By Keibock Lee, Park Systems Inc.
Introduction
One of the most challenging issues in the semiconductor industry is the failure analysis (FA) investigation of devices with enduringly shrinking geometries down to single digit nanometer trench widths. The complexity of such semiconductors means that the task of identifying sources of failure is extremely refined, sophisticated and difficult to perform. The most utilized current systems, such as scanning electron microscopes (SEMs) and focused ion beams (FIBs),have a limited measurement and property characterization modes beyond imaging of the sample, and destructive in high resolution. Therefore, the alternate solution of higher spatial resolution that can provide electrical property information has been more of a necessity than an option. The powerful conductive atomic force microscopy (AFM) is one of the technological advances in the field of AFM and provides an effective solution to the respective FA problems. However, the conventional conductive AFM technique requires a tradeoff between the spatial resolution and the current signal. Therefore, there was a need for the introduction of an innovative method that would allow FA engineers to overcome the disadvantages of conventional Conductive AFM. The introduction and the subsequent implementation of PinPoint Conductive AFM (PinPoint iAFM)by Park Systems offers frictionless conductivity scanning and excellent high spatial resolution and sensitivity with a very high signal-to-noise ratio. In this paper, I will explain how Conductive AFM can solve the FA problems faced by engineerswith respect to SRAM (static random access memory) cells as one of application examples. Subsequently,I will discuss the limitations of the conventional Conductive AFM and how the novel PinPoint iAFM effectively solves those issues, thereby providing the optimum solution for the FA engineers.
Failure Analysis in SRAM cells
Failure analysis is the process of investigating and determining/confirming the cause of failure in devices by means of electric measurements or physical and chemical analysis techniques. However, the continuously shrinking dimensions of semiconductors call for new solutions that can adequately and effectively clarify the failure mode (root cause) of such devices. A characteristic example that exhibits such a technological challenge is the failure analysis in SRAM cells. SRAM is a high-speed semiconductor memory that has extreme read-write endurance. A single SRAM cell contains several transistors that must all operate properly. Despite the fact that a dysfunctional cell within the embedded memory can be localized by using conventional methods such as SEM and FIB, the cause of failure itself is difficult, and in some cases impossible, to identify. This is because these techniques cannot ‘look’ inside the cell and thereby identify the faulty transistor.
In addition to the aforementioned localization problem, SEM probes expose both the tip and the sample to high mechanical stress because there is no control (feedback) mechanism (Grützner, 2005). Furthermore, the user cannot see the respective contact point between the tip and the sample. This translates into unavoidable errors once the analysis is performed in relatively small geometries. On the other hand, the electron beam may result in surface contamination and subsequently contact problems.Whereas with FIB, the sample preparation is very time-consuming and the access to the circuit nodes somewhat limited. The Conductive AFM meets this challenge for a more precise and effective failure analysis.
Conductive AFM in SRAM Failure Analysis
Conductive AFM is a measurement technology that provides data regarding both the topography and the electrical properties of the sample under investigation. Its principle of operation is very similar to conventional AFM, where a conductive cantilever probe scans the region of interest of the sample. A bias voltage is applied between the tip and the sample, and the tunnelling current between the two is measured. Therefore, the AFM’s tip deflection signal provides the topographical map, whilethe electric current amplifier measures the respective electric conductivity, as shown in Figure 2.
It should be noted that Conductive AFM sample preparation is relatively simple. Moreover, the contact size can be as small as a few nanometres, and hence the local variation of the electric property can be clearly observed (in contrast to the conventional macroscopic techniques). This automatically allows for an extremely fast and accurate characterization of single individual transistors within the SRAM cell. Therefore, this AFM technique extends the range of measurable geometries beyond their former limits. A typical Conductive AFM image of SRAM at -0.2 V is shown in Figure 3.
Limitations of conventional Conductive AFM
Conventional Conductive AFM suffers from several intrinsic problems, which means the user has two alternatives. The first is theacquisition of images with compromised spatial resolution due to the fast wearing of the metal-coated tip during contact mode topography.However, this also underlines the need for frequent AFM tipchange, hence adding more time and cost. The second alternative involves the sacrifice of current because of the short and limited time of contact between the tip and the sample. Nevertheless, in conventional Conductive AFMthere is a clear discrepancy between the forward and the backward scans. This ultimately provokesa poor reproducibility of the data,whichnow lack accuracy and precision, especially when it comes to consecutive measurements. Moreover, both the asymmetric signal,which is obtained due to the existence of lateral forces, and the small signal-to-noise ratio degrade the acquired measurements to a significant degree. Therefore, the introduction of PinPoint iAFM by Park Systems offers FA engineers an innovative solution to these difficult problems.
PinPoint iAFM – The innovative solution
Figure 4 depicts the actual implementation and apparatus of the Park Systems PinPoint iAFM that is operated in contact mode. The application of a bias between the tip and the sample produces a current flow that generates the respective AFM images.
The probe is monitoring the respective feedback signal, and hence follows an approach-and-retract function at each pixel, as shown in Figure 5. The tip will approach the surface of the sample until a predefined threshold point is reached.
It will then measure the height of the Z scanner while rapidly retracting the probe.This process ensures a frictionless mode of operation, and thusit provides a non-destructive operation for both the tip and the sample, which combines the best spatial resolution and current measurement (whereas, in conventional mode the user needs to compromise one of them). Moreover, the electrical contact between the surface of the sample and the AFM tip is very well defined and controllable. It can control both the contact force and the contact time. Additionally, the motion of the XY scanner stops during the current acquisition making the scan virtually frictionless. An example of a respective AFM image is shown in Figure 6.
Conclusions
Park SystemsPinPoint iAFM is an innovative conductive measurement technology for nano-scale measurement of electrical current flow that solves the limitations of SEM, FIB and conventional Conductive AFM for device failure analysis. The advantages offered by PinPoint iAFM are very significant because they overcome, and even eliminate, the respective difficulties that are present in conventional Conductive AFM. This novel AFM technique offers the lowest current noise level (< 0.1 pA), the maximum current available in the industry (10 mA), and the highest gain selection in the industry (it covers approximately seven orders of magnitude (106 – 1012). Furthermore, the controllable data acquisition time allows for a very high signal-to-noise ratio.Park’s PinPoint AFM is an extremely effective tool for the characterization of electrical defects in SRAM cells for failure analysis.
What are typical scan speed, bandwidth of measurement, contact time and force accuracy control ?
What are the improvement comparing existing solutions like Peak Force TUNA and or Peak Force SSRM ?
Regards,
Denis
What are typical scan speed, bandwidth of measurement, contact time and force accuracy control ?
> The typical scan speed would be about 2um/s at 1hz bandwidth. The contact time is adjustable from ms range and up. Force accuracy is in the range nN. The staying time for measuring the current from a specific site on a sample is precisely controlled.
What are the improvement comparing existing solutions like Peak Force TUNA and or Peak Force SSRM ?
> Unlike Peak Force Tuna or SSRM, there is no dragging or friction force with Park Systems’s PinPoint conductive AFM. That means with PinPoint, there is no degradation of tip sharpness and the measurement is accurate to that specific location with very high signal to noise ratio.
When trying to make electrical measurements using Peak Force you have an undefined contact. The unfiltered electrical data from Peak Force is very noise and changes greatly from tap to tap. PinPoint provides a simpler and more intuitive way to define the electrical contact between the tip and the sample.
I hope that helps.
Here’s a website link for more information:
http://www.parkafm.com/index.php/products/general-afm/park-nx20/afm-modes