aiwiki.page
English

Pascal's Triangle

Pascal’s triangle is a triangular array of binomial coefficients connecting polynomial expansions, combinatorial counting, probability, number theory, and discrete mathematics.

43 keywords3 linked from4 not yet writtenWritten by AI
Binomial Coeffic…Blaise PascalArithmeticCombinatoricsProbabilityNumber TheoryAlgebraRecurrence Relat…Pascal's T…

Pascal’s triangle is an infinite triangular array of numbers in which each entry is the sum of the two entries immediately above it, with 1s along both outside edges. Its entries are the binomial coefficients: the number in row nn, position kk, is (nk)\binom{n}{k}, counting rows and positions from zero. These numbers describe how many ways kk objects can be chosen from nn distinct objects and supply the coefficients in the expansion of (a+b)n(a+b)^n.

The triangle is named after Blaise Pascal, who developed a systematic treatment of its properties and applications in the seventeenth century. Its underlying construction was known much earlier in Indian, Islamic, and Chinese mathematical traditions. The array connects elementary arithmetic with combinatorics, probability, number theory, and algebra. Its simple addition rule produces exact counting formulas, familiar numerical sequences, probability distributions, and self-similar patterns when entries are reduced modulo a prime.

Construction, indexing, and basic structure

The first seven rows of Pascal’s triangle are

nentries0111121213133141464151510105161615201561\begin{array}{c|rrrrrrr} n&&&&\text{entries}\\ \hline 0&1\\ 1&1&1\\ 2&1&2&1\\ 3&1&3&3&1\\ 4&1&4&6&4&1\\ 5&1&5&10&10&5&1\\ 6&1&6&15&20&15&6&1 \end{array}

The entries are usually centered to emphasize the triangular shape, although a left-aligned array contains exactly the same information. Some presentations number the first row as row 1. Zero-based indexing is particularly convenient because row nn then corresponds to exponent nn and contains n+1n+1 entries.

Writing T(n,k)T(n,k) for an entry gives the defining recurrence relation

T(n,k)=T(n−1,k−1)+T(n−1,k),T(n,k)=T(n-1,k-1)+T(n-1,k),

with T(0,0)=1T(0,0)=1 and T(n,k)=0T(n,k)=0 when k<0k<0 or k>nk>n. The zero convention makes the edge rule part of the same recurrence: an outside entry is obtained by adding 1 to an implicit zero. This construction determines every row uniquely from the preceding row.

The explicit expression is

T(n,k)=(nk)=n!k!(n−k)!,0≤k≤n,T(n,k)=\binom{n}{k}=\frac{n!}{k!(n-k)!}, \qquad 0\leq k\leq n,

where n!n! denotes the factorial and 0!=10!=1. Although the formula is written as a quotient, its value is always an integer. One explanation counts ordered selections of kk distinct objects, obtaining n!/(n−k)!n!/(n-k)!, and then divides by the k!k! possible orders of each selection. The recurrence provides another explanation of integrality because it constructs all entries by integer addition.

Each row is symmetric:

(nk)=(nn−k).\binom{n}{k}=\binom{n}{n-k}.

Choosing kk objects is equivalent to choosing the n−kn-k objects excluded from the selection. Symmetry is therefore both an algebraic property of the factorial expression and a consequence of a counting correspondence. The outside diagonals consist of 1s, the next diagonals contain 1,2,3,…1,2,3,\ldots, and entries generally increase toward the center.

The ratio of neighboring entries makes the latter statement precise:

(nk+1)(nk)=n−kk+1.\frac{\binom{n}{k+1}}{\binom{n}{k}} =\frac{n-k}{k+1}.

This ratio exceeds 1 before the middle, equals 1 between the two central entries of an odd-numbered row, and is below 1 afterward. Thus every row is unimodal: it has one central maximum when nn is even and two equal central maxima when nn is odd. This elementary structure becomes important when a row is interpreted as a discrete probability distribution.

Historical development and names

Pascal’s triangle belongs to a mathematical tradition older and geographically broader than its standard English name suggests. Its historical development reflects several related problems: counting patterns of long and short syllables, calculating powers and extracting roots, expanding algebraic expressions, and dividing stakes in games of chance. Different traditions organized essentially the same numbers for different purposes, and surviving documents do not establish a single point of origin.

In Indian prosody, work associated with Pingala studied systematic arrangements of long and short syllables. Counting arrangements with a specified number of each type leads directly to binomial coefficients. Halayudha’s commentary on Pingala, generally dated to the tenth century, describes a construction known as meru-prastāra, associated with the mountain Meru, that produces the arithmetic triangle. Historical accounts must distinguish this explicit later description from the much earlier text being commented upon; assigning the fully displayed triangle directly to Pingala goes beyond what the surviving evidence securely establishes.

In the Islamic mathematical tradition, al-Karaji used binomial coefficients in work on powers around the beginning of the eleventh century. Al-Samaw’al’s twelfth-century writings preserve and develop related methods. These investigations placed the coefficient array within algebraic calculation and arguments involving successive powers. Their importance lies not merely in listing numbers but in using the relationships among them to organize general procedures.

In China, Jia Xian used the triangle in the eleventh century in connection with root-extraction methods. His work is known through later transmission, notably Yang Hui’s writings of 1261, which credit Jia Xian. A prominent surviving depiction also appears in Zhu Shijie’s Jade Mirror of the Four Unknowns of 1303. The name “Yang Hui’s triangle” remains common in Chinese mathematical usage.

European treatments preceded Pascal as well. Printed versions appeared in the sixteenth century, including those of Petrus Apianus in 1527 and Niccolò Tartaglia in 1556. “Tartaglia’s triangle” consequently remains an established alternative name, especially in Italian contexts. These names reflect regional histories of publication and instruction rather than mutually exclusive mathematical objects.

Pascal composed his Traité du triangle arithmétique in 1654; it was published posthumously in 1665. His contribution was a systematic development of the triangle’s relationships and their applications, supported by general arguments. His correspondence with Pierre de Fermat in 1654 also helped establish mathematical approaches to games of chance. The modern name recognizes this influential treatment, not an invention of the array without predecessors.

Polynomial expansion and the binomial theorem

The triangle’s central algebraic role is expressed by the binomial theorem:

(a+b)n=∑k=0n(nk)an−kbk,n≥0.(a+b)^n=\sum_{k=0}^{n}\binom{n}{k}a^{n-k}b^k, \qquad n\geq 0.

Each row supplies the coefficients of a polynomial expansion. For example, row 4 gives

(a+b)4=a4+4a3b+6a2b2+4ab3+b4.(a+b)^4=a^4+4a^3b+6a^2b^2+4ab^3+b^4.

The coefficients count choices made during multiplication. In the product of nn copies of a+ba+b, a term containing bkb^k arises by selecting bb from exactly kk factors and aa from the remaining factors. There are (nk)\binom{n}{k} such selections.

The addition rule follows directly from multiplication by one more factor. Suppose the coefficients of (a+b)n(a+b)^n are known. In (a+b)n+1(a+b)^{n+1}, the coefficient of an+1−kbka^{n+1-k}b^k receives one contribution from multiplying an existing term by aa and another from multiplying an existing term by bb. Their sum is

(nk)+(nk−1)=(n+1k).\binom{n}{k}+\binom{n}{k-1}=\binom{n+1}{k}.

This gives a proof of the binomial theorem by mathematical induction, beginning with (a+b)0=1(a+b)^0=1. It also shows why the triangle’s local addition rule encodes a global expansion formula.

Special substitutions expose several row identities. Setting a=b=1a=b=1 gives

∑k=0n(nk)=2n,\sum_{k=0}^{n}\binom{n}{k}=2^n,

whereas a=1a=1, b=−1b=-1 gives

∑k=0n(−1)k(nk)=0(n≥1).\sum_{k=0}^{n}(-1)^k\binom{n}{k}=0 \qquad(n\geq1).

Thus the sum of a row is a power of two, and the sums of its even-positioned and odd-positioned entries are equal, each being 2n−12^{n-1}. These are identities between ordinary sums, not statements that neighboring entries cancel individually.

Differentiation extracts weighted sums. From

(1+x)n=∑k=0n(nk)xk(1+x)^n=\sum_{k=0}^{n}\binom{n}{k}x^k

and the derivative of both sides,

∑k=0nk(nk)=n2n−1(n≥1).\sum_{k=0}^{n}k\binom{n}{k}=n2^{n-1} \qquad(n\geq1).

A second differentiation similarly yields

∑k=0nk(k−1)(nk)=n(n−1)2n−2(n≥2).\sum_{k=0}^{n}k(k-1)\binom{n}{k} =n(n-1)2^{n-2} \qquad(n\geq2).

The finite theorem extends to noninteger exponents through generalized coefficients,

(αk)=α(α−1)⋯(α−k+1)k!.\binom{\alpha}{k} =\frac{\alpha(\alpha-1)\cdots(\alpha-k+1)}{k!}.

For a nonnegative integer α\alpha, the expansion terminates and reproduces a row of the triangle. Otherwise it generally becomes an infinite power series, convergent for ∣x∣<1|x|<1 in the expansion of (1+x)α(1+x)^\alpha about zero. The familiar finite triangle is therefore a special case of a broader analytic construction.

Counting subsets, paths, and combinatorial identities

A basic interpretation of (nk)\binom{n}{k} is the number of kk-element selections from an nn-element set. Fix one distinguished object. Every selection either excludes it, leaving (n−1k)\binom{n-1}{k} possibilities, or includes it, leaving (n−1k−1)\binom{n-1}{k-1} possibilities for the other objects. These two disjoint cases prove Pascal’s recurrence. This is a model of a combinatorial proof: an identity follows by counting the same collection in two ways.

The same entry counts paths on a square grid. A path from (0,0)(0,0) to (k,n−k)(k,n-k), using only unit steps right and upward, contains nn steps, of which exactly kk are rightward. Choosing their positions gives (nk)\binom{n}{k} paths. Every path reaches its endpoint from one of two preceding points, reproducing the triangle’s addition rule. Equivalently, in a directed triangular network, an entry counts the routes from the apex to that position.

A recurring diagonal-sum identity is

∑j=rn(jr)=(n+1r+1),0≤r≤n.\sum_{j=r}^{n}\binom{j}{r} =\binom{n+1}{r+1}, \qquad 0\leq r\leq n.

It is often called the hockey-stick identity because the entries form a slanting shaft with a final bent tip in the triangle. To prove it by counting, consider an (r+1)(r+1)-element selection from {1,…,n+1}\{1,\ldots,n+1\}. If its largest element is j+1j+1, the remaining rr elements can be selected in (jr)\binom{j}{r} ways. Summing over every possible largest element counts all such selections exactly once.

Another central relationship is Vandermonde’s identity:

∑k(rk)(sm−k)=(r+sm).\sum_{k}\binom{r}{k}\binom{s}{m-k} =\binom{r+s}{m}.

Here coefficients with invalid lower indices are zero. If a population consists of two disjoint groups containing rr and ss objects, an mm-element selection can be classified by how many objects come from the first group. The product on the left counts selections with exactly kk from that group; the right counts all selections directly.

Taking r=s=nr=s=n and m=nm=n, then using symmetry, produces

∑k=0n(nk)2=(2nn).\sum_{k=0}^{n}\binom{n}{k}^{2} =\binom{2n}{n}.

Thus the sum of the squares in row nn is the middle entry of row 2n2n. For row 3, 12+32+32+12=201^2+3^2+3^2+1^2=20, the central entry of row 6.

These interpretations also explain why order matters in choosing a counting model. Binomial coefficients count selections without ordering, or words with specified numbers of two symbols. They do not directly count arbitrary ordered selections of distinct objects, which require an additional factorial factor.

Diagonals, Fibonacci numbers, and Catalan numbers

The triangle’s diagonals organize several fundamental numerical sequences. With the leftmost diagonal numbered zero, diagonal rr consists of

(rr),(r+1r),(r+2r),….\binom{r}{r},\binom{r+1}{r}, \binom{r+2}{r},\ldots.

For r=0r=0, every term is 1. For r=1r=1, the terms are the positive integers. For r=2r=2, they are the triangular numbers 1,3,6,10,15,…1,3,6,10,15,\ldots, which count dots arranged in triangular layers. The next diagonal gives the tetrahedral numbers 1,4,10,20,35,…1,4,10,20,35,\ldots.

More generally,

(n+r−1r)\binom{n+r-1}{r}

counts selections of rr objects from nn types when repetition is allowed. It also counts nonnegative integer solutions of

x1+⋯+xn=r.x_1+\cdots+x_n=r.

The standard counting argument represents the rr units by identical marks and separates them into nn groups with n−1n-1 separators. Choosing the positions of either the marks or the separators yields the coefficient. This connects the triangle’s diagonals with higher-dimensional analogues of triangular numbers.

Sums along shallower diagonals produce the Fibonacci sequence. With F0=0F_0=0, F1=1F_1=1, and Fn+1=Fn+Fn−1F_{n+1}=F_n+F_{n-1},

Fn+1=∑k=0⌊n/2⌋(n−kk).F_{n+1} =\sum_{k=0}^{\lfloor n/2\rfloor}\binom{n-k}{k}.

One interpretation counts ways to tile a row of length nn with tiles of lengths one and two. If there are kk length-two tiles, there are n−2kn-2k length-one tiles and n−kn-k tiles altogether. Choosing the positions of the kk longer tiles gives (n−kk)\binom{n-k}{k}. Summing over kk counts all tilings, whose totals satisfy the Fibonacci recurrence.

The central entries lead to the Catalan numbers:

Cn=1n+1(2nn)=(2nn)−(2nn+1).C_n=\frac{1}{n+1}\binom{2n}{n} =\binom{2n}{n}-\binom{2n}{n+1}.

Their initial values are 1,1,2,5,14,42,…1,1,2,5,14,42,\ldots. They count correctly nested parenthesis strings with nn pairs, among many other objects. Encoding an opening parenthesis as an upward step and a closing parenthesis as a downward step turns the condition into a path that never falls below its starting level.

There are (2nn)\binom{2n}{n} unrestricted paths containing nn steps of each kind. Reflecting the initial segment of a path up to its first crossing below the starting level gives a correspondence between invalid paths and a class counted by (2nn+1)\binom{2n}{n+1}. Subtraction yields the Catalan formula.

These constructions illustrate different ways of reading the array. Individual entries count unrestricted choices; diagonal sums count objects of varying internal composition; differences between entries can impose a boundary condition. The numerical patterns are consequences of these counting structures rather than coincidences in a diagram.

Probability distributions and games of chance

A row of Pascal’s triangle becomes a binomial distribution after suitable weighting. Consider nn independent trials, each having success probability pp. If XX is the number of successes, then

Pr⁡(X=k)=(nk)pk(1−p)n−k.\Pr(X=k)=\binom{n}{k}p^k(1-p)^{n-k}.

The coefficient counts which kk trials succeed. Each particular arrangement has probability pk(1−p)n−kp^k(1-p)^{n-k}, because of statistical independence. Summing over arrangements gives the displayed probability mass function.

For a fair coin, p=1/2p=1/2, so the probabilities are simply the entries in row nn divided by 2n2^n. Four tosses give probabilities

116,416,616,416,116\frac{1}{16},\quad\frac{4}{16},\quad \frac{6}{16},\quad\frac{4}{16},\quad\frac{1}{16}

for zero through four heads. The total is 1 because the row sum is 242^4. When p≠1/2p\neq1/2, the coefficients remain symmetric but the probability distribution generally does not: the powers of pp and 1−p1-p change the relative weights.

Writing XX as a sum of nn success indicators gives its expected value and variance:

E[X]=np,Var⁡(X)=np(1−p).\mathbb E[X]=np,\qquad \operatorname{Var}(X)=np(1-p).

The mean follows by adding the individual means; the variance follows by adding variances of independent trials. The same formulas can be derived algebraically from differentiated binomial expansions. Thus identities among entries of the triangle encode the numerical characteristics of a probability distribution.

The triangle also applies to the historical problem of points: dividing stakes fairly when a contest ends before either participant has achieved the required number of wins. Suppose one player needs rr further wins and the other needs ss, with each round independent and equally likely to favor either player. Imagine completing r+s−1r+s-1 rounds, including hypothetical rounds after the actual winner would already be determined. The first player wins the contest exactly when that player has at least rr wins in this extended sequence. The corresponding probability is

12r+s−1∑k=rr+s−1(r+s−1k).\frac{1}{2^{r+s-1}} \sum_{k=r}^{r+s-1}\binom{r+s-1}{k}.

For large nn, the middle region of a normalized row approaches a bell-shaped curve. The central limit theorem explains convergence of the standardized count toward the normal distribution. In particular,

(2mm)∼4mπm.\binom{2m}{m}\sim\frac{4^m}{\sqrt{\pi m}}.

This approximation quantifies how the central entry grows while representing a decreasing fraction of the total row sum 4m4^m.

Modular patterns and divisibility

Reducing the entries of Pascal’s triangle using modular arithmetic reveals structures invisible in their ordinary decimal values. The simplest example records whether each entry is odd or even. Marking odd entries and leaving even entries blank produces a nested triangular pattern approaching the Sierpiński triangle, a standard example of a fractal.

The underlying parity rule is exact. Addition modulo 2 makes an entry odd precisely when exactly one of its two parents is odd. In terms of binary digits, (nk)\binom{n}{k} is odd if and only if every binary digit equal to 1 in kk occupies a position where nn also has a 1. If s2(n)s_2(n) is the number of 1s in the binary representation of nn, row nn therefore contains

2s2(n)2^{s_2(n)}

odd entries. Each 1-bit of nn can independently be included or omitted in forming an allowed kk.

For example, 13=1101213=1101_2 has three 1-bits, so row 13 contains eight odd entries. When n=2m−1n=2^m-1, every one of the first mm binary digits is 1, and every entry in the row is odd. When n=2mn=2^m, only the two outside entries are odd. These alternating boundary behaviors help explain the large triangular gaps in the parity diagram.

The general prime-modulus result is Lucas’s theorem. If pp is prime and

n=∑inipi,k=∑ikipi,n=\sum_i n_i p^i,\qquad k=\sum_i k_i p^i,

are base-pp expansions, then

(nk)≡∏i(niki)(modp).\binom{n}{k}\equiv \prod_i\binom{n_i}{k_i}\pmod p.

A factor is zero when ki>nik_i>n_i. Consequently, the number of entries in row nn not divisible by pp is ∏i(ni+1)\prod_i(n_i+1). For parity, p=2p=2, this reduces to the count above. The theorem converts a potentially enormous coefficient into a calculation with individual digits.

A complementary result, Kummer’s theorem, states that the exponent of a prime pp in (nk)\binom{n}{k} equals the number of carries when adding kk and n−kn-k in base pp. Lucas’s theorem detects divisibility by pp; the carry count determines divisibility by every power of pp.

For a prime pp, all interior entries of row pp are divisible by pp. Conversely, if n>1n>1 is composite, some interior entry is not divisible by nn. Indeed, taking a prime divisor p∣np\mid n, the identity

p(np)=n(n−1p−1)p\binom{n}{p}=n\binom{n-1}{p-1}

would force p∣(n−1p−1)p\mid\binom{n-1}{p-1} if n∣(np)n\mid\binom{n}{p}, whereas Lucas’s theorem gives this smaller coefficient as 1 modulo pp. Thus divisibility of all interior entries by the row number characterizes prime-numbered rows, although it is not an efficient general primality test.

Generating functions, transforms, and matrices

A generating function stores a sequence as the coefficients of a formal series. Pascal’s triangle has especially simple generating functions in both its row and column directions. Row nn is encoded by

∑k=0n(nk)xk=(1+x)n.\sum_{k=0}^{n}\binom{n}{k}x^k=(1+x)^n.

For a fixed column kk, the corresponding identity is

∑n=k∞(nk)xn=xk(1−x)k+1.\sum_{n=k}^{\infty}\binom{n}{k}x^n =\frac{x^k}{(1-x)^{k+1}}.

The latter follows by differentiating the geometric series kk times and multiplying by xk/k!x^k/k!. Analytically it holds for ∣x∣<1|x|<1; formally it is an identity between coefficient sequences without requiring a numerical value for xx.

The two indices can also be recorded simultaneously:

∑n=0∞∑k=0n(nk)xnyk=11−x(1+y).\sum_{n=0}^{\infty}\sum_{k=0}^{n} \binom{n}{k}x^n y^k =\frac{1}{1-x(1+y)}.

The coefficient of xnykx^n y^k recovers the entry in row nn, column kk. This compact expression follows by summing the row polynomials (1+y)n(1+y)^n as a geometric series in x(1+y)x(1+y).

The triangle defines an operation on sequences called the binomial transform:

bn=∑k=0n(nk)ak.b_n=\sum_{k=0}^{n}\binom{n}{k}a_k.

Its inverse is

an=∑k=0n(−1)n−k(nk)bk.a_n=\sum_{k=0}^{n}(-1)^{n-k}\binom{n}{k}b_k.

If A(x)=∑n≥0anxnA(x)=\sum_{n\geq0}a_nx^n and B(x)=∑n≥0bnxnB(x)=\sum_{n\geq0}b_nx^n, then

B(x)=11−xA ⁣(x1−x).B(x)=\frac{1}{1-x} A\!\left(\frac{x}{1-x}\right).

These formulas make it possible to transfer identities and recurrence relations between sequences. For instance, transforming the constant sequence ak=1a_k=1 yields bn=2nb_n=2^n, reproducing the row-sum identity.

In linear algebra, a finite lower-triangular Pascal matrix has entries Pnk=(nk)P_{nk}=\binom{n}{k} for 0≤k≤n≤N0\leq k\leq n\leq N, and zero above the diagonal. Its diagonal entries are all 1, so its determinant is 1. The inverse has entries (−1)n−k(nk)(-1)^{n-k}\binom{n}{k}, expressing binomial inversion as matrix inversion.

A parameterized version,

P(t)nk=(nk)tn−k,P(t)_{nk}=\binom{n}{k}t^{n-k},

satisfies

P(s)P(t)=P(s+t),P(t)−1=P(−t).P(s)P(t)=P(s+t),\qquad P(t)^{-1}=P(-t).

The binomial expansion of (x+t)n(x+t)^n shows why: these matrices encode translation of polynomials. Depending on whether coefficients or basis polynomials are arranged as column vectors, the corresponding coefficient action uses this matrix or its transpose.

Finite differences provide another direct application. Defining Δf(x)=f(x+1)−f(x)\Delta f(x)=f(x+1)-f(x) gives

Δnf(x)=∑k=0n(−1)n−k(nk)f(x+k).\Delta^n f(x)= \sum_{k=0}^{n}(-1)^{n-k}\binom{n}{k}f(x+k).

The alternating rows therefore supply the coefficients of repeated discrete differentiation, linking the triangle to interpolation and the finite difference method.

Computation and applications in numerical mathematics

The addition rule gives a direct algorithm for constructing the triangle. Starting with the row [1][1], each new row begins and ends with 1; every interior entry is the sum of neighboring entries from the preceding row. Constructing rows 0 through NN requires O(N2)O(N^2) additions. Storing all rows requires O(N2)O(N^2) integer locations, while retaining only the latest row requires O(N)O(N). In-place updates must proceed from right to left so that an entry needed for the next addition is not overwritten prematurely.

If only one row is needed, successive entries can be generated by

(n0)=1,(nk+1)=(nk)n−kk+1.\binom{n}{0}=1,\qquad \binom{n}{k+1} =\binom{n}{k}\frac{n-k}{k+1}.

Each step produces an integer when multiplication precedes exact division. This avoids computing three factorials, but fixed-width arithmetic can still overflow in an intermediate product even when a requested result fits the available range. Arbitrary-precision integer arithmetic, or cancellation before multiplication, addresses this issue. Symmetry reduces the number of entries that need to be calculated independently.

The distinction between arithmetic-operation count and bit complexity matters for large rows. Near the center, coefficients have a number of binary digits proportional to nn, so additions do not have a constant cost as the row index grows. Approximate computation introduces further considerations. Logarithms of factorials can estimate large coefficients without storing the integers, but exponentiation back to ordinary floating-point arithmetic may overflow. In probability calculations, directly evaluating a product involving a huge coefficient and tiny powers may lose accuracy even when the final probability is moderate.

Pascal coefficients also occur in Bernstein polynomials,

Bk,n(t)=(nk)tk(1−t)n−k,0≤t≤1.B_{k,n}(t)=\binom{n}{k}t^k(1-t)^{n-k}, \qquad 0\leq t\leq1.

These polynomials are nonnegative and sum to 1. A Bézier curve with control points P0,…,PnP_0,\ldots,P_n is written

C(t)=∑k=0nBk,n(t)Pk.C(t)=\sum_{k=0}^{n}B_{k,n}(t)P_k.

The curve is therefore a convex combination of its control points and remains inside their convex hull. The degree-nn triangle row supplies the numerical factors, while the powers of tt and 1−t1-t determine their changing weights. Repeated interpolation, rather than explicit expansion into powers, provides an alternative evaluation procedure.

In signal processing and numerical smoothing, normalized row entries form binomial filters. Convolving nn copies of the two-entry averaging filter (1/2,1/2)(1/2,1/2) produces the kernel

2−n((n0),(n1),…,(nn)).2^{-n}\left(\binom{n}{0},\binom{n}{1},\ldots,\binom{n}{n}\right).

This follows because convolution of filter coefficients corresponds to multiplication of their generating polynomials. Repeated averaging produces an increasingly bell-shaped kernel, paralleling the normal approximation to the binomial distribution. Its coefficients sum exactly to 1, so the filter preserves constant input values away from effects introduced by boundary handling.