Vous êtes sur la page 1sur 11

Co-parallelogram Rule for One-forms

by Gabe Czobel
The parallelogram rule for the addition of vectors, represented as arrows in 2 dimensions, is a well
known visualisation aid. The magnitude of a vector is represented as the length of the arrow and,
of course, its direction is self-evident. The parallelogram rule is entirely intuitive when the vectors
are taken as displacements. It is clear that the total displacement of two displacement vectors is
along the diagonal of the parallelogram, from the tail of one vector to the head of the other.
But what of the addition of one-forms or covectors? Typically, these are not represented as arrows
but as a series of surfaces. The magnitude of a one-form is inversely proportional to the
separation of the surfaces in this representation. That is, the closer the surfaces are to each other, the
larger the represented one-form. There is no self-evident direction to the one form, although the
measure of the separation, being taken perpendicular to the surfaces, makes some sense as a
direction characteristic. In addition, the sense of positive direction along this perpendicular is
intuitive when considering one-forms as gradients of scalar fields, where the positive direction is
along the maximum increase in the gradient.
The following diagram depicts the vector parallelogram rule in part (a) and the depiction of oneforms as surfaces in part (b). The little notches on the surfaces indicate the direction of increase and
the perpendicular a depicts the separation of the surfaces, the inverse of which is the magnitude of
the one-form.

The addition of one-forms can be represented geometrically as a co-parallelogram rule. This rule is
not intuitive at all, especially considering that the magnitude of each one-form is represented by the
inverse of the separation of the surfaces. Adding inverse magnitudes is not geometrically intuitive.
For vectors, the special case where the vectors are collinear results in a degenerate parallelogram,
but simply placing the vectors tail to head, one can still easily add (or subtract) the lengths. The
corresponding situation for one-forms, where the surfaces of the two one-forms being added are
parallel, calls for special treatment.
The following diagrams represent a general visualisation for the addition of one-forms where the
angle between their positive directions ranges from greater than 0 to less than . For angles that
are greater than , the symmetric case holds where the diagram is flipped horizontally. We will
look at the special collinear case later.

(a) Angle between directions is acute.

(b) Perpendicular angle between directions.

(c) Angle between directions is obtuse.


The co-parallelogram rule in the preceding diagrams goes like this. Take the intersection of two of
the surfaces, one from each of the one-forms being added, such as along OR for a and OP for b
intersecting at point O . Take the next surface in the positive direction for each one-form, such as
along PQ and QR , and form the parallelogram OPQR . The surfaces for the resulting one-form
c = a + b are those that pass through the diagonal PR opposite to vertex O , pass through vertex O
parallel to the diagonal PR , and others along corresponding diagonals in the lattice of
parallelograms formed from the surfaces. The positive direction is from vertex O to the diagonal
PR .
To justify that this co-parallelogram rule actually represents the addition of two one-forms in 2
dimensions, we need to go through a number of simple geometrical arguments that confirm that the
one-form that represents the sum of the other two has the correct magnitude and positive
direction according to its surfaces.

The first step is to confect some manner of adding inverse magnitudes geometrically. One
geometric technique that assists in accomplishing this is the inversion of points in a circle. The
technique is shown in the diagrams below.

The diagrams show two cases where a point P is inverted in a circle of radius r to find the inverse
point P' such that the relationship OP . OP' = r2 . As shown in the diagram, this can be constructed
by means of ruler and compass for the cases where P is either inside or outside the circle. By
means of similar isosceles triangles OPS and OSP', it may be shown that OP/OS = OS/OP' or
OP . OP' = OS2 = r2

For the case where OP < r/2 and S can't be obtained, a slightly more complex technique, still using
ruler and compass, is available. The important fact for our purposes is that OP . OP' = r2 always
holds.
This fits the bill perfectly since the distance of the inverse point P' from the center of the circle is
proportional to 1/distance of P from the center. This converts inverse magnitudes into direct
magnitudes which can then be easily added geometrically. Even more, the inverse of the inverse
point itself is the original point. Hence, direct magnitudes can be converted back to inverse ones.
This is very much like the duality relationship between vectors and covectors in that there is a
unique vector that is the dual of a covector and this relationship is bidirectional.
The radius of the circle is immaterial in this method as it only determines a constant of
proportionality between a point and its inverse. Inverting a particular point and then inverting the
result, still gets back to the original point, regardless of the radius of the circle used in the inversion,
as long as the same circle is used going in both directions.
Now we can use inversion in a circle to geometrically add two covectors by first converting them to
their dual vectors, adding the dual vectors using the parallelogram rule, and inverting the vector
result back into its dual covector. This process is shown in the following diagrams for both acute
and obtuse angles between the covectors.

We see two one-forms a and b represented by two surfaces each. The magnitude of a is the inverse
of the perpendicular distance OP between the surfaces and for b it is the inverse of OQ. We choose
a circle of radius OP for inversion of points P and Q, so that OQ . OQ' = OP . OP' = OP2 . Having
performed these inversion, we get vectors OP' and OQ' as duals of the one-forms a and b
respectively. Adding these vectors using the parallelogram rule, we get vector OR' and note the
alternate angles and at vertices O and R' as shown.
To get the direction and magnitude of the one-form that is the sum of a and b , we invert point R' at
the tip of vector OR' in the circle to get point R . Then OR is the inverse magnitude of the oneform c = a + b and the surfaces of c may be taken as perpendicular to OR as shown. We have thus
constructed the surfaces of the one form c = a + b geometrically. But this method is tedious.
Let's see what relationships we can derive about the construction. OR . OR' = OP2 on account of
the inversion. Hence, OR'/OP = OP/OR which, along with the common angle , implies that
OPR' is similar to ORP . Hence angle OPR is equal to as shown. Since OQ . OQ' = OP2
then OQ . OQ' = OR . OR' . Hence, OQ/OR = OR'/OQ' . With the common angle , this implies
that OQ'R' is similar to ORQ and angle OQR is equal to as shown.

In the next step, we dispense with the cumbersome geometric inversion in a circle and use a simple
technique of intersecting diagonals of parallelograms. Since OQ . OQ' = OP2 above, then
OQ/OP = OP/OQ' . Looking at parallelogram OLMN in the following diagram, where ON = OQ,
and OL = OP, and using OQ/OP = OP/OQ', we conclude that ON/OL = OP/OQ' where OQ'
relates to the parallelogram OQ'R'P in the above diagrams . Given the common angle Q'OP is the
same as angle LON, the parallelogram OQ'R'P is similar to the parallelogram OLMN. By
similarity, angle LOM is , and angle MON is . Rotating
parallelogram OLMN
counterclockwise until LO is colinear with OP and translating to the right so that vertex L coincides
with vertex O , we get parallelogram POM'N' with the corresponding angles and as shown.
With the intersection of the diagonals of these parallelograms at point R we have managed to
recreate ORP as on the diagram for inversions in a circle. In the obtuse case, we actually need to
extend diagonal PM' to find the intersect point R.And thus we have c = a + b as before.
The parallelograms are easily formed using the perpendiculars between the surfaces of the oneforms, OQ and OP, that are a measure of the inverse of the one-form magnitudes as stated earlier.

Finally, we can show that the technique of intersecting parallelogram diagonals is itself equivalent
to the co-parallelogram rule presented first above where we used only parallelograms formed from
the surfaces representing the covectors, without the need to involve the perpendiculars between the
surfaces.
In the following diagram the upper part shows the same two parallelograms as in the diagonal
intersection method above. A perpendicular OS is taken to the diagonal M'P . In the right OSP
we have + + = /2 . The lower part of the diagram depicts the co-parallelogram rule. We
again have the perpendiculars OQ and OP between the one-form surfaces, corresponding to the
same line segments in the upper part.
It is clear that angles AOQ and POC are equal to as + + = /2 again. Hence angles OAQ
and PCO equal + . Now OA = OQ/cos and OQ = OM', hence OA = OM'/cos . Also, OC =
OP/cos . Thus, parallelogram OABC is similar to parallelogram OM'N'P. Since diagonal AC
corresponds to diagonal M'P we also have the corresponding angles and as shown at vertex A
and C.

If we drop a line segment OR' to diagonal AC with the angles and at vertex O as shown, then
we know that AR'O is a right triangle since + + = /2, or OR' is a perpendicular to
diagonal AC. And by the corresponding angles and , line segment OR' is in the same direction
as line segment OR. Since OR is the perpendicular between the surfaces of the one-form a + b as
derived before, we just need to show that OR = OR' to establish the desired equivalence.
In OSR, OR = OS/cos . But OS = OM' sin = (OA cos ) sin = ([OR'/ sin ] cos ) sin =
OR' cos
Hence OR = OS/cos = (OR' cos )/cos = OR', as needed.

When the covectors are collinear, no parallelograms can be formed and the co-parallelogram rule
breaks down. In this case, the earlier technique of inverting points in a circle still allows a
geometrical solution.

2015 by Gabe Czobel

Vous aimerez peut-être aussi