PDA

View Full Version : POVM


Mike Stay
May12-04, 01:41 PM
<jabberwocky><div class="vbmenu_control"><a href="jabberwocky:;" onClick="newWindow=window.open('','usenetCode','toolbar=no, location=no,scrollbars=yes,resizable=yes,status=no ,width=650,height=400'); newWindow.document.write('<HTML><HEAD><TITLE>Usenet ASCII</TITLE></HEAD><BODY topmargin=0 leftmargin=0 BGCOLOR=#F1F1F1><table border=0 width=625><td bgcolor=midnightblue><font color=#F1F1F1>This Usenet message\'s original ASCII form: </font></td></tr><tr><td width=449><br><br><font face=courier><UL><PRE>I\'m trying to understand the concept of a POVM. I think it\'s\nsomething like classical probabilities applied to a complete set of\nprojection operators. I.e. first pick an orthonormal basis, and then\nfor each basis state, you pick an arbitrary classical probability\ndistribution over the set of possible results. That way, the mean\nvalue of any element of a POVM is the probability of getting that\nresult.\n\nIs that right?\n\nIs this a POVM for a qubit:\n\na=| 0.5 0 |,\n| 0 1 |\n\nb=| 0.1 0 |,\n| 0 0 |\n\nc=| 0.4 0 |?\n| 0 0 |\n\n\nIf so, the set {a^2, b^2, c^2} is not a POVM; how do you calculate the\nstandard deviation of elements of a POVM?\n\nThanks,\n--\nMike\n\n</UL></PRE></font></td></tr></table></BODY><HTML>');"> <IMG SRC=/images/buttons/ip.gif BORDER=0 ALIGN=CENTER ALT="View this Usenet post in original ASCII form">&nbsp;&nbsp;View this Usenet post in original ASCII form </a></div><P></jabberwocky>I'm trying to understand the concept of a POVM. I think it's
something like classical probabilities applied to a complete set of
projection operators. I.e. first pick an orthonormal basis, and then
for each basis state, you pick an arbitrary classical probability
distribution over the set of possible results. That way, the mean
value of any element of a POVM is the probability of getting that
result.

Is that right?

Is this a POVM for a qubit:

a=| .5 |,
| 1 |

b=| .1 |,
| |

c=| .4 |?
| |


If so, the set {a^2, b^2, c^2} is not a POVM; how do you calculate the
standard deviation of elements of a POVM?

Thanks,
--
Mike

Arnold Neumaier
May17-04, 02:02 AM
<jabberwocky><div class="vbmenu_control"><a href="jabberwocky:;" onClick="newWindow=window.open('','usenetCode','toolbar=no, location=no,scrollbars=yes,resizable=yes,status=no ,width=650,height=400'); newWindow.document.write('<HTML><HEAD><TITLE>Usenet ASCII</TITLE></HEAD><BODY topmargin=0 leftmargin=0 BGCOLOR=#F1F1F1><table border=0 width=625><td bgcolor=midnightblue><font color=#F1F1F1>This Usenet message\'s original ASCII form: </font></td></tr><tr><td width=449><br><br><font face=courier><UL><PRE>Mike Stay wrote:\n&gt; I\'m trying to understand the concept of a POVM. I think it\'s\n&gt; something like classical probabilities applied to a complete set of\n&gt; projection operators.\n\nNo. In a POVM you generally don\'t have any projection operators around.\n\nIn case of a discrete measurement, a POVM is a family of Hermitian positive\nsemidefinite operators P_r summing to 1 (the identity operator). The index\nr represents the possible measurement results. The probability that upon\nmeasuring a system with a detector described by this POVM you get the\nresult r is p_r=&lt;P_r&gt; when the system is in the state &lt;.&gt;. These are\nclassical probabilities, while the state is quantum. For example, for\na pure state represented by the unit vector psi, &lt;f&gt;=psi^* f psi.\n\nIn the special case where the P_r are idempotent, they must be projection\noperators, in this case one recovers von Neumann\'s description of\nideal measurements. The POVM description is a much more realistic version\nof the latter.\n\n\nFor some background, see my paper\nEnsembles and experiments in classical and quantum physics,\nInt. J. Mod. Phys. B 17 (2003), 2937-2980.\nquant-ph/0303047,\n\n\nArnold Neumaier\n\n</UL></PRE></font></td></tr></table></BODY><HTML>');"> <IMG SRC=/images/buttons/ip.gif BORDER=0 ALIGN=CENTER ALT="View this Usenet post in original ASCII form">&nbsp;&nbsp;View this Usenet post in original ASCII form </a></div><P></jabberwocky>Mike Stay wrote:
> I'm trying to understand the concept of a POVM. I think it's
> something like classical probabilities applied to a complete set of
> projection operators.

No. In a POVM you generally don't have any projection operators around.

In case of a discrete measurement, a POVM is a family of Hermitian positive
semidefinite operators P_r summing to 1 (the identity operator). The index
r represents the possible measurement results. The probability that upon
measuring a system with a detector described by this POVM you get the
result r is p_r=<P_r> when the system is in the state <.>. These are
classical probabilities, while the state is quantum. For example, for
a pure state represented by the unit vector \psi, <f>=\psi^* f \psi.

In the special case where the P_r are idempotent, they must be projection
operators, in this case one recovers von Neumann's description of
ideal measurements. The POVM description is a much more realistic version
of the latter.


For some background, see my paper
Ensembles and experiments in classical and quantum physics,
\Int. J. Mod. Phys. B 17 (2003), 2937-2980.
http://www.arxiv.org/abs/quant-ph/0303047,


Arnold Neumaier

Robert Tucci
May18-04, 03:33 AM
<jabberwocky><div class="vbmenu_control"><a href="jabberwocky:;" onClick="newWindow=window.open('','usenetCode','toolbar=no, location=no,scrollbars=yes,resizable=yes,status=no ,width=650,height=400'); newWindow.document.write('<HTML><HEAD><TITLE>Usenet ASCII</TITLE></HEAD><BODY topmargin=0 leftmargin=0 BGCOLOR=#F1F1F1><table border=0 width=625><td bgcolor=midnightblue><font color=#F1F1F1>This Usenet message\'s original ASCII form: </font></td></tr><tr><td width=449><br><br><font face=courier><UL><PRE>\nSome highly advanced civilizations in the far reaches of our galaxy\nhave been referring for countless millennia to a POVM as a RINNO\n(Resolution of Identity by Non-Negative Operators)\n</UL></PRE></font></td></tr></table></BODY><HTML>');"> <IMG SRC=/images/buttons/ip.gif BORDER=0 ALIGN=CENTER ALT="View this Usenet post in original ASCII form">&nbsp;&nbsp;View this Usenet post in original ASCII form </a></div><P></jabberwocky>Some highly advanced civilizations in the far reaches of our galaxy
have been referring for countless millennia to a POVM as a RINNO
(Resolution of Identity by Non-Negative Operators)

Mike Stay
May20-04, 11:52 AM
<jabberwocky><div class="vbmenu_control"><a href="jabberwocky:;" onClick="newWindow=window.open('','usenetCode','toolbar=no, location=no,scrollbars=yes,resizable=yes,status=no ,width=650,height=400'); newWindow.document.write('<HTML><HEAD><TITLE>Usenet ASCII</TITLE></HEAD><BODY topmargin=0 leftmargin=0 BGCOLOR=#F1F1F1><table border=0 width=625><td bgcolor=midnightblue><font color=#F1F1F1>This Usenet message\'s original ASCII form: </font></td></tr><tr><td width=449><br><br><font face=courier><UL><PRE>Arnold Neumaier &lt;Arnold.Neumaier@univie.ac.at&gt; wrote in message news:&lt;40A77B89.2020606@univie.ac.at&gt;...\n&gt; In case of a discrete measurement, a POVM is a family of Hermitian positive\n&gt; semidefinite operators P_r summing to 1 (the identity operator). The index\n&gt; r represents the possible measurement results. The probability that upon\n&gt; measuring a system with a detector described by this POVM you get the\n&gt; result r is p_r=&lt;P_r&gt; when the system is in the state &lt;.&gt;. These are\n&gt; classical probabilities, while the state is quantum. For example, for\n&gt; a pure state represented by the unit vector psi, &lt;f&gt;=psi^* f psi.\n\nOK, so while the example I gave was a POVM, the definition wasn\'t\ninclusive enough.\n\nSince they\'re Hermetian, the operators are observables, so I guess you\ncould calculate the standard deviation of one like you would for any\nother observable.\n\n--\nMike Stay\nhttp://www.cs.auckland.ac.nz/~msta039\n\n</UL></PRE></font></td></tr></table></BODY><HTML>');"> <IMG SRC=/images/buttons/ip.gif BORDER=0 ALIGN=CENTER ALT="View this Usenet post in original ASCII form">&nbsp;&nbsp;View this Usenet post in original ASCII form </a></div><P></jabberwocky>Arnold Neumaier <Arnold.Neumaier@univie.ac.at> wrote in message news:<40A77B89.2020606@univie.ac.at>...
> In case of a discrete measurement, a POVM is a family of Hermitian positive
> semidefinite operators P_r summing to 1 (the identity operator). The index
> r represents the possible measurement results. The probability that upon
> measuring a system with a detector described by this POVM you get the
> result r is p_r=<P_r> when the system is in the state <.>. These are
> classical probabilities, while the state is quantum. For example, for
> a pure state represented by the unit vector \psi, <f>=\psi^* f \psi.

OK, so while the example I gave was a POVM, the definition wasn't
inclusive enough.

Since they're Hermetian, the operators are observables, so I guess you
could calculate the standard deviation of one like you would for any
other observable.

--
Mike Stay
http://www.cs.auckland.ac.nz/~msta039

Arkadiusz Jadczyk
May21-04, 04:12 AM
<jabberwocky><div class="vbmenu_control"><a href="jabberwocky:;" onClick="newWindow=window.open('','usenetCode','toolbar=no, location=no,scrollbars=yes,resizable=yes,status=no ,width=650,height=400'); newWindow.document.write('<HTML><HEAD><TITLE>Usenet ASCII</TITLE></HEAD><BODY topmargin=0 leftmargin=0 BGCOLOR=#F1F1F1><table border=0 width=625><td bgcolor=midnightblue><font color=#F1F1F1>This Usenet message\'s original ASCII form: </font></td></tr><tr><td width=449><br><br><font face=courier><UL><PRE>\n\n\nOn Wed, 12 May 2004 18:41:09 +0000 (UTC), staym@datawest.net (Mike Stay)\nwrote:\n\n&gt;If so, the set {a^2, b^2, c^2} is not a POVM; how do you calculate the\n&gt;standard deviation of elements of a POVM?\n\nLike in the classical measure theory we calculate standard deviations\nof random variables, not of measures. Here you need to have an\nobservable as well, for instance\n\nEq.1: A= 1a+2b+3c\n\nThat means, in a given state psi you will have\n&lt;A&gt;_psi = 1&lt;psi,a psi&gt;+2&lt;psi,b psi&gt;+ 3&lt;psi,c psi&gt;\n\nWhen we represent A as above, we mean that the experimental procedure is\nsuch that, when measuring repeatedly "A" in the state psi we will be\ngetting as the results 1,2,3 with probabilities &lt;psi,a psi&gt;,&lt;psi,b\npsi&gt;,&lt;psi,c psi&gt; respectively. It follows that &lt;A^2&gt;_psi should\nbe calculated as\n\n&lt;A^2&gt;_psi=1&lt;psi, a psi&gt;+4&lt;psi, b psi&gt;+ 9&lt;psi, c psi&gt;\n\nark\n\n\n\n--\n\nArkadiusz Jadczyk\nhttp://www.cassiopaea.org/quantum_future/homepage.htm\n\n--\n</UL></PRE></font></td></tr></table></BODY><HTML>');"> <IMG SRC=/images/buttons/ip.gif BORDER=0 ALIGN=CENTER ALT="View this Usenet post in original ASCII form">&nbsp;&nbsp;View this Usenet post in original ASCII form </a></div><P></jabberwocky>On Wed, 12 May 2004 18:41:09 +0000 (UTC), staym@datawest.net (Mike Stay)
wrote:

>If so, the set {a^2, b^2, c^2} is not a POVM; how do you calculate the
>standard deviation of elements of a POVM?

Like in the classical measure theory we calculate standard deviations
of random variables, not of measures. Here you need to have an
observable as well, for instance

Eq.1: A= 1a+2b+3c

That means, in a given state \psi you will have
<A>_psi = 1<\psi,a \psi>+2<\psi,b \psi>+ 3<\psi,c \psi>

When we represent A as above, we mean that the experimental procedure is
such that, when measuring repeatedly "A" in the state \psi we will be
getting as the results 1,2,3 with probabilities <\psi,a \psi>,<\psi,b\psi>,<\psi,c \psi> respectively. It follows that <A^2>_psi should
be calculated as

<A^2>_psi=1<\psi, a \psi>+4<\psi, b \psi>+ 9<\psi, c \psi>

ark



--

Arkadiusz Jadczyk
http://www.cassiopaea.org/quantum_future/homepage.htm

--

Arnold Neumaier
May22-04, 04:49 AM
<jabberwocky><div class="vbmenu_control"><a href="jabberwocky:;" onClick="newWindow=window.open('','usenetCode','toolbar=no, location=no,scrollbars=yes,resizable=yes,status=no ,width=650,height=400'); newWindow.document.write('<HTML><HEAD><TITLE>Usenet ASCII</TITLE></HEAD><BODY topmargin=0 leftmargin=0 BGCOLOR=#F1F1F1><table border=0 width=625><td bgcolor=midnightblue><font color=#F1F1F1>This Usenet message\'s original ASCII form: </font></td></tr><tr><td width=449><br><br><font face=courier><UL><PRE>Mike Stay wrote:\n&gt; Arnold Neumaier &lt;Arnold.Neumaier@univie.ac.at&gt; wrote in message news:&lt;40A77B89.2020606@univie.ac.at&gt;...\n&gt;\n&gt;&gt;In case of a discrete measurement, a POVM is a family of Hermitian positive\n&gt;&gt;semidefinite operators P_r summing to 1 (the identity operator). The index\n&gt;&gt;r represents the possible measurement results. The probability that upon\n&gt;&gt;measuring a system with a detector described by this POVM you get the\n&gt;&gt;result r is p_r=&lt;P_r&gt; when the system is in the state &lt;.&gt;. These are\n&gt;&gt;classical probabilities, while the state is quantum. For example, for\n&gt;&gt;a pure state represented by the unit vector psi, &lt;f&gt;=psi^* f psi.\n&gt;\n&gt;\n&gt; OK, so while the example I gave was a POVM, the definition wasn\'t\n&gt; inclusive enough.\n&gt;\n&gt; Since they\'re Hermetian, the operators are observables, so I guess you\n&gt; could calculate the standard deviation of one like you would for any\n&gt; other observable.\n\nYes; it will give you the unavoidable uncertainty in their determination\nby a single measurement (where you get a 0 or 1). For the unavoidable\nuncertainty in the average of N independent measurements, divide by sqrt(N).\n\nArnold Neumaier\n\n</UL></PRE></font></td></tr></table></BODY><HTML>');"> <IMG SRC=/images/buttons/ip.gif BORDER=0 ALIGN=CENTER ALT="View this Usenet post in original ASCII form">&nbsp;&nbsp;View this Usenet post in original ASCII form </a></div><P></jabberwocky>Mike Stay wrote:
> Arnold Neumaier <Arnold.Neumaier@univie.ac.at> wrote in message news:<40A77B89.2020606@univie.ac.at>...
>
>>In case of a discrete measurement, a POVM is a family of Hermitian positive
>>semidefinite operators P_r summing to 1 (the identity operator). The index
>>r represents the possible measurement results. The probability that upon
>>measuring a system with a detector described by this POVM you get the
>>result r is p_r=<P_r> when the system is in the state <.>. These are
>>classical probabilities, while the state is quantum. For example, for
>>a pure state represented by the unit vector \psi, <f>=\psi^* f \psi.
>
>
> OK, so while the example I gave was a POVM, the definition wasn't
> inclusive enough.
>
> Since they're Hermetian, the operators are observables, so I guess you
> could calculate the standard deviation of one like you would for any
> other observable.

Yes; it will give you the unavoidable uncertainty in their determination
by a single measurement (where you get a or 1). For the unavoidable
uncertainty in the average of N independent measurements, divide by \sqrt(N).

Arnold Neumaier