15.08.2013 Views

shot noise in mesoscopic conductors - Low Temperature Laboratory

shot noise in mesoscopic conductors - Low Temperature Laboratory

shot noise in mesoscopic conductors - Low Temperature Laboratory

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

24 Ya.M. Blanter, M. Bu( ttiker / Physics Reports 336 (2000) 1}166<br />

The current is taken to be positive if it #ows from the reservoir towards the <strong>mesoscopic</strong> structure.<br />

For the average current <strong>in</strong> the two-term<strong>in</strong>al geometry, we have I #I "0. We emphasize that<br />

<br />

current conservation must hold not only on the average but at each <strong>in</strong>stant of time. In particular,<br />

current conservation must also hold for the #uctuation spectra which we discuss subsequently. In<br />

general, for time-dependent currents, we have to consider not only contacts which permit carrier<br />

exchange with the conductor, but also other nearby metallic structures, for <strong>in</strong>stance gates, aga<strong>in</strong>st<br />

which the conductor can be polarized. The requirement that the results are gauge <strong>in</strong>variant means<br />

<strong>in</strong> this context, that no current arises if voltages at all reservoirs are simultaneously shifted by the<br />

same value (and no temperature gradient is applied). For the average currents (see Eqs. (39), (47))<br />

both properties are a direct consequence of the unitarity of the scatter<strong>in</strong>g matrix.<br />

For the conductance matrix G current conservation and gauge <strong>in</strong>variance require that the<br />

<br />

elements of this matrix <strong>in</strong> each row and <strong>in</strong> each column add up to zero,<br />

G " G "0 . (48)<br />

<br />

<br />

Note that for the two term<strong>in</strong>al case this implies G,G "G "!G "!G . In the two<br />

<br />

term<strong>in</strong>al case, it is thus su$cient to evaluate one conductance to determ<strong>in</strong>e the conductance<br />

matrix. In multi-probe samples the number of elements one has to determ<strong>in</strong>e to "nd the conductance<br />

matrix is given by the constra<strong>in</strong>ts (48) and by the fact that the conductance matrix is<br />

a susceptibility and obeys the Onsager}Casimir symmetries<br />

G (B)"G (!B) .<br />

<br />

In the scatter<strong>in</strong>g approach the Onsager}Casimir symmetries are aga<strong>in</strong> a direct consequence of the<br />

reciprocity symmetry of the scatter<strong>in</strong>g matrix under "eld reversal.<br />

In the stationary case, the current conservation and the gauge <strong>in</strong>variance of the results are<br />

a direct consequence of the unitarity of the scatter<strong>in</strong>g matrix. In general, for non-l<strong>in</strong>ear and<br />

non-stationary problems, current conservation and gauge <strong>in</strong>variance are not automatically<br />

ful"lled. Indeed, <strong>in</strong> ac-transport a direct calculation of average particle currents does not yield<br />

a current conserv<strong>in</strong>g theory. Only the <strong>in</strong>troduction of displacement currents, determ<strong>in</strong>ed by the<br />

long-range Coulomb <strong>in</strong>teraction, leads to a theory which satis"es these basic requirements. We will<br />

discuss these issues for <strong>noise</strong> problems <strong>in</strong> Section 3.<br />

2.4. General expressions for <strong>noise</strong><br />

We are concerned with #uctuations of the current away from their average value. We thus<br />

<strong>in</strong>troduce the operators IK (t),IK (t)!I .Wede"ne the correlation function S (t!t) of the<br />

<br />

current <strong>in</strong> contact and the current <strong>in</strong> contact as<br />

S (t!t),IK (t)IK (t)#IK (t)IK (t) . (49)<br />

<br />

Note that several de"nitions, di!er<strong>in</strong>g by numerical factors, can be found <strong>in</strong> the literature. The one we use<br />

corresponds to the general de"nition of time-dependent #uctuations found <strong>in</strong> Ref. [11]. We de"ne the Fourier transform<br />

with the coe$cient 2 <strong>in</strong> front of it, then our normalization yields the equilibrium (Nyquist}Johnson) <strong>noise</strong> S"4k ¹G<br />

and is <strong>in</strong> accordance with Ref. [1], see below. The standard de"nition of Fourier transform would yield the<br />

Nyquist}Johnson <strong>noise</strong> S"2k ¹G. Ref. [9] de"nes the spectral function which is multiplied by the width of the<br />

frequency <strong>in</strong>terval where <strong>noise</strong> is measured.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!