# Chaos bound
The **chaos bound** is a bound on the growth of [[0482 Out-of-time-order correlator|OTOC]] or more precisely the [[0466 Lyapunov exponent|Lyapunov exponent]] that characterises the [[0008 Quantum chaos|chaotic]] properties of the quantum system. It follows from unitarity and analyticity. There is no such analogue for classical chaos.
Defining the OTOC as $
F(t)=\operatorname{tr}[y V(0) y W(t) y V(0) y W(t)], \quad y^4=\frac{e^{-\beta H}}{\operatorname{tr}\left[e^{-\beta H}\right]},
$[[2015#Maldacena, Shenker, Stanford|MSS]] started by making some assumptions about this function, $F(t)$, which are believed to be true for quite general physical systems:
1. $F(t)$ is analytic in the half strip $\left\{t \in \mathbb{C} \mid \operatorname{Re} t \geq t_0\right.$ and $|\operatorname{Im} t| \leq \frac{\beta}{4}\}$;
2. $F(t)$ is bounded by the factorised correlator $|F(t)|\le F_d$ on the boundary of the strip;
3. $(F(t))^*=F\left(t^*\right)$ (the Schwarz reflection condition),
where the factorised correlator, $F_d$, is defined as $
F_d=\operatorname{tr}\left[y^2 V(0) y^2 V(0)\right] \operatorname{tr}\left[y^2 W(t) y^2 W(t)\right],
$and $t_0$ is called the factorisation time, defined to be when $\operatorname{tr}\left[y^2 W(t) V(0) y^2 V(0) W(t)\right] \approx F_d,$i.e., when the OTOC factorises.
From these assumptions, it follows that $\left(\frac{2 \pi}{\beta}-\partial_t\right)\left(F_d-F(t)\right) \geq 0,$which leads to the bound on the [[0466 Lyapunov exponent|Lyapunov exponent]] for [[0008 Quantum chaos|chaotic]] systems because $F_d-F(t) \sim e^{\lambda_L t}$ for such systems. See [[2015#Maldacena, Shenker, Stanford|MSS]] for the derivation.
According to [[2021#Kundu (Sep, a)|Kundu]], however, that was not the full story: the MSS bound did not fully exploit the properties/assumptions above. Kundu defines a local moment of the OTOC as follows: $\mu_J(t)=e^{\frac{4 \pi J}{\beta} t} \int_{t-i \frac{\beta}{4}}^{t+i \frac{\beta}{4}} d t^{\prime} e^{-\frac{2 \pi}{\beta}\left(t^{\prime}-i \frac{\beta}{4}\right)(2 J+1)}\left(F\left(t^{\prime}\right)-F_d\right).$Then a stronger set of constraints follow from the assumptions above:
1. (Positivity and boundedness): $0<\mu_J(t)<\frac{2 \beta F_d}{\pi(2 J+1)} e^{-\frac{2 \pi}{\beta} t}$;
2. (Monotonicity): $\mu_{J+1}(t)<\mu_J(t)$;
3. (Log-convexity): $\mu_{J+1}(t)^2 \leq \mu_J(t) \mu_{J+2}(t)$.
## Refs
- leading bound [[2015#Maldacena, Shenker, Stanford]] (main ref)
- subleading (infinitely many) [[2021#Kundu (Sep, a)]][](https://arxiv.org/pdf/2109.03826.pdf)
- extremal chaos: saturate all bounds [[Kundu202109b]]
- [[2021#Kundu (Apr)]]: constraining bulk higher derivative couplings using chaos bound
- causality bounds [[HashimotoSugiura2022]][](https://arxiv.org/pdf/2205.13818.pdf)
- velocity dependent bound [[2019#Mezei, Sarosi]] (and from a [[0179 Pole skipping|pole-skipping]] perspective: [[2020#Choi, Mezei, Sarosi]])
- from stability: [[Yoon2019]][](https://arxiv.org/pdf/1905.08815.pdf)
- bounds on the *generalised* [[0466 Lyapunov exponent|Lyapunov exponent]] from fluctuation-dissipation theorem: [[2022#Pappalardi, Kurchan]]
- relation to [[0040 Eigenstate thermalisation hypothesis|ETH]]: [[2019#Murthy, Srednicki (Jun, b)]]
- relation to [[0091 Boundary causality|boundary causality]]: [[2017#Afkhami-Jeddi, Hartman, Kundu, Tajdini]]
## Saturation
- holographic systems in the classical gravity limit
- SYK-like systems at low temperatures