ψ (r ) sin θ d θd φ 1 r 3 2 E.θ. Derive Green’s theorem (III.7) and (III. In this case.( i .8) from Green’s first identity (III. the result simplifies to R →∞ ˆ ∫∫ A ⋅ nda = ∫ ∫ lim ⎡⎣(A ⋅ r ) r 0 0 r →∞ π 2π 2 ⎤ sin θ d θd φ ⎦ (III. (III. ∂S ∂ S × . Eq.13) (III.11) ∫ Ψ (r ) dr = 1 .3) is the analogue of the fundamental theorem of calculus. Then ∂u ∂ v (III.6) ˆ where r is the unit vector in the outward-looking radial direction. Similarly important are Green’s first identity 2 ∫∫∫ φ (r ) ∇ ψ (r ) + ∇φ (r ) ⋅ ∇ψ (r )dr ≡ ∫∫ φ (r ) ∇ψ (r ) ⋅ nda and Green’s theorem 2 2 ∫∫∫ φ (r ) ∇ ψ (r ) − ψ (r ) ∇ φ (r )dr ≡ (III. To see this. v ) ⋅ N (u. 2 4 . and we denote this with the notation R → ∞ on the right-hand-side of Eq.7) Equations (III.

Equation (III. 5 . as many physicists now believe. equivalent. (III. If. Show that ∫∇ 2 Ψ (r ) dr = 0 .18) is the same as (III. then Eq. etc. (III. Eq. In three dimensions.17) must hold. Eqs. G. 1 ∇2 ( r ) = −4πδ (r ) (III. show that (in any number of dimensions) 1 ∇( r ) = − At some level.16) is needed to prove the desired results. θ and φ . the analogue of Eq. (III. 2 (III.17) by showing that ∫∫∫ ∇ 2 1 r dr = −4 π (III. First. where it is indeterminate. then.16) Equation (III. and we might observe deviations from inverse-square forces.21) I.17) would not represent the potential due to an elementary charged particle. membranes.10). consequence would be that photons would have mass.17) would not be valid and we might observe this effect by an as-yet unobserved perturbation away from inverse-square forces. Equation (III. (A second. but is normalized to −4π . our universe has more than 3 “spatial” dimensions. the Laplacian of a function that does not depend on the angular coordiantes.20) r3 In three dimensions. we should observe that gravity and electromagnetic forces are slightly weaker (slightly shorter ranged) than their commonly supposed r12 dependence.18) Ωd = ∫ ∫ … ∫ ∫ sin (θ ) sin 1 0 0 0 0 2 (θ2 ) sind −2 (θd −2 )d θd −2d θd −3 …d θ2d φ (III.). (III. (III. (III. r . (III.9) and (III. H.17) This equation is often cited as the “reason” electronic and gravitational forces are inverse-square forces (and thus 1 have r potentials). can be written ∂ ∇2 f (r ) = ( r12 ) ∂r r 2 ∂∂r (III. if even the most elementary particles in the universe are not point-charges and point-masses (many physicists believe they look like high-dimension strings. In fact. Show that ∇2 ( r1 ) = 0 unless r = 0 .19) If the physicists are right.17) in d ≥ 3 dimensions is ∇2 ( r d1−2 ) = where 2π π π π −1 d −2 Ωd δ (r ) (III.) Similarly.∂ Ψ(r ) 2 ∂r < ∂ ( ) 1 r3 ∂r < 1 r4 (III.17) when d = 3 .17) is the reason our universe acts the way it does. Show that Eq.22) 1 Since ∇2 ( r ) = 0 except when r = 0 . Complete the derivation of Eq.15) at long range. This can be used to F. J. We should derive this result.

IV. Then. U ({Rα }) . {Rα })) ψeBO ({ri }.R (IV. (IV. The Born-Oppenheimer Approximation (. By comparison.R indicates that the integration variables include both the electronic and the nuclear coordinates. (IV. here the notation.R ˆ ≥ Ψexact Tn + U BO ({Rα }) Ψexact r . using the fact that Eexact = ˆ ˆ Ψexact Tn + Te + V ({ri }.7) and so ˆ Ψ Te + V ({ri }.10) (Tn + U BO ({Rα })) χnBO ({Rα }) = E BO χnBO ({Rα }) .{Rα }) ψe ψe ψe r r = U ({Rα }) (IV. we have that ˆ Ψ Te +V ({ri }. and we could use the variational principle to determine the potential energy surface. We can consider Ψexact ({ri }. {Rα } is constant. (IV.5). {Rα }) . (IV. Ψ Ψ r . Ψexact ({ri }. 6 .{Rα }) Ψ Ψ Ψ r r ≥ ˆ ψe Te +V ({ri }.1) From the variational principle. {Rα }) χn ({Rα }) and determine ψe using the “electronic Schrödinger equation” (T e + V ({ri }. Show that Eexact Ψexact Ψexact r .3) Ψ BO ≡ ψeBO ({ri }. {Rα }) Ψ r .R Eexact = min . from Eq. in the Born-Oppenheimer approximation. U BO ({Rα }) = min ψe ˆ ψe Te + V ({ri }. r .7) The goal of this exercise is to show that the ground state total energy computed using the BornOppenheimer approximation is a lower bound to the true energy. {Rα }) Ψexact Ψexact Ψexact r .R . (IV. Let’s consider the exact ground state wave function. which solves ˆ (T n ˆ + Te + V ({ri }. {Rα }) ψe ψe ψe r r (IV.R r .) Now. (IV.6) (IV.7) and (IV. {Rα }) Ψ r ≥ U ({Rα }) Ψ Ψ r . recall that the nuclear wave function in the Born-Oppenheimer approximation is determined from the Schrödinger equation. {Rα }) .2).R Ψ (IV. the ground state energy could also be found by ˆ ˆ Ψ Tn + Te + V ({ri }.9) (Hint: use Eqs. {Rα }) to be an “approximate wave function” for the electronic Hamiltonian. Now.4) For any given set of nuclear positions.8). {Rα })) Ψexact = Eexact Ψexact . indicates that the expression is only integrated with respect to electronic coordinates.2) The notation r . {Rα }) = U BO ({Rα }) ψeBO ({ri }. ˆ (IV.8) A.5) Unlike the case in Eq. we approximate the wave function with BO (IV.

χm. etc. kmµ . Since we are interested in “correcting” the Born-Oppenheimer approximation. (IV. U m ({Rα }) and ψm ({ri }.1 ({Rα }) and Em . {R }) ˆ (V.µ ({Rα }) form a complete orthonormal set.4) The notation here is the same as that in problem IV: r denotes that the integration is performed only over the electronic variables (the nuclear coordinates are fixed). {R })⎤ ψ BO ({r }.2). {Rα } . C. For any given position of the atomic nuclei.R . Here. (V. {Rα }α=1 denote the positions of the atomic nuclei.10). we consider the eigenfunctions of the BornOppenheimer Hamiltonians. Write the variational principle corresponding to Eq. of course) for the m th electronic excited state of the BO BO system. any wave function can be expanded in terms of a complete set of basis functions. so that exact BO BO Ψexact ({ri }. {Rα }) is the ground state electronic wave function. {ri }i =1 denote the positions of the electrons.0) Sometimes the Born-Oppenheimer approximation just isn’t good enough.R denotes integration over both electronic and nuclear coordinates. and so on. the nuclear wave functions are determined by solving the nuclear Schrödinger BO BO equation.µ ({Rα }) . {Rα }) ≡ ∑ ∑ km µ ⋅ ψm ({ri }. {Rα }) is the electronic wave function of the first excited state.1 denote the wave function and energy of the first nuclear excited state on the BO potential energy surface of the m th electronic excited state. which is a “diagonal correction” to the Born-Oppenheimer approximation. In that case. ⎡T + V ({r }. for any given set of nuclear positions. ψ1BO ({ri }.0 denote the ground state nuclear wave function and the ground state total energy (in the Born-Oppenheimer approximation. We choose to expand the exact wave function in this basis. BO ψ0 ({ri }. one often uses the Born-Huang approximation.9) and the result from part B to show that Eexact ≥ E BO . can be determined by evaluating exact BO BO km µ = ψm χm. ψm ({ri }.1) i i m m i α α α α ⎢⎣ e ⎥⎦ m and BO BO BO BO ˆ (V. V ({ri }. 7 . (IV. In this problem we’ll derive the Born-Huang approximation and discuss its implications for chemical reactivity.B. {Rα }) denotes the sum of the electron-electron. In these equations. R denotes integration over only the nuclear coordinates (the electronic coordinates are fixed). {R }) = U BO ({R }) ψ BO ({r }.µ ({Rα }) µ= 0 m =0 ∞ ∞ (V. {Rα }) and BO χm. Taken together.. That is. The Born-Huang Approximation: the importance of nonadiabatic effects (1. As you know.µ χm.µ Ψexact r . nuclear-electron. and r .2) (Tn + U m ({Rα })) χm. {Rα }) denote the energy and N P wave function of the m th excited state of the system.µ ({Rα }) = Em. Eq. (V. For any given electronic state. {Rα }) χm . Use Equation (IV. χm.11) V.0 ({Rα }) and Em . and nuclear-nuclear potential BO energy.3) where the expansion coefficients.

we can find exact wave function for the system by minimizing the energy.4). In particular. What used to be the “continuum” electronic and nuclear wave functions now resemble the states of the particle in a box.µ } m . µ=0 ∑ ∞ km . One way is to go through the (somewhat complicated) mathematics necessary to treat “continuum wave functions”—wave functions that cannot be normalized. albeit not the electrons). When the electronic energy is greater than zero. Now. Derive Eq.µ k BO BO ˆ BO BO BO χn .ν m . which can not be normalized. This construction is the wave the wave functions used here should be considered. we can write the summand in the numerator of Eq. There are several ways to solve these problems. (V. the part of the nuclear wave function associated with the center of mass is exactly that due to a particle in a box. ∞ ∞ (V.R {km . the likelihood of any electron or nucleus being close to the boundary of the “box” is very small. because the “entire molecule” can move freely.5) Ψ Ψ r . we recover the exact result. ({ri }) ∼ e ik⋅ri . {Rα }) ) and the potential outside the sphere is infinite for both the electrons and the nuclei. (V. {Rα }) χm.µ 2 (V.R Eexact ≡ min . Put the molecule in a large sphere of radius L . BO BO Ψ ({ri }.6) µ=0 m =0 we find that Eexact ≡ min ∑ ∑ ∑ ∑k µ=0 m =0 ν =0 n =0 ∞ ∞ ∞ ∞ * n .µ ({Rα }) = e i(k⋅RCOM ) = cos (k ⋅ RCOM ) + i sin (k ⋅ RCOM ) . There is a subtle point related to the above treatment. so ˆ ˆ Ψ Tn ({Rα }) + Te ({ri }) + V ({ri }. chemists tend to take a “pragmatic” approach. so the chemical properties of the system are unaffected by this change.1) .7).A. then.µ ({Rα }) .7) as 8 . Using the result from Eq. all the wave functions in the problem can be normalized. However. where RCOM is the position of the center of mass of the particle. {Rα }) ≡ ∑ ∑ km µ ⋅ ψm ({ri }. M α α=1 P Recall that the nuclear kinetic energy operator is (V.7) B. if L is taken to be very large. (V. ˆ Tn ({Rα }) ≡ ∑ 2−1α ∇2 . Finally. where the potential inside the sphere is just that due to the electrons and nuclei (V ({ri }. In addition.8) where P is the number of nuclei and M α is the mass number for the αth nucleus. Using the variational principle.µ ψm r . (V.ν ψn Tn ({Rα }) + U m ({Rα }) χm . With this ansatz. One can also “separate out” the motion of the center of mass from the nuclear equation (which solves the problem for the nuclei. ψ BO m Derive Eq. the nuclear wave functions have a component (corresponding to the free translation of the center of mass) like BO χm.R Ψ If we expand Ψ in terms of the Born-Oppenheimer states. {Rα }) Ψ r . (III. taking the limit L → ∞ .

Unlike the BornBO Oppenheimer approximation.9) is called the non-adiabatic term. The interpretation is that the motion of the nuclei can “induce” a transition from the ground electronic state to an electronic excited state.µ =0 ∑ km .µ ⋅ ∇α ψm )⎥ dr1 …drN dR1 …dRP ⎢⎣ ⎥⎦ α=1 ∫ (χ BO n .µTn ψm ⎤⎥ dr1 …drN dR1 …dRP ⎦ ⎡ BO BO * P −1 ⎤ BO BO ∫ ⎢⎢⎣(χn.10) δij = ⎪ ⎨1 i = j. is quite useful.) So let us ignore the last term in Eq.11) ∞ {km .µ 2 Equation (V. µ } m .µ ∑ ∑ ∫∫∫ ∫ (χn . (Hint: remember that the eigenfunctions are all orthonormal.µ ⋅ ∇ψm )⎥⎥⎦ dr1 …drN dR1 …dRP α=1 r . Eq. (This of course.µ + (χn .ν ψn U m ({Rα }) χm.ν ψn ) χm . ⎪ ⎪ ⎩ Before going any further.ν ψnBO )* χm . (V. Noting that the energy of the nuclear energy states is essentially the energy of molecular vibrations. but it is difficult to evaluate because the second term in the numerator couples different electronic states of the molecule.9).ν <µ .µ + ∫∫∫ BO * BO ˆ BO BO BO * BO ˆ BO ψn ) ψm Tn χm.) Born and Huang derived a simpler approximation by assuming that BO * ˆ BO (V.R (V. one cannot separate the motion of the electrons from the motion of the nuclei.µ 2 E m .ν ψn ) ∑ M α (∇α χm. BO BO the nuclear wave function must move from an excited state to the ground state χ0. (V.7) then becomes ∞ ∞ ∞ ∞ ⎛ ⎞ BO BO BO ˆ BO ⎜ km . (Remember. since it means that the “physics” of the minimal adiabatic approximation is correct.ν +∫∫∫ BO BO BO BO BO + χn. (V. energy is BO BO conserved.µ + kn*. When it is included.11) gives Golden’s minimal adiabatic approximation. In particular.µ → χm>0. R2. if the molecule moves from the electronic ground state to an electronic excited state ψ0 → ψm>0 . (V. That is.9).) The last term in Eq.µTn ψm dr1 …drN dR1 …dRP ⎟ ⎟ ∑ ∑⎜ ⎟ ⎜ ⎠ µ =0 m =0 ⎝ ν =0 n =0 Eexact ≡ min (V. This is why it is only when the separation between nuclear energy levels (and the largest separations are associated with molecular vibration) and the separation between electronic energy levels (which are usually a hundred times greater than the nuclear energy spacings) are similar in size that non-adiabatic effects have chemically important consequences. {Rα }) are coupled to the BO ground state.ν BO BO * BO ˆ ψn ) χm. ψm>0 ({ri }. It is the most accurate “adiabatic” approximation known. let’s consider the interpretation of the last two terms in Eq. So.9).11) predicts the Jahn-Teller distortion of degenerate ground states. we cannot separate variables. (V. we expect this “transfer of energy” from the nuclei to the electrons can only occur when two electronic states are very close in energy. so that the “gap” between two electronic states is about the same size as the spacing between the vibrational energy levels in the molecule. …RP ) δnm 9 .µ ψm BO = δnm δνµEm. unlike the simpler approximation we shall consider.12) ∫∫∫ ∫ (ψn ) Tn ψm dr1 …drN ≡ Tm (R1. {Rα }) .∫∫∫ ∫ ⎢⎣(χ ⎡ BO n . in these terms the electronic excited states. ψ0 ({ri }.µTn ψm dr1 …drN dR1 …dRP C.ν km . Derive Eq.ν ψn ) ∑ Mα (∇α χm. (Because both the nuclear and the electronic wave functions are operated upon. Equation (V.9) +∫∫∫ ∫ ⎡ BO BO * P −1 ⎤ BO BO ⎢(χn. δij .) The Kronecker delta. is just the identity matrix. i≠j ⎧0 ⎪ (V.

(V. Use the variational principle to show that we can find the Born-Huang approximation by solving the Born-Huang nuclear Schrödinger equation.18) along the reaction path.µ ({Rα }) . (V.µ=0 BO Now. {Rα }) = ψm ({ri }.0 ≥ E 0. Recall that the primary error in both the Born-Oppenheimer and the Born-Huang approximation is that we do not account for the ability for the nuclear motion to cause a transition in the electronic state of the molecule.0 . as BH BH BH χm. so the total Born-Huang wave function is BH BO BH Ψ m. D. the curves typically look like 10 .ν χm. BH exact BO E 0. This is clearly incorrect. assuming the Born-Huang approximation. they assumed that the kinetic energy operator does not cause transitions between electronic states.ν ({Rα }) . so that the spacing between nuclear energy levels and electronic energy levels is similar. is valid.That is. unlike the Born-Oppenheimer approximation. because {χm .12).14) µ=0 E. we can write the nuclear wave BH function in the Born-Huang approximation. ∞ ∞ ∑ ∞ km . A + B −C → A−B + C .µ.µ {Rα }}µ=0 is a complete orthonormal set of functions.17) Proving this result is much more difficult that problem IV.ν ({Rα }) = Em .0 ≥ E 0. If you’re interested. ˆ (T n BO BH BH BH + U m ({Rα }) + Tm ({Rα })) χm.ν ({Rα }) . µ 2 (V.µ } ∑ ∑ ∑k ∞ ∞ ∞ * m. so I won’t ask you do so. Show that. χm.ν ({Rα }) = ∑ km . the total Born-Huang energy of the ground state is an upper bound to the true energy. (V.15) The electronic wave function in the Born-Huang approximation is just the Born-Oppenheimer electronic wave function. but it is often a good approximation. This approximation was judged to be especially severe when the spacing between electronic states was small. ask me and I’ll show you how it’s done.µ k BO ˆ BO BO χm. Eq. µ R m. then E BH ≡ min µ = 0 m = 0 ν = 0 {km .ν χm.13) (V. If we plot the energy of the ground state and the first excited state of a substitution reaction.ν ({ri }.ν ({Rα }) . (V.ν m . That is.ν Tn ({Rα }) + U m ({Rα }) + Tm ({Rα }) χm.16) One of the reasons the Born-Huang approximation is useful is because. {Rα }) χm. (V.

7). ∆BO ({Rα }) = U BO ({Rα }) − excited state. (V. Assuming this to be the case ˆ Ψexact Te +V ({ri }.(The idea that more stable states have larger gaps to the first excited state is called the maximum hardness principle. we can approximate the Born-Oppenheimer error as 11 . (IV. (V. ∆BH ({Rα }) < ∆BO ({Rα }) . and was proposed by Pearson in 1988. Where might the approximation be particularly problematic? Will the Born-Oppenheimer approximation tend to overestimate or underestimate reaction rates? What about the Born-Huang approximation? ∆BH ({Rα }) = U BO ({Rα }) + T ({Rα }) − ˆ Ψexact Te +V ({ri }. B = H + ). we have ≥ 0. For a proton-transfer reaction (referring to Eq.20) Usually. reactants. we can define the error in the Born-Oppenheimer approximation.{Rα }) Ψexact Ψexact Ψexact r r ≤ 0.) Using Eq. That is. Compare the accuracy of the Born-Oppenheimer approximation for the transition state. the Born-Huang approximation is usually more accurate than the Born-Oppenheimer approximation. and products.{Rα }) Ψexact Ψexact Ψexact r r In the Born-Huang approximation. (V.19) It is usually approximately true that ∆ ({Rα }) is inversely proportional to the gap between the ground and first F.18).

they can be very significant near the transition state. Especially interesting are “stationary” or “stability” principles.) kcal mol . when Eq. or minimum in a function.1) Equation (VI. (VI.21) where η is the energy gap between the ground and first excited states. maximum. (V. the energy will only change in a way 2 proportional to the norm (= size) of δΨ . if we take a wave function.9) You were undoubtedly told that the Schrödinger equation (or something like unto it) was the fundamental basis for quantum mechanics. δΨ .3) is of the order dr ⋅ dr —in the limit of very small dr such terms are negligible. if the error in energy is greater than 1 kcal .22) calculate how small η ({Rα }) must be before the error in the approximations exceeds 1 kcal . (At room temperature. (VI. The Calculus of Variations and Stationary States (.4 ∆BO ({Rα }) ≈ − 4 ( mep ) η ({Rα }) = − 3⋅1832 3 m 1 η {Rα } (V. via the formula dg (r ) ≡ g (r + dr ) − g (r ) = ∇g (r ) ⋅ dr + O (dr ⋅ dr ) . the average kinetic energy of molecules in a gas is about 1 we miss the activation energy by this much. In particular. namely 2 f (x + dx ) − f (x ) ∝ (dx ) (VI. dr to changes in g (r ) . provided they disappear into the vacuum sufficiently quickly). we make a significant error in the reaction rate. Using the approximations in Eqs.1) holds we can say that the functional derivative of E with respect to Ψ is zero. variational principles are considered to be much more fundamental.22) Based on the preceding figure. we may surmise that it is quite unlikely that non-adiabatic effects are important in the reactant or product states. the excited states of an isolated molecule would have an infinite lifetime. Recall the definition of the gradient of a function.2) is seen to imply that the derivative of f with respect to x is zero. (VI. which is precisely the familiar condition for stability.3) Here. In fact. (V. ∇g (r ) is defined as the vector that maps changes in r . Mathematically. predictions of the rates of chemical reactions could be off by mol an order of magnitude. Equation (VI. ∆BH ({Rα }) ≈ 4 3 m 4 ( m ) η ({Rα }) = 3⋅(1832) e p 2 2 1 η {Rα } . and the Schrödinger equation can be derived from such principles. that maps changes in Ψ . mol VI. Most modern physicists would disagree with this viewpoint. in Hartree. so if G.1) is analogous to the condition for a “point of inflection”. δA[ Ψ ] δΨ( τ) .21) and (V. (VI. Rather. were it not for an exotic phenomena called vacuum polarization (which has to do with the ability of empty space to spontaneously generate photons. That is. then Eq. We define the functional derivative in the same way—the functional derivative of A [ Ψ ] with respect to Ψ is defined as that function. Similarly. However.2) If you write f (x ) as a Taylor series. we say that E [ Ψ + δΨ ] − E [ Ψ ] ∝ δΨ δΨ (VI. dg . dA [ Ψ ] via the formula dA [ Ψ ( τ)] ≡ A [ Ψ ( τ) + δΨ ( τ)] − A [ Ψ ( τ)] = ∫ δA[ Ψ ] δΨ( τ) δΨ ( τ)d τ + O ( δΨ ( τ) δΨ ( τ) ) .1) is the reason that excited states of molecules are metastable. The Born-Huang error can be approximate in similar fashion. O (dr ⋅ dr ) denotes that the error in Eq. δΨ to changes in A [ Ψ ] .4) 12 . (VI. Ψ and change it by some small amount. The simplest such principle is the statement that the energy of a “stationary state” is stable with respect to small changes in the wave function.

7) so that δΨ ( τ) = ∑ dkn Φn ( τ) .10) B.” For instance. The coefficient of δΨ ( τ) is the functional derivative. (VI. (VI. just a “function of a function. (VI.4). but the argument of a simple function in parentheses. This is a more fundamental “derivation” of the Schrödinger equation than most (though one can do much by forcing a functional called the “action” to be stationary instead 13 . we can write the energy as a function of the coefficients of a basis function expansion: using (VI. We should say a few words about what a functional is.13) can be derived from the requirement that = 0 . (VI. and turns a functional into a function.10)). we denote the argument of a functional with in brackets.12) * l m The Schrödinger equation. For instance.3) and (VI. Show that δA[ Ψ ] δΨ( τ0 ) ⎛dA ⎡ Ψ + εδ ( τ − τ 0 )⎤ ⎞ ⎦⎟ ⎟ ⎜ =⎜ ⎣ (VI. because it can be computed from the wave function using ˆ ΨH Ψ . Show that Eq. Choose δΨ ( τ) = ε ⋅ δ ( τ − τ 0 ) in Eq. as should be clear from the above.5) is a useful formula for computing functional derivatives. ˆ H Ψ k = Ek Ψ k δE [ Ψ ] δΨ( τ) (VI. A functional is. the energy is a functional of the wave function. (VI. Yet another method follows when we write expand the wave function in a basis set. n =0 ∞ (VI. introducing a basis set for the wave function reveals that the functional derivative is really just a generalization of the gradient (compare Eqs.6).11) E [Ψ ] ≡ Ψ Ψ To remove possible ambiguity. Functionals and functions are closely related. = lim A ⎡⎣ Ψ + εδ ( τ − τ 0 )⎤⎦ − A [ Ψ ] Ψ ( τ) ≡ ∑ kn Φn ( τ) Ψ ( τ) + δΨ ( τ) ≡ ∑ (kn + dkn ) Φn ( τ) n =0 n =0 ∞ ∞ (VI. Eq.A.4). Φl Φm (VI.10) satisfies the definition of the functional derivative. ⎟ ⎟ ⎠ n = 0 ∂ kn (VI.5) ⎟ ⎜ ⎟ ⎜ ε dε ⎟ ε→ 0 ⎝ ⎠ε=0 Equation (VI.6) (VI. Another method follows directly from the definition: compute A [ Ψ + δΨ ] − A [ Ψ ] and write the result in the form of Eq.9) one finds that ∞ ∞ ⎛ ( ) δA [ Ψ ] ∂A ⎛ δkn ⎞ ⎟ = ∑ ∂A ⎜ δ Ψ τ Φn ⎜ =∑ ⎟ ⎜ ( )⎠ ⎜ δΨ τ ⎟ n =0 ∂k ⎜ δΨ ( τ) ⎜ δΨ ( τ) n =0 ∂kn ⎝ n ⎝ (VI. For instance. ∞ ∞ ˆ ∑ ∑ k *k Φ H Φ E (k ) = E ({kn }n =0 ) ≡ ∞ l m l m l = 0 m =0 ∞ ∞ ∑ ∑k k l =0 m =0 .8) Using the relationship A (k + dk ) − A (k ) = ∇k A (k ) ⋅ dk + O (dk ⋅ dk ) ∞ ∂A =∑ dkn + O (dk ⋅ dk ) n = 0 ∂k n ∞ ⎞ ∂A ⎟= ⎟ ∑ (Φn ) .4). (VI.

5).) δΨ ( τ) (I know of two straightforward ways to perform the derivation. or stable. and ˆ = φH φ . C.C allows one to perform the analysis. namely ⎛k ⎞ k (k − 1) 2 k ⎜ ⎟ (1 + x ) ≡ 1 + k ⋅ x + x + … + ⎜n ⎟ x n + ….16) where Ek is the energy of the stationary state.15) What is a stationary state? The following analysis might be helpful. This is important. The following figure depicts the situation for the first excited state of a system. but the k th excited state is a true saddle point—one can lower the energy by moving in a k directions (take φ ( τ) = ∑ cn Ψ n ( τ) ) but other directions (take φ ( τ) = n =0 k −1 n =k +1 ∑cΨ n ∞ n ( τ) ) raise the energy. (VI. Consider what happens when we change the wave function of a stationary state. III. Show that the Schrödinger Equation is equivalent to principle of stationary energy. D.17) The interpretation is that stationary states of systems are “saddle points”. (Read the following hint. However. the analysis is a bit complicated because of the kinetic energy operator. The key is to use Green’s theorem (cf. but this is complicated by the fact that the kinetic energy operator will operate on the delta function.8 in problem set 1) to write integrals like Ψ ∇2δΨ in terms of integrals like ∇2 Ψ δΨ . Show that 2 E [ Ψ k + εφ ] ≡ Ek + ε2 (Eφ − Ek ) − ε 4 (Eφ − Ek ) + … Eφ ≡ ˆ φH φ φ φ (VI. states of the system. 14 . Ψ k . In expanding the denominator you might want to use the geometric series. since it reveals that the eigenvectors and eigenvalues of the Hamiltonian operator represent stationary. by a small amount. but the second order change is negative if φ ( τ) has a lower energy than Ψ k ( τ) but positive if φ ( τ) has a larger energy than Ψ k ( τ) . δE [ Ψ ] = 0 . Ψ k . and not a maximum).focusing on the energy). The most straightforward approach is to calculate E [ Ψ + δΨ ] − E [ Ψ ] and rearrange the result into a form like E [ Ψ + δΨ ] − E [ Ψ ] = ∫ A[ Ψ ] δΨ (τ)d τ + O ( δΨ (τ) ) 2 (VI. (VI. The ground state is the absolute minimum energy state (and is thus a simple minimum.14) Again. Then (you will need to use the geometric series again). using the three-dimension generalization of problem I. One is to use Eq. We can assume that φ ( τ) and Ψ k ( τ) are orthogonal and normalized. Eq. the first order change in energy is zero. εφ ( τ) . ⎜ ⎟ 2! ⎝ ⎠ (VI.

(VII. ˆ We are given an unperturbed Hamiltonian. from Eq.1) When a set of degenerate orbitals (or wave functions) is placed in an electromagnetic field. the degeneracy is usually partially removed.The stability of a stationary state is related to “how broad” the saddle point is. (VII.5) 15 . (VI. not only for describing the splitting of spectral lines. but the same principle quantum number. H . VII. let’s consider the “perturbed” Hamiltonian. ˆ H ∑ ci Ψ i = Ek +1 ∑ ci Ψ i i =1 i =1 g (VII. We will derive the first order perturbation theory results for degenerate states in this problem. l .16). would be proportional to (Ek − El ) . This gives the right trend (higher excited states are inherently less stable and transitions from excited states to lower-energy states are favored. Thus. Now. ˆ ˆ ˆ H perturbed (λ ) ≡ H + λV . But the assumption of a linear dependence is wrong (for rather subtle reasons). Perturbation Theory for Degenerate States (1. {ci }i =1 . In addition. but also for describing the removal of the degeneracy of orbitals with different values of the orbital angular momentum quantum number. to a lower-energy stationary state. and we label the states using the Schrödinger equation. so ˆ H Ψ n ≡ En Ψ n . Ek . (VII. the lifetime of excited 3 states is proportional to (Ek − El ) . n . g ˆ Ψ n H Ψ n ≠ Ek +1 (VII. we assume that the functions. since this determines “how big” a perturbation is needed to a system away from its “stationary state” and force a transition to a lower-energy state.4) That is. El <k . one might expect that the rate at which a molecule spontaneously decays from an excited state. For this reason. the perturbation theory of degenerate states is a key tool. any linear combination of a set of degenerate eigenvectors is also an eigenvector of the Hamiltonian.1) ∞ We assume that {Ψ n }n =0 are orthonormal and form a complete set of functions. so that ˆ Ψ H Ψ n n = Ek +1 g k +1≤ n ≤k +g otherwise.3) Note that for any choice of the coefficients.2) form a g-fold degenerate state.

2). but still large enough to remove the degeneracy. very. eigenfunctions. ˆ ˆ ˆ H (0) = H + µV .16) i =0 As is usual in perturbation theory. the perturbation is very. (VII.12). (VII. That is.14) are our working equations. even when 0 < λ 1 —that is. small—the wave functions of the perturbed the system might not resemble those of the unperturbed system. but hopefully easier to follow. To get around this issue. That is. You can find standard treatments in most good quantum texts. In particular. namely.6) (VII. (VII. because of the assumed g-fold degeneracy in Eq. E λ λ En = En =0 + λ ∂∂λn λ λ =0 + 1 λ2 2 1 + 2 λ2 λ ∂ 2 En ∂λ2 λ =0 ∂2 Ψ λ n ∂λ 2 λ =0 +… +… (VII. (VII.14) n n where. ˆ ˆ ˆ H (µ) ≡ H (0) − µV .Unless the perturbation has a special symmetry. ˆ Because the eigenfunctions of H (µ) form a complete set. The usual procedure is to write the energy and wave function as a Taylor series in λ . we can write Ψ (0) as (0 Ψ (0) = ∑ cin ) Ψ n ∞ n (µ ) i . let’s turn the problem on its head. for λ ≠ 0 the degeneracy is broken.10) (The perturbation is assumed to be very small.4)) to use for λ k +g the zeroth order approximation to {Ψ n }n =k +1 .4)) that is not an eigenvector of the perturbed Hamiltonian.13) ˆ V = H (0) + µ (− ˆ) ˆ is just the original.7) ˆ ˆ (H + λV ) Ψ λ = E λ Ψλ is the Schrödinger equation describing the perturbed system. Hamiltonian. there is almost always some linear combination of the eigenvectors of the previous system (cf. but a bit complicated. For clarity. we expand quantities in Taylor series.9) λ Ψ λ = Ψ n =0 + λ ∂Ψn n ∂λ λ λ =0 The problem is that we don’t know which of the many possible unperturbed wave functions (cf. (VII.) For this reason. ( ( E 0µ) < E1µ) < < Ek(µ)1 = Ek(µ)2 = = Ek(µ)g < Ek(µ)g +1 . (VII. The eigenfunctions and eigenvalues of H (µ) are known since they are just the eigenfunctions of the original system. 16 . The way around this issue is straightforward. degenerate. ( ˆ H (0) Ψ (0) = En0) Ψ (0) (VII. (VII. (VII. The choice of notation is such that a superscript (µ) denotes the Hamiltonian. ( ˆ H (µ) Ψ (µ) = Enµ) Ψ (µ) (VII.11) and (VII.11) and (VII.11) n n and ( ( E 00) < E10) < < Ek(0)1 < Ek(0)2 < . ˆ ˆ ˆ Ψ λ+1 H perturbed (λ ) Ψ λ+1 ≠ Ψ λ+2 H perturbed (λ ) Ψ λ+2 ≠ ≠ Ψ λ+g H perturbed (λ ) Ψ λ+g k k k k k k where (VII. (VII. and what follows is a bit nonstandard. Eq. Consider a perturbation of the Hamiltonian which totally removes the degeneracy. Eq. This means that. we adopt a naming convention similar to that in Eqs.8) (VII.12) + + The “perturbed Hamiltonian”. and eigenvectors of the original degenerate system.15) + + + + Equations (VII. while a superscript ˆ (0) denotes the corresponding quantities for the nondegenerate system described with H (0) .

18) (VII. on the left-hand-side and right-hand-side of Eq. cin ) = δin . then k+ k+ k+ (µ (µ ˆ H (µ) ∑ cin ) Ψ (µ) = Ek(µ)1 ∑ cin ) Ψ(µ) . That is. This is because in order for f (x ) = ∑ an µn n =0 ∞ (VII. is zero i n unless its arguments are equal ( i = j ).) Thus.23) ( Let us first consider the case where Enµ) is a non-degenerate state. (VII. (Otherwise f (x ) > 0 most places. must all be equal. ( When Enµ) is a degenerate state. µ . ….ˆ (λ) ˆ ˆ H (µ) = H (0) + µ ∂Hλ ∂ λ =µ (VII. For µ 0 = 1 . We now substitute these expressions into Eq. (VII.25) (µ (µ This indicates that we can replace cin ) with cin )δEi En . all the an must equal zero. any linear combination of the degenerate eigenfunctions will also be an ˆ eigenfunction of H (µ) . cin = 0 unless Ψ (µ) and Ψ (µ) have the same energy.19) λ =µ + 2 (λ) µ2 ∂ cin 2 ∂λ 2 λ =µ + ….24) That is. we can summarize the consequences of Eq. f ( µ ) ≡ ∑ ln µ = n n =0 ∞ ∑r µ n n =0 2 ∞ n ≡ g (µ ) (VII. etc.22) we must have that l µ = u µ .11). So if the states Ψ (µ)1.4).21) to be zero everywhere. µ . i i i =0 i =0 ∞ ∞ (VII.20) must be equal. Then we must have that (to within an arbitrary (µ (µ ( complex phase factor). Recalling that the Kronecker delta.23) with (µ ( (µ ˆ H (µ) ∑ δEi En cin ) Ψ (µ) = Enµ) ∑ δEi En cin ) Ψ(µ) .17) λ =µ ( ( n Enµ) = En0) + µ ∂Eλ ∂ (µ (0 cin ) = cin ) + µ (λ ∂cin ) ∂λ (λ) λ =µ + ( µ2 ∂2Enλ) 2 ∂λ2 +… (VII. (VII. δij . + i i i =k +1 i =k +1 k +g k +g (VII. µ2 . so ∞ (λ) 2 (λ) 2 ⎛ ⎞ ∞ ( ( ˆ ˆ ⎟ ⎜ (H (µ) + µV ) ∑ cin0)Ψ(iµ) = ⎜En(µ) − µ ∂Eλn λ=µ − µ2 ∂∂Eλn2 λ=µ − …⎠ ∑ cin0)Ψ(iµ) ⎟ ∂ ⎟ ⎝ i =0 i =0 ∞ (λ 2 (λ) ⎛ (µ ) ⎞ (µ ) ∂cin ) µ2 ∂ cin (µ ) ˆ ˆ ⎟ (VII. Ψ(µ)2 .) For this reason. cin ) = 0 unless i = n when Enµ) is a non-degenerate state. cf.20) (H + µV ) ∑ ⎜cin − µ ∂λ λ=µ − 2 ∂λ2 λ=µ − …⎠ Ψi ⎟ ⎜ ⎟ ⎜ ⎝ i =0 (λ) 2 (λ) 2 2 ⎞ ⎛ ( ⎞ ∞ ⎛ (µ ∂c(λ) ∂2c(λ) n ⎟ ⎜ = ⎜Enµ) − µ ∂Eλ − µ2 ∂∂En2 − …⎟ ∑ ⎜cin ) − µ ∂in − µ2 ∂λin − …⎟ Ψ (µ) 2 ⎟ ⎜ ∂ λ ⎜ λ ⎟ ⎝ ⎟ ⎝ ⎠ i =0 ⎠ i λ =µ λ =µ λ =µ λ =µ The standard perturbation theory idea is that the coefficients of µ 0 = 1 . Ψ (µ)g are degenerate. (VII. etc. 17 . in order for two functions to be equal everywhere. µ . (Otherwise f (µ) − g (µ) = ∑ (ln − rn ) µn would not equal zero. this gives the equation (µ ( (µ ˆ H (µ) ∑ cin ) Ψ (µ) = Enµ) ∑ cin ) Ψ (µ) i i i =0 i =0 ∞ ∞ 0 (VII. n =0 ∞ we must have that the coefficients of µ = 1 . Eq.

20).19). ⎟ ⎟ λ =µ ⎠ (VII. In this case.28) (if E (µ) = E (µ) ) ∑ Ψ(µ) V Ψ(µ) c(µ) = − ∂En c(µ) m i in ∂λ i =0 λ =µ mn m n (µ ( Because cin ) = 0 unless Ei(µ) = Enµ) .33) 18 . The second term on the right-hand-side then vanishes.n . Show that if cn ≡ ⎡⎢⎣ck(µ)1. (VII.27).29).32) and ∞ ⎛ (µ ⎞ (µ ) ∂c(λ) ⎟ Ψi Ψ (0) = ∑ ⎜cin ) − µ ∂in ⎟ n λ ⎜ ⎟ ⎜ ⎝ λ =µ ⎠ i =0 (VII. λ ⎜ ∂ ⎟ ⎜ ⎝ i =0 i =0 ∞ ⎛ n(λ) ⎞ ∂c(λ) (µ ( ⎟ ⎜ ∂ cin ) + (Enµ) − Ei(µ) ) ∂in (VII. (VII. ck(µ)2. B. (VII. and so Eq. but is otherwise determined by Eq. this becomes i =k +1 ∑ k +g (λ) (µ n ˆ Ψ (µ) V Ψ (µ) cin ) = − ∂Eλ m i ∂ λ =µ (µ cmn) (if E ( ) = E ( ) ) µ m µ n (VII. show that you obtain the equation ∞ (λ) ⎛ n(λ) (µ ˆ (µ ) (µ ) (µ) ∂cin cin )V Ψ (µ) = −∑ ⎜ ∂Eλ ∑ i ⎜ ∂ λ=µ cin + (En − Ei ) ∂λ ⎜ ⎝ i =0 i =0 ∞ The “meat” of our treatment is based on the equating the coefficients of µ1 = µ in Eq. Enµ) = Emµ) . In the first case. …. m ≤ k + g .26) The easiest way to solve Eq. We can now substitute our results back into our original expressions.n .30) then Eq. ( ( n En0) ≈ Enµ) − µ ∂Eλ ∂ (λ) λ =µ (VII. with k + 1 ≤ i. cmn) = 0 .29) This equation is solved by finding the eigenvalues and eigenvectors of the matrix whose elements are ˆ Ψ (µ) V Ψ (µ) .27) simplifies to ∞ ( ˆ ∑ Ψ(mµ) V Ψ(iµ) cinµ) ∂cmn) (λ i =0 (VII. Using Eqs.A. m i k + 1 ≤ i. We obtain m ∞ ∞ (λ ) ⎛ (λ) ⎞ ( ( ˆ ⎟ ⎟ ∑ cinµ) Ψ(mµ) V Ψ(iµ) = −∑ ⎜ ∂Eλn λ=µ cinµ) + (En(µ) − Ei(µ) ) ∂∂cin λ=µ ⎠ Ψ(mµ) Ψ(iµ) . m ≤ k + g (VII.18) and (VII.n ⎤⎥⎦ is an eigenvector of + + + ˆ V ≡ Vmi = Ψ (µ) V Ψ (µ) .26) is to multiply both sides by Ψ (µ) and integrate. (VII.27) arises when Emµ) ≠ Enµ) . (VII. Show that Eq. λ ⎟ ⎜ ⎝ λ =µ λ =µ ⎠ i =0 ⎛ n(λ) ⎞ c(λ) (µ ( ( ⎟ cmn) + (Enµ) − Emµ) ) ∂∂mn = − ⎜ ∂Eλ ⎟ ∂ λ ⎜ ⎟ ⎝ λ =µ λ =µ ⎠ ( ( There are two cases in Eq. m i C. ck(µ)g . and we get ∞ (λ) ˆ (VII. When you equate the coefficients of µ in Eq. (VII.25) actually agrees with the conclusions in the preceding two paragraphs for both non-degenerate and degenerate states. (VII.29) is satisfied. ( ( (µ The second case in Eq.20).31) = ( ( ∂λ λ = µ Emµ) − Enµ) (µ where cin ) = 1 if Ei ≠ En . (VII. (VII. ⎞ (µ ) ⎟ Ψi .27) = −∑ ⎜ ∂Eλ ⎟ δmi . (VII.

forming cations and electrons: NH3 Na ⎯⎯⎯ Na + + e − → (VIII.) We can use the variational principle to estimate the ground state wave function and ground state energy of an electron confined by this potential (cf. (VIII. Show that Eqs.4) 19 .) A. Suppose H (µ) has no degenerate states. m Ψ ( 0) n = Ψn + ∑ (µ ) n =0 n ≠m ˆ Ψ (µ) µV Ψ (µ) m n En − Em (µ ) (µ ) (VII. (Hint: Use spherical coordinates and separation of variables. α ) = N (α)e −αr .29). where the blue and bronze phase have no perceptible difference in density). Precipitation of the “bronze” solution gives alkalide salts. which are believed to be about 3-3.1) The electrons occupy cavities in the ammonia (the solution is of much lower density than usual alkali salt solutions). ˆ D. and (VII. To model this state. (VII. adding sodium metal to ammonia causes the sodium to ionize (the sodium cation is stabilized by ammonia molecules with the lone pair on nitrogen “pointing” at the cation). floating above the “blue phase” until the entire solution is bronze. Electrons in Ammonia (2. In the rare cases where crystals have been obtained from the blue solution.31). (VIII.0) When a small amount of an alkali metal is dissolved in liquid ammonia. 2 ψ0 (r . that is. ( ( ˆ En0) = Enµ) + Ψ (µ) µV Ψ (µ) n n ∞ (VII.25). (VII. Write the Hamiltonian for an electron confined by this 3-dimensional potential.4 Å in diameter. That is.2) where A > 0 approximates the depth of the energy well. problem IID). a bronze-colored phase separates out (except for Cesium. We choose. as an approximate wave function for the ground state. there is no anion present in the crystal structure. with special emphasis on sodium metal. they will be very important in this course!! VIII.34) Ψ (µ ) .3) This wave function must be normalized.where the coefficients are determined using Eqs. (It should be noted that this is a model potential (designed so you can do the integrals easily!)—don’t take the results too literally! The model does reproduce some qualitative features: the well is somewhat “flat” on the bottom and the potential becomes positive as you near the “boundary molecules” confining the electron. The blue color is because a very broad absorption maximum (centered at 1500 nm ) spreads into the red portion of the spectrum. Then.35) It is important to understand these results. a blue solution results. (VII. 1= ∫ (N (α)e ) −αr 2 0 ∞ 2 4πr 2dr .32) and (VII. The blue solution is believed to result from free electrons dissolved in the ammonia solution. so that blue light is absorbed less weakly (reflected more strongly) than red light. we can use a potential of the form V (r ) ≡ A (3 − r 2 )e − r3 2 (VIII. In the following we will explore this system. as the solution becomes less dilute.33) reduce to the usual formulae from first-order perturbation theory.

α ) H ψ0 (r. F. find N (α) . Show that the wave function ψ1 (r . Find the value of α for which Eq. What is the experimentally observed energy of the transition. J. (VIII. Using the result from part G.2) causes the theoretical excitation energy. β )dr . β ) H ψ (r. (VIII. L. what is the approximate ground state energy? What is the approximate ground state wave function? The observed absorbance can be attributed to an analogue of the 1s → 2p transition in Hydrogen.4). 0 1 The “bronze” phase is generally attributed to the “disproportionation reaction”. Using Eq. (VIII. . you should “look ahead” to the last few pieces of this problem. α ) and ψ0 (r . we can approximate the first excitation energy (cf.5) C. According to the variational principle (problem IID). 1 1 0 ∞ (VIII. where h is Planck’s constant and ν = where c is the speed of light and. E ↑ ? 0 1 What value of A in Eq. M (β ) . This is not true in the gas phase. (VIII. in this case. since this is the lowest-lying non-forbidden transition for this system. (VIII. To do this. 2 (VIII. Find the normalization constant. β ) . Find the value of β for which Eq. calculate the expectation value of the Hamiltonian from part A.7) H. to equal the experimental one. but here this is not true: the disproportionation reaction in Eq. Using the result from part C. (VIII. E1 − E 0 . but 20 .7) takes its minimum value. β ) = M (β )(r cos θ )e −βr is orthogonal to ψ0 (r. we should choose the parameter. I. calculate the expectation value of the Hamiltonian E 1 (β ) ≡ ˆ ∫ ψ (r. D.5) takes its minimum value. α ) are orthogonal. Because ψ1 (r . For relevant integrals. → (VIII. problem IIE) by minimizing the energy with respect to β .6) G. NH Na ⎯⎯ 3⎯ Na + + Na − . ∞ E 0 (α) ≡ ∫ψ 0 0 ˆ (r . first ionization potential of an atom or molecule was less than its second ionization potential.8) can only occur if the first ionization potential of the sodium anion (the electron affinity of the sodium atom) is less than the second ionization potential of the anion (the ionization potential of sodium). what is the approximate excited state energy? What is the approximate excited state wave function? Is β smaller or larger than α ? What does this say about “how delocalized” the electron is in its ground and first excited states? c λ The energy of the transition can be considered to be hν . K.8) You’ll recall that the first day of class. That is. so that the energy is minimized.B. λ ≈ 1500 nm . E ↑ . I said that “in general”. E. α )dr . α .

9) Suppose the new absorption maximum is at 1700 nm . so that light ionizes the anion and shifts an electron to the cation. What is the value of Zeff ? Is this reasonable? As should be evident from the preceding. Note that 0 ∞ dJ dα = ∫ 0 −r 2e −αr dr 2 (VIII. Such phenomena are beyond the scope of this course. the possibility that we are observing an excited state of the anion. M. we model the anion as 2 electrons moving in some effective nuclear 11 . (VIII. One is that this is a “charge transfer” band. Since the first 10 electrons in Sodium occupy a closed shell. we may deduce that the absorbed wavelength for two electrons moving in the effective nuclear charge. The 1s → 2p transition in Helium has an energy of 584.10) ∞ (I (α)) = = 2 −∞ ∞ ∞ ∫ e dx ∫ e −αy dy −∞ 2 −∞ −∞ ∫ ∫e 2π ∞ −α(x 2 +y 2 ) (VIII. The bronze color of this solution may be taken as an indication that yellow light is no longer strongly absorbed. and so we might expect that the absorption maximum in the bronze solution has been shifted to higher wavelength (lower energy).occurs in solution because a polar solvent (ammonia) is much better at stabilizing ions (especially the cation) than neutral atoms/molecules.13) (VIII.3 Å .7 nm 2 Zeff . The following techniques are very useful in this regard. There are two possible models.. we often need to evaluate integrals of Gaussian functions.43 nm ⋅ ( Zeff ) = Z 2 223. let J = I 1 2 (α ) = −αr ∫ e dr .11) dxdy We then evaluate the integral using spherical coordinates.15) 21 . (VIII. will be approximately given by He λ = 58.14) ∞ d 2J d α2 = ∫ 0 r 4e −αr dr 2 etc. ∞ ∫r 0 2n −αr 2 e dr ≡ (2n −1)(2n −3) (1) 2(2 α)n π α . namely. Because atomic energy charge. We can consider. however. So ∞ −∞ ∫e −αx 2 −αx 2 dx (VIII. (I (α )) π α ∞ 2 = ∫ 0 2 −αr ∫ e rdrd θ 0 2 (VIII.12) N. To evaluate ∞ I (α) = we first square the integral. Zeff . Zeff level spacings change in approximate proportion to Z 2 . The result is the formula. Show that I (α) = (α ) Now.

(VIII.17) fund. since it can be done using the fundamental theorem of calculus. On the one hand.The first term in the numerator is sometimes called a “double” or “two-step” factorial. one has K (α) = = = ∫ re 0 ∞ −1 2α −αr 2 dr 2 ∫ 0 (−2αr )e −αr dr d dr −αr 2 −1 2α ∫ (e 0 ∞ (VIII. −αr 2 2 ⎤ − e −α(0) ⎥ ⎥⎦ 1 → )dr ⎯⎯⎯⎯⎯⎯ −α ⎡⎢⎢⎣ lim e 2 r →∞ = 21α .15) for the n = 1 case. By evaluating dK (α) dα . P. differentiating K (α) with respect to α gives higher order terms in the series. Using Eq. you won’t get stuck. For n = 0 .13).) On the other hand.16) for n = 1 . I have no problem if you jot these integrals down somewhere so you never have to do them again. O. of calc. ∞ The result for odd powers of r is ∫ 0 ∞ r 2n +1e −αr dr = 2 n! 2 αn +1 . confirm Eq. Now. if you ever get stranded on a deserted island and decide to while away the hours scribbling solutions to theoretical chemistry problems in the sand. confirm the formula (VIII. 22 . that once you know how to do them. they aren’t that hard. (VIII. I want you to know how to do them. though. (VIII.16) The validity of this formula is even easier to establish. and is often denoted (2n − 1) !! . th. (I think you’ll agree. That way.