MANSFIELD, Circuit Judge:
National Micronetics, Inc. (“National”) appeals that portion of a judgment of the District Court for the Southern District of New York, Robert J. Ward, Judge, holding valid certain claims of U.S. Patent No. 3,246,383 (“the Peloschek Patent”). Appellant does not challenge the district court’s additional finding that it infringed the patent.
The plaintiff, U.S. Philips1 Corporation, commenced this action against National on March 2,1971, seeking damages and injunctive relief as assignee of three allegedly infringed patents. National denied infringement, challenged the validity of all three patents, and asserted counterclaims for patent misuse and antitrust violations against plaintiff and its assignors, North American Philips and N.V. Philips Gloeilarnpenfabriken. Prior to trial plaintiff withdrew one infringement claim and Judge Ward severed and stayed the counterclaims pending the outcome of the infringement action.
On January 27, 1976, after a trial without a jury, the district court issued a 50-page opinion finding one of the patents invalid and that claims 1-A, 6 and 8-11 of the Peloschek patent were valid and infringed. National’s appeal is limited to the issue of the validity of the Peloschek patent under 35 U.S.C. § 103, which states that:
“A patent may not be obtained . if the differences between the subject matter sought to be patented and the prior art are such that the subject matter as a whole would have been obvious at the time the invention was made to a person having ordinary skill in the art to which said subject matter pertains.”
The Peloschek Patent
The Peloschek patent’s subject matter is a process for the manufacture of a magnetic recording head, which is that part of a tape recorder or computer that translates electrical impulses into magnetic patterns on a tape for the purposes of storage. In reverse, the recording head can “read” the magnetic patterns from the tape, meaning that it retranslates them into electrical impulses which then become sounds, pictures, or a computer printout.
A magnetic recording head consists of two basic parts, the first constructed of magnetic material and the second of a nonmagnetic material bonded to the first as a “gap” which permits the electrical impulses flowing through the recording head to create magnetized patterns on the tape opposite the gap. The magnetic heads at issue in this case generally have a magnetic portion constructed of two ferrite pieces, bonded together by a non-magnetic portion, usually of a glass-type material. (See Exhibit A hereto). The process for which the Pelos-chek patent was obtained describes a method of bonding the two ferrite pieces and the gap material to create a single-piece magnetic head with a minute non-magnetic gap.
The recording head and the gap in it must be of near-microscopic dimensions in order to maximize the amount of information which can be stored on a given portion of tape. The gap, which determines the size of the magnetized spot on the tape, generally measures no more than 50 to 100 microinches or millionths of an inch. (An average sheet of paper is about 3,000 mi-croinches thick.) The evidence shows that for a given recording head to be able accurately to read or retranslate information which has originally been stored on a tape through use of a different recording head, the gap dimensions in each head must be virtually identical, with allowable deviations or tolerances of no more than 10 to 20 millionths of an inch for the larger heads, and much less in smaller recording heads. [718]*718Since numerous recording heads are used in a single computer disc file, it is of vital importance to have all heads conform to prescribed gap dimensions within a few millionths of an inch. Otherwise some of the essential recording heads will not function as the system requires.
For nearly a decade prior to the issuance of the Peloschek patent the recorder-manufacturing industry was concerned over the inability to produce glass-bonded heads which would uniformly meet the precise specifications and conform to the narrow tolerances that were essential to successful performance and interchangeability. The Peloschek patent addresses itself to this problem. Filed May 3, 1963, and issued April 19, 1966, the patent states its objective as follows:
“Magnetic heads with very short gaps having lengths between 1 and 20 microns are difficult to manufacture since close tolerances are usually imposed on the length of the gap and the non-magnetic material in the gap must have a good resistance to detrition and must be capable of being readily processed; in addition, the process of manufacture should be as economical and simple as possible. The gap material also must be as homogeneous as possible. It is the primary object of the invention to provide a method of manufacturing magnetic heads with very short gap lengths which is comparatively simple to perform while achieving close tolerances for gap length.” (Emphasis added).
The claims which were sustained by the district court are:
“1. A method of manufacturing portions of magnetic heads composed of two magnetic circuit parts consisting of sin-tered oxidic ferromagnetic material and having confronting gap surfaces with a gap therebetween filled with a nonmagnetic material bonding the circuit parts together, comprising: placing spacing members having a thickness equal to the desired gap length at opposite ends of a first polished gap surface of one circuit part, placing a corresponding polished gap surface of a second circuit part on said spacing members in confronting relationship with said first surface thereby forming a gap between said surfaces, placing a quantity of nonmagnetic material adjacent to the gap, said nonmagnetic material having a melting temperature below that of said ferromagnetic material, and heating the resulting assembly to the melting temperature of said nonmagnetic material, whereby said nonmagnetic material melts, fills the gap by capillary action, and bonds the circuit parts together.
“2. A method according to claim 1, wherein said nonmagnetic material is glass.
“3. A method according to claim 1, wherein said magnetic material is enamel.
“4. A method according to claim 1, wherein pressure is applied to the assembly during'the heating step.”
Claims 6 and 8 describe the use of the process in making a multiple of recording heads, and claims 10-11 omit the use. of spacing members in the process.
The claimed manufacturing process for the bonding of the magnetic and nonmagnetic portions of the recording heads may be summarized as: (1) presetting the gap or space between two ferrite pieces at a desired length by placing the surfaces of the ferrite pieces opposite each other, inserting spacing members or shims of precise given dimensions between them, and bringing the ferrite surfaces firmly together against the spacers or shims, which serve to fix the gap length at a predetermined distance; (2) placing the nonmagnetic material (e. g., glass) adjacent to the space between the ferrite pieces; (3) heating the assembly to the melting point of the nonmagnetic material (glass); and (4) permitting the melted nonmagnetic material (glass) to flow into and fill up the gap, with the result that the two ferrite pieces are bonded together by nonmagnetic material of a predetermined thickness. The flowing action is created by capillarity, a principle of physics which is defined by Webster’s New Collegiate Die-[719]*719tionary (7th ed.) as “The action by which the surface of a liquid where it is in contact with a solid is elevated or depressed depending on the relative attraction of the molecules of the liquid for each other and for those of the solid.”
The district court, in a well-reasoned decision, held that the subject matter of the Peloschek patent would not have been obvious to one skilled in the art of glass bonding at the time it was made, observing
“The crucial feature ... is the use of capillary action to fill a preset gap of precise, reproducible dimensions. Nowhere in the prior art . is there any disclosure which would make it clear to one skilled in the art that such an application of capillary action would be successful.”
The district court refused to accord weight to National’s showing that capillary action had been disclosed in patents in the electronics and recording industry. “The prior art indicates the widespread use of capillary action to fill minute gaps, but in none is the precise size of the gap of great importance or predetermined.”
Appellant claims that the capillary process was “obvious” under 28 U.S.C. § 103 and that the district court erred in inferring from the evidence that the Peloschek patent claimed a level of gap precision and uniformity greater than that found in the prior art. We disagree, and affirm the district court’s finding of validity.
Scope of Review
To resolve the issue whether the invention was obvious within the meaning of 35 U.S.C. § 103, we must examine the scope and content of the prior art, the differences between the prior art and the claims at issue, and the level of ordinary skill in the pertinent art. Graham v. John Deere Co., 383 U.S. 1, 17, 86 S.Ct. 684, 15 L.Ed.2d 545 (1966).1 In doing so, we recognize that the ultimate issue of the validity of a patent is a question of law, and that the district court’s findings of fact will not be disturbed unless clearly erroneous, except that where they are based on undisputed facts or documentary evidence, we may substitute our own inferences and conclusions for those of the district judge. Maclaren v. B-I-W Group, Inc., 535 F.2d 1367, 1371 (2d Cir.), petition for cert. filed, - U.S. -, 97 S.Ct. 531, 50 L.Ed.2d 612 (1976).
The Prior Art
Prior art processes for the manufacture of magnetic recording heads were aimed principally at the discovery or development of a manufacturing process which would produce uniformity in the non-magnetic gap lengths within the tolerances or permitted ranges of deviation that are essential to the proper functioning of recording heads. Another aim was to insure that the nonmagnetic material used to fill the gap would not suffer from irregularities, such as bubbles or other defects that would impair the magnetic effectiveness of the ferrite pieces or core. The prior art history, as disclosed in a series of patents issued during the period 1954 to 1962, is one of repeated efforts to solve these problems, which were for the most part unsuccessful. Although some improvements were achieved in the quality of the recording heads produced, the processes generally resulted in low yields of acceptable heads, ranging from 5% to 50%.
The earliest of the patents for manufacture of magnetic recording heads introduced at trial was No. 2,919,312, issued to Georg Rosenberger, et al., on December 29, 1959, on an application filed March 16, 1954. It claimed a process of manufacturing such heads with a “limited and well defined very narrow working gap,” through the cementing of non-magnetic filler material between the ferrite pieces. In 1956, seven years prior to the filing of the Peloschek patent, Simon Duinker and Jules Bos filed a patent [720]*720application entitled “Glass Gap Spacer for Magnetic Heads.” This application, for which patent No. 3,024,318 (Duinker ’318) was issued in 1962, accurately disclosed the state of the art of manufacturing magnetic recording heads at the time. The patent substituted a glass foil material in place of other gap materials. In discussing “the requirements to be satisfied by the width of the gap,”2 the application stated that “in the present state of the art with respect to magnetic recordings, these requirements have become comparatively exacting and the disadvantages attendant on the use of such [foil spacing plates] during manufacture are becoming increasingly marked.” Duinker ’318, by its own terms, sought to achieve the “correct gap width” by placing a thin glass foil between the ferrite surfaces, heating the assembly, and applying pressure to bond the ferrite and the glass, forming the gap. This process became known colloquially as the “sandwich” technique, as distinguished from Rosenberger, which simply used cement.
In patent No. 3,094,772 (Duinker ’772), entitled “Method of Producing Magnetic Heads with Accurately Predetermined Gap Heights,” Duinker once again sought to attain a “well-defined” gap through revision of the technique employed in ’318. He criticized the prior art on the grounds that “the gap height . . . cannot be determined within narrow limits” and stated “[i]t is also difficult if not impossible to produce small gap widths.” Essentially Duinker ’772 claimed that it improved the accuracy of the gap lengths through use of more accurately placed and finished ferrite pieces.
In 1963 patent No. 3,079,470, for which application had been made in 1959, was issued to Marvin Camras of the Armour Research Foundation of the Illinois Institute of Technology. The object of this process, as with its predecessors, was to construct a magnetic recording head having “a very precise stable gap structure capable of maintaining precise gap dimensions.” This time, the inventor introduced a low melting point gap material and proposed the use of an “evaporated film” as a preliminary gap spacer in the process. The bonding procedure under this patent continued to employ a sandwich-like pressure of the heated gap materials to the ferrite.
In 1959 Duinker, still trying to find a way of solving the problem of producing uniform gaps of non-magnetic material, applied for a patent which further refined but failed to depart from the sandwich technique. Duinker now proposed to add “shims” or spacers made of mica to help attain a “correct gap length.” In discussing the disadvantages of the prior art, the application, for which patent No. 3,117,367 was issued on January 14, 1964, stated that “the adjustment of the gap length to the correct value is not possible when the glass flows away.” 3 However, the spacers were largely ineffective in achieving their objective because the glass foil inserted between the faces of the ferrite pieces was thicker than the shims so that when heat and pressure were applied, the gap length was sometimes altered by the presence of glass between the ferrite and the spacers or by the movement of the spacers.
A patent issued to Robert Pfost of the Ampex Corporation on November 8, 1966, on an application filed April 17, 1961, sought a process for the manufacture of “ferrite core heads having accurate gap dimensions.” It described the prior art techniques as “tedious, time-consuming, and uneconomical because [the heads] must be manufactured individually to provide proper gap dimensions.” The claimed process in [721]*721this patent consisted essentially of placing spacer strips on one of the ferrite surfaces, a layer of glass on the other surface, and joining them together under high temperature and pressure.
Finally, in an application filed in July, 1962, less than a year prior to the filing of the Peloschek patent, James S. Hanson, an employee of I.B.M., claimed the invention of a process capable of producing a “more predictable gap,” and “permitting manufacture of gaps which are uniform throughout the entire dimension thereof.” Hanson’s patent, No. 3,217,305, sought to improve the distribution of the glass bonding material within the gap by employing grooves in the ferrite surface. His basic process, like those of his predecessors, was the sandwich technique.
Thus the picture of the art of manufacturing magnetic recording heads prior to Peloschek is one in which, although there were some advancements, each of the processes immediately preceding Peloschek employed variations on a method which bonded the magnetic and non-magnetic materials through the use of heat and application of pressure in a sandwich configuration, with several of the methods using spacers and shims in an effort to set the desired gap size. Each patent criticized the prior art as unable to attain the reproducibility of uniformly accurate gap dimensions. None of these prior art processes, moreover, was able uniformly to meet the stringent gap-length tolerances required by manufacturers of magnetic recording heads. Due to the higher pressures used in the sandwich method the ferrite gap surfaces moved toward one another, resulting in variations in the gap. Although spacers or shims were used in an effort to hold the ferrite faces apart, the spacers could not uniformly control the gap length because the glass foil inserted between the ferrite faces was thicker than the spacers and the pressure applied to the ferrite pieces varied. As a result, viscous glass would sometimes enter between the spacers and the ferrite pieces or the spacers would move outward rather than remain in place, causing unacceptable variations in the gap length.
The Peloschek patent differs significantly from these prior art processes used to manufacture recording heads. First, the Pelos-chek patent places the glass bonding material outside of the pre-set gap rather than between the pieces of ferrite. Second, it departs from the use of pressure to accomplish bonding4 and relies instead upon the force of capillarity, an element previously unused in the art of manufacturing magnetic recording heads. Additionally, undisputed testimony by plaintiff’s expert witness and by the co-inventor of the Pelos-chek process established that, for the first time, yields of 90% accurate gap dimensions were obtained in an economically feasible production process, as compared to only 5% to 50% under prior art processes.
As might be expected of one challenging the validity of a patent on grounds of obviousness under § 103, National seeks to broaden the scope of the prior art'beyond processes for the manufacture of recording heads to those using the principle of capillarity for other manufacturing purposes, such as to cause molten metal or liquid epoxy cement to flow into random irregular spaces or into voids between metal parts held in contact with one another. National contends that these processes show that the use of capillarity was predictable and obvious to one ordinarily skilled in the art of manufacturing recording heads.
Whether these latter processes, which do not involve use of glass or ferrite and are not aimed at creating minute non-magnetic gaps of precise, predetermined size, are pertinent prior art for present purposes is extremely doubtful. However, we need not resolve that issue for the reason that even if these processes are considered to be prior art, the proof is overwhelming that they do not suggest to one ordinarily skilled in the development of processes for the manufac[722]*722ture of magnetic recording heads that capillarity might be used in combination with existing processes as a means of producing very minute gap dimensions within the narrow tolerances demanded by the trade. The patent issued to Grant (No. 2,500,748), for instance, is heavily relied on by National because it claims the use of capillary action to form non-magnetic gaps in a magnetic structure. Despite the superficial similarity of Grant to Peloschek, however, the differences between the two patents make it readily apparent that the capillarity concept would not have been apparent to one searching for a means of filling a pre-set gap of precise, very minute, dimensions.
Grant was merely concerned with filling random metal voids with epoxy or metal, not with near-microscopic manufacture of exactly reproducible gap lengths of fixed dimensions, with only a few millionths of an inch variance. Indeed, in Grant there is no such thing as pre-fixed spacing or an effort to create a predetermined gap, much less one of specific dimensions. Moreover, the prior art (Zinke, “Technologie der Glas Ver-schurelzungen”) taught that molten glass was unsuitable for capillary action because it tended to remain at the area of application and not to flow into narrow interstices. The apparent inappropriateness of capillary action as a means of solving the problem faced by Peloschek and his predecessors is further evidenced by the fact that, although Grant was issued in 1947, and numerous persons skilled in the art of recording head manufacture, including Duinker, had sought over a period of more than ten years to solve the problem of creating reproducible minute non-magnetic gaps of precise predetermined size, there is no evidence that any attempt was made to use capillarity
DISCUSSION
As we recently said in Timely Products Corporation v. Arron, 523 F.2d 288, 294 (2d Cir. 1975):
“We can conceive of no better way to determine whether an invention would have been obvious to persons of ordinary skill in the art at the time than to see what such persons actually did or failed to do when they were confronted with the problem in the course of their work. If the evidence shows that a number of skilled technicians actually attempted, over a substantial period, to solve the specific problem which the invention overcame and failed to do so, notwithstanding the availability of all the necessary materials, it is difficult to see how a court could conclude that the invention was ‘obvious’, to such persons at the time.”
In Timely Products we found that “no such evidence exists.” Here, in contrast, the evidence clearly demonstrates that the industry’s search for a process which would accurately reproduce minute gap dimensions had been the subject of continuing experimentation for nearly a decade, and that the object was successfully accomplished only when Peloschek abandoned the sandwich techniques in favor of the use of capillarity. Moreover, in contrast to Ma-claren v. B-I-W Group, Inc., supra, this innovation was not foreshadowed by any of the preceding patents for the manufacture of recording heads.
Accordingly, we conclude that the Pelos-chek patent is significantly different from the prior art and represents a distinct advancement in the level of skill in the art. The district court’s conclusion that the Pe-loschek claims were not obvious is therefore correct. Moreover, it is consistent with the terms of 35 U.S.C. § 103 and the cases interpreting its provisions. Even assuming that patents disclosing capillary action, such as Grant, were deemed part of the pertinent prior art, so that the Peloschek patent would represent a combination of existing elements rather than a pure innovation,5 the invention would still satisfy the “rigorous” standards of § 103. Lemelson v. Top[723]*723per Corp., 450 F.2d 845, 848 (2d Cir. 1971); Maclaren v. B-I-W Group, Inc., supra.
The existence of an important problem in the art which has remained unsolved for a long period, despite continued efforts and a series of refinements of the art, until a new combination of concepts produces a solution, is evidence that the combination was not obvious. Shaw v. E. B. & A. C. Whiting Co., 417 F.2d 1097, 1104 (2d Cir. 1969), cert. denied, 397 U.S. 1076, 90 S.Ct. 1518, 25 L.Ed.2d 811 (1970). This case falls under the rule that
“[i]f those skilled in the art are working in a given field and have failed after repeated efforts to discover a particular new and useful improvement, the person who first makes the discovery does more than make the obvious improvement which would suggest itself to a mechanic skilled in the art, and is entitled to protection as an inventor.” McCullough Tool v. Well Surveys, Inc., 343 F.2d 381, 399 (10th Cir. 1965), cert. denied, 383 U.S. 933, 86 S.Ct. 1061, 15 L.Ed.2d 851 (1966).
See also A. E. Staley Manufacturing Co. v. Harvest Brand, Inc., 452 F.2d 735, 738 (10th Cir. 1971), cert. denied, 406 U.S. 974, 92 S.Ct. 2415, 32 L.Ed.2d 674 (1972).
The cases upon which appellant relies are clearly distinguishable. In Anderson’s-Black Rock, Inc. v. Pavement Salvage Co., Inc., 396 U.S. 57, 60, 90 S.Ct. 305, 24 L.Ed.2d 258 (1969), the Supreme Court rejected a combination of old elements to solve a problem in “blacktop” paving processes on the grounds that the combination was “not critical or essential” to curing the problem and did not produce a “new or different function.” The Court rested its recent decision in Sakraida v. Ag Pro, Inc., 425 U.S. 273, 96 S.Ct. 1532, 47 L.Ed.2d 784 (1976), on this same finding with respect to a barn-cleaning combination patent. In contrast to those cases, the evidence in the present case is unrebutted that the Peloschek patent ■produced a result unobtainable with the prior processes. For the first time a process functioned to produce high yields of accurate gap dimensions, curing a problem in the art.
In Dann v. Johnston, 425 U.S. 219, 96 S.Ct. 1393, 47 L.Ed.2d 692 (1976), the Court found that the gap between the prior art and the claimed invention in systems used to record a breakdown of banking customers’ transactions by the nature of the transaction was not great enough to justify pat-entability of the respondent’s system because it failed to meet the non-obviousness standard. There, all systems sorted out the transactions, albeit by slightly different methods. In the present case, on the other hand, the gap between the prior art and the Peloschek patent is enormous — the difference between success and failure. In lieu of an unsuccessful re-refinement of an existing process, Peloschek reveals a flash of .'brilliance which found a solution, however simple, by departing from the norm. It is • this type of advancement that has traditionally been rewarded with patent rights, and to deny those rights here might unjustifiably deter industry members from seeking to invest, innovate and experiment.6
In its most favorable light, National’s evidence merely shows that the elements of the Peloschek patent existed in the prior art of manufacturing electronic and magnetic devices. We have consistent- : ly held that such a showing alone is inadequate to demonstrate obviousness when the [724]*724combination of those pre-existing elements results in novel, unanticipated or long-sought results. See Koppers Co., Inc. v. S & S Corrugated Paper Machinery Co., Inc., 517 F.2d 1182, 1188 (2d Cir. 1975); Carter-Wallace v. Otte, 474 F.2d 529, 539-40 (2d Cir. 1972), cert. denied, 412 U.S. 929, 93 S.Ct. 2753, 37 L.Ed.2d 156 (1973); Reiner v. I. Leon Co., 285 F.2d 501, 503 (2d Cir. 1960).
The judgment of the district court is therefore affirmed.
EXHIBIT A
Drawing of Parts of Peloschek Magnetic Recording Head, Shown in Block Form, Before Bonding and Slicing