Holographic Realization of the Prime Number Quantum Potential

Donatella Cassettari,1 Giuseppe Mussardo,2 and Andrea Trombettoni3, 2

1
SUPA School of Physics & Astronomy, University of St. Andrews,
North Haugh, St. Andrews KY16 9SS, UK
2
SISSA and INFN, Sezione di Trieste, Via Bonomea 265, I-34136 Trieste, Italy
3
Department of Physics, University of Trieste, Strada Costiera 11, I-34151 Trieste, Italy
We report the first experimental realization of the prime number quantum potential VN (x), defined
as the potential entering the single-particle Schrödinger Hamiltonian with eigenvalues given by
the first N prime numbers. We use holographic optical traps and, in particular, a spatial light
modulator to tailor the potential to the desired shape. As a further application, we also implement
arXiv:2202.03446v1 [quant-ph] 7 Feb 2022

a potential with lucky numbers, a sequence of integers generated by a different sieve than the
familiar Eratosthenes’s sieve used for the primes. Our results pave the way towards the realization of
quantum potentials with arbitrary sequences of integers as energy levels and show, in perspective, the
possibility to set up quantum systems for arithmetic manipulations or mathematical tests involving
prime numbers.

I. INTRODUCTION

Mathematics is a golden mine of surprises, starting from its very basic branch: arithmetics. Consider
as major examples the sets of the natural numbers

N = {1, 2, 3, 4, 5, . . .} (1)

and of the prime numbers

P = {2, 3, 5, 7, 11, . . .}. (2)

While the pattern of natural numbers is obvious, since no matter which one you pick, it is straightforward
to determine what the next one is, the answer is instead highly non-trivial for the set of prime numbers,
whose intriguing and sometimes apparently erratic properties never ceased to intrigue mathematicians,
physicists, scientists, and curious people in general [1–9].
The sequences of integers and prime numbers are mathematical objects: one could say that they are
the mathematical objects par excellence, being at the roots of arithmetics and therefore at the roots of
entire mathematics. However, a useful point of view - particularly relevant for computational purposes –
is to see them as quantities which emerge from physical operations performed in the physical world. The
rationale is to have a physical system, an abacus, on which one performs the desired operations acting
on the elements of certain desired sequences of integers. Since ultimately all aspects of the world around
us can be explained using quantum mechanics, we would like to have a quantum abacus and it is natural
to employ a Hamiltonian whose eigenvalues are the elements of the desired quantum abacus.
Adopting this point of view, discrete sequences of numbers may be seen as spectra of some quantum
Hamiltonians. In this paper we focus on the one-dimensional quantum Hamiltonian of a single particle
of mass m, expressed in the standard form as
2
b = p + V (x) ,
H (3)
2m
where p is the momentum operator and V (x) is taken to be a continuous function in a given interval
J (which can be the entire real axis). If a potential VN (x) is such that the eigenvalues En of the
time-independent Schrödinger equation

~2 d2
 
Hψn = −
b + VN (x) ψn = En ψn (4)
2m dx2

are the first N prime numbers, we call VN (x) a N prime number quantum potential. In saying that the
eigenvalues of (4) are the prime numbers, or any other sequence of integers, we are actually referring to
the eigenvalues en of Hb defined in dimensionless units: in physical units, the eigenvalues En are equal
to their dimensionless counterpart en multiplied by a constant having the dimension of an energy, which
 2

depends on ~, m, and a length characteristic of the potential V (x) itself. To fix the notation, hereafter
the ground state in (4) corresponds to n = 0 (and therefore, in the case of the primes, e0 = 2). For
the sake of precision, let’s point out that, in addition to the discrete part of the spectrum featuring the
desired first N primes, the potential VN will also have a continuous spectrum for energies larger than the
highest considered prime pN .
The experimental realization of VN (x) will provide the first ingredient for the implementation of a
quantum abacus, in which arithmetic operations can be translated into physical operations on the quan-
tum particle. In this regard, light sculpting techniques provide versatile tools to dynamically control and
engineer optical potentials of any desired form. These techniques are based on devices such as spatial light
modulators (SLMs), digital micromirror devices (DMDs), and scanning acousto-optic deflectors (AODs)
[10]. In particular SLMs underpin computer-generated holography, in which a spatial phase modulation
is applied to the trapping light such that a desired intensity distribution is realized in the far field. Holo-
graphic optical traps provide a flexible tool to tailor the potential experienced by neutral atoms, and have
been employed in experiments ranging from single atoms to Bose-Einstein condensates [11]. Therefore
holographic techniques are a natural tool to implement the prime number quantum potential.
The aim of this paper is to present the first experimental realization of such a potential. As discussed
in more detail in the next sections, the problem presents a number of simple but interesting points which
touch on key theoretical and experimental questions in quantum mechanics and, at the same time, have
fascinating outputs in number theory. The paper is organised as follows: in Section II we recall some
basic features of the prime numbers, pointing out the main challenges one has to face for setting up a
potential V (x) which has them as spectrum. In Section III we provide a brief insight into holographic
techniques. In Section IV we present the theoretical framework of Supersymmetric Quantum Mechanics
[13–15] which leads to the exact expression of VN (x), and we also discuss the pros and cons of this
approach compared to a semi-classical determination of the prime number potential [16]. In Section V we
present the experimental prime number potential suitable for atom trapping, and we assess the feasibility
of a subsequent implementation with ultracold atoms. In Section VI, to demonstrate the flexibility of
computer-generated holography, we discuss the realization of another quantum potential associated to a
different discrete sequence of integers, the so-called lucky numbers, an interesting set of integers which
may be regarded as “cousins” of the prime numbers, generated by a slightly different sieve than the
familiar Eratosthenes’s sieve which gives rise to the prime numbers [23]. Our conclusions can be found
in Section VII.

II. THE INTRIGUING PRIME NUMBERS

A fundamental theorem of arithmetic states that every natural number greater than 1 is either a prime
number or can be represented as product of prime numbers. Hence, the prime numbers may be regarded
as the atoms of arithmetic but, in contrast with the finitely many chemical elements, the number of
primes is instead infinite, as shown by a classic argument by Euclid dated more than 2000 years ago.
Besides this fundamental role in arithmetic, what makes the prime numbers intriguing is their bipolar
personality, i.e. in the realm of mathematics they are the perfect Dr. Jackill and Mr. Hyde. Such a
mental insanity emerges by looking at the short and large distance scales of these numbers. Indeed, at
short scale, their appearance along the sequence of the integers is completely unpredictable but, on a large
scale, their coarse graining properties, and in particular how many prime numbers there are below any
real number x, is an aspect which can be controlled with a great precision. In other words, while there
is no known simple function f (n) which gives the n-th prime number pn (and the actual determination
of prime numbers can only be done by means of the familiar Eratosthenes’s sieve [5–9]), thanks to the
insights of many prominent mathematicians (in particular Riemann), we have instead perfect knowledge
of the inverse function π(x) which counts the number of primes below the real number x [1–9, 17–22].
Such a function has a staircase behaviour (since it jumps by 1 each time that x crosses a prime) but
becomes smoother and smoother for increasing values of x. Its first estimate was empirically obtained
by Gauss and Legendre
x
π(x) ∼ , (5)
ln x
and, even though this formula may be considered just a coarse approximation of π(x), it is nevertheless
able to capture the asymptotic behaviour of π(E) – a result which constitutes the content of the “Prime
 3

Number Theorem” [17–20]

π(x) ln x
lim = 1 . (6)
x→∞ x
R x dt
A more precise version of this estimate is given by π(x) ' li(x) ≡ 2 ln t
, while a further refinement was
provided by Riemann [21, 22] in terms of the series
∞
X µ(n)  1/n 
π(x) ' R(x) = li x , (7)
n=0
n

with the Moebius numbers µ(m) defined by


 1 if n = 1
µ(n) = 0 if n is divisible by a square of a prime
(−1)k otherwise


where k is the number of prime divisors of the integer n. It is worth stressing that R(x) is the smooth
function which approximates π(x) more efficiently and it is well known that to reproduce the actual
staircase jumps of π(x) one needs to employs the zeros of the Riemann zeta-function [7, 21, 22].
Knowing π(x) helps us to estimate the growth behaviour of the n-th prime number. Indeed, by setting
pn = π −1 (n) and inverting at the lowest order the function π(x) (for instance using Gauss formula (5)),
we have the following scaling law for the n-th prime number
pn ' n log n . (8)
However, the true unpredictable nature of the primes becomes particularly evident if we focus our atten-
tion on their gaps: for every prime pn , let g(pn ) be the number of composite numbers between pn and
the next prime pn+1 , so that
pn+1 = pn + g(pn ) + 1 . (9)
With this definition, g(pn ) is the size of the gap between pn and pn+1 . Using the scaling law (8), we
expect the average gap g(pn ) between pn and pn+1 to go as g(pn ) ∼ log n, but the interesting question
is: how wide is the range of values of these gaps? There is an extensive literature on this topic, see
for instance [25–29], and hereafter we only underline some basic features which are important for our
subsequent considerations.
The minimum value of g(p) is 1 and is obtained for the twin primes, i.e. the pairs such as (17, 19) or
(29, 31), etc. which differ by 2. Presently it is not known whether or not there are infinitely many twin
primes, although there are strong reason to believe that the number of twin primes is indeed infinite (see,
for instance, the heuristic arguments presented in [6]). On the other hand, it is quite easy to show that
g(pn ) can be arbitrarily large, so that
lim sup g(pn ) = ∞ (10)
n→∞

To prove such a result, consider an arbitrary integer N > 1 and the associated sequence of integers
N ! + 2, N ! + 3, N ! + 4, N ! + 5, ..., N ! + N
These (N − 1) consecutive numbers are all composite and therefore, if p is the largest prime less than
N ! + 2 we have g(p) > N − 1. Since we can send N → ∞, we arrive to the result (10). In summary, the
sequence of primes shows a pattern of the gaps which is not at all regular, for instance one does not have
any clue where the smallest gaps may appear.
These features make evident the irregular behaviour of the primes and lead to the conclusion that
the quantum potential VN (x) that encodes them should be a rather peculiar function. Given that it
has N bound states with energies equal to the prime numbers and strong variations in the energy gap
between consecutive levels, we expect VN (x) to display a rich structure of maxima and minima which
depends on N . Hence the experimental technique to realize VN (x) must be sufficiently flexible in order
to accurately reproduce this structure. As shown in the next Section, experimental techniques such as
computer-generated holography start with sampling VN (x) over a number of points (”pixels”). With
more pixels available, it is possible to increase the complexity of VN (x), hence the number of energy
levels.
 4

FIG. 1: Experimental setup, showing the phase profile imprinted by the SLM and the resulting light intensity
profile in the trapping plane. The zoomed-in image shows the signal region, i.e. the region of the output plane in
which the field is programmed by the conjugate gradient minimization algorithm. In this case the signal region
contains a generic 1D trapping potential. The circular intensity distribution delimitates the boundary with the
region of the output plane where the intensity is left unconstrained. The scale bar is 100 µm.

III. HOLOGRAPHIC TECHNIQUES

Before discussing how to obtain the analytic expression of VN (x), we review the experimental techniques
which permit the optical realization of the prime number potentials. Our optical potentials are suitable
for trapping ultracold atoms in a one-dimensional geometry via the optical dipole force. The optical dipole
potential is proportional to the intensity of the light [30], hence here we shape the intensity profile of an
incoming laser beam using holographic techniques. In computer-generated holography, a liquid-crystal
SLM spatially modulates the phase of the light. The phase pattern on the SLM acts as a generalised
diffraction grating, so that in the far field we have Fraunhofer diffraction and an intensity pattern is
formed, which can be used to implement VN (x). The SLM acts effectively as a computer-generated
hologram, and the light field in the output plane is the Fourier transform of the light field in the SLM
plane. The calculation of the appropriate phase modulation to give the required output field is a well-
known inverse problem which, in general, requires numerical solutions. Here we use a conjugate gradient
minimization technique which efficiently minimizes a specified cost function [31, 32]. The cost function is
defined to reflect the requirements of the chosen light field in the output plane. In addition to specifying
the intensity profile of the field, which gives VN (x), we also constrain the phase of the light in the output
plane. Namely, a uniform phase is programmed across the whole intensity profile. Controlling the phase
this way leads to a well-maintained intensity profile as the light propagates out of the output plane.
Figure 1 is a schematic of the experimental setup. Our SLM (Hamamatsu LCOS-SLM X10468) is
illuminated by laser light with wavelength 1064 nm. The light diffracted by the SLM is focussed on the
output plane by a f = 75 mm achromatic doublet (Thorlabs AC508-075-B), and detected by a CCD
camera (Thorlabs DCU224M). As shown in the figure, in order to accurately reproduce a given target
light profile, we program only a small subset of the output plane (the ”signal region” SR), whereas the
field is left unconstrained in the rest of the plane [35]. We use the following cost function [32]:
 2
X n o
∗ e out
C = 10d 1 − Re τep,q Ep,q  (11)
p,q∈SR

out
where p and q denote the output plane coordinates. Here τp,q is the target electric field, Ep,q is the
output electric field, linked to the SLM electric field via a Fourier transform, and the over-tilde denotes
out
normalization over the signal region. This cost function minimizes the discrepancy between τp,q and Ep,q
in the parameter space of all the different phase distributions that the SLM can generate. The prefactor
10d , where d = 9 for the results shown here, increases the steepness of the cost function to improve
convergence time and accuracy.
In the algorithm, both the SLM plane and the output plane have a size of 2m × 2m pixels, where m × m
is the number of pixels in the SLM array: namely, the SLM plane is “padded” with zeroes to increase its
 5

size from m × m to 2m × 2m. The purpose of this is to fully resolve the output plane [35]. We find that
with 512 × 512 SLM pixels available, we achieve 1D potentials that are up to 100-pixel long in the output
plane. This amounts to about 1/10 of the linear dimension of the 1024×1024 output plane. If we increase
the length of the potential beyond this, we lose light-utilization efficiency, i.e.the light intensity in the
pattern becomes too low. With our holographic method, we can obtain any smoothly varying intensity
profile over this 100-pixel interval. However, given the non-trivial behaviour of the potential described
at the end of the previous section, this 100-pixel maximum interval limits the number of energy levels of
the potential, i.e. the number of primes. Therefore we envisage that to increase the number of primes
contained in the potential beyond what we present in this work, it is necessary to increase the number
of SLM pixels.
It will be possible to realize prime number potentials also with light sculpting techniques which are
alternative to the liquid-crystal SLM we use in this work. For instance, AODs have been used to realize
a wide range of optical potentials [33, 34]. While the AOD potentials realised so far are simpler than the
prime number potentials presented here, it is possible in principle to increase their level of complexity.
Another possibility is the use of DMDs. A DMD is a matrix of individually addressable mirrors which
can be used as an intensity mask that can be directly imaged on the atoms. The projected image from
a DMD is intrinsically binary, due to the individual mirrors being either “on” or “off”, however there
are methods that overcome this limitation and allow the realization of intensity gradients: half-toning
and time-averaging [10]. Half-toning relies on the finite optical resolution of the imaging optical system,
whereby multiple mirrors contribute to each resolution spot in the projected plane. This provides a
number of possible intensity levels in each resolution spot which is given by the number of contributing
mirrors. In [36] half-toning has been used to demonstrate 1D potentials with a high degree of control. If
half-toning is combined with time-averaging, in which a time-averaged potential is achieved with high-
speed modulation of the mirrors, intensity control can be further improved. With this combination of
approaches, it will be possible to use DMDs to realize the prime number potentials presented here. The
DMD dimension required for this is comparable to the dimension of the liquid-crystal SLM we use here,
making the two techniques equivalent.

IV. DISCRETE SEQUENCES AND QUANTUM POTENTIALS

Let’s now address the problem of how to design, in general, a quantum potential V (x) in such a way
that a given sequence of real numbers

E = {e0 , e1 , e2 , e3 , . . .} (12)

coincides with the set of eigenvalues of the Schrödinger equation (4). It is important to distinguish two
cases:
• the sequence E is infinite. In this case, a necessary condition for the existence of a continuous
potential V (x) able to support the sequence E as spectrum is that asymptotically the en ’s satisfy
the bound [16]

en ≤ A n2 , n→∞ (13)

where A is a positive real number.

• the sequence E is, on the contrary, finite, i.e. it only consists of a finite number N of terms,
e0 , e1 , e2 , . . . , eN −1 , plus possibly a continuous part of the spectrum with energies larger than the
maximum value of the en ’s. In this case there is no obstacle to the existence of a potential V (x)
which supports such a spectrum, and the explicit form of this potential can indeed be found using
methods of Supersymmetric Quantum Mechanics [13], as discussed below.
A familiar example of the first case is provided by the infinite set N of all natural numbers, whose
corresponding potential V (x) ∝ x2 gives rise to the well-known Hamiltonian of the harmonic oscillator
[24]. Another example is provided by the sequence E = {1, 4, 9, 16, . . .} of squared integers, which can
be realized as a quantum spectrum in terms of a properly tuned infinite-well potential [24]. It is im-
portant to underline that, besides these known cases and very few others, it is in general not known
a universal procedure for engineering a potential V (x) with exactly all elements of an infinite sequence
E = {e0 , e1 , e2 , e3 , . . .} as eigenvalues. The best one can do in such a case is to identify a semiclassical
 6

potential Vsc (x): it is worth to stress, however, that this quantity is only able to capture the scaling
growth of the eigenvalues rather their actual values since is determined by the formula
ZVsc
~ dE
x(Vsc ) = √ , (14)
2m dE √
E0 Vsc − E
dn
which depends upon the density of states dE/dn rather than the individual energy levels en ’s. Taking,
for instance, the infinite set P of the primes, it is easy to see that these numbers satisfy the bound (13)
(see eq. (8)). The corresponding semiclassical potential has been determined in [16] by substituting in
eq. (14) the density of states coming from (7)
 −1 ∞
dE dπ 1 X µ(m) (1−m)/m
= ' E . (15)
dn dE log E m=1 m

Let’s now consider the second case where the sequence E = {e0 , e1 , e2 , e3 , . . .} is instead finite: here
there exists the general procedure of Supersymmetric Quantum Mechanics (SQM) [13] for engineering a
potential V(x) with the en ’s as its exact spectrum. It is well known that there are many potentials which
share the same spectrum [12] and to identify uniquely one of these potentials, in the following we impose
the additional condition V(x) = V(−x). Using the methods of SQM, one sets up a chain of potential
Vk (x) (k = N, N − 1, . . . 0), as those shown in Figure 2, with the property that the potential Vk−1 (x) has
the same spectrum of the previous one Vk (x) except its ground state. In other words, climbing down in
the label k of these potentials Vk (x), there is a depletion, one by one, of the lowest level of the previous
potential. This chain of potentials is determined by a system of differential equations and, as we shall see
below, this structure is at the root of the exact reconstruction of the potential with a given set of energy
levels. Indeed, it is sufficient to reverse the procedure and adjust, one by one, all the desired eigenvalues!
With a finite set of discrete eigenvalues, the final potential has a finite limit at x → ±∞ and therefore
also has a continuum part of the spectrum: however this is not relevant for our purposes and will not be
discussed further. In more detail, the top-down procedure works as follows [13]:
1. First of all, we subtract from all the eigenvalues en the highest one eN , so that the new set of
numbers {E en }

ek = eN −k − eN ,
E k = 0, 1, . . . , N , (16)

will be considered as the new spectrum. The E ek ’s are of course the (negative) gaps computed from
the highest eigenvalue eN . Notice that, consistently, they are enumerated starting from the top to
the bottom, so E e0 = 0, E e2 the second gap, and so on. A potential V(x) where
e1 is the first gap, E
its only eigenvalue is E0 = 0 is of course V0 (x) = 0. This potential is used as input for the Riccati
e
equation for the super-potential W1 (x)

W10 (x) − W12 (x) + V0 (x) = E

e1 (17)

FIG. 2: Sequence of supersymmetric partner potentials Vk (x), where Vk shares all the spectrum of the previous
one Vk+1 except its ground state energy.
 7

with boundary condition W1 (0) = 0.

2. Once such a function W (x) has been obtained, one can construct another potential V1 (x) as
e1 + 2W 2 (x) − V0 (x) .
V1 (x) = 2E (18)
1

This potential is then substituted into (17) (i.e., V1 (x) → V0 (x)), substituting as well also E
e1 → E
e2 ,
so that one has a differential equation for another super-potential W2 (x)
W20 (x) − W22 (x) + V1 (x) = E
e2 , (19)

3. Proceeding iteratively in this way, one has a recursive sequence of differential equations
Wk0 (x) − Wk2 (x) + Vk (x) = E
ek ,
(20)
ek + 2W 2 (x) − Vk−1 (x)
Vk (x) = 2E k

all of them solved with the boundary condition Wk (0) = 0, which ensures that the final potential
V(x) = VN (x) is an even function. This recursive system is continued until all the gaps have been
taken into account. Hence, solving (in general numerically) the differential equations (20), one
arrives to the Hamiltonian which has exactly the spectrum {en }
d2
H = − + VN (x) + eN , (21)
dx2
These are indeed the theoretical steps which lead to the potential VN (x) with exactly the first N prime
numbers [14, 15], in other words, our ”quantum abacus” with a finite number of beads. From what
discussed above, the pros and cons of the SQM method are fairly evident:
• pros: the method is exact, i.e. given a finite set of numbers {en }, this method provides the exact
potential which has these numbers as the exact eigenvalues. If this set is made of the first N primes,
the potential exactly accommodates these prime numbers in the spectrum.
• cons: the method is not smoothly scalable, in the sense that if we have determined the potential
which has as eigenvalues a set of N values en , adding a new value eN +1 we have to start again from
scratch and determine altogether another potential which will accommodate exactly the (N + 1)
eigenvalues. We will see below some examples of this feature.
In comparison, the semi-classical potential Vsc (x) (e.g. for the primes) has somehow opposite pros and
cons: it is scalable, in the sense that once that it has been implemented, all its eigenvalues are fixed but,
on the other hand, it is not exact, i.e. its eigenvalues are not exactly the prime numbers.
In closing this section, it is important to stress two important mathematical properties of the system
of differential equations (20). The first property is that there is one and only one family of potentials
which provides a close analytic solutions of the system of the differential equations and which recursively
reproduce themselves at each step of the procedure. This is the family of the Pöschl-Teller potentials
given by
1 N (N + 1)
VbN (x) = − , (22)
2 cosh2 x
associated to the exact sequence of gaps
2
en = − n
E , n = 0, 1, 2, . . . N (23)
2
The Pöschl-Teller potentials VbN (x) do not have oscillations and present the typical shape of an inverted
bell, see Figure 3(a). The second property is that any finite sequence E e of gaps other than the one given
by (23) is expected to give rise to a potential with oscillations. This is true even for very straightforward
sequences, as for instance the potential which has exactly the sequence of the first 20 negative integers
as its energy gaps, see Figure 3(b). This feature is also pretty evident in our realizations of the prime
potentials: in Figure 4 we show, for instance, the potentials V10 (x) and V15 (x) which have, as eigenvalues,
exactly the first 10 and 15 prime numbers respectively: these potentials have a number of oscillations
which scales as the number of accommodated eigenvalues. These oscillations are expected to be more
pronounced in correspondence to more irregular sequences of numbers chosen to be energies. This is
certainly the case for the prime numbers, as underlined in Section II.
 8

FIG. 3: (a) Sequence of Pöschl-Teller potential VbN (x) obtained by solving the recursive differential equations (20)
to accommodate the energy gaps shown on the right hand side. (b) Potential V (x) which has exactly the first
(negative) 20 natural numbers as energy gaps.

V. EXPERIMENTAL PRIME NUMBER POTENTIALS

We use the apparatus shown in Figure 1 to realize the prime number potentials V10 (x) and V15 (x). The
experimental light intensity profiles and the corresponding potentials are shown in Figure 4(a,b). The
potentials are meant to be implemented with light that is red-detuned relative to the atomic transition,
so that regions of higher intensity correspond to a lower value of the potential [30]. The potentials are
re-scaled so that they are plotted in dimensionless units alongside their theoretical counterparts. The
conversion between dimensionless eigenvalues en and physical eigenvalues En is given by:
 2
En ~2 l
= (24)
en m L
where l and L are the lengths of the potential in dimensionless and physical units respectively. Assuming
87
Rb atoms, we obtain En /en = h × 0.029 Hz for V10 (x), and En /en = h × 0.026 Hz for V15 (x). The
potential depth can be calculated with the same conversion formula, leading to a depth of kB × 47 pK

(c) Prime numbers 2 3 5 7 11 13 17 19 23 29 31 37 41 43 47

en for V10 (x) 1.58 3.31 5.40 7.33 10.9 13.2 16.9 19.4 23.2 29.3
en for V15 (x) 1.58 3.21 5.00 7.22 11.3 13.2 16.6 19.4 22.9 28.8 31.4 36.9 40.6 43.4 47.1

FIG. 4: (a,b) Experimental prime number potentials V10 (x) and V15 (x). The images show the light intensity
profiles, where the red scale bar is 50 µm, applicable to both images. The plots show the corresponding potentials,
re-scaled to dimensionless units, alongside their theoretical counterparts. The experimental eigenvalues are shown
as horizontal lines on the plots and are also tabulated in (c), alongside the first 15 prime numbers for comparison.
 9

Lucky en for
(b) (L)
numbers V10 (x)
1 0.85
3 3.16
7 7.45
9 9.12
13 12.6
15 15.1
21 20.7
25 24.6
31 30.9
33 33.2

(L)
FIG. 5: (a) Experimental lucky number potential V10 (x), where the red scale bar is 50 µm. The corresponding
eigenvalues are tabulated in (b) alongside the first 15 lucky numbers.

for V10 (x) and of kB × 69 pK for V15 (x). These figures show that we have shallow traps with very low
trap frequencies, and that we would need to work at extremely cold temperatures, beyond what has been
experimentally realized [37, 38].
However trap depths and trap frequencies can be increased by increasing the numerical aperture of the
optical system, i.e. with stronger focusing in conjunction with a larger SLM physical size. Our current
optical system has a resolution spot of 10 µm. This determines the typical distance between peaks in
the potentials, which is about three times the resolution spot. If in future we use a resolution of 1 µm,
e.g. with a quantum gas microscope [39, 40], then the distance between peaks and the overall physical
length of the potentials can be scaled down accordingly. Given that energies scale with the square of the
potential length, it will be possible to achieve trap depths of several nK, therefore improving experimental
feasibility.
The experimental eigenvalues, expressed in dimensionless units, approximate the primes well: namely
they match the primes if rounded to the nearest integer, as shown in Figure 4(c). The differences arise
from the small discrepancies between the theoretical and experimental potentials, visible in Figure 4(a,b),
which in turn are caused by an imperfect SLM response and by aberrations in the optical setup. The
root-mean-square (r.m.s.) fractional discrepancy between the theoretical and the experimental potentials
is 10% for V10 (x) and 7% for V15 (x), giving an r.m.s. fractional discrepancy between the eigenvalues
and the primes of 8% and 6% respectively. In future these errors can be reduced with error-correction
algorithms as shown in [10]. Another source of experimental uncertainty is due to the optical power on
the SLM fluctuating over time, as this will change the eigenvalues. Specifically, a 1% fluctuation in the
optical power leads to a ∼ 0.5% change in the relative position of the eigenvalues. Hence to achieve the
three-digit precision shown in Figure 4(c), it is necessary to stabilise the optical power to 1%, which is
feasible with active stabilisation techniques.

VI. THE LUCKY QUANTUM POTENTIAL

The method presented in this paper can be straightforwardly extended to other sequences of integers.
(L)
As a significant example related to the primes, we present here the potential VN (x) having as eigenvalues
the so-called lucky numbers

L = {1, 3, 7, 9, 13, 15, 21, 35, 31, 33, . . .} . (25)

These numbers, introduced in the 50’s by Gardiner, Lazarus, Metropolis and Ulam [23], are obtained
with a sieve (known as the sieve of Josephus Flavius) different from the sieve of Eratosthenes used for
the primes. Briefly, to obtain the prime numbers one notoriously eliminates from the list of integers
the multiples of 2 (the even numbers), then the multiples of 3, then the multiples of 5, and so on. On
the contrary, for the lucky numbers one eliminates numbers based on their position in the remaining
set, instead of their original value, i.e. their position in the initial set of natural numbers. So, one
eliminates every second number (again the even numbers), then, rescaling the remaining set, every third
 10

FIG. 6: (a) The G-apparatus filters numbers that are both lucky and prime numbers. The devices A and B are
(L)
made of the potentials VM (x) and VN (x) for the M lucky numbers and N prime numbers respectively, with an
energy cutoff 0 . (b) The transmission amplitude T (E) versus the (dimensionless) energy E from the G-apparatus,
with sharp resonance peaks in correspondence of those values of E that are both lucky and prime numbers.

number, then every fourth number, and so on. As for the primes, there are infinitely many lucky numbers.
Moreover, the prime and the lucky numbers share many properties, including the asymptotic behaviour
according to the prime number theorem. A ”lucky prime” is a lucky number that is prime, and it has
been conjectured that there are infinitely many lucky primes.
Proceeding as in Sections IV and V, in Figure 5 we present, as an example, the experimental holographic
(L)
realization of the potential V10 (x) for the first 10 lucky numbers. Notice that, using the transmission
and reflection properties of a quantum potential, it is possible to set up a simple physical experiment,
shown in Figure 6, to test whether a given number w is both a lucky and a prime number. It involves
a generalization of the proposal originally made in [16] for checking the primality of a number: in the
(L)
present case, let’s imagine that in the box A we have realized the lucky potential VM with a number of
levels M large enough so that LM  w, while in the box B we have instead realized the prime number
potential VN (x) with pN  w. Both potentials can be rounded and truncated at an energy cutoff 0
(which can be controlled by an external handle) in such a way that the original energy levels are essentially
left unperturbed but there are now asymptotic free states. Hence, we can take advantage of the typical
resonance phenomena of quantum mechanics. We send on the composite apparatus G, made of A and B,
a wave-packet from the left (x → −∞) with dimensionless energy w. If the number w is a lucky number,
it will be completely transmitted through box A, and if it is also a prime number, it will be completely
transmitted through box B as well. Therefore, if the particle with energy w is observed coming out the
apparatus G, then the number w is both a lucky and a prime number. This way one could implement a
experimental setup to test whether or not any given number w is a lucky prime.

VII. CONCLUSIONS

In this paper we have provided the first experimental realization of the prime number quantum potential
VN (x), whose single-particle quantum Hamiltonian has the lowest N prime numbers as eigenvalues. The
exact theoretical shape of such a potential has been determined using supersymmetric quantum mechanics
 11

and experimentally implemented by means of holographic techniques. As a proof of principle, we have

experimentally realized the potential VN (x) with N = 10 and N = 15, finding a good agreement of the
eigenvalues of these potentials with the first 15 prime numbers. We have also discussed how this procedure
can be successfully used to implement potentials having other sequences of integers as eigenvalues: this
(L)
is the case of the “lucky” potential VN (x), i.e. the potential which has the first N lucky numbers as
eigenvalues.
The present results provide a physical setup for a quantum mechanical manipulations of discrete se-
quences of numbers. This paves the way towards using these potentials for a variety of mathematical
tests (such as the primality test) and arithmetic manipulations (such as prime factorization) by means of
quantum experiments. It will be interesting to populate the energy levels with neutral atoms (bosonic or
fermionic) and to induce transitions between levels by “shaking” the potential, either in terms of varying
its overall strength or its center of mass, using a periodic drive. A compelling aspect is to determine
whether it is better to employ for such manipulations either fermionic or bosonic atoms. Preliminary
results seem to favour the latter, in absence of sizable atomic interactions, and further work is currently
in progress. Equally interesting is to address other important open problems related to temperature
effects and to the role played by atomic interactions, in view of the efficient implementation of any given
arithmetic operation one wishes to implement on integers.

Acknowledgements: We would like to thank G. D. Bruce for technical assistance.

[1] G.H. Hardy and E.M. Wright, An Introduction to Theory of Numbers (Oxford University Press, 1979).
[2] T.M. Apostol, Introduction to Analytic Number Theory, 5th ed. (Springer, New York, 1998).
[3] T. Tao, Structure and Randomness in the Prime Numbers in An Invitation to Mathematics, eds. D. Schleicher
and M. Lackmann (Springer 2011).
[4] O. Ore, Number Theory and its History, McGraw–Hill (New York), 1948.
[5] P. Ribenboim, The New Book of Prime Number Records (Springer-Verlag, Berlin-New York, 1996).
[6] ) M.R. Schroeder, Number Theory in Science and Communication, Springer–Verlag (Berlin), 1990.
[7] D. Zagier, Math. Intelligencer 0 (1977), 7.
[8] Andrew Granville and Greg Martin, Prime Number Races, American Mathematical Monthly. 113 (1): 1–33.
2006.
[9] H.E. Rose, A Course in Number Theory, Oxford Science Publications (1994).
[10] G. Gauthier et al., Dynamic high-resolution optical trapping of ultracold atoms, Advances In Atomic, Molec-
ular, and Optical Physics (Elsevier) 70, 1-101 (2021).
[11] L. Amico et al., Roadmap on Atomtronics: State of the art and perspective, AVS Quantum Sci. 3, 039201
(2021).
[12] R. Jost and W. Kohn, Equivalent Potentials, Phys. Rev. 88 (1952), 382.
[13] F. Cooper, A. Khare, U. Sukhatme, Supersymmetry and Quantum Mechanics, Phys.Rept.251:267-385,1995.
[14] A. Ramani, B. Grammaticos and E. Caurier, Fractal potentials from energy levels, Phys. Rev. E 51, 6323
(1995).
[15] B.P. van Zyl and D.A. Hutchinson, Riemann zeros, prime numbers, and fractal potentials, Phys. Rev. E 67,
066211 (2003).
[16] G. Mussardo, The Quantum Mechanical Potential for the Prime Numbers , arXiv:cond-mat/9712010.
[17] J. Hadamard, Sur la distribution des zéros de la fonction zeta(s) et ses conséquences arithmétiques, Bull. Soc.
math. France 24, 199-220, 1896.
[18] C.J. de la Vallée Poussin, Recherches analytiques la théorie des nombres premiers, Ann. Soc. scient. Bruxelles
20, 183-256, 1896.
[19] A. Selberg, An Elementary Proof of the Prime Number Theorem, Ann. Math. 50, 305-313, 1949.
[20] P. Erdős, Démonstration élémentaire du théorème sur la distribution des nombres premiers, Scriptum 1,
Centre Mathématique, Amsterdam, 1949.
[21] G.F. B. Riemann, Über die Anzahl der Primzahlen unter einer gegebenen Grösse, Monatsber. Königl. Preuss.
Akad. Wiss. Berlin, 671-680, Nov. 1859.
[22] H.M. Edwards, Riemann Zeta Function, Academic Press, New York, 1974.
[23] Verna Gardiner, R. Lazarus, N. Metropolis and S. Ulam, On Certain Sequences of Integers Defined by Sieves,
Mathematics Magazine, Vol. 29, No. 3 (1956), pp. 117-122.
[24] L. D. Landau and E. M. Lifshitz, Quantum Mechanics: Non-Relativistic Theory, Pergamon Press (1977).
[25] H. Cramér, On the order of magnitude of the difference between consecutive prime numbers, Acta. Arith. 2,
23 (1936).
[26] A.E. Ingham, On the difference between consecutive primes, Quarterly Journal of Mathematics. Oxford Series.
 12

8 (1): 255–266 (1937).

[27] J. Maynard, Small gaps between primes, Annals of Mathematics. 181 (1): 383–413, (2015), arXiv:1311.4600;
J. Maynard Large gaps between primes, Ann. of Math. 183 (3): 915–933 (2016) arXiv:1408.5110
[28] K. Ford, B. Green, S. Konyagin, T. Tao, Large gaps between consecutive prime numbers, Ann. of Math. 183
(3): 935–974 (2016) arXiv:1408.4505
[29] K. Ford, B. Green, S. Konyagin, J. Maynard, T. Tao, Long gaps between primes, J. Amer. Math. Soc. 31 (1):
65–105 (2018), arXiv:1412.5029
[30] R. Grimm, M. Weidemüller, and Y. B. Ovchinnikov, Optical dipole traps for neutral atoms, Advances In
Atomic, Molecular, and Optical Physics (Elsevier) 42, 95–170 (2000).
[31] T. Harte, G. D. Bruce, J. Keeling, and D. Cassettari, Conjugate gradient minimisation approach to generating
holographic traps for ultracold atoms, Opt. Express 22, 26548–26558 (2014).
[32] D. Bowman et al., High-fidelity phase and amplitude control of phase-only computer generated holograms
using conjugate gradient minimisation, Opt. Express 25, 11692-11700 (2017).
[33] D. Trypogeorgos et al., Precise shaping of laser light by an acousto-optic deflector, Opt. Express 21, 24837-
24846 (2013).
[34] C. Ryu, E. C. Samson, and M. G. Boshier,Quantum interference of currents in an atomtronic SQUID, Nature
Comm. 11, 3338 (2020).
[35] M. Pasienski and B. DeMarco, A high-accuracy algorithm for designing arbitrary holographic atom traps,
Opt. Express 16, 2176-2190 (2008).
[36] M. Tajik et al., Designing arbitrary one-dimensional potentials on an atom chip, Opt. Express 27, 33474-
33487 (2019).
[37] A. E. Leanhardt et al., Cooling Bose-Einstein Condensates Below 500 Picokelvin, Science 301, 1513 (2003).
[38] P. Medley et al., Spin Gradient Demagnetization Cooling of Ultracold Atoms, Phys. Rev. Lett. 106, 195301
(2011).
[39] E. Haller et al., Single-atom imaging of fermions in a quantum-gas microscope, Nature Physics 11, 738–742
(2015).
[40] Ph. Zupancic et al., Ultra-precise holographic beam shaping for microscopic quantum control, Opt. Express
24, 13881-13893 (2016).