🔬 physics

Academic collaborations and movements towards successful careers in physics

By analyzing 35,708 physicists over 15 years, this study reveals that career success and mobility are maximized by evolving from loosely connected networks into moderately clustered structures with expanding size, a pattern that converges toward an optimal balance of cohesion and diversity by mid-career.

Original authors: Mingrong She, Jan Bachmann, Fariba Karimi, Leto Peel

Published 2026-02-26

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Mingrong She, Jan Bachmann, Fariba Karimi, Leto Peel

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a physicist's career not as a solitary climb up a mountain, but as navigating a vast, shifting ocean. To survive and thrive, you need two things: a boat (your collaboration network) and the courage to sail to new waters (mobility).

This paper, written by researchers from Maastricht University and others, looks at the careers of nearly 36,000 physicists over 15 years to figure out which navigation strategies lead to the biggest, most successful voyages.

Here is the story of their findings, told in simple terms.

The Two Key Ingredients of a Career

The researchers focused on two main things about a scientist's "boat" (their network of co-authors):

Network Size: How many different people are on your boat? (Are you sailing alone, with a small crew, or a massive fleet?)
Clustering: How well do the people on your boat know each other?
- High Clustering: Everyone on the boat is best friends who have sailed together for years. They trust each other deeply, but they might all know the same things.
- Low Clustering: You have many different people on board, but they don't know each other. You are the bridge connecting different groups.

The Winning Strategy: "The Expanding Circle"

The study found that the most successful scientists didn't just stick to one strategy. They followed a specific, dynamic rhythm:

Early Career (The Loose Net): Successful scientists started with loose networks. They didn't have a tight-knit, closed circle of friends yet. Instead, they cast a wide net, meeting many different people from different places.
Mid-Career (The Tightening & Growing): As they moved into their prime years (roughly years 5 to 15), they did two things at once:
- They expanded their network (got more people on the boat).
- They tightened their relationships (made sure the people on the boat started working together and trusting each other).

The Analogy: Think of it like building a sports team.

The Losers were either the "Lone Wolves" (too small a team) or the "Closed Clubs" (a tiny group of friends who never let anyone new in).
The Winners started by recruiting players from all over the world (loose network). Then, once they had a big roster, they started running drills together to build chemistry (tightening), while still recruiting new talent (growing).

The "Sweet Spot" of Friendship

One of the most interesting findings is about the "Goldilocks" zone of friendship.
By the time these scientists reached year 15, the most successful ones all converged on a moderate level of closeness.

If your team is too tight (everyone knows everyone), you get stuck in an echo chamber. You stop hearing new ideas.
If your team is too loose (everyone is a stranger), you lack trust and can't get deep work done.
The Winner: A team where you have a core group of trusted collaborators, but you also keep your doors open to new, diverse connections.

The Role of Travel (Mobility)

The paper also looked at mobility—moving to different countries or universities.

The Connection: The scientists who traveled the most were the ones who built the best networks.
Why? Moving to a new place forces you to meet new people (expanding the network). It breaks up old, tight circles and forces you to build new bridges.
The Surprise: Even if you account for the network changes, simply moving still helped. It's like having a "traveler's luck" factor. Being mobile seems to signal that you are adaptable and open to new opportunities, which leads to more publications and citations.

The Takeaway for Everyone

Whether you are a scientist, a business owner, or just trying to build a career, the lesson is the same:

Don't stay in your comfort zone too early. In the beginning, cast a wide net. Meet people from different backgrounds. Don't just stick with your college friends.
Build depth as you grow. Once you have a wide network, start deepening those relationships. Turn acquaintances into a reliable team.
Keep moving. Changing your environment (physically or professionally) is a powerful way to refresh your network and find new opportunities.
Find the balance. The best teams aren't closed clubs, and they aren't random crowds. They are diverse groups that have learned to work together.

In short: Start wide, grow deep, and keep moving. That is the secret recipe for a successful career in science (and likely in many other fields too).

1. Problem Statement

While academic collaboration networks and international mobility are widely recognized as critical drivers of scientific success, existing literature lacks a systematic examination of how these factors co-evolve over time. Most studies analyze network structures at single time points or examine mobility without linking it to long-term career trajectories. Key unanswered questions include:

Do distinct patterns of network evolution exist across researchers?
How do these evolutionary trajectories (specifically regarding network size and clustering) relate to long-term success (e.g., becoming a Principal Investigator, publication output, citation impact)?
How does mobility interact with these network patterns at different career stages?

2. Methodology

Data Source and Selection:

Dataset: The study utilizes the American Physical Society (APS) dataset, covering 678,916 publications from 19 Physical Review journals (1893–2020).
Sample: After applying strict inclusion criteria, the final dataset comprises 35,708 physicists.
- Criteria: Career length $\ge$ 15 years; exclusion of authors who became Principal Investigators (PIs) within 5 years (to avoid rapid outliers); minimum of 2 co-authors (to calculate clustering); minimum of 3 publications.
PI Definition: An author is defined as a PI if they have at least three last-author publications (excluding alphabetic ordering).

Variables and Metrics:

Network Evolution (Years 5–15): Two key structural metrics were tracked annually:
1. Network Size: Cumulative count of unique collaborators.
2. Clustering Coefficient: The degree to which collaborators are interconnected.
Productivity Control: To isolate structural evolution from mere productivity growth, the authors regressed network size and clustering coefficients against the number of publications for each career year. Standardized residuals were used as the adjusted metrics for analysis.
Mobility: Measured using an "international mobility score" based on country transitions in affiliations (early-career: years 1–5; mid-career: years 6–15).
Scientific Outcomes: Four metrics were evaluated:
1. Likelihood of becoming a PI.
2. Time to PI attainment.
3. Total publication output.
4. Citation impact (normalized citations per paper).

Analytical Approach:

Recursive Time Series Clustering: The authors applied Time Series K-Means to the 11-year trajectories (years 5–15) of adjusted network size and clustering.
Recursive Splitting: A hierarchical splitting strategy was used. The sample was split into two clusters, and a Monte Carlo permutation test ( $p < 0.001$ ) determined if the split resulted in significant differences in scientific outcomes. This process was repeated recursively until no further significant splits could be made, resulting in 10 distinct network evolution clusters.

3. Key Contributions

Identification of 10 Distinct Evolutionary Patterns: The study moves beyond static network analysis to categorize careers into 10 unique trajectories based on how network size and clustering change relative to productivity.
Convergence Hypothesis: The paper demonstrates that despite diverse starting points (tight vs. loose networks), the most successful career paths converge toward a state of moderate clustering by career year 15.
Decoupling Mobility and Network Effects: The study isolates the effect of mobility, showing that while mobility correlates with specific network patterns, it retains a positive association with scientific outcomes even after controlling for network evolution.
Career-Stage Specificity: It reveals that the optimal network strategy differs by career stage (e.g., early mobility aids network expansion, while mid-career tightening aids stability).

4. Key Results

A. Optimal Network Evolution Patterns

The "Growing and Tightening" Pattern: The most successful cluster (Cluster 111) is characterized by authors who start with loose networks (low clustering) in early career and progressively tighten their networks (increasing clustering) while simultaneously expanding their network size during mid-career.
Performance: This pattern yields the highest rates of PI attainment, total publications, and citation impact.
Convergence: By career year 15, all high-performing clusters (regardless of initial tightness) converge toward a moderate clustering coefficient. This suggests an optimal balance between core team cohesion and diverse external connections.
Sub-optimal Patterns:
- Shrinking/Loosening: Associated with faster PI attainment but lower citation impact.
- Shrinking/Tightening: High clustering but limited network growth leads to lower overall output compared to growing patterns.

B. Relationship with Mobility

Early Career: Authors with initially tight networks exhibit lower early-career mobility. Conversely, those with loose initial networks show higher early-career mobility, which facilitates network expansion.
Mid-Career: "Growing" patterns (increasing network size) generally correlate with higher mid-career mobility.
Independence: Regression analyses within clusters confirm that mobility remains positively associated with scientific outcomes (PI status, citations) even when network evolution patterns are controlled for.

C. Gender Analysis

Statistical tests (Chi-squared and Cramér's V) revealed that while gender distribution varied slightly across clusters, the effect size was negligible ( $V = 0.037$ ), indicating that the identified network patterns apply relatively equally across genders in this dataset.

5. Significance and Implications

Theoretical Insight: The findings challenge the notion that "more connections" or "tighter teams" are universally better. Instead, they support a dynamic balance theory: successful careers require an initial phase of broad exploration (loose networks, high mobility) followed by a consolidation phase (tightening relationships) while maintaining growth.
Policy and Funding:
- Mobility Programs: Supports the efficacy of programs like Marie Skłodowska-Curie Actions, as mobility is linked to the "loose-to-tight" successful trajectory.
- Mentorship: Suggests mentors should encourage broad networking early in careers but assist in building stable core teams during mid-career.
- Funding Structures: Highlights the need for long-term funding to allow researchers to navigate these complex evolutionary trajectories without being forced into premature specialization.
Limitations & Future Work: The study is limited to Physics (APS data) and relies on publication data, potentially missing informal mentorship or grant-based networks. Future research should explore causality and cross-disciplinary generalizability.

Conclusion:
The paper provides the first systematic evidence that network evolution dynamics, rather than static network properties, are predictive of long-term academic success. The "sweet spot" for a physicist's career involves starting with a diverse, loose network, leveraging mobility to expand connections, and gradually tightening these relationships into a cohesive, moderately dense core by mid-career.