Coulomb drag near the metal-insulator transition in two dimensions

We studied the drag resistivity between dilute two-dimensional hole systems, near the apparent metal-insulator transition. We find the deviations from the $T^2$ dependence of the drag to be independent of layer spacing and correlated with the metalliclike behavior in the single-layer resistivity, suggesting they both arise from the same origin. In addition, layer-spacing dependence measurements suggest that while the screening properties of the system remain relatively independent of temperature, they weaken significantly as the carrier density is reduced. Finally, we demonstrate that the drag itself significantly enhances the metallic $T$ dependence in the single-layer resistivity.

Figure 1

${\rho }_D∕T^2$ vs $T$ at $p_m=2.5\times {10}^{10}{\mathrm{cm}}^{-2}$, for different $d$. For clarity, data from the $d=450\mathrm{\AA }$ sample has been multiplied by a factor of 6.5. The dashed line marks the peak position. Open circles measured in the dilution refrigerator.

Figure 2

${\rho }_D$ vs $p_m$ on log-log scale, at $T=300\mathrm{mK}$. Data from samples A $(d=300\mathrm{\AA })$ and D $(d=275\mathrm{\AA })$ are shown by closed and open circles, respectively. The solid line is a fit with slope $-5$. Inset (a): ${\rho }_D$ vs $p_1$ on log-log scale, with $p_2=2.15\times {10}^{10}{\mathrm{cm}}^{-2}$, at $T=80$ and $300\mathrm{mK}$. Solid lines are fits with slope close to $-2.5$. Dashed line indicates matched density. Inset (b): ${\rho }_D$ vs $V_{\mathit{\text{bias}}}$. $p_m=2.1\times {10}^{10}{\mathrm{cm}}^{-2}$ and $T=230\mathrm{mK}$.

Figure 3

(Color online) $\rho \left(T\right)∕\rho \left(T_0\right)$ vs $T$ at $p=2.5\times {10}^{10}{\mathrm{cm}}^{-2}$ for different $d$. Inset: $[\rho \left(T\right)-{\rho }_D\left(T\right)]∕[\rho \left(T_0\right)-{\rho }_D\left(T_0\right)]$ vs $T$. Dashed line marks the peak position in ${\rho }_D∕T^2$ vs $T$. $T_0=400\mathrm{mK}$.

Figure 4

${\rho }_D$ vs $d$, on log-log scale, for $p_m=1.5$, 2.0, and $2.5\times {10}^{10}{\mathrm{cm}}^{-2}$ at $T=1.0\mathrm{K}$. Solid lines are the best linear fits of each data set.

Figure 5

$\alpha$ vs $T$ for $p_m=1.5$, 2.0, and $2.5\times {10}^{10}{\mathrm{cm}}^{-2}$. $\alpha$ deduced from fitting ${\rho }_D$ vs $d$ to a $d^{-\alpha }$ fit.