Maclaurin's Inequality¶
The descending chain of root-normalized elementary symmetric means of a finite nonnegative real vector.
Core Idea¶
For nonnegative real numbers \(a_1,\ldots,a_n\), let \(e_k\) be the sum of every product of \(k\) distinct entries, and let \(s_k=e_k/\binom nk\). Maclaurin's inequality says that the positive-degree root means descend:
The division by \(\binom nk\) averages over all \(k\)-element subsets; taking the \(k\)-th root then puts products of different degrees on the original input scale. Both normalizations are essential to the stated comparison. The ends are the arithmetic and geometric means, so their comparison is a consequence of the full chain rather than its definition.[1]
Newton's inequality \(s_{k-1}s_{k+1}\leq s_k^2\) is one route to the chain when inputs are nonnegative. The two statements are related but not identical: Newton's inequality is an adjacent log-concavity relation among the \(s_k\); Maclaurin compares their degree roots across the entire index range.[1]
Structural Signature¶
Sig role-phrases: nonnegative finite vector → all subset-products → binomially averaged sums → degree roots → nonincreasing chain → qualified equality boundary.
- Admissible vector: the operands are a finite list of nonnegative real numbers. Signed roots do not inherit this exact classical chain merely because their polynomial is real-rooted.[1]
- Elementary symmetric family: \(e_k=\sum_{i_1<\cdots<i_k}a_{i_1}\cdots a_{i_k}\) incorporates every distinct \(k\)-subset, not one chosen product or a power sum.[1]
- Combinatorial average: \(s_k=e_k/\binom nk\) removes the changing number of terms. The unnormalized \(e_k\) are not the means in this theorem.[1]
- Degree-root comparison: \(s_k^{1/k}\) restores the scale of one input after multiplying \(k\) of them; the index \(k\) then measures aggregation order rather than units.[1]
- Monotone guarantee: if \(k<\ell\), \(s_k^{1/k}\geq s_\ell^{1/\ell}\). A reversed comparison signals a mistaken calculation or a violated hypothesis, not an exception to the theorem.[1]
- Equality boundary: equal inputs make every link equal. Equality of the whole chain for \(n>1\) also forces equal inputs through the arithmetic–geometric endpoint. But with permitted zeros, an isolated later link can read \(0=0\) for unequal inputs; its equality alone is not an all-equal test.[1]
What It Is Not¶
The result is not an arbitrary statement that “averaging smooths values.” It names a particular family of symmetric sums, two explicit normalizations, an admissible input domain, and an indexed order relation. The polynomial coefficient identity is a useful equivalent encoding of the same \(s_k\), not a requirement that a polynomial be presented as the input.[1]
It is not Newton's inequality under another name. Newton constrains neighboring unnormalized-degree positions via \(s_{k-1}s_{k+1}\leq s_k^2\); Maclaurin constrains degree-root means. Nor does the classical statement hold for arbitrary signed reals: Tao gives an even-dimensional balanced \(+1/-1\) family in which an intermediate symmetric mean vanishes while a later one does not. Naively putting absolute-value bars around the classical chain does not repair that case.[1]
Finally, the seed's unqualified claim “equality exactly when all entries are equal” is too broad if it refers to each adjacent inequality on the nonnegative domain. For \((1,0,0)\), \(s_2=s_3=0\), hence \(s_2^{1/2}=s_3^{1/3}\), although the inputs differ. The whole-chain equality statement and single-link equality statement must be kept separate.
Scope of Application¶
In inequality analysis, the theorem compares all orders of elementary symmetric means for the same nonnegative vector. It can bound a later-order product statistic when an earlier-order mean is known, with the hypotheses carried along. At \(n=3\), the first link gives \(e_2/3\leq(e_1/3)^2\), equivalently \(e_2\leq e_1^2/3\). This is a derived special case, not a new theorem with no conditions.[1]
In polynomial coefficient analysis, write \(\prod_i(z-a_i)=\sum_{k=0}^n(-1)^k\binom nk s_k z^{n-k}\). When all roots \(a_i\) are nonnegative real, the normalized coefficient magnitudes obey the same descending root chain. A polynomial that is merely real-rooted, or one with roots of mixed sign, needs a different argument; Tao's signed-variable work develops such alternatives rather than extending the exact classical chain by fiat.[1]
Combinatorial extensions of Maclaurin-type inequalities exist, but their extra graph or extremal hypotheses must be checked on their own terms. Their names do not make every graph invariant an instance of the classical vector statement.[2]
Clarity¶
The theorem resolves which of several “symmetric means” is being compared. \(e_2\) for three numbers sums three pair products, whereas \(s_2=e_2/3\) averages them; only after taking \(\sqrt{s_2}\) can that quantity be compared directly with \(s_1\), an ordinary arithmetic mean. Forgetting either normalization may produce an apparent violation that is just a type or scale error.[1]
It also separates weak ordering from strictness. On \((1,0,0)\), the theorem remains valid, but the zero tail has equality. A declaration that equality at one link means all entries are equal would reject a valid case. The correct question is which link is equal and whether its means are positive or degenerate.
Manages Complexity¶
For each \(k\), \(e_k\) contains \(\binom nk\) terms. The chain compresses many separate coefficient comparisons into one ordered sequence indexed by degree. Once nonnegativity and the two normalizations are verified, a bound on \(s_k^{1/k}\) automatically bounds every higher-order degree root. One need not estimate every subset product independently.[1]
That compression is conditional. A raw polynomial coefficient list may come from negative roots, and the signs can create cancellation. In that case the smallness of one coefficient need not constrain later coefficients in the classical monotone manner. Applying the compact chain without checking root signs turns a useful reduction into a false claim.[1]
Abstract Reasoning¶
Take \(a=(1,2,3)\). The symmetric sums are \(e_1=6\), \(e_2=1\cdot2+1\cdot3+2\cdot3=11\), and \(e_3=6\). After binomial normalization, \((s_1,s_2,s_3)=(2,11/3,6)\). The comparable degree roots are \(2\), \(\sqrt{11/3}\), and \(\sqrt[3]{6}\), in strictly descending order. The raw numbers \(2,11/3,6\) do not descend; the theorem never claimed that they did. This calculation exhibits exactly what the degree-root step contributes.[1]
For a general nonnegative vector, Newton's adjacent log-concavity provides a proof route. With \(s_0=1\), nonnegative symmetric means and the inequalities \(s_{k-1}s_{k+1}\leq s_k^2\), the adjacent ratios do not increase where defined; equivalently, the averages of their logarithms yield the descending \(k\)-th roots. Zero cases follow by continuity or direct interpretation of the weak inequality. This explains why a local constraint on neighboring coefficients generates a global chain, while preserving the distinction between the two results.[1]
Knowledge Transfer¶
The transfer from a vector of nonnegative numbers to a polynomial with exactly those numbers as roots is literal: its normalized alternating coefficients are the same \(s_k\). A researcher can move between the two representations without changing the inequality or its hypotheses.[1]
The transfer to combinatorial generalizations is not automatically literal. Nikiforov develops an extension from classical symmetric-function inequalities to combinatorial settings, but the extension must state its own carrier and constraints. Outside mathematics, claims that “higher-order combinations are smaller” are at most analogies unless an actual elementary-symmetric normalization and nonnegative input vector have been identified.[2]
Examples¶
Numerical inequality check. Mapped back: admissible vector = \((1,2,3)\); elementary symmetric family = \((e_1,e_2,e_3)=(6,11,6)\); combinatorial averages = \((2,11/3,6)\); degree roots = \((2,\sqrt{11/3},\sqrt[3]{6})\); monotone guarantee = each listed root is smaller than the preceding one; equality boundary = positive unequal inputs make this displayed chain strict. This is an instance of the theorem, not just an illustration of its notation.[1]
Polynomial coefficient bound. For \(p(z)=(z-1)^2(z-4)=z^3-6z^2+9z-4\), Mapped back: admissible vector = the nonnegative roots \((1,1,4)\); elementary symmetric sums = \((6,9,4)\); binomial averages = \((2,3,4)\); degree roots = \((2,\sqrt3,\sqrt[3]4)\); monotone guarantee = \(2\geq\sqrt3\geq\sqrt[3]4\). The inequality bounds the normalized later coefficients from the earlier ones because this polynomial has the required roots.[1]
Zero boundary. Mapped back: admissible vector = the unequal nonnegative list \((1,0,0)\); elementary symmetric sums = \((e_1,e_2,e_3)=(1,0,0)\); binomial averages = \((1/3,0,0)\); degree roots = \((1/3,0,0)\); monotone guarantee = the sequence remains weakly descending; equality boundary = the last link is \(0=0\) even though the whole chain is not equal. The all-equal criterion cannot be applied to that isolated tail link.
Structural Tensions¶
Strong coefficient bound versus broader root class. Nonnegative roots permit a clean chain of exact comparisons. Allowing signed roots broadens the polynomial class but permits cancellation and failure of that chain; a weaker signed-variable estimate must replace it. Diagnostic: Are all the roots nonnegative, or has a different theorem been proved for this coefficient set?[1]
Short equality slogan versus degenerate cases. “Equality only for equal inputs” is memorable for the full chain and for the usual positive-input setting, but it misstates a single zero-to-zero link. Reporting the equality position and checking whether both compared means vanish costs a little extra analysis; skipping that step gives a false characterization. Diagnostic: Is the assertion about equality of the endpoints, every link, or one isolated adjacent pair?
Structural–Framed Character¶
Evaluative weight. The ordering is a mathematical theorem, not a preference for one distribution; whether its bound is useful depends on the problem. Human-practice dependence. Mathematicians choose notation and proof strategy, but once \(s_k\) and nonnegative inputs are fixed, the relation is not settled by social convention. Institutional origin. The theorem has a historical name and literature, yet no institution makes a violating nonnegative vector valid.[1]
Vocabulary travel. “Mean” and “balance” travel broadly, but this exact chain travels only where elementary symmetric sums and degree roots are present. Import versus recognition. A new field literally imports the theorem only by identifying nonnegative numeric inputs and proving the relevant quantity is the normalized symmetric mean, not by observing a superficial decline across levels.
Its character: strongly structural inside finite-dimensional inequality theory, yet domain-specific in the Encyclopedia. The exact algebraic carrier and hypotheses remain essential even when the reasoning is represented as polynomial coefficients.
Structural Core vs. Domain Accent¶
Portable skeleton. A family of many combinatorial quantities is normalized to a common scale and constrained by an ordered relation. That broad idea might suggest a higher-order abstraction, but this entry does not assert that all such families obey Maclaurin's theorem.
Domain accent. The carrier is a finite nonnegative real vector; the sums are elementary symmetric polynomials; the normalization uses \(\binom nk\) and a \(k\)-th root; the conclusion is a specific weak chain. These are not decorative domain details but the conditions under which the theorem is true.[1]
Why not prime. The numerical and polynomial settings are equivalent mathematical representations of one vector relation. They do not independently demonstrate the named theorem in three unrelated domains. The broader normalization-and-ordering skeleton remains a separate abstraction question rather than an asserted parent edge.
Instantiates / Related Primes¶
The live named inequality entries checked by exact surface and semantic-neighbor retrieval have different carriers and claims. Newton's inequalities are a mathematically related proof ingredient, but there is no verified current live typed node for an edge to them here. The arithmetic–geometric mean comparison is the first-to-last consequence, not a direct parent inserted solely to give this entry a place in the DAG.
Neighborhood in Abstraction Space¶
Maclaurin's Inequality sits in a sparse region of the domain-specific corpus (78th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Polynomials & Algebraic Invariants (20 abstractions)
Nearest neighbors
- AM–GM Inequality — 0.86
- Gram Matrix — 0.85
- Semiperfect Number — 0.82
- Daniell Integral — 0.81
- Maharam Algebra — 0.81
Computed from structural-signature embeddings · 2026-10-08
Not to Be Confused With¶
Newton's inequalities compare adjacent symmetric means through multiplicative log-concavity; they help prove, but are not identical to, the root-monotone chain. The arithmetic–geometric mean inequality compares only the two endpoints and omits the intermediate symmetric means. Tao's Maclaurin-type inequality is a different signed-variable bound and should not be cited as if its conclusion were the classical chain for arbitrary real entries. A generic polynomial coefficient inequality need not have the nonnegative-root hypothesis.[1]
References¶
[1] Terence Tao, “A Maclaurin type inequality” (2023), abstract and §1, especially equations (1.1)–(1.4). This article explicitly states the classical inequality before developing a distinct signed-variable variant. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s ↩t ↩u ↩v ↩w ↩x
[2] Vladimir Nikiforov, “An extension of Maclaurin's inequalities” (2006; v3 2013), abstract; cited only to document the existence and distinctness of a combinatorial extension. registry ↩a ↩b