Generating monomials in dimensions three and four (Q1295766)

The paper provides an alternative proof and an extension of results of \textit{F. J. Curtis} [J. Pure Appl. Algebra 104, No. 2, 161-167 (1995; Zbl 0854.13011)]. For \(n,d \in \mathbb Z_{>0}\) let \[ S_n(d)=\{(l_1, \dots, l_n) \in \mathbb Z_{\geq 0}^n: \sum l_i=d \}. \] Let \(e_i=(0, \dots,0, 1,0, \dots,0)\) be the \(i\)-th standard vector; \(i=1, \ldots, n\). Call a family \(C\subset S_n(d-1)\) a cover of \(S_n(d)\) if \(S_n(d)= \bigcup_{e\in C} e+\{e_1, \dots, e_n\}.\) Define \(\rho_n(d)=\) minimal cardinality of a cover of \(S_n(d).\) Corollary 2.2 (Curtis): For \(d\geq 10,\) \(\rho_3 (d)=\lfloor {1\over 3} {{d+4} \choose 2} \rfloor-3\). Corollary 3.2: For \(d\geq 25,\) \(\rho_4 (d)\geq (d^3+10d^2-865 d+230)/24.\) Theorem 4.1 and 4.2: For odd \(d\geq 5,\) \(\rho_4(d) \leq (d^3+15d^2-61d +261)/24;\) for even \(d\geq 6,\) \(\rho_4(d) \leq (d^3+15d^2-34d +240)/24.\) The lower bounds of corollary 2.2 and corollary 3.2 are found by reducing the enumeration problems in a subtle and interesting way to relatively large linear programming problems. The upper bounds are easier to establish; indeed the authors give explicit covers and short proofs not relying on the computer. The bounds improve for \(n=3, 4\) the general ones established by \textit{A. V. Geramita}, \textit{D. Gregory} and \textit{L. Roberts} [J. Pure Appl. Algebra 40, 33-62 (1986; Zbl 0586.13015) and Rend. Semin. Mat., Torino 42, No. 2, 1-10 (1984; Zbl 0584.14001)] in connection with questions from algebraic geometry where their impact stems from the bijection \((l_1, \ldots, l_n) \leftrightarrow x_1^{l_1} \cdots x_n^{l_n}.\)