Ion irradiation of the native oxide/silicon surface increases the thermal boundary conductance across aluminum/silicon interfaces

Caroline S. Gorham, Khalid Hattar, Ramez Cheaito, John C. Duda, John T. Gaskins, Thomas E. Beechem, Jon F. Ihlefeld, Laura B. Biedermann, Edward S. Piekos, Douglas L. Medlin, and Patrick E. Hopkins

Phys. Rev. B 90, 024301 – Published 3 July 2014

Abstract

The thermal boundary conductance across solid-solid interfaces can be affected by the physical properties of the solid boundary. Atomic composition, disorder, and bonding between materials can result in large deviations in the phonon scattering mechanisms contributing to thermal boundary conductance. Theoretical and computational studies have suggested that the mixing of atoms around an interface can lead to an increase in thermal boundary conductance by creating a region with an average vibrational spectra of the two materials forming the interface. In this paper, we experimentally demonstrate that ion irradiation and subsequent modification of atoms at solid surfaces can increase the thermal boundary conductance across solid interfaces due to a change in the acoustic impedance of the surface. We measure the thermal boundary conductance between thin aluminum films and silicon substrates with native silicon dioxide layers that have been subjected to proton irradiation and post-irradiation surface cleaning procedures. The thermal boundary conductance across the Al/native oxide/Si interfacial region increases with an increase in proton dose. Supported with statistical simulations, we hypothesize that ion beam mixing of the native oxide and silicon substrate within $\sim 2.2 nm$ of the silicon surface results in the observed increase in thermal boundary conductance. This ion mixing leads to the spatial gradation of the silicon native oxide into the silicon substrate, which alters the acoustic impedance and vibrational characteristics at the interface of the aluminum film and native oxide/silicon substrate. We confirm this assertion with picosecond acoustic analyses. Our results demonstrate that under specific conditions, a “more disordered and defected” interfacial region can have a lower resistance than a more “perfect” interface.

Received 9 February 2014

DOI:https://doi.org/10.1103/PhysRevB.90.024301

Authors & Affiliations

Caroline S. Gorham¹, Khalid Hattar², Ramez Cheaito¹, John C. Duda^1,*, John T. Gaskins¹, Thomas E. Beechem², Jon F. Ihlefeld², Laura B. Biedermann², Edward S. Piekos², Douglas L. Medlin³, and Patrick E. Hopkins^1,†

¹Department of Mechanical and Aerospace Engineering, University of Virginia, Charlottesville, Virginia 22904, USA
²Sandia National Laboratories, Albuquerque, New Mexico 87123, USA
³Sandia National Laboratories, Livermore, California 94550, USA

^*Current address: Seagate Technology, Bloomington, Minnesota 55435, USA.
^†phopkins@virginia.edu

Article Text (Subscription Required)

Click to Expand

References (Subscription Required)

Click to Expand

Issue

Vol. 90, Iss. 2 — 1 July 2014

Reuse & Permissions

Access Options

Author publication services for translation and copyediting assistance advertisement

Images

Figure 1
Thermal boundary conductance as a function of proton dose for the various Al/Si interfaces that are examined in this work. The surface treatments are indicated in the figure (no cleaning: open squares, alcohol clean: open diamonds, alcohol clean followed by 30 min plasma clean: filled triangles, alcohol clean followed by 60 min plasma clean: filled circles). In the nonirradiated samples, the factor of nearly 2 increase in $h_{K}$ when the substrate is cleaned indicates the importance of cleaning as-received wafers to remove contaminants that can affect $h_{K}$ . The decrease in $h_{K}$ observed in the ion-irradiated samples that were not subjected to plasma cleaning is due to a build up of carbon contamination on the silicon surface [23, 24]. When the carbonaceous layer is removed via plasma cleaning, an increase in $h_{K}$ is observed. We attribute this increase to ion beam mixing of the native oxide layer with the silicon, which is discussed in detail in the text. The inset shows the measured thermal conductivities of the various silicon substrates as a function of ion dose.
Reuse & Permissions
Figure 2
(a) HAADF-STEM images of the interfacial layer between the Al and Si for the (i) nonirradiated alcohol cleaned, (ii) irradiated with $10^{14}$ protons $cm^{- 2}$ and alcohol cleaned, and (iii) irradiated with $10^{14}$ protons $cm^{- 2}$ and plasma cleaned samples. (b) The thicknesses of the interfacial layers, as measured from our TEM analysis, depends on the ion dose when a plasma clean is not used. This is consistent with our observation of carbon contamination building up at the interface with ion dose. When the samples are subjected to a plasma clean post-irradiation, no dose dependence is observed in the interfacial layer thicknesses. (c) Representative EDS profile that shows an enrichment in oxygen at the interfacial layer. This data set was taken on the sample that was irradiated with $10^{14}$ protons $cm^{- 2}$ and plasma cleaned. The onset of the oxygen signal also coincides with the onset of the aluminum signal, indicating the formation of an aluminum-oxide-rich layer from interaction of the Al with the $SiO_{2}$ native oxide on the surface of the silicon substrate.
Reuse & Permissions
Figure 3
TDTR data focused on the region of the data sets from which picosecond acoustics are monitored to lend insight into the interfacial quality and compositional changes from the different surface treatments (offset for clarity). At the Al/Si interface, a decrease (dip) then increase (hump) is expected due to a strain reflection from the Al off of the native oxide layer and then off of the Si substrate. A dip corresponds to an acoustically softer layer than the Al, originating from either the native oxide, the contamination on the as-received sample (open triangles), or the carbon buildup from the irradiation process (filled circles). Note that this negative trough is diminished with an alcohol clean on the as-received sample (filled triangles), but still observable consistent with the presence of a native oxide. At the plasma cleaned and irradiated interface (filled squares), the amplitude of the strain wave reflection is diminished, which we attribute to a smearing of the native oxide layer and an acoustic impedance of the mixing layer that is closer to that of Al and Si.
Reuse & Permissions
Figure 4
(a) Picosecond acoustic residuals as a function of pump-probe delay time from our TDTR data on the plasma cleaned samples subjected to various proton doses. The various proton doses are indicated in the figure; the control sample was not subjected to any ion irradiation. An increase in proton dose leads to a decrease in the reflected amplitude of the acoustic wave off of the Al/oxide/Si interface. This indicates a shift in the low-frequency vibrational properties of the silicon near the surface. (b) Peak-to-peak amplitude of the picosecond acoustic residuals shown in (a) as a function of ion dose.
Reuse & Permissions

Physical Review B

covering condensed matter and materials physics