Artificial Intelligence is increasingly pervasive across domains, with ever more complex models delivering impressive predictive performance. This fast technological advancement however comes at a concerning environmental cost, with state-of-the-art models - particularly deep neural networks and large language models - requiring substantial computational resources and energy. In this work, we present the intuition of Green AI dynamic model selection, an approach based on dynamic model selection that aims at reducing the environmental footprint of AI by selecting the most sustainable model while minimizing potential accuracy loss. Specifically, our approach takes into account the inference task, the environmental sustainability of available models, and accuracy requirements to dynamically choose the most suitable model. Our approach presents two different methods, namely Green AI dynamic model cascading and Green AI dynamic model routing. We demonstrate the effectiveness of our approach via a proof of concept empirical example based on a real-world dataset. Our results show that Green AI dynamic model selection can achieve substantial energy savings (up to ~25%) while substantially
Africa's green industrialization imperative demands reliable infrastructure for monitoring air quality. We present a satellite-reanalysis PM2.5 fusion system trained on 2,068,901 records from 404 monitoring locations in 29 African countries (OpenAQ, 2017-2022), combining LightGBM with leakage-resistant spatial cross-validation and conformal prediction to quantify predictions and their geographic applicability limits. Under 5-fold location-grouped spatial cross-validation, LightGBM achieves RMSE = 30.83 +/- 5.07 ug/m3, MAE = 14.54 +/- 1.66 ug/m3, R2 = 0.134 +/- 0.023, and macro F1 = 0.336 +/- 0.018. This R2 is substantially below random-split benchmarks (>0.90) but reflects true geographic generalisation difficulty rather than model failure. Split conformal prediction targeting 90% marginal coverage reveals severe East Africa degradation (actual PICP = 65.3% vs. nominal 90%), consistent with medium-strength covariate shift (humidity KS = 0.2237, sat_pblh KS = 0.2558). We operationalise these findings through regional reliability flags (High/Medium/Low/Unreliable) and a monitor prioritisation score directing infrastructure expansion toward highest-burden unmonitored populations, d
The green valley represents the population of galaxies that are transitioning from the actively star-forming blue cloud to the passively evolving red sequence. Studying the properties of the green valley galaxies is crucial for our understanding of the exact mechanisms and processes that drive this transition. The green valley does not have a universally accepted definition. The boundaries of the green valley are often determined by empirical lines that are subjective and vary across studies. We present an unambiguous definition of the green valley in the colour-stellar mass plane using the entropic thresholding. We first divide the galaxy population into the blue cloud and the red sequence based on a colour threshold that minimizes the intra-class variance and maximizes the inter-class variance. Our method splits the region between the mean colours of the blue cloud and the red sequence into three parts by maximizing the total entropy of that region. We repeat our analysis in a number of independent stellar mass bins to define the boundaries of the green valley in the colour-mass diagram. Our method provides a robust and natural definition of the green valley.
The immense technological progress in artificial intelligence research and applications is increasingly drawing attention to the environmental sustainability of such systems, a field that has been termed Green AI. With this contribution we aim to broaden the discourse on Green AI by investigating the current status of approaches to both environmental assessment and ecodesign of AI systems. We propose a life-cycle-based system thinking approach that accounts for the four key elements of these software-hardware-systems: model, data, server, and cloud. We conduct an exemplary estimation of the carbon footprint of relevant compute hardware and highlight the need to further investigate methods for Green AI and ways to facilitate wide-spread adoption of its principles. We envision that AI could be leveraged to mitigate its own environmental challenges, which we denote as AI4greenAI.
For any point $\mathrm{X}$ in the cluster complex $\mathrm{Cpx}(\mathcal{C})$ of a 2-Calabi-Yau category $\mathcal{C}$, we introduce $\mathrm{X}$-evolution flow on $\mathrm{Cpx}(\mathcal{C})$. We show that such a flow induces a piecewise linear one-dimensional $\mathrm{X}$-foliation with two singularities, the unique sink $\mathrm{X}$ and the unique source $\mathrm{X}[1]$. Moreover, we show that evolution flows on cluster complexes are continuous refinement/generalization of green mutations on cluster exchange graphs. For the cluster category of a Dynkin or Euclidean quiver $Q$, we prove that the $\mathrm{X}$-foliation is compact or semi-compact, for various choices of $\mathrm{X}$. As an application, we show that $\mathrm{Cpx}(\mathcal{C})$ is spherical (Dynkin case) or contractible (Euclidean case). As a byproduct, we show that the fundamental group of the cluster exchange graph of $Q$ is generated by squares and pentagons.
This is a sequel to arXiv:2401.02087. We prove the Green function rigidity conjecture in arXiv:2401.02087 for conformal Laplacian in dimension $n\geq 3$. For the Paneitz operator, we prove the Green function rigidity conjecture when $n eq 4k+2, k\geq 2$. Important ingredients in our proof are the positive mass theorem and the positive energy theorem for Paneitz operator. As a byproduct, we also obtain a new formula for the ADM mass of an asymptotically flat hypersurface that allows for a non-entire graph.
Progressing digitalization and increasing demand and use of software cause rises in energy- and resource consumption from information and communication technologies (ICT). This raises the issue of sustainability in ICT, which increasingly includes the sustainability of the software products themselves and the art of creating sustainable software. To this end, we conducted an analysis to gather and present existing literature on three research questions relating to the production of ecologically sustainable software ("Green Coding") and to provide orientation for stakeholders approaching the subject. We compile the approaches to Green Coding and Green Software Engineering (GSE) that have been published since 2010. Furthermore, we considered ways to integrate the findings into existing industrial processes and higher education curricula to influence future development in an environmentally friendly way.
AI is demanding an evergrowing portion of environmental resources. Despite their potential impact on AI environmental sustainability, the role that programming languages play in AI (in)efficiency is to date still unknown. With this study, we aim to understand the impact that programming languages can have on AI environmental sustainability. To achieve our goal, we conduct a controlled empirical experiment by considering five programming languages (C++, Java, Python, MATLAB, and R), seven AI algorithms (KNN, SVC, AdaBoost, decision tree, logistic regression, naive bayses, and random forest), three popular datasets, and the training and inference phases. The collected results show that programming languages have a considerable impact on AI environmental sustainability. Compiled and semi-compiled languages (C++, Java) consistently consume less than interpreted languages (Python, MATLAB, R), which require up to 54x more energy. Some languages are cumulatively more efficient in training, while others in inference. Which programming language consumes the most highly depends on the algorithm considered. Ultimately, algorithm implementation might be the most determining factor in Green AI,
With the wireless internet access being increasingly popular with services such as HD video streaming and so on, the demand for high data consuming applications is also rising. This increment in demand is coupled with a proportional rise in the power consumption. It is required that the internet traffic is offloaded to technologies that serve the users and contribute in energy consumption. There is a need to decrease the carbon footprint in the atmosphere and also make the network safe and reliable. In this article we propose a hybrid system of RF (Radio Frequency) and VLC (Visible Light Communication) for indoor communication that can provide communication along with illumination with least power consumption. The hybrid network is viable as it utilizes power with respect to the user demand and maintains the required Quality of ServiceQoS and Quality of Experience QoE for a particular application in use. This scheme aims for Green Communication and reduction in Electromagnetic EM Radiation. A comparative analysis for RF communication, Hybrid RF+ VLC and pure VLC is made and simulations are carried out using Python, Scilab and MathWorks tool. The proposal achieves high energy effici
With the ever-growing adoption of AI, its impact on the environment is no longer negligible. Despite the potential that continual learning could have towards Green AI, its environmental sustainability remains relatively uncharted. In this work we aim to gain a systematic understanding of the energy efficiency of continual learning algorithms. To that end, we conducted an extensive set of empirical experiments comparing the energy consumption of recent representation-, prompt-, and exemplar-based continual learning algorithms and two standard baseline (fine tuning and joint training) when used to continually adapt a pre-trained ViT-B/16 foundation model. We performed our experiments on three standard datasets: CIFAR-100, ImageNet-R, and DomainNet. Additionally, we propose a novel metric, the Energy NetScore, which we use measure the algorithm efficiency in terms of energy-accuracy trade-off. Through numerous evaluations varying the number and size of the incremental learning steps, our experiments demonstrate that different types of continual learning algorithms have very different impacts on energy consumption during both training and inference. Although often overlooked in the con
The retarded Green function of a wave equation on a 4-dimensional curved background spacetime is a (generalized) function of two spacetime points and diverges when these are connected by a null geodesic. The Hadamard form makes explicit the form of this divergence but only when one of the points is in a normal neighbourhood of the other point. In this paper we derive a representation for the retarded Green function for a scalar field in Schwarzschild spacetime which makes explicit its {\it complete} singularity structure beyond the normal neighbourhood. We interpret this representation as a sum of Hadamard forms, the summation being taken over the number of times the null wavefront has passed through a caustic point: the sum of Hadamard forms applies to the non-smooth contribution to the full Green function, not only the singular contribution. (The term non-smooth applies modulo the causality-generating step functions that must appear in the retarded Green function.) The singularity structure is determined using two independent approaches, one based on a Bessel function expansion of the Green function, and another that exploits a link between the Green functions of Schwarzschild sp
We present a characteristic initial value approach to calculating the Green function of the Regge-Wheeler and Zerilli equations. We combine well-known numerical methods with newly derived initial data to obtain a scheme which can in principle be generalised to any desired order of convergence. We demonstrate the approach with implementations up to sixth-order in the grid spacing. By combining the results of our numerical code with late-time tail expansions and methods of subtracting the direct part of the Green function, we show that the scalar self-force in Schwarzschild spacetime can be computed to better accuracy than previous Green-function based approaches. We also demonstrate agreement with frequency-domain methods for computing the Green function in the gravitational case. Finally, we apply the Regge-Wheeler and Zerilli Green functions to the computation of the gravitational energy flux.
There is incremental growth in adopting self-reconfigurable robots in automating manufacturing conventional product lines. Using this class of robots adapting themselves with ever-changing environmental conditions has been acclaimed as a promising way of reducing energy consumption and environmental impact and thus enabling green manufacturing. Whilst the majority of existing research focuses on highlighting the efficacy of self-reconfigurable robots in energy reduction with technical driven solutions, the research on exploring the salient factors in design and development self-reconfigurable robots that directly enable or hinder green manufacturing is non-extant. This interdisciplinary research contributes to the nascent body of the knowledge by empirical investigation of design-time, run-time, and hardware aspects which should be contingently balanced when developing green-aware self-reconfigurable robots. Keywords Green manufacturing, self-reconfigurable robots, robot design, green awareness
We study existence and uniqueness of Green functions for the Cheeger $Q$-Laplacian in metric measure spaces that are Ahlfors $Q$-regular and support a $Q$-Poincaré inequality with $Q>1$. We prove uniqueness of Green functions both in the case of relatively compact domains, and in the global (unbounded) case. We also prove existence of global Green functions in unbounded spaces, complementing the existing results in relatively compact domains proved recently in [BBL20].
At present, analytical lab-on-chip devices find their usage in different facets of chemical analysis, biological analysis, point of care analysis, biosensors, etc. In addition, graphene has already established itself as an essential component of advanced lab-on-chip devices. Graphene-based lab-on-chip devices have achieved appreciable admiration because of their peerless performance in comparison to others. However, to accomplish a sustainable future a device must undergo Green-Screening to check its environmental compatibility. Thus, extensive research is carried out globally to make the graphene-based lab-on-chip green, though it is yet to be achieved. Nevertheless, as a ray of hope, there are few existing strategies that can be stitched together for feasible fabrication of environment-friendly green graphene-based analytical lab-on-chip, and those prospective pathways are reviewed in this paper.
Reduction of unnecessary energy consumption is becoming a major concern in wired networking, because of the potential economical benefits and of its expected environmental impact. These issues, usually referred to as "green networking", relate to embedding energy-awareness in the design, in the devices and in the protocols of networks. In this work, we first formulate a more precise definition of the "green" attribute. We furthermore identify a few paradigms that are the key enablers of energy-aware networking research. We then overview the current state of the art and provide a taxonomy of the relevant work, with a special focus on wired networking. At a high level, we identify four branches of green networking research that stem from different observations on the root causes of energy waste, namely (i) Adaptive Link Rate, (ii) Interface proxying, (iii) Energy-aware infrastructures and (iv) Energy-aware applications. In this work, we do not only explore specific proposals pertaining to each of the above branches, but also offer a perspective for research.
A diagrammatic Monte Carlo evaluation of the ladder series contributions to the correlation potential (self energy) of a positron in the field of a molecule is presented. The $GW$@TDHF, virtual-positronium ($T$-matrix), and positron-hole Goldstone ladder series contributions are stochastically sampled order-by-order within the Tamm-Dancoff approximation, which is exact for the latter two classes, with Ces{á}ro-Riesz resummation used to extrapolate to infinite order. Gaussian bases are employed and Coulomb matrix elements are represented via density fitting, with the three centre integrals the largest arrays required to be stored in memory. The stochastic approach thus realizes a reduction in memory of the largest arrays required on the order of the number of molecular orbitals in the basis $N\sim$10$^2$--10$^3$ compared to the exact deterministic solution of Bethe-Salpeter equations [J. Hofierka, B. Cunningham, C. M. Rawlins, C. H. Patterson and D. G. Green, Nature {\bf 606}, {688} (2022)]. Benchmark results for lithium hydride show quantitative agreement with exact diagonalisation, notably demonstrating the successful stochastic summation of the virtual-positronium infinite electr
Massive scalar fields on black hole backgrounds generally admit two families of modes: quasi-bound states (QBS) and quasinormal modes (QNM). We demonstrate the orthogonality between the two mode families with respect to a relativistic product. We also find that, although the two families appear on different Riemann sheets of the Green's function of massive scalar perturbations, they can be brought to a single sheet with an appropriate redefinition of the frequency variable. In this variable, it is more natural to see how both mode families can be excited by initial data, and to approximate the Green's function with saddle points. Finally, we investigate the QNM emission from boson clouds - the latter effectively consisting of a single QBS - driven by the tidal perturbation of a second compact object. We show that while the resonant emission of QNMs is generally suppressed, QNM transitions may be more prominent when the interaction with the perturber is non-resonant, such as in the dynamical capture of unbound objects, and when the perturber transits close to the light ring.
Bipolar nanoporous membranes and bipolar nanochannels are used in water desalination and energy-harvesting systems that provide clean water and green energy, respectively. The growing need for both requires continuous improvement of their performance. However, the underlying physics of these complex systems is still not fully understood, making empirical optimization slow and inefficient. In this work, we combine theoretical analysis and numerical simulations to develop a unified framework for improving the design of nanofluidic devices. We show that the system response is governed by the interplay between the applied voltage and a parameter $η$, which depends on the ratio of geometry and surface charge densities of both charged regions. At low voltages, the response is mostly determined by $η$, allowing its dependence to be represented by a simplified phase space. At high voltages, this phase space becomes oversimplified. To demonstrate the framework's robustness, we scan a range of configurations, from unipolar channels (single charged region) to bipolar channels (positive and negative segments). We compare the numerically simulated current-voltage responses with three theoretica
Doppler-broadened $γ$-ray spectra for positron annihilation on molecules are calculated using many-body theory. By employing Gaussian bases for the electron and positron wavefunctions, a computable expression that involves a four-centre integral over the two-annihilation-photon momenta is derived for the $γ$ spectra in the independent particle model approximation to the annihilation vertex, and implemented in the open-source {\tt EXCITON+} code. The influence of electron-positron correlations on the $γ$ spectra is examined through \textit{ab initio} treatment of the positron wavefunction, whilst corrections to the annihilation vertex are treated approximately via enhancement factors previously calculated [D. G. Green and G. F. Gribakin, Phys.~Rev.~Lett.~{\bf 114}, 093201 (2015)] exactly for atoms. Calculated $γ$ spectra for furan and acetonitrile are presented for annihilation from the positron bound state with electrons of individual molecular orbitals. For such annihilation from the positron-molecule bound state, it is found that the magnitude of the partial contribution to the $γ$ spectra from individual molecular orbitals depends not just on the orbital energies, but also on th