Operator Formalism I 6 6.1 Momentum operator In eqs. 5/(47-48) found Z Z 2 dx|ψ(x)| = dp|f (p)|2 (1) suggesting that if |ψ(x)|2 is the probability density for finding the particle in spatial range [x, x + dx], then |f (p)|2 is prob. density for finding particle’s momentum in range [p, p + dp]. Plane wave ψp = eipx/h̄ is state of definite momentum. Define momentum operator ∂ (2) p̂ = −ih̄ , ∂x verify p̂ψp = pψp, note difference between operator p̂ and eigenvalue p. Compare with Hψ = Eψ for state of definite energy. 6.2 Expectation values or, how the probabilistic nature of ψ affects physically measurable quantities. Consider statistical ensemble of systems with 1 particle described by ψ. Prob. particle in d3r at r is dP = |ψ(r, t)|2d3r. 1 Calculating averages Calculate avg. value by summing over all possible values of something, weighted with prob. of each. For example, mean value of particle’s position: Z < r >= ψ ∗rψ d3r (3) if ψ normalized. Called expectation value of r because it is at < r > you would expect to find particle, on avg. Expectation val. of any operator  is just Z <  >= ψ ∗Â(r)ψ d3r (4) Avg. value of momentum (need to use momentum prob. dens. |f (p)|2: <p> = = Z Z d3p f (p)∗pf (p) (5) d3pd3r1d3r2 ∗ −ip·(r2 −r1 )/h̄ ψ (r )ψ(r )pe (6) 1 2 (2πh̄)3 Simplify by writing pe−ip·(r2−r1)/h̄ = +ih̄∇2e−ip·(r2−r1)/h̄ (7) where ∇2 grad wrt r2. Integrate by parts in 3D: note for 2 functions g(r), h(r) which fall off suff. rapidly at |r| → ∞, Z Z 3 d rg∇h = − d3r(∇g)h (8) since the surface term gh|∞ vanishes. So rewrite (6) as d3pd3r1d3r2 ∗ −ip·(r2 −r1 )/h̄ ψ (r )(−ih̄∇ )ψ(r )e (9) < p >= 1 2 2 (2πh̄)3 Z 2 Now perform int. over p, gives (2πh̄)3δ(r2 − r1), =⇒ Z < p > = d3rψ ∗(r)p̂ψ(r), p̂ ≡ −ih̄∇ (10) (11) looks just like eqn. (3) for position operator. Have found a way to represent physical observable momentum in posiR tion representation. Can show further < p2 >= d3rψ ∗p̂2ψ, R in general < f (p) >= d3rψ ∗f (p̂)ψ, provided f (p) can be expanded in a power series. (check!). 6.3 Inner products Generalize representation of momentum, position, energy as operators on function ψ to any physical observable. Formally, operator Ô maps ψ onto some new function ψ 0, and is linear: Ô(αψ + βψ 0) = αÔψ + β Ôψ 0 (12) 3 ψ and ψ 0 are elements of a general linear function space (like a vector space). We define a norm and scalar product on this space just as for ordinary vectors. Definition. Inner product of two functions ψ, φ: Z (ψ, φ) ≡ d3rψ ∗(r)φ(r) (13) has following properties: (ψ, αφ + βχ) = α(ψ, φ) + β(ψ, χ) (αψ, φ) = α∗(ψ, φ) (ψ, φ)∗ = (φ, ψ) (14) (15) (16) Notation allows us to collect previous results in succinct way: normalization condition 1 expectation values: < r > <p> <H> 6.4 = = = = .. (ψ, ψ) (ψ, rψ) (ψ, p̂ψ) (ψ, Hψ) (17) (18) (19) (20) Hermitian adjoint Q: under what conditions can we expect general linear operator Ô to represent a physical observable? First must demand expectation value (ψ, Ôψ) must be real. Definition. Adjoint Ô† of operator Ô obeys (Ô†ψ, φ) = (ψ, Ôφ) (21) 4 Examples: 1. α a const. =⇒ α† = α∗ 2. Ô = ∂ ∂x : ∂ψ ∗ 3 ψ dr = − φd r ∂x ∂x ∂ ∂ so ( )† = − ∂x ∂x Z ∗ ∂φ 3 Z (22) (23) 3. (αÔ)† = α∗Ô† † ∂ ∂ ∂ then p̂† = ih̄ ∂x = −ih̄ ∂x = p̂. p̂ is said so if p̂ = −ih̄ ∂x to be self-adjoint or Hermitian. 4. (P̂ + Ô)† = P̂ † + Ô† 5. (P̂ Ô)† = Ô†P̂ † 6. If (ψ, Ôψ) = (Ôψ, ψ)∀ψ, then (ψ, Ôφ) = (Ôψ, φ) so Ô is self-adjoint. Check this! 7. From (16) get (ψ, Ôφ)∗ = (Ôφ, ψ) = (φ, Ô†ψ) (24) (25) (ψ, Ôψ)∗ = (ψ, Ô†ψ) (26) so in particular so that if Ô is self-adjoint, Ô† = Ô, then (ψ, Ôψ)real . All observables wil be represented by self-adjoint operators. 5 6.5 Eigenvalues and eigenfunctions Let Q̂ be a linear operator, and ψ an eigenfunction of Q̂ with eigenvalue q, i.e. Q̂ψ = qψ. If Q̂ is Hermitian, Q̂ = Q̂† then several properties follow: 1. All eigenvalues q are real! Start with (Q̂ψ, ψ) = (ψ, Q̂ψ) Q̂ψ = qψ, (27) (28) (qψ, ψ) = (ψ, qψ) q ∗(ψ, ψ) = (ψ, ψ)q =⇒ (29) (30) then q∗ = q (31) 2. Eigenfunctions belonging to different eigenvalues are orthogonal. If Q̂ψ1 Q̂ψ2 then (Q̂ψ1, ψ2) q1(ψ1, ψ2) = q1ψ1 = q2ψ2, (32) (33) = (ψ1, Q̂ψ) = q2(ψ1, ψ2), (34) (35) so if q1 6= q2, then (ψ1, ψ2) = 0 and the two fctns said to be orthogonal. 6 3. Linearly independent eigenfunctions belonging to the same eigenvalue q can be made orthogonal. (Example: recall we studied eigenvalues of 3D SHO in problem set, showed degeneracy—number of eigenfunctions corresponding to given eigenvalue– was N 2/2+3N/2+ 1. ) Functions ψ1...ψm are linearly ind. if only soln. to Pm n=1 ψn = 0 is cn = 0. Suppose Q̂ψn = qψn, n = 1, . . . m. (36) It might be that (ψm, ψn) 6= δmn for this choice of the ψn, but can always find new set of φn obeying Q̂φn = qφn for which (φm, φn) = δmn using Gram-Schmidt orthonormalization (as in linear algebra!): Let φ1 φ2 with α chosen s.t. (φ1, φ2) let φ3 and require (φ1, φ3) = ψ1 = ψ2 + αψ1 (37) (38) = (ψ1, ψ2) + α(ψ1, ψ1) = 0 (39) = ψ3 + βφ2 + γψ1 (40) = (φ2, φ3) = 0. (41) Last line is two equations determining 2 unknowns β and γ. And so on. . . 4. If P̂ and Q̂ commute, [P̂ , Q̂] = 0, can find set of simultaneous eigenfunctions of P̂ and Q̂. 7 Simple case: suppose every eigenfctn. of P̂ corresponds to different eigenvalue (nondegenerate). For all such eigenfctns., P̂ ψ = pψ so Q̂(P̂ ψ) = Q̂(pψ) if [P̂ , Q̂] = 0 ↓ ↓ P̂ (Q̂ψ) = p(Q̂ψ). (42) (43) (44) So Q̂ψ is eigenfctn. of P̂ with eigenvalue p. But there is only one such lin. ind. fctn ψ by assumption, so must have Q̂ψ ∝ ψ, means ψ also eigenfctn. of Q̂! If 2 or more eigenfctns {ψ}n correspond to same p, more complicated construction necessary to find set {ψn} which are simultaneous eigenfctns. of P̂ and Q̂. Not terribly enlightening (see Peebles for details) but be aware it can always be done. Point: if operators P̂ and Q̂ commute, observable quantities p and q can be determined with arbitrary precision simultaneously! Not true in general, e.g. [p̂i, xj ] = −ih̄. Example: SHO wavefunctions Use wave functions we’re somewhat familiar with, 1D SHO energy eigenfctns given by 1 x − 21 xx 2 √ 1/2 Hn( )e 0 ψn(x) = n (45) (2 n!x0 π) x0 s where x0 = h̄ mω is the length scale we have encountered 8 before, and the Hn are specific polynomials which arise in the solution of the S.-eqn. for higher n. H0 = 1 and H1 = 2x. 1. Check: energy eigenfunctions belonging to different eigenvalues are orthogonal. (ψ1, ψ0) = 0 but (ψ0, ψ0) = 1, etc. (46) 2. Check: are these momentum eigenfunctions too? ∂ − 12 xx 2 (47) p̂ψ0 = −ih̄ ψ0 ∼ xe 0 ∂x so no, it’s not simultaneously an eigenfunction of momentum. We don’t expect this, since p̂2 1 [H, p̂] = [ + mω 2x2, p̂] = mω 2[x2, p̂] 6= 0 2m 2 (48) 3. Linearly independent eigenfunctions belonging to the same eigenvalue q can be made orthogonal. The 1D SHO wavefunctions aren’t degenerate =⇒ can’t discuss this case. 9 But 3D SHO homework problem showed energies are EN = h̄ω(n1 + n2 + n3 + 3/2), N = n1 + n2 + n3 (49) and eigenfunctions are ψn1,n2,n3 = ψn1 (x)ψn2 (y)ψn3 (z) (50) i.e., one for each possible combination n1, n2, n3. Consider ψ100, ψ010, and ψ001. Each has N = 1, corresponding to an energy E1 = 5h̄ω/2. Physically they correspond to oscillator with one quantum of excitation energy vibrating in x, y, or z directions, respectively. For isotropic potential, energies must be equal (degenerate), so degeneracy is 3 for the eigenvalue E1. They are already orthogonal, e.g. Z d3rψ1∗(x)ψ0(x)ψ0∗(y)ψ1(y)|ψ0(z)|2 (51) = (ψ1, ψ0)1D (ψ0, ψ1)1D (ψ0, ψ0)1D = 0 (ψ100, ψ010) = because the 1st two factors vanish. 10