{"id":"708472ed-d371-4152-8ff4-62b782716e93","arxiv_id":"1908.04763","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For uniformly completely controllable discrete time-varying linear systems, any prescribed union of disjoint closed intervals can be realized as the dichotomy spectrum of the closed-loop system by bounded feedback.","lead":"This paper proves that uniform complete controllability of a discrete time-varying linear control system lets a designer choose the dichotomy spectrum of the closed loop as any finite union of disjoint intervals, using bounded linear feedback. The result extends an earlier pole-placement theorem for Lyapunov spectra to a stronger spectral object that captures forward and backward exponential behavior.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof of Theorem 5 depends on Theorem 10, a two-sided upper-triangularization stated only as a 'slight modification' of [5, Theorem 4.6] with proof omitted; if that extension fails, the central claim is unsupported.","rationale":"The reader's weakest_assumption identifies exactly the same load-bearing concern: Theorem 10 is asserted without proof and the p_i^0=0 entry violates the positivity and Lyapunov conditions. My pass found no additional independent objection; in particular, the block structure of (13) does give dichotomy spectrum [a_i,b_i] once p_i^0 is set to a positive value, because the sequence contains arbitrarily long runs at rate a_i and b_i. The omitted two-sided extension, however, is not a cosmetic gap: the one-sided proof in [5] may rely on forward recursion and controllability from nonnegative initial times, while two-sided kinematic equivalence requires boundedness and invertibility on all of Z. Without this step the central theorem is unproven. The p_i^0=0 issue is almost certainly repairable, and the two-sided upper-triangularization may also be true, so the appropriate disposition remains the reader's CONDITIONAL rather than REJECT or ACCEPT. Hence I set verdict_should_be to UNCHANGED and agreement_with_reader to agree.","tokens_in":8780,"tokens_out":19310,"duration_ms":206807,"concrete_test":"Formally derive Theorem 10 from [5, Theorem 4.6], paying attention to the two-sided extension. A concrete test case: take d=2, A_n = [[0,1],[0,0]], B_n = (1,0)^T, which is uniformly completely controllable, and prescribe constant diagonal entries p^1=2, p^2=3. Run the one-sided construction from [5] on n≥0 and try to continue the same recursion to n<0. Verify whether a bounded feedback U_n and a Lyapunov transformation T_n satisfying (A_n+B_nU_n)T_n = T_{n+1}C_n exist for all n∈Z with ‖T_n‖, ‖T_n^{-1}‖ uniformly bounded. If the backward extension cannot be made, Theorem 10 is doubtful; if it can, the proof is salvageable after fixing p_i^0.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is Theorem 5: uniform complete controllability of (2) implies assignability of the dichotomy spectrum. The only mechanism the proof gives is Theorem 10, which converts the controlled system into a kinematically equivalent upper-triangular system with prescribed positive Lyapunov diagonal sequences. Everything downstream — Proposition 9, the scalar spectra (15), and the final equality Σ_ED(A+BU)=Σ_ED(C)=∪[a_i,b_i] — depends on that equivalence. But Theorem 10 is not proved. Its proof is the single line 'See [5, Theorem 4.6]', and [5, Theorem 4.6] is a one-sided result. Extending it to Z is nontrivial: for every n∈Z one needs a bounded feedback U_n and a Lyapunov transformation T_n satisfying (A_n+B_nU_n)T_n = T_{n+1}C_n. The one-sided construction typically solves this recursively forward from n=0; extending to negative n requires a boundary condition at 0 and bounded invertibility in both time directions. The manuscript supplies no argument that this is possible for all uniformly completely controllable two-sided systems. If the two-sided version is false, the main theorem has no proof. Secondary, and independent of Theorem 10, the sequences (13) assign p_i^0=0, so they are not positive Lyapunov sequences, C_n is not invertible at n=0, and the hypotheses of Theorem 10 and Proposition 9 are violated. This last point is likely a typo, but as written the construction is internally inconsistent.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims that for a discrete time-varying linear control system x_{n+1}=A_n x_n+B_n u_n on Z, uniform complete controllability implies assignability of the dichotomy spectrum: for any prescribed union of ℓ disjoint closed intervals there is a bounded feedback U such that the closed-loop system lies in LLya and its dichotomy spectrum equals that union. The proof strategy is to triangularize the closed-loop system via feedback, then compute the spectrum of the triangular system from its diagonal entries. Section 3.1 proves an inclusion for upper-triangular systems (Theorem 8), Section 3.2 states a two-sided upper-triangularization result (Theorem 10) as a 'slight modification' of [5, Theorem 4.6], and Section 3.3 constructs diagonal sequences p_i^n with prescribed dichotomy spectra and applies these two results.","tokens_in":9136,"tokens_out":11439,"duration_ms":116999,"significance":"If fully proved, the main theorem would substantially strengthen the Lyapunov spectrum assignment result of Babiarz, Czornik, Makarov, Niezabitowski, and Popova, because the dichotomy spectrum is a coarser object that contains the Lyapunov spectrum, and Remark 6 correctly explains the implication. The paper's high-level mechanism is natural: use uniform complete controllability to put the closed-loop system into upper-triangular form with prescribed diagonal sequences, then compute the dichotomy spectrum of the triangular system. The manuscript is clearly organized and the intended reduction is transparent. However, as written the proof rests on three load-bearing gaps, described in the major comments; the central claim is plausible but is not established by the present text.","major_comments":[{"comment":"Theorem 10 is stated for two-sided systems, but its proof is the single sentence 'See [5, Theorem 4.6]'. The cited result is explicitly described in the text as a one-sided theorem. The two-sided extension is not a routine corollary: the one-sided construction iterates forward from 0 and does not address the additional requirement that the Lyapunov transformation T_n be bounded and invertible for all n∈Z, including negative times, nor the consistency condition at n=0 for the feedback and the transformation. Since Theorem 10 is the only mechanism in the proof of Theorem 5 that produces the upper-triangular closed-loop system, the main theorem is not proved as written. The authors need to supply a complete proof of the two-sided version or cite a reference that actually states it.","section":"Section 3.2, Theorem 10"},{"comment":"The sequences p_i^n are defined by p_i^n=e^{a_i}, p_i^n=e^{b_i}, and '0, for n=0'. This makes p_i^0=0, so the sequences are not positive and do not belong to LLya. Theorem 10 requires arbitrary scalar positive Lyapunov sequences, and Proposition 9 requires Lyapunov sequences satisfying the symmetry condition (9). Moreover, the scalar system (14) is not invertible at n=0, so its two-sided evolution operator is not defined in the sense used throughout the paper. Thus the construction in the proof of Theorem 5 violates the hypotheses of the results it invokes. This is likely a typo, but as written it is internally inconsistent and must be fixed, for example by assigning a positive value at n=0 and checking that the dichotomy spectrum remains [a_i,b_i].","section":"Section 3.3, Eq. (13)"},{"comment":"The proof of Theorem 8 identifies the discrete system (6) with the 1-time discretization of the piecewise-constant continuous system dot{x}=W(t)x with W(t)=D_n on [n,n+1). By the definition in the Appendix, the coefficient of the associated 1-time discrete system is A_n=Φ_W(n+1,n), which in this piecewise-constant case equals e^{D_n}, not D_n in general. For a diagonal scalar example, the diagonal entries of the discretized system are e^{a_n} and e^{b_n}, not a_n and b_n. Consequently the equalities in (8) are not justified by the given argument. Theorem 8 may be true, but the proof as written needs to be replaced by a direct discrete-time argument or by a correct continuous embedding that actually produces the ordered evolution over [n,n+1).","section":"Section 3.1, proof of Theorem 8"}],"minor_comments":[{"comment":"In the statement of Theorem 10, the feedback is written as U=(U_n)_{n∈N}; the index set should be Z to be consistent with the surrounding definitions and with the requirement U∈L∞(Z,R^{s×d}).","section":"Section 3.2, Theorem 10"},{"comment":"The sentence 'The structure of the paper is follows' should read 'The structure of the paper is as follows'.","section":"Introduction"},{"comment":"Definition 11 refers to 'the dichotomy spectrum of (1)' when defining the continuous-time spectrum; the equation number should refer to the continuous system (16), not the discrete system (1).","section":"Appendix, Definition 11"}],"recommendation":"major_revision","confidential_remarks":"The most serious obstacle is the unproved two-sided Theorem 10 and the incorrect continuous embedding in the proof of Theorem 8. Both are load-bearing and require substantial additions or replacements in a revision. The p_i^0=0 inconsistency is easy to repair but should not be dismissed, since it currently invalidates the construction as written. I have not independently verified whether the two-sided extension of [5, Theorem 4.6] is true; the burden is on the authors to provide a complete proof. If they cannot, the main theorem remains unsupported."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: this paper has a genuinely interesting claim and a plausible route, but as written the main theorem is not proved. The proof depends on a two-sided upper-triangularization result (Theorem 10) that is stated with no proof, just a reference to a one-sided theorem in [5]; and the scalar sequences used in the construction set the n=0 term to zero, which violates the positivity and Lyapunov assumptions the argument requires. Both issues look fixable, but they are real.\n\nWhat's new: the claim that uniform complete controllability can assign the whole dichotomy spectrum, not just the Lyapunov spectrum, is a natural strengthening of the 2017 pole-placement result. The symmetric dyadic construction for scalar sequences with prescribed interval spectrum is a neat idea, and I think it works once the n=0 slip is corrected. The discrete version of the Battelli-Palmer triangular spectral inclusion (Proposition 9) is also a solid piece of work, and the appendix lemma linking continuous and sampled spectra is clean.\n\nThe soft spots: Theorem 10 is the load-bearing wall. Extending [5, Theorem 4.6] from one-sided to two-sided time is not a 'slight modification' you can hand-wave; you need to construct the feedback and the Lyapunov transformation on all of Z, with a boundary condition at 0 and uniform bounds in both directions. The manuscript doesn't supply that. If the two-sided version is false, the main theorem has no support. This isn't a small gap; it's the central step. The zero at n=0 in (13) is smaller—likely a typo—but as written it breaks the LLya requirement and the positivity condition, and it makes the closed-loop matrix singular at one time. Fix it and that part goes away.\n\nThe citation pattern is fine; the single self-citation is in a supporting discretization lemma, not in the main line. This is honest work, not a disguised triviality.\n\nWho should read it: anyone working on nonautonomous linear control and spectral assignability. It deserves a serious referee—the problem is worth solving and the approach is promising. My recommendation: send it to review, but ask the referee to insist on a full proof of Theorem 10 and the corrected definition of p_i^0, and to check whether Proposition 9's symmetry assumption is really enough for the two-sided dichotomy spectrum.","headline":"A promising paper with a genuine strengthening of Lyapunov spectrum assignability, but the proof currently rests on an unproved two-sided triangularization theorem and a small zero-at-zero slip that breaks the construction's own hypotheses.","tokens_in":9610,"tokens_out":3973,"would_cite":false,"duration_ms":38355,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["34A30","34D09","34D08","34H05"],"pacs":[],"model":"deepseek-v4-flash","headline":"For discrete time-varying linear control systems, uniform complete controllability implies that any prescribed dichotomy spectrum—any finite union of disjoint closed intervals—can be realized by bounded linear feedback.","keywords":["time-varying linear systems","dichotomy spectrum","assignability","uniform complete controllability","bounded linear feedback","Lyapunov spectrum","upper-triangular systems","kinematical equivalence"],"falsifier":"Run the construction in dimension one, where no triangularization step is needed and the closed-loop coefficient is forced to be the sequence $p_n$ itself: with $p_0=0$, the closed-loop matrix at $n=0$ is not invertible, so the system is not in the Lyapunov class on which the definition of dichotomy spectrum rests. Computing the dichotomy spectrum of this scalar system—or replacing $p_0$ by a positive value and checking whether the spectrum remains $[a_i,b_i]$—would directly show whether the paper's claimed assignment is realized as written.","tokens_in":8552,"feed_emoji":"🎛️","tokens_out":13054,"duration_ms":115957,"temperature":0.7,"pith_summary":"For a discrete time-varying linear control system $x_{n+1}=A_nx_n+B_nu_n$, this paper claims that uniform complete controllability is enough to assign, by a bounded linear feedback $u_n=U_nx_n$, any prescribed dichotomy spectrum made of finitely many disjoint closed intervals $[a_1,b_1],\\dots,[a_\\ell,b_\\ell]$. This is a spectrum-level strengthening of the earlier pole-placement theorem that only assigned the Lyapunov spectrum, a finite set of growth rates; here the whole intervals between those rates are under control. If the claim is correct, one controllability hypothesis gives full freedom over the exponential growth and decay rates of all solutions of the closed-loop system, so the controlled dynamics can be shaped qualitatively at will.","feed_headline":"Uniform controllability makes dichotomy spectra fully assignable","feed_subtitle":"A single controllability condition lets bounded feedback place the closed-loop spectrum on any prescribed intervals.","key_machinery":"The machinery has three load-bearing pieces. Kinematical equivalence is the conjugacy relation that lets the paper replace the closed-loop system by an upper-triangular one without changing the dichotomy spectrum: two systems are equivalent when a bounded invertible transformation $T_n$ satisfies $A_nT_n=T_{n+1}C_n$. The two-sided upper-triangularization theorem (Theorem 10), adapted from [5, Theorem 4.6], is what turns uniform complete controllability into the existence of a bounded feedback placing the system in that triangular form with prescribed diagonal sequences. Finally, Proposition 9 computes the spectrum of an upper-triangular system whose diagonal entries are scalar Lyapunov sequences that are even under $n\\mapsto -n$: the dichotomy spectrum is exactly the union of the scalar spectra, and a scalar sequence switching between $e^{a_i}$ and $e^{b_i}$ on alternating dyadic blocks has spectrum $[a_i,b_i]$.","core_discovery":"The paper's central claim is Theorem 5: if system (2) is uniformly completely controllable, then for any disjoint closed intervals $[a_1,b_1],\\dots,[a_\\ell,b_\\ell]$ there exists a bounded feedback $U$ such that $A+BU$ is a Lyapunov sequence and $\\Sigma_{\\mathrm{ED}}(A+BU)=\\bigcup_{i=1}^\\ell[a_i,b_i]$. The proof works by reducing the feedback design to a triangularization problem: Theorem 10 asserts that a two-sided uniformly completely controllable system can be fed back so that it becomes kinematically equivalent to an upper-triangular system whose diagonal entries are prescribed positive scalar Lyapunov sequences, with the proof presented as a slight modification of the one-sided result in [5, Theorem 4.6] and deferred to that reference. Once the closed loop is in triangular form, Proposition 9 computes its dichotomy spectrum as the union of the scalar spectrum of each diagonal sequence, and the diagonal sequences are chosen to alternate between $e^{a_i}$ and $e^{b_i}$ on dyadic blocks, symmetric under $n\\mapsto -n$, so that each scalar spectrum is exactly $[a_i,b_i]$.","pith_inferences":["A testable extension the authors do not state: the same dyadic construction should assign the one-sided dichotomy spectra $\\Sigma^+_{\\mathrm{ED}}$ and $\\Sigma^-_{\\mathrm{ED}}$ on the positive and negative half-lines, since each half of the symmetric sequence (13) independently yields the interval $[a_i,b_i]$.","Because Theorem 10 requires positive Lyapunov diagonal sequences while (13) sets $p_i^0=0$, a minimal repair would be to replace that single zero by a positive value; checking whether the scalar dichotomy spectrum remains $[a_i,b_i]$ would test whether the construction is coherent or needs modification.","Through the 1-time discretization of Lemma 12, the discrete assignment theorem suggests a route to assigning the dichotomy spectrum of sampled continuous-time control systems, a problem the paper leaves open."],"forward_implications":["Any finite union of disjoint closed intervals can be realized as the dichotomy spectrum of a closed-loop system, provided the open-loop system is uniformly completely controllable.","By taking $a_i=b_i=\\lambda_i$, the dichotomy spectrum and the Lyapunov spectrum of the closed-loop system both equal $\\{\\lambda_1,\\dots,\\lambda_\\ell\\}$; the dichotomy-spectrum result therefore recovers the Lyapunov-spectrum pole placement theorem of [5] as a corollary.","The upper-triangular spectral formula (Theorem 8) shows that in the triangular normal form, off-diagonal couplings do not enlarge the dichotomy spectrum beyond the union of the diagonal spectra.","The construction isolates a one-dimensional phenomenon: a scalar coefficient that switches between $e^{a}$ and $e^{b}$ on long dyadic blocks has dichotomy spectrum exactly $[a,b]$, which is the atomic building block for assigning arbitrary intervals."],"supporting_citations":[{"why":"This is the one-sided upper-triangularization theorem that Theorem 10 says it extends to two-sided systems; it also contains the Lyapunov-spectrum assignability result this paper strengthens.","marker":"[5]"},{"why":"Provides the exponential dichotomy criterion for triangular continuous-time systems, which the paper extends to discrete time as Theorem 8.","marker":"[7]"},{"why":"Establishes that the dichotomy spectrum of a discrete nonautonomous system is a union of at most d intervals.","marker":"[13]"},{"why":"Introduces the dichotomy spectrum for linear time-varying systems, the object whose assignability is studied.","marker":"[16]"},{"why":"Proves the Lyapunov spectrum is contained in the dichotomy spectrum, used in Remark 6.","marker":"[9]"},{"why":"Sets out a spectral theory for nonautonomous difference equations used to reconcile the paper's definition with the existing one.","marker":"[3]"}],"fun_headline_variants":["Controllability fully assigns dichotomy spectra","Dichotomy spectrum: fully assignable via controllability","Uniform control yields complete dichotomy spectrum assignment","Controllability lets you place dichotomy spectra anywhere","Dichotomy spectra: arbitrary assignment from one condition"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument depends on the two-sided upper-triangularization theorem (Theorem 10), which is asserted as a 'slight modification' of the one-sided result in [5, Theorem 4.6] with its proof omitted, and on the diagonal sequences in (13) being positive Lyapunov sequences although those sequences set $p_i^0=0$, so if either the unproved extension fails or the zero cannot be repaired, the construction collapses.","fun_headline_variants_meta":{"raw":{"variants":["Controllability fully assigns dichotomy spectra","Dichotomy spectrum: fully assignable via controllability","Uniform control yields complete dichotomy spectrum assignment","Controllability lets you place dichotomy spectra anywhere","Dichotomy spectra: arbitrary assignment from one condition"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000158,"raw_usage":{"total_tokens":1178,"prompt_tokens":854,"completion_tokens":324,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":470,"completion_tokens_details":{"reasoning_tokens":254}},"tokens_in":470,"tokens_out":324,"duration_ms":3668,"temperature":1.0,"reasoning_tokens":254,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:53:12.528258+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the construction in dimension one, where no triangularization step is needed and the closed-loop coefficient is forced to be the sequence $p_n$ itself: with $p_0=0$, the closed-loop matrix at $n=0$ is not invertible, so the system is not in the Lyapunov class on which the definition of dichotomy spectrum rests. Computing the dichotomy spectrum of this scalar system—or replacing $p_0$ by a positive value and checking whether the spectrum remains $[a_i,b_i]$—would directly show whether the paper's claimed assignment is realized as written.","supporting_citations":[{"cited_title":"Babiarz, A","cited_arxiv_id":null,"evidence_quote":"This is the one-sided upper-triangularization theorem that Theorem 10 says it extends to two-sided systems; it also contains the Lyapunov-spectrum assignability result this paper strengthens."},{"cited_title":"Battelli and K.J","cited_arxiv_id":null,"evidence_quote":"Provides the exponential dichotomy criterion for triangular continuous-time systems, which the paper extends to discrete time as Theorem 8."},{"cited_title":"P¨ otzsche.Geometric Theory of Discrete Nonautonomous Dynamical Systems","cited_arxiv_id":null,"evidence_quote":"Establishes that the dichotomy spectrum of a discrete nonautonomous system is a union of at most d intervals."},{"cited_title":"Sacker and G.R","cited_arxiv_id":null,"evidence_quote":"Introduces the dichotomy spectrum for linear time-varying systems, the object whose assignability is studied."},{"cited_title":"Johnson, K.J","cited_arxiv_id":null,"evidence_quote":"Proves the Lyapunov spectrum is contained in the dichotomy spectrum, used in Remark 6."},{"cited_title":"Aulbach and S","cited_arxiv_id":null,"evidence_quote":"Sets out a spectral theory for nonautonomous difference equations used to reconcile the paper's definition with the existing one."}],"review_version":1}