Dieser Abschnitt ist noch im Entstehen und noch nicht offizieller Bestandteil des Buchs. Gib den Autoren Zeit, den Inhalt anzupassen!
We have already seen the vector space of linear maps between two -vector spaces and . We will now consider the case where the vector space corresponds to the field .
Consider the following example: We want to buy apples and pears. An apple costs $ and a pear $. If is the number of apples and is the number of pears, how much do we have to pay in total? The formula for the total price is . We can express this equation as -linear map
Let's assume that the prices increase by half. To get the formula that gives the new total price, we need to multiply the old formula by . The formula that gives this price would then be . The corresponding linear map is
We thus recognize that .
Suppose now that the price of apples increases by $ and the price of pears by $. We obtain the corresponding formula for the total price by adding to the original formula, i.e. . This can be understood as the addition of linear maps. We define by and . Then holds true. So in this example, we simpy added linear maps from to and multiplied them by scalars.
The total price is indicated by linear maps from . Such a map assigns a value, namely the price, to each vector. In other words, we can say that the mapping "measures" these vectors. This is why we call linear maps from to linear measurement functions. We have seen above that sums and scalar multiples of such maps are again linear maps. In other words, linear combinations of linear maps are again linear maps. So also on the set of linear maps on , we can find a vector space structure.
What about other vector spaces? Let's look at the -vector space of complex polynomials of degree at most . There are a number of simple measurement functions here. These can, for example, assign to a polynomial its value at a point :
Alternatively, we can assign to a polynomial the value of its derivative at the point :
Since the coefficients of polynomials are scalars, we can use them to define further measurement functions. For example, for , consider the mappings defined by and . Then . We can also see here that sums of measurement functions are again measurement functions.
In general, we can also consider the space of linear measurement functions over an arbitrary -vector space . We will see that, as in the previous examples, this is a vector space. This space is called the dual space of .
The following theorem states that the dual space is a vector space.
Theorem ( is a vector space)
Let be a vector space over a field . Then with the two relations
and
a -vector space.
Examples of vectors in the dual space[Bearbeiten]
Example (Characterization of )
The dual space of is the vector space of all linear maps from to . Each such linear map is given by multiplication with a (1x2) matrix, the representing matrix, and is therefore of the form
for certain . Thus, the elements in the dual space of are described by linear equations of the form .
More generally, an element of is represented by a (1xn) matrix or a linear equation of the form with coefficients .
Example (Integral)
Let be the space of continuous functions . Consider the mapping
which sends a continuousfunction on to its integral over this interval. As an example, for ,
We verify by direct calculation that the mapping is linear: For and the following applies
This follows from the properties of the integral. So is an element of .
We now know what the dual space of a -vector space is: It consists of all linear maps from to . Intuitively, we can understand these maps as linear maps that measure vectors from . This is why we sometimes call elements of the dual space "(linear) measurement functions" in this article.
Motivated by this intuitive notion of "measurements", we ask ourselves: Is there a subset of measurement functions that can be used to uniquely determine vectors? In other words, is there a subset so that we can find a measurement function with for every choice of vectors with ?
Let's first consider what this means using an example:
Example (Unique determination of vectors using measurement functions)
Let us consider . Then the dual space is the vector space of all linear maps . Consider the linear maps with
If , we cannot use these functions to determine vectors uniquely: For and , we have , but .
Even with the measurement functions in , the vectors and cannot be distinguished: We also have .
However, if we consider the subset of measurement functions instead, then vectors in are uniquely determined by the measurements in : Let and be any vectors with . Assume that and apply. From we obtain . Together with , we would then also get , i.e. . This would mean that , which is a contradiction to our assumption. Therefore, or (or both) applies. Hence, for each choice of different vectors in , at least one of the two measurements in provides different values for and . Vectors are therefore uniquely determined by the measurements in .
In sumary, our question is: Does there exist a subset such that applies to all vectors: If applies to all measurements , then must be true.
We will first try to answer this question in .
Measurement functions for unique determination of vectors[Bearbeiten]
A vector is uniquely determined by its entries . If we select measurement functions from in such a way that their values provide us with the entries of a vector, then we have ensured that a vector is already uniquely determined by these values. Let us therefore consider the following mappings for
You can check that the maps are linear. In addition, holds for every . The map therefore provides the -th entry of vectors in . A vector is already uniquely determined by the values of : Suppose we have vectors and in with equal function values among the , i.e., with for all . Then applies for all and therefore . Thus, if with for all , then follows.
It is also intuitively clear that we cannot omit any of the measurement functions in order to uniquely determine a vector by its measurement values. For example, if we omit , , then for
we may have for all measurement functions with , but nevertheless . The measurement functions with therefore no longer uniquely determine a vector.
So the with form a set of measurement functions that uniquely determine vectors from . Further, they are minimal because we cannot omit any of the functions.
Can we generalize this to a general vector space ? In we have used the fact that a vector is uniquely determined by its entries . Now, the are precisely the coordinates of with respect to the standard basis :
In a general vector space , we do not have a standard basis. However, as soon as we have chosen any basis , we can speak of the coordinates of a vector with respect to in the same way as in . Just as in with the standard basis, in with the selected basis , a vector is uniquely determined by its coordinates with respect to . As soon as we have chosen a basis, we can try to proceed in the same way as in .
In the following, we assume that is finite-dimensional, i.e. . Let be a basis of . Then every vector is of the form
with uniquely determined coordinates . Analogous to , we now define the linear measurement functions for in
One of the measurement functions therefore determines the -th coordinate of vectors with respect to the basis . Thus,
for every vector .
Warning
Note that the definition of depends on the selected basis .
Since vectors in are already uniquely determined by their coordinates, they are also already uniquely determined by the values of . In other words, for all we have
For the same reason as with , none of the can be omitted: If the -th measurement function , , is missing, then any two vectors for which only the -th coordinate with respect to differs, can no longer be distinguished.
Question: Which two vectors can you choose here?
We choose an example analogous to and set
and
Then holds for all , but nevertheless . If the -th measurement function is omitted, then vectors are no longer uniquely determined by the function values of .
The measurement functions form a basis[Bearbeiten]
Let be a vector space with a fixed basis and let the be defined as above. If you want to determine vectors uniquely using the values of , you cannot do without any of the . The reason for this is that the result of a measurement (the -th coordinate of with respect to ) cannot be deduced from the other measurements. That means, we cannot represent any of the measurement functions as a linear combination of the other (). In other words, the measurement functions are linearly independent.
On the other hand, the values of already tell us everything there is to know about a vector : Its coordinates with respect to the selected basis . Can all other measurement functions from therefore be combined from ? Any measurement function from is already uniquely determined by its values on the basis vectors according to the principle of linear continuation. For , let be these values. Furthermore, and apply for and all . By inserting the we obtain that
assume the same values on the basis vectors. According to the principle of linear continuation, the two linear maps are therefore identical. Thus, every can be written as a linear combination of . In other word, the measurement functions form a generating system of .
Hence, is a basis of the dual space and we can prove the following theorem:
Theorem (Existence of a dual basis)
Let be a finite dimensional vector space and a basis of . Then there exists a unique basis of such that
is true for all .
Proof (Existence of a dual basis)
Proof step: Existence and uniqueness of the .
According to the principle of linear continuation, the linear maps exist and are uniquely determined by their values on the basis vectors of .
Proof step: The are linearly independent.
Let with . Let further . Because and for , we obtain the following by plugging in :
Because was arbitrary, we conclude .
Proof step: The form a generating system.
Let be arbitrary. For we define and set . Then, proceeding as in the proof of linear independence, we obtain
for each . Because applies to all and because a linear map is already uniquely determined by the images of its basis vectors, we have . The therefore form a generating system.
We call the uniquely determined basis the dual basis with respect to and denote its basis vectors by .
Warning
Note that depends on the basis chosen for . Furthermore, you cannot "dualize" individual vectors from , but only entire bases.
What happens in the infinite dimension?[Bearbeiten]
Above, we only considered the case . Can we proceed analogously if is infinite dimensional? To define the measurement functions , we must first choose a basis of . Let be a basis of , where is an (infinite) index set. The principle of linear continuation also applies in infinite dimensions: For given values , , there is exactly one linear map with for all . Just as in the finite-dimensional case, we can therefore define the map for using the rule
We can then show that is also a linearly independent subset of in infinite dimensions. The proof is analogous to the proof of linear independence in the theorem on the dual basis.
However, in infinitely many dimensions, cannot be a generating system of : One can consider the function
which assumes the value 1 on all basis vectors. This function cannot be represented as a finite linear combination of .
So in infinitely many dimensions, the "dual basis" is not a basis of the dual space.
How to get to the proof? (Duale Basisvektoren und ihre Kerne bestimmen)
Wir müssen eine lineare Abbildung konstruieren. Das ist genau ein Element von . Nach dem Prinzip der linearen Fortsetzung, können wir lineare Abbildungen konstruieren, indem wir angeben, was sie auf einer Basis tun.
Um das zu nutzen, ist es praktisch eine Basis von zu haben. Noch praktischer ist es, eine Basis von zu haben, die als Basisvektor enthält.
Eine solche Basis können wir mithilfe des Basisergänzungssatzes konstruieren: Nach dem Basisergänzungssatz hat eine Basis mit . Damit können wir mit dem Prinzip der linearen Fortsetzung eine lineare Abbildung konstruieren, die nicht auf schickt. Zum Beispiel können wir das wählen, das alle auf schickt und für auf .
Das ist genau der duale Basisvektor der dualen Basis zu .
Exercise (Duale Basis bestimmen)
- Betrachte die Basis von . Bestimme die zu duale Basis , d.h. bestimme für die explizite Funktionsvorschrift
- Betrachte die Basis von . Bestimme die zu duale Basis , d.h. bestimme für die explizite Funktionsvorschrift
- Betrachte die Basis von . Bestimme die zu duale Basis , d.h. bestimme für die explizite Funktionsvorschrift
Solution (Duale Basis bestimmen)
Solution sub-exercise 1:
Setze , und . Wir suchen lineare Abbildungen , deren Werte wir nur auf den Basisvektoren kennen. Wir müssen für allgemeine definieren.
Per Definition der dualen Basis kennen wir schon die Funktionswerte jedes auf den Basisvektoren in . Nach dem Prinzip der linearen Fortsetzung können wir daraus alle Funktionswerte bestimmen: Weil eine Basis ist, gibt es für jedes Koordinaten sodass . Mithilfe der Linearität folgt
Die Werte kennen wir per Definition der dualen Basis. Wir müssen also nurnoch die Koordinaten eines beliebigen Vektors bzgl. bestimmen. Danach können wir die hinschreiben.
Proof step: Koordinaten eines beliebigen Vektors bzgl. bestimmen
Wir wollen die Koordinaten bzgl. von einem beliebigen Vektor bestimmen. Seien also . Wir schreiben
Die Koordinaten von bzgl. der Standardbasis sind also einfach , und . Wenn wir für die Koordinatenabbildung schreiben, bedeutet das
Wir können diese in Koordinaten bzgl. umrechnen, indem wir den Koordinatenvektor bzgl. von links mit der Basisübergangsmatrix von nach multiplizieren. Es gilt also
Um die Basisübergangsmatrix zu bestimmen, berechnen wir die Koordinaten der Standardbasisvektoren bzgl. . Diese bilden die Spalten von .
Wir beginnen mit : Wir suchen sodass
gilt. Wir lösen also das lineare Gleichungssystem
und erhalten , und . Genauso bestimmen wir die Koordinaten von bzgl. und die Koordinaten von bzgl. . Also gilt
Beachte: Wir hätten auch alle drei Gleichungssysteme auf einmal lösen können, indem wir die "rechten Seiten" spaltenweise zusammenfassen, d.h. indem wir die Inverse von
bestimmen. Das macht Sinn, denn diese Matrix ist die Basiswechselmatrix von in die Standardbasis. Ihre Inverse ist somit die gesuchte Basisübergangsmatrix von nach .
Die Koordinaten von bzgl. sind also
Es ist natürlich auch in Ordnung, die Koordinaten von bzgl. durch genaues Hinsehen zu erraten, ohne Gleichungssysteme zu lösen.
Proof step: Ergebnis für
Wir können nun ein beliebiges schreiben als
Mit der Linearität der und der Definition der dualen Basis erhalten wir
Genauso berechnen wir und . Insgesamt haben wir also die drei Basisvektoren der dualen Basis bestimmt:
Solution sub-exercise 2:
Wir wissen, was die Abbildung auf den Basisvektoren macht. Um herauszufinden, wie die auf einem allgemeinen Vektor agiert, können wir ihn in der Basis ausdrücken:
Damit können wir die Funktionsvorschriften ausrechnen. Für haben wir
Für bekommen wir
Die Funktionsvorschrift von ist
Für erhalten wir
Zusammengefasst erhalten wir für die Funktionsvorschriften
Solution sub-exercise 3:
Wir kennen die Werte von jedem auf den Basisvektoren und wollen den Wert für eine beliebige Matrix bestimmen. Dafür drücken wir als Linearkombination der aus:
Mithilfe der Definition der dualen Basis und der Linearität der können wir nun die Lösung angeben: Es gilt für und , also folgt
How to get to the proof? (Elemente des Dualraums und ihr Kern)
Für die Elemente im Kern von und gilt für alle . Das heißt, das gesuchte hängt nur von den ab, die nicht im Kern von und liegen. Um das genauer zu verstehen, betrachten wir zunächst die Dimension des Kerns. Mit der Dimensionsformel erhalten wir
und somit gilt . Nun ist ein Untervektorraum von . Weil eindimensional ist, erhalten wir dass die Dimension vom Bild von entweder oder ist. Somit ist oder .
Nun haben wir ; das heißt, sie haben beide die gleiche Dimension. Wenn ist, haben sie die gleiche Dimension wie . Somit gilt und und sind die Nullabbildung. Also gilt und wir können wählen.
Es bleibt noch der Fall übrig. In diesem Fall haben wir tatsächlich Vektoren, bei denen eine Rolle spielt. Um die Abbildungen zu vergleichen, bietet es sich an, sie auf einer Basis zu betrachten, da wir nach dem Prinzip der linearen Fortsetzung wissen, dass und durch ihr Verhalten auf einer Basis schon vollkommen bestimmt sind. Um das zu tun, lohnt es sich eine Basis von zu wählen, bei der wir schon viel über unsere Abbildungen und wissen. Wir wissen schon, was beider auf . Sei eine Basis von . Dann können wir mit dem Basisergänzungssatz diese Basis zu einer Basis von fortsetzen.
Weil ist, wissen wir das und gilt. Weiter wissen wir für . Wir brauchen nun einen Kandidaten für . Da von Elementen aus abhängt, die nicht auf abgebildet werden, ergibt es Sinn für den Kandidaten zu verwenden. Mit erhalten wir .
Um zu sehen, ob für alle gilt, reicht es nun wieder nach dem Prinzip der linearen Fortsetzung, dies auf unserer Basis zu überprüfen. Für wissen wir dies bereits, und für mit haben wir . Damit haben wir die Aussage bewiesen.
Solution (Elemente des Dualraums und ihr Kern)
Die Funktion ist eine lineare Abbildung zwischen zwei endlich dimensionalen Vektorräumen. Aus dem Dimensionssatz folgt
Weil das Bild ein Untervektorraum des -Vektorraums ist, gilt . Außerdem gilt . Damit können wir folgern
Also gilt . Andererseits ist , weil der Kern ein Untervektorraum von ist. Deshalb gibt es nur zwei Möglichkeiten:
- Die Dimension von ist .
- Die Dimension von ist .
Genauso können wir folgern, dass die Dimension vom Kern von entweder oder ist.
Wir nehmen an, dass und zeigen, dass es dann ein gibt mit .
Nun betrachten wir die zwei Fälle und .
Fall 1:
Fall 2:
In diesem Fall folgt aus dem Dimensionssatz
Sei eine Basis von . Wegen ist es auch eine Basis von . Wegen dem Basisergänzungssatz können wir ergänzen zu einer Basis von : .
Wir definieren und . Der Vektor liegt nicht in , folglich gilt . Definiere . Wir zeigen, dass . Wegen dem Prinzip der linearen Fortsetzung reicht es, diese Gleichheit auf der Basis zu zeigen.
Wir betrachten zuerst mit . Weil , gilt
Für den Basisvektor gilt
Für jeden Basisvektor stimmen und überein. Also gilt .
Solution (Duale Basis und Hyperebenen)
Solution sub-exercise 2:
Nach dem Prinzip der linearen Fortsetzung ist eine lineare Abbildung dadurch bestimmt, was sie auf einer Basis macht. Um dieses verwenden zu können, wählen wir zunächst eine Basis von . Der Basisergänzungssatz liefert uns nun einen Vektor , sodass eine Basis von ist.
Nach dem Prinzip der linearen Fortsetzung, können einen Kandidaten für die lineare Abbildung definieren, indem wir sagen, was auf einer Basis von passiert. Die Vektoren sind Elemente von . Da der Kern von sein soll, müssen wir für fordern. Der letzte Basisvektor ist nicht in . Damit darf nicht im Kern von liegen. Das heißt, wir können beispielsweise fordern. Zusammengefasst definieren wir als die lineare Abbildung mit
Da von erzeugt wird, ist . Wir müssen also nur noch Zeigen, dass gilt. Dafür sei . Weil eine Basis von , finden wir mit . Nun wissen wir
Somit ist und . Das heißt, wir haben .
Solution sub-exercise 3:
Die Abbildung ist nicht Eindeutig: Wir wissen, dass , weil . Somit existiert mit . Weil gilt, gibt es ein Element mit . Somit ist . Wenn wir nun die lineare Abbildung . Diese hat den gleichen Kern, weil genau dann gilt, wenn gilt. Dies ist genau dann der Fall, wenn gilt, weil .
Weiter ist , weil gilt. Somit ist die lineare Abbildung aus dem zweiten Teil nicht eindeutig.
In der letzten Aufgabe haben wir gefordert, weil wir im Beweis ein Element benötigt haben, das weder noch ist. Der Körper besteht nur aus den Elementen und . Das heißt, wenn wir eine lineare Abbildung konstruieren wollen, die einen -dimensionalen Untervektorraum als Kern hat, dann müssen wir sie als
definieren. Diese Abbildung ist linear, weil es eine lineare Abbildung gibt, deren Kern ist und die einzige Möglichkeit eine Abbildung Kern hinzuschreiben, diese Abbildungsvorschrift ist. Insbesondere kommen wir bei der letzten Teilaufgabe zu einem anderen Ergebnis: Die Abbildung ist eindeutig.
Exercise (Basis vom Kern von )
Sei ein -Vektorraum, eine Basis und die zu duale Basis. Zeige: Für jedes gilt
Insbesondere ist eine Basis von .
Solution (Basis vom Kern von )
Per Definition der dualen Basis gilt für . Es gilt also für alle und da der Kern ein Unterraum ist, gilt auch
Da gilt, ist nicht die Nullabbildung. Mit der vorherigen Aufgabe folgt somit . Da linear unabhängig sind, gilt , und da dieser Spann im Kern von enthalten ist, folgt die Gleichheit der beiden Unterräume.
Exercise
Betrachte die Basis
von .
- Bestimme die zu duale Basis mit für .
- Bestimme den Kern und zeichne ihn im für .
Solution
Solution sub-exercise 1:
Die darstellende Matrix einer linearen Abbildung bzgl. der Standardbasen von und von ist die eindeutig bestimmte Matrix sodass
für alle gilt.
Wir suchen die Funktionsvorschrift der linearen Abbildungen , . Wir bestimmen also die drei dazugehörigen darstellenden Matrizen bzgl. der Standardbasen. Per Definition der dualen Basis soll gelten
und analog für . Fassen wir diese Gleichungen in Matrixform zusammen erhalten wir
Wir müssen also eine Inverse der Matrix auf der linken Seite der Gleichung bestimmen, die die Basisvektoren in als Spalten hat.
To-Do:
Das macht Sinn, weil man dann die Basisübergangsmatrix von der Standardbasis zu bestimmt (Zusammenhang von dualer Basis und Kooridnaten)
Die Inverse ist
Die Zeilen sind die gesuchten darstellenden Matrizen der dualen Basisvektoren. Wir haben also
Solution sub-exercise 2:
Aus der vorherigen Aufgabe wissen wir, dass , und gilt. Eingezeichnet in erhalten wir jeweils eine von den beiden Vektoren aufgespannte Ebene im .
Anstatt die vorherige Aufgabe zu nutzen, können wir auch die Kerne der Matrizen berechnen:
Proof step:
Der Kern von enthält alle mit , d.h. mit . Also gilt
Beachte, dass ist, also stimmt das Ergebnis für den Kern mit dem aus der vorherigen Aufgabe überein.
Proof step:
Der Kern von enthält alle mit , d.h. mit . Also gilt
Auch hier gilt gilt, also stimmt das Ergebnis mit dem vorherigen überein.
Proof step:
Der Kern von enthält alle mit , d.h. mit . Also gilt
Wegen stimmt das mit dem vorher bestimmten Ergebnis überein.
Solution (Duale Abbildung)