Adjoint-based A Posteriori Error Analysis for Semi-explicit Index-1 and Hessenberg Index-2 Differential-Algebraic Equations
Jehanzeb Chaudhary, Owen Lewis, Md Al Amin Molla · 2025
Read Adjoint-based A Posteriori Error Analysis for Semi-explicit Index-1 and Hessenberg Index-2 Differential-Algebraic Equations by Jehanzeb Chaudhary, Owen Lewis, Md Al Amin Molla (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-03712.txt.
Introduction
Three authors develop better error checks for a class of hard equations in this 2025 math paper. The equations, called differential-algebraic equations, mix rates of change with fixed constraints, and they appear in circuit, chemical, and motion models.
The new method uses an adjoint, a helper equation run in reverse, to estimate the error in a chosen output after a computer solves the main equations in time steps. The team gives formulas for many kinds of outputs and tries two routes to build the adjoint. Tests on nonlinear and larger problems showed the error estimates were close to the true errors, which helps users trust and refine their simulations.
Catalog record
| CREATOR | Jehanzeb Chaudhary, Owen Lewis, Md Al Amin Molla |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 77 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-03712
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Adjoint-based A Posteriori Error Analysis for Semi-explicit Index-1 and Hessenberg Index-2 Differential-Algebraic Equations Jehanzeb Chaudhary, Owen Lewis, Md Al Amin Molla Abstract In this work we develop and analyze adjoint-based analyses for a posteriori error estimation for the temporal discretization of differential-algebraic equations (DAEs) of special type: semi-explicit index-1 and Hessenberg index-2. Our technique quantifies the error in a Quantity of Interest (QoI), which is defined as a bounded linear functional of the solution of a DAE. We derive representations for errors various trypes of QoIs (depending on the entire time interval, final time, algebraic variables, differential variables etc.). We form two analyses: one that defines the adjoint to the DAE system, and one that first converts the DAE to an ODE system and then applies classical a posteriori analysis techniques. A number of examples are presented, including nonlinear and non-autonomous DAEs, as well as discretized partial differential-algebraic equations (PDAEs). Numerical results indicate a high degree of accuracy in the error estimation. † † equal-contributors: These authors contributed equally to this work. † † equal-contributors: These authors contributed equally to this work. † † equal-contributors: These authors contributed equally to this work. 1 Introduction Differential-algebraic equations (DAEs) are a broad class of mathematical models that arise in a wide range of scientific and engineering applications where constraints are intrinsic to the system under consideration. Such systems arise in optimization, model simplification, optimal control, uncertainty quantification, and many other fields to investigate physical phenomena Cao et al. 2003 ; Brenan et al. 1995 ; Hairer and Wanner 2010 . DAEs are a generalization of ordinary differential equations (ODEs). Unlike ODEs, DAEs combine both differential dynamics and algebraic constraints, which introduce unique challenges for their analysis and numerical simulation. In this article we focus on DAEs containing only ordinary derivatives, as opposed to partial derivatives. However, discretization of PDEs with constrains often leads to DAEs Brenan et al. 1995 . Numerical techniques for solving DAEs must handle complexities like stiffness and nonlinearity inherent to ODEs, but also the additional difficulties introduced by the algebraic structure, index, hidden constraints and consistent initial conditions. The index of a DAE, which represents the complexity of its constraint structure, plays an important role in determining the appropriate numerical method. There are a number of techniques for the numerical solution of DAEs (see section 2.3 and the references therein). Solving DAEs numerically invariably involves error in the computed solution. This error in the computed solution must be quantified if DAEs are to be used in a robust and reliable manner in science and engineering applications. There are many cases where scientists and engineers are interested in quantifying the error in a bounded linear functional of the solution to DAEs, the so called quantity of interest (QoI). Adjoint based a posteriori error estimation plays a key role in accurate estimation of the error in the computed value of a QoI, and is the focus of this article. Adjoint based methods and analyses are used in numerous applications, e.g., sensitivity analysis Cao et al. 2003 ; Cheng et al. 2020 ; Serban and Hindmarsh 2005 ; Cao et al. 2002 , optimization Hammond et al. 2021 ; Zhang et al. 2024 ; Su et al. 2023 , optimal control Manzoni et al. 2021 ; Gerdts 2023 , large-scale DAE solver: SUNDIALS Hindmarsh et al. 2005 ; Gardner et al. 2022 , neural network Kharazmi et al. 2021 ; Haber and Ruthotto 2017 ; Tahersima et al. 2019 , uncertainty quantification Mitra and Hlavacek 2019 , and error estimation Eriksson et al. 1995 ; Prudhomme and Oden 1999 ; Becker and Rannacher 2001 ; Giles and Süli 2002 . Although a well-developed theoretical framework for adjoint based a posteriori error estimation is available for ODEs as well as certain classes of partial differential equations (PDEs) Eriksson et al. 1995 ; Cao and Petzold 2004 ; Ainsworth and Oden 1997 ; Bangerth and Rannacher 2003 ; Barth 2004 ; Becker and Rannacher 2001 ; Carey et al. 2009 ; Chaudhry et al. 2015 ; Chaudhry et al. 2021 ; Chaudhry 2018 ; Chaudhry et al. 2017 ; Chaudhry et al. 2015 ; Chaudhry et al. 2016 ; Chaudhry et al. 2013 ; Chaudhry et al. 2019 ; Collins et al. 2015 ; Collins et al. 2014 ; Estep 1995 ; Estep 2009 ; Estep et al. 2012 ; Estep et al. 2002 ; Estep et al. 2000 ; Johansson et al. 2015 ; Chaudhry et al. 2021 , no such analysis has been carried out for DAEs to the best of our knowledge. However, DAEs have been analyzed in other contexts of error estimation and adjoint analysis. For example, error estimation for collocation solutions to linear index-1 DAEs has been studied in Auzinger et al. 2007 ; Auzinger et al. 2009 , while local error control has been investigated in Sieber 2005 . Various a priori error analyses have been carried out in Hindmarsh and Petzold 1995 ; Hairer et al. 1988 ; Boscarino 2007 ; Linh et al. 2011 ; Arnold et al. 1995 ; Leimkuhler 1986 ; Karasözen and Somali 1996 . The adjoint system for index-1 DAEs has been analyzed in Balla and März 2000 . It was shown that the original DAE and its associated adjoint system have the same index März 2000 . Adjoint based sensitivity analysis for DAEs has been carried out in Cao et al. 2003 . In this article we carry out a thorough adjoint based analysis for semi-explicit index-1 and Hessenberg index-2 DAEs. In particular, this involves defining the associated adjoint DAEs with consistent initialization depending on the structure of the QoI, and using the adjoint solutions to quantify the error in the computed QoI. The theoretical results in this article derive accurate error estimates for various QoIs ( e.g. depending on final time, average over time intervals, involving either differential or algebraic variables, or both). We develop two error analyses in this article. The first approach is based on defining the adjoint directly to the DAE system, and requires careful setting of the adjoint initial conditions. The second approach defines an adjoint problem to the ODE corresponding to the DAE, which in turn allows the use of classical error analyses techniques. The accuracy of the resulting error estimates is demonstrated through a number of numerical examples including nonlinear, non-autonomous DAEs as well as a semi-discretized PDE. The remainder of this paper is organized as follows: In section 2 we discuss the background of DAEs with two special types of DAEs; semi-explicit index-1 and Hessenberg index-2 DAEs. Further we discuss their numerical treatment and also introduce quantities of interest in this section. In sections 3 and 4 we develop two analyses to quantify the error in the QoI for DAEs of type semi-explicit index-1 and Hessenberg index-2. The first technique, called Ajoint DAE, is based on forming an adjoint to the DAE system and is presented in section 3 . The second technique, called Adjoint ODE, first forms an ODE system corresponding to the DAE and then utilizes the adjoint to the ODE system. This second technique is discussed in section 4 . In section 5 we discuss the implementation aspects and