Quantum limits on the time-bandwidth product of an optical resonator

Mankei Tsang1, 2,
1
Department of Electrical and Computer Engineering,
National University of Singapore, 4 Engineering Drive 3, Singapore 117583
2
Department of Physics, National University of Singapore, 2 Science Drive 3, Singapore 117551
(Dated: September 29, 2017)
A thought-provoking proposal by Tsakmakidis et al. [Science 356, 1260 (2017)] suggests that
the time-bandwidth product of a passive optical resonator can be made arbitrarily large. Here
I consider textbook quantum models of an optical resonator and show that quantum mechanics
imposes a limit on its time-bandwidth product, or else requires extra noise to be added in the same
fashion as amplified spontaneous emission in an active resonator. I also use thermodynamics to
arXiv:1709.10073v1 [quant-ph] 28 Sep 2017

argue that extra noise must be present in their proposed nonreciprocal device.

The simplest quantum model of a lossy optical res- where A (t) is the annihilation operator of another field
onator [1, 2] obeys the quantum Langevin equation given that obeys [A (t), A (t )] = (tt ) and 0 is the cor-
by responding coupling rate. To retrieve the classical model
in Ref. [3], the mean field hA (t)i should be zero. The
da(t) 1 commutator is preserved when ( + ) = 2, meaning
= i0 a(t) a(t) + A(t), (1)
dt that 2.
> 2 implies that there is more quantum fluctua-
where a(t) is the annihilation operator of the resonator tion from the input than dissipation. To keep the system
mode in the Heisenberg picture, A(t) is the annihilation linear and the classical correspondence intact, any mod-
operator of a field at the input of the resonator, 0 is ification of Eqs. (1) and (4) should be an additive term
the decay time of the resonator mode, and 0 is the with zero mean, and to make > 2, the commuta-
input coupling rate. Taking the expectation of Eq. (1) tor of this additive term should be negative to preserve

and substituting (t) = ha(t)i results in the classical [a(t), a (t)]. These requirements can be satisfied by in-
equation assumed in Ref. [3]. Ref. [3] defines the time- troducing a Bogoliubov coupling of a(t) to B (t) of yet
bandwidth product (TBP) as and claims that their another field [1, 2], viz.,
proposed nonreciprocal device can make arbitrarily
large. da(t) 1 p
= i0 a(t) a(t) + A(t) + A (t)
 Assuming the standard bosonic commutation relation dt


A(t), A (t ) = (t t ) for the input field, it can be + B (t), (5)
shown that
where [B (t), B(t )] = [B(t ), B (t)] = (t t ) and
 
a(t), a (t) = e2t/ + 1 e2t/ . 0 is the coupling rate for the third field. The com-
 
(2)
2 mutator becomes
Since the Heisenberg-picture operator a(t) evolves as ( + )  
a(t), a (t) = e2t/ + 1 e2t/ ,
 
a(t) = U (t)a(0)U
 (t), where
 U is a unitary operator, 2
the commutator a(t), a (t) = 1 must be preserved at (6)
all times. This requirement forces the TBP to be [1, 2]
and now
= 2. (3) = 2 + (7)

In other words, the decay of the commutator at a rate
of 2/ must be compensated exactly by coupling to a
quantum field at a rate of = 2/ . This is a quantum
version of the dissipation-fluctuation relation.
Coupling the resonator to more fields in the same fash-
ion can only make the total dissipation rate 2/ larger
and the TBP smaller. For a specific example, suppose
da(t) 1 p
= i0 a(t) a(t) + A(t) + A (t), (4)
dt
mankei@nus.edu.sg
can exceed 2 and become arbitrarily large. It is well
known [1, 2] that such a model is equivalent to a res-
onator with an internal amplifier, in which case is the
gain coefficient introduced by the amplifier and the net
dissipation rate 2/ = + is reduced. The pres-
ence of B (t), however, must introduce extra noise [1, 2],
known as amplified spontaneous emission in the context
of amplifiers. This result implies one of the following
conclusions:
1. If the model here describes the device proposed in
Ref. [3], it is equivalent to an active resonator and
must suffer from extra noise. There is no practi-
cal need for the device when active resonators, an
established technology, can do the same job.

2. If the device can make > 2 without the ex-
tra noise, new physics beyond the textbook models
here will be required to explain it.
In the context of the nonreciprocal surface magneto-
plasmons (SMPs) considered in Ref. [3], thermodynamics
can provide another reason for the extra noise. In their
classical model, only the right-propagating SMP mode
exists and the left-propagating mode is forbidden, mean-
ing that energy can transfer only from the left to the
right. Assume a cold bath on the left, a hot bath on
the right, and the two being coupled via the SMPs only.
The one-way energy transfer leads to heat flowing from
the cold bath to the hot one, violating the second law of
thermodynamics. To restore the second law, there must
[1] Crispin W. Gardiner and Peter Zoller, Quantum Noise
(Springer-Verlag, Berlin, 2004).
[2] Hermann A. Haus, Electromagnetic Noise and Quantum
Optical Measurements (Springer-Verlag, Berlin, 2000).
[3] K. L. Tsakmakidis, L. Shen, S. A. Schulz, X. Zheng, J. Up-
ham, X. Deng, H. Altug, A. F. Vakakis, and R. W.
2
be coupling to another bath that is neglected in the clas-
sical model.
Similarly, the nonreciprocal resonator hypothesized in
Ref. [3], with its large input coupling rate and arbitrarily
small output rate, is analogous to a black hole that sucks
energy in one direction; extra noise akin to Hawking ra-
diation should be present to uphold the second law. The
active-resonator model offers a natural way to extend the
analogy to the quantum domain: the type of Bogoliubov
coupling used in Eq. (5) also underpins the quantum the-
ory of black-hole radiation [4] and ensures consistency
with both quantum mechanics and thermodynamics.
This work is supported by the Singapore Ministry of
Education Academic Research Fund Tier 1 Project R-
263-000-C06-112.
Boyd, Breaking Lorentz reciprocity to overcome the time-
bandwidth limit in physics and engineering, Science 356,
12601264 (2017).
[4] Stephen W. Hawking, Black hole explosions? Nature
248, 3031 (1974).