News & views

Complex networks https://doi.org/10.1038/s41567-024-02507-y

Critical delay accumulation

Jari Saramäki Check for updates

In many schedule-based systems, timeliness is condensed matter, they have become an important topic in diverse
areas of science (Fig. 1).
paramount. A recent study provides insights One area where phase transitions are of particular importance is
into delay accumulation in networks, revealing network science6. Many natural and man-made systems can be mod-
elled as networks, and both their structure and the dynamics that
a phase transition with connections to the they support can display critical phenomena. For example, whether
physics of interfaces. networks are physical such as power grids, or more immaterial such
as social networks, their connectivity is governed by the percolation
There are many real-world systems where something flows through transition, which is in turn closely related to crossing the epidemic
complex networks: supply chains distribute food, goods, and compo- threshold in models of disease spreading on networks.
nents1, and transport networks move people from one place to another2. The novel phase transition reported by Moran and colleagues
In such systems, the timeliness of operations is crucial. However, belongs to the dynamical category. The authors investigated a general
a system-level picture of their resilience to delays has largely been model of schedule-based systems in which delays propagated and
missing. Now, writing in Nature Physics, José Moran and colleagues accumulated from one network node to another3 through a tempo-
have reported a phase transition in schedule-based systems between ral network whose links changed in time7. The nodes were endowed
two phases where delays are absorbed or grow depending on buffer with a temporal buffer that could absorb a finite amount of delay; in a
size3, which has connections to the physics of pinning4. real-world network, this would be analogous to, for example, an inven-
The general theory of phase transitions5 is a triumph of statisti- tory of necessary components held by a manufacturer. The authors
cal physics, explaining how physical systems suddenly move from showed that when the buffer was large, delays were absorbed by the
one phase to another as an external control parameter is varied — system, but when it was small, delays grew without bound (Fig. 1). At
for example, how magnetization or superconductivity appears at the boundary between these two phases — that is, at criticality — delay
low temperature. An intriguing feature of phase transitions is their avalanches of all sizes and durations appeared.
universality: systems in the same universality class cross the criti- This timeliness transition has interesting connections to other
cal boundary between two phases in the same way, regardless of areas of physics, highlighting the universality of phase transitions.
their exact microscopic details. It is therefore not too surprising that The behaviour of the critical exponents is reminiscent of glasses8.
even though phase transitions were once confined to the study of Moreover, the model has a close relationship with the bounded

B > Bc B = Bc B < Bc

Time = 0

Time = 1

Time = 2

Time = 3
Average delay

Average delay

Average delay

Time Time Time

Fig. 1 | Accumulation of delays in a temporal network. In the left column, the law size distribution. On the right, the buffer is too small, and the average delay
nodes have a temporal buffer B large enough to eventually absorb delays. In the grows at a constant rate. Node size reflects the delay accumulated at the node.
middle, the temporal buffer is critical, resulting in delay avalanches with power-

nature physics Volume 20 | August 2024 | 1228–1229 | 1228

News & views

Kardar–Parisi–Zhang4,9 equation which describes the time evolution and duration. As the world is still plagued by shortages of components
of a driven interface in the presence of a wall that confines it to one half owing to the Covid-19 pandemic, these findings ought to have some
of the space. Depending on the drive, the interface is either trapped in repercussions among policymakers. How can we, in preparation for
the vicinity of the wall or becomes unbound and drifts away from the future crises, ensure that our critical systems are resilient against
wall toward infinity, much like the delays in the model presented by sudden disruptions?
Moran and colleagues.
The delay accumulation model and the criticality it displays will Jari Saramäki
keep network scientists busy for a while. The authors showed that the Department of Computer Science, School of Science, Aalto University,
results hold when the model is run on a real-world temporal-network Espoo, Finland.
data set, but just how general are the results? If the network structure e-mail: jari.saramaki@aalto.fi
is varied, is the critical buffer size sensitive to some features? Is there a
way to construct optimally delay-resilient networks for a given buffer? Published online: 20 June 2024
Modellers of supply chains and transport networks also have an obvi-
ous interest in this work. Of course, it remains to be seen whether the References
1. Pichler, A. et al. Science 382, 270 (2023).
stylized, general model is directly applicable to the kinds of real-world 2. Guihaire, V. & Hao, J.-K. Transp. Res. A 42, 1251 (2008).
systems discussed above, each of which has its own idiosyncratic devia- 3. Moran, J. et al. Nat. Phys. https://doi.org/10.1038/s41567-024-02525-w (2024).
4. Hinrichsen, H. Adv. Phys. 49, 815 (2010).
tions from the model’s assumptions. 5. Goldenfeld, N. Lectures on Phase Transitions and the Renormalization Group (Westview
More broadly, the research of Moran and colleagues adds to the Press, 1992).
growing body of theoretical results showing that in networks full of 6. Newman, M. E. J. Networks 2nd edn (Oxford University Press, 2018).
7. Holme, P. & Saramäki, J. Phys. Rep. 519, 97 (2012).
interdependencies10, Adam Smith’s invisible hand might in fact lead 8. Götze, W. Complex dynamics of glass-forming liquids: A mode-coupling theory (Oxford
the system to its ruin. What might appear optimal from the myopic University Press, 2009).
point of view of each player might in fact be detrimental to the system 9. Muñoz, M. A. & Hwa, T. Europhys. Lett. 41, 147 (1998).
10. Battiston, S. et al. J. Econ. Dyn. Control 36, 1121 (2012).
as whole. In the language of supply chains1, if everyone keeps their
inventory as low as they can, the system can become vulnerable: delays Competing interests
may grow or, at criticality, result in avalanches of unpredictable size The author declares no competing interests.

nature physics Volume 20 | August 2024 | 1228–1229 | 1229