Non-volatile memory is routinely used in computer systems for program and data storage. Extraction of information from these devices is used for software debugging, hardware failure analysis, chip reverse engineering where patent infringement is analyzed or for forensic purposes.
EEPROMS are one type of non-volatile memory consisting of arrays of memory transistors, which can either be electrically charged or uncharged, representing the two values of a binary digit – a Bit (1/0). In a scanning electron microscope electrically charged or un-charged transistors, create different image signals, based on the passive voltage contrast mechanism.
Here, we present a first study for bitwise reading of EEPROM memory by using a multi-beam scanning electron microscope and image processing algorithms.
A commercial off-the-shelf EEPROM chip was programmed with an alternating 0/1 Bit pattern and afterwards de-processed by a de-layering backside approach. Silicon substrate was removed by polishing and wet etching until the thin tunnel oxide layer of the floating gate transistors was exposed. Images were acquired using a 61-beam scanning electron microscope, the ZEISS MultiSEM 505, operated under strict low-dose conditions. Probe current was set to 20 pA per beam at 2.5 keV landing energy, pixel size was 15 nm and pixel dwell time was 100 ns. Sub-images of charged and uncharged memory cells were selected manually, aligned and averaged resulting in two class averages used for localizing cells and threshold definition. Cell localization was done by correlation, classification by applying a grey value threshold operation at each localized cell with the result that the initially set bit value pattern was confirmed.
With this feasibility study, we demonstrate that large area memory cell readout is in reach and will, in combination with automated frame averaging and classification algorithms, allow direct access to memory content and information stored on memory chips.

TitleDirect readout of EEPROM memory cells by passive voltage contrast multi-beam SEM imaging
Image caption

Readout of EEPROM chip a-c) localizing memory cells by correlation in multi-beam SEM passive voltage contrast image d-f) identification of charged (1/red) and uncharged cells (0/blue) by thresholding.

First nameStephan
Last nameNickell

Microscopy Business Group, Business Development mSEM, Carl Zeiss Microscopy GmbH, ZEISS Group, Carl-Zeiss-Straße 22, 73447 Oberkochen, Germany

Additional authors
Session12. Sensors and cameras II
Other Entries

No entries match your request.