Where Does the Slit Interference Pattern Come From?

— Following the derivation in lecture notes (9), (10) and (11) (for second- and third-year undergraduates)

The simulator draws the distribution of landing points on the screen with the formula |ψ(x)|2=4A02a2(sin⁡αxαx)2 (the double-slit tab draws the formula in Eqs. (9) and (10) of §3). Below we derive where this formula comes from, following the flow of the notes. The only real calculation is a single integral of an exponential function.

The conclusion first If you add up the phases of the waves emitted from each point of the slit, the answer takes the form sin⁡αx/αx — and the key to everything is that the quantity α=pa/ℏL that appears in it is proportional to the mass. As the mass increases, the pattern is squeezed sideways, and in the end it can no longer be told apart from “particles flying straight.”

0. Setup

As in the figure in lecture notes (9), a slit of width 2a is struck perpendicularly by an electron beam of momentum p, and we find the distribution on plane A (the x axis), a distance L away. Let y be the position on the slit and x the position on the screen.

SymbolMeaningValue used in the notes
LDistance from the slit to the screen1.0 m
aHalf-width of the slit (the width of a single slit is 2a). For the double slit (§3) it is the half-width of the central barrier, and the openings are a≤|y|≤a+b1.0×10−3 m
bWidth of each slit of the double slit1.0×10−3 m
vSpeed of the particle (p=mv)1.0 m/s

The relation between the wave number and the momentum is, as written in the margin of the notes, p=hλ=h2π⋅2πλ=ℏk so that k=p/ℏ.

1. Intuition First (the 30-Second Version)

You cannot say which point of the slit the particle passed through. Since that is undetermined, we add up the contributions of all the paths that “pass through the point y and arrive at the point x on the screen,” as complex arrows (probability amplitudes). The direction of each arrow is determined by the phase kr, which is proportional to the path length r.

The following puts this “line up or spin around” into a formula. What determines how fast the arrows spin is α=pa/ℏL — and the larger the momentum, the faster they spin, so the pattern becomes finer and narrower.

2. The Single Slit — Lecture Notes (9)

2.1 Adding up the probability amplitudes

We add up the contribution from each point y of the slit, attaching the phase eikr to each. Since y is continuous, the sum becomes an integral.

(1) ψ(x)= A0 ∫−aa eikrdy , r=L2+(x−y)2

2.2 Expanding the distance (the Fraunhofer approximation)

r cannot be integrated with the square root as it is. Assuming |x−y|≪L and using (1+ε)1/2≃1+ε/2 we obtain

(2) r=L1+(x−yL)2 ≃ L+x22L −xyL +y22L⏟dropped

Dropping the last term corresponds to the part of the notes that says “y is a small value, and y2/L is smaller still, so it is almost 0.” The condition under which it may be dropped is not that the length is small but that the phase is small, namely that pa2/2ℏL≪1 holds. For the setting in the notes, this quantity is ≃4.3×10−3 so there is no problem.

Of the three remaining terms, L and x2/2L do not contain y, so they can be taken outside the integral.

(3) ψ(x)= A0 eipℏ(L+x22L)⏟absolute value 1 ∫−aa e−ipxℏLy dy

The first factor has absolute value 1, so it does not affect |ψ|2 at all. Only the remaining integral determines the pattern.

2.3 Carrying out the integral

With u≡px/ℏL defined, the integral is nothing but the integral of an exponential function.

(4) ∫−aa e−iuydy = [e−iuy−iu]y=−ay=a =2sin⁡uau =2asin⁡αxαx
Here α appears α≡paℏL =2πmvahL All we did was the substitution ua=αx, but this α is the only parameter. Every length that appears from here on is a constant multiple of 1/α.

2.4 The probability distribution, and a caution about normalization

From Eqs. (3) and (4), the probability density is

(5) |ψ(x)|2 =4a2A02 (sin⁡αxαx)2
The notes write the preceding coefficient together as A02a2 — but since A0 is a constant fixed at the end by normalization, the form is the same; the only difference is where the factor 4 is included.

For normalization we use ∫−∞∞(sin⁡t/t)2dt=π (this is the formula written at the left edge of the notes). Setting t=αx, we obtain

(6) ∫−∞∞ (sin⁡αxαx)2 dx=πα ⟹ |ψ|2= απ (sin⁡αxαx)2
⚠ What “normalizing gives 1/π” means The notes state that the normalization constant is 1/π — which seems to conflict with α/π in Eq. (6). This is a difference in the variable with respect to which the distribution is normalized to 1. The horizontal axis of the graph in the notes is αx (dimensionless), and measured in that variable it is indeed 1/π. If it is measured in x [m], a factor of α appears from the change of variables. The value of a probability density is undefined until you say “per what.” — Exactly the same pitfall exists in Planck’s formula, where it appears as the fact that “the peak seen in λ” and “the peak seen in ν” are different things.

The first dark fringe (the first solution of sin⁡αx=0 with x≠0) is

(7) x1=πα =πℏLpa =λL2a

This has exactly the same form as optical diffraction: “wavelength × distance ÷ slit width.” For the setting in the notes (an electron), α=8.638 m−1, x1=36.4 cm. The full width of the central bright region is a little over 70 cm, a spread that is clearly visible to the eye.

3. The Double Slit — Lecture Notes (11)

3.1 Why it becomes a “subtraction”

In lecture notes (11) there are two integration ranges, −a−b≤y≤−a and a≤y≤a+b. If you look at this as subtracting rather than adding, the calculation finishes in one line.

(8) ∫−a−b−a + ∫aa+b = ∫−(a+b)a+b⏟one slit of half-width a+b − ∫−aa⏟central barrier

In other words, a double slit is one large slit minus the part blocked in the middle. Both integrals have already been computed in Eq. (4), so the answer is the difference of two single-slit expressions. With β≡p(a+b)/ℏL defined, we obtain

(9) |ψ(x)|2 =4a2A02 ( sin⁡βx−sin⁡αx αx )2

This is the last equation of lecture notes (11), and it is what the double-slit tab of the simulator draws. Once you see that the “difference” is really the subtraction of the blocked part, there is no need to memorize the sign.

3.2 Splitting into fringes and an envelope

The sum-to-product formula sin⁡B−sin⁡A=2cos⁡A+B2sin⁡B−A2 splits Eq. (9) into a product of two factors. Letting the center-to-center distance be d=2a+b, we have

(10) |ψ|2∝ cos2⁡(pdx2ℏL)⏟fine fringes (set by the separation d) ⋅ (sin⁡(pbx/2ℏL)pbx/2ℏL)2⏟envelope (set by the width b)

The fringe spacing and the first dark fringe of the envelope are, respectively,

(11) fringe spacing=λLd , envelope dark fringe=λLb

The simulator readouts (242.5 mm and 727.4 mm) are these two quantities. The fineness of the fringes is set by the slit separation, and the spread of the whole pattern by the slit width — the two can be read separately.

4. What Happens When the Mass Changes? — Lecture Notes (10)

4.1 α is proportional to the mass

From α=2πmva/hL we see that, with the speed v (and the half-width a of the slit and the distance L) held fixed, α is proportional to the mass, and the width of the pattern x1=π/α is inversely proportional to the mass. Placing the two examples from lecture notes (10) side by side gives:

Particlem [kg] α=pa/ℏL [m−1] x1Appearance
Electron9.109×10−318.638 36.4 cmSpreads widely
Heavy particle (notes)6.626×10−32π×1028 5×10−29 mA single point (about 5×10−14 times the size of a proton)

This agrees with the notes, which state “for the heavy particle, α=2π×1028.” x1=5×10−29 m is more than 13 orders of magnitude smaller even than the radius of a proton. Whether it can be measured is beside the point: it has no meaning.

4.2 Convergence to the δ function

The notes say, “Only at x=0 does this equal 1/π and elsewhere it is (almost) 0 (a δ function).” Stated precisely, it goes like this. The normalized distribution of Eq. (6) is

(12) limα→∞ απ (sin⁡αxαx)2 =δ(x)

The area stays at 1 regardless of α, while only the width collapses to 0 as 1/α — this is exactly the condition for a sequence to converge to the δ function (a δ-sequence) (this form is the continuous version of the Fejér kernel). Quantum mechanics turns into classical mechanics not because the formula changes, but because this limit takes effect.

Computing the probability of landing within a window of ±1 µm shows how fast the convergence is.

m [kg]x1 [m] P(|x|<1 µm)
10−303.3×10−10.000006
10−273.3×10−40.006
10−243.3×10−70.966
10−213.3×10−100.99997
10−183.3×10−131.000000

Around a mass of 10−24 kg (roughly one C60 fullerene molecule), it already begins to look as if the particles “hit a single point.” Try it out in the figure below.

Left: the normalized |ψ|2. The horizontal axis is reset to ±3x1 each time — the shape does not change at all; only the scale does. Right: how that scale x1=π/α changes with the mass (log–log).

5. What Does the Simulator Calculate?

What we found above is the distribution. The simulator draws the landing points from it one at a time. It does only three things.

  1. Random sampling (rejection method). Taking the normalized Eq. (6) for the single slit and the normalized Eq. (9) for the double slit as the probability density, the simulator draws x by the rejection method with a Cauchy distribution as the proposal distribution. Because it does not cut off at a window, samples come from the true distribution, tails included. The acceptance rate is 1/π≃32 % for the single slit and 1/2π≃16 % for the double slit.
  2. Matching the scales. To put the histogram and the theoretical curve on the same vertical axis, the simulator uses the analytic value of π/α=λL/2a in Eq. (6) (λL/2b for the double slit). That is why the two do not merely look similar “by eye” but agree quantitatively.
  3. Testing the agreement. If the χ2/ν shown at the bottom of the simulator window is about 1, it means that the histogram is statistically indistinguishable from the theoretical distribution.
About “which slit did it pass through?” In the double-slit tab you can display straight lines after the slits, but these are a visual guide only. As the derivation of Eq. (8) shows, interference appears because the two integrals are added first and the absolute value is taken afterward, and if you fix “which one it passed through” (that is, take the absolute values first and then add), the fringes disappear. In the default display, a particle that has passed the slits is drawn as a wave spreading from both openings, and only its landing point on the screen is recorded as a single point. The straight lines that appear when you check the box are drawn for convenience, simply connecting up to that one recorded point.

6. How Far Do the Approximations Hold?

Two things were dropped in Eq. (2). They matter in different ways, so we look at them separately.

What was droppedCondition for dropping itFor the setting in the notes
y2/2L (Fraunhofer condition) pa2/2ℏL≪1 4.3×10−3 ✓
Terms of second and higher order in the expansion of the square root ((x−y)4/8L3 onward; the paraxial condition) x1/L≪1 0.36 ⚠
⚠ The setting in the notes is actually just outside the paraxial approximation An electron with v=1 m/s has a de Broglie wavelength as long as 727 µm, so x1/L=0.36 — the first dark fringe appears at an angle of as much as 20° when you look up at the screen. Here the accuracy of the expansion in Eq. (2) itself drops. The simulator always displays this value and shows a ⚠ when it exceeds 0.2. This actually makes good teaching material, because it lets you keep an eye on the range of validity of a formula yourself.
Conversely, what about the heavy-particle side? As the mass increases, the paraxial condition x1/L≪1 is satisfied better and better, but the Fraunhofer condition pa2/2ℏL=αa/2 is, on the contrary, increasingly violated. Therefore Eq. (12) in §4.2 does not say that the distribution of a heavy particle literally becomes zero in width. Treated rigorously, what a heavy particle produces is the geometric shadow of a slit of width 2a. Please read Eq. (12) as an idealization of the fact that the diffraction spread x1=λL/2a is negligible compared with the slit width.

7. Checking with Numbers

All the manipulations above were done by hand, so they have been checked independently by numerical computation (integrals by Simpson’s rule, comparing at several chosen values of x).

What was checkedAgreement
The integral in Eq. (4): numerical integration vs. closed formRelative difference < 3×10−14
Eq. (8): sum over the two intervals vs. “wide slit − central barrier”Relative difference < 4×10−14
Eq. (10): decomposition by the sum-to-product formulaRelative difference < 1×10−15
Eq. (7): π/α=λL/2aRelative difference 2×10−16
Eq. (11) and the simulator readouts242.5 / 727.4 mm agree
The notes’ values pa/ℏL=8.638, p(a+b)/ℏL=17.28 8.6380 / 17.2760

8. Summary

  1. Add up the contributions from every point of the slit together with their phases. That is the only starting point, Eq. (1).
  2. Expanding the distance up to L+x2/2L−xy/L gives an exponent in which y appears only to first order, so the integral can be carried out (Eqs. (3) and (4)).
  3. The answer is sin⁡αx/αx. The only parameter is α=pa/ℏL.
  4. The double slit is “a wide slit minus the central barrier.” Hence the difference of two sines, Eqs. (8) and (9).
  5. Splitting with the sum-to-product formula, we can read off that the fringes are determined by the separation d and the envelope by the width b (Eqs. (10) and (11)).
  6. Since α∝m, raising the mass makes the distribution converge to a δ function (Eq. (12)). Classical mechanics emerges as a limit of quantum mechanics.

Mochizuki Research Group, Tokyo University of Science / from lecture notes (9) “Electrons passing through a slit,” (10) “Exploring the wave nature of particles,” and (11) “Observing de Broglie waves, Part 2 — The double-slit experiment” (in Japanese). The numerical checks use Simpson’s rule and the series and asymptotic expansions of the sine integral Si(x). The equations are typeset in the browser’s native MathML (no external libraries or network communication are used at all).