Steps Towards Constructing a Finite-Dimensional

Holographic Scalar Quantum Field

A thesis submitted in partial fulfillment of

requirements for the award of the degree of

Master of Science (MS)

by

Varun Kushwaha
(19MS040)
5th Year BS-MS
Department of Physics
Indian Institute of Science Education and Research-Kolkata, India

Supervised by
Dr. Oliver Friedrich
Fakultät für Physik
Ludwig-Maximilians-Universität München

Co-supervised by
Prof. Daniel Grün
Fakultät für Physik
Ludwig-Maximilians-Universität München

Coordinated by
Prof. Supratim Sengupta
Indian Institute of Science Education and Research-Kolkata, India
 Declaration
I, Varun Kushwaha, roll number 19MS040, a student of the Department of Physical Sci-
ences of the BS-MS program of IISER Kolkata, hereby declare that this thesis is my own work
and, to the best of my knowledge, it neither contains materials previously published or written
by any other person, nor has it been submitted for any degree/diploma or any other academic
award anywhere before. I have used the originality checking service to prevent inappropriate
copying.

I also declare that all copyrighted material incorporated into this thesis is in compliance with
the Indian Copyright Act, 1957 (amended in 2012) and that I have received written permission
from the copyright owners for my use of their work.

I hereby grant permission to IISER Kolkata to store the thesis in a database which can be
accessed by others.

..............................
Date: 16th May, 2024
(Signature of author)
Place: IISER Kolkata, Mohanpur
Varun Kushwaha
IISER, Kolkata
 Certificate

This is to certify that the thesis titled “Steps Towards Constructing a Finite-Dimensional
Holographic Scalar Quantum Field” submitted by Varun Kushwaha, a student of the De-
partment of Physical Sciences of the BS-MS program of IISER Kolkata, is based upon his own
research work under our supervision. We also certify, to the best of our knowledge, that neither
the thesis nor any part of it has been submitted for any degree/diploma or any other academic
award anywhere before.

.............................. ..............................
(Signature of supervisor) (Signature of co-supervisor)
Oliver Friedrich Daniel Grün
LMU, München LMU, München

Date: 16th May, 2024 Date: 16th May, 2024

Place: LMU, München Place: LMU, München
 Acknowledgements

Let’s just say, I’ve never been the best at predicting how much effort something will require,
be it in my personal or professional life. My final year at IISER-K was a perfect example. Little
did I know, it would be a rollercoaster ride of emotions, with more twists and turns than I
could have ever imagined. But through every challenge, there were incredible people who made
this journey not just possible, but truly enriching.

First and foremost, my deepest gratitude goes to my incredible supervisor, Oliver. He’s a rare
gem – someone who fosters encouragement instead of judgment. Even when I wrestled with
the most basic concepts, his patience never wavered. He believed in the power of “what ifs,”
reminding me that the best things in life rarely unfold exactly according to plan. He instilled in
me a spirit of perseverance, whispering, “Keep moving, keep learning, keep trying. Trust the
process – it might just surprise you.” Our meetings were filled with insightful discussions and
laughter, and for that, I’m eternally grateful. Oliver, your guidance and support have been the
bedrock of this thesis, and I couldn’t have imagined a better person to walk this path with.

My heart overflows with love for my friends at IISER-K. Sai, you were my rock, the cheer-
leader who lifted my spirits during the inevitable lows. Those late-night walks around campus
were more therapeutic than any textbook could ever be. Thank you for your unwavering sup-
port and patience with my emotional rollercoaster rides. Shashwat, your quiet wisdom kept
me grounded. When the world felt like it was crumbling, your reminder that even broken
things can be rebuilt in ways we never imagined offered a beacon of hope. Bipradeep, you were
my technical knight in shining armor, always ready to decipher code and lend a hand. Your
unwavering belief in me, even when I doubted myself, meant the world. Aditya, your genuine
interest in my work was a constant source of motivation, and your kindness during my low
points was a balm to my soul. Isha and Nivedaa, your quiet presence and subtle gestures of
support were never lost on me.

An honorary shout-out goes to Rudra, the most magnificent sloth a friend could ask for.
Thank you for listening to my rants and offering those much-needed hugs that squeezed away
the anxieties.
 5

Juee, you were my grounding force, the one who gently tightened the screws of my self-
esteem whenever they came loose. Your reminders of my past successes reignited my confidence
during moments of self-doubt.

They say life is what happens when you’re busy making other plans. Well, sometimes the
magic happens in the form of unexpected connections. A big thank you to the wonderful
23MS batch – an unexpected delight during my time at IISER-K. Teaching them “Electricity
and Magnetism” became a delightful part of my routine, a welcome break from the intensity of
the thesis. Their perception of me as an approachable and friendly senior filled my days with
laughter and joy. Sharing silly mistakes and random musings with them brought lightness to
even the most monotonous days. Who knew a group of energetic students could be such a
source of happiness?

Finally, a note of gratitude to Dr. Ritesh K. Singh and Dr. Siddhartha Lal for the insight-
ful discussions we had about my project. I have always carried positive memories from our
interactions.

To the developers behind the amazing Python packages NumPy, SciPy, CuPy, and Mat-
plotlib – a heartfelt thank you! Your tireless work laid the foundation for this thesis, and I am
deeply grateful for your contributions.

This thesis is a testament not just to my own efforts, but to the incredible network of support
that carried me through. To each and every one of you, thank you from the bottom of my heart.

And lastly, a special thank you to Bill Watterson, the creator of Calvin and Hobbes comics.
Their sense of humor and exploration resonated with me throughout my thesis journey, offer-
ing moments of levity and reminding me to embrace the wonder of discovery. It’s important
to note that while I have incorporated some illustrations inspired by Calvin and Hobbes to
enhance the presentation of this thesis, none of them infringe upon copyright. All the images
were sourced from royalty-free websites.
 Contents

List of Figures 17

1 Introduction 19
1.1 Outline of the Chapters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2 Why Finite Dimensional Hilbert Space? 23

2.1 Hints from the black hole thermodynamics . . . . . . . . . . . . . . . . . . . 23
2.2 Beyond the Black Hole: Cosmological Event Horizons . . . . . . . . . . . . . 25
2.3 Strong Case from the Holographic Principle . . . . . . . . . . . . . . . . . . 26
2.4 Connecting the Dots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3 The Standard Way of Doing Things 33

3.1 Scalar field in Minkowski Background . . . . . . . . . . . . . . . . . . . . . . 33
3.2 Quantization of the Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4 Finite Dimensional Field 39

4.1 Representation of Weyl Relation . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.2 The commutator of Q and P . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.3 Finite dimensional scalar field . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.3.1 Finite Dimensional Harmonic Oscillator . . . . . . . . . . . . . . . . 48
4.3.2 Finite Dimensional “Ladder” Operators . . . . . . . . . . . . . . . . . 57

5 Scaling of Degrees of Freedom 61

5.1 Finite-dimensional field holographic or not? . . . . . . . . . . . . . . . . . . 62
5.2 Overlapping Degrees of Freedom . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.3 Non-local Commutation Relation . . . . . . . . . . . . . . . . . . . . . . . . 66

6 Overlapping Quantum Scalar Field 67

6.1 Overlapping Position and Momentum Operators . . . . . . . . . . . . . . . . 68
6.2 How good are these operators? . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6.3 Limiting behaviour of Overlapping Operators . . . . . . . . . . . . . . . . . 76
7 Consequences of the Construction 83
7.1 Breaking of Lorentz Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . 85
7.2 Dimension of the Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
7.3 How Small is the Overlap? . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
7.4 Comment on Low Dimensional Simulation . . . . . . . . . . . . . . . . . . . 92
7.5 Vacuum Energy of Overlapping Field . . . . . . . . . . . . . . . . . . . . . . 93
7.6 Overlapping Wavefunctions . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
7.7 Overlapping “Raising” Operator . . . . . . . . . . . . . . . . . . . . . . . . . 102
7.8 Lifetime of Plane Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

8 Discussion and future goals 109

8.1 Summary of results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
8.2 Potential future directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

A Appendix 123
A.1 No Finite-Dimensional Representation of CCR . . . . . . . . . . . . . . . . . 123
A.2 Representation of Q and P in Asymptotic Limit . . . . . . . . . . . . . . . . 124
A.3 Resolution Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

B Appendix 127
B.1 Johnson-Lindenstrauss Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 127
B.2 Moments of the Random Overlaps . . . . . . . . . . . . . . . . . . . . . . . . 129
 Abstract

This thesis explores a novel approach to constructing a finite-dimensional holographic scalar

field theory. Inspired by the holographic principle, which suggests a finite information content
bound by area, not its volume. This implies a finite number of degrees of freedom, challenging
the conventional infinite-dimensional framework of quantum field theory.

We achieve finite dimensionality by replacing standard position and momentum operators

with Generalized Pauli Operators (GPOs). These GPOs exhibit properties similar to their
infinite-dimensional counterparts, especially at low energies. However, this construction still
exhibits a volume-scaling of the degrees of freedom, violating the holographic principle.

To overcome this, we introduce the concept of “overlapping degrees of freedom”, where

a large number of operators are squeezed into a smaller Hilbert space. We construct these
overlapping operators as linear combinations of the GPOs, weighted by random coefficients
drawn from a standard normal distribution. This method guarantees near-orthogonality of
the vectors, enabling us to achieve area-scaling of the degrees of freedom while maintaining a
controlled overlap. Intriguingly, the low-energy behavior of the overlapping Hamiltonian re-
sembles the standard case, and the overlapping raising operator can effectively excite individual
modes, despite the presence of overlaps suggesting the persistence of the particle interpretation.

Our construction, however, introduces potential violations of Lorentz symmetry due to the
non-local nature of mode overlaps. We investigate the lifetime of plane wave states in our theory,
finding a decrease in lifetime with increasing overcrowding in the Hilbert space. Furthermore,
we found no suppression in the vacuum energy of the overlapping field.

This thesis paves the way for a comprehensive finite-dimensional holographic theory. Future
work will address limitations, such as obtaining analytical solutions and implementing Lorentz-
invariant methods. Additionally, exploring alternative operator constructions and expanding to
curved backgrounds are crucial for fully realizing the holographic principle’s phenomenology.
 List of Figures

4.1 Ground state energy of the finite-dimensional harmonic oscillator Hamilto-

nian plotted against the eigenspacing αk . Parameters are M = 1 and Ωk = 10
for all plots. (Top panel) The ground state energy is plotted for a range of
αk values with a fixed dimension dk = 21. (Bottom panel) The ground state
energy is plotted for the same range of αk values but for three choices of
dimension: dk = 21, 51, 81. . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

4.2 Eigenvalues of the Hamiltonian Hˆk plotted against eigennumbers for param-
eters M = 1, Ωk , and dk = 101. The finite-dimensional spectrum (dashed
line) is compared to the corresponding “vanilla spectrum” (solid line) of the
infinite-dimensional harmonic oscillator. A clear deviation between the two
spectra is observed after the marked eigenindex. We have also marked the
maximum number of eigennumbers up to which the infinite-dimensional and
finite-dimensional wavefunctions exhibit good agreement. . . . . . . . . . . . 51

4.3 Ground state wavefunction of the finite-dimensional Hamiltonian Ĥk (orange

crosses) for parameters M = 1, Ωk = 10, and dk = 101. The correspond-
ing ground state wavefunction of the infinite-dimensional harmonic oscillator
(solid line) is also plotted. The residual between the two wavefunctions is
shown below the main plot. The residual is of the order of ∼ 10−15 , indicating
excellent agreement between the finite-dimensional and infinite-dimensional
ground states. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4.4 First excited state wavefunction of the finite-dimensional Hamiltonian Ĥk

(orange crosses) for parameters M = 1, Ωk = 10, and dk = 101. The corre-
sponding first excited state wavefunction of the infinite-dimensional harmonic
oscillator (solid line) is also plotted. The residual between the two wavefunc-
tions is shown below the main plot. The residual is of the order of ∼ 10−15 ,
indicating excellent agreement between the finite-dimensional and infinite-
dimensional ground states. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
12 LIST OF FIGURES

4.5 Second excited state wavefunction of the finite-dimensional Hamiltonian Ĥk

(orange crosses) for parameters M = 1, Ωk = 10, and dk = 101. The corre-
sponding second excited state wavefunction of the infinite-dimensional har-
monic oscillator (solid line) is also plotted. The residual between the two
wavefunctions is shown below the main plot. The residual is of the order of
∼ 10−15 , indicating excellent agreement between the finite-dimensional and
infinite-dimensional ground states. . . . . . . . . . . . . . . . . . . . . . . . . 54

4.6 The wavefunction Zk |0⟩ (orange crosses) is plotted alongside the state |0⟩ of
the finite-dimensional Hamiltonian Hˆk (solid line). Parameters are M = 1,
Ωk = 10, and dk = 101. The residual between the two wavefunctions is shown
below the main plot. The small residual, on the order of ∼ 10−14 , suggests
that Zk acts almost like the identity operator on the vacuum state |0⟩. . . . 54

4.7 The wavefunction Zk |1⟩ (orange crosses) is plotted alongside the state |1⟩ of
the finite-dimensional Hamiltonian Hˆk (solid line). Parameters are M = 1,
Ωk = 10, and dk = 101. The residual between the two wavefunctions is shown
below the main plot. The small residual, on the order of ∼ 10−14 , suggests
that Zk acts almost like the identity operator on the first excited state |1⟩. . 55

4.8 The wavefunction Zk |2⟩ (orange crosses) is plotted alongside the state |2⟩ of
the finite-dimensional Hamiltonian Hˆk (solid line). Parameters are M = 1,
Ωk = 10, and dk = 101. The residual between the two wavefunctions is shown
below the main plot. The small residual, on the order of ∼ 10−14 , suggests
that Zk acts almost like the identity operator on the second excited state |2⟩. 55

4.9 Expectation value of Zk plotted against the eigennumbers of the finite-dimensional

Hamiltonian Hˆk for parameters M = 1, Ωk = 10, and dk = 101. The expec-
tation value ⟨Ψ|Zk |Ψ⟩ is approximately equal to one up to the point where the
finite-dimensional wavefunctions match well with the corresponding infinite-
dimensional ones. Beyond this region, the expectation value deviates from
unity, possibly due to the eigenspacing αk becoming too large to accurately
resolve the high-energy wavefunctions. . . . . . . . . . . . . . . . . . . . . . 56

4.10 Action of the finite-dimensional creation operator â†k on the ground state. The
resulting excited state (orange crosses) is compared to the first excited state
of the finite-dimensional Hamiltonian Hˆk (solid line). Parameters are M = 1,
Ωk = 10, and dk = 101. The small residual between the two wavefunctions,
on the order of ∼ 10−15 , suggests that â†k effectively acts like a raising operator
on the ground state, |0⟩. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
 LIST OF FIGURES 13

4.11 Action of the finite-dimensional creation operator â†k twice on the ground
state. The resulting excited state (orange crosses) is compared to the second
excited state of the finite-dimensional Hamiltonian Hˆk (solid line). Parame-
ters are M = 1, Ωk = 10, and dk = 101. The small residual between the two
wavefunctions, on the order of ∼ 10−14 , suggests that â†k effectively acts like
a raising operator on the low energy states of Hˆk . . . . . . . . . . . . . . . . 59

5.1 Top: Standard commutation (non-overlapping circles, large Hilbert space).

Bottom: Reduced dimensionality with controlled commutativity (overlapping
circles, smaller space). This approach targets area scaling while respecting
holographic and particle physics constraints. . . . . . . . . . . . . . . . . . . 65

6.1 This cartoon is adapted from Friedrich et al. (2024). (Left sketch) The scalar
field can be decomposed into a collection of harmonic oscillators represented
by points k in Fourier space. In three dimensions, the number of oscillators
within a small Fourier space shell s (red crossed region in the upper sketch)
scales as ks2 ∆s , where ks is the radius and ∆s is the width of the shell. To
achieve holographic scaling of the effective degrees of freedom in the field
theory, we aim to embed these qubits into a finite-dimensional Hilbert space
that is strictly smaller than the tensor product of the individual oscillator
Hilbert spaces (Right sketch). . . . . . . . . . . . . . . . . . . . . . . . . . . 70

6.2 The plot compares the eigenvalues of the non-overlapping position operator
Qk (blue solid line) with those of the overlapping position operator Q̃k (or-
ange solid line). The overlapping operator is constructed with the following
parameters: ns = 3, Ns = 4, dk = 19, Ωk = 10, and M = 1. The eigenval-
ues of the non-overlapping operator exhibit degeneracy, which is lifted after
the introduction of overlaps, as evident in the eigenvalues of the overlapping
operator Q̃k . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

6.3 Average eigenvalues of finite-dimensional position operators. This plot shows

the average eigenvalues of the non-overlapping position operator Qk (blue
dots) and the overlapping position operator Q̃k (orange dots), calculated over
the eigennumbers where the non-overlapping eigenvalues remain degenerate
(refer to Figure 6.2 for reference). The overlapping operator parameters are
ns = 3, Ns = 4, dk = 19, Ωk = 10, and M = 1. We observe that the
average eigenvalues coincide in the bulk (central region) but deviate towards
the edges, indicating the influence of the introduced overlaps. . . . . . . . . . 75
14 LIST OF FIGURES

7.1 The plot shows the eigenspectra of the Hamiltonian Hˆk for three different
dimensions: dk = 101 orange), dk = 201 (green), and dk = 401 (red). Other
parameters are fixed at M = 1 and Ωk = 10. As the dimension dk increases,
the agreement between the “vanilla spectrum” (blue) (expected spectrum of
the infinite-dimensional harmonic oscillator) and the finite-dimensional spec-
trum improves. Furthermore, for larger values of dk , the action of Zk on the
eigenstates of Ĥk becomes increasingly similar to that of the identity operator
for a wider range of states. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

7.2 Bound on the commutator due to finite-dimensional embedding. This plot

shows the bound on the commutator ∥[Q
e , Pe ′ ]|Ψ⟩∥ for non-coincident modes
k k

k ̸= k′ with similar wavenumbers |k|≈ k ≈ |k′ |. The embedding used to

achieve a finite-dimensional Hilbert space is the Johnson-Lindenstrauss em-
bedding (see Chapter 6 for details). The vertical small arrows indicate typical
energy scales of the Large Hadron Collider (LHC) and everyday phenomena.
We observe that the bound is significantly smaller than the operator norms
∥[Q
e , Pe ′ ]|Ψ⟩∥ at these energy scales. This suggests that for such energies,
k k

any two Fourier modes behave almost like independent degrees of freedom. . 91

7.3 The plot shows the ground state wavefunction of the overlapping Hamiltonian
constructed from ns = 3 effective non-overlapping oscillators (orange crosses)
compared to the ground state wavefunction of the corresponding infinite-
dimensional system (solid blue line). The overlapping Hamiltonian consists
of Ns = 4 oscillators, each with dimension dk = 17 and parameters M = 1
and Ωk = 10. The orange crosses represent the ground state after taking the
“effective trace” of the complete wavefunction. The residual between the two
wavefunctions is on the order of ∼ 10−7 , indicating excellent agreement. This
result is achieved even though it represents a scenario that minimally exploits
the full potential of our embedding scheme. This suggests that our asymptotic
limit results for the ground state hold even in this extreme low-dimensional
case. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

7.4 The plot shows the closest first state wavefunction of the overlapping Hamil-
tonian constructed from ns = 3 effective non-overlapping oscillators (orange
crosses) compared to the first excited state wavefunction of the corresponding
infinite-dimensional system (solid blue line). The overlapping Hamiltonian
consists of Ns = 4 oscillators, each with dimension dk = 17 and parameters
M = 1 and Ωk = 10. We have again taken “effective trace.” The residual
between the two wavefunctions is on the order of ∼ 10−6 , indicating excellent
agreement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
 LIST OF FIGURES 15

7.5 The plot compares the closest second state wavefunction of the overlapping
Hamiltonian constructed from ns = 3 effective non-overlapping oscillators
(orange crosses) compared to the second excited state wavefunction of the
corresponding infinite-dimensional system (solid blue line). The overlapping
Hamiltonian consists of Ns = 4 oscillators, each with dimension dk = 17 and
parameters M = 1 and Ωk = 10. As before, the orange crosses represent
the state after taking the “effective trace” of the complete wavefunction. The
residual between the two wavefunctions is on the order of ∼ 10−2 , suggesting
a noticeable impact of overlaps on this state compared to the ground state.
However, the key features of the wavefunction are still well-preserved. . . . . 98

7.6 The plot compares the closest third state wavefunction of the overlapping
Hamiltonian constructed from ns = 3 effective non-overlapping oscillators
(orange crosses) compared to the third excited state wavefunction of the cor-
responding infinite-dimensional system (solid blue line). The overlapping
Hamiltonian consists of Ns = 4 oscillators, each with dimension dk = 17 and
parameters M = 1 and Ωk = 10. As before, the orange crosses represent
the state after taking the “effective trace” of the complete wavefunction. The
residual between the two wavefunctions is on the order of ∼ 10−2 , suggesting
a noticeable impact of overlaps on this state compared to the ground state.
However, the key features of the wavefunction are still well-preserved. . . . . 99

7.7 The plot compares the closest fourth state wavefunction of the overlapping
Hamiltonian constructed from ns = 3 effective non-overlapping oscillators
(orange crosses) compared to the fourth excited state wavefunction of the
corresponding infinite-dimensional system (solid blue line). The overlapping
Hamiltonian consists of Ns = 4 oscillators, each with dimension dk = 17 and
parameters M = 1 and Ωk = 10. As before, the orange crosses represent
the state after taking the “effective trace” of the complete wavefunction. The
residual between the two wavefunctions is on the order of ∼ 10−2 , suggesting
a noticeable impact of overlaps on this state compared to the ground state.
However, the key features of the wavefunction are still well-preserved. . . . . 100
16 LIST OF FIGURES

7.8 The plot compares the closest fifth state wavefunction of the overlapping
Hamiltonian constructed from ns = 3 effective non-overlapping oscillators
(orange crosses) compared to the fifth excited state wavefunction of the cor-
responding infinite-dimensional system (solid blue line). The overlapping
Hamiltonian consists of Ns = 4 oscillators, each with dimension dk = 17 and
parameters M = 1 and Ωk = 10. As before, the orange crosses represent
the state after taking the “effective trace” of the complete wavefunction. The
residual between the two wavefunctions is on the order of ∼ 10−2 , suggesting
a noticeable impact of overlaps on this state compared to the ground state.
However, the key features of the wavefunction are still well-preserved. . . . . 101

7.9 The plot compares the wavefunction obtained after applying the creation
operator ã†k to the ground state of the overlapping Hamiltonian (orange
crosses) with the first excited state wavefunction of the corresponding infinite-
dimensional system (solid blue line). The overlapping Hamiltonian is con-
structed from ns = 3 effective non-overlapping oscillators and consists of
Ns = 4 individual oscillators, each with dimension dk = 17 and parameters
M = 1 and Ωk = 10. As before, the orange crosses represent the state after
taking the “effective trace” of the complete wavefunction. The small residual
between the two wavefunctions, on the order of ∼ 10−7 , suggests excellent
agreement between the two. . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

7.10 The plot compares the wavefunction obtained after applying the creation op-
erator ã†k to the closest first excited state of the overlapping Hamiltonian (or-
ange crosses) with the second excited state wavefunction of the corresponding
infinite-dimensional system (solid blue line). The overlapping Hamiltonian is
constructed from ns = 3 effective non-overlapping oscillators and consists of
Ns = 4 individual oscillators, each with dimension dk = 17 and parameters
M = 1 and Ωk = 10. As before, the orange crosses represent the state af-
ter taking the “effective trace” of the complete wavefunction. The residual
between the two wavefunctions is on the order of ∼ 10−2 , which is larger
compared to the ground state case (see Figure 7.9). This suggests a more
significant impact of overlaps on excited states. However, the key features of
the wavefunction are still preserved. . . . . . . . . . . . . . . . . . . . . . . . 104
 LIST OF FIGURES 17

7.11 The plot compares the wavefunction obtained after applying the creation op-
erator ã†k to the closest second excited state of the overlapping Hamiltonian
(orange crosses) with the third excited state wavefunction of the correspond-
ing infinite-dimensional system (solid blue line). The overlapping Hamil-
tonian is constructed from ns = 3 effective non-overlapping oscillators and
consists of Ns = 4 individual oscillators, each with dimension dk = 17 and
parameters M = 1 and Ωk = 10. As before, the orange crosses represent
the state after taking the “effective trace” of the complete wavefunction. The
residual between the two wavefunctions is on the order of ∼ 10−2 , which is
larger compared to the ground state case (see Figure 7.9). This suggests
a more significant impact of overlaps on excited states. However, the key
features of the wavefunction are still preserved. . . . . . . . . . . . . . . . . 105
7.12 Scrambling time versus number of overlapping operators. The plot shows
the scrambling time, Tscramble , as a function of the number of overlapping
operators, Ns , included in the overlapping Hamiltonian. Results are presented
for two different dimensions of the individual oscillators: dk = 9 (blue crosses)
and dk = 15 (orange dots). Other parameters are fixed at ns = 3, M = 1, and
Ωk = 10. Due to the use of dimensionless quantities in the simulations, the y-
axis is unitless, and the specific numerical values hold less significance. Each
data point represents the average scrambling time over 1200 simulations, with
error bars indicating the standard deviation of the mean of these 7 points.
The key observation is a decreasing trend in scrambling time with increasing
Ns , which is explained in detail in Section 7.8. . . . . . . . . . . . . . . . . . 108
 Introduction
1
One of the most significant challenges in contemporary physics lies in reconciling quantum
mechanics and general relativity, the two pillars of modern physics, into a unified framework
known as “quantum gravity” (Padmanabhan, 2014; Rendall, 2005). This quest remains a
formidable task, with limited experimental guidance due to the extremely high energies in-
volved. However, the holographic principle offers a powerful new perspective on the infor-
mation content of spacetime, potentially shedding light on quantum gravity (see e.g., Smolin
(2001); Hořava & Minic (2000); Banks (2001)).

The holographic principle, a profound conjecture arising from black hole thermodynam-
ics, suggests a fundamental limitation on the information capacity of a region within spacetime.
Counterintuitively, it posits that the maximum entropy, and therefore the information content,
of a region is proportional to the area of its boundary, not its volume. This implies that the de-
grees of freedom describing all of physics in the region, reside on its boundary, challenging the
traditional view of Local Quantum Field Theory (LQFT) where these degrees of freedom are
assumed to exist at every point in spacetime (Bousso, 2002). While the holographic principle
remains unproven and could potentially be an emergent phenomenon, its lack of counterex-
amples and its ability to bridge seemingly disparate concepts like spacetime geometry and the
number of quantum states are highly significant. Current frameworks struggle to explain this
bound, highlighting the potential of the holographic principle to advance our understanding of
quantum gravity.
20 Introduction

The covariant entropy bound, the most general formulation of the holographic principle,
states a fundamental relationship between the information content and the geometry of space-
time. As put forward by Bousso (2002), the number of independent quantum states describing
matter on light-sheets of any surface is bounded by the exponential of the surface area:

N [L(B)] ≤ eA(B)/4 . (1.1)

This bound has profound implications for our understanding of the nature of reality. It suggests
that quantum field theory, with its assumption of infinitely many degrees of freedom per unit
volume, might be an effective description that breaks down at a fundamental energy scale. The
holographic principle compels us to seek a new theory that inherently embodies this bound,
where the number of fundamental degrees of freedom is intrinsically linked to the area of en-
closing surfaces in spacetime (for a detailed review, see e.g., Bousso (2002)).

In this thesis, we embark on a constructive exploration of holographic phenomenology

by focusing on a real scalar quantum field in Minkowski spacetime analogous to Friedrich
et al. (2024). While a de Sitter background, often considered more physically relevant to our
universe, would be ideal, the computational complexity associated with it necessitates a sim-
pler starting point. This choice of Minkowski spacetime allows us to investigate the effects
of finite dimensionality and holography in a more tractable setting, paving the way for future
advancements in this area.

Our approach involves minimally modifying conventional scalar field theory to comply with
the holographic principle and its ramifications. Building upon the work of Cao et al. (2019);
Singh & Carroll (2020); Friedrich et al. (2024, 2022), we seek to construct a finite-dimensional
holographic scalar field that exhibits area-law scaling of its degrees of freedom. This endeavor
necessitates addressing two key challenges:

1. Constructing a finite-dimensional representation of position and momentum operators

that respects the fundamental principles of quantum mechanics.

2. Implementing a mechanism for “overlapping degrees of freedom” to achieve the desired

area scaling.

1.1 Outline of the Chapters

Chapter 2 begins with an extensive discussion on the theoretical motivations for consider-
ing finite-dimensional Hilbert spaces in quantum gravity. We draw insights from black hole
thermodynamics, cosmological event horizons, and the holographic principle.
 1.1 Outline of the Chapters 21

Chapter 3 provides a brief review of the standard, infinite-dimensional construction of a

real scalar field. Chapter 4 delves into the intricacies of finite-dimensional quantum mechan-
ics. We introduce Generalized Pauli Operators (GPOs) as our finite-dimensional counterparts
to position and momentum operators. These operators, while satisfying a modified commuta-
tion relation, exhibit properties that closely resemble their infinite-dimensional counterparts,
particularly in the low-energy regime. We investigate the eigenspectrum and eigenstates of the
finite-dimensional harmonic oscillator Hamiltonian, highlighting the deviations from the stan-
dard infinite-dimensional case.

Chapter 5 analyzes the degrees of freedom associated with our finite-dimensional field and
demonstrates that it still exhibits volume scaling, violating the holographic principle. This
motivates the introduction of the concept of “overlapping degrees of freedom” in Chapter 6.

We construct “overlapping” position and momentum operators as linear combinations of

their non-overlapping counterparts, weighted by random coefficients derived from normalized
vectors drawn from a standard normal distribution. We leverage the Johnson-Lindenstrauss
theorem, which guarantees the existence of nearly orthogonal vectors in high-dimensional
spaces. This method allows us to pack a large number of operators into a smaller Hilbert
space while maintaining a controlled overlap. We analyze the properties of these overlapping
operators, deriving their commutation relations and demonstrating their convergence to the
standard canonical commutation relation in the infinite-dimensional limit.

Chapter 7 explores the consequences of our overlapping construction. We investigate the

“overlapping Hamiltonian,” constructed from these overlapping operators, and analyze its eigen-
spectrum and eigenstates. Numerical computations reveal a remarkable resemblance between
the low-energy wavefunctions of the overlapping Hamiltonian and their standard counterparts.
We introduce an “effective trace” procedure to average over the randomness inherent in our
construction, further highlighting this correspondence. We also explore the behavior of the
overlapping “raising operator” and its ability to effectively excite individual modes. This find-
ing suggests that the particle interpretation of excitations, a crucial aspect of standard QFT,
persists in our holographic model, at least for low-energy states.

We acknowledge, however, that our construction also introduces potential violations of

Lorentz symmetry due to the non-local nature of mode overlaps. We present a preliminary
analysis of the lifetime of plane wave states in our theory and conduct low-dimensional simula-
tions to explore its dependence on the degree of operator squeezing. We discuss the implications
of these findings and outline potential avenues for future work to address this challenge.

Finally, Chapter 8 summarizes our key results and discusses the limitations of our approach,
highlighting several open questions and potential future directions. These include the need for
analytical expressions for the eigenspectrum and eigenstates of the overlapping Hamiltonian,
a more rigorous treatment of plane wave lifetime, the implementation of Lorentz-invariant
22 Introduction

regularization schemes, and the exploration of alternative constructions of conjugate operators.

We also emphasize the importance of extending our construction to an expanding background
and investigating its implications for the cosmological constant problem.

This thesis serves as a stepping stone towards constructing a comprehensive finite-dimensional

holographic scalar field theory. By addressing the limitations and open questions outlined in
Chapter 8, we can gain deeper insights into the implications of finite dimensionality and holog-
raphy for quantum field theory, ultimately taking significant steps toward realizing the phe-
nomenology of the holographic principle.
 Why Finite
2
Dimensional Hilbert Space?

2.1 Hints from the black hole thermodynamics

Motivated by the flourishing developments in classical general relativity during the 1970s,
significant progress was made in understanding black holes. Two particularly noteworthy the-
orems emerged. The first, the area theorem (Hawking, 1971), which says that the area of a
black hole’s event horizon is non-decreasing with time. Furthermore, if multiple black holes
coalesce, the resulting black hole possesses an event horizon with an area greater than the sum
of the area of the individual black holes’ horizons. The second, the no-hair theorem (Carter,
1971; Hawking, 1972; Israel, 1967, 1968), postulates that a stationary black hole can be fully
characterized by just three quantities: its charge (QH ), mass (MH ), and angular momentum
(JH ).

However, the no-hair theorem presents a paradoxical situation. From an external observer’s
perspective, it appears to violate the second law of thermodynamics. The phase space, encom-
passing all possible microstates of a system, seemingly undergoes a drastic reduction. The final
state – the black hole itself – possesses zero entropy, in stark contrast to the arbitrarily high
entropy a collapsing system may exhibit. This implies a potential information loss, as various
24 Why Finite Dimensional Hilbert Space?

initial conditions could seemingly converge to an indistinguishable final state.

Bekenstein proposed a solution to this paradox (Bekenstein, 1972, 1973, 1974). He in-
troduced the concept of black hole entropy, SBH = ξA, where ξ is a constant of order
unity and A represents the horizon area. This assignment ensures that the total entropy,
Stotal = Smatter + SBH , where Smatter is the matter entropy, always satisfies the non-decreasing
entropy condition. This later became the “Generalized Second Law of Thermodynamics.”

Interestingly, since the no-hair theorem dictates that a black hole formed through gravita-
tional collapse is characterized by only three parameters: mass (MH ), angular momentum (JH ),
and charge (QH ) therefore, a black hole with a specific set of these parameters can possess a vast
number of indistinguishable internal configurations, reflecting the different initial conditions
of the collapsing matter. Classically, this number would be infinite as a black hole could form
from an infinitely large collection of particles with infinitesimally small masses. However, quan-
tum mechanics restricts the particles’ energies – for collapse to occur, their wavelengths must
be smaller than the black hole’s size. This suggests a finite number of internal configurations.
Consequently, an entropy, SBH , can be associated with the black hole, representing the loga-
rithm of this number (Bekenstein, 1973, 1974; Hawking, 1976). This underpins the rationale
behind Bekenstein’s assignment of finite entropy to black holes.

For consistency, a black hole with entropy SBH would radiate like a body with a temperature
of: −1
∂SBH
 
TH = G2N . (2.1)
∂MH JH ,QH

Furthermore, the Einstein equations would yield:

κ A
 
dMH = d = TH dSBH , (2.2)
2π 4

where κ is the black hole’s surface gravity. Here, the analogy with thermodynamics seems
stretched. Classically, a black hole with non-zero temperature should radiate; however, nothing
can escape a black hole, implying a temperature of zero.

Hawking clarified this issue through a semi-classical calculation (Hawking, 1974; Hawking,
1975). He demonstrated that a distant observer would detect a thermal spectrum of parti-
cles emanating from the black hole, with a temperature directly proportional to κ/(2π). This
groundbreaking discovery of Hawking radiation illuminated the physical significance of the
thermodynamic description applied to black holes. What previously appeared as a superficial
analogy was now recognized as a genuine physical property. Consequently, the entropy and
temperature associated with a black hole are as real as its mass. Notably, Hawking’s findings
corroborated Bekenstein’s earlier proposition, solidifying the notion that black hole entropy
(SBH = A/4 in Planck units) is a genuine contribution to the total entropy of the universe.
2.2 Beyond the Black Hole: Cosmological Event Horizons

Building on the application of thermodynamics and particle creation near black holes, Gib-
bons and Hawking in 1977 (Gibbons & Hawking, 1977) explored the extension of these con-
cepts to cosmological event horizons. In a black hole, the immense gravitational pull traps even
light, preventing it from escaping to a distant observer. This inaccessible region is demarcated
by the event horizon. Similarly, cosmological models with a positive cosmological constant
(Λ) exhibit event horizons. The repulsive nature of Λ drives an accelerated expansion, causing
certain regions to become causally disconnected from a specific observer. The boundary of this
inaccessible region defines the observer’s cosmological event horizon.

Asymptotically, these Λ-dominated models approach de Sitter space at late times. In de Sitter
space, the concept of future infinity becomes spacelike (Penrose, 2011; Hawking & Ellis, 2023).
This implies that each observer on a timelike worldline has an event horizon, separating their
observable universe from the forever unseen region. Intriguingly, this cosmological event hori-
zon displays remarkable parallels to a black hole event horizon. It exhibits behaviors analogous
to the zeroth, first, and second laws of black hole mechanics in classical gravity (Bardeen et al.,
1973). Similar to black holes, it demarcates a region where particles can exhibit negative
energy relative to the observer.

Consequently, particle creation near these cosmological horizons is a logical possibility.

Indeed, Gibbons and Hawking demonstrated that observers would detect isotropic thermal
radiation with a characteristic wavelength linked to the Hubble radius, corresponding to an
extremely low temperature (around 10−28 K) (Gibbons & Hawking, 1977). While negligible
from a practical standpoint, this discovery conceptually highlights the broader applicability of
thermodynamic arguments to the entire cosmos and strengthens the connection between event
horizons, gravity, and thermodynamics – a connection already observed in black holes.

However, the notion of observer-independent particles breaks down in the context of event
horizons. Instead, we must consider the perspective of an observer with a particle detector
following a timelike geodesic. Such an observer would detect an isotropic thermal background
q
with a temperature related to the surface gravity (κC = Λ/3) of their own cosmological
event horizon. Interestingly, different observers in relative motion would still measure the
same temperature, indicating that they are not detecting the same particles. This underscores
the observer-dependent nature of particles in this context (for a more detailed discussion, see
section VI of Gibbons & Hawking (1977)).

In essence, the area of the cosmological event horizon can be interpreted as a measure of an
observer’s limited knowledge about the universe beyond their causal patch.
2.3 Strong Case from the Holographic Principle

The holographic principle proposes a profound limitation on the information content of

spacetime regions. This principle has evolved over time, culminating in its current, most general
form known as the “covariant entropy bound”. This bound postulates that a fundamental law
of physics, likely a unified quantum theory of gravity, must govern the information content
of spacetime. This theory would necessitate the emergence of Lorentzian geometries and their
matter content such that the number of independent quantum states describing matter on light-
sheets of any surface (denoted by B) is demonstrably bounded by the exponential of the surface
area (Bousso, 2000):
N [L(B)] ≤ eA(B)/4 . (2.3)

The holographic principle asserts that this bound is not merely coincidental, but rather a
consequence of a new, underlying theory. This bound bridges seemingly disparate concepts:
spacetime geometry (reflected in the exponent) and the number of quantum states associated
with matter. This suggests that any theory incorporating the holographic principle must unify
matter, gravity, and quantum mechanics, forming a quantum theory of gravity that transcends
the limitations of quantum field theory and general relativity. This expectation is bolstered by
the strong connection between the covariant entropy bound and the semi-classical properties
of black holes. Furthermore, string theory results offer limited, yet supportive evidence for the
bound’s validity (Strominger & Vafa, 1996; Peet, 2001).

The holographic principle implies a dramatic reduction in the degrees of freedom tra-
ditionally used to describe nature. It exposes quantum field theory, with its assumption of
degrees of freedom at every point in space, as an effective description that obscures the true,
reduced number of degrees of freedom.

However, a note of caution is warranted. The covariant entropy bound, while compelling,
could potentially be proven incorrect or simply an accidental pattern. Even if the bound does
stem from a fundamental theory, the connection might be indirect, rendering the holographic
principle less insightful into the theory’s core tenets. Nevertheless, the holographic principle
remains a significant conceptual advancement needed for progress in quantum gravity research.
The bound’s validity has been upheld across various frameworks, and no realistic counterex-
amples have been identified (Bousso, 2002).

The bound’s non-triviality lies in its defiance of naive expectations. Intuitively, one might
expect the maximum entropy to scale with the volume of a region. However, the bound
dictates that the area of a surface sets the limit. This bound pertains to statistical entropy,
implying a fundamental restriction on the number of degrees of freedom in nature, independent
of microscopic matter properties. Additionally, current physical laws fail to explain the
 2.3 Strong Case from the Holographic Principle 27

bound. Black hole thermodynamics alone cannot account for a bound encompassing both the
deep interior of black holes and cosmological contexts (Bousso, 2000). These factors collectively
suggest that the bound is an imprint of a more fundamental theory.

The holographic principle compels us to formulate a theory that inherently embodies the
covariant entropy bound. How can such a holographic theory be constructed? Conventional
quantum field theories (QFTs), for instance, posit degrees of freedom at every point in space-
time. Even with a finite ultraviolet cutoff, the information content within a spatial region would
seemingly scale with its volume. Conversely, the holographic principle suggests a relationship
between the number of fundamental degrees of freedom and the area of enclosing surfaces in
spacetime. Reconciling this apparent contradiction is a central challenge.

One key aspect to consider is the diverse ways information can be encoded in physics. Ex-
amples include quantum states of a Conformal Field Theory (CFT) or a lattice of spins. Despite
its precise predictions regarding the information content within a spacetime region, the holo-
graphic principle remains agnostic about the nature of these fundamental degrees of freedom.
While it dictates the overall information quantity, its specific form remains elusive.

The notion that all of physics could be underpinned by pure, abstract information is a fasci-
nating proposition. However, this very abstractness presents a significant hurdle in the design
of concrete models that successfully integrate the holographic principle (Bousso, 2002).

Two primary approaches have emerged to address the challenges posed by the holographic
principle, each framing the issue in a distinct way.

One approach emphasizes retaining locality within a holographic framework. This could
be achieved by identifying a specific gauge symmetry within a local theory, effectively reducing
the number of physical degrees of freedom to align with the covariant entropy bound. However,
implementing such a comprehensive and unconventional gauge symmetry presents a significant
challenge (see e.g., Hooft (1999); ’t Hooft (2000); Hooft (2001); ’t Hooft (2001a,b,c); van de
Bruck et al. (2000)).

The second approach views locality as an emergent property, not a fundamental princi-
ple (Cao & Carroll, 2018; Cao et al., 2017; Carroll, 2021; Carroll & Singh, 2018; Cotler et al.,
2019; Dyson et al., 2002). In this perspective, holographic data holds primacy. Understanding
its generation and evolution becomes a central challenge, alongside explaining how this under-
lying data translates, under suitable conditions, into a classical spacetime populated by local
quantum fields. A successful theory in this vein would necessarily shape spacetime geometry
and distribute matter in a way that adheres to the covariant entropy bound. Since holographic
data naturally associates with the area of surfaces, a significant hurdle lies in comprehending
how locality can emerge within this framework.
28 Why Finite Dimensional Hilbert Space?

The AdS/CFT correspondence offers some support for the second approach (for a more
general reference on AdS/CFT, see, e.g., Nastase (2007)). However, its applicability is limited
to asymptotically AdS universes due to several specific features, providing minimal guidance for
broader applications of this approach. Additionally, perturbative string theory only partially
embodies the holographic principle. Effects linked to holography include the wave function’s
independence of the longitudinal coordinate in the light cone frame and the growth of state size
with momentum (Klebanov & Susskind, 1988; Kallosh & Linde, 1996).

Several studies have investigated the extent of non-locality exhibited by string theory in
the context of the holographic principle (Lowe et al., 1994, 1995). These investigations are
closely tied to understanding black hole evaporation unitarity from a string theory perspective,
particularly through the principle of black hole complementarity.

The one bit per Planck area entropy bound is not explicitly manifest in perturbative string
theory. Susskind (1995) demonstrated that the perturbative expansion breaks down before the
bound is violated. A full manifestation of the holographic principle is likely to require a
non-perturbative formulation of string theory.

Since the inception of the holographic principle, non-perturbative string theory definitions
have indeed emerged, such as the AdS/CFT correspondence, where the number of degrees of
freedom demonstrably aligns with the holographic principle (Susskind & Witten, 1998).

The holographic principle might not only aid the search for additional non-perturbative
string theory definitions but could also contribute to a background-independent formulation,
potentially illuminating the core concepts underlying string theory.

The preference for one approach over the other hinges largely on which aspect one finds
more challenging: eliminating most degrees of freedom or recovering locality. This is not
necessarily a strict dichotomy, and the two approaches are not mutually exclusive. A successful
theory might admit multiple equivalent formulations, thereby reconciling both perspectives.

The holographic principle typically restricts the number of degrees of freedom, N , relative
to a specified surface. However, there are spacetimes (e.g., asymptotically de Sitter spacetime,
arguably representing our Universe) where the covariant entropy bound suggests an absolute
upper limit on N . In such universes with a positive cosmological constant (Λ), the observ-
able entropy is bounded by 3π/Λ (Bousso, 2002). Physically, these universes lack the “room”
to generate entropy exceeding N . Notably, they cannot accommodate black holes with an
area exceeding an order of N . The finiteness of de Sitter entropy has particularly strong im-
plications, as noted by Banks and Fischler (Banks, 2001; Banks & Fischler, 2001a,b). These
implications are discussed in more detail in the next section.
2.4 Connecting the Dots

A key distinction between black hole horizons and cosmological horizons lies in their ob-
server dependence. In asymptotically flat spacetimes (or their Anti-de Sitter counterparts), a
black hole horizon encompasses the region causally accessible to any observer at infinity. Con-
versely, in cosmological backgrounds like de Sitter space, different observers can influence and
observe distinct spacetime regions, leading to unique horizons. Notably, the high symmetry of
de Sitter space ensures that all observers experience horizons with the same area. This might
not hold true in more general cosmological models.

De Sitter gravity plays a crucial role in modern cosmology due to the possibility that our
universe is currently undergoing a de Sitter (dS) phase (Klemm & Vanzo, 2004). Observations
of Type Ia supernovae suggest an accelerating cosmic expansion, hinting at a non-zero, positive
cosmological constant (Perlmutter et al. (1999); Riess et al. (1998)). This aligns with indepen-
dent evidence from Cosmic Microwave Background (CMB) anisotropy measurements (COBE,
Boomerang, WMAP) which indicate a spatially flat universe (Spergel et al., 2003). However, a
flat geometry necessitates a total energy density significantly exceeding that of ordinary matter.
This implies the existence of a dominant, unseen form of “dark energy.”

The simplest and most compelling candidate for dark energy is a small, positive cosmological
constant denoted by Λ (see, e.g., Peebles & Ratra (2003) for a general discussion). Recent WMAP
data suggests that roughly 73% of the universe’s energy density is attributed to dark energy, with
dark matter contributing approximately 22% and baryons at 4.4% (Aghanim et al., 2020). While
the value of Λ is constantly refined, current estimations for its associated energy density can be
expressed as:
Λ
ρΛ = ≤ (10−3 eV)4 ≃ 103 eV · cm−3 . (2.4)
8πG
This translates to a mass density of 10−29 g/cm3 . It is crucial to note that this value is significantly
smaller than the vacuum energy predicted by Standard Model fields using the Planck scale as a
cutoff, differing by a factor of approximately 10122 (see, e.g., Martin (2012) for a discussion).
This discrepancy is known as the cosmological constant problem (Weinberg, 1989).

Supersymmetry has been proposed as a potential solution, as it can cancel vacuum fluc-
tuations between bosons and fermions (Zumino, 1977). However, such a solution would ne-
cessitate an extremely precise fine-tuning mechanism, cancelling virtual-particle energies to an
exceptional degree. A detailed discussion of the cosmological constant problem is beyond the
scope of this work, and readers are referred to the extensive literature available on the subject
(see e.g., Weinberg (1989), Carroll (2001), Ellwanger (2003), Peebles & Ratra (2003)).

The applicability of de Sitter gravity extends beyond the current epoch of cosmic expansion.
30 Why Finite Dimensional Hilbert Space?

It also finds relevance in the inflationary epoch, a period in the universe’s early history that is
widely believed to have been dominated by a de Sitter phase (Klemm & Vanzo, 2004).

While black hole entropy remains a topic of intense research, the entropy of cosmological
horizons, like the de Sitter horizon, presents a particularly profound challenge (Almheiri et al.,
2021). A prominent hypothesis suggests that black hole entropy corresponds to the logarithm
of a quantum Hilbert space’s dimension, which is necessary to describe a black hole from the
perspective of an observer outside the horizon. How can this concept be applied to cosmo-
logical horizons? An observer in de Sitter space, studying its horizon, serves as a potential
analog to an observer outside a black hole horizon. The causally accessible region of de Sitter
space for such an observer is referred to as a “static patch.” Several authors have proposed a
de Sitter complementarity principle analogous to the black hole version (Susskind et al., 1993;
’t Hooft, 1995; Banks & Fischler, 2001b; Banks, 2001; Bousso, 2000; Parikh & Verlinde, 2005;
Dyson et al., 2002). While specifics may differ, all versions posit that the physics within a single
causal patch of de Sitter space can be described as a finite-temperature, isolated quantum system,
leading to a finite thermal entropy for the system.

Drawing an analogy to the black hole entropy hypothesis, this perspective suggests that
de Sitter entropy reflects the logarithm of the quantum Hilbert space dimension an observer
within the static patch can utilize to account for inaccessible degrees of freedom beyond the
horizon. Proposals somewhat along these lines have been put forth (e.g., Bousso (2000, 2001);
Banks (2001); Banks & Fischler (2001b); Banks (2005); Banks et al. (2006); Banks & Fischler
(2018); Susskind (2021a,b)), albeit with a subtle distinction.

We now explore the relationship between finite entropy and the dimensionality of the
Hilbert space in de Sitter space. A finite entropy implies discrete energy levels, with a finite
number of levels existing below any given energy threshold Goheer et al. (2003). This is the
crucial point. There is, however, additional reason to believe an upper bound on the energy
levels might exist Banks & Fischler (2001b); Banks (2001). Furthermore, Bousso (2000) has
demonstrated that no experiment in a spacetime with a positive cosmological constant can
extract more information than this.

To illustrate this, consider a de Sitter space containing ordinary particles, fields, and poten-
tially even small black holes, rather than the initially empty space. In this scenario, following
Bekenstein, we define a generalized entropy encompassing both the horizon entropy (A/4GN )
and the ordinary entropy (Sout ) of the matter accessible to the observer:

A
Sgen = + Sout . (2.5)
4GN

For a black hole in asymptotically flat spacetime, both terms (A/4GN and Sout ) can be arbitrar-
ily large. This necessitates an infinite-dimensional Hilbert space to describe possible world states
 2.4 Connecting the Dots 31

as seen by an external observer. This isn’t necessarily true for de Sitter space. Here, the entropy
of a black hole is bounded from above by the entropy of the Nariai solution (Nariai & Ishihara,
1983), the largest black hole that can fit within the cosmological horizon. Consequently, one
cannot have excitations with arbitrarily large entropy (Klemm & Vanzo, 2004).

It’s been argued that the maximum possible value for Sgen coincides with the value for empty
de Sitter space (Bousso, 2000, 2001; Maeda et al., 1998). While Sout can certainly increase by
considering a non-empty static patch state, the claim is that this reduction in A/4GN outweighs
the rise in Sout . The proposition, therefore, suggests that empty de Sitter space maximizes the
entropy of any static patch state. Here, “empty de Sitter space” refers to the generalization,
incorporating gravity, of the natural de Sitter invariant state for quantum fields in de Sitter space
(often called the Bunch-Davies state) (Chernikov & Tagirov, 1968; Allen, 1985; Mottola, 1985).

Consequently, for small GN , the maximum achievable entropy of a static patch state, includ-
ing particles, fields, potentially contained black holes, and degrees of freedom somehow linked
to the cosmological horizon, is AdS /4GN , where AdS represents the horizon area of empty de
Sitter space (for the one-loop correction to the AdS /4GN formula for empty de Sitter space
entropy, (see e.g., Anninos et al. (2022)). The existence of a maximum entropy state in de Sitter
space allows for an interpretation of de Sitter entropy that lacks a direct analog for black hole
entropy.

This perspective suggests that a Hilbert space with a dimension roughly equal to ∼ exp(AdS /4GN )
suffices to describe all possible states of the static patch, including any matter and black holes it may
contain, along with the cosmological horizon (Bousso, 2000, 2001; Banks, 2001; Banks & Fischler,
2001b; Banks, 2005; Banks et al., 2006; Banks & Fischler, 2018; Susskind, 2021a,b). Empty de
Sitter space would then be described by the maximally mixed state on this finite-dimensional
Hilbert space (Dong et al., 2018; Lin & Susskind, 2022). Conversely, in the case of a black
hole, the Bekenstein-Hawking entropy (A/4GN ) potentially determines the size of a Hilbert
space that describes the black hole from an external perspective, but this Hilbert space does not
encompass particles and fields situated outside the black hole horizon.

It is also worth noting that there are model-independent arguments suggesting that the
Hilbert space of quantum gravity is locally finite-dimensional. In other words, in the true
theory of nature, which includes gravity, any local region (when it is properly defined) is char-
acterized by a finite-dimensional factor of Hilbert space (Bao et al., 2017b).

All of this hints that our understanding of spacetime symmetries may not be fundamen-
tal, but rather emergent phenomena arising from a more fundamental, purely quantum theory
(e.g., Cao & Carroll (2018); Cao et al. (2017); Carroll (2021); Carroll & Singh (2018); Cotler
et al. (2019); Dyson et al. (2002)). This perspective challenges the current paradigm where
these symmetries are considered foundational. If true, then the familiar symmetry groups we
currently utilize might be effective descriptions at a specific energy scale, potentially breaking
32 Why Finite Dimensional Hilbert Space?

down or acquiring new features at different scales.

Carroll & Singh (2021) suggests that within this emergent framework, quantum fields could
be interpreted as effective descriptions of pointer observables. These pointer observables would
emerge from the way a quantum system interacts with its environment. The system-environment
split would be optimized to maximize notions of locality, predictability, and robustness against
decoherence (see also related ideas in Cotler et al. (2019)). This perspective offers a potential
explanation for why these concepts, which are crucial for our macroscopic understanding of
the world, appear so fundamental.

However, translating these intriguing ideas into concrete models, particularly in areas like
cosmological physics, remains an ongoing challenge. Existing models often rely heavily on
the notion of spacetime symmetries. New frameworks are needed to incorporate emergent
concepts into these models.

The approaches outlined in Cao et al. (2019), Singh & Carroll (2020), and Friedrich et al.
(2022) for constructing finite-dimensional quantum fields provide a promising starting point for
such model development (see also alternative approaches in, e.g., Cao & Lackey (2021); Bao et al.
(2017a)). These approaches offer potential avenues for exploring how spacetime symmetries and
other familiar concepts might emerge from a more fundamental quantum description.

Further research is necessary to explore the viability of these emergent frameworks and their
implications for our understanding of space, time, and the fundamental laws of physics.
 The Standard Way
3
of Doing Things

3.1 Scalar field in Minkowski Background

While a de Sitter background would be ideal for studying the finite-dimensional quantum
field, the complexity of calculations necessitates a simpler starting point. Investigating a real
scalar quantum field in a non-expanding Minkowski background offers a valuable foundation
for exploring the effects of holography and finite-dimensional Hilbert spaces. As will be demon-
strated later, even this seemingly straightforward scenario presents significant analytical chal-
lenges. Nonetheless, this initial analysis serves as a stepping stone for future advancements in
this area.

In the following, we provide a concise overview of the standard, infinite-dimensional con-

struction of a real scalar field. We begin with the action:

1Z 4
 
S= d x η αβ ϕ,α ϕ,β − m2 ϕ2 , (3.1)
2

where η αβ denotes the components of the inverse Minkowski metric tensor and m represents
34 The Standard Way of Doing Things

the field’s mass. We employ the following form for the metric:

ds2 = dt2 − dx2 where, x ≡ (x1 , x2 , x3 ) . (3.2)

Applying this metric, the action becomes:

1Z 3 ∂ϕ
Z   
S= dt d x ϕ̇2 − (∇ϕ)2 − m2 ϕ2 where, ϕ̇ = . (3.3)
2 ∂t

The expression within the brackets represents the Lagrangian:

1Z 3
 
L = d x ϕ̇2 − (∇ϕ)2 − m2 ϕ2 . (3.4)
2

To derive the Hamiltonian of the field, we begin by obtaining the conjugate momentum field
∂L
from the Lagrangian. It follows directly that π = = ϕ̇. The Hamiltonian then takes the
∂ ϕ̇
form:
" #
Z
π 2 (∇ϕ)2 + m2 ϕ2
H = 3
dx + . (3.5)
2 2

It will be better if we rewrite the Hamiltonian in terms of the Fourier modes of the fields:
Z
d3 k 1 X
ϕ(x) = 3
ϕk eik·x ≈ 3 ϕk eik·x ,
(2π) L ⃗
k
Z
d3 k 1 X
π(x) = 3
πk eik·x ≈ 3 πk eik·x , (3.6)
(2π) L ⃗
k

where we substituted d3 k → ∆k 3 = (2π/L)3 and the summation extends over all k = (k1 , k2 , k3 )
with ki ∈ {2πn/L|n ∈ Z}.

The horizon size introduces a natural infrared cutoff. Since we’re studying the field within the
observer’s static patch, we approximate its finiteness by considering a box of physical length
L. We acknowledge that a spherical boundary for the Universe is a more realistic scenario,
achievable through expansion in terms of 3D Zernike polynomials (Janssen, 2015). However,
we anticipate minimal qualitative discrepancies arising from this simplification. While a rigor-
ous examination based on these polynomials merits future work, for now, we can treat L as the
radius of a spherically bounded Universe.

Since ϕ(x) and π(x) are real scalar fields, their Fourier modes satisfy certain reality con-
ditions. Specifically, ϕ†k = ϕ−k and πk† = π−k (Peskin & Schroeder, 1995). Expressing the
Hamiltonian in terms of these Fourier modes yields:
" # " #
Z
d3 k |πk |2 (|k|2 +m2 )|ϕk |2 X 1 |πk |2 (|k|2 +m2 )|ϕk |2
H = + ≈ + . (3.7)
(2π)3 2 2 k L
3 2 2
 3.2 Quantization of the Field 35

This form resembles the Hamiltonian of a collection of harmonic oscillators. To strengthen

this analogy, we define ϕk = Ak + iBk and πk = Ck + iDk , where Ak , Ck are even and Bk , Dk
are odd functions of k (Padmanabhan, 2014). Subsequently, we introduce new fields based on
these definitions:

√ Ak ,

for k1 ≤ 0
qk = 2 (3.8)
B ,
k for k1 > 0

√ C k ,

for k1 < 0
pk = 2 (3.9)
D
k, for k1 ≥ 0


This leads to the following expression for the Hamiltonian:

" #
1 p2k (|k|2 +m2 )qk2
H =
X
+ . (3.10)
k L3 2 2

For future calculations, it’s convenient to introduce dimensionless versions of pk and qk :

Qk = qk /L2 ,
Pk = pk /L . (3.11)

Incorporating these definitions, the Hamiltonian of the scalar field becomes:

" #
Pk2 L(|k|2 +m2 )Q2k
H =
X
+ ,
k 2L 2
" #
Pk2 M Ω2k Q2k
=⇒ H =
X
+ . (3.12)
k 2M 2
q
Here, we define M = L and Ωk = |k|2 +m2 to make the Hamiltonian formally resemble a set
of harmonic oscillators, each corresponding to a specific Fourier mode k.

3.2 Quantization of the Field

The standard procedure for obtaining the quantum theory of the scalar field involves pro-
moting Pk and Qk to Hermitian operators acting on a Hilbert space. These operators then
satisfy the following Canonical Commutation Relation (CCR) (Peskin & Schroeder, 1995):

[Q̂k , P̂k′ ] = iδk,k′ . (3.13)

36 The Standard Way of Doing Things

The dynamics of the quantized field is governed by the Hamiltonian operator:

" #
P̂k2 M Ω2k Q̂2k
Hˆ (t) =
X
+ . (3.14)
k 2M 2

An intriguing quantity is the vacuum (or zero-point) energy of this field. The minimum
eigenvalue (or energy) of the Hamiltonian at any time t is:
q
|k|2 +m2
λmin [Hˆ ] =
X
. (3.15)
k 2

This represents the sum of zero-point energies (i.e., Ωk /2) of individual oscillators associated with
each mode k. The corresponding vacuum energy density of the field is:
q
X 1 |k|2 +m2 1 Z d3 k q 2
ϵvac = ≈ |k| +m2 . (3.16)
k L3 2 2 (2π)3

The integral in Equation (3.16) is evidently divergent. A common approach is to introduce an

ultraviolet momentum cutoff, ΛU V , representing the scale at which the effective theory becomes
inapplicable. This integral then diverges as Λ4U V , known as the quartic divergence of zero-point
energy (Peskin & Schroeder, 1995; Padmanabhan, 2014). The specific value of ΛU V remains
unknown; it could be the Planck scale, the string scale, or even the supersymmetry breaking
scale. Nevertheless, assuming ΛU V ∼ MP l , the highest known energy scale, we obtain ϵvac ∼
MP4 l ∼ 2.22 × 1076 GeV .

This is where the crux of the problem lies. While the vacuum energy acts as a constant
additive term in any system and holds minimal significance in non-gravitational physics (since
only energy differences are observable), it plays a critical role when gravity is considered.
Here, the energy itself, not the difference, acts as the source of gravity. Furthermore, the
vacuum energy is incorporated into the Einstein equations as a cosmological constant term.
Cosmological observations suggest that the observed vacuum energy density, ϵvac, obs ∼ 0.7ρc ≃
4.3 × 10−47 GeV, is significantly lower than the critical density of the Universe, ρc (Riess et al.,
1998). This represents a discrepancy of approximately 123 orders of magnitude compared to field-
theoretic calculations! This is the essence of the ‘‘cosmological constant problem” discussed
earlier.

The sharp momentum cutoff approach to regularize the energy expression has limitations.
This method breaks Lorentz invariance, a fundamental symmetry of the underlying theory
(Akhmedov, 2002; Martin, 2012). When a spatial momentum cutoff is applied, the vacuum
expectation values of the field’s energy density, ⟨ρϕ ⟩, and pressure, ⟨pϕ ⟩, no longer satisfy the
expected relation ⟨ρϕ ⟩ = −⟨pϕ ⟩ arising from Lorentz symmetry (Martin, 2012).
 3.2 Quantization of the Field 37

Therefore, a regularization scheme that preserves Lorentz invariance is necessary for a mean-
ingful evaluation of the zero-point energy density (Akhmedov, 2002; Ossola & Sirlin, 2003;
Koksma & Prokopec, 2011). One option is dimensional regularization. However, for formulat-
ing a finite-dimensional quantum theory, assuming the continuum theory’s symmetries (sup-
ported only in an infinite-dimensional Hilbert space) also hold at the finite-dimensional level is
not entirely justified. We should primarily focus on ensuring crucial physical properties, like
Lorentz symmetry, emerge in the final theory (see e.g., Barceló et al. (2016); Güijosa & Lowe
(2004); Mathur (2020)).

For simplicity and to maintain tractable calculations, we retain the sharp momentum cutoff
here. Consequently, the modified Hamiltonian operator and its minimum eigenvalue become:
" #
P̂k2 M Ω2k Q̂2k
Hˆ (t) =
X
+ , (3.17)
k<ΛU V
2M 2
q
|k|2 +m2
λmin [Hˆ ] =
X
. (3.18)
k<ΛU V
2

It’s important to note that even though we’ve discretized the Fourier modes by restricting
the field within the horizon and imposing a cutoff on the highest mode (resulting in a finite
number of harmonic oscillators), the theory remains infinite-dimensional due to the Canon-
ical Commutation Relation (CCR) in Equation (3.13). No finite-dimensional representation
of the CCR exists in any Hilbert space H. This means there are no bounded operators P̂k and
Q̂k on H satisfying the relation [Q̂k , P̂k′ ] = i (see Appendix A.1 for details).

The question then arises: how can we construct a finite-dimensional quantum theory for the
scalar field if the conjugate operators cannot be accommodated in a finite-dimensional Hilbert space?
This is precisely the subject of the next chapter, where we explore the Weyl commutation rela-
tion, an equivalent form of the CCR that admits a finite-dimensional representation. General-
ized Pauli Operators (GPOs) serve as the essential replacements for the position and momentum
operators in this context.
 Finite Dimensional Field
4
As discussed in the previous chapter (see also Appendix A.1), the Canonical Commutation
Relation (CCR) (Equation (3.13)) does not admit a finite-dimensional representation within an
arbitrary Hilbert space H. This can be demonstrated by taking the trace of both sides in Equation
(3.13) and exploiting the cyclic property of the trace map. In a finite-dimensional representation,
the trace on the left-hand side would be zero, while the right-hand side would represent the
dimension of the Hilbert space, leading to a clear contradiction.

The Stone-von Neumann theorem guarantees the equivalence (up to unitary equivalence)
of any two irreducible representations of the CCR (Hall, 2013). However, as highlighted in Ap-
pendix A.1, operators satisfying the CCR are necessarily unbounded. This necessitates careful
consideration of the operator domains when working with the CCR.

Specifically, if ∆ψ O denotes the uncertainty of operator O in state ψ, it’s possible to con-

struct counterexamples where operators satisfy the CCR (i.e., [A, B] = i) but violate the
uncertainty principle, i.e., (∆ψ A)(∆ψ B) = 0 for some state ψ. This contradicts the corollary
1
stating that (∆ψ A)(∆ψ B) ≥ 2
(see Section 12.2 of Hall (2013)).

Therefore, we require an additional condition to distinguish “good” and “bad” cases of the
CCR. One approach is to utilize the Weyl relations, which express the CCR in an exponential
form. These exponentiated operators are both bounded and unitary. However, the equivalence
between the Weyl relations and the standard CCR is not mathematically rigorous due to the
40 Finite Dimensional Field

unbounded nature of the CCR operators.

Consider the Canonical Commutation Relation (CCR) for self-adjoint operators Qk and Pk
acting on separable Hilbert spaces (up to unitary equivalence): [Qk , Pk ] = i. Stone’s theorem
guarantees a one-to-one correspondence between self-adjoint operators and strongly continu-
ous one-parameter groups (Hall, 2013). This allows us to express the CCR in terms of the
corresponding unitary groups eitQk and eisPk :

eitQk eisPk = e−ist eisPk eitQk . (4.1)

Conversely, for any two one-parameter unitary groups U (t) and V (s) satisfying the braiding
relation:
U (t)V (s) = e−ist V (s)U (t) ∀s, t (4.2)

one can show by differentiating at t = 0 that the infinitesimal generators satisfy the CCR. This
formulation of the CCR using one-parameter unitary groups is known as the Weyl form of the
CCR.

However, due to the unbounded nature of the operators in the CCR (A.1), constructing
counterexamples is possible where operators satisfy the CCR but violate the Weyl relation,
requiring careful consideration of domain assumptions. Hall (2013) provides an example of
such a scenario (see Example 14.5). In well-behaved (“good”) cases, though, we expect operators
satisfying the CCR to also satisfy the Weyl relations.

We will now explore how the representation of relation (4.2) can be used to construct finite-
dimensional versions of position and momentum operators.

4.1 Representation of Weyl Relation

Building on the equivalence between the Weyl form (Equation (4.2)) and the CCR (Equa-
tion (3.13)) established earlier (with the caveat of proper domain considerations), we now focus
on the Weyl form. Unlike the CCR, the Weyl form admits a unique finite-dimensional repre-
sentation (up to unitary equivalence) as guaranteed by the Stone-von Neumann theorem (Hall,
2013). This section aims to obtain that representation and justify its use for finite-dimensional
position and momentum operators.

Inspired by prior work on finite-dimensional representations of the Weyl form (Santhanam

& Sinha, 1976; Jagannathan et al., 1981; Jagannathan & Santhanam, 1982; Singh & Carroll,
2020), we leverage the framework of Generalised Clifford Algebra (GCA) to achieve this
representation. We provide a brief overview of the key results relevant to our finite-dimensional
field theory.
 4.1 Representation of Weyl Relation 41

Consider a finite-dimensional Hilbert space Hk with dimension dk < ∞. We equip the space
of linear operators L(Hk ) acting on Hk with two unitary operators, Âk and B̂k , satisfying the
following algebra:
Âk B̂k = ω −1 B̂k Âk , (4.3)
2πi
 
where ω = exp , the dk -th primitive root of unity. Iterating this relation, we can show
dk
that Âkk and B̂kl satisfy:
Âkk B̂kl = ω −kl B̂kl Âkk . (4.4)

This implies that Âdkk commutes with B̂k and vice versa. Furthermore, if the representation is
irreducible, Schur’s lemma dictates that:

Âdkk = 1k , B̂kdk = 1k . (4.5)

By choosing an appropriate normal coordinate system, we can diagonalize B̂k :

B̂k = diag(1, ω, ω 2 , · · · , ω dk −1 ) . (4.6)

Consequently, the spectra of Âk and B̂k are identical:

spec(Âk ) = spec(B̂k ) = {1, ω, ω 2 , · · · , ω dk −1 } . (4.7)

While specifying the dimension of the Hilbert space Hk is sufficient to determine the commu-
tation relation and operator spectra, the Generalised Clifford Algebra (GCA) can be applied
to both even and odd dimensional spaces. For convenience in labeling the eigenvalues symmet-
rically around zero, we choose an odd dimension dk = 2lk + 1 where lk ∈ Z+ . This allows
us to use indices i, j, k ∈ {−lk , −lk + 1, . . . , 0, . . . , lk − 1, lk } instead of {1, 2, . . . , dk }. The
calculations performed here for the odd dimension can be adapted to even dimensions with a
corresponding adjustment in the labeling scheme.

With this new indexing, the eigenspectra of the operators become:

spec(Âk ) = spec(B̂k ) = {ω −lk , ω −lk +1 , . . . , ω −1 , 1, ω, . . . , ω lk −1 , ω lk } (4.8)

This indexing scheme provides a more streamlined presentation.

42 Finite Dimensional Field

When B̂k is diagonal in a particular basis, as described by Jagannathan (2010), the matrix
representations of Âk and B̂k become:
   
0 0 0 ··· 1 ω −lk 0 0 ··· 0
   
1

0 0 · · · 0


 0 ω −lk +1 0 ··· 0

   
Âk = 0

1 0 ··· 0  , B̂k =

 0 0 ω −lk +2 · · · 0 
 . (4.9)
 .. .. .. .. ..  .. .. .. .. ..
   
. . . . . . . . . .
 
 
   
0 0 ··· 1 0 dk ×dk
0 0 0 · · · ω lk dk ×dk

If {|bki ⟩} represents the eigenbasis of B̂k , the matrix elements can be expressed concisely as:

[Âk ]ij = ⟨bki |Âk |bkj ⟩ = δi,j+1 ; [B̂k ]ij = ⟨bki |B̂k |bkj ⟩ = ω i δi,j . (4.10)

Here, indices i and j range from −lk to lk , and δi,j is the Kronecker delta function.

Now, we have the eigenvalue equation of B̂k as:

B̂k |bkj ⟩ = ω j |bkj ⟩ . (4.11)

With the above in place, we now consider the action of Âk on the eigenstates of B̂k . For the
state Âk |bkj ⟩, we have:

B̂k (Âk |bkj ⟩) = (B̂k Âk )|bkj ⟩ = (ω Âk B̂k )|bkj ⟩ = ω j+1 Âk |bkj ⟩ . (4.12)

We can infer from the previous equation that Âk |bkj ⟩ is also an eigenstate of B̂k . In fact, it maps
precisely to the state |bkj+1 ⟩. Furthermore, since Âdkk = 1k (identity matrix), we have:

|bklk +1 ⟩ = Âdkk |bk−lk ⟩ = |bk−lk ⟩ . (4.13)

The inherent symmetry in the algebra between Âk and B̂k implies a similar action of B̂k on the
eigenstates of Âk . Consequently, these operators establish a cyclic structure: each operator
induces a cyclic shift in the eigenstates of the other.

This characteristic is significant because it provides us with a set of operators that exhibit
behavior analogous to the conjugate variables qk and pk in the infinite-dimensional Hilbert
space (e.g., e−ix0 pk |x⟩ = |x + x0 ⟩ and eip0 qk |p⟩ = |p + p0 ⟩).

To further emphasize the connection between these operators, we recognize that Âk and
B̂k are related by a Discrete Fourier transform (DFT). Specifically, if Sk represents the DFT
matrix, also known as Sylvester’s circulant matrix, then we have:

Sk Âk Sk−1 = B̂k . (4.14)

 4.1 Representation of Weyl Relation 43

The matrix elements of Sk in the eigenbasis of B̂k are given explicitly by:
q
[Sk ]ij = ω jk / dk . (4.15)

Now, it is clear that building on the established properties of Âk and B̂k , we can introduce con-
jugate variables in the finite-dimensional Hilbert space. These conjugate variables are defined
as two self-adjoint operators where each generates a shift in the eigenstates of the other.
As our analysis has revealed, Âk and B̂k exhibit precisely this behavior. Therefore, let’s define
Q̂k and P̂k as conjugate operators on the finite-dimensional Hilbert space:
   
Âk ≡ exp −iαk P̂k , B̂k ≡ exp iβk Q̂k . (4.16)

Here, αk and βk are non-zero real parameters that determine the eigenspacing of Q̂k and P̂k ,
respectively. Since Âk and B̂k are unitary operators, we can infer that the resulting operators Q̂k
and P̂k are also self-adjoint. Furthermore, they act as generators of translations in each other’s
eigenspaces.

The exponential relationship between P̂k and Q̂k with Âk and B̂k implies that they share the
same eigenstates. However, for convenience, we will denote the eigenstates of Q̂k as |qjk ⟩ and
those of P̂k as |pkj ⟩, with j ∈ {−lk , . . . , lk }.

Leveraging the results from Equations (4.8) and (4.16), we can arrive at the following:
!
2π
Q̂k |qjk ⟩ =j |qjk ⟩, (4.17)
dk βk
2π
 
P̂k |pkj ⟩ = j |pkj ⟩ . (4.18)
dk αk

We now investigate the matrix representation of Q̂k within its eigenbasis. From (4.17), this
representation takes the following form:
!
2π
⟨qjk |Q̂k |qjk′ ⟩ =j δj,j ′ . (4.19)
dk βk

The connection between Âk , B̂k , and the DFT matrix Sk can be leveraged to establish a rela-
tionship between P̂k and Q̂k (as demonstrated in Singh & Carroll (2020)). Since Âk = Sk B̂k Sk−1 ,
taking the logarithm of both sides yields:

log Âk = Sk log B̂k Sk−1 . (4.20)

44 Finite Dimensional Field

This equation allows us to arrive at the following relationship between the operators:
!
−βk
P̂k = Sk−1 Q̂k Sk , (4.21)
αk
!
−αk
Q̂k = Sk P̂k Sk−1 . (4.22)
βk

By exploiting this relationship and the eigenbasis of Q̂k , we can compute the matrix represen-
tation of P̂k :

lk
2πi(j − j ′ )n
! !
2π
⟨qjk |P̂k |qjk′ ⟩
X
= n exp , (4.23)
d2k αk n=−lk dk

0,

 if j = j ′
= 2πlk (j − j ′ ) (4.24)
!
iπ

 cosec , if j ̸= j ′
dk αk dk


It’s important to remember that the eigenstates of both Q̂k and P̂k form orthonormal bases for the
finite-dimensional Hilbert space Hk .

Our previous analysis revealed that Q̂k and P̂k exhibit behavior analogous to conjugate vari-
ables (position and momentum) in the infinite-dimensional Hilbert space. We can now establish
a formal connection between these operators and their infinite-dimensional counterparts. How-
ever, it’s crucial to recognize that as of yet, there’s no established association with a physical
length scale for the finite-dimensional operators Q̂k and P̂k .

Justification for this identification goes beyond simply exhibiting conjugate behavior. We
will achieve this justification in two steps. First, we will demonstrate that the commutation
relation between these finite-dimensional position and momentum operators approaches the
standard Canonical Commutation Relation as the dimension of the Hilbert space Hk tends to
infinity. Second, we will construct a finite-dimensional version of the harmonic oscillator and
show that its eigenstates align with those of the infinite-dimensional harmonic oscillator up to
a specific limit. A perfect match throughout wouldn’t be ideal, as we would anyways desire to
retain the ability to distinguish between the finite and infinite-dimensional cases and explore finite-
dimensional effects.
4.2 The commutator of Q̂k and P̂k

It’s well-known that the CCR doesn’t admit a finite-dimensional representation. This fact
motivated us to explore the Weyl form of the CCR, which does allow for a finite-dimensional
counterpart. In the previous section, we explicitly constructed this finite-dimensional represen-
tation using Generalized Pauli Operators. These operators provided a natural way to implement
conjugacy within the finite-dimensional Hilbert space.

However, as highlighted earlier, the commutator of these finite-dimensional operators

cannot be the identity operator. This deviation from the CCR necessitates further investiga-
tion. In this section, we will analyze the commutation matrix of Q̂k and P̂k and explore the
qualitative implications for our theory.

Previously, only the dimension of the Hilbert space, dk , was crucial for obtaining our re-
   
sults. However, examining the expressions for Âk = exp −iαk P̂k and B̂k = exp iβk Q̂k , we
observe the presence of two free parameters, α and β, which determine the eigenspacing of Q̂k
and P̂k .

Building upon the identification in Equation (4.16), we can expand the left-hand side of
the relation in Equation (4.3) using the Baker-Campbell-Hausdorff Lemma. To ensure that the
commutator of Q̂k and P̂k converges to i1k in the infinite-dimensional limit (as required by the
CCR), Singh & Carroll (2020) derived the following constraint on α and β (refer to their work
for details):
2π
αk βk = . (4.25)
dk
Now, with this constrain and using the expressions (4.19) and (4.24) we can calculate the matrix
representation of [Q̂k , P̂k ] in the eigenbasis of Q̂k to be

lk
4π 2 (j − j ′ ) X 2πi(j − j ′ )n
 
⟨qjk |[Q̂k , P̂k ]|qjk′ ⟩ = n exp dk , (4.26)
d3k αk βk n=−lk
lk
2π(j − j ′ ) X 2π(j − j ′ )n
 
= n exp . (4.27)
d2k n=−lk dk

and upon performing the summation, we obtain:


0,

 if j = j ′
⟨qjk |[Q̂k , P̂k ]|qjk′ ⟩ = iπ(j − j ′)

2πlk (j − j ′ )
 (4.28)

 cosec , if j ̸= j ′
dk dk


While the constraint in Equation (4.25) restricts the product of αk and βk , we retain the freedom
to choose one of them independently. As we will demonstrate later, this choice of eigenspacing
46 Finite Dimensional Field

significantly impacts the ground state energy of our finite-dimensional scalar field.

For ease of reference, let’s denote the finite-dimensional commutator as:

[Q̂k , P̂k ] = iẐk . (4.29)

It’s evident from Equation (4.28) that Ẑk deviates from the identity operator. Many core prin-
ciples in quantum mechanics and quantum field theory hinge on the commutator of conjugate
operators being a complex number. As a consequence, several familiar quantum mechanics re-
sults will diverge due to this non-central commutator Peskin & Schroeder (1995); Padmanabhan
(2014).

Throughout the following chapters, the specific elements of Ẑk won’t be crucial for most
discussions, except when running simulations. However, it’s valuable to explore the properties
of this matrix. As observed in Equation (4.28), Ẑk is Hermitian with real-valued elements.
Furthermore, as expected for any finite-dimensional commutator, its trace, Tr(Ẑk ) is zero.

Interestingly, Ẑk exhibits a specific structure: a symmetric Toeplitz matrix. In such matri-
ces, each descending diagonal from left to right maintains a constant value, in our case the main
diagonal consists entirely of zeros. While Toeplitz matrices have a simple structure, determining
their eigenspectrum and eigenbasis can be quite complex (Dai et al., 2009; Böttcher et al., 2017).
This complexity necessitates numerical calculations in the finite-dimensional case to interpret
the results effectively.

We can now demonstrate that the standard canonical commutation relation (CCR) is re-
covered in the infinite-dimensional limit. Let’s consider the case where lk → ∞, with the
eigenspacing of Q̂k and P̂k becoming infinitesimally small while still adhering to the constraint
αk βk = 2π/dk . It’s important to acknowledge that finite-dimensional Hilbert spaces with
dk → ∞ are not isomorphic to infinite-dimensional ones (even for countably infinite dimen-
sions). However, a path exists to retrieve the CCR as dk → ∞.

In the expression for the commutator in Equation (4.28), we substitute n/dk with a contin-
uous variable x ∈ R and replace the summation with an integral with measure dx ≡ 1/dk . This
leads to:
Z ∞
⟨qjk |[Q̂k , P̂k ]|qjk′ ⟩ = 2π(j − j ′ ) dx x exp(2πi(j − j ′ )x) . (4.30)
−∞
 4.3 Finite dimensional scalar field 47

In the infinite-dimensional case, labels j and j ′ become continuous. We can therefore rewrite
the above as:
Z ∞
1 d
 
⟨qjk |[Q̂k , P̂k ]|qjk′ ⟩ = 2π(j − j ′ ) ′
dx exp(2πi(j − j ′ )x) , (4.31)
2πi d(j − j ) −∞
d
= −i(j − j ′ ) δ(j − j ′ ), (4.32)
d(j − j ′ )
= iδ(j − j ′ ) [∵ sδ ′ (s) = −δ(s)] . (4.33)

The final line aligns with the standard Canonical Commutation Relation. This assures
us that our finite-dimensional conjugate operators satisfy the CCR in the infinite-dimensional
limit (i.e., Ẑk → 1k as dk → ∞). Additionally, as demonstrated in Appendix A.2, the infinite-
dimensional limit of P̂k recovers the familiar representation of −id/dx in the eigenbasis of Q̂k . This
strongly supports our treatment of Q̂k and P̂k as finite-dimensional position and momentum
operators, respectively. In the next section, we will quantize our scalar field using the finite-
dimensional commutation relation (4.29).

4.3 Finite dimensional scalar field

Having established finite-dimensional counterparts for position and momentum operators,

a compelling case has been made for their utilization. Consequently, we quantize our field based
on the relation in Equation (4.29) rather than Equation (3.13). The calculations remain largely
unchanged, with the key difference being the inclusion of finite-dimensional Q̂k and P̂k in the
Hamiltonian expression of Equation (3.17), replacing their infinite-dimensional counterparts.

It’s important to note that we use the same notation for both finite and infinite-dimensional
position and momentum operators for convenience, as we will frequently switch between them.
However, the context should clarify which ones are being referred to. The Hamiltonian of our
finite-dimensional field represents the sum of Hamiltonians for finite-dimensional harmonic os-
cillators with a specific frequency, Ωk . Understanding the behavior of individual Hamiltonians
allows us to draw inferences about our scalar field.

Similar to the standard approach in field theory, we aim to determine the eigenspectrum and
eigenvectors of the Hamiltonian governing the field’s dynamics. To achieve this concretely, we
need to fix the free parameters in our construction: dk , αk , and βk for each mode k. However,
due to the constraint in Equation (4.25), fixing only αk suffices.

We will adopt the choice presented in Friedrich et al. (2022). They justify their choice in two
ways. Firstly, they demonstrate that their selection of αk maximizes the ground state energy of
the harmonic oscillator Hamiltonian, making their construction conservative by minimizing
48 Finite Dimensional Field

finite-dimensionality effects. Secondly, they require the ground state wavefunction to exhibit
equal resolution in both “position” and “momentum” space. In this section, we will revisit
their first motivation, as a similar calculation can be applied to our analogous construction.
The second motivation is presented in Appendix A.3.

4.3.1 Finite Dimensional Harmonic Oscillator

We consider the Hamiltonian acting on the Hilbert space factor, Hk , associated with the
mode k (as introduced in Equation (3.14)):

P̂k M Ω2k
Hˆk = + Q̂k . (4.34)
2M 2

As discussed earlier, Q̂k and P̂k are finite-dimensional operators. It’s convenient to express them
as:

Q̂k = αk diag(−lk , · · · , lk ), (4.35)

P̂k = −βSk−1 diag(−lk , · · · , lk )Sk . (4.36)

where Sk denotes the Sylvester’s matrix and dk = dim(Hk ) = 2lk + 1. Introducing L̂k ≡
diag(−lk , · · · , lk ), the Hamiltonian can be rewritten as:

1 (2π)2 −1 2 2
2 M Ωk 2
Hˆk = 2 S L̂ Sk + α L̂k , (4.37)
αk 2M d2k k k k
2
1 (2π)2 2 2
2 M Ωk −1 2
= 2 L̂ + αk Sk L̂k Sk . (4.38)
αk 2M d2k k 2

Here, we’ve utilized the constraint on eigenspacings (refer to Equation (4.25)). Since Sk is
unitary, the eigenvalues of S −1 Hˆk Sk remain unchanged. Furthermore, using the relations in
k
Eq. (4.22), it can be demonstrated that Sk−2 L̂2k Sk2 = L̂2k . Applying this transformation yields the
following form for the Hamiltonian:

1 (2π)2 2 2
2 M Ωk 2
Hˆk = 2 L̂ + α L̂k . (4.39)
αk 2M d2k k k
2

Analysing the two equivalent forms of the Hamiltonian (Equation (4.38) and Equation (4.39)),
we observe that the eigenspectrum remains invariant under the following substitution:

M Ω2k 1 (2π)2
αk → α̃k with α̃k = 2 . (4.40)
2 αk 2M d2k
 4.3 Finite dimensional scalar field 49

This transformation possesses a fixed point where αk = α̃k . Solving for this fixed point value,
we obtain:
s s
2π 2πM Ωk
αk, fix = =⇒ βk, fix = . (4.41)
M Ω k dk dk

Since the eigenvalues are identical for distinct eigenspacings αk and α̃k related by Equation
(4.40), it follows that a plot of the eigenvalue versus the eigenspacing αk will exhibit mirror
symmetry around the fixed point αk, fix with the small caveat that the plot will be compressed on one
side due to the inverse relationship between αk and α̃k . Notably, the fixed point must extremize
each eigenvalue of Hˆk , and consequently, the minimum eigenvalue as well. While a rigorous
proof for the nature of this extremum for the minimum eigenvalue remains elusive, numerical
simulations suggest that αk, fix actually maximizes the minimum eigenvalue. This behavior is
illustrated in the top panel of Figure (4.1).

0.6
α̃ k αk
0.5 dk = 21
λmin, k /Ω k

0.4
0.3
0.2
0.1
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Eigenspacing, α

0.5
0.4
λmin, k /Ω k

0.3 dk = 21
dk = 51
0.2 dk = 81
0.1
0.00 0.05 0.10 0.15 0.20 0.25 0.30
Eigenspacing, α
Figure 4.1: Ground state energy of the finite-dimensional harmonic oscillator Hamiltonian plot-
ted against the eigenspacing αk . Parameters are M = 1 and Ωk = 10 for all plots. (Top panel) The
ground state energy is plotted for a range of αk values with a fixed dimension dk = 21. (Bottom
panel) The ground state energy is plotted for the same range of αk values but for three choices of
dimension: dk = 21, 51, 81.
50 Finite Dimensional Field

We illustrate the minimum eigenvalue of Hˆk for a specific choice of parameters: M = 1,

Ωk = 10, and dimension dk = 21. While this particular dimension might seem small, numeri-
cal verification suggests that the qualitative behavior of the plot remains consistent for signifi-
cantly larger dimensions (and even lower ones). As evident from the plot, the maximum value
approaches Ωk /2, which corresponds to the standard vacuum energy of an infinite-dimensional
harmonic oscillator (Padmanabhan, 2014). The lower panel of Figure (4.1) depicts the mini-
mum eigenvalue as a function of the eigenspacing, αk . It demonstrates that as the dimensionality
of the operators increases, the minimum eigenvalue of the Hamiltonian approaches the standard
value across a wider range of eigenspacings. In simpler terms, finite dimensionality leads to a
suppression of the oscillator’s vacuum energy.

Friedrich et al. (2022) delve deeper into this concept, providing an approximate expression
for the minimum energy of a finite-dimensional harmonic oscillator that closely aligns
with numerical results (Singh & Carroll, 2020). They leverage this finite-dimensional suppres-
sion to construct a finite-dimensional scalar field with a vacuum energy suppressed by a factor
of ∼ 10−60 compared to the standard field theory estimate with a sharp UV cutoff (Equation
(3.2)). Unlike our approach, their scalar field resides in a uniformly expanding background, and
each mode k is initialized in its ground state at a specific time tk with the eigenspacing of the
operators set to αk, fix at that time. As time progresses, the Hamiltonian associated with mode
k changes due to time-dependent M and Ωk . This implies that at later times t, when M (t)
and Ωk (t) deviate sufficiently from their values at tk , the initially chosen eigenspacing (which
was a fixed point for the transformation in Equation (4.40) corresponding to the Hamiltonian
Hˆk (tk )) is no longer optimal for maximizing the vacuum energy of the Hamiltonian Hˆk (t) at
later times. Consequently, the field’s vacuum energy gets suppressed. Through careful selection of
modeling parameters, their construction achieves a suppression factor of ∼ 10−60 . While this
certainly falls short of solving the cosmological constant problem, it hints at potential observ-
able consequences of finite-dimensional Hilbert spaces for quantum fields.

In our construction, we don’t consider an expanding background. Therefore, this fixed-

point choice of the eigenspacing will maximize the minimum energy of our Hamiltonians asso-
ciated with the Fourier modes. We will tackle the question of the Hilbert space dimension, dk ,
in later chapters and for now treat it as a free parameter of our model.

With fixed model parameters, we can now discuss the eigenspectrum of the Hamiltonian
for each Fourier mode k. In the infinite-dimensional case, the eigenvalues and eigenvectors of
the harmonic oscillator Hamiltonian can be obtained analytically (Peskin & Schroeder, 1995;
Padmanabhan, 2014). Utilizing the ladder operators of the infinite-dimensional Hamiltonian
and their commutation relations, one arrives at the following expression for the eigenvalues:

1
 
λvanilla = Ωk n + , n ∈ {0, 1, 2, · · ·} . (4.42)
2
 4.3 Finite dimensional scalar field 51

We refer to this infinite-dimensional spectrum as the ‘‘vanilla spectrum.” Unfortunately,

we were unable to find an analytical expression for the eigenspectrum of our finite-dimensional
Hamiltonian (Equation (4.34)). However, Figure 4.2 presents the eigenspectrum of the Hamil-
tonian (Equation (4.34)) computed for M = 1, Ωk = 10, and dk = 101, with the eigenspacing
of the operators given by Equation (4.41).

140
Vanilla spectrum
120 dk = 101
Eigenvalues, λk /Ωk

100 Spectrum matches till here

80
60
40 Wavefunction matches till here

20
0
0 20 40 60 80 100
Eigen number
Figure 4.2: Eigenvalues of the Hamiltonian Hˆk plotted against eigennumbers for parameters
M = 1, Ωk , and dk = 101. The finite-dimensional spectrum (dashed line) is compared to the
corresponding “vanilla spectrum” (solid line) of the infinite-dimensional harmonic oscillator. A
clear deviation between the two spectra is observed after the marked eigenindex. We have also
marked the maximum number of eigennumbers up to which the infinite-dimensional and finite-
dimensional wavefunctions exhibit good agreement.

Our analysis of the finite-dimensional and infinite-dimensional harmonic oscillator

eigenspectra reveals distinct characteristics, summarized below. These discrepancies can be
attributed to the non-central nature of the commutation relation, [Q̂k , P̂k ] ̸= i1k .

1. The lower eigenvalues of the finite-dimensional oscillator exhibit good agreement with
their counterparts in the infinite-dimensional case. However, the higher-energy portions of
the spectra diverge significantly, as highlighted in Figure 4.2.

2. Unlike the infinite-dimensional oscillator with its unbounded spectrum, the finite-dimensional
one appears to possess a maximum eigenvalue that scales roughly linearly with the Hilbert space
dimension (Singh & Carroll, 2020).
52 Finite Dimensional Field

3. While the infinite-dimensional oscillator boasts a uniformly spaced eigenspectrum, this

property is not preserved in the finite-dimensional counterpart. Beyond a specific eigen-
value (indicated in the figure), the finite-dimensional spectrum exhibits pronounced non-
linearity in eigenvalue spacing.

These discrepancies in the high-energy eigenspectra suggest that probing the discreteness of
space, if such exists, might necessitate high-energy states.

Extensive numerical simulations across various dimensions and parameters provide an em-
pirical observation: the eigenvalues of the finite-dimensional oscillator closely match those of the
infinite-dimensional counterpart up to an eigenindex of approximately 0.8dk . While the cause of
this agreement remains unclear (potentially a property of the construction or a numerical coin-
cidence), we treat it as an empirical result for now.

Given the alignment of the low-energy spectra for both oscillator types, one might antic-
ipate a corresponding match between their eigenfunctions. This expectation is indeed con-
firmed! The eigenfunctions of the finite-dimensional oscillator demonstrate good agreement
with those of the infinite-dimensional oscillator, not precisely up to the eigenindex where the
spectra match, but very close to it. We define “good agreement” as a residual (the difference
between the two wavefunctions) on the order of ∼ 1/10th of the wavefunction’s amplitude.
The point up to which this agreement holds is also indicated in Figure 4.2, and as observed,
it aligns closely with the point of spectral agreement. Figures 4.3, 4.4, and 4.5 further illus-
trate the concordance between the two wavefunctions for the first few states of the respective
Hamiltonians.

We previously established in Section 4.2 that the commutator of Q̂k and P̂k deviates from
the identity operator. However, as the dimension of the Hilbert space, dim(Hk ), approaches
infinity, the commutator converges to the identity operator (i1k ). Interestingly, our numerical
calculations of Zk reveal that even for finite dim(Hk ), the action of Zk on the low-energy
eigenstates of Hˆk closely resembles that of the identity operator! Figures 4.6, 4.7, and 4.8
illustrate the action of Zk (computed for dimension dk = 101) on the first few excited states of
the Hamiltonian whose eigenspectrum is depicted in Figure 4.2.

Inspecting the residual plots in Figures 4.6, 4.7, and 4.8, we can infer that Zk |Ψ⟩ ≈ |Ψ⟩. To
first-order series expansion, there might exist small correction terms expressible as ϵk |Ψ⊥ ⟩, where
ϵk ≪ 1 and ⟨Ψ|Ψ⊥ ⟩ = 0. Numerical evaluation reveals that this “near identity” behavior of Zk
extends up to the eigenindex where the finite-dimensional and infinite-dimensional eigenfunc-
tions exhibit agreement. Figure 4.9 portrays this behavior of Zk by plotting the expectation
value ⟨Ψ|Zk |Ψ⟩ as a function of the eigenindex.

Beyond this “critical” eigenindex, the expectation value ⟨Ψ|Zk |Ψ⟩ exhibits oscillatory be-
havior and diverges with the Hilbert space dimension, dk . This intriguing behavior of Zk can
 4.3 Finite dimensional scalar field 53

®
0.3 |0 , ∞ dim
Wavefunction, Ψ

®
|0 , dim(H k ) = 101
0.2

0.1

0.0
4 3 2 1 0 1 2 3 4
1e 15
2.5
Residual

0.0
4 3 2 1 0 1 2 3 4
x
Figure 4.3: Ground state wavefunction of the finite-dimensional Hamiltonian Ĥk (orange
crosses) for parameters M = 1, Ωk = 10, and dk = 101. The corresponding ground state wavefunc-
tion of the infinite-dimensional harmonic oscillator (solid line) is also plotted. The residual between
the two wavefunctions is shown below the main plot. The residual is of the order of ∼ 10−15 , indi-
cating excellent agreement between the finite-dimensional and infinite-dimensional ground states.

®
0.2 |1 , ∞ dim
Wavefunction, Ψ

®
|1 , dim(H k ) = 101
0.0

0.2

4 3 2 1 0 1 2 3 4
1e 15
Residual

0
5 4 3 2 1 0 1 2 3 4
x
Figure 4.4: First excited state wavefunction of the finite-dimensional Hamiltonian Ĥk (orange
crosses) for parameters M = 1, Ωk = 10, and dk = 101. The corresponding first excited state
wavefunction of the infinite-dimensional harmonic oscillator (solid line) is also plotted. The resid-
ual between the two wavefunctions is shown below the main plot. The residual is of the order of
∼ 10−15 , indicating excellent agreement between the finite-dimensional and infinite-dimensional
ground states.
54 Finite Dimensional Field

®
0.2 |2 , ∞ dim
Wavefunction, Ψ

®
|2 , dim(H k ) = 101

0.0

0.2
4 3 2 1 0 1 2 3 4
1e 15
5
Residual

0
5
4 3 2 1 0 1 2 3 4
x
Figure 4.5: Second excited state wavefunction of the finite-dimensional Hamiltonian Ĥk (orange
crosses) for parameters M = 1, Ωk = 10, and dk = 101. The corresponding second excited
state wavefunction of the infinite-dimensional harmonic oscillator (solid line) is also plotted. The
residual between the two wavefunctions is shown below the main plot. The residual is of the order
of ∼ 10−15 , indicating excellent agreement between the finite-dimensional and infinite-dimensional
ground states.

®
0.3 |0 , dim(Hk ) = 101
Wavefunction, Ψ

®
Zk |0
0.2

0.1

0.0
4 3 2 1 0 1 2 3 4
1e 14
1
Residual

0
1
4 3 2 1 0 1 2 3 4
x
Figure 4.6: The wavefunction Zk |0⟩ (orange crosses) is plotted alongside the state |0⟩ of the
finite-dimensional Hamiltonian Hˆk (solid line). Parameters are M = 1, Ωk = 10, and dk = 101.
The residual between the two wavefunctions is shown below the main plot. The small residual, on
the order of ∼ 10−14 , suggests that Zk acts almost like the identity operator on the vacuum state
|0⟩.
 4.3 Finite dimensional scalar field 55

®
0.2 |1 , dim(Hk ) = 101
Wavefunction, Ψ

®
Zk |1
0.0

0.2

4 3 2 1 0 1 2 3 4
1e 14
1
Residual

0
1
4 3 2 1 0 1 2 3 4
x
Figure 4.7: The wavefunction Zk |1⟩ (orange crosses) is plotted alongside the state |1⟩ of the
finite-dimensional Hamiltonian Hˆk (solid line). Parameters are M = 1, Ωk = 10, and dk = 101.
The residual between the two wavefunctions is shown below the main plot. The small residual, on
the order of ∼ 10−14 , suggests that Zk acts almost like the identity operator on the first excited
state |1⟩.

0.2 ®
|2 , dim(Hk ) = 101
Wavefunction, Ψ

®
Zk |2
0.0

0.2

4 3 2 1 0 1 2 3 4
1e 14
1
Residual

0
1 4 3 2 1 0 1 2 3 4
x
Figure 4.8: The wavefunction Zk |2⟩ (orange crosses) is plotted alongside the state |2⟩ of the
finite-dimensional Hamiltonian Hˆk (solid line). Parameters are M = 1, Ωk = 10, and dk = 101.
The residual between the two wavefunctions is shown below the main plot. The small residual, on
the order of ∼ 10−14 , suggests that Zk acts almost like the identity operator on the second excited
state |2⟩.
56 Finite Dimensional Field

2.0
Wavefunction matches till here
1.5
1.0
0.5
®
Ψ|Zk |Ψ

0.0
Zk behaves almost like 1 k
0.5
­

1.0
1.5
2.0 0 20 40 60 80 100
Eigen number
Figure 4.9: Expectation value of Zk plotted against the eigennumbers of the finite-dimensional
Hamiltonian Hˆk for parameters M = 1, Ωk = 10, and dk = 101. The expectation value ⟨Ψ|Zk |Ψ⟩
is approximately equal to one up to the point where the finite-dimensional wavefunctions match
well with the corresponding infinite-dimensional ones. Beyond this region, the expectation value
deviates from unity, possibly due to the eigenspacing αk becoming too large to accurately resolve
the high-energy wavefunctions.

be understood by examining its matrix structure. As a symmetric Toeplitz matrix, knowledge

of its first row (or column) suffices to determine all other elements. Referring to Eq. (4.28),
we observe that the first-row (or column) elements oscillate due to the cosec(x) function, with
divergence occurring towards the row’s (or column’s) end as the dimension dk increases. In
simpler terms, the bottom-left and top-right corners of the Zk matrix contain elements with
significant magnitudes.

However, since these large-amplitude elements are concentrated near the corners, any wavefunc-
tion that decays rapidly towards the edges (i.e., as x → ∞ ) remains largely unaffected by them.
Consequently, if the bulk of the wavefunction remains relatively constant or oscillates over a
scale larger than the eigenspacing of the operators, the small oscillating elements within the
bulk of Zk have minimal influence on the wavefunction, effectively leaving it unchanged. This
explains why Zk loses its near-identity behavior for higher-energy wavefunctions.

We know from the infinite-dimensional harmonic oscillator that higher-energy eigenstates

oscillate rapidly across position values. If this oscillation width becomes smaller than the eigenspac-
ing of our finite-dimensional operators, then for the aforementioned reasons, Zk demonstrably fails to
act like an identity on them. This is evident in the “tomato-colored” region of Figure 4.9. This
breakdown of the “near-identity” property for high-energy states can be viewed as further sup-
porting our earlier claim that probing the discreteness of space necessitates high-energy states.
 4.3 Finite dimensional scalar field 57

After all, it is only at high energies that we observe significant deviations from the predictions
of standard (infinite-dimensional) quantum mechanics.

4.3.2 Finite Dimensional “Ladder” Operators

The concept of ladder operators is central to the quantum harmonic oscillator. In the
infinite-dimensional case, these operators, denoted by âk and â†k , satisfy the commutation rela-
tion:

[âk , â†k ] = 1k . (4.43)

These operators are defined as follows (in the infinite-dimensional case):

s
M Ωk i
âk := Q̂k + √ P̂k ,
2 2M Ωk
s
M Ωk i
â†k := Q̂k − √ P̂k . (4.44)
2 2M Ωk

However, arguably more significant are the following commutation relations:

[Hˆk , â†k ] = Ωk â†k , [Hˆk , âk ] = −Ωk âk . (4.45)

These relations demonstrate that â†k excites a state to the next energy level, while âk de-excites
it to the previous level, justifying the nomenclature “ladder operators.”

In our finite-dimensional setting, a natural question arises: what are the corresponding
ladder operators? It turns out that the naively defined operators â†k and âk (shown in Equa-
tion (4.44)) do not fulfill the role of ladder operators for the finite-dimensional Hamiltonian. This
becomes evident from the finite-dimensional commutation relations:

[âk , â†k ] = Zk . (4.46)

Ωk Ωk
[Hˆk , â†k ] = Ωk â†k Zk + [Zk , â†k ], [Hˆk , âk ] = −Ωk Zk âk + [Zk , âk ] . (4.47)
2 2

Recall that, in the limit dim(Hk ) → ∞, Zk becomes the identity operator, and we recover
the standard relations (4.43, 4.45). Conversely, the deviation of Zk from the identity operator,
Zk ̸= 1k , is precisely why the finite-dimensional âk and â†k fail to act strictly as ladder operators.

However, as previously observed, for low-energy states, we can treat the finite-dimensional
operators as if they were infinite-dimensional with minimal error in calculations. Therefore,
58 Finite Dimensional Field

for such states where Zk |Ψn ⟩ ≈ |Ψn ⟩, we can approximate the commutation relations (4.47) as
follows:

Ωk
[Hˆk , â†k ]|Ψn ⟩ = Ωk â†k Zk |Ψn ⟩ + [Zk , â†k ]|Ψn ⟩,
2
Ωk
≈ Ωk â†k |Ψn ⟩ + (Zk ↠|Ψn ⟩ − ↠Zk |Ψn ⟩),
2
† Ωk
≈ Ωk âk |Ψn ⟩ + (Zk |Ψn+1 ⟩ − ↠|Ψn ⟩),
2
Ωk
≈ Ωk â†k |Ψn ⟩ + (|Ψn+1 ⟩ − |Ψn+1 ⟩) (if Zk |Ψn+1 ⟩ ≈ |Ψn+1 ⟩),
2
≈ Ωk ↠|Ψn ⟩ . (4.48)

Following a similar logic, if Zk |Ψn−1 ⟩ ≈ |Ψn−1 ⟩ then:

[Hˆk , âk ]|Ψn ⟩ ≈ −Ωk âk |Ψn ⟩ . (4.49)

These approximated commutation relations (4.49, 4.48) allow us to employ â†k and âk as ladder
operators for the low-energy eigenstates of Hˆk . Figures 4.10 and 4.11 illustrate the action of ↠k
on the vacuum state of the finite-dimensional harmonic oscillator, up to the first two excited
states. As expected, the residual plot aligns well with our assumptions.

®
0.2 |1 , dim(Hk ) = 101
Wavefunction, Ψ

®
ak |0
0.0

0.2

4 3 2 1 0 1 2 3 4
1e 15
2.5
Residual

0.0
2.5
4 3 2 1 0 1 2 3 4
x
Figure 4.10: Action of the finite-dimensional creation operator â†k on the ground state. The
resulting excited state (orange crosses) is compared to the first excited state of the finite-dimensional
Hamiltonian Hˆk (solid line). Parameters are M = 1, Ωk = 10, and dk = 101. The small residual
between the two wavefunctions, on the order of ∼ 10−15 , suggests that â†k effectively acts like a
raising operator on the ground state, |0⟩.
 4.3 Finite dimensional scalar field 59

0.2 ®
|2 , dim(Hk ) = 101
Wavefunction, Ψ
®
(ak ) 2 |0
0.0

0.2

4 3 2 1 0 1 2 3 4
1e 14
Residual

0
2
4 3 2 1 0 1 2 3 4
x
Figure 4.11: Action of the finite-dimensional creation operator â†k twice on the ground state.
The resulting excited state (orange crosses) is compared to the second excited state of the finite-
dimensional Hamiltonian Hˆk (solid line). Parameters are M = 1, Ωk = 10, and dk = 101. The
small residual between the two wavefunctions, on the order of ∼ 10−14 , suggests that â†k effectively
acts like a raising operator on the low energy states of Hˆk .

The insights gained from the finite-dimensional operators, Hamiltonian, and commutator
will prove valuable in our upcoming discussion on overlapping constructions. As we will dis-
cover, analytical calculations become increasingly challenging in that context. The discussions
presented here regarding the assumptions under which we can approximate the action of Zk
with the identity operator, the matching between low-energy eigenspectra and eigenstates of
the finite-dimensional and infinite-dimensional harmonic oscillators, and the ladder operator-
like behavior of the finite-dimensional â†k and âk for low-energy states, will significantly simplify
our analysis.

By leveraging generalized Pauli operators, we can construct a finite-dimensional scalar field

that closely resembles an infinite-dimensional field in the low-energy regime, with some modifi-
cations arising from the finite dimensionality of the operators (e.g., suppression of the vacuum
energy) (Friedrich et al., 2022; Cao et al., 2019). Our choice of generalized Pauli operators
(GPOs) stems from their ability to mimic the concept of conjugate operator pairs within the
finite-dimensional Hilbert space. As emphasized in Carroll & Singh (2021), such pairs are cru-
cial for the emergence of quasi-classical Hilbert space factorizations. However, we have not yet
explored whether the GPO construction is the sole method for achieving this duality in finite
dimensions. Additionally, we do not expect the emergent pointer observables identified in Car-
roll & Singh (2021) to be directly expressible in terms of exact GPOs. (See, for instance, Section
V.A of Carroll & Singh (2021); they define the candidate pointer observable as the observable
60 Finite Dimensional Field

of a given Hilbert space factorization that has the smallest commutator with the Hamiltonian.
There is no guarantee that this will be a GPO.) Nevertheless, GPOs provide a foundation for
our investigations into the effects of finite-dimensional Hilbert spaces on scalar quantum fields.

With a finite-dimensional scalar field in hand, the next question naturally arises: does the
maximum number of degrees of freedom of this field scale with the horizon area, as ex-
pected from the Holographic Principle?
 5
Scaling of
Degrees of Freedom

In the previous chapter, we presented our finite-dimensional construction of the scalar field. This
involved initially confining the field within a box of length L, which represents the horizon
radius. This discretization transforms the field’s Fourier modes into a discrete but still infinite
set. To address the resulting divergent vacuum energy, a sharp ultraviolet (UV) cutoff, ΛU V , was
introduced. This step yielded a finite number of Fourier modes for the scalar field.

Finally, to achieve a finite-dimensional theory, we quantized the conjugate pairs, Qk and Pk ,

using the finite-dimensional commutation relation (Equation (4.29)). This relation relies on
Generalized Pauli Operators, which enable the concept of conjugate operators within a finite-
dimensional Hilbert space. This approach was necessary because the standard canonical com-
mutation relation (CCR) lacks a finite-dimensional representation. Furthermore, we observed
that the finite-dimensional harmonic oscillators closely resemble their infinite-dimensional coun-
terparts in the low-energy regime, suggesting the potential to explore the ramifications of finite
dimensionality in high-energy regimes.

However, our objective extends beyond constructing a finite-dimensional scalar field. We

also aim to create a field with a maximum number of degrees of freedom that scales with the
horizon area, as predicted by the holographic principle (Equation (2.3)). Indeed, by construct-
ing a finite-dimensional field, we have significantly reduced the field’s degrees of freedom (from
62 Scaling of Degrees of Freedom

infinite to a finite value). In the following section, we calculate the degrees of freedom associated
with our finite-dimensional field and assess whether they scale with the horizon area. If not, we
will explore potential modifications to achieve this scaling behavior.

5.1 Finite-dimensional field holographic or not?

The maximum number of degrees of freedom of a quantum system is defined as the loga-
rithm (or logarithm with base 2) of the minimum dimension of the Hilbert space required for
the complete description of the system (Bousso, 2002). The way we understand the motivation
behind defining it like this is as follows − Entropy is a measure of our ignorance about the micro-
scopic details of the system. Therefore, more entropy means there’s more we don’t know about
the details of the system, in other words, there are more “parameters” about whose precise
value (or nature) we don’t know. These parameters are what collectively known as the “degrees
of freedom” of the system. Thus, number of degrees of freedom is proportional to the entropy.
The entropy associated to any quantum state of the field is given by S = −kB T r(ρ log ρ),
where ρ is the density matrix of the state and the density matrix of the state which has the
largest possible entropy is given by:

1
ρmax = 1. (5.1)
dim(H)

with entropy, Smax = kB log(dim(H)) and H represent the Hilbert space of the quantum sys-
tem. From the aforementioned reasoning we have the maximum number of degrees of freedom
of a system equal to ln(dim(H)).

Our finite-dimensional scalar field, constructed as described in Chapter 4, possesses a Hilbert

space:
O
H= Hk . (5.2)
k

Each factor Hk represents the Hilbert space of a finite-dimensional harmonic oscillator with
dimension dk . If we consider N Fourier modes of the field within the UV cutoff ΛU V , the
dimension of H becomes:
!
O
dim(H) = dim Hk ,
k
Y
= dim(Hk ),
k
Y
= dk . (5.3)
k
 5.2 Overlapping Degrees of Freedom 63

Assuming all finite-dimensional harmonic oscillators have the same dimension, i.e., dk = d
for all k, the maximum number of degrees of freedom for our field is:
! !
N = ln
Y Y
dk = ln d ,
k k
 
N
= ln d ,
= N ln d, (5.4)
4πΛ3U V /3
≈ ln d,
(2π/L)3
(LΛU V )3
= ∝V. (5.5)
6π 2

While the degrees of freedom scale with the horizon volume, this is not the desired behavior.
How can we modify our field to achieve an area scaling for the maximum number of degrees of
freedom?

5.2 Overlapping Degrees of Freedom

Examining Equation (5.4), the volume scaling arises because N , representing the number of
Fourier modes within ΛU V , scales with the volume. If N could instead scale with the area, then
N would also exhibit the area scaling expected by the holographic principle.

What does it imply for N to scale with area? In essence, it suggests that our theory
effectively possesses n degrees of freedom (where n ∝ Ahorizon , the horizon area) that collec-
tively produce the same effects as if we had N degrees of freedom (where N ∝ V). In simpler
terms, we need a way to represent the physics that, in standard field theory (or our current
finite-dimensional case), would require a Hilbert space of dimension dN using states from a
lower-dimensional Hilbert space with dimension dn .

Our initial assumption was that the dimension d of the Hilbert space for each finite-dimensional
harmonic oscillator is constant across all Fourier modes k. This limited our approach to manip-
ulating N to achieve area scaling for the degrees of freedom. However, there’s no fundamental
justification for assuming uniform dimensions across all harmonic oscillators.

An alternative path to area scaling involves modifying ln d (with units of length) as a function
of L by a factor of ξ (where ξ also has units of length), such that N ln d ∝ Ahorizon (as suggested
in Friedrich et al. (2022); Cao et al. (2019)). Additionally, N and ln d could scale independently,
as long as their product scales with the area. Exploring these possibilities is a subject of future
work.
64 Scaling of Degrees of Freedom

For now, let’s focus on the scenario where the Hilbert space dimension is identical and
denoted by the free parameter d in our theory. In Equation (5.4), we substitute N with n ∝
Ahorizon .

This assumption of constant dimension across Fourier modes aligns our problem with the
task of embedding N qubits into a smaller Hilbert space of dimension 2n (where n < N ), as
discussed in Chao et al. (2017). Leveraging the concept of “overlapping qubits,” Friedrich et al.
(2024) constructed creation and annihilation operators, ĉ†k and ĉk , for the Weyl fermionic field,
satisfying the following anticommutation relations:

{ĉ†k , ĉk } = 1k . (5.6)

|{ĉ†k , ĉk′ }| < ϵ . (5.7)

Here, ϵ represents a small number that decreases as k increases. We aim to achieve a similar
outcome for our scalar field. In the infinite-dimensional limit, instead of having pairwise com-
muting position and momentum operators, [Q̂k , P̂k′ ] = iδk,k′ , acting on a Hilbert space of
dimension dN , we introduce a controlled deviation from this condition. The resulting opera-
tors, [Q̂k , P̂k′ ] ̸= iδk,k′ , will exhibit non-pairwise commutation but with a small parameter ϵ.
Crucially, these operators will act on a smaller Hilbert space of dimension dn .

The following illustration visually depicts this concept (see Figure 5.1). Imagine circles repre-
senting operators and a rectangle symbolizing the Hilbert space of states they act upon. In the top
panel, non-overlapping circles reside comfortably within the rectangle, reflecting the standard
commutation relation shown on the right. The rectangle’s size signifies the dimension of the
Hilbert space. Our aim is to introduce a controlled departure from pairwise commutativity,
visualized by overlapping circles compressed into a smaller rectangle, representing a Hilbert
space of dimension dn . Importantly, we seek to minimize the deviation between commutators
for operators associated with distinct Fourier modes. These commutator values should be
sufficiently large to satisfy the holographic bound but remain small enough to maintain
established particle physics behaviors.

Intuitively, cramming more Q̂k and P̂k operators into a smaller Hilbert space likely increases
the overlap value, ϵ. A crucial question arises: how small can ϵ be?

Conceptually, we can embed all N operators (acting on a dN dimensional Hilbert space) into
a smaller space where they act on states in a dn dimensional space, with n ∝ A. However,
this is an extreme scenario, resulting in non-zero pairwise commutations for all k satisfying
|k|≤ ΛU V . Our goal is to minimally modify the theory.

Consider a shell-wise embedding approach. We can decompose Fourier space into a series
of non-overlapping shells with width ∆s (which can vary between shells) labeled by s. Each
shell contains Ns ≈ 4πks2 ∆s /(2π/L)3 modes (and operators), where ks represents the shell’s
 5.2 Overlapping Degrees of Freedom 65

Non-Overlapping case

Overlapping case

Figure 5.1: Top: Standard commutation (non-overlapping circles, large Hilbert space). Bottom:
Reduced dimensionality with controlled commutativity (overlapping circles, smaller space). This
approach targets area scaling while respecting holographic and particle physics constraints.

mean radius. These Ns operators are squeezed into a Hilbert space of dimension dns , where
ns ≈ 2πks ∆s /(2π/L)2 .

In this strategy, only operators within a specific shell are affected by the embedding, lead-
ing to non-commutativity among themselves but commutativity with operators from different
shells. This avoids breaking commutativity for all operators.

While embedding Ns operators (typically requiring a dNs dimensional space) into a dns di-
mensional space might seem like overcrowding for larger ks , but Cao et al. (2017) suggests that
we can embed many qubits into a smaller space while still maintaining very small pairwise
commutator norms. This is supported by the Johnson-Lindenstrauss theorem, which guaran-
tees the existence of exponentially many nearly orthogonal vectors in high-dimensional spaces
with small pairwise overlaps (for more details, see Appendix B.1). Notably, even exponential
overcrowding is not necessary. A simple calculation shows:

Ns
∝ ks2 ; =⇒ ; Ns ∝ n2s . (5.8)
∆s

Therefore, only polynomial overcrowding is required. We can effectively compress the oper-
ators into a lower-dimensional space while maintaining very small pairwise commutators.
5.3 Non-local Commutation Relation

Our embedding construction, which will be detailed in the next chapter, introduces a poten-
tial concern regarding locality violation in Local Quantum Field Theory (LQFT). In LQFT,
causality is enforced by requiring spacelike separated operators to commute. Mathematically,
for two field operators O1 (x) and O2 (y) at spacetime points x and y, their commutation en-
sures that the order of measurement does not affect the outcome. This makes sense because if
x and y are spacelike separated, there’s insufficient time for a signal to propagate information
about the first measurement to the other point, thereby keeping the state measurement indepen-
dent. This locality principle implies that measuring O1 (x) cannot influence the measurement
of O2 (y), and vice versa. Additionally, at the same time (tx = ty ), operators must commute
unless their spatial positions coincide.

However, our overlapping construction explicitly breaks this microcausality condition. We

encounter non-zero commutators [Q̂k , P̂k′ ] ̸= 0 even for distinct momenta k ̸= k′ . This seem-
ingly introduces non-locality into our framework.

Intriguingly, such behavior is anticipated for gauge-invariant observables in quantum

gravity, as discussed in Donnelly & Giddings (2016a,b). The authors constructed leading-
order gauge-invariant observables in GN and explicitly calculated commutators to obtain leading-
order corrections. These operators exhibited deviations from local commutation relations due
to “gravitational dressings,” significantly departing from LQFT commutators in regions pre-
viously constrained by locality bounds established in Giddings & Lippert (2001, 2004); Gid-
dings (2006). This characteristic appears to be a universal weak-field feature of any quantum
gravity theory that recovers Einstein’s theory in the weak-field regime, potentially providing
valuable insights into the complete theory of quantum gravity (Berglund et al., 2023; Donnelly
& Giddings, 2016a).

It’s important to clarify that we are not claiming a direct correspondence between the
nature of our commutator and a specific feature of quantum gravity. Instead, we leverage
the expectation of non-commutativity in QG observables to explore the physical implications
of our non-commuting operators and investigate their potential connection to non-locality in
quantum gravity.
 Overlapping
6
Quantum Scalar Field

This chapter builds upon the concept of overlapping degrees of freedom introduced in Chap-
ter 5. We leverage the idea of overlapping qubits presented in Cao et al. (2017), where N qubits
were embedded in a Hilbert space of dimension 2n (n < N ).

Recall that for N qubits in a Hilbert space, a minimum dimension of 2N is required for
each qubit to be independent. This ensures the anti-commutation relation {Si , Ti } = 0 for
all i and the pairwise commutation relation [Si , Tj ] = 0 for all i ̸= j. However, Cao et al.
(2017) constructed N qubits in a lower-dimensional space (2n , n < N ) while preserving the
anti-commutation relation for each qubit ({S̃i , T̃i } = 0 ∀i) but allowing for almost pairwise
commutation among them (||[S̃i , T̃j ]||≤ ϵ). This “failure to pairwise commute” defines over-
lapping qubits.

Friedrich et al. (2024) utilized overlapping qubits to suppress degrees of freedom and construct
a finite-dimensional Weyl field satisfying the cosmic Bekenstein-entropy bound. They then explored
the consequences of such a fermionic field. In this thesis, we extend their work to bosons and
develop the necessary tools for further investigation.

Bosons introduce unique complexities. In the fermionic case, the Hamiltonian expressed
in terms of Fourier modes can be recast as the Hamiltonian of non-interacting spins in a k-
68 Overlapping Quantum Scalar Field

dependent magnetic field (Friedrich et al., 2024). Each Fourier mode is analogous to a spin,
similar to how a scalar field can be represented as a collection of harmonic oscillators for each k
mode. However, the spin Hilbert space dimension is 2, resulting in a factor of 4 (including both
particle and anti-particle sectors) for each Fourier mode’s Hilbert space. Unlike the scalar field,
we have a finite product of finite-dimensional Hilbert spaces here. For N Fourier modes (k <
ΛU V ) with a 4-dimensional Hilbert space factor Hk associated with each mode, the total Hilbert
NN
space of the theory is Htotal = k=1 Hk , with a dimension of 4N . This is a finite-dimensional
theory to begin with, unlike the infinite-dimensional theory arising from the infinite dimensionality
of each quantum harmonic oscillator’s Hilbert space (as discussed in Chapter 3). Chapter 4 addressed
this issue by making the theory finite-dimensional.

Furthermore, as highlighted in Chapter 5, even the finite-dimensional theory exhibits de-

grees of freedom that do not conform to the Bekenstein-entropy bound. This is also true for
the standard fermionic field for reasons detailed in Chapter 5. To address this volume scaling of
degrees of freedom, Friedrich et al. (2024) employed overlapping qubits. Since each Fourier mode
in the fermionic field corresponds to a spin (or qubit), reducing the field’s degrees of freedom
involved lowering the Hilbert space dimension. They achieved this by “squeezing” the qubits
present in a Fourier space shell with radius ks (say Ns qubits) into a Hilbert space of dimension
4ns (where ns is the effective number of qubits required for area scaling of the degrees of free-
dom). We will adopt a similar approach for our scalar field to enforce the Bekenstein-entropy
bound, albeit with some modifications. The construction of the “overlapping” position and
momentum operators, and their incorporation into the field to achieve area scaling of the effec-
tive degrees of freedom, will be explained in detail below.

6.1 Overlapping Position and Momentum Operators

In chapter 4, we used Generalised Pauli Operators to model the finite-dimensional version

of position and momentum operators. In the following we are going to use these to construct
what we call “overlapping” position and momentum operators with which we will replace the
non-overlapping Q̂k and P̂k operators in the Hamiltonian (3.17) to obtain the Hamiltonian for
our holographic scalar field. Below are the steps of our construction (also see Figure 6.1)):

1. In the discretized Fourier space, consider a shell of width, ∆s with central radius being
ks . These shells labelled by {s} covers the entirety of Fourier space with no overlapping
of the adjacent shells. Let Ns denote the number of Fourier modes present in it. This
 6.1 Overlapping Position and Momentum Operators 69

number can be approximated as:

(2π)3

Ns ≈ 4πks2 ∆s (6.1)
L3

It is because of this factor, ks2 ∆s that we obtained the volume scaling of our “indepen-
dent” degrees of freedom (see Section 5.1). We would want to have a scaling like ks ∆s .
For this, we will squeeze our position and momentum operators into a Hilbert space of
dimension, dns , where ns is approximately given by:

(2π)2

ns ≈ 2πks ∆s (6.2)
L2

Nns
2. Consider the Hilbert space, Hs = i=1 Hi , where each factor, Hi represents the Hilbert
space of a finite-dimensional harmonic oscillator. The s in the subscript of Hs labels the
Fourier space shell we are referring to and therefore, the notation Hs denotes the effective
Hilbert space corresponding to the shell-s. Note, that later as we will see we may not be
able to factorise, Hs as we can now but we will keep using this notation to refer to the
Hilbert space of the shell, quite generally. Let {qi , pi }ni=1
s
represents the corresponding
position and momentum operators (matrices) in the factors, Hi . Clearly, in Hs these
operators will look like the following:

P1 = p1 ⊗ 1 ⊗ · · · ⊗ 1
P2 = 1 ⊗ p2 ⊗ · · · ⊗ 1
..
.
Pns = 1 ⊗ · · · ⊗ 1 ⊗ pns (6.3)

and similarly,

Q1 = q1 ⊗ 1 ⊗ · · · ⊗ 1
Q2 = 1 ⊗ q2 ⊗ · · · ⊗ 1
..
.
Qns = 1 ⊗ · · · ⊗ 1 ⊗ qns (6.4)

3. In the space Rns , we draw Ns random vectors from a standard normal distribution and
normalize them. Let’s represent these normalized random vectors as {vk }. The label, k
means this vector is associated with the mode, k.
70 Overlapping Quantum Scalar Field

4. We define the “overlapping” position and momentum operators as:

ns
X
Q̃k = ⟨vk |ei ⟩ Qi (6.5)
i=1
Xns
P̃k = ⟨vk |ei ⟩ Pi (6.6)
i=1

here, {ei }ni=1

s
are some orthonormal basis of Rns . For the sake of simplicity we will take
them as standard basis vectors of Rns .

5. These operators can serve as position and momentum operators corresponding to the
Fourier modes in the shell-s
 forour holographic field. We can stitch the operators in
different shells as P̂k = / 1 ⊗ P̃k , i.e., to consider a tensor product of the operator
N
k∈s

P̃k (which is defined on the Hilbert space of the shell that contains k) with the unit
operators in all shells that do not contain k.

Figure 6.1: This cartoon is adapted from Friedrich et al. (2024). (Left sketch) The scalar field
can be decomposed into a collection of harmonic oscillators represented by points k in Fourier
space. In three dimensions, the number of oscillators within a small Fourier space shell s (red
crossed region in the upper sketch) scales as ks2 ∆s , where ks is the radius and ∆s is the width of
the shell. To achieve holographic scaling of the effective degrees of freedom in the field theory, we
aim to embed these qubits into a finite-dimensional Hilbert space that is strictly smaller than the
tensor product of the individual oscillator Hilbert spaces (Right sketch).
6.2 How good are these operators?

We have described the steps to construct what we call “overlapping” position and momen-
tum operators. In this section we will derive some of their properties. Let us first start with
the algebra which these operators follow. We have written our overlapping operators, Q̃k and
P̃k as a weighted linear sum of non-overlapping Qi and Pi operators. Recall that these non-
overlapping operators obey the following commutation relation

Ai = exp(−iαi Pi ); Bi = exp(iβi Qi ) . (6.7)

Ai Bj = ω −δij Bj Ai ,
− 2πi δij
=e dk Bj Ai . (6.8)

where, αi and βi ’s are the eigenspacing of the operators Qi and Pi , respectively and dk is the
dimension of the Hilbert space associated with each mode, k. We have already decided to keep
the dimension of Hilbert same for all the modes but we only keep the eigenspacing of the
operators associated with modes, k within the shell-s same i.e., αi = αs and βi = βs for all
i ∈ {1, · · · , ns }. The reason for this will become clear in the next chapter. Note that all the
operators associated on different Fourier shells by construction commute therefore, we only need to
look at the algebra of operators corresponding to the modes within a shell.

We can exponentiate Q̃k and P̃k to obtain

 
Ãk = exp −iαs P̃k ,
 ns
X 
= exp − iαs ⟨vk |ei ⟩Pi ,
i=21
ns ns
exp(−iαs pi )⟨vk |ei ⟩ .
Y O
= exp(−iαs ⟨vk |ei ⟩Pi ) = (6.9)
i=1 i=1

where in the third step we have used BCH formula and the fact that [Pi , Pj ] = 0 ∀ i, j.
Similarly, using [Qi , Qj ] = 0 ∀ i, j, we will get the following.
ns ns
exp(iβs qi )⟨vk |ei ⟩ .
Y O
B̃k = exp(iβs ⟨vk |ei ⟩Qi ) = (6.10)
i=1 i=1

Since, Pi ’s and Qi ’s are all Hermitian operators so it follows that Ãk and B̃k are still unitary
operators. One can easily see from (6.9), (6.10) and (6.8) that our overlapping operators satisfy
the following commutation relation.
ns
−ζ
X
Ãk B̃k = ω B̃k Ãk , ∀k ∈ s where, ζ = ⟨vk |ei ⟩ . (6.11)
i=1
72 Overlapping Quantum Scalar Field

Clearly, this is not the Generalised Clifford Algebra which the non-overlapping operators
follows. The commutation relation has been “modified” because of the random factors of our
construction. Unfortunately, there does not exist a nice closed form expression for the commutation
between Ãk and B̃l but it is evident from the expressions (6.9) and (6.10) that they also do not com-
mute, which is actually what we expected. Using the results on the moments of ⟨vk |ei ⟩ as derived
in Appendix B.2 we can see that:
ns
X  ns
X
E(ζ) = E ⟨vk |ei ⟩ = E(⟨vk |ei ⟩) = 0 ∵ E(⟨vk |ei ⟩) = 0 ∀vk , ei . (6.12)
i=1 i=1

and
ns
X 
2 2
X
V ar(ζ) = E(ζ ) = E ⟨vk |ei ⟩ + ⟨vk |ei ⟩⟨vk |ej ⟩ ,
i=1 i̸=j
ns
E(⟨vk |ei ⟩2 ) +
X X
= E(⟨vk |ei ⟩⟨vk |ej ⟩),
i=1 i̸=j
ns
X 1 δij
= +0 ∵ E(⟨vk |ei ⟩⟨ej |vk ⟩) = ,
i=1 ns ns
=1. (6.13)

Even though Ãk and B̃k don’t follow the GCA, and since ζ is a random variable so, a naive
guess would be that its mean value must be 1 with its variance falling off as at least ∼ 1/ns . But
the above moments of ζ tells us that this is sadly not the case. We will explore the severity of this
commutation relation on our construction in the next chapter.

In finite-dimensional Hilbert space, Generalised Pauli Operators provides us a way to have

the conjugate operators. Now, that we have overlapped these operators and clearly the new
operators do not obey the GCA, are they still conjugate operators? If not then we have no
motivation to move ahead with this construction. Fortunately, as it turned out that Q̃k and P̃k
forms a conjugate pair. To see this, consider the eigenstates of B̃k :
ns
O  ns ns
O 
⟨vk |ei ⟩
|qjkl ⟩ |qjkl ⟩
O
B̃k = exp(iβs qi ) ,
l=1 i=1 l=1
ns
exp(iβs ql )⟨vk |el ⟩ |qjkl ⟩,
O
=
l=1
ns
ω ⟨vk |el ⟩jl |qjkl ⟩
O
= using (4.11),
l=1
ns
P
ns
⟨vk |el ⟩jl O
= ω l=1 |qjkl ⟩ . (6.14)
l=1
 6.2 How good are these operators? 73

and look at the action of Ãk on these states:

 ns  ns
O 
|qjkl ⟩ = B̃k Ãk |qjkl ⟩ ,
O
B̃k Ãk
l=1 l=1
ns
P
⟨vk |el ⟩ ns
O 
=ω l=1 Ãk B̃k |qjkl ⟩ using (6.11),
l=1
ns
P ns
P
⟨vk |el ⟩ ⟨vk |el ⟩jl ns
O 
=ω l=1 ω l=1 Ãk |qjkl ⟩ ,
l=1
ns
P
⟨vk |el ⟩(jl +1)  ns 
|qjkl ⟩ .
O
= ω l=1 Ãk (6.15)
l=1
ns ns
|qjkl ⟩ = |qjkl +1 ⟩ .
O O
=⇒ Ãk (6.16)
l=1 l=1

The above relation tells us that Ãk shifts the eigenstate of B̃k and the eigenvalue is changed
ns
precisely by the amount ζ = ⟨vk |el ⟩ (the exponent of ω in the braiding relation of Ãk and
P
l=1
B̃k ). Notice, that Al would have shifted only the state |qjkl ⟩ but Ãk shifts all the states by a unit.
In other words, unlike Al , Ãk “talks” with every state in all the tensor product factors. Fur-
thermore, Ãk still maintains the cyclicity of translations since, Ãdkk = 1 (follows immediately
from Adl k = 1k ). Thus, Ãk generates cyclic translations in the eigenstates of B̃k and vice versa.

At the same time, it must also not be forgotten that due to “overlapping” Ãk′ does not
commute with B̃k either. So, even the action of Ãk′ on the eigenstates of B̃k is expected to have
non-trivial effects, however, since, we couldn’t find a neat closed form expression for their com-
mutation relation (at least in finite-dimensions), the precise details of how it affects remains
obscure. But this discussion and (6.16) establishes that Ãk and B̃k forms a conjugate pair, a
property we desire.

Using Ãk and B̃k one can obtain the entire spectrum of Q̃k and P̃k . The eigenvalue equation
of these two are as follows:
ns
O  ns
X ns
 O 
Q̃k |qjkl ⟩ = αs ⟨vk |el ⟩jl |qjkl ⟩ . (6.17)
l=1 l=1 l=1
Ons  Xns  Ons 
P̃k |pkjl ⟩ = βs ⟨vk |el ⟩jl |pkjl ⟩ . (6.18)
l=1 l=1 l=1

The eigenvalues of Qi ’s are discrete and goes from −αs lk to αs lk (in unit steps), each with
degeneracy of dnks −1 that comes from the factor of identity, 1 which is in tensor product with qi
(see the definition (6.4)). However, after the overlapping, as is evident from the eigenvalues of
Q̃k (6.17), this degeneracy has been broken because of the random factors. As a consequence,
new eigenvalues has “emerged” in between the eigenvalues of Qi ’s. This is because our “over-
lapping” operators don’t follow the same algebra which the non-overlapping ones follow. To
74 Overlapping Quantum Scalar Field

make this point even clear in figure (6.2), we have plotted the eigenvalues of Q1 and of Q̃k for
some choice of our model parameters. From the plot one can see that the eigenvalues of Q̃k
fills the gap between the eigenvalues of Q1 . Although, we have shown the plot of eigenvalues of
position operator only but the eigenvalues of momentum operator behaves in the exact same way.

2 Eigenvalues of Q
Eigenvalues of Q̃k
1
Eigen values

2
0 1000 2000 3000 4000 5000 6000 7000
Eigen number
Figure 6.2: The plot compares the eigenvalues of the non-overlapping position operator Qk (blue
solid line) with those of the overlapping position operator Q̃k (orange solid line). The overlapping
operator is constructed with the following parameters: ns = 3, Ns = 4, dk = 19, Ωk = 10, and
M = 1. The eigenvalues of the non-overlapping operator exhibit degeneracy, which is lifted after
the introduction of overlaps, as evident in the eigenvalues of the overlapping operator Q̃k .

Our construction involving “overlapping” operators introduces randomness into the eigen-
values. To address this, we seek an averaging scheme to provide a clearer picture beyond the
randomness. A straightforward approach would be to average each eigenvalue. However, since
E(⟨vk |ei ⟩) = 0, the expected value of each eigenvalue is zero, reflecting the fact that the expec-
tation of our “overlapping” operators is a null operator. Consequently, this averaging scheme
wouldn’t be effective.

An alternative approach is to average the eigenvalues of Q̃k over the interval where the eigen-
values of non-overlapping operators Qi are degenerate. This can be interpreted as a “coarse-
graining” procedure that removes the non-degenerate fluctuations in the eigenvalues caused
by the overlapping operators. Figure 6.3 depicts the outcome of this “coarse-graining.” The
averaged eigenvalues of Q̃k exhibit a closer resemblance to the eigenvalues of non-overlapping
position operators. Numerical simulations suggest that this correspondence improves with
increasing the dimension dk of the operators, providing support for the validity of our “coarse-
graining” method.
 6.2 How good are these operators? 75

1.5 Eigenvalues of Q, dk = 19
1.0 Avg. eigenvalues of Q̃k
Avg. Eigen values

0.5
0.0
0.5
1.0
1.5
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
Eigen number
Figure 6.3: Average eigenvalues of finite-dimensional position operators. This plot shows the
average eigenvalues of the non-overlapping position operator Qk (blue dots) and the overlapping
position operator Q̃k (orange dots), calculated over the eigennumbers where the non-overlapping
eigenvalues remain degenerate (refer to Figure 6.2 for reference). The overlapping operator param-
eters are ns = 3, Ns = 4, dk = 19, Ωk = 10, and M = 1. We observe that the average eigenvalues
coincide in the bulk (central region) but deviate towards the edges, indicating the influence of the
introduced overlaps.
6.3 Limiting behaviour of Overlapping Operators

Our previous section established that the non-commutativity of Ãk and B̃k arises from the
overlapping nature of the operators. Furthermore, deriving a closed-form expression for this
commutator violation in the finite-dimensional case proved challenging. As discussed in de-
tail in Chapter 4, many problems simplify when taking the infinite-dimensional limit. Addi-
tionally, for most low-energy states, results obtained from finite-dimensional operators closely
match those from infinite-dimensional ones. In this section, we will investigate the infinite-
dimensional limit of the commutation relations for our overlapping operators. The behavior of
low-energy states of the overlapping Hamiltonian will be explored in the next chapter.

In the infinite-dimensional limit, we recall the well-known commutation relation [Qi , Pi ] =

i1. Utilizing equations (6.9) and (6.10), we can now examine that:
ns
Y ns
Y
Ãk B̃ =
k′ exp(−iαs ⟨vk |ei ⟩Pi ) exp(iβs ⟨vk′ |ei ⟩Qi ),
i=1 i=1
Yns
= exp(−iαs ⟨vk |ei ⟩Pi ) exp(iβs ⟨vk′ |ei ⟩Qi ) (since, [Qi , Pj ] = 0 ∀ i ̸= j),
i=1
ns
1
Y  
= exp −iαs ⟨vk |ei ⟩Pi + iβs ⟨vk′ |ei ⟩Qi + αs βs ⟨vk |ei ⟩⟨ei |vk′ ⟩[Pi , Qi ] (using BCH formula),
i=1 2
ns
i
Y  
= exp −iαs ⟨vk |ei ⟩Pi + iβs ⟨vk′ |ei ⟩Qi − αs βs ⟨vk |ei ⟩⟨ei |vk′ ⟩ ,
i=1 2
ns
i
Y  
= exp −iαs ⟨vk |ei ⟩Pi + iβs ⟨vk′ |ei ⟩Qi + αs βs ⟨vk |ei ⟩⟨ei |vk′ ⟩ − iαs βs ⟨vk |ei ⟩⟨ei |vk′ ⟩ ,
i=1 2
ns
Y
= exp(−iαs βs ⟨vk |ei ⟩⟨ei |vk′ ⟩) exp(iβs ⟨vk′ |ei ⟩Qi ) exp(−iαs ⟨vk |ei ⟩Pi ),
i=1
ns
!
X
= exp −iαs βs ⟨vk |ei ⟩⟨ei |vk′ ⟩ B̃k′ Ãk ,
i=1
ns
2π
 
−⟨vk |vk′ ⟩
X
=ω B̃k′ Ãk ∵ |ei ⟩⟨ei |= 1 & αs βs = . (6.19)
i=1 dk

Clearly, in the infinite-dimensional case, for k = k′ , Ãk and B̃k obeys the Generalised Clif-
ford Algebra (or the Weyl form of CCR). Furthermore, at least in the infinite-dimensional limit
we are able to obtain the closed form expression of the commutation relation between Ãk and
B̃k′ and evidently Ãk and B̃k′ do not commute with each other with the overlapping amount
equal to ⟨vk |vk′ ⟩ which we expect to be very small. However, for most of our calculations a
more readily useful version of these commutation relations is in terms of Q̃k and P̃k . Using
 6.3 Limiting behaviour of Overlapping Operators 77

expressions (6.6) and (6.5) one can obtain:

ns
X
[Q̃k , P̃k′ ] = ⟨vk |ei ⟩⟨ej |vk′ ⟩[Qi , Pj ] .
i,j=1

(6.20)

defining, Zi := 1 ⊗ · · · ⊗ Z ⊗ · · · ⊗ 1, where the operator Z = −i[q, p] (as defined in (4.28))

is inserted at the i-th factor. In terms of Zi ’s we can write the above relation as:
ns
X
[Q̃k , P̃k′ ] = ⟨vk |ei ⟩⟨ei |vk′ ⟩iZi = iZ̃k,k′ . (6.21)
i=1

This is the finite-dimensional commutation relation followed by our “overlapping” opera-

tors. Z̃k,k′ is a trace-less symmetric operator and unlike the finite-dimensional commutator
operator (4.29) it does not have a Toeplitz structure. This is because Toepltiz structure is not
preserved under tensor product with identity, 1. Now, if we take the infinite-dimensional limit
then Zi ’s will become identities and the above relation will become
ns
X
[Q̃k , P̃k′ ] = ⟨vk |ei ⟩⟨ei |vk′ ⟩i1 = i⟨vk |vk′ ⟩1 . (6.22)
i=1

Again, clearly when k = k′ , our overlapping operators satisfies the canonical commutation
relation and for all k ̸= k′ these operators fail to commute by the amount ⟨vk |vk′ ⟩ which is
also what we noted from the braiding relations of their exponentiated operators (6.19). These
asymptotic commutation relations of our “overalapping” operators give us good confidence that in the
large dimension limit we won’t be making significant errors as all the nice commutation properties
that we desire seems to work out. Furthermore, in chapter 4 we established that the action of
Zi ’s on the low energy states is like that of an identity therefore, at least for those states these
asymptotic relations holds even in the finite-dimensions.
ns
X
⟨Ψ|[Q̃k , P̃k′ ]|Ψ⟩ = ⟨vk |ei ⟩⟨ei |vk′ ⟩i⟨Ψ|Zj |Ψ⟩ ≈ i⟨vk |vk′ ⟩ . (6.23)
j=1

Note that, in the infinite-dimensional limit our “overlapping” operators Q̃k and P̃k
obeys both the canonical commutation relations and exponentiated commutation rela-
tions which is a sufficient condition for these operators to be unitarily equivalent to Schrodinger’s
position and momentum representation as guaranteed by Stone-von Neumann theorem (Hall,
2013). However, it remains to explicitly show this calculation, just like we showed it for our
non-overlapping operators in the Appendix A.2. At this point it also important to mention that
this equivalence only holds individually i.e., if a unitary map, Uk that transforms Uk Ãk Uk−1 to
e−iαs P̂ and Uk B̃k Uk−1 to eiβs X̂ , where, X̂ and P̂ are Schrodinger’s position and momentum op-
78 Overlapping Quantum Scalar Field

erators then, it will not do the same for operators that are labelled with k′ ̸= k as it’d do for
the non-overlapping operators (see theorem 14.8 of Hall (2013)). This is because of the braiding
relation Ãk B̃k′ = ω ⟨vk |vk′ ⟩ B̃k′ Ãk , as now the invariant subspaces of {Ãk , B̃k } and {Ãk′ , B̃k′ }
are not disjoint and hence, the Hilbert space can’t be decomposed as an orthogonal direct sum
of invariant subspaces {Vk } ∀ k (see chapter 14 of Hall (2013)). This is expected since each
overlapping operator now “talks” to every other operator which makes it impossible to have
“clean” mutually exclusive invariant subspaces as required. However, this was possible for non-
overlapping operators because were literally in tensor product with each other but now this
neat tensor product factorization is not possible.

While Friedrich et al. (2024) obtained a convenient commutation relation for their opera-
tors, unfortunately, we were unable to find a similar relation like equation (6.19) in the finite-
dimensional case. Despite this limitation, our construction presented in equations (6.6) and (6.5)
offers a natural interpretation.

In the infinite-dimensional scenario, each Pi (and similarly Qi ) acts independently, corre-

sponding to measurements of “momentum” (and “position”) along orthogonal axes in Rns .
In this context, a natural concept of overlapping operators would involve measuring “mo-
mentum” (or “position”) along axes that are nearly orthogonal to each other. As shown in
Appendix B.1, the Johnson-Lindenstrauss theorem guarantees the existence of exponentially
many such nearly orthogonal directions (unit vectors). Therefore, employing operators that
measure along these random directions allows us to construct the required number of overlap-
ping operators within a smaller Hilbert space.

Since we are essentially measuring along different axes, conjugate operators measuring along the
same axis will maintain the standard commutation relation. However, conjugate operators mea-
suring along different axes will have contributions arising from the overlap between their cor-
responding axes. Our approach adopts this concept but with finite-dimensional position and
momentum operators. Unfortunately, due to the non-centrality of the finite-dimensional com-
mutation relation, operators measuring along the same direction no longer necessarily follow
the same commutation relation as those along independent directions. Nevertheless, our con-
struction appears to be the most natural choice within this framework.

Given the random nature of our construction due to the choice of nearly orthogonal ran-
dom vectors, we can characterize them using moments, similar to any random variable. The
first two moments provide insights into how closely the behavior of our “overlapping” opera-
tors resembles the desired behavior. In the following derivations, we will leverage the results
established in Appendix B.2.
 6.3 Limiting behaviour of Overlapping Operators 79

1. Expectation of the commutation relation [Q̃k , P̃k ]:

ns
X 
E([Q̃k , P̃k ]) = E ⟨vk |ej ⟩⟨ej |vk ⟩iZj ,
j=1
ns
X
= E(⟨vk |ej ⟩⟨ej |vk ⟩)iZj ,
j=1
ns
1 1
 
2
X
= iZj ∵ E(⟨vk |ej ⟩ ) = . (6.24)
j=1 ns ns

and in the infinite-dimensional limit we recover the CCR

ns
X 1 d→∞
E([Q̃k , P̃k ]) = iZj −−−→ i1 . (6.25)
j=1 ns

2. Variance of the commutation relation [Q̃k , P̃k ]:

V ar([Q̃k , P̃k ]) = E{([Q̃k , P̃k ] − E([Q̃k , P̃k ]))† ([Q̃k , P̃k ] − E([Q̃k , P̃k ]))},
ns ns
1
 X X 
=E ⟨vk |ej ⟩⟨ej |vk ⟩Zj − Zj
j=1 j=1 ns
| {z }
ξ
ns ns
1
X X 
× ⟨vk |el ⟩⟨el |vk ⟩Zl − Zl ,
l=1 l=1 ns
| {z }
ξ
ns
 X ns ns 
2 2 2 2 2
X X
=E ⟨vk |ej ⟩ ⟨vk |el ⟩ Zj Zl − ξ · ⟨vk |ej ⟩ Zj − ξ · ⟨vk |el ⟩ Zl + ξ ,
j,l=1 j=1 l=1
ns
E(⟨vk |ej ⟩2 ⟨vk |el ⟩2 )Zj Zl − ξ 2 ,
X
=
j,l=1
ns
E(⟨vk |ej ⟩4 )Zj2 + E(⟨vk |ej ⟩2 ⟨vk |el ⟩2 )Zj Zl − ξ 2 .
X X
=
j=1 j̸=l

(6.26)

Using the result (B.22) and (B.17), we get:

ns ns 2
3 1 1
X
Zj2
X X
V ar([Q̃k , P̃k ]) = + Zj Zl − Zj ,
j=1 ns (ns + 2) j̸=l ns (ns + 2) j=1 ns
ns 
3 1 1 1
  
− 2 Zj2 +
X X
= − 2 Zj Zl ,
j=1 ns (ns + 2) ns j̸=l ns (ns + 2) ns
ns
X 2(ns − 1) 2 X 2
= 2
Zj − 2
Zj Zl . (6.27)
j=1 ns (ns + 2) j̸=l ns (ns + 2)
80 Overlapping Quantum Scalar Field

and in the infinite-dimensional limit we see that:

ns
X 2(ns − 1) 2 X 2
V ar([Q̃k , P̃k ]) = 2
Zj − 2
Zj Zl ,
j=1 ns (ns + 2) j̸=l ns (ns + 2)
ns
d→∞ X 2(ns − 1) X 2
−−−→ 2
− 2
,
j=1 ns (ns + 2) j̸=l ns (ns + 2)
2(ns − 1) 2
= 2
· ns − 2 · ns (ns − 1),
ns (ns + 2) ns (ns + 2)
=0. (6.28)

This is expected since, our operators obeys the CCR in the infinite-dimensional limit with-
out involving any random factors.

3. Expectation of the commutation relation [Q̃k , P̃k′ ]:

ns
X 
E([Q̃k , P̃k ]) = E ⟨vk |ej ⟩⟨ej |vk′ ⟩iZj ,
j=1
ns
X
= E(⟨vk |ej ⟩⟨ej |vk′ ⟩)iZj ,
j=1
δij
 
=0 ∵ E(⟨vk |ei ⟩⟨ej |vk′ ⟩) = . (6.29)
ns

4. Variance of the commutation relation [Q̃k , P̃k′ ]:

V ar([Q̃k , P̃k′ ]) = E([Q̃k , P̃k′ ]† [Q̃k , P̃k′ ]),

ns
 X ns
 X 
=E ⟨vk |el ⟩⟨el |v ⟩Zl
k′ ⟨vk |ej ⟩⟨ej |v ⟩Zj
k′ ,
l=1 j=1
ns
X
= E(⟨vk |ej ⟩⟨ej |vk′ ⟩⟨vk |el ⟩⟨el |vk′ ⟩)Zl Zj ,
j, l=1
Xns
= E(⟨vk |ej ⟩⟨el |vk ⟩)E(⟨vk′ |ej ⟩⟨el |vk′ ⟩)Zl Zj (∵ vk & vk′ are independent),
j, l=1
ns
E(⟨vk |ej ⟩⟨el |vk′ ⟩)2 Zj Zl ,
X
=
j, l=1
Xns
Zj2
= 2
. (6.30)
j=1 ns

and in the infinite dimensional limit the above expression becomes

Zj2 d→∞ 1
ns
X
V ar([Q̃k , P̃k′ ]) = 2
−−−→ 1 . (6.31)
j=1 ns ns
 6.3 Limiting behaviour of Overlapping Operators 81

This means that the overlaps that we have in the infinite-dimensional commutation relation
becomes smaller and smaller with increasing the number of operators that we overlap. In
chapter 5 we showed that ns ∝ ks so, we can expect that the overlaps becomes smaller as we
go to higher energy Fourier shells. This will be one of our results that we shall discuss in the next
chapter.
 Consequences of
7
the Construction

In the previous chapter we discussed the construction of our overlapping “position” and “mo-
mentum” operators and also looked at the commutation relation it obeys. We noted that the
commutation relation is not “neat” in the sense that at least in the finite-dimensions these over-
lapping operators don’t obey the Generalised Clifford Algebra. They instead obey the “modified”
algebra (6.11) which involves the random factors even when the labels are same for the operators (i.e.,
when k = k′ ). An immediate consequence of this “modified” algebra is that the degeneracy in
the eigenspectrum of non-overlapping operators is broken and we now have almost continu-
ous eigenspectrum of Q̃k and P̃k . Furthermore, we also described a coarse-graining “averaging
scheme” which we expect that it smooths out these randomness and help us see “on-average”
behaviour of our overlapping operators. In this procedure we do the averaging over indices over
which non-overlapping position (or momentum) operators suggests no change in their eigen-
value. It is like taking a trace and therefore, we can think of this scheme as taking an “effective
trace.” Figure (6.3) further reinforces our confidence in this scheme.

In the finite-dimensions the algebra of overlapping operators differ from the GCA because
of the presence of the random factor, ζ (6.11). Interestingly, it’s expectation value is 0 with with
variance 1. These values at first makes our construction less attractive. But, we can modify our
construction such that ζ becomes exactly 1. Instead of taking considering ⟨vk |ei ⟩, if we take
84 Consequences of the Construction

ns
⟨vk |ei ⟩2 in the definition of Q̃k and P̃k then one can show that ζ will become ⟨vk |ei ⟩2 = 1.
P
j=1
Thus, by taking quadratic random factors we can have our overlapping conjugate oper-
ators obey GCA! However, there are reasons that couldn’t convince us to proceed with this
construction. Firstly, with these quadratic random factors one can show the following
ns
⟨vk |ej ⟩2 ⟨ej |vk′ ⟩2 iZj .
X
[Q̃k , P̃k′ ] = (7.1)
j=1

This implies, for k = k′ ,

ns ns
d→∞
⟨vk |ej ⟩4 iZj ⟨vk |ej ⟩4 ̸= i1 .
X X
[Q̃k , P̃k ] = −−−→ i (7.2)
j=1 j=1

and,
ns ns
X
4
X 3 d→∞ 3i
E([Q̃k , P̃k ]) = E(⟨vk |ej ⟩ )iZj = iZj −−−→ . (7.3)
j=1 j=1 ns (ns + 2) ns + 2

Similarly, for k ̸= k′ ,
ns ns
X
2 2
X 1 d→∞ i
E([Q̃k , P̃ ]) =
k′ E(⟨vk |ej ⟩ ⟨ej |v ⟩ )iZj =
k′ 2
iZj −−−→ . (7.4)
j=1 j=1 ns ns

Clearly, in the asymptotic case this construction won’t reproduce the standard canonical
commutation relation unlike our construction with linear random factors. Along the same
lines as in (6.19) one can show that it using quadratic coefficients won’t lead to GCA in the
infinite-dimensional limit. It seems like quadratic construction has more issues than the (only)
solution it offers.

Secondly, towards the end of chapter 5 we discussed the interpretation of our construction
with linear random factors. And we noted that arguably it is one of the most natural way one
can construct overlapping operators. This interpretation is not feasible for the construction with
quadratic random factors. Thus, even though it satisfies the GCA it fails on many aspects on
which our original construction offers better results and understanding. It is for these reasons
we discard the construction based on quadratic random factors and adopt the linear one.

In this chapter we take a deeper look at the consequences of our construction. Particularly,
how these overlaps affects the states of the “overlapping Hamiltonian” and also how severe
the failure to obey GCA is, in the finite-dimensions, as in the infinite-dimensions everything
works out.
7.1 Breaking of Lorentz Symmetry

In Chapter 2, we argued that symmetries of the continuum field theory are not necessarily
inherited by its finite-dimensional counterpart. We emphasized that these symmetries, or
their consequences, should emerge in appropriate energy regimes. However, this does not imply
introducing assumptions that blatantly violate Lorentz symmetry (even in the infinite-dimensional
limit) for the sake of computational convenience. We should strive to minimize such assumptions
whenever possible.

The construction we have discussed so far breaks Lorentz symmetry for several reasons,
although the underlying principles remain similar to those discussed in Friedrich et al. (2024)
for holographic fermionic fields. The first reason is the assumption of a sharp ultraviolet
(UV) cutoff, which we elaborated on in Chapter 3. As noted in Martin (2012) and Akhmedov
(2002), dimensional regularization offers a Lorentz-invariant approach to UV regularization of
quantum fields. However, Mathur (2020) criticizes this scheme as merely a mathematical trick and
lacking concrete physical meaning.

An alternative, also mentioned in Friedrich et al. (2024), is to employ “super-holographic”

overlaps. Here, instead of simply discarding modes above a momentum scale ΛU V , we impose a
stronger squeezing on Fourier shells with mean radius ks exceeding ΛU V . In simpler terms, we
attempt to squeeze ∼ ks2 ∆s modes in shell-s into a significantly smaller number than ∼ ks ∆s of
physical modes. With sufficiently strong squeezing, we can potentially maintain a total number
of degrees of freedom comparable to what we obtain with a sharp UV cutoff, ΛU V , while still
satisfying the cosmic Bekenstein bound. This approach essentially introduces a transition from
the standard holographic mode density for modes below ΛU V to a “super-holographic” mode density
for higher Fourier modes.

Second, as alluded to in our discussion of the holographic principle (Section 2.3), it provides
a bound on the number of degrees of freedom required to describe matter on light-sheets of
any area. However, in this work, we have employed a simplified approach that is only strictly
valid for the horizon area. We suspect that our results might be approximate for other area
choices.

To explore this limitation in future work, we aim to directly quantize our field on a cosmic
light-sheet. By imposing holographic mode overlaps on this light-sheet, we hope to achieve a
more refined description. This approach might also address the need for an infrared (IR) cutoff,
as the cosmic light-sheet naturally terminates at the cosmological horizon.

Lastly, our use of non-local mode overlaps in constructing the holographic scalar field in-
troduces potential violations of Lorentz symmetry. While the full ramifications of this remain
86 Consequences of the Construction

an open question within this thesis, some initial steps towards understanding them are dis-
cussed in Section 7.8. A more comprehensive treatment of this issue was presented in Friedrich
et al. (2024). Their work demonstrated that plane waves in their holographic field exhibit a
finite lifetime. Specifically, they showed that the energy carried by a wave with wave vector
k is isotropically distributed over time to other modes k′ with similar absolute wave numbers
|k′ |≈ |k|. This breaking of Lorentz symmetry potentially connects to concepts of modified dis-
persion relations for relativistic particles, which have been previously explored as a generic fea-
ture of quantum gravity (see e.g., Amelino-Camelia et al. (1998); Alves Batista et al. (2023)). Ad-
ditionally, they showed that the plane wave lifetime can be extended to cosmological timescales
by choosing an ultraviolet (UV) cutoff well below the Planck scale, yet still significantly higher
than the energies probed by current particle physics experiments. We plan to perform a similar
calculation for our holographic scalar field in our future work.

7.2 Dimension of the Oscillator

Thus far, we have treated the dimension of the Hilbert space for each harmonic oscillator as a
free parameter, with a constant value across all modes. However, we can introduce a constraint
on this dimension. We can calculate the maximum number of degrees of freedom associated
with our overlapping field and demand that it respects the cosmic Bekenstein bound, which
limits the information content within our universe’s horizon.

In Chapter 5, we defined the maximum number of degrees of freedom as N := ln(dim(H)),

where dim(H) represents the dimension of the Hilbert space. An alternative, but equivalent,
definition utilizes the base-2 logarithm. With this alternative, the maximum number of degrees
of freedom contributed by our field is given by:

N =
X X
log2 (dim(Hs )) = ns log2 d,
shell, s shell, s

L2 X
= log2 d · ks ∆s ,
2π shell, s
L2 Z ΛU V
≈ log2 d · dk k,
2π 0
(LΛU V )2
= log2 d . (7.5)
4π

here, we have kept the Fourier shell width, ∆s as the minimum, 2π/L. Clearly, the degrees of
freedom of your field scales with area. Now, if we demand that this is less than the maximum
 7.2 Dimension of the Oscillator 87

allowed number of degrees of freedom then:

(LΛU V )2
log2 d ≲ π(LΛplanck )2 ,
4π
2
2 Λplanck

=⇒ log2 d ≲ 4π . (7.6)
ΛU V

Note that our choice of UV cutoff is explicitly distinct from the Planck scale. If we hypoth-
esize that the degrees of freedom provided by our field saturate the cosmic Bekenstein bound,
then the dimension of the Hilbert space for each harmonic oscillator becomes:

Λplanck 2
4π 2
d∼2 ΛU V . (7.7)

With ΛU V = Λplanck , this simplifies to d ∼ 1011 . However, our universe undoubtedly contains
additional fields. A more accurate estimate would incorporate their contributions. Assuming an
area law scaling for the degrees of freedom of all fields (including spacetime degrees of freedom),
we arrive at:
2
Λplanck Ndof, ϕ

2
log2 d ≲ 4π . (7.8)
ΛU V Ndof, total

The standard model has 90 fermionic and 28 bosonic degrees of freedom (Husdal, 2016). This
means:

Ndof, ϕ log2 d
≈ . (7.9)
Ndof, total 90 + 28 log2 d

Using this in equation (7.8) then implies

2
Λplanck log2 d

2
log2 d ≲ 4π · ,
ΛU V 90 + 28 log2 d
 2
2 Λplanck
4π ΛU V
− 90
=⇒ log2 d ≲ . (7.10)
28
2
Λplanck

2
4π − 90
ΛU V
=⇒ d ≲ 2ϱ where, ϱ = . (7.11)
28

Our initial assumption of ΛU V = ΛP lanck leads to an unphysical result of d ∼ 0.28. Even a

more modest choice of ΛU V ≈ 0.1ΛP lanck yields a very large value of d ∼ 1041 .

To address this better, we introduce an additional parameter B into the expression for ns .
(2π)2

This modified expression reads: ns = B · 2πks ∆s . With this modification, Equation
L2
88 Consequences of the Construction

(7.8) becomes:
2
Λplanck Ndof, ϕ

2
B · log2 d ≲ 4π . (7.12)
ΛU V Ndof, total

This modified expression leads to a revised estimate for the dimension:

4π 2 Λplanck
 2
− 90
ϱ′ B ΛU V
d≲2 where, ϱ′ = . (7.13)
28

!
ΛP lanck
The parameter B acts as a suppression factor for the term in the exponent, for
ΛU V
any ΛU V < ΛP lanck .

As highlighted in Chapter Ch-6, our approach involves two key steps: finite-dimensionalization
of the original infinite-dimensional theory and subsequent mode overlapping to achieve
area scaling of the degrees of freedom. There exists a trade-off between smaller overlap val-
ues and operators that effectively mimic their infinite-dimensional counterparts. We will show
1 √
in the next section that ||[Q̃k , P̃k′ ]||≲ ϵ ∝ √ , and since ns ∝ B, this implies ϵ ∝ 1/ B.
ns
Therefore, if we strive for finite-dimensional operators that closely resemble their infinite-
dimensional counterparts (according to Equation (7.13)), B needs to be small, which in turn
weakens the mode overlap. Conversely, prioritizing the independent behavior of the operators
(i.e., reducing overlap by squeezing ∼ ks2 ∆s number of modes into a larger than ∼ ks ∆s number
of modes, implying B >> 1), necessitates working with a smaller Hilbert space dimension for
the oscillators. We defer the determination of the optimal balance between B and d to future work.

7.3 How Small is the Overlap?

Our approach to reducing the degrees of freedom of the finite-dimensional field involves
compressing non-overlapping operators into a smaller Hilbert space, as detailed in Chapter 5.
We define “overlapping” operators as those that fail to commute. To minimize deviations from
standard field theory, we aim for a small overlap, i.e., ||[Q̃k , P̃k′ ]||≈ ϵ, to preserve known
particle physics behavior.

Due to the randomness inherent in our construction, we employed probabilistic measures

(expectation and variance) in the previous chapter to quantify this “smallness” in the asymptotic
limit. Our results aligned with expectations: for identical labels k = k′ , the standard canonical
commutation relation is reproduced with zero variance, indicating an exact result. For distinct
labels k ̸= k′ , the expectation value is zero, and the variance decreases inversely with respect to
 7.3 How Small is the Overlap? 89

ns (6.31).

While this behavior is encouraging, we seek a stronger guarantee – a high probability that
each realization of these random overlaps remains small. It’s conceivable that a specific mode
k might exhibit a large overlap, but the probability of this occurrence should be exceedingly
low. To achieve a more robust quantification of the “smallness” of the commutation relation
between operators associated with different modes, we propose utilizing the operator norm.

From (6.21) we can write,

ns
X ns
X
||[Q̃k , P̃k′ ]||= ⟨vk |ej ⟩⟨ej |vk ⟩iZj ≤ |⟨vk |ej ⟩⟨ej |vk ⟩|||Zj || . (7.14)
j=1 j=1

Since, ||Zj ||= ||Z||·||1||ns −1 = ||Z|| ∀j, therefore,

||[Q̃k , P̃k′ ]||≤ |⟨vk |vk′ ⟩|||Z|| . (7.15)

Again, because of the peculiar (Toeplitz) structure of Z the analytical calculation of its operator
norm wasn’t possible. We resort to numerical estimation of it and it suggests that the norm
of Z scales almost linearly with the dimension of the Hilbert space. This could have been
inferred from the Figure (4.9) as for higher energy states Zk behaves significantly different from
an identity operator and its norm diverges with higher and higher energy states. But the critical
thing is that increasing the dimension of the operator makes the even the higher energy states
match with the infinite-dimensional ones as shown in Figure (7.1).

This is because for a fixed dimension of the Hilbert space (and Ωk , M ) the eigenspacing and
thus the resolution of the wavefunction is fixed. As a consequence the higher energy wavefunc-
tions are simply not properly resolved and thus it leads to a deviation of the action of Zk on
those states from that of an 1k . Hence, for a sufficiently large dimension of the Hilbert space, so
as to have the resolution fine enough to have the predictions match with the standard field the-
ory one, up to the energies probed by the LHC (∼ 13 TeV), we can consider the operator norm
of Z only with respect to the states for which its action is similar to that of 1. We interpret this
as the following - of course there will be states for which the norm of Z will deviate significantly
from 1 but we suspect that those states won’t represent the classical reality and as indicated by the
Figure (7.1) for large dk the norm of Z with respect to almost all the relevant states is ∼ 1. Any-
ways, as shown in Section 4.2, in the asymptotic limit of dk → ∞, Zk becomes identity which
basically reinforces our numerical observation that as dk approaches infinity the eigen number
up to which ⟨Ψ|Z|Ψ⟩ ∼ 1, also approaches infinity, essentially covering the entire spectrum
of physical states, owing to the infinite resolution we’d get in the infinite-dimensional Hilbert
space.
90 Consequences of the Construction

600
Vanilla spectrum
500 dk = 101
Eigenvalues, λk /Ωk

dk = 201 Spectrum matches till here

400 dk = 401
300
200 Zk acts like 1 k till here
100
0
0 50 100 150 200 250 300 350 400
Eigen number
Figure 7.1: The plot shows the eigenspectra of the Hamiltonian Hˆk for three different dimensions:
dk = 101 orange), dk = 201 (green), and dk = 401 (red). Other parameters are fixed at M = 1
and Ωk = 10. As the dimension dk increases, the agreement between the “vanilla spectrum” (blue)
(expected spectrum of the infinite-dimensional harmonic oscillator) and the finite-dimensional spec-
trum improves. Furthermore, for larger values of dk , the action of Zk on the eigenstates of Ĥk
becomes increasingly similar to that of the identity operator for a wider range of states.

Thus, using this effective operator norm we finally obtain an estimate on the “smallness” of
our commutator to be

||[Q̃k , P̃k′ ]|| ≲ |⟨vk |vk′ ⟩| (7.16)

Now, the final step is to determine the magnitude of the random overlap |⟨vk |vk′ ⟩|. Using
the result of the Appendix B.1, we have for all k ̸= k′ in the Fourier shell-s the norm of the
commutator
s
8 Ns
 
||[Q̃k , P̃k′ ]|| ≲ ϵ(ks ) ≈ ln √ ∀ |k|≈ |k′ |≈ ks with probability ≥ 1 − δ (7.17)
ns δ

Using the values of Ns and ns from the previous section we obtain the bound to be:
s
8 k 2 L2
 
||[Q̃k , P̃k′ ]|| ≲ ϵ(ks ) ≈ ln s√ . (7.18)
ks L π δ

Figure 7.2 depicts the dependence of ϵ(ks ) on ks (eV). It reveals a significant decrease in
ϵ(ks ) at higher energies. Even at everyday scales of ∼ 1 cm, the bound reaches 10−27 . At
 7.3 How Small is the Overlap? 91

Holography induces® non-zero commutator for modes k k0 , |k| ≈ k ≈ |k0 |

BUT: |[Q̂k , P̂k ]|Ψ(x) | < ²(k) for all pairs of modes (99% certainty in JL)
0
10 15

10 19
~ 1cm
10 23

10 27

10
²(k)

31

10 35

10 39 ~ LHC

10 43

10 14 10 10 10 6 10 2 102 106 1010 1014

k [eV]
Figure 7.2: Bound on the commutator due to finite-dimensional embedding. This plot shows the
bound on the commutator ∥[Q e k , Pe ′ ]|Ψ⟩∥ for non-coincident modes k ̸= k′ with similar wavenum-
k
bers |k|≈ k ≈ |k′ |. The embedding used to achieve a finite-dimensional Hilbert space is the Johnson-
Lindenstrauss embedding (see Chapter 6 for details). The vertical small arrows indicate typical
energy scales of the Large Hadron Collider (LHC) and everyday phenomena. We observe that the
bound is significantly smaller than the operator norms ∥[Q e k , Pe ′ ]|Ψ⟩∥ at these energy scales. This
k
suggests that for such energies, any two Fourier modes behave almost like independent degrees of
freedom.
92 Consequences of the Construction

LHC energies (∼ 13 TeV), it further reduces to 10−43 . This indicates that at high energies, the
operators exhibit near-independence despite residing in a smaller Hilbert space.

A smaller norm at any ks value is desirable, but it becomes particularly important at higher
ks . This is because the Hamiltonian frequency, Ωk , for modes in Fourier shell-s increases
with ks . Consequently, the ground state (and subsequent low-energy excitations) narrows as
1
σ0 ∝ √ . As previously emphasized, narrower wavefunctions are more sensitive to operator
M Ωk
eigenspacings. Therefore, minimizing overlap effects on higher modes is crucial to avoid signifi-
cant deviations from the desired behavior. It is important to note that our chosen eigenspacing
ensures a well-resolved ground state wavefunction (see Appendix (A.3)). Recalling Equation
1
(4.41), αk also scales as ∼ √ (since dk is constant for all k). This guarantees that at least
M Ωk
the ground state and potentially other lower oscillator excitations will always be well-resolved.
Furthermore, numerical verification demonstrates that with our fixed-dimension Hilbert space
and chosen eigenspacing, varying Ωk has no impact on the eigennumber up to which the wave-
function is well-resolved. These observations bolster our confidence in the choices made for
our finite-dimensional holographic scalar field construction. The near-independence of operators
at high energies, even after overlapping, is indeed an encouraging result.

7.4 Comment on Low Dimensional Simulation

A minor limitation of our construction is that the finite-dimensional overlapping operators

do not satisfy the Generalized Clifford Algebra. However, in the infinite-dimensional limit,
they recover the standard canonical commutation relations. Ideally, we would like our overlap-
ping operators to also be generalized Pauli operators, similar to the non-overlapping ones, but
with the caveat that they would not commute for different modes. If we could achieve a relation
like Equation (6.19) in the finite-dimensional case, we could guarantee that the eigenspectrum
of the overlapping position and momentum operators remains unchanged, eliminating the need
for the “effective trace” scheme. The algebra satisfied by these overlapping operators leads to
random and non-degenerate eigenvalues, making result interpretation challenging.

Nevertheless, we still proceed with this and examine the low-energy wavefunctions of the
Hamiltonian constructed from Q̃k and P̃k (the “overlapping Hamiltonian”). We desire, at least
the low-energy wavefunctions to remain consistent with the standard non-overlapping ones. As
discussed in Chapter 4, analytical calculations of eigenfunctions and other properties proved in-
tractable even for the non-overlapping case in the finite-dimensional Hamiltonian. We therefore
resort to numerical computations here as well.

However, the construction method for the overlapping operators leads to an exponential
growth in the matrix dimension with respect to the Hilbert space dimension. Due to resource
 7.5 Vacuum Energy of Overlapping Field 93

limitations, we performed calculations for low-dimensional cases and unrealistic values of Ns

and ns . We constructed four overlapping operators from three non-overlapping Q and P oper-
ators. We set the Hilbert space dimension for each mode to dim(Hk ) = d = 17 (recognizing
that the actual value will likely be much larger). Utilizing realistic values of M and Ωk would
1
result in a very small eigenspacing, since, αk ∝ √ , and thus, it’d effectively be treated as
M Ωk
zero by the computer. To address this, we chose M = 1 and Ωk = 10 for our calculations. It is
important to note that we work with dimensionless quantities, as including proper dimensions
would render these values very small and compromise computational accuracy due to limita-
tions in the decimal precision of Python’s NumPy library. We await analytical results to obtain
more realistic estimates.

We acknowledge that such small values of Ns and ns do not fully exploit the power of the
Johnson-Lindenstrauss theorem. In essence, we are simulating a worst-case scenario for our
construction.

One approach to analytical calculations would involve taking the asymptotic limit of the
Hilbert space dimension approaching infinity (since we have a better understanding of infinite-
dimensional operators) and simultaneously taking Ns and ns large. We can then compare the
outcomes from low-dimensional simulations and the asymptotic analysis, with reality likely
residing somewhere between these extremes. As we will demonstrate, even the low-dimensional
numerical results are remarkably encouraging and support the validity of our construction.

In the following section, we will show that the vacuum energy of the overlapping Hamilto-
nian can be calculated analytically in the asymptotic limit, and that the ground state remains
Gaussian. Since we only perform overlapping within Fourier shells, it suffices to demonstrate
these results for the Hamiltonian corresponding to a single shell. Hamiltonians for different
shells are in tensor product with each other, so the vacuum energy will simply add, and the
total wavefunction will simply be product of the wavefunction corresponding to each shell.

7.5 Vacuum Energy of Overlapping Field

Our chosen eigenspacing for the non-overlapping operators remains applicable to the
overlapping operators as well. This can be verified by repeating the calculations from Section
4.3.1, recognizing that the overlapping Q̃k and P̃k operators are linear combinations of the
non-overlapping Q and P operators (see their definition in 6.5 & 6.6). Consequently, the same
choice of αk and βk will still maximize the ground state of the finite-dimensional overlapping
Hamiltonian. We retain this eigenspacing choice, and numerical calculations reveal that the
vacuum energy of the overlapping Hamiltonian matches that of the non-overlapping infinite-
dimensional field. Below we will derive this for the asymptotic case. But still this is surprising
94 Consequences of the Construction

because we anticipated a suppression in the vacuum energy due to the reduction in overall field
degrees of freedom, similar to the fermionic field discussed in Friedrich et al. (2024).

In the asymptotic limit, the vacuum energy can be calculated as follows. Consider the Hamil-
tonian corresponding to shell-s:

X  P̃k2 M Ω2k

H˜s = + Q̃k . (7.19)
k∈s 2 2

Since all modes within the shell have similar magnitudes, their frequencies, Ωk , will be ap-
proximately equal. Let’s denote this common frequency by Ω. Using the definitions of P̃k
and Q̃k from Equations (6.6) and (6.5), respectively, and relabeling the modes with a numerical
index j, we can rewrite the above equation as:
   
P1 Q1
   
Ns P  Ns Q )
(
2
1  X  2 MΩ  X  2
H˜s = P1 P2 · · · Pns |vj ⟩⟨vj | 
 ..  +
 Q1 Q2 · · · Qns |vj ⟩⟨vj | 
 .. 
 .
2 j=1  .  2 j=1  . 
   
Pn s Qns
(7.20)

The random matrices |vj ⟩⟨vj | for all j are rank-one, symmetric matrices. Additionally, each
PNs
pair of vectors vi and vj is nearly orthogonal. Therefore, the rank of the matrix j=1 |vj ⟩⟨vj |
will be ns (given Ns > ns ). Due to its symmetric nature, the matrix is diagonalizable, and its
full rank guarantees ns non-zero eigenvalues. We can express this as follows:

Ns
|vj ⟩⟨vj |= OT DO .
X
M= (7.21)
j=1

Here, O is an orthogonal matrix, and D is a diagonal matrix with its diagonal entries represent-
ing the eigenvalues of M. We denote these eigenvalues as {λl }nl=1
s
. Now, let’s define a new set
of operators as follows:
ns
X
P̂i = Oij Pj . (7.22)
j=1
Xns
Q̂i = Oij Qj . (7.23)
j=1

In terms of these new operators, the Hamiltonian can be written as:

ns
P̂ 2 M Ω2 2
 
H˜s = λl l +
X
Q̂l . (7.24)
l=1 2 2
 7.6 Overlapping Wavefunctions 95

While P̂l and Q̂l are not Generalized Pauli Operators due to their non-compliance with the
GCA, they satisfy the following commutation relation in the infinite-dimensional limit:
X
[Q̂i , P̂j ] = Oik Ojl [Qk , Pl ] ,
j, k
X
= Oik Ojl iδjk Zk ,
j, k
ns
d→∞
Oil OljT Zl −−−→ iδij .
X
=i (7.25)
l=1

This matches the commutation relation obeyed by the infinite-dimensional non-overlapping

position and momentum operators. Furthermore, it can be straightforwardly shown that these
operators also satisfy the Weyl form of the Canonical commutation relation.

Therefore, in the infinite-dimensional limit, we can express the overlapping Hamiltonian as

a sum of Hamiltonians for ns non-overlapping harmonic oscillators, weighted by the (random)
eigenvalues of M. The vacuum energy of this total Hamiltonian becomes the weighted sum of
the vacuum energy of each individual Hamiltonian. Since all eigenvalues of M are positive (be-
cause M is simply the sum of outer products of vectors), the vacuum energy of our overlapping
Hamiltonian is:
ns
Ω X Ω Ns Ω
λ̃0 = · λl = · T r(M) = . (7.26)
2 l=1 2 2

N
Ps N
Ps
Here, we have used that T r(M) = T r(|vj ⟩⟨vj |) = 1 = Ns . Notably, we observe no sup-
j=1 j=1
pression in the vacuum energy with our overlapping construction, even in the infinite-dimensional
case.

7.6 Overlapping Wavefunctions

From Equation (7.24), we can also argue that the ground state wavefunction will be Gaus-
sian. In the infinite-dimensional case, the ground state wavefunction of Equation (7.24) will
evidently be Gaussian in the eigenbasis of Q̂l ’s. The relationship between Q̃k and Q̂j ’s is as
follows

Q̃k = vk,i OijT Q̂j with ||vk · OT ||= 1 . (7.27)

This equation expresses Q̃k as a linear combination of the new operators Q̂j . Because we have
||vk · OT ||= 1, we are essentially viewing the same wavefunction but from a transformed coordinate
system. Since all the oscillators have the same frequency, the ground state wavefunction in the
96 Consequences of the Construction

eigenbasis of Q̂l will be a symmetric Gaussian. Consequently, the ground state wavefunction in
the eigenbasis of Q̃k will also appear as a symmetric Gaussian.

However, this argument based on transformation of coordinates is not directly applicable

to the finite-dimensional case. Nevertheless, we can examine the low-dimensional ground state
wavefunction of the overlapping Hamiltonian using the parameter choices discussed earlier (see
Figure (7.3)). The figure visually confirms that the ground state in this case closely resembles
the corresponding infinite-dimensional ground state. It is important to note that the ‘‘effective
trace” procedure was again employed to generate this plot.

ns = 3, Ns = 4, dk = 17
0.6
®
dim(H k ) = ∞; |0
0.5 dim(H k ) = 17; |0
®

0.4
Wavefunction, Ψ

0.3

0.2

0.1

0.0
1.5 1.0 0.5 0.0 0.5 1.0 1.5
1e 7
0
Residual

2
1.5 1.0 0.5 0.0 0.5 1.0 1.5
x
Figure 7.3: The plot shows the ground state wavefunction of the overlapping Hamiltonian con-
structed from ns = 3 effective non-overlapping oscillators (orange crosses) compared to the ground
state wavefunction of the corresponding infinite-dimensional system (solid blue line). The over-
lapping Hamiltonian consists of Ns = 4 oscillators, each with dimension dk = 17 and parameters
M = 1 and Ωk = 10. The orange crosses represent the ground state after taking the “effective
trace” of the complete wavefunction. The residual between the two wavefunctions is on the order
of ∼ 10−7 , indicating excellent agreement. This result is achieved even though it represents a sce-
nario that minimally exploits the full potential of our embedding scheme. This suggests that our
asymptotic limit results for the ground state hold even in this extreme low-dimensional case.

As previously argued, if the ground state wavefunction matches in these two extreme scenar-
ios (infinite-dimensional and low-dimensional but numerical computation), then the wavefunc-
tion for realistic values of d, M , Ω, ns , and Ns can also be expected to exhibit a good degree of
similarity.
 7.6 Overlapping Wavefunctions 97

While we have successfully explained the ground state energy and wavefunction of the over-
lapping Hamiltonian in the infinite-dimensional limit, a complete understanding of its full eigen-
spectrum remains elusive. However, numerical comparisons can be made between the first few
excited states of our overlapping Hamiltonian and their standard counterparts. These compar-
isons, presented in Figures (7.4, 7.5, 7.6, 7.7, & 7.8), reveal a remarkable agreement.

ns = 3, Ns = 4, dk = 17
®
0.4 dim(H k ) = ∞; |1
®
dim(H k ) = 17; |1
0.2
Wavefunction, Ψ

0.0

0.2

0.4

1.5 1.0 0.5 0.0 0.5 1.0 1.5

1e 6

1
Residual

1.5 1.0 0.5 0.0 0.5 1.0 1.5

x
Figure 7.4: The plot shows the closest first state wavefunction of the overlapping Hamiltonian
constructed from ns = 3 effective non-overlapping oscillators (orange crosses) compared to the
first excited state wavefunction of the corresponding infinite-dimensional system (solid blue line).
The overlapping Hamiltonian consists of Ns = 4 oscillators, each with dimension dk = 17 and
parameters M = 1 and Ωk = 10. We have again taken “effective trace.” The residual between the
two wavefunctions is on the order of ∼ 10−6 , indicating excellent agreement.
98 Consequences of the Construction

ns = 3, Ns = 4, dk = 17
0.4 dim(H k ) = ∞; |2
®
®
dim(H k ) = 17; |2
0.2
Wavefunction, Ψ

0.0

0.2

0.4
1.5 1.0 0.5 0.0 0.5 1.0 1.5
0.00

0.01
Residual

0.02

0.03
1.5 1.0 0.5 0.0 0.5 1.0 1.5
x
Figure 7.5: The plot compares the closest second state wavefunction of the overlapping Hamil-
tonian constructed from ns = 3 effective non-overlapping oscillators (orange crosses) compared to
the second excited state wavefunction of the corresponding infinite-dimensional system (solid blue
line). The overlapping Hamiltonian consists of Ns = 4 oscillators, each with dimension dk = 17
and parameters M = 1 and Ωk = 10. As before, the orange crosses represent the state after taking
the “effective trace” of the complete wavefunction. The residual between the two wavefunctions is
on the order of ∼ 10−2 , suggesting a noticeable impact of overlaps on this state compared to the
ground state. However, the key features of the wavefunction are still well-preserved.
 7.6 Overlapping Wavefunctions 99

ns = 3, Ns = 4, dk = 17
0.4 dim(H k ) = ∞; |3
®
®
dim(H k ) = 17; |3
0.2
Wavefunction, Ψ

0.0

0.2

0.4
1.5 1.0 0.5 0.0 0.5 1.0 1.5
0.04
0.02
Residual

0.00
0.02
0.04
1.5 1.0 0.5 0.0 0.5 1.0 1.5
x
Figure 7.6: The plot compares the closest third state wavefunction of the overlapping Hamiltonian
constructed from ns = 3 effective non-overlapping oscillators (orange crosses) compared to the
third excited state wavefunction of the corresponding infinite-dimensional system (solid blue line).
The overlapping Hamiltonian consists of Ns = 4 oscillators, each with dimension dk = 17 and
parameters M = 1 and Ωk = 10. As before, the orange crosses represent the state after taking
the “effective trace” of the complete wavefunction. The residual between the two wavefunctions is
on the order of ∼ 10−2 , suggesting a noticeable impact of overlaps on this state compared to the
ground state. However, the key features of the wavefunction are still well-preserved.
100 Consequences of the Construction

ns = 3, Ns = 4, dk = 17
0.5
®
0.4 dim(H k ) = ∞; |4
®
dim(H k ) = 17; |4
0.3
Wavefunction, Ψ

0.2
0.1
0.0
0.1
0.2
0.3
1.5 1.0 0.5 0.0 0.5 1.0 1.5

0.04
0.02
Residual

0.00
0.02
0.04
1.5 1.0 0.5 0.0 0.5 1.0 1.5
x
Figure 7.7: The plot compares the closest fourth state wavefunction of the overlapping Hamil-
tonian constructed from ns = 3 effective non-overlapping oscillators (orange crosses) compared to
the fourth excited state wavefunction of the corresponding infinite-dimensional system (solid blue
line). The overlapping Hamiltonian consists of Ns = 4 oscillators, each with dimension dk = 17
and parameters M = 1 and Ωk = 10. As before, the orange crosses represent the state after taking
the “effective trace” of the complete wavefunction. The residual between the two wavefunctions is
on the order of ∼ 10−2 , suggesting a noticeable impact of overlaps on this state compared to the
ground state. However, the key features of the wavefunction are still well-preserved.
 7.6 Overlapping Wavefunctions 101

ns = 3, Ns = 4, dk = 17
0.4 ®
dim(H k ) = ∞; |5
0.3 dim(H k ) = 17; |5
®

0.2
Wavefunction, Ψ

0.1
0.0
0.1
0.2
0.3
0.4
1.5 1.0 0.5 0.0 0.5 1.0 1.5

0.05
Residual

0.00

0.05
1.5 1.0 0.5 0.0 0.5 1.0 1.5
x
Figure 7.8: The plot compares the closest fifth state wavefunction of the overlapping Hamiltonian
constructed from ns = 3 effective non-overlapping oscillators (orange crosses) compared to the
fifth excited state wavefunction of the corresponding infinite-dimensional system (solid blue line).
The overlapping Hamiltonian consists of Ns = 4 oscillators, each with dimension dk = 17 and
parameters M = 1 and Ωk = 10. As before, the orange crosses represent the state after taking
the “effective trace” of the complete wavefunction. The residual between the two wavefunctions is
on the order of ∼ 10−2 , suggesting a noticeable impact of overlaps on this state compared to the
ground state. However, the key features of the wavefunction are still well-preserved.
102 Consequences of the Construction

It is important to note that due to the violation of the Generalized Commutation Relation
(GCA) by the overlapping operators Q̃k and P̃k , the eigenspectrum of the overlapping Hamil-
tonian exhibits less degeneracy compared to the non-overlapping case. Consequently, there are
states corresponding to eigenvalues not present in the non-overlapping system. Therefore, the fig-
ures depict comparisons between eigenstates of the overlapping Hamiltonian that closely
resemble the standard states. As expected, the agreement deteriorates for higher energy states.
This can be attributed to the limited overlap (ϵ ∼ 0.2) resulting from the low dimensionality
and small values chosen for Ns and ns . Despite the diminishing resemblance, it is noteworthy that
the wavefunction features remain somewhat discernible.

7.7 Overlapping “Raising” Operator

Another intriguing observation concerns the behavior of the overlapping “raising opera-
tor.” Similar to the non-overlapping case, the overlapping Hamiltonian lacks true ladder op-
erators. However, as previously demonstrated, constructing ladder operators analogous to the
infinite-dimensional ones proves effective, at least for low-energy eigenstates. We applied this
approach to the overlapping case, and the results are presented in Figures (7.9, 7.10, & 7.11).
Notably, applying the overlapping raising operator ã†k only once gives an excited state. Repeated
applications (twice or more) do not yield genuine excitations but rather different states. We sus-
pect this behavior could be due to the significant overlap among the operators and the small
dimensionality of the Hilbert space considered for the computation.

To gain deeper insights into the behavior of the overlapping raising operator, we can analyze
its action in the infinite-dimensional limit or on states where the commutator behaves like the
identity. It can be straightforwardly demonstrated that the overlapping ladder operators satisfy
the following commutation relation:

⟨0|[ãk , ã†k′ ]|0⟩ = ⟨vk |vk′ ⟩ . (7.28)

 7.7 Overlapping “Raising” Operator 103

®
0.4 dim(H k ) = ∞; |1
®
dim(H k ) = 17; ã |0

0.2
Wavefunction, Ψ

0.0

0.2

0.4

1.5 1.0 0.5 0.0 0.5 1.0 1.5

1e 7
5
Residual

0
5
1.5 1.0 0.5 0.0 0.5 1.0 1.5
x
Figure 7.9: The plot compares the wavefunction obtained after applying the creation operator
ã†k to the ground state of the overlapping Hamiltonian (orange crosses) with the first excited state
wavefunction of the corresponding infinite-dimensional system (solid blue line). The overlapping
Hamiltonian is constructed from ns = 3 effective non-overlapping oscillators and consists of Ns = 4
individual oscillators, each with dimension dk = 17 and parameters M = 1 and Ωk = 10. As
before, the orange crosses represent the state after taking the “effective trace” of the complete
wavefunction. The small residual between the two wavefunctions, on the order of ∼ 10−7 , suggests
excellent agreement between the two.
104 Consequences of the Construction

0.4

0.2
Wavefunction, Ψ

0.0

0.2
®
dim(H k ) = ∞; |2
®
dim(H k ) = 17; ã |1
0.4
1.5 1.0 0.5 0.0 0.5 1.0 1.5
0.00
Residual

0.01

0.02
1.5 1.0 0.5 0.0 0.5 1.0 1.5
x
Figure 7.10: The plot compares the wavefunction obtained after applying the creation operator
ã†k to the closest first excited state of the overlapping Hamiltonian (orange crosses) with the sec-
ond excited state wavefunction of the corresponding infinite-dimensional system (solid blue line).
The overlapping Hamiltonian is constructed from ns = 3 effective non-overlapping oscillators and
consists of Ns = 4 individual oscillators, each with dimension dk = 17 and parameters M = 1
and Ωk = 10. As before, the orange crosses represent the state after taking the “effective trace” of
the complete wavefunction. The residual between the two wavefunctions is on the order of ∼ 10−2 ,
which is larger compared to the ground state case (see Figure 7.9). This suggests a more signifi-
cant impact of overlaps on excited states. However, the key features of the wavefunction are still
preserved.
 7.7 Overlapping “Raising” Operator 105

®
0.4 dim(H k ) = ∞; |3
®
dim(H k ) = 17; ã |2

0.2
Wavefunction, Ψ

0.0

0.2

0.4

1.5 1.0 0.5 0.0 0.5 1.0 1.5

0.025
Residual

0.000
0.025
1.5 1.0 0.5 0.0 0.5 1.0 1.5
x
Figure 7.11: The plot compares the wavefunction obtained after applying the creation operator
ã†k to the closest second excited state of the overlapping Hamiltonian (orange crosses) with the
third excited state wavefunction of the corresponding infinite-dimensional system (solid blue line).
The overlapping Hamiltonian is constructed from ns = 3 effective non-overlapping oscillators and
consists of Ns = 4 individual oscillators, each with dimension dk = 17 and parameters M = 1
and Ωk = 10. As before, the orange crosses represent the state after taking the “effective trace”
of the complete wavefunction. The residual between the two wavefunctions is on the order of
∼ 10−2 , which is larger compared to the ground state case (see Figure 7.9). This suggests a more
significant impact of overlaps on excited states. However, the key features of the wavefunction are
still preserved.
106 Consequences of the Construction

Now, consider the state ã†k |0⟩. Applying the overlapping Hamiltonian H˜s and exploiting
the fact that Z|0⟩ ≈ |0⟩ for the ground state, we obtain:
( )
X † 1
H˜s ã†k |0⟩ = Ω ãk′ ãk′ + ã†k |0⟩ ,
k′
2
X † ã†k

=Ω ãk′ ã k′ ã†k + |0⟩ ,
k′
2
X ã†k

=Ω ⟨vk′ |vk ⟩ã†k′ |0⟩ + † †
ãk ãk′ ãk′ |0⟩ + |0⟩ ,
k′
| {z } 2
0
Ω †
 
⟨vk′ |vk ⟩ã†k′ |0⟩ ,
X X
= Ω+ ãk |0⟩ + Ω
k′
2 k̸=k′
Ω
 
ã†k |0⟩ + Ω ϵk′ ã†k′ |0⟩ .
X X
= Ω+ (7.29)
k′
2 k̸=k′
| {z } | {z }
vacuum energy other mode excitations

The final line reveals that ã†k excites not only the mode k but also other modes within the
overlapping shell. However, these additional excitations are significantly suppressed by the
overlap factors (ϵk′ ≪ 1). This allows us to neglect them for realistic values of ns and Ns , as
shown in Figure 7.2, where the maximum mode overlap is demonstrably small. This suggests
that even after overlapping, ã†k can still be considered an effective raising operator, at least
for low-energy states. In our low-dimensional simulations with significant overlap, the action
of the raising operator might be hindered, leading to unreliable excitations beyond the first
excitation.

7.8 Lifetime of Plane Waves

Our ultimate goal is to leverage our concrete model of a holographic field to explore observable
phenomena arising from the holographic principle. As highlighted in Friedrich et al. (2024), the
lifetime of plane wave states in our theory presents a potential avenue for investigating such
phenomena. In standard field theory, plane wave states are stable and have an infinite lifetime.
However, when degrees of freedom are overlapped in an attempt to render the field holographic,
these states become unstable and acquire a finite lifetime.

This instability can be understood by examining the behavior of occupation number opera-
tors within each shell. These operators, denoted by Ñk = ã†k ãk and Ñk′ = ã†k′ ãk′ , correspond
to different Fourier modes k and k′ . Unlike standard theory, they no longer commute. Con-
sequently, the commutator [H˜ , Ñk ] is non-zero, implying that the occupation number in a
specific Fourier mode is no longer conserved. In standard theory, plane wave states (ã†k |0⟩) are
eigenstates of the Hamiltonian and exhibit trivial time evolution. This is no longer true in our
 7.8 Lifetime of Plane Waves 107

holographic version of the scalar field. As a result, time evolution leads to a redistribution of
the occupation number in mode k among the other modes. This behavior aligns with our ob-
servation in Equation (7.29), where ã†k excites all other modes within the shell alongside mode
k.

Within our fiducial construction, the overlaps between different modes are independent of
their relative direction. We interpret this modified time evolution as an isotropic scrambling of
plane waves within their respective Fourier space shells. This phenomenon suggests a breaking
of Lorentz symmetry (particularly, momentum conservation), which was briefly discussed in
Section 7.1.

To quantify the severity of this scrambling effect, Friedrich et al. (2024) considered the plane
wave state in their overlapping fermionic theory. They estimated its characteristic lifetime by
calculating the ensemble average of the second derivative of the expectation value of the num-
ber operator with respect to the lowest energy state in the eigenspace of the number operator
(characterized by the +1 eigenvalue). This state was identified as the closest equivalent within
their theory to the standard plane wave state c̃†k |0⟩ (where c̃†k is the creation operator in the
standard Weyl field theory) for the non-overlapping Weyl field.

This definition is not directly applicable to our case due to the commutation relation pre-
sented in Equation (6.11). As a consequence, the eigenvalues of the number operator Ñk = ã†k ãk
do not exhibit a discrete integer spectrum but rather an almost continuous one. We propose an
alternative approach to estimate the lifetime, which is as follows:
 2
1 d

2
≡ −E 2
⟨0|ãk Ñk ã†k |0⟩ (7.30)
Tscramble dt

We acknowledge that the definition of lifetime presented above lacks a rigorous justifi-
cation. However, it represents our initial attempt to conduct an analysis similar to that of
Friedrich et al. (2024). Based on the relation in Equation (7.29), we expect ã†k to primarily excite
mode k, resulting in a unit occupation number for that mode.

While Friedrich et al. (2024) were able to derive an analytical expression for the scrambling
time in their overlapping fermionic theory, we have not yet achieved this for our model. There-
fore, we resort to low-dimensional simulations to gain insights into the lifetime trend as we
increase the number of operators squeezed into a fixed, smaller Hilbert space.

Intuitively, overcrowding the Hilbert space should significantly disrupt the near-independence
of the operators. Consequently, our “plane wave state” would become progressively less stable,
leading to a decrease in lifetime with increasing Ns . Figure 7.12 aligns with this interpretation.
Due to the use of dimensionless quantities in the simulations, the y-axis lacks units, and the
specific values hold less significance.
108 Consequences of the Construction

The key takeaway is the decreasing trend observed with increasing Ns .

M = 1, Ω k = 10, ns = 3
0.25 dim(H k ) = 9
dim(H k ) = 15
0.20
Tscramble

0.15

0.10

0.05

2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0

Ns
Figure 7.12: Scrambling time versus number of overlapping operators. The plot shows the
scrambling time, Tscramble , as a function of the number of overlapping operators, Ns , included in
the overlapping Hamiltonian. Results are presented for two different dimensions of the individual
oscillators: dk = 9 (blue crosses) and dk = 15 (orange dots). Other parameters are fixed at ns = 3,
M = 1, and Ωk = 10. Due to the use of dimensionless quantities in the simulations, the y-axis
is unitless, and the specific numerical values hold less significance. Each data point represents the
average scrambling time over 1200 simulations, with error bars indicating the standard deviation
of the mean of these 7 points. The key observation is a decreasing trend in scrambling time with
increasing Ns , which is explained in detail in Section 7.8.

Our future work aims to establish an analytical expression for the scrambling time in our
field. High-dimensional simulations are currently infeasible due to computational resource lim-
itations. This figure serves as an initial step, analogous to the approach taken by Friedrich et al.
(2024).
 Discussion and
8
future goals

In chapters 2-7, we started with the motivation for our work from the perspective that a Hilbert
space with a dimension roughly equal to ∼ exp(AdS /4GN ) suffices to describe all possible states
of the static patch, including any matter and black holes it may contain, along with the cosmo-
logical horizon (Bousso, 2000, 2001; Banks, 2001; Banks & Fischler, 2001b; Banks, 2005; Banks
et al., 2006; Banks & Fischler, 2018; Susskind, 2021a,b). This approach draws inspiration from
the holographic principle, which suggests that the boundary degrees of freedom are sufficient
to encapsulate the physics within a specific volume. As discussed in Chapter 2 (see Section 2.3
for a specific reference), the holographic principle implies a dramatic reduction in the degrees
of freedom traditionally used in quantum field theory.

Motivated by these considerations, we embarked on a constructive exploration of holo-

graphic phenomenology. Our goal was to minimally modify quantum field theory (QFT) to
comply with the holographic principle and its ramifications, such as local finite-dimensionality.
Prior work by Friedrich et al. (2024) demonstrated this approach for the Weyl field. This project
aimed to achieve the same for a scalar field, with the detailed implementation outlined in the
preceding chapters.
8.1 Summary of results

We summarize our construction and key findings in the following points.

1. The requirement for a finite-dimensional Hilbert space to describe our Universe poses a
challenge to conventional scalar field theory. The latter inherently necessitates an infinite-
dimensional Hilbert space due to the absence of a finite-dimensional representation of po-
sition and momentum operators that satisfy the Canonical Commutation Relation (CCR)
(see Appendix A.1). We address this challenge by introducing Generalised Pauli Opera-
tors (GPOs) (see Chapter 4). These operators possess appealing properties that make
them well-suited for representing position and momentum within a finite-dimensional
framework.

2. The key elements of this finite-dimensional formalism are: (1) The dimensions dk of the
individual mode Hilbert spaces Hk , which characterize the size of each finite-dimensional
oscillator. (2) The eigenvalue spacings αk and βk of the finite-dimensional conjugate field
operators Q̂k and P̂k , respectively (defined through Equations (4.19) and (4.24).

3. For the analysis of finite-dimensional position and momentum operators, specific eigenspac-
ings (αk and βk ) were chosen based on two criteria:

(a) Maximizing Ground State Energy: A specific value (αk,fix ) was found to max-
imize the ground state energy of the finite-dimensional harmonic oscillator, thus
minimizing the effect of finite dimensionality and bringing it closer to the infinite-
dimensional case (detailed in Chapter 4).

(b) Equal Resolution in Position and Momentum Space: The ground state wavefunc-
tion was required to have equal resolution in both position and momentum space.
By enforcing this constraint, the same αk,fix value was obtained, along with the cor-
responding βk,fix (see Appendix A.3). This choice minimizes deviations from the
expected behavior of the field, especially for low-energy states.

4. However, our finite-dimensional field, while significantly reducing the degrees of freedom
compared to the standard infinite-dimensional theory, still exhibited a volume-scaling of
the degrees of freedom, violating the area-law dictated by the holographic principle (see
Chapter 5). To address this discrepancy, we introduced the concept of “overlapping de-
grees of freedom,” inspired by the overlapping qubit construction of Cao et al. (2017) and
its application to the Weyl field in Friedrich et al. (2024). This method involves squeezing
a larger number of non-overlapping operators into a smaller Hilbert space, effectively in-
troducing non-commutativity among operators associated with different Fourier modes.
 8.1 Summary of results 111

5. We explicitly constructed overlapping position and momentum operators as linear com-

binations of their non-overlapping counterparts, weighted by random coefficients derived
from normalized vectors drawn from a standard normal distribution (see definitions 6.5
& 6.6). The Johnson-Lindenstrauss theorem guarantees that such random vectors can be
nearly orthogonal, allowing us to pack a large number of operators into a smaller space
while maintaining a controlled overlap. We derived the commutation relations for these
overlapping operators, both in the finite-dimensional and infinite-dimensional limits, and
demonstrated that they recover the standard canonical commutation relation for the same
mode and exhibit a small, controllable overlap for different modes (see Chapter 6).

6. We find that each pair of Fourier modes of that field still behaves like a pair of almost
independent, non-overlapping degrees of freedom (||[Q̃k , P̃k ]||∼ 10−43 at LHC energies,
cf. Figure 7.2).

7. By overlapping degrees of freedom, we successfully construct a scalar field whose max-

imum number of degrees of freedom scales with the area of the horizon, in accordance
with the holographic principle. The cosmic Bekenstein bound imposes an upper limit on
the dimension, dk , of the Hilbert space for each harmonic oscillator. This bound depends
on the UV cutoff, ΛU V , and the total number of degrees of freedom in the theory. We
also introduce another parameter, B, which modifies the effective number of modes in
each shell (see Section 7.2). This introduces a trade-off: smaller B leads to weaker overlap
and larger dk , while larger B results in stronger overlap and smaller dk .

8. Our analysis of the “overlapping Hamiltonian,” constructed from these overlapping op-
erators, revealed intriguing properties. We found (numerically) that the eigenspectrum of
the Hamiltonian exhibits less degeneracy compared to the non-overlapping case, leading
to the emergence of new energy levels. The low-energy wavefunctions of the overlapping
Hamiltonian, despite the random nature of our construction, remarkably resemble their
standard counterparts. We introduced an “effective trace” procedure to average over the
randomness, which revealed a clear correspondence between the low-energy states of the
overlapping and non-overlapping oscillators (see Section 7.6).

9. Interestingly, the eigenspacing used in the non-overlapping finite dimensional operators

remained applicable even for the overlapping operators. This is because they are con-
structed as linear combinations of the non-overlapping ones. Consequently, this simplifi-
cation not only benefits the analysis but also leads to a maximized vacuum energy for the
overlapping field as well. In the infinite-dimensional limit, we showed that the vacuum
energy of the overlapping Hamiltonian for the shell-s is Ns Ω/2, exhibiting no suppression
(see Section 7.5).

10. Furthermore, we explored the behavior of the overlapping “raising operator” and ob-
served its ability to effectively excite individual modes, despite the presence of overlaps.
112 Discussion and future goals

This finding suggests that the particle interpretation of excitations, a crucial aspect of stan-
dard QFT, persists in our holographic model, at least for low-energy states (see Section
7.7).

11. However, our investigation also uncovered potential consequences of the overlapping con-
struction, particularly the breaking of Lorentz symmetry. We attributed this breaking to
the non-local nature of mode overlaps and the resulting non-conservation of occupation
numbers in individual Fourier modes. We proposed a preliminary definition for the
lifetime of plane wave states in our theory and conducted low-dimensional simulations to
explore its dependence on the degree of operator squeezing. The results suggest a decrease
in lifetime with increasing overlap, as expected from the intuition that overcrowding the
Hilbert space disrupts the near-independence of the operators (see Section 7.8).

8.2 Potential future directions

This thesis lays the groundwork for a finite-dimensional holographic scalar field theory,
but acknowledges several limitations and open questions that necessitate further investigation.
Below is a compilation of these points, along with suggested improvements:

1. Analytical Expressions for Eigenvalues and Eigenstates: The absence of analytical ex-
pressions for the eigenspectrum and eigenstates of the overlapping Hamiltonian hinders a
complete understanding of the field’s behavior (see Chapter 7, Section 7.6). Therefore, a
potential avenue for future work would be to develop methods for obtaining such expres-
sions, either exactly or approximately. This goal may seem ambitious, considering the
lack of a similar analytical understanding even for the eigenspectrum and eigenstates of
a finite-dimensional harmonic oscillator Hamiltonian. However, we can aim for a more
attainable objective: to achieve at least a partial understanding of how the eigenspectrum
and eigenstates of the overlapping Hamiltonian relate to those of the non-overlapping
one. This approach would mirror the strategy employed to understand the ground state
energy and wavefunction, as detailed in Section 7.5.

2. Rigorous Treatment of Lifetime: The proposed definition of plane wave lifetime, while
intuitive, lacks rigorous justification (see Chapter 7, Section 7.8). Therefore, a potential
area for future work would be to develop a more robust definition and explore its theo-
retical and observational implications for a deeper understanding of Lorentz symmetry
breaking. This should include an analytical derivation of the scrambling time, similar to
the approach taken in Friedrich et al. (2024).

3. Beyond the Sharp UV Cutoff: While convenient, using a sharp UV cutoff breaks
Lorentz invariance and lacks physical justification (see Chapters 3 and 7). Therefore, a
 8.2 Potential future directions 113

potential future direction would be to implement alternative, Lorentz-invariant regular-

ization schemes, such as dimensional regularization (Akhmedov, 2002; Martin, 2012) or
the “super-holographic” overlaps proposed in Friedrich et al. (2024). The latter involves
squeezing modes above a certain momentum scale into a smaller number of physical
modes, effectively transitioning to a “super-holographic” mode density for higher Fourier
modes.

4. Quantization on Cosmic Light-Sheets: The current construction utilizes a simplified

approach that is strictly valid only for the horizon area, not fully capturing the covariant
entropy bound (see Chapter 7, Section 7.1). Therefore, a potential area of future work
would be to quantize the field directly on a cosmic light-sheet. This approach could
potentially at least, makes the choice of the IR scale more natural, because it will be the
radius where the light sheet meets a horizon (such a horizon however only appears in an
expanding Universe).

5. Exploring the interplay between B and dk : We initially assumed a constant dimension,

dk , for the Hilbert space of each harmonic oscillator associated with the Fourier modes.
This simplification benefits the analysis by allowing us to focus on the overlapping mecha-
nism as the primary source of area-scaling degrees of freedom. However, we acknowledge
that the assumption of constant dimension might be overly restrictive. We also raise the
possibility that the dimension, dk , might vary across different Fourier modes, perhaps
scaling in a way that contributes to the area-law scaling of degrees of freedom.

To introduce further flexibility, we introduced another parameter, B, which modifies the

effective number of modes, ns , in each Fourier shell. This effectively scales the dimension
of the Hilbert space for each shell and affects the overlap values (Equation 7.13). There-
fore, a potential avenue for future exploration would be to systematically explore the
parameter space of B and dk to find the optimal balance between minimizing overlap and
maintaining near-infinite-dimensional behavior for the operators. We may investigate po-
tential physical constraints on the dimension dk arising from cosmological observations.
In other words, this would require connecting the abstract Hilbert space dimension to
physically meaningful quantities and exploring the implications for the model’s predic-
tions.

Alternative Constructions of the Conjugate Operators: GPOs emulate the concept of

conjugate operator pairs, which, as highlighted in Carroll & Singh (2021), play a crucial
role in achieving quasi-classical Hilbert space factorizations. However, we have not yet
explored whether the GPO construction is the sole approach to achieve this duality in fi-
nite dimensions. Additionally, we do not anticipate that the emergent pointer observables
of Carroll & Singh (2021) can be expressed entirely in terms of exact GPOs. Therefore,
this area offers a promising avenue for further investigation.
114 Discussion and future goals

6. Implementing our Construction in an Expanding Background: We have so far as-

sumed a fixed background. In an expanding scenario, we suspect that the vacuum energy
might exhibit suppression, similar to the findings in Friedrich et al. (2022). This is because
the time-dependence of the Hamiltonian due to the expansion could render the initially
optimal eigenspacing suboptimal at later times. However, an opposing viewpoint exists.
As discussed in the section on the dimension of the oscillator, the Hilbert space dimension
of each oscillator is already very high. It is therefore conceivable that even if the param-
eters of the Hamiltonian, M (t), and Ωk (t), change with time, the value of αk (or βk )
might not deviate significantly from the fixed-point value due to the large dimensionality
of the harmonic oscillator’s Hilbert space. Nevertheless, we acknowledge that the chosen
eigenspacing might not be the sole viable option. Further investigation into alternative
choices and their implications remains a valuable area for future exploration.

7. Calculation of Real-Space Commutator: We now aim to obtain the real-space commu-

tation relation between the conjugate fields, ϕ(x) and π(y), and investigate the non-local
deviation that arises here, analogous to the analysis performed in Friedrich et al. (2024).

8. Understanding the Action of Overlapping Raising Operators: Repeated application

of the overlapping raising operator does not yield genuine excitations, potentially due to
significant overlap and low dimensionality in the simulations (see Chapter 7, Section 7.7).
Therefore, we must systematically understand this behavior. Furthermore, numerical
results show that the ground state of the overlapping Hamiltonian constructed from a
single overlapping position and momentum operator is not Gaussian, as expected for a
single harmonic oscillator. However, the ground state of a collection of overlapping oper-
ators is Gaussian. This inconsistency is concerning because the Hamiltonian constructed
from a single Q̃k and P̃k operator is the closest equivalent to a standard harmonic oscilla-
tor Hamiltonian. This problem necessitates further investigation.

9. Investigating the “Modified” Algebra of Overlapping Operators: In finite dimen-

sions, overlapping operators deviate from the Generalized Clifford Algebra, introducing
randomness and non-degenerate eigenvalues (see Chapter 6 and 7). This motivates the
exploration of alternative constructions that could lead to a “neat” commutation relation
in finite dimensions, ideally resembling Equation 6.19. Achieving this would simplify the
analysis and potentially eliminate the need for the “effective trace” procedure.

One possibility involves using Haar-random unitary matrices to construct the overlap-
ping position and momentum operators. However, the challenge would then lie in quan-
tifying the “smallness” of the overlap and determining the feasibility of analytical calcu-
lations. Resource limitations often restrict simulations to low dimensions and unrealistic
numbers of overlapped operators.
 8.2 Potential future directions 115

10. Alternative Density of Mode Depletion: As argued in Cohen et al. (1999); Banks &
Draper (2020); Blinov & Draper (2021), the effective degrees of freedom underlying ef-
fective field theory (EFT) are expected to exhibit a depletion in their number that scales
with energy even faster than the naive area scaling of ns ∼ ks ∆s assumed here. It is im-
portant to note that Blinov & Draper (2021) explores a variety of potential observational
consequences of EFT mode depletion, which have not been incorporated into the present
analysis. Nonetheless, a successful modification of our construction might serve as a suit-
able framework to embody the concepts proposed in Banks & Draper (2020); Blinov &
Draper (2021).

These points collectively highlight the exciting avenues for future research arising from this
thesis. By addressing these limitations and open questions, we can gain a deeper understanding
of the implications of finite dimensionality and holography for quantum field theory, ulti-
mately taking significant steps towards constructing a comprehensive model that realizes the
phenomenology of the holographic principle.
 Bibliography

Aghanim N., et al., 2020, Astronomy &amp; Astrophysics, 641, A6

Akhmedov E. K., 2002, Vacuum energy and relativistic invariance (arXiv:hep-th/0204048)

Allen B., 1985, Phys. Rev. D, 32, 3136

Almheiri A., Hartman T., Maldacena J., Shaghoulian E., Tajdini A., 2021, Reviews of
Modern Physics, 93

Alves Batista R., et al., 2023

Amelino-Camelia G., Ellis J., Mavromatos N. E., Nanopoulos D. V., Sarkar S., 1998, Nature,
393, 763–765

Anninos D., Denef F., Law Y. T. A., Sun Z., 2022, Journal of High Energy Physics, 2022

Banks T., 2001, International Journal of Modern Physics A, 16, 910–921

Banks T., 2005, Some Thoughts on the Quantum Theory of Stable de Sitter Space
(arXiv:hep-th/0503066)

Banks T., Draper P., 2020, Physical Review D, 101

Banks T., Fischler W., 2001a, An Holographic Cosmology (arXiv:hep-th/0111142)

Banks T., Fischler W., 2001b, M-theory observables for cosmological space-times
(arXiv:hep-th/0102077)

Banks T., Fischler W., 2018, International Journal of Modern Physics D, 27, 1846005

Banks T., Fiol B., Morisse A., 2006, Journal of High Energy Physics, 2006, 004–004

Bao N., Cao C., Carroll S. M., McAllister L., 2017a, Quantum Circuit Cosmology: The
Expansion of the Universe Since the First Qubit (arXiv:1702.06959)

Bao N., Carroll S. M., Singh A., 2017b, International Journal of Modern Physics D, 26,
1743013
118 BIBLIOGRAPHY

Barceló C., Carballo-Rubio R., Di Filippo F., Garay L. J., 2016, Journal of High Energy
Physics, 2016

Bardeen J. M., Carter B., Hawking S. W., 1973, Commun. Math. Phys., 31, 161

Bekenstein J. D., 1972, Lett. Nuovo Cim., 4, 737

Bekenstein J. D., 1973, Phys. Rev. D, 7, 2333

Bekenstein J. D., 1974, Phys. Rev. D, 9, 3292

Berglund P., Geraci A., Hübsch T., Mattingly D., Minic D., 2023, Classical and Quantum
Gravity, 40, 155008

Bhargava A., Kosaraju S., 2005. pp 396–408, doi:10.1007/11534273_35

Blinov N., Draper P., 2021, Densities of States and the CKN Bound (arXiv:2107.03530)

Bousso R., 2000, Journal of High Energy Physics, 2000, 038–038

Bousso R., 2001, JHEP, 04, 035

Bousso R., 2002, Reviews of Modern Physics, 74, 825–874

Böttcher A., Bogoya J. M., Grudsky S. M., Maximenko E. A., 2017, Sbornik: Mathematics,
208, 1578

Cai T., Fan J., Jiang T., 2013, Distributions of Angles in Random Packing on Spheres
(arXiv:1306.0256)

Cao C., Carroll S. M., 2018, Physical Review D, 97

Cao C., Lackey B., 2021, Journal of High Energy Physics, 2021

Cao C., Carroll S. M., Michalakis S., 2017, Physical Review D, 95

Cao C., Chatwin-Davies A., Singh A., 2019, International Journal of Modern Physics D, 28,
1944006

Carroll S. M., 2001, Living Reviews in Relativity, 4

Carroll S. M., 2021, Reality as a Vector in Hilbert Space (arXiv:2103.09780)

Carroll S. M., Singh A., 2018, Mad-Dog Everettianism: Quantum Mechanics at Its Most
Minimal (arXiv:1801.08132)

Carroll S. M., Singh A., 2021, Physical Review A, 103

Carter B., 1971, Phys. Rev. Lett., 26, 331

 BIBLIOGRAPHY 119

Chao R., Reichardt B. W., Sutherland C., Vidick T., 2017. Schloss Dagstuhl – Leibniz-
Zentrum für Informatik, doi:10.4230/LIPICS.ITCS.2017.48, https://drops.dagstuhl.
de/entities/document/10.4230/LIPIcs.ITCS.2017.48

Chernikov N. A., Tagirov E. A., 1968, Annales Henri Poincar&eacute;, 9, 109

Cohen A. G., Kaplan D. B., Nelson A. E., 1999, Physical Review Letters, 82, 4971–4974

Cotler J. S., Penington G. R., Ranard D. H., 2019, Communications in Mathematical

Physics, 368, 1267–1296

Dai H., Geary Z., Kadanoff L. P., 2009, Journal of Statistical Mechanics: Theory and
Experiment, 2009, P05012

Dasgupta S., Gupta A., 2003, Random Structures & Algorithms, 22, 60

Dong X., Silverstein E., Torroba G., 2018, Journal of High Energy Physics, 2018

Donnelly W., Giddings S. B., 2016a, Physical Review D, 93

Donnelly W., Giddings S. B., 2016b, Physical Review D, 94

Dyson L., Kleban M., Susskind L., 2002, Journal of High Energy Physics, 2002, 011–011

Ellwanger U., 2003, in String Phenomenology. WORLD SCIENTIFIC,

doi:10.1142/9789812704917_0015, http://dx.doi.org/10.1142/9789812704917_0015

Folland G. B., 2001, The American Mathematical Monthly, 108, 446

Friedrich O., Singh A., Doré O., 2022, Classical and Quantum Gravity, 39, 235012

Friedrich O., Cao C., Carroll S. M., Cheng G., Singh A., 2024, Holographic phenomenology
via overlapping degrees of freedom (arXiv:2402.11016)

Gibbons G. W., Hawking S. W., 1977, Phys. Rev. D, 15, 2738

Giddings S. B., 2006, Physical Review D, 74

Giddings S. B., Lippert M., 2001, Physical Review D, 65

Giddings S. B., Lippert M., 2004, Physical Review D, 69

Goheer N., Kleban M., Susskind L., 2003, Journal of High Energy Physics, 2003, 056–056

Güijosa A., Lowe D. A., 2004, Phys. Rev. D, 69, 106008

Hall B., 2013, Quantum Theory for Mathematicians. Graduate Texts in Mathematics,
Springer New York, https://books.google.co.in/books?id=bYJDAAAAQBAJ
120 BIBLIOGRAPHY

Hawking S. W., 1971, Phys. Rev. Lett., 26, 1344

Hawking S. W., 1972, Commun. Math. Phys., 25, 152

Hawking S. W., 1974, , 248, 30

Hawking S. W., 1975, Commun. Math. Phys., 43, 199

Hawking S. W., 1976, Phys. Rev. D, 13, 191

Hawking S. W., Ellis G. F. R., 2023, The Large Scale Structure of Space-Time.
Cambridge Monographs on Mathematical Physics, Cambridge University Press,
doi:10.1017/9781009253161

Hooft G. 1999, Classical and Quantum Gravity, 16, 3263–3279

Hooft G. 2001, in Basics and Highlights in Fundamental Physics. WORLD SCIENTIFIC,

doi:10.1142/9789812811585_0005, http://dx.doi.org/10.1142/9789812811585_0005

Hořava P., Minic D., 2000, Physical Review Letters, 85, 1610–1613

Husdal L., 2016, Galaxies, 4, 78

Israel W., 1967, Phys. Rev., 164, 1776

Israel W., 1968, Commun. Math. Phys., 8, 245

Jagannathan R., 2010, On generalized Clifford algebras and their physical applications
(arXiv:1005.4300)

Jagannathan R., Santhanam T. S., 1982, Int. J. Theor. Phys., 21, 351

Jagannathan R., Santhanam T. S., Vasudevan R., 1981, Int. J. Theor. Phys., 20, 755

Janssen A. J. E. M., 2015, Generalized 3D Zernike functions for analytic construction of

band-limited line-detecting wavelets (arXiv:1510.04837)

Johnson W. B., Lindenstrauss J., 1984, Contemporary mathematics, 26, 189

Kallosh R., Linde A., 1996, Physical Review D, 53, 5734–5744

Klebanov I. R., Susskind L., 1988, Nucl. Phys. B, 309, 175

Klemm D., Vanzo L., 2004, Journal of Cosmology and Astroparticle Physics, 2004, 006–006

Koksma J. F., Prokopec T., 2011, The Cosmological Constant and Lorentz Invariance of the
Vacuum State (arXiv:1105.6296)

Lin H., Susskind L., 2022, Infinite Temperature’s Not So Hot (arXiv:2206.01083)
 BIBLIOGRAPHY 121

Lowe D. A., Susskind L., Uglum J., 1994, Physics Letters B, 327, 226–233

Lowe D. A., Polchinski J., Susskind L., Thorlacius L., Uglum J., 1995, Physical Review D,
52, 6997–7010

Maeda K., Koike T., Narita M., Ishibashi A., 1998, Physical Review D, 57, 3503–3508

Martin J., 2012, Comptes Rendus. Physique, 13, 566–665

Mathur S. D., 2020, International Journal of Modern Physics D, 29, 2030013

Mottola E., 1985, Phys. Rev. D, 31, 754

Nariai H., Ishihara H., 1983

Nastase H., 2007

Ossola G., Sirlin A., 2003, The European Physical Journal C, 31, 165–175

Padmanabhan T., 2014, Gravitation: Foundations and frontiers. Cambridge University Press

Parikh M., Verlinde E., 2005, Journal of High Energy Physics, 2005, 054–054

Peebles P. J. E., Ratra B., 2003, Reviews of Modern Physics, 75, 559–606

Peet A. W., 2001, in Strings, Branes and Gravity. WORLD SCIENTIFIC,

doi:10.1142/9789812799630_0003, http://dx.doi.org/10.1142/9789812799630_0003

Penrose R., 2011, General Relativity and Gravitation, 43, 901

Perlmutter S., et al., 1999, The Astrophysical Journal, 517, 565–586

Peskin M. E., Schroeder D. V., 1995, An Introduction to Quantum Field Theory. Westview
Press

Rendall A. D., 2005, in , 100 Years of Relativity. WORLD SCIENTIFIC, pp 76–92,

doi:10.1142/9789812700988_0003, https://doi.org/10.1142%2F9789812700988_0003

Riess A. G., et al., 1998, , 116, 1009

Santhanam T., Sinha K., 1976, Foundations of Physics, 6, 583

Singh A., Carroll S. M., 2020, Modeling Position and Momentum in Finite-Dimensional
Hilbert Spaces via Generalized Pauli Operators (arXiv:1806.10134)

Smolin L., 2001, Nuclear Physics B, 601, 209–247

Spergel D. N., et al., 2003, The Astrophysical Journal Supplement Series, 148, 175–194

Strominger A., Vafa C., 1996, Physics Letters B, 379, 99–104

122 BIBLIOGRAPHY

Susskind L., 1995, Journal of Mathematical Physics, 36, 6377–6396

Susskind L., 2021b, Black Holes Hint Towards De Sitter-Matrix Theory

(arXiv:2109.01322)

Susskind L., 2021a, De Sitter Holography: Fluctuations, Anomalous Symmetry, and Worm-
holes (arXiv:2106.03964)

Susskind L., Witten E., 1998, The Holographic Bound in Anti-de Sitter Space
(arXiv:hep-th/9805114)

Susskind L., Thorlacius L., Uglum J., 1993, Physical Review D, 48, 3743–3761

Thordsen E., Schubert E., 2020, ABID: Angle Based Intrinsic Dimensionality. Springer In-
ternational Publishing, p. 218–232, doi:10.1007/978-3-030-60936-8_17, http://dx.doi.
org/10.1007/978-3-030-60936-8_17

Weinberg S., 1989, Rev. Mod. Phys., 61, 1

Zumino B., 1977, in EPS Conference on Particle Physics..

’t Hooft G., 1995, Quantum information and information loss in General Relativity
(arXiv:gr-qc/9509050)

’t Hooft G., 2000, Determinism and Dissipation in Quantum Gravity, Erice lecture
(arXiv:hep-th/0003005)

’t Hooft G., 2001a, Determinism in Free Bosons (arXiv:hep-th/0104080)

’t Hooft G., 2001b, How Does God Play Dice? (Pre-)Determinism at the Planck Scale
(arXiv:hep-th/0104219)

’t Hooft G., 2001c, Quantum Mechanics and Determinism (arXiv:hep-th/0105105)

van de Bruck C., Dorca M., Brandenberger R. H., Lukas A., 2000, Physical Review D, 62
 Appendix
A
A.1 No Finite-Dimensional Representation of CCR

To prove: For any Hilbert space H there are no bounded operators P̂k and Q̂k which satisfy
[Q̂k , P̂k ] = i

Proof: Let’s prove this by contradiction. Suppose that there exists bounded operators P̂k and
Q̂k such that Q̂k P̂k − P̂k Q̂k = i1.

Then for any n ∈ N the following will hold:

Q̂nk P̂k − P̂k Q̂nk = inQ̂kn−1 ̸= 0 . (A.1)

The above claim can be proven by induction as - for n = 1 it obviously holds and then if we
assume it holds for some n then it’ll hold for n + 1 as well, since:

Q̂n+1 n+1
k P̂k − P̂k Q̂k = Q̂nk (Q̂k P̂k ) + (Q̂nk P̂k − P̂k Q̂nk )Q̂nk = i(n + 1)Q̂nk . (A.2)

Hence, by induction, Q̂nk ̸= 0, which establishes the claim. From the equality Q̂nk P̂k − P̂k Q̂nk =
124 Appendix

inQ̂n−1
k , we have:

n||Q̂n−1 n n n n n−1
k ||= ||Q̂k P̂k − P̂k Q̂k ||≤ ||Q̂k P̂k ||+||P̂k Q̂k ||≤ 2||Q̂k ||||Q̂k ||||P̂k || . (A.3)

∵ for any n, ||Q̂n−1 n−1

k ||̸= 0, we can cancel ||Q̂k || on both sides and obtain:

n ≤ 2||Q̂k ||||P̂k || . (A.4)

Since, this holds for any n ∈ N, it leads to the contradiction that ||Q̂k ||, ||P̂k ||< ∞.

A.2 Representation of Q̂k and P̂k in Asymptotic Limit

We have the finite-dimensional version of position and momentum operators as denoted by

Q̂k and P̂k , respectively. With reference to (4.19) the eigenvalue equation of Q̂k is

Q̂k |qjk ⟩ = αk j|qkk ⟩ . (A.5)

Just like we label position eigenstates in the infinite-dimensions as |x⟩ let’s label in the same
manner for the finite-dimensional case too. This means the components of any wavefunction,
|Ψ⟩ in the position eigenbasis is represented by {⟨αk j|Ψ⟩|j = −lk , · · · , lk }, where lk is such
that dk = 2lk + 1. Therefore, in the finite-dimensional case Q̂k operator acts like

⟨αk j|Q̂k Ψ⟩ = αk ⟨αk j|Ψ⟩ . (A.6)

and P̂k operator acts like

lk lk
1 X 2πis(j − j ′
 
⟨αk j ′ |Ψ⟩ .
X
⟨αk j|P̂k Ψ⟩ = sβk exp (A.7)
dk j ′ =−lk s=−lk dk

Now, when dk → ∞, lk will also become infinitely large and αk and βk will become infinites-
imally small such that the constraint (4.25) is always obeyed. In such a limit we can regard αk j,
αk j ′ and βk s as almost continuously varying variables from −∞ to ∞. For proper indexing we
replace these quasicontinuous variables by qk , qk′ , and pk , respectively. Then the relations (A.6)
and (A.7) along with (4.25) becomes

⟨qk |Q̂k Ψ⟩ = qk ⟨qk |Ψ⟩ = ⟨qk |q̂k Ψ⟩ . (A.8)

 A.3 Resolution Criteria 125

and,

lk lk
1 X 2πis(j − j ′ )
 
⟨αk j ′ |Ψ⟩ ,
X
⟨qk |P̂k Ψ⟩ = sβk exp
dk j ′ =−lk s=−lk dk
α k lk βX
k lk
1 2πisβk (αk j − αk j ′
 
⟨αk j ′ |Ψ⟩ ,
X
= αk βk sβk exp
dk αk βk ′
αk j =−αk lk βk s=−βk lk d k αk βk

1 Z∞ ′ Z∞
≈ dqk dpk pk exp(ipk (qk − qk′ ))⟨qk′ |Ψ⟩ ,
2π −∞ −∞
d
= −i ⟨qk |Ψ⟩
dqk
= ⟨qk |p̂k Ψ⟩ . (A.9)

Clearly, from (A.8) and (A.9) we can say that in the infinite dimensional limit our finite-
dimensional operators Q̂k and P̂k are equivalent to the corresponding Schrodinger representa-
tion.

⟨qk |q̂k Ψ⟩ = qk ⟨qk |Ψ⟩ . (A.10)

d
⟨qk |p̂k Ψ⟩ = −i ⟨qk |Ψ⟩ . (A.11)
dqk

in accordance with the Canonical Commutation Relation, [q̂k , p̂k ] = i.

A.3 Resolution Criteria

We want to choose the eigenspacing of our finite-dimensional “position” and “momentum”

operators such that our construction resembles as closely as possible to the infinite-dimensional
one. We motivated one choice of eigenspacing αk and βk in chapter 4 by maximizing the ground
state energy of the Hamiltonian of the finite-dimensional harmonic oscillator. Here, we try to
motivate the same value from a different point-of-view. This is also discussed in Friedrich et al.
(2022).

From the constraint relation (4.25), it is clear that reducing the eigenspacing of say position
operator increases the eigenspacing of the momentum operator (and vice versa). In other words,
if by choosing a very small eigenspacing for the position operator, αk we have a wavefunction
well resolved in the position space then it won’t be well resolved in the momentum space, be-
cause βk would then be large. Anyways, the high energy states are sensitive to the eigenspacing
of these operators and would demand an incredible resolution in both position and momentum
space which is not possible at the same time owing to the constraint (4.25). However, in the
infinite-dimensional oscillator both the conjugate operators are equally and infinitely well re-
solved. So, we can at least demand the ground state of the finite-dimensional harmonic oscillator
126 Appendix

|0k ⟩ to be equally resolved in both the spaces i.e.,

⟨0k |Q̂2k |0k ⟩ ⟨0k |P̂k2 |0k ⟩

= . (A.12)
αk2 βk2

⟨0k |Q̂2k |0k ⟩ ⟨0k |P̂k2 |0k ⟩ 2

=⇒ = dk ,
αk4 (2π))2
⟨0k |Q̂2k |0k ⟩ (2π)2
=⇒ αk4 = . (A.13)
⟨0k |P̂k2 |0k ⟩ d2k

We couldn’t obtain the analytical expression of the above expectation values. But we can
approximate them using their infinite-dimensional values. Therefore, we have

1 M Ωk
⟨0k |Q̂2k |0k ⟩ ≈ ; ⟨0k |P̂k2 |0k ⟩ ≈ ,
2M Ωk 2
s s
2π 2πM Ωk
=⇒ αk ≈ ; βk ≈ . (A.14)
M Ωk dk dk

These are the same values that we obtained in (4.41). This reinforces our confidence that this
choice of eigenspacing indeed make our finite-dimensional construction to closely resemble the
infinite-dimensional one.
 Appendix
B
B.1 Johnson-Lindenstrauss Theorem

Our overlapping construction and much of its properties are based upon this theorem. Inter-
ested readers can find more thorough details and proofs in (Dasgupta & Gupta, 2003; Johnson
& Lindenstrauss, 1984; Bhargava & Kosaraju, 2005). In this appendix we simply quote the main
result that we will require in the bulk of our thesis. This is section is based on the appendix of
Friedrich et al. (2024) and we describe it here for the sake of completeness.

Consider a random matrix Md×n whose columns are random vectors drawn from a normal
multivariate distribution N (0, 1d ) and then normalised to unity. Then for all

8 n
 
d≥ 2
ln √ , (B.1)
ϵ δ

and for any n vectors {e1 , · · · , en } ∈ Rn , we will have for all pairs ei , ej the following,

(1 − ϵ)|ei − ej |2 ≤ |M ei − M ej |2 ≤ (1 + ϵ)|ei − ej |2 , (B.2)

with probability ≥ 1 − δ. In other words, with probability more than 1 − δ the pairwise
distances between any pair of vectors ei and ej will remain same up to an error of ±ϵ even after
128 Appendix

their transformation to M ei and M ej , respectively.

This can be proved by noting that the probability that the equation (B.2) is not satisfied for
some pair of vectors is (Dasgupta & Gupta, 2003)

|ei − ej |2 −|M ei − M ej |2 dϵ2

   
P > ϵ ≤ 2 exp − . (B.3)
|ei − ej |2 4

Using (B.1) we can re-write this as

|ei − ej |2 −|M ei − M ej |2 n 2δ
    
P > ϵ ≤ 2 exp − 2 ln √ = . (B.4)
|ei − ej |2 δ n2

The above is for some pair not satisfying (B.2), therefore, the probability of having at least one
pair not satisfying (B.2) is

|ei − ej |2 −|M ei − M ej |2 2δ
 
n
P > ϵ for at least one pair ≤ C2 · ≤δ. (B.5)
|ei − ej |2 n2

This implies the probability that all pairs satisfy (B.2) will simply be ≥ 1 − δ.

Furthermore, Bhargava & Kosaraju (2005) showed that we can have for all pairs ei , ej the
following

ei · ej − ϵ ≤ M ei · M ej ≤ ej · ej + ϵ , (B.6)

with probability ≥ 1 − δ. And with an orthonormal choice of vectors {ei }ni=1 , we obtain
the version of Johnson-Lindenstrauss theorem that we actually used in the main text i.e, |M ei ·
M ej |≤ ϵ.

In our construction, we identify vk and vk′ with M ek and M ek′ with the setting that d = ns
and n = Ns . Thus, we will have for all modes k ̸= k′ we will have

||[Q̃k , P̃k′ ]||≤ ϵ , (B.7)

with probability ≥ 1 − δ, if we choose

8 Ns
 
ns ≥ ln √ . (B.8)
ϵ2 δ

Therefore, for a given value of ns , Ns and δ, the smallest value of the overlap in the commutation
relation is given by
s
8 Ns
 
ϵbest = ln √ . (B.9)
ns δ
B.2 Moments of the Random Overlaps

Our construction involves drawing random vectors from a multivariate standard normal
distribution N (0, 1) and then normalising them (see Chapter 6). Notation wise, let’s denotes
the random draws by {ṽk } ∈ Rns and the normalised draws by {vk }. Furthermore, let {ei }ni=1
s

be the standard basis of Rns . At many places we require the value of the moments of components
of these vectors or the moments of overlaps of these vectors. In the following we derive some
of these for easy reference.

We know that, if ṽk is normally distributed then it follows that vk is uniformly distributed
over the unit sphere in Rns (Thordsen & Schubert, 2020). Therefore, the distribution of the
overlap between any two vectors, vk and k′ simply means the distribution of cosine of the
angle between them, say θ. In fact, Thordsen & Schubert (2020) has obtained this distribution
by extending the work of Cai et al. (2013) on the distribution of angle between two random
vectors drawn uniformly and independently on a unit sphere in Rd . They showed that cos θ
is distributed as 2X − 1, where X ∼ B(α, β), a beta distribution defined on [0, 1] with
d−1
α=β= 2
.

From this and the moments of a beta distribution we can derive the moments of the overlap
that we’ll require as:

α 1
 
E(⟨vk |vk′ ⟩) = E(cos θ) = 2E(X) − 1 = 0 ∵ E(X) = = . (B.10)
α+β 2
E(⟨vk |vk′ ⟩2 ) = E(cos2 θ) = 1 + 4E(X 2 ) − 4E(X) ,
αβ αβ 1
 
2
=1+4 −2, ∵ E(X ) = +
(α + β)2 (α + β + 1) (α + β)2 (α + β + 1) 4
1 1
= = . (∵ d = ns in our construction) (B.11)
d ns
1 4 3 3
  
4 4
E(⟨vk |vk′ ⟩ ) = E(cos θ) = 16E X − = = . (B.12)
2 d(d + 2) ns (ns + 2)
−6 3
   
2
∵ 4th moment of X is + 3 [V ar(X)] =
3 + 2α 16d(d + 2)

We not only require the moments of the overlaps between the normalised random vectors but
also the moments of its components. This is because we have chosen {ei } as the standard normal
basis of Rns and therefore, ⟨vk |ei ⟩ is nothing but the ith component of vk i.e., vki . Since, for any
draw of two vectors on a unit sphere, we can always take one of them along any of {ei }. From
130 Appendix

this it immediately follows that:

E(⟨vk |ei ⟩) = 0 . (B.13)

1
E(⟨vk |ei ⟩2 ) = . (B.14)
ns

Furthermore, knowing the distribution of ⟨vk |ei ⟩2 will be useful for our calculations. It is not
hard either to obtain its distribution. See that :
2
ṽk,i
⟨vk |ei ⟩2 = ns , (B.15)
ṽ 2
P
ṽk,i + k,j
j̸=i

ns
2
and ṽk,i ∼ N (0, 1), this implies ṽk,i ∼ χ2 with 1-dof and ṽ 2 ∼ χ2 with (ns − 1) dof ’s.
P
k,i
j̸=i
Then from the ratio of distributions it follows immediately that:

1 ns − 1
 
2
⟨vk |ei ⟩ ∼ B , . (B.16)
2 2

From this we can calculate the 4th moment of ⟨vk |ei ⟩ as

1 ns − 1
  
4
E(⟨vk |ei ⟩ ) = V ar B , + [E(⟨vk |ei ⟩2 )]2 ,
2 2
1 (ns −1)
· 2
 2
2 1
= 2  + .
1 (ns −1) 1 (ns −1) ns
2
+ 2 2
+ 2 +1

3
=⇒ E(⟨vk |ei ⟩4 ) = . (B.17)
ns (ns + 2)

And finally, we need the moments of the kind (⟨vk |ei ⟩⟨ej |vk ⟩)k with i ̸= j. Let’s denote each
realisation of ṽk,i as xi then these moments can be computed as follows:
k 
ṽk,i ṽk,j

µk := E[(⟨vk |ei ⟩⟨ej |vk ⟩)k ] = E 2 2
,
ṽk,1 + · · · + ṽk,n s
k
1 xi xj
Z 
2 2
= n /2 2
e−(x1 +···+xns )/2 dx1 · · · dxns ,
(2π) s n
R s x1 + · · · + xns
2
Z ∞  Z 
ns −1 −r2 /2
= r e ωik ωjk dσ(ω) . (B.18)
0 Sns −1

Here, ωi = √ xi
and Sns −1 := {x ∈ Rns : ||x||= 1} is the unit sphere in Rns and σ is the
x21 +···+x2ns
 B.2 Moments of the Random Overlaps 131

surface measure on Sns −1 . Now, using the integration result of Folland (2001) i.e.,

Z 0,

is some αi is odd,
ω1α1 · · · ωnαn dσ(ω) = (B.19)
Sns −1  2Γ(β1 )···Γ(βn ) ,

if all αi are even and βi = 21 (αi + 1)
Γ(β1 +···+βn )

we can simply the integration to obtain,






0, if k is odd,
µk =  (1·3·5···(k−1))2
, if k is even (B.20)
 k−1
Q
(n+2l)


l=0

From this it is clear that

E(⟨vk |ei ⟩⟨ej |vk ⟩) = 0 . ∵ k = 1 (odd) (B.21)

1
E(⟨vk |ei ⟩2 ⟨ej |vk ⟩2 ) = . ∵ k = 2 (even) (B.22)
ns (ns + 2)