PETAL PROJECTIONS, KNOT COLORINGS & DETERMINANTS

ALLISON HENRICH AND ROBIN TRUAX

Abstract. An übercrossing diagram is a knot diagram with only one crossing that may involve more than
arXiv:2004.00148v2 [math.GT] 13 Aug 2021

two strands of the knot. Such a diagram without any nested loops is called a petal projection. Every knot has
a petal projection from which the knot can be recovered using a permutation that represents strand heights.
Using this permutation, we give an algorithm that determines the p-colorability and the determinants of
knots from their petal projections. In particular, we compute the determinants of all prime knots with
crossing number less than 10 from their petal permutations.

1. Background
There are many different ways to define a knot. The standard definition of a knot, which can be found in
[1] or [4], is an embedding of a closed curve in three-dimensional space, K : S 1 → R3 . Less formally, we can
think of a knot as a knotted circle in 3-space. Though knots exist in three dimensions, we often picture them
via 2-dimensional representations called knot diagrams. In a knot diagram, crossings involve two strands of
the knot, an overstrand and an understrand. The relative height of the two strands at a crossing is conveyed
by putting a small break in one strand (to indicate the understrand) while the other strand continues over
the crossing, unbroken (this is the overstrand). Figure 1 shows an example of this.

Two knots are defined to be equivalent if one can be continuously deformed (without passing through
itself) into the other. In terms of diagrams, two knot diagrams represent equivalent knots if and only if they
can be related by a sequence of Reidemeister moves and planar isotopies (i.e. “wiggling”). In Figure 2, we
illustrate the three Reidemeister moves. To learn more, see [1] or [4].

If a knot is equivalent to a simple, crossingless circle, it is called a trivial knot, or the unknot. Two examples
of the unknot are provided in Figure 3.

If a knot cannot be deformed into the unknot, it is said to be non-trivial —that is, there’s no way to
untangle it without passing it through itself. One of the central problems of knot theory is to find ways to
show that a knot is non-trivial. To do this, much of knot theory is based on finding and understanding knot
invariants—functions defined on knots that give the equivalent outputs (in the form of numbers, polynomials,
etc.) if the two input knots are equivalent. Examples of knot invariants include colorability, the knot

2010 Mathematics Subject Classification. 57M27.

Key words and phrases. petal projection, knot determinant, colorability.

Figure 1. A knot diagram.

1
2 ALLISON HENRICH AND ROBIN TRUAX

Figure 2. The three Reidemeister moves.

Figure 3. Two diagrams of the unknot.

determinant, the Jones polynomial, the Alexander polynomial, and Khovanov homology. It is the first two of
these invariants that will be the focus of this paper.

2. Colorability
This section relies heavily on an important definition in knot theory: an “arc.”
Definition 1. An arc is a segment of a knot in a diagram which begins and ends at adjacent undercrossings
(crossings such that the segment passes under the crossing).
We may decompose a knot diagram into a collection of arcs. Figure 4 demonstrates an example. It’s also
important to note that there are always the same number of arcs as there are crossings, with the exception of
the crossingless diagram of the unknot. We can see this because at every crossing, exactly one of the strands
must go under another, thereby starting a new arc. This will become important later, as it lets us construct
a matrix from a knot diagram (with rows corresponding to crossings and columns corresponding to arcs) that
is guaranteed to be square.
2.1. Tricolorability. To understand the general notion of p-colorability—one of the knot invariants we are
most interested in for the purposes of this paper—we should first describe the specific case when p = 3:
tricolorability. Colorability is a well-studied concept in knot theory (see [5] for a further description of all the
topics covered in this section), and tricolorability is a particularly accessible type of colorability to study.
 PETAL PROJECTIONS, KNOT COLORINGS & DETERMINANTS 3

Figure 4. A decomposition of the trefoil into 3 arcs (dashes used to distinguish different
arcs, not actual breaks in the knot).

Figure 5. The unknot (left) is not tricolorable, while the trefoil (right) is.

Definition 2. A knot diagram is tricolorable when each of the arcs in the knot diagram can be assigned
one of three colors such that:
• more than one color is used, and
• at any given crossing, all of the arcs are either the same color or all different colors.
For example, the usual diagram of the unknot is not tricolorable, since it has only one arc and therefore
can only use one color. In contrast, the usual diagram of trefoil knot is tricolorable; see Figure 5. Not only is
it true that tricolorability is a knot invariant – that is, any two diagrams of a knot are either both tricolorable
or both not tricolorable – but the number of ways to tricolor a knot is also invariant under all Reidemeister
moves. Therefore, any two equivalent knot diagrams must be tricolorable in the same number of ways. This
means that showing that there are a different number of ways to tricolor two knots is enough to show that
they’re not equivalent. Note that the converse is not true—two knots that are tricolorable in the same number
of ways may still be distinct, so the invariant isn’t complete.

Before we talk about generalizing tricolorability, let us see how determining if a knot is tricolorable can be
done by solving a system of linear equations. Suppose we assign a variable to each arc in a knot diagram. At
a crossing, the overstrand might be called ak while the understrands are labeled with ai and aj . A “coloring”
of these arcs can be viewed as assignments of the numbers 0, 1, or 2 to our variables. This is not just an
arbitrary choice of symbols: it turns out that the equation 2ak − ai − aj = 0 holds mod 3 if and only if the
three “colors” (or numbers) are all the same or all different. Recall that arithmetic mod 3 means that any
number n is considered equivalent to any other number with the same remainder as n when divided by 3. For
example, 1 ≡ 4 ≡ 7 (mod 3) and −1 ≡ 2 ≡ 14 (mod 3). Another notation we can use to describe a number n
and its equivalents mod 3 is n. The set of these equivalence classes is called Z/3Z = {0, 1, 2} and has a ring
structure. This leads us to recall the following classical result:
Theorem 1. A knot K is tricolorable if and only if for any diagram of K, the arcs of the diagram (labeled
with the variables a1 , a2 , a3 , ..., an ) can be assigned values from the ring Z/3Z = {0, 1, 2} (the integers mod 3)
4 ALLISON HENRICH AND ROBIN TRUAX

such that at least two elements are used, and at every crossing, we have that
2ak − ai − aj ≡ 0 (mod 3)
where ak is the overstrand label and ai and aj correspond to the understrands at the crossing.
Note that, in the theorem above, it is possible for ak , ai , and aj to not all be distinct, as is the case for a
reducible crossing. This case does not pose a problem, since assigning the same “color” to all three variables
satisfies the equation.
2.2. Knot p-colorability and the knot determinant. Generalizing the notion of tricolorability to p-
colorability is as simple as replacing “Z/3Z” in Theorem 1 with Z/pZ. For completeness, we will state the
resulting definition:
Definition 3. A knot K is p-colorable if and only if for any diagram of K, the arcs of the diagram (labeled
with the variables a1 , a2 , a3 , . . . , an ) can be assigned values from the ring Z/pZ = {0, 1, 2, . . . , p − 1} such that
at least two elements are used, and at every crossing, we have that
2ak − ai − aj ≡ 0 (mod p)
where ak is the overstrand label and ai and aj correspond to the understrands at the crossing.
As an example, we can 5-color the figure-eight knot as shown in Figure 6.

Figure 6. The figure-eight knot.

Notice that if we index the crossings in a knot diagram using the integers {1, . . . , n} and index the arcs in
the diagram using the same set, we can create an n × n matrix M with elements in Z/pZ that represents the
system of linear equations generated by the p-colorability definition. In this matrix, each row corresponds to
a crossing, and each column corresponds to an arc. Since there are always at least p solutions (the trivial
ones where we “color” all arcs the same number), the determinant of this matrix will always be 0. To get
a more useful invariant, we take a first minor of M (that is, we strike a row and column from our matrix
and take the determinant of the resulting submatrix) – if it is 0, the knot is p-colorable. Otherwise, it is not.
This determinant is independent of both the diagram chosen to represent the knot and its labeling. Thus, we
see a nice connection between colorability and linear algebra.

However, we would like to extend this connection by finding a way to test knots for all primes p at once.
To do this, we construct the above matrix as a matrix in Z and take a first minor to get an invariant called
the determinant of the knot. By our work in the previous paragraph, a knot is p-colorable if and only if p
divides the determinant of the knot (since then the minor goes to 0 mod p).

A final note for this section: p-colorability is only defined for ordinary, double-crossing diagrams of knots,
since our definition only allows for three arcs interacting at a single crossing. We are about to venture into a
realm of nonstandard projections of knots to which the definitions above don’t obviously apply. One of the
 PETAL PROJECTIONS, KNOT COLORINGS & DETERMINANTS 5

Figure 7. A petal projection of the trefoil

main goals of this paper is to describe a process for determining colorability from these nonstandard knot rep-
resentations. Before we can describe this process, we must describe the projections we’re interested in studying.

3. Petal projections, split petal projections, and their properties

3.1. Petal projections. In [2], Adams et al. introduce the concept of a petal projection—a special type of
diagram of a knot involving a single crossing, called an übercrossing diagram.
Definition 4. A petal projection is a knot projection with only one crossing—that has many strands of
the knot passing through it—and no nested loops. Each strand passing through the crossing is assigned a
number 1, 2, ..., n that indicates the height of the strand, where 1 is assigned to the topmost strand and n is
assigned to the bottom strand. Here, n is called the petal number of the petal projection. A petal projection
with petal number n is also called an n-petal projection.
We see an example of a petal projection of a trefoil in Figure 7. The trefoil is the only non-trivial knot
that requires just 5 petals to draw.

Petal projections are typically described by a permutation that describes the relative ordering of the
strands, taken from the projection’s labeling. There are many petal permutations that are equivalent. Since
we can begin traveling around the petal projection at any petal, recording the heights of the strands we
encounter along the way, we can cyclically permute the permutation that describes the projection. For
example, the trefoil in Figure 7 is defined by the permutation (1, 3, 5, 2, 4), but it could also be defined as
(3, 5, 2, 4, 1) or (5, 2, 4, 1, 3). For this reason, we follow the convention in [2] to always write petal permutations
starting with the 1.

We can also imagine turning the petal projection upside down, which doesn’t change the knot type or the
number of petals in the projection. This is equivalent to making the highest strand the lowest, the second
highest the second lowest, and so on. Therefore, the trefoil could also be written: (5, 3, 1, 4, 2) (and then
shifted to (1, 4, 2, 5, 3)). Notice that while 1 is at the start of both this representation and the original one,
they are still two different petal permutations describing the same knot.

Furthermore, since we can always move the bottom strand around to the top of the pile of übercrossing
strands, we can start recording the order of the strands at any height. So, the permutation (1, 6, 4, 2, 5, 7, 3)
could also be written as (2, 7, 5, 3, 6, 1, 4) by adding 1 to each entry (mod 7), which can be rewritten as
(1, 4, 2, 7, 5, 3, 6).

Another important property of petal permutations is that if any two consecutive numbers in the permuta-
tion are consecutive integers, we may pull the petal with the consecutive integers through the multi-crossing,
merging three petals into one. This is possible because, as the two integers on the petal correspond to heights,
the petal labeled with two consecutive integers occupies its own vertical slice of the multi-crossing, and
therefore can be pulled through without interference. This is difficult to visualize without a diagram, so one
is provided in Figure 8. Notice that the petal labeled with 3 and 4 is the one that is removed, so when the
6 ALLISON HENRICH AND ROBIN TRUAX

Figure 8. Simplifying a petal projection.

Figure 9. Any petal projection with an even number of petals splits into a (possibly
linked) linked union of unknots.

heights are relabeled, all of the heights greater than 4 are reduced by 2.

This result is very interesting, as it significantly reduces the number of possible irreducible petal projections
we need to consider. This alone, with the properties discussed above, shows that the trefoil is the only knot
with petal number 5. Indeed, there exists a more complete classification of when two petal permutations
describe the same knot, given in the recent paper A Reidemeister type theorem for petal diagrams of knots
[6]. This is an exciting development, as it gives more credence to the theory that petal projections provide a
new, useful language from the geometric to the purely algebraic.

The last property of petal projections that we observe is that all petal projections of knots must have an
odd number of petals. This is a result of the fact that if we draw a petal projection with an even number of
petals (say, 2n), we end up with an n-component link. Figure 9 provides an example.
3.2. Split petal projections.
Definition 5. A split petal projection is a redrawing of a petal projection in such a way that:
• There are only double-crossings.
• The split petal projection has the same symmetry as the original petal projection.
• There are no reducible crossings.
• The split petal projection is uniquely identified by the same petal permutation.
The notion of split petal projections briefly appears in [2] under the name “star patterns”, but is not explored.
For example, in Figure 10, we see a 5-petal projection of the trefoil knot, and its corresponding split petal
projection.
To see how to “split” a petal projection, it is helpful to work backwards—we instead visualize how a split
petal projection can be reformed into a petal projection, although the process is invertible.
 PETAL PROJECTIONS, KNOT COLORINGS & DETERMINANTS 7

Figure 10. A petal projection and split petal projection of the trefoil.

In Figure 11, we provide an example showing how to assemble a petal projection from a split petal
projection. Here, we’ll work with the knot shadow for simplicity. In the figure, we’re using the example of a
split petal projection with 7 petals, but the process is analogous for any number of petals. Simply fold each
strand (a portion of the knot extending from one of the outer-most points of the diagram to another) into
the center, one at a time.

Figure 11. Repairing a split petal projection.

8 ALLISON HENRICH AND ROBIN TRUAX

Definition 6. The petal number of a split petal projection is the petal number of the petal projection that
arises from the folding process described in Figure 11.
Split petal projections have some particularly nice properties. For example,
p(p−3)
Proposition 2. A split petal projection with p petals has 2 crossings and the same number of arcs.
Proof. Split petal projections can be obtained by perturbing all the arcs at central crossing and then untwisting
all the resulting loops, as seen in Figure 11 (starting at the end of the sequence and working backwards).
Now, initially each of the p strands going through the übercrossing intersects each other strand exactly
once. However, there is only one crossing because all of these intersections happen at a single point. In
contrast, after the initial perturbation of the arcs off of the central intersection point, each of the p strands
still intersects each of the other p − 1 strands, but now we only have double-crossings. This results in p(p−1)
2
crossings in total. Then, the untwisting of all the loops (to avoid reducible crossings) reduces the number of
crossings by p, giving p(p−3)
2 crossings, as desired. Finally, we note that, in any knot diagram with at least
one crossing, the number of crossings equals the number of arcs.
In a petal projection, a strand of the knot passing through the übercrossing has a certain height, described
by the petal permutation. The key fact about split petal projections is that the crossings “inherit” these
heights in a symmetric way, so that each strand height from the petal projection helps to determine the
crossing information at p − 3 crossings in the split petal projection.

One way to think about this is that a strand of the knot associated to a given height intersects the strands
associated to each of the other heights exactly once when strands are initially perturbed to avoid the central
übercrossing point, resulting in p − 1 crossings associated to a given height. The two crossings at the ends of
the strand are removed when loops are untwisted, resulting in a total of p − 3 crossings that are labeled by a
given height.

There is another way to come to this answer. The key is to notice that the number and arrangement
of crossings is independent of the heights; another way to say this is that the projection of a split petal
projection is independent of its petal permutation. Therefore, any given height will appear at the same
number of crossings, by symmetry. This gives the following equation:
# of crossings × # of heights at each crossing
= # of crossings per height
# of heights
Filling in the requisite information, we again find that a given height appears at p − 3 crossings:
(p(p − 3)/2) · 2
=p−3
p
This style of reasoning will be critical in the next section, when we develop a completely algebraic formula for
the Gauss code of a split petal projection.

4. Gauss Codes
A Gauss code for a knot is a sequence of integers that describes the order in which you encounter crossings
in a knot diagram when traveling along the knot. To generate a Gauss code from a knot diagram, we first
number the crossings in the diagram 1 through n (where n is the total number of crossings) in any order.
Then, starting at any point on the knot and choosing a direction to travel, we traverse the knot by following
the strand as it weaves through the knot. Whenever we pass through a crossing, we record either the crossing’s
number (if we travel “over” the crossing) or the crossing number’s negative (if we travel “under” the crossing).

Visually, calculating the Gauss code of a simple knot is easy. See Figure 12 for an example. While each
knot may be represented by many different Gauss codes (generated by different diagrams, numberings, base
points, etc.), a given Gauss code determines a knot type uniquely (up to mirror image).

Now, our main goal is to evaluate the knot determinant for general split petal projections, so we need an
algorithm to generate the Gauss code for any split petal projection from a petal permutation. In short, we’d
 PETAL PROJECTIONS, KNOT COLORINGS & DETERMINANTS 9

Figure 12. A Gauss code of the 819 knot: −1, 2, −3, −8, 5, 1, −2, 6, −7, 3, 4, −5, −6, 7, 8, −4.

Figure 13. A naive attempt to generate the unsigned Gauss code of a split petal projection:
1, 2, 3, 4, 5, 6, 7, 8, 9, 1, 10, 3, 11, 5, 12, 7, 13, 9, 14, 10, 2, 11, 4, 12, 6, 13, 8, 14.

like to find a function with a petal permutation as an input and a Gauss code—preferably one that inherits
certain properties from the symmetry of the petal projection—as the output.

4.1. Gauss Codes for Split Petal Projections. Before we begin describing an algorithm to generate
Gauss codes for split petal projections, let’s generate a simpler, unsigned Gauss code. The advantage of
beginning with an unsigned Gauss code (that doesn’t record over-under information about crossings) is that
we don’t need to know in advance the information about relative heights of strands that will be provided by
the petal permutation. We simply need to know the petal number.

Let’s consider a naı̈ve attempt to create an algorithm to find an unsigned Gauss code for a given petal
number. We will use the example of the 7-split petal projection with petal permutation (1, 3, 5, 2, 7, 4, 6), the
figure-eight knot. One way to start might be to label each crossing in the order in which we first encounter it.
This starts simply, giving an unsigned Gauss code that begins 1, 2, 3, . . . for the first few numbers. However,
this method comes with its own problems; while the Gauss code starts simple, it quickly gets complicated, as
shown in Figure 13.

A better way to generate a useful unsigned Gauss code starts by considering the structure of the split petal
projection. Notice that in general, there are p−3
2 “layers” of crossings, each with p symmetrically distributed
10 ALLISON HENRICH AND ROBIN TRUAX

Figure 14. Layer 0 is the set of crossings lying in the innermost (red) circle, Layer 1 is the
set of crossings in the middle (green) circle, and Layer 2 is the set of crossings in the
outermost (yellow) circle.

Figure 15. The unsigned Gauss code of a 7-split petal projection is:
(0, 1), (1, 1), (1, 2), (0, 3), (0, 4), (1, 4), (1, 5), (0, 6), (0, 7), (1, 7), (1, 1), (0, 2), (0, 3), (1, 3),
(1, 4), (0, 5), (0, 6), (1, 6), (1, 7), (0, 1), (0, 2), (1, 2), (1, 3), (0, 4), (0, 5), (1, 5), (1, 6), (0, 7)

crossings. These layers appear in concentric circles, with the first layer (Layer ‘0’) being all the crossings that
touch the central region, and each next layer being the next concentric circle out. For instance, when the
petal number p = 9, there are 9−32 = 3 layers, as shown in in Figure 14. Layer 0 consists of all crossings at
the vertices of the central nonagon. Layer 1 consists of the nine crossings lying just outside the nonagon,
and Layer 2 contains the remaining nine crossings along the boundary of the exterior region of the diagram.
Another example, this one for p = 7, can be seen in Figure 15.

Now, instead of labeling the crossings directly, we begin by labeling each crossing with an ordered pair (L, n),
where L denotes the layer that the crossing is in, and n is the “index” of the crossing in the particular layer.
We determine the index of each crossing by letting the first crossing we encounter in each layer be index 1, and
number the rest in the layer clockwise. Numbering the crossings this way more clearly shows a general pattern
in the unsigned Gauss code. This can be verified with the example of the 7-petal projection, shown in Figure 15.
 PETAL PROJECTIONS, KNOT COLORINGS & DETERMINANTS 11

Figure 16. The unsigned Gauss code of a 9-split petal projection is:
(0, 1), (1, 1), (2, 1), (2, 2), (1, 3), (0, 4), (0, 5), (1, 5), (2, 5),
(2, 6), (1, 7), (0, 8), (0, 9), (1, 9), (2, 9), (2, 1), (1, 2), (0, 3),
(0, 4), (1, 4), (2, 4), (2, 5), (1, 6), (0, 7), (0, 8), (1, 8), (2, 8),
(2, 9), (1, 1), (0, 2), (0, 3), (1, 3), (2, 3), (2, 4), (1, 5), (0, 6),
(0, 7), (1, 7), (2, 7), (2, 8), (1, 9), (0, 1), (0, 2), (1, 2), (2, 2),
(2, 3), (1, 4), (0, 5), (0, 6), (1, 6), (2, 6), (2, 7), (1, 8), (0, 9).

As you can see, the first entries in the ordered pairs appear in a repeating palindrome: 0, 1, 1, 0. For larger
knots, this palindrome will grow exactly as expected, first to 0, 1, 2, 2, 1, 0, then to 0, 1, 2, 3, 3, 2, 1, 0, and so
on. This result generalizes into Definition 8. This palindrome repeats in blocks of length p − 3. From the
block perspective, the second number’s pattern is just as predictable. In a 7-petal projection, each block is 4
numbers long, and the numbers come in the form n, n, n + 1, n + 2. Note that we reduce each number in
the block, replacing it with the smallest positive integer congruent to it mod p. In a 9-petal projection, the
blocks have length 6, and the numbers come in the form n, n, n, n + 1, n + 2, n + 3. Again, this result will
generalize into Definitions 9 and 10.

To demonstrate our algorithm, we consider the Stevedore knot, a 9-petal knot with petal permutation
(1, 3, 5, 2, 8, 4, 6, 9, 7). See Figure 16.

If we can explicitly generate this pattern for any size petal permutation, we will have effectively generated
an unsigned Gauss code for each petal number, since we can simply map (L, n) → L · p + n (where p is the
petal number). This map is a bijection from the set of pairs (L, n) to the integers 1 through p(p−3) 2 . We
know this because n ranges between 1 and p (since there are p crossings in each layer), and L varies from 0
to p−3 p−3
2 − 1, since there are 2 layers. This bijection inspires Proposition 3, which completes the algorithm.
In Figure 17, we give the example for the Stevedore knot.

Now, our goal is to take our unsigned Gauss code and use the petal permutation to determine where to
insert the negative signs that denote underpasses. To achieve this, we use the fact that a strand in the petal
projection associated with a given height is mapped to a strand in the split petal projection passing through
p − 3 crossings. This allows us to make a table mapping a crossing label from the unsigned Gauss code to the
height of the over- or under-strand at that crossing. After making this table, we can determine if the strand
associated to a number in the unsigned Gauss code is going over or under at the crossing to make the signed
Gauss code. Again, we will use the example of the Stevedore knot, which has a 9-split petal projection with
petal permutation (1, 3, 5, 2, 8, 4, 6, 9, 7):
12 ALLISON HENRICH AND ROBIN TRUAX

Figure 17. The unsigned Gauss code of a 9-split petal projection is:
1, 10, 19, 20, 12, 4, 5, 14, 23, 24, 16, 8, 9, 18, 27, 19, 11, 3,
4, 13, 22, 23, 15, 7, 8, 17, 26, 27, 10, 2, 3, 12, 21, 22, 14, 6,
7, 16, 25, 26, 18, 1, 2, 11, 20, 21, 13, 5, 6, 15, 24, 25, 17, 9.

Index 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Gen. Gauss Code 1 10 19 20 12 4 5 14 23 24 16 8 9 18 27 19 11 3
P.P. Height 1 1 1 1 1 1 3 3 3 3 3 3 5 5 5 5 5 5
Index 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
Gen. Gauss Code 4 13 22 23 15 7 8 17 26 27 10 2 3 12 21 22 14 6
P.P. Height 2 2 2 2 2 2 8 8 8 8 8 8 4 4 4 4 4 4
Index 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53
Gen. Gauss Code 7 16 25 26 18 1 2 11 20 21 13 5 6 15 24 25 17 9
P.P. Height 6 6 6 6 6 6 9 9 9 9 9 9 7 7 7 7 7 7

Table 1. The unsigned Gauss code of a 9-split petal projection (which may resolve, for
example, into the Stevedore knot when we sign the entries).

From the table, we can finally create the signed Gauss code for the Stevedore knot as follows. Note, for
instance, that the first “1” in the Gauss code is positive because the height of the associated instance of “1” in
the table is 1, while the height of the second instance of “1” in the table is 6. Since lower numbered arcs are
taken to be above higher numbered arcs, this tells us that the first pass through crossing 1 should be an overpass.

Stevedore Knot Gauss Code:

1, 10, 19, 20, 12, 4, 5, 14, −23, 24, 16, 8, 9, 18, 27, −19, 11, −3, −4, 13, 22, 23, 15, 7, −8, −17, −26,
−27, −10, 2, 3, −12, 21, −22, −14, 6, −7, −16, 25, 26, −18, −1, −2, −11, −20, −21, −13, −5, −6,
−15, −24, −25, 17, −9

We can use this to draw the split petal projection with crossing information clearly marked, as in Figure
18.When we create a general algorithm to generate the Gauss code of a split petal projection from a petal
permutation, breaking it down into these individual steps allows for much easier computation. At this point,
we are ready to explicitly describe how to determine the unsigned Gauss code for a given petal number. To
begin, we will establish several notational definitions.
 PETAL PROJECTIONS, KNOT COLORINGS & DETERMINANTS 13

Figure 18. The Stevedore Knot’s Split Petal Projection.

Definition 7. Let rn (k) be the smallest non-negative integer congruent to k mod n, i.e., the remainder
when k is divided by n. For example, r2 (5) = 1, and r3 (9) = 0.
Definition 8. Let p be a petal number. Then Lp is the repeating palindromic sequence
p−5 p−5 p−5 p−5
0, 1, 2, . . . , , , . . . , 2, 1, 0, 0, 1, 2, . . . , , , . . . 2, 1, 0, . . .
2 2 2 2
which we will denote by Lp0 , Lp1 , . . . . When the p is understood, we may suppress the superscripts and denote
our sequence by L0 , L1 , . . . .
Note that the periodicity of Lp is p − 3 = p−5 p−5
2 + 2 + 2.

This definition encodes the “layers” visited as we follow the knot, as discussed earlier.
Definition 9. Let p be a petal number. Then ap is the sequence
p−3 p−3
0, . . . , 0 , 1, 2, . . . , , 0, . . . , 0 , 1, 2, . . . , ,...
| {z } 2 | {z } 2
p−3 p−3
2 times 2 times

denoted by ap0 , ap1 , . . . . As before, when p is understood, we may drop the superscripts and write a0 , a1 , . . . .
Just as with the sequence Lp , the periodicity of ap is p − 3 for all p.
Definition 10. Define npk to be the quantity given by
   
p−1 k p
rp · + ak + 1.
2 p−3
Similarly, this definition encodes what we called the “index in a particular layer.” Now, we are ready to
describe a general formula for the unsigned Gauss code of a petal projection with p petals.
Proposition 3. An unsigned Gauss code for a split petal projection with p petals is given by the sequence
c0 , . . . cp(p−3)−1
where ck = p · Lk + nk . We call this specific Gauss code the petal unsigned Gauss code.
Definition 11. In a (signed or unsigned) Gauss code C corresponding to a petal projection with p petals,
the index k petal block corresponds to the p − 3 crossings ck(p−3) , . . . , c(k+1)(p−3)−1 . Notice that a petal
projection with p petals will have exactly p petal blocks, each corresponding to a height in the petal permutation.
14 ALLISON HENRICH AND ROBIN TRUAX

Consider, for example, Table 1. In this table, there are a total of 9 “blocks” which all have the same
height in the petal permutation, and all of these blocks have the same length (namely 9 − 3 = 6). To help
illustrate the previous definitions, we will again compute the unsigned Gauss code of a 9-split petal projection
in Table 2. Compare this to Table 1 – note that the row ck in Table 2 is identical to the unsigned Gauss code
computed in Table 1.

Index (k) 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Lp (see Def. 8) 0 1 2 2 1 0 0 1 2 2 1 0 0 1 2 2 1 0
ap (see Def. 9) 0 0 0 1 2 3 0 0 0 1 2 3 0 0 0 1 2 3
np (see Def. 10) 1 1 1 2 3 4 5 5 5 6 7 8 9 9 9 1 2 3
ck (see Prop. 3) 1 10 19 20 12 4 5 14 23 24 16 8 9 18 27 19 11 3
Index (k) 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
Lp (see Def. 8) 0 1 2 2 1 0 0 1 2 2 1 0 0 1 2 2 1 0
ap (see Def. 9) 0 0 0 1 2 3 0 0 0 1 2 3 0 0 0 1 2 3
np (see Def. 10) 4 4 4 5 6 7 8 8 8 9 1 2 3 3 3 4 5 6
ck (see Prop. 3) 4 13 22 23 15 7 8 17 26 27 10 2 3 12 21 22 14 6
Index (k) 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53
Lp (see Def. 8) 0 1 2 2 1 0 0 1 2 2 1 0 0 1 2 2 1 0
ap (see Def. 9) 0 0 0 1 2 3 0 0 0 1 2 3 0 0 0 1 2 3
np (see Def. 10) 7 7 7 8 9 1 2 2 2 3 4 5 6 6 6 7 8 9
ck (see Prop. 3) 7 16 25 26 18 1 2 11 20 21 13 5 6 15 24 25 17 9

Table 2. The unsigned Gauss code c0 , . . . , cp(p−3)−1 of a 9-split petal projection, computed
directly using the definitions created above.

4.2. The Signed Gauss Code. From here, we seek to produce a signed Gauss code, which depends not
only on the number of petals, but also on our knot’s petal permutation. Converting to the signed Gauss code
from the unsigned Gauss code is not difficult: it simply involves negating each ck that is associated with an
understrand at a crossing.

To do this, we need to find the heights of the two strands that make up a crossing. For this, it will be
helpful to have the following definition, which we will draw on in the following sections:
Definition 12. Suppose k is an index in the petal unsigned Gauss code c0 , . . . , cp(p−3)−1 . Then define k 0 to
be the unique index of the petal unsigned Gauss code such that k 6= k 0 but ck = ck0 .
Note that k and k 0 correspond to one another, meaning that (k 0 )0 = k. For example, if we are working
with the Stevedore knot, and take k = 0, then k 0 = 41, since c0 = c41 = 1, and similarly if we take k = 41,
then k 0 = 0 for the same reason. Then, our table allows us to see that the height of c41 is 6 but the height
of c0 is 1, so c0 is the overstrand and stays positive in the signed Gauss code, while c41 ’s sign is flipped.
Note that in general, it suffices to identify the petal block that ck0 is in, since each block corresponds to a
single number in the petal permutation. For instance, for each j ∈ {36, 37, 38, 39, 40, 41}, cj in the Stevedore
unsigned Gauss code has height 6. So in fact, we don’t need to know that k 0 = 41 specifically. We simply
need to know that 36 ≤ k 0 < 42.
0
k
Thus, we seek a simple expression for b p−3 c, the index of the petal block that ck0 is in (between 0 and
p − 1, inclusive). If we use the previous example of k = 0 and k 0 = 41, we can find the petal block that k 0 is
in (the 7th block, with index 6—coincidentally the same number as the height associated to this index) by
plugging in k 0 = 41 and p = 9 to get b 9−3
41
c = 6. To get the most out of identifying heights using petal blocks
0
rather than k itself, we seek a more general expression for block indices that doesn’t rely on a brute-force
search to find k 0 .
 PETAL PROJECTIONS, KNOT COLORINGS & DETERMINANTS 15

Lemma 4. Given that rn (k) denotes the smallest non-negative integer congruent to k mod n, then
 0    
k k
= rp + 2(ak0 − ak ) .
p−3 p−3
where, as before, k 0 is the unique integer such that ck = ck0 but k 0 6= k.
Proof. First note that ck0 = ck implies that both Lk0 = Lk and nk0 = nk , by Proposition 3. We may use the
definition of nk , Definition 10, to get the following.
   0 
p−1 k p−1 k
· + ak ≡ · + ak0 mod p
2 p−3 2 p−3
We also use the fact that (p − 1) is congruent to −1 mod p and multiply through by −2.
   0 
k k
− 2ak ≡ − 2ak0 mod p
p−3 p−3
j k j 0 k
k k
Finally, we rearrange and note that both p−3 and p−3 are bounded by 0 and p − 1, giving us the
desired result:

k0
    
k
= rp + 2(ak0 − ak ) .
p−3 p−3

This will prove to be a useful formula, but we still lack a crucial piece: it is difficult to determine what
ak0 will be without knowing k 0 , rendering this formula in its current form less useful. Therefore, we seek
an expression for ak0 − ak that is dependent only on k. To find this, note that both an = arp−3 (n) and
Ln = Lrp−3 (n) for all n. These follow from Definitions 8 and 9, the definitions of a and L, since both sequences
were defined to be periodic with period p − 3.
Lemma 5. Given that rn (k) denotes the smallest non-negative integer congruent to k mod n, then
rp−3 (k) + rp−3 (k 0 ) = p − 4
for each k and the corresponding k 0 .
This lemma can be intuitively understood as follows. Recall that k and k 0 denote the indices of a given
crossing as it appears twice in an unsigned Gauss code. Because the two indices correspond to the two
distinct strand heights associated to the same crossing, they must be in different petal blocks. What this
lemma tells us is that they appear in their respective blocks in complementary positions. For instance, if
k appears as the first index in its block, then k 0 appears as the last index in its block. If k appears as the
second index in its block, then k 0 appears as the second-to-last index in its block, and so forth. Thus, the
sum of the remainders of the two indices when divided by block length equals p − 4.

Proof. Consider that ck = ck0 implies that Lk = Lk0 . Because of the palindromic nature of Lk , this implies
that either rp−3 (k) = rp−3 (k 0 ) or rp−3 (k) + rp−3 (k 0 ) = p − 4.

To prove that we are in the second case, we assume that the first equality holds and derive a contradiction.
If rp−3 (k) = rp−3 (k 0 ), then ak = ak0 , by the periodicity of the a sequence. Combined with the fact that
nk = nk0 , this forces k and k 0 to be in the same petal block. In this case, the equality of rp−3 (k) = rp−3 (k 0 )
implies that k = k 0 , contradicting the definition of k 0 . Thus, rp−3 (k) + rp−3 (k 0 ) = p − 4, as desired.

Lemma 6. The quantity ak0 − ak is given by

(
p−3 p−3
ak0 − ak = p−3 2 − rp−3 (k) if rp−3 (k) < 2 ,

2 − rp−3 (k) − 1 otherwise.

Proof. By Lemma 5, k and k 0 appear at complementary points in the sequence a. Thus, either ak = 0 or
ak0 = 0. This allows us to characterize ak0 − ak as follows.
16 ALLISON HENRICH AND ROBIN TRUAX

• Case 1: Suppose that ak = 0. Then we are in the first half of a’s cycle, so rp−3 (k) < p−3
2 . In this
case, by Definition 9 and Lemma 5,
p−3 p−3 p−3
ak0 − ak = ak0 = rp−3 (k 0 ) + 1 − = p − 3 − rp−3 (k) − = − rp−3 (k).
2 2 2
p−3
• Case 2: Suppose that ak0 = 0, so rp−3 (k) ≥ 2 . In this case, by Definition 9,
p−3
ak0 − ak = −ak = − rp−3 (k) − 1.
2
These two cases combine to give us our desired piecewise function.
Thus, we have the following definition which will enable us to identify the petal blocks of corresponding
crossing labels in the unsigned Gauss code, given in Definition 14.
Definition 13. Define the following function.
(
p − 3 − 2rp−3 (k) rp−3 (k) < p−3
2 ,
d(k) = 2(ak0 − ak ) =
p − 5 − 2rp−3 (k) otherwise.
j k
Definition 14. The petal block containing k is given by b(k) = p−3 k
, and the petal block containing k 0 is
j k 
given by b0 (k) = rp p−3
k
+ dk .

Using the machinery developed above, we now can determine the signed Gauss code from a petal permutation
as follows.
Theorem 7. Let P = h0 , h1 , . . . hp−1 be a petal permutation describing a given knot in terms of a p-petal
projection. Let C = c0 , . . . cp(p−3)−1 denote the corresponding petal unsigned Gauss code, as defined in
Definition 11. Then a signed Gauss code corresponding to P is given by the following: C 0 = c00 , . . . c0p(p−3)−1 ,
where c0i = ci if hb(k) < hb0 (k) and c0i = −ci otherwise. We call this specific Gauss code the petal signed
Gauss code for P .
Consider, as an illustration of this theorem, Table 3. Table 3 lists the unsigned Gauss code, the k 0
corresponding to each k (which depends only on the unsigned Gauss code by definition), the heights derived
from the petal permutation of the Stevedore knot, and uses them to sign the unsigned Gauss code of a 9-split
petal projection corresponding to the Stevedore knot along the lines of Theorem 7.

Index (k) 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
ck 1 10 19 20 12 4 5 14 23 24 16 8 9 18 27 19 11 3
k0 41 28 15 44 31 18 47 34 21 50 37 24 53 40 27 2 43 30
P.P. Height 1 1 1 1 1 1 3 3 3 3 3 3 5 5 5 5 5 5
0
ck 1 10 19 20 12 4 5 14 -23 24 16 8 9 18 27 -19 11 -3
Index (k) 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
ck 4 13 22 23 15 7 8 17 26 27 10 2 3 12 21 22 14 6
k0 5 46 33 8 49 36 11 52 39 14 1 42 17 4 45 20 7 48
P.P. Height 2 2 2 2 2 2 8 8 8 8 8 8 4 4 4 4 4 4
0
ck -4 13 22 23 15 7 -8 -17 -26 -27 -10 2 3 -12 21 -22 -14 6
Index (k) 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53
ck 7 16 25 26 18 1 2 11 20 21 13 5 6 15 24 25 17 9
k0 23 10 51 26 13 0 29 16 3 32 19 6 35 22 9 38 25 12
P.P. Height 6 6 6 6 6 6 9 9 9 9 9 9 7 7 7 7 7 7
c0k -7 -16 25 26 -18 1 -2 -11 -20 -21 -13 -5 -6 -15 -24 -25 17 -9

Table 3. The signed Gauss code c00 , . . . , c0p(p−3)−1 of the Stevedore knot, calculated directly
using the algorithm outlined above.
 PETAL PROJECTIONS, KNOT COLORINGS & DETERMINANTS 17

We can now describe how this particular signed Gauss code for a split petal projection can help us to
compute a knot’s determinant from its petal permutation. We’ll implement an algorithm for computing the
knot determinant of split petal projections in Python. We’ll test the results of our program against the those
results about petal projections of all prime knots with fewer than 10 crossings given in [2]. We’ll also test
it against non-trivial knots of various sizes, from 5-split petal projections up to 50-split petal projections.
These will appear in our appendices and will include run times over multiple trials.

5. An algorithm to determine the colorability of a split petal projection

5.1. Description of the algorithm. To determine the colorability of a knot from one of its petal permuta-
tions, P , we begin by labeling the arcs in the corresponding split petal projection of the knot with elements
of Z/pZ. We then form a matrix from these labelings, as described Section 2.2. Using linear algebra, we can
determine the dimension of the linear system of equations’ solution space (which represents the p-colorability
of the knot). Since the trivial solutions form a space of dimension 1, the knot is p-colorable if the solution
space has dimension at least 2. Let us now consider how exactly to construct this matrix, MP .

For concreteness, we will first explore the specific case of the Stevedore knot. We use the signed Gauss
code to number the arcs. Recall that an arc begins and ends at an understrand – thus arcs begin and end at
negative numbers in the signed Gauss code, as follows:
Stevedore Knot Gauss Code:
1, 10, 19, 20, 12, 4, 5, 14, −23, 24, 16, 8, 9, 18, 27, −19, 11, −3, −4, 13, 22, 23, 15, 7, −8, −17, −26,
−27, −10, 2, 3, −12, 21, −22, −14, 6, −7, −16, 25, 26, −18, −1, −2, −11, −20, −21, −13, −5, −6,
−15, −24, −25, 17, −9

The Arc Start/End Points in the Stevedore Knot’s Gauss Code:

(−23, −19), (−19, −3), (−3, −4), (−8, −17), (−17, −26), (−26, −27), (−27, −10), (−10, −12),
(−12, −22), (−22, −14), (−14, −7), (−7, −18), (−18, −1), (−1, −2), (−11, −20), (−20, −21),
(−21, −13), (−13, −5), (−5, −6), (−6, −15), (−15, −24), (−24, −25), (−25, −9), (−9, −23)
We will examine a single row of this matrix, corresponding to the crossing labeled 1. Each of the arcs
is represented by one of the columns in the matrix, numbered in the order shown above. We will place a
−1 in the column associated with the two arcs which are understrands at crossing 1, and a 2 in the column
representing the overstrands.
row 1 of MP = 0 0 0 0 0 0 0 0 0 0 0 0 −1 −1 0 0 0 0 0 0 0 0 0 0 0 0 2

The bold arcs are the three which involve crossing 1. Notice that the two arcs that form the understrands
have crossing 1 as their start or end point. However, the arc that passes over crossing 1 does not appear
explicitly. Since one can check that 1 is between −9 and −23 in the signed Gauss code, the arc (−9, −23) is
the arc that passes over crossing 1.

At this point, for completeness, we will state the process we for finding the Gauss code’s associated matrix
in full generality:
(1) We start with a petal signed Gauss code of length p(p − 3), and define M to be a p(p−3) 2 × p(p−3)
2
matrix comprised entirely of 0s.
(2) We split the Gauss code into arcs by ending the previous arc and starting a new one whenever we
encounter a negative number, just as we did earlier with the Stevedore knot.
(3) We include both endpoints of the arc in our “arc-sets” (which is why we say “split” instead of
“partition”).
(4) We label the arcs we’ve created with the numbers 1 to p(p−3) 2 . For example, in the case of the
Stevedore knot, the arc (−23, −19) is Arc 1, the arc (−19, −3) is Arc 2, etc.
(5) For each number n ∈ {1, . . . , p(p−3)
2 }, we determine which arc is the overstrand at crossing n (say
Arc i) and which arcs are the understrands at crossing n (say Arcs j and j + 1).
(6) We set the nth row, ith column element to 2, and the nth row, jth and j + 1th column elements to
−1, just we did when creating the first row of MP for the Stevedore knot.
18 ALLISON HENRICH AND ROBIN TRUAX

Now, we take the first minor of this matrix (striking out the first row and column), treating the elements
as members of Z (instead of how they were originally defined in Z/pZ), and see if this determinant is divisible
by particular p. This approach allows us to test the p-colorability of the knot for all p at once. Furthermore,
we recall the classical result that the number of ways in which the knot is p-colorable is equal to the product
of all p in the prime factorization of this resulting minor.

Definition 15. Let ordp (n), where p and n are positive integers, be the largest power of p that divides n. In
other words, k = ordp (n) is the unique nonnegative integer such that pk | n but pk+1 - n.

The following is a well-known result [5].

Theorem 8. Let M be the first minor of a coloring matrix for a knot K. This minor, called the determinant
of the knot is an invariant. Furthermore, K is colorable using p-colors in pordp (|M |)+1 ways. Since every knot
has p trivial colorings (one for each color), there are pordp (|M |)+1 − p nontrivial p-colorings of the knot.

We now turn our attention to directly implementing the algorithm for determining a knot’s determinant
from one of its petal permutations in Python.

5.2. Implementing the algorithm in Python. The first step in implementing the algorithm is to create
a library of functions: one for determining the unsigned petal Gauss code from the petal number of the
split petal projection, one for deriving the (signed) petal Gauss code from the unsigned code and the
petal permutation, one for generating the petal Gauss code matrix from the petal Gauss code, and one for
evaluating the knot determinant from this matrix. Figure 19 visually shows the way these functions fit together.

How these functions are written can vary. Ours are written for readability first and speed second. We
have included our full code (with documentation) in Appendix A. We compute the determinants of all prime
knots with fewer than 10 crossings from their petal permutations, given in [3], with runtime in Appendix
B. We were also interested in discovering how our algorithm performs as we increase the size of the petal
projection. Current bounds on the relationship between crossing number and petal number will mean that
our experiments will deal with knots of very large size, pushing the code to its limits. This is further explored
in Appendix C. Finally, we will do some experimental testing of the distribution of p-colorability of random
knots using randomly selected permutations in Appendix D.

6. Conclusion
Currently, the fastest determinant algorithms have complexity ≈ O(n2.373 ) and n is O(p2 ), where p is
the petal number of the knot. This means that our algorithm should have complexity ≈ O(p4.746 ), since
everything else we do has strictly less complexity. We do not know how this is related to crossing number,
the standard measure of complexity for knots. However, Colin Adams et. al [2] have proven that the
2
crossing number, c(K), is bounded above by p(K) −2p(K)−3
4 . Indeed, torus knots of the form Tr,r+1 realize
the inequality, showing that this bound is tight. This is good news: it suggests that, at least for some torus
knots, our algorithm can have sub-cubic complexity with respect to crossing number.

There are still more interesting questions we can ask in this setting. For instance,
(1) Is there a faster (less computationally complex) way to compute the knot determinant than evaluating
the first minor of the petal Gauss code matrix that takes further advantage of the symmetry of petal
projections?
(2) Can one replicate this work using a wider class of objects (than elements of Z/pZ) to label the arcs of
the split petal projection? For instance, can quandles be used in conjunction with petal projections
to produce interesting knot invariants?
We hope that this paper encourages continued interest in the possibilities of petal projections. Viewing
knots via their petal projections has the potential lead to more connections between knot theory, combinatorics,
and algebra, just as it has led to connections between knot theory and probability theory in [7].
 PETAL PROJECTIONS, KNOT COLORINGS & DETERMINANTS 19

Figure 19. A visualization of the algorithm.

Acknowledgements
The authors would like to thank the Summer Institute of Mathematics at the University of Washington
(SIMUW) for bringing us together to work on this project. In particular, we would like to thank Sándor
Kovács, Ron Irving, Jim Morrow, Kelly Emmrich, Shea Engle, Guangqiu Liang, Wilson Ly, and Andreas Quist
for their leadership and support. We would also like to thank Andrew Lebedinsky and Gregory Baimetov,
SIMUW participants, for contributing early ideas to the project. The authors would also like to thank the
Simons Foundation (#426566, Allison Henrich) for their support of this research. Finally, the authors would
like to thank their anonymous reviewer from Involve for valuable feedback and detailed suggestions.

References
[1] Adams, Colin. The Knot Book. The American Mathematical Society, 2004: 1-16.
20 ALLISON HENRICH AND ROBIN TRUAX

[2] Adams, Colin, Orsola Capovilla-Searle, Jesse Freeman, Daniel Irvine, Samantha Petti, Daniel Vitek, Ashley Weber, and
Sicong Zhang. “Bounds on übercrossing and petal numbers for knots.” Journal of Knot Theory and Its Ramifications. 24,
no. 02 (2015): 1550012.
[3] Adams, Colin, Thomas Crawford, Benjamin DeMeo, Michael Landry, Alex Tong Lin, MurphyKate Montee, Seojung Park,
Saraswathi Venkatesh, and Farrah Yhee. “Knot projections with a single multi-crossing.” Journal of Knot Theory and Its
Ramifications. 24, no. 03 (2015): 1550011.
[4] Johnson, Inga, Allison Henrich. An Interactive Introduction to Knot Theory. Dover Publications, 2017: 26-27, 39-40.
[5] Livingston, Charles. Knot theory. Vol. 24. Cambridge University Press, 1993: 32-48.
[6] Colton, Leslie, Cory Glover, Mark Hughes, and Samantha Sandberg. “A Reidemeister type theorem for petal diagrams of
knots.” Topology and its Applications. 267, (2019): 106896.
[7] Even-Zohar, Chaim, Joel Hass, Nati Linial, and Tahl Nowik. “Invariants of random knots and links.” Discrete & Computational
Geometry. 56, no. 2 (2016): 274-314.

Appendix A. Code
Any of the code we used in calculations for creating our appendices is included here. All of our code will
also be available on Github (clickable) at the URL github.com/RobinTruax/PetalProjections.
Listing 1. Contains an implementation of our algorithm
#Importing the math library, SymPy (for prime factorization of natural numbers), and
,→ Numpy (for determinant calculation)
import math
import numpy
from sympy.ntheory import factorint

#Auxiliary Function 1: r_n(k)

def r(k,n):
return k%n

#Auxiliary Function 2: d(k), which depends on the petal number |P|

def d(k, petal_number):
if 2*r(k, petal_number-3) < (petal_number-3):
return petal_number - 3 - 2*r(k, petal_number-3)
return petal_number - 5 - 2*r(k, petal_number-3)

#Auxilary Function 3: b(k), which depends on the petal number |P|

def b(k, petal_number):
return math.floor(k/(petal_number - 3))

#Auxiliary Function 4: b’(k), which depends on the petal number |P|

def bPrime(k, petal_number):
return r(b(k, petal_number)+d(k, petal_number),petal_number)

#Building Block Function 1: Generates the unsigned Gauss Code, depends only on the petal
,→ number |P|
def unsignedGaussCode(petal_number):
L_array = [i for i in range(int((petal_number-3)/2))] + [int((petal_number-5)/2) - i
,→ for i in range(int((petal_number-3)/2))]
a_array = [0 for i in range(int((petal_number-3)/2))] + [i+1 for i in range(int((
,→ petal_number-3)/2))]
n_array = [r(int((petal_number-1)/2)*math.floor(k/(petal_number-3)) + a_array[r(k,
,→ petal_number-3)],petal_number)+1 for k in range(petal_number*(petal_number-3))]

unsigned_gauss_code = [petal_number*L_array[r(k, petal_number -3)] + n_array[k] for k

,→ in range(petal_number*(petal_number-3))]
return unsigned_gauss_code
 PETAL PROJECTIONS, KNOT COLORINGS & DETERMINANTS 21

#Building Block Function 2: Generates the signed Gauss Code, requires unsigned Gauss code
,→ and petal permutation
def signGaussCode(Gauss_code, petal_permutation):
petal_number = len(petal_permutation)
for k in range(petal_number*(petal_number - 3)):
if(petal_permutation[b(k, petal_number)] < petal_permutation[bPrime(k,
,→ petal_number)]):
pass
else:
Gauss_code[k] *= -1
return Gauss_code

#Building Block Function 3: Splits up a signed Gauss code into an aray of arcs
def splitSignedGaussCode(signed_Gauss_code):
#Preparing to split up the signed Gauss code into arcs
arcs = []
current_arc = []

#Creating a list to scan through that starts with the first negative number
trawl = list(range(len(signed_Gauss_code)))
for i in range(len(signed_Gauss_code)):
if signed_Gauss_code[i] < 0:
trawl = trawl[i:] + trawl[:i]
break

#Splitting the signed Gauss code up into arcs

current_arc.append(signed_Gauss_code[trawl[0]])
for i in trawl[1:]:
current_arc.append(signed_Gauss_code[i])
if signed_Gauss_code[i] < 0:
arcs.append(current_arc)
current_arc = []
current_arc.append(signed_Gauss_code[i])
current_arc.append(signed_Gauss_code[trawl[0]])
arcs.append(current_arc)

return(arcs)

#Building Block Function 4: Generates the crossing matrix, requires the signed Gauss code
,→ and petal permutation
def createCrossingMatrix(signed_Gauss_code, petal_permutation):
#Initializing everything
petal_number = len(petal_permutation)
matrix_size = math.floor(petal_number*(petal_number-3)/2)
crossing_matrix = [[0 for i in range(matrix_size)] for i in range(matrix_size)]

arcs = splitSignedGaussCode(signed_Gauss_code)

for crossing in range(1, len(crossing_matrix)+1):

for arc in range(len(arcs)):
if crossing in arcs[arc]:
crossing_matrix[crossing-1][arc] = 2
if -1*crossing in arcs[arc]:
crossing_matrix[crossing-1][arc] = -1
22 ALLISON HENRICH AND ROBIN TRUAX

return crossing_matrix

#Building Block Function 5: Evaluates the first minor of the crossing matrix
def evaluateKnotDeterminant(crossing_matrix):
#This part slices off the first crossing (first step of getting the minor)
crossing_matrix = crossing_matrix[1:]

#This part slices off the first arc in each crossing (second step of getting the
,→ minor)
for crossing in range(len(crossing_matrix)):
crossing_matrix[crossing] = crossing_matrix[crossing][1:]

#This part converts our 2D array into a numpy array and then evaluates the
,→ determinant
crossing_numpy_matrix = numpy.array(crossing_matrix)
return abs(int(round(numpy.linalg.det(crossing_numpy_matrix))))

#This function directly generates the knot determinant of a petal projection with a given
,→ petal permutation
def knotDeterminant(petal_permutation):
unsigned = unsignedGaussCode(len(petal_permutation))
signed = signGaussCode(unsigned, petal_permutation)
crossing_matrix = createCrossingMatrix(signed, petal_permutation)
return evaluateKnotDeterminant(crossing_matrix)

#This function presents what the knot determinant implies about the possible colorings of
,→ the knot
def presentDeterminant(determinant):
print("The knot’s determinant is " + str(determinant) + ".")
factorization = factorint(determinant)
for i in factorization:
print("Since " + str(i) + " appears " + str(factorization[i]) + " time(s) in the
,→ prime factorization of " + str(determinant) + ", there are " + str(i) + "^(
,→ " + str(factorization[i]) + "+1) - " + str(i) + " = " + str(pow(i,
,→ factorization[i]+1)-i) + " nontrivial " + str(i) + "-colorings of the knot.
,→ ")
print("These are all nontrivial colorings of the knot, though there are p trivial
,→ colorings for every prime p.")

To test it, we use the following code:

Listing 2. Code for testing the algorithm by using the Stevedore knot.
#Here, we import the code from Listing 1.
import spp_library

#The petal permutation of the Stevedore knot.

petal_permutation = [1, 3, 5, 2, 8, 4, 6, 9, 7]
petal_number = len(petal_permutation)
unsigned = spp_library.unsignedGaussCode(petal_number)
signed = spp_library.signGaussCode(unsigned, petal_permutation)
arcs = spp_library.splitSignedGaussCode(signed)
crossing_matrix = spp_library.createCrossingMatrix(signed, petal_permutation)
knotDeterminant = spp_library.evaluateKnotDeterminant(crossing_matrix)
 PETAL PROJECTIONS, KNOT COLORINGS & DETERMINANTS 23

print("")
print("You asked for analysis of the knot with petal permutation " + str(
,→ petal_permutation))
spp_library.presentDeterminant(knotDeterminant)
print("")
We also include the following code, which we used with minor modifications in the calculations for our
other appendices. This code tests a list of petal projections and returns their knot determinant and the
runtime (in ms) of each calculation.
Listing 3. Code for testing the algorithm on sets of petal projections
#Here, we import the code from Listing 1 and an algorithm timing library in Python.
import spp_library
import timeit
petal_projections_to_test = [[1, 3, 5, 2, 4],[1, 3, 5, 2, 7, 4, 6],[1, 3, 5, 2, 8, 4, 6,
,→ 9, 7]]
number_of_runs_each = 100

def mean(runtimes):
return sum(runtimes) / len(runtimes)

for petal_projection in petal_projections_to_test:

#Compute the knot determinant
petal_projection_determinant = spp_library.knotDeterminant(petal_projection)

alg_runtimes = []
#Run the petal projection calculation number_of_runs_each times and get a runtime
,→ list
for i in range(number_of_runs_each):
start = timeit.default_timer()
spp_library.knotDeterminant(petal_projection)
alg_runtimes.append(1000*(timeit.default_timer() - start))

#Print everything
print("Petal Projection: " + str(petal_projection) + " | " + str(
,→ petal_projection_determinant) + "&" + str(round(min(alg_runtimes),4)) + "&" +
,→ str(round(mean(alg_runtimes),4)))
Note: all tests were conducted using the same computer, which uses a Ryzen 5 1600 CPU and GTX 1060
graphics card. Given an optimized and/or more powerful computer, a quicker language (ex. C/C++), or
both, all of these runtimes could be significantly reduced. Additionally, this code is not optimized to make
use of multithreading – an implementation which did take advantage of multithreading could be orders of
magnitude faster.

Appendix B. The p-colorability of All Prime Knots with Fewer Than 10 Crossings
Again, we source our list of the prime knots with fewer than 10 crossings in petal projection from [3]. The
minimum and average runtimes are collected from a total of 100 attempts per knot in the list.
24 ALLISON HENRICH AND ROBIN TRUAX

Knot p(K) A minimal petal representation Determinant Min. runtime (ms) Avg. runtime (ms)
31 5 (1, 3, 5, 2, 4) 3 0.0737 0.0769
41 7 (1, 3, 5, 2, 7, 4, 6) 5 0.1931 0.1990
51 7 (1, 3, 6, 2, 5, 7, 4) 5 0.1928 0.2009
52 7 (1, 3, 6, 2, 4, 7, 5) 7 0.1992 0.2040
61 9 (1, 3, 5, 2, 8, 4, 6, 9, 7) 9 0.5329 0.5821
62 9 (1, 3, 5, 2, 8, 4, 7, 9, 6) 11 0.5295 0.5968
63 9 (1, 3, 5, 2, 9, 7, 4, 8, 6) 13 0.5341 0.6007
71 9 (1, 8, 4, 9, 5, 3, 6, 2, 7) 7 0.5417 0.5917
72 9 (1, 3, 6, 9, 7, 2, 4, 8, 5) 11 0.5366 0.5845
73 9 (1, 3, 6, 9, 7, 2, 5, 8, 4) 13 0.5317 0.5887
74 9 (1, 3, 6, 4, 8, 2, 5, 9, 7) 15 0.5298 0.5768
75 9 (1, 3, 6, 4, 8, 2, 7, 9, 5) 17 0.5414 0.5953
76 9 (1, 3, 6, 4, 9, 7, 2, 8, 5) 19 0.5343 0.6008
77 9 (1, 3, 7, 9, 4, 6, 2, 8, 5) 21 0.5268 0.5884
81 11 (1, 3, 5, 2, 8, 11, 9, 4, 6, 10, 7) 13 1.5281 2.9710
82 11 (1, 3, 5, 2, 9, 4, 7, 11, 8, 10, 6) 17 1.5175 2.8166
83 11 (1, 3, 6, 2, 9, 5, 11, 4, 7, 10, 8) 17 1.5194 2.8753
84 11 (1, 3, 5, 8, 6, 2, 10, 4, 7, 11, 9) 19 1.4199 3.0731
85 11 (1, 3, 5, 8, 6, 11, 9, 2, 10, 4, 7) 21 1.5375 3.0680
86 11 (1, 3, 5, 2, 8, 6, 10, 4, 9, 11, 7) 23 1.5763 3.8458
87 11 (1, 3, 5, 2, 10, 7, 4, 8, 11, 9, 6) 23 1.5653 3.5296
88 11 (1, 3, 5, 2, 8, 6, 11, 9, 4, 10, 7) 25 1.4279 3.3398
89 11 (1, 3, 5, 9, 2, 7, 11, 6, 4, 10, 8) 25 1.6110 3.6660
810 11 (1, 3, 5, 2, 9, 7, 11, 8, 4, 10, 6) 27 1.5062 3.5207
811 11 (1, 3, 5, 2, 8, 11, 9, 4, 7, 10, 6) 27 1.6855 3.5786
812 11 (1, 3, 5, 2, 9, 11, 8, 6, 10, 4, 7) 29 1.4598 3.4152
813 11 (1, 3, 5, 2, 10, 7, 4, 9, 11, 8, 6) 29 1.5943 3.3468
814 11 (1, 3, 5, 2, 10, 8, 11, 6, 9, 4, 7) 31 1.8390 3.8781
815 11 (1, 3, 5, 2, 8, 11, 7, 9, 4, 10, 6) 33 1.5407 3.5597
816 11 (1, 3, 5, 8, 6, 2, 11, 9, 4, 10, 7) 35 1.6293 3.7273
817 11 (1, 3, 5, 8, 6, 2, 10, 4, 9, 11, 7) 37 1.4885 3.4082
818 11 (1, 3, 7, 4, 10, 2, 8, 6, 11, 9, 5) 45 1.6567 3.6826
 PETAL PROJECTIONS, KNOT COLORINGS & DETERMINANTS 25

Knot p(K) A minimal petal representation Determinant Min. runtime (ms) Avg. runtime (ms)
819 7 (1, 4, 7, 3, 6, 2, 5) 3 0.1811 0.1994
820 9 (1, 3, 5, 8, 2, 6, 9, 4, 7) 9 0.5331 0.5799
821 9 (1, 3, 5, 8, 2, 7, 4, 9, 6) 15 0.5282 0.5863
91 11 (1, 10, 5, 11, 6, 4, 7, 3, 8, 2, 9) 9 1.5855 3.3514
92 11 (1, 3, 6, 10, 7, 2, 4, 8, 11, 9, 5) 15 1.4888 3.0756
93 11 (1, 3, 7, 5, 9, 2, 6, 11, 8, 10, 4) 19 1.4289 3.0935
94 11 (1, 3, 6, 10, 7, 2, 5, 8, 11, 9, 4) 21 1.5295 3.4236
95 11 (1, 3, 6, 4, 8, 11, 9, 2, 5, 10, 7) 23 1.4775 2.4631
96 11 (1, 3, 6, 4, 9, 2, 7, 11, 8, 10, 5) 27 1.4985 2.6101
97 11 (1, 3, 6, 10, 7, 2, 4, 9, 11, 8, 5) 29 1.6424 3.4747
98 11 (1, 3, 6, 10, 8, 4, 11, 7, 2, 9, 5) 31 1.5704 3.2222
99 11 (1, 3, 6, 10, 7, 2, 5, 9, 11, 8, 4) 31 1.4988 3.2292
910 11 (1, 3, 7, 5, 8, 11, 9, 2, 6, 10, 4) 33 1.4525 3.3888
911 11 (1, 3, 6, 4, 10, 7, 2, 8, 11, 9, 5) 33 1.6525 3.2743
912 11 (1, 3, 6, 10, 5, 7, 2, 8, 11, 9, 4) 35 1.6505 3.5714
913 11 (1, 3, 6, 4, 9, 2, 5, 11, 8, 10, 7) 37 1.5017 3.437
914 11 (1, 3, 7, 10, 5, 2, 9, 11, 8, 4, 6) 37 1.6414 3.6194
915 11 (1, 3, 6, 4, 10, 8, 2, 7, 11, 9, 5) 39 1.5063 3.4034
916 11 (1, 3, 7, 4, 10, 2, 9, 11, 6, 8, 5) 39 1.6249 3.6076
917 11 (1, 3, 7, 10, 4, 6, 2, 9, 11, 8, 5) 39 1.5607 3.9702
918 11 (1, 3, 6, 4, 8, 11, 9, 2, 7, 10, 5) 41 1.4809 3.4053
919 11 (1, 3, 7, 5, 9, 11, 4, 8, 2, 10, 6) 41 1.5197 3.4281
920 11 (1, 3, 6, 4, 10, 8, 2, 9, 5, 11, 7) 41 1.4670 3.3930
921 11 (1, 3, 6, 4, 10, 7, 2, 9, 11, 8, 5) 43 1.4145 3.1027
922 11 (1, 3, 6, 4, 9, 7, 2, 10, 5, 11, 8) 43 1.7543 3.6143
923 11 (1, 3, 6, 4, 9, 11, 7, 2, 8, 10, 5) 45 1.5742 3.742
924 11 (1, 3, 6, 11, 5, 7, 2, 9, 4, 10, 8) 45 1.5370 3.3379
925 11 (1, 3, 6, 4, 8, 11, 7, 9, 2, 10, 5) 47 1.5600 3.6257
926 11 (1, 3, 7, 5, 10, 6, 2, 9, 11, 4, 8) 47 1.6188 3.7474
927 11 (1, 3, 6, 4, 11, 7, 2, 8, 10, 5, 9) 49 1.7437 3.6646
928 11 (1, 3, 6, 11, 5, 7, 2, 8, 10, 4, 9) 51 1.6539 3.8004
929 11 (1, 3, 6, 4, 10, 7, 2, 8, 5, 11, 9) 51 1.5392 3.9865
930 11 (1, 3, 6, 4, 10, 8, 2, 7, 11, 5, 9) 53 1.7512 4.2598
931 11 (1, 3, 6, 10, 5, 7, 2, 9, 11, 4, 8) 55 1.4685 3.7721
932 11 (1, 3, 6, 4, 9, 11, 7, 2, 10, 5, 8) 59 1.6178 3.5188
933 11 (1, 3, 6, 4, 10, 2, 7, 11, 9, 5, 8) 61 1.4888 3.6090
934 13 (1, 3, 7, 9, 13, 5, 11, 8, 2, 4, 6, 10, 12) 69 4.5353 11.4473
935 11 (1, 3, 10, 6, 2, 9, 11, 8, 5, 7, 4) 27 1.4939 2.7999
936 11 (1, 3, 6, 4, 9, 7, 11, 8, 2, 10, 5) 37 1.5997 4.0018
937 11 (1, 3, 7, 10, 4, 6, 2, 8, 11, 9, 5) 45 1.7504 3.8878
938 11 (1, 3, 6, 4, 9, 2, 7, 11, 5, 10, 8) 57 1.6299 3.6175
939 11 (1, 3, 6, 4, 10, 2, 7, 9, 5, 11, 8) 55 1.5260 3.7586
940 13 (1, 11, 7, 5, 13, 2, 10, 8, 6, 12, 4, 9, 3) 75 5.2471 12.1151
941 11 (1, 3, 7, 11, 4, 8, 10, 6, 2, 9, 5) 49 1.5291 3.5573
942 9 (1, 3, 6, 2, 9, 5, 8, 4, 7) 7 0.5351 0.5776
943 9 (1, 3, 6, 9, 5, 8, 2, 7, 4) 13 0.5262 0.6009
944 9 (1, 3, 6, 9, 4, 7, 2, 8, 5) 17 0.5248 0.5718
945 9 (1, 3, 7, 4, 9, 6, 2, 8, 5) 23 0.5282 0.5999
946 9 (1, 3, 6, 9, 5, 2, 8, 4, 7) 9 0.5362 0.6032
947 11 (1, 3, 5, 7, 10, 4, 9, 6, 2, 11, 8) 27 1.5552 3.4711
948 11 (1, 3, 5, 2, 9, 11, 7, 4, 10, 6, 8) 27 1.6477 3.7317
949 11 (1, 3, 5, 2, 7, 11, 8, 4, 10, 6, 9) 25 1.5622 3.8258
26 ALLISON HENRICH AND ROBIN TRUAX

Appendix C. Finding the Limits of the Algorithm

First, we test our program 1000 times against the petal projection (1, 2, . . . , n) for various (odd) n. Note
that the result will always be 1, as this petal projection degenerates into the unknot via petal cancellation.

n Min. runtime (ms) Avg. runtime (ms)

11 1.3715 2.0532
25 34.1978 60.7494
51 501.0886 549.8304
75 2255.2766 2341.1118
101 7693.6962 7919.2274
Next, we test our program against 1000 random petal permutations with petal number n – that is, random
elements of Sn for various (odd) n. Here we expect a variety of determinants, but we cannot list all of them here.

n Min. runtime (ms) Avg. runtime (ms)

11 1.5544 3.7661
25 45.0358 89.2748
51 513.6894 576.6921
75 2320.5901 2425.1528
101 8086.1688 8339.2420
Clearly, this algorithm suffices to compute the determinant of any “reasonably-sized” petal projection in a
fairly short amount of time.

Appendix D. Distribution of p-colorability Amongst Randomly Generated Petal

Projections
Finally, we will offer an empirical description of the proportion of petal permutations with petal number
n that are p-colorable for p ≤ 23. For example, the n = 5, p = 3 entry is 8.3%, so approximately 8.3% of
the 5-petal projections are 3-colorable. The “NC” column refers to the approximate percentage of petal
permutations with petal number n that are not p-colorable for any p.

For sufficiently small n (i.e. n such that n! < 1000), we simply test every permutation of {1, . . . , n}.
Otherwise, we select 1000 random permutations of {1, . . . , n}, find the knot determinant of the corresponding
petal projection, and calculate what percentage are p-colorable for each p that appears in the prime
factorization of any of them. This is a very rough approximation: we simply use this for intuition and to
generate further questions related to petal projection determinants.
 PETAL PROJECTIONS, KNOT COLORINGS & DETERMINANTS 27

n NC 3 5 7 11 13 17 19 23
1 100 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
3 100 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
5 91.7 8.3 0.0 0.0 0.0 0.0 0.0 0.0 0.0
7 64.5 22.4 6.3 5.0 0.2 0.2 0.0 0.0 0.0
9 50.2 25.0 10.2 7.3 2.7 1.1 0.6 0.2 0.2
11 38.8 30.0 16.0 7.6 4.6 2.6 1.7 1.1 0.8
13 24.6 32.3 17.6 11.6 5.2 4.1 2.8 2.1 1.4
15 20.3 37.0 20.3 12.7 6.4 4.7 4.3 3.0 2.4
17 11.8 33.1 17.7 14.1 8.6 6.5 4.9 4.6 3.6
19 8.7 34.5 20.1 14.6 8.1 8.4 4.2 4.6 2.5
21 5.5 36.2 20.3 16.3 8.0 6.5 5.8 4.1 3.1
23 2.7 37.5 21.2 15.2 9.4 8.2 5.1 4.9 4.6
25 1.8 35.7 22.3 12.5 8.4 6.4 4.3 4.1 3.7
27 0.5 35.6 22.2 15.8 9.2 8.2 5.3 5.1 5.7
29 0.5 37.2 21.6 16.9 8.8 8.3 6.5 5.8 4.1
31 0.2 35.4 19.1 17.4 8.4 8.3 6.0 5.1 4.0
33 ∼ 0.0 35.4 21.1 12.0 9.6 9.0 6.0 5.4 4.6
35 ∼ 0.1 37.0 21.5 15.2 9.5 8.3 5.3 4.8 4.0
37 ∼ 0.0 36.8 22.2 15.3 10.8 5.9 5.5 3.9 4.8
39 ∼ 0.0 35.9 20.6 16.7 9.5 7.2 6.5 3.8 4.6
41 ∼ 0.0 34.8 19.5 16.0 10.5 6.3 5.4 6.1 4.9
43 ∼ 0.0 38.3 21.5 13.8 8.9 9.3 5.6 5.1 4.4
45 ∼ 0.0 35.7 20.1 13.5 8.4 6.7 6.1 5.0 4.2
47 ∼ 0.0 35.1 20.7 13.5 8.5 8.3 5.3 5.2 3.7
49 ∼ 0.0 36.7 22.2 14.2 10.1 8.2 5.5 5.6 4.2
Department of Mathematics, Seattle University, 901 12th Avenue, Seattle, WA 98122
Email address: [email protected]

Stanford University, Stanford, CA 94305

Email address: [email protected]