About Chern number of U(1) principal bundle on a torus

  • A
  • Thread starter lichen1983312
  • Start date
  • Tags
    Torus
In summary, the U(1) bundle on a torus is a crucial mathematical setup for many physics problems. A one-form ##A## can be defined over the whole torus, and to define a connection one-form for this bundle, a Lie-algebra valued one-form is needed on the torus. The local curvature two-form is ##F = diA + iA \wedge iA = idA##, and if there is no continuous section found, both ##iA## and ##F## are only well-defined on local charts. The transition function for the overlapping parts is ##{t_{ij}}(p) = \exp [i\Lambda (p)]##, causing the transition ##iA \to
  • #1
lichen1983312
85
2
The U(1) bundle on a torus is a important math setup for a lot of physics problems. Since I am awkward on this subject and many of the physics material doesn't give a good introduction. I like to put some of my understanding here and please help me to check whether they are right or wrong.
1. A one-form ##A## can be defined over the whole torus.
2. To define a connection one-form for this bundle, we need a Lie-algebra valued one-form on the torus. So I can simply define this form by adding an ##i## to ##A## as ##iA##.
3. So the Lie-algebra valued local curvature two-form is ##F = diA + iA \wedge iA = idA##
4. If there is no continuous section can be found. Both ##iA## and ##F## can only be well defined on local charts. In the overlapping parts ##{U_i} \cap {U_j}## the transition function is ##{t_{ij}}(p) = \exp [i\Lambda (p)]##. So we have the transition ##iA \to iA + id\Lambda (p)## and ##F \to t_{ij}^{ - 1}Ft_{ij}^{ - 1} = F## and the Chern number is found as
##\int_T {\frac{i}{{2\pi }}trF = } \int_T {\frac{i}{{2\pi }}F} \in Z##

So ##F## is exactly ##idA## everywhere? is this right ?

If so, by directly appling stokes theorem
##\int_T {F = - i} \int_T {dA = } \int_{\partial T} {A = 0} ##
I found the Chern number must be 0? What is wrong here? Please help.
 
Last edited:
Physics news on Phys.org
  • #2
For a general ##U(1)## bundle there is not a global gauge field although the Chern class is always a global 2 form. The field strength is exact only locally so one can not use Stokes Theorem to conclude that its integral over the torus is zero. If the bundle is trivial, that is if it has a global section, then the gauge field is a global Lie algebra valued 1-form and the Chern class is zero.

A basic example of this type of phenomenon is the global 1 form ##dθ## on the circle. This form is not exact although locally it is the differential of the angle function ##θ##. However, ##θ## cannot be defined continuously on the entire circle.

- The Chern class of the tangent bundle of the torus is zero since it is a trivial bundle. But there are infinitely many ##U(1)## bundles on the torus that are not trivial.
 
Last edited:
  • Like
Likes fresh_42
  • #3
lavinia said:
For a general ##U(1)## bundle there is not a global gauge field although the Chern class is always a global 2 form. The field strength is exact only locally so one can not use Stokes Theorem to conclude that its integral over the torus is zero. If the bundle is trivial, that is is has a global section, then the gauge field is a global Lie algebra value 1-form and the the Chern class is zero.

A basic example of this type of phenomenon is the global 1 form ##dθ## on the circle. This form is not exact although locally it is the differential of the angle function ##θ##. However, ##θ## cannot be defined continuously on the entire circle.

- The Chern class of the tangent bundle of the torus is zero since it is a trivial bundle. But there are infinitely many ##U(1)## bundles on the torus that are not trivial.
Thanks very much.
You said "Chern class is zero" do you really mean Chern number ?

So what is the reason ##\theta ## is not exact? Just because it cannot be globally defined?
Also, can you explain more about how two U(1) bundles on the same base space are different?
 
Last edited:
  • #4
lichen1983312 said:
Thanks very much.
You said "Chern class is zero" do you really mean Chern number ?

Same thing, The Chern number is zero if and only if the Chern class is zero. Otherwise put, the Chern number is zero if and only if the Chern form is exact and this is true if and only if its cohomology class is zero in ##H^2(##Torus;##R)##. When one says that the Chern class is zero, one means that its cohomology class is zero.
 
Last edited:
  • #5
lichen1983312 said:
So what is the reason ##\theta ## is not exact? Just because it cannot be globally defined?

##dθ## is not exact ( even though it is written as an exterior derivative. But this is just notation). One way to see that it is not exact is to observe that its integral over the circle is equal to ##2π##. If it were exact its integral would be zero - by Stokes Theorem.
 
Last edited:
  • Like
Likes fresh_42
  • #6
lavinia said:
##dθ## is not exact ( even though it is written as an exterior derivative. But this is just notation). One way to see that it is not exact is to observe that its integral over the circle is equal to ##2π##. If it were exact its integral would be zero - by Stokes Theorem.

Thanks, I am really confused about something here. Can I say that ##dθ## is not exact is because the circle cannot be parameterized completely by ##θ##?
 
  • #7
lichen1983312 said:
Also, can you explain more about how two U(1) bundles on the same base space are different?

##U(1)## bundles are characterized by their Chern class. Two bundles have the same Chern class(equivalently Chern number) if and only if they are bundle isomorphic.

For example, on the 2 sphere, two non-isomorphic ##U(1)## bundles are the tangent bundle and the Hopf bundle. The tangent bundle has Chern number 2 and the Hopf bundle has Chern number 1.

Split the 2 sphere into two hemispheres along the equator. Since each hemisphere is a topological disk and since ##U(1)## is a topological circle, the bundle restricted to each hemisphere is homeomorphic to the Cartesian product ##D^2×S^1## which is a solid torus so the ##U(1)## bundle splits into two solid tori. These are pasted together along their boundaries which are just two regular tori (not solid). These boundary tori lie above the equator of the sphere. So a ##U(1)## bundle over the 2 sphere is obtained from two solid tori that are pasted together along their boundaries. Different ways of pasting these boundaries together produce different ##U(1)## bundles.

It is a good exercise to examine these pasting maps for the cases of the tangent bundle and the Hopf bundle.
 
Last edited:
  • #8
lichen1983312 said:
Thanks, I am really confused about something here. Can I say that ##dθ## is not exact is because the circle cannot be parameterized completely by ##θ##?
Yes. ##dθ## is the 1 form that assigns a value of 1 to the positively oriented unit tangent vectors of the circle at each point. So integrating over an arc gives the arc length. Arc length does not define a continuous function on the entire circle since if the arc goes once around the circle the integral is ##2π## but if the arc goes nowhere, the integral is zero.
 
Last edited:
  • #9
lavinia said:
Yes. ##dθ## is the 1 form that assigns a value of 1 to the positively oriented unit tangent vectors of the circle at each point. So integrating over an arc gives the the arc length. Arc length does not define a continuous function on the entire circle since if the arc goes once around the circle the integral is ##2π## but if the arc goes nowhere, the integral is zero.
So, can I say that I can always have a globally defined one-form ##A## for an arbitrary manifold, as long as in the overlapping area of different charts it transform according to the transition function? Its exterior derivative ##dA## should be also globally defined? right ?
 
  • #10
lavinia said:
##U(1)## bundles are characterized by their Chern class. Two bundles have the same Chern class(equivalently Chern number) if and only if they are bundle isomorphic.

For example, on the 2 sphere, two non-isomorphic ##U(1)## bundles are the tangent bundle and the Hopf bundle. The tangent bundle has Chern number 2 and the Hopf bundle has Chern number 1.

Split the 2 sphere into two hemispheres along the equator. Since each hemisphere is a topological disk and since ##U(1)## is a topological circle, the bundles restricted to each hemisphere is homeomorphic to the Cartesian product ##D^2×S^1## which is a solid torus so the ##U(1)## bundle splits into two solid tori. These are pasted together along their boundaries which are just two regular tori (not solid). These boundary tori lie above the equator of the sphere. So a ##U(1)## bundle over the 2 sphere is obtained from two solid tori that are pasted together along their boundaries. Different ways of pasting these boundaries together produce different ##U(1)## bundles.

It is a good exercise to examine these pasting maps for the cases of the tangent bundle and the Hopf bundle.
Thanks very much.
 
  • #11
lichen1983312 said:
So, can I say that I can always have a globally defined one-form ##A## for an arbitrary manifold, as long as in the overlapping area of different charts it transform according to the transition function? Its exterior derivative ##dA## should be also globally defined? right ?

One can always have a globally defined form of any degree just by defining a form on a coordinate neighborhood and extending it to the whole manifold by multiplying it by a smooth function that equals zero outside of the neighborhood.
 
  • #12
lavinia said:
One can always have a globally defined form of any degree just by defining a form on a coordinate neighborhood and extending it to the whole manifold by multiplying it by a smooth function that equals zero outside of the neighborhood.
Yes, that is generally possible right?
 
  • #13
lichen1983312 said:
Yes, that is generally possible right?
Yes. The 1 form that is identically zero everywhere is global as well

Here is an example. Choose a non-negative smooth function on the circle that is zero outside of a small interval around 1 and which is strictly positive on a subinterval on non-zero length. The graph of this function looks like a little bump surrounding 1 that flattens out completely outside of the interval. Multiply ##dθ## by this function. This produces a new 1 form that is identically zero outside of the interval. However, the integral of this 1 form over the circle is not zero since the function is positive on an interval of non-zero length. Therefore by Stokes Theorem this form is not exact.
 
Last edited:
  • #14
lavinia said:
Yes. The 1 form that is identically zero everywhere is global as well.
Sorry I was trying to say a every where non-zero form. So if I go back to my original post, and let ##A## be such a globally defined one-form on the torus. Since the bundle is not necessarily trivial. On certain charts the gauge potential has the form ##iA + id{\Lambda _1} + id{\Lambda _2} \cdots ##, but the field strength is globally defined as ##d(iA + id{\Lambda _1} + id{\Lambda _2} \cdots ) = idA##.
is this right ?
 
  • #15
lichen1983312 said:
Sorry I was trying to say a every where non-zero form. So if I go back to my original post, and let ##A## be such a globally defined one-form on the torus. Since the bundle is not necessarily trivial. On certain charts the gauge potential has the form ##iA + id{\Lambda _1} + id{\Lambda _2} \cdots ##, but the field strength is globally defined as ##d(iA + id{\Lambda _1} + id{\Lambda _2} \cdots ) = idA##.
is this right ?

This is how I understand ##U(1)## gauge potentials. One has a connection 1 form ##ω## defined globally on the tangent space to the ##U(1)## bundle itself - not on the tangent space to the manifold. This is a Lie algebra valued 1 form. Given a local section of the bundle one can pull the connection 1 form back to the tangent bundle of the manifold to get what is called a gauge field. This field in general is only defined locally since there may not be a global section of the bundle. Given a different local section one gets a different gauge field over a possibly different local neighborhood.

The exterior derivative of the connection 1 form on the ##U(1)## bundle is the curvature 2 form of the connection. Pulling this form back by the local section gives the field strength of the gauge potential. One can show - and I think you did - that the different field strengths over different local neighborhoods piece together to give a global 2 form. This form represents the Chern class of the ##U(1)## bundle and its integral over the torus - or for that matter any complete orientable surface without boundary e.g. the sphere - is the Chern number of the bundle.

Notice though that the gauge fields themselves may not piece together to give a global 1 form and this is where I think your question comes from. You said in the original post

"So ##F## is exactly ##idA## everywhere? is this right ?"

In general this is wrong because the gauge potential ##A## may not be defined globally but only locally. If however, ##A## is in fact defined globally then you are right to conclude that the Chern number is zero.

- BTW; Notice that this is the same kind of situation as with ##dθ##. Locally one has an angle function ##θ## whose exterior derivative equals ##dθ## but there is no global function on the circle whose exterior derivative equals ##dθ##. In the same way one has a global 2 form ##dA## defined on the torus which is locally the exterior derivative of a 1 form. But globally it is not an exterior derivative unless the ##U(1)## bundle is trivial.
 
Last edited:
  • Like
Likes WWGD
  • #16
lavinia said:
This is how I understand ##U(1)## gauge potentials. One has a connection 1 form ##ω## defined globally on the tangent space to the ##U(1)## bundle itself - not on the tangent space to the manifold. This is a Lie algebra valued 1 form. Given a local section of the bundle one can pull the connection 1 form back to the tangent bundle of the manifold to get what is called a gauge field. This field in general is only defined locally since there may not be a global section of the bundle. Given a different local section one gets a different gauge field over a possibly different local neighborhood.

The exterior derivative of the connection 1 form on the ##U(1)## bundle is the curvature 2 form of the connection. Pulling this form back by the local section gives the field strength of the gauge potential. One can show - and I think you did - that the different field strengths over different local neighborhoods piece together to give a global 2 form. This form represents the Chern class of the ##U(1)## bundle and its integral over the torus - or for that matter any complete orientable surface without boundary e.g. the sphere - is the Chern number of the bundle.

Notice though that the gauge fields themselves may not piece together to give a global 1 form and this is where I think your question comes from. You said in the original post

"So ##F## is exactly ##idA## everywhere? is this right ?"

In general this is wrong because the gauge potential ##A## may not be defined global but only locally. If however, ##A## is in fact defined globally then you are right to conclude that the Chern number is zero.

- BTW; Notice that this is the same kind of situation as with ##dθ##. Locally one has an angle function ##θ## whose exterior derivative equals ##dθ## but there is no global function on the circle whose exterior derivative equals ##dθ##. In the same way one has a global 2 form ##dA## defined on the torus which is locally the exterior derivative of a 1 form. But globally it is not an exterior derivative unless the ##U(1)## bundle is trivial.

Thanks for your explanation, that is very helpful. I think the mistake in my example is that I tried to build a globally defined gauge potential (one-form) on the base space first, then I used it to define the connection one-form on the bundle. This is not valid at all. One should always start from a connection-one form on the bundle first and then push it back to the base manifold.
Is this right?
 
  • #17
lichen1983312 said:
Thanks for your explanation, that is very helpful. I think the mistake in my example is that I tried to build a globally defined gauge potential (one-form) on the base space first, then I used it to define the connection one-form on the bundle. This is not valid at all. One should always start from a connection-one form on the bundle first and then push it back to the base manifold.
Is this right?

Yes. You got it.

BTW: This helped me to understand the Chern class of a ##U(1)## bundle. ##U(1)## topologically is a circle so it is a 1 dimensional sphere - as a manifold. A ##U(1)## bundle is an example of a sphere bundle, a fiber bundle whose fiber is homeomorphic to a sphere - in this case a 1 sphere. The Chern class of a ##U(1)## bundle is a special case of a characteristic class that is defined for oriented sphere bundles where the fiber sphere can be of any dimension. This characteristic class is called the Euler class of the bundle.

For smooth bundles over smooth manifolds, the Euler class is characterized by the following construction: First one can show that for an oriented k-sphere bundle there is a k-form ##ω## defined on the total space of the bundle whose integral over each fiber sphere is equal to 1. For ##U(1)## bundles ##ω## can be chosen to be the connection 1 form appropriately normalized. The exterior derivative of ##ω## for any k-sphere bundle is a smooth ##k+1## form and remarkably it is always the pull back under the bundle projection map of a globally defined ##k+1## form on the base manifold. That is: ##dω = π^{*}C## and ##C## is called the Euler form of the sphere bundle. For a ##U(1)## bundle ##dω## is the curvature 2 form of the connection and ##C## is the Chern form.

Notice that ##C## may not be exact even though its pull back to the total space is equal to ##dω##. But for any local section ##s## of the bundle ##C = ds^{*}ω## so ##C## is locally exact but in general not globally. We saw this with the Chern form of the ##U(1)## bundle.

Now suppose that there actually is a global section of the sphere bundle. Then ##C = s^{*}dω = ds^{*}ω## so ##C## is exact. So the Euler class is an obstruction to a section of the sphere bundle. That is: if there is a section then the Euler class must be homologous to zero i.e. any differential form that represents it must be exact.

In general, one can not integrate the Euler class over the base manifold since ##C## is a ##k+1## form and the dimension of manifold may not be equal to ##k+1##, for instance a ##U(1)## bundle over the 5 sphere. However if the manifold's dimension equals ##k+1## then it can be integrated over the manifold to yield the Euler number of the bundle. For the case of a ##U(1)## bundle over a 2 dimensional surface the Euler number is the same as the Chern number.

For the case of the tangent bundle of a smooth manifold, the Euler number is a topological invariant called the Euler characteristic.

Notice that if there are two ##k## forms ##ω_1## and ##ω_2## then their difference defines a global ##k## form on the base manifold. It follows that the difference of the corresponding Euler forms ##C_1## and ##C_2## is exact so the Euler class is independent of the choice of ##ω##. For a ##U(1)## bundle this means that the Chern class is independent of the connection. So it is an invariant of the bundle and is independent of the bundle's geometry.
 
Last edited:
  • #18
lavinia said:
Yes. You got it.

BTW: This helped me to understand the Chern class of a ##U(1)## bundle. ##U(1)## topologically is a circle so it is a 1 dimensional sphere - as a manifold. A ##U(1)## bundle is an example if a sphere bundle, a fiber bundle whose fiber is homeomorphic to a sphere - in this case a 1 sphere. The Chern class of a ##U(1)## bundle is a special case of a characteristic class that is defined for oriented sphere bundles where the fiber sphere can be of any dimension. This characteristic class is called the Euler class of the bundle.

For smooth bundles over smooth manifolds, the Euler class is characterized by the following construction: First one can show that for an oriented k-sphere bundle there is a k-form ##ω## defined on the total space of the bundle whose integral over each fiber sphere is equal to 1. For ##U(1)## bundles ##ω## can be chosen to be the connection 1 form appropriately normalized. The exterior derivative of ##ω## for any k-sphere bundle is a smooth ##k+1## form and remarkably it is always the pull back under the bundle projection map of a globally defined ##k+1## form on the base manifold. That is: ##dω = π^{*}C## and ##C## is called the Euler form of the sphere bundle. For a ##U(1)## bundle ##dω## is the curvature 2 form of the connection and ##C## is the Chern form.

Notice that ##C## may not be exact even though its pull back to the total space is equal to ##dω##. But for any local section ##s## of the bundle ##C = ds^{*}ω## so ##C## is locally exact but in general not globally. We saw this with the Chern form of the ##U(1)## bundle.

Now suppose that there actually is a global section of the sphere bundle. Then ##C = s^{*}dω = ds^{*}ω## so ##C## is exact. So the Euler class is an obstruction to a section of the sphere bundle. That is: if there is a section then the Euler class must be homologous to zero i.e. any differential form that represents it must be exact.

In general, one can not integrate the Euler class over the base manifold since ##C## is a ##k+1## form and the dimension of manifold may not be equal to ##k+1##, for instance a ##U(1)## bundle over the 5 sphere. However if the manifold's dimension equals ##k+1## then it can be integrated over the manifold to yield the Euler number of the bundle. For the case of a ##U(1)## bundle over a 2 dimensional surface the Euler number is the same as the Chern number.

For the case of the tangent bundle of a smooth manifold, the Euler number is a topological invariant called the Euler characteristic.

Notice that if there are two ##k## forms ##ω_1## and ##ω_2## then their difference defines a global ##k## form on the base manifold. It follows that the difference of the corresponding Euler forms ##C_1## and ##C_2## is exact so the Euler class is independent of the choice of ##ω##. For a ##U(1)## bundle this means that the Chern class is independent of the connection. So it is an invariant of the bundle and is independent of the bundle's geometry.

Lavinia, can you recommend a good source for this material, (diff. topology, I guess)? Maybe Bott and Tu?
 
  • #19
WWGD said:
Lavinia, can you recommend a good source for this material, (diff. topology, I guess)? Maybe Bott and Tu?
Yes Bott and Tu for Euler classes using differential forms. For Characteristic classes derived from connections on vector bundles - Milnor's Characteristic Classes Appendix C - For gauge fields try this link

https://empg.maths.ed.ac.uk/Activities/GT/Lect1.pdf

Bott and Tu do the Euler class for smooth oriented sphere bundles using de Rham theory and Czech cohomology. Really good. They show examples of sphere bundles whose Euler class is zero but have no section.

For the Euler class using differential geometry on a Riemannian manifold Chern's original paper. I think it is called "A Simple Intrinsic Proof of the Gauss-Bonnet Theorem." Also Milnor's Appendix C. If the metric is not positive definite I am not sure. The Pfaffian form probably has some analogue for non-positive definite metrics.
 
Last edited:
  • Like
Likes WWGD
  • #20
lavinia said:
Yes. You got it.

BTW: This helped me to understand the Chern class of a ##U(1)## bundle. ##U(1)## topologically is a circle so it is a 1 dimensional sphere - as a manifold. A ##U(1)## bundle is an example if a sphere bundle, a fiber bundle whose fiber is homeomorphic to a sphere - in this case a 1 sphere. The Chern class of a ##U(1)## bundle is a special case of a characteristic class that is defined for oriented sphere bundles where the fiber sphere can be of any dimension. This characteristic class is called the Euler class of the bundle.

For smooth bundles over smooth manifolds, the Euler class is characterized by the following construction: First one can show that for an oriented k-sphere bundle there is a k-form ##ω## defined on the total space of the bundle whose integral over each fiber sphere is equal to 1. For ##U(1)## bundles ##ω## can be chosen to be the connection 1 form appropriately normalized. The exterior derivative of ##ω## for any k-sphere bundle is a smooth ##k+1## form and remarkably it is always the pull back under the bundle projection map of a globally defined ##k+1## form on the base manifold. That is: ##dω = π^{*}C## and ##C## is called the Euler form of the sphere bundle. For a ##U(1)## bundle ##dω## is the curvature 2 form of the connection and ##C## is the Chern form.

Notice that ##C## may not be exact even though its pull back to the total space is equal to ##dω##. But for any local section ##s## of the bundle ##C = ds^{*}ω## so ##C## is locally exact but in general not globally. We saw this with the Chern form of the ##U(1)## bundle.

Now suppose that there actually is a global section of the sphere bundle. Then ##C = s^{*}dω = ds^{*}ω## so ##C## is exact. So the Euler class is an obstruction to a section of the sphere bundle. That is: if there is a section then the Euler class must be homologous to zero i.e. any differential form that represents it must be exact.

In general, one can not integrate the Euler class over the base manifold since ##C## is a ##k+1## form and the dimension of manifold may not be equal to ##k+1##, for instance a ##U(1)## bundle over the 5 sphere. However if the manifold's dimension equals ##k+1## then it can be integrated over the manifold to yield the Euler number of the bundle. For the case of a ##U(1)## bundle over a 2 dimensional surface the Euler number is the same as the Chern number.

For the case of the tangent bundle of a smooth manifold, the Euler number is a topological invariant called the Euler characteristic.

Notice that if there are two ##k## forms ##ω_1## and ##ω_2## then their difference defines a global ##k## form on the base manifold. It follows that the difference of the corresponding Euler forms ##C_1## and ##C_2## is exact so the Euler class is independent of the choice of ##ω##. For a ##U(1)## bundle this means that the Chern class is independent of the connection. So it is an invariant of the bundle and is independent of the bundle's geometry.

Can you help me to check this example? It is one of the reasons I post this thread. It is the magnetic monopole problem.
If there is a magnetic monopole, it should produce a magnetic field ##\vec B = g\vec r/{r^3}## so that ##\int_S {\vec B \cdot d\vec S} = 4\pi g##, where ##4\pi g## is a constant. It is also known that in physics there should be a vector potential ##{\vec A}## such that ##\vec B = \nabla \times \vec A##. As you can expect, there is no globally defined ##{\vec A}## can be found. So people use two vector potentials ##{{\vec A}_N}##and ##{{\vec A}_S}##.
##{{\vec A}_N}## is defined everywhere except the south pole with components
##A_N^x = \frac{{ - gy}}{{r(r + z)}}\,##, ##A_N^y = \frac{{gx}}{{r(r + z)}}##, and ##A_N^z = 0##.
##{{\vec A}_S}## is defined everywhere except the north pole with components
##A_S^x = \frac{{gy}}{{r(r - z)}}\,\,##, ##A_S^y = \frac{{ - gx}}{{r(r - z)}}## and ##A_S^z = 0##
So if think ##A## is a one-form
##{A_N} = A_N^xdx + A_N^ydy## and ##{A_S} = A_S^xdx + A_S^ydy##
by switching to spherical coordinates
##{A_N} = g(1 - \cos \theta )d\varphi ##
##{A_S} = - g(1 + \cos \theta )d\varphi ##
and the curvature is globally defined
##d{A_N} = d{A_S} = g\sin \theta d\theta \wedge d\varphi ##
so the books says if we put an Lie algebra factor "i" in front of "A" , "iA" becomes a Lie algebra valued one-form. "idA" becomes the field strength.
##i{A_N} - i{A_S} = id(2g\varphi )##
is also a valid transition property.
So the books says we have are dealing with a ##U(1)## bundle and in this case it is non-trivial.

So my question is
Is this a case that we construct a U(1) bundle from properties from the base manifold?
Can I just conclude that if I have
1, locally defined Lie algebra-valued one-form ##{A_i}## on the base manifold.
2, ##{A_i}## have a transitions property ##{A_i} = {A_j} + id\phi##.
3, ##dA## is globally defined
there must be a well defined U(1) bundle corresponding to these properties?
 
  • #21
lichen1983312 said:
Can you help me to check this example? It is one of the reasons I post this thread. It is the magnetic monopole problem.
If there is a magnetic monopole, it should produce a magnetic field ##\vec B = g\vec r/{r^3}## so that ##\int_S {\vec B \cdot d\vec S} = 4\pi g##, where ##4\pi g## is a constant. It is also known that in physics there should be a vector potential ##{\vec A}## such that ##\vec B = \nabla \times \vec A##. As you can expect, there is no globally defined ##{\vec A}## can be found. So people use two vector potentials ##{{\vec A}_N}##and ##{{\vec A}_S}##.
##{{\vec A}_N}## is defined everywhere except the south pole with components
##A_N^x = \frac{{ - gy}}{{r(r + z)}}\,##, ##A_N^y = \frac{{gx}}{{r(r + z)}}##, and ##A_N^z = 0##.
##{{\vec A}_S}## is defined everywhere except the north pole with components
##A_S^x = \frac{{gy}}{{r(r - z)}}\,\,##, ##A_S^y = \frac{{ - gx}}{{r(r - z)}}## and ##A_S^z = 0##
So if think ##A## is a one-form
##{A_N} = A_N^xdx + A_N^ydy## and ##{A_S} = A_S^xdx + A_S^ydy##
by switching to spherical coordinates
##{A_N} = g(1 - \cos \theta )d\varphi ##
##{A_S} = - g(1 + \cos \theta )d\varphi ##
and the curvature is globally defined
##d{A_N} = d{A_S} = g\sin \theta d\theta \wedge d\varphi ##
so the books says if we put an Lie algebra factor "i" in front of "A" , "iA" becomes a Lie algebra valued one-form. "idA" becomes the field strength.
##i{A_N} - i{A_S} = id(2g\varphi )##
is also a valid transition property.
So the books says we have are dealing with a ##U(1)## bundle and in this case it is non-trivial.

So my question is
Is this a case that we construct a U(1) bundle from properties from the base manifold?
Can I just conclude that if I have
1, locally defined Lie algebra-valued one-form ##{A_i}## on the base manifold.
2, ##{A_i}## have a transitions property ##{A_i} = {A_j} + id\phi##.
3, ##dA## is globally defined
there must be a well defined U(1) bundle corresponding to these properties?

The Lie algebra of any Lie group is the tangent space at the identity. For ##U(1)## this is the 1 dimensional vectors space of all a vectors ##ai## where ##a## is a real number and ##i## is the square root of ##-1##. The way to see this is to note that a tangent vector to a point on a circle is perpendicular to the circle and ##i## is perpendicular to ##1##. So multiplying by ##i## puts you in the Lie Algebra of ##U(1)##.
 
Last edited:
  • #22
lavinia said:
The Lie algebra of any Lie group is the tangent space at the identity. For ##U(1)## this is the 1 dimensional vectors space of all a vectors ##ai## where ##a## is a real number and ##i## is the square too of ##-1##. The way to see this is to note that a tangent vector to a point on a circle is perpendicular to the circle and ##i## is perpendicular to ##1##. So multiplying by ##i## puts you in the Lie Algebra of ##U(1)##.
Thanks very much.
 
  • #23
A couple of points.

The transition rule for two gauge fields from one local section to another is ##A_{N} = g_{NM}A_{M}g^{-1}_{NM} - dg_{NM}g_{NM}^{-1}##

##U(1)## is abelian so this reduces to ##A_{N} = A_{M} - dg_{NM}g_{NM}^{-1}##

The problem here is that ##\varphi## returns to the same point in space with an increment of ##2π## so as is, the transition function ##g_{NM}## is multivalued. The way around this is to write ##g_{NM} = e^{iΨ(\varphi)}## but then ##dg_{NM}g_{NM}^{-1} = idΨ## and is purely imaginary. This can be fixed by multiplying ##A_{N,M}## by ##i##.

One already knows that ##A_{N}-A_{M} = 2gd\varphi## so ##dΨ=2gd\varphi##
 
Last edited:
  • #24
lavinia said:
A couple of points.

The transition rule for two gauge fields from one local section to another is ##A_{N} = g_{NM}A_{M}g^{-1}_{NM} - dg_{NM}g_{NM}^{-1}##

##U(1)## is abelian so this reduces to ##A_{N} = A_{M} - dg_{NM}g_{NM}^{-1}##

The problem here is that ##ϑ## returns to the same point in space with an increment of ##2π## so as is the transition ##g_{NM}## function is multivalued. The way around this is to write ##g_{NM} = e^{iφ(ϑ)}## but then ##dg_{NM}g_{NM}^{-1} = idφ## and is purely imaginary. This can be fixed by multiplying ##A_{N,M}## by ##i##.

One already knows that ##A_{N}-A_{M} = 2gdϑ## so ##dφ=2gdϑ##
This is the same thing as I wrote in my previous post, right? if ##{g_{NM}} = \exp (i\phi (p))##, then ##{A_N} = {A_M} + id\phi (p)##.
 
  • #25
lichen1983312 said:
This is the same thing as I wrote in my previous post, right? if ##{g_{NM}} = \exp (i\phi (p))##, then ##{A_N} = {A_M} + id\phi (p)##.

Right. I was just trying to motivate why you need to go to the Lie algebra of ##U(1)##. This is so that the transition function will be well defined.
 
  • #26
lavinia said:
Right. I was just trying to motivate why you need to go to the Lie algebra of ##U(1)##. This is so that the transition function will be well defined.
Thanks very much.
 
  • #27
Also it is clear that ##g## must be a half integer in order for the transition function to be well defined since as ##φ## increases by ##2π##, ##Ψ## increases by ##4πg## so twice ##g## has to be an integer to come back to the same value after a full circuit. For each integer ##n## one gets a different ##U(1)## bundle.

One is constructing all of the ##U(1)## bundles over the 2 sphere, one for each integer value of ##2g##.

I don't know the physics but its seems from a little web reading that ##g## is a magnetic charge that is used in a generalization of Maxwell's equations. This derivation seems to show that magnetic charge is quantized. Can you elaborate on the physics a little more?
 
Last edited:
  • #28
lavinia said:
Also it is clear that ##g## must be an integer in order for the transition function to be well defined since as ##φ## increases by ##2π##, ##Ψ## increases by ##4πg## so ##g## has to be an integer to come back to the same value after a full circuitr. For each integer ##n## one gets a different ##U(1)## bundle.

From this point of view, one is constructing all of the ##U(1)## bundles over the 2 sphere, one for each integer value of ##g##.

I don't know the physics but its seems from a little web reading that ##g## is a magnetic charge that is used in a generalization of Maxwell's equations. This seems to show that magnetic charge is quantized. Can you elaborate on the physics a little more?

You are right the limitation on "g" lead to a quantization of the monopole value. I honestly cannot explain well about this issue, which is very deep and is not my focus of research. One of the naive motivations is that the electric fields have source ##\nabla \cdot \vec E = \rho ##, where ##\rho ## is charge density. However the magnetic field is sourceless ##\nabla \cdot \vec B = 0##. This make the electric part and magnetic part of Maxwell's equations asymmetric. Introducing a magnetic particle (monopole) would make the theory more beautiful. However the monople is still a fictionary particle, never experimentally confirmed.

I study this example because there are a bunch of other topics, which are more related to me, whose mathmatical setup are almost identical to the monopole problem.
 
  • #29
lichen1983312 said:
You are right the limitation on "g" lead to a quantization of the monopole value. I honestly cannot explain well about this issue, which is very deep and is not my focus of research. One of the naive motivations is that the electric fields have source ##\nabla \cdot \vec E = \rho ##, where ##\rho ## is charge density. However the magnetic field is sourceless ##\nabla \cdot \vec B = 0##. This make the electric part and magnetic part of Maxwell's equations asymmetric. Introducing a magnetic particle (monopole) would make the theory more beautiful. However the monople is still a fictionary particle, never experimentally confirmed.

I study this example because there are a bunch of other topics, which are more related to me, whose mathmatical setup are almost identical to the monopole problem.

I saw on line that the curvature ##d{A_N} = d{A_S} = g\sin \theta d\theta \wedge d\varphi## is actually the magnetic monopole written in spherical coordinates.

In Cartesian coordinates it is ##B=(g/ 2πr^3)(xdy∧dz + ydz∧dx + zdx∧dy)##

The motivation seems to have been that since ##dB=0## to try to express the magnetic field as the curl of a vector potential as in the usual Maxwell's equations. This can not be done globally because of the singularity at the origin. So one splits it into two vector potentials and glues them via a gauge transformation along the intersection of their domains of definition.
 
Last edited:

What is the Chern number of a U(1) principal bundle on a torus?

The Chern number of a U(1) principal bundle on a torus is an integer that characterizes the topology of the bundle. It is a topological invariant that remains unchanged as long as the bundle is continuously deformed.

How is the Chern number of a U(1) principal bundle calculated?

The Chern number can be calculated using the curvature of the bundle, which is a 2-form on the base space (in this case, the torus). The integral of this curvature over any closed 2-dimensional surface on the torus will give the Chern number.

What does the Chern number tell us about the U(1) principal bundle on a torus?

The Chern number provides information about the topology of the bundle. It tells us how many times the bundle "wraps" around the torus, and provides insight into the behavior of particles moving within the bundle.

Why is the Chern number important in physics?

The Chern number is important in physics because it is a topological invariant that is related to the quantization of physical quantities. In condensed matter physics, it is used to describe the properties of topological insulators and superconductors.

Can the Chern number of a U(1) principal bundle on a torus change?

As a topological invariant, the Chern number remains unchanged under continuous deformations of the bundle. However, it can change if the bundle undergoes a discontinuous change, such as a phase transition in condensed matter systems.

Similar threads

Replies
4
Views
2K
  • Differential Geometry
Replies
3
Views
2K
  • Differential Geometry
Replies
9
Views
419
Replies
3
Views
2K
Replies
12
Views
6K
  • Differential Geometry
Replies
1
Views
2K
  • Differential Geometry
Replies
12
Views
2K
Replies
6
Views
876
Replies
16
Views
3K
  • Differential Geometry
Replies
11
Views
3K
Back
Top