Pith. sign in

REVIEW 3 major objections 5 minor 22 references

Solving a fractional parabolic-hyperbolic free boundary problem which models the growth of tumor with drug application using finite difference-spectral method

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper proves that a hybrid finite difference-spectral scheme for the fractional parabolic-hyperbolic tumor-growth model is unconditionally convergent and stable, with error $O((\Delta t)^{2-\alpha/2})$ plus a vanishing spectral term.

desk verdict A substantial but fatally incomplete numerical analysis paper: the convergence proof covers only the zero-data subsystem and the validation example does not satisfy the model. read the letter →

arxiv 1908.07386 v1 pith:X3TVMKOV submitted 2019-08-19 math.NA cs.NA

Sakine Esmaili , F. Nasresfahani , M.R. Eslahchi This is my paper · ORCID
classification math.NAcs.NA MSC 65M7065M1235K2035L03
keywords fractionalparabolic-hyperbolicsystemfreeboundaryproblemtumorgrowthmodeldrugapplicationspectralmethodfinitedifferenceunconditionalconvergencestability
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper takes a radially symmetric free-boundary model of tumor growth in which two diffusion equations describe nutrient and drug concentration and three hyperbolic equations track proliferative, quiescent, and dead cell densities, and replaces the diffusion equations with fractional time derivatives so that subdiffusion is represented. It then constructs an approximate solver that discretizes the fractional time derivative with a finite difference formula, uses spectral (Legendre) polynomials in the radial variable, integrates the cell equations along characteristics, and updates the tumor radius from the cell-velocity relation. The paper's central claim is that this hybrid method is unconditionally convergent and stable: the maximum error in the cell densities and radius, and the weighted $H^1$ error in nutrient and drug concentrations, are bounded by $C e^{MT}((\Delta t)^{2-\alpha/2}+K(N_1))$, where $K(N_1)\to 0$ as the polynomial degree grows. A sympathetic reader would care because a proven error bound is what turns simulations of the fractional model into reliable statements about how nutrient supply and drug application move the tumor boundary.

What carries the argument

The load-bearing object is the finite-difference approximation of the Riemann--Liouville fractional derivative: $$\frac{\partial^\$\alpha$ u}{\partial t^\$\alpha$}(\rho,t_n) \approx u(\rho,0)\frac{$t_n^{{-\alpha}}$}{\Gamma(1-\$\alpha$)}+\sum_{k=0}^{n-1} a_k\big(u(\rho,t_{n+1-k})-u(\rho,t_{n-k})\big),$$ with monotone weights $a_k=\big((k+1)^{1-\alpha}-k^{1-\alpha}\big)/(\Delta t)^\alpha\Gamma(2-\alpha)$, whose truncation error is $O((\Delta t)^{2-\alpha})$. This is combined with orthogonal polynomial trial functions that satisfy the zero Neumann condition at $\rho=0$ and the zero Dirichlet condition at $\rho=1$, the characteristic/midpoint treatment of the three cell equations, and the radius update $R^{n+1}=R^{n-1}\exp(2\Delta t \int_0^1 \rho^2 h(c^n,w^n,p^n,q^n,d^n)\,d\rho)$. The error analysis runs on weighted $L^2$ and $H^1$ norms, spectral interpolation and projection estimates, and the discrete Gronwall lemma.

What would settle it

Implement the proposed scheme on the original model (16)--(26) with nonzero constant boundary supplies $\bar c(t),\bar w(t)$ and nonzero initial profiles, using a manufactured exact solution or a fine-grid reference, and compare the observed temporal rate with the claimed $O((\Delta t)^{2-\alpha/2})$; if the error does not follow the bound or the proof cannot be rerun without the zero-data reduction, the 'without loss of generality' premise is what fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is a convergence-and-stability theorem for a specific numerical recipe applied to the fractional model (16)--(26). Theorem 3.1 states that if the exact nutrient and drug concentrations are smooth enough, then for every time step $\Delta t$ and every spectral degree $N_1$ the global error satisfies $$\max_{0\le k\le n+1} E_k \le C_4^* $e^{{MT}}$\left((\$\Delta$ t)^{2-\$\alpha$/2}+K_4^*(N_1)\right),$$ with an analogous bound for the weighted $H^1$ errors $e_k$, where $K_4^*(N_1)\to 0$ as $N_1\to\infty$ and equals $1/N_1^m$ when the relevant second derivatives are $C^m$. Theorem 4.1 extends the same conclusion to the perturbed problem (69)--(79): if six small forcing or data perturbations of size $\epsilon_1$ are added, the error contains an additional $\epsilon_1$ term. The proof is inductive, closing the coupling between the parabolic unknowns, the hyperbolic cell densities, and the moving radius through a discrete Gronwall argument.

Load-bearing premise

The load-bearing premise is the step in Section 3 that says 'without loss of generality' one may set $c=w=0$ and $c_0=w_0=0$: if that reduction is not actually lossless, because the reaction terms $f(c,p,q)$ and $g(w,p,q)$ depend nonlinearly on the nutrient and drug concentrations, then the convergence and stability theorems apply to the homogeneous version of the problem rather than to the full tumor model with nonzero boundary supplies.

Editorial extensions

If this is right

  • For the homogeneous problem the method needs no CFL restriction: the error bound holds for every time step, so simulations can use large $\Delta t$ when accuracy permits.
  • The temporal order $2-\alpha/2$ lies between $1.5$ and $2$ for $0<\alpha<1$, so the fractional order visibly slows the convergence of the diffusion part, and the numerical tables confirm a rate near this value.
  • With $C^m$ smooth solutions the space error decays as $1/N_1^m$, so spectral accuracy in the radial direction improves rapidly as the polynomial degree grows.
  • Small perturbations in the forcing terms or initial data change the solution by an amount controlled by the same exponential factor plus $\epsilon_1$, giving stability in the sense the theorem states.
  • The same algorithm generates the full time history of the moving boundary radius $R(t)$ along with the cell densities, which is what a user needs to simulate the drug response of a radially symmetric tumor.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the zero-data reduction is not harmless, the practical claim about the drug-supply model is stronger than the theorems establish; a separate proof that keeps $\bar c$, $\bar w$, $c_0$, and $w_0$ nonzero would be needed to cover the original problem.
  • The same L1-spectral template could be carried over to related fractional moving-boundary problems, for example nutrient-only spheroids or therapy models with different kinetic laws, whenever the reaction terms are Lipschitz and the boundary data are smooth.
  • The exponential factor $e^{MT}$ suggests that for long time horizons the guaranteed error bound may become pessimistic; the numerical tests shown use moderate $T$, so the theorem does not by itself certify very-long-time simulation.
  • Because the theorem controls concentrations in a weighted $H^1$ norm and cell densities in the $\infty$-norm, the practical error seen at the moving boundary may be smaller than the worst-case bound, and a boundary-focused error analysis is a natural next step.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes a combined finite-difference/spectral method for a fractional parabolic-hyperbolic free boundary system modeling radially symmetric tumor growth with nutrient and drug concentrations and three cell populations. The model (16)-(26) consists of two reaction-diffusion equations with a Riemann-Liouville fractional derivative applied to the diffusion term, three hyperbolic transport equations for the cell densities, and an ODE for the tumor radius. The time discretization is an L1-type scheme, the spatial discretization for the nutrient and drug is a Legendre spectral method, and the hyperbolic equations are treated along characteristics with midpoint-type finite differences. The main theoretical claim, stated in the abstract and Theorem 3.1, is unconditional convergence and stability with error O((Δt)^{2-α/2}+K(N)) for the cell-density errors, the radius error, and weighted H1 errors of the nutrient and drug. One numerical example is presented to justify the theoretical results.

Significance. If the main theorem were valid, the paper would provide a useful convergence and stability analysis for a coupled fractional moving-boundary problem, and the spectral basis (53) is a sensible construction adapted to the stated boundary conditions. The paper also makes a serious attempt at a full energy-estimate proof, and the stability analysis in Theorem 4.1 is a natural extension of the convergence argument. However, the central claim is not supported as written: the convergence proof is carried out only for a homogeneous version of the problem, and the numerical example used for validation does not satisfy the governing fractional equations or all of the stated initial and boundary conditions. These are load-bearing deficiencies, not presentation issues.

major comments (3)
  1. [Section 3, after Eq. (29)] The reduction 'without loss of generality we can suppose that c = w = 0 and c0(ρ) = w0(ρ) = 0' is not valid for the nonlinear problem (16)-(26). Since f(c,p,q) = K1(c)p + K2(c)q and g(w,p,q) = K3(w)p + K4(w)q depend nonlinearly on c and w, the solution map is not additive: knowing the zero-data solution does not determine the solution with nonzero boundary supplies c̄(t), w̄(t) and nonzero initial profiles c0, w0, and no lifting or fixed-point argument is provided. All subsequent estimates, including Lemma 3.1, Theorem 3.1, and Theorem 4.1, are proved for the homogeneous discrete equations (46)-(47) with zero boundary and initial data. The abstract's claim to solve the original model (16)-(26) is therefore not established; this is confirmed by Example 1, which tests only c(1,t)=w(1,t)=0 and c0=w0=0.
  2. [Section 5, Eqs. (86)-(87), (94)-(95)] The stated exact solutions do not satisfy the fractional diffusion equations. For example, w(t,ρ)=8t(1-ρ²) gives w_t=8(1-ρ²), while the diffusion term in (87) is a Riemann-Liouville fractional derivative of a quantity proportional to t R(t)^{-2}. Since D_t^α t = t^{1-α}/Γ(2-α), this term is not constant in time, so the two sides of the equation cannot agree for all t; the same obstruction occurs for c. Thus Tables 1-3 do not measure the error of the method against a true solution of the fractional PDE.
  3. [Section 5, Eqs. (91)-(95) and Eq. (26)] The numerical example is internally inconsistent. The initial condition in (91) sets d(ρ,0)=4ρ, while the exact d in (95) has d(ρ,0)=((2ρ-1)²+1)+1=(2ρ-1)²+2; these agree only at isolated points such as ρ=1/2 and otherwise differ. The boundary condition in (92) sets v(0,t)=-ε e^{-1}/(t+1)², contradicting the required v(0,t)=0 in (26), and ε is never defined. In addition, p0(ρ)=-((2ρ-1)²+1) is negative, violating the model assumption p0≥0 in (27). These inconsistencies invalidate the numerical verification.
minor comments (5)
  1. [Section 2, Eq. (16)] The phrase 'where ∂α/∂ρ = 0D_t^α' appears to contain a typo; the symbol should be ∂α/∂t, and the definition of the Riemann-Liouville derivative should be stated explicitly and consistently throughout.
  2. [Section 3, Eqs. (29), (33), (45), Lemma 3.1] The temporal truncation-error exponents are stated inconsistently: Eq. (29) gives O((Δt)^{2-α}), Eq. (33) gives O((Δt)^{3-α}), while Lemma 3.1 uses (Δt)^{2+(1-α)/2}; the derivation of the final rate in Theorem 3.1 should reconcile these estimates.
  3. [Section 3, Eqs. (38)-(40) and (48)-(50)] Equation (40) contains an apparent typo: d(ρ,t_{n+1}) = d(..., t_{n-1}), t_{n-1}) + ... has an extra argument, and the same formula is repeated in Eq. (50).
  4. [Section 5, Table 1] Several entries in Table 1 are garbled, for example '0.2435 2e-06' and '1.9342 9e-06', which makes the reported errors unreadable.
  5. [Section 3, Eqs. (46)-(52)] The paper does not specify how the coupled discrete system is solved at each time step, nor whether a nonlinear iteration is required; this should be stated for reproducibility.

Circularity Check

1 steps flagged · score 4.0 of 10

Zero-data reduction makes the advertised convergence theorem apply to a homogeneous surrogate problem.

  1. self definitional [Section 3, after Eq. (29); Eqs. (30)-(31), Lemma 3.1, Theorem 3.1, and final paragraph of Section 3 after Eq. (68)]
    "In the following without loss of generality we can suppose that c =w = 0 and c0(ρ) = w0(ρ) = 0. ... Thus from Theorem 3.1, we can conclude that the sequence {(cap n ,w ap n ,p ap n ,q ap n ,d ap n ,R ap n )}∞ n=0 converges to the exact solution of problem (16)–(26) on [0 , 1] × [0,T ]."

    The convergence analysis is carried out entirely after imposing c = 0, w = 0, c0(ρ) = 0, w0(ρ) = 0, so Lemma 3.1 and Theorem 3.1 estimate the error against the zero-data version of the problem encoded in Eqs. (30)-(31), not against the original problem (16)-(26) with positive boundary supplies c̄(t), w̄(t) and nonzero initial profiles. The phrase 'without loss of generality' is the load-bearing step: no lifting argument, superposition principle, or fixed-point construction is supplied to pass from the zero-data solution to a general solution. This is not available here because f(c,p,q) = K1(c)p + K2(c)q and g(w,p,q) = K3(w)p + K4(w)q depend nonlinearly on c and w, while R(t) and v(ρ,t) also couple through h(c,w,p,q,d).

full rationale

The core energy estimates in Lemma 3.1 and Theorem 3.1 are standard consistency-plus-stability arguments: they bound the error at each step by previous errors, explicit truncation terms, and spectral projection terms, then apply the discrete Gronwall lemma. The proof does not appear to assume the desired convergence bound as an input. The cited references [17], [19], and [21] are standard external texts and articles, not self-citations by the present authors, and no uniqueness theorem is imported from the authors' own prior work. Thus there is no fitted-input-called-prediction circularity and no load-bearing self-citation chain. The central circularity-like defect is the 'without loss of generality' reduction to zero nutrient/drug data: because the reduction is invalid for this nonlinear coupled system, the theorem that is actually proved concerns Eqs. (30)-(31), while the abstract and the final conclusion claim convergence for the original model (16)-(26). The numerical experiment also uses the same zero-data specialization, so the claimed positive-supply case is never exercised. This is a definitional substitution in the derivation chain rather than a merely external correctness gap, which warrants a moderate score rather than a higher one.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the fractional model itself, which is an ad hoc modification of the integer-order model, and on several analytical assumptions (smoothness, zero-data reduction, standard spectral inequalities) that are not derived in the paper. No fitted parameters appear in the method; the constants in the error bounds are existential and not estimated.

assumptions (4)
  • domain assumption The change of variables ρ = r/R(t) maps the moving domain to a fixed domain and yields equations (16)-(26).
    Section 2, Eqs. (15)-(26). The transformation is standard, but the derivation is not shown in detail and the fractional modification is introduced without biological justification.
  • domain assumption The exact solution of the fractional free boundary problem exists and is sufficiently smooth, with ∂²c/∂ρ² being C¹ or C^m.
    Assumed in Lemma 3.1 and Theorem 3.1; no existence or regularity theorem for the fractional model is proved.
  • ad hoc to paper The reduction c=w=0 and c0=w0=0 is without loss of generality.
    Section 3, after Eq. (29). This is not WLOG because the equations are nonlinear and boundary supplies c̄(t), w̄(t) are nonzero.
  • standard math The L1 discretization formula (28) has the stated truncation error O((Δt)^{2-α}).
    Eqs. (28)-(29) citing [21]. The paper later uses O((Δt)^{3-α}) in (33), which is inconsistent with the final convergence rate.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Solving a fractional parabolic-hyperbolic free boundary problem which models the growth of tumor with drug application using finite difference-spectral method." pith.science (2026). https://pith.science/paper/X3TVMKOV

@misc{pith2026190807386,
  author       = {Pith},
  title        = {Pith review of: Solving a fractional parabolic-hyperbolic free boundary problem which models the growth of tumor with drug application using finite difference-spectral method},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/X3TVMKOV}},
  note         = {Machine review of arXiv:1908.07386}
}
read the original abstract

In this paper, a free boundary problem modelling the growth of tumor is considered. The model includes two reaction-diffusion equations modelling the diffusion of nutrient and drug in the tumor and three hyperbolic equations describing the evolution of three types of cells (i.e. proliferative cells, quiescent cells and dead cells) considered in the tumor. Due to the fact that in the real situation, the subdiffusion of nutrient and drug in the tumor can be found, we have changed the reaction-diffusion equations to the fractional ones to consider other conditions and study a more general and reliable model of tumor growth. Since it is important to solve a problem to have a clear vision of the dynamic of tumor growth under the effect of the nutrient and drug, we have solved the fractional free boundary problem. We have solved the fractional parabolic equations employing a combination of spectral and finite difference methods and the hyperbolic equations are solved using characteristic equation and finite difference method. It is proved that the presented method is unconditionally convergent and stable to be sure that we have a correct vision of tumor growth dynamic. Finally, by presenting some numerical examples and showing the results, the theoretical statements are justified.

Figures

Figures reproduced from arXiv: 1908.07386 by the authors.

Figure 1
Figure 1. Error functions e c,M,N n , e w,M,N n , e p,M,N n , e q,M,N n , e d,M,N n , e c,M,N n for N=5 and M=200 and α = 0.1. Error of M=100 M=1000 M=2000 M=3000 M=4000 M=5000 c 2.22198e-03 3.50646e-05 9.31974e-06 4.25763e-06 0.24352e-06 1.57668e-06 w 2.85697e-03 3.04010e-05 7.66240e-06 3.42437e-06 1.93429e-06 1.24213e-06 p 2.89789e-02 2.85568e-04 7.14999e-05 3.18273e-05 1.79269e-05 1.14866e-05 q 1.84616e-03 1.81339e-05 4.54… view at source ↗
Figure 2
Figure 2. Error functions e c,M,N n , e w,M,N n , e p,M,N n , e q,M,N n , e d,M,N n , e c,M,N n for N=10 and M=300 and α = 0.1. 1000 1500 2000 2500 3000 3500 4000 4500 5000 10-6 10-5 10-4 10-3 [PITH_FULL_IMAGE:figures/full_fig_p019_2.png] view at source ↗
Figure 3
Figure 3. Maximum time-errors with N=20 and various M and α = 0.1 [PITH_FULL_IMAGE:figures/full_fig_p019_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: The behaviour of time-errors in LogLog scale with α = 0.1. Error of N=10 N=20 N=40 N=80 N=100 c 1.81574e-05 4.72446e-06 1.17313e-06 2.59655e-07 1.48707e-07 w 849925e-06 2.21145e-06 5.49129e-07 1.21541e-07 6.96080e-08 [PITH_FULL_IMAGE:figures/full_fig_p020_4.png]
Figure 5
Figure 5. Figure 5: Maximum time-error function for M=200 and various N and α = 0.1. also proved the method is unconditionally convergent and stable, which leads us to trust the obtained solution. Finally by giving the numerical results, the theoretical statements are justified. Appendix …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 20 canonical work pages

  1. [1]

    Kim, Regulation of cell proliferation and migration in glioblastoma: new therapeutic approach, Front

    Y. Kim, Regulation of cell proliferation and migration in glioblastoma: new therapeutic approach, Front. Oncol. 3 (2013) 359–371

    work page 2013

  2. [2]

    J. C. L. Alfonso, K. Talkenberger, M. Seifert, B. Klink, A. Hawkins-Daarud, K. R. Swanson, H. Hatzikirou, A. Deutsch, The biology and mathematical modelling of gliom a invasion: a review, J. R. Soc. Interface (2017) DOI: 10.1098/rsif.2017.0490

    arXiv 2017

  3. [3]

    Duman, N.S

    L. Duman, N.S. Gezer, P. Balci, c. Altay, I. Basara, M.G. Du rak, A.I. Sevinc, Differentiation between phyllodes tumors and fibroadenomas based on mammographic sonographic and MRI features, Breast Care 11 (2016) 123– 127

    work page 2016

  4. [4]

    T. L. Jackson, Vascular tumor growth and treatment: Conse quences of polyclonality, competition and dynamic vascular support, J. Math. Biol. 44 (2002) 201–226

    work page 2002

  5. [5]

    Zhao, A parabolic-hyperbolic free boundary problem mo deling tumor growth with drug application, Electron

    J. Zhao, A parabolic-hyperbolic free boundary problem mo deling tumor growth with drug application, Electron. J. Differ. Eq. 2010 (2010) 1–18

    work page 2010

  6. [6]

    T. L. Jackson, H. M. Byrne, A mathematical model to study th e effects of drug resistance and vasculature on the response of solid tumors to chemotherapy, Math. Biosci. 164 (2000), 17–38

    work page 2000

  7. [7]

    Y. Tao, M. Chen, An elliptic-hyperbolic free boundary pro blem modelling cancer therapy, Nonlinearity, 19 (2006), 419–440

    work page 2006

  8. [8]

    Khaitan, S

    D. Khaitan, S. Chandna, M.B. Arya, B.S. Dwarakanath, Esta blishment and characterization of multicellular spheroids from a human glioma cell line: implications for tu mor therapy, J. Transl. Med. 4 (2006) 12–25 30

    work page 2006

Show all 22 references
  1. [9]

    Tao, A free boundary problem modeling the cell cycle and cell movement in multicellular tumor spheroids, J

    Y. Tao, A free boundary problem modeling the cell cycle and cell movement in multicellular tumor spheroids, J. Diff. Eq. 247 (2009) 49–68

    2009

  2. [10]

    C. Geng, H. Paganetti, C. Grassberger, Prediction of tre atment response for combined chemo and radiation therapy for non-small cell lung cancer patients using a bio- mathematical model, Sci Rep. 7 (2017) 13542 doi:10.1038/s41598-017-13646-z

    2017 doi

  3. [11]

    Abernathy, Z

    K. Abernathy, Z. Abernathy, K. Brown, C. Burgess, R. Hoeh ne, Global dynamics of a colorectal cancer treat- ment model with cancer stem cells, Heliyon. 3 (2017)e00247 d oi: 10.1016/j.heliyon.2017.e00247

    2017 doi

  4. [12]

    Enderling, M.A

    H. Enderling, M.A. Chaplain, Mathematical modeling of t umor growth and treatment. Curr. Pharm. Des. 20 (2014) 4934–4940

    2014

  5. [13]

    Ansarizadeh, M

    F. Ansarizadeh, M. Singh, D. Richards, Modelling of tumo r cells regression in response to chemotherapeutic treatment, Appl. Math. Model. 48 (2017) 96-112

    2017

  6. [14]

    Ates and P

    I. Ates and P. A. Zegeling, A homotopy perturbation metho d for fractional order advection-diffusion-reaction boundary-value problems, Appl. Math. Model. 47 (2017) 425– 441

    2017

  7. [15]

    Sohail, S

    A. Sohail, S. Arshad, S. Javed, K. Maqbool, Numerical ana lysis of fractional-order tumor model, Int. J. Biomath. 8 (2015) 1550069

    2015

  8. [16]

    Veeresha, D.G

    P. Veeresha, D.G. Prakasha, H.M. Baskonus, New numerica l surfaces to the mathematical model of cancer chemotherapy effect in Caputo fractional derivatives, Chao s 29 (2019) 013119. doi:10.1063/1.5074099

    2019 doi

  9. [17]

    Quarteroni, A

    A. Quarteroni, A. Valli, Numerical Approximation of Par tial Differential Equations, Springer, Berlin, 1997

    1997

  10. [18]

    Canuto, M.Y

    C. Canuto, M.Y. Hussaini, A. Quarteroni, T. A. Zang, Spec tral Methods Fundamentals in Single Domains, Springer, Berlin, 2006

    2006

  11. [19]

    J. Shen, T. Tang, L. W ang, Spectral Methods, Algorithms, Analysis and Applications, Springer-Verlag Berlin Heidelberg, 2011

    2011

  12. [20]

    Z. W u, J. Yin, C. W ang, Elliptic and Parabolic Equations, W orld Scientific, Singapore, 2006

    2006

  13. [21]

    Lin, C.J

    Y.M. Lin, C.J. Xu, Finite difference/spectral approxima tions for the time-fractional diffusion equation, J. Comput. Phys. 225 (2007) 1533-1552

    2007

  14. [22]

    Huang, Z

    C. Huang, Z. Zhang, Q. Song, Spectral methods for substan tial fractional differential equations, J. Sci. Comput. 74 (2018) 1554–1574

    2018

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.