Pascal's triangle: Difference between revisions
imported>Olier Raby (→Normal Curve: Ordered differently two references.) |
mNo edit summary |
||
(37 intermediate revisions by 11 users not shown) | |||
Line 1: | Line 1: | ||
{{subpages}} | {{subpages}} | ||
'''Pascal's triangle''' is a convenient tabular representation of the [[binomial coefficient]]s. Already known in the 11th century,<ref>[http://www-gap.dcs.st-and.ac.uk/~history/Biographies/Al-Karaji.html ''Abu Bekr ibn Muhammad ibn al-Husayn Al-Karaji''], School of Mathematics and Statistics, University of St Andrews. Consulted 2005-09-03.</ref> it was adopted in the [[Western world]] under this name after [[Blaise Pascal]] published his ''Traité du triangle arithmétique'' ("Treatise on the Arithmetical Triangle") in 1654. | |||
Pascal's triangle appears under different formats. Here is its most common: | |||
Pascal's triangle appears under different formats. Here is its most common : | |||
<center> | <center> | ||
Line 19: | Line 17: | ||
</math> | </math> | ||
</center> | </center> | ||
We can use Pascal's triangle to compute the [[binomial expansion]] of <math>\scriptstyle (x + y)^n</math>. For instance, | We can use Pascal's triangle to compute the [[binomial expansion]] of <math>\scriptstyle (x + y)^n</math>. For instance, | ||
:<math> | :<math> (x+y)^4 = 1 \times x^4 + 4 \times x^3y + 6 \times x^2y^2 + 4 \times xy^3 + 1 \times y^4 </math> | ||
The triangle shows the [[coefficient]]s <math>\scriptstyle 1, 4, 6, 4, 1</math> on the fifth row. | The triangle shows the [[coefficient]]s <math>\scriptstyle 1, 4, 6, 4, 1</math> on the fifth row. | ||
Line 30: | Line 27: | ||
== History == | == History == | ||
{{Image|Pascal's triangle - Yanghui triangle.PNG|right|200px|Drawing of Pascal's Triangle published in 1303 by Zhu Shijie (1260-1320), in his Si Yuan Yu Jian. It was called Yanghui Triangle by the Chinese, after the mathematician Yang Hui.<br>The fourth entry from the left in the second row from the bottom appears to be a typo (34 instead of 35, correctly given in the fifth entry in the same row).}} | |||
The earliest explicit depictions of a triangle of [[binomial coefficients]] occur in the 10th century in commentaries on the ''Chandas Shastra'', an ancient Indian book on [[Sanskrit]] [[prosody]] written by [[Pingala]] between the 5th century BC and 2nd century BC. While Pingala's work only survives in fragments, the commentator [[Halayudha]], around 975, used the triangle to explain obscure references to ''Meru-prastaara'', the "Staircase of Mount Meru". It was also realised that the shallow diagonals of the triangle sum to the [[Fibonacci number]]s. The Indian mathematician Bhattotpala (c. 1068) later gives rows 0-16 of the triangle. | The earliest explicit depictions of a triangle of [[binomial coefficients]] occur in the 10th century in commentaries on the ''Chandas Shastra'', an ancient Indian book on [[Sanskrit]] [[prosody]] written by [[Pingala]] between the 5th century BC and 2nd century BC. While Pingala's work only survives in fragments, the commentator [[Halayudha]], around 975, used the triangle to explain obscure references to ''Meru-prastaara'', the "Staircase of Mount Meru". It was also realised that the shallow diagonals of the triangle sum to the [[Fibonacci number]]s. The Indian mathematician Bhattotpala (c. 1068) later gives rows 0-16 of the triangle. | ||
At around the same time, it was discussed in [[Persia]] by the mathematician [[Al-Karaji]] (953–1029) and the poet-astronomer-mathematician [[Omar Khayyám]] (1048-1131); thus the triangle is referred to as the "Khayyam triangle" in | At around the same time, it was discussed in [[Persia]] by the mathematician [[Al-Karaji]] (953–1029) and the poet-astronomer-mathematician [[Omar Khayyám]] (1048-1131); thus the triangle is referred to as the "Khayyam triangle" in Iran. Several theorems related to the triangle were known, including the [[binomial theorem]]. It seems that Khayyam used a method of finding ''n''th roots based on the binomial expansion, and therefore on the binomial coefficients. | ||
In 13th century, [[Yang Hui]] (1238-1298) presented an arithmetic triangle, which was the same as Pascal's Triangle. Today, Pascal's triangle is called "Yang Hui's triangle" in | In 13th century, [[Yang Hui]] (1238-1298) presented an arithmetic triangle, which was the same as Pascal's Triangle. Today, Pascal's triangle is called "Yang Hui's triangle" in China. In [[Italy]], it is known as "Tartaglia's triangle", named for the Italian algebraist [[Niccolo Fontana Tartaglia]] who lived a century before Pascal. | ||
In 1655, [[Blaise Pascal]] published its ''Traité du triangle arithmétique'' ("Treatise on arithmetical triangle"), wherein he collected several results then known about the triangle, and employed them to solve problems in [[Probability theory|probabilities]]. The triangle was later named after Pascal by [[Pierre Raymond de Montmort]] (1708) and [[Abraham de Moivre]] (1730). | In 1655, [[Blaise Pascal]] published its ''Traité du triangle arithmétique'' ("Treatise on arithmetical triangle"), wherein he collected several results then known about the triangle, and employed them to solve problems in [[Probability theory|probabilities]]. The triangle was later named after Pascal by [[Pierre Raymond de Montmort]] (1708) and [[Abraham de Moivre]] (1730). | ||
After that, even if it was useful in many areas of | After that, even if it was useful in many areas of mathematics, most research was done within its descendants, like probabilities and [[combinatorics]]. | ||
== Some Properties == | == Some Properties == | ||
Each term in the triangle is the sum of the two terms above it<ref name="not_for_ones">This rule does not apply to the ones bordering the triangle. We just insert them.</ref>. For instance, <math>\scriptstyle 6 = 3 + 3</math>. The binomial coefficients relate to this construction by Pascal's rule, which states that if | Each term in the triangle is the sum of the two terms above it<ref name="not_for_ones">This rule does not apply to the ones bordering the triangle. We just insert them.</ref>. For instance, <math>\scriptstyle 6 = 3 + 3</math>. The binomial coefficients relate to this construction by Pascal's rule, which states that if | ||
Line 71: | Line 67: | ||
</math> | </math> | ||
</div> | </div> | ||
Each coefficient is the sum of the coefficient exactly over it and its left neighbour<ref name="not_for_ones"/>. For instance, <math>\scriptstyle 3 + 3 = 6</math>. Let's call this rule the "addition rule". | Each coefficient is the sum of the coefficient exactly over it and its left neighbour<ref name="not_for_ones"/>. For instance, <math>\scriptstyle 3 + 3 = 6</math>. Let's call this rule the "addition rule". | ||
Using this format, it is easy to apply an index to each row : | Using this format, it is easy to apply an index to each row : | ||
<math> | <math> | ||
Line 91: | Line 85: | ||
\end{array} | \end{array} | ||
</math> | </math> | ||
Starting the indices at zero facilitates many calculations. | Starting the indices at zero facilitates many calculations. | ||
Line 98: | Line 91: | ||
Since there is a formula for summing a row, then maybe there is one for a column ? It is the case. This time, we index the columns : | Since there is a formula for summing a row, then maybe there is one for a column ? It is the case. This time, we index the columns : | ||
<math> | <math> | ||
Line 114: | Line 106: | ||
\end{array} | \end{array} | ||
</math> | </math> | ||
Anyone familiar with the [[factorial]] function can easily find the general formula. The sum of a column <math>\scriptstyle c</math> ending at row <math>\scriptstyle r</math> is | Anyone familiar with the [[factorial]] function can easily find the general formula. The sum of a column <math>\scriptstyle c</math> ending at row <math>\scriptstyle r</math> is | ||
Line 120: | Line 111: | ||
:<math>\frac{(r+1) \times r \, \times \cdots \times (r-c+1)}{(c + 1)!} \text{ with } r, c \in \N^*</math>. | :<math>\frac{(r+1) \times r \, \times \cdots \times (r-c+1)}{(c + 1)!} \text{ with } r, c \in \N^*</math>. | ||
There is another method to compute this sum, see <ref>Suppose we wish to add the terms in row 3, i.e. the fourth column, until row 6. The sum is given by multiplying four terms at [[numerator]], starting at <math>\scriptstyle (r + 1) = 6 + 1 = 7</math>, and four terms at the [[denominator]] starting at <math>\scriptstyle (c + 1) = (3 + 1) = 4</math>. It | There is another method to compute this sum, see <ref>Suppose we wish to add the terms in row 3, i.e. the fourth column, until row 6. The sum is given by multiplying four terms at [[numerator]], starting at <math>\scriptstyle (r + 1) = 6 + 1 = 7</math>, and four terms at the [[denominator]] starting at <math>\scriptstyle (c + 1) = (3 + 1) = 4</math>. It is equal to <math>\scriptstyle \frac{ 7 \times 6 \times 5 \times 4 }{ 4 \times 3 \times 2 \times 1 } = 35</math>. In short, fourth column, four terms at numerator, four terms at denominator, all decreasing.</ref>. | ||
Up until now, we added along the rows and the columns. We can add along the diagonals. Doing so from left to right and from bottom to top gives : | Up until now, we added along the rows and the columns. We can add along the diagonals. Doing so from left to right and from bottom to top gives : | ||
<math> | <math> | ||
\begin{array}{ccccccccccr} | \begin{array}{ccccccccccr} | ||
1 | 1 & & & & & & & = & 1 \\ | ||
1 | 1 & & & & & & & = & 1 \\ | ||
1 & + & 1 | 1 & + & 1 & & & & & = & 2 \\ | ||
1 & + & 2 | 1 & + & 2 & & & & & = & 3 \\ | ||
1 & + & 3 & + & 1 | 1 & + & 3 & + & 1 & & & = & 5 \\ | ||
1 & + & 4 & + & 3 | 1 & + & 4 & + & 3 & & & = & 8 \\ | ||
1 & + & 5 & + & 6 & + & 1 | 1 & + & 5 & + & 6 & + & 1 & = & 13 \\ | ||
1 & + & 6 & + &10 & + & 4 | 1 & + & 6 & + &10 & + & 4 & = & 21 \\ | ||
... | ...& & & & & & & = &... \\ | ||
\end{array} | \end{array} | ||
</math> | </math> | ||
The numbers on the right are the [[Fibonacci number]]s. | The numbers on the right are the [[Fibonacci number]]s. | ||
Line 144: | Line 133: | ||
== One Row at a Time == | == One Row at a Time == | ||
We can build any row if we know the row before just by adding its terms two at a time. However, it is possible to build a row directly. We will build the row 4 in order to discover the rule. Each row starts with 1 : | We can build any row if we know the row before just by adding its terms two at a time. However, it is possible to build a row directly. We will build the row 4 in order to discover the rule. Each row starts with 1 : | ||
<math> | <math> | ||
Line 153: | Line 141: | ||
:<math> 1 \times \frac{4 - 0}{1 + 0} = 4 ~</math> | :<math> \scriptstyle 1 \times \frac{4 - 0}{1 + 0} = 4 ~</math> | ||
:<math> 4 \times \frac{4 - 1}{1 + 1} = 6 ~</math> | :<math> \scriptstyle 4 \times \frac{4 - 1}{1 + 1} = 6 ~</math> | ||
:<math> 6 \times \frac{4 - 2}{1 + 2} = 4 ~</math> | :<math> \scriptstyle 6 \times \frac{4 - 2}{1 + 2} = 4 ~</math> | ||
:<math> 4 \times \frac{4 - 3}{1 + 3} = 1 ~</math> | :<math> \scriptstyle 4 \times \frac{4 - 3}{1 + 3} = 1 ~</math> | ||
:<math> 1 \times \frac{4 - 4}{1 + 4} = 0 ~</math> | :<math> \scriptstyle 1 \times \frac{4 - 4}{1 + 4} = 0 ~</math> | ||
:<math> \cdots ~</math> | :<math> \scriptstyle \cdots ~</math> | ||
Line 169: | Line 157: | ||
Once the row and the column indices are know, we can compute the neighbours, either right or left, of any term. Because it only applies to a row, we call it the "row rule". | Once the row and the column indices are know, we can compute the neighbours, either right or left, of any term. Because it only applies to a row, we call it the "row rule". | ||
== Polynomials with coefficients in triangle == | |||
* [[Binomial theorem]] | |||
* [[Fibonacci polynomials]] | |||
== Newton's Binomial Coefficients == | == Newton's Binomial Coefficients == | ||
[[Isaac Newton]] studied the triangle's properties and discovered two remarkable generalizations.<ref>Eli Maor, ''e: The Story of a Number'', Princeton University Press, 1994, p.71. ISBN 0-691-05854-7.</ref> | [[Isaac Newton]] studied the triangle's properties and discovered two remarkable generalizations.<ref>Eli Maor, ''e: The Story of a Number'', Princeton University Press, 1994, p.71. ISBN 0-691-05854-7.</ref> | ||
He extended the triangle along the negative axis. To better understand how he achieved this, let us write the triangle using another format : | He extended the triangle along the negative axis. To better understand how he achieved this, let us write the triangle using another format : | ||
<math> | <math> | ||
Line 191: | Line 181: | ||
</math> | </math> | ||
[[Jacob Bernoulli]] is the "father" of this mathematical object named the "figurate number triangle"<ref>Eric W. Weisstein, ''CRC Concise Encyclopedia of Mathematics'', CRC Press, 1999, p. 636. ISBN 0-8493-9640-9</ref>. To ease the understanding in the following text, we will call it the "Bernoulli triangle", even if this [[matrix]] does not resemble a triangle anymore ! | |||
[[Jacob Bernoulli]] is the "father" of this mathematical object named the "figurate number triangle"<ref>Eric W. Weisstein, ''CRC Concise Encyclopedia of Mathematics'', CRC Press, 1999, p. 636. ISBN 0-8493-9640-9</ref>. To ease the understanding in the following text, we will call it the "Bernoulli triangle", even | |||
Using the row rule, let's compute the row -1 : | Using the row rule, let's compute the row -1 : | ||
<math> | <math> | ||
Line 209: | Line 197: | ||
\end{array} | \end{array} | ||
</math> | </math> | ||
There is an infinity of <math>\scriptstyle 1</math> and <math>\scriptstyle -1</math>. This is another hint that this object is a matrix. Each term in this row obeys the addition rule and the row rule. Some calculations should convince you. | There is an infinity of <math>\scriptstyle 1</math> and <math>\scriptstyle -1</math>. This is another hint that this object is a matrix. Each term in this row obeys the addition rule and the row rule. Some calculations should convince you. | ||
We can further extend the triangle along the negative axis. | We can further extend the triangle along the negative axis. | ||
<math> | <math> | ||
Line 232: | Line 218: | ||
\end{array} | \end{array} | ||
</math> | </math> | ||
As we wrote earlier, Newton did not stop there. He asked himself if there were rows with fractional index, like <math>\scriptstyle \frac{1}{2}</math>. And the answer is yes ! | As we wrote earlier, Newton did not stop there. He asked himself if there were rows with fractional index, like <math>\scriptstyle \frac{1}{2}</math>. And the answer is yes ! | ||
In the preceding example, we computed the terms of row 4. Let's use that row again : | In the preceding example, we computed the terms of row 4. Let's use that row again : | ||
<math> | <math> | ||
Line 253: | Line 237: | ||
:<math> \scriptstyle \cdots ~</math> | :<math> \scriptstyle \cdots ~</math> | ||
What is the first term of row <math>\scriptstyle \frac{1}{2}</math>? By definition, it is 1. What are the terms after that ? We will use the row rule to compute them ! | What is the first term of row <math>\scriptstyle \frac{1}{2}</math>? By definition, it is 1. What are the terms after that ? We will use the row rule to compute them ! | ||
Line 269: | Line 252: | ||
No term goes to zero, even if each comes closer as it is further away from the first. | No term goes to zero, even if each comes closer as it is further away from the first. | ||
In the following | In the following Bernoulli triangle, using the row rule, we added the rows <math>\scriptstyle \frac{1}{2}</math> and <math>\scriptstyle \frac{3}{2}</math> : | ||
<math> | <math> | ||
Line 288: | Line 270: | ||
\end{array} | \end{array} | ||
</math> | </math> | ||
Some calculations confirm that the addition rule works, as expected. After some tests, we see that it works only if the row indices have a difference of 1. For instance, we cannot easily build row 3.25 from row 3, but we can easily build row <math>\scriptstyle \frac{3}{2}</math> from row <math>\scriptstyle \frac{1}{2}</math>. Up to a point, there are two matrices in the preceding Bernoulli triangle, one for the integers and one for the multiples of <math>\scriptstyle \frac{1}{2}</math>. | Some calculations confirm that the addition rule works, as expected. After some tests, we see that it works only if the row indices have a difference of 1. For instance, we cannot easily build row 3.25 from row 3, but we can easily build row <math>\scriptstyle \frac{3}{2}</math> from row <math>\scriptstyle \frac{1}{2}</math>. Up to a point, there are two matrices in the preceding Bernoulli triangle, one for the integers and one for the multiples of <math>\scriptstyle \frac{1}{2}</math>. | ||
Line 309: | Line 290: | ||
Suppose we replace the exponent 2, an [[integer]], by the exponent <math>\frac{1}{2}~</math>, a fraction. When the exponent value is 2, we get the square, when it is <math>\frac{1}{2}~</math>, we get the square root. This a rule of the [[exponentiation]]. Thus, we can use the triangle's terms to compute a square root. | Suppose we replace the exponent 2, an [[integer]], by the exponent <math>\frac{1}{2}~</math>, a fraction. When the exponent value is 2, we get the square, when it is <math>\frac{1}{2}~</math>, we get the square root. This a rule of the [[exponentiation]]. Thus, we can use the triangle's terms to compute a square root. | ||
This observation is technically correct, but there are some details that hugely simplify the calculations. If the exponent is a fraction, then <math>x~</math> SHOULD be equal to 1 and, in this case, <math>y~</math> MUST be equal to or less than 1. More | This observation is technically correct, but there are some details that hugely simplify the calculations. If the exponent is a fraction, then <math>x~</math> SHOULD be equal to 1 and, in this case, <math>y~</math> MUST be equal to or less than 1. More precisely, <math>|y| < 1~</math><ref>The reason of this limitation is outside the scope of this article, it has to do with [[series]] convergence.</ref>. | ||
Let's compute <math>(2+3)^{\frac{1}{2}}~</math>, i.e. <math>\sqrt{5}~</math>. | Let's compute <math>(2+3)^{\frac{1}{2}}~</math>, i.e. <math>\sqrt{5}~</math>. | ||
<math> | <math> | ||
Line 321: | Line 301: | ||
= 2 \, (1 + 0.25)^{\frac{1}{2}} | = 2 \, (1 + 0.25)^{\frac{1}{2}} | ||
</math> | </math> | ||
In the last parenthesis, <math>x~</math> is equal to 1, and <math>|y| = 0.25~</math> is lower than 1. Here is how we proceeded. Compute the perfect square lower than <math>(2 + 3)~</math>, here it is 4. Change the sum within the parenthesis in order to display the perfect square. Get it outside the parenthesis, while applying the square root operation on both terms. Extract the square root from both, one requiring to use the binomial coefficients. | In the last parenthesis, <math>x~</math> is equal to 1, and <math>|y| = 0.25~</math> is lower than 1. Here is how we proceeded. Compute the perfect square lower than <math>(2 + 3)~</math>, here it is 4. Change the sum within the parenthesis in order to display the perfect square. Get it outside the parenthesis, while applying the square root operation on both terms. Extract the square root from both, one requiring to use the binomial coefficients. | ||
To compute the root square of the parenthesis, we use the row <math>{\frac{1}{2}}~</math> in the triangle. | To compute the root square of the parenthesis, we use the row <math>{\frac{1}{2}}~</math> in the triangle. | ||
<math> | <math> | ||
(1 + 0.25)^{\frac{1}{2}} | (1 + 0.25)^{\frac{1}{2}} | ||
= 1 + \frac{1}{2} \times 0.25 | = 1 + \frac{1}{2} \times 0.25 | ||
- \frac{1}{8} \times 0.25 | - \frac{1}{8} \times 0.25^2+ | ||
\cdots | \cdots | ||
</math> | </math> | ||
Line 342: | Line 316: | ||
<math> | <math> | ||
(1 + 0.25)^{\frac{1}{2}} | (1 + 0.25)^{\frac{1}{2}} | ||
\approx 1 + \frac{1}{2} \times 0.25 | |||
- \frac{1}{8} | - \frac{1}{8} \times 0.25^2=1.1171875 | ||
</math> | </math> | ||
Using three terms, the square root estimate of <math>(2 + 3)~</math> is <math>2 * 1.1171875 = 2.234375</math>. The square of this value is <math>4.9924316\dots</math>. Compared to the exact value, the error of calculating square root is close to 0.1%. Using more terms will sharpen the result. | |||
Using | |||
We can compute any root with this method. However, it is not used in practice, since there are faster convergent methods, like the [[Newton-Raphson algorithm]]. | We can compute any root with this method. However, it is not used in practice, since there are faster convergent methods, like the [[Newton-Raphson algorithm]]. | ||
Line 373: | Line 339: | ||
:<math>(1+z)^{-1} = 1 - z + z^2 - z^3 + z^4 - z^5 + z^6 +\cdots~</math> | :<math>(1+z)^{-1} = 1 - z + z^2 - z^3 + z^4 - z^5 + z^6 +\cdots~</math> | ||
If we [[integration|integrate]] both sides of the previous equation<ref> | If we [[integration|integrate]] both sides of the previous equation<ref>From [[calculus]], we know that : | ||
From [[calculus]], we know that : | |||
:<math> \ | :<math> \scriptstyle \cdots~</math> | ||
:<math> \scriptstyle \int{z^1 dz}= \frac{z^2}{2} + C,\,\,C\,\hbox{is an arbitrary constant}~</math> | |||
:<math> \int{z^ | :<math> \scriptstyle \int{z^0 dz}= \frac{z^1}{1} + C~</math> | ||
:<math> \scriptstyle \int{z^{-1} dz} = \int{\frac{dz}{z} } = \ln{z} + C~</math> | |||
:<math> \int{z^{- | :<math> \scriptstyle \int{z^{-2} dz} = \frac{z^{-1}}{-1} + C~</math> | ||
:<math> \scriptstyle \int{z^{-3} dz} = \frac{z^{-2}}{-2} + C ~</math> | |||
:<math> \cdots~</math></ref>, we get the natural logarithm of a number<ref>William H Beyer (ed.), ''Standard Mathematical Tables and Formulae, 29th edition'', CRC Press, p.279. ISBN 0-8493-0629-9</ref> : | :<math> \scriptstyle \cdots~</math></ref>, we get the natural logarithm of a number<ref>William H Beyer (ed.), ''Standard Mathematical Tables and Formulae, 29th edition'', CRC Press, p.279. ISBN 0-8493-0629-9</ref> : | ||
:<math>\ln{(1+z)} = z - \frac{z^2}{2} + \frac{z^3}{3} - \frac{z^4}{4} + \frac{z^5}{5} - \frac{z^6}{6} + \frac{z^7}{7} +\cdots \text{ with } z \in \R \text{ and } -1 < z \leq 1 ~</math> | :<math>\ln{(1+z)} = z - \frac{z^2}{2} + \frac{z^3}{3} - \frac{z^4}{4} + \frac{z^5}{5} - \frac{z^6}{6} + \frac{z^7}{7} +\cdots \text{ with } z \in \R \text{ and } -1 < z \leq 1 ~</math> | ||
Line 454: | Line 413: | ||
\end{array} \right] </math> | \end{array} \right] </math> | ||
To get bigger figurate number triangles, use bigger square matrices, filling the subdiagonal under the main diagonal with <math>\scriptstyle 1, 2, 3, \cdots</math>, the rest is set to zero. | |||
In [[combinatorics]], the Pascal's triangle is simply defined by <math>\scriptstyle \binom{n}{k} = \frac{n!}{k!(n-k)!}</math> with <math>\scriptstyle k, n \in \N</math>. | In [[combinatorics]], the Pascal's triangle is simply defined by <math>\scriptstyle \binom{n}{k} = \frac{n!}{k!(n-k)!}</math> with <math>\scriptstyle k, n \in \N</math>. This compact equation provides the basis for many identities. However, it is not easy to manipulate through [[function]]s. The matrix exponential opens new territories, since it gives a new way to manipulate the triangle using results from the [[matrix theory]] and the function theory. For instance, is the triangle a solution of <math> \scriptstyle x^2 + 3x - 4I = 0 </math> ? What is its inverse ? What is its logarithm ? | ||
== Normal Curve == | == Normal Curve == | ||
Pascal's triangle lends itself to graphical representation. If you draw an histogram of the | Pascal's triangle lends itself to graphical representation. If you draw an histogram of the 19th row, you would get something like the following picture. | ||
{{Image|Pascal's Triangle, Histogram of binomial coefficients at row 19.png|center|300px|Histogram of the 19th row within the Pascal's triangle}} | |||
Suppose we compute the terms of the | Suppose we compute the terms of the 1000th row in Pascal's triangle. If we divide each term by 2<sup>1000</sup> and consider each value as a point on a ''continuous'' curve, we would get the following curve. | ||
[[Image:Pascal's Triangle, Normal Curve.png|center|thumb|300px|The [[Normal distribution|normal curve]], a limit case of any high-order row within the Pascal's Triangle.]] | [[Image:Pascal's Triangle, Normal Curve.png|center|thumb|300px|The [[Normal distribution|normal curve]], a limit case of any high-order row within the Pascal's Triangle.]] | ||
[[Carl Friedrich Gauss|Gauss]], the famous German mathematician, studied the [[Normal distribution|normal curve]]<ref name="boursin">{{fr}} Jean-Louis Boursin, ''Les Structures du hasard'', Seuil, coll. Les Rayons de la science, 1966, p.129.</ref>. It is the commonest [[statistics|statistical]] law encountered in natural phenomenons<ref name="boursin"/> | [[Carl Friedrich Gauss|Gauss]], the famous German mathematician, studied the [[Normal distribution|normal curve]]<ref name="boursin">{{fr}} Jean-Louis Boursin, ''Les Structures du hasard'', Seuil, coll. Les Rayons de la science, 1966, p.129.</ref>. It is the commonest [[statistics|statistical]] law encountered in natural phenomenons<ref name="boursin"/><ref>Paul Westbrook, ''Math Smart for Business: Cultivating a Six-Figure Vocabulary'', Princeton Review, 1997, p. 70. ISBN 0679783911. Consulted 2007-11-19. View an extract at [http://books.google.com/books?id=viAvQxqSMPAC&pg=PA70&lpg=PA70&dq=%22normal+curve%22+statistics+%22most+common+curve%22&source=web&ots=3-VbF7Y-KP&sig=HEJAwB8Ct5KTb1aWG8usApJU5Go].</ref>. For this reason, the properties of many statistical curves are compared to the normal curve properties. In its simplest form, its equation is <math>\scriptstyle P(z) = e^{-z^2/2}</math>. | ||
== Identities == | |||
We already presented some useful equations for building the Pascal's triangle. Since they are true within a [[field]] (a commutative [[Ring (mathematics)|ring]]), they are called [[identity|identities]] in mathematics. Below, we list some more, but be aware there are literaly thousands of them. | |||
'''1.''' | |||
For <math> \scriptstyle x, y \in \R</math> and <math> \scriptstyle n \in \N </math>, | |||
<math> | |||
(x+y)^n = \sum_{k=0}^n \binom{n}{k} x^k y^{n-k} | |||
</math> | |||
with | |||
<math> \binom{n}{k} = \frac{n!}{k!(n-k)!} </math> | |||
'''2.''' | |||
For <math> \scriptstyle n \in \N</math>, | |||
<math> | |||
\sum_{k=0}^n \binom{n}{k} = 2^n | |||
</math> | |||
'''3.''' | |||
For <math> \scriptstyle n \in \N</math>, | |||
<math> | |||
\sum_{k=0}^n (-1)^k \binom{n}{k} = 0 | |||
</math> | |||
'''4.''' | |||
For <math> \scriptstyle n \in \N</math>, | |||
<math> | |||
\sum_{k=0}^n 2^k \binom{n}{k} = 3^n | |||
</math> | |||
'''5.''' | |||
(''Pascal's identity'') For <math> \scriptstyle n \in \N</math>, | |||
<math> | |||
\binom{n + 1}{k} = \binom{n}{k - 1} + \binom{n}{k} | |||
</math> | |||
'''6.''' | |||
(''Vandermonde's identity'') For <math> \scriptstyle m, n, r \in \N^+</math>, <math> \scriptstyle r \le m</math>, <math> \scriptstyle r \le n</math>, | |||
<math> | |||
\binom{m + n}{r} = \sum_{k=0}^r \binom{m}{r - k} \binom{n}{k} | |||
</math> | |||
'''7.''' | |||
For <math> \scriptstyle n \in \N</math>, | |||
<math> | |||
\binom{2n}{n} = \sum_{k=0}^n \binom{n}{k}^2 | |||
</math> | |||
'''8.''' | |||
For <math> \scriptstyle n, r \in \N^+</math>, <math> \scriptstyle r \le n</math>, | |||
<math> | |||
\binom{n + 1}{r + 1} = \sum_{j=r}^n \binom{j}{r} | |||
</math> | |||
The previous identities remain true if the variables are complex (e.g., <math> \scriptstyle (3 - 7 i)</math>). Some identities relating to Pascal's triangle are true when <math> \scriptstyle m</math>, <math> \scriptstyle n</math>, and <math> \scriptstyle r</math> are fractional. | |||
'''9.''' ('' Newton's identity'') For <math> \scriptstyle x, y \in \R</math> and <math> \scriptstyle r \in \R</math>, | |||
<math> | |||
(x+y)^r=\sum_{k=0}^\infty \binom{n}{k} x^k y^{r-k} | |||
</math> | |||
with | |||
<math> | |||
\binom{n}{k} = \frac{1}{k!} \prod_{i=0}^{n-1}(n-k) | |||
= \frac{n(n-1)\cdots(n-k+1)}{k!} | |||
</math> | |||
When <math> \scriptstyle r</math> is an integer, we fall back on the first identity. The number of terms is still infinite, but almost all of them are equal to zero (those in the tail). | |||
== Unexpected Relations == | == Unexpected Relations == | ||
; Prime number multiples | |||
Ones excepted, all the terms in any row where the second term is a [[prime number]] are multiple of the prime number. For instance, if a row starts with <math>\scriptstyle 1, 23, 253, \cdots </math>, then all its terms, ones excepted, are multiple of 23.<ref>See [http://binomial.csuhayward.edu/Identities.html#5.1 ''Pascal's Triangle From Top to Bottom''], California State University, East Bay. Consulted 2007-11-28.</ref> | |||
; Odd numbers within a row | ; Odd numbers within a row | ||
Each row of order <math>\scriptstyle 2^n-1, n \in \N</math>, only contains odd numbers. No other row has this property. | Each row of order <math>\scriptstyle 2^n-1, n \in \N</math>, only contains odd numbers. No other row has this property. | ||
Line 549: | Line 589: | ||
</center> | </center> | ||
Dividing the underlined numbers by 1, 2, 3, 4,..., we get 1, 1, 2, 5, 14,... The Catalan numbers appear in solutions of many combinatorial problems.<ref>Eric W. Weisstein, ''CRC Concise Encyclopedia of Mathematics'', CRC Press, 1999, p. 200. ISBN 0-8493-9640-9</ref> There is another way to compute these numbers : in the triangle, | Dividing the underlined numbers by 1, 2, 3, 4,..., we get 1, 1, 2, 5, 14,... The Catalan numbers appear in solutions of many combinatorial problems.<ref>Eric W. Weisstein, ''CRC Concise Encyclopedia of Mathematics'', CRC Press, 1999, p. 200. ISBN 0-8493-9640-9</ref> There is another way to compute these numbers : in the triangle, subtract the neighbour from each underlined number. | ||
There is a surprising way to compute them. Square the terms in each row and add them : | There is a surprising way to compute them. Square the terms in each row and add them : | ||
Line 564: | Line 604: | ||
<math> \scriptstyle ... </math> | <math> \scriptstyle ... </math> | ||
These are the underlined values in the previous triangle. | |||
; Geometric counting | ; Geometric counting | ||
Line 600: | Line 642: | ||
== References == | == References == | ||
{{reflist|2}}[[Category:Suggestion Bot Tag]] |
Latest revision as of 16:01, 1 October 2024
Pascal's triangle is a convenient tabular representation of the binomial coefficients. Already known in the 11th century,[1] it was adopted in the Western world under this name after Blaise Pascal published his Traité du triangle arithmétique ("Treatise on the Arithmetical Triangle") in 1654.
Pascal's triangle appears under different formats. Here is its most common:
We can use Pascal's triangle to compute the binomial expansion of . For instance,
The triangle shows the coefficients on the fifth row.
Pascal's triangle has applications in algebra and in probabilities. We can use it to compute the Fibonacci numbers and to create the Sierpinski triangle. After studying it, Isaac Newton expanded the triangle and found new methods to extract the square root and to calculate the natural logarithm of a number.
History
The earliest explicit depictions of a triangle of binomial coefficients occur in the 10th century in commentaries on the Chandas Shastra, an ancient Indian book on Sanskrit prosody written by Pingala between the 5th century BC and 2nd century BC. While Pingala's work only survives in fragments, the commentator Halayudha, around 975, used the triangle to explain obscure references to Meru-prastaara, the "Staircase of Mount Meru". It was also realised that the shallow diagonals of the triangle sum to the Fibonacci numbers. The Indian mathematician Bhattotpala (c. 1068) later gives rows 0-16 of the triangle.
At around the same time, it was discussed in Persia by the mathematician Al-Karaji (953–1029) and the poet-astronomer-mathematician Omar Khayyám (1048-1131); thus the triangle is referred to as the "Khayyam triangle" in Iran. Several theorems related to the triangle were known, including the binomial theorem. It seems that Khayyam used a method of finding nth roots based on the binomial expansion, and therefore on the binomial coefficients.
In 13th century, Yang Hui (1238-1298) presented an arithmetic triangle, which was the same as Pascal's Triangle. Today, Pascal's triangle is called "Yang Hui's triangle" in China. In Italy, it is known as "Tartaglia's triangle", named for the Italian algebraist Niccolo Fontana Tartaglia who lived a century before Pascal.
In 1655, Blaise Pascal published its Traité du triangle arithmétique ("Treatise on arithmetical triangle"), wherein he collected several results then known about the triangle, and employed them to solve problems in probabilities. The triangle was later named after Pascal by Pierre Raymond de Montmort (1708) and Abraham de Moivre (1730).
After that, even if it was useful in many areas of mathematics, most research was done within its descendants, like probabilities and combinatorics.
Some Properties
Each term in the triangle is the sum of the two terms above it[2]. For instance, . The binomial coefficients relate to this construction by Pascal's rule, which states that if
is the kth binomial coefficient in the binomial expansion of , then
By convention, the binomial coefficient is set to zero if is either less than zero or greater than .
To better understand some properties, the triangle is presented differently :
Each coefficient is the sum of the coefficient exactly over it and its left neighbour[2]. For instance, . Let's call this rule the "addition rule".
Using this format, it is easy to apply an index to each row :
Starting the indices at zero facilitates many calculations.
The sum of any row is , with being the row index : For instance, the sum of row 4 is .
Since there is a formula for summing a row, then maybe there is one for a column ? It is the case. This time, we index the columns :
Anyone familiar with the factorial function can easily find the general formula. The sum of a column ending at row is
- .
There is another method to compute this sum, see [3].
Up until now, we added along the rows and the columns. We can add along the diagonals. Doing so from left to right and from bottom to top gives :
The numbers on the right are the Fibonacci numbers.
One Row at a Time
We can build any row if we know the row before just by adding its terms two at a time. However, it is possible to build a row directly. We will build the row 4 in order to discover the rule. Each row starts with 1 :
Once the row and the column indices are know, we can compute the neighbours, either right or left, of any term. Because it only applies to a row, we call it the "row rule".
Polynomials with coefficients in triangle
Newton's Binomial Coefficients
Isaac Newton studied the triangle's properties and discovered two remarkable generalizations.[5]
He extended the triangle along the negative axis. To better understand how he achieved this, let us write the triangle using another format :
Jacob Bernoulli is the "father" of this mathematical object named the "figurate number triangle"[6]. To ease the understanding in the following text, we will call it the "Bernoulli triangle", even if this matrix does not resemble a triangle anymore !
Using the row rule, let's compute the row -1 :
There is an infinity of and . This is another hint that this object is a matrix. Each term in this row obeys the addition rule and the row rule. Some calculations should convince you.
We can further extend the triangle along the negative axis.
As we wrote earlier, Newton did not stop there. He asked himself if there were rows with fractional index, like . And the answer is yes !
In the preceding example, we computed the terms of row 4. Let's use that row again :
What is the first term of row ? By definition, it is 1. What are the terms after that ? We will use the row rule to compute them !
No term goes to zero, even if each comes closer as it is further away from the first.
In the following Bernoulli triangle, using the row rule, we added the rows and :
Some calculations confirm that the addition rule works, as expected. After some tests, we see that it works only if the row indices have a difference of 1. For instance, we cannot easily build row 3.25 from row 3, but we can easily build row from row . Up to a point, there are two matrices in the preceding Bernoulli triangle, one for the integers and one for the multiples of .
Newton's generalizations opened routes to areas not primarily connected to integers, namely numeric roots (like the square roots) and logarithms.
Computing a Root Square
Armed with the knowledge of the previous section, we are ready to compute the square root of a number.
We start with this equation :
- ,
If we replace with 2 and with 3, then we get
- ,
which is equal to 25, or . This is quite common knowledge, not to say boring.
Suppose we replace the exponent 2, an integer, by the exponent , a fraction. When the exponent value is 2, we get the square, when it is , we get the square root. This a rule of the exponentiation. Thus, we can use the triangle's terms to compute a square root.
This observation is technically correct, but there are some details that hugely simplify the calculations. If the exponent is a fraction, then SHOULD be equal to 1 and, in this case, MUST be equal to or less than 1. More precisely, [7].
Let's compute , i.e. .
In the last parenthesis, is equal to 1, and is lower than 1. Here is how we proceeded. Compute the perfect square lower than , here it is 4. Change the sum within the parenthesis in order to display the perfect square. Get it outside the parenthesis, while applying the square root operation on both terms. Extract the square root from both, one requiring to use the binomial coefficients.
To compute the root square of the parenthesis, we use the row in the triangle.
Using three terms, the square root estimate of is . The square of this value is . Compared to the exact value, the error of calculating square root is close to 0.1%. Using more terms will sharpen the result.
We can compute any root with this method. However, it is not used in practice, since there are faster convergent methods, like the Newton-Raphson algorithm.
Computing the Logarithm of a Number
Starting with this equation :
- ,
we replace some terms within it :
- ,
which is good for the square of . To compute the logarithm, we use the terms in row -1.
If we integrate both sides of the previous equation[8], we get the natural logarithm of a number[9] :
Matrix Exponential
At the beginning of the 21st century, someone found a remarkable equality between the Pascal's triangle and a power series of square matrices.[10]
As you probably know, we can compute the exponential function like this : Moreover, maybe you remember that we can multiply a square matrix of rank with any other square matrix of the same rank. Combining these two facts, we get
We are ready to compute the figurate number triangle of order 3 :
To get bigger figurate number triangles, use bigger square matrices, filling the subdiagonal under the main diagonal with , the rest is set to zero.
In combinatorics, the Pascal's triangle is simply defined by with . This compact equation provides the basis for many identities. However, it is not easy to manipulate through functions. The matrix exponential opens new territories, since it gives a new way to manipulate the triangle using results from the matrix theory and the function theory. For instance, is the triangle a solution of ? What is its inverse ? What is its logarithm ?
Normal Curve
Pascal's triangle lends itself to graphical representation. If you draw an histogram of the 19th row, you would get something like the following picture.
Suppose we compute the terms of the 1000th row in Pascal's triangle. If we divide each term by 21000 and consider each value as a point on a continuous curve, we would get the following curve.
Gauss, the famous German mathematician, studied the normal curve[11]. It is the commonest statistical law encountered in natural phenomenons[11][12]. For this reason, the properties of many statistical curves are compared to the normal curve properties. In its simplest form, its equation is .
Identities
We already presented some useful equations for building the Pascal's triangle. Since they are true within a field (a commutative ring), they are called identities in mathematics. Below, we list some more, but be aware there are literaly thousands of them.
1. For and ,
with
2. For ,
3. For ,
4. For ,
5. (Pascal's identity) For ,
6. (Vandermonde's identity) For , , ,
7. For ,
8. For , ,
The previous identities remain true if the variables are complex (e.g., ). Some identities relating to Pascal's triangle are true when , , and are fractional.
9. ( Newton's identity) For and ,
with
When is an integer, we fall back on the first identity. The number of terms is still infinite, but almost all of them are equal to zero (those in the tail).
Unexpected Relations
- Prime number multiples
Ones excepted, all the terms in any row where the second term is a prime number are multiple of the prime number. For instance, if a row starts with , then all its terms, ones excepted, are multiple of 23.[13]
- Odd numbers within a row
Each row of order , only contains odd numbers. No other row has this property.
- and
We saw the normal curve, a way to represent the binomial coefficients as a continuous curve. In its simplest form, its equation and the area under that curve is . The equation of this curve involves , while its area involves . How can this be possible, since the binomial coefficients are obtained by adding natural numbers, and we divided those coefficients by a power of 2 ?
- Constructible polygons
Applying modulo 2 to every number in Pascal's triangle, we get
1
1 1
1 0 1
1 1 1 1
1 0 0 0 1
1 1 0 0 1 1
1 0 1 0 1 0 1
...
Viewing these ones and zeroes as binary digits, we get 1, 3, 5, 15, 17, 51,... in base ten. They are the number of sides found in constructible polygons with odd number of sides. This is true up to row 32.[14]
- Computing
We can compute from the Pascal's triangle : . For instance, outputs . This formula is not used in practice since must be very large to compute to any decent precision.
- Stacking marbles
If you stack marbles in the simplest 1D figure (along a line), then you arrange 1, 2, 3, 4,... marbles. If you stack marbles in the simplest 2D figure (triangle), then you arrange 1, 3, 6, 10,... marbles. If you stack marbles in the simplest 3D figure (a tetrahedron or 4-sided pyramid), then you arrange 1, 4, 10, 20,... marbles. If you stack marbles in the simplest 4D figure (a 4-simplex), then you arrange 1, 5, 15, 35,... marbles. And so on.
These numbers appear within the columns of the figurate number triangle, the column 1 contains the classic and basic counting numbers : 1, 2, 3, 4,... The column 2 contains the triangular numbers : 1, 3, 6, 10, 15, 21,... The column 3 contains the tetrahedral numbers : 1, 4, 10, 20, 35, 56,... The column 4 contains the pentatope numbers : 1, 5, 15, 35, 70, 126,... And so on.
- Catalan numbers
The Pascal's triange hides the Catalan numbers.
1
1 1
1 2 1
1 3 3 1
1 4 6 4 1
1 5 10 10 5 1
1 6 15 20 15 6 1
1 7 21 35 35 21 7 1
1 8 28 56 70 56 28 8 1
...
Dividing the underlined numbers by 1, 2, 3, 4,..., we get 1, 1, 2, 5, 14,... The Catalan numbers appear in solutions of many combinatorial problems.[15] There is another way to compute these numbers : in the triangle, subtract the neighbour from each underlined number.
There is a surprising way to compute them. Square the terms in each row and add them :
These are the underlined values in the previous triangle.
- Geometric counting
Somebody found a strange property relating geometry to the numbers within the Pascal's triangle.
A dot has one vertex (1). A segment has two vertices and one line (2,1). A triangle has three vertices, three edges and one surface (3, 3, 1). A triangular pyramid has four vertices, six edges, four surfaces and one volume (4, 6, 4, 1). And so on. Thus, the element count of a minimal regular geometric object (build from regular shapes) in 1D, 2D, 3D, 4D, etc., is specified by a row in the Pascal's triangle.
- Bernoulli numbers
Bernoulli numbers, named in honor of Jacob Bernoulli who studied them intensively, arise in a wide variety of analytical and combinatorial contexts. They are[16]
Knowing that and are the two first terms of this series, we can compute all the terms :
, then
, then
, then
, then
, then
, then
is computed twice, but this arrangement clearly shows the binomial coefficients.[17]
References
- ↑ Abu Bekr ibn Muhammad ibn al-Husayn Al-Karaji, School of Mathematics and Statistics, University of St Andrews. Consulted 2005-09-03.
- ↑ 2.0 2.1 This rule does not apply to the ones bordering the triangle. We just insert them.
- ↑ Suppose we wish to add the terms in row 3, i.e. the fourth column, until row 6. The sum is given by multiplying four terms at numerator, starting at , and four terms at the denominator starting at . It is equal to . In short, fourth column, four terms at numerator, four terms at denominator, all decreasing.
- ↑ Equally, we can compute any triangle's term using , but it may exceed the calculator capacity !
- ↑ Eli Maor, e: The Story of a Number, Princeton University Press, 1994, p.71. ISBN 0-691-05854-7.
- ↑ Eric W. Weisstein, CRC Concise Encyclopedia of Mathematics, CRC Press, 1999, p. 636. ISBN 0-8493-9640-9
- ↑ The reason of this limitation is outside the scope of this article, it has to do with series convergence.
- ↑ From calculus, we know that :
- ↑ William H Beyer (ed.), Standard Mathematical Tables and Formulae, 29th edition, CRC Press, p.279. ISBN 0-8493-0629-9
- ↑ For instance, see Gottfried Helms, "Some Properties of the Binomial-("Pascal"-)-triangular matrix", 2005-08-31, version 3.
- ↑ 11.0 11.1 Template:Fr Jean-Louis Boursin, Les Structures du hasard, Seuil, coll. Les Rayons de la science, 1966, p.129.
- ↑ Paul Westbrook, Math Smart for Business: Cultivating a Six-Figure Vocabulary, Princeton Review, 1997, p. 70. ISBN 0679783911. Consulted 2007-11-19. View an extract at [1].
- ↑ See Pascal's Triangle From Top to Bottom, California State University, East Bay. Consulted 2007-11-28.
- ↑ Eric W. Weisstein, CRC Concise Encyclopedia of Mathematics, CRC Press, 1999, p. 1399. ISBN 0-8493-9640-9
- ↑ Eric W. Weisstein, CRC Concise Encyclopedia of Mathematics, CRC Press, 1999, p. 200. ISBN 0-8493-9640-9
- ↑ William H Beyer (ed.), Standard Mathematical Tables and Formulae, 29th edition, CRC Press, p.394. ISBN 0-8493-0629-9
- ↑ Eric W. Weisstein, CRC Concise Encyclopedia of Mathematics, CRC Press, 1999, p. 112. ISBN 0-8493-9640-9