Extended bra-ket notation

Extended Bra-Ket notation, or EBK for short, is a potential name for the notation system for regular temperament theory (RTT) objects that has gradually developed over time, with contributions from various theoreticians.

Several stylistic variations are possible, and no formal style has yet been established, but it can be safely said that EBK always involves enclosing lists of values in sets of brackets, with pointed brackets used to distinguish different types of lists.

Per the name, EBK extends bra-ket notation, which is used in quantum mechanics. The use of bra-ket notation for RTT was originally proposed by Gene Ward Smith in February 2002.

Basics

The name "bra-ket" notation refers to its two basic structures: the bra and the ket. A bra is a list which has a pointed bracket on its left side, and a ket is a list which has a pointed bracket on its right side. Bras are used for covectors, and kets are used for vectors (for more information on the difference between covectors and vectors, see variance).

The usage patterns described in this section are the same as the standard usage of bra-ket notation in quantum mechanics.

Common applications

In RTT, covectors are used most frequently for maps and tuning maps. For example, the map for 7-ET is notated as ⟨7 11 16], and a tuning map for it might be ⟨1209.682 1900.930 2764.988].

The most common types of vectors used in RTT are prime-count vectors (PC-vectors) and generator-count vectors (GC-vectors). For example, the PC-vector for 45/32 is [-5 2 1⟩, and the GC-vector for ~45/32 in porcupine temperament is [2 -11⟩.

Combining

When a bra and a ket are put together, the dot product is taken, and the result is a scalar (a single number). In order to be compatible in this way, the bra and ket must at least have the same number of entries, so that each of their entries will pair up for a product.

So mapping 45/32 with 7-ET would be notated as ⟨7 11 16][-5 2 1⟩ = 7·-5 + 11·2 + 16·1 = -35 + 22 + 16 = 3.

Or mapping the GC-vector for porcupine's ~45/32 with a generator tuning map like ⟨1200 162.737] would look like ⟨1200 162.737][2 -11⟩ = 1200·2 + 162.737·-11 = 2400 + -1790.107 = 609.893.

Extensions

EBK extends standard bra-ket notation using two different kinds of nesting:

1. alternating, to allow for notating matrices, and
2. repeating, to allow for notating multivectors.

Alternating, for matrices

Vectors (including covectors) are used to represent musical objects of various dimensionality. For example, the PC-vector [1 -2 1⟩ represents the interval 10/9, which would be found as a point in the three-dimensional space of 5-limit JI lattice, while the PC-vector [0 -1 1 1 -1⟩ represents the interval 35/33, which would be found as a point in the five-dimensional space of the 11-limit JI lattice. However, regardless of the dimensionality of the musical object represented, a vector itself will always be a one-dimensional structure, in the sense that it is a simple list of numbers. Due to this, vectors are always fairly easy to embed in similarly one-dimensional strings of text or data cells of tables.

But RTT uses a number of two-dimensional structures as well — i.e. numbers arranged in a grid of rows and columns — which are called matrices. Having the ability to present these objects one-dimensionally can be quite helpful too, and so the first way in which EBK extends bra-ket notation is designed to provide that.

Advantage

One catch is that some matrices in RTT are more naturally thought of in terms of their rows, while other matrices are more naturally thought of in terms of their columns. EBK offers a good solution to this problem, though: it allows us to notate matrices either in terms of their rows or in terms their columns.

EBK is capable of doing this because of how vectors and covectors are interpreted as matrices. Specifically, a vector can be thought of as a [math](x,1)[/math]-shaped matrix (many rows, one column), while a covector can be thought of as a [math](1,x)[/math]-shaped matrix (one row, many columns).

So, one way to think of a matrix would be as a single vector column which contains a number of covector rows. Another way to think of a matrix would be as a single covector row which contains a number of vector columns. Either way is possible, but in most cases, one of these two ways will be easier to understand.

Examples

For example, the mapping for meantone temperament is a matrix [math]M[/math] that looks like this:

[math] \left[ \begin{array} {rrr} 1 & 0 & -4 \\ 0 & 1 & 4 \\ \end{array} \right] [/math]

RTT mappings are typically thought of in terms of their rows. This mapping [math]M[/math] has two rows; following mathematical conventions, let's call them [math]𝒎_1[/math] and [math]𝒎_2[/math]. And so to notate this mapping in EBK, we can first imagine capturing the rows as bras like we would normally: [math]𝒎_1[/math] = ⟨1 0 -4] and [math]𝒎_2[/math] = ⟨0 1 4]. Then, to put them together, we can think of this matrix as a single column containing these two rows, or in other words, a ket containing the two bras: [[math]𝒎_1[/math] [math]𝒎_2[/math]⟩, or fully written out, [⟨1 0 -4] ⟨0 1 4]⟩.

For another example, the canonical comma basis for 7-ET consists of the two commas 2187/2048 and 135/128, with PC-vectors [-11 7⟩ and [-7 3 1⟩, respectively. As a matrix [math]\mathrm{C}[/math], we'd see this as:

[math] \left[ \begin{array} {rrr} -11 & -7 \\ 7 & 3 \\ 0 & 1 \\ \end{array} \right] [/math]

RTT comma bases, as opposed to mappings, are more naturally thought of in terms of their columns. This comma basis has two columns; let's call them [math]\textbf{c}_1[/math] and [math]\textbf{c}_2[/math] (in this case, that conveniently works out as both 'c' for "column" and 'c' for "comma"!). And so to notate this comma basis in EBK, we can first imagine capturing the columns as kets like we would normally: [math]\textbf{c}_1[/math] = [-11 7 0⟩ and [math]\textbf{c}_2[/math] = [-7 3 1⟩. Then, to put them together, we can think of this matrix as a single row containing these two columns, or in other words, a bra containing the two kets: ⟨[math]\textbf{c}_1[/math] [math]\textbf{c}_2[/math]], or fully written out, ⟨[-11 7 0⟩ [-7 3 1⟩].

History

The alternating-for-matrices extension was contributed by Graham Breed, dating at least from March 2008 when he used it in his paper "Provably Complete Rank 2 Temperament Searches".

Usage

The alternating-for-matrices extension is used in Graham Breed's writings and web app, as well as the writings on the wiki of Douglas Blumeyer and Dave Keenan.

Alternatives

In many wiki writings, mappings and comma bases are provided as lists of vectors, notated using square brackets on both sides and commas between entries, like this: [a, b, c...]. So meantone mapping would look like [⟨1 0 -4], ⟨0 1 4]], and a comma basis for 7-ET would look like [[-11 7 0⟩, [-7 3 1⟩]. This notation is completely sufficient and unambiguous, but — for better or worse — does not emphasize the matrix-like structure of the data quite as strongly.

Repetition, for multivectors

The second extension of EBK from standard bra-ket notation is the repetition of brackets, allowing for the representation of multivectors.

Some advanced practitioners of RTT use multivectors, such as wedgies, as an alternative way to represent temperaments (besides mappings and comma bases). These multivectors come in various grades, such as 2-vectors and 3-vectors. In fact, ordinary vectors are simply 1-vectors. In order to distinguish a [math]g[/math]-vector from a 1-vector, the brackets that would normally be used can be repeated [math]g[/math] times, where [math]g[/math] is the grade.

For example, the 2-vector (bivector) representing meantone temperament uses two sets of brackets: [[1 4 4⟩⟩. The 3-covector representing 7-limit 31-ET ⟨⟨⟨-87 72 -49 31]]].

The repetition-for-multivectors extension was developed in November of 2003 by Dave Keenan.

For more information, see Douglas Blumeyer and Dave Keenan's Intro to exterior algebra for RTT#From vectors to multivectors.

Stylistic variations

Greater than and less than signs versus angle brackets

Standard bra-ket notation uses angle brackets:

symbol	Unicode	description
⟨	U27E8	MATHEMATICAL LEFT ANGLE BRACKET
⟩	U27E9	MATHEMATICAL RIGHT ANGLE BRACKET

Graham Breed's RTT web app uses these, as do his tuning-related papers, and the templates on this wiki itself (as of 2020 per Dave Keenan and Douglas Blumeyer's suggestion). In ASCII-only environments, the greater-than and less-than signs (available on standard keyboards) are a reasonable substitute, and these are used in many pages on this wiki that haven't been updated to the templates, and in Paul Erlich's Middle Path paper.

Square brackets versus pipes

Standard bra-ket notation uses pipes on the opposite side of bras and kets from the angle brackets. Graham's papers do this too. His temperament finder, though, uses square brackets. Paul's Middle Path paper uses pipes when bras and kets are combined, but square brackets when they are separate, which is a reasonable style that was discussed on the Yahoo Groups tuning lists.

The motivation for the square bracket deviation from standard bra-ket is largely for visual clarity, due to JI and RTT's differing needs from quantum mechanics. With quantum mechanics, the values typically enclosed by bra-ket notation are variables, which are easily distinguished from the simple vertical line appearance of the pipe. But in RTT we frequently represent actual numerals in the EBK notation, and the numeral 1 (the most common numeral) looks very similar to the pipe. So the square brackets are easier to distinguish at a glance.

Commas

Originally, it was common for entries of bras and kets to be separated by (punctuation) commas, like [-4, 4, 1⟩. Nowadays, this is almost never seen.

Separated sections

A method for visually breaking up the entries of longer bras and kets was proposed by George Secor in November of 2003. The proposal was for a comma to be placed after the entry for prime 3, and thereafter a comma is placed after every third prime. So for example, 13/8 would be given as [-3 0, 0 0 0, 1⟩.

A further variation is to drop sequences of three zeros thus enclosed by commas, so that 13/8 would appear as [-3 0, , 1⟩.

The stylistic reasoning is that it resembles the use of commas in large numbers to separate every third digit (e.g. 1,000,000). The first comma being after prime 3 rather than after the third prime (prime 5) is due to the notational specialness of primes 2 and 3, those being the primes capable of being encoded by standard sheet music (prime 2 by clefs and staff positions and prime 3 by Pythagorean nominals A,B,C,D,E,F,G and sharp and flat symbols).

Variant including curly and square brackets

Dave Keenan and Douglas Blumeyer propose that it may be helpful to distinguish objects with distinctive shapes, such as generator tuning maps and generator-count vectors, by using curly brackets {...] [...} in place of angle brackets ⟨...] [..⟩, wherever the height or width of a vector or matrix is equal to the rank of the temperament, [math]r[/math]. A mnemonic for the curly bracket is that it resembles the tilde (~) which is commonly used to mark approximated or tempered intervals, e.g. ~3/2 is an approximation of 3/2.

For example, while the PC-vector representing 5/4 would be written [-2 0 1⟩, the mapped version of this in meantone could be written [-2 4}. And while the tuning map for quarter-comma meantone might be written ⟨1200.000 1896.578 2786.314], the generator tuning map could be written {1200.000 696.578].

We further propose that the use of the normal angle bracket could be restricted to matrix widths and heights equal only to the dimensionality of the temperament, [math]d[/math], and any other width or height besides [math]d[/math] and [math]r[/math] would be given with plain square brackets [...]. So, for example, a comma basis could be written [[4 -4 1⟩ [7 0 -3⟩] because its width is equal to the nullity of the temperament, [math]n[/math]. This is consistent with the fact that it is common for linear algebra texts to treat a nullspace basis not as a matrix but as a mere list of vectors.

Instead of curly brackets, we considered using round brackets (parentheses), (...] [...), however it is sometimes necessary to include parenthesized expressions as entries in the bra or ket, which would be difficult to parse if we had round bra or ket brackets.

We also considered using curved angle brackets ⧼...] [...⧽, (Unicode U+29FC and U+29FD), however in some fonts these are almost indistinguishable from ordinary angle brackets. If you see distinctly-curved angle brackets in the preceding sentence, it is only because we specified that they be shown in the font 'Cambria Math' (Windows) or 'STIX Two Math' (Mac). Here's an ordinary angle bracket followed by a curved angle bracket without special treatment: ⟩⧽. If this situation should change in future, a mnemonic for the curved angle brackets is that they can also be called "rounded" angle brackets, and a rounded number is an approximate number in the same way that a tempered interval is an approximation of a just interval. Curly brackets would still be used as a substitute in ASCII-only environments.