Ion temperature gradient (ITG) and electron temperature gradient (ETG) modes
are possible sources of anomalous transport in fusion plasmas. Despite
their similarity in the linear regime, their nonlinear behavior differs
significantly and is full of surprises. Comparing the characteristics of
these two basic types of plasma dynamics, novel aspects about the nature
of temperature gradient driven turbulence are revealed. Recent developments
in this area will be presented, based on nonlinear gyrokinetic Vlasov
simulations.

First, fully developed ITG and ETG turbulence will be characterized in
terms of its statistical properties like amplitudes, spectra, correlations,
and energy transfer paths. Pattern formation will be briefly discussed,
highlighting the nature of high-amplitude radial streamers and the role
of self-generated zonal flows and fields. Based on these simulation results,
an analytic model is developed which captures several properties of
adiabatic ETG and ITG turbulence reasonably well. A similar approach
may well turn out to be useful in a variety of other turbulence problems.

About 25 years ago, Ohkawa speculated about the role of turbulent
fluctuations on collisionless skin depth scales, triggering a large
number of studies in that area. Among other things, it was proposed that
ETG modes might lead to significant transport by magnetic flutter at
just those scales. However, nonlinear gyrokinetic simulations have
shown that this scenario is not realized, i.e., ExB
transport tends to clearly dominate. On the other hand, Ohkawa scaling might also
be obtained via finite beta effects on the ETG turbulence itself. It will be
shown that the latter result from a subtle competition between long-wavelength
primary modes and strong secondary instabilities driven by poloidal gradients
in the primary's parallel electron velocity component.

In the last part of this presentation, issues concerning the cross-scale
coupling of various forms of plasma turbulence will be addressed. To this
aim, we have performed the first gyrokinetic studies including both ion
and electron spatio-temporal scales in a single simulation. Several direct
and indirect (i.e., zonal flow mediated) coupling mechanisms are investigated
and discussed. The simulation of cross-scale coupling represents a demanding
computational task, pushing both the software and the hardware to its limits.
The employed GENE code runs efficiently on multiple platforms, achieving,
e.g., 280 MFlop/s per processor on a Hitachi SR-8000 and more than 500 MFlop/s
per processor on an IBM Regatta system. Nevertheless, well-resolved two-scale
runs require are of the available resources of present-day supercomputers.