Access to this full-text is provided by Springer Nature.
Content available from Scientific Reports
This content is subject to copyright. Terms and conditions apply.
Scientic Reports | (2021) 11:5666 |
www.nature.com/scientificreports
Multilayer representation
of collaboration networks
with higher‑order interactions
E. Vasilyeva1,2, A. Kozlov1, K. Alfaro‑Bittner3,4*, D. Musatov1,5,6, A. M. Raigorodskii1,6,7,8,
M. Perc9,10,11 & S. Boccaletti1,3,12,13
Collaboration patterns oer important insights into how scientic breakthroughs and innovations
emerge in small and large research groups. However, links in traditional networks account only for
pairwise interactions, thus making the framework best suited for the description of two‑person
collaborations, but not for collaborations in larger groups. We therefore study higher‑order scientic
collaboration networks where a single link can connect more than two individuals, which is a natural
description of collaborations entailing three or more people. We also consider dierent layers of these
networks depending on the total number of collaborators, from one upwards. By doing so, we obtain
novel microscopic insights into the representativeness of researchers within dierent teams and their
links with others. In particular, we can follow the maturation process of the main topological features
of collaboration networks, as we consider the sequence of graphs obtained by progressively merging
collaborations from smaller to bigger sizes starting from the single‑author ones. We also perform the
same analysis by using publications instead of researchers as network nodes, obtaining qualitatively
the same insights and thus conrming their robustness. We use data from the arXiv to obtain results
specic to the elds of physics, mathematics, and computer science, as well as to the entire coverage
of research elds in the database.
Scientic collaboration networks are an important subset of complex social networks1–4. ey document patterns
of collaboration that we have formed to do research, and to arrive at new scientic discoveries and breakthroughs
that drive technological progress and innovation in our societies. e outstanding importance of science and
progress for the wellbeing of modern human societies, together with the consistent denition of scientic col-
laboration that is accurately documented in published research5, has given rise to a rich plethora of research
dedicated to the determination of structure and function of scientic collaboration networks6–12. Along the same
lines, citation networks13–15, bipartite author-publication networks16–19, hypergraphs of scientic output20, as well
as simplicial descriptions of publications and corresponding topological methods21,22, have also been considered
and studied in much detail.
However, despite the fact that traditional complex networks have come a long way in improving our under-
standing of economic, infrastructural, technological, as well as social and computer networks23–26, the past decade
has witnessed the rise of the narrative that the majority of these networks do not exist in isolation. Rather, many
are coupled together and therefore should be best described as interdependent or multilayer networks27,28. Indeed,
it has been shown that even tiny changes or a failure in one network layer can lead to a catastrophic cascade of
much more signicant failures across many other network layers29. It was a seminal discovery, and while some
OPEN
Russia.
Russia.
China.
Institute of Mathematics and Computer Science, Buryat State University, ul. Ranzhurova,
Italy. *
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol:.(1234567890)
Scientic Reports | (2021) 11:5666 |
www.nature.com/scientificreports/
argued that processes in dierent network layers could simply be added up and described as a conglomerate
process on a single-layer network, it soon became clear that, as is in general true for complex systems, the whole
is not simply the sum of its parts30–32. Multilayer networks have since found applications for better understanding
epidemic spreading33,34, vaccination35, evolution of cooperation36, and biological organization at dierent scales37.
We here map published scientic papers in a multilayer network: scientists are nodes in all layers, and a link
between two nodes in the
jth
(
j>1
) layer stands for the participation of the corresponding two scientists in a
publication jointly written by j co-Authors. is way, single-author publications form the rst layer, two-author
publications form the second layer, three-author publications form the third layer, and so on. In doing so, the
layers themselves already hold important information about the collaboration. It is namely easy to argue that
two researchers that are the only two authors on a publication have a much stronger link than two researchers
that have co-written a paper that has several hundred authors, as is oen the case in high-energy physics publica-
tions. Multilayer collaboration networks dened in this way thus naturally take into account the problems that
are commonly associated with unweighted single-layer collaboration networks12,38–41. Moreover, if we aggregate
all the layers, we simply obtain the complete scientic collaboration network, but with the added value that, as
we coalesce the layers obtained with ever larger collaboration sizes, we obtain novel microscopic insights into
the representativeness of researchers within dierent teams and their links with others, and we can follow the
maturation of topological features and the relevance each particular layer has in this process.
Another important distinction of our research to traditional scientic collaboration networks is that we
consider higher-order interactions to describe the networks. is is irrelevant for the rst and second layer, but
becomes theoretically much more convenient for the subsequent layers, where three or more coauthors are natu-
rally connected by a single higher-order link—a hyperlink – rather than a series of 2nd-order links connecting
pairs of researchers consecutively with one another. Although the value of higher-order interactions has been
recognized already in the early 70s by Atkin42,43 and Berge44, the interest peaked only recently with mounting
inability to converge on what constitutes a group or how to dene it consistently in the realm of social network
analysis45–49, and the interested reader can nd a comprehensive account on the role of higher-order interactions
in networked systems in Ref.49.
Here we use the formalisms of multilayer and higher-order networks, oen also called hypergraphs, to study
the maturation of dierent topological characteristics of collaboration networks in physics, mathematics, and
computer science by using the arXiv database50. And we also consider the entire coverage of research elds in
the same database. e question that we seek to answer is, how many layers does one need to obtain a proper
and robust description of the collaboration network? Or equivalently, is it possible to describe the collaboration
network by taking into account publications with only a couple of authors, for example up to layer four or ve?
Results
We refer to the information publicly available in the arXiv (https ://arxiv .org/, https ://githu b.com/mattb ierba um/
arxiv -publi c-datas ets/) database50. Data parsing was also made according to50. From the database, metadata on
1,679,779 articles were downloaded. en, information about 1,068,043 unique authors was parsed.
Let N be the number of authors in the database. e main idea is to represent the data-set as a primal
H=(V,EH)
co-authorship hypergraph, in which
V={v1,...,vN}
is the set of nodes (authors) and
EH
is a set of
hyperedges accounting for articles. In this representation, an article co-authored by d authors corresponds then
to an hyperedge grouping the d authors of the paper, as it is schematically depicted in Fig.1a. In Fig.1a nodes
are therefore labeled with the name of the authors, whereas coloured hyperlinks are labeled by the correspond-
ing paper identier in the arXiv (with dierent colours, moreover, standing for dierent numbers of coauthors).
Notice that this representation allows to distinguish the case of two (or a limited group of ) researchers that are
the only authors of a publication and therefore they supposedly have a strong ties, from that of two (or a limited
group of) researchers that just participate in huge collaboration projects giving rise to papers that have several
hundred authors.
Moreover, the primal hypergraph
H=(V,EH)
can be associated to a dual hypergraph
H∗=(V∗,E∗
H)
in
which
V∗
is the set of articles and
E∗
H
groups papers written by the same author (in collaboration with others, or
individually), as schematically depicted in Fig.1b. One can also introduce a kind of “pairwise approximation” of
H, given by a undirected graph
G=(V,EG)
where an edge between authors reects the existence of a joint paper
(independently on the number of coauthors). erefore, each hyperedge of H corresponds to a clique in G. With
the same spirit, the dual graph
G∗=(V∗,E∗
G)
is the pairwise approximation of
H∗
where nodes correspond to
articles and existence of an edge indicates that two articles have at least one joint author.
e hypergraph H (and its dual
H∗
) as well as the graphs G and
G∗
can be viewed as multilayer networks
with layering index dened by the number of article’s coauthors, and represented in Fig.1 by dierent colours
assigned to dierent papers (yellow denoting Manuscripts authored by a single scholar, green papers co-authored
by two scholars, etc...). en, one can operate a progressive fusion of such layers, and obtain the hypergraph
(graph, dual hypergraph and dual graph respectively) H(n) (G(n),
H∗(n)
,
G∗(n)
), where only papers with no
more than n coauthors are considered. Let
¯n
be the number of maximal layer in the statistics, and let us simplify
the notations further by writing
H(¯n)
(
G(¯n)
,
H∗(¯n)
,
G∗(¯n)
) as H (G,
H∗
,
G∗
). H, G,
H∗
and
G∗
are the “asymptotic”
graphs and they are actually the “classical” representations given to collaborations’ data, where all level of co-
authorship (as much those implying just a few scholars as those implying instead thousands of scholars) are
mixed together, and whose main properties have been largely characterized by the denition and calculation of
a wealth of topological measures.
Our idea is, instead, that such topological measures are actually maturating as one progressively fuse the
distinct layers. In other words, we suggest that there exists a given
˜n
at which each specic network’s topologi-
cal property maturates, i.e. it assumes the asymptotic value which is calculated on H, G,
H∗
and
G∗
. Obviously,
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol.:(0123456789)
Scientic Reports | (2021) 11:5666 |
www.nature.com/scientificreports/
such maturation level may be dierent for dierent elds of cooperation (as processes of scientic collaboration
formation vary from eld to eld) and for dierent topological measures as well, and it is of great interest to
study how distinct topological properties emerge at distinct levels of fusions (i.e. taking into account only proper
subsets of the original data, where the number of coauthors of a given Manuscript is limited).
Finally, it has to be noticed that all articles in the database are related to eight main areas (physics, mathemat-
ics, quantitative biology, computer science, quantitative nance, statistics, electrical engineering and systems
science, economics) and in the present study we give our new representation of co-authorship networks for the
following elds (in parentheses we report the notation for each one of the obtained asymptotic graphs):
• physics (
Hphys
,
Gphys
and the dual ones),
• math (
Hmath
,
Gmath
and the dual ones),
• computer science (
Hcs
,
Gcs
and the dual ones),
• all eight areas together (H, G and the dual ones).
A rst characterization of the hypergraphs. A rst rough characterization of the primal and dual
graphs is shown in Fig.2, where we report the complementary cumulative distribution function (CCDF) for
H(
Hphys,Hmath ,Hcs
) in panel (a) and for G(
Gphys,Gmath ,Gcs
) in panel (b).
e CCDF is dened with the following expression:
where F(x) is the cumulative distribution function. If the tail of the distribution is tting a the power-law, then
where
xm
is a proper parameter, and
γ
can be estimated as the slope of the linear t in a log–log scale. In Fig.2a,b
we report the CCDF for nodes’ and hyperedges’ degree distributions of H,
Hphys,Hmath ,Hcs
and of their dual
graphs. From the gures it is apparent that the dierent graphs deviate from a power law in their tails. e
distributions in physics (red curves) can be seen as consisting of two dierent parts which actually seems to
correspond to dierent power law exponents. Most likely, such a property is due to experimental works in huge
collaborations. Hyperedges’ degree distributions in math and CS deviate from the power law only in tails. e
(1)
CCDF(x)=1−F(x),
CCDF(x)∼x−(γ +1),x>xm,γ>0
Figure1. Schematic illustration of the co-authorship hypergraph (a) and of the dual hypergraph (b). In panel
(a) nodes are authors, and hyperlinks are co-authored Manuscript. e hyperlinks are labeled with letters and
colours. e legend at the bottom of the Figure reports for each letter the corresponding Manuscript’s identier
in the ArXiv. In the legend, moreover, Manuscripts are grouped in coloured boxes, and dierent colours stand
for a dierent number of coauthors: yellow papers are authored by a single Scholar, whereas green, red and blue
Manuscripts are co-authored by two, three and four Scholars, respectively. Panel (b) contains a sketch of the dual
representation, where nodes are now papers [labeled with the same colours and letters than in panel (a)], and
links are labeled with the name of the authors who participated in the co-authorship of the Manuscripts.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol:.(1234567890)
Scientic Reports | (2021) 11:5666 |
www.nature.com/scientificreports/
distributions corresponding to the entire database display the same features as those in physics, as papers related
to this science prevail in the arXiv collection.
e results shown in Fig.2 point to the fact that there are papers with an extremely high number of coauthors.
However, as already discussed in the Introduction, real patterns of authors’ interactions are unlikely to be deter-
mined by such huge collaborations. erefore, it seems reasonable to analyse how papers with large numbers
of authors aect the network properties, or equivalently to analyse the maturation properties of the multilayer
networks dened in the previous sub-section.
The maturation process of topological features in the multilayer graph. e main objective of
our study is to compare stabilization and maturation patterns of co-authorship networks describing scientic
cooperation in dierent elds. To this purpose, we analyze how dierent topological properties change when the
layer index n changes.
Let x(n) be some property (i.e., some topological measure) of a graph G(n). To simplify the notations, we
omit the argument for the case of the maximal layer
¯n
, and we write
x=x(¯n)
. We say that the specic property
x(n) is maturated at the layer
˜n(x)
if:
where
ε
is a small constant accounting for an acceptable accuracy (i.e., a tolerable dierence). In all our calcula-
tions, we use
ε=0.05
.
In order to illustrate the concept of maturation, Fig.3 anticipates some of the major points and conclusions
of our Manuscript, and reports three panels, each one displaying the maturation behavior (or the absence of
maturation) of important topological features, as the fusion index n of layers increases.
Precisely, Fig.3a compares the behavior of the average degree
k
versus
n/¯n
for the areas of mathematics (light
red curve) and computer science (light blue curve). Normalization in the horizontal axis is needed because the
two areas have actually distinct maximum numbers
¯n
of layers. It is clearly seen that
k
maturates rather early
in the area of mathematics:
k(n/¯n)
is a monotonically increasing curve which attains its asymptotic value (the
value at
n=¯n
) already at layer
˜n=8
. e horizontal light red bar in panel (a), indeed, stands for the (plus or
minus)
ε=0.05
error around the asymptotic value
k(¯n)
, and it is evident that the curve
k(n/¯n)
stays inside
the error area for all values of
˜n≤n≤¯n
. At variance, the average degree never maturates in the area of computer
science, as witnessed by the light blue line in Fig.3a: once again the horizontal light blue bar indicates the (plus
or minus)
ε=0.05
error around the asymptotic value
k(¯n)
, but now the curve
k(n/¯n)
never enters the error
area before attaining its asymptotic value at
n=¯n
.
Dierent topological features may maturate at dierent values of
˜n
, as illustrated in panel (b) of Fig.3. Namely,
the upper (lower) part of panel (b) reports the evolution of the diameter d (of the shortest path L) in the areas
of mathematics (red curve) and computer science (light blue curve). d maturates at layer 3 in the area of math-
ematics and at layer 10 in the area of computer science; L instead maturates at layer 4 in mathematics and again
at layer 8 in computer science. It is seen, moreover, that dierent fusion stages at which maturation in dierent
(2)
˜n
(x)=arg min
n
n:∀k≥n−→ |x(k)−x|
x
≤ε
,
Figure2. (a) Complementary cumulative distribution functions (CCDF, see text for denition) for the
primal graphs obtained from the data-set. e distributions are functions of the nodes’ degree distributions for
H(
Hphys,Hmath ,Hcs
) and of hyperedges’ degree distributions for the respective dual hypergraphs. (b) CCDF
for the dual graphs, which are functions of the hyperedges’ degree distribution in H(
Hphys,Hmath ,Hcs
) and of
the nodes degree distribution in the respective dual hypergraphs. Curves are coloured according to the dierent
speciality from which papers are extracted from the data-set (see the colour code at the top right of each panel).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol.:(0123456789)
Scientic Reports | (2021) 11:5666 |
www.nature.com/scientificreports/
areas takes place is not the simple consequence of the normalization of the fusion index n to the relative maxi-
mum number of layers in the area.
Finally, panel (c) of Fig.3 anticipates another important conclusion of our study: in some cases dual graphs,
where hyperlinks connect publications instead of coauthors, may represent a better rendering of collaboration
networks, in that some topological features maturate in dual graphs, whilst they never maturate in the direct
graphs. is is illustrated with reference to the average degree
k
in the area of physics: it is clearly seen that the
curve
k(n)
for the direct graph
Gphys
(light blue line) does not display any maturation feature, whereas
k(n)
(light red line) maturates at layer 574 in
G∗
phys
.
It is essential to remark that the calculation of some graph’s topological measure for all layers n may have an
associated very high computational demand. erefore, the networks G(n) and
G∗(n)
are here analysed using a
rather sparse grid, and aer that the dependencies are interpolated using splines. e procedure, however, do
not aect (nor distorts) the conclusions which we are oering below.
General properties. e natural starting point is the analysis of the networks’ global substructures.
Let CC be a set of network’s connected components, and LCC be the set of vertices in the largest connected
component. e following notation can be introduced:
•
m=|EG|
,
•
NCC =|CC|
,
•
sLCC
=
|LCC|
N
.
Table1 reports the maximal number of layers (
¯n
), the number of nodes (N), the number of edges (m), the number
of connected components (
NCC
), the relative size of the largest connected component (
sLCC
), and the maturation
layer’s numbers for all these features (
˜n(·)
).
e rst signicant feature which should be noticed is the dierentiation in
¯n
for the dierent disciplines.
Namely, physics corresponds to the highest value of
¯n
(2831 layers). Moreover, the number of nodes N in physics
maturates quite late if compared with math and CS, therefore a consistent number of Scholars in this eld write
papers only in rather big collaborations. In contrast, in math one has see the smallest number of layers (67), and
not only N. Furthermore, not only N maturates early (already at level 5) in this eld, but even the edges’ number
maturates at level eight, which implies that focusing only on papers with no more than eight authors one has
an almost complete description of the graph representing the math discipline. For the other graphs, one sees
instead that the number of edges signicantly changes at all levels of the fusion process, up to the nal layers.
Another notable feature which appear from Table1 is related to the number of connected components. is
property maturates relatively early for all elds, as well as for the whole graph. erefore, besides the largest
connected component, the general backbone of the other part of the graph is formed by many clusters (con-
nected components) each one containing a relatively small number of papers. On the other hand, the largest
connected component consists of about 80% of nodes for the elds of math and CS and 93% of nodes in physics.
e relative size of the LCC in the whole graph is 90%, which means that the LCC of the whole graph contains
all authors from the LCC’s of the dierent elds’ graphs. is notion follows from the fact that if we suppose
that the smallest LCC from Table1 (the one of math) is not included into LCC of the whole graph than size of
Figure3. Illustration of the maturation process of dierent topological features. Panel (a): the average degree
k
vs. the normalized fusion index
n/¯n
(see text for denitions), for the areas of mathematics (light red curve)
and computer science (light blue curve). e horizontal light red and light blue bars stand for the (plus or
minus)
ε=0.05
errors around the respective asymptotic values
k(¯n)
. Panel (b): the upper (lower) sub-panel
reports the evolution of the diameter d (of the shortest path L) in the areas of mathematics (light red curve) and
computer science (light blue curve). d maturates at layer 3 in the area of mathematics and at layer 10 in the area
of computer science; L instead maturates at layer 4 in mathematics and again at layer 8 in computer science.
Notice that dierent topological features maturate at dierent fusion stages. Panel (c): the average degree
k
in
the area of physics vs. the fusion index n, for the direct graph
Gphys
(light blue line) and for the dual graph
G∗
phys
(light red line).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol:.(1234567890)
Scientic Reports | (2021) 11:5666 |
www.nature.com/scientificreports/
LCC of the whole graph should be not more than
(1.07 −0.76 ·0.21)/1.07 =0.85
. As a conclusion, there is an
important role of interdisciplinary links connecting Authors from dierent elds. In Table1 we also report the
same properties for the dual graphs. e remarkable result is that even the edge number maturates for all elds
in the dual graphs. Most likely this occurs because an extremely large number of authors is much more frequent
than an extremely large number of papers written by a particular author, and moreover such papers have to have
dierent sets of coauthors in order to contribute to the number of edges. erefore, papers with large number
of authors contribute large cliques in G but not in its dual graph, and therefore, in this context, the dual graph
constitutes a better representation of the social collaborations than its primal counterpart.
Degree distribution. e second step of our analysis is the description of the local networks’ properties,
and we start with the study of the degree distributions. Let
ki(n),i=1, ...,N(n)
be the degree of node i in G(n).
Our results show that, for all graphs analysed in the current study, the probability distribution functions (PDFs)
of the degree k are fat-tailed, with tails well described by a power law scaling with exponent
γ
:
Table2 reports the values of the mean degree in the four graphs studied (
k
) and the estimated tail exponents
γ
for the corresponding degree distributions. e only eld in which we see a maturation of the mean degree is
(3)
p
(k)∼
1
k
γ
,
Table 1. Maturation indices and maturation values of the main general properties of primal and dual graphs.
All notations and denitions are reported in the text. e symbol “–” reects the fact that the property does
not maturate, implying that signicant changes in the property’s value occur at all fusion indices, up to the nal
layer (the reported values are therefore the “asymptotic” ones obtained by fusing all layers).
All Math CS Phys
¯n
2831 67 427 2831
G
N,×106
1.07 0.21 0.28 0.71
˜n(N)
26 5 9 44
m,×107
4.11 0.05 0.13 3.96
˜n(m)
– 8 – –
NCC,×104
5.18 2.74 2.08 2.48
˜n(NCC)
11 4 6 23
sLCC
0.90 0.76 0.79 0.93
˜n(sLCC)
7 4 6 7
G∗
N,×106
1.68 0.44 0.26 1.08
˜n(N)
8 4 6 10
m,×107
10.11 0.82 0.59 8.27
˜n(m)
522 5 9 574
NCC,×104
5.18 2.7 2.08 2.48
˜n(NCC)
11 4 6 23
sLCC
0.94 0.86 0.86 0.95
˜n(sLCC)
4 3 5 4
Table 2. Maturation indices and maturation values of the degree distribution’s properties for the primal and
dual graphs. All notations and denitions are reported in the text. e symbol “–” reects the fact that the
property does not maturate, implying that signicant changes in the property’s value occur at all fusion indices,
up to the nal layer (the reported values are therefore the “asymptotic” ones obtained by fusing all layers).
All Math CS Phys
¯n
2831 67 427 2831
G
k
77.01 4.62 8.97 111.51
˜n(k)
– 8 – –
γ
1.7 3.6 2.6 1.6
˜n(γ)
498 – – 1411
G∗
k
120.54 36.65 44.33 153.54
˜n(k)
522 5 8 574
γ
2.8 3.3 3.9 2.6
˜n(γ)
756 2 7 495
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol.:(0123456789)
Scientic Reports | (2021) 11:5666 |
www.nature.com/scientificreports/
math, which is also characterized by the highest tail exponent. e fat-tailed nature of the degree distribution is
the most likely reason for the absence of maturation in the mean degree, as well as in the tail exponent estima-
tion. In the cases of the whole graph and of the graph of physics, one sees that maturation, however, occurs at a
very high value of the fusion index. Even if such distribution maturates, sample estimations of such values are
oen very sensitive to additional observations or data. Table2 reports also the results for the dual graphs. One
immediately sees that the mean degrees of all graphs under consideration maturate and, moreover, the exponents
of the respective power law distributions are signicantly higher. erefore, once again the dual graphs estimates
seem to provide a more accurate characterization.
Network clustering. One of the most important graph’s properties is clustering. Such a measure, indeed,
accounts for networks’ transitivity, and in the context of co-authorship graph it describes how oen two coau-
thors of one particular author are coauthors themselves in other papers. Quantication of clustering’s eects can
be obtained by measuring two dierent coecients: the global and the local clustering ones. e global cluster-
ing coecient is dened by the following expression:
where
#K3
is the number of triangles in the graph and
#P2
is the number of connected chains of length two.
e local clustering coecient of vertex i is instead calculated as
where
EG
is the set of edges of graph G,
Ni
is the set of i’s neighbors. I.e. local clustering coecient measures the
fraction of connected triples around node i. e overall graph clustering property
c
can be obtained by averag-
ing the local clustering coecient of Eq.(5) over all nodes:
One can easily see that the expression (Eq. (4)) can be rewritten as
FromEq. (7) it follows that in calculating the global clustering coecient the higher is the degree of the nodes
the higher its weight in the average, whereas
c
takes all nodes equivalently. erefore, the higher the dierence
between C and
c
is, the higher is the non-uniformity of clustering distribution between nodes.
Table3 shows the clustering coecients estimation and maturation for all primal and dual graphs. e rst
notable feature is that the global clustering coecient never maturates, while the averages of the local clustering
coecient always do. is naturally follows from the fact that papers from the last layers are associated with
larger numbers of additional triangles, and they also contribute a huge number of edges, thus enlarging nodes’
degrees signicantly, which are then used to calculate weights in the average of the global clustering coecient
[see Eq.(7)]. e smallest values of the clustering coecients are in the eld of math, which also can be distin-
guished for signicant dierence between C and
c
. Namely, in math global clustering is two times less than the
averaged local one. erefore in maths nodes with high degree are less clustered then the ones with small degree.
In dual graphs, both global and local clustering coecients maturate. Moreover, the averaged local clustering
coecients maturate earlier than the ones calculated in the primal graphs. Furthermore, the levels of maturation
(4)
C
=
3#K
3
#P2
,
(5)
c
i=
|{j,k∈E
G
:j,k∈N
i
}|
C2
|Ni|
,
(6)
c=
1
N
i∈V
ci
.
(7)
C
=i∈VC
2
|Ni|ci
i∈V
C2
|Ni|
.
Table 3. Maturation indices and maturation values of the graphs’ clustering properties. All notations and
denitions are reported in the text. e symbol “–” reects the fact that the property does not maturate,
implying that signicant changes in the property’s value occur at all fusion indices, up to the nal layer (the
reported values are therefore the “asymptotic” ones obtained by fusing all layers).
All Math CS Phys
¯n
2831 67 427 2831
G
C0.57 0.24 0.70 0.57
˜n(C)
– – – –
c
0.65 0.48 0.69 0.68
˜n(c)
10 5 6 15
G∗
C0.27 0.78 0.62 0.26
˜n(C)
546 4 7 522
c
0.72 0.76 0.68 0.70
˜n(c)
5 2 3 7
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol:.(1234567890)
Scientic Reports | (2021) 11:5666 |
www.nature.com/scientificreports/
in the whole graph, in physics and in maths are the same as those corresponding to the maturation of the number
of edges (
˜n(m)
in Table1). In CS, maturation of the global clustering occurs at the 8-th level, while number of
edges maturates at the 9-th level. In the physics dual graph, there is a signicant dierence between values of local
and averaged global clustering: the former is more than two times less than the latter. Most likely, this property
is the consequence of the existence of collaborative papers with large degree connected with other papers writ-
ten by extremely large number of authors. However, such “connecting” authors may have not a close relation of
collaboration between each other, and therefore papers authored by them are not necessarily neighbors in the
dual graph.
Diameter and characteristic path length. e essential measure describing closeness between two par-
ticular authors (papers) is the shortest path. Based on this measure two important characteristics of a graph can
be calculated. e rst is the diameter (d)—the maximum shortest path for all pairs of nodes in the LCC. e
second is the characteristic path length (L)—the mean shortest path for all pairs of nodes in the LCC.
e maturation analysis for d and L are presented in Table4. e characteristic path length properties in
physics (and, as a consequence, in the whole graph) dier signicantly from all other elds: the value of L is
less than half those in math and CS. However, this value changes signicantly on the last layers, therefore, this
property is highly dependent on collaborative papers. Interestingly, graphs’ diameters maturate in all elds.
e maturation indices in math and CS are close to the values obtained for the number of nodes. erefore, in
these elds papers with relatively large number of authors are basically joint with those who are already in the
same community. e dierence in physics, instead, indicates that large collaborative papers may inuence the
network’s community structure.
Similar conclusions can be drawn from the results of the dual graphs, for which even in the case of physics
(and the whole network) the characteristics path length maturates. Its maturation appears quite late, but it should
be noted that it happens much earlier than edges number maturates. In CS, maturation of both the diameter and
the characteristic path length appears earlier than in the primal one. e same is true for the diameter in the
eld of math. However, characteristic path length in math dual graph maturates later than in the primal graph
of this eld.
Centrality and eciency. As nodes in the networks have very dierent importance or relevance, various
measures of nodes’ centrality have been proposed in the literature. As the distribution of nodes’ centralities in
the network (the so-called centrality vector) contains very relevant information on the graphs structure and
function, maturation of the centrality vectors is an important signal of the network maturation as a whole. We
here report the maturation properties of the mean betweenness and closeness centrality measures, which will
be dened momentarily. On the other hand, we also focus here on network’s eciency, which in real social
networks describes the so called “small-world” property—the fact that information transfer is very ecient in
such networks51.
Node i’s betweenness centrality
bi
is dened as
where |P(j,k)| is the total number of shortest paths between nodes j and k, and |P(j,k,i)| is the number of shortest
paths between j and k which pass through node i. Mean betweenness
b
of the graph is obtained by averaging
over all nodes, and in the paper we calculate it only for nodes belonging to the LCC.
Node i’s closeness centrality
qi
is dened as
(8)
b
i=
2
(N−1)(N−2)
j
�=
i,k
�=
i
|P(j,k,i)|
|P(j,k)|
,
Table 4. Maturation indices and maturation values of the graphs’ diameter and characteristic path length. All
notations and denitions are reported in the text. e symbol “–” reects the fact that the property does not
maturate, implying that signicant changes in the property’s value occur at all fusion indices, up to the nal
layer (the reported values are therefore the “asymptotic” ones obtained by fusing all layers).
All Math CS Phys
¯n
2831 67 427 2831
G
d21 25 26 21
˜n(d)
436 3 10 425
L3.1 7.3 6.1 2.8
˜n(L)
– 4 8 –
G∗
d21 24 26 20
˜n(d)
402 4 6 434
L5.4 8.8 5.2 4.7
˜n(L)
329 9 8 430
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol.:(0123456789)
Scientic Reports | (2021) 11:5666 |
www.nature.com/scientificreports/
where d(i,j) is the length of the shortest path between i and j. Once again, the mean closeness
q
is obtained by
averaging over all nodes, and limiting ourselves to the set of nodes in the LCC.
Network’s eciency is dened by
Table5 shows the results for E,
q
and
b
. In co-authorship graphs of math and CS papers with extremely large
number of authors do not aect the values of the listed properties and, moreover, maturation appears relatively
early in both disciplines. is is in agreement with the results of the previous sub-section, where characteristics
path length’s maturation was analysed. Moreover, the maturation levels of eciency and betweenness for these
two elds are close to
˜n(L)
. In the dual graphs, the same conclusion can be made only for math and physics.
In the case of CS, the dual graph does not instead maturate, and this is the only case in which the dual graph
representation seems to provide a less accurate representation of the data. It has to be noticed that, for the CS
dual graph and the one for all elds, maturation of centrality and eciency (not reported here) occurs when
ε
is slightly increased (i.e., when
ε=0.1
).
Discussion
In summary, we have studied patterns of collaboration in the arXiv database by using the formalism of multilayer
higher-order networks, where each layer corresponds to the number of collaborators on publications that are
considered for that layer. For layer three, corresponding to three-author publications, and onwards, we have
also used higher-order links to connect groups of authors as a much more convenient and theoretically elegant
description of group interactions. By doing so, we were able to monitor separately how each relevant topologi-
cal feature of the network matures toward the value that was measured for the complete classical collaboration
network. We have also demonstrated that our representation reveals the true nature of collaborations among
researchers, which is fundamentally dierent when they coauthor a paper in a small group, implying an intense
and meaningful research relationship, as opposed to a collaboration in a huge group of coauthors were only very
few actually share any noteworthy contact.
In terms of implications for specic research elds, our research shows that dierent topological features
mature at dierent fusion indices for dierent research elds. Earlier for elds where the number of authors on
a particular publication is traditionally low, as in mathematics, and later for elds where large collaborations are
more common, as in physics. Either way, our representation allows us to progressively follow how the nal values
that determine the topological features of collaboration networks emerge as the fusion index, i.e., the number
of layers that have been fused together, increases. is thus oers a completely new and fresh microscopic view
into the collaboration patterns of researchers across dierent disciplines and depth of contact.
It is also worth noting that our research conrms, in line with previous research20,52, that the alternative
representation of collaboration networks, where hyperlinks connect publications instead of coauthors, yield a
better representation in that for these type of collaboration networks all topological features eventually mature
as layers are coalesced, whilst in the classical representation some topological feature never mature.
(9)
q
i=
1
j
∈
V,j
�=
i
d(i,j)
,
(10)
E
=
1
N(N−1)
i,j
∈
V,i
�=
j
1
d(i,j)
.
Table 5. Maturation indices and maturation values of the graphs’ centrality and eciency indicators. All
notations and denitions are reported in the text. e symbol “–” reects the fact that the property does not
maturate, implying that signicant changes in the property’s value occur at all fusion indices, up to the nal
layer (the reported values are therefore the “asymptotic” ones obtained by fusing all layers).
All elds Math CS Physics
¯n
2831 67 427 2831
G
E0.40 0.14 0.17 0.43
˜n(E)
– 4 7 –
q
3.4 ·10−3
8.8 ·10−7
7.4 ·10−7
2.5 ·10−3
˜n(q)
– 6 15 –
b
2.4 ·10−5
4.0 ·10−5
2.3 ·10−5
1.9 ·10−5
˜n(b)
– 4 18 –
G∗
E0.24 0.13 0.21 0.27
˜n(E)
– 9 – 540
q
3.2 ·10−3
4.5 ·10−4
4.5 ·10−4
2.4 ·10−3
˜n(q)
– 9 – 444
b
4.7 ·10−4
9·10−4
4.7 ·10−4
3.8 ·10−4
˜n(b)
– 9 – 447
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol:.(1234567890)
Scientic Reports | (2021) 11:5666 |
www.nature.com/scientificreports/
ese insights create many possible directions for future research. For example, one viable avenue worth
exploring is to customize growth models of hypergraphs that would take into account the fact that a given
topological feature must mature at a given stage of fusion. We would thereby obtain a more apt theoretical
description of scientic collaboration, which would in turn promise a better understanding of this vital process
that upkeeps modern human societies. It would also be interesting to look at the maturation of other network
properties, such as the community structure and various centrality measures. Lastly, it would also be worth while
exploring how the proposed multilayer higher-order network formalism works in other forms of documented
collaboration, such as on patents and legal proceedings. We hope our research will prove inspirational towards
this goals in the near future.
Received: 28 December 2020; Accepted: 19 February 2021
References
1. Newman, M. E. J. Networks: An Introduction (Oxford University Press, 2010).
2. Estrada, E. e Structure of Complex Networks: eory and Applications (Oxford University Press, 2011).
3. Barabási, A.-L. Network Science (Cambridge University Press, 2016).
4. Latora, V., Nicosia, V. & Russo, G. Complex Networks: Principles, Methods and Applications (Cambridge University Press, 2017).
5. Newman, M. E. J. e structure of scientic collaboration networks. Proc. Natl. Acad. Sci. U.S.A. 98, 404 (2001a).
6. Newman, M. E. J. Scientic collaboration networks. I. Network construction and fundamental results. Phys. Rev. E 64, 016131
(2001b).
7. Newman, M. E. J. Scientic collaboration networks. II. Shortest paths, weighted networks, and centrality. Phys. Rev. E 64, 016132
(2001c).
8. Newman, M. E. J. Assortative mixing in networks. Phys. Rev. Lett. 89, 208701 (2002).
9. Fan, Y. et al. Network of econophysicists: A weighted network to investigate the development of econophysics. Int. J. Mod. Phys.
B 18, 2505 (2004).
10. Perc, M. Growth and structure of Slovenia’s scientic collaboration network. J. Informetrics 4, 475 (2010).
11. Krumov, L., Fretter, C., Müller-Hannemann, M., Weihe, K. & Hütt, M.-T. Motifs in co-authorship networks and their relation to
the impact of scientic publications. Eur. Phys. J. B 84, 535 (2011).
12. Pan, R. K. & Saramäki, J. e strength of strong ties in scientic collaboration networks. EPL 97, 18007 (2012).
13. Redner, S. How popular is your paper? An empirical study of the citation distribution. Eur. Phys. J. B 4, 131 (1998).
14. Lehmann, S., Lautrup, B. & Jackson, A. D. Citation networks in high energy physics. Phys. Rev. E 68, 026113 (2003).
15. Kuhn, T., Perc, M. & Helbing, D. Inheritance patterns in citation networks reveal scientic memes. Phys. Rev. X 4, 041036 (2014).
16. Goldstein, M. L., Morris, S. A. & Yen, G. G. Group-based Yule model for bipartite author-paper networks. Phys. Rev. E 71, 026108
(2005).
17. Peltomäki, M. & Alava, M. Correlations in bipartite collaboration networks. J. Stat. Mech. 6, P01010 (2006).
18. Tian, L., He, Y., Liu, H. & Du, R. A general evolving model for growing bipartite networks. Phys. Lett. A 376, 1827 (2012).
19. Zhou, Y.-B., Lü, L. & Li, M. Quantifying the inuence of scientists and their publications: distinguishing between prestige and
popularity. New J. Phys. 14, 033033 (2012).
20. Lung, R. I., Gaskó, N. & Suciu, M. A. A hypergraph model for representing scientic output. Scientometrics 117, 1361 (2018).
21. Moore, T. J., Drost, R. J., Basu, P., Ramanathan, R., Swami, A. Analyzing collaboration networks using simplicial complexes: A
case study. In 2012 Proceedings IEEE INFOCOM Workshops 238–243 (IEEE, 2012).
22. Patania, A., Petri, G. & Vaccarino, F. e shape of collaborations. EPJ Data Sci. 6, 18 (2017).
23. Albert, R. & Barabási, A.-L. Statistical mechanics of complex networks. Rev. Mod. Phys. 74, 47 (2002).
24. Newman, M. E. J. e structure and function of complex networks. SIAM Rev. 45, 167 (2003).
25. Boccaletti, S., Latora, V., Moreno, Y., Chavez, M. & Hwang, D. Complex networks: Structure and dynamics. Phys. Rep. 424, 175
(2006).
26. Fortunato, S. Community detection in graphs. Phys. Rep. 486, 75 (2010).
27. Boccaletti, S. et al. e structure and dynamics of multilayer networks. Phys. Rep. 544, 1 (2014).
28. Kivelä, M. et al. Multilayer networks. J. Complex Netw. 2, 203 (2014).
29. Buldyrev, S. V., Parshani, R., Paul, G., Stanley, H. E. & Havlin, S. Catastrophic cascade of failures in interdependent networks.
Nature 464, 1025 (2010).
30. Gómez, S. et al. Diusion dynamics on multiplex networks. Phys. Rev. Lett. 110, 028701 (2013).
31. De Domenico, M. et al. Mathematical formulation of multilayer networks. Phys. Rev. X 3, 041022 (2013).
32. De Domenico, M., Solé-Ribalta, A., Omodei, E., Gómez, S. & Arenas, A. Ranking in interconnected multilayer networks reveals
versatile nodes. Nat. Commun. 6, 6868 (2015).
33. Pastor-Satorras, R., Castellano, C., Van Mieghem, P. & Vespignani, A. Epidemic processes in complex networks. Rev. Mod. Phys.
87, 925 (2015).
34. de Arruda, G. F., Rodrigues, F. A. & Moreno, Y. Fundamentals of spreading processes in single and multilayer complex networks.
Phys. Rep. 756, 1 (2018).
35. Wang, Z. et al. Statistical physics of vaccination. Phys. Rep. 664, 1 (2016).
36. Wang, Z., Wang, L., Szolnoki, A. & Perc, M. Evolutionary games on multilayer networks: A colloquium. Eur. Phys. J. B 88, 124
(2015).
37. Gosak, M. et al. Network science of biological systems at dierent scales: A review. Phys. Life Rev. 24, 118 (2018).
38. Barrat, A., Barthelemy, M., Pastor-Satorras, R. & Vespignani, A. e architecture of complex weighted networks. Proc. Natl. Acad.
Sci. U.S.A. 101, 3747 (2004).
39. Opsahl, T., Colizza, V., Panzarasa, P. & Ramasco, J. J. Prominence and control: e weighted rich-club eect. Phys. Rev. Lett. 101,
168702 (2008).
40. Ramasco, J. J. & Morris, S. A. Social inertia in collaboration networks. Phys. Rev. E 73, 016122 (2006).
41. Ke, Q. & Ahn, Y.-Y. Tie strength distribution in scientic collaboration networks. Phys. Rev. E 90, 032804 (2014).
42. Atkin, R. H. From cohomology in physics to Q-connectivity in social science. Int. J. Man-Mach. Stud. 4, 139 (1972).
43. Atkin, R. H. Mathematical Structure in Human Aairs (Heinemann Educational Publishers, 1974).
44. Berge, C. Graphs and Hypergraphs (North-Holland Pub. Co., 1973).
4 5. Estrada, E. & Rodríguez-Velázquez, J. A. Subgraph centrality and clustering in complex hyper-networks. Physica A 364, 581 (2006).
46. Benson, A. R., Gleich, D. F. & Leskovec, J. Higher-order organization of complex networks. Science 353, 163 (2016).
47. Perc, M. et al. Statistical physics of human cooperation. Phys. Rep. 687, 1 (2017).
48. Alvarez-Rodriguez, U. et al. Evolutionary dynamics of higher-order interactions in social networks. Nat. Hum. Behav. 1, 1 (2020)
((in press)).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Vol.:(0123456789)
Scientic Reports | (2021) 11:5666 |
www.nature.com/scientificreports/
49. Battiston, F. et al. Networks beyond pairwise interactions: structure and dynamics. Phys. Rep. 874, 1 (2020).
50. Clement, C. B., Bierbaum, M., O’Keee, K. P., & Alemi, A. A. On the Use of ArXiv as a Dataset. arXiv :1905.00075 (2019).
51. Latora, V. & Marchiori, M. Ecient behavior of small-world networks. Phys. Rev. Lett. 87, 198701 (2001).
52. Gaskó, N., Lung, R. I. & Suciu, M. A. A new network model for the study of scientic collaborations: Romanian computer science
and mathematics co-authorship networks. Scientometrics 108, 613 (2016).
Acknowledgements
E.V. acknowledges the project “Post-crisis world order: challenges and technologies, competition and coopera-
tion” supported by Ministry of Science and Higher Education of the Russian Federation (agreement number
075-15-2020-783). M.P. was supported by the Slovenian Research Agency (Grant Nos. P1-0403 and J1-2457). e
research of D. Musatov and A. M. Raigorodskii was supported by the Russian Federation Government (Grant
number 075-15-2019-1926).
Author contributions
S.B. conceived the study; D.M. and A.M.R. suggested to consider both primal and dual hypergraphs; E.V. and
A.K. performed all data analyses; E.V., A.K., and K. A.-B. made all graphical representations; K. A.-B., D.M.,
A.M.R., M.P., and S.B. discussed and analyzed the results. All authors drew the main conclusions and wrote the
manuscript.
Competing interests
e authors declare no competing interests.
Additional information
Correspondence and requests for materials should be addressed to K.A.-B.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional aliations.
Open Access is article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. e images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.
© e Author(s) 2021
Content courtesy of Springer Nature, terms of use apply. Rights reserved
1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com